Drosophila PSI controls circadian period and the phase of circadian behavior under temperature cycle via tim splicing

  1. Lauren E Foley
  2. Jinli Ling
  3. Radhika Joshi
  4. Naveh Evantal
  5. Sebastian Kadener
  6. Patrick Emery  Is a corresponding author
  1. University of Massachusetts Medical School, United States
  2. Hebrew University of Jerusalem, Israel
  3. Brandeis University, United States

Abstract

The Drosophila circadian pacemaker consists of transcriptional feedback loops subjected to post-transcriptional and post-translational regulation. While post-translational regulatory mechanisms have been studied in detail, much less is known about circadian post-transcriptional control. Thus, we targeted 364 RNA binding and RNA associated proteins with RNA interference. Among the 43 hits we identified was the alternative splicing regulator P-element somatic inhibitor (PSI). PSI regulates the thermosensitive alternative splicing of timeless (tim), promoting splicing events favored at warm temperature over those increased at cold temperature. Psi downregulation shortens the period of circadian rhythms and advances the phase of circadian behavior under temperature cycle. Interestingly, both phenotypes were suppressed in flies that could produce TIM proteins only from a transgene that cannot form the thermosensitive splicing isoforms. Therefore, we conclude that PSI regulates the period of Drosophila circadian rhythms and circadian behavior phase during temperature cycling through its modulation of the tim splicing pattern.

Introduction

Circadian rhythms are the organism’s physiological and behavioral strategies for coping with daily oscillations in environment conditions. Inputs such as light and temperature feed into a molecular clock via anatomical and molecular input pathways and reset it every day. Light is the dominant cue for entraining the molecular clock, but temperature is also a pervasive resetting signal in natural environments. Paradoxically, clocks must be semi-resistant to temperature: they should not hasten in warm summer months or lag in the winter cold (this is called temperature compensation), but they can synchronize to the daily rise and fall of temperature (temperature entrainment) (Pittendrigh, 1960). Not only can temperature entrain the clock, it also has a role in seasonal adaptation by affecting the phase of behavior (see for example Majercak et al., 1999).

Molecular circadian clocks in eukaryotes are made up of negative transcriptional feedback loops (Dunlap, 1999). In Drosophila, the transcription factors CLOCK (CLK) and CYCLE (CYC) bind to E-boxes in the promoters of the clock genes period (per) and timeless (tim) and activate their transcription. PER and TIM proteins accumulate in the cytoplasm where they heterodimerize and enter the nucleus to feedback and repress the activity of CLK and CYC and thus downregulate their own transcription (Hardin, 2011). This main loop is strengthened by a scaffolding of interlocked feedback loops involving the transcription factors vrille (vri), PAR domain protein 1 (Pdp1) and clockwork orange (cwo). Post-translational modifications are well-established mechanisms for adjusting the speed and timing of the clock (Tataroglu and Emery, 2015).

Increasing evidence indicates that post-transcriptional mechanisms controlling gene expression are also critical for the proper function of circadian clocks in many organisms. In Drosophila, the post-transcriptional regulation of per mRNA has been best studied. per mRNA stability changes as a function of time (So and Rosbash, 1997). In addition, per contains an intron in its 3’UTR (dmpi8) that is alternatively spliced depending on temperature and lighting conditions (Majercak et al., 1999; Majercak et al., 2004). On cold days, the spliced variant is favored, causing an advance in the accumulation of per transcript levels as well as an advance of the evening activity peak. This behavioral shift means that the fly is more active during the day when the temperature would be most tolerable in their natural environment. The temperature sensitivity of dmpi8 is due to the presence of weak non-canonical splice sites. However, the efficiency of the underlying baseline splicing is affected by four single nucleotide polymorphisms (SNPs) in the per 3’UTR that vary in natural populations and form two distinct haplotypes (Low et al., 2012; Cao and Edery, 2017). Also, while this splicing is temperature-sensitive in two Drosophila species that followed human migration, two species that remained in Africa lack temperature sensitivity of dmpi8 splicing, (Low et al., 2008). Furthermore, Zhang et al. (2018) recently demonstrated that the the trans-acting splicing factor B52 enhances dmpi8 splicing efficiency, and this effect is stronger with one of the two haplotypes. per is also regulated post-transcriptionally by the TWENTYFOUR-ATAXIN2 translational activation complex (Zhang et al., 2013; Lim et al., 2011; Lim and Allada, 2013a; Lee et al., 2017). This complex works by binding to per mRNA as well as the cap-binding complex and poly-A binding protein. This may enable more efficient translation by promoting circularization of the transcript. Interestingly, this mechanism appears to be required only in the circadian pacemaker neurons. Non-canonical translation initiation has also been implicated in the control of PER translation (Bradley et al., 2012). Regulation of PER protein translation has also been studied in mammals, with RBM4 being a critical regulator of mPER1 expression (Kojima et al., 2007). In flies however, the homolog of RBM4, LARK, regulates the translation of DBT, a PER kinase (Huang et al., 2014). miRNAs have emerged as important critical regulators of circadian rhythms in Drosophila and mammals, affecting the circadian pacemaker itself, as well as input and output pathways controlling rhythmic behavioral and physiological processes (Tataroglu and Emery, 2015; Lim and Allada, 2013b).

RNA-associated proteins (RAPs) include proteins that either bind directly or indirectly to RNAs. They mediate post-transcriptional regulation at every level. Many of these regulated events – including alternative splicing, splicing efficiency, mRNA stability, and translation – have been shown to function in molecular clocks. Thus, to obtain a broad view of the Drosophila circadian RAP landscape and its mechanism of action, we performed an RNAi screen targeting 364 of these proteins. This led us to discover a role for the splicing factor P-element somatic inhibitor (PSI) in regulating the pace of the molecular clock through alternative splicing of tim.

Results

An RNAi screen for RNA-associated proteins controlling circadian behavioral rhythms

Under constant darkness conditions (DD) flies have an intrinsic period length of about 24 hr. To identify novel genes that act at the post-transcriptional level to regulate circadian locomotor behavior, we screened 364 genes, which were annotated in either Flybase (FB2014_03, Thurmond et al., 2019) or the RNA Binding Protein Database (Cook et al., 2011) as RNA binding or involved in RNA associated processes, using period length as a readout of clock function (Supplementary file 1: RAP Screen Dataset). We avoided many, but not all, genes with broad effects on gene expression, such as those encoding essential splicing or translation factors. When possible, we used at least two non-overlapping RNAi lines from the TRiP and VDRC collections. RNAi lines were crossed to two different GAL4 drivers: tim-GAL4 (Kaneko et al., 2000) and Pdf-GAL4 (Renn et al., 1999) each combined with a UAS-dicer-2 transgene to enhance the strength of the knockdown (Dietzl et al., 2007). These combinations will be abbreviated as TD2 and PD2, respectively. tim-GAL4 drives expression in all cells with circadian rhythms in the brain and body (Kaneko et al., 2000), while Pdf-GAL4 drives expression in a small subset of clock neurons in the brain: the PDF-positive small (s) and large (l) LNvs (Renn et al., 1999). Among them, the sLNvs are critical pacemaker neurons that drive circadian behavior in DD (Renn et al., 1999; Stoleru et al., 2005). In the initial round of screening, we tested the behavior of 4–8 males for each RNAi line crossed to both TD2 and PD2 (occasionally, fewer males were tested if a cross produced little progeny). We also crossed some RNAi lines to w1118 (+) flies (most were lines selected for retest, see below). We noticed that RNAi/+ control flies for the TRiP collection were 0.3 hr shorter than those of the VDRC collection (Figure 1A). Furthermore, the mean period from all RNAi lines crossed to either PD2 or TD2 was significantly shorter for the TRiP collection than for the VDRC collection (Figure 1A) (0.2 hr, TD2 crosses; 0.5 hr, PD2 crosses). We also found that many of the VDRC KK lines that resulted in long period phenotypes when crossed to both drivers contained insertions in the 40D locus (VDRC annotation), although this effect was stronger with PD2 than TD2. It has been shown that this landing site is in the 5’UTR of tiptop (tio) and can lead to non-specific effects in combination with some GAL4 drivers, likely due to misexpression of tio (Vissers et al., 2016; Green et al., 2014). Indeed, when we crossed a control line that contains a UAS insertion at 40D (40D-UAS) to PD2, the progeny also had a ca. 0.6 hr longer period relative to the PD2 control (Figure 1B). Thus, in order to determine a cutoff for candidates to further investigate, we analyzed the data obtained in our screen from the TRiP, VDRC, and the 40D KK VDRC lines independently (Figure 1C). These data are represented in two overlaid histograms that show period distributions: one for the TD2 crosses (blue) and one for the PD2 crosses (magenta). We chose a cutoff of two standard deviations (SD) from the mean period length for each RNAi line set. RNAi lines were selected for repeat if knockdown resulted in period lengths above or below the 2-SD cutoff. We also chose to repeat a subset of lines that did not pass the cutoff but were of interest and showed period lengthening or shortening, as well as lines that were highly arrhythmic in constant darkness (DD) or had an abnormal pattern of behavior in a light-dark cycle (LD). After a total of three independent experiments, we ended up with 43 candidates (Table 1) that passed the period length cutoffs determined by the initial screen; 31 showed a long period phenotype, while 12 had a short period. One line showed a short period phenotype with PD2 but was long with TD2 (although just below the 2-SD cutoff). Although loss of rhythmicity was also observed in many lines (Supplementary file 1), we decided to focus the present screen on period alterations to increase the probability of identifying proteins that regulate the circadian molecular pacemaker. Indeed, a change in the period length of circadian behavior is most likely caused by a defect in the molecular pacemaker of circadian neurons, while an increase in arrhythmicity can also originate from disruption of output pathways, abnormal development of the neuronal circuits underlying circadian behavioral rhythms, or cell death in the circadian neural network, for example.

An RNAi screen of RNA associated proteins identifies long and short period hits.

(A–B) Background effect of TRiP and VDRC collections on circadian period length. Circadian period length (hrs) is plotted on the y axis. RNAi collection and genotypes are labeled. Error bars represent SEM. (A) Left group (black bars): Patterned bars are the average of period lengths of a subset of RNAi lines in the screen crossed to w1118 (TRiP/+ N = 17 crosses, VDRC/+ N = 46 crosses, 40D KK VDRC/+ N = 20 crosses). Solid bar is the w1118 control (N = 20 crosses). Middle group (blue bars): Patterned bars are the average of period lengths of all RNAi lines in the screen crossed to tim-GAL4, UAS-Dicer2 (TD2) (TRiP/TD2 N = 151 crosses, VDRC/TD2 N = 340 crosses, 40D KK VDRC/TD2 N = 61 crosses). Solid bar is the TD2/+ control (N = 35 crosses). Right group (magenta bars): Patterned bars are the average of period lengths of all RNAi lines in the screen crossed to Pdf-GAL4, UAS-Dicer2 (PD2) (TRiP/PD2 N = 176 crosses, VDRC/PD2 N = 448 crosses, 40D KK VDRC/PD2 N = 69 crosses). Solid bar is the PD2/+ control (N = 36 crosses). One-way ANOVA followed by Tukey’s multiple comparison test: *p<0.05, ***p<0.001, ****p<0.0001. Note that the overall period lengthening, relative to wild-type (w1118), when RNAi lines are crossed to TD2 or PD2 is a background effect of our drivers (see main text), while the period differences between the TRiP (shorter) and VDRC (longer) collections is most likely a background effect of the RNAi lines themselves. There is also a lengthening effect of the 40D insertion site in the VDRC KK collection that cannot be explained by a background effect, as it is not present in the RNAi controls (Left panel). Instead the lengthening was only observed when these lines were crossed to our drivers. A modest effect was seen with TD2 (middle panel) and a larger effect was seen with PD2 (right panel). (B) The period lengthening effect of the VDRC 40D KK lines is likely due to overexpression of tio, as we observed lengthening when a control line that lacks a RNAi transgene, but still has a UAS insertion in the 40D (40D-UAS) locus was crossed to PD2. N = 32 flies per genotype, ****p<0.0001, Unpaired Student’s t-test. (C) Histogram of period lengths obtained in the initial round of screening. Number of lines per bin is on the y axis. Binned period length (hrs) is on the x axis. Bin size is 0.1 hr. TD2 crosses are in blue and PD2 crosses are in magenta. Dashed lines indicate our cutoff of 2 standard deviations from the mean. Number of crosses that fell above or below the cutoff is indicated. Top panel: TRiP lines. 0 lines crossed to TD2 and 2 lines crossed to PD2 gave rise to short periods and were selected for repeats. four lines crossed to TD2 and 10 lines crossed to PD2 gave rise to long periods and were selected for repeats. Middle panel: VDRC lines. eight lines crossed to TD2 and 5 lines crossed to PD2 gave rise to short periods and were selected for repeats. 12 lines crossed to TD2 and 20 lines crossed to PD2 gave rise to long periods and were selected for repeats. Bottom panel: VDRC 40D KK lines. one line crossed to TD2 and 1 line crossed to PD2 gave rise to short periods and were selected for repeats. two lines crossed to TD2 and 3 lines crossed to PD2 gave rise to long periods and were selected for repeats.

Table 1
Circadian behavior in DD of screen candidates
GeneRNAi LineDrivern% of
Rhythmic Flies
Period Average
±SEM
Power Average
±SEM
Atx-1GD11345TD2247526 ± 0.161.5 ± 4.1
PD2177626.4 ± 0.150.7 ± 5.6
KK108861TD2247925.7 ± 0.149.1 ± 4.7
PD2237426.2 ± 0.161.8 ± 4.5
barcGD9921PD2207526.5 ± 0.246.9 ± 5.6
KK101606**TD268325.3 ± 0.555.4 ± 12.7
PD2167527 ± 0.443.9 ± 5.1
bsfJF01529TD2248825.8 ± 0.168.4 ± 4.6
PD2246725.7 ± 0.147.6 ± 4.1
CG16941GD9241PD280
HMS00157PD224423.428.3
KK102272PD280
CG32364HMS03012PD2248825.7 ± 0.158.9 ± 3
CG42458KK106121TD2233526.5 ± 0.238.3 ± 4.9
PD2228226.2 ± 0.171 ± 4.1
CG4849KK101580TD210
PD2246327.3 ± 0.248.8 ± 4.1
CG5808KK102720*TD2237027.4 ± 0.145.3 ± 5.1
PD2245428.5 ± 0.634.8 ± 2.7
CG6227GD11867TD210
PD2166326.7 ± 0.251.4 ± 7
KK108174TD240
PD2203024.2 ± 0.430.9 ± 3.5
CG7903KK103182*TD224823.626.3
PD2247526.4 ± 0.249.1 ± 3.7
CG8273GD13870TD2248325.9 ± 0.147.3 ± 4.6
PD21410025.4 ± 0.151.2 ± 4.8
KK102147TD2245825.5 ± 0.141.1 ± 5
PD22310025.7 ± 0.164.3 ± 3.9
CG8636GD13992PD2125026.9 ± 0.236 ± 6.4
KK110954TD210
PD2196326.3 ± 0.351.4 ± 5.6
CG9609HMS01000PD2244626.3 ± 0.246.1 ± 6.5
KK109846TD2237825.3 ± 0.148.5 ± 4.2
PD2239126.3 ± 0.156.4 ± 3.9
Cnot4JF03203TD2232623.7 ± 0.139.8 ± 6
PD2317723.9 ± 0.151.1 ± 3.2
KK101997TD2324723.9 ± 0.137.3 ± 2.9
PD2279325 ± 0.148 ± 4.1
Dcp2KK101790TD2226426 ± 0.149.7 ± 5.3
PD2249225.9 ± 0.162.5 ± 4.1
eIF1KK109232*PD224423.268.9
eIF3lKK102071TD2242126 ± 0.228.9 ± 2.4
PD22310025.7 ± 0.162.5 ± 3.9
Hrb98DEHMS00342PD2229125.8 ± 0.160.2 ± 4.1
l(1)G0007GD8110PD2246326.3 ± 0.242.4 ± 3.7
KK102874TD2241726.9 ± 0.432.6 ± 5.5
PD2234826.7 ± 0.248 ± 6.1
LSm7GD7971PD2223628 ± 0.443.5 ± 5.6
ncmGD7819PD280
KK100829*PD2193223.3 ± 0.134.4 ± 5.6
Nelf-AKK101005TD2246326.4 ± 0.152.9 ± 4.4
PD2237424.8 ± 0.159.4 ± 4.5
Not1GD9640PD223422.643.6
KK100090PD2103023.8 ± 0.339.4 ± 4.7
Not3GD4068PD280
KK102144PD2211423.6 ± 0.130.8 ± 2.1
Patr-1KK104961*TD2233026.3 ± 0.233.6 ± 3
PD2246327.1 ± 0.238.3 ± 3.6
Pcf11HMS00406PD28132420.1
KK100722PD2242123.3 ± 0.135.4 ± 5
pcmGD10926TD2166325.7 ± 0.136.6 ± 4.1
PD2205526.3 ± 0.240.4 ± 3.8
KK108511TD2242125.7 ± 0.240.7 ± 7.8
PD2241727.7 ± 0.632.9 ± 6.1
PsiGD14067TD2487923.7 ± 0.0749.6 ± 3.0
PD2328424.2 ± 0.153.3 ± 4.1
HMS00140TD22410024 ± 0.161.8 ± 4.2
PD2208524.5 ± 0.152.9 ± 5.6
JF01476TD2249224 ± 0.164.7 ± 4.9
PD2249224.3 ± 0.153.2 ± 4
KK101882TD2357723.6 ± 0.0661.9 ± 3.7
PD2478924.7 ± 0.0656.3 ± 3.4
RgaGD9741TD2242126.2 ± 0.132.8 ± 3.2
PD2223625.4 ± 0.236.1 ± 4.7
RpS3GD4577PD2145726.4 ± 0.248.9 ± 5.9
JF01410PD2245025.6 ± 0.234.9 ± 2.3
KK109080PD283826 ± 1.334.5 ± 6.3
Rrp6GD12195PD2101024.527.2
KK100590PD2211023.643.2
sbrHMS02414TD2138526.8 ± 0.248.7 ± 5.3
PD22110024.9 ± 0.157.2 ± 4.6
Set1GD4398TD2209025.8 ± 0.152.1 ± 4.2
PD2137725.3 ± 0.142.1 ± 5.5
HMS01837TD2237825.6 ± 0.147.9 ± 3.6
PD2249224.8 ± 0.150 ± 3.8
SmBGD11620PD2136926.2 ± 0.152.1 ± 8
HM05097PD2245825.6 ± 0.145.2 ± 4.4
KK102021PD2210025.667.1
SmEGD13663PD2245825.7 ± 0.337.3 ± 3.3
HMS00074PD2810024.5 ± 0.155.1 ± 7.4
KK101450PD2156726.5±51.3 ± 7.8
SmFJF02276PD2247525.8 ± 0.146.3 ± 3.9
KK107814PD2215727.3 ± 0.345.4 ± 4.2
smgGD15460PD2245826.5 ± 0.239 ± 3.5
Smg5KK102117TD2235223.7 ± 0.138.9 ± 3.7
PD2247923.9 ± 0.158.5 ± 4.3
SmnJF02057TD236724.225.9
PD2245425.7 ± 0.147.2 ± 3.6
KK106152TD2246725.3 ± 0.139.7 ± 3.5
PD2249626.3 ± 0.248.7 ± 2.7
snRNP-U1-CGD11660PD2118225.7 ± 0.156.5 ± 6.1
HMS00137PD2249225.8 ± 0.155.9 ± 4.1
SpxGD11072PD2146426.5 ± 0.256.1 ± 7.4
KK108243TD2410024 ± 0.247.5 ± 10.2
PD2197926.9 ± 0.356.4 ± 5
Srp54kGD1542PD250
KK100462PD2241723.7 ± 0.431.3 ± 6
Zn72DGD11579TD2288926.3 ± 0.146.1 ± 4.6
PD2228226.4 ± 0.159.4 ± 6.9
KK100696TD2267326.8 ± 0.157 ± 3.6
PD2248326 ± 0.157 ± 4.5
  1. *Line contains insertion at 40D.

    ** Unknown if line contains insertion at 40D.

Among the 43 candidate genes (Tables 1 and 2), we noticed a high proportion of genes involved or presumed to be involved in splicing (17), including five suspected or known to impact alternative splicing. Perhaps not surprisingly, several genes involved in snRNP assembly were identified in our screen. Their downregulation caused long period phenotypes. We also noticed the presence of four members of the CCR4-NOT complex, which can potentially regulate different steps of mRNA metabolism, including deadenylation, and thus mediate translational repression. Their downregulation mostly caused short period phenotypes and tended to result in high levels of arrhythmicity. Rga downregulation, however, resulted in a long period phenotype, suggesting multiple functions for the CCR4-NOT complex in the regulation of circadian rhythms. Interestingly, two genes implicated in mRNA decapping triggered by deadenylation, were also identified, with long periods observed when these genes were downregulated. Moreover, POP2, a CCR4-NOT component, was recently shown to regulate tim mRNA and protein levels (Grima et al., 2019). Another gene isolated in our screen, SMG5, was also recently found to impact circadian behavior (Ri et al., 2019). This validates our screen.

Table 2
Predicted or known functions of screen candidates
GeneMolecular function (based on information from Flybase) (Thurmond et al., 2019)
Atx-1RNA binding
barcmRNA splicing; mRNA binding; U2 snRNP binding
bsfmitochondrial mRNA polyadenylation, stability, transcription, translation; polycistronic mRNA processing; mRNA 3'-UTR binding
CG16941/Sf3a1alternative mRNA splicing; RNA binding
CG32364/tuttranslation; RNA binding
CG42458mRNA binding
CG4849mRNA splicing; translational elongation
CG5808mRNA splicing; protein peptidyl-prolyl isomerization; regulation of phosphorylation of RNA polymerase II C-terminal domain; mRNA binding
CG6227alternative mRNA splicing; ATP-dependent RNA helicase activity
CG7903mRNA binding
CG8273/SonmRNA processing; mRNA splicing; RNA binding
CG8636/eIF3g1translational initiation; mRNA binding
CG9609transcription; proximal promoter sequence-specific DNA binding
Cnot4CCR4-NOT complex
Dcp2deadenylation-dependent decapping of mRNA; cytoplasmic mRNA P-body assembly; RNA binding
eIF1ribosomal small subunit binding; RNA binding; translation initiation
eIF3ltranslational initiation
Hrb98DEtranslation; alternative mRNA splicing; mRNA binding
l(1)G0007alternative mRNA splicing; 3'−5' RNA helicase activity
LSm7mRNA splicing; mRNA catabolic process; RNA binding
ncmmRNA splicing; RNA binding
Nelf-Atranscription elongation; RNA binding
Not1translation; poly(A)-specific ribonuclease activity; CCR4-NOT complex
Not3translation; transcription; poly(A)-specific ribonuclease activity; CCR4-NOT complex
Patr-1cytoplasmic mRNA P-body assembly; deadenylation-dependent decapping of mRNA; RNA binding
Pcf11mRNA polyadenylation; transcription termination; mRNA binding
pcmcytoplasmic mRNA P-body assembly; 5'−3' exonuclease activity
Psialternative mRNA splicing; transcription; mRNA binding
Rgatranslation; transcription; poly(A)-specific ribonuclease activity; CCR4-NOT complex
RpS3DNA repair; translation; RNA binding; structural constituent of ribosome
Rrp6chromosome segregation; mRNA polyadenylation; nuclear RNA surveillance; 3'−5' exonuclease activity
sbrmRNA export from nucleus; mRNA polyadenylation; RNA binding
Set1histone methyltransferase activity; nucleic acid binding; contains an RNA Recognition Motif
SmBmRNA splicing; RNA binding
SmEmRNA splicing; spliceosomal snRNP assembly
SmFmRNA splicing; spliceosomal snRNP assembly; RNA binding
smgRNA localization; translation; mRNA poly(A) tail shortening; transcription; mRNA binding
Smg5nonsense-mediated decay; ribonuclease activity
Smnspliceosomal snRNP assembly; RNA binding
snRNP-U1-CmRNA 5'-splice site recognition; mRNA splicing, alternative mRNA splicing
SpxmRNA splicing; mRNA binding
Srp54kSRP-dependent cotranslational protein targeting to membrane; 7S RNA binding
Zn72DmRNA splicing; RNA binding

Knockdown of Psi shortens the period of behavioral rhythms

A promising candidate to emerge from our screen was the alternative splicing regulator PSI (Labourier et al., 2001; Siebel et al., 1992). Knockdown of Psi with both TD2 and PD2 crossed to two non-overlapping RNAi lines from the VDRC collection (GD14067 and KK101882) caused a significant period shortening, compared to the TD2/+ and PD2/+ controls (Figure 2A–E, Table 3), which the experimental flies need to be compared to since the GAL4 drivers in the TD2 and PD2 combination cause a previously reported dominant ca. 0.8 hr period lengthening (Figure 2C, left panel (TD2/+ compared to w1118); Kaneko et al., 2000; Renn et al., 1999; Zhang and Emery, 2013; Zhang et al., 2013). Importantly, the RNAi lines did not cause period shortening on their own (Figure 2C left panel, Table 3). While most experiments were performed at 25°C, we noticed that at 30°C, TD2/+ control had a period of ca. 24 hr (Figure 2C). We could thus meaningfully compare TD2/RNAi flies to both RNAi/+ and TD2/+ control at that temperature. The period of the experimental flies was significantly shorter than both controls (Figure 2C). Two additional lines from the TRiP collection (JF01476 and HMS00140) also caused period shortening when crossed to TD2 (Table 1). Interestingly, HMS00140 targets only the Psi-RA isoform, indicating that the RA isoform is important for the control of circadian period (Figure 2A). Since four RNA lines caused a similar phenotype and only two of them partially overlapped (Figure 2A), we are confident that the period shortening was not caused by off-target effects. Moreover, both the KK101882 and GD14067 lines have been shown to efficiently downregulate Psi (Guo et al., 2016), and we confirmed by quantitative Real-Time PCR (qPCR) that the RNAi line KK101882, which gave the shortest period phenotype with TD2, significantly reduced Psi mRNA levels in heads (Figure 2—figure supplement 1). This line was selected for most of the experiments described below as it gave the strongest period phenotype.

Figure 2 with 2 supplements see all
Expression level of Psi affects the circadian behavior period length and circadian rhythmicity.

(A) Schematic of Psi isoforms and position of the long and short hairpins used in this study. Adapted from Ensembl 94 (Zerbino et al., 2018). (B–E) Knockdown of Psi shortens the behavioral period. (B) Double-plotted actograms showing the average activities during 3 days in LD and 5 days in DD. Left panel: TD2/+ (control) flies. Right panel: TD2/PsiRNAi (Psi knockdown) flies. Note the short period of Psi knockdown flies. n = 8 flies/genotype. (C–E) Circadian period length (hrs) is plotted on the y axis. Genotypes are listed on the x axis. Error bars represent SEM. Solid black bar is w1118 (WT) control; solid blue, magenta and gray bars are driver controls; patterned bars are Psi knockdown with two non-overlapping RNAi lines: GD14067 (PsiRNAiGD) and KK101882 (PsiRNAiKK). *p<0.05, ***p<0.001, ****p<0.0001, one-way ANOVA followed by Tukey’s multiple comparison test (C) Dunnett’s multiple comparison test (D and E). (C) Knockdown in all circadian tissues. Left panel 25°C, right panel 30°C. Note that even at 25°C, the experimental flies are shorter than their respective RNAi/+ control, despite the dominant period lengthening caused by TD2 (D) Knockdown in PDF+ circadian pacemaker neurons. (E) Knockdown in PDF- circadian tissues. In D and E, only the driver controls are shown, since they are the controls which the experimental flies need to be compared to because of the dominant period lengthening caused by PD2 and TD2. (F–H) Overexpression of Psi lengthens the behavioral period and decreases rhythmicity. Left panels: Circadian period length (hrs) is plotted on the y axis. Error bars represent SEM. Right panels: Percent of flies that remained rhythmic in DD is plotted on the y axis. Both panels: Genotypes are listed on the x axis. Not significant (ns)p>0.05, *p<0.05, ****p<0.0001, one-way ANOVA followed by Tukey’s multiple comparison test. (F) Overexpression of Psi in all circadian tissues lengthened the circadian period and decreased the percent of rhythmic flies. (G) Overexpression of Psi in PDF+ circadian pacemaker neurons caused a slight but non-significant period lengthening compared to the driver control (PG4/+), which is the relevant comparison because of the dominant period lengthening caused by PG4. Rhythmicity was slightly reduced compared to PG4/+ but not compared to UAS-Psi/+. (H) Overexpression of Psi in PDF- circadian tissues lengthened the circadian period and decreased rhythmicity.

Table 3
PSI affects circadian behavior
GenotypePeriod
±SEM
Power
±SEM
n% of
Rhythmic Flies
Psi downregulation and overexpression at 25°C
TD2/+24.8 ± 0.0448.2 ± 2.37182
TD2/PsiRNAiGD23.7 ± 0.0749.6 ± 3.04879
TD2/PsiRNAiKK23.6 ± 0.0661.9 ± 3.73577
PD2/+24.9 ± 0.0450.4 ± 2.17783
PD2/PsiRNAiGD24.2 ± 0.0653.3 ± 4.13284
PD2/PsiRNAiKK24.7 ± 0.0656.3 ± 3.44789
TD2/+; PdfGAL80/+24.5 ± 0.0749.4 ± 2.84075
TD2/PsiRNAiGD; PdfGAL80/+23.8 ± 0.1745.8 ± 5.52450
TD2/PsiRNAiKK; PdfGAL80/+24.0 ± 0.0571.9 ± 4.03995
w111824.1 ± 0.0384.8 ± 2.57099
PsiRNAiGD/+24.2 ± 0.0458.9 ± 2.96394
PsiRNAiKK/+24.0 ± 0.0467.1 ± 3.75596
TG4/+25.2 ± 0.0552.5 ± 2.26888
TG4/+; UAS-Psi/+25.9 ± 0.0731.3 ± 1.230216
PG4/+25.0 ± 0.0566.0 ± 3.52696
PG4/+; UAS-Psi/+25.2 ± 0.0744.0 ± 2.74877
TG4/+; PdfGAL80/+24.6 ± 0.0642.8 ± 2.83784
TG4/+; PdfGAL80/UAS-Psi24.9 ± 0.1931.3 ± 2.811611
UAS-Psi/+24.2 ± 0.0446.4 ± 1.88079
Psi downregulation at 20°C
TD2/+24.9 ± 0.1042.0 ± 3.13959
TD2/PsiRNAiGD23.6 ± 0.0752.2 ± 4.74466
TD2/PsiRNAiKK23.7 ± 0.0843.8 ± 5.54436
PsiRNAiGD/+24.0 ± 0.0946.0 ± 3.73272
PsiRNAiKK/+23.8 ± 0.0839.1 ± 4.93238
Psi downregulation at 30°C
TD2/+23.7 ± 0.0748.2 ± 2.93987
TD2/PsiRNAiGD23.1 ± 0.1338.3 ± 3.84240
TD2/PsiRNAiKK22.8 ± 0.1543.1 ± 4.24141
PsiRNAiGD/+23.6 ± 0.0443.2 ± 3.43275
PsiRNAiKK/+23.5 ± 0.0363.0 ± 3.73190
TIM-HA suppression of PSI's effect on circadian behavior
TG4/PsiRNAiKK; UAS-Dcr2/+23.4 ± 0.0459.5 ± 4.35775
TG4/+; UAS-Dcr2/+24.9 ± 0.0459.4 ± 3.13692
tim0,TG4/tim0; UAS-Dcr2/timHA24.9 ± 0.0744.3 ± 4.02875
tim0,TG4/tim0,PsiRNAiKK; UAS-Dcr2/timHA24.8 ± 0.0650.0 ± 2.93879

The phenotype caused by Psi downregulation was more pronounced with TD2 than with PD2 (Figure 2C–D, Table 3). This was unexpected since the sLNvs - targeted quite specifically by PD2 - determine circadian behavior period in DD (Stoleru et al., 2005; Renn et al., 1999). This could happen if PD2 is less efficient at downregulating Psi in sLNvs than TD2, or if the short period phenotype is not solely caused by downregulation of Psi in the sLNvs. To distinguish between these two possibilities, we used Pdf-GAL80 combined with TD2 to inhibit GAL4 activity specifically in the LNvs (Stoleru et al., 2004), while allowing RNAi expression in all other circadian tissues. With this combination, we also observed a significant period shortening compared to TD2/+; Pdf-GAL80/+ controls, but the period shortening was not as pronounced as with TD2 (Figure 2E, Table 3). We therefore conclude that both the sLNvs and non-PDF cells contribute to the short period phenotype caused by Psi downregulation (see discussion).

Psi overexpression disrupts circadian behavior

Since we observed that downregulating Psi leads to a short period, we wondered whether overexpression would have an inverse effect and lengthen the period of circadian behavior. Indeed, when we overexpressed Psi by driving a UAS-Psi transgene (Labourier et al., 2001) with the tim-GAL4 (TG4) driver, the period length of circadian behavior increased significantly by about 0.7 hr compared to the TG4/+ control (Figure 2F, Table 3). Interestingly, we also observed a severe decrease in the number of rhythmic flies. When we overexpressed Psi with Pdf-GAL4 (PG4), period was not statistically different from control (PG4/+), and rhythmicity was not reduced compared to the UAS-Psi/+ control (Figure 2G). Overexpression of Psi with the tim-GAL4; Pdf-GAL80 combination caused a severe decrease in rhythmicity but caused only a subtle period lengthening compared to TG4/+; Pdf-GAL80/+ controls (Figure 2H, Table 3). The effect of Psi overexpression on period is in line with the knockdown results, indicating that PSI regulates circadian behavioral period through both PDF+ LNvs and non-PDF circadian neurons. However, the increase in arrhythmicity observed with Psi overexpression is primarily caused by non-PDF cells.

Psi downregulation also shortens the period of body clocks

We wanted to further examine the effect of Psi knockdown on the molecular rhythms of two core clock genes: period (per) and timeless (tim). To do this, we took advantage of two luciferase reporter transgenes. We downregulated Psi with the TD2 driver in flies expressing either a TIM-LUCIFERASE (ptim-TIM-LUC) or a PER-LUCIFERASE (BG-LUC) fusion protein under the control of the tim or per promoter, respectively. We estimated period of luciferase activity rhythms over the first two days in DD, because oscillations rapidly dampened. Fully consistent with our behavioral results, the period of LUC activity was significantly shortened by about 1–1.5 hr compared to controls when Psi was downregulated in ptim-TIM-LUC flies (Figure 2—figure supplement 2A and B). Knockdown of Psi in BG-LUC flies resulted in a similar trend, although differences did not reach statistical significance (Figure 2—figure supplement 2C and D). Period was however shorter in experimental flies compared to both control genotypes in all four independent experiments performed with BG-LUC (and all six with ptim-TIM-LUC). Since the luciferase signal in these flies is dominated by light from the abdomen (Lamba et al., 2018; Stanewsky et al., 1997), this indicates that Psi knockdown, shortens the period of circadian clocks in peripheral tissues as well as in the brain neural network that controls circadian behavior.

Alternative splicing of two clock genes, cwo and tim, is altered in Psi knockdown flies

PSI has been best characterized for its role in alternative splicing of the P element transposase gene in somatic cells (Labourier et al., 2001; Siebel et al., 1992). However, it was recently reported that PSI has a wider role in alternative splicing (Wang et al., 2016). Wang et al. reported an RNA-seq dataset of alternative splicing changes that occur when a lethal Psi-null allele is rescued with a copy of Psi in which the AB domain has been deleted (PSIΔAB). This domain is required for the interaction of PSI with the U1 snRNP, which is necessary for PSI to mediate alternative splicing of P element transposase (Labourier et al., 2002). Interestingly, Wang et al. (2016) found that PSIΔAB affects alternative splicing of genes involved in complex behaviors such as learning, memory and courtship. Intriguingly, we found four core clock genes listed in this dataset: tim, cwo, sgg and Pdp1. We decided to focus on cwo and tim, since only one specific splicing isoform of Pdp1 is involved in the regulation circadian rhythm, (Pdp1e) (Zheng et al., 2009), and since the sgg gene produces a very complex set of alternative transcripts. After three days of LD entrainment, we collected RNA samples at four time points on the first day of DD and determined the relative expression of multiple isoforms of cwo and tim in Psi knockdown heads compared to driver and RNAi controls.

CWO is a basic helix-loop-helix (bHLH) transcriptional factor and is part of an interlocked feedback loop that reinforces the main loop by competing with CLK/CYC for E-box binding (Matsumoto et al., 2007; Lim et al., 2007; Kadener et al., 2007; Richier et al., 2008). There are three mRNA isoforms of cwo predicted in Flybase (Figure 3—figure supplement 1A) (Thurmond et al., 2019). Of the three, only cwo-RA encodes a full-length CWO protein. Exon two is skipped in cwo-RB, and in cwo-RC there is an alternative 3’ splice site in the first intron that lengthens exon 2. Translation begins from a downstream start codon in cwo-RB and -RC, because exon two skipping or lengthening, respectively, causes a frameshift after the start codon used in cwo-RA. The predicted start codon in both cwo-RB and cwo-RC would produce an N-terminal truncation of the protein, which would thus be missing the basic region of the bHLH domain and should not be able to bind DNA. The cwo-RB and cwo-RC isoforms may therefore encode endogenous dominant negatives.

We found that the level of the cwo-RB isoform was significantly reduced compared to both controls at CT 9 (Figure 3—figure supplement 1C). The cwo-RA isoform was also reduced compared to both controls at CT9 (Figure 3—figure supplement 1B). This reduction was significant compared to the TD2/+ control (p=0.0002) but was just above the significance threshold compared to the PsiRNAiKK/+ control (p=0.0715). Conversely, cwo-RC isoform expression was significantly increased at CT 15 (Figure 3—figure supplement 1D). The overall expression of all cwo mRNAs in Psi knockdown fly heads was significantly reduced at both CT 9 and CT 15, indicating that the RC isoform’s contribution to total cwo mRNA levels is quite modest (Figure 3—figure supplement 1E).

We then analyzed alternative splicing of tim in Psi knockdown heads compared to controls. Specifically, we looked at the expression of three temperature-sensitive intron inclusion events in tim that all theoretically lead to C-terminal truncations of the protein (Figure 3A). The tim-cold isoform, which is not annotated in Flybase (Thurmond et al., 2019), is dominant at low temperature (18°C) and arises when the last intron is retained (Boothroyd et al., 2007). We found that tim-cold is elevated in Psi knockdown heads at peak levels under 25°C conditions (CT15, Figure 3D). Similarly, we found that another intron inclusion event, tim-sc (tim-short and cold) which has also been shown to be elevated at 18°C and is present in the tim-RN and -RO isoforms (Martin Anduaga et al., 2019), is significantly increased at 25°C in Psi knockdown heads at CT15 (Figure 3B). Thus, interestingly, two intron inclusion events that are upregulated by cold temperature are also both upregulated in Psi knockdown heads at 25°C. In contrast, we found that an intron included in the tim-RM and -RS isoforms (tim-M, for tim-Medium) and shown to be increased at high temperature (29°C, Martin Anduaga et al., 2019; Shakhmantsir et al., 2018) is significantly decreased at CT 9, 15 and 21 in Psi knockdown heads at 25°C (Figure 3F). In the case of tim-sc, it should be noted that the intron is only partially retained, because a cleavage and poly-adenylation signal is located within this intron, thus resulting in a much shorter mature transcript (Martin Anduaga et al., 2019). Based on PSI function, the most parsimonious explanation is that PSI reduces production of tim-sc by promoting splicing of the relevant intron. However, we cannot entirely exclude that PSI regulates the probability of premature cleavage causing the RNA polymerase to undergo transcription termination soon after passing the poly-adenylation signal.

Figure 3 with 3 supplements see all
Knockdown of Psi increases the expression of cold induced tim isoforms and decreases the expression of a warm induced tim isoform.

(A) Schematic of tim isoforms. Flybase transcript nomenclature on left, intron retention events studied here on right (tim-L refers to tim transcripts that do not produce C-terminal truncations of TIM via intron retention). Arrows indicate the location of retained introns: blue, upregulated at cold temperature; red, upregulated at warm temperature. The retained intron that gives rise to the tim-cold isoform is not annotated in Flybase (Thurmond et al., 2019). It is possible that multiple tim-cold transcripts may exist due to alternative splicing and alternative transcription/translation start sites in the 5’ region of the gene (dashed box). However, for simplicity, we depict this region of tim-cold using the most common exons. Adapted from Ensembl 94 (Zerbino et al., 2018). (B, D, F) Relative expression of tim mRNA isoforms at 25°C (normalized to the average of all Psi knockdown time points) in heads on the y axis measured by qPCR. Circadian time (CT) on the x axis. Error bars represent SEM. Gray line: driver control. Black line: RNAi control. Dashed line: Psi knockdown. Controls, N = 3. Psi knockdown, N = 5 (3 technical replicates per sample). Both driver and RNAi control compared to Psi knockdown, two-way ANOVA followed by Tukey’s multiple comparison test: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. (C, E, G) Relative expression of tim mRNA isoforms at 18°C and 29°C (normalized to the average of all Psi knockdown time points). Solid line: RNAi control. Dashed line: Psi RNAi knockdown. Blue indicates flies were transferred to 18°C at CT0 (start of subjective day) on the first day of DD. Red indicates flies were transferred to 29°C. N = 3 (3 technical replicates per sample). 18°C samples compared to 29°C samples, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, two-way ANOVA followed by Tukey’s multiple comparison test. (C) Blue asterisks refer to RNAi control compared to Psi knockdown.

Collectively, these results indicate that, in wild-type flies, PSI shifts the balance of tim alternative splicing events toward a warm temperature tim RNA isoform profile at an intermediate temperature (25°C). This could be achieved either by altering the temperature sensitivity of tim introns, or by promoting a ‘warm temperature splicing pattern’ independently of temperature. We therefore also measured tim splicing isoforms at 18°C and 29°C (Figure 3C,E,G). We entrained flies for 3 days in LD at 25°C to maintain similar levels of GAL4 expression and thus of Psi knockdown (the GAL4/UAS system’s activity increases with temperature, Duffy, 2002). We then shifted them to either 18°C or 29°C at CT 0 on the first day of DD and collected samples at CT 3, 9, 15 and 21. We found that both the tim-cold intron and the tim-sc introns were elevated at 18°C in both Psi knockdown heads and controls (Figure 3C and E). Thus, Psi knockdown does not block the temperature sensitivity of these introns. tim-M levels were unexpectedly variable in DD, particularly in the Psi knockdown flies, perhaps because of the temperature change. Nevertheless, we observed a trend for the tim-M intron retention to be elevated at 29°C (Figure 3G), further supporting our conclusion that Psi knockdown does not affect the temperature sensitivity of tim splicing, but rather determines the ratio of tim mRNA isoforms, and it does this at all temperatures.

As expected from these results, Psi downregulation did not affect the ability of flies to adjust the phase of their evening and morning peak to changes in temperature (Figure 3—figure supplement 2). We also tested whether Psi knockdown flies responded normally to short light pulses, since TIM is the target of the circadian photoreceptor CRY (Emery et al., 1998; Stanewsky et al., 1998; Lin et al., 2001; Busza et al., 2004; Koh et al., 2006). These flies could both delay or advance the phase of their circadian behavior in response to early or late-night light pulses, respectively (Figure 3—figure supplement 3). We noticed however a possible slight shift of the whole Phase Response Curve toward earlier times. This would be expected since the pace of the circadian clock is accelerated.

PSI controls the phase of circadian behavior under temperature cycle

Since PSI regulates thermosensitive tim splicing events, we wondered whether it might have an impact on circadian behavioral responses to temperature. As mentioned above, Psi downregulation does not affect Drosophila’s ability to adjust the phase of their behavior to different constant ambient temperatures, under a LD cycle (Figure 3—figure supplement 2). Psi knockdown did not appear to affect temperature compensation, as these flies essentially responded to temperature in a similar way as their TD2/+ control, with shorter period at 29°C (Figure 4—figure supplement 1). However, we found a striking phenotype in flies with Psi downregulation under temperature cycle (29/20°C). Once flies had reached a stable phase relationship with the entraining temperature cycle (Busza et al., 2007), the phase of the evening peak of activity was advanced by about 2.5 hr in TD2/PsiRNAi, compared to controls, and this with two non-overlapping dsRNAs (Figure 4). Controls included TD2/+ or TD2/VIE-260B (KK host strain), RNAi/+, as well as TD2 crossed to a KK or GD RNAi line that did not produce a circadian phenotype. Importantly, no such phase advance was observed under LD (Figure 3—figure supplement 2), indicating that the short period phenotype does not account for the evening-peak advanced phase under temperature cycle. Rather, the phase advance is specific to temperature entrainment. The morning peak was difficult to quantify as it tended to be of low amplitude.

Figure 4 with 1 supplement see all
Knockdown of Psi advances the phase of circadian behavior under temperature cycle.

(A) Eductions showing the average activity of flies during 4 days of 12:12 29°C(red)/20°C(blue) temperature entrainment (days 7–10) in DD. Top panels: (driver controls) TD2/+ (left), TD2/VIE-260B (right). Middle panels: (RNAi controls) PsiRNAiGD/+ (left), PsiRNAiKK/+ (right). Bottom panels: (Psi knockdown) TD2/PsiRNAiGD (left), TD2/PsiRNAiKK (right). Note that, Psi knockdown flies advance the phase of their evening activity by about 2.5 hr relative to controls. (C–D) Evening peak phase relative to an internal control in each run (w1118) (hrs) is plotted on the y axis. Genotypes are listed on the x axis. Error bars represent SEM. ***p<0.001, ****p<0.0001, one-way ANOVA followed by Tukey’s multiple comparison test. N = 3–5 runs (C) Quantification of PsiRNAiGD knockdown and controls. Note additional RNAi controls: larpRNAiGD/+ (black bar, gray border) and TD2/larpRNAiGD (patterned bar, gray border). larpRNAiGD (GD8214) is an RNAi line from the GD collection that targets a RAP from our screen that was not a hit. (D) Quantification of PsiRNAiKK knockdown and controls. Note additional RNAi controls: VIE260B/+ (white bar, black border), TD2/VIE260B (gray bar), Rbp9RNAiKK/+ (black bar, gray border) and TD2/Rbp9RNAiKK (patterned bar, gray border). VIE260B is a KK collection host strain control containing the 30B transgene insertion site. Rbp9RNAiKK (KK109093) is an RNAi line from the KK collection targeting a RAP from our screen that was not a hit.

tim splicing is required for PSI’s regulation of circadian period and circadian behavior phase under temperature cycle

Because tim is a key element of the circadian transcriptional feedback loop and its splicing pattern is determined by the ambient temperature, we wondered whether PSI might be regulating the speed of the clock and the phase of the evening peak through its effects on tim splicing. We therefore rescued the amorphic tim allele (tim0) with a tim transgene that lacks the known temperature sensitive alternatively spliced introns as well as most other introns (timHA) (Figure 5A) (Rutila et al., 1998). Importantly, the tim0 mutation is a frame-shifting deletion located upstream of the temperature-sensitive alternative splicing events (Myers et al., 1995), and would thus truncate any TIM protein produced from the splice variants we studied. Strikingly, we found that knockdown of Psi in timHA rescued tim0 flies had no impact on the period of circadian behavior (Figure 5B–C, Table 3). Likewise, the evening peak phase under temperature cycles was essentially insensitive to Psi knockdown in timHA rescued tim0 flies (Figure 5D–E). This indicates that PSI controls circadian period in DD and the phase of the evening peak under temperature cycle through tim splicing.

The short period and temperature cycle phase advance effects of Psi knockdown are dependent on tim introns.

(A) Schematic of timHA transgene. The tim promoter is fused upstream of the transcription start site (TSS). Two introns remain in the 5’UTR, upstream of the start codon; however, they are not, to our knowledge, temperature sensitive. A C-terminal HA tag is fused to full length tim cDNA, which lacks any of the introns that are known to be retained at high or low temperatures. (B) Knockdown of Psi with tim-GAL4 and a UAS-dcr2 transgene inserted on the 3rd chromosome also causes period shortening. We used this insertion to more easily generate stocks in a tim0 background, since the tim gene is on the second chromosome, instead of the TD2 combination that has both the tim-GAL4 and UAS-dcr2 transgenes on the 2nd chromosome. ****p<0.0001, Student’s t-test. (C) Period shortening in response to Psi knockdown with tim-GAL4 and UAS-dcr2 is abolished in tim0, ptim-timHA flies that can only produce the full length tim isoform. ns, p=0.1531, Student’s t-test. (B, C) Circadian period length (hrs) is plotted on the y axis. Genotypes are listed on the x axis. Error bars represent SEM. (D) Knockdown of Psi with tim-GAL4 and a UAS-dcr2 3rd chromosome transgene also causes a phase advance in a 12:12 29°C/20°C temperature cycle. (E) The phase advance is abolished in tim0, ptim-timHA flies that can only produce the full length tim isoform. (D, E) Evening peak phase relative to an internal control in each run (w1118) (hrs) is plotted on the y axis. Genotypes are listed on the x axis. Error bars represent SEM. **p<0.01, one-way ANOVA followed by Tukey’s multiple comparison test. N = 3 runs.

Figure 5—source data 1

Psi downregulation in a tim0; timHA background – behavioral period length in DD and temperature cycle phase.

https://cdn.elifesciences.org/articles/50063/elife-50063-fig5-data1-v2.xlsx
Figure 5—source data 2

Figure statistics – Figure 5.

https://cdn.elifesciences.org/articles/50063/elife-50063-fig5-data2-v2.xlsx

Discussion

Our results identify a novel post-transcriptional regulator of the circadian clock: PSI. PSI is required for the proper pace of both brain and body clock, and for proper phase-relationship with ambient temperature cycles. When Psi is downregulated, the circadian pacemaker speeds up and behavior phase under temperature cycles is advanced by 3 hr, and these phenotypes appear to be predominantly caused by an abnormal tim splicing pattern. Indeed, the circadian period and behavior phase of flies that can only produce functional TIM protein from a transgene missing most introns is insensitive to Psi downregulation. We note however that cwo’s splicing pattern is also affected by Psi downregulation, and we did not study sgg splicing pattern, although it might also be controlled by PSI (Wang et al., 2016). We therefore cannot exclude a small contribution of non-tim splicing events to PSI downregulation phenotypes, or that in specific tissues these other splicing events play a greater role than in the brain.

Interestingly, Psi downregulation results in an increase in intron inclusion events that are favored under cold conditions (tim-sc and tim-cold), while an intron inclusion event favored under warm conditions is decreased (tim-M). However, the ability of tim splicing to respond to temperature changes is not abolished when Psi is downregulated (Figure 3C,E,G). This could imply that an as yet unknown factor specifically promotes or represses tim splicing events in a temperature-dependent manner. Another possibility is that the strength of splice sites or tim’s pre-mRNA structure impacts splicing efficiency in a temperature–dependent manner. For example, suboptimal per splicing signals explain the lower efficiency of per’s most 3’ splicing event at warm temperature (Low et al., 2008).

How would the patterns of tim splicing affect the pace of the circadian clock, or advance the phase of circadian behavior under temperature cycles? In all splicing events that we studied, intron retention results in a truncated TIM protein. It is therefore possible that the balance of full length and truncated TIM proteins, which may function as endogenous dominant-negatives, determines circadian period. For example, truncated TIM might be less efficient at protecting PER from degradation, thus accelerating the pacemaker, or affecting its phase. Consistent with this idea, overexpression of the shorter cold-favored tim isoform (tim-sc) shortens period (Martin Anduaga et al., 2019). Strikingly, Psi downregulation increases this isoform’s levels and also results in a short phenotype. Shakhmantsir et al. (2018) also proposed that production of tim-M transcripts (called tim-tiny in their study) delays the rate of TIM accumulation. Such a mechanism could also contribute to the short period we observed when Psi is downregulated, since this reduces tim-M levels, which may accelerate TIM accumulation. Another interesting question is how PSI differentially affects specific splice isoforms of tim. One possibility is that the execution of a specific tim splicing event negatively influences the probability of the occurrence of other splicing events. For example, PSI could downregulate tim-sc and tim-cold by enhancing splicing and removal of the introns whose retention is necessary for production of these isoforms. This could indirectly reduce splicing of the intron that is retained in the warm tim-M isoform and result in tim-M upregulation. Conversely, PSI could directly promote tim-M intron retention and indirectly downregulate production of tim-sc and tim-cold.

Other splicing factors have been shown to be involved in the control of circadian rhythms in Drosophila. SRm160 contributes to the amplitude of circadian rhythms by promoting per expression (Beckwith et al., 2017), while B52/SMp55 and PRMT5 regulate per’s most 3’ splicing, which is temperature sensitive (Zhang et al., 2018; Sanchez et al., 2010). Loss of PRMT5 results in essentially arrhythmic behavior (Sanchez et al., 2010), but this is unlikely to be explained by its effect on per’s thermosensitive splicing. B52/SMp55 knockdown flies show a reduced siesta, which is controlled by the same per splicing (Zhang et al., 2018). With the identification of Psi, we uncover a key regulator of tim alternative splicing pattern and show that this pattern determines circadian period length, while per alternative splicing regulates the timing and amplitude of the daytime siesta. Interestingly, a recent study identified PRP4 kinase and other members of tri-snRNP complexes as regulators of circadian rhythms (Shakhmantsir et al., 2018). Downregulation of prp4 caused excessive retention of the tim-M intron. PSI and PRP4 might thus have complementary functions in tim mRNA splicing regulation, working together to maintain the proper balance of tim isoform expression.

An unexpected finding is the role played by both PDF neurons and other circadian neurons in the short period phenotype observed with circadian locomotor rhythms when we knocked-down Psi. Indeed, it is quite clear from multiple studies that under constant darkness, the PDF-positive sLNvs dictate the pace of circadian behavior (Stoleru et al., 2005; Yao and Shafer, 2014). Why, in the case of Psi downregulation, do PDF negative neurons also play a role in period determination? The explanation might be that PSI alters the hierarchy between circadian neurons, promoting the role of PDF negative neurons. This could be achieved by weakening PDF/PDFR signaling, for example.

While we focused our work on PSI, several other interesting candidates were identified in our screen (Tables 1 and 2). We note the presence of a large number of splicing factors. This adds to the emerging notion that alternative splicing plays a critical role in the control of circadian rhythms. We have already mentioned above several per splicing regulators that can impact circadian behavior. In addition, a recent study demonstrated that specific classes of circadian neurons express specific alternative splicing variants, and that rhythmic alternative splicing is widespread in these neurons (Wang et al., 2018). Interestingly, in this study, the splicing regulator barc, which was identified in our screen and which has been shown to causes intron retention in specific mRNAs (Abramczuk et al., 2017), was found to be rhythmically expressed in LNds. Moreover, in mammals, alternative splicing appears to be very sensitive to temperature, and could explain how body temperature rhythms synchronize peripheral clocks (Preußner et al., 2017). Another intriguing candidate is cg42458, which was found to be enriched in circadian neurons (LNvs and Dorsal Neurons 1) (Wang et al., 2018). In addition to emphasizing the role of splicing, our screen suggests that regulation of polyA tail length is important for circadian rhythmicity, since we identified several members of the CCR4-NOT complex and deadenylation-dependent decapping enzymes. Future work will be required to determine whether these factors directly target mRNAs encoding for core clock components, or whether their effect on circadian period is indirect. Interestingly, the POP2 deadenylase, which is part of the CCR4-NOT complex, was recently shown to regulate tim mRNA levels post-transcriptionally (Grima et al., 2019). It should be noted that while our screen targeted 364 proteins binding or associated with RNA, it did not include all of them. For example, LSM12, which was recently shown to be a part of the ATXN2/TYF complex (Lee et al., 2017), was not included in our screen because it had not been annotated as a potential RAP when we initiated our screen.

In summary, our work provides an important resource for identifying RNA associated proteins regulating circadian rhythms in Drosophila. It identifies PSI is an important regulator of circadian period and circadian phase in response to thermal cycles, and points at additional candidates and processes that determine the periodicity of circadian rhythms.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Gene (Drosophila melanogaster)PsiFLYB:FBgn0014870Flybase name:P-element somatic inhibitor
Gene (Drosophila melanogaster)timFLYB:FBgn0014396Flybase name: timeless
Gene (Drosophila melanogaster)tioFLYB:FBgn0028979Flybase name: tiptop
Gene (Drosophila melanogaster)perFLYB:FBgn0003068Flybase name: period
Gene (Drosophila melanogaster)cwoFLYB:FBgn0259938Flybase name:clockwork orange
Gene (Drosophila melanogaster)RpL32FLYB:FBgn0002626qPCR control
Flybase name:Ribosomal protein L32
Gene (Drosophila melanogaster)larpFLYB:FBgn0261618Flybase name:
La related protein
Gene (Drosophila melanogaster)Rbp9FLYB:FBgn0010263Flybase name:RNA-binding protein 9
Gene (Drosophila melanogaster)Dcr-2FBgn0034246Flybase name:
Dicer-2
Genetic reagent (D. melanogaster)tim-GAL4Kaneko et al., 2000FLYB:FBtp0010385
Genetic reagent (D. melanogaster)Pdf-GAL4Renn et al., 1999FLYB:FBtp0011844
Genetic
reagent (D. melanogaster)
Pdf-GAL80, Pdf-GAL80Stoleru et al., 2004
Genetic reagent (D. melanogaster)UAS-Dcr2Dietzl et al., 2007FLYB:FBti0100275
RRID:BDSC_24650
Chromosome 2
Genetic reagent (D. melanogaster)UAS-Dcr2Dietzl et al., 2007FLYB:FBti0100276Chromosome 3
Genetic reagent (D. melanogaster)PsiRNAi KK101882FLYB:FBal0231542
Genetic reagent (D. melanogaster)PsiRNAi GD14067Dietzl et al., 2007FLYB:FBst0457756
Genetic reagent (D. melanogaster)UAS-PsiLabourier et al., 2001
Genetic reagent (D. melanogaster)BG-LUCStanewsky et al., 1997
Genetic reagent (D. melanogaster)ptim-TIMLUCLamba et al., 2018
Genetic reagent (D. melanogaster)timHARutila et al., 1998FLYB:FBal0143160
Genetic reagent (D. melanogaster)tim0Sehgal et al., 1994FLYB:FBal0035778
Genetic reagent (D. melanogaster)VIE260BVDRC_ID:
60100
 genetic reagent (D. melanogaster)larpRNAi
GD8214
Dietzl et al., 2007VDRC_ID:
17366
Genetic reagent (D. melanogaster)Rbp9RNAi
KK109093
VDRC_ID:
101412
Genetic reagent (D. melanogaster)w1118VDRC_ID:
60000
Genetic reagent (D. melanogaster)40D-UASVDRC_ID:
60101
Sequence-based reagentRpL32-forwardDubruille et al., 2009PCR primersATGCTAAGCTGTCGCACAAA
Sequence-based reagentRpL32-reverseDubruille et al., 2009PCR primersGTTCGATCCGTAACCGATGT
Sequence-based reagentpsi-forwardThis paperPCR primersGGTGCCTTGAATGGGTGAT
Sequence-based reagentpsi-reverseThis paperPCR primersCGATTTATCCGGGTCCTCG
Sequence-based reagenttim-M-forwardThis paperPCR primersTGGGAATCTCGCCCGAAAC
Sequence-based reagenttim-M-reverseThis paperPCR primersAGAAGGAGGAGAAGGAGAGAGG
Sequence-based reagenttim-sc-forwardThis paperPCR primersACTGTGCGATGACTGGTCTG
Sequence-based reagenttim-sc-reverseThis paperPCR primersTGCTTCAAGGAAATCTTCTG
Sequence-
based reagent
tim-cold-forwardThis paperPCR primersCCTCCATGAAGTCCTCGTTCG
Sequence-based reagenttim-cold-reverseThis paperPCR primersATTGAGCTGGGACACCAGG
Sequence-based reagentcwo-fowardThis paperPCR primersTTCCGCTGTCCACCAACTC
Sequence-based reagentcwo-reverseThis paperPCR primersCGATTGCTTTGCTTTACCAGCTC
Sequence-based reagentcwoRA-forwardThis paperPCR primersTCAAGTATGAGAGCGAAGCAGC
Sequence-based reagentcwoRA-reverseThis paperPCR primersTGTCTTATTACGTCTTCCGGTGG
Sequence-based reagentcwoRB-forwardThis paperPCR primersGTATGAGAGCAAGATCCACTTTCC
Sequence-based reagentcwoRB-reverseThis paperPCR primersGATGATCTCCGTCTTCTCGATAC
Sequence-based reagentcwoRC-forwardThis paperPCR primersGTATGAGAGCCAAGCGACCAC
Sequence-based reagentcwoRC-reverseThis paperPCR primersCCAAATCCATCTGTCTGCCTC
Commercial assay or kitDirect-zol RNA MiniPrep kitZymo ResearchZymo Research: R2050
Commercial assay or kitiSCRIPT cDNA synthesis kitBio-RADBio-RAD: 1708891
Commercial assay or kitiTaq Universal SYBR Green SupermixBio-RADBio-RAD: 1725121
Chemical compound, drugD-Luciferin, Potassium SaltGoldbioGoldbio: LUCK-1G
Chemical compound, drugTRIzol ReagentInvitrogenThermoFisher Scientific:15596026
Software, algorithmFaasX softwareGrima et al., 2002http://neuro-psi.cnrs.fr/spip.php?article298&lang=en
Software,
algorithm
MATLAB (MathWorks) signal-processing toolboxLevine et al., 2002MATLAB RRID: SCR_001622
Software, algorithmMS ExcelRRID: SCR_016137
Software, algorithmGraphPad Prism version 7.0 c for Mac OS XGraphPad Software, La Jolla, CA USARRID: SCR_002798www.graphpad.com

Fly stocks

Request a detailed protocol

Flies were raised on a standard cornmeal/agar medium at 25°C under a 12 hr:12 hr light:dark (LD) cycle. The following Drosophila strains were used: w1118 -- w; tim-GAL4, UAS-dicer2/CyO (TD2) (Dubruille et al., 2009) -- y w; Pdf-GAL4, UAS-dicer2/CyO (PD2) (Dubruille et al., 2009) -- y w; Tim-GAL4/CyO (TG4) (Kaneko et al., 2000) -- y w; Pdf-GAL4 (PG4) (Renn et al., 1999) -- w;; UAS-dcr2 (Dietzl et al., 2007) -- y w;; timHA (Rutila et al., 1998) -- yw; TD2; Pdf-Gal80, Pdf-GAL80 (Zhang and Emery, 2013). The following combinations were generated for this study: y w; TG4; Pdf-GAL80, Pdf-GAL80 -- w; tim-GAL4/CyO; UAS-dicer2/TM6B -- tim0,TG4/CyO; UAS-Dcr2/TM6B -- tim0, PsiRNAiKK/CyO; timHA/TM6B. TD2, ptim-TIM-LUC and TD2, BG-LUC transgenic flies expressing a tim-luciferase and per-luciferase fusion gene respectively, combined with the TD2 driver, were used for luciferase experiments. The TIM-LUC fusion is under the control of the tim promoter (ca. 5 kb) and 1st intron (Lamba et al., 2018), BG-LUC contains per genomic DNA encoding the N-terminal two-thirds of PER and is under the control of the per promoter (Stanewsky et al., 1997). RNAi lines (names beginning with JF, GL, GLV, HM or HMS) were generated by the Transgenic RNAi Project at Harvard Medical School (Boston, MA) and obtained from the Bloomington Drosophila Stock Center (Indiana University, USA). RNAi lines (names beginning with GD or KK) and control lines (host strain for the KK library containing landing sites for the RNAi transgenes, VIE-260B, and tio misexpression control strain, 40D-UAS) were obtained from the Vienna Drosophila Stock Center. UAS-Psi flies were kindly provided by D. Rio (Labourier et al., 2001).

Behavioral monitoring and analysis

Request a detailed protocol

The locomotor activity of individual male flies (2–5 days old at start of experiment) was monitored in Trikinetics Activity Monitors (Waltham, MA). Flies were entrained to a 12:12 LD cycle for 3–4 days at 25°C (unless indicated) using I-36LL Percival incubators (Percival Scientific, Perry IA). After entrainment, flies were released into DD for five days. Rhythmicity and period length were analyzed using the FaasX software (courtesy of F. Rouyer, Centre National de la Recherche Scientifique, Gif-sur-Yvette, France) (Grima et al., 2002). Rhythmicity was defined by the criteria – power >20, width >1.5 using the χ2 periodogram analysis. Actograms were generated using a signal-processing toolbox implemented in MATLAB (MathWorks), (Levine et al., 2002). For phase-shifting experiments, groups of 16 flies per genotype were entrained to a 12:12 LD cycle for 5–6 days at 25°C exposed to a 5 min pulse of white fluorescent light (1500 lux) at different time points on the last night of the LD cycle. A separate control group of flies was not light-pulsed. Following the light pulse, flies were released in DD for six days. To determine the amplitude of photic phase shifts, data analysis was done in MS Excel using activity data from all flies, including those that were arrhythmic according to periodogram analysis. Activity was averaged within each group, plotted in Excel, and then fitted with a 4 hr moving average. A genotype-blind observer quantified the phase shifts. The peak of activity was found to be the most reliable phase marker for all genotypes. Phase shifts were calculated by subtracting the average peak phase of the light-pulsed group from the average peak phase of non-light pulsed group of flies. Temperature entrainment was performed essentially as described in Busza et al. (2007). Flies were entrained for 4–5 days in LD followed by 11 days in an 8 hr phase advanced temperature cycle. Behavior was analyzed between day 7 and day 10 of the temperature cycle. Actograms were used to ensure that all genotypes had reached – as expected from Busza et al. (2007) – a stable phase relationship with the temperature cycle. The phase of the evening peak of activity was determined as described for the phase response curve above. Because, under a LD cycle, the evening peak tend to be truncated by the light off transition, we used the approach described in Harrisingh et al. (2007), which compares the percent of activity between ZT17.5–23.5 that occurs between ZT20.5–23.5 (Morning anticipation phase score), or the percent of activity between ZT5.5–11.5 that occurs between ZT8.5–11.5 (Evening anticipation phase score). If phase is advanced, and activity increases earlier than normal, this percent will decrease.

Statistical analysis

Request a detailed protocol

For the statistical analysis of behavioral and luciferase period length, Student’s t-test was used to compare means between two groups, and one-way analysis of variance (ANOVA), coupled to post hoc tests, was used for multiple comparisons. Tukey’s post hoc test was used when comparing three or more genotypes and Dunnett’s post hoc test was used when comparing two experimental genotypes to one control. For the statistical analysis of qPCR and the behavioral phase-shifting experiments, two-way ANOVA, coupled to Tukey’s post hoc test, was used for multiple comparisons. Statistical analyses were performed using GraphPad Prism version 7.0 c for Mac OS X, GraphPad Software, La Jolla California USA, www.graphpad.com. P values and 95% Confidence Intervals are reported in data source files ‘Figure statistics’.

Luciferase experiments

Request a detailed protocol

The luciferase activity of whole male flies on Luciferin (Gold-biotech) containing agar/sucrose medium (170 μl volume, 1% agar, 2% sucrose, 25 mM luciferin), was monitored in Berthold LB960 plate reader (Berthold technologies, USE) in l-36LL Percival incubators with 90% humidity (Percival Scientific, Perry IA). Three flies per well were covered with needle-poked Pattern Adhesive PTFE Sealing Film (Analytical sales and services 961801). The distance between the agar and film was such that the flies were not able to move vertically. Period length was determined from light measurements taken during the first two days of DD. The analysis was limited to this window because TIM-LUC and BG-LUC oscillations severely dampened after the second day of DD. Period was estimated by an exponential dampened cosinor fit using the least squares method in MS Excel (Solver function).

Real-time quantitative PCR

Request a detailed protocol

Total RNA from about 30 or 60 fly heads collected at CT 3, CT9, CT15 and CT21 on the first day of DD were prepared using Trizol (Invitrogen) and Zymo Research Direct-zol RNA MiniPrep kit (R2050) following manufacturer’s instructions. 1 μg of total RNA was reverse transcribed using Bio-RAD iSCRIPT cDNA synthesis kit (1708891) following manufacturer’s instructions. Real-time PCR analysis was performed in triplicate (three technical replicates per sample) using Bio-RAD iTaq Universal SYBR Green Supermix (1725121) in a Bio-RAD C1000 Touch Thermal Cycler instrument. A standard curve was generated for each primer pair, using RNA extracted from wild-type fly heads, to verify amplification efficiency. Data were normalized to RpL32 (Dubruille et al., 2009) using the 2-ΔΔCt method. Primers used: RpL32-forward ATGCTAAGCTGTCGCACAAA; RpL32-reverse GTTCGATCCGTAACCGATGT; psi-forward GGTGCCTTGAATGGGTGAT; psi-reverse CGATTTATCCGGGTCCTCG; tim-M-forward TGGGAATCTCGCCCGAAAC; tim-M-reverse AGAAGGAGGAGAAGGAGAGAGG; tim-sc-forward ACTGTGCGATGACTGGTCTG; tim-sc-reverse TGCTTCAAGGAAATCTTCTG; tim-cold-forward CCTCCATGAAGTCCTCGTTCG; tim-cold-reverse ATTGAGCTGGGACACCAGG; cwo-foward TTCCGCTGTCCACCAACTC; cwo-reverse CGATTGCTTTGCTTTACCAGCTC; cwoRA-forward TCAAGTATGAGAGCGAAGCAGC; cwoRA-reverse TGTCTTATTACGTCTTCCGGTGG; cwoRB-forward GTATGAGAGCAAGATCCACTTTCC; cwoRB-reverse GATGATCTCCGTCTTCTCGATAC; cwoRC-forward GTATGAGAGCCAAGCGACCAC; cwoRC-reverse CCAAATCCATCTGTCTGCCTC.

Data availability

All source data are included in this submission.

The following previously published data sets were used
    1. Wang Q
    2. Tallatero M
    3. Rio D
    (2016) NCBI Gene Expression Omnibus
    ID GSE79916. The PSI-U1 snRNP interaction regulates male mating behavior in Drosophila.

References

Decision letter

  1. Mani Ramaswami
    Reviewing Editor; Trinity College Dublin, Ireland
  2. Ronald L Calabrese
    Senior Editor; Emory University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

A transcriptional feedback loop comprising of PER/TIMELESS and CLOCK/CYCLE underlies periodic oscillation of the circadian clock and diurnal rhythms. However, the activity of these core components is modulated by post-transcriptional and post-translational processes that remain to be fully understood. This work, by Foley and coauthors, begins to address post-transcriptional control mechanisms by screening Drosophila genes encoding RNA-binding and RNA-associated proteins for their role in the control of circadian rhythms. From 364 genes screened a total of 43 candidates (12%) appear to alter period length. The unexpectedly high hit rate suggests that RNA regulation plays a significant role in clock function. The authors then focus on one of these proteins, PSI, and reveal that it encodes a splicing factor that produces timeless splice isoforms described and characterised in a companion paper by Anduaga et al. in this issue of eLife. Several observations support a model in which PSI is required to mediate alternative splicing of tim isoforms, which in turn determines how the phase of the circadian cycles adapts in response to temperature variation. Observations made here suggest that PSI is required for appropriate tim mRNA splicing in both cold and warm conditions, and acts downstream of a primary temperature-sensing mechanism. Additional work is needed to identify how exactly temperature-specific splicing patterns of TIM are determined.

Decision letter after peer review:

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "PSI controls tim splicing and

circadian period in Drosophila" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by Mani Ramaswami as Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

The reviewers concur that your careful screen of RNA-binding proteins is well done and broadly useful. However, the consensus view is that the behavioural evidence to support the model that PSI mediated splicing of Tim contributes to locomotor rhythms is too subtle to be convincing. It also remains possible that thermal adaption affected by psi mutations have nothing to do with locomotor rhythms but perhaps to with do other non-rhythmic phenotypes. Thus, the feeling is that despite the interest of the subject and the overall quality of the work, the manuscript falls short of biological insight required for acceptance.

Reviewer #1:

This is an interesting paper focusing on a splicing factor PSI that regulates timeless isoforms. The screening and experimental work is done well with care taken to exclude background effects. One of the disappointing aspects of the study, and this is no fault of the authors, is the lack of any striking disruption of temperature adaptations with the PSI KD. With all these changes in timeless and cwo splicing, particularly of the temperature sensitive tim isoforms, the locomotor rhythms of the psi KD flies in LD cycles is normal without the kind of dramatic effects seen on siestas when per 3' splicing is disturbed. So I was left scratching my head as to what disruption of psi does phenotypically – small changes in period, sure, but these are not reflected in the LD profiles of Figure 3—figure supplement 2.

As the period seems to change with psi manipulation, I was surprised that the authors did not perform a simple temperature compensation experiment (or did I miss it?). I realize that gal4 is temperature sensitive but the direction of any period change between 18-30° C might have been interesting. Are there compensatory changes of tim transcripts in psi KD? The accompanying paper hinted at that with the cold-sensitive isoforms.

The loss of rhythmicity in Figure 2E and 2G on psi overexpression is striking. Any idea what happens to tim isoforms under these conditions? Clearly something in the dorsal neurons is having a major effect on the rhythmicity of PDF neurons. This is interesting as the DNs do not by themselves generate locomotor rhythms. Surprisingly, the Discussion does not mention this most dramatic result of the manuscript.

Reviewer #2:

This is a very interesting paper from the Kadener and Emery labs about a genetic screen for RNA-binding and RNA-associated genes involved in circadian clock control. After screening 364 genes, they found a total of 43 candidates (12%) that alter period length. Compared to standard genetic screens, this is a very high hit rate, indicating that RNA regulation plays a significant role in clock regulation. If correct, this is an important discovery revealing yet another regulatory level of the already very complex circadian clock mechanism.

The paper is well written and experiments conducted are of high quality and according to the standard in the field. The authors focus on one gene identified in their screen (psi), which they show causes period shortening when down-regulated and arrhythmicity when overexpressed in the entire clock circuit. Effects with more narrow down regulation (Pdf-gal4) are much less pronounced and overall it remains unclear in which parts of the circuit Psi plays a role. Moreover important controls are lacking in the behavior experiments (outlined below). Because the effects on period are rather mild, these controls are important. Also in light of the fact that psi has no measurable effect on the light PRC or normal LD entrainment it seems important to really nail the effect on period. Molecularly it seems clear that psi affects tim-splicing, but the mechanism remains elusive and it is not clear how psi can promote intron retention in certain tim transcripts at cooler temperature and that in other tim transcripts at warmer temperatures.

1) The period shortenings after Psi-knockdown are relatively mild (0.2-1.2 hr depending on the driver used) and the same applies for the lengthening observed after psi-overexpression (0.1 – 0.7 hr) (Figure 2 and Table 3). Given that these period changes are key to the message of the paper (psi controls circadian period), I think it is important to perform a statistical analysis comparing the various genotypes. Also, some key controls are missing: the psi RNAi lines have not been tested without a driver. Data are shown for the total of RNAi-lines from a given collection over + in Figure 1A, but due to the large variability between the different collections with regard to period length, the values for the individual RNAi lines 'over +' must be given. I also checked the source file (Dataset 1) and found some data, which I think shows the RNAi/+ data. From these data it looks the effect of the psi knockdown is even smaller (psiRNAi-kk/+: 24.0, TD2>kk: 23.6, PD2>kk:24.7 and psiRNAi-GD/+ 24.2, TD2>GD: 23.8, PD2>GD: 24.2). So no effect with the Pdf driver on period at all, and a small 0.4 hr effect with TD2. Moreover, no effects with 2 other RNAi lines shown in this Data set with either tim or Pdf drivers. I am not sure if the authors have more control data and if the Data set table represents only the actual screening data, but clearly additional controls are required. This is an important point, because no other behavioral phenotypes after psi knockdown could be observed (LD behavior at different temperatures and light-PRC, Figure 3—figure supplements 2 and 3).

2) The authors see stronger effects with tim-gal4 (both knock-down and overexpression of psi) compared to Pdf-gal4. They aim to distinguish between potentially stronger tim-gal4 expression in the s-LNv, or a period-determining function of non-s-LNv neurons by applying a tim-gal4/Pdf-gal80 combination, which should eliminate gal4 from the s-LNv but not all the other clock cells. Unfortunately, this experiment did not allow for a clear distinction, as the period shortening after knockdown was in between that of tim and Pdf drivers alone. Overexpression with the tim-gal4/Pdf-gal80 combination recapitulated the high percentage of arhythmicity seen with tim alone, but not the period-lengthening, whereas overexpression with Pdf-gal4 only had no effect. Overall the results suggest a role for non-s-LNv neurons in setting period length and controlling rhythmicity, but I think this should be repeated with other, non-s-LNv drivers to get a clearer picture (e.g., it cannot be ruled out that Pdf-gal80 is completely blocking gal4 in the s-LNv). I suggest using the splitE cell-gal4 and drivers specific for the DN1 to solve this issue.

3) psi overexpression with tim-gal4 and tim-gal4/Pdf-gal80 produces very high levels of arrhythmicity and the few rhythmic flies (n=5 and n=4, respectively) have low average power values. Considering the low number of rhythmic flies and the low power values, I find it problematical to claim that overexpression with tim-gal4 causes period-lengthening. Could the authors show actograms that clearly show the longer period in these flies compared to the tim-gal4/Pdf-gal80 flies. Lacking additional support (such as convincing actograms and/or higher n's), I think it is not OK to conclude that the overexpression results are in line with the knockdown results (which are problematical in itself, see major point 1 above).

4) The peripheral clock data in Figure 3 do not look very convincing, due to the poor rhythmicity of the luciferase oscillations in DD. As correctly mentioned in the Materials and methods part, it is expected and was reported previously that rhythms of these reporters dampen rapidly in DD. But I am not sure if it is valid to calculate period values from 48 hr only, particularly if one the 2nd day the oscillations are very low amplitude. Perhaps it would be better to look in LD (where the reporters cycle with high-amplitude) to see if the rhythms in psi knock down flies are slightly phase-advanced? Or use dissected peripheral tissues in hope of stronger rhythms in DD (whole body rhythms could be further dampened due to internal desynchronization between tissues).

5) Figure 5 and accompanying Results text. I think this is the key figure about the molecular effects of reducing psi function (the effects on cwo are relatively weak and much less convincing). The data in Figure 5 clearly show that tim introns that are usually retained at cold temperatures (due to no splicing at the usual splice sites, I assume) and lead to higher mRNA levels of these particular transcripts, are also elevated at 25°C when psi is being knocked down. So it seems that psi is responsible for splicing out these introns at warmer temperatures. In contrast, an intron normally retained at high temperatures (29°C) and resulting in high tim-M transcript levels is spliced out in psi knockdown heads at 25°C, leading to lower tim-M levels both at 25°C and 29°C. This suggests that for this transcript, psi seems to promote intron retention (so to reduce splicing) at warmer temperatures. How can this be explained by a common molecular mechanism? Also, it doesn't help that the authors refer to an accompanying paper (not available for this reviewer) for the nature of the different tim transcripts and splicing events. Without a map explaining these mainly totally novel alternative tim transcripts it is impossible to follow what is going on. So, a map seems mandatory, also because the reader should not be forced to swap to a different paper in order to understand the current one.

6) Figure 6 supports the idea that psi regulates period by tim splicing. I am bit confused by the genotype description used in the figure and Table 3. In the other parts of the paper tim-gal4 UAS-dicer2 was abbreviated as TD2, but here 'TG4/+;UAS-Dcr2/+ was used. Are these the same flies or constructs on different chromosomes. Please explain. Also, as in Figure 2, the controls for RNAi/+ are missing.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for submitting your article "Drosophila PSI controls circadian period and the phase of circadian behavior under temperature cycle via tim splicing" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by Mani Ramaswami as Reviewing Editor and Ronald Calabrese as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This is a resubmitted version of a previously declined article that has been substantially improved. The work described begins with a genetic screen for RNA-binding and RNA-associated genes involved in circadian clock control. Of 364 genes screened, a total of 43 candidates (12%) appear to alter period length. The unexpectedly high hit rate compared to standard genetic screens is significant and interesting, suggesting that RNA regulation plays a significant role in clock regulation. The authors go on to focus on one of these, PSI, a splicing factor that regulates timeless isoforms. Various observations support a model in which PSI controlled temperature-dependent-splicing regulates the circadian period and determines how the phase of the circadian cycles adapts in response to temperature cycles. However, there remain several ambiguities, missing data, controls and issues that need to be addressed.

Essential revisions:

1) Figure 2C to H. At 30°C there are clear period-shortening effects (panel 2C). But from Table 3 and it's difficult for the reader to extract from these numbers whether the UAS-RNAi, and UAS-Psi constructs are significantly different from the experimentals at 25°C. Where are the UAS-Psi RNAi and UAS-Psi overexpression controls for 25°C work? Without these it's difficult to assess how valid are the period changes at 25°C. These should be included. Some of the statements being presented from Figure 2 are possibly incorrect e.g. panel 2D. The authors need to present the 25°C UAS/+ results graphically in Figure 2 and perform the appropriate statistical comparisons. In fact the UAS psi/+ result would actually strengthen their case for period lengthening for panels F and H.

2) Figure 3A and B isn't so convincing visually because there is only one cycle, even though the cosinor method used is appropriate. Might this be relegated to a supplementary figure as it adds very little to the manuscript?

3) Further in Figure 3: It is still very difficult to see the period shortening in the bioluminescence traces (B, D). In particular it is difficult to distinguish between the black circles and black squares (RNAi/+ and test flies, respectively). It would help to at least use colored symbols to help distinguishing the different genotypes.

4) Figure 4 does not add much to this manuscript, where two cwo splice forms appear to be reduced and one is increased. The results are described briefly then immediately dropped – this could be relegated to a supplementary figure.

5) The tim splicing results are very confusing and need to be clarified. The cold tim isoforms (tim-cold and tim-sc) are elevated at warm temperatures in the psi mutant so normally psi would enhance splicing at cold temperatures. The warmer tim isoform (tim-M) is reduced at 25°C in the psi mutant so we'd presume that psi normally enhances this isoform at warm temperatures. Therefore psi normally has opposite effects on the tim isoforms, so a general temperature-sensitive psi mechanism appears to be excluded.

6) Figure 4—figure supplement 1 the UAS RNAi controls are not shown here. Are they temperature compensated? The Td2 control is not temperature compensated and has a relatively large change in period at hot temperature.

7) 'Collectively, these results indicate that PSI shifts the balance toward a warm temperature tim RNA isoform profile at an intermediate temperature (25°C).' This sentence makes no sense – no general mechanism can emerge as PSI appears to have opposite effects on the cold/warm tim isoforms. It enhances the warm forms and reduces the cold forms. Do the authors mean wild-type PSI or the psi-mutant? Please clarify.

https://doi.org/10.7554/eLife.50063.sa1

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

We would like to thank the two reviewers for their insightful comments. We believe that by addressing them with additional experiments and improved discussion, we have very significantly improved our manuscript. The most important additions are the followings. First, we have discovered that PSI downregulation advances the phase of circadian behavior under temperature cycles. This phenotype is remarkably strong and reveal a specific role for PSI in circadian thermal response. Second, we have also significantly strengthened the data supporting a role for PSI in period control.

Reviewer #1:

This is an interesting paper focusing on a splicing factor PSI that regulates timeless isoforms. The screening and experimental work is done well with care taken to exclude background effects. One of the disappointing aspects of the study, and this is no fault of the authors, is the lack of any striking disruption of temperature adaptations with the PSI KD. With all these changes in timeless and cwo splicing, particularly of the temperature sensitive tim isoforms, the locomotor rhythms of the psi KD flies in LD cycles is normal without the kind of dramatic effects seen on siestas when per 3' splicing is disturbed. So I was left scratching my head as to what disruption of psi does phenotypically – small changes in period, sure, but these are not reflected in the LD profiles of Figure 3—figure supplement 2.

As the period seems to change with psi manipulation, I was surprised that the authors did not perform a simple temperature compensation experiment (or did I miss it?). I realize that gal4 is temperature sensitive but the direction of any period change between 18-30°C might have been interesting. Are there compensatory changes of tim transcripts in psi KD? The accompanying paper hinted at that with the cold-sensitive isoforms.

We thank the reviewer for his interest in our work. We did perform a temperature compensation experiment but did not observe a phenotype. We have added these results to the manuscript. However, we have now also performed temperature entrainment experiments and observed a striking phenotype.The phase of circadian behavior is advanced by several hours under a temperature cycle. It is however not advanced under a light/dark cycle. Moreover, in flies that cannot splice tim in a temperature dependent manner, the phase of circadian behavior is not advanced. Therefore, PSI specifically regulates the phase of circadian behavior in the presence of a temperature cycle, through regulation of tim splicing.

The loss of rhythmicity in Figure 2E and 2G on psi overexpression is striking. Any idea what happens to tim isoforms under these conditions? Clearly something in the dorsal neurons is having a major effect on the rhythmicity of PDF neurons. This is interesting as the DNs do not by themselves generate locomotor rhythms. Surprisingly, the Discussion does not mention this most dramatic result of the manuscript.

Reviewer 2 also showed interest in this phenotype, which we tried to attribute to a specific group of circadian neurons. Unfortunately, the results are not straightforward, and suggest that arrhythmicity is the sum of PSI downregulation in multiple groups of circadian neurons. Also, as we discussed in the original manuscript, arrhythmicity is more difficult to interpret than a period phenotype. We would thus prefer not to include these data in the present manuscript, particularly since we have now identified a striking temperature-dependent phenotype. However, if the editors and reviewers concur that we should add these data, we will provide them.

Reviewer #2:

[…] The paper is well written and experiments conducted are of high quality and according to the standard in the field. The authors focus on one gene identified in their screen (psi), which they show causes period shortening when down-regulated and arrhythmicity when overexpressed in the entire clock circuit. Effects with more narrow down regulation (Pdf-gal4) are much less pronounced and overall it remains unclear in which parts of the circuit Psi plays a role. Moreover important controls are lacking in the behavior experiments (outlined below). Because the effects on period are rather mild, these controls are important. Also in light of the fact that psi has no measurable effect on the light PRC or normal LD entrainment it seems important to really nail the effect on period. Molecularly it seems clear that psi affects tim-splicing, but the mechanism remains elusive and it is not clear how psi can promote intron retention in certain tim transcripts at cooler temperature and that in other tim transcripts at warmer temperatures.

1) The period shortenings after Psi-knockdown are relatively mild (0.2-1.2 hr depending on the driver used) and the same applies for the lengthening observed after psi-overexpression (0.1 – 0.7 hr) (Figure 2 and Table 3). Given that these period changes are key to the message of the paper (psi controls circadian period), I think it is important to perform a statistical analysis comparing the various genotypes. Also, some key controls are missing: the psi RNAi lines have not been tested without a driver. Data are shown for the total of RNAi-lines from a given collection over + in Figure 1A, but due to the large variability between the different collections with regard to period length, the values for the individual RNAi lines 'over +' must be given. I also checked the source file (Dataset 1) and found some data, which I think shows the RNAi/+ data. From these data it looks the effect of the psi knockdown is even smaller (psiRNAi-kk/+: 24.0, TD2>kk: 23.6, PD2>kk:24.7 and psiRNAi-GD/+ 24.2, TD2>GD: 23.8, PD2>GD: 24.2). So no effect with the Pdf driver on period at all, and a small 0.4 hr effect with TD2. Moreover, no effects with 2 other RNAi lines shown in this Data set with either tim or Pdf drivers. I am not sure if the authors have more control data and if the Data set table represents only the actual screening data, but clearly additional controls are required. This is an important point, because no other behavioral phenotypes after psi knockdown could be observed (LD behavior at different temperatures and light-PRC, Figure 3—figure supplements 2 and 3).

Indeed, at first sight, it might appear that the PSI RNAi phenotypes are quite modest. Because the tim-GAL4 and pdf-GAL4 drivers cause a ca. 1hr period lengthening on their own (this has been observed since these drivers were first reported in the late 90s by the Hall lab), the only meaningful period comparison is between driver/+ flies and drivers/RNAi flies, which is what we showed on the figures. Importantly PSI came out of a screen of over 600 RNAi lines and was one of the rare genes to show a short period phenotype when downregulated, so it is clear that a 1.5 hour period shortening compared to driver control is highly significant, and specific. What the RNAi/+ control in dataset1 show is that the RNAi transgenes, on their own, do not shorten period. This is now explained more thoroughly in the manuscript. We would like to mention that in the late stages of preparing our original manuscript, a portion of the text that contained much of these explanations was inadvertently deleted. We apologize for this oversight, which probably contributed to the concerns of the reviewer.

We were however able to further support our claim that PSI regulates period. When we monitored circadian behavior at 30°C instead of 25°C, we observed that the circadian period of the timGAL4/+ controls was not significantly different from the RNAi/+ controls. We could therefore meaningfully compare both RNAi/+ and timGAL4/+ controls to the experimental flies. The results are clear, period is shorter in the experimental flies than in the controls. This was added to Figure 2C.

We also considerably increased the number of PSI overexpressing flies tested, and confirmed that period is indeed long in these flies.

Thus, we have clearly established that PSI levels are critically important for the period length of circadian behavior.

2) The authors see stronger effects with tim-gal4 (both knock-down and overexpression of psi) compared to Pdf-gal4. They aim to distinguish between potentially stronger tim-gal4 expression in the s-LNv, or a period-determining function of non-s-LNv neurons by applying a tim-gal4/Pdf-gal80 combination, which should eliminate gal4 from the s-LNv but not all the other clock cells. Unfortunately, this experiment did not allow for a clear distinction, as the period shortening after knockdown was in between that of tim and Pdf drivers alone. Overexpression with the tim-gal4/Pdf-gal80 combination recapitulated the high percentage of arhythmicity seen with tim alone, but not the period-lengthening, whereas overexpression with Pdf-gal4 only had no effect. Overall the results suggest a role for non-s-LNv neurons in setting period length and controlling rhythmicity, but I think this should be repeated with other, non-s-LNv drivers to get a clearer picture (e.g., it cannot be ruled out that Pdf-gal80 is completely blocking gal4 in the s-LNv). I suggest using the splitE cell-gal4 and drivers specific for the DN1 to solve this issue.

As described above in response to one of Reviewer #2’s comments, we tried to map arrhythmicity to specific neurons, but the results are not straightforward, and an arrhythmic phenotype is more difficult to interpret than a change in period We would thus prefer to leave these data out of the manuscript to keep the focus on the most important and solid results, but if the editors and reviewers consider it necessary for us to add these data, we will do so.

3) psi overexpression with tim-gal4 and tim-gal4/Pdf-gal80 produces very high levels of arrhythmicity and the few rhythmic flies (n=5 and n=4, respectively) have low average power values. Considering the low number of rhythmic flies and the low power values, I find it problematical to claim that overexpression with tim-gal4 causes period-lengthening. Could the authors show actograms that clearly show the longer period in these flies compared to the tim-gal4/Pdf-gal80 flies. Lacking additional support (such as convincing actograms and/or higher n's), I think it is not OK to conclude that the overexpression results are in line with the knockdown results (which are problematical in itself, see major point 1 above).

We agree that the number of rhythmic flies was low. We have repeated these experiments and considerably increased the Ns. We confirmed our initial results.

4) The peripheral clock data in Figure 3 do not look very convincing, due to the poor rhythmicity of the luciferase oscillations in DD. As correctly mentioned in the Materials and methods part, it is expected and was reported previously that rhythms of these reporters dampen rapidly in DD. But I am not sure if it is valid to calculate period values from 48 hr only, particularly if one the 2nd day the oscillations are very low amplitude. Perhaps it would be better to look in LD (where the reporters cycle with high-amplitude) to see if the rhythms in psi knock down flies are slightly phase-advanced? Or use dissected peripheral tissues in hope of stronger rhythms in DD (whole body rhythms could be further dampened due to internal desynchronization between tissues).

We have not been very successful with dissected organs in our lab in DD. In LD, the effect would be expected to be minimal based on behavior (see Figure 3—figure supplement 2), and we think it would be unlikely to observe a significant effect. However, we are confident that the results presented are solid. Phase changes are clearly visible on the graphs, and period was reproducibly shorter in experimental flies in all experiments, even though for one reporter we were slightly above a P value of 0.05. In addition, the period shortening observed with luciferase is very similar to that observed with locomotor behavior.

5) Figure 5 and accompanying Results text. I think this is the key figure about the molecular effects of reducing psi function (the effects on cwo are relatively weak and much less convincing). The data in Figure 5 clearly show that tim introns that are usually retained at cold temperatures (due to no splicing at the usual splice sites, I assume) and lead to higher mRNA levels of these particular transcripts, are also elevated at 25°C when psi is being knocked down. So it seems that psi is responsible for splicing out these introns at warmer temperatures. In contrast, an intron normally retained at high temperatures (29°C) and resulting in high tim-M transcript levels is spliced out in psi knockdown heads at 25°C, leading to lower tim-M levels both at 25°C and 29°C. This suggests that for this transcript, psi seems to promote intron retention (so to reduce splicing) at warmer temperatures. How can this be explained by a common molecular mechanism? Also, it doesn't help that the authors refer to an accompanying paper (not available for this reviewer) for the nature of the different tim transcripts and splicing events. Without a map explaining these mainly totally novel alternative tim transcripts it is impossible to follow what is going on. So, a map seems mandatory, also because the reader should not be forced to swap to a different paper in order to understand the current one.

We regret that the reviewer could not access the accompanying paper. However, we agree that a map of tim splicing should be added to our manuscript and we have done so. In terms of mechanisms for differential effects on different splicing events, it is possible that one splicing event regulates the probability of the other. We are now mentioning this possibility in the Discussion.

6) Figure 6 supports the idea that psi regulates period by tim splicing. I am bit confused by the genotype description used in the figure and Table 3. In the other parts of the paper tim-gal4 UAS-dicer2 was abbreviated as TD2, but here 'TG4/+;UAS-Dcr2/+ was used. Are these the same flies or constructs on different chromosomes. Please explain. Also, as in Figure 2, the controls for RNAi/+ are missing.

Yes, we used a UAS-Dcr2 transgene on a different chromosome, because we needed to be in a tim0background for the experiments shown on Figure 7. tim is on the second chromosome, so to build the flies it was easier to use a 3rd chromosome UAS-Dcr2 insertion. This is more clearly explained in the figure legend.

[Editors' note: the author responses to the re-review follow.]

Essential revisions:

1) Figure 2C to H. At 30°C there are clear period-shortening effects (panel 2C). But from Table 3 and it's difficult for the reader to extract from these numbers whether the UAS-RNAi, and UAS-Psi constructs are significantly different from the experimentals at 25°C. Where are the UAS-Psi RNAi and UAS-Psi overexpression controls for 25°C work? Without these it's difficult to assess how valid are the period changes at 25°C. These should be included. Some of the statements being presented from Figure 2 are possibly incorrect e.g. panel 2D. The authors need to present the 25°C UAS/+ results graphically in Figure 2 and perform the appropriate statistical comparisons. In fact the UAS psi/+ result would actually strengthen their case for period lengthening for panels F and H.

We have added the PSI RNAi/+ controls to Figure 2A. They show that the RNAi lines, on their own, do not shorten period, compared to wild-type flies. In addition, the experimental flies are statistically shorter than these controls. However, the difference is smaller than against tim-GAL4, Uas-dicer2 (TD2)/+ controls, because tim-GAL4 causes a well-known ca. 0.8 hr period lengthening at 25°C. Because this period lengthening is dominant, the key comparison to understand the impact of PSI downregulation on circadian behavior is TD2/+ vs. TD2/Psi RNAi. This is the case for the PD2 combination as well, because pdf-GAL4 similarly lengthens period in a dominant manner. The effect on period is weaker when RNAi expression is restricted to PDF cells with PD2, or to non-PDF cells with TD2/-pdf-GAL80. As a result, while the experimental flies are statistically significantly shorter than PD2/+ or TD2/+; pdfGAL80/+ control flies, they are not shorter than PSI-RNAI/+ controls. This does not invalidate our conclusions that both PDF and non-PDF cells are implicated in the PSI phenotype. The presence of the RNAi transgenes, on their own, has no effect on period (new Figure 2A), and the comparison that has to be done, to evaluate the contribution of PSI-RNAi expression, is experimental flies vs. the driver controls (PD2/+ or TD2/+; pdfGAL80/+) because of their dominant effect on period. To make sure that the figure is not confusing to the reader, we only showed the key genotype comparisons in Figure 2B and C, but in the figure legend and the main text we made every effort to make clear to the reader why these are the key comparisons, and that the RNAi on their own do not impact period.

We have now also included the UAS-PSI/+ controls in Figure 2F-H. As expected, experimental flies are statistically longer than these control flies, but again, the key comparison is actually experimental flies vs. GAL4 driver controls.

2) Figure 3A and B isn't so convincing visually because there is only one cycle, even though the cosinor method used is appropriate. Might this be relegated to a supplementary figure as it adds very little to the manuscript?

As suggested, we have now moved this figure to the supplementary figures (Figure 2—figure supplement 2A, B). Please not however, that there are two cycles on the figures, not just one.

3) Further in Figure 3: It is still very difficult to see the period shortening in the bioluminescence traces (B, D). In particular it is difficult to distinguish between the black circles and black squares (RNAi/+ and test flies, respectively). It would help to at least use colored symbols to help distinguishing the different genotypes.

Thank you for this suggestion, we have improved the presentation of this figure.

4) Figure 4 does not add much to this manuscript, where two cwo splice forms appear to be reduced and one is increased. The results are described briefly then immediately dropped – this could be relegated to a supplementary figure.

As suggested, we have moved this figure to the supplementary figures (Figure 3—figure supplement 1).

5) The tim splicing results are very confusing and need to be clarified. The cold tim isoforms (tim-cold and tim sc) are elevated at warm temperatures in the psi mutant so normally psi would enhance splicing at cold temperatures. The warmer tim isoform (tim-M) is reduced at 25°C in the psi mutant so we'd presume that psi normally enhances this isoform at warm temperatures. Therefore psi normally has opposite effects on the tim isoforms, so a general temperature-sensitive psi mechanism appears to be excluded.

We have made additional efforts in the text (Results and Discussion) to make clear the impact of PSI on tim splicing. Briefly, PSI promotes splicing favored at warm temperature, while inhibiting those favored at cold temperature, and it does this at any temperature. As we discussed, while PSI determines the ratio of tim isoforms, their temperature sensitivity might be encoded through splice site strength, as for per 3’UTR intron retention event. In the Discussion, we proposed that splicing events might be co-regulated, and that therefore favoring one can, at the same time, indirectly influence the probability of another (excerpt from the previous Discussion: “Another interesting question is how PSI affects differentially specific splice tim isoforms. One possibility is that the execution of a specific tim splicing event influences that of another”). We have expended this point of Discussion to make it clearer. We believe that with this addition and a few minor text corrections in the relevant Results section, the impact of PSI on tim splicing should be entirely clear to the reader.

6) Figure 4—figure supplement 1 the UAS RNAi controls are not shown here. Are they temperature compensated? The Td2 control is not temperature compensated and has a relatively large change in period at hot temperature.

They are temperature compensated. We had left them out because of the dominant impact of TD2, but we have now added them.

7) 'Collectively, these results indicate that PSI shifts the balance toward a warm temperature tim RNA isoform profile at an intermediate temperature (25°C).' This sentence makes no sense – no general mechanism can emerge as PSI appears to have opposite effects on the cold/warm tim isoforms. It enhances the warm forms and reduces the cold forms. Do the authors mean wild-type PSI or the psi-mutant? Please clarify.

Please see our response above on the possibility that a splicing event influences another. Also, we have added “in wild-type flies”. We hope that this sentence now makes sense.

https://doi.org/10.7554/eLife.50063.sa2

Article and author information

Author details

  1. Lauren E Foley

    Department of Neurobiology, University of Massachusetts Medical School, Worcester, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7635-7338
  2. Jinli Ling

    Department of Neurobiology, University of Massachusetts Medical School, Worcester, United States
    Contribution
    Conceptualization, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Radhika Joshi

    Department of Neurobiology, University of Massachusetts Medical School, Worcester, United States
    Contribution
    Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Naveh Evantal

    Hebrew University of Jerusalem, Jerusalem, Israel
    Contribution
    Resources, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Sebastian Kadener

    1. Hebrew University of Jerusalem, Jerusalem, Israel
    2. Brandeis University, Waltham, United States
    Contribution
    Resources, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0080-5987
  6. Patrick Emery

    Department of Neurobiology, University of Massachusetts Medical School, Worcester, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Investigation, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    Patrick.Emery@umassmed.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5176-6565

Funding

National Institute of General Medical Sciences (1R35GM118087)

  • Patrick Emery

National Institute of General Medical Sciences (1R01GM125859)

  • Sebastian Kadener

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are particularly grateful to Vincent van der Vinne for his help with the analysis of the luciferase recording. We also express our gratitude to Monika Chitre for help with qPCR, Pallavi Lamba for help with PRC, Elaine Chang for help with luciferase recordings, and Diana Bilodeau-Wentworth, Dianne Szydlik, Chunyan Yuan and Vinh Phan for technical assistance. We also thank Dr. Donald Rio as well as the Bloomington and Vienna Drosophila Resource Centers for fly stocks. This work was supported by MIRA award 1R35GM118087 from the National Institute of General Medicine Sciences (NIGMS) to PE, and NIGMS grant 1R01GM125859 to SK.

Senior Editor

  1. Ronald L Calabrese, Emory University, United States

Reviewing Editor

  1. Mani Ramaswami, Trinity College Dublin, Ireland

Publication history

  1. Received: July 10, 2019
  2. Accepted: November 7, 2019
  3. Accepted Manuscript published: November 8, 2019 (version 1)
  4. Version of Record published: December 3, 2019 (version 2)

Copyright

© 2019, Foley et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,403
    Page views
  • 229
    Downloads
  • 14
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Lauren E Foley
  2. Jinli Ling
  3. Radhika Joshi
  4. Naveh Evantal
  5. Sebastian Kadener
  6. Patrick Emery
(2019)
Drosophila PSI controls circadian period and the phase of circadian behavior under temperature cycle via tim splicing
eLife 8:e50063.
https://doi.org/10.7554/eLife.50063

Further reading

    1. Neuroscience
    Jiawen Huang, Isabel Velarde ... Christopher Baldassano
    Research Article

    When forming a memory of an experience that is unfolding over time, we can use our schematic knowledge about the world (constructed based on many prior episodes) to predict what will transpire. We developed a novel paradigm to study how the development of a complex schema influences predictive processes during perception and impacts sequential memory. Participants learned to play a novel board game ('4-in-a-row') across six training sessions, and repeatedly performed a memory test in which they watched and recalled sequences of moves from the game. We found that participants gradually became better at remembering sequences from the game as their schema developed, driven by improved accuracy for schema-consistent moves. Eye tracking revealed that increased predictive eye movements during encoding, which were most prevalent in expert players, were associated with better memory. Our results identify prediction as a mechanism by which schematic knowledge can improve episodic memory.

    1. Neuroscience
    Tarek Amer, Lila Davachi
    Review Article

    Pattern separation, or the process by which highly similar stimuli or experiences in memory are represented by non-overlapping neural ensembles, has typically been ascribed to processes supported by the hippocampus. Converging evidence from a wide range of studies, however, suggests that pattern separation is a multistage process supported by a network of brain regions. Based on this evidence, considered together with related findings from the interference resolution literature, we propose the ‘cortico-hippocampal pattern separation’ (CHiPS) framework, which asserts that brain regions involved in cognitive control play a significant role in pattern separation. Particularly, these regions may contribute to pattern separation by (1) resolving interference in sensory regions that project to the hippocampus, thus regulating its cortical input, or (2) directly modulating hippocampal processes in accordance with task demands. Considering recent interest in how hippocampal operations are modulated by goal states likely represented and regulated by extra-hippocampal regions, we argue that pattern separation is similarly supported by neocortical–hippocampal interactions.