Thermosensitive alternative splicing senses and mediates temperature adaptation in Drosophila

Circadian rhythms are generated by the cyclic transcription, translation, and degradation of clock gene products, including timeless (tim), but how the circadian clock senses and adapts to temperature changes is not completely understood. Here, we show that temperature dramatically changes the splicing pattern of tim in Drosophila. We found that at 18°C, TIM levels are low because of the induction of two cold-specific isoforms: tim-cold and tim-short and cold. At 29°C, another isoform, tim-medium, is upregulated. Isoform switching regulates the levels and activity of TIM as each isoform has a specific function. We found that tim-short and cold encodes a protein that rescues the behavioral defects of tim01 mutants, and that flies in which tim-short and cold is abrogated have abnormal locomotor activity. In addition, miRNA-mediated control limits the expression of some of these isoforms. Finally, data that we obtained using minigenes suggest that tim alternative splicing might act as a thermometer for the circadian clock.


INTRODUCTION 42
Circadian rhythms organize most physiological and behavioral processes to 24 hours cycles 43 influence on their degradation rates (Hardin and Panda, 2013;Ozkaya and Rosato, 2012). 59 Modification of PER may additionally influence its transcriptional repressor activity or the 60 timing of its activity (Hardin and Panda, 2013;Ozkaya and Rosato, 2012). 61 Circadian clocks are extraordinarily robust systems; they are able to keep time accurately 62 without timing cues. In addition, and despite their biochemical nature, they are resilient to large 63 variations in environmental conditions. The robustness of the circadian system is likely the 64 result of multiple layers of regulation that assure accurate timekeeping and buffering of 65 stochastic changes into the molecular clockwork. These levels of regulation are physically 66 and/or functionally interconnected and, importantly, extend even beyond the single-cell level. 67 Circadian neurons in the brain are organized in a network that is believed to synchronize the 68 individual neuronal oscillators thereby contributing to a coherent and robust behavioral output 69 higher, expression than the canonical isoform tim-L. This suggests the existence of previously 238 unknown aspects of tim regulation even in canonical conditions. In comparison to the big 239 changes driven by temperature observed in tim alternative splicing, the well-characterized 240 changes in per alternative splicing are very modest (Low et al., 2008) (Figure2D). To assess 241 if the observed changes in tim RNA processing upon temperature changes also happen in the 242 fly brain, we dissected the brains of flies entrained at 18, 25 and 29°C and measured the levels 243 a Holarctic species, is also geographically temperate. Although all species displayed some 258 degree of behavioral adaptation to temperature, we could detect differences between species 259 when analyzing the ratio between light and dark activity in all three temperatures ( Figure 3A). 260 As expected, D.melanogaster displays a temperature-dependent gradient in the day/night 261 activity ratio, with highest ratios at 18°C (Figure 3A  We followed by generating whole-transcript polyA + RNA-seq datasets from the heads of all 272 three species entrained at the 3 different temperatures. When examining the splicing pattern 273 of tim, we could see that all species displayed dynamic changes in tim-cold and tim-M in 274 response to temperature changes, with tim-cold increasing and tim-M decreasing at lower 275 temperatures ( Figure 3B and Figure S6). This suggests that these changes might be important 276 for the behavioral changes shared by all species. However, we only detected tim-sc in 277 D.simulans and D. melanogaster, which had very similar temperature-dependent alternative 278 splicing profiles ( Figure 3B and Figure S6). As mentioned above, this isoform appears only at 279 18°C, strongly suggesting that this isoform is responsible for the adaptation to cold 280 temperatures, which is shared between D.simulans and D.melanogaster. The results 281 displayed above showed a strong correlation among the different fly species between the 282 changing on tim mRNA isoforms and the capacity to adapt to colder temperatures. We also 283 looked at changes in per 3' UTR splicing in the different species between temperatures. 284 D.virilis displayed no changes in per splicing, while the changes in D.Yakuba were mild, in 285 both cases the spliced isoform dominated ( Figure 3B and Figure S6B). 286 287

Cold temperature decreases TIM-L levels by two independent mechanisms 288
As previously described (Majercak et al., 1999), we found that the levels of TIM-L protein are 289 significant downregulated at 18°C compared to 25°C or 29°C. We could not clearly detect or 290 differentiate the other isoforms by Western Blot (data not shown). Nevertheless, TIM-COLD 291 has been reported to be expressed by Boothroyd et al. (2007a). tim-M mRNA is very abundant 292 but we could not see a protein of the predicted size in the western blotts, which strongly 293 suggests that that this isoform is poorly and/or not translated. Hence, expression of this 294 isoform could serve as a way to regulate the amount of tim-L production. This agrees with a 295 recent publication that identified this mRNA isoform as non-coding (Shakhmantsir et al., 2018). 296 We also could not univocally detect TIM-SC. While we observed bands of TIM-SC expected 297 size in the TIM immunoblot, these bands are present at all temperatures (data not shown). We 298 believe that these bands might represent canonical TIM degradation products of similar size 299 than TIM-SC (see below). In addition, the utilized TIM antibody was raised against the whole 300 protein and detects TIM-SC with low efficiency (see below). 301 Interestingly, the lower levels of TIM-L at 18°C are not due to changes in total tim mRNA, as 302 observed in our datasets (Figure S4) and reported by others (Boothroyd et al., 2007b). This 303 suggests that the lower TIM levels are due to post-transcriptional and/or post-translational 304 regulation. miRNA-mediated repression is one of the most common mechanisms of post-305 transcriptional control. Moreover, recent work demonstrated that tim mRNA is regulated by 306 miR-276 (Chen and Rosbash, 2016). We then decided to determine whether tim is regulated 307 by miRNAs in a temperature dependent way. To do so, we tested the genome-wide 308 association of mRNAs with the ARGONAUTE 1 (AGO1, the only miRNA-RISC effector protein 309 in Drosophila (Förstemann, Horwich, Wee, Tomari, & Zamore, 2007) at 18°C, 25°C and 29°C. 310 Briefly, we performed AGO1 immunoprecipitation (IP) from fly heads followed by hybridization 311 to oligonucleotide microarrays. In these datasets, we can only assess overall tim levels (as 312 these microarrays cannot distinguish between tim-L, tim-cold and tim-M). We first confirmed 313 that the overall distribution of AGO1-associated mRNAs (and hence miRNA-mediated 314 regulation) is similar at all three temperatures ( Figure S7). Interestingly, we observe changes 315 in the association of tim to AGO1 ( Figure 4A). Binding of tim to AGO1 is very strong at 18°C 316 while there is very little or no association at 25 or 29°C. This patter is tim-specific and, when 317 examining other core circadian clock components, we could not see such temperature-driven 318 AGO1 association changes. Some are bound at all temperatures (like Clk or vri) and some do 319 not bind at all (like cry, cyc and per; Figure 4A). Importantly, the probes in the oligonucleotide 320 microarray do not allow distinguishing between tim-L, tim-cold and tim-M. To determine 321 whether the association of the different tim transcripts to the miRNA-effector machinery is 322 temperature dependent, we performed isoform-specific qPCRs from newly generated AGO1-323 IP samples. While the mRNA encoding the transcription factor CBT (a known miRNA-324 regulated RNA (Kadener et al., 2009;Lerner et al., 2015)) was strongly bound to AGO1 in all 325 the assayed temperatures, we discovered that the association of tim-L/M, tim-cold and tim-M 326 to AGO1 was significantly higher at 18°C ( Figure 4B). 327 To determine the consequences of AGO1 binding on TIM expression, we generated flies 328 carrying luciferase reporters fused to the different 3' UTRs of tim. All transgenes were targeted 329 and inserted in the same genomic location, and we expressed these UAS-transgenes using 330 the tim-gal4 driver. To minimize the effects of temperature on the GAL4 transgene, we 331 performed the experiment at 25°C. The Luciferase levels of the transgenes carrying the 3' 332 UTRs of tim-L and tim-sc were high, suggesting little (or no) post-transcriptional control, at 333 least at 25°C. Interestingly, the reporter carrying the 3' UTR of tim-cold showed a strong 334 decrease of Luciferase levels compared to the rest of the reporters, suggesting that the 335 presence of this 3' UTR strongly diminishes the translation potential of this isoform either by 336 affecting the stability of the mRNA or, directly, by preventing its translation ( Figure 4C). 337 Similarly, the tim-M 3' UTR reporter displayed lower luciferase levels, although the effect was 338 milder than the one observed with tim-cold 3' UTR. 339 To further understand the post-transcriptional regulation of tim, we analyzed the different 3' 340 UTRs for potential miRNA binding sites using Target Scan (Agarwal, Bell, Nam, & Bartel, 341 2015). Three of the isoforms, tim-L, tim-cold and tim-M contain a large number of miRNA 342 binding sites, many of which are evolutionary conserved and highly expressed in fly heads 343 (Table S3). Tim-sc contains only 5 miRNA binding sites and was not bound to AGO1 at any 344 temperature demonstrating it is not regulated by miRNAs ( Figure 4D). 345 The miRNAs responsible for the association of tim isoforms with AGO1 could target only a 346 temperature-specific isoform, more than one isoform, be expressed in a temperature-347 dependent way or a combination of these. We therefore sequenced AGO1-associated 348 miRNAs from heads of flies entrained to 12:12 LD cycles at 18, 25 and 29°C. While the 349 expression of most miRNAs was not affected by temperature, a few of them were ( Figure 4E, 350 Table S4). We first focus on the miRNAs targeting the 3' UTR region common to tim-L, tim-351 cold and tim-M. Of the 17 miRNAs predicted to target this region, only two change with 352 temperature. One of them (miR-33) is modestly increased at 29°C in comparison with 18C. 353 Notably miR-969, which contains a strong binding site in this 3' UTR, is strongly (~3 fold) 354 In sum, these results, strongly suggest that, at 18°C, changes in the canonical TIM protein are 366 due to the deviation of timeless transcription towards the production of a short isoform (tim-367 sc) and another isoform under strong post-transcriptional regulation (tim-cold). In addition, at 368 this temperature even tim-L is subjected to a stronger post-transcriptional control by miRNAs. 369 Last, tim-M is also regulated post-transcriptionally at all temperatures (similar to tim-cold). 370 371

Overexpression of the different TIM isoforms alters circadian behavior differently 372
We next sought to determine the functionality of the different TIM isoforms. To do so we 373 generated plasmids expressing the different TIM protein isoforms fused to a C-terminal FLAG impossible) to detect this isoform univocally by western blot. We followed by generating flies 382 containing these plasmids by targeting them into the same attB insertion site, in order to obtain 383 flies that can express the proteins at similar levels. We then over-expressed each of these 384 TIM protein isoforms using the tim-gal4 driver and determined the locomotor activity patterns 385 of these flies at 25°C. When kept under a 12:12 LD cycle, we observed a clear advance in the 386 start of the evening activity peak in the tim-sc overexpression flies compared to the tim-Gal4 387 control ( Figure 5A

Elimination of tim-sc results in changes in tim processing and locomotor activity 399
To definitively establish whether tim-sc is functional, we generated flies in which the cleavage 400 and polyadenylation site used for the generation of tim-sc mRNA is mutated by CRISPR (40A 401 flies, Figure 6A). As expected, these flies do not express tim-sc at any temperature (see qPCR 402 in Figure 6B). We followed by assessing the locomotor behavior of 40A mutants and their 403 isogenic controls. Control flies presented two different responses when transferred to 18°C: 404 an advance of the evening peak as well as a decreased night activity in comparison to the 405 flies maintained at 25°C ( Figure 6C, 6D and 6E). In addition, the activity patterns of the 40A 406 flies are strikingly similar at the different temperatures. At 18 and 25°C, 40A mutants display 407 lower activity than control flies only during the night while at 29°C, these 40A flies were less 408 active both in the light and dark periods ( Figures 6D and S9). Even more importantly, and 409 opposite to TIM-SC overexpression, 40A mutants display a significant delay in the time of 410 evening activity onset specially at 18°C but also more mildly at 25°C (Figures 6C and 6E). 411 Finally, we measured the levels of the other tim isoforms in control and 40A flies at 18, 25 and 412 29°C in order to assess the molecular consequences of the disruption of tim-s production.  Figure 6E). tim-L/M levels also are increased in the mutant, but 417 only at 25°C. 418 These results strongly suggest that both TIM-SC protein and tim-sc production can regulate 419 the daily pattern of locomotor activity and the response to temperature changes. Importantly, 420 we still observe some degree of temperature adaptation in 40A mutants, which we postulate 421 are mediated by the increased production of tim-cold. 422 423

Tim alternative splicing is regulated directly by temperature in-vitro and in-vivo 424
To get insights into the mechanism by which temperature regulates tim alternative splicing, 425 we determined if tim 01 and per 01 mutants also change tim alternative splicing in a temperature-426 dependent way. Hence, we entrained control and both mutant lines in LD conditions at 18°C 427 and at 29°C, collected flies every four hours, extracted RNA from fly heads from a mix of all 428 time points and performed qPCR. We observed a similar trend in all three fly lines, namely: 429 an increase in tim-cold and tim-sc and a decrease in tim-M at 18°C ( Figure 7A). This 430 demonstrates that, as previously shown for per temperature-dependent splicing, tim 431 thermosensitive splicing events are independent of the circadian clock. 432 The splicing sites flanking alternatively spliced exons are usually weaker and temperature is 433 known, at least in vitro, to strongly influence the efficiency of splicing. Hence, we determined 434 the strengths of the different splice sites in tim using a publicly available software (Reese et 435 al., 1997). Almost all tim constitutive introns exhibited strong 5' and 3' splice site strengths 436 (above 0.6 scores, Figure S10). Interestingly, the introns retained in tim-M and in tim-cold have 437 strong 5' splice sited and a weak 3' splice site. This suggests that at least part of the 438 mechanism underlying the production of these isoforms is due to the presence of weak 3' 439 splice sites. However, the splice sites of the intron associated with tim-sc bear strong 5' and  So, we transfected these cells with a plasmid driving expression of Clk. We used an inducible 454 vector in order to bypass any transfection variation that might arise due to different 455 temperature conditions (Lerner et al., 2015). After the induction of Clk, we cultured the cells 456 either at 25°C or 18°C for 24 hours. We then collected the cells, extracted RNA and 457 characterized the splicing pattern of tim by qPCR. For each of the alternative isoforms, we 458 measured the ratio between the unspliced and spliced variants and compared these ratios 459 between the two temperatures. Indeed, we completely reproduce the results obtained in vivo, 460 namely: tim-cold and tim-sc were higher at 18°C, while tim-M showed the opposite trend 461 ( Figure 7B). However, we did not observe any temperature-induced changes of a constitutive 462 tim exon (Control, Figure 7B) demonstrating that this effect is specific for the thermosensitive 463

introns. 464
One intriguing possibility is tim splicing per se functioning as a thermometer. This could be 465 accomplished if temperature could, for example, expose binding sites for a specific splicing 466 factor or if the splicing of these introns is temperature dependent. To test this possibility, we 467 In this study we show that temperature dramatically and specifically changes the splicing 494 pattern of the core circadian component timeless. We found that the lower levels of the 495 canonical TIM (TIM-L) protein at 18°C are due to the induction of two cold-specific splicing 496 isoforms (tim-cold and tim-short&cold). Tim-cold encodes a protein very similar to TIM-L but it 497 is under strong post-transcriptional control, as showed by AGO1-IP and in vivo luciferase 498 reporters. Moreover, tim-sc encodes a short TIM isoform which results in an advanced phase 499 of the circadian clock when overexpressed. Interestingly, these changes in tim splicing 500 patterns are conserved across several Drosophila species and correlate well with the capacity 501 of the species to adapt their activity to temperature changes. We then generated flies in which 502 the production of tim-sc is abrogated. These flies display altered patterns of locomotor activity 503 at 18°C and 25°C as well as altered expression of the remaining tim isoforms demonstrating 504 the importance of tim-sc production. Last, we showed that the temperature-dependent 505 changes in tim alternative splicing are independent of the circadian clock. Moreover, we could 506 reproduce them in Drosophila S2 cells either promoting endogenous tim expression using the 507 p-Act-Clk plasmid or utilizing splicing minigenes. The latter results strongly suggest that tim 508 intronic sequences themselves are the temperature sensor for these splicing changes. 509 The fact that despite being the most abundant RNA isoform, we could not detect TIM-M 510 suggests that this RNA is poorly (or not) translated. Alternatively, the protein could be quickly 511 degraded. However, the protein is fairly stable as we could detect large amounts of TIM-M 512 upon overexpression ( Figure S8). Moreover, our luciferase reporter experiments suggest that 513 little protein is produced from this RNA, likely by regulation by miRNAs. It is also possible that 514 this RNA isoform is not even exported from the nucleus and nuclear/cytoplasmic fractionation 515 experiments could be useful to test this possibility. Regarding tim-cold, this isoform is strongly 516 regulated by mIRNAs at all temperatures and was reported (Montelli et al., 2015) to display 517 weaker binding to CRY in yeast. Our results with the tim-sc null mutant provide support for a 518 specific (and even dominant) function of TIM-COLD. We observed that tim-sc mutants have 519 increased levels of tim-cold mRNA at 25°C. At this temperature these flies behave as if they 520 were at lower temperature (low daytime activity and advanced evening peak of activity). 521 Similar to tim-cold, tim-sc seems to work both at the RNA and protein levels. On one hand, 522 the increase in the production of tim-sc at 18°C helps regulate the amount of tim-L and tim-523 cold produced at this temperature ( Figure S8). In addition, overexpression of TIM-SC 524 advances the phase and shortens the period of the circadian clock, suggesting that expression 525 of this isoform might mediate some of the changes seen upon introduction of flies at 18°C. An 526 accompanying manuscript (Foley et al., Submitted) shows that the splicing factor psi regulates 527 tim alternative splicing in the same direction as when transferring the flies to colder 528 temperatures. However, psi expression and activity are not modulated by temperature, 529 suggesting that the temperature is sensed by a different system. 530 While tim-cold and tim-M are strongly post-transcriptionally regulated at all temperatures, the 531 RISC binds also stronger to tim-L at 18°C. Indeed, tim has been reported to be regulated by 532 miRNAs276a (Chen and Rosbash, 2016). Our miRNA-seq experiments suggest that miR-969 533 might be responsible for this temperature-specific regulation. However, it is possible that other 534 miRNAs are involved and/or that other RNA binding proteins enhance the use of some 535 miRNAs in a temperature-dependent way. CRISPR experiments targeting the miR-969 site 536 could help distinguish between these possibilities. The miRNA profiling and AGO-IP 537 experiments also suggest that other mRNAs might be regulated by miRNAs in a temperature-538 dependent way. It will be really interesting to match those datasets in order to understand how 539 the transcriptional and post-transcriptional expression programs are regulated by temperature. 540 Despite the large number of RNAs and miRNAs that are regulated in a temperature-dependent 541 manner, the impact of temperature in alternative splicing seems to be quite restricted. 542 Although we performed only a superficial analysis of the data, only tim, per and Hsf show 543 evident temperature-dependent regulation of splicing (data not shown). Splicing is strongly 544 temperature dependent, at least in vitro. Flies might have mechanisms that make the outcome 545 of splicing temperature independent. This could be achieved by compensatory changes in 546 chromatin structure, RNA pol II elongation rate or RNA editing which is known to be altered 547 by temperature (Buchumenski et al., 2017). In addition, the specificity of the splicing strongly 548 suggests that the changes in tim splicing are unlikely to be due to the expression or activation 549 of a specific splicing factor. This possibility is supported by the findings reported in the 550 accompanying manuscript by Foley et al. showing that while psi regulates tim splicing, this 551 factor on its own cannot explain the temperature sensitivity of these alternative splicing events. 552 Importantly, our results demonstrate that the introns in isolation are still able to sense and 553 respond to the temperature changes (Figure 7). These results suggest that RNA structure 554 likely plays a key role in making this splicing events temperature sensitive. Based on these and previous results we propose the following model in which miRNA-577 mediated control imposes temperature-dependent thresholds for protein expression for the 578 different tim isoforms (Figure 8). At 25°C, tim-L and tim-M are produced. Both are moderately 579 regulated by miRNAs, due to sequences present in their last exon. Tim-cold production is low 580 at 25°C and the miRNA-mediated control is enough to abolish most (or all) protein expression 581 from this isoform at this temperature. In addition, tim-M is target of many additional miRNAs, 582 some of which are not differentially regulated by temperature, while others are higher at 29°C 583 or 18°C. Therefore, we predict that no TIM-M is produced at any temperature, and the 584 production of this isoform has a regulatory role (as recently suggested by Shakhmantsir et al. 585 (2018)). When the temperature is decreased, the strong increase in tim-cold overcomes 586 miRNA-mediated repression and some TIM-COLD is produced. In addition, tim-sc RNA and 587 protein are produced. We hypothesize that both TIM-COLD and TIM-SC contribute to the 588 phase advance and lower night activity observed at 18°C. TIM-SC lacks the cytoplasmic 589 retention domain and increased levels of this protein might lead to the phase advance 590 observed at 18°C. At 29°C, Tim-M is increased but TIM amounts are maintained constant 591 likely due to higher production rates of TIM (due to a more modest association of tim-L with 592 AGO1). In this way, tim co and post-transcriptional regulation can modulate the amounts and 593 type of TIM proteins to facilitate temperature adaptation. It is possible that these changes in 594 alternative splicing could regulate other situations in which the circadian clock requires 595 adjustments of TIM levels and/or activity independently of CLK-driven transcription. These 596 could include entrainment by temperature or light as well as cell to cell synchronization. 597 Moreover, as tim is being alternatively spliced, it is even possible that some of these protein 598 isoforms could be used by the cell as temperature sensors. 599 In sum, in this study we determine that tim is extensively regulated by alternative splicing in a 600 temperature dependent way. Moreover, we show that this regulation is, at least partially, 601 responsible for temperature adaptation. We propose that this complex regulation of alternative 602 splicing and miRNA-mediated control provides a mechanism for altering the relationship 603 between TIM and PER proteins. This would allow the modulation of the circadian phase 604 without affecting the period (which is strongly dependent on CLK and PER protein levels and 605 activity). Last but not least, our data suggest that tim alternative splicing might act as a 606 thermometer for the cell and might regulate temperature responses not related to the circadian 607 clock. Transgenic lines for the overexpression of different FLAG-tagged tim isoforms or luciferase 667 fused to their 3'UTRs were generated by injecting the plasmids (described below) in a site-668 specific manner into the pattP2 site using the PhiC31 integrase-mediated transgenesis system 669 (Best Gene Drosophila Embryo Injection Services). These transgenic flies were crossed to 670 tim-Gal4 driver (Blau and Young, 1999; Kaneko and Hall, 2000). 671 All crosses were performed and raised at 25°C. Newborn adults were either maintained at 672 25°C or switched to colder (18°C) or warmer (29°C) temperatures as described in the text.  independent replicates (n=4) and the differential gene expression analysis was done using a 786 negative binomial model using DeSeq2 package on R. A miRNA with p adjusted value less 787 than 0.05 and an absolute log2(fold change) more than 1 was considered differentially 788

expressed. 789
For the DGE data, the circadian analysis was performed using the package MetaCycle (Wu 790 et al., 2016). For each temperature and circadian timepoint, two replicates were analyzed 791 (n=2). To normalize over different library preparation, after normalizing by library size the 792 counts were divided by the maximum in each replicate. Genes with more than two cero counts 793 in any timepoint was discarded from further analysis. The amplitude for each replicate was 794 then calculated as the maximum divided the minimum for each gene. JTK algorithm was used 795 for the circadian analysis. A gene was considered as cycling if the JTK pvalue was less than 796 0.05 and the amplitude was more than 1.5. For these genes, the phase shift was calculated 797 as the phase at 25ºC minus the phase at 18ºC or 29ºC. 798 799 TargetScan analysis of putative miRNA binding to each tim 3'UTR 800 As not all of the 3'UTRs for the different tim isoforms are annotated, TargetScanFly version 801 6.1 was run locally (which allows manual entry of the sequences of interest). UTR sequences 802 were downloaded from UCSC dm6 27way conservation. This way we identified several 803 putative miRNAs binding to each of the isoforms. We continued by manually looking at the 804 abundance of this miRNAs (from the AGO1-IP followed by smallRNA sequencing) in order to 805 determine which of those putative miRNAs are being expressed in the fly heads. Additionally, 806 we defined as miRNAs that change in a temperature-dependent manner those miRNAs that 807 had more than 2-fold change between 18 and 29°C and in which this difference was 808 statistically significant (pval<0.05). 809 810

Real Time PCR analysis 814
Total RNA was extracted from adult fly heads (or brains) at the mentioned timepoints using 815 TRI Reagent (Sigma) and treated with DNase I (NEB) following the manufacturer's protocol. 816 cDNA was synthesized from this RNA (using iScript and oligodT primers, Bio-Rad) and diluted 817 1:60 prior to performing the quantitative real-time PCR using SYBR green (Bio-Rad) in a 818 C1000 Thermal Cycler Bio-Rad. Primers used for amplifying each isoform were: tim-sc (5'-   The presented data includes the aggregated data from the 3'-seq datasets presented in Figure  1111 1 (upper traces) as well as full transcript polyA + RNA seq datasets (lower traces). The latter 1112 includes 2 timepoints at 25°C and three timepoints at 18°C and 29°C (see text for details). The 1113 arrows indicate the alternative splicing events that are regulated by temperature. B. A scheme 1114 of the alternatively spliced tim isoforms. In grey are constitutive exons, in red sequences found 1115 mainly at high temperatures and in blue, sequences found mainly at low temperatures. A zoom 1116 on the exons surrounding each non-canonical isoform is represented in the rectangle.

Figure 3
A.

C.
Luciferase (counts)                  Figure S10 Table S1. Circadian oscillation analysis. The circadian analysis was performed using the MetaCycle algorithm (see main text for details). In the table the p-value, phase and amplitude are reported for each gene along with its normalized expression at each circadian timepoint.