How does temperature affect splicing events? Isoform switching of splicing factors regulates splicing of LATE ELONGATED HYPOCOTYL (LHY)

Abstract One of the ways in which plants can respond to temperature is via alternative splicing (AS). Previous work showed that temperature changes affected the splicing of several circadian clock gene transcripts. Here, we investigated the role of RNA‐binding splicing factors (SFs) in temperature‐sensitive AS of the clock gene LATE ELONGATED HYPOCOTYL (LHY). We characterized, in wild type plants, temperature‐associated isoform switching and expression patterns for SF transcripts from a high‐resolution temperature and time series RNA‐seq experiment. In addition, we employed quantitative RT‐PCR of SF mutant plants to explore the role of the SFs in cooling‐associated AS of LHY. We show that the splicing and expression of several SFs responds sufficiently, rapidly, and sensitively to temperature changes to contribute to the splicing of the 5′UTR of LHY. Moreover, the choice of splice site in LHY was altered in some SF mutants. The splicing of the 5′UTR region of LHY has characteristics of a molecular thermostat, where the ratio of transcript isoforms is sensitive to temperature changes as modest as 2 °C and is scalable over a wide dynamic range of temperature. Our work provides novel insight into SF‐mediated coupling of the perception of temperature to post‐transcriptional regulation of the clock.

Circadian clocks play crucial roles in regulating physiology and behaviour by anticipating environmental changes, principally predictable alterations in light:dark, and concomitant changes in temperature.
The framework of the plant circadian clock consists of interlocking gene expression feedback loops (Harmer, 2009;Hsu & Harmer, 2014;Millar, 2016;Nagel & Kay, 2012;Pruneda-Paz & Kay, 2010), similar in concept to the feedback loops in other eukaryotic clocks (Dunlap, 1999). The central loop of the plant transcriptional core oscillator features a set of two dawn-expressed, closely related, and partially redundant Myb transcription factors, CIRCADIAN CLOCK ASSOCIATED 1 (CCA1), and LATE ELONGATED HYPOCOTYL (LHY) (Alabadi, Yanovsky, Mas, Harmer, & Kay, 2002;Mizoguchi et al., 2002). Temperature-dependent AS of LHY-in particular cooling associated retention of the 5′UTR intron 1 (I1R, event UAS4 in James, Syed, Bordage, et al., 2012) and the inclusion of exon 5a (event AS5 in James, Syed, Bordage, et al., 2012)-contribute to a notable decline in LHY transcript abundance (James, Syed, Bordage, et al., 2012). Mutations in the SICKLE (SIC) gene, which encodes a nuclear protein implicated in the control of AS, markedly stimulate the accumulation of splice variants of LHY and other clock genes at cool temperatures (Marshall, Tartaglio, Duarte, & Harmon, 2016). Moreover, sic mutants and mutants in other splicing-related genes, including for the spliceosomal components GEMIN2 and SKIP, affect the period of the circadian clock (Hernando, Sanchez, Mancini, & Yanovsky, 2015;Jones et al., 2012;Marshall et al., 2016;Sanchez et al., 2010;Schlaen et al., 2015;Wang et al., 2012). Notably, sic, skip, and gemin2 mutants are also impaired in temperature compensation, a defining feature of circadian clocks whereby the pace of the clock is largely unaffected across a range of physiologically relevant temperatures (Pittendrigh, 1960).
Here, we demonstrate that PTB and U2AF65A transcripts undergo cold-induced AS isoform switching (i.e., reversals in abundance of two isoforms) such that the balance between functional and non-functional transcripts is temperature dependent. Experiments with mutant lines suggest that PTB1-U2AF65A-SUA represent part of a network involved in the perception and transduction of prevailing temperature fluctuations to the clock via the splicing of the 5′UTR region of LHY. This splicing factor-LHY module is sensitive to temperature changes as modest as 2°C and is scalable over a wide dynamic range of temperature.
We therefore identify some elements of the machinery that links the perception of temperature to the circadian clock and to temperature-dependent outputs of the clock.
RT-PCR was performed using cDNAs and GoTaq Green DNA polymerase (Promega). Products were co-electrophoresed on 1.5% agarose, 0.5 × TBE gels with 100 bp markers (Roche Diagnostics) and stained with SYBR safe DNA gel stain (ThermoFisher Scientific). All primer sequences are provided in Table S1. Schematics of regions amplified are detailed in Note S2.

| High resolution RT-PCR determination of relative isoform levels
High resolution (HR) RT-PCR was carried out as described previ- overlapping RT-PCR primer pairs (each with one 6-FAM fluorescently labelled primer, see Table S1) were designed to cover the relevant area of gene sequence. RT-PCR reactions were carried out for 24 cycles as described previously (Simpson et al., 2008), and the products detected on an ABI3730 automatic DNA sequencer along with GeneScan™ 1200 LIZ size standard.
Amplicons were accurately sized and mean peak areas calculated using GeneMapper software (Applied Biosystems, Life Technologies). To measure changes in AS of particular isoforms, the peak areas were normalized relative to the peak area values for two reference genes, UBC21 (At5g25760) and PP2A (PP2AA3; At1g13320).

| Analysis and presentation of transcript abundance data
Gene and transcript models for PTB1, PTB2, U2AF65A, and SUA reported in AtRTD2  are shown in Figure S2. The functional and the main non-functional premature termination codon-containing (PTC + ) PTB transcripts are referred to as fully spliced (FS) and alternative exon (AE), respectively. There are two PTB2 AE isoforms differing in the length of the AE by only three nucleotides (P2 and P4). Two U2AF65A transcripts contain a spliced exitron (ID3 and JC2). Two SUA transcripts with full coding potential differ by an intron retention event in the 5′UTR. Figure S2 shows details of the individual highly abundant transcripts (low abundance transcripts were removed), whereas other figures have aggregated some transcripts with similar properties (e.g., PTC + ) for simplicity, as described in the figure legends.
The PTB AE transcripts are targeted for degradation by NMD so their levels detected by qPCR or RNA-seq underestimate their relative abundance in newly synthesized transcripts prior to NMD. We, therefore, assessed levels of NMD-degraded transcripts using the strong NMD-impaired mutant line upf1-3 also carrying the pad4-1 mutation, which overcomes the lethality of the upf1-3 mutation (Riehs-Kearnan, Gloggnitzer, Dekrout, Jonak, & Riha, 2012; Figure S3a). Measured levels of PTB AE transcripts were compensated for NMD by applying the fold increase in abundance in upf1-3 pad4-1 to that in Col-0 (3.5-fold for PTB1 and fivefold for PTB2). We present PTB transcript data as FS, AE (i.e., not compensated for NMD), and cAE (compensated AE). The LHY transcripts from which intron 1 has been excised are referred to as FS; those retaining the intron are referred to I1R.
Using the upf1-3 pad4-1 line, we found that I1R transcript levels appear to be relatively insensitive to NMD ( Figure S3b) so they are presented without adjustment.
In some figures, LHY data is presented as the splice ratio (SpR), defined as FS/(FS + I1R). The SpR at dawn at 20°C was estimated to be 0.9 from our previous work (James, Syed, Bordage, et al., 2012). PTB data are shown as SpR or cSpR (SpR after compensation for NMD), defined as FS/(FS + AE) and FS/(FS + cAE), respectively.
The cSpR values for PTB1 and PTB2 at dawn at 20°C were estimated to be 0.5 and 0.33, respectively, using RNA-seq data (Figure 2a,b) with AE abundances compensated for NMD. U2AF65A has five transcripts: The functional U2AF65A transcript is referred to as FS; the other four all introduce PTCs. P2 has retention of intron 11, P3 has an alternative 3′ splice site (Alt3′ss) in exon 11, ID3 has splicing of an exitron, and JC2 has both the Alt 3′ ss and spliced exitron. These PTC + forms have been combined to compare with the protein-coding isoform and the SpR is defined as FS/(FS + aggregated PTC + ).

| Immunoblotting
Immunoblots for LHY were performed as previously described (James et al., 2008). Plant growth conditions were as for gene expression studies (see Section 2.1). In brief, Col-0 plants were grown hydroponically and leaf tissue harvested 5 weeks after sowing (9-13 plants harvested and pooled per condition/time point, and tissue immediately flash frozen in liquid N 2 ). The anti-LHY antibody was raised in rabbits against His-tagged LHY (Kim, Song, Taylor, & Carre, 2003). Two bands specific for LHY protein (James et al., 2008) were quantified using ImageJ software (see legend to Figure S8).

| Temperature-dependent AS of LHY
The 5′UTR of LHY contains two introns separated by a short miniexon of 26 nt (exon 2; Figure 1a). Inspection of the LHY 5′UTR region for potential pY-rich PTB and SUA cis-consensus sequences ("UUCU"; Singh et al., 1995 and "UCUUCU[U/C]C"; Marquez et al., 2015, respectively) showed that intron 2 contains multiple PTB binding sites, whereas intron 1 contains only one, whereas exon 1 contains a region rich in potential SUA binding elements that includes 4 tandem repeats of the sequence "UCUUCUUC" (Figure 1a, Note S1). U2AF65 binding site has not been determined but the RNA recognition motifs 1 and 2 of human U2AF65 bind to variable pyrimidine-rich sequences to allow recognition of degenerate polypyrimidine tracts (Jenkins et al., 2013). Splicing in the LHY 5′UTR region is temperature dependent (James, Syed, Bordage, et al., 2012). Figure  3.2 | Temperature-dependent AS of PTB1, PTB2, U2AF65A, SUA, and LHY Preliminary analysis by RT-PCR showed that cooling rapidly affects AS of both PTB1 and PTB2 transcripts, giving higher levels of the FS and lower levels of the AE isoforms ( Figure S4a). To obtain quantitative data for PTB1, PTB2, SUA, and U2AF65A, we examined transcript isoform-specific expression profiles from an RNA-seq experiment for Arabidopsis shoot tissue harvested across 3 days before, during, and after a low temperature transition (20 to 4°C; Figure S1). For all of the genes under study, the majority of introns are efficiently spliced at both 20°C and 4°C with AS affecting only specific regions of the genes ( Figure S2). The RNA-seq data for PTB1 and PTB2 showed the rapid switching of isoform abundance upon cooling (Figure 2a,b), and qPCR analysis of the same samples for PTB1 isoforms gave a similar picture to the RNA-seq profile ( Figure S4b). Thus, three approaches -high resolution RT-PCR, qPCR, and RNA-seq-all demonstrate the temperature dependence of AS of PTB transcripts.
The response of U2AF65A to cooling is shown in Figure 2c. The was much more abundant than P5, which retains intron 1 in the 5′ UTR (see Figure S2d). P1 declined with cooling, whereas PTC + iso- Because the cooling-induced abundance of LHY I1R transcripts declines after 3 days of exposure to 4°C (Figures 1b and S5), we assessed the effect of cold adaptation on the AS of the splicing factors. Figure S6 shows that the splice ratios of PTB1, PTB2, and with cooling, to the denoted temperature, initiated at dusk and plant tissue harvested at dawn 12 hr later. Middle; PTB1 cSpR and LHY SpR and lower; U2AF65A and SUA transcript isoforms. Individual replicates are plotted (n = 3) for middle plots and means ± SEM (n = 3) for lower plots. Transcript abundances were measure by qPCR. For LHY, the abundances at dawn and 20°C were taken to represent an SpR of 0.9, and for PTB1, they were taken to represent a cSpR of 0.5 (see Section 2); the ratios for other data points were calculated by comparison of the isoform abundances to those at dawn and 20°C. Only the Δ°C where the change first becomes significant is labelled. For U2AF65A and SUA, isoform transcripts were expressed relative to the abundances of the forms at dawn and 20°C. Significant p values (unpaired Student's t test) are reported for the splice ratios 0 versus Δ2°C, *p < .05, **p < .01 (b) LHY transcripts were measured by qPCR and LHY protein by western blotting (see Figure S8). Values are expressed relative to those observed at dawn and 20°C, presented as means ± SD, n = 3, except for LHY protein (mean ± SD, n = 2 U2AF65A all recover partially after this adaptation period. This is consistent with the view that these factors may be involved in the regulation of transient AS of the LHY 5′UTR in response to cooling.

| AS of LHY and splicing factors is sensitive to both small and brief temperature changes
We next assessed the sensitivity of these AS events to different extents or durations of cooling. AS of PTB1 and LHY showed remarkable sensitivity to small temperature changes illustrated by modulation of the relative abundance of individual PTB1 and LHY splice isoforms. For example, PTB1 cAE and LHY I1R levels were sensitive to temperature changes of only 2°C (presented as Δ2°C; Figure 3 a; for clarity, full data with statistical analysis is shown in Figure   S7) for clarity full data with statistical analysis is shown in Figure S9).
Thus, specific AS transcript isoforms coded by splicing factor genes are sufficiently rapid and sensitive to cooling that they could comprise part of a primary response network to reductions in temperature.  Wang & Okamoto, 2009). We, therefore, used an artificial microRNA (ami) knockdown of PTB1 and PTB2, in the Col-0 background, where the knockdown reduces both mRNAs to around 40% of wild-type transcript levels . We found that PTB1 and PTB2 FS isoform levels were reduced to 47% and 68%, respectively, in the knockdown line compared with Col-0 ( Figure   S10). In transiently cooled plants, but not at 20°C, the levels of both LHY FS and I1R transcripts were higher in the amiPTB1&2 knockdown line than in Col-0 (Figure 4a,b). This suggests that the PTBs can affect levels of LHY gene expression, perhaps indirectly, in a temperature-dependent fashion. SUA FS levels were markedly diminished in the amiPTB1&2 line at ambient temperature (Figure 4c, left), but not in transiently cooled plants (Figure 4c, right). We found that U2AF65A

| LHY expression and AS is affected in splicing factor mutants
FS levels were only marginally elevated in the amiPTB1&2 line in both temperature conditions implying that there is either no or only a weak association between U2AF65A and PTB1/2 (Figure 4d).  Figure S11). Overall, the data show that single mutations in SUA and U2AF65A alter the way that AS of the 5′UTR of LHY transcripts responds to temperature.

| PTB1 splicing is regulated by light quantity
Because the circadian clock is sensitive to both temperature and light, we next asked whether light quantity could influence the splicing of PTB1 and LHY transcripts. Plants were exposed to constant light at 12°C for 48 hr in order to allow either increases or decreases in AS to be detected and then shifted to higher or lower light intensities (150 to 300 or 75 μmol.m −2 .s −1 ; Figure 5a). Increasing the light intensity decreased the cSpR for PTB1 (Figure 5b

| DISCUSSION
Recent work has identified temperature-dependent AS of a number of genes in the Arabidopsis circadian clock and emphasized the importance of these AS events in clock function (reviewed by Romanowski and Yanovsky, 2015). Hence, the nature of the signal transduction mechanism(s) through which temperature changes are detected and how specific splicing events are regulated is of great interest. Cooling associated retention of intron 1 in the 5′UTR of LHY is of particular note due to its prevalence with temperature transitions and thus may be synonymous with temperature fluctuations found in nature.
AS of LHY, therefore, provides a read-out of the effects of temperature on the plant circadian clock, and we have sought to investigate the relation between this read-out and the PTB, U2AF65A, and SUA splicing factors.
It is well established that plant PTBs are involved in splicing/AS and are themselves subject to control by AS (Simpson et al., 2014;Stauffer et al., 2010;Wachter et al., 2012). Here, we found that the non-functional AE forms of the Arabidopsis PTBs, which comprise a small proportion of the total transcripts at 20°C, are reduced even further on cooling. In particular, the relative abundances of the PTB1 FS and AE isoforms, after compensation for the effects of NMD, show a dramatic isoform switch on cooling to increase the FS protein-coding isoform after cooling. The functional FS and non-functional PTC + forms of U2AF65A transcripts show a similar pattern.
For these three key splicing factors, it is clear that temperature radically affects the levels of functional mRNAs by AS rather than at SUA and U2AF65A data were normalized to the values at dawn and 20°C. PTB1 FS values were expressed relative to the value in Col-0 at dawn and 20°C, taking this to be 50% of total transcripts. Data are means ± SEM, n = 3; *p < .05, **p < .01, ***p < .001 the level of gene transcription. The splicing of both PTB1 and U2AF65A responded sensitively to cooling, with changes detectable after a transition of only 2°C, from 20 to 18°C (Figure 3). The AS of SUA was also temperature dependent but responded less sensitively to the extent of cooling. In addition, changes in AS of both PTB1 and LHY were detected after only 1 hr of cooling ( Figure 3).

The level of sensitivity exhibited by these changes in AS in
Arabidopsis approaches that observed in mammals, where a 1°C change in body temperature is sufficient to induce a concerted splicing switch and changes in splicing are detectable within 30 min of the onset of the temperature change (Preussner et al., 2017).
We provide several lines of evidence that suggest that these rapid and sensitive changes in the splicing of PTBs, U2AF65A, and SUA contribute to the temperature-dependent splicing of the LHY 5′UTR. First, evidence from mutant lines (Figure 4) shows that the effects of the  Figure   S5). Second, the sensitivity and speed of the changes in the splicing of PTBs and U2AF65a seem sufficient to account for changes in the AS of  (Streitner et al., 2013). Thus, our data provide another example of combinatorial control of splice site selection in AS, which is already well established in other systems (Barash et al., 2010;Lee & Rio, 2015;Smith & Valcarcel, 2000).  James, Syed, Bordage, et al., 2012). It is also possible that temperature-dependent remodelling of the structure of LHY pre-mRNA contributes to the changes in the abundance of LHY I1R on cooling (James, Sullivan, & Nimmo, 2018), but whether this could account for transient effects is less clear.

FIGURE 5
White light intensity consolidates temperature associated AS for PTB1 but not for LHY. (a) Schematic showing experimental regime for examining the effects of light quality on the splice ratios of (b) PTB1 (left) and LHY (right). Plants were subjected to a cool (12°C) LL regime (light and subjective dark, white and grey hatched bars, respectively). White light quantity was adjusted at 60 hr (from 150 to 300 μE or 75 μE). Samples were harvested within the subjective day before and after the changes in light intensity (denoted by * in the schematic). Extended profiles are provided in Figure S12. Transcript expression levels were measured using qPCR and splice ratios calculated as in Figure 3a. Data are means ± SEM, n = 3. p value reports significance, separately for the pre-and post-light shifted data, by one-way analysis of variance, ns; nonsignificant (see Note S4) [Colour figure can be viewed at wileyonlinelibrary.com] We have focussed on the LHY I1R event as a convenient readout in which large changes in the relative abundance of splice variants are observed. The functional effect of this event on LHY protein is not clear but could include effects on translatability (Hinnebusch, Ivanov, & Sonenberg, 2016). The temperature-dependent inclusion of exon 5a introduces a premature termination codon and leads to NMD, and we have already shown that these two splicing changes together contribute significantly to the reduction in the levels of LHY functional transcripts and protein on cooling to 4°C (James, Syed, Bordage, et al., 2012). The defect in temperature compensation in the SICKLE mutant (Marshall et al., 2016) suggests that splicing is involved in this process and hence that the temperature-dependent splicing of LHY may contribute to temperature compensation. In this respect, the changes in LHY 5′UTR splicing with temperature are interesting. The LHY SpR displays a relationship to temperature akin to a molecular thermostat-it is sensitive to modest changes in temperature, as low as 2°C, and is also scalable over a wide dynamic range of temperature, both primary considerations of molecular thermometers (McClung & Davis, 2010).
The circadian clock controls many diverse downstream processes and is itself affected by many environmental factors. Here, we also show that changes in light intensity elicit some changes in splicing that are similar to those induced by cooling. AS is also involved in responses of the clock to other biotic and abiotic stresses (Filichkin et al., 2015) and in other signalling pathways such as responses to nutrient deficiency and to ABA (Nishida, Kakei, Shimada, & Fujiwara, 2017;Zhu et al., 2017). Whether these responses involve the same set of splicing factors or different or overlapping splicing networks involved in responses to temperature remains to be investigated and highlights the need to identify plant splicing regulators and their binding sequences. Hugh G. Nimmo http://orcid.org/0000-0003-1389-7147