SRR1 is essential to repress flowering in non-inductive conditions in Arabidopsis thaliana

Summary SRR1 regulates flowering time in Arabidopsis by integrating photoperiodic and photoperiod-independent signals. By promoting expression of several repressors of FT, SRR1 represses flowering in non-inductive conditions.


Introduction
Due to their sessile lifestyle, plants need to be able to adapt to their local environment. In particular, the transition from a vegetative to a reproductive state is carefully timed to maximize reproductive success. An intricate system of proteins that relay environmental and physiological stimuli forms a network of signalling pathways that converge at a small number of 'floral pathway integrator genes' including FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) (Srikanth and Schmid, 2011). These in turn activate 'floral meristem identity genes' such as APETALA 1 (AP1) and LEAFY to trigger formation of flowers (Abe et al., 2005;Wigge, 2005).
Arabidopsis thaliana is a facultative long-day (LD) plant. Increasing daylength and temperature in spring promote flowering by antagonizing inhibitory effects of FLOWERING LOCUS C (FLC) (Amasino, 2010;Andrés and Coupland, 2012). The photoperiod is sensed in the leaves by an endogenous timekeeper, the circadian clock. The circadian clock consists of transcriptional feedback loops through which clock proteins generate their own 24 h oscillations (McClung, 2011;Staiger et al., 2013). In Arabidopsis, the core clock loop consists of two Myb transcription factors LATE ELONGATED HYPOCOTYL (LHY) and CIRCADIAN CLOCK ASSOCIATED 1 (CCA1) peaking at dawn, and TIMING OF CAB EXPRESSION1 (TOC1) peaking at dusk, that reciprocally repress each other (Wang et al., 1997;Schaffer et al., 1998;Strayer et al., 2000;Alabadi et al., 2002). Interlocked with this core loop is the morning-phased loop comprising PSEUDO-RESPONSE REGULATOR 7 (PRR7) and PRR9 that are connected to CCA1 and LHY (Matsushika et al., 2000;Locke et al., 2006;Zeilinger et al., 2006). Further, an evening-phased loop comprises GIGANTEA (GI), TOC1, and the evening complex components EARLY FLOWERING 3 (ELF3), ELF4, and LUX ARRHYTHMO (Fowler et al., 1999;Kolmos et al., 2009;Pokhilko et al., 2010;Nusinow et al., 2011;Herrero et al., 2012), completing the basic structure of the circadian clock of interlocked central, morning, and evening loops. The circadian clock regulates the key component of the photoperiodic pathway, the zinc finger protein CONSTANS (CO) (Putterill et al., 1995;Suarez-Lopez et al., 2001). The CO mRNA undergoes circadian oscillations with a peak 8-10 h after dawn in non-inductive short days (SDs). As the CO protein is degraded in darkness, it does not accumulate to significant levels in SDs. Under inductive LDs, the CO mRNA peaks 12-16 h after dawn. The clock protein GI and FLAVIN KELCH F BOX 1 (FKF1), an F box ubiquitin ligase, that both peak 10-14 h after dawn in LDs, undergo a light-dependent interaction. The GI-FKF1 complex promotes degradation of the CYCLING DOF FACTORS (CDFs) that repress CO promoter activity (Imaizumi et al., 2005;Sawa et al., 2007;Fornara et al., 2009;Song et al., 2012). In particular, CDF1 has been proven to repress CO promoter activity and also to repress FT activity by binding to the FT promoter in the morning (Song et al., 2012). CDFs are thus important to prevent early-day accumulation of CO as well as FT and, by extension, premature flowering.
Accumulation of CO mRNA in the light phase of LDs allows CO protein to accumulate and stimulate FT transcription in the companion cells. FT protein then moves through the phloem to the shoot apex to induce flower formation (Corbesier et al., 1996;Jaeger and Wigge, 2007;Mathieu et al., 2007).
Photoperiodic flowering can also be triggered in a CO-independent manner, via GI regulation of miR172 processing (Jung et al., 2007). Furthermore, GI can bind directly to FT (Sawa and Kay, 2011).
In addition to LDs, an extended period of cold enables Arabidopsis plants to flower. This vernalization response prevents inappropriate flowering during LDs in the autumn and instead promotes flowering in the spring (Alexandre and Hennig, 2008;Kim et al., 2009). Vernalization leads to down-regulation of the key repressor FLC, a MADS domain transcription factor that binds to the promoters of FT in the leaf and SOC1 in the apical meristem (Helliwell et al., 2006;Searle et al., 2006). Apart from very low temperatures, moderate changes in ambient growth temperature influence floral transition (Blazquez et al., 2003;Balasubramanian et al., 2006;Lee et al., 2007). A temperature rise by 4 °C accelerates flowering in non-inductive SDs to the same extent as extension of the photoperiod to LDs. Photoreceptors are believed to be part of this thermosensory pathway. For example, phyB mutants flower earlier than wild-type plants at 22 °C but not at 16 °C (Halliday et al., 2003). A complex of the MADS domain transcription factor SHORT VEGETATE PHASE (SVP) and a splice isoform of the FLC-related MADS transcription factor FLOWERING LOCUS M (FLM), FLM-β, represses transcription of FT to prevent flowering at colder temperatures (Lee et al., 2013;Pose et al., 2013). At higher temperatures, FLM-β is down-regulated at the expense of the splice isoform FLM-δ that forms a complex with SVP that cannot interact with DNA (Pose et al., 2013). Additionally, SVP is degraded at higher temperatures (Lee et al., 2013). This combined regulation of FLM alternative splicing and SVP protein stability favours flowering at higher temperatures.
Furthermore, the phytohormone gibberellic acid (GA) is required for flowering in SDs, and external application of GA on SD-grown plants results in a LD-like flowering behaviour (Wilson et al., 1992). The action of the GA pathway has long been thought to be largely masked by the photoperiodic pathway under LD conditions (Reeves and Coupland, 2001). More recently, GAs have been assigned a role in floral induction in response to inductive LDs through activation of FT transcription in leaves and of the SQUAMOSA PROMOTER BINDING PROMOTER LIKE genes in the shoot apical meristem (Porri et al., 2012). GA signalling has also been linked to photoperiodic regulation via the TEMPRANILLO (TEM) transcription factors (Osnato et al., 2012). TEM1 binds to a regulatory region in the first exon of the GA biosynthesis gene GA3oxidase 1 (GA3ox1) to repress its expression and can thus control the amount of active GA in the plant and in this way influence flowering (Osnato et al., 2012). Furthermore, the TEM1 and TEM2 proteins can directly repress FT to counteract CO activity, acting redundantly to each other (Castillejo and Pelaz, 2008;Osnato et al., 2012) A recent study combining genome-wide association and quantitative trait loci (QTL) mapping measured flowering time in ecologically realistic environmental conditions (Brachi et al., 2010). In this field experiment, several genes associated with the circadian clock were identified, including SENSITIVITY TO RED LIGHT REDUCED (SRR1). The loss-of-function srr1 mutation has previously been shown to affect multiple outputs of the Arabidopsis circadian clock, including leaf movement rhythms and gene expression (Staiger et al., 2003). The oscillations of morning-and evening-phased output genes as well as components of the core clock show a short period and reduced amplitude in srr1. srr1 also exhibits reduced hypocotyl and petiole elongation in red light, showing that SRR1 is involved in phytochrome B (phyB) signalling. However, as circadian rhythms are affected in light-dark cycles, continuous light, and continuous darkness, SRR1 activity is probably required for normal clock function independently of a function in light input (Staiger et al., 2003). Furthermore, srr1 flowers early in a short photoperiod of 9 h light. SRR1 is a pioneer protein whose sequence is very well conserved among a wide range of species, including mammals. Work performed in yeast showed that the Saccharomyces cerevisiae counterpart BER1 was involved in microtubule stability (Fiechter et al., 2008), but its mode of action in plants is not known.
In this study, the role of SRR1 in flowering time control is characterized. The presented data show that SRR1 regulates expression of CO, FT, and CDF1 in the photoperiodic pathway to inhibit flowering specifically in SDs. Furthermore, SRR1 connects the expression of the TEM1 and TEM2 transcription factors to the circadian clock. In addition, SRR1 can also repress flowering independently of photoperiod, demonstrated by the genetic relationships between SRR1 and CO and GI. SRR1 is thus acting as an integrator between photoperiodic regulation and other pathways to maintain repression of flowering in conditions not suitable for reproduction.

Plant materials and growth conditions
The T-DNA mutant srr1 in the Col-7 background has been described (Staiger et al., 2003) and has now been renamed srr1-1. Additional srr1 alleles from the SALK T-DNA collection (SALK_132099 and SALK_077868) have been characterized here and were named srr1-2 and srr1-3, respectively. co-9 was obtained from D. Weigel (Balasubramanian et al., 2006). gi-2 and phyB-9 were acquired from the NASC stock centre. Mutations were confirmed using PCR, with the primers listed in Supplementary Table S1 available at JXB online.
All seeds were stratified for 3 d at 4 °C before put on soil. Seeds grown on plates were surface-sterilized and stratified for 3 d before they were sown on agar-solidified half-strength MS (Murashige and Skoog) medium (Duchefa) supplemented with 0.5% sucrose and 0.5 g MES l -1 .
For GA treatment, plants growing on soil were sprayed with 100 μM GA 3 in the middle of the light period [Zeitgeber time (ZT) 4-6] once a week starting at day 10 after stratification. Mock treatment was performed by spraying with 0.1% dimethylformamide (DMF)/0.02% Tween-20. For paclobutrazol treatment, plants growing on soil were watered with 5 μM paclobutrazol in the middle of the light period (ZT4-6) once a week starting at day 10 after stratification. Mock treatment was performed by watering with 0.1% DMF/0.02% Tween-20. Vernalization treatment was performed as previously described (Streitner et al., 2008). Plants were grown in Percival incubators AR66-L3 (CLF laboratories) in 150 μmol m -2 s -1 light intensity, with the light-dark and temperature conditions as indicated.
Flowering experiments Seeds were germinated as described above and grown on soil in a randomized fashion in Percival incubators AR66-L3 (CLF Laboratories). For ambient temperature flowering experiments, plants were grown at 16 °C or 20 °C before being shifted to 20 °C or 27 °C, respectively. Flowering time was determined by counting the rosette leaves once the bolt was 0.5 cm tall (Steffen et al., 2014). Mean values ±SD were calculated.

Transcript analysis
Above-ground material, or leaves without the apex and apically enriched material, respectively, was harvested separately, as indicated, and immediately frozen in liquid N 2 . A green safe light was used for sampling in the dark periods. Samples were ground in a bead mill (Retsch MM400, www.retsch.com, last accessed 24 July 2014) using stainless steel beads. Total RNA was extracted using the Universal RNA kit (Roboklon, Berlin, Germany) and reverse transcribed. Quantitative PCR was performed as described (Streitner et al., 2012) using the iTaq kit (Bio-Rad, www.bio-rad.com, last accessed 24 July 2014) on a Bio-Rad CFX-96 Realtime Detection System. C T values were determined and relative expression levels were calculated based on non-equal efficiencies for each primer pair (Pfaffl, 2001). Data were normalized to PP2A (At1g13320) and expressed as the mean expression levels ±SE based on three biological replicates. Primers used are listed in Supplementary Table S1 at JXB online.

SRR1 represses flowering in SDs
As flowering time control in response to the photoperiod depends on the circadian clock measuring the daylength, loss or misregulation of proteins involved in the circadian clock often results in a flowering phenotype (Schaffer et al., 1998;Somers et al., 1998;Wang and Tobin, 1998;Kim et al., 2005). Similarly, the srr1 T-DNA mutant shows impaired circadian rhythms and flowers very early in 9 h light-15 h dark cycles (Staiger et al., 2003).
To characterize the photoperiodic response of srr1 in detail, srr1 and Col-7 wild-type (wt) plants were grown in photoperiods of different length. Flowering was accelerated with increasing daylength, with wt plants forming about half the number of rosette leaves in 12 h light-12 h dark compared with SDs (8 h light-16 h dark) and again forming about half the number of rosette leaves in LDs (16 h light-8 h dark) compared with 12 h light-12 h dark (Supplementary Fig. S1 at JXB online). In contrast, the acceleration of flowering in srr1 in 12 h light-12 h dark compared with SDs was only moderate, and an additional extension by 4 h to LDs resulted in only a small further acceleration. Thus, srr1 responded much more weakly to increasing daylength than the wt.
To obtain independent confirmation of the flowering phenotype, additional T-DNA alleles from the SALK collection were characterized. The position of the T-DNA was confirmed using PCR, and homozygous lines were identified (Supplementary Fig. S2A at JXB online). The line SALK 132099, named srr1-2, has a T-DNA insertion in the promoter region of SRR1, 400 bp upstream of the ATG. SRR1 transcript levels in srr1-2 were reduced to ~60% of the wt levels ( Supplementary Fig. S2B). srr1-2 flowered moderately earlier compared with the wt in SDs ( Supplementary Fig. S2C). Another line, SALK 077868, named srr1-3, has a T-DNA insertion in the 5′-untranslated region, 271 bp upstream of the ATG. SRR1 transcript levels were unchanged and flowering was unaffected compared with the wt (Supplementary Fig. S2B, C). srr1 in the Col-7 background, which does not express SRR1 transcript at all (Staiger et al., 2003) and showed the most pronounced early flowering phenotype, was renamed srr1-1 and used in all subsequent experiments.
To show that the flowering phenotype of the mutant is caused by the loss of SRR1, srr1-1 plants transformed with a construct where the SRR1 coding sequence and a green fluorescent protein tag was expressed from the endogenous SRR1 promoter (Staiger et al., 2003) were assayed for flowering time. Independent transformants displayed wt-like flowering in both SDs and LDs ( Supplementary Fig. S2D, E at JXB online). Thus, SRR1 complements the srr1-1 flowering phenotype.

SRR1 inhibits flowering through regulation of photoperiod components
Since the response to increasing daylength in srr1-1 plants was severely reduced, the functionality of the photoperiodic pathway in srr1-1 was examined. To do this, the srr1-1 mutation was introduced into the co-9 mutant background by crossing. While the co-9 mutation greatly delayed flowering in LDs (52 ± 5.2 leaves), the srr1-1 co-9 double mutant displayed an intermediate phenotype (30.9 ± 4.5 leaves), flowering later than srr1-1 (10.6 ± 1.4 leaves) but earlier than co-9 ( Fig. 1A) (Student's t-test, P<0.01). In SDs, the co-9 mutant flowered in the same way as the wt (64 ± 2.2 leaves versus 63.6 ± 1.8 leaves), while the srr1-1 co-9 double mutant flowered like the srr1-1 single mutant (29.4 ± 3.3 leaves versus 26.8 ± 2.6 leaves) (Fig. 1A). No difference in leaf numbers was observed between SDs and LDs in the double mutant.
To test whether SRR1 affects the CO-independent branch of the photoperiodic pathway, double mutants with gi-2 were also generated. The gi-2 mutation strongly delayed flowering in LDs (68.5 ± 2.2 leaves), while the srr1-1 gi-2 double mutant displayed an intermediate phenotype (42 ± 2.8 leaves), flowering later than srr1-1 but earlier than gi-2 (Fig. 1B). In SDs, the srr1-1 gi-2 double mutant flowered somewhat later than srr1-1 (43.6±2.3 leaves versus 26.8 ± 2.6 leaves), but still earlier than both gi-2 (82.9 ± 2.3 leaves) and the wt (63.6 ± 1.8 leaves) (Student's t-test, P<0.01). Again there was no difference in leaf numbers between SDs and LDs for the srr1-1 gi-2 double mutant. The intermediate flowering phenotypes in srr1-1 co-9 and srr1-1 gi-2 compared with the respective single mutants in LDs suggests that SRR1 has a dual mode of action to repress flowering, partly through the photoperiodic pathway but also in a photoperiod-independent manner. This is supported by the observation that a loss of SRR1 accelerates flowering in the co-9 background in non-inductive SD conditions, where the photoperiodic pathway is not active.
To substantiate this behaviour further, transcript patterns of CO and FT were analysed in wt and srr1-1 plants sampled every 3 h, starting at ZT1 (1 h after lights on), in SDs and LDs. CO levels were elevated in SDs in srr1-1 compared with the wt, at the end of the light phase (ZT7), but not in LDs ( Fig. 2A,  B). Furthermore, FT levels, while low at all time points in the wt in SDs, were strongly elevated at the beginning of the dark phase, with a peak around ZT10 in srr1-1 (Fig. 2C, D). FT thus displayed an LD-like transcript pattern, with the peak at the beginning of the dark period (ZT10 in SDs versus ZT16 in LDs). This strong and early accumulation of FT supports the flowering phenotype under non-inductive photoperiods. In LDs, FT levels were only moderately elevated in srr1-1 compared with the wt. This could explain the moderate early flowering phenotype of srr1-1 in LDs.
CDF1 is an important repressor of CO and FT expression during the morning (Imaizumi et al., 2005;Sawa et al., 2007;Fornara et al., 2009). CDF1 transcript levels were reduced in srr1-1 compared with the wt at the end of the night and throughout the light phase in SDs (Fig. 2E, F). Thus, earlier accumulation of CO correlates with lower CDF1 at a time when CO is repressed in the wt (Fig 2A, E).
GI and FKF1 form a complex that degrades CDF proteins in the second half of the light period, mainly in LDs (Song et al., 2012). GI levels were largely unchanged in srr1-1 compared with the wt in SDs, while a somewhat narrower peak of transcript accumulation could be seen in LDs (Fig. 2G, H). FKF1 transcript levels were reduced in srr1-1 in SDs between ZT1 and ZT7 and in LDs at ZT10 (Fig. 2I, J). A lower peak of FKF1 expression could also be observed in LDs, similar to what could be seen for GI (Fig. 2H, J). Lower levels of FKF1 should hypothetically lead to higher CDF protein levels. However, since CDF degradation via FKF1 occurs later in the day, it may be of little significance for the srr1-1 flowering phenotype in SDs, partly due to the already reduced CDF1 transcript levels.
Thus, in srr1-1, both the reduction in CDF1 peak transcript levels and the early increase in CO levels most probably contribute to the rhythmic FT transcript pattern seen in SDs, by reduced repression and increased activation of FT, respectively.

srr1-1 responds only weakly to vernalization
srr1-1 early flowering correlates with elevated levels of the floral integrator FT that is reciprocally regulated by CO and Fig. 1. Flowering time of srr1-1 co-9 and srr1-1 gi-2 double mutants. Flowering time of Col-7 wt, co-9, srr1-1 and srr1-1 co-9 (A) and Col-7 wt, gi-2, srr1-1 and srr1-1 gi-2 (B) in SDs and LDs. Data represent means of rosette leaves ±SD (n >10). Statistical significance was tested using a two-tailed Student's t-test. Asterisks indicate P-values of <0.01 between the wt and mutant in SDs. Crosses indicate P-values of <0.01 between the wt and mutant in LDs. Experiments were performed three times with similar results. FLC. The FLC transcript level was strongly reduced in srr1-1 compared with the wt (Fig. 3A). It was further reduced by vernalization to levels similar to those in vernalized wt plants. Accordingly, vernalized srr1-1 plants flowered with fewer leaves than untreated plants, but the vernalization response was much weaker than in wt plants (Fig. 3B). Thus, even when the wt and srr1-1 have comparable low levels of FLC, srr1-1 flowers earlier than the wt. Furthermore, the low FLC level in srr1-1 at 20 °C probably limits the effect of the vernalization treatment on flowering time in this mutant.

srr1-1 plants respond to ambient temperature changes
An increase in ambient temperature accelerates flowering in Arabidopsis (Balasubramanian et al., 2006). To examine the behaviour of srr1-1 in different temperatures, plants were grown at 16, 20, and 27 °C. After 4 weeks, a subset of these plants was shifted from 16 °C to 20 °C and another subset from 20 °C to 27 °C. Both wt and srr1-1 plants flowered earlier, with fewer leaves when grown at a constant temperature of 20 °C compared with 16 °C, and earlier when grown at a constant temperature of 27 °C compared with 20 °C ( Fig. 4A-C). srr1-1 consistently flowered earlier than the wt, with about half the number of leaves. A shift from 16 °C to 20 °C or from 20 °C to 27 °C promoted flowering in both wt and srr1-1 plants, compared with the plants kept at constant 16 °C or 20 °C, respectively (Fig. 4A, B, D, E). SRR1 has been implicated in phyB signalling and, because PhyB protein levels in srr1-1 are similar to those in the wt, SRR1 probaby acts downstream of phyB (Staiger et al., 2003). phyB mutants lose their early flowering phenotype when grown at 16 °C (Halliday et al., 2003). To investigate whether SRR1 mediates phyB signals to control flowering, a srr1-1 phyB-9 double mutant was generated and grown at different ambient temperatures. At 16 °C, where mutations in phyB have no effect on flowering, the srr1-1 phyB-9 double Flowering of plants shifted from 16 °C to 20 °C (D) and from 20 °C to 27 °C (E). Temperature-shifted plants were grown at their initial temperature for 4 weeks, before being moved to a higher growth temperature. Data represent means of rosette leaves ±SD (n >10). Statistical significance was tested using a two-tailed Student's t-test. Asteriks indicate P-values of <0.01 between the wt and mutant, or between different genotypes, as indicated by the bars. Experiments were performed three times with similar results. mutant flowered with the same number of leaves as the srr1-1 mutant (Fig. 4A). At 20 °C the effect of the two mutations was additive, with the double mutant flowering earlier than the srr1-1 mutant (Fig. 4B). In plants grown at 27 °C, both the phyB-9 and srr1-1 phyB-9 mutants responded very strongly to the high temperature by flowering with only ~6.7 ± 0.7 leaves and earlier than srr1-1 (9.2 ± 1.4 leaves) (Fig. 4C). The srr1-1 phyB-9 plants shifted from 16 °C to 20 °C or from 20 °C to 27 °C also responded with accelerated flowering, with an additive phenotype compared with srr1-1 (Fig. 4D, E). In conclusion, SRR1 is not affected by lack of phyB at 16 °C, while at 20 °C both proteins contribute to repression of flowering, and at 27 °C SRR1 seems to depend on phyB for its control of flowering.

Regulation of flowering time components by SRR1
To identify downstream targets of SRR1 in flowering time control, the expression of known flowering time genes was compared in srr1-1 and wt plants under non-inductive and inductive conditions. Plants were grown in SD conditions at 20 °C for 3 weeks and subsequently shifted to 27 °C for 5 d. Leaf material and apically enriched material was harvested separately in the second half of the photoperiod (ZT6).
The FT level was higher in leaves of srr1-1 plants compared with wt plants under non-inductive conditions (Fig. 5A). Upon transfer to 27 °C, FT increased, reaching similar levels in wt plants and srr1-1 plants. The weak signal in apically enriched material probably reflects expression in residual leaf material. In concert with elevated FT levels in the leaf, the meristem identity gene AP1 was more strongly expressed in the apically enriched material of srr1-1 compared with the wt at 20 °C. AP1 was strongly induced at 27 °C and showed a higher level of expression in srr1-1 compared with the wt (Fig. 5B). The increased levels of FT in the leaves in srr1-1 under non-inductive growth conditions correlate with the early flowering phenotype in SDs, and the response to an inductive treatment is consistent with the weaker flowering phenotype in srr1-1 under inductive conditions. The floral integrator gene SOC1 was expressed at somewhat higher levels in srr1-1 compared with the wt at 20 °C in the leaves ( Supplementary Fig. S3A at JXB online). Transcript levels of SOC1 decreased somewhat after the shift to 27 °C in leaves, but expression levels were too low to draw any conclusions about changes in apically enriched material, where SOC1 activity is important for flowering. FLC acts as a repressor of flowering by binding to the FT promoter in the leaves and repressing FD and SOC1 in the shoot apical meristem (Searle et al., 2006). In srr1-1, FLC levels were decreased in both leaves and apically enriched material compared with wt plants, which probably contributes to early flowering of srr1-1 (Figs 3B, 5C). Little difference could be seen in FLC levels upon transfer to 27 °C, similar to earlier findings (Edwards et al., 2006). FLC has been shown to have a role in suppressing thermal induction in ecotypes with high FLC levels (Balasubramanian et al., 2006), but it has also been noted that FLC is not a major player in ambient temperature-responsive flowering in the Col ecotype (Blazquez et al., 2003).
The levels of SVP, a key component of the ambient temperature pathway, were similar in srr1-1 and the wt ( Supplementary Fig. S3B at JXB online). FLM transcript levels decreased in response to the increased temperature in both srr1-1 and the wt (Fig. 5D). Ratios between the repressive FLM-β isoform and the competing FLM-δ isoform decreased from ~5 at 20 °C to ~2 at 27 °C in the leaves and from ~4 at 20 °C to ~2 in 27 °C in the apically enriched material, in both srr1-1 and the wt ( Supplementary Fig. S3C-E). This is similar to what has previously been reported (Pose et al., 2013). Since no differences were observed between srr1-1 and the wt, the temperature response pathway was not consistently altered in srr1-1.
The transcription factors TEM1 and TEM2 are direct repressors of FT and have been shown to antagonize CO activation of FT in a redundant manner (Castillejo and Pelaz, 2008). Moreover, they have recently been shown to establish and control the length of juvenility and also repress CO expression (Sgamma et al., 2014). Down-regulation of both TEM1 and TEM2 expression is necessary for a plant's ability to respond to inductive photoperiods, through accumulation of FT. Both TEM1 and TEM2 were expressed at reduced levels in leaves of srr1-1 compared with the wt at 20 °C (Fig. 5E,  F). Upon transfer to 27 °C, TEM1 and TEM2 levels were strongly reduced in the wt and further reduced in srr1-1, in correlation with derepression of flowering in response to an ambient temperature increase. In the apically enriched material, only very low expression was detected, which was not significantly different between the wt and srr1-1 or between 20 °C and 27 °C. Expression of TEM1 and TEM2 transcripts is thus temperature sensitive.
The SRR1 transcript itself was expressed in both leaves and apically enriched material. It was not up-regulated by increased ambient temperature ( Supplementary Fig. S3F at JXB online). Thus, SRR1 does not respond to increases in temperature.

GA biosynthesis components are changed in srr1-1
To further examine the behaviour of TEM1 and TEM2 in srr1-1, their expression was tested in srr1-1 and the wt throughout the day. Both TEM1 and TEM2 showed peaks after dusk in SDs and LDs (Fig. 6A-D), correlating with a previous report (Osnato et al., 2012). The srr1-1 mutation led to somewhat lower levels of TEM1 and TEM2 in SDs (Fig. 6A, C) and LDs (Fig. 6B, D). The peak of TEM1 and TEM2 in SDs around ZT12 has been proposed to be important for repression of FT in SDs (Osnato et al., 2012), suggesting that lower TEM1 and TEM2 expression contributes to derepression of FT in srr1-1.
TEM1 and TEM2 have also been shown to regulate GA metabolism in Arabidopsis in concert with photoperiod (Castillejo and Pelaz, 2008;Osnato et al., 2012). TEM1 directly represses the expression of the GA biosynthetic genes GA3oxidase1 (GA3ox1) and GA3ox2 (Osnato et al., 2012). To examine whether the effect of the srr1-1 mutation on TEM1 and TEM2 transcript levels also affected expression of genes involved in GA biosynthesis, transcript levels of GA3ox1 and GA3ox2 were tested in plants grown at 20 °C in SDs. GA3ox1 levels were elevated in srr1-1 compared with the wt, while no difference could be observed in GA3ox2 levels (Fig. 6E, F). Previously, a higher up-regulation of GA3ox1 expression than of GA3ox2 was observed in tem1-1 tem2-1 loss-of-function mutants (Osnato et al., 2012). The smaller effect in the srr1-1 mutant, compared with tem1-1 tem2-1, is most probably due to the fact that TEM1 and TEM2 are still expressed in srr1-1, although at lower levels. Also the level of the GA biosynthesis gene GA20ox2, which is down-regulated in response to overexpression of TEM1, was somewhat higher in srr1-1 ( Supplementary Fig. S4A at JXB online). The catabolic enzyme GA2ox2 was unchanged between srr1-1 and the wt, suggesting that deactivation of GA is unaffected ( Supplementary Fig. S4B). Thus, through regulation of TEM1 and TEM2, SRR1 can affect FT repression and GA biosynthesis, which in both cases influence flowering.
To examine whether the overall GA response in srr1-1 was affected, srr1-1 and wt plants grown in SDs were treated with the bioactive GA 3 . This strongly promoted flowering in the wt ( Supplementary Fig. S4C at JXB online). srr1-1 plants reacted almost as strongly to exogenous GA 3 as wt plants, flowering with about half the leaves of untreated srr1-1 plants. This suggests that the GA pathway is functional in srr1-1. Treatment of SD-grown and LD-grown plants with the GA biosynthesis inhibitor paclobutrazol delayed flowering in both the wt and srr1-1, in line with the importance of GA also in LD conditions (Porri et al., 2012). srr1-1 still flowered earlier than the wt ( Supplementary Fig. S4D, E).

Discussion
Flowering in Arabidopsis is triggered by environmental factors such as increasing daylength (Andrés and Coupland, 2012) and temperature (Blazquez et al., 2003). It is, however, not only important for plants to respond to environmental changes that are suitable for flowering. Equally important is the ability to accumulate sufficient resources before the transition to reproductive growth. Floral repressors have an important role in this as safeguards against premature transition to flowering (Yant et al., 2009). Srr1-1 plants flower very early in SDs compared with the wt. Lengthening of the photoperiod greatly advances floral transition in the wt but has only a small promotive effect in srr1-1 (Supplementary Fig.  S1 at JXB online), suggesting that SRR1 is more important in non-inductive conditions.

SRR1 can affect flowering in ways both dependent on and independent of the photoperiodic pathway
The clock-controlled CO transcript oscillation with a peak in the dark determines the flowering response to LDs (Suarez-Lopez et al., 2001;Valverde et al., 2004). In srr1-1 plants, CO transcript levels were increased and CO started to accumulate already during the light period in SDs, possibly due to an advanced phase resulting from the srr1-1 clock phenoype ( Fig. 2A). This correlated with lower peak transcript levels of the CO repressor CDF1, compared with the wt (Fig. 2E, F). Thus, lower CDF levels most probably result in higher CO protein levels, which can promote FT expression. In addition, CDF1 can directly repress FT, and the lower CDF1 levels probably lead to derepression of FT.
As a result of these changes in CO and CDF1 transcript levels, the expression of FT, which normally is repressed at all time points in non-inductive conditions, has an LD-like pattern in srr1-1 in SDs, with a peak of expression at the beginning of the dark period. It thus seems as if the srr1-1 mutation unmasks an underlying rhythm of FT expression, showing that SRR1 has an important role as an inhibitor of flowering in non-inductive conditions, ensuring that the photoperiodic response is not triggered.
Despite this, the introduction of srr1-1 into the photoperiodic mutants gi-2 and co-9 resulted in accelerated flowering compared with gi-2 and co-9 single mutants, respectively, but delayed flowering compared wiht the srr1-1 single mutant in LDs (Fig. 1A, B). In SDs, the srr1-1 co-9 double mutant flowered in the same way as srr1-1 in SDs (Fig. 1A). Interestingly, both the srr1-1 gi-2 and srr1-1 co-9 double mutants flowered with the same number of leaves in SDs as in LDs, suggesting that the moderately earlier flowering phenotype of srr1-1 plants in LDs is dependent on the photoperiodic pathway. This is most probably a result of decreased repression in the absence of SRR1 and less promotion of flowering throughout the photoperiodic pathway, rendering the double mutants photoperiod independent. The accelerated flowering by srr1-1 in the co-9 and gi-2 background in both SDs and LDs does however suggest that SRR1 clearly can act independently of the photoperiod to regulate flowering. Thus, SRR1 can repress flowering in a dual mode, both through the photoperiodic pathway that is controlled by the circadian clock and in a photoperiod-independent manner.

SRR1 regulates several transcription factors that are repressors of FT
FT levels were higher in srr1-1 compared with the wt at 20 °C and increased in response to a flowering-inducing temperature shift from 20 °C to 27 °C, in both srr1-1 and the wt (Fig. 5A). This confirmed that FT is derepressed under noninductive conditions in srr1-1.
The decreased transcript levels of the FT repressors TEM1 and TEM2 in srr1-1 compared with the wt (Fig. 5E, F) and their diurnal expression profiles ( Fig. 6A-D) with a lower peak of expression in the dark phase in SDs explains part of the derepression of FT. Partial suppression of TEM1 and TEM2 may not fully explain the early flowering of srr1-1, due to redundancy of the single tem1 and tem2 mutants; however, as both TEM1 and TEM2 expression is reduced in srr1-1, an effect on FT levels is likely.
The peaks of TEM1 and TEM2 in the dark phase have been proposed to be important not only for FT repression but also for regulation of GA biosynthesis components (Osnato et al., 2012). Consequently, the TEM1/TEM2 targets in the GA biosynthesis pathway, GA3ox1 and GA3ox2, were somewhat increased (Fig. 6E, F).
Lower TEM1 and TEM2 levels in srr1-1 led to less repression of FT and thus accelerated flowering and indirect (positive) effects on flowering through increased activity in the GA biosynthesis pathway. In addition, this further connects the circadian clock and TEM1 and TEM2, where SRR1 helps to maintain TEM1 and TEM2 levels and in this way inhibits FT accumulation and flowering. GI has also been shown to interact with TEM1 and TEM2 on the protein level (Sawa and Kay, 2011). GI levels are, however, unchanged in the srr1-1 background in SDs. A possibility is that SRR1 promotes TEM1 and TEM2 expression and that GI in turn interacts with the TEM1 and TEM2 proteins to regulate them. TEM1 and TEM2 are reported to counteract CO promotion of FT in a developmental manner, decreasing with increasing age of the plants, and also repress CO expression (Castillejo and Pelaz, 2008;Sgamma et al., 2014). Since SRR1 can influence the transcript patterns of both CO and TEM1 and TEM2, it possibly acts to balance the expression between CO on the one hand and TEM1 and TEM2 on the other hand to sustain vegetative growth until both environmental and developmental factors favour transition to flowering.

SRR1 represses flowering over a wide range of temperatures
srr1-1 flowered earlier than the wt under all tested temperatures, and responded to increases in temperature by accelerated flowering, showing that the temperature response in srr1-1 is functional (Fig. 4). The srr1-1 phyB-9 double mutant behaved like srr1-1 at 16 °C, where the phyB-9 mutation has no effect on flowering, while the effect of the two mutations was additive at 20 °C (Fig. 4A, B). At 27 °C, the double mutant flowered like phyB-9, with a very strong flowering response to the temperature. The consistent early flowering of srr1-1 at all tested temperatures shows that SRR1 is necessary to prevent premature flowering in a wide temperature range. Moreover, the changing impact of the phyB-9 mutation on the srr1-1 flowering phenotype in different temperatures, ranging from no effect on the flowering phenotype of the srr1-1 phyB-9 mutant at 16 °C to an additive effect at 20 °C and phyB-9like flowering at 27 °C, suggests that the relationship between phyB and SRR1 could be temperature dependent (Fig. 4).
Furthermore, srr1-1 plants showed a much weaker vernalization response than wt plants, probably because FLC transcript, encoding a key floral repressor, is already much lower in srr1-1 than in wt plants before vernalization (Fig. 3). SRR1 can thus prevent flowering in a photoperiod-independent manner by promoting FLC expression.

SRR1 integrates photoperiod-dependent and photoperiod-independent information to repress flowering in non-inductive conditions
The presented data reveal that SRR1 affects expression of several repressors of FT. Among those are transcription factors of different classes including the MADS domain protein FLC, the RAV (RELATED TO ABI3/VP1) family TEM1 and TEM2, and the Dof (DNA-binding with one finger) protein CDF1. This includes genes with no rhythmic expression (FLC), and rhythmic genes with an expression peak in the morning (CDF1) as well as with an expression peak in the dark (TEM1/TEM2). A working model for SRR1's role in flowering time control is described in Fig. 7.
With srr1-1 responding to all tested flowering-promoting treatments, the role of SRR1 seems not to be restricted to a specific signalling pathway, but rather to maintaining a basal level of repressive elements in non-inductive conditions. The early flowering phenotype of srr1-1 plants in SDs thus seems to be the result of the combined effect of loss of expression of several transcription factors that act as direct repressors of FT expression, leading to an LD-like expression pattern of FT. Under Fig. 7. Conceptual model of SRR1 function. SRR1 has a function in setting the period of the circadian clock, as well as in phyB signalling. Expression of FT repressors involved in photoperiodic regulation of flowering-CDF1, TEM1, and TEM2-as well as the photoperiod-independent FLC is promoted by SRR1. TEM1 and TEM2 are also connected to developmental control of flowering, since their expression levels decrease with age. In this way, SRR1 prevents premature flowering in non-inductive SDs. In the srr1-1 mutant, decreased repression of FT and early accumulation of CO leads to an LD-like expression pattern of FT and early flowering, especially in non-inductive environmental conditions. (This figure is available in colour at JXB online.) inductive conditions such as LDs, other activating factors overcome the effect of SRR1 to trigger a flowering response. This could explain why SRR1 was identified as an important regulator of flowering in a genome-wide association and QTL mapping study on a plant population grown in field conditions over two seasons (Brachi et al., 2010), since screens for flowering time regulators in laboratory conditions have been performed in conditions optimized for flowering that seldom occur in a realistic environment. Further, an SRR1 homologue in Brassica rapa was recently shown to be associated with flowering time control in a study combining flowering QTL analysis and wholegenome transcript variation (Xiao et al., 2013). SRR1 appears to be a focal point of several pathways, necessary to synchronize photoperiodic regulation with other factors to maintain vegetative growth under non-inductive conditions. Thus, SRR1 is an upstream regulator of reproduction, preventing flowering until other factors signal that the time is suitable to shift from vegetative to reproductive growth.

Supplementary data
Supplementary data are available at JXB online. Figure S1. Flowering time of srr1 in different photoperiods. Fgure S2. Characterization of T-DNA insertion lines in SRR1. Figure S3. Transcript analysis of SVP, FLM-β, FLM-δ, SOC1, and SRR1. Figure S4. Transcript analysis of GA20ox2 (A) and GA2ox2 (B) in plants grown at 20 °C and subsequently shifted to 27 °C. Table S1. List of primers used in this study.