Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Plants regulate their time to flowering by gathering information from the environment. Photoperiod and temperature are among the most important environmental variables. Sub-optimal, but not near-freezing, temperatures regulate flowering through the thermosensory pathway, which overlaps with the autonomous pathway. Here we show that ambient temperature regulates flowering by two genetically distinguishable pathways, one requiring TFL1 and another requiring ELF3. The delay in flowering time observed at lower temperatures was partially suppressed in single elf3 and tfl1 mutants, whereas double elf3 tfl1 mutants were insensitive to temperature. tfl1 mutations abolished the temperature response in cryptochrome mutants that are deficient in photoperiod perception, but not in phyB mutants, which have a constitutive photoperiodic response. In contrast to tfl1, elf3 mutations were able to suppress the temperature response in phyB mutants, but not in cryptochrome mutants. Gene expression profiles revealed that the tfl1 and elf3 effects are due to the activation of different sets of genes, and identified CCA1 and SOC1/AGL20 as being important cross-talk points. Finally, genome-wide gene expression analysis strongly suggests a general and complementary role for ELF3 and TFL1 in temperature signalling.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Plants compute variables such as light and temperature to finetune flowering onset, integrating environmental information to determine whether the appropriate flowering season is about to arrive (Boss et al., 2004). In Arabidopsis thaliana, day-length detection is accomplished through the photoperiod pathway, whereas responses to low temperatures, associated with winter time, require the vernalization pathway.
As a facultative long-day (LD) plant, Arabidopsis flowers earlier under LD conditions than under short-day (SD) conditions. CONSTANS (CO) is the central component of the photoperiod pathway, because it is essential in discriminating between LD and SD due to its regulation at different levels. First, expression of CO at the mRNA level is tightly regulated by the circadian clock and other components, such that maximum expression levels occur during the night under SD conditions or late in the light period under LD conditions (Suarez-Lopez et al., 2001; Yanovsky and Kay, 2002; Valverde et al., 2004; Imaizumi et al., 2005; Laubinger et al., 2006). Second, its expression is post-translationally regulated by light; photoreceptors phytochrome A (phyA) and cryptochrome 2 (cry2) promote CO stability, whereas phytochrome B (phyB) antagonizes this stabilization (Valverde et al., 2004; Laubinger et al., 2006; Jang et al., 2008). These results account for the late flowering of phyA and cry2 mutants and the early flowering of the phyB mutants. As a result of both types of regulation, CO levels are only high enough under LD conditions to induce the expression of FLOWERING LOCUS T (FT), a promoter of flowering that acts as an integrator of various flowering pathways (Boss et al., 2004). After induction, FT protein moves to the apical meristem to switch on genes required for reproductive development (Corbesier et al., 2007; Jaeger and Wigge, 2007; Mathieu et al., 2007; Tamaki et al., 2007).
In turn, the vernalization pathway present in over-wintering accessions promotes flowering by stably down-regulating a flowering repressor, the MADS box transcription factor FLOWERING LOCUS C (FLC) (Boss et al., 2004; He and Amasino, 2005). FLC represses the expression of FT and another important flowering-time integrator, the MADS box transcription factor SUPPRESSOR OF OVEREXPRESSION OF CO1 (SOC1, also known as AGL20) (Searle et al., 2006).
Despite advances in understanding of the vernalization pathway, flowering responses to sub-optimal temperatures are mostly unknown (Samach and Wigge, 2005; Lee et al., 2008; Penfield, 2008). Low temperatures can regulate development in Arabidopsis (Mazzella et al., 2000), and it has been proposed that a thermosensory pathway regulates flowering time in response to ambient temperature by FLC-independent mechanisms (Blazquez et al., 2003). FVE and FCA, two genes that were previously classified as part of the autonomous pathway, are also part of the thermosensory pathway (Blazquez et al., 2003). More recently, the flowering repressor SHORT VEGETATIVE PHASE (SVP) was shown to be important for the ambient temperature response by directly regulating FT expression (Lee et al., 2007). On the other hand, the early-flowering phenotype of phyB mutants was shown to be temperature-dependent (Halliday and Whitelam, 2003; Halliday et al., 2003), suggesting that at least some interactions between light and temperature may be expected. These findings suggest that temperature signalling may occur by various pathways or mechanisms; however, no evidence has been presented to date to support this view. Of the known flowering repressors, only SVP was shown to be involved in temperature signalling (Lee et al., 2007).
We decided to study the role of flowering repressors in the ambient temperature response. TERMINAL FLOWER 1 (TFL1) and EARLY FLOWERING 3 (ELF3) appeared as interesting candidates for our research. TFL1 is a member of the FT family, but acts in an antagonistic manner to FT. The antagonistic effects of FT and TFL1 have been mapped to a single amino acid position (Hanzawa et al., 2005; Ahn et al., 2006). tfl1 mutants display an early-flowering phenotype that is ameliorated by growth under SD conditions (Shannon and Meeks-Wagner, 1993). TFL1 has been placed genetically downstream of the autonomous pathway genes FVE and FCA (Ruiz-Garcia et al., 1997; Page et al., 1999; Soppe et al., 1999), genes that are also involved in the thermosensory pathway (Blazquez et al., 2003). These results suggest that TFL1 is an interesting candidate for a temperature-signalling component.
Here, we present evidence that ambient temperature regulates flowering by two genetic pathways: one that is closely associated with the photoperiod pathway and requires ELF3, and another that requires TFL1 and is related to the autonomous pathway. Furthermore, we show by microarray analysis that both ELF3 and TFL1 play important and complementary roles in temperature signalling.
Ambient temperature affects the photoperiodic response
To investigate whether photoperiod signals affect the response to ambient temperature, we grew Arabidopsis plants under SD (8 h light/16 h dark) and LD (16 h light/8 h dark) conditions at either 16 or 23°C. As previously reported, growth at 16°C produced a delay in flowering time (Blazquez et al., 2003; Halliday and Whitelam, 2003; Halliday et al., 2003; Lee et al., 2007). However, the delay was most obvious under LD conditions compared with SD conditions (Figure 1a), and a significant interaction between photoperiod and temperature was observed (two-way anova, P <0.01). The effects of low temperature were also observed under continuous light (CL) (Figure 1b).
tfl1 and elf3 mutants display a reduced response to temperature
The observed interaction between photoperiod and temperature suggests that the photoperiod pathway may be antagonized by low temperatures or promoted by higher ones. We reasoned that if low temperatures delay flowering by antagonizing the photoperiod pathway, at least some of the early-flowering mutants could have a reduced responsiveness to ambient temperature. Only a few of the known early-flowering mutants have been evaluated for flowering time at low temperatures (Blazquez et al., 2003; Halliday et al., 2003; Lee et al., 2007). We chose two early-flowering mutants, elf3 and tfl1, and compared their behaviour with that of phyB mutants. We measured the flowering time of elf3-7, elf3-9, tfl1-1 and tfl1-14 mutants and compared them side by side with the wild-type (WT) and phyB mutants (Figure 2). phyB mutants flowered at an earlier stage than the WT plants at 23°C (i.e. when they had approximately four fewer leaves), but no significant differences were observed at 16°C (Figure 2), as reported for plants grown under SD conditions (Halliday and Whitelam, 2003; Halliday et al., 2003). Because phyB mutants flowered early at 23°C, the temperature effect was even stronger in phyB mutants compared with WT controls. In contrast to phyB, elf3 and tfl1 mutants showed a reduced response to temperature. In other words, low temperatures were less efficient in restraining flowering in elf3 and tfl1 mutants. These results were not due to allele-specific effects, because we found similar behaviours with the various alleles (Figure 2).
elf3 but not tfl1 mutations impair the responsiveness to temperature of phyB mutants
The contrasting responses observed in the early-flowering mutants prompted us to study the epistatic relationships among tfl1, elf3 and phyB. We grew phyB, tfl1 and elf3 single and double mutants under continuous light at 16 or 23°C (Figure 3). The genetic interactions among tfl1, elf3 and phyB were strikingly different. The elf3 phyB double mutants flowered earlier than single mutant parents at both temperatures, which is consistent with phyB and ELF3 acting independently in flowering (Reed et al., 2000). However, elf3 was epistatic to phyB with respect to temperature sensitivity, a conclusion that is supported by two facts. First, loss of ELF3 function rendered a phyB mutant hyposensitive to temperature (Figure 3, compare phyB mutants with phyB elf3 double mutants). Second, a phyB mutant was early flowering at 16°C in the elf3 genetic background (compare WT versus phyB and elf3 versus phyB elf3 at 16°C). In other words, the elf3 mutation not only reduced the response to temperature, but also changed the behaviour of phyB mutants at low temperatures. In contrast to elf3, the tfl1 effect was mostly additive to the phyB effect. tfl1 and elf3 mutants still showed responsiveness to temperature, but double elf3 tfl1 plants flowered essentially at the same developmental time at either 16 or 23°C. Taken together, these results suggest that ELF3 and TFL1 regulate the flowering response to ambient temperature by different pathways: ELF3, but not TFL1, acts in a photoperiod- and phyB-related pathway.
The tfl1 mutation suppresses the temperature response in cryptochrome-deficient plants
The three photoreceptors that promote flowering under LD conditions in Arabidopsis are cry2, phyA and cry1, with cry2 being the most important based on the phenotype of cry2 mutants (Mockler et al., 1999, 2003). It was previously reported that, under LD conditions, the delay in flowering imposed by low temperatures was exaggerated in cry2 mutants (Blazquez et al., 2003). We decided to test whether the tfl1 mutation was able to suppress the delayed flowering onset at 16°C reported in cryptochrome-deficient mutants. We grew WT plants and tfl1, cry2, tfl1 cry2, cry1 cry2 and tfl1 cry1 cry2 mutants under CL at either 16 or 23°C (Figure 4). Lack of cryptochromes resulted in a delay in the time to flowering, irrespective of the presence of the tfl1 mutation. These results imply that expression of the tfl1 phenotype requires the presence of cryptochromes, as recently reported (Buchovsky et al., 2008). However, the delay in flowering time produced by the low temperature in the cry2 mutant disappeared in the tfl1 cry2 double mutant. A similar but weaker effect was observed in the tfl1 cry1 cry2 triple mutant (Figure 4). These results show that the loss of temperature sensitivity in tfl1 mutants (Figure 2) is not due to saturation of the flowering-promoting pathways in the tfl1 genetic background, because a similar loss of temperature responsiveness still occurs in the late-flowering tfl1 cry2 double mutant (Figure 4).
The effect of the elf3 mutation on the temperature response requires a functional photoperiod pathway
In phyB mutants, in which the photoperiod pathway is constitutively activated (Valverde et al., 2004), an elf3 mutation severely impaired the sensitivity to temperature (Figure 3). When we compared the effect of elf3 on the temperature response in cryptochrome-deficient plants, we observed a different behaviour to that of the tfl1 mutants. Although elf3 cry2 double mutants flowered earlier when compared with tfl1 cry2 mutants (Figure 4), the elf3 mutation did not suppress the response to temperature in a cry2 background, as observed with tfl1. Progressive accumulation of mutations in the other photoperiod photoreceptors, cry1 and phyA, rendered an elf3 mutation relatively ineffective at inducing early flowering at both 16 and 23°C, but even in the extreme case of the triple phyA cry1 cry2 mutant background, elf3 was not able to suppress the response to temperature (Figure 4). A similar effect was observed under LD conditions (Figure S1).
A phyB mutation accelerated flowering in the cry2 single and phyA cry1 cry2 triple mutant backgrounds at 23°C, consistent with previous reports (Mockler et al., 1999, 2003). However, the effect of the phyB mutant was suppressed by low temperatures, rendering the phyB cry2 double and phyA phyB cry1 cry2 quadruple mutants hypersensitive to temperature (Figure S1).
TFL1 and ELF3 regulate the flowering response to temperature at different points
To try to understand the molecular mechanisms underlying the involvement of TFL1 and ELF3 in modulating flowering in response to temperature, we analysed the gene expression profile of WT and the elf3 and tfl1 mutants. As a first approach, we focused on a list of flowering-time genes selected from the literature (Table S1). We found that CIRCADIAN CLOCK ASSOCIATED 1 (CCA1), LATE ELONGATED HYPOCOTYL (LHY), GIGANTEA (GI), CO, FT, SOC1 and FLC are differentially expressed in elf3 mutants (Figure 5 and Figure S2; see Table S1 for q values for the elf3 effect). Despite the complex behaviour of gi mutants, genetic evidence suggests that GI acts downstream of CCA1 and LHY in the regulation of flowering time (Mizoguchi et al., 2005; Niwa et al., 2007). GI, CO, FT and SOC1 appear to act in a linear pathway (Suarez-Lopez et al., 2001; Yanovsky and Kay, 2002; Mizoguchi et al., 2005; Yoo et al., 2005). These data strongly suggest that early flowering at low temperatures in elf3 mutants is due, at least in part, to activation of an important set of photoperiod pathway genes by the low mRNA levels of the circadian core components CCA1 and LHY. The changes observed in the expression patterns of CYCLING DOF FACTOR 1 (CDF1), CRY1, CONSTANS-LIKE 2 (COL2), ACTIN RELATED PROTEIN 2 (ATARP4) and ELF4 are either relatively small or the direction of change is such that it does not account for the early-flowering phenotype of the elf3 mutants (Ledger et al., 2001; Doyle et al., 2002; Imaizumi et al., 2005; Kandasamy et al., 2005), but we cannot rule out a small contribution from HY5 HOMOLOG (HYH) (Holm et al., 2002). Interestingly, the tfl1 expression profile showed a completely different pattern to that of elf3, with SOC1 as the only flowering-time gene that was differentially expressed (Figure 5 and Figure S2; see Table S1 for q values for the tfl1 effect). The slightly elevated CDF3 levels are not expected to affect flowering time (Imaizumi et al., 2005).
When elf3 × temperature interactions were analysed, CCA1 was at the top of the list of flowering-time genes (q <0.05). Although LHY showed a similar expression pattern (Figure 5a and Figure S2), the interactions were not statistically significant. Interestingly, SOC1 was down-regulated by lower temperatures (q <0.05), indicating that, in elf3 and tfl1 mutants, the effect of temperature is partially suppressed by up-regulation of SOC1 (Figure 5c).
The results of transcriptome profiling suggest a role for circadian clock components in temperature signalling. In addition to their early-flowering phenotype, elf3 mutants are arrhythmic under CL (McWatters et al., 2000; Covington et al., 2001; Hicks et al., 2001). We decided to test CCA1 over-expressors, which are also arrhythmic, but, unlike elf3 mutants, are late-flowering (Wang and Tobin, 1998). We grew CCA1 over-expressor lines and gi mutants under CL at 16 and 23°C. As expected, both genotypes were late flowering compared with WT, but displayed reduced sensitivity to low temperatures (Figure 6).
Another important, well-known flowering gene that is mis-expressed in efl3 mutants is FLC. FLC mRNA levels were threefold lower in elf3 mutants (Figure S2). However, consistent with previous reports (Blazquez et al., 2003; Lee et al., 2007), we observed only a relatively minor loss of sensitivity to temperature in flc mutants (Figure 6b). The levels of both FLC and FT were not clearly regulated by temperature in the WT as previously observed (Blazquez et al., 2003; Lee et al., 2007). We believe that this is due to the different conditions used; unlike the previous authors, we used CL in our microarray experiments.
The data presented so far are consistent with a model in which the ambient temperature can regulate flowering by two separate pathways, one that requires TFL1 and is independent of the photoperiod pathway, and another that shows at least some degree of interaction with the photoperiod pathway and depends on ELF3 activity.
ELF3 and TFL1 play a general role in temperature signalling
We reasoned that if ELF3 or TFL1 plays a more general role in temperature responses, the set of temperature-responsive genes should be enriched in genes affected by the mutant genetic backgrounds. After filtering of the data (see Experimental procedures), we generated a list of temperature-responsive genes in WT and a list of elf3- and tfl1-responsive genes at 23°C (t test, P <0.05). Of the 2473 temperature-regulated genes and 478 elf3-regulated genes, 235 were found in both groups. This intersection is highly significant (P =2.66 × 10−16, Fisher exact test) (Figure S3). Furthermore, 219 of the 235 shared genes (93%) changed in the same direction after a decrease in temperature or presence of theelf3 mutation. Of the 629 tfl1-regulated genes at 23°C (t test, P <0.05), 175 were also temperature-regulated, which is significant (P =0.0006, Fisher exact test); however, only 80 (46%) changed in the same direction. Twenty-six genes were found to be present in all three groups, i.e. elf3-, tfl1- and temperature-regulated (P =2.95 × 10−5) (Figure S3).
Next, we compared the list of temperature-responsive genes in the WT with the list of elf3- and tfl1-responsive genes at 16°C rather than 23°C (Figure S4). Of 1263 elf3-regulated genes, 341 were also temperature-responsive (P =1.3 × 10−5), but only 231 (68%) changed in the same direction. Of 1761 tfl1-regulated genes, 615 were also temperature-responsive (P =7.94 × 10−21), but only 46 (7%) changed in the same direction, meaning that 93% of the shared genes changed in the opposite direction after a decrease in temperature or presence of the tfl1 mutation at 16°C. These results show that a common set of genes is affected by temperature and elf3 and/or tfl1, and also suggest that there is a concordance between the effect of lower temperatures on gene expression and the effect of the elf3 and tfl1 genotypes at 23 and 16°C, respectively.
These results were confirmed using gene set enrichment analysis (GSEA) (Subramanian et al., 2005). This method uses a list of genes ranked by the effect of one factor (temperature in the WT), and then questions whether the genes affected by the other factor (elf3 or tfl1 genetic background) are randomly distributed in the former list or clustered at the top or bottom. A running sum statistic is calculated for the list of genes ranked by the temperature effect; the magnitude increases when the gene belongs to the group of genes affected by the genotype and decreases when it does not. The maximum deviation from zero is the enrichment score (ES). The ES was highly significant in each case (Figure S5), showing that the group of temperature-regulated genes is enriched in genes affected by the tfl1 and elf3 mutations at both temperatures.
The previous results confirm that ELF3 and TFL1 share common target genes with temperature signalling. However, when the targets are also similarly affected by the different treatments, a functional connection is revealed. We used a modified GSEA analysis (concordance GSEA), that has previously been used to query the Connectivity Map (Lamb et al., 2006), to infer functional connections among ELF3, TFL1 and temperature signalling. The concordance GSEA takes into account both the significance and the direction of changes in gene expression. Genes in the WT were ordered from the most down-regulated to the most up-regulated by low temperature. This list was queried using the top 200 differentially expressed genes for the four treatments, i.e. elf3 at 23°C, elf3 at 16°C, tfl1 at 23°C and tfl1 at 16°C; the running sum was computed separately for the down-regulated and up-regulated genes (Figure 7). At 23°C, genes down-regulated in the elf3 genotype (Figure 7a, blue vertical lines) were enriched among the genes down-regulated by temperature, whereas genes that were up-regulated in the elf3 genotype (Figure 7a, red vertical lines) were enriched among the genes up-regulated by temperature. The ES was highly significant (0.870, P <0.0001), confirming a direct concordance between the changes in gene expression produced by the elf3 mutation at 23°C and those occurring at low temperatures. In other words, the elf3 mutants grown at 23°C mimic WT plants grown at 16°C. Conversely, we did not find any significant correlation between tfl1-regulated genes at 23°C and the temperature-response expression profile (Figure 7c). At 16°C, the results changed dramatically. The concordance between the elf3 genotype at 16°C and the effects of low temperature was much weaker, although significant (Figure 7b; ES = 0.276, P <0.001), whereas a strong inverse concordance was found for the tfl1 genotype (Figure 7d; ES = 0.843, P <0.0001). The genes up-regulated in the tfl1 genotype at 16°C, represented by red vertical lines (Figure 7d), were enriched among the genes down-regulated by low temperature, and genes down-regulated in the tfl1 genotype were enriched among the genes up-regulated by low temperature (blue vertical lines, Figure 7d).
Finally, we compared our set of data with those previously reported (Balasubramanian et al., 2006). These datasets were used to generate a list of genes ranked by the responsiveness to temperature (25°C to 16°C) and the concordance with the differentially expressed genes of our data analysed as above (Figure S6). Interestingly, despite comparison of samples from plants grown under very different conditions (5-week-old SD-grown apices), we obtained similar results (compare Figure 7a with Figure S6a and Figure 7d with Figure S6d).
The inverse concordance between tfl1 and the low-temperature response strongly suggests that TFL1 is a positive regulator of the responses to low temperature. On the other hand, the direct concordance between elf3 and low-temperature effects suggests a negative role for ELF3. This result appears to be contradictory because elf3 mutants are early flowering and low temperatures delay flowering. This apparent contradiction can be explained by the effects of interactions between the elf3 genotype and temperature on gene expression. The flowering behaviour of elf3 mutants may be explained by its effects on CCA1 expression, which decreased in a temperature-dependent manner in elf3 mutants (Figure 5 and Figure S2). The interactions between the elf3 genotype and temperature are not restricted to CCA1 expression. Seven genes from the phenlypropanoid pathway were significantly affected in the elf3 mutant background (Figure S7 and Table S2, P <0.001), including two transcription factors, AT4G09820 (TT8) and AT1G22640 (MYB3), which encode positive (Nesi et al., 2000) and negative regulators of the pathway, respectively. It is noteworthy that four of the seven selected genes showed significant elf3 x temperature interactions (Table S4 and Figure S7). Their expression was significantly affected at 16°C, but not at 23°C, suggesting that the pathway is more active in elf3 mutants grown at 16°C. Furthermore, the genes that showed a significant elf3 x temperature interaction (Table S4, uncorrected P value <0.01) were enriched in those that experienced higher effects of elf3 at 16°C than at 23°C (P =0.0022, Fisher exact test), which supports a global role for ELF3 in the response to ambient temperature.
In the present paper, we investigated the role of TFL1 and ELF3 in the regulation of flowering time at low ambient temperatures. Lower temperatures (16°C) delay flowering (Figure 1), and this delay is severely impaired in both tfl1 and elf3 single and double mutants (Figures 2 and 3). Two sets of experiments strongly suggest that TFL1 and ELF3 regulate the sensitivity to temperature by different genetic pathways. First, whereas tfl1 and elf3 single mutants displayed reduced sensitivity to temperature (Figure 2), the flowering time of the elf3-9 tfl1-1 double mutant was almost unaffected by temperature (Figure 3). The elf3-9 allele is likely to be null because of an early stop codon (Hicks et al., 2001), whereas the tfl1-1 allele has undetectable TFL1 levels due to a change in a conserved amino acid (Bradley et al., 1997; Page et al., 1999; Conti and Bradley, 2007). Thus, it seems unlikely that the behaviour of single and double mutants is due to a combination of weak alleles. The second set of experiments also supports the notion of two separate pathways. Whereas tfl1 and phyB effects were additive, elf3 was mostly epistatic to phyB with respect to temperature sensitivity (Figure 3). On the other hand, tfl1 was effective in reducing the temperature response in cry2 mutants, but elf3 was not (Figure 4).
The genome-wide gene expression data showed that elf3 mutations affect a set of photoperiod pathway genes (Table S1, Figure 5 and Figure S2). Low levels of CCA1 and LHY expression and a concomitant rise in GI, CO, FT and SOCI/AGL20 levels were observed, consistent with reported effects of elf3 and circadian clock mutations on gene expression (Suarez-Lopez et al., 2001; Kim et al., 2005b; Mizoguchi et al., 2005). In contrast, the tfl1 mutation antagonized the effects of temperature on SOC1 expression, but did not show clear mis-regulation of upstream photoperiod pathway genes (Table S1, Figure 5 and Figure S2).
Our genetic and gene expression data collectively show that ambient temperature regulates flowering time by at least two pathways. One pathway appears to work regardless of the state of the photoperiod pathway and requires TFL1. This pathway is more likely to be related to the thermosensory pathway reported previously (Blazquez et al., 2003). This is consistent with the proposition that TFL1 might work genetically as a negative regulator of events that occur downstream of FVE and FCA (Ruiz-Garcia et al., 1997; Page et al., 1999; Soppe et al., 1999), the autonomous pathway genes that are required for the thermosensory pathway (Blazquez et al., 2003). Interestingly, TFL1 appears to act genetically in the same pathway as EARLY FLOWERING IN SHORT DAYS (EFS), which was also placed downstream of FVE and FCA (Soppe et al., 1999). However, EFS is involved in histone methylation, and efs mutations are highly pleiotropic, leading to early flowering in several late-flowering backgrounds (Soppe et al., 1999; El-Assal et al., 2003), probably due to high levels of SOC1 and FT mRNA (Zhao et al., 2005).
The second pathway requires ELF3 and is likely to be associated with the photoperiod pathway. Several results support this view. First, the elf3 mutation suppressed the low-temperature response in phyB mutants (Figure 3), which are known to have a constitutively activated photoperiod pathway (Valverde et al., 2004), but not in photoperiod pathway-impaired mutants (Figure 4). Second, the elf3 mutation affects expression of photoperiod- and clock-related genes such as GI and CCA1 (Figure 5), and over-expression of CCA1 or loss of function of GI impairs the response to temperature (Figure 6). These results also raise the possibility that the role of ELF3 may be exerted through the circadian clock. elf3 mutants are arrhythmic in the light, and we confirmed that this is also the case at 16°C. After entrainment under 12 h light/12 h darkness photoperiods, we moved 8-day-old seedlings to continuous light at either 16 or 23°C. Under these free-running conditions, CCA1 mRNA did not cycle in elf3 mutants at either temperature (data not shown). The acclimation response to freezing temperatures, controlled by the CBF regulon, is gated by the circadian clock (Fowler et al., 2005). Understanding of the ambient temperature response at a similar level awaits the development of gene expression markers for the acute response to ambient temperature changes. These advances will allow the study of the interactions between ELF3 and the state of the oscillator, especially in darkness, when the elf3 oscillator is functional.
Our data are consistent with TFL1 and ELF3 acting in different organs: ELF3 interacting with the photoperiod pathway and clock components in the leaves and TFL1 regulating SOC1 in the apex. However, we cannot rule out the possibility that they may act in common tissues. ELF3 is widely expressed, including in the apex (Hicks et al., 2001), whereas TFL1 is mostly expressed in the apex, but also in the inflorescence (Bradley et al., 1997). Similarly, SOC1 is highly expressed in the apex but is also found in leaves (Lee et al., 2000; Kim et al., 2005a), and we cannot rule out the possibility that TFL1 might regulate SOC1 expression beyond the apex (Conti and Bradley, 2007).
Finally, GSEA analysis strongly suggests that ELF3 and TFL1 play more general roles in the responses of plants to ambient temperature. The inverse concordance between the effect of low temperature and the effects of tfl1 at 16°C, but not at 23°C (Figure 7c,d), strongly suggests that TFL1 plays a positive role in the response to low ambient temperature. These results are more interesting in light of recent findings that TFL1 is associated with membranes (Sohn et al., 2007) and the role of membrane processes in the perception of temperature in non-plant systems (Mansilla et al., 2004). On the other hand, the direct concordance between low temperature and elf3 effects on gene expression (Figure 7a) suggests a negative role for ELF3 in modulating the response to temperature. In principle, this proposition appears contradictory, because, if ELF3 plays a negative role in temperature signalling, we would expect elf3 mutants to flower later than WT plants at lower temperatures. However, early flowering of elf3 mutants at low temperatures correlates with its effects on the expression of CCA1 and other photoperiod pathway genes (Figure 5 and Figure S2).
SVP plays an important role in the flowering response to ambient temperature (Lee et al., 2007) and was recently shown to regulate SOC1 transcription directly (Li et al., 2008), which is interesting given that our microarray experiments revealed that SOC1 is responsive to temperature. Whether SVP acts in a similar pathway to TFL1 or ELF3 is still unclear. We did not find differences in SVP expression in the tfl1 or elf3 mutants that could account for the behaviour of the tfl1 and elf3 mutants at lower temperatures. Conversely, we did not find significant effects of temperature on TFL1 or ELF3 mRNA levels in svp and autonomous pathway mutants (Figure S8). The NAC-family transcription factor LONG VEGETATIVE PHASE 1 (LOV1) has been shown to play a dual role, repressing flowering within the photoperiod pathway and positively regulating the cold response, suggesting that LOV1 is a link between cold responses and flowering (Yoo et al., 2007). In contrast to our results, LOV1 appears to act independently of CCA1, LHY and GI, negatively regulating CO expression. We did not find relevant changes in LOV1 expression in our microarray experiments. The fact that GI positively regulates CO in the photoperiod pathway, but that gi mutants are hypersensitive to cold (Cao et al., 2005), whereas co mutants are tolerant to freezing (Yoo et al., 2007), underscores the complexity of the interactions between flowering-time genes and temperature responses.
The use of Arabidopsis as a model system has elucidated the role of new players in the temperature-signalling network in plants. However, unlike several light and hormonal receptors, the nature of the thermosensor is still not well understood. Part of the difficulty is the enormous diversity of biological macromolecules whose activities are temperature-sensitive. Whatever the nature of the thermosensor(s), when spring approaches, the increase in day length induces flowering in LD plants, such as Arabidopsis. However, the variation in ambient temperature may still be drastic as seasons change. The ability of Arabidopsis to antagonize flowering promotion when temperatures are still sub-optimal ensures that seed set occurs under more benign conditions.
The tfl1-1 cry2-1, elf3-7 cry2-1 and elf3-9 cry2-1 double mutants were obtained by crossing single mutant parents, selecting for late-flowering cry2 homozygotes in the F2, and screening for either tfl1-1, elf3-7 or elf3-9 homozygous plants in the F3 using dCAPs (see Appendix S1 for more details on genotyping various alleles).
The tfl1-1 cry2-1 hy4B104 and elf3-7 cry2-1 hy4B104 triple mutants were obtained by crossing the double mutants tfl1-1 cry2-1 and elf3-7 cry2-1 with the cry2-1 hy4B104 double mutants and selecting for tall plants under blue light in the F2. Genotypes were confirmed by PCR using dCAPs (Neff et al., 1998) as described in Appendix S1.
The elf3-7 phyA412 cry1-304 cry2-1 quadruple mutants were obtained by crossing elf3-7 to the phyA412 cry1-304 cry2-1 triple mutant. F2 seedlings were grown under far-red light, and tall plants were transplanted to soil. Because phyA and cry2 are linked, the selected plants were also late flowering. F2 plants heterozygous for elf3-7 were harvested. F3 siblings showing consistent long hypocotyls under blue light (and therefore cry1 homozygous candidates) were transplanted to soil and genotyped to search for elf3-7 homozygotes.
In all cases, the double mutants were genotyped again, and the hypocotyl length was checked under far-red, blue and white light to confirm the genotypes before flowering-time experiments. The phyA412 phyB9 cry1-304 cry2-1 quadruple mutant was a generous gift from Todd Mockler (Department of Botany and Plant Pathology, Oregon State University, OR).
Seeds were sterilized with chlorine in the vapour phase, and plants were grown on a 1:1:1 mix of peat moss, vermiculite and perlite. Every two weeks, plants were fertilized with a 0.1% solution of Hakaphos (Compo Agricultura, http://www.compo.es). Photoperiods were as indicated for each experiment, with a light intensity of 80 μmol m−2 sec-1 produced by cool white fluorescent tubes. The total leaf number, rosette plus cauline leaves, was determined at the time of flowering. Experiments were repeated at least once for consistency of the results. When a two-way anova was used (Figure 1), the data were log10 transformed to achieve normality and homoscedasticity.
Seeds were sterilized and sown on plates with Murashige & Skoog salts and 1.2% plant agar (Duchefa Biochemie, http://www.duchefa.com). After 3 days of stratification at 4°C, plants were incubated in growth chambers under the same conditions used for flowering experiments for 10 more days. Seedlings were harvested, weighed and frozen in liquid N2. Total RNA was prepared using a plant RNAeasy kit (Qiagen, http://www.qiagen.com), and 5 μg was used to prepare the cRNA that was hybridized to the Affymetrix expression arrays (ATH1-121501) as described by the manufacturer (Affymetrix, http://www.affymetrix.com). The expression set (18 chips, three biological replicates per treatment) was obtained after Robust Multi-Array Average (RMA) normalization and elimination of the probe sets with signals that were not significantly higher than the background using the affy package implemented in the R system (Irizarry et al., 2003; Gautier et al., 2004; R-Development Core Team, 2008). The genotype, temperature and interaction coefficients were obtained by fitting the data to the linear model: yi = μi + αi + βi + αix βi + ε, where μi is the mean for gene i, αi is the genotype effect on gene i, βi is the temperature effect on gene i, and αix βi is the interaction effect between genotype and temperature. ε is the zero mean normally distributed error. Moderated P values for the null hypothesis that the coefficients are equal to zero were estimated for each gene using the limma algorithm (Smyth, 2005). We used False Discovery Rate (FDR) to correct P values for multiple hypothesis testing, and corrected values are reported as q values.
The gene set enrichment analysis (GSEA) was implemented in the R system. Redundant probe sets mapping to the same gene were eliminated by keeping only the one that showed the highest dynamic range (CV) across the complete expression set. The expression set was rank-sorted using the P value complement: rank(1 − P), where P is the P value, and the enrichment score (ES) was estimated using an exponent score of 1 as previously described (Subramanian et al., 2005). The P value for the ES was computed from a null distribution obtained by permuting the gene labels 10 000 times. For concordance GSEA, the running sums were computed independently for the up-regulated and down-regulated genes of the querying gene set using the same expression profile sorted from the most down-regulated to the most up-regulated gene. A combined running sum was computed by inverting the order of the running sum for the up-regulated genes and subtracting it from the running sum for the down-regulated genes. The maximum and minimum values of this combined running sum were added to obtain the ES.
We are grateful to the ABRC Stock Center (Ohio State University, Columbus, OH) and José A. Jarillo (INIA-UAU, Madrid, Spain), Ji-Hoon Ahn (Korea University, Seoul, Korea) and Miguel Blázquez (Universidad Politécnica de Valencia, Valencia, Spain) for seed stocks, Todd Mockler (Oregon State University, Corvallis, OR) for the phyA412 phyB9 cry1-304 cry2-1 quadruple and phyA412 cry1-304 cry2-1 triple mutants, Joanne Chory for the phyB9 and elf3-7 mutants and the CCA1 over-expressor, Jorge Casal (IFEVA-CONICET, Buenos Aires, Argentina), Marcelo Yanovsky (IFEVA-CONICET) and Santiago Mora-García for useful comments on the manuscript, Edith Trejo and Matías Rugnone for technical assistance, and other lab members for their support. The earlier phases of this work were supported by a National Institutes of Health grant to Joanne Chory (The Salk Institute for Biological Sciences, La Jolla, California), and afterwards by grants PICT 01-14287 (ANPCyT) and CRP/ARG05-02 (ICGEB) to P.D.C. B.S. was supported by a fellowship from the YPF foundation.