Potent Induction of Arabidopsis thaliana Flowering by Elevated Growth Temperature

1The postembryonic developmental program of plants is extraordinarily flexible and can change dramatically in response to many environmental factors. Examples of such variables are light quantity (day length), light quality (red–far-red ratio), vernalization (exposure to winter temperatures for several weeks) and ambient growth temperature. In addition, flowering is also affected by nitrous oxide and various stresses, including biotic (e.g., pathogens) and abiotic stress. Extensive genetic analysis of laboratory-induced mutations as well as naturally occurring genetic variants has identified at least four distinct pathways controlling flowering in the reference plant Arabidopsis thaliana. The photoperiodic pathway receives inputs from the circadian clock and day length, and the nuclear protein CONSTANS (CO) integrates its effects. Vernalization promotes flowering by enabling stable repression of FLOWERING LOCUS C (FLC), a potent suppressor of flowering. In winter annual accessions of Arabidopsis, FLC levels are high due to activation by FRIGIDA (FRI). Loss of function of FRI or attenuation of FLC contributes to a rapid-cycling behavior in many Arabidopsis accessions, including the commonly used laboratory strains Landsberg erecta and Columbia. A series of autonomous pathway genes promote flowering in a photoperiod-independent manner, also via suppression of FLC levels. Finally, the hormone gibberellin (GA) is essential in a fourth pathway, which controls flowering redundantly with the photoperiod pathway. All of these pathways appear to converge on a small number of integrators, the flowering-time genes FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1), and the floral meristem identity gene LEAFY.

2In contrast to light- and vernalization-dependent flowering, less is known about how other factors modulate the transition to flowering. Some progress has been made in understanding how moderate changes in growth temperature affect flowering. Plants with a defect in the red-light receptor phytochrome B (PHYB) flower early at 23 °C, but not at 16 °C. In contrast, higher temperatures ameliorate the late-flowering phenotype of plants that lack the blue light receptor CRYPTOCHROME 2 (CRY2), apparently because the redundantly acting far-red-light receptor PHYTOCHROME A (PHYA) is not able to promote flowering at lower temperatures. Such a flowering behavior is found in many wild accessions, which cluster in their response to different environments with cry2 mutants. The differential flowering phenotype of phyB and cry2 mutants contrasts with the temperature-insensitive late flowering of autonomous pathway mutants.

3Here, we show that a modest increase of 2 °C to 4 °C in ambient growth temperature beyond the common laboratory condition of 23 °C potently triggers flowering in the absence of photoperiodic cues. There is extensive natural variation in this response, and we identify loci that contribute to this response in wild accessions of Arabidopsis thaliana. We demonstrate that the floral repressors FLC and FLM differ in their effects on thermal induction, and that the thermal response is integrated downstream of CO at FT. Microarray analyses confirm that the genomic responses to photoperiodic and thermal induction differ, and identify unique sets of genes that are activated or repressed in response to either cue.

4Arabidopsis thaliana is commonly cultivated in the laboratory at temperatures of 20 °C to 23 °C, even though its natural distribution is largely in areas that have a lower mean temperature. While growing wild-type Landsberg erecta (Ler) and Columbia (Col) plants at a range of temperatures in short days, we noticed that a modest increase in temperature from 23 °C to either 25 °C or 27 °C had a dramatic effect on flowering (Figure 1A). As reported before, the two common laboratory strains Ler and Col flowered much later in 23 °C 8-h short days than 16 °C 16-h long days. Ler flowered at about the same time in 25 °C or 27 °C short days as in 16 °C long days, while Col flowered only slightly earlier in 16 °C long days compared to 25 °C and 27 °C short days (Figure 1B). The similar behavior is seen both when measuring days to flowering (chronological time) as well as total leaf number produced before the first flower on the main stem (physiological age), indicating that this is not simply due to differences in growth rate (unpublished data). Since it is known that flowering-time mutants and wild strains of Arabidopsis vary in their long-day flowering behavior under different temperature regimens, we assessed the thermal response of about 50 mutants and 52 wild accessions in short days. As with many other environmental responses, there is extensive variation in thermal response (Table S1). However, strains flowered on average earlier at 27 °C compared to 23 °C short days, and there was a significant difference in mean total leaf number between 23 °C and 27 °C (Figure 2A). This difference was also seen when total leaf number was partitioned into rosette and cauline leaf numbers.

5FLC is a potent repressor of flowering. Together with its upstream regulator FRI, FLC plays an important role in natural variation of Arabidopsis thaliana flowering [7–9,11,12,18–21]. Since the effect of FRI/FLC is reduced, but still significant in short days, we first assessed whether strains that differ in FRI/FLC status vary in thermal response. We found a significant difference in the mean total leaf number at 23 °C versus 27 °C only among the strains that have lesions at FRI or FLC (Figure 2A). Since flc-3 knockout mutants can still respond to thermal induction in a manner similar to the parental Col line, thermal induction cannot be simply mediated by suppression of FLC. We therefore asked whether elevated FLC levels could attenuate thermal induction. A well-known way to increase FLC levels even in the absence of functional FRI is through mutations in the autonomous pathway. Mutants with defects in this pathway did not respond to thermal induction of flowering (Figure 2B). If higher FLC levels are indeed responsible for the failure of autonomous pathway mutants to respond to elevated temperatures, a mutation in FLC should suppress their nonresponsiveness to thermal induction. Indeed, double mutants of flc-3 with autonomous pathway mutants in the Col background showed a response similar to that of flc-3 single mutants, flowering substantially earlier in 27 °C than 23 °C short days (Figure 2B). In addition, plants that overexpress FCA, an autonomous pathway gene, responded well to thermal induction (Figure 2B).

6It has recently been demonstrated that FLC mediates natural variation in temperature compensation of the circadian clock. FLC lengthens the period of the clock at higher temperatures, which should lead to a delay in flowering, consistent with our observations. If FLC indeed modulates sensitivity to temperature, one might expect that the mean sensitivity to temperature will vary between the lines with functional or nonfunctional alleles at FRI/FLC. We found this to be the case, with the lines that have lesions at FRI/FLC being more sensitive than the lines with putatively functional FRI/FLC (ANOVA, p = 0.02). Taken together, these results suggest a role of FLC in suppressing thermal induction. In addition, these results indicate that the failure of autonomous pathway mutants to respond to thermal induction appears to be largely due to their elevated FLC levels.

7We next asked in several ways whether accessions differ in their response to temperature. The reaction norms and the thermal sensitivities of accessions indicated that there is considerable variation in their response to thermal induction (Figure 2C and Table S2). We then ranked the accessions based on their flowering time at 23 °C and 27 °C. While there is a general correlation between flowering times in the two conditions, some strains ranked very differently in 23 °C versus 27 °C (Table S2). These findings confirm that there is a significant genotype × environment (G × E) interaction in thermal response. Finally, by comparing flowering times at 23 °C and 27 °C, we identified strains that do not respond to thermal treatment.

8Since FLC is a potent suppressor of thermal induction, we looked for nonresponsive accessions among those that carry nonfunctional alleles at either FRI or FLC (Figure 2D). Of six unresponsive strains, Uk-3, Pog-0, and Cvi-0 have high FLC levels in spite of an FRI deletion. Because the higher FLC levels likely explain the thermal insensitivity of these three accessions, we focused our further analysis on other temperature non-responsive strains.

9Since recombinant inbred lines were available for Col-0 crossed to Nd-1, a temperature-insensitive strain with low FLC levels, we decided to perform quantitative trait locus (QTL) mapping using Niederzenz-1/Col (NdC) recombinant inbred lines (RILs). QTL mapping experiments have led to the identification of a deletion of the floral repressor FLOWERING LOCUS M (FLM)/MADS AFFECTING FLOWERING1 (MAF1) as a major cause for early flowering of Nd-1 in short days at 23 °C. Since FLM is similar in sequence to FLC, which represses thermal induction, we expected that FLM would also inhibit temperature responsiveness. Contrary to our expectations, the effect of FLM, the major effect QTL in 23 °C short days, was masked in 27 °C short days, and the QTL was no longer detectable (Figure 3A). Consistently, there is no significant difference in the mean flowering time of plants with or without the FLM deletion at 27 °C (not shown), indicating that FLM modulates the sensitivity to temperature. To confirm this assumption, we calculated the thermal sensitivities of each of the RILs and asked whether a QTL for thermal response colocalizes with the FLM locus, which was indeed the case (Figure 3A). In contrast to FLC, where strains with low expression levels respond more strongly to thermal induction, lines without FLM respond less well to temperature compared to FLM wild-type strains (Figure 3B). This finding suggests that temperature acts in the same genetic cascade as FLM. Temperature might suppress the repressive effect of FLM on flowering. Alternatively, temperature might act downstream of FLM to bypass the repressive effect of FLM.

10The response to temperature is reduced, but not eliminated in the NdC lines with the FLM deletion, compared to lines with the wild-type allele, indicating that other factors contribute to thermal response as well (Figure 3B). We identified Ei-6, another strain that flowers early in short days, as having the same deletion as Nd-1 (not shown). Ei-6 also has a reduced thermal response (Figure 3B). However, Ei-6 flowers even faster than Nd-1 in 23 °C short days. Our earlier analysis of F2 populations derived from a cross between Ei-6 and Col had pointed to a complex genetic basis of the early flowering behavior of Ei-6, consistent with the hypotheses that there are natural modifiers for FLM effects.

11FLM functions independently of the autonomous and vernalization pathways, but interacts with the photoperiod pathway. Therefore, we tested the effects of photoperiod mutants, gigantea (gi), co, cry1, and cry2. GI encodes a nuclear protein with several roles in light-mediated and stress responses, while CO functions as the principal output of the photoperiod pathway by integrating circadian with light information. The CRY2 photoreceptor functions primarily in flowering control, while CRY1 controls both seedling and flowering responses to blue light. A strong thermal response is retained in plants that have mutations in the CRY1 or CRY2 photoreceptors (Figure 4A). Like the cry mutants, phyA, phyB, and phyC mutations do not interfere with the thermal response (unpublished data). co mutants in either the Col or Ler background flower considerably earlier in 27 °C compared to 23 °C short days (Figure 4A). Similarly, the gi-6 allele induced in Ler shows a thermal response, although this is not the case for the gi-2 allele induced in Col. To determine whether this difference is due to a direct role of GI in thermal response or due to an allele specific effect of gi-2, we tested another strong allele, gi-3, in the Ler background. gi-3 mutants flowered earlier and similar to gi-6 at 27 °C (Table S1). Thus, the absence of a thermal response in gi-2 plants could be due to natural variation between Ler and Col with respect to GI function, or caused by an allele specific effect of gi-2, rather than indicating a direct role for GI in thermal response. An allele-specific effect is consistent with the recent observation that gi-2 differs from other gi alleles in its effects on the circadian clock. However, we cannot rule out the possibility that the response of the Ler alleles is due to remaining partial GI activity in gi-3 or gi-6.

12Flowering pathways converge at the level of the so-called integrators. Among these, FT and SOC1 have crucial roles in flowering time, while LEAFY functions primarily in floral identity. In addition to ft and soc1 mutants, we assayed fwa mutants, which have a reduced response to FT activity, along with plants with a mutation in FD, which mediates FT activity, as well as fe mutants, which have a similar genetic behavior as ft. All these mutants showed a reduced response to 27 °C short days (Figure 4B), suggesting that thermal induction, like the other known floral induction pathways, acts upstream of the integrators.

13soc1–1 ft-7 double mutants flowered in 27 °C short days at a similar time as wild-type did in 23°C short days, but much later than wild-type in 27 °C short days (Figure 4B). It has recently been suggested that the available ft alleles in the Ler background are hypomorphic alleles, and that the additive effect of soc1–1 ft-7 double mutants is simply due to ft-7 not being a null allele. Consistent with this hypothesis we found that the RNA null allele ft-10 in the Col background is completely insensitive to temperature (Figure 4B), indicating that thermal induction is mediated primarily by FT.

14All co alleles tested, co-2 and co-8 in Ler, and co-9 in Col, have a pronounced response to temperature, indicating that thermal induction is clearly independent of CO. Thus, temperature must affect FT in at least in two ways. First, temperature modulates the effects of floral repressors such as FLC. Second, temperature leads to a photoperiod independent activation of FT. Consistent with this hypothesis, we find mRNA levels of FT to be more than ten-fold at 27 °C compared to 23 °C short days (Figure S1). Circadian oscillation of FT was not affected; its relative levels were higher at all time points. The increase in FT expression levels could be at least partially responsible for the thermal induction of flowering, consistent with the effects of an ft mutation on this process.

15GA acts redundantly with the photoperiod pathway in promoting flowering, and they are especially important in short days. Plants with a dominant mutation in the GA response factor GA INSENSITIVE (GAI), which have a reduced response to GA, flower much later than wild-type under normal short day conditions. Short-day flowering of gai-D plants was still accelerated by 27 °C (Table S1). GA-deficient ga1–3 mutants, which cannot flower at all in regular short days, also did not respond to thermal induction. Thus, thermal induction cannot overcome the requirement for GA in short days. Given that the ga1–3 phenotype can be suppressed by overexpression of FT, this observation suggests either that thermal induction of FT is not sufficiently strong to bypass the GA requirement, or that thermal induction does not act exclusively through FT, at least in the Ler background. Consistent with this, a small thermal response was retained by ft mutants in Ler background (Figure 4B).

16Because temperatures in the native range of Arabidopsis are on average well below 25 °C, we were curious whether flowering upon thermal induction could be due to a stress response. Many stress responses in plants are promoted by the hormone abscisic acid (ABA). ABA has recently been shown to act through the FLC regulator FCA. ABA, however, represses flowering, which is inconsistent with a positive role of ABA in accelerating flowering in response to higher temperatures.

17Another stress hormone that has recently been implicated in flowering control is salicylic acid (SA), which is required for activation of flowering by UV-C light stress. The effects of SA signaling on known flowering regulators are complex. To determine whether thermal induction of flowering might be caused by the SA pathway, we tested the thermal response of mutants in which SA-dependent responses are blocked. Mutations in ENHANCED DISEASE SUSCEPTIBILITY1 (EDS1) were initially isolated because they are required for certain types of pathogen resistance. More recently, EDS1 has been shown to play important roles in responses to several abiotic stresses as well (reviewed in). The SA INDUCTION-DEFICIENT2 (SID2) locus, also known as EDS16, encodes an enzyme required for SA synthesis. Both eds1-2 and eds16-2 behaved similar to the parental lines and flowered early at 27 °C (Figure 4A and Table S1), indicating that the SA pathway does not mediate thermal induction of flowering.

18Since genetic analysis indicated that temperature acts independently of the photoperiod pathway, we went on to study molecular changes caused by thermal induction. To assess whether the genomic responses to thermal and photoperiodic induction are different, we first grew plants in 16 °C short days for 5 wk, after which they were still vegetative. We then changed either the photoperiod or the temperature; one set of plants was transferred to 16 °C long days, while the other set was transferred to 25 °C short days. Using a similar design as described, we analyzed changes in the transcriptome of shoot apices with Affymetrix ATH1 arrays on days 2, 5, and 9 after the transfer.

19Several genes, including SOC1, FRUITFUL (FUL), APETALA1 (AP1), CAULIFLOWER (CAL), LEAFY, SQUAMOSA PROMOTER BINDING PROTEIN LIKE 3 (SPL3), and SPL4, are robust early markers for flower development. On day 9, expression levels of these markers were substantially elevated, with several of them responding earlier and more strongly to thermal than to photoperiodic treatment (Figure 5A).

20We employed several approaches to test how the genomic responses to thermal and photoperiodic induction differ. First, we used principal component analysis, which showed that transcriptomes of plants assayed at different time points in 25 °C short days were more similar to each other than those of plants in 16 °C long days, and that the environmental variable was more important than the temporal component (Figure 5B). Hierarchical clustering of conditions supported this finding (unpublished data). Because many genes responded to both treatments, we repeated this analysis with genes that were different on day 9 in both conditions compared to day 0. Even then the thermal and photoperiodic shift conditions clustered independently (unpublished data), confirming that the kinetics of induction of common targets is different.

21By comparing the transcriptome after either thermal or photoperiodic induction, we identified several genes whose profiles responded specifically to only one of the treatments (Figure 5C). We used the GeneMerge tool to determine whether gene products involved in specific biological processes or molecular functions, as defined by their Gene Ontology annotations, preferentially responded to one of the treatments. Because early floral markers, such as SPL3, were already induced on day 2 in 25 °C short days, we focused on this timepoint.

22In 16 °C long days, there was a significant enrichment of genes encoding zinc-binding proteins (p = 0.0067 after correcting for multiple testing), with two upregulated and four downregulated genes. All of them belong to B-box type zinc finger transcription factors. Several members of this family are clock-regulated, suggesting that this enrichment is due to an altered circadian profile in response to longer photoperiods (Figure 5D). This raised the question whether all the genes that are altered during light shift could simply be circadian-regulated genes. Therefore, we visually inspected the genes that show an altered profile during light shift for their behavior in a diurnal dataset. More than three-quarters of the genes that had responded to photoperiod change by day 2 showed diurnal oscillations. This fraction was reduced to about two-thirds by day 9. In contrast, only about a quarter of the genes that responded to the thermal shift showed diurnal oscillations.

23A typical heat shock does not induce flowering. In addition, our plants flowered early regardless of whether they were transferred at the adult stage to higher temperatures, or were grown from germination at elevated temperatures. In the latter case, plants should have become acclimated by the age that they are responsive to floral induction, and one would therefore not expect a heat shock effect on flowering. Nevertheless, we wanted to know whether heat shock–responsive genes were affected by our treatment.

24An analysis of the microarray data showed that both the thermal and the photoperiodic shifts affected a small number of heat shock genes. However, none of the timepoints showed an enrichment for heat shock genes in thermally induced samples compared to photoperiod-induced samples. Among the genes highly induced in the thermal samples was At3g12580, one of the 14 HSP70 genes encoded in the genome. Although At3g12580 is stress responsive, it also has a complex developmentally regulated pattern of RNA expression. Furthermore, it is well known that HSP70 genes are involved in many different biological processes apart from heat shock responses.

25Because heat shock genes typically respond rapidly and strongly to elevated temperatures, but return within hours to normal levels, our first microarray timepoint (48 h) might not have been appropriate for the analysis of heat shock genes. We therefore specifically compared the expression of several genes encoding heat shock proteins (HSPs) or heat shock transcription factors in samples shifted from 16 °C to 25 °C or 37 °C after 2 h. The selection of these genes was based on a previous detailed analysis of their response to heat shock. As expected, several genes were strongly induced by 37 °C, while 25 °C had only minor effects (Figure S2), confirming that the thermal induction we find is not due to a typical heat shock response.

26In 25 °C short days, we found a significant enrichment of genes encoding proteins involved in RNA processing (p = 0.007 after correcting for multiple testing; Figure 5E). This group included 11 factors associated with splicing or having RNA recognition motifs. Six genes encoding SR proteins, which are thought to control splice site selection and alternative splicing both in Arabidopsis as well as rice were upregulated. Some of these factors have tissue-specific expression profiles and are expressed in the shoot apical meristem. In addition, overexpression of RSZ33 and SR30 has pleiotropic effects that include variable alterations in flowering time. RSZ33 interacts with RSZ21, another splicing factor, which was also upregulated.

27The enrichment of RNA-processing related gene products responding to thermal induction suggests that temperature might affect RNA processing in Arabidopsis. This is particularly interesting, since there are several flowering time regulators with alternatively spliced transcripts. Therefore, in a first step, we assessed FCA, MAF2, and FLM, genes known to have different splice forms. We detected different splice forms for all three genes. Using regular RT-PCR experiments, no obvious changes in the abundance of different FCA splice forms were detected (Figure S3), but more sensitive real-time RT-PCR experiments using splice-form–specific primers showed that there appears to be a subtle change in the relative abundance of the FCA splice forms, with a specific increase in the beta form during thermal shift (unpublished data). However, this form has no obvious function.

28For MAF2, two splice forms of similar abundance were detected at 16 °C, before and after transfer to long days (Figure S3). In contrast, the larger splice form was more abundant in 25 °C short days. These differences in splicing patterns were consistent in three independent shift experiments (unpublished data).

29For FLM, larger splice forms, albeit of minor abundance, appeared specifically after thermal shift, which seemed to be accompanied by a reduction in the levels of the major splice form (Figure S3). While it is tempting to speculate that this reduction could be a mechanism through which temperature overcomes the repressive effect of FLM, the function of the different splice forms is not yet known. Nevertheless, the observed changes in FLM are consistent with a role of FLM in this process. In addition, flm mutants flower earlier than flc mutants in 23 °C short days, suggesting that the repressive effects of FLM may be more important for the later flowering of plants grown in 23 °C compared to 25 °C or 27 °C short days. Conversely, the observed changes in FCA splice forms in response to elevated temperature may not be that important, since the FCA target FLC is still a potent repressor of flowering under these conditions. Further experiments are required to understand the specific effects of the different FLM and MAF2 splice forms. Similarly, it remains to be tested whether the observed change in expression of SR protein genes directly affects these splicing patterns.

30We have shown that elevated temperature has a strong inductive effect on flowering, even in the absence of photoperiodic cues. Among wild accessions, there is tremendous variation in this response, which is partially attributable to the suppressive effect of FLC on flowering. Both mutant analysis and QTL mapping demonstrate that thermal induction has a genetic basis and acts in the same genetic cascade as FLM. Importantly, the genomic response to thermal induction differs from that of photoperiodic induction. Preliminary analyses suggest that altered splicing may be a component of thermal response in plants.

31Seeds were obtained from the Nottingham Arabidopsis Stock Centre and from colleagues. The stock numbers of accessions and mutants used are given in Table S1. T-DNA insertion lines for CO (co-9) were obtained from Syngenta (Garlic-24-H04.b.1a.Lb3Fa; Basel, Switzerland). A SALK T-DNA line (Stock Centre number N641971) for FLM was isolated (flm-3) and verified to be an RNA-null allele (Min-Chul Kim and DW, unpublished data). Plants were cultivated in paired incubators (Percival Scientific, Perry, Iowa, United States) or growth rooms. Short days were 8 h of light/16 h of dark; long days were 16 h of light/8 h of dark. Ten to 12 plants per genotype were grown in a completely randomized design, in order to minimize environmental variation, and scored for their flowering time, which was measured by counting total leaf number (TLN, partitioned into Rosette [RLN] and Cauline [CLN] leaf number). Thermal induction was robust at 25 °C in incubators, but required 27 °C in the growth rooms. Therefore, induction experiments in the growth rooms were done at 27 °C. Growth chamber experiments were done at 25 °C, including microarray studies. There were no strong differences in the spectral quality between the chambers and the growth rooms, indicating that the observed differences could possibly be due to small temperature fluctuations in the chambers compared to the better ventilated growth rooms, which have only little variation (±0.1 °C).

32Plants were grown for 5 wk at 16 °C in short days, and then transferred to either 16 °C long days or 25 °C short days. Twenty-five apices per genotype/condition/replicate were dissected and flash frozen on the day of the transfer (day 0) and days 2, 5, and 9 after transfer. RNA extracted from two independent biological replicate samples was hybridized to ATH1 arrays (Affymetrix, Santa Clara, California, United States) as previously described. Data were normalized using the gcRMA algorithm (bioconductor.org), a modification of the robust multiarray analysis (RMA) algorithm, and visualized using Gene Spring 7 (Agilent Technologies, Palo Alto, California, United States). Pairwise comparisons were performed using logit-transformed probe-level testing. Microarray data has been deposited with the ArrayExpress database (Supplemental Information). In addition to the RNA samples used for the microarray experiment, samples collected independently using a similar design were used for RT-PCR analysis to verify variations in splicing patterns. The primers used for the analysis of splicing patterns are given in Table S3. For circadian profiling of FT, aerial parts of 2-wk-old seedlings were collected over a 24-h time period in biological replicates. For the analysis of heat shock proteins, plants grown in 16 °C long days were shifted to 25 °C or 37 °C, and leaf material was collected at 30-min intervals for 2 h after the shift.

33Statistical analysis was done using the JMP package (SAS Institute, Cary North Carolina, United States), the statistical package R (http://www.r-project.org), and Microsoft Excel. The NdC population has been described. Twelve plants per RIL were grown at 27 °C short days in growth rooms in a completely randomized design. Broad-sense heritability (H2) was calculated as between-line variance (VG) divided by total variance. The total variance was partitioned into between-line variance and the residuals in a one-way ANOVA model using the genotype as a single factor of random effect and the TLN as the response. The estimated heritability was 0.62. QTL mapping was performed using R-qtl. Likelihood of odds thresholds were determined using 1,000 permutations. Sensitivity to temperature was assessed through regression of the sample mean on the environmental mean as previously described. Temperature sensitivity was calculated for each of the RILs, mutants and the accessions. The sensitivity measures obtained for the RILs were then used in QTL mapping as a phenotype to identify a QTL for thermosensitivity. The phenotypic measurements and the genotypic data used for generating the QTL map are available in CSV format (NdC.csv).

34The ArrayExpress (http://www.ebi.ac.uk/arrayexpress) accession number for the temperature/photoperiodic shift microarray experiment is E-MEXP-728.