Molecular basis of flowering under natural long-day conditions in Arabidopsis.

Plants sense light and temperature changes to regulate flowering time. Here, we show that expression of the Arabidopsis florigen gene, FLOWERING LOCUS T (FT), peaks in the morning during spring, a different pattern than we observe in the laboratory. Providing our laboratory growth conditions with a red/far-red light ratio similar to open-field conditions and daily temperature oscillation is sufficient to mimic the FT expression and flowering time in natural long days. Under the adjusted growth conditions, key light signalling components, such as phytochrome A and EARLY FLOWERING 3, play important roles in morning FT expression. These conditions stabilize CONSTANS protein, a major FT activator, in the morning, which is probably a critical mechanism for photoperiodic flowering in nature. Refining the parameters of our standard growth conditions to more precisely mimic plant responses in nature can provide a powerful method for improving our understanding of seasonal response.

Main Text:

Many plants utilize day-length (=photoperiod) and temperature information to control various seasonal responses for survival and reproduction. Among the seasonal responses, flowering regulation in Arabidopsis is the most characterized response at the molecular level[1]. Photoperiod and temperature information is processed through circadian clock-dependent mechanisms to induce the expression of the florigen gene, FT, around dusk in long days (LD)[2, 3]. This LD specific FT induction occurs in leaf phloem companion cells. Once synthesized in LD, FT protein is transferred from the leaves to the shoot apical meristem to trigger the transition from vegetative to reproductive development[4].

Many components in the Arabidopsis photoperiodic pathway are highly conserved in angiosperms (including major crops such as rice, wheat, barley, and potato) to regulate seasonal responses. For instance, genes identified through quantitative trait locus (QTL) analyses on flowering time, yield, or other domestication traits (often tied with loss/reduction of photoperiod sensitivity) in many crops frequently turned out to be homologs of the Arabidopsis photoperiodic flowering components[5]. The photoperiodic sensing mechanism originally characterized in Arabidopsis was found to already exist in bryophytes to regulate photoperiodic reproductive development[6]. This indicates that incorporating photoperiodic information into developmental regulation has been important for land plant survival.

Thus far, Arabidopsis research has been instrumental in not only identifying the components involved in photoperiodic flowering, but also understanding how these components function in this pathway under well-controlled lab settings. However, it remains unknown whether the current model of photoperiodic flowering regulation can recapitulate the seasonal flowering mechanisms in complicated natural LD environments. Here we show the presence of a previously uncharacterized regulation of florigen induction in Arabidopsis plants grown in natural LD, and our subsequent attempt to elucidate its regulatory mechanism using lab growth conditions optimized to plant responses in nature.

Results

Flowering regulation under natural LD conditions

Light (day length and light quality) and temperature are major environmental parameters that control flowering time[2, 3, 7]. The day length and temperature conditions of the summer solstice in Seattle, WA, USA, (47°36′N; day length, 15 hours 59 min; average high temperature from 1971 to 2000, 21.1 °C) were similar to our lab LD conditions (16 hours, constant 22 °C). In addition, ecological studies showed that summer annuals of wild Arabidopsis plants grown in similar latitudes to Seattle germinate and flower within a roughly one month period between March and July in both Europe and North America[8, 9, 10, 11, 12, 13, 14, 15]. Therefore, we tested how accurately photoperiodic flowering regulation that occurs in controlled lab environments can represent flowering regulation under similar natural LD conditions.

We grew wild-type (WT: Col-0) plants outside in Seattle in June and harvested them around the summer solstice (Fig. 1a and Supplementary Fig. 1). We analyzed the expression of genes important for photoperiodic flowering regulation (Fig. 1b, c and Supplementary Fig. 2)[3]. The expression patterns of six circadian clock genes, clock output flowering genes, and floral repressor genes were relatively similar to those already described in the lab-grown samples (Supplementary Fig. 2)[16, 17, 18, 19]. In addition, the night-peaking CO expression profile was similar to the one in lab LD with cooler nights (Fig. 1b)[19]. These results indicate that simplified lab conditions recapitulate the natural gene expression profiles of those genes. However, in the plants grown outside, FT showed a bimodal expression pattern with peaks in the morning and around dusk (Fig. 1c), which clearly differed from the typical FT pattern peaking near dusk in lab LD[17, 20]. We also observed a similar bimodal expression pattern in TWIN SISTER OF FT (TSF), a related florigen gene (Supplementary Fig. 2l)[20].

Fig. 1:

The florigen FT gene is induced in the morning in natural LD.

a, Changes in light intensity and temperature on the days near the summer solstice in 2013 when the samples were harvested. For outside conditions, Zeitgeber time 0 (ZT0) was set as the sunrise time (i.e. 5 AM in Seattle from 6/23/13 to 6/25/13). Light intensity results are means ± SEM from different growth areas (n=3). Temperature data were obtained from a nearby weather station. b, c, Expression profiles of CO (b) and FT (c) under the conditions shown in (a). All gene expression results (means ± SEM) in this manuscript were normalized against IPP2 and PP2A (n=3 biologically independent samples). d, Flowering time results of plants grown outside in June and in lab LD. Each box is located between the upper and the lower quartiles, and the whiskers indicate 1.5-times interquartile ranges. The thick horizontal lines in the boxes represent the median, and open diamonds represent the mean. Outliers are indicated by circles. (12≤n≤100, *** p<0.001, ns: non-significant, linear models or generalized linear models were used throughout the manuscript. See detail statistical information in Supplementary Table 3). e, FT expression profiles in wild-type (WT: Col-0) plants, fkf1, and gi-2 mutants grown outside around the summer solstice. f, FT expression profiles in WT grown at different times in spring. g, FT expression profiles in WT grown around the summer solstice in Seattle and Zürich. (for e-g, n=3 biologically independent samples). h, Flowering phenotypes of WT plants grown in different months and locations in spring. The details of the box plots are the same as those in Fig. 1d (n≥11, ** p<0.01, *** p<0.001, ns: non-significant, statistical information in Supplementary Table 3).

Since FT levels strongly correlate with flowering time[19, 21], we analyzed the flowering time of Col-0 plants grown in natural LD. During the last five years, even though temperatures around the summer solstice varied (Supplementary Fig. 1), Col-0 plants all flowered at similar developmental times with fewer leaves than plants grown in lab LD (Fig. 1d). In addition, we repeatedly observed similar CO and FT expression patterns in samples harvested around the last five summer solstices (Fig. 1b, c, e-g and Supplementary Fig. 3 and 5). We also grew other accessions, such as a common lab accession, Ler, and another WT accession, Vancouver-0 (Van-0, isolated from Vancouver, BC, Canada; 49°15′N) in natural LD. We speculated that Van-0 is adapted to an environment similar to Seattle. Both accessions flowered earlier than Col-0 in lab LD, even slightly earlier in natural LD (Fig. 1d), and showed bimodal FT expression patterns in natural LD (Supplementary Fig. 4).

We also analyzed the phenotypes of some non-transgenic alleles of photoperiodic flowering mutants, such as flavin-binding, kelch repeat, f-box 1 (fkf1) and gigantea (gi-2), in natural LD. The flowering time of these late-flowering mutants was significantly earlier than that in lab LD (Fig. 1d). The fkf1 mutant flowered at almost the same time as Col-0 plants in natural LD, suggesting that some regulation that takes place outside but not in the lab may trigger earlier flowering in the fkf1 mutant. When we analyzed FT expression patterns in fkf1 and gi-2 grown in natural LD, the gi-2 mutant lost FT expression as expected[22], but the morning FT expression was clearly observed in fkf1 (Fig. 1e). These results suggest that morning FT expression in fkf1 is likely the cause of the early flowering phenotype of fkf1 in natural LD.

As our results suggested a functional contribution of morning FT expression on flowering in natural LD, we analyzed the expression patterns of FT in samples grown in different times in spring (April, May, and June), which is the growth season of summer annuals[8, 9, 10, 11, 12, 14]. In Seattle, the days are already lengthening in April (approximately 14 hours); however, the ambient temperature in April was colder than in May and June (Supplementary Fig. 1). Col-0 plants flowered later in April than in May and June (Fig. 1h). In all samples grown during spring in the last two years, FT peaked in the morning with different levels (lower in April than in May and June) without changing CO patterns (Fig. 1f and Supplementary Fig. 5).

Next, we tested whether the FT morning peak is observed in two native locations of Arabidopsis. We grew Col-0 plants in Zürich (47°37’N, similar latitude to Seattle), and in Edinburgh, (55°57’N) in June. Although the temperatures (and day length in Edinburgh) were different than in Seattle (Supplementary Fig. 6), Col-0 plants grown in both locations flowered at a similar developmental timing to the ones grown in Seattle (Fig. 1h). Both samples also showed the morning expression of FT, although the afternoon FT expression levels differed (Fig. 1g and Supplementary Fig. 7). These results indicate that WT plants grown in natural LD induce FT in the morning and possibly around dusk to induce early flowering.

Reconstitution of lab growth conditions that reflect natural conditions for flowering

Our results obtained from plants grown in natural LD demonstrated that our current lab LD conditions are not sufficient to reproduce all important flowering regulation. To more precisely study these mechanisms, we adjusted our current growth conditions using FT expression patterns as a proxy for the flowering regulation in nature. We first hypothesized that the daily light intensity changes might alter the FT pattern, compared to step changes (=light on/off) under lab conditions. The light intensity changes did not drastically alter the FT pattern (or CO and TSF) (Supplementary Fig. 8a-e), indicating that light on/off conditions might be sufficient. Daily temperature changes affect FT expression patterns[19, 23]. Therefore, we analyzed the effect of daily temperature fluctuation on FT expression. When Col-0 plants were grown in lab LD with daily temperature oscillations based on the average changes that occurred around the summer solstice (Supplementary Fig. 8f, g), CO was strongly induced at the end of the night (Supplementary Fig. 8h). The afternoon FT levels (but not the TSF levels) were severely repressed by daily temperature changes (Supplementary Fig. 8i, j). These results suggest that temperature oscillation is not enough to induce FT in the morning, although it can repress FT in the afternoon.

The red to far-red (R/FR) ratio in an open field (including our outside conditions) is approximately 1 (ref [24]), but it varies from approximately 2 (our lab conditions) to 13 (ref [24]) under fluorescent lamps. Plant shade conditions (=very low R/FR ratios) highly induce the FT expression even in the morning[25, 26]. We wondered whether the morning FT expression could be induced under R/FR ratio=1 conditions. To test this, we supplemented our fluorescent lamps (R/FR=2) with dim FR LEDs to adjust the R/FR ratio to 1 (Supplementary Fig. 9a). Merely adjusting the R/FR ratio from 2 to 1 was sufficient to induce FT (and TSF) in the morning, without affecting CO expression patterns (Fig. 2a and Supplementary Fig. 9b, c). Under these conditions (named LD+FR), the levels of both morning and afternoon FT levels were higher than in lab LD (Fig. 2a). However, the afternoon FT peak was still slightly higher than the morning peak, different from the FT patterns in natural LD. We also tested whether R/FR=1 induces FT in the morning independent of photoperiod changes. We analyzed FT expression in short days (SD) with R/FR=1 (SD+FR). FT was not induced in SD+FR, similar to regular SD (Supplementary Fig. 10), implying that the morning induction of FT under LD+FR conditions is long-day specific.

Fig. 2:

Adjusting the R/FR ratio to 1 and changing the daily temperature of the lab growth conditions are sufficient to recreate the FT profiles and flowering of plants grown in natural LD.

a, b, FT expression profiles in LD and LD+FR (a), and in LD, LD+FR+temp, and outside in 2014 (b). The results represent means ± SEM (n=3 biologically independent samples). c, Flowering phenotypes of WT accessions and photoperiodic mutants in LD+FR+temp. Each box is located between the upper and the lower quartiles, and the whiskers indicate 1.5-times interquartile ranges. The thick horizontal lines in the boxes represent the median, and open diamonds represent the mean. Outliers are indicated by circles. (n≥11, * p<0.05, ** p<0.01, *** p<0.001, ns: non-significant, statistical information in Supplementary Table 3). d, Spatial expression patterns of FT in LD+FR+temp. FT:GUS plants were grown in LD+FR+temp for two weeks and harvested at ZT4 (n=4–5 independent plants, repeated twice biologically). As a comparison, the FT:GUS plants were grown in LD and harvested at ZT4 (n=5 independent plants). The staining patterns of GUS activity in the LD-grown samples harvested in ZT4 resembled those in the ones harvested at the end of the day (ZT16) (Supplementary Fig. 13), most likely due to the very stable nature of the GUS protein[63]. Scale bar=1 mm.

As temperature oscillation reduced afternoon FT levels (Supplementary Fig. 8i), we hypothesized that combining the R/FR=1 conditions with daily temperature changes may cause a similar FT pattern to that observed outside. Incorporating these two parameters in the simplified lab conditions (LD+FR+temp) was sufficient to generate similar FT expression patterns (and CO and TSF) to that observed in nature (Fig. 2b and Supplementary Fig. 11). We then analyzed the flowering time of WT accessions and photoperiodic mutants including co, ft, ft tsf and fkf1 under the LD+FR+temp conditions. WT plants and fkf1 alleles (fkf1 and fkf1–2) flowered earlier in LD+FR+temp than in lab LD (Fig. 2c). In addition, the FT expression profile in fkf1–2 was similar to that in fkf1 grown in natural LD (Supplementary Fig. 12), validating that the simplified LD+FR+temp conditions captured the major environmental parameters to recreate the FT expression patterns and flowering time responses of the plants grown in natural LD. Importantly, the co, ft single and ft tsf double mutants still showed a similar late flowering phenotype under all experimental conditions (Fig. 2c). In addition, the ft-1 tsf-1 double mutants flowered later than the ft-1 mutant (Fig. 2c), indicating that changes in the expression patterns of both FT and TSF may contribute to flowering time in LD+FR+temp.

We also analyzed whether the LD+FR+temp conditions changed the spatial expression patterns of FT to induce its morning peak. The tissue-specific GUS activity patterns in the FT:GUS plants were similar in LD, LD+FR, and LD+FR+temp (Fig. 2d and Supplementary Fig. 13), indicating that the adjustment of the R/FR ratio and temperature mainly affected the temporal expression pattern of FT.

To explore the similarities between LD+FR+temp and natural LD conditions on a whole transcriptome scale, we performed RNA-sequencing analysis using WT plants grown in lab LD, LD+FR+temp, and two years (2013 and 2014) of natural LD conditions. The samples were harvested at Zeitgeber time 4 (ZT: time after light onset) when morning FT peaks. Compared with lab LD conditions, 57 genes were consistently upregulated in the morning in two different years of natural LD and LD+FR+temp conditions (Fig. 3a and Supplementary Table 1). GO term enrichment analysis showed that genes involved in light (UV-B, FR, and R) responses were enriched (Fig. 3a and Supplementary Table 1). Among these 57 genes, only four, including FT and TSF, were identified as flowering genes, based on FLOR-ID[27] (Fig. 3b). In the downregulated genes common among the three conditions compared with the lab LD condition, environmental stress-related genes were enriched (Supplementary Fig. 14 and Supplementary Table 1). However, there were no downregulated flowering genes overlapping among the three conditions (Supplementary Fig. 14). These results indicate that FT and TSF induction levels might be the major difference important for flowering time regulation between LD and LD+FR+temp as well as natural LD conditions.

Fig. 3:

Morning induction of florigen expression occurs under both natural LD and LD+FR+temp conditions, and is a common response in wild-type accessions.

a, The upregulated genes of RNA-seq results in two-week-old samples harvested at ZT4 in 2013, 2014, and LD+FR+temp conditions compared with the ZT4 samples in LD (n=3 biologically independent samples). The GO term categories enriched in the 57 genes are shown. The p-values represent one-tail Fisher Exact Probability Values. See Supplementary Table 1 for the actual values. b, Flowering-related genes in FLOR-ID were extracted from the dataset shown in (E). c, FT expression levels in the morning (ZT4) and at dusk (ZT16) in 20 Arabidopsis WT accessions (Supplementary Fig. 15) in LD and LD+FR+temp.

Flowering time is a critical adaptive trait within WT accessions[28]. Our data showed that the generation of the FT morning peak was closely related to the early flowering phenotypes in natural LD (Fig. 1c-h). We asked whether this mechanism is widely conserved in WT accessions. To test this, we compared FT levels between morning and evening among 20 summer annual accessions[29] originating from different latitudes (Supplementary Fig. 15a) grown in lab LD and LD+FR+temp. In LD, FT levels in all accessions were significantly higher in the afternoon than in the morning (Fig. 3c and Supplementary Fig. 15b). However, in LD+FR+temp, the differences in FT levels between morning and evening were much reduced (Fig. 3c and Supplementary Fig. 15c). These results suggest that the mechanisms that induce morning FT expression in LD+FR+temp are largely conserved across Arabidopsis accessions.

Components important for flowering time regulation in nature

As our LD+FR+temp conditions reproduced FT expression profiles similar to those in natural LD (Fig. 2b), we next investigated whether any known components in the flowering and light signaling pathways are involved in the regulation of morning FT expression. Because CO is a chief activator of FT[1], we first analyzed FT expression in the co mutant. FT levels in the co mutant were very low throughout the day in LD+FR+temp, LD, and LD+FR (Fig. 4a-c), implying that CO function is essential for FT induction even under conditions more similar to the natural environment. We then analyzed FT expression patterns in photoreceptor and light signaling mutants, circadian clock mutants, and mutants in the ambient temperature flowering pathway in LD+FR+temp (Fig. 4a-f and Supplementary Fig. 16)[1, 3, 30]. Compared with WT plants, FT expression in the morning was specifically reduced in the phytochrome A (phyA) mutant (phyA-211) in LD+FR+temp (Fig. 4a). This phenotype was also pronounced in LD+FR (Fig. 4b, c). These results prompted us to analyze the flowering phenotype of the phyA mutant in natural LD and LD+FR+temp conditions. To grow the phyA mutant outside, we utilized non-transgenic phyA-201 allele (Ler background), since phyA-211 possesses a transgene[31]. phyA-201 flowered later than Ler plants outside (Supplementary Fig. 17a). Previous studies posited that a certain amount of FR light is required to observe phyA-dependent effects on flowering, as phyA mutants only showed a late flowering phenotype in LD with lower R/FR ratio in the afternoon or in continuous FR light conditions[7, 32]. In LD+FR+temp conditions, both phyA-201 and phyA-211 flowered later than their parental accessions (Supplementary Fig. 17a). These results indicate that both natural LD and LD+FR+temp conditions (R/FR=1) contain enough FR light to observe the phyA contribution to flowering induction. We also analyzed FT expression in phyA-201, and found that FT levels were lower in both morning and afternoon than Ler plants (Supplementary Fig. 17b, c). This result indicates that, although phyA is clearly involved in FT induction in the morning, its contribution to afternoon FT expression may differ in either different backgrounds and/or alleles. We further assessed the significance of the phyA signaling using the far-red elongated hypocotyl 1 (fhy1) fhy1-like (fhl) mutant (Col-0 background) in which phyA signaling is severely attenuated due to impairment of phyA nuclear transport[33]. Although the phenotype was weaker than the phyA-211 mutant, fhy1 fhl also showed a reduction in morning FT expression in LD+FR+temp (Supplementary Fig. 16a, b). This result further supports the notion that phyA signaling is involved in flowering regulation through inducing FT in natural LD.

Fig. 4:

phyA and ELF3 are involved in the regulation of morning FT expression in LD+FR+temp.

a-c, FT expression profiles in WT plants, co-101, and phyA-211 mutants in LD+FR+temp (a), LD (b), and LD+FR (c). d-f, FT expression profiles in WT plants, phyB-9, and elf3–1 mutants in LD+FR+temp (d), LD (e), and LD+FR (f). g, h, FT levels in WT plants, phyA-211, and elf3–1 mutants in LD with different R/FR ratios. The levels of FT in these plants in the morning, ZT4 (g), and at dusk, ZT16 (h). For a-h, the results represent means ± SEM (n=3 biologically independent samples). i, Daily accumulation patterns of phyA protein in LD and LD+FR+temp. j, Daily accumulation patterns of ELF3 protein in ELF3:ELF3–6H3F plants in LD and LD+FR+temp. For both (i) and (j), the representative blot images are shown. Actin was used as a loading control. The protein quantification results (relative values against the loading control) represent means ± SEM (n=6 biologically independent samples).

phyA functionally antagonizes phyB in flowering[34, 35]. In phyB mutant, FT levels were higher than in WT plants in LD+FR+temp (Fig. 4d). In early flowering 3 (elf3), which is a phenocopy of the phyB mutants[36], FT levels were even higher than in the phyB mutant (Fig. 4d-f). The difference between phyB and elf3 mutants was more pronounced in LD and LD+FR (Fig. 4d-f). These results suggest that ELF3 may regulate not only phyB signaling but also other signaling pathways important for FT induction.

Despite being downstream signaling components of phyB and ELF3[37, 38], PHYTOCHROME INTERACTING FACTOR 1 (PIF1), PIF3, PIF4, and PIF5 might not be important for FT induction in LD+FR+temp, as FT profiles in the pif1 pif3 pif4 pif5 (pifq) mutant resembled that in WT (Supplementary Fig. 16c, d). The FT levels in both constitutive photomorphogenic 1 (cop1) and the suppressor of phyA-105 1 (spa1) spa3 spa4 triple mutants were higher without changing CO mRNA patterns (Supplementary Fig. 16c, d). As COP1 and SPAs directly control CO protein degradation[39, 40], this indicates that CO protein stability regulation is still important in LD+FR+temp. In the cryptochrome 1 (cry1) cry2 double mutant, FT expression occurred just in the morning (Supplementary Fig. 16e, f). This result prompted us to analyze the flowering phenotype in LD+FR+temp. The cry1 cry2 mutant flowered significantly earlier in LD+FR+temp than in LD (Supplementary Fig. 18), which resembles the flowering phenotypes of fkf1 mutants. This result further indicates that the morning expression of FT contributes to flowering time regulation.

Circadian clock components often regulate FT expression in LD[1]. FT levels were depressed in both gi and pseudo response regulator7 (prr7) prr9 mutants in LD+FR+temp (Supplementary Fig. 16e-h). In the circadian clock associated 1 (cca1) late elongated hypocotyl (lhy) double mutant, FT levels during the afternoon were strongly increased in LD+FR+temp (Supplementary Fig. 16g, h). Based on these mutant phenotypes, our results suggest that GI and PRRs are important for the induction of FT throughout the day[22, 41], while morning clock genes, CCA1 and LHY, strongly repressed FT mainly in the afternoon[1, 42].

In the ambient temperature pathway mutants, such as short vegetative phase (svp), the triple mutant of svp flowering locus m (flm) flowering locus c (flc), and high expression of osmotically responsive genes 1 (hos1)[29, 43], the difference in morning FT levels looked greater than that in the afternoon (Supplementary Fig. 16i, j). This could be due to lower temperatures in the morning, which activate the ambient temperature pathway. In summary, based on our results, several known components are involved in morning FT expression regulation.

To further investigate the mechanisms of morning FT induction, we studied possible interactions between phyA and ELF3. Our results showed that phyA functions as an FT activator, while ELF3 is a FT repressor in the morning in LD+FR+temp (Fig. 4a, d). In addition, phyA was identified as one of the proteins co-immunoprecipitated with ELF3, indicating that phyA and ELF3 may exist in the same protein complex[44]. First, as FT levels are highly increased under lower R/FR ratios, and FT levels are the major determinants of plant shade-induced flowering timing[25, 26, 45], and because the R/FR=1 condition is enough to induce FT in the morning, we investigated the more comprehensive relationship between R/FR ratios and FT levels. Also, we analyzed whether the phyA and elf3 mutations affect FT levels in LD with different R/FR ratios. In WT plants, there is nearly a linear relationship between the decrease in the R/FR ratios and the increase in FT levels in the morning (ZT4) and the afternoon (ZT16) (Fig. 4g, h). In the phyA-211 mutant, morning FT induction was severely reduced under a wide range of R/FR ratios, while morning FT levels in the elf3–1 mutant were constantly high (Fig. 4g). In both phyA-211 and elf3–1 mutants, FT levels over different R/FR ratios stayed at nearly similar levels at ZT4 (Fig. 4g), indicating that the function of both proteins is required to tune FT levels in response to R/FR ratio changes during the morning. In the afternoon, the lack of elf3 made the plants more sensitive to the R/FR ratio changes with a large increase in FT expression under lower R/FR ratios (Fig. 4h). There was only a small phyA-211 mutation effect on afternoon FT expression levels (Fig. 4h). These results suggest that both phyA and ELF3 have time-dependent functions in light quality-controlled FT level regulation.

We next studied the genetic relationship between PHYA and ELF3 in this regulation. There was an intermediate level of FT expression in the phyA-211 elf3–1 double mutant compared to FT levels in each mutant under all conditions (Supplementary Fig. 19), indicating that phyA and ELF3 function antagonistically on FT regulation. As a biochemical study indicated the presence of a phyA-ELF3 complex[44], we examined whether our modified LD conditions influence the amount of phyA co-immunoprecipitated with ELF3 protein. In LD, phyA protein dissociated from the ELF3 complex as soon as the light was turned on, while in both LD+FR and LD+FR+temp conditions, similar amounts of phyA were co-immunoprecipitaed with ELF3 at later time points during the morning (Supplementary Fig. 20). These results suggest that the prolonged presence of the phyA-ELF3 complex in LD+FR+temp may change the expression levels and/or activity of phyA and/or ELF3 proteins.

We also analyzed the phyA and ELF3 protein expression patterns in LD+FR+temp. The accumulation levels of phyA protein in LD+FR+temp were higher during the morning than in LD, although they eventually reached trough level by the end of the day (Fig. 4i). This is likely controlled by posttranslational regulation, since PHYA transcript levels under these conditions were very similar (Supplementary Fig. 21a). In contrast, the ELF3 protein levels in LD+FR+temp were lower throughout the day than in LD (Fig. 4j). The ELF3 transcript levels in LD+FR+temp were also slightly lower than in LD (Supplementary Fig. 21b). Higher expression of FT at ZT4 in LD+FR+temp is consistent with the higher level of its activator phyA and lower levels of its repressor ELF3 under these conditions.

We further analyzed whether ELF3 affects phyA protein patterns or vice versa. In LD+FR+temp, there was no difference in PHYA levels between WT and the elf3 mutant (Supplementary Fig. 22a). The phyA protein levels were slightly higher in elf3 than in WT, although the difference was not statistically significant (Supplementary Fig. 22c). There was less ELF3 protein in phyA than in WT without affecting transcript levels (Supplementary Fig. 22b, d), suggesting that phyA may regulate ELF3 protein levels posttranscriptionally in LD+FR+temp. However, since ELF3 is a repressor of FT, the reduction of ELF3 levels in the phyA mutants cannot be the major cause of the reduction of FT levels in phyA. Previous work showed that ELF3 forms a complex with a large number of light signaling and circadian clock components, many of which require functional phyB to physically associate with the ELF3 complex[44]. In addition, more phyA protein was co-immunoprecipitated with ELF3 after dawn in LD+FR+temp (Supplementary Fig. 20). We therefore hypothesized that phyA may affect ELF3 function by directly modulating the interaction of ELF3 with other factors in a light/temperature-dependent manner.

To assess whether phyA influences the composition of the ELF3 complex, we harvested ELF3:ELF3–6H3F samples with/without the phyA mutation in the morning of LD+FR+temp conditions, and identified peptides co-immunoprecipitated with ELF3 using mass spectrometry analysis. We included ELF3:ELF3–6H3F/phyB as a reference. When we compared our peptide list of ELF3:ELF3–6H3F samples with the previous ones harvested in the afternoon[44], we noticed that our list did not contain peptides from ELF3-associated circadian clock proteins (Supplementary Table 2), suggesting that ELF3 does not assemble with the same Evening Complex in the morning. We identified peptides derived from COP1 but not SPA1. phyB is still important for ELF3 complex formation in the morning. The loss of phyA did not seem to drastically change the composition of the ELF3 complex, although fewer peptides from COP1 and phyE were detected in the phyA background (Supplementary Table 2). These results indicate that phyA may affect the interaction of a small number of components in the ELF3 complex. However, these results are not sufficient to evaluate whether those changes may affect either ELF3 function or FT transcription in the phyA mutant in LD+FR+temp. Further analysis is required to elucidate the exact mechanism by which phyA and ELF3 antagonistically regulate FT levels in LD+FR+temp.

As CO is required for the FT morning peak (Fig. 4a), and the cop1 and spa triple mutants showed increased FT levels in LD+FR+temp (Supplementary Fig. 16c, d), we hypothesized that CO protein levels may increase under these conditions. To test this, we analyzed the diurnal expression profile of CO protein in CO:HA-CO plants[46] in LD and LD+FR+temp. The overall accumulation patterns of CO protein in LD and LD+FR+temp were similar (Fig. 5a). However, CO protein increased more in LD+FR+temp than in LD at the ZT4 time point, when the FT morning peak was induced (Fig. 5a). Therefore, we analyzed a fine scale time course of CO profiles during the morning. In LD, CO protein acutely accumulated just after dawn (ZT0.5–1) but quickly degraded by ZT2 (Fig. 5b). In LD+FR+temp, CO protein levels kept increasing until ZT1, and then decreased more gradually during the morning. The levels of CO protein expressed in the morning of LD+FR+temp was similar to the levels around dusk (Fig. 5a, b), suggesting that the elevation of CO protein levels in LD+FR+temp might contribute to morning FT peak generation.

Fig. 5:

CO protein stability was increased in LD+FR+temp during the morning.

a, b, CO protein accumulation patterns in CO:HA-CO plants in LD and LD+FR+temp. Histone H3 was used as a loading control. The quantification results represent means ± SEM [n=5 (a) and n=3 (b) biologically independent samples]. c, Coimmunoprecipitation analysis of ELF3 and CO proteins. 35S:ELF3–6H3F, 35S:3HA-CO, and 35S:3HA-CO/35S:ELF3–6H3F plants were grown in LD, LD+FR (labeled as FR), or LD+FR+temp (FR+temp) and harvested in the morning (ZT4). The experiments were repeated three times independently, and similar results were obtained. d, CO protein accumulation patterns in CO:HA-CO and CO:HA-CO/elf3–1 plants grown in LD+FR+temp. The quantification results represent means ± SEM (n=5 biologically independent samples). e, f, A model for CO-dependent FT regulation under natural LD conditions. This model shows temporal expression patterns of CO protein (top) and FT transcripts (bottom) under lab LD (e) and natural LD (f) conditions. e, Under artificial lab LD conditions in which the R/FR ratio is equal to or greater than 2 and the temperature is constant, CO protein appears to immediately accumulate after light onset and then rapidly degrade, resulting in low levels of CO protein in the morning and early afternoon. During this period, ELF3 protein inhibits FT expression through an unknown mechanism. CO protein peaks again at the end of the day, which directly activates FT transcription under these conditions. f, Under natural LD conditions, the R/FR ratio is 1 and the ambient temperature oscillates throughout the day. The amount of phyA protein increases in the morning, whereas the amount of ELF3 protein decreases (Fig. 3i, j). CO protein accumulates rapidly at high levels after sunrise, and CO protein degrades more slowly under natural LD conditions than under lab LD conditions. This CO accumulation might be important for morning induction of FT. In addition to the CO protein stability changes, there might be other factors (depicted as “X”) involved in the induction of morning FT under natural LD conditions. The phyA signal is positively involved in FT induction under these conditions. ELF3 negatively acts on FT regulation under these conditions. In addition, the temperature oscillations strongly repress FT transcription in the evening. Therefore, although CO protein abundance is high even at dusk, the levels of FT expression remain relatively low around dusk compared to morning. We showed that we can recreate these FT expression profiles in the lab by simply adjusting the R/FR ratio of light source and temperature conditions.

CO protein levels are controlled by several E3 ubiquitin ligases, such as the COP1/SPA complex[39, 40]. Both phyA and ELF3 physically interact with the COP1/SPA1 complex to regulate its function[44, 47, 48]. We therefore investigated whether phyA and/or ELF3 mediate FT regulation through regulation of CO protein stability. We found that ELF3 is in the same CO protein complex in planta and in vivo (Fig. 5c and Supplementary Fig. 23). In addition, FT expression and the early flowering phenotype of elf3–1 were largely dependent on functional CO, as FT expression levels are very low in both co-101 and elf3–1 co-101, and elf3–1 co-101 flowered just slightly earlier than the late flowering co-101 mutant (Supplementary Fig. 24). Although the FT levels in the elf3–1 co-101 mutant are similar to those in the previously characterized elf3–1 co-1 mutant[49], the elf3–1 co-1 mutant showed an intermediate flowering phenotype between the co-1 and elf3–1 mutants[49]. This flowering time difference might be caused by the difference in the genetic backgrounds (elf3–1 co-1 in F3 segregants between Col-0 and Landsberg[49] cross, vs. elf3–1 co-101 in Col-0) and/or co alleles. We next analyzed whether ELF3 influences CO stability, and found that CO protein was more abundant in the elf3 mutant, including at ZT4 (Fig. 5d and Supplementary Fig. 25). Together with our results that showed lower levels of ELF3 protein in LD+FR+temp, these results indicate that ELF3 may negatively influence CO stability in the morning, and that LD+FR+temp conditions in part reduce the amount of the negative regulator to increase CO protein stability, consistent with increased expression of its target FT.

Discussion

Our results indicate that the difference in R/FR ratios and daily temperature are the main causes of the difference in flowering time between natural LD and lab LD conditions. Mechanistically, this difference in growth conditions likely causes different expression levels of the florigen genes, FT and TSF, especially in the morning. Previous work indicated that FT induced between ZT12 and ZT20 in SD was more effective for floral induction than FT induced during other time windows[21]. How does morning-expressed FT affect flowering compared to evening-expressed? The uploading of FT proteins into the phloem and the unloading of them into the shoot apical meristem are actively regulated, at least in cucurbit plants[50, 51]. Phloem flux and the concentration of major transport sugars in phloem sap exhibit diurnal and developmental changes in some plants[52, 53, 54]. Therefore, the efficiency of florigen movement may change depending on growth conditions, time of day, and plant age. Although it is beyond the scope of our current research, it would be of interest to assess whether the timing of FT expression during the morning has some mechanical advantages compared to evening in natural LD.

Our results indicate that phyA and ELF3 are involved in the regulation of the morning expression of FT in natural LD (Fig. 5e, f). In addition, CO protein is likely more stable in the morning of natural LD than in regular lab LD. This may contribute to higher induction of FT in the morning. However, as CO protein interacts with several other transcription factors to regulate FT in the morning[55], we assume that there are still other factors that participate in controlling FT levels in natural LD morning (Fig. 5e, f). We therefore think the findings presented here are a starting point to understanding the mechanisms of the previously uncharacterized florigen induction that takes place in natural LD.

With the external coincidence model for explaining photoperiodic response as a basis[56], molecular mechanisms that consist of complex interplay between light signaling and the circadian clock have been proposed to explain LD specific dusk FT expression[1]. Although our results indicate the involvement of some known flowering regulators in morning FT induction, the current model cannot explain how these factors induce FT in the morning in natural LD. Investigating these mechanisms will help us to understand how Arabidopsis plants flower in spring in nature.

Behavioral rhythms in model animals (Drosophila, mouse, and golden hamster) differed between natural and lab conditions[57, 58, 59]. In Drosophila, transcriptional levels of clock genes were altered between these two conditions[60]. Even for Arabidopsis, previous work reported discrepancies in predicted flowering phenotypes when flowering mutants were grown outside[13], although the molecular mechanism that caused this was unknown. Based on our work, the discrepancies might be partly caused by the difference in light quality and temperature between lab and natural conditions. To understand plant response at molecular levels in nature, recent functional genomic approaches in molecular ecology have successfully revealed certain mechanisms by which plants sense specific environmental stimuli in complex natural environments[61, 62]. However, these approaches still have geographical and environmental limitations. Our approach for optimizing simplified lab conditions based on plant response in nature will be widely feasible. Studying plant responses under refined lab conditions that more closely reflect natural conditions will likely fill the current gap between genetics and ecology and facilitate interdisciplinary communication between them in order to more holistically understand the underlying mechanisms of ever-changing phenological response in plants.

Methods:

Plant materials and growth conditions.

Except where indicated, all Arabidopsis thaliana plants, wild type (WT), fkf1(ref [17]), fkf1–2 (ref [17]), gi-2 (ref [22]), ft-1 and ft-1 tsf-1 (ref [64]), ft-101 and co-101 (ref [65]), phyA-211 (ref [31]), phyB-9 (ref [66]), elf3–1 (ref [67]), fhy1–3 fhl-1 (ref [33]), cop1–6 (ref [68]), spa1–3 spa3–1 spa4–1 (ref [69]), pif1 pif3 pif4 pif5 (pifq)[70], cry1 (hy4–2.23N) cry2–1 (ref [71]), cca1–1* lhy-Null (ref [72]), prr7–11 prr9–10 (ref [73]), hos1–3 (ref [43]), svp-32 and svp-32 flm-3 flc-3 (ref [29]), ELF3:ELF3–6H3F and ELF3:ELF3–6H3F/phyB-9 (ref [44]), CO:HA-CO and 35S:3HA-CO[46], and FT:GUS [65], used in this study are in the Columbia-0 (Col-0) background. phyA-201 is in the Landsberg erecta (Ler) background[31]. The ft-1 and ft-1 tsf-1 seeds were kindly provided by Dr. Motomu Endo. The fhy1–3 fhl-1 seeds were kindly provided by Dr. Mathias Zeidler. The cop1–6 seeds were kindly provided by Dr. Xing Wang Deng. The spa1–3 spa3–1 spa4–1 seeds were kindly provided by Dr. Ute Hoecker. The cca1–1* lhy-Null seeds were kindly provided by Drs. Rachel Green and Stacy Harmer. The prr7–11 prr9–10 seeds were kindly provided by Dr. Takafumi Yamashino. The svp-32 and svp-32 flm-3 flc-3 seeds were kindly provided by Dr. Ji Hoon Ahn. Wild-type Arabidopsis accessions, Oy-1, RLD-1, Mh-0, An-1, Nos-0, Ma-1, Rd-0, Nd-1, En-1, Jl-3, Kz-9, Di-G, Wei-0, Ka-0, Sei-0, Mt-0, and Van-0 were all obtained from the ABRC stock center at Ohio State University. The phyA-211 elf3–1 double mutant was generated by a genetic cross between phyA-211 and elf3–1. The co-101 elf3–1 double mutant was generated by a genetic cross between co-101 and elf3–1.

To generate 35S:ELF3–6H3F transgenic lines, the pENTR/D-TOPO vector harboring the full length of ELF3 cDNA without a stop codon[44] was transferred to pB7HFC binary vector[44]. The 35S:ELF3–6H3F construct in pB7HFC was transformed into elf3–1 plants. For ELF3:ELF3–6H3F/phyA-211 lines, the ELF3:ELF3–6H3F construct in pK7HFC vector[44] was introduced to phyA-211 plants. To generate 35S:3HA-CO/35S:ELF3–6H3F lines, the 35S:3HA-CO construct in pH7WG2 (ref [46]) was transformed into the 35S:ELF3–6H3F line. The transgenic plants were selected based on the expression levels of both CO and ELF3 genes. For CO:HA-CO/elf3–1 lines, the CO:HA-CO construct in pPZP221 binary vector[46] was transformed into elf3–1 plants.

All plants were grown either on soil in standard flats with inserts (STF-1020-OPEN and STI-0804, T.O. Plastics; for plants grown in Seattle or similar flats/inserts for plants grown in Zürich and Edinburgh) or in sterile 1X Linsmaier and Skoog (LS) agar media (Caisson) without sucrose. The soil (Sunshine Mix #4, Sun Gro Horticulture) contained a slow release fertilizer (Osmocote 14–14-14, Scotts Miracle-Gro) and a pesticide (Systemic Granules, Bonide). After seeds were sown onto soil or growth media, they were stratified in a 4°C room for at least three days and then transferred to outside growth areas or growth chambers. Only non-transgenic plants were used for the outdoor experiments, following institutional, national, and international restrictions on handing genetically modified organisms (transgenic plants were only used in certified lab settings). For outside experiments, the flats containing stratified seeds were transferred onto a platform in a low tunnel equipped with a shading filter in our caged plant growth areas (University of Washington, University of Zürich, and University of Edinburgh). To avoid shading effects from neighboring plants, seeds were sown at a low density, and when necessary, younger seedlings were thinned to let individuals grow separately. To prevent potential light stress from excess direct sunlight exposure (which can be stronger than 1000 μmol/m[2]/s), the cage was covered with double layers of Reemay Garden Blanket (Reemay) to reduce sunlight intensity without changing the red/far-red ratio (R/FR). The R/FR ratio was measured by LightScout Red/Far Red meter (Spectrum Technologies), as well as UV-VIS Spectrometer (StellarNet Inc). The light intensity changes around the summer solstice were measured using LI-250A light meter (LI-COR). Temperature was directly monitored by HOBO Pendant Temperature/Light 64K Data Loggers (Onset) for Edinburgh as well as Seattle. Air temperature was obtained from nearby weather station data, https://www.ed.ac.uk/geosciences/weather-station/weather-station-data for Edinburgh, http://www.meteoswiss.admin.ch/home/research-and-cooperation/nccs.html for Zurich, and http://www-k12.atmos.washington.edu/k12/grayskies/nw_weather.html for Seattle, as shown in Supplementary Fig. 1. Information regarding sunrise time (ZT0) and day length in Seattle was obtained from http://aa.usno.navy.mil/data/docs/RS_OneYear.php.

Normal lab LD and SD conditions were described previously[46]. For FR light supplement in LD+FR and LD+FR+temp conditions, weak 730 nm far-red LED light (RAY “PfrSpec”, Fluence Bioengineering, previously referred to as BML Horticulture) was provided together with full-spectrum white fluorescent light (F017/950/24”, Octron Osram Sylvania) in order to set the R/FR=1. To obtain dim far-red light, we used a dimmer (Fluence Bioengineering/BML Horticulture) with the LED light source and also wrapped the LED light with a single layer of regular white copy paper. The R/FR ratio was adjusted using LightScout Red/Far Red meter (Spectrum Technologies), and confirmed using UV-VIS spectrometer.

To apply LD+light intensity conditions, light intensity changes during the daytime shown in Supplementary Fig. 8a were set based on averages of three day-long light intensity measurements shown in Fig. 1a. Specific settings in growth chambers were as follows: ZT0, 0 μmol/m[2]/s; ZT1, 29 μmol/m[2]/s; ZT4, 84 μmol/m[2]/s; ZT7, 173 μmol/m[2]/s; ZT10 148 μmol/m[2]/s; ZT13, 81 μmol/m[2]/s; and ZT16, 0 μmol/m[2]/s. The light intensity between two settings was gradually changed in a ramping mode. The light intensity changes in the chamber were confirmed using LI-250A light sensor (LI-COR).

For temperature fluctuation settings in LD+temp and LD+FR+temp conditions, temperature data for seven days around the summer solstice of 2013, from June 21st to 27th, in Seattle was obtained from a website (http://www-k12.atmos.washington.edu/k12/grayskies/nw_weather.html) and averaged. Based on the average temperature data, a multi-step program shown in Supplementary Fig. 8g was set in ramping mode as follows: ZT0, 15.9°C; ZT2, 16.2°C; ZT4, 17.9°C; ZT5.8, 19.4°C; ZT9, 22.2°C; ZT10.4, 22.6°C; ZT11, 22.8°C; ZT12, 22.1°C; ZT16, 19.8°C; ZT17, 18.3°C; ZT20, 16.6°C; ZT23, 15.8°C. The temperature changes were confirmed using HOBO Pendant Temperature/Light 64K Data Loggers (Onset).

Flowering time was measured by the number of rosette and cauline leaves on the main stem when inflorescence reached 1–5 cm high as described previously[46]. Flowering time experiments were performed with 12 individual plants at a minimum. All flowering time results in this manuscript are means ± standard errors of means (SEM).

RNA preparation and gene expression analyses.

For gene expression analyses, 14-day-old seedlings grown on soil (all outside grown samples) or LS agar plates (samples grown in the incubators) were harvested every 3 hours during a 24-h period and were used for RNA extraction. RNA extraction, cDNA synthesis, Q-PCR conditions and normalization by IPP2+PP2A were described previously[74]. Primers and PCR conditions for CCA1, LHY, PRR9, PRR7, PRR5, TOC1, CDF1, FKF1, GI, CO, FT, FLC, SVP, ISOPENTENYL PYROPHOSPHATE/DIMETHYLALLYL PYROPHOSPHATE ISOMERASE (IPP2), and SERINE/THREONINE PROTEIN PHOSPHATASE 2A (PP2A) were previously described[18, 19, 74, 75]. All expression results were normalized using averages of IPP2 and PP2A values. The remaining primer sequences used for analyzing gene expression profiles are the following: 5’-GCACAGACTGATTAAGGTTCAAAAAC-3’ and 5’- CTTCACTGGATAGCTTTTAGCAG-3’ for ELF3; 5’- AATCTAGAGATCAGGTTAACGC-3’ and 5’-CTTCTTCTGACACATCTTCCT-3’ for PHYA; 5’-CTCGGGAATTCATCGTATTG-3’ and 5’-CCTCTGGCAGTTGAAGTAAG-3’ for TSF. Q-PCR for CCA1, LHY, PRR9, PRR7, PRR5, TOC1, CDF1, GI, SVP, and IPP2 was done using the following program: 1 min at 95°C, followed by 40–50 cycles of 10 sec at 95°C and 20 sec at 60°C. Q-PCR for FKF1, CO, FT, TSF, ELF3, PHYA, and PP2A was done using the following program: 1 min at 95°C, followed by 40–50 cycles of 10 sec at 95°C, 15 sec at annealing temperature, and 15 sec at 72°C. Annealing temperature for each primer set was 55°C for FKF1 and CO, 64°C for FT, 59°C for TSF, 61°C for ELF3, 64.3°C for PHYA, and 64°C for PP2A.

Whole transcriptome RNA-sequencing (RNA-seq) analysis.

Wild-type plants were grown on soil under LD, LD+FR+temp and natural LD conditions and harvested at ZT4 on day 14. The “2013 outside” samples were harvested on 6/25/13, and the “2014 outside” samples were harvested on 7/7/14. After mRNA was purified using NEB Next Poly(A) mRNA magnetic isolation kit (New England Biolabs), RNA-seq libraries were prepared using the YourSeq 3’-Digital Gene Expression RNAseq Library Kit (Amaryllis Nucleics). A Bioanalyzer 2100 (Agilent, High Sensitivity DNA Kit) was used for library quality control, to determine average library size, and together with concentration data from a Qubit 2.0 Fluorometer (Life Technologies, dsDNA High Sensitivity Assay Kit) to determine individual library molarity and pooled library molarity. Pooled libraries were sequenced on a NextSeq 500 (Illumina, High Output v2 75 cycle kit) to yield single-read 80 bp reads.

FASTQ sequence files were preprocessed in two steps. A Python library (clipper.py, https://github.com/mfcovington/clipper) was used to trim off the first 8 nucleotides of each read to remove potential mismatches to the reference sequence caused by annealing of a random hexamer required for library synthesis. Trimmomatic v0.36; http://www.usadellab.org/cms/?page=trimmomatic] was used to remove adapter sequences and trim or filter reads based on quality. The parameters used for Trimmomatic were “ILLUMINACLIP:TruSeq3-PE-2.fa:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:50”.

Preprocessed reads were mapped to the Arabidopsis thaliana TAIR10 cDNA reference sequence (ftp://ftp.ensemblgenomes.org/pub/plants/release-34/fasta/arabidopsis_thaliana/cdna/Arabidopsis_thaliana.TAIR10.cdna.all.fa.gz) using bowtie2 with the "--norc" parameter to enforce strand-specific alignment. Read counts for each transcript in the cDNA reference were calculated using a Perl script (simple_counts.pl, https://github.com/mfcovington/read_counter).

The R package edgeR[76] was used to identify differentially expressed transcripts between samples grown in lab LD conditions and samples grown in LD+FR+temp conditions, outdoor samples from 2013 and 2014. Transcripts were retained for analysis if they had more than two counts per million in at least three samples. After normalization factors were calculated and dispersion estimated, pairwise comparisons were performed using edgeR’s exact test. Differentially expressed genes were then filtered using a false discovery rate (FDR) cutoff of 0.05 and a minimum log2 fold change of 1. FDRs were calculated by adjusting P-values for multiple comparisons using the Benjamini–Hochberg procedure[77].

Differential gene expression results were annotated using TAIR10 gene and transcript descriptions (https://www.arabidopsis.org/download_files/Genes/TAIR10_genome_release/gene_description_20131231.txt.gz). Gene ontology analysis was performed using DAVID[78].

GUS staining.

For histochemical staining of GUS activity for tissue-specific expression of the FT gene, 14-day-old FT:GUS plants grown under LD, LD+FR and LD+FR+temp conditions were harvested either at ZT4 (LD+FR and LD+FR+temp grown samples) or at both ZT4 and ZT16 (LD grown samples), and immediately treated with 90% pre-chilled acetone on ice for 10–15 min to fix and extract chlorophylls. After washing three times with 100 mM Na-phosphate pH7.0, whole plant tissues were submerged in the staining solution (100 mM Na-phosphate pH7.0, 10 mM EDTA, 0.5 mM Potassium ferricyanide, 0.5 mM Potassium ferrocyanide, 0.1% Triton X-100, and 1 mM X-Gluc). After 4-hour staining, the tissues were washed and dehydrated with ethanol series 30%, 50%, 80%, and 100%.

Tandem affinity purification coupled mass spectrometry (TAP-MS) analysis.

Fourteen-day-old ELF3:ELF3–6H3F, ELF3:ELF3–6H3F/phyA-211 and ELF3:ELF3–6H3F/phyB-9 lines grown on LS agar plates under the LD+FR+temp conditions were harvested at ZT4. Procedures for Tandem FLAG and His-immunoprecipitations (IP), protein digestion, and liquid chromatography-tandem mass spectrometry (LC-MS/MS) were performed according to Huang et al.[44].

Statistical analysis.

Statistical analyses for flowering time experiments were done using R Statistical Computing software (v3.2.3; R Core Team, 2015). The effect of conditions on flowering time was tested using linear models (lm) when the assumptions were met in ‘gvlma’ function in the ‘gvlma’ package. When the assumptions were not met, generalized linear models (glm) with poisson error distribution (faily=poisson) were used. For more than two groups, pairwise comparisons were conducted with Tukey’s multiple comparisons adjustment using ‘glht’ function in the ‘multcomp’ package.

Immunoblot analysis and protein quantification.

For analyzing diurnal expression profiles of phyA, ELF3, and CO proteins, 14-day-old Arabidopsis seedlings grown on LS agar media under LD or LD+FR+temp conditions were harvested at each time point, frozen in liquid nitrogen, and stored at −80 ºC. Total proteins were extracted using extraction buffer [50 mM Na-phosphate pH7.4, 100 mM NaCl, 10% glycerol, 5 mM EDTA, 1 mM DTT, 1% NP-40, 0.5% SDS, 0.5% Sodium deoxycholate, 50 μM MG-132, 2 mM NaVO4, 2 mM NaF, and protease inhibitor tablets-EDTA free (Pierce)], and nuclei samples were prepared using CelLytic Plant Nuclei Isolation/Extraction Kit (Sigma) based on manufacturer protocol.

To detect proteins, total protein extract for phyA and ELF3, or nuclear extract for CO, were resolved in 9% or 11–12% SDS-PAGE gels, respectively, and transferred to nitrocellulose membranes (Bio-rad). phyA, ELF3–6H3F, and HA-CO proteins were detected using a monoclonal anti-phyA antibody[79] kindly provided by Dr. Akira Nagatani, anti-FLAG (A8592, Sigma), and anti-HA (3F10, Roche) antibodies. Actin or Histone H3 proteins were used for internal loading controls of total protein or nuclear extract, respectively, and detected by anti-actin (C4, Millipore) and anti-histone H3 (ab1791, Abcam) antibodies, respectively.

For protein quantification, immunoreactive proteins on immunoblotted membranes were visualized with SuperSignal West Pico Luminol/Enhanced Solution (Thermo) and/or ECL Select Western Blotting Detection Reagent (Amersham) and imaged by ChemiDoc Touch (Bio-rad). The image was used for quantification with the Image Lab program (Bio-rad). Relative protein abundance was normalized against Actin or Histone H3.

Co-immunoprecipitation experiments.

To analyze in vivo interactions, the ELF3:ELF3–6H3F, 35S:ELF3–6H3F[44], 35S:3HA-CO, and 35S:3HA-CO 35S:ELF3–6H3F lines grown under LD, LD+FR, or LD+FR+temp conditions were harvested at ZT0, ZT2, or ZT4 on day 14, frozen in liquid nitrogen, and stored at −80 ºC. For analyzing in planta interactions, the 35S:ELF3–6H3F, 35S:3HA-CO[46], and 35S:CO-TAP[46] constructs were infiltrated into 3-week-old Nicotiana benthamiana leaves as described[46].

The method for coimmunoprecipitation (Co-IP) assays was described previously[46]. Briefly, proteins were extracted from 1 ml volume of ground tissues using Co-IP buffer [50 mM Na-phosphate pH7.4, 150 mM NaCl, 10% glycerol, 5 mM EDTA, 1 mM DTT, 0.1% Triton X-100, 50 μM MG-132, 2 mM NaVO4, 2 mM NaF, and protease inhibitor tablets-EDTA free (Pierce)] and incubated with Protein G-coupled magnetic beads (Dynabeads Protein G, Invitrogen) that captured anti-FLAG (F1804, Sigma) antibody at 4 ºC for 10 minutes under dim light. After washing three times, precipitated proteins were eluted with 2X SDS sample buffer at 80 ºC for three minutes. Fifty percent of the eluted proteins and 1.5% of the total extract as an input were resolved in 9% SDS-PAGE gels. ELF-6H3F and endogenous phyA proteins were detected by western blot using anti-FLAG (Sigma) and anti-phyA antibodies, respectively.