Variation in shade-induced flowering in Arabidopsis thaliana results from FLOWERING LOCUS T allelic variation

Plants have evolved developmental mechanisms to ensure reproduction when in sub-optimal local environments. The shade-avoidance syndrome is one such mechanism that causes plants to elongate and accelerate flowering. Plants sense shade via the decreased red:far-red (R:FR) ratio that occurs in shade. We explored natural variation in flowering behavior caused by a decrease in the R:FR ratio of Arabidopsis thaliana accessions. A survey of accessions revealed that most exhibit a vigorous rapid-flowering response in a FR-enriched environment. However, a subset of accessions appeared to be compromised in the accelerated-flowering component of the shade-avoidance response. The genetic basis of the muted response to FR enrichment was studied in three accessions (Fl-1, Hau-0, and Mir-0). For all three accessions, the reduced FR flowering-time effect mapped to the FLOWERING LOCUS T (FT) region, and the FT alleles from these accessions are expressed at a lower level in FR-enriched light compared to alleles from accessions that respond robustly to FR enrichment. In the Mir-0 accession, a second genomic region, which includes CONSTANTS (CO), also influenced flowering in FR-enriched conditions. We have demonstrated that variation in the degree of precocious flowering in shaded conditions (low R:FR ratio) results from allelic variation at FT.


Introduction
Plants have evolved the ability to alter morphology and the timing of flowering as an adaptation to changes in the local environment. One such mechanism is the shade-avoidance response, which allows plants to compete with neighboring vegetation for sunlight. When challenged with a shaded environment, plant stems and petioles elongate in an attempt to grow above the canopy and capture more light. Moreover, shaded plants often reproduce earlier than non-shaded plants to ensure that at least some limited seed set occurs in the suboptimal environment (for review see [1,2]).
Changes in light quality are perceived by the photoreceptors including PHYTOCHROMES A-E (PHYA-E), CRYPTOCHROMES 1 and 2, and PHOTOTROPINS 1 and 2 (for review see [3]). In shade, the ratio of red (R) to far-red (FR) light is decreased because neighboring foliage absorbs R light, but reflects FR light. PHYB is the primary phytochrome that elicits a response to decreasing R:FR ratio, with PHYD and PHYE having mostly redundant roles with PHYB in light perception [4]. Natural variation in PHYB can result in different responses to changes in the ratio of R:FR light, providing an avenue for local adaptation [5]. The active form of PHYB elicits a response to changes in light quality by entering the nucleus and affecting transcription of a number of genes [6].
The circadian clock and phytochromes are part of the system that monitors day-length and mediates changes in the expression of gene networks. The circadian clock regulates GIGANTEA expression, and GIGANTEA is required for CONSTANS (CO) expression [7]. CO expression oscillates during a 24 hour period, and stability of the CO protein is light dependent, with degradation occurring in the dark [8]. CO promotes flowering by activating expression of downstream floral integration genes, such as FT and SUPPRESSOR OF CONSTANS 1, and FT expression is a key inducer of flowering [9]. Thus, both light duration and quality, perceived by phytochromes, influence the timing of reproductive development.
FT was initially genetically identified by Koornneef et al. [10] as a delayed-flowering mutant. The FT gene, more recently called FLOWERING LOCUS T, was cloned in 1999 (11). FT is activated by CO in long-day photoperiods in long-day plants, such as Arabidopsis [11]. FT is now recognized as "florigen," a mobile signal that promotes flowering (for review see [12]). The ancient CO/FT module is highly conserved among plant species, and facilitates several developmental responses, such as the onset of dormancy, in response to changes in photoperiod, an annually consistent environmental cue [13][14][15].
The control of FT expression is an integration point at which multiple environmental signals allow plants to fine tune flowering time as a function of environmental conditions. In addition to photoperiod, FT expression is influenced by vernalization (extended cold), ambient growth temperature, and hormone levels [16][17][18][19][20]. Recent work has demonstrated conserved regions within the FT promoter that are likely to be key determinants for expression via binding of different proteins or protein complexes [21,22]. FLC (FLOWERING LOCUS C), a potent floral repressor silenced by vernalization, as well as the repressor SHORT VEGE-TATIVE PHASE, bind directly to the FT promoter [23,24]. CO and GA (gibberellic acid) responsive proteins activate expression of FT [9,25]. If there is allelic variation at FT it is likely to be a result of cis-regulatory element variation, because the FT amino acid sequence is quite highly conserved among Arabidopsis accessions that are highly variable for flowering time.
In this study, we evaluate natural variation for flowering time in FR-enriched conditions that mimic shade. The majority of Arabidopsis accessions flower rapidly in these conditions, with the exception of two groups of accessions, which appear to be "blind" to the increased FR light. Previous work has demonstrated that FR enrichment overrides the repressive effect of FLC and alters the timing of expression of GI and CO [26][27][28][29]. Here we find FR responsiveness maps to FT and is correlated with FT expression levels. For one group, FT is the only locus identified as causing FR insensitivity, whereas for a second group, both FT and CO are candidate genes.

Growth conditions
All growth-chamber experiments were conducted in reach-in chambers (Percival Scientific I-60LX) fitted with T12 fluorescent bulbs (Philips Lighting; R:FR approximately 5, PPFD approximately 60-70 umol m -2 s -1 ) set for long days (LD; 16h of light/8h of dark) or short days (SD; 8h of light/16h of dark). All experiments were conducted at 22C, except for the cycling experiments in which the 16h light period was at 22C and the 8h night period was at 16C. Certain chamber shelves also contained arrays of FR LEDs (Plasma Ireland; wavelengths approximately 735-740 nm) that lowered the R:FR ratio from 5 to 0.15 (low FR) or 0.04 (high FR). Vernalization treatment was done in soil in a coldroom in the dark at 4C. Plants grown outdoors were first grown in growth chambers for 3 weeks and then moved outside to a partially shaded (70%) location in Madison WI beginning in early June. All light measurements were made with a wideband portable spectroradiometer (International Light; RPS900-R).
Plants were phenotyped by counting Rosette Leaf Numbers (RLN) on the primary shoot prior to flowering, with the exception of plants grown outdoors where Days to Flowering (DTF) was used due to excessive growth of secondary shoots. Genotypes were determined by SSLP or CAPS markers using TaKaRa ExTaq RR001A.

Primers, markers, and strain polymorphisms
See S6 Table. FT expression Plants were genotyped after one week of growth in LD conditions. Plants heterozygous for FT were grown for 4 weeks in LD and then shifted to LD + FR or kept in LD conditions.
To monitor FT expression with and without FR enrichment, we carried out Semi-quantitative RT-PCR of allele-specific transcripts. Total RNA was isolated from each accession, and cDNA was generated [33]. We used FT specific primers, 5'-CTCAGGAACTTCTATACTTTGG TTATG-3' and 5'-CTGACAATTGTAGAAAACTGCG-3' to amplify FT via 28 PCR cycles. The loading control of each reaction was carried out using UBQ10 primers, 5'-CTACCGTGA TCAAGATGCAGATC-3' and 5'-TTGTCGATGGTGTCGGAGCTTTC-3'. To check allele-specific expression, we monitored FT expression and its digestion pattern in various heterozygous accessions. To check FT expression in Col FRI/Fl-1, we amplified a region of FT using a polymorphism between the two accessions (BsmBI does not cut-Col FRI, BsmBI cuts Fl-1). Similarily, FT expression in the other heterozygous accessions was monitored and digested with Cac8I (no Cac8I site: Fl-0 and Mir-0, Cac8I cuts Hau-0 and Col FRI) using primer set (5'-CAATCAACACAGAGAAACCACCTG-3' and 5'-CTCGCGAGTGTTGAAGTTCTG -3').

Surveying natural accessions for FR flowering responsiveness
Our studies of late-flowering, vernalization-insensitive accessions revealed that most of these accessions flower relatively rapidly when exposed to FR-enriched light. However, a small subset does not exhibit accelerated flowering with FR enrichment. To characterize the FRflowering phenotype and investigate how it relates to other known pathways that accelerate flowering, we grew 36 accessions under 5 conditions, including FR enrichment. The accessions (S1 Table; Experiment 1), mostly winter annuals, were biased towards those that respond poorly to a standard vernalization treatment (40 days at 4˚C). As expected many accessions did not respond with greatly accelerated flowering after vernalization, but only a small subset of the accessions did not respond to FR-enrichment with accelerated flowering (Fig 1; Experiment 1). This experiment demonstrates that the vernalization and FR response were distinct in . The x-axis shows Arabidopsis accessions (S1 Table) sorted according to flowering time in LD. Experiment 1. Initial screen for flowering behavior using one or two plants per accession. FR was on for 24hrs, white light was on for 16hrs. Outdoors experiments were conducted in a partially (70%) shaded plot. All the y-axes are from 0-80RLN, except for FR, which is from 0-20RLN. Experiment 2. Secondary screen for flowering behavior. promoting flowering (r = 0.406), and thus may be acting through different floral-promoting pathways. Likewise, treatment with GA was relatively loosely correlated with the FR flowering phenotype (r = 0.571). However, the FR phenotype did correlate well with flowering time recorded outside at a site shaded by foliage approximately 70% of the day (r = 0.878; Fig 1).
To further characterize the FR phenotype, we characterized another set of 36 accessions (Fig 1; Experiment 2). Eight accessions in this set were vernalization insensitive, flowering with more than 30 rosette leaves even after an extended cold exposure (70 days at 4˚C). The vernalization phenotype was poorly correlated with the FR phenotype (Fig 1; Experiment 2 (r = 0.19)). However, once again the FR phenotype correlated well with outdoor flowering time (r = 0.941). Interestingly, combining vernalization and FR treatments caused all accession to flower with fewer than 20 rosette leaves (average = 11.2 leaves), demonstrating a synergistic relationship between the two conditions, consistent with FR and vernalization acting through separate floral initiation pathways.
To assay the effect of photoperiod in promoting flowering, we characterized another set of 36 accessions (S1 Table; SD), biased for spring annuals, in short days (SD) +/-FR (Fig 2). The addition of FR had no significant effect on flowering time (SD average = 37.07 RLN (Rosette Leaf Number); SD + FR average = 35.74 RLN), demonstrating that the FR response is acting through the photoperiod pathway, which is inactive in SD conditions. We also assayed a set of 43 accessions (S1 Table; Cyc) for flowering time under realistic light (22C) and dark (14C) temperature cycling conditions +/-FR. Most accessions flowered earlier with the addition of FR, demonstrating that temperature cycling did not eliminate or enhance the FR effect, suggesting that the two treatments operate through separate pathways, and that the cycling temperature regime we used does not substitute for FR (Fig 2). In fact, temperature cycling had little effect on flowering time in this experiment as demonstrated by the high correlation coefficient in LD and LD + cycling (r = 0.931).
All the above experiments indicate that the FR effect on flowering is dependent on the photoperiod pathway, and the high correlation with flowering time recorded in a partially shaded site outdoors demonstrates that the FR flowering phenotype we observed is not a laboratory anomaly, but is likely to be ecologically relevant.

Physiology and genetics of FR responsiveness
To further explore the genetic and molecular components underlying the FR effect on flowering, 10 accessions representing much of the flowering-time variation observed in the initial survey experiments, plus the laboratory control strains (Col and ColFRI), were chosen for further study and grown in 10 different conditions (S2 Table). Fig 3 shows the phenotypes of the accessions in the 5 most relevant growth conditions, and correlations between all conditions are shown in S3 Table. As found in the previously described experiments, the flowering phenotypes in FR-enriched conditions were poorly correlated with flowering time after vernalization or with exogenous GA application, and GA largely eliminates the photoperiod effect on flowering in these accessions (S1 Fig). We tested two different intensities of FR, high (R:FR Flowering time, shade, and FLOWERING LOCUS T ratio = 0.04) and low (R:FR = 0.16). Some accessions responded robustly to high intensity FR, but not to low intensity FR. Based upon the phenotypes, we were able to group the 10 accessions, plus the laboratory controls Col and ColFRI, into 5 groups (Table 1). Accessions were either vernalization sensitive or insensitive, while FR response fell into three categories, sensitive, semi-sensitive, and insensitive.

Characterization and mapping of FR sensitivity in segregating populations
To determine the genetic basis of the FR-flowering time differences, crosses were performed among the 10 accessions and controls with different FR-flowering behaviors.   Evaluation of F2 populations derived from crosses between FR semi-sensitive and insensitive strains to ColFRI demonstrated tight linkage of FR sensitivity to the region of chromosome 1 containing FT. For some crosses, multiple independent F2 populations were analyzed, and the results were reproducible (data not shown). As already discussed, the association of the FT region and FR-enriched flowering in the Hau/ColFRI and the Fl-1/ColFRI F2 populations is strongly supported, whereas the FLC/CO region shows no support. In addition, for crosses between Sue-0 (semi-sensitive) and Et-0 (semi-sensitive) to ColFRI, flowering time was linked in the F2 to the FT region (p = 0.001, p = 0.074, respectively). However, one population, Mir-0/ColFRI, has relatively lower support at the FT region and significant support for the FLC/CO region in comparison to the other F2 populations, indicating natural variability in which locus/loci confers FR sensitivity (this population is discussed in more detail below). In addition, an F2 population from a FR insensitive accession (Mir-0) and a FR semi-sensitive accession (Sue-0) had significant linkage between FR-enhanced flowering and the FT region (p = 0.02), as would be expected. Likewise, when Fl-1 (insensitive) was crossed to Hau-0 (semi-sensitive), the F2 FR flowering phenotype was dependent on the FT region (p = 0.01).

The role of CO in FR sensitivity
The Mir-0 accession displays some unique behavior compared to the other FR-insensitive accessions. First, it is not extremely late flowering without FR (50 RLN), and flowers with approximately the same number of leaves as with or without FR enrichment when grown side by side. Second, Mir-0 is extra sensitive to vernalization, with even short cold treatments causing a reduction in flowering time. To pursue in more detail the behavior of this accession, we grew 96 F2 plants of Mir-0/ColFRI in FR-enriched conditions. The earliest and latest 10 plants were genotyped at a tightly linked FT marker (F5I14). The earliest plants were enriched for ColFRI alleles (17/20) and the latest plants were enriched for Mir-0 alleles (15/20). Heterzygous plants were present in both the earliest and latest classes, suggestive of an additional causative locus. Using progeny lines fixed for FT (ColFRI) and FLC (ColFRI), but heterozygous for CO we observed a differential response to FR flowering attributable to the CO region (p-value <0.0001). Thus, the almost complete FR insensitivity of Mir-0 is likely to be caused by natural variation at both FT and CO. A second accession, Br-0, was also different from the other FR insensitive accession, and displayed identical flowering behavior as Mir-0. Both accessions are also glabrous, suggesting possible relatedness. Thus Br-0 was included in the co and ft mutant analysis described below.

FR sensitivity is decreased in FT and CO reduced-function alleles
Given the compelling data for a role of FT in FR sensitivity, we evaluated flowering time in an F2 population from a Col ft mutant crossed to ColFRI (Fig 5), and the results clearly demonstrate that FT is necessary for accelerated flowering in FR-enriched conditions. In plants containing one or two copies of the FRI gene, homozygous ft plants did not flower in this experiment (80+ RLN), whereas one functional copy of FT was sufficient for FR sensitivity and relatively early flowering (<28 RLN). We also evaluated the co mutant in the ColFRI background and it also flowered extremely late in FR-enriched conditions, nearly identical to the ft mutant in the same genetic background (data not shown).
To further our understanding of the role of FT and CO in FR responsiveness, we crossed multiple accessions to ft or co mutants in the ColFRI background and phenotyped the F1 plants for FR sensitivity (S5 Table). It is important to note that interpretation of results for flowering time from these F1 populations can be complicated by the introduction of "functional" dominant FT and/or semi-dominant CO alleles. Mir-0 and Br-0 when crossed to Col-FRI(ft-1) flowered substantially later (20+ RLN) than the parental accession, demonstrating an inability of the Mir-0 and Br-0 FT alleles to complement the FT mutant. In addition, when Br-0 and Mir-0 were crossed to a CO mutant, the F1 plants were equally late, comparable to the crosses with ft, showing a lack of co complementation. The mutant analysis combined with the previously described experiments support the idea that Mir-0 and Br-0 likely have alleles of CO and FT which do not respond to FR enrichment.
For the majority of accessions analyzed, FT appears to be critical for FR sensitivity. The semi-sensitive accession (Hau-0) crossed to ColFRI(ft-1) had an F1 FR-flowering phenotype equivalent to Hau-0. This demonstrates that the FT allele in Hau-0 retains some FR-sensitive FT function; however, the F1 from Hau-0/ColFRI(ft-1) is later than ColFRI, indicating that the FT allele from Hau-0 is less capable of responding to FR compared to the FT allele from Col-FRI (S5 Table). As expected due to FT dominance, F1 plants from crosses between the ft mutant and FR-sensitive accessions (Yo-0, HOG, and Nok-3) had a flowering phenotype equivalent to the FR-sensitive accession (S5 Table). Two semi-sensitive accessions (Et-0 and Sue-0) with moderately late FR-flowering times (43 and 48 RLN, respectively) flowered later when crossed to the CO mutant (56 and 58.5 RLN, respectively,) likely due to the dosagedependent behavior of CO, and suggesting a functional CO in these accessions. Et-0 (semi-sensitive) crossed to FR-sensitive (sensitive) accessions (HOG and Nok-3) had flowering times equivalent to the FR-sensitive accessions, suggesting that the Et-0 CO allele is capable of acting on the FT alleles from those two accessions. In addition, the Et-0/ColFRI F2 showed no linkage at chromosome 5, demonstrating CO functionality in this accession (S5 Table).
The only accession with an amino acid substitution in FT, Fl-1, flowers extremely late in FR-enriched conditions (>80 RLN). The Fl-1/ColFRI(ft-1) F1 population flowers as late as Fl-1 and ColFRI(ft-1), demonstrating that the Fl-1 FT allele has an altered function. Fl-1/ColFRI (co) F1 flowering time data (54 RLN) shows that the Fl-1 CO allele appears functional and is capable of acting on the ColFRI FT allele, further supporting the notion that Fl-1 FR insensitivity of Fl-1 results exclusively from an altered functional FT allele. Therefore, the data support the notion that FT allelic variability in natural accessions is a major factor in reduced FR sensitivity for flowering time.
Using a Recombinant Inbred Lines (RIL) population, the Est-1 accession was shown to have a hypo-active FT allele, which mapped to the FT promoter region [32]. A NIL (Near-Isogenic Line) used to isolate the Est-1 FT region in a Col background was crossed to ColFRI to determine if this allele of FT was compromised in FR-dependent flowering in a winter-annual (FRI+) background. The FR-flowering phenotype of the F2 population was dependent on the FT region with plants containing Col FT (14 RLN) flowering substantially earlier than plants

FT Expression
We evaluated FT expression in 3 of our well-characterized accessions compromised in FRenriched flowering and ColFRI, +/-FR (Fig 6A). Without FR addition, the only accession/ strain that flowered was ColFRI, and FT expression is only detected in that background. After 1 week in FR-enriched conditions, FT expression is observed in ColFRI, Hau-0, and Mir-0, but not Fl-1, which is consistent with their flowering times under FR conditions (15RLN,  32RLN, 50RLN, and DNF, respectively). After 5 weeks of FR treatment, all accessions/strains were expressing FT with ColFRI and Hau-0 having relatively high and equivalent expression, which correlates well with the FR responsiveness. Both Mir-0 and Fl-1 had not flowered after 5 weeks of FR and this correlates with their lower level of FT expression.
To explore whether or not the FT alleles would be expressed at different levels, we evaluated F3 plants that are heterozygous at FT from 3 populations for allele-specific expression ( Fig 6B). As expected from the above results, FT expression correlated with flowering time with ColFRI alleles having the highest/earliest expression followed by Hau-0, Mir-0, and Fl-1 alleles, respectively. These results demonstrate that in an identical genetic background, the four different FT alleles are differentially expressed under FR conditions.

Discussion
We identified a number of accessions that had an attenuated flowering response to FR. The genetic pathway promoting flowering under FR-enriched conditions is likely to be different than the pathway that influences hypocotyl elongation, because comparing FR and GA sensitivity for hypocotyl elongation from Maloof et al. [35] to accessions which are insensitive to FR-induced flowering showed no correlation (S4 Table). In addition, FR-dependent petiole elongation occurs in accessions that are sensitive, semi-sensitive and insensitive for FRinduced flowering (Fig 7). Thus, phenotypic responses to FR or shade likely utilize different downstream pathways to elicit different responses. A similar conclusion was reached by evaluating flowering time and petiole elongation in phytochrome mutants [36]. Whereas petiole elongation in shaded conditions may be more advantageous regardless of location at the beginning of the plants life cycle, precocious flowering could have deleterious effects on reproductive success.
An interesting aspect of this study is that the phenotype in laboratory FR-enriched conditions mirrored the phenotype of plants grown outside in a partially shaded site. However, the R:FR ratio in these two situations was substantially different. In the laboratory, we utilized two levels of FR enrichment, with R:FR ratios of 0.04 and 0.16. As measured by a spectroradiometer, the R:FR ratio outside was never below 0.60; however, overall light intensity was 3X higher. Thus, it is possible that at higher intensities, a higher R:FR ratio is effective in promoting flowering. Another possibility is that temperature cycling outdoors acts synergistically with FR to promote flowering. We did not, however, observe an effect of temperature cycling in this work, but it may have been masked by the effect of the low R:FR ratio we used in the laboratory.
Most accessions that do not show a rapid acceleration of flowering under FR-enriched conditions are from northern Europe (UK, Finland, Sweden, and Denmark). It is tempting to speculate that this behavior is not just a consequence of population structure, but provides an adaptive advantage to accessions at higher latitudes that experience a long winter. One possibility is that because northern latitudes experience longer durations of twilight which is enriched in FR, insensitivity to FR could prevent flowering at inappropriate times of the growing season [37,38]. Another possibility is simply genetic drift; because these accessions reproducibly encounter cold yearly, there would be little selective pressure to maintain an additional pathway (photoperiod) to promote flowering.
Two accessions from southern Europe, Mir-0 (Miramare, Italy, 45.7025 N, 13.7125E) and Br-0 (Brunn, Czech Republic 49.1166 N, 16.6083), were also FR insensitive for flowering. Both accessions are glabrous, have identical flowering behavior, and were collected 378 miles apart, indicating the possibility of relatedness. These two accessions are also unique in that both CO and FT allelic variation is likely to contribute to insensitivity to FR-induced flowering. Why FR insensitivity is present in accessions from these two regions is unknown, but the relatively short vernalization requirement shared by these two accessions may render the lack of a FR response ecologically relevant, providing an additional layer of protection to prevent precocious flowering.
In summary, we demonstrate that FT expression is intimately associated with FR-induced flowering. In Arabidopsis, FT is one of the most highly conserved proteins with only a single amino acid variation across the species; thus, variability associated with FT must reside in expression of the FT gene, as demonstrated in this study. FT has both negative regulators (FLC, SVP, and TEMPRANILLO) as well as positive regulators (CO, GI, and SQUAMOSA PRO-MOTER BINDING PROTEIN-LIKE 3), all of which bind to the promoter region or the first intron [7,39,40]. Phylogenetic shadowing has identified conserved blocks in the region upstream of FT that could serve as binding sites for regulatory proteins that control FT transcription [21,41]. Future studies evaluating differences in proteins and their interactions with promoter motifs will be useful in understanding the cooperative or competitive interactions controlling FT expression. In addition to regulators and promoter motifs, a role for chromatin modification has been demonstrated providing an additional layer of regulation [21]. The unusual genomic spacing of FT, with no ORFs 7.5kb upstream and 9.5kb downstream, may be related to regulation by higher-order chromatin structure. Thus, FT regulation is complex as might be expected for an integrator that receives signals from multiple pathways (photoperiod, temperature, light quality, and hormone levels). We have shown that FR enrichments leads to accelerated flowering in some accessions via variability in FT expression. FT, or florigen, is central to environmentally induced flowering and an attractive candidate for engineering plants for a changing environment.  Table. Correlations of flowering times for 10 accessions in various growth conditions. Correlations among the ten conditions for the 10 accessions from SD S2 Table. (XLSX) S4  Table. Oligos and markers used in this study. Over 100 accessions were scored for multiple polymorphisms around FT, with two being within the ORF of FT, which was used for the allele-specific expression experiments (BsmBI and Cac8I). Only two accessions (Fl-1 and Bsch-2) contained the polymorphism responsible for the BsmBI site. In addition, the table shows data for an indel upstream of FT (1.2, 1.4, and 2.3kb; Warthmann personal communication). The 1.2kb version is relatively rare being present in only 6 accessions, including two (Et-0 and Br-0), accessions used extensively in this study. (XLSX) Methodology: C. J. Schwartz, Joohyun Lee, Richard Amasino.

Supporting information
Project administration: C. J. Schwartz.