ELF3 polyQ variation in Arabidopsis thaliana reveals PIF4-independent role in thermoresponsive flowering

Plants have evolved elaborate mechanisms controlling developmental responses to environmental stimuli. A particularly important stimulus is temperature. Previous work has identified the interplay of PIF4 and ELF3 as a central circuit underlying thermal responses in Arabidopsis thaliana. However, thermal responses vary widely among strains, possibly offering mechanistic insights into the wiring of this circuit. ELF3 contains a polyglutamine (polyQ) tract that is crucial for ELF3 function and varies in length across strains. Here, we use transgenic analysis to test the hypothesis that natural polyQ variation in ELF3 is associated with the observed natural variation in thermomorphogenesis. We found little evidence that the polyQ tract plays a specific role in thermal responses beyond modulating general ELF3 function. Instead, we made the serendipitous discovery that ELF3 plays a crucial, PIF4-independent role in thermoresponsive flowering under conditions more likely to reflect field conditions. We present evidence that ELF3 acts through the photoperiodic pathway, pointing to a previously unknown symmetry between low and high ambient temperature responses. Moreover, in analyzing two strain backgrounds with vastly different thermal responses, we demonstrate that responses may be shifted rather than fundamentally rewired across strains. Our findings tie together disparate observations into a coherent framework in which multiple pathways converge in accelerating flowering in response to temperature, with some such pathways modulated by photoperiod.


INTRODUCTION 22
photomorphogenesis [26]. Under light conditions, degradation of PIFs is mediated by 1 direct interactions with photoactivated phyB [22]. PIF4 is distinct from the other PIFs in 2 having specific roles in temperature sensing and flowering [27]. pif4 null mutants show 3 short hypocotyls with photomorphogenic attributes even in the dark [28]. 4 At elevated ambient temperatures (27º-29º) the wiring of these signaling 5 pathways changes. Several independent studies have recently found that elevated 6 temperatures, specifically during dark periods [29], inhibit the activity of the EC by an 7 unknown mechanism [14-16], leading to increased expression of PIF4 and its targets 8 [11,27]. This increased PIF4 activity leads to several morphological temperature 9 responses through various signaling pathways [13,27]. PIF4 is also required for the 10 acceleration of flowering at 27ºC under short photoperiods [9,29], though these 11 observations have been disputed [30,31]. While PIF4 action alone (among PIFs) is 12 essentially sufficient for most described thermomorphogenic responses [11,27] inhibition by phyB. Under longer photoperiods and higher temperature a flowering 20 acceleration still exists [7,11], which suggests a PIF4-independent thermoresponsive 21 flowering pathway. Nonetheless, recent reviews of the literature tend to emphasize the 22 primacy of PIF4 in this response [10, 32,33], although the condition of elevated 1 temperature with short photoperiods is probably rare in the field. 2 Recent studies have identified ELF3 as a plausible upstream regulator of PIF4 in 3 thermal responses [14][15][16][17][18]. However, others have implicated different candidates, such 4 as FCA [13], and mathematical modeling has suggested that ELF3/EC complex 5 regulation alone is insufficient to explain PIF4 thermal regulation [14,34]. The exact 6 mechanisms of this response have yet to be unraveled. 7 Specifically, the mechanism by which EC/ELF3 activity is reduced under elevated 8 temperatures ("temperature sensing") is not known. We recently used transgenic 9 experiments to demonstrate that ELF3 function is dependent on the unit copy number of 10 its C-terminal polyglutamine (polyQ) tract [35]. This domain is likely disordered, and 11 disordered domains evince structural changes in response to physical parameters such 12 as temperature [36]. Thermal remodeling of this polyQ tract is a plausible mechanism by 13 which ELF3 activity could be modulated through temperature. This polyQ tract also 14 shows substantial natural variation [35], potentially serving as a factor underlying natural 15 variation in thermoresponsive phenotypes. For example, in flies, variable repeats are 16 associated with local temperature compensation adaptations [37]. In short, the ELF3-17 polyQ is an attractive candidate for adaptive variation in the ecologically relevant trait of 18 temperature response [38]. 19 In this study, we used transgenic polyQ variants of ELF3 in two A. thaliana 20 genetic backgrounds to dissect the contribution of the polyQ tract to temperature 21 response. We show that polyQ repeat copy number modulates temperature sensing by 22 affecting overall ELF3 function. Surprisingly, we found that ELF3's role in 23 1 Fig. 1. Response to elevated temperature (27º, relative to 22º) among transgenic lines 2 expressing ELF3-polyQ variants. Mean response and error were estimated by 3 regression, based on two independently-generated transgenic lines for each genotype, 4 with n >= 30 seedlings of each genotype in each condition (Table S1). WT = Ws, elf3 = 5 elf3 mutant+vector control, 0Q = elf3 mutant+ELF3 transgene lacking polyQ, etc. Error Bonferroni-corrected p < 0.01, **: Bonferroni-corrected p < 0.05, . : Bonferroni-corrected 10 p < 0.1 in testing the interaction term (different response from WT, Ws or Col). (C): 11 Temperature response is a function of ELF3 functionality (repression of hypocotyl 1 elongation at 22º). Simple means of 22º hypocotyl length, regression estimates of 2 temperature response. PCC = Pearson correlation coefficient; p-value is from a Pearson 3 correlation test. 4

5
Expression of PIF4 and PIF4 targets as a function of temperature and ELF3. 6 To evaluate the hypothesis that the thermal response defects in the transgenic lines 7 was due to up-regulation of PIF4 and PIF4 targets, we measured transcript levels of 8 PIF4 and its target AtHB2 in seedlings of selected lines from both backgrounds at 22ºC 9 and 27ºC (Fig. S1). Like others [15,16], we observed an inverse relationship between 10 ELF3 functionality and transcript levels of PIF4 and AtHB2, with larger effects on PIF4 11 expression. The ELF3 lines with the strongest thermal response (e.g. 16Q in the Ws 12 background) showed the most robust de-repression of PIF4 at elevated temperature. 13 However, elf3 null mutants retained some PIF4 up-regulation under these conditions, 14 especially in the Ws background. We conclude that ELF3-mediated de-repression of 15 PIF4 is involved in thermal responses as suggested by prior studies [15,16]; however, 16 de-repression of PIF4 and its targets may not be sufficient to explain the entirety of 17 thermal response defects in elf3 null mutants.  Following the expectation that ELF3's thermal response acts through PIF4, we 21 reasoned that ELF3 should also play a role in other PIF4-dependent thermal responses. 22 One well-known response to elevated temperature is adult petiole elongation. pif4 23 mutants fail to show this response when grown at elevated temperatures [11]. We 1 measured petiole length in the ELF3 polyQ transgenic lines, expecting that, due to 2 general PIF4 de-repression, poorly-functioning ELF3 polyQ lines would show no 3 response (perhaps due to constitutively elongated petioles, similar to hypocotyls; Fig.  4 2). In stark contrast to this expectation, we found that all lines had a robust petiole 5 response to temperature ( Fig. 2A, B). This effect was apparent in both Ws ( Fig. 2A) and 6 Col backgrounds (Fig. 2B). Moreover, this response was actually accentuated in elf3 7 null mutants and in poorly-functioning ELF3 polyQ variants ( Fig. 2A, B).  (Table S2) transcriptional regulators of FT such as SVP [31]. 23 Unlike Col, Ws lacks a robust flowering response to elevated temperature under 1 these conditions [42], and indeed, variants in the Ws background generally showed no 2 thermoresponsive flowering (Fig. 2C). Thus, ELF3 polyQ variation does not suffice to 3 enhance the negligible thermoresponsive flowering in the Ws background under these 4 conditions. In light of this data, the roles of ELF3 and PIF4 in the elevated temperature 5 response appear to be independent of one another under these experimental conditions 6 and for these traits. These results are intriguing, given that the PIF4 pathway is the 7 best-recognized mechanism for thermoresponsive flowering at high temperatures 8 [9,10,32,33]. Therefore, we suggest that ELF3 acts in a PIF4-independent pathway for 9 thermoresponsive flowering at high temperatures. 10 11 ELF3 regulates thermoresponsive flowering under long days, and is not required for 12

PIF4-dependent adult thermomorphogenesis. 13
Our results with ELF3-polyQ variants suggested that ELF3 dysfunction does not 14 meaningfully affect PIF4-dependent traits, but does affect PIF4-independent traits in 15 adult plants in long days. However, these results may be due to subtle differences in 16 conditions between our approach and those used by previous investigators. We 17 therefore directly addressed the relationship of ELF3 and PIF4 in adult 18 thermoresponsive phenotypes by growing pif4 and elf3 mutants with various thermal 19 treatments. Previous experiments with pif4 mutants used different conditions from ours, 20 specifically a later transfer to elevated temperature [11]. Hence, it was possible that the 21 observed inconsistencies between elf3 and pif4 effects on adult thermoresponsive 22 phenotypes were a trivial consequence of experimental conditions. Specifically, the 23 effects of elevated temperature during the early seedling stages (the conditions we use) 1 may induce pathways irrelevant to treatments at later, vegetative stages. Thus, we 2 tested both transfer conditions under long days (Fig. 3). We found that the effect of 3 different experimental conditions is negligible, though the earlier 27ºC treatment showed 4 a slightly stronger morphological response (Fig. 3A, B). Thus, the timing of the 27ºC 5 treatment (early seedling vs. vegetative stage) does not substantially affect adult 6 thermoresponsive traits. Further, our results under long days were similar to previous 7 observations under continuous light [11], showing that PIF4 is essential for petiole 8 elongation ( Fig. 3B), but dispensable for thermoresponsive flowering (Fig. 3C). Our 9 PIF4 results were in direct contrast to ELF3, which was dispensable for petiole 10 elongation (Fig. 3B), but essential for thermoresponsive flowering (Fig. 3C). These 11 results indicate the apparent independence of ELF3 and PIF4 in these specific 12 responses, and suggest that seedling thermomorphogenesis, adult 13 thermomorphogenesis, and thermoresponsive flowering constitute three independent 14 developmental responses.  Table S3. In each case, **: Bonferroni-9 corrected p < 0.01, **: Bonferroni-corrected p < 0.05, . : Bonferroni-corrected p < 0.1 in 10 testing the interaction term (different response from Col). One open question was whether the dispensability of ELF3 for petiole elongation 13 reflected increased importance of other inputs to PIF4, such as FCA, which is involved 14 in PIF4-dependent thermoresponsive petiole elongation in 7-day-old seedlings [13]. We 1 therefore measured adult thermoresponsive petiole elongation in fca mutants ( Fig.  2 S2A), and unexpectedly found no substantial difference between fca mutants and WT 3 Col. Regulatory rewiring across development may remove FCA and ELF3 as inputs to 4 PIF4-dependent thermomorphogenesis in 25-day-old adult plants. constitute independent temperature responses, with ELF3 controlling the former and 20 PIF4 controlling the latter in additive fashions (Fig. 4). That is, elf3 pif4 double mutants 21 showed negligible thermoresponsive flowering like elf3, and a negligible petiole 22 response like pif4. Additionally, elf3 pif4 flowered slightly later than elf3 at 22º, while 23 maintaining a negligible thermal response in flowering, indicating that elf3 mutants are 1 not simply restricted by a physiological limit of early flowering. The additivity of these 2 phenotypes establishes that, under these conditions, ELF3 and PIF4 likely operate in 3 separate thermal response pathways. Consequently, our results support the previously-suggested dominance of 5 thermomorphogenesis by PIF4 rather than other PIFs, and the irrelevance of PIF4 (and 6 most likely other PIFs as well) to thermoresponsive flowering under LD. 7 Overall, the strong photoperiod-dependence of PIF4-related thermoresponsive 8 flowering necessitates the existence of some pathway or pathways independent of PIF4 9 under long days, given the persistence of the phenomenon under these conditions. 10 Based on our data, ELF3 acts in one such pathway.  (Fig. 5A). We found that GI is strongly up-regulated in elf3 null mutants of Col and 19 Ws backgrounds, confirming previous reports in Col [39,46]. Further, wild-type Ws 20 showed higher basal GI levels compared to Col, which did not increase at higher 21 temperatures. In contrast, Col showed very low basal GI levels that increased at higher 22 temperatures to approximately the same levels as Ws. CO levels, however, were not 23 substantially increased by either elf3 mutation or increased temperature, consistent with 1 previous reports [8,46]. Thus, robust thermoresponsive flowering was correlated with 2 low basal levels of GI, and with temperature-dependent GI up-regulation, as observed in 3 Col. The ELF3-dependent thermal responsiveness of GI expression confirms previous 4 reports [15], though the among-strain variation in responsiveness appears to be novel 5 and correlated specifically with flowering induction (but not hypocotyl or petiole 6 elongation, Figures 1 and 2). High basal GI levels in Ws may be associated with other We attempted to measure FT transcript levels in these samples, expecting that 12 they would be elevated in the early-flowering elf3 and 27ºC conditions ( Figure S4).  Edges with increased weight indicate relative increases of influence between conditions. 7 Pathways are indicated, along with other important actors reported elsewhere. 8 9 If the photoperiodic pathway contributes to thermoresponsive flowering at 10 elevated ambient temperatures in long days (LD), we would expect mutants in this 11 pathway to show abrogated thermal responses, as they do under short days (SD), along 12 with members of the autonomous pathway [7]. These two pathways also contribute 13 independently to thermoresponsive flowering at low temperatures (16ºC vs. 23ºC) [6,8]. 14 Altogether, we would expect that a photoperiodic thermoresponsive flowering pathway 15 would operate independently of both PIF4 and the autonomous pathways in long days. 16 It is not clear whether the autonomous pathway would be independent of PIF4, given 17

known interactions between FCA and PIF4 [13]. 18
To evaluate whether these past results under other conditions also apply to long 19 days and elevated temperatures, we measured flowering time at 22ºC and 27ºC in 20 mutants in the photoperiodic pathway (gi, co, Fig. 5B). We also tested mutants of the 21 gibberellin pathway (spy), and a terminal floral integrator (soc1), which we do not expect 22 to be necessary for thermoresponsive flowering. We found robust thermal responses in 23 all mutants except elf3 and gi, similar to previous results under different conditions 1 [7,8,45,46]. These results emphasize once again that differences in thermoresponsive 2 flowering are not generalizable between photoperiods, as it has recently been shown 3 that co mutants have a partial flowering acceleration defect under SD. These results 4 implicate GI (but not CO) as an actor in thermoresponsive flowering at elevated 5 temperatures. Collectively, these experiments suggest that the photoperiod pathway is 6 necessary in promoting thermoresponsive flowering in long days, and expression data 7 in this and other studies suggests that ELF3 is likely to act within this pathway. In previous work, we demonstrated that polyQ variation in ELF3 is (i) common, (ii) 4 affects many known ELF3-dependent phenotypes, and (iii) is dependent on the genetic 5 background [35]. Following the recent discoveries that ELF3 is involved with thermal 6 response [14-16], we confirmed that ELF3 polyQ variation also affects thermal 7 response phenotypes in a background-dependent fashion. However, we found little 8 support for the hypothesis that the polyQ tract has a special role in temperature 9 sensing. Instead, as was the case for other ELF3-dependent phenotypes, ELF3 polyQ 10 variation appeared to affect overall ELF3 functionality, with less functional ELF3 variants 11 lacking robust temperature responses. However, a more exhaustive series of polyQ 12 variants may be required for revealing polyQ-specific effects, in particular because the 13 molecular mechanism(s) by which polyQ variation affects ELF3 functionality remain 14 27ºC. Here, we show that this acceleration requires ELF3, like the elevated temperature 11 acceleration in Col. Another example of differential mutational effects among strains is 12 that gi mutants in the Ler background display robust thermoresponsive flowering [6,7] We propose a model of thermoresponsive flowering, in which PIF4, ELF3, the 7 photoperiodic pathway, and other pathways interact depending upon condition and 8 genetic background (Fig. 5D). Under short days or other short photoperiods, phyB 9 activity is down-regulated, leading to up-regulation of PIF4 [22,[50][51][52], which at high that the first two pathways are necessary but not sufficient for thermoresponsive 12 flowering, and that the third (PIF4) is sufficient but not necessary for thermoresponsive 13 flowering. Further study will be necessary in understanding the interdependencies of the 14 three pathways. For instance, it has been suggested that PIF4 binding to the FT 15 promoter is dependent on cooperativity with a second photoperiod-controlled actor [34]. 16 In conclusion, we observe that ELF3 is involved in the hypocotyl response to 17 elevated temperature as reported previously, and that this response can be abrogated 18 by poorly-functioning ELF3 polyQ variants. We further demonstrate that ELF3 has little 19 effect on the petiole temperature response, and is necessary for the flowering 20 temperature response, suggesting that it functions independently of PIF4, potentially in 21 the photoperiodic pathway. These results reiterate the complexity of these crucial 22 environmental responses in plants, and can serve as a basis for further development of 23 our understanding of how plants respond to elevated temperatures. In the context of 1 climatic changes, this understanding will serve those attempting to secure the global 2 food supply. 3 4

MATERIALS AND METHODS 5
Plant materials and growth conditions. All mutant lines (except pif4-2 elf3-200) were 6 either described previously or obtained as T-DNA insertions from the Arabidopsis 7 Biological Resources Center at Ohio State University [53,54], and are described in 8   Table S11. pif4-2 elf3-200 was obtained via crossing and genotyping. T-DNA insertions 9 were confirmed with primers described in Table S10. For hypocotyl assays, seedlings 10 were grown for 15d in incubators set to SD (8h light : 16h dark days, with light supplied 11 at 100 µmol· m -2 ·s -1 by cool white fluorescent bulbs) on vertical plates as described 12 previously [35]. All plates were incubated at 22º for one day, after which one replicate 13 arm was transferred to an incubator set to 27º, with another replicate arm maintained at Where temperature responses are reported, they consist of the β T + β GxT terms and 2 associated errors ( ) where σ T is the standard error for β T and σ GxT is the 3 standard error for β GxT ), and thus are corrected for systematic experimental variation 4 and temperature-independent genotype effects. Where p-values are reported for GxT 5 interaction effects, they have been subjected to a Bonferroni correction to adjust for 6 multiple comparisons. Analysis scripts and data are provided at 7 https://figshare.com/articles/elf3_pif4_data_code_v2/3398353. 8 9 10 Gene expression analyses. Seedlings were grown for 1d under LD at 22º, after which 11 one replicate arm was transferred to LD at 27º, with another replicate arm remaining at 12 22º, and all seedlings were harvested 6d later at indicated times. At harvest, ~30mg 13 aerial tissue of pooled seedlings was flash-frozen immediately in liquid nitrogen and 14 stored at -80º. RNA extraction, cDNA synthesis, and real-time quantitative PCR were 15 performed as described previously [35], using primers in Table S10 Table S8.  Tissue was collected from 7d seedlings at ZT0. Error bars indicate SEM across three 1 biological replicates. 2 3 Table S1. Regression analysis of hypocotyl elongation temperature response 4 among Col and Ws transgenic lines. 5 Table S2. Regression analysis of petiole : leaf length ratio and rosette leaf 6 number at flowering temperature response among Col and Ws transgenic lines. 7 Table S3. Regression analysis of rosette leaf number at flowering and petiole : 8 leaf length ratio temperature responses in elf3 and pif4. 9 Table S4. Regression analysis of rosette leaf number at flowering temperature 10 response in Ws and elf3-4. 11 Table S5. Regression analysis of petiole : leaf length ratio temperature response 12 in Col and fca mutants. 13 Table S6. Regression analysis of rosette leaf number at flowering temperature 14 response in elf3 pif4 double mutants. 15 Table S7. Regression analysis of rosette leaf number at flowering temperature 16 response in pif4 pif5 double mutants. 17 Table S8. Regression analysis of the petiole elongation temperature response in 18 pif4 pif5 double mutants. 19 Table S9. Regression analysis of rosette leaf number at flowering and petiole : 20 leaf length ratio temperature responses in flowering pathway mutants. 21 Table S10. Primers used in this study. 22 Table S11. Mutant lines used in this study. 23