Photoperiod sensing of the circadian clock is controlled by ELF3 and GI

ELF3 and GI are two important components of the Arabidopsis circadian clock. They are not only essential for the oscillator function but are also pivotal in mediating light inputs to the oscillator. Lack of either results in a defective oscillator causing severely compromised output pathways, such as photoperiodic flowering and hypocotyl elongation. Although single loss of function mutants of ELF3 and GI have been well-studied, their genetic interaction remains unclear. We generated an elf3 gi double mutant to study their genetic relationship in clock-controlled growth and phase transition phenotypes. We found that ELF3 and GI repress growth during the night and the day, respectively. We also provide evidence that ELF3, for which so far only a growth inhibitory role has been reported, can also act as a growth promoter under certain conditions. Finally, circadian clock assays revealed that ELF3 and GI are essential Zeitnehmers that enable the oscillator to synchronize the endogenous cellular mechanisms to external environmental signals. In their absence, the circadian oscillator fails to synchronize to the light-dark cycles even under diurnal conditions. Consequently, clock-mediated photoperiod-responsive growth and development is completely lost in plants lacking both genes, suggesting that ELF3 and GI together convey photoperiod sensing to the central oscillator. Since ELF3 and GI are conserved across flowering plants and represent important breeding and domestication targets, our data highlight the possibility of developing photoperiod-insensitive crops by manipulating the combination of these two key genes. One sentence summary ELF3 and GI are essential for circadian clock mediated photoperiod sensing. Author Contributions M.U.A., S.J.D. and M.Q. conceived the project. M.U.A. and A.D. performed the experiments. M.U.A. wrote the article with contributions of all authors. Funding information The funding for this work was provided by a Biotechnology and Biological Sciences Research Council grant to SJD (BBSRC grant code BB/N018540/1), a grant by the Deutsche Forschungsgemeinschaft to MQ (Qu 141/6–1), and the Leibniz Association.


Abstract 24
ELF3 and GI are two important components of the Arabidopsis circadian clock. They are not only 25 essential for the oscillator function but are also pivotal in mediating light inputs to the oscillator. Lack of 26 either results in a defective oscillator causing severely compromised output pathways, such as 27 photoperiodic flowering and hypocotyl elongation. Although single loss of function mutants of ELF3 and 28 GI have been well-studied, their genetic interaction remains unclear. We generated an elf3 gi double 29 mutant to study their genetic relationship in clock-controlled growth and phase transition phenotypes. 30 We found that ELF3 and GI repress growth during the night and the day, respectively. We also provide 31 evidence that ELF3, for which so far only a growth inhibitory role has been reported, can also act as a 32 growth promoter under certain conditions. Finally, circadian clock assays revealed that ELF3 and GI are 33 essential Zeitnehmers that enable the oscillator to synchronize the endogenous cellular mechanisms to 34 external environmental signals. In their absence, the circadian oscillator fails to synchronize to the light-35 dark cycles even under diurnal conditions. Consequently, clock-mediated photoperiod-responsive 36 growth and development is completely lost in plants lacking both genes, suggesting that ELF3 and GI 37 together convey photoperiod sensing to the central oscillator. Since ELF3 and GI are conserved across 38 flowering plants and represent important breeding and domestication targets, our data highlight the 39 possibility of developing photoperiod-insensitive crops by manipulating the combination of these two 40 key genes. 41

Introduction: 42
Rotation of the earth around its axis results in rhythmic oscillations in light and temperature during a 24-43 hour day/night cycle. As a consequence of evolving under these predictable changes, organisms have 44 developed internal timekeeping mechanisms known as the circadian clock that enables them to 45 anticipate periodic changes in their surrounding environment (de Montaigu et al., 2010; Anwer and 46 . Circadian clocks consist of three pathways: inputs, core oscillators, and outputs. Input 47 pathways deliver external cues (also known as Zeitgeber, German for time-givers), such as ambient light 48 and temperature, to circadian oscillators. The timing information from the Zeitgeber is received by core-49 oscillator components known as Zeitnehmer (German for time-takers) that help to reset and synchronize 50 the clock with the local environment (entrainment). Once entrained, the oscillators generate a ~24h 51 rhythmicity that can be sustained for long periods; even in the absence of environmental cues (i.e., free-52 running conditions, such as constant light and temperature conditions) Oakenfull and 53 Davis, 2017). After synchronizing with the external environment, oscillators link to various processes to 54 rhythmically regulate the levels of genes, proteins, and metabolites. This allows organisms to anticipate 55 and adapt to the changing environment, such as seasonal changes in day length (photoperiod). The 56 circadian clock thereby regulates various output pathways including photosynthesis, growth, disease 57 resistance, starch metabolism, and flowering time (Andres and Coupland, 2012; Müller 58 et al., 2014). 59 The central part of the clock, the oscillators, are composed of transcriptional-translational feedback 60 loops . The Arabidopsis thaliana (Arabidopsis) oscillator 61 consists of three such loops: a morning loop, an evening loop and a central oscillator. The central 62 oscillator is comprised of two partially redundant myb-like transcription factors CIRCADIAN CLOCK 63 ASSOCIATED 1 (CCA1) and LATE ELONGATED HYPOCOTYL (LHY), and a member of the PSEUDO-64 RESPONSE REGULATOR (PRR) family TIMING OF CAB EXPRESSION 1 (TOC1/PRR1). This is a dual negative 65 feedback loop where respective morning and evening expression of CCA1/LHY and TOC1 repress each 66 other Alabadı́ et al., 2001;. In the morning, the core-67 oscillator components CCA1/LHY activate PRR7 and PRR9, which later repress CCA1/LHY, together 68 constituting the morning loop . The 69 evening expression of TOC1 represses GIGENTEA (GI), which in turn activates TOC1 and formulates the 70 evening loop . Besides these three fundamental 71 loops, a complex of three evening phased proteins (known as evening complex or EC), consisting of 72 EARLY FLOWERING 4 (ELF4), ELF3 and LUX ARRYTHMO (LUX), have been identified as an essential part of 73 the core oscillator . The EC is 74 connected to all three loops of the oscillator. By direct binding to their promoters, the EC represses the 75 transcription of PRR9 and GI Ezer et al., 76 2017). A direct repression of ELF3 by CCA1 connects the EC with the central oscillator 77 Kamioka et al., 2016). 78 ELF3 is one focus of this study and it encodes a multifunctional protein that regulates several 79 physiological and developmental processes. Consistently, elf3 null mutants display pleiotropic 80 phenotypes such as long hypocotyl, accelerated flowering, elongated petioles, and arrhythmia under 81 free-running conditions, suggesting that several important pathways are disrupted 82 Kolmos et al., 2011;Anwer et al., 2014;Box et al., 2014). In addition to its role as a 83 member of the EC in the core oscillator, it functions as a Zeitnehmer in the light input pathway. 84 4 Therefore, plants lacking ELF3 display severe light gating defects . A physical 85 interaction of ELF3 and PHYTOCHROME B (PhyB) establishes a direct link between the oscillator and 86 photoreceptors . For the regulation of rhythmic growth, ELF3 mainly 87 relies on the EC binding to the promoters of major growth regulators PHYTOCHROME-INTERACTING 88 FACTOR 4 (PIF4) and PIF5, causing their transcriptional repression during the night 89 Raschke et al., 2015). However, ELF3 can also inhibit PIF4 by sequestering it from its targets (Nieto et al., 90 2014). Consistently, the lack of PIF4/PIF5 repression in elf3 mutants results in accelerated growth during 91 the night Box et al., 2014). In addition to growth, ELF3 controls flowering time by 92 acting on the major floral activator FLOWERING LOCUS T (FT) via direct repression of GI (Mizuno et al., 93 2014;Ezer et al., 2017). Interestingly, ELF3 repression of FT does not require CONSTANS (CO) (Kim et al., 94 2005). Taken together, functional presence of ELF3 is essential for both plant growth and development. 95 The second protein in the focus of this study is GI, a large, preferentially nuclear-localized protein with 96 domains of unknown functions . The gene's transcription is controlled by 97 the circadian clock. Furthermore, it is post-transcriptionally regulated by light and dark (Fowler et al., 98 1999;David et al., 2006). GI regulates diverse developmental and physiological pathways. The role of GI 99 in the control of photoperiodic flowering is well documented. Here, GI acts as a major activator of FT 100 expression, either by directly binding to its promoter or by inducing the expression of CO (Fornara et al., 101 2009;. Moreover, GI physically interacts with both red and blue light 102 photoreceptors PhyB and ZEITLUPE (ZTL), respectively, indicating a functional role also in 103 photomorphogenesis . Consistently, gi mutants are defective in 104 proper light responses and display elongated hypocotyls under both red and blue lights (Huq et al., 105 2000;. Although the underlying molecular mechanism of hypocotyl growth 106 regulation is not fully understood, it relies at least partially on PIF4, since the growth promoting effect of 107 gi mutations was fully masked by the absence of PIF4 (de Montaigu et al., 2014;Fornara et al., 2015). 108 The EC subunit ELF4 is epistatic to GI in regulating hypocotyl length, suggesting that the GI effect on PIF4 109 is EC dependent . However, ELF4 masking of GI is specific to growth regulation because 110 in flowering time control the genetic hierarchy between these two is reversed. Here, GI is epistatic to 111 ELF4. To make the interaction between these two players even more interesting, both are working 112 additively or synergistically in the control of the circadian clock . GI plays a pivotal role 113 in generating robust circadian rhythms under natural conditions in a way that daily rhythms of its 114 expression respond to day length that depends on the latitude of origin of Arabidopsis accessions (de 115 Montaigu and Coupland, 2017). 116 Interestingly, GI co-localizes with the EC components ELF4, ELF3 and LUX in nuclear bodies (Yu et al., 117 2008;, where it physically interacts with ELF4 and ELF3 (Yu et al., 2008;Kim et al., 118 2013). ELF4 regulates GI subcellular localization and modulates its DNA binding ability by sequestering it 119 from the nucleosome . Further, GI and ELF4 have differentially dominant influences on 120 circadian physiological outputs at dusk and dawn, respectively . The functional 121 importance of ELF3-GI interaction is unknown. However, it is reported that ELF3 regulates diurnal 122 protein accumulation of GI by facilitating its degradation during darkness by a CONSTITUTIVE 123 PHOTOMORPHOGENIC 1 (COP1) mediated proteasomal mechanism (Yu et al., 2008). Consistent with the 124 finding that ELF3 binds to the GI promoter and represses its transcription , all 125 components of the EC were found to bind the GI promoter in a CHIP-Seq experiment, demonstrating a 126 direct relationship between GI and the EC (Ezer et al., 2017). 127 5 As mentioned above, the genetic hierarchy between ELF4 and GI is relatively well understood (Kim et al., 128 2012). Based on the observations that mutations in EC components exhibit similar defects (Herrero et 129 al., 2012), a conserved genetic relationship between GI and other EC components seems reasonable. On 130 the other hand, the finding that ELF3 likely functions also independently of the EC (Nieto et al., 2014) 131 opens the possibility for a different pattern of genetic interactions between ELF3 and GI. 132 In this study, we provide genetic support for the biochemical evidence of an EC independent function of 133 ELF3. We furthermore demonstrate that ELF3 and GI are essential clock Zeitnehmers that are required to 134 synchronize endogenous signals with the external environment. In their absence the circadian clock fails 135 to respond to light signals, resulting in the breakdown of the photoperiod sensing mechanism. From an 136 applied perspective, this interaction has the potential to generate photoperiod-independent crops, 137 possibly allowing the cultivation of numerous day light sensitive species in currently non-permissive 138 latitudes. 139 6 Results: 140 ELF3 and GI are essential for photoperiod responsive growth and development 141 ELF3 and GI are two important factors involved in photoperiod responsive flowering (Andres and  142 Coupland, 2012; . A previous report has suggested that under long days (LD, 16h light/8 h 143 dark) GI is epistatic to ELF3 (Chou and Yang, 1999). GI is also epistatic to ELF4, another component of EC,144 further suggesting that flowering time control of the EC acts through GI . However, it is 145 unclear whether the suggested genetic hierarchy between ELF3 and GI is universally applicable under a 146 range of photoperiods. To investigate the environmental sensitivity of these genetic interactions in 147 detail, we generated an elf3-4 gi-158 double mutant (hereafter designated as elf3 gi) and measured 148 flowering time in comparison to the corresponding single mutants elf3-4 (hereafter designated as elf3) 149 and gi-158 (hereafter designated as gi), and the Ws-2 wild type (WT) under long day (LD, 16h light/8 h 150 dark), short day (SD, 8/16), and neutral day (ND, 12/12) photoperiods. Consistent with reported 151 phenotypes of elf3 and gi null mutants Fowler et al., 1999;, gi and 152 elf3 flowered later and earlier, respectively, than WT under all photoperiods tested ( Figure 1A). 153 Furthermore, similarly to WT, both single mutant alleles flowered earlier in longer photoperiods than in 154 shorter photoperiods, therefore displaying an intact response to the length of the light period. 155 Interestingly, such a photoperiodic response was completely lost in the elf3 gi double mutant, where 156 flowering time was unaffected by the photoperiod ( Figure 1A). Moreover, while under LD and ND 157 flowering time of elf3 gi was similar to gi, it was similar to elf3 under SD ( Figure 1A). Thus, unlike ELF4, 158 where GI is epistatic under both LD and SD , no clear genetic hierarchy was observed 159 between ELF3 and GI, suggesting independent roles in flowering-time control. 160 Since transition from the vegetative to the reproductive phase is only one of several developmental 161 processes influenced by the photoperiod, we next sought to determine whether elf3 gi is also insensitive 162 to photoperiod during the early growth phase. A classic phenotypic output for vegetative growth is 163 elongation of the juvenile stem (hypocotyl), which, like flowering time, is also determined by the length 164 of the light period. In WT, the length of the photoperiod is inversely proportional to the length of the 165 hypocotyl. However, this relationship is not linear. Until a critical photoperiod (14-16 h light) is reached, 166 the growth inhibitory effect of the increased photoperiod remains intact. After this time point, a further 167 increase in the photoperiod has almost no effect on growth . To investigate the role of 168 ELF3 and GI in photoperiod growth control, we measured hypocotyl length of WT, elf3, gi, and elf3 gi 169 seedlings grown under a range of photoperiods, from 24 hours darkness (DD), with a gradual increase of 170 2 hour light periods, to 24 hours light (LL) ( Figure 1B, S1A-B, Tables S1-S2). In confirmation of Niwa et al. 171  , an intact response to photoperiod was observed in WT with plants responding to an 172 increase in day length with a decrease in hypocotyl length until the 16h photoperiod. After 16h, no 173 significant decrease in hypocotyl length was observed. Albeit with an overall longer hypocotyl, WT-like 174 response to the changing photoperiod was also observed in gi ( Figure 1B, S1A-B, Tables S1-S2). 175 Interestingly, both elf3 and elf3 gi did not display an intact photoperiod response of growth inhibition. 176 Unlike WT, the repressive action of longer photoperiods continued even after 16h. Notably, the effect of 177 light repression was discontinued after 20h photoperiod in elf3, whereas, in elf3 gi it continued until LL 178 ( Figure 1B, S1A-B, Tables S1-S2). Thus, our data indicate a previously not recognized additive function of 179 ELF3 and GI in photoperiod sensing, which only becomes visible in the absence of both genes. 180 7 The EC controls hypocotyl elongation by regulating the expression of PIF4 . Under 181 LD and SD, the length of the elf4 gi double mutant is similar to elf4, indicating that ELF4 is epistatic to GI 182 . Since ELF3, like ELF4, is also a component of the EC, a similar genetic hierarchy could 183 also be expected between ELF3 and GI. If so, hypocotyl length of elf3 gi and elf3 should be similar. 184 However, we found that under both LD and SD elf3 gi was significantly longer than elf3 ( Figure 1B, Light 185 periods 8 and 16), suggesting an additive function of ELF3 and GI. Together, these data demonstrate 186 that both ELF3 and GI are essential for photoperiod sensing at both juvenile and adult stages of plant 187 development. 188

ELF3 promotes growth under blue light 189
Previously, ELF3 was reported solely as an inhibitor of growth under a range of light quantities and 190 qualities Doyle et al., 2002). In our light-period growth analysis, 191 we observed that under LL elf3 is significantly shorter than WT ( Figure  significantly shorter hypocotyl compared to WT ( Figure 1C). Although contradictory to the accepted 204 understanding of being a general negative regulator of growth, these observations reveal a previously 205 unknown growth promoting role of ELF3 specifically under BB. 206 The growth inhibitory role of ELF3 in white light is known to be exerted at least in part via PIF4 (Nusinow 207 et al., 2011). To better understand the elf3 growth behavior under BB and to dissect the possibility of 208 antagonistic action of ELF3 on PIF4 under these conditions, we measured the expression of PIF4 and its 209 direct targets IAA29 and YUC8. Interestingly, under BB, albeit a higher PIF4 expression in elf3, the levels 210 of its target genes were lower than in WT ( Figure 1D). This indicates that in the absence of ELF3 alone, 211 PIF4 fails to fully induce the expression of its targets under BB, resulting in short hypocotyls. Provided 212 that ELF3 affects the expression levels of these PIF4 target genes by acting on PIF4 itself suggests that 213 under BB ELF3 exerts a growth promoting effect by positively influencing PIF4 activity. Also in agreement 214 with their extended growth phenotypes under BB, elf3 gi double mutants express higher levels of the 215 growth promoting PIF4 target genes ( Figure 1D). A positive effect of ELF3 on PIF4 activity contradicts the 216 previously described negative role of ELF3 in the regulation of PIF4 activity . It is 217 therefore possible that under BB ELF3 does not directly act on PIF4, but rather affects one of its many 218 negative regulators . 219 ELF3 and GI repress growth during night and day, respectively 220 Under diurnal conditions, the elongation of hypocotyl is gated by the circadian clock, allowing maximum 221 growth to occur at dawn under LD . By repressing growth during the night, ELF3 222 8 functions as an important factor in clock gating. Consistently, elf3 mutants have been reported to lose 223 the normal gating response, resulting in maximum growth during the night Box et 224 al., 2014). The role of GI in clock-controlled growth, however, remains largely unknown. The additive 225 growth phenotype of elf3 gi ( Figure 1B), reveals two possibilities: first, both ELF3 and GI work 226 cooperatively at a similar time of day. If so, the loss of both in the elf3 gi double mutant results in an 227 increased growth at that particular time. Alternatively, both repress growth at a different time of the 228 day-night cycle, resulting in an enhanced growth in elf3 gi at separate times. To dissect these 229 possibilities, we measured growth rate of WT, elf3, gi and elf3 gi every hour for two days under LD using 230 infrared imaging, which allowed growth monitoring also in darkness (Figure 2A-D). As reported 231 previously , maximum growth in WT was observed during the early morning at 232 around ZT4 (Figure 2A). In elf3, the growth rate was overall increased with maximum elongation 233 detected during the night ( Figure 2B, Table S3), confirming the night-specific repressive function of ELF3 234 in elongation growth. The gi mutant displayed a broader growth peak during the afternoon with 235 maximum growth observed at ZT8-10 ( Figure 2C, Table S3). In elf3 gi, growth was pronounced during 236 the night. However, in contrast to WT and both single mutants, growth rates did not peak at a specific 237 time of day, but instead remained on a rather constant level. Compared to WT and the single mutants, 238 the rate of elongation growth was increased during both day and night ( Figure 2D, Table S3). Taken  239 together, while we can confirm the previously described growth-repressive role of ELF3 during the night, 240 our results reveal an unknown role of GI in repressing growth specifically during day times. For effective 241 gating of clock-controlled growth, both ELF3 and GI are essential. 242

ELF3 and GI work independently in the circadian clock 243
Since ELF3 and GI are important components of the circadian clock (Mizoguchi et al., 2005;Anwer et al., 244 2014), we asked whether the photoperiod insensitivity of elf3 gi, as revealed by growth and flowering 245 behavior shown above, could be attributed to a malfunctional oscillator. To investigate the interactive 246 role of ELF3 and GI in the clock, we monitored the expression of the CCR2:LUC reporter under constant 247 light (LL) in WT, elf3, gi and elf3 gi plants that were previously entrained under LD, ND or SD (Figure 3). 248 As expected for a functional oscillator, WT displayed a robust rhythm. In contrast, no rhythmic 249 expression of the reporter was detected in elf3 and elf3 gi. The gi mutant was also rhythmic albeit with 250 lower amplitude ( Figure 3A). Moreover, the levels of CCR2:LUC in elf3 gi were higher than the WT, and 251 the single mutants elf3 and gi ( Figure 3A-B), indicating an independent repressive function of ELF3 and 252 GI in the clock. 253 Using the same data, we next calculated the free-running period of the aforementioned lines. 254 Irrespective of the photoperiod provided for entrainment, we found that the WT displayed a similar 255 free-running period ( Figure 3C). Compared to WT, an acceleration in clock speed was observed in gi 256 ( Figure 3C). Like WT, the photoperiod used during entrainment had no effect on gi periodicity ( Figure  257 3C). Consistent with their arrhythmic phenotypes, no regular pattern of periodicity response to 258 photoperiod was detected in elf3 and elf3 gi. While elf3 displayed an overall deceleration in circadian 259 periodicity after all entrainment photoperiods, the elf3 gi response was more random, with a long and 260 short period after LD and ND entrainment, respectively. After SD entrainment, the period of elf3 gi was 261 similar to that of the WT ( Figure 3C). 262 Next we assessed the precision of the oscillator by calculating the relative amplitude error (RAE). An RAE 263 value of "0" represents a perfect rhythm, whereas an RAE of "1" typifies no rhythm (Anwer et al., 2014). 264 9 A general cutoff value of 0.5 is normally used to distinguish between a robust and a dysfunctional 265 oscillator. As expected for a fully functional clock, the WT displayed a very low RAE after all 266 entrainments ( Figure 3D). The RAE measured for gi was significantly higher than the WT but lower than 267 0.5, suggesting a compromised but functional clock ( Figure 3D). Consistent with their arrhythmic 268 phenotype, the RAEs of elf3 and elf3 gi were extremely high (RAE>0.6), indicating a dysfunctional 269 oscillator. Collectively, a dysfunctional oscillator along with an increased CCR2:LUC expression in elf3 gi 270 indicate an additive/synergistic role ELF3 and GI in the clock. 271 Clock entrainment to light signals requires both a functional ELF3 and GI 272 Several clock mutants that are arrhythmic under free-running conditions, display robust oscillations 273 under diurnal conditions, suggesting that the oscillator is still capable of reacting to persistent 274 environmental changes . The complete lack of response of the elf3 gi double 275 mutant to photoperiod ( Figure 1A-B and S1A-B), however, prompted us to think otherwise. Specifically, 276 we hypothesized that the oscillator in elf3 gi might not be responsive to light signals even under diurnal 277 conditions. To test this hypothesis, we monitored the expression of major central-oscillator genes CCA1, 278 TOC1, PRR9, GI and ELF3 under diurnal conditions (ND) ( Figure 4A-E). In WT, the expression profiles of all 279 these genes were consistent with previous data Anwer et al., 2014), with CCA1 and 280 PRR9 peaking in the morning, TOC1 and GI peaking in the evening, whereas ELF3 peaks in the night 281 ( Figure 4A-E). In gi, the expression of TOC1 and ELF3 was higher than the WT, whereas the levels of 282 PRR9 was lower than WT ( Figure 4B-C,E), consistent with previous reports for gi null mutants (Fowler et 283 al., 1999;. No obvious difference in CCA1 expression was detected in gi ( Figure 4A). Also 284 in agreement with published data, expression of TOC1, PRR9 and GI in elf3 was higher than in WT, while 285 CCA1 expression was lower (Hall et al., 2003;Anwer et al., 2014) ( Figure 4A-D). 286 Importantly, in both elf3 and gi, albeit differences in expression levels, the overall shape of the 287 expression patterns of all genes tested was similar to WT ( Figure 4A-E). These data thus indicate an 288 aberrant but functional oscillator in elf3 and gi single mutants, which is capable of responding to 289 environmental signals and generating robust rhythms under diurnal conditions. In elf3 gi double mutant, 290 however, no detectable response to diurnal light signals were observed ( Figure 4A-E). The expression 291 profile of all clock genes tested were completely different from both single mutants and WT. Specifically, 292 the overall expression of PRR9 and ELF3 was higher than the other genotypes tested. CCA1 levels were 293 almost non-detectable. The overall expression of TOC1 was increased compared to WT and gi but 294 decreased compared to elf3. The GI abundance was higher and lower in WT and elf3, respectively 295 ( Figure 4A-E). Most importantly, the characteristic peaks of expression of these genes, which were 296 clearly detectable in WT, elf3 and gi, were absent in elf3 gi. Most of the genes displayed a constant 297 higher or lower expression, which was irresponsive to changes in the light during a diurnal cycle ( Figure  298 4A-E). These data demonstrate that only in the absence of both ELF3 and GI, the circadian oscillator is 299 insensitive to persistent light-input cues. Thus, ELF3 and GI are essential Zeitnehmers that are required 300 for clock entrainment to external light cycles. 301 ELF3 and GI are essential to establish endogenous and light signaling links 302 Once entrained, the circadian clock regulates several key endogenous processes such as gene expression 303 and ensures their precise synchronization with the external environment. This internal-external signaling 304 synchronization is vital for several clock-controlled pathways such as flowering time and hypocotyl 305 elongation. Since the oscillator in elf3 gi failed to establish a link with the external light signals, such a 306 synchronization could potentially be lost in elf3 gi, explaining its photoperiod-insensitive flowering and 307 growth. This could be tested by monitoring the diurnal expression of key clock-regulated genes that are 308 involved in photoperiod-responsive flowering and hypocotyl elongation as a proxy. 309 To investigate the functional ability of the elf3 gi oscillator to regulate its target genes, we first 310 monitored the expression of the key flowering-time genes GI, CO and FT under ND ( Figure 5A-C). 311 Consistent with previous reports, we detected a rhythmic expression of GI, CO and FT in WT (Fowler et 312 al., 1999;, with GI expressing during the day with the peak levels at ZT8, CO 313 showing dual peaks, a smaller one at ZT8 and another one at ZT16-20. The maximum levels of FT were 314 detected at dusk, at ZT12 ( Figure 5A-C). Consistent with the late flowering phenotype of the gi null 315 mutant, the expression of CO and FT was barely detectable in gi ( Figure 5B,C). In elf3, the expression of 316 GI was higher at almost all time points ( Figure 5A), consistent with the direct repression of GI by ELF3 317 Ezer et al., 2017). The expression of CO was higher during the early day and again 318 during the night, whereas FT expression was only elevated during the day at ZT4 (Figure 5A-C). The 319 expression pattern of CO and FT in elf3 gi was similar to gi. Notably, no diurnal peak of expression was 320 observed in the elf3 gi double mutant for any of the genes tested, with the overall expression hardly 321 fluctuating over the entire diurnal cycle ( Figure 5A-C). 322 We further validated these results by monitoring the expression of the major growth promoter PIF4 323 under ND ( Figure 5D). Consistent with their long hypocotyls, an overall higher expression of PIF4 was 324 observed in elf3 and gi ( Figure 5D). Furthermore, in gi, PIF4 followed a similar clock-regulated diurnal 325 pattern as that of WT, albeit with marginally but consistently higher levels ( Figure 5D). PIF4 expression in 326 elf3 also followed a diurnal pattern. However, it showed a characteristic light regulated profile, with a 327 gradual decrease in expression during the light period and a gradual increase during the dark period 328 ( Figure 5D). Interestingly, the PIF4 expression in the elf3 gi double mutant was completely different 329 from the diurnal patterns in WT and single mutants. Compared to WT, the level of PIF4 was higher in 330 elf3 gi at almost all time points, explaining for example its extreme growth phenotype shown in Figure 2. 331 Further, elf3 gi displayed neither the clock regulated PIF4 profile as observed for gi, nor the light 332 regulated expression as observed in elf3 ( Figure 5D). A closer examination revealed that the PIF4 levels 333 remained almost similar throughout the diurnal cycle with the exception of ZT16 where expression 334 levels were increased in comparison to other time points ( Figure 5D). Collectively, these data 335 demonstrate that both ELF3 and GI are required for clock entrainment and thereby for the generation of 336 rhythmic endogenous processes synchronized with the external signals. 337

Discussion: 338
The circadian clock is an important time keeping mechanism that synchronizes the internal cellular 339 mechanism to the external environment. Light is the primary cue that provides timing information to the 340 clock . While light sensing by the photoreceptors is well 341 understood, it remains unclear how this information is perceived by the central oscillator. Here, we 342 show that clock components ELF3 and GI are essential to perceive light input into the clock and thereby 343 for the measurement of the photoperiod. Absence of these components results in a dysfunctional 344 oscillator, even under diurnal conditions, failing to regulate photoperiod-responsive growth and 345 development. 346 Single loss of function mutants of individual EC components exhibit similar clock, hypocotyl and 347 flowering time phenotypes, indicating that they work cooperatively Herrero et al., 348 2012). Recent biochemical data has suggested that ELF3 can also function independently of the EC 349 . However, conclusive genetic evidence supporting the biochemical data is lacking. 350 Previous data reported a clear genetic hierarchy between ELF4 and GI with ELF4 being epistatic to GI in 351 control of hypocotyl elongation. Vice versa, GI is epistatic to ELF4 in flowering time regulation (Kim et al.,352 2012). In our study, we did not observe such genetic relationships for ELF3 and GI. Taking into account 353 that ELF3 and ELF4 function together in the EC , this is 354 somewhat surprising, supporting the proposed EC independent function for ELF3 . 355 The phenotypes we observed in single and double mutants for hypocotyl elongation suggest an additive 356 function of ELF3 and GI in controlling elongation growth ( Figure 1B, S1A-F), whereas in flowering time 357 regulation ELF3 and GI were epistatic to each other under SD and LD, respectively. In circadian clock 358 control, elf3 gi displayed similar additive/synergistic phenotypes ( Figure 3A-B) as reported for ELF4 and 359 GI. Collectively, in agreement with the biochemical data, our genetic analyses demonstrate that ELF3 360 function is not solely dependent on the EC. 361 ELF3 has been established as a repressor of growth that mainly works by acting on PIF4 (Nusinow et al., 362 2011;. Under diurnal conditions, the role of ELF3 as a growth inhibitor is undisputed. 363 However, under constant light, contradictory phenotypes of elf3 mutants were reported. Under LL, elf3 364 mutants displayed either similarly long or slightly longer hypocotyls Kim et al., 2005) 365 compared to WT (Doyle et al., 2002;. Consistent with previous data 366 Kolmos et al., 2011;Anwer et al., 2014;Box et al., 2014;Raschke et 367 al., 2015), under a range of photoperiods and light spectra, we consistently observed an elongated 368 hypocotyl of elf3 ( Figure 1B, S1A-F). However, under LL, elf3 was significantly shorter than WT ( Figure  369 1B, S1C). Further experiments under different light spectra revealed that the growth promoting function 370 of ELF3 was photoreceptor dependent. Specifically, the elf3 was shorter than WT under BB ( Figure 1C). 371 Interestingly, under these conditions PIF4 levels were still increased in elf3, but it failed to induce the 372 expression of its target genes IAA29 and YUC8, indicating the possibility of a decreased PIF4 activity 373 ( Figure 1E, S2B). Collectively, while our data consolidate the known growth inhibitory role of ELF3 in 374 PhyB mediated hypocotyl elongation, we propose a novel function of ELF3 as a growth enhancer under 375 BB. The underlying molecular mechanism of ELF3 mediated growth promotion remains unknown. 376 However, based on its known transcription/activity repressor function Ezer et al., 377 2017), it seems likely that ELF3 inhibits the function of a growth repressor under BB. If so, CRY1 would 378 represent an attractive candidate. In support of this hypothesis, CRY1 and PIF4 have been shown to 379 physically interact and bind to the same promoter regions Pedmale et al., 2015). This 380 12 binding decreases PIF4 transcriptional activity in a blue light dependent manner , which 381 could be explained by a competitive repressor-of-the-repressor model. The molecular mechanism by which GI controls growth is not fully understood. An elongated hypocotyl 386 of gi mutants under red and blue light suggested a repressive role in photoreceptor mediated growth 387 inhibition . Recent data demonstrated that GI requires PIF4 388 for growth regulation (de Montaigu et al., 2014;Fornara et al., 2015). Since the EC regulates PIF4 389  and ELF4 is epistatic to GI , a role of GI upstream of the EC in 390 growth regulation has been proposed (de Montaigu et al., 2014). However, our data, especially an 391 additive hypocotyl phenotype and increased levels of PIF4 in elf3 gi ( Figure 1B-E, 5D), advocate an 392 independent repressive action of GI on PIF4. 393 As growth and developmental phenotypes investigated in this study depend on the circadian clock, we 394 asked whether ELF3's and GI's function in the clock might be able to explain the observed effects. By a 395 "gating" mechanism the clock ensures that maximum growth happens at the correct time of day. In WT, 396 under LD growth rates peak in the early morning coinciding with the maximum expression of PIF4 397 .To coordinate this timing of growth rates, TOC1 and EC components including ELF3 398 repress growth during the late-evening and night, respectively Box et al., 2014;Zhu 399 et al., 2016). In this study, we demonstrate that GI is also essential for clock mediated gating. GI 400 represses growth during mid-day to late afternoon, thereby contributing to restricting growth peaks to 401 the morning, resulting in normal rhythmic growth. Consistently, the loss of day and night time gating 402 response in elf3 gi double mutants results in uncontrolled elongation growth ( Figure 2D). Based on these 403 observations we propose a model of rhythmic growth incorporating ELF3 and GI. In that model ELF3 and 404 GI gate growth mainly by repressing PIF4 during the night and late afternoon, respectively, allowing it to 405 accumulate only during the early morning under LD. The morning accumulation of PIF4 induces its 406 downstream targets that consequently trigger cellular growth ( Figure 5E). 407 The gating properties of the circadian clock are mainly dependent on its ability to synchronize internal 408 cellular mechanisms with the external environment. Although after entrainment the clock maintains the 409 same rhythm in the absence of the external input, in nature these free-running conditions almost never 410 exist. Thus, proper clock responses to consistent external cues during a diurnal cycle are crucial for the 411 synchronization of endogenous and environmental signals. Interestingly, arrhythmic clock genotypes, 412 such as null mutants of the EC members ELF3, ELF4 and LUX, as well as overexpressors of CCA1 and 413 TOC1, exhibit a non-functional oscillator under free-running conditions, but they are fully capable of 414 generating robust rhythms under diurnal conditions (Fowler et al., 1999;Hall et al., 415 2003;. Even higher order clock mutants including cca1-1 lhy-11 416 toc1-2, which lack the entire central oscillator, can generate rhythms under cycling conditions 417 . The data presented in this study demonstrate that the absence of the two 418 components ELF3 and GI is sufficient to make the oscillator arrhythmic under both free-running and 419 even under diurnal conditions ( Figure 4A-E). We demonstrated that ELF3 and GI serve as important 420 Zeitnehmers that are essential for clock entertainment. In their absence, the oscillator cannot perceive 421 external timing cues provided by cycling light conditions and thus fails to generate rhythmic oscillation 422 of the downstream endogenous outputs. A closer look at the transcriptional profile of the major core-423 13 oscillator genes and the clock-regulated output genes under diurnal conditions in elf3 gi suggests that 424 the entire clock-regulated transcriptome seems arrested ( Figure 4A-E). As such, even changes in the 425 environmental conditions during a diurnal cycle had no effect on the oscillator and were unable to 426 release the clock-regulated transcriptome from its arrested state ( Figure 5A-D). This should rationally 427 lead to a breakdown of any clock-control output pathway. Consistently, photoperiod-responsive 428 flowering and growth was disrupted in elf3 gi ( Figure 1A-B). Notably, light regulated processes that are 429 independent of the circadian clock seem to be intact in elf3 gi. A continuous inhibition of hypocotyl 430 length under increasing photoperiod ( Figure 1B, Tables S1-S2) along with marked differences in growth 431 rate during the light and dark phase in elf3 gi support this notion ( Figure 2D, Table S3). Collectively, our 432 data demonstrate that ELF3 and GI control the circadian clock Zeitgeber-Zeitnehmer interface, enabling 433 the oscillator to synchronize internal cellular mechanisms to the external environment. 434 Orthologues of ELF3 and GI have been identified in several higher plants. Both genes have been prime 435 breeding targets in crops for flowering time (Faure et al., 2012;Bendix et al., 2015;Panigrahi and 436 Mishra, 2015;. The elf3 gi double mutants develop rather normally and 437 flower at the same time irrespective of the photoperiod ( Figure 1A-B). If similar genetic and functional 438 relationships between ELF3 and GI exist in economically important crops as reported here for 439 Arabidopsis, breeders could develop photoperiod-insensitive varieties lacking ELF3 and GI that would be 440 independent of latitudinal photoperiodicity (Soyk et al., 2016). 441 14

Plant material 443
All genotypes used were in Ws-2 genetic background. The elf3-4 null mutant  was 444 previously described in . The gi-158 mutant was obtained in an 445 ENU (N-ethyl-N-nitrosourea) genetic screen and will be explained elsewhere. The gi-158 is possibly a null 446 mutant that contains a premature stop codon resulting in a truncated protein of 146/1173 amino acids. 447 The flowering time and hypocotyl phenotypes of gi-158 were very similar to the gi-11 null mutant 448 (Fowler et al., 1999) (Figure S2A-B). The double mutant elf3-4 gi-158 was generated by crossing elf3-4 449 and gi-158, and was confirmed by genotyping ( Figure S2C). Marker used for genotyping were: gi-158, 450 (forward ACTCATTACAACCGTCCCATCTA, reverse, GCGCATGAACACATAGAAGC (XbaI) elf3-4 (forward 451 TGCAGATAAAGGAGGGCCTA, reverse, ATGGTCCAGGATGAACCAAA. 452

Growth conditions 453
For luciferase assays, seeds were surface-sterilized and plated on MS medium containing 3% sucrose. 454 Following ∼3 days stratification at 4°C, seedlings were entrained for 7 days, either under LD, ND, SD 455 cycles (∼100µmol m-2 s -1 ) with constant temperature of 20°C (LD). The bioluminescence measurement 456 and data analysis was performed as described . For hypocotyl assays, seedlings 457 were grown on ATS medium, as described previously . Hypocotyl length was 458 determined for seedlings grown under varying photoperiod for 7 days or under RR or BB (light intensity  459 white fluorescent light, 90 µmol m -2 s -1 ; light intensity RR and BB: monochromatic LED, 20 µmol m -2 s -1 ). 460 The correct spectrum and intensities of red and blue light was confirmed by a spectrometer (UPRtek® 461 MK350S). Seedlings were imaged, and hypocotyl elongation was measured using the Rootdetection 462 program (http://www.labutils.de/rd.html). (1 cm above rosette leaves) as the total number of days to bolt. For all experiments, data loggers were 466 used to monitor the growth conditions. 467 Infra-red photography for growth rate measurement 468 Seedlings were grown as described above with the following exception: to facilitate imaging 469 unobstructed in air, seedlings were grown vertically on an agar ledge formed by removing part of agar in 470 a square petri plate. Seeds were placed in small ridges on top of the agar. Imaging was started as soon 471 as the cotyledons emerged from the seed coat. Photographs were taken every 60 minutes for 48 hours 472 in LD cycles (white fluorescent light: 30 µmol m -2 s -1 , constant 20°C). To image growth in day-night cycles 473 we built an infrared imaging platform consisting of a modified camera with IR long pass 830 nm cut 474 filters (Panasonic G5). Illumination was achieved using 880 nm IR backlights (Kingbright BL0106-15-29). 475 Image stacks were analyzed using ImageJ (Wayne Rasband, National Institutes of Health, USA, 476 http://rsb.info.nih.gov/ij). Data loggers were used to monitor the growth conditions. 477

Expression Analysis 478
Total RNA was isolated with NucleoSpin® RNA Plant (Macherey-Nagel) following the manufacturer's 479 protocol from 1-week-old seedlings entrained in 12L:12D (90 µmol m -2 s -1 , constant 20°C). Light 480 intensities for BB were 20 µmol m -2 s -1 . Quantitative RT-PCR, and primer sequences were previously 481 15 described ) with following modifications: ABsolute Blue qPCR SYBR Green 482 (ThermoFisher®) was used instead of iQ SYBR Green (Biorad). Agilent Mx3005P or AriaMx realtime 483 system (Agilent®) were used instead of BioRad. Data loggers were used to monitor the growth 484 conditions. 485 Error bars represent standard deviation (StD). n≥24. Letters above the bars represent statistically 490 significant differences calculated using one-way ANOVA (ANalysis Of VAriance) with post-hoc bars represent the standard error of the mean (SEM) of three biological replicates. Significance as 504 described above, P<0.05. See also Figure S1, Table S1, and Table S2. 505 taken every one hour using a modified Infra red camera. To measure new growth, the time-lapsed 508 images were imported into ImajeJ and hypocotyl length was measured (please see materials and 509 methods for details). Non-shaded are in the graph represents light period (day), and shaded area 510 represents dark period (night). Error bars represent standard error of the mean (S.E.M.), n≥8. 511 Experiment was repeated at least three times with similar results. See also Table S3. 512 Error bars represent standard deviation, n≥24. (B) Hypocotyl length of  Significance as described in Figure 1 within a specific photoperiod. (C) confirmation of the elf3 gi mutant 549 by genotyping. Two independent double mutant lines were obtained after crossing: elf3 gi (1) and elf 3 550 gi (2). After genotypic and phenotypic confirmation, only one line elf3 gi (1)   , ND (12h light: 12h darkness), and SD (8h light: 16h darkness). Flowering time was counted as number of days to 1cm bolt. Error bars represent standard deviation (StD). n≥24. Letters above the bars represent statistically significant differences calculated using one-way ANOVA (ANalysis Of VAriance) with post-hoc Tukey HSD (Honestly Significant Difference) Test, p<0.01. (B) Hypocotyl length of Ws-2, elf3, gi and elf3 gi under different photoperiods. Numbers at X-axis represent the length of the light period. For instance '8' typify (8h light: 16h darkness), and 12 (12h light: 12h darkness). DD and LL represent constant darkness and constant light, respectively. Seedling were grown for seven days under the respective photoperiod at constant 20°C. Error bars represent standard deviation. n≥18. Letters above the bars represent statistically significant differences among four genotypes under the specified photoperiod (ANOVA with post-hoc Tukey HSD Test, p<0.01). (C) Hypocotyl length of Ws-2, elf3, gi and elf3 gi under constant red (RR) or constant blue (BB) light. Plants were grown for 7 days under monochromatic red or blue light at constant 20°C before the pictures were taken and hypocotyl length was measured. Significance as described in (B) calculated separately for RR and BB. Experiment was repeated at least three times with similar results. Starting from the third day photographs were taken every one hour using a modified Infra red camera. To measure new growth, the time-lapsed images were imported into ImajeJ and hypocotyl length was measured (please see materials and methods for details). Non-shaded are in the graph represent light period (day), and shaded area represents dark period (night). Error bars represent standard error of the mean (S.E.M.), n≥8. Experiment was repeated at least three times with similar results. Time (