Chimeric Activators and Repressors Define HY5 Activity and Reveal a Light-Regulated Feedback Mechanism

The first exposure to light marks a crucial transition in plant development. This transition relies on the transcription factor HY5 controlling a complex downstream growth program. Despite its importance, its function in transcription remains unclear. Previous studies have generated lists of thousands of potential target genes and competing models of HY5 transcription regulation. In this work, we carry out detailed phenotypic and molecular analysis of constitutive activator and repressor HY5 fusion proteins. Using this strategy, we were able to filter out large numbers of genes that are unlikely to be direct targets, allowing us to eliminate several proposed models of HY5's mechanism of action. We demonstrate that the primary activity of HY5 is promoting transcription, and that this function relies on other, likely light-regulated, factors. In addition, this approach reveals a molecular feedback loop via the COP1/SPA E3-ubiquitin ligase complex suggesting a mechanism which maintains low HY5 in the dark, primed for rapid accumulation to reprogram growth upon light exposure. Our strategy is broadly adaptable to the study of transcription factor activity. Lastly, we show that modulating this feedback loop can generate significant phenotypic diversity in both Arabidopsis ( Arabidopsis thaliana ) and tomato ( Solanum lycopersicum ). show that HY5-VP16 and HY5ΔN77-VP16, but not HY5ox or HY5ΔN77ox , can de-etiolation in the dark. Furthermore, HY5-VP16 activates reporter expression in insect cells and leads to phenotypic changes in tomatoes, while HY5 alone cannot. Taken these provide definitive evidence for the requirement of interacting partners for HY5 activity. our finding that HY5 can promote de-etiolation in the dark when stabilized and fused to an activation domain suggests that endogenous HY5-interacting transcriptional activators mainly in the light. is consistent with a model in which HY5 the light-dependent growth program by binding target gene them as targets for light-dependent transcriptional activators. Identification characterization of the complexes that are needed for HY5 activity will future It bZIP the DNA when all three factors HY5 TFs form


INTRODUCTION
Plants continuously monitor the environment to tune development and optimize performance. They sense changes in factors such as temperature, water availability, gravity, nutrients, and many aspects of light, the latter of which are sensed by a diverse array of photoreceptors and feed into a variety of growth programs (Chen et al., 2004).
De-etiolation is one of the best-understood light-mediated developmental switches, activated very early in development, immediately following germination (Arsovski et al., 2012). When light is a limiting factor, the seedling will use etiolated growth, characterized by an elongated embryonic stem (hypocotyl) with the apical hook and tightly-closed embryonic leaves (cotyledons) protecting the meristematic tissue.
Rapid elongation during this stage allows the plant to quickly emerge from the soil and gain access to light.
Upon sensing sufficient light, it transitions to de-etiolated growth. This is characterized by a deceleration of hypocotyl growth, promotion of cotyledon expansion and accumulation of chlorophyll. For the purposes of this article, we define photomorphogenesis as germination and growth in light, without a soil-like darkgrowth stage. De-etiolation and photomorphogenesis are both triggered by photoreceptors transducing light signals to a set of key transcription factors (TFs) that change the expression of thousands of genes (Leivar et al., 2009).
One of these TFs, the basic domain/leucine zipper (bZIP) TF HY5, is a master regulator of deetiolation and photomorphogenesis. hy5 was identified genetically in the now classic long hypocotyl screen carried out by Koornneef et al. (1980). Of the five loci identified in that study, hy5 was the only transcription factor. This key factor has since been shown to control a network consisting of thousands of genes as plants modify their growth programs to fit the environment. Plants lacking HY5 are characterized by long hypocotyls under all light conditions, suggesting that HY5 responds to signals from several photoreceptor families to inhibit hypocotyl elongation (Koornneef et al., 1980;Chory, 1992). Studies of HY5 mutant and over-expression plants have implicated HY5 in regulating chlorophyll and anthocyanin biosynthesis, primary and lateral root development, shade and high temperature responses, flowering time, and many other processes (Oyama et al., 1997;Ang et al., 1998;Holm et al., 2002;Andronis et al., 2008;Delker et al., 2014;Nozue et al., 2015;Gangappa and Botto, 2016). HY5 interacts with many light-and growthrelated TFs (Holm et al., 2002;Datta et al., 2007;Singh et al., 2012;Jang et al., 2013;Abbas et al., 2014;Gangappa and Botto, 2016). It is worth noting that the transcription of most of these factors is also regulated by light.
HY5 is directly regulated at the protein level by another key control element in the light response: the E3 ubiquitin ligase complex comprised of the CONSTITUTIVELY PHOTOMORPHOGENIC 1 (COP1) and SUPPRESSOR OF PHYA-105 (SPA) proteins (COP1/SPA). This complex degrades many photomorphogenesispromoting factors in the dark and is deactivated upon light perception (Deng et al., 1992;Lau and Deng, 2012).
How HY5 regulates the transcription of its target genes remains an open question. Genomic studies have found that HY5 can bind to thousands of genes (Lee et al., 2007;Zhang et al., 2011;Kurihara et al., 2014). It has been characterized as a transcriptional activator, a repressor, or both (Ang et al., 1998;Lee et al., 2007;Ruckle et al., 2007;Kindgren et al., 2012;Delker et al., 2014;Gangappa and Botto, 2016;Noren et al., 2016;Xu et al., 2016;Gangappa and Kumar, 2017;Nawkar et al., 2017;Zhang et al., 2017). HY5 has also been reported to feed back to promote its own transcription (Abbas et al., 2014;Binkert et al., 2014). Complicating matters, in vitro work suggests that HY5 does not have its own activation or repression domain (Ang et al., 1998). Current hypotheses regarding HY5 activity include: 1) HY5 controls growth not only by its own DNA-binding, but by regulating the DNA binding of its interacting partners (Ram et al., 2014); 2) HY5 may have activator or repressor activity depending on the binding partner (Ruckle et al., 2007;Kindgren et al., 2012); 3) HY5 may compete with other TFs for binding sites on DNA (Toledo-Ortiz et al., 2014;Gangappa and Kumar, 2017;Nawkar et al., 2017). These conflicting hypotheses thus require a new approach to clarify the exact molecular function of HY5.
Chromatin-Immunoprecipitation-sequencing (ChIP-seq) and RNA-sequencing (RNA-seq), experiments often end up with thousands of differentially expressed genes between genotypes or conditions. While this may provide an interesting mine for hypothesis generation, there is also the risk of false positives or indirect effects on transcription through downstream transcription factor networks. Additionally, while many studies combine RNA-seq with ChIP-seq to find target genes, the multi-layered nature of transcription control in the context of epigenetic modifications, diverse signaling mechanisms, and higher-order protein complexes means that transcription factor binding does not necessarily equal activity at a given locus (Eeckhoute et al., 2009;Schacht et al., 2014). Identifying the most likely direct targets thus requires the development of a higher-order filtering strategy to determine sites of physiologically relevant transcription regulation.
We generated chimeric constitutive repressor and activator forms of HY5 to understand the function of HY5 during early seedling development. We show that HY5 promotes transcription of target genes. Using these new reagents, in combination with RNA-seq and ChIP-seq, we identify a highconfidence set of direct HY5 target genes, as well as genes likely to be indirectly controlled by HY5dependent transcriptional networks. We found SPA1, SPA4, and COP1 among the direct targets of HY5.
We propose that this transcriptional control, coupled with COP1/SPA-mediated HY5 degradation in the dark, is a mechanism underlying the plant's ability to quickly switch from etiolated to de-etiolated growth upon light perception. In addition, we provide in planta evidence that HY5 lacks its own activation domain, supporting the hypothesis that the observed transcriptional activation is driven by preferential interaction with activating TFs. We propose that this multiplexed strategy may be generalizable to the study of transcription factor function.

Chimeric HY5 Proteins Have Distinct Effects on De-Etiolation
To determine the dominant transcriptional regulation activity of HY5 during de-etiolation, we generated chimeric HY5 variants, by adding a transcriptional silencing motif (EAR repressor motif of Arabidopsis SUPERMAN gene-SRDX) (Ohta et al., 2001;Hiratsu et al., 2002) or the activation domain from VP16 (Triezenberg et al., 1988). The SRDX and VP16 domains have been shown to effectively repress or activate gene expression, respectively, when fused to plant transcription factors (Shani et al., 2009;Fujiwara et al., 2014). We expressed HY5-SRDX, HY5-VP16, and the native form of HY5 (HY5ox), with an N-terminal FLASH tag in the hy5 mutant background ( Figure 1A) (see Methods). We reasoned that the fusion proteins would specifically repress (HY5-SRDX) or activate (HY5-VP16) the transcription of HY5 direct targets.
Since HY5 was shown to regulate its own promoter (Abbas et al., 2014;Binkert et al., 2014), we expressed these HY5 variants under the control of the constitutive 35S promoter. While overexpression under this promoter may increase the false positive rate, it provides a useful search space in our sorting strategy.
Phenotypic and molecular observations in these lines were filtered using the differences between the Col-0 background control and the hy5 loss-of-function mutant as a primary benchmark.
We tested the effects of these fusion proteins on a series of hy5 related phenotypes during deetiolation, including hypocotyl elongation, cotyledon expansion, and chlorophyll content (as a proxy for chloroplast maturation). Expression of HY5ox and HY5-VP16 rescued the long hypocotyl of hy5 mutant seedlings, whereas HY5-SRDX caused a dramatic increase in hypocotyl length in continuous simulated white light (sWlc -blue, red, and far-red LED lights) as well as monochromatic red, blue, and far-red light conditions (Figures 1B and 1C;Supplemental Figure 1). In addition, HY5-SRDX increased and HY5-VP16 decreased cotyledon size, and HY5-SRDX strongly reduced chlorophyll content, beyond that observed in the hy5 mutant ( Figures 1D and 1E). Together, these data show that the constitutive transcriptional activator HY5-VP16 phenotypically follows the wild-type and HY5ox lines, while the constitutive repressor HY5-SRDX produces exaggerated phenotypes in the same direction as the hy5 loss-of-function mutant. This indicates that HY5 mainly works as an activator during de-etiolation.
We then examined processes beyond de-etiolation, finding that expression of HY5-VP16 and HY5ox rescued the hy5 mutant phenotypes in root growth, flowering time, shade-induced hypocotyl elongation and anthocyanin level, while HY5-SRDX either did not change or exacerbated the mutant phenotype (the HY5-SRDX line 46 did not always follow this pattern due to its severe growth defect) Supplemental Figures 2A and 2B). Recent studies have reported that HY5 is required as a transcriptional repressor to inhibit hypocotyl elongation at ambient temperature, and that this activity is reduced in warm temperature conditions (Delker et al., 2014;Gangappa and Kumar, 2017). Surprisingly, and contrary to the published model in which HY5 repression activity inhibits growth, we observed that 5 seedlings expressing the HY5-SRDX constitutive repressor had longer hypocotyl than the controls, while those expressing HY5ox and HY5-VP16 had shorter hypocotyls in 20°C and 28°C (Figures 2A and 2B).
The phenotype of the chimeric HY5-SRDX suppressor suggests that releasing HY5 transcriptional repression activity is unlikely to be the major cause of the high temperature-mediated growth. To test this further, we examined the expression of the growth-related genes SAUR19, IAA19, and EXP8 which were previously suggested to be repressed by HY5 (Jing et al., 2013;Delker et al., 2014;Gangappa and Kumar, 2017). We hypothesized that if HY5 directly regulates their expression, these transcripts will be reduced in HY5-SRDX and increased in HY5-VP16 transgenic lines. All three genes were upregulated in response to high temperature in wild-type and hy5 plants. However, the expression in HY5ox and HY5-VP16 plants was lower than in wild type, and higher than in wild type in HY5-SRDX plants at both 20°C and 28°C ( Figure   2C-E). The observed repression of these growth-related genes is likely due to factors acting downstream of HY5, rather than direct HY5 action. These results are in agreement with the phenotypes we observed, and suggests that HY5 works as an activator in many processes.

Identifying High Confidence Direct Targets of HY5 Using Activator and Repressor Chimeras
All previous genomic studies searching for HY5 direct targets arrived at the same conclusion: HY5 can be both an activator and a repressor (Lee et al., 2007;Zhang et al., 2011;Kurihara et al., 2014). However, our results so far indicate that HY5 works mainly as an activator. To resolve this discrepancy, we performed RNA-seq and ChIP-seq with plants expressing our HY5 fusion proteins (using line number 5 from each construct). We reasoned that the expression of a direct HY5 target would respond to VP16 and SRDX by increasing or decreasing expression, respectively. These gene expression patterns could then be compared to those of HY5ox and hy5, and correlated with HY5 binding sites identified with ChIP-seq analysis, yielding a high-quality and high confidence list of direct HY5 target genes.
Our ChIP-seq data recovered the ACGTG motif (bases 2-6 of the canonical G-box), the strongest binding motif for all HY5 fusion proteins, suggesting that the fusion of HY5 to VP16 or SRDX did not change the specificity of the DNA binding domain ( Figures 3A and 3B).
We then validated the transcriptional regulation activity of the chimeric HY5-VP16 and HY5-SRDX constructs by RNA-seq. There was no clear directional regulation of the differentially expressed genes (upregulated or downregulated) in HY5-VP16 and HY5-SRDX transgenic lines relative to hy5 mutant plants ( Figure 3C-left; Supplemental Data Set 1). We filtered this list by selecting only the genes with HY5 peaks in all of our ChIP-seq samples (high confidence HY5 binding sites). In this new list 79% were upregulated in HY5-VP16 and downregulated in HY5-SRDX lines, indicating that HY5-VP16 activates and HY5-SRDX repress gene expression as expected (Figure 3C-right; Supplemental Data Set 1). Only 12 genes had the opposite response, and 10 of them were also downregulated in HY5ox. This strongly suggests that the binding at these loci is not the determining factor for expression and supports the hypothesis that these 6 genes are being controlled by a repressor that is itself transcriptionally activated by HY5. The expression of the 15 remaining genes changed in the same direction in both HY5-SRDX and HY5-VP16 lines.
Transcriptional regulation of this group is likely more complicated, possibly influenced by factors other than or further downstream of HY5. Overall, the majority of gene expression profiles track well with the activator and repressor activity of VP16 and SRDX, respectively.
Having confirmed that the VP16 and SRDX fusion proteins regulate gene expression in the expected directions at the level of the whole transcriptome, we began our search for the crucial set of genes that are regulated directly by HY5. To increase the stringency and specificity of this analysis, we looked for genes that changed expression in hy5 and in HY5ox (either up or down), were upregulated in HY5-VP16, and downregulated in HY5-SRDX. 297 genes satisfied these criteria ( Figure 3D; Supplemental Data Set 2).
All but two genes on this list were downregulated in hy5 compared to wild type, suggesting that they are positively regulated by HY5. We found that 81% of the genes were bound by HY5 in at least four samples and only 3% lack binding in all eight samples. We propose that these are the most likely direct HY5 targets in vivo. In contrast, previous studies searching for potential HY5 direct targets by comparing expression data with chromatin occupancy found that only 12% (Kurihara et al., 2014), 19% (Lee et al., 2007 or 43% (Zhang et al., 2011) of the differentially expressed genes in hy5 also had a HY5 ChIP peak.
The targets identified here represent both characterized and novel facets of HY5 physiology, including previously described HY5 targets, such as HY5-HOMOLOG (HYH), B-BOX (BBX22, BBX24, BBX32), and LONG HYPOCOTYL IN FAR-RED (HFR1). Importantly, we filtered out growth regulated genes previously suggested to be repressed by HY5, including SAUR19, IAA19, and EXP8. These three genes were upregulated in HY5-SRDX and downregulated in HY5-VP16 as we shown earlier (Figures 2C-2E, Supplementa Data Set 3). We consider it highly unlikely that these strong activator and repressor domains would act in an opposite role (activator becomes a repressor, or repressor becomes an activator) at these specific loci, and therefore conclude that they are probably regulated by factors downstream of HY5. In addition, we found that HY5 can directly regulate the expression of SCHLAFMÜTZE ( While the binding of HY5 to the COP1 promoter was previously described (Huang et al., 2012;Binkert et al., 2014;Xu et al., 2016), we wanted to confirm the binding to the SPAs in vitro. To this end, we tested HY5 activity on these promoters using a luciferase assay in the Drosophila S2 system. We confirmed HY5 transcriptional activation on the SPA1 promoter, but not on the COP1, SPA3, or SPA4 promoters (Supplemental Figure 5B). This result suggests that either the binding of HY5 to the SPA1 promoter is stronger than the binding to the other promoters, or that the COP1, SPA3, and SPA4 promoters are partially repressed in S2 cells. In addition, we found in a recently published ChIP-seq data set that HY5 binds to the promoters of SPA1, SPA3, SPA4, and COP1 when expressed under its native promoter (Hajdu et al., 2018).
These findings increase our confidence that HY5 binds to these promoters in vivo, and that our results are most likely not an artifact of the 35S promoter driving expression of our transgenes.
Our initial RNA-seq was done on whole seedlings, so we examined the contribution of organspecific expression of the 297 HY5-regulated genes during a 12-hour (12h) de-etiolation time course in wild-type hypocotyls and cotyledons (Supplemental Figure 5C; Supplemental Data Set 4). The majority (96%) were upregulated, in at least one time point, in both the hypocotyl and the cotyledon. Thus, we conclude that HY5 promotes transcriptional activation near-uniformly in multiple sub-regions of the plant.

HY5 Regulates the Expression of COP1/SPA Complex Components
Previous studies on hy5 mutants or overexpression of HY5 in wild-type seedlings did not describe modified growth patterns in the dark unless combined with hormone treatments or mutants which promote de-etiolation in the dark (Ang and Deng, 1994;Ang et al., 1998;Alabadi et al., 2008;Stracke et al., 2010;Li and He, 2016;Shi et al., 2018). Consistent with these reports, our HY5ox lines did not show any notable phenotypes in etiolated seedlings. However, one of the 12 HY5-VP16 lines (#5), was partially de-etiolated in the dark, with shorter stature and an open apical hook (Figures 4A-4C; Supplemental Figure 1F). The HY5 protein and RNA levels in our HY5ox lines were higher than in HY5-VP16 or HY5-SRDX, and all chimeric proteins were similarly stabilized in the light (Supplemental Figure 6), suggesting that the dark phenotypes of HY5-VP16 #5 are due to activity rather than expression level.
Two converging observations indicate that phenotypes induced by HY5-VP16 are highly sensitive to dosage or expression level. First, we noticed that many of the HY5-VP16 lines had a tendency to be silenced. Second, we found that HY5 directly promotes SPA1, SPA3, SPA4, and COP1 expression, possibly limiting HY5 protein accumulation ( Figures 3B and 3E; Supplemental Data Set 2). These observations might explain the lack of dark phenotypes in medium-or weakly-expressing HY5ox or HY5-VP16 lines. We hypothesized that the COP1/SPA-mediated degradation may limit HY5-VP16 activity, therefore, we overexpressed HY5 lacking the COP1-interacting domain (HY5∆N77) with or without VP16 (Ang et al., 1998).
Without this domain, HY5 cannot be degraded by COP1, thereby increasing its abundance and activity (Ang et al., 1998). None of the 10 HY5∆N77ox lines tested showed notable phenotypes in the dark; however, five out of 10 lines expressing HY5∆N77-VP16 showed a partly de-etiolated phenotype in the dark ( Figures   4D-4F). In support of our interpretation, HY5∆N77-VP16 line #2 had very low expression of the chimeric protein, and phenotypically resembled the wild type (Figures 4D-4F and 4I; Supplemental Figure 7A). The partial de-etiolation of HY5∆N77-VP16 suggests that HY5 may control its own degradation via COP1/SPA transcription. We therefore examined SPA1, SPA3, SPA4, and COP1 RNA levels in HY5∆N77-VP16 and HY5∆N77ox dark-grown seedlings. The expression of each of these components in the dark was higher in HY5∆N77-VP16 than in HY5∆N77ox, supporting our hypothesis that HY5 activates their transcription (Figures 4G and 4H;Supplemental Figure 7B and 7C). Consistent with this, we also found that COP1 and SPA1 protein levels were higher in HY5∆N77-VP16 than in HY5∆N77ox (Figures 4I and 4J;Supplemental Figures 7D and 7E). This result also suggests that HY5 depends on other TFs, potentially other targets of the COP1/SPA complex in the dark, to promote transcription. We did not observe any difference in gene expression between HY5-VP16 and HY5ox dark-grown seedlings in our RNA-seq data (Supplemental Figure 7F). While the expression of SPA1, SPA3, SPA4 was higher than wild-type in HY5ox and HY5-VP16 etiolated seedlings, this degree of over-expression evidently was not enough to promote de-etiolation.
Together, these data indicate that the negative feedback loop between HY5 and COP1/SPAs can be uncoupled by disrupting the HY5-COP1 interaction.
These plants were very small, with short petioles and small leaf blades, demonstrating the importance of tight regulation over HY5 in all stages of development ( Figure 5A, Supplemental Figure 8A). There was a clear gradient in plant size and petiole length of the adult plants, from HY5∆N77-VP16 (the smallest) to HY5-SRDX (the biggest), with HY5ox and hy5 being similar to wild type. HY5∆N77ox was slightly smaller than wild type, but larger than HY5∆N77-VP16. While the protein levels of HY5∆N77 were lower than those of full-length HY5, in both native and VP16 constructs, the mature HY5∆N77 plants were smaller, indicating that HY5∆N77 is a hyperactive form of HY5. This is consistent with a previous study (Ang et al., 1998) (Figures 5A and 5B; Supplemental Figure 8A). We therefore conclude that in mature plants, as in etiolated seedlings, HY5-dependent phenotypes are controlled by a HY5-COP1/SPA feedback loop.

HY5 Transcriptional Control Activity Requires Additional Factors
HY5-VP16 has stronger phenotypic effects then HY5ox in the dark and at the adult stage, suggesting that HY5 is dependent on the expression of other proteins to activate transcription. To demonstrate this further, we compared HY5 activity in tobacco with the ability of HY5 to activate gene expression in Drosophila S2 cells, which lack the plant proteins that interact with HY5 to form the activation complex. Using luciferase reporter assays, we tested HY5's ability to activate a synthetic promoter made from four tandem repeats of the G-box (4x G-box), which HY5 has been shown to bind in vitro (Shi et al., 2018), and the CHS promoter (CHSp), which was previously reported to be a direct target promoter of HY5 (Ang et al., 1998). HY5 did not change the basal luciferase expression from either the promoters in S2 cells, while HY5-VP16 strongly activated expression from both (Supplemental Figure 8B). Of interest, co-expression of HY5 and HY5-VP16 reduced luciferase activation from both promoters, supporting the hypothesis that HY5 lacks its own activation domain. By contrast, HY5 can weakly activate the expression of CHSp without VP16 in tobacco leaves (Supplemental Figure 8C). This suggests that plant-specific transcriptional activators are required for HY5 activity.

Chimeric Arabidopsis HY5 Can Modulate Growth in Tomato
Early seedling establishment is crucial for survival in many crops. For example, tomato plants with long hypocotyls are quite fragile, which can lead to the loss of many seedlings during mechanical planting (Brigard et al., 2006). Conversely, early elongation may aid competition against weeds. HY5 plays key roles in regulating hypocotyl growth in tomato (Liu et al., 2004). Recent work in tomato has described strategies for generating phenotypic gradients (Rodriguez-Leal et al., 2017;Israeli et al., 2019). Given our observation of HY5-dependent size gradients in Arabidopsis, we tested the utility of chimeric Arabidopsis HY5 proteins with an N-terminal DOF (HA-YFP-HA) tag, in generating phenotypic diversity in tomato plants. Similar to Arabidopsis, AtHY5-VP16 dramatically inhibited hypocotyl growth in tomato seedlings and generally slowed growth in later stages, resulting in bushy and dwarf plants ( Figures 5C-5E).
Overexpression of AtHY5-SRDX, however, led to longer hypocotyls and seedling lethality. Interestingly, while it was recently shown that overexpression of the tomato HY5 (SlHY5) increased anthocyanin levels (Liu et al., 2018), AtHY5ox tomato plants did not visibly differ from the wild type. All together, these results suggest that Arabidopsis HY5 may be able to accomplish some, but not all functions of tomato HY5.
Nevertheless, the phenotypic changes we observed in AtHY5-VP16 and AtHY5-SRDX plants indicate that modification of HY5 activity can achieve dynamic control of growth. Fine tuning temporal and spatial control over HY5 expression will be the next critical steps in optimizing this process.

DISCUSSION
HY5, a bZIP transcription factor, plays a pivotal role in the transcriptional response of plants to changes in the local light conditions. Despite its importance to the biology of plants, HY5's mechanism of action has been difficult to define. In this study, we used very stringent criteria to show that HY5 promotes transcriptional activation of its direct targets. Moreover, we define a negative feedback loop whereby HY5 promotes expression of its negative regulators in the COP1/SPA complex. This use of a number of stringent criteria allowed us to filter out a large number of genes that were unlikely to be direct targets of HY5, thereby allowing us to eliminate several proposed models of HY5's mechanism of action.

A Multiplex Strategy Using Constitutive Chimeras Defines Transcriptional Activity of HY5
Determining a causal relationship between TF binding and changes in gene expression is very challenging. In this study, we improve upon the existing RNA-seq/ChIP-seq pipeline. We directed HY5 to work as a repressor (HY5-SRDX) or as an activator (HY5-VP16) and combined RNA-seq and ChIP-seq with rigorous phenotypic analysis of these chimeric constructs. This approach allowed us to define highconfidence direct target genes and revealed that transcriptional activation is the major physiologically significant function of HY5. We identified a relatively small number of HY5 direct targets in comparison with previous studies (Lee et al., 2007;Zhang et al., 2011;Kurihara et al., 2014). This disparity is likely due to two reasons: (1) we collected our samples soon after the shift to light, while previous studies examined plants that had been grown in light for several days, which will lead to many indirect effects due to the phenotypic differences. (2) The use of the chimeric HY5 constructs enabled us to filter out many indirect targets. In support of this, we found that about 10% of our HY5 direct targets are TFs, indicating sequential layers of transcriptional control and potentially explaining the downregulation of genes that were thought to be direct targets of HY5.
Deeper examination of transcription factor function will require that these methods be further refined. Our use of the constitutive 35S promoter was one solution to the problem of autoactivation by HY5. Nevertheless, the plants were sampled at steady state. This issue could be resolved using an inducible system to transiently express TFs of interest and examine short-time-scale changes in gene expression in a pulse-chase-like format. In the case of TFs that have strong intrinsic activation or repression activity, it may be necessary to identify and mutate these domains to an inactive form prior to fusing the protein to known activator and repressor domains in the strategy we have described here.

Refining Models of HY5 Function
Several models of HY5 function have been put forward in recent years, primarily relying on coanalysis of chromatin binding and gene expression. In light of our strict analysis, we examine areas of agreement and disagreement among these models. First, several groups have suggested that HY5 can act as a transcriptional repressor. This idea was supported by data suggesting that HY5 can compete with other TFs for DNA binding sites (Li et al., 2010;Xu et al., 2016), and that it can lead to down-regulation of gene expression through interaction with chromatin remodeling factors (Jing et al., 2013;Zhao et al., 2019). We cannot rule out the possibility that HY5 may contribute to transcriptional repression in some rare or specific cases. However, the contributions of these events to the tested phenotypes are not detectable in our conditions, and the proposed target genes examined in these studies do not respond in the predicted directions to the VP16 or SRDX fusion constructs.
It was also suggested that HY5 and PIFs work antagonistically, and compete for binding at the same target genes, leading to opposing phenotypes during de-etiolation at elevated temperature (Toledo-Ortiz et al., 2014;Gangappa and Kumar, 2017). However, we find only a few common targets between HY5 and PIFs, suggesting that this model is not sufficient to explain the combined functions of these TFs during deetiolation (Supplemental Figure 9). Both of these proposed models suggest that HY5 acts as a repressor, however these are incompatible with our findings demonstrating the phenotypic and molecular requirement for HY5-based transcriptional activation. In addition, we demonstrated that the growth-related genes SAUR19, IAA19 and EXP8, which were shown to be directly activated by PIF4 and PIF5, are probably not directly regulated by HY5 (Franklin et al., 2011;Hornitschek et al., 2012;Sun et al., 2013;Ma et al., 2016;Gangappa and Kumar, 2017) (Figures 2C-2E). We therefore suggest that many of the cases in which HY5 is considered a repressor are in fact likely to be effects of the downstream transcriptional network. For example, HFR1, which represses PIF4 activity and was among our high-confidence HY5 direct targets, may account for the antagonistic activity of HY5 and PIF4.
Another potential interaction between HY5 and PIF activities is highlighted by the finding that COP1-SPA1 promotes PIF1 rapid degradation in the light and SPA1 directly phosphorylates PIF1 (Zhu et al., 2015;Paik et al., 2019). This is consistent with our identification of SPA1 as a high-confidence HY5 direct target. The observation that HY5∆77 is hyperactive relative to full-length HY5 provides some insight into this issue, in that the increased activity of degradation-resistant HY5∆77 leads to more SPA1. This increased population of SPA1 would then promote the degradation of PIF1, thus reducing PIF-dependent growth while maintaining the HY5 pool. We suggest a path in which, in wild-type plants, HY5 promotes the expression of SPA1 in the light, thus promoting degradation of PIF1 and providing a mechanism for maintaining a balance between HY5 and PIFs.

HY5 Requires Interacting Partners to Promote Gene Activation
Here we show that HY5-VP16 and HY5ΔN77-VP16, but not HY5ox or HY5ΔN77ox, can promote deetiolation in the dark. Furthermore, HY5-VP16 activates reporter expression in insect cells and leads to phenotypic changes in tomatoes, while HY5 alone cannot. Taken together, these observations provide definitive evidence for the requirement of interacting partners for HY5 activity. Additionally, our finding that HY5 can promote de-etiolation in the dark when stabilized and fused to an activation domain suggests that endogenous HY5-interacting transcriptional activators are expressed mainly in the light. This is consistent with a model in which HY5 directs the light-dependent growth program by binding target gene promoters, thereby marking them as targets for light-dependent transcriptional activators. Identification and characterization of the complexes that are needed for HY5 activity will be of future interest. It was shown that the human bZIP TFs FOS and JUN, together with the TF NFAT, form a complex that binds strongly to the DNA only when all three factors are present (Macian et al., 2000). HY5 may similarly require TFs from different families to form a complex that strongly binds to DNA and promotes transcription.

Feedback between HY5 and the COP1/SPA Complex Readies Seedlings for De-Etiolation
A seedling's ability to switch from etiolation mode to de-etiolation upon light exposure is an important factor in determining its success in development. HY5 is a key activator of this process and, as such, is tightly regulated. We show here that HY5 directly promotes the expression of SPA1, SPA4, COP1 (as previously described), and to some extent SPA3 by directly binding to their promoters. As a result, accumulation of HY5 in the dark leads to an increase in the COP1/SPA complex, and thus its own degradation, and forms a negative feedback loop between HY5 and COP1/SPA, balancing the level of HY5.
This negative feedback is consistent with a previous study showing that plants expressing VP16-HY5 are similar to wild-type (Stracke et al., 2010), as reaching an active threshold concentration while remaining vulnerable to COP1-mediated degradation is likely a rare event. By using HY5ΔN77-VP16, we were able to simultaneously uncouple this feedback loop and eliminate the need for HY5 interacting partners, thus demonstrating the potential detrimental effect of HY5 accumulation in the dark. For example, if a seedling switches to a de-etiolated growth mode before emerging from the soil, it will never reach a light source, and if it activates the chloroplast maturation program prior to seeing light, it sensitizes itself to damaging reactive oxygen species when it is finally exposed (Seluzicki et al., 2017). We propose a model, based on our results and the existing literature, in which HY5 is regulated at the transcript level by a number of factors including itself. At the protein level, HY5 is controlled by COP1/SPA, multiple photoreceptors, and by the availability of other light-dependent HY5-interacting protein(s) (Figure 6) (Ang et al., 1998;Lau and Deng, 2012;Zheng et al., 2013;Abbas et al., 2014;Binkert et al., 2014;Lu et al., 2015;Yang et al., 2018;Lau et al., 2019;Ponnu et al., 2019;Xu, 2019). In the dark, HY5 and its known partners are degraded, consistent with the observation that overexpression of HY5 alone cannot promote de-etiolation. The negative feedback loop between HY5 and COP1/SPA in the dark restricts HY5 activity, limiting it to a small active pool. Upon exposure to light, the expression of COP1/SPA increases but their activity on HY5 is inhibited, allowing HY5 protein levels to rapidly increase, along with other interacting proteins. As they accumulate together, these proteins form complexes with HY5 which bind to DNA and activate gene expression. Importantly, none of the single or double mutants knocking out known HY5-interacting TF genes can mimic the hy5 long hypocotyl phenotype in all light conditions (Holm et al., 2002;Datta et al., 2008;Kushwaha et al., 2008;Jang et al., 2013;Zhang et al., 2017). This suggests that HY5 acts through multiple parallel interactions with other TFs to run this light-dependent growth program.
The identification of a HY5-COP1-SPA negative feedback loop reveals an intervention point for directed modulation of growth properties. Here we show that by combinatorially directing transcription activity and enhancing protein stability we can generate a phenotypic gradient in plant size. This raises another hypothesis, which remains to be tested, whereby the negative feedback is important for the 13 maintenance of COP1/SPA levels during the day, becoming active at dusk and degrading HY5 and other proteins. We also show that chimeric HY5 proteins can modulate growth in tomato. We therefore propose that this feedback loop represents a distinct access point to tune desirable growth properties by modulating the strength of transcriptional activation and stability via the interaction with the COP1/SPA complex.

Plant Materials, Growth Conditions, and Seedling Measurements
All

DNA Constructs and Plant Transformation
The multisite Gateway system (Invitrogen) was used. To generate HY5ox, HY5-VP16, and HY5-SRDX gene constructs, full length HY5 coding sequence (with or without stop codon) was PCR amplified from Col-0 cDNA and recombined into pDONR-P2RP3. KpnI and XhoI were used to introduce the VP16 (PCR product) and SRDX (primers synthesized and annealed) coding sequences to the C terminus of HY5.
HY5ΔN77 was generated by PCR from plasmids containing HY5ox and HY5-VP16 and cloned into pDONR-P2RP3. Different HY5 versions were cloned downstream of 2x35Spro in pDONR-P4P1R and FLASHv2 (the 2xStrepII-6xHis-protease cleavage site-3xFlag) in pDONR-P221, respectively. Agrobacterium tumefaciens GV3101 carrying the various constructs using the floral dip method (Clough and Bent, 1998). Segregation analyses of antibiotic resistance were used to isolate single-insertion homozygous lines except for 35Spro:FLASH-HY5ΔN77-VP16, for which heterozygous lines were used.

RNA Analysis
For quantitative RT-qPCR, RNA was extracted using the RNeasy Micro Kit (Qiagen). cDNA synthesis was performed using Maxima first-strand cDNA synthesis kit (Thermo Fisher Scientific) with 1 g of RNA.
RT-qPCR analysis was carried out using CFX384 Real-Time PCR Detection System (Bio-Rad), with Premix Ex Taq II (TaKaRa, #RR820A). Levels of mRNA were calculated relative to IPP2 as an internal control as described in (Shleizer-Burko et al., 2011). Primers used for the RT-qPCR analysis are detailed in Supplemental Data Set 5.

RNA-seq Experiments and Analysis
HY5 RNA-seq: Line number 5 used for all constructs. Three biological replicates (two biological replicates collected for HY5-SRDX), of whole seedlings grown in the dark for three days (3d), or 3d dark plus 1.5h in simulated white light (Supplemental Figure 10)  Sequencing (NGS) Core Facility. Raw reads were aligned to TAIR10 genome using TopHat (v2.0.11) with default parameter except library type set to firststrand and intron length was set to 40 minimum, 15 2000 maximum (Kim et al., 2013). Mapped reads were counted by HTSeq (v0.6.0) with default parameters except "-m intersection-strict -s reverse" (Anders et al., 2015) and analyzed by edgeR (v3.16.5) (Robinson et al., 2010). Only genes with CPM (counts-per-million) values greater than 1 in at least two samples were kept for differential expression analysis. A design matrix was set up where each treatment (dark or light) for each genotype (hy5, WT, HY5ox, HY5-VP16, HY5-SRDX) at each time point (1.5h or 4h) is one group, and a generalized linear model was fitted to the read counts with this design matrix. For differential gene expression, contrasts were set up between each of the genotypes and hy5 (at the relevant time point). Genes with p-value of less than 0.05 following Benjamini and Hochberg correction and Log2 fold change <-0.5 or >0.5 were considered differentially expressed. List of genes cpm and fold change can be found in Supplemental Data Set 3.
De-etiolation RNA-seq: Seedling were grown vertically in the dark on 0.5 LS plates with nylon mash (100 micron, 44% open area, 40 inch width-ELKO filtering). After three days in dark the seedlings moved to white light or stayed in the dark and at the indicated time the roots were removed with a scalpel and two biological replicates of the remaining seedlings were collected into ice-cold 80% acetone. In different day the cotyledons and the hypocotyl were separated under a stereomicroscope as described in (Park et al., 2012) and RNA was isolated as described above. 500ng of total RNA was used to prepare RNA-seq libraries, using TruSeq Library prep kit-v2 (Illumina). Multiplexed libraries were sequenced on an Illumina Hi-Seq 2500. Raw reads were aligned to TAIR10 genome using TopHat (v2.0.8b) with default parameter except intron length was set to 40 minimum, 2000 maximum. Cuffdiff (v2.1.1) was used to calculate the expression change between the different time points in the light with the dark sample. Genes with FDR less than 0.05 in at least one time point were included in the heatmap in Supplemental Figure 5C and Supplemental Data Set 4. We replace FPKM values of 0 with 0.01.

ChIP-seq Experiments and Analysis
Three biological replicates of HY5ox #5 and HY5-VP16 #5 and two for HY5-SRDX #5, of seedling grown in the dark for 3d and moved to simulated white light for 2.5h were used for ChIP-seq analysis. ChIP-seq experiment was performed as previously described (Kaufmann et al., 2010). Monoclonal αFLAG M2 antibody (Sigma-Aldrich #F1804) was used for immunoprecipitation. After elution, reversing crosslinks, and DNA purification, Illumina TruSeq libraries were constructed according to manufacturer's protocols.
HY5ox_2 biological replicate was collected and processed for ChIP-seq separately. Multiplexed libraries were sequenced on an Illumina Hi-Seq 2500. Bowtie2 (v2.3.0) used with default parameters, to mapped sequencing reads to the TAIR10 genome (Langmead and Salzberg, 2012). HOMER (v4.9.1) Findpeaks script (Heinz et al., 2010) was used to call peaks using "-style factor -F 3 -P 0.0001 -L 2 -LP 0.001". IP from wild-type (Col-0) extract using anti-FLAG antibody was used as a background control. For HY5ox_2, IP from hy5 (collected and ChIP-seq together), used as a background control, and for HY5ox_3 and HY5-VP16_3 Col-0_3 used as a background control (collected and ChIP-seq together). The bedtools (v2.26.0) intersectBed function (Quinlan and Hall, 2010) was used to associate peaks with TAIR10 annotated genes within 2000 bp upstream of the transcription start site or with gene bodies. HOMER mergePeaks script was used to identify genes with binding in all six samples ( Figure 3C-right). The full list of genes and peaks can be found in Supplemental Data Set 6. For motifs shown in Figure 3A, the peaks of the biological replicates were merged using HOMER mergePeaks script and then findMotifsGenome script used with default parameter except "-size 50". Peaks were visualized using IGV browser (Thorvaldsdottir et al., 2013).

Chlorophyll and Anthocyanin Measurements
For chlorophyll measurements, ~50mg of 13d-old seedlings, grown in LD condition at 21°C, were collected. Fresh weights were measured after quick blotting to dry, and the tissue was immediately frozen in liquid nitrogen. After homogenization, the chlorophyll was extracted twice in 1ml ice-cold 80% acetone and debris was removed by centrifugation at 10,000g for 20 min at 4 °C. Chlorophyll was measured with a spectrophotometer and levels were calculated according to (Woodson et al., 2015). For anthocyanin measurements, 7d-old plants, grown in constant light (100 µmol m −2 s −1 ) at 21°C were transferred to high light (1080 µmol m −2 s −1 ) for 24h. Anthocyanin was extracted and measured as described in (Nakata and Ohme-Takagi, 2014).

Dual-Luciferase Assay in S2 Cells
Schneider's Drosophila S2 cells were maintained at high density in Schneider's Drosophila medium (Gibco) supplemented with 10% Fetal Bovine Serum (GIBCO) and 10 mg per ml of penicillin-streptomycin (Sigma-Aldrich). S2 cells at a density of 1X10 6 were plated in 24-well tissue culture plates. After 24h the cells were transfected using Effectene (Qiagen), according to the manufacturer's instructions. In all transformations 30 ng pAC5.1:RENILLA was used as an internal transfection control, and empty pAC5.1/V5-HIS vector (Invitrogen) was added to a total of 220 ng plasmid DNA in combination with 5 μl Effectene. Dual-Luciferase Reporter Assay System (Promega) was used according to the manufacturer's instructions. Luciferase activity was measured with a TECAN safire 2 plate reader and normalized to the Renilla luciferase activity.
EcoRI and XhoI were used to introduce the HY5 and HY5-VP16 (using PCR on pDONR-P2RP3 plasmids contain HY5 or HY5-VP16) to PAC5.1/V5-HIS. XhoI and HindIII were used to introduce the HSP70 basal promoter, PCR amplified from the CEX3:LUC plasmid (So et al., 2000) (a gift from Dr. Ravi Allada), to create the pGL2-HSP70:LUC plasmid. The 4xG-box fragment was introduced by annealing two oligos and PCR amplifying to introduce the pGL2-HSP70:LUC homology sequence. The CHS, COP1, SPA1, SPA3 promoters were PCR amplified from Col-0 DNA and the SPA4 promoter (1450 bp upstream to the ATG) was synthesized (Genewiz), to introduce the pGL2-HSP70:LUC homology sequence. XhoI was used to open the pGL2-HSP70:LUC plasmid and Gibson assembly was used to insert the different promoters.

Dual-Luciferase Assay in Nicotiana benthamiana
Leaves of N. benthamiana plants were co-infiltrated with A. tumefaciens cultures containing the effector and reporter constructs as described in (Burger et al., 2017). 48h after infiltration a 0.5 cm leaf discs were collected and the Dual-Luciferase Reporter Assay System (Promega) was used according to the manufacturer's instructions using 500 µl lysis buffer. Luciferase activity was measured with a TECAN Safire 2 plate reader and normalized to Renilla luciferase activity. The 4xG-box promoter did not show any expression in tobacco with any of our HY5 plasmids.

Accession Numbers
Sequence data from this article can be found in the GenBank/EMBL data libraries under the following