Time-series transcriptomics reveals a BBX32 -directed control of acclimation to high light in mature Arabidopsis leaves

SUMMARY The photosynthetic capacity of mature leaves increases after several days’ exposure to constant or intermittent episodes of high light (HL) and is manifested primarily as changes in chloroplast physiology. How this chloroplast-level acclimation to HL is initiated and controlled is unknown. From expanded Arabidopsis leaves, we determined HL-dependent changes in transcript abundance of 3844 genes in a 0 – 6 h time-series transcriptomics experiment. It was hypothesized that among such genes were those that contribute to the initiation of HL acclimation. By focusing on differentially expressed transcription (co-)factor genes and applying dynamic statistical modelling to the temporal transcriptomics data, a regulatory network of 47 predominantly photoreceptor-regulated transcription (co-)factor genes was inferred. The most connected gene in this network was B-BOX DOMAIN CONTAINING PROTEIN32 ( BBX32 ). Plants overexpressing BBX32 were strongly impaired in acclimation to HL and displayed perturbed expression of photosynthesis-associated genes under LL and after exposure to HL. These observations led to demonstrating that as well as regulation of chloroplast-level acclimation by BBX32 , CRYPTOCHROME1, LONG HYPOCOTYL5 , CONSTITUTIVELY PHOTOMORPHOGENIC1 and SUPPRESSOR OF PHYA-105 are important. In addition, the BBX32 -centric gene regulatory network provides a view of the transcriptional control of acclimation in mature leaves distinct from other photoreceptor-regulated processes, such as seedling photomorphogenesis.


INTRODUCTION
The exposure of plants to increased light intensities can lead to the development of enhanced photosynthetic capacity [here defined as high-light (HL) acclimation], is an important determinant of plant fitness or crop yield, is under genetic as well as environmental control and includes changes in the expression of many genes (Athanasiou et al., 2010;Eberhard et al., 2008;Murchie and Horton, 1997;Murchie et al., 2005;Oguchi et al., 2003;van Rooijen et al., 2015;Schottler and Toth, 2014;Vialet-Chabrand et al., 2017;Walters et al., 1999).In young expanding leaves, acclimation to HL brings about increased photosynthetic capacity by eliciting changes in both leaf morphology, such as altered leaf and vascular diameter in minor veins and thickness of the lamina (Adams et al., 2014;Oguchi et al., 2003;Terashima et al., 2011;Vialet-Chabrand et al., 2017).This developmental acclimation includes changes to chloroplast physiology such as adjustments to the composition of reaction centres and light harvesting antennae (Drozak and Romanowska, 2006;Murchie and Horton, 1997;Murchie et al., 2005;Vialet-Chabrand et al., 2017;Walters et al., 1999).In contrast, in mature leaves, exposure to sustained or episodic HL brings about changes primarily in chloroplast physiology that raise the light use efficiency of photosynthesis, which can reflect increased rates of photosynthesis and/or a decreased number of photosystem II (PSII) reaction centres and is termed dynamic acclimation (Athanasiou et al., 2010;Murchie et al., 2005;van Rooijen et al., 2015;Vialet-Chabrand et al., 2017;Walters et al., 1999).
For both dynamic and developmental acclimation, it is not known how HL exposure initiates the process at the level of the chloroplast.However, an important lead is provided from an early study (Walters et al., 1999).This was a comparison of Arabidopsis photoreceptor signalling the photosynthetic capacity of mutants and PSII efficiency grown under two different light intensities [photosynthetically active photon flux densities (PPFDs)] of 100 and 400 µmol m À2 sec À1 .From this study, it was proposed that a PHYTOCHROMEA (PHYA), PHYB, and CRYPTOCHROME1 (CRY1) photoreceptor driven CONSTI-TUTIVELY PHOTOMORPHOGENIC/DE-ETIOLATED1/ FUSCA (COP/DET/FUS) regulatory module could transmit signals from the nucleus to chloroplasts to participate in establishing increased photosynthetic capacity (Walters et al., 1999).In support of this, photosynthesisassociated nuclear genes are among the most enriched gene classes subject to control from photoreceptors in de-etiolating seedlings (Chory and Peto, 1990;Ganguly et al., 2015;Holtan et al., 2011;Li et al., 2015;Pham et al., 2018;Shikata et al., 2014).Various combinations of the 11 COP/DET/FUS loci (Lau and Deng, 2012), in conjunction with other regulatory genes, control the integration of signals from photoreceptors and are central to many plant-light environment interactions including seedling photomorphogenesis, the shade avoidance response, stomatal opening and development, the timing of flowering and cross-talk between phytohormone and light signalling (Dong et al., 2014;Huang et al., 2014;Lau and Deng, 2012;Pham et al., 2018).
In this study, we hypothesized that in the first hours of exposure of fully expanded leaves to HL, processes are initiated that eventually lead, several days later, to acclimation manifested as increased photosynthetic capacity.This hypothesis of an early initiation of HL acclimation processes was an extension of an earlier proposal regarding the temporal order of events leading to protection against oxidative stress-induced photoinhibition and the restructuring of light harvesting antennae and PSI/PSII reaction centres (Eberhard et al., 2008).We set out to test this hypothesis by identifying genes that would have a role in both determining immediate responses to HL and the capacity to acclimate.

Gene Ontology analysis of time-series transcriptomics of HL-exposed leaves provides insights into the initiation of acclimation
The starting point for this study was the development of a HL time-series transcriptomics experiment.Our plan was to subject groups of time-resolved differentially expressed genes (DEGs) to Variational Bayesian State Space Modelling (VBSSM; see Experimental procedures), which requires highly resolved time-series data (Beal et al., 2005;Bechtold et al., 2016;Penfold and Buchanan-Wollaston, 2014;Penfold and Wild, 2011).Therefore, we opted for 30-min sampling over a 6 h HL period beginning 1 h after subjective dawn.We chose this time period because it spans the initiation of both the short-term and long-term acclimation to HL proposed by Eberhard et al. (2008).
Full transcriptome profiles using CATMA microarrays (Sclep et al., 2007) were obtained from leaf 7 of HLexposed plants along with parallel LL controls.Leaf 7 from 35 days post-germination (dpg) plants was chosen because under our growth conditions (see Experimental procedures) this leaf had ceased expansion, although the rosette continued to increase in area and biomass (Bechtold et al., 2016).Microarray analysis of variance (MAANOVA; Wu et al., 2003; see Experimental procedures) was used to extract expression values from each probe for every treatment for each technical and biological replicate.To determine DEGs that showed a significant difference between HL-exposed leaves and the LL controls over all or part of the time period, a Gaussian process two-sample test (GP2S; Stegle et al., 2010) was applied and 4069 probes were selected with a Bayes factor score >10, which corresponded to 3844 DEGs (Data S1).The full dataset is deposited with Gene Expression Omnibus (GEO; GSE78251).
To obtain further insight into the overall response to HL at the molecular level, hierarchical co-cluster analysis of the 3844 DEGs was carried out using SPLINECLUSTER (Heard et al., 2005).We reasoned that groups of DEGs that display similar temporal patterns of expression could be coregulated and clustering would be useful in identifying groups of genes for dynamic modelling.Based on comparing temporal gene expression patterns in both the HLexposed and control LL leaves, the 3844 DEGs were divided into 43 temporal clusters (Figure 1a and/or display a downward pattern over the diel (Figure 1a; e.g.clusters 3 and 13 in Figure S1).This pattern changes progressively with increasing degree of expression being higher in HL than LL but against a descending diel pattern in clusters 14-20 (Figure 1a; e.g.cluster 18 in Figure S1), followed by transient but progressively increasingly greater differential transcript levels in HL samples compared with LL in clusters 21-26 (Figure 1a; e.g.cluster 23 in Figure S1) to progressively sustained periods of higher expression in HL compared with LL from clusters 27 to 43 against a background of level or increasing transcript levels across the diel (Figure 1a; e.g.clusters 33 and 43 in Figure S1).
To gain a better view of the timings of differential expression in response to HL, the DEGs from the time-local GP2S (Data S1) were used to identify intervals of differential expression as described by Windram et al. (2012).A histogram of the time of first differential expression (HL compared with LL) is shown in Figure 1(b) and indicated that the response to HL was rapid with >700 genes becoming differentially expressed by 1 h into the HL time course.Nevertheless, it was also clear that changes in transcript abundance were being initiated for significant numbers of genes up to 4 h HL.In summary, the response of the leaf 7 transcriptome to HL entails changed expression in response to the stimulus, with changes occurring across the time of the experiment against a backdrop of complex changes in transcript abundance across 6 h of an 8-h photoperiod.
Gene Ontology (GO) analysis showed that clusters 22, 23, and 25 were highly enriched for generic abiotic stressdefensive genes (P ≤ 0.1, Bonferroni corrected; Data S2).In contrast, some of the other clusters displayed a different set of GO function enrichments (Data S2).These multiple enriched sets were consistent with a readjustment to cellular metabolism.For example, in clusters 39 and 41-43 with generally higher expression in HL compared with LL, there was over-representation of genes associated with amino acid and protein synthesis respectively.Among the clusters showing a lowered expression in HL compared with LL, there was enrichment for genes associated with cell wall metabolism (callose deposition, cell wall thickening cluster 1), phenylpropanoid and glucosinolate metabolism (clusters 1 and 10 respectively), basal resistance to infection (cluster 3), and chromatin re-modelling (cluster 10).
Assessing the effect of a temperature increase accompanying HL exposure The HL exposure raised leaf temperature by 5°C within 5 min of exposure that remained at this level for the remainder of the experiment (Gorecka et al., 2014).To determine the effect of this raised temperature (and the accompanying change in vapour pressure deficit) on the wider leaf transcriptome, a microarray analysis was carried out on plants exposed to HL for 30 min, or 27°C under LL for 30 min (LL/27°C) compared with LL/22°C control plants.There were 609 DEGs [1.5-fold change; false discovery rate (FDR) <0.05] that responded to HL and/or LL/27°C (Data S3; see also GSE87755 and GSE87756).Of these DEGs, 73 responded to the temperature increase alone (Data S3) but were not removed from the time-series data.
In conclusion, the elevated irradiance was the major environmental factor contributing to changes in transcript abundance.The 4-5°C temperature rise in fully expanded leaves accompanying the HL did not cause irreversible photoinhibition (Balfag on et al., 2019;Huang et al., 2019).

Induction of acclimation by repeated exposure to HL
To test our interpretation of the HL time-series data, we determined if HL acclimation could be induced by exposing a plant every day to 4 h HL (see Experimental procedures).This period of HL exposure was chosen as most differential expression had been initiated by this time (Figure 1b).Other than being shortened to 4 h, the environmental conditions were the same as for the time-series transcriptomics experiment (see Experimental procedures).The daily HL regime brought about a stepwise increase in the operating efficiency of PSII (Fq 0 /Fm 0 ; Baker, 2008) of fully expanded leaves (Figure 2a; Data S5).By day 5, the PSII operating efficiency had increased substantially (e.g.78% at 800 µmol m À2 sec À1 actinic PPFD; Figure 2b; see also Figure 4b).This pattern was followed by equivalent changes in Fv 0 /Fm 0 and Fq 0 /Fv 0 (Figure S2a; Data S5).Fv 0 / Fm 0 indicates the maximum operating efficiency of PSII at a given PPFD and a rise in this parameter indicates a decline in NPQ (Baker, 2008).Fq 0 /Fv 0 is the PSII efficiency factor and it is mathematically identical to the coefficient for PQ (q p ) and indicates increased capacity to drive electron transport (Baker, 2008).LL-grown plants of the same age as the plants subjected to five daily HL treatments did not show these changes in chlorophyll fluorescence (CF) parameters (Figure S2b; Data S5).
The first exposure to HL (day 1) did not result in irreversible photoinhibition (Figure S3a) or significant tissue damage (Figure S3b).This was confirmed in the HL timeseries data, which used the same PPFDs, in which steady levels of transcripts for genes considered to be markers for H 2 O 2 (APX2 and FER1; Ball et al., 2004;Gadjev et al., 2006) rose but those associated with 1 O 2 -induced signalling (AAA-ATPase and BAP1; Ramel et al., 2013) remained unchanged or declined (Figure S3c).The changes in expression of these marker genes indicated the HL treatment used in the time-series transcriptomics experiments, also did not elicit photodamage and provided conditions that could promote HL acclimation.
The increased operating efficiency of PSII (Fq 0 /Fm 0 and Fq 0 /Fv 0 ) after the 5-day HL treatment (Figure 2a; Figure S2a; Data S5) could have reflected enhanced photosynthetic capacity.To test this possibility, gas-exchange measurements for photosynthesis were carried out (see (a) Plants were exposed daily to 4 h HL and Fq 0 /Fm 0 determined for mature leaves.After the HL, plants were dark adapted and imaged under increasing actinic photosynthetically active photon flux densities (PPFDs) from 200 to 1400 µmol m À2 sec À1 in 200 µmol m À2 sec À1 increments every 5 min.Data were collected as chlorophyll fluorescence images and processed digitally to collect values from mature leaves.Plants were treated in this way daily for 5 days: day 1 (blue), day 2 (red), day 3 (olive green), day 4 (purple), and day 5 (light blue).Data (mean AE SE) correspond to 38 plants at 24-28 days post-germination (dpg) over six experiments, and the asterisks show differences in chlorophyll fluorescence parameters between days 1 and 5 were significant (P ≤ 0.001; ANOVA and Tukey HSD).Full statistical data comparing all days of HL exposure are provided in Data S5.(b) Daily changes in Fq 0 /Fm 0 plotted from the data in A and Data S5 (right panel).Fq 0 /Fm 0 values are from the same plants over the daily HL exposures showing the increase in photosystem II (PSII) operating efficiency at 800 µmol m À2 sec À1 PPFD actinic light over the 5 days of the experiments.(c) Photosynthesis plotted as CO 2 assimilation rate (A) as a function of actinic PPFD in mature leaf 7 (mean AE SE; n = 8 plants for each treatment; 49 dpg).Measurements were taken the day after 1 (dashed lines) and 5 days (solid lines) of daily 4 h HL exposures (blue lines) along with the low light (LL) control plants (red lines) not subjected to this treatment.(d) Photosynthesis plotted as CO 2 assimilation rate (A) as function of leaf internal CO 2 concentration (C i ) in mature leaf 7 (mean AE SE; n = 8 plants for each treatment; 49 dpg).Measurements were taken the day after 5 days of daily 4 h HL exposures (blue line) along with the LL control (red line).A was determined by infra-red gas analysis (see Experimental procedures).Asterisks indicate significant differences (P < 0.02; covariant T and two-tailed F tests) between LL-and HLexposed plants.
© 2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384Experimental procedures).The same experiment was repeated and, in the photoperiod following the last HL treatment, measurements were taken of CO 2 assimilation rates (A) over a range of light intensities in fully expanded leaf 7 of these plants (Bechtold et al., 2016).This showed that the light-saturated photosynthetic rate (A sat ) was significantly greater (P < 0.001) by 64% compared with the LL control plants (Figure 2c).In contrast, after a single 4 h HL exposure, followed by photosynthesis measurements in the next photoperiod, no increase in A sat was observed (Figure 2c).In a separate series of experiments, the measurement of A over a range of internal leaf CO 2 concentrations (Ci) also showed that the maximal CO 2 -saturated rate of photosynthesis (A max ) increased by 31% (P < 0.002) after five daily HL exposures (Figure 2d).This confirmed that repeated HL exposures did not solely affect stomatal behaviour but brought about an increase in foliar photosynthetic capacity.The changes in CF parameters by day 5 of HL treatments observed in the previous experiments (Figure 2a) occurred also in larger older leaves that were required for gas exchange measurements (Figure S2c; see Experimental procedures).
In summary, increased A sat and A max after 5 days of repeated HL exposure (Figure 2c,d) was accompanied by a highly significant increase in Fq 0 /Fm 0 (Figure 2a; Figure S2c; P < 0.0001, ANOVA and Tukey HSD; Data S5).Therefore, a substantial (>40%; typically using the median 800 µmol m À2 sec À1 actinic PPFD value) change in Fq 0 /Fm 0 between days 1 and 5 of repeated HL was subsequently used as a more convenient image-based measurement of the establishment of increased photosynthetic capacity, which was taken as indicative of HL acclimation.

Dynamic statistical modelling infers a BBX32-centric HL gene regulatory network
The HL time-series data were used to infer gene regulatory networks (GRNs) using VBSSM (Beal et al., 2005;Penfold and Wild, 2011).We chose VBSSM because it has been demonstrated to infer known GRNs from temporal gene expression data and to infer novel GRNs whose highly connected genes (nodes) have subsequently been shown experimentally to have a novel and important function (Beal et al., 2005;Bechtold et al., 2016;Breeze et al., 2011;Penfold and Buchanan-Wollaston, 2014;Penfold and Wild, 2011;Windram and Denby, 2015).However, due to the limited number of time points, we opted to infer networks for about 100 genes or probes to avoid overfitting by constraining network size (Allahverdiyeva et al., 2015;Beal et al., 2005;Bechtold et al., 2016;Windram and Denby, 2015).To accommodate this limitation, we focused on DEG coding for transcription regulatory genes such as transcription (co-)factors (TFs).We reasoned that regulatory networks composed of such genes would control the expression of a wide network of genes and by inferring GRNs this would allow us to identify and focus on the most connected of them, often termed hub genes (Albihlal et al., 2018;Windram and Denby, 2015).Consequently, we reasoned that such regulatory genes would control the expression, directly and indirectly, of a sufficiently large number of genes to influence whole leaf HL responses and acclimation phenotypes.Therefore, the intention was to screen highly connected candidate regulatory hub genes directly for their impact upon HL acclimation measured as changes in photosynthetic efficiency.
It was estimated that there were 371 HL DEGs coding for TFs or (co-)TFs (Data S6).To narrow our selection further, comparisons were made between the 43 HL temporal clusters (Figure 1a; Data S1) and 14 publicly available transcriptomics datasets or meta-analyses of such data for HL treatments or mutants perturbed in chloroplast-to-nucleus and reactive oxygen species-mediated signalling (Data S4).On a cluster-by-cluster basis, the highest number of significant (P < 0.00001) overlaps in clusters 1, 2, 3, 5, 6, 9, 10, 14, 16, 17, and 27 were encountered with phyA/phyB DEGs (Data S4; Shikata et al., 2014).This observation suggested that photoreceptor-mediated regulation of HL-responsive genes was highly represented in the time-series transcriptomics dataset.Therefore, we examined whether photoreceptor-regulated (co-)TF genes (Dong et al., 2014;Shikata et al., 2014) were also over-represented in the HL dataset.This was the case with 91 photoreceptor-and light-regulated (co-)TF DEGs identified irrespective of which temporal cluster they were drawn from (P = 1.4E-06;Hypergeometric Distribution Test, Data S6).The HL timeseries expression data from these 91 genes were used to infer networks with VBSSM.
The first inferred network for HL revealed LATE ELON-GATED HYPOCOTYL (LHY) as the most highly connected gene (Figure S4a).However, mutant lhy-21 plants were not perturbed in HL acclimation (Figure S4b).Therefore, the VBSSM modelling was reiterated but omitted the LHY expression data.This inferred a 47 node-HL network centred on BBX32 (Figure 3).The transcript levels of the 12 most connected nodes (≥3 edges) across the time series, under LL and HL conditions, is shown in Figure S5 and shows the diversity of expression patterns derived from the temporal clusters (Data S1; Figure 1a,b).

BBX32 is a negative regulator of photosynthetic capacity and HL acclimation
Acclimation was tested in two independent BBX32 overexpressing (BBX32-OE) genotypes (BBX32-10 and BBX32-12) and a T-DNA insertion mutant (bbx32-1; see Experimental procedures and Holtan et al., 2011).BBX32-OE plants showed some inhibition of Fq 0 /Fm 0 after day 1 of HL (Data S7) but a highly significant impairment of the increase in PSII operating efficiency at the end of the 5-day serial HL exposure indicative of an inhibition of HL acclimation (Figure 4a; Data S7).In contrast, bbx32-1 plants showed a weak but significant accelerated acclimation phenotype (Figure 4b; Data S7).We define an accelerated acclimation phenotype as a significant enhancement of PSII operating efficiency over one or more days in the 5-day serial HL treatment.The strong negative impact of BBX32 overexpression on acclimation was confirmed subsequently by showing a significant inhibition of photosynthetic capacity (A sat ) after the 5 days of daily 4 h HL (Figure 4c).

Transcriptomics provides a partial verification of the BBX32 HL TF network
To explore further the connections depicted in the network model (Figure 3), massively parallel RNA sequencing (RNA-seq) was carried out (see Experimental procedures; GEO; GSE158898) to profile the foliar transcriptome of fully expanded leaves of Col-0 and BBX32-OE plants exposed to 3.5-h HL in comparison with LL controls.From these data, the transcript levels of 25 of 47 constituent genes in the inferred network were significantly altered by constitutive BBX32 overexpression compared with Col-0 plants in LL and/or HL (Figure 5; Data S8), partly validating the GRN.

Transcriptome of BBX32OE plants links initial responses to HL with dynamic acclimation
The impaired ability of BBX32-OE plants to photosynthesize (Figure 4a,c; Data S7) prompted an analysis of the RNA-seq data on the impact of BBX32 overexpression on the transcript levels of photosynthesis-associated genes (PhAGs).There was a clear influence of BBX32 overexpression under LL and HL on the transcript levels of a range of transcripts coding for LH Antenna proteins, Calvin-Benson cycle enzymes, and components of photosynthetic electron transport, PSI and PSII (Figure 6a; Data S8).We concluded that these and other transcripts affected in BBX32-OE plants might reflect their perturbed photosynthetic physiology.The establishment of dynamic acclimation (see Introduction) requires the expression of GLUCOSE-6-PHOSPHATE/PHOSPHATE TRANSLOCATOR2 (GPT2; see Discussion; Athanasiou et al., 2010).HL differential GPT2 transcript levels were evident (Figure 6b,c) placing it in temporal cluster 30 (Data S1).In addition, this change in GPT2 transcript levels was strongly inhibited in BBX32-OE plants exposed to 3.5-h HL (Figure 6c).This disruption of PhAG transcript levels led us to examine the impact of BBX32 overexpression on other cellular processes.Of the 2903 genes whose transcript levels were HL responsive (P < 0.05; ≥ 2-fold differentially expressed; Data S9), BBX32 overexpression perturbed the transcript levels of 32% and 15% of them in LL and HL conditions respectively (Figure 7a; Data S9).The HL/LL Col-0 DEGs were enriched for 35 GO BP terms and 26 of them were significantly over-represented in the BBX32-OE/Col-0 LL and BBX32-OE/Col-0 HL DEGs (Data S10).These shared GO groups all describe responses to various abiotic and biotic stresses or response to endogenous stimuli such as  and S11).Although the number of overlapping genes was lower (Figure 7b), 256 BBX32-OE HL DEGs again confirmed enrichment for a range of GO terms that describe generic responses to environmental stress (Figure 7c; Data S11).However, the 408 BBX32-OE LL DEGs also differentially expressed in the HL time-series dataset, also revealed significant enrichment (FDR <0.05) of a range of additional functions (Figure 7c; Data S11) including glucosinolate and glycosinolate metabolism (GO:0019760, GO:0050896, GO:006143, GO:0019757, GO:0016144, GO:0019761, GO:0019758), cell wall thickening (GO:0052543, GO:0052386), and callose deposition (GO:0052543, GO:0052545).Downregulation of these groups of genes in the HL time-series data (Data S1 and S2) may reflect a redistribution of resources towards HL acclimation and away from basal immunity (see above and Discussion).The observations also reinforce that BBX32 influences immediate responses before or during a single exposure to HL.
CRY1 and HY5 control of photosynthetic efficiency and acclimation BBX32 has been proposed to be a negative regulator of the integration of light signals from phytochromes (PHYs) and cryptochromes (CRYs) during photomorphogenesis (Gangappa and Botto, 2014;Holtan et al., 2011).BBX32-OE seedlings display long hypocotyls in the light phenocopying photoreceptor mutants and mutations in LONG HYPO-COTYL5 (HY5; Holtan et al., 2011).Notably, HY5 is a member of the BBX32-centric GRN (Figures 3 and 5) and along with CRY1, has also been implicated in influencing the expression of HL-inducible gene expression (Chen et al., 2013;Kleine et al., 2007;Shaikhali et al., 2012).Furthermore, PHYA-, PHYB-, and CRY1-mediated signalling was proposed to regulate photosynthetic capacity in plants grown in a range of PPFDs (Walters et al., 1999; see Introduction).These considerations prompted us to test HL acclimation in photoreceptor-defective and hy5 mutants.
No significant impact of PHYA or PHYB on acclimation was observed (Figure S6a,b).In contrast, cry1 mutants almost completely failed to undergo any acclimation (Figure 8a,b), whereas cry2-1 was not impaired (Figure S6c).One of the cry1 mutants shown (cry1-M32; Figure 8b) arose serendipitously from a screening of T-DNA insertion mutants in genes coding for 7-transmembrane proteins that had been postulated to be implicated in HL-mediated G protein signalling (Galvez-Valdivieso et al., 2009;Gorecka et al., 2014).However, the one mutant recovered from this screening, was shown subsequently to be deficient in HL acclimation due to a disabling second site mutation in CRY1 (see Experimental procedures).As the defective acclimation phenotype was identified before knowing the identity of the causal mutation, we took this to be forward genetic evidence of the importance of CRY1 in setting PSII operating efficiency in mature leaves.
The light environment used to grow plants for this study and subject to HL was enriched for blue wavelengths (Figure S7; see Discussion).Therefore, we considered the possibility that a role for PHYs in dynamic acclimation could be obscured, favouring a predominance of CRY1 under our growth conditions.To test this notion, a mutant harbouring a constitutively active form of PHYB, phyBY276H (YHB) in a Col-0 background (Jones et al., 2015) was tested for HL acclimation (Figure 8c).This mutant exhibited a higher PSII operating efficiency than Col-0 after 1 day of HL exposure consistent with an accelerated acclimation phenotype.
Mutants defective in HY5 function were strongly impaired in HL acclimation (Figure 8d,e) consistent with being a member both of a BBX32-centric GRN (Figures 3  and 5, Data S8).
Cop1-4 plants, despite a severely dwarfed shoot morphology (Figure 9a; Deng and Quail, 1992;Gangappa and Kumar, 2018), displayed an accelerated acclimation phenotype (Figure 9b) such as the HL response of YHB plants (Figure 8c).In contrast, despite a similar dwarf shoot morphology (Figure 9a), det1-1 displayed no defect in HL acclimation (Figure 9d).This suggests that the HL acclimation response of chloroplasts is independent of shoot size and that these two traits are not coupled.Furthermore, spa1/ spa2/spa3 (spa1,2,3) plants also displayed accelerated HL acclimation (Figure 9c).Therefore, it was concluded that one or more type of the COP1/SPA complex (Hoecker, 2017;Huang et al., 2014) are negative regulators of HL acclimation and that DET1 plays no role in this process.
There is a high degree of redundancy among the PIF family and therefore a quadruple null mutant of PIF1, 3, 4, and 5 (hereafter called pifq; Leivar et al., 2008) was tested and shown to display significant inhibition of HL acclimation (Figure 9e).In contrast, the HL acclimation of a single mutant allele of PIF4 (pif4-2) was normal (Figure S6d).

Time-series HL transcriptomics data indicate the initiation of acclimation processes
The exposure to a 7.5-fold increase in PPFD (HL) presents both a threat and an opportunity to the plants in this study.The threat comes from the possibility that the PPFD will continue to increase and render the plant susceptible to irreversible photoinhibition.The opportunity comes from enhancing photosynthetic capacity and consequently acclimating to the HL (Figure 2a-d), accompanied by a lowered reliance on the dissipation of excitation energy using NPQ (Figure S2a), which can limit plant productivity (Kromdiijk et al., 2016).
The adaptation to a potential increase in photo-oxidative stress and photoinhibition (see Introduction) is the early (≤ 1 h), strong, but transient change in transcript abundance of 257 genes in clusters 21-26, upon exposure to HL. Clusters 22, 23, 25, and 26 include among them 64 known genes that promote abiotic stress tolerance (Figure 1a,b; cluster 23 in Figure S1; Figure S2c; Data S1 and S2).The transiently enhanced expression of these genes presumably allows the plant to overcome any potential initial detrimental effects of the HL exposure, as many other studies have reported (e.g.Balfag on et al., 2019;Ball et al., 2004;Crisp et al., 2017;Gadjev et al., 2006;Huang et al., 2019;Ramel et al., 2012Ramel et al., , 2013;;Willems et al., 2016).
Coordinated alteration in specific biological processes was evident in some clusters.Downregulated clusters include those collectively associated with aspects of basal or innate resistance to pathogens (Piasecka et al., 2015;Underwood, 2012).Examples include genes coding for cell wall modifications and callose deposition (cluster 1), defence response to bacteria (cluster 3), and glucosinolate/ glycosinolate biosynthesis (cluster 10).In this study, plants were grown at a PPFD below their light saturated rate of photosynthesis (A sat ; Figure 2c; see Experimental procedures).Plants grown under such light-limiting conditions may initially reallocate resources away from some cellular processes to begin acclimation and take advantage of a sustained increase in PPFD.Photosynthetically active expanded but not senescing leaves, such as leaf 7 used here (Bechtold et al., 2016), may maintain a higher degree of poising of immunity to respond to biotic stress compared with abiotic stress (Berens et al., 2019).Therefore, in a converse situation where a potential abiotic stress threat emerges, the diversion of resources from defence against pathogens may be an appropriate response.Meanwhile, among the DEG time-series clusters whose transcript levels increased at various points in the experiment, are those that could be preparing the leaf to increase its photosynthetic and metabolic capacity to begin acclimation (Dietz, 2015 ;Eberhard et al., 2008).Genes in overrepresented GO BP terms included those involved in macromolecule synthesis and particularly translation (clusters 41-43) and related metabolic processes such as enhanced amino acid and organic acid biosynthesis (cluster 39).The HL induction of transcript levels of one gene, GPT2, in temporal cluster 30 (Data S1), is noteworthy (Figure 5b).This gene is required for dynamic acclimation (Athanasiou et al., 2010) and this change in expression may indicate initiation of HL acclimation processes.

BBX32 connects a range of cellular processes during the response to HL
Of all the comparisons carried out with relevant transcriptomics datasets, the most extensive overlap with timeseries HL DEGs was with those from dark-germinated phyA/phyB seedlings exposed to red light (Data S4; Shikata et al., 2014).While this was initially surprising because of the very different experimental conditions, earlier studies had shown a strong influence of photoreceptor genes (CRYs and PHYs) on photosynthetic capacity in Arabidopsis grown at a range of PPFDs (Walters et al., 1999) and photoreceptor-directed signalling on the induction of HLresponsive genes (Guo et al., 2016;Huang et al., 2019;Kleine et al., 2007;Shaikhali et al., 2012).
The above analysis prompted a selection of 91 light-and PHYA/B-regulated (co-)TF genes (Data S6).The HL time-series expression data from these genes were subjected to VBSSM, which after two iterations, inferred a highly interconnected BBX32-centric (co-)TF GRN (Figure 3; Figure S4; see Results).In the GRN, >50% of the nodes (genes) were subsequently confirmed by RNA-seq to be influenced significantly in their expression by BBX32 (Figures 3 and 5; Data S8).
BBX32 showed a greater transcript abundance over LL controls at any point onwards from 2 h HL.Nevertheless, its transcript abundance was on a downward trend through the diel, paralleling its LL pattern of expression (Figure S5).Interestingly, while BBX32-OE plants displayed a 66-fold elevated BBX32 transcript level in LL, this reduced to 33-fold after 3.5 h HL (Data S9).The enhanced BBX32 expression in these plants is driven by the CaMV 35S promoter (Holtan et al., 2011); therefore, the decline in transcript abundance over a diel could indicate that a temporal post-transcriptional control operates on BBX32 expression.
The overexpression of BBX32 strongly influenced the immediate responses of plants to HL across a range of cellular processes (Figure 6c; Data S8 and S9) and in their photosynthetic physiology (Figure 4a-d).Most prominently, under LL, BBX32-OE plants displayed perturbed expression of genes with basal immunity, including multiple GO designations for glucosinolate/glycosinolate metabolism, callose deposition, and responses to chitin and to pathogens (Figure 7c; Data S10 and S11).This observation is consistent with enrichment of the same processes in downregulated HL temporal clusters in Col-0 (see above; Data S2) and supports our suggestion that in wild-type plants, downregulation of basal immunity may be a necessary prerequisite for successful adjustment to elevated light intensities and that BBX32 is a negative regulator of this process.
BBX32 overexpression under LL and HL conditions perturbed the expression of PhAGs (Figure 6a the combined effects of such disruption would be consistent with a modestly significant (P < 0.1) depressed PSII quantum efficiency that both BBX32-OE lines displayed after a single 4 h of HL (Data S7).

BBX32 and the control of photosynthetic capacity and HL acclimation
There was an inhibitory effect of BBX32 overexpression in HL on GPT2 transcript levels (Figure 6c) and on those of LHCB4.3 in BBX32-OE HL and LL plants (Figure 6a; Data S8).GPT2 is required for dynamic acclimation (Athanasiou et al., 2010) and levels of LHCB4.3 correlate with the degree of long-term acclimation to HL (Albanese et al., 2016).This indicates that processes that would lead to acclimation had been initiated during this first exposure to HL and that BBX32 is involved in their regulation.However, a single exposure to 4 h HL is not sufficient to induce HL acclimation and increased photosynthetic capacity.This requires, under our conditions, a further three daily episodes of 4 h HL for this to begin to occur (Figure 2a-d).Our experience is consistent with a previous study where dynamic acclimation took about 5 days to be fully manifested and 2-3 days to discern any change in photosynthesis rates after a permanent shift from a PPFD of 100-400 µmol m À2 sec À1 (Athanasiou et al., 2010).In contrast to the weak but significant effects on photosynthesis of BBX32 overexpression during a single 4 h HL exposure (Data S7), there was a strongly significant negative impact upon HL acclimation after 5 days of daily 4 h HL (Figure 4a,c).This suggests that BBX32 exerted a negative control on HL acclimation that was stronger than its impact on photosynthesis at growth PPFD.Similarly, a gpt2 mutant showed wild-type levels of maximal photosynthetic capacity when grown at two different PPFDs (100 and 400 µmol m À2 sec À1 ) but lowered dynamic acclimation going from the lower to the higher PPFD (Athanasiou et al., 2010).In summary, we propose that BBX32 exerts control over a large number of genes in the first hours of HL exposure and the acclimationassociated increase in photosynthetic capacity that occurs several days later.Therefore, BBX32 provides a link between these temporally distinct events that establish acclimation to HL (Eberhard et al., 2008;see Introduction).
Negative regulation of HL acclimation by BBX32 (Figure 4a) suggested that a defective allele ought to confer a converse elevated phenotype.The mutant bbx32-1 (see Results; Holtan et al., 2011), displayed a weakly significant trend of enhanced PSII operating efficiency compared with Col-0 between days 2 and 4 of the 5 days of 4-h HL exposure (Figure 4b; Data S7).However, this genotype is unlikely to be a null mutant.The mutagenic T-DNA is inserted such that the first 172 amino acid residues of BBX32 would still be produced and a truncated transcript spanning this region has been detected in bbx32-1 seedlings (Holtan et al., 2011).The retained N-terminal region coded by this allele harbours the single B-Box zinc finger domain of BBX32 (Gangappa and Botto, 2014) and downstream sequences to residue 88, capable of binding at least the transcription regulator EMBRYONIC FLOWER1 (EMF1; Park et al., 2011).The possibility of a partially functional truncated BBX32 may explain the weak phenotype of bbx32-1 with respect to this acclimation phenotype (Figure 4b; Data S7) and its mild constitutive photomorphogenic phenotype in seedlings (Holtan et al., 2011).

Establishment of HL acclimation involves BBX32-centric GRN members
The VBSSM that led us to BBX32 also led us to HY5 (Figure 3) and was subsequently reinforced by its known interaction with BBX32 in seedling photomorphogenesis (Gangappa and Botto, 2016;Holtan et al., 2011).HY5 was shown to be a strong positive regulator of acclimation in mature leaves (Figure 8d,e).Therefore, these observations reveal new functions for BBX32 and HY5, extending their role to a further dimension in the interaction of the plant with its light environment.In seedlings, HY5 controls chlorophyll content and transcript levels of PhAGs in cool temperatures (Toledo-Ortiz et al., 2014) and the control of chloroplast development during photomorphogenesis (Ruckle et al., 2007), which suggests, along with data shown here (Figure 6; Data S8), that control of these photosynthesis-associated processes by a BBX32/HY5regulatory module is retained throughout the life of the plant.
SPA1, PIF4, and PIF7 also were incorporated into the BBX32-centric GRN by VBSSM (Figure 3).This reinforced the comparison between the control of seedling photomorphogenesis and HL acclimation in fully expanded leaves, which was extended beyond the GRN by establishing that CRY1 (and possibly PHYB) along with and one or more members of the PIF family are positive regulators of HL acclimation (Figures 8a,c and 9e), while COP1 and one or more SPA genes are negative regulators (Figure 9b,c).
We suggest that COP1 and SPA genes act together to suppress HL acclimation under LL by enabling the ubiquitin-mediated degradation of HY5 and therefore coupling photosynthetic capacity to the prevailing PPFD.In HL, this suppression would be reversed by CRY1 physically interacting with and inhibiting the action of COP1/SPA (Gangappa and Botto, 2016;Hoecker, 2017;Huang et al., 2014;Lau and Deng, 2012;Lau et al., 2019;Laubinger et al., 2004;Lian et al., 2011;Pham et al., 2018).Consequently, CRY1 would cause HY5 to be redirected to HL acclimation.However, a further adaptation may be required to retard or accelerate acclimation.For example, to fine tune the establishment of HL acclimation in a fluctuating light environment in order to balance source-sink relationships.We suggest under HL, when HY5 is free of negative regulation by COP1/SPA, that BBX32 is the important additional   2).Note that because of the size of the cop1-4, pifq, and det1-1 plants, data were collected from whole rosettes rather than from mature leaves.CF parameter values were collected at a range of actinic PPFDs at the end of days 1 and 5 of HL.Fq 0 /Fm 0 values at day 1 (black lines) and day 5 (red lines) for mutant plants (dashed line) and Col-0 (solid line) of the HL treatments for (b) cop1-4, (c) spa1,2,3, (d) det1-1, and (e) pifq.Asterisks (e) indicate significant difference between mutant compared with Col-0 at day 5 (P < 0.01; ANOVA and Tukey HSD).Upward arrows (b,c) indicate significant difference between mutants and Col-0 at day 1 (P < 0.01; ANOVA and Tukey HSD).
The opposing regulation of HL acclimation by BBX32 and HY5 could mean that some form of interaction between these genes drives its establishment in a manner similar to their respective negative and positive regulation of photomorphogenesis (Datta et al., 2007;Gangappa and Botto, 2016;Holtan et al., 2011;Xu et al., 2014).BBX32 does not bind DNA and has been proposed to act as (co-) TF in complexes with several TFs, such as the BBX32-BBX21-HY5 tripartite complex involved in the control of photomorphogenesis (Datta et al., 2007;Gangappa and Botto, 2016;Holtan et al., 2011;Park et al., 2011;Tripathi et al., 2017;Xu et al., 2014).Therefore, there may also be a post-translational control of HY5 by BBX32 during HL acclimation.
The proposed need for both a CRY1/COP1/SPA-and a BBX32-mediated control of photosynthetic capacity and acclimation comes also from considerations about light quality and intensity.First, the fluence of blue light in the HL exposure used in this study would exceed the saturation of CRY1 activation, which is approximately 32-40 µmol m À2 sec À1 blue light (Hoang et al., 2008;Liu et al., 2020).Therefore, while CRY1 signalling would need to be activated (i.e., on) for acclimation to happen, further signalling input may be required from other sources via BBX32 and its GRN to modulate the degree of response.A second factor is that at high fluence, CRY1 may produce H 2 O 2 in the nucleus (Consentino et al., 2015).H 2 O 2 for HL signalling is primarily synthesized and exported from chloroplasts and is dependent upon active photosynthetic electron transport (Exposito-Rodriguez et al., 2017;Mullineaux et al., 2018).However, this does not exclude the possibility that HL-dependent accumulation of H 2 O 2 in nuclei for signalling may be augmented from other sources such as photo-saturated CRY1.
In contrast to Arabidopsis grown at differing PPFDs but using similar fluorescent lighting to this study (Walters et al., 1999; see Experimental procedures), no influence of PHYA or PHYB was observed on HL acclimation (Figure S6a,b).This could have been a consequence of the degree of blue light used in both growth conditions and in HL exposure (9% and 58% of the total PPFD respectively; Figure S7; see Experimental procedures).This range of wavelengths in artificial lighting is typical of many controlled environment conditions (Naznin et al., 2019) and may have favoured a response mediated by CRY1.The observation that plants harbouring a constitutively active PHYB allele (YHB) displayed a partially accelerated acclimation phenotype (Figure 8c) means that PHYs could also control HL acclimation under some light environments and modify or interact with a CRY1-dependent signalling pathway (Ahmad et al., 2002;Yu et al., 2010).

Conundrum of the control of photosynthetic capacity and the type of HL acclimation
In interpreting the data from the mutants and BBX32-OE genotypes used in this study, the question can be asked: Is the mutants' altered HL acclimation phenotype a consequence of limited development or functioning of the photosynthetic apparatus such that maximal photosynthetic capacity could never be attained?This question can be answered in two parts: first and as stated above, the effect of BBX32 overexpression is more marked in the increase in photosynthetic efficiency between days 1 and 5 of daily HL exposures (i.e.HL acclimation) than in the starting photosynthetic efficiencies at day 1 (Figure 4a,c; Data S7).The same pattern can be observed in the hy5 mutants (Figure 8d,e; Data S7).However, the cry1 mutants and pifq showed no difference from Col-0 on day 1 HL but a strong difference by day 5 HL (Figures 8a,b and 9e).Second, mutants such as bbx32-1, YHB, cop1-4, and spa1,2,3 displayed an accelerated chloroplast-level HL acclimation phenotype, which showed that their photosynthetic apparatus was set at a level higher than it should have been for their growth PPFD and for the number of days of 4-h HL exposure (Figures 4b,.This means that HL acclimation can become uncoupled from the prevailing light intensity.However, in such mutants this phenotype cannot be due to partially disabled photosynthesis but is a feature of acclimated wild-type plants having been exposed to HL for a longer period.Therefore, in summary, we conclude that BBX32, members of its GRN and CRY1 exert both negative and positive control over the setting of photosynthetic capacity and the extent of chloroplast-level acclimation to HL. HL acclimation in fully expanded leaves strongly suggests BBX32 regulates dynamic acclimation (see Introduction) possibly, but not exclusively, through the control of GPT2 expression (Figure 6c; Athanasiou et al., 2010).However, we cannot rule out that BBX32, its GRN and CRY1directed signalling influence developmental acclimation (see Introduction) as many of the mutants used in this study have altered growth and development phenotypes  (Gangappa and Kumar, 2018;Holtan et al., 2011;Jones et al., 2015;Laubinger et al., 2004;Leivar et al., 2008;Ruckle et al., 2007;Figure 9a).However, such an effect would probably not influence chloroplast-level acclimation as this property was unaffected in det1-1 plants despite their severe dwarf shoot morphology (Figure 9a,d).
In summary, this study uncovering a BBX32-centric GRN provides the outline for a highly sensitive and flexible system of adjusting photosynthetic capacity and points to how chloroplast-level acclimation is influenced not only by light intensity and quality but also many other environmental and internal cues.

Growth conditions
Plants were grown in an 8-h photoperiod (short day) at a PPFD of 150 (AE10) µmol m À2 sec À1 under fluorescent tubes (Philips TLD 58W, 830 (warm whites); the spectrum of the light source is shown in Figure S7), 22 AE 1°C, 1-kPa vapour pressure deficit and cultivation conditions as described previously (Bechtold et al., 2016;Windram et al., 2012).Unless stated otherwise, all plants were used from 35 to 40 dpg.

Identification of the cry1M32 mutant
Based upon earlier research in our laboratory (Galvez-Valdivieso et al., 2009;Gorecka et al., 2014) in which we studied a possible role for heterotrimeric G protein-mediated HL signalling, we set out to identify candidate genes coding for seven transmembrane proteins that may have a role as G protein-coupled receptors.A collection of 59 T-DNA insertion mutants in genes coding for putative 7-transmembrane proteins (Moriyama et al., 2006; a kind gift from Professor Alan Jones, University of North Carolina) was screened for perturbed CF in response to HL exposure (see below).The screening revealed that the insertion line Sail_1238_E12 (hereafter termed M32) was deficient in HL acclimation (Figure 8b).The information available on T-DNA flanking sequences indicated that this was a T-DNA insertion in the first exon of At4g21570, a gene encoding a transmembrane protein of unknown function.However, complementation of M32 by transformation with the wild-type At4g21570 gene did not restore a wild-type phenotype.
Besides being defective in dynamic acclimation, M32 was impaired in blue light inhibition of hypocotyl elongation under both low and high blue light fluence, accumulated less chlorophylls and anthocyanins than Col-0 under blue light, and presented delayed flowering time when grown in short day photoperiod.(Figure S8b-e).Later and in light of our subsequent hypothesis that CRY1-mediated signalling controls HL acclimation in Arabidopsis (see Results and Discussion), we realized that M32 resembled the phenotype of known cry1 mutants.Therefore, we tested if CRY1 was altered in this mutant.CRY1 was amplified from its genomic DNA and the PCR product was Sanger sequenced on both strands.Col-0 CRY1 amplicon was also sequenced.The analysis of the sequence showed that in M32, CRY1 contains a single point mutation (G?A), which caused a substitution of Gly 347 Arg mutation in CRY1 (Figure S8f).This mutation was previously identified in a screening of EMSmutagenized Arabidopsis seedlings (Ahmad et al., 1995) and designated as hy4-15, and affects the domain comprising the photolyase signature sequence.Consequently, hy4-15 plants produce a wild-type amount of full-length CRY1, but the protein is not functional.Therefore, we concluded that the M32 mutant is in fact a cry1 mutant that we named cry1M32.

HL exposures
The HL exposure was a PPFD of 1100 (AE100) µmol m À2 sec À1 from a white light emitting diode (LED) array (Isolight 4000; Technologica Ltd, Colchester, UK) as described previously (Gorecka et al., 2014) and permitted the simultaneous exposure of nine plants.The spectrum of the LED array is shown in Figure S7.
For the HL time-series transcriptomics, two consecutive sowings, 24 h apart, were grown to 35 dpg on the same growth room shelf and randomized across the shelf every day.Leaf 7 (Bechtold et al., 2016) was tagged at 30 dpg.We used this staging of plant growth and 3 LED Isolight arrays to treat 27 plants each day.The HL exposure began 1 h after subjective dawn and was completed 1 h before subjective dusk.Each set of tagged leaves (four) at each HL time point and their LL controls (four) were sampled within 5 min at time 0.5 h and each 0.5 h interval for the 6 h exposure.Two HL experiments were conducted with duplicate samplings of a full range of time points on each day.In addition, four time zero samples were processed for the 0 h time point.Both HL experiments provided, in total, 100 samples for RNA extraction.These were four biological replicates (i.e.four sampled leaves) per time point per HL treatment (48 samples) and LL control (48 samples) plus four 0 time point samples.
To elicit HL acclimation, plants were subjected to 4 h HL, followed by a 0.5 h dark adaptation and then exposed to a range of actinic PPFDs (over 50 min) to collect CF data (see below).This HL treatment was repeated daily and CF data collected from the same plants for 5 consecutive days or on days 1 and 5 only as stated.

CF measurements and imaging
During the time-series HL experiments, CF measurements were taken from leaf 7 of one plant in situ under each isolight using PAM-2000 portable modulated fluorimeters (PAM-2000; Walz GmbH, Effeltrich, Germany).At the end of each experiment the dark-adapted CF parameter Fv/Fm was determined for the same plants and LL controls and then again 24 h after being returned to growth conditions.
For the HL acclimation experiments, photosynthetic efficiency was estimated with a CF imaging system (Fluorimager; Technologica Ltd), exposing the plants to increasing actinic PPFD from 200 to 1400 µmol m À2 sec À1 in 200 µmol m À2 sec À1 steps every 5 min as described previously (Barbagallo et al., 2003;Gorecka et al., 2014).Whole rosette CF images were collected at each PPFD and processed using software (Technologica Ltd) to collect numerical data typically from fully expanded leaves (≥ 4 per plant) for Fq 0 /Fm 0 , Fv 0 /Fm 0 and Fq 0 /Fv 0 (Baker, 2008;Barbagallo et al., 2003;Gorecka et al., 2014) (Gorecka et al., 2014).The fluorimager software produces average data of all leaf pixel values.CF parameters were represented as mean AE SE from a minimum of four plants, and statistical significance was estimated with ANOVA followed by a post-hoc Tukey HSD test.

Measurement of photosynthesis
A was measured on leaf 7 of plants at 49 dpg using an infrared gas exchange system (CIRAS-1; PP Systems, Amesbury, MA, USA).The response of A to changes in the intercellular CO 2 concentration (C i ) was measured under a saturating PPFD, provided by a combination of red and white LEDs (PP Systems, Amesbury, MA, USA).In addition, the response of A to changes in PPFD from saturating to subsaturating levels was measured using the same light source at the current atmospheric CO 2 concentration (390 µmol mol À1 ).All gas analysis was made at a leaf temperature of 20 AE 1°C and a VPD of 1 AE 0.2 kPa.Plants were sampled between 1 and 4 h after the beginning of the photoperiod.For each leaf, steady-state rates of A at current atmospheric [CO 2 ] were recorded at the beginning of each measurement.

Relative ion leakage
The method described by Overmyer et al. ( 2008) was followed.Briefly, leaves were collected from plants and placed in 5 ml deionized water, incubated with rotary shaking (100 rpm) for 4 h, and the conductivity of the solution determined with a conductivity meter (Mettler Toledo, Leicester, UK) calibrated according to the manufacturer's instructions.Leaves were frozen overnight, thawed, and conductivity measured again.Relative ion leakage was expressed as conductivity after 4 h/conductivity after freezethawing.

RNA extraction, labelling and hybridization to microarrays
For the time-series HL experiment, RNA was extracted from leaf 7 samples, labelled and hybridized to CATMA (a Complete Arabidopsis Transcriptome MicroArray) microarrays (v3; Sclep et al., 2007), as described by Breeze et al. (2011).Two technical replicates were used per biological replicate.Four biological replicates with, in total, 13 time points per treatment (HL and LL) were analysed in this way, resulting in a highly replicated high-resolution time series of expression profiles.The experimental procedure for the hybridization of labelled cDNA samples for the HL and LL time series followed a statistically randomized loop design (Figure S9), which enabled expression to be determined at different time points both within and between treatments.After hybridization and washing, microarrays were scanned for Cy3 and Cy5 fluorescence and analysed as below.The raw and processed data are deposited with NCBI GEO (GSE78251).

Analysis of microarray data
This has been described in detail previously (Breeze et al., 2011;Windram et al., 2012).Briefly, a mixed model analysis using MAA- NOVA (Breeze et al., 2011;Wu et al., 2003) was used with the same random (dye and array slides) and fixed variables (time point, treatments, and biological replicate) to test the interaction between these factors for the analysis of time-series microarray data for senescing, Botrytis cinerea-infected, Pseudomonas syringae-infected and drought-stressed leaves (Bechtold et al., 2016;Breeze et al., 2011;Lewis et al., 2015;Windram et al., 2012).Predicted means were calculated for each gene probe for each of the combinations of treatment, biological replicate, and time point, and for each of the combinations of treatment and time point from averages of the biological replicates.A GP2S Bayes' factor (Stegle et al., 2010) was used to rank probes and genes in order of likelihood of differential expression over the whole of the time series.Inspection of selected probes from the rank order of likelihood of differential expression was used to identify significant changes in expression with a Bayes' factor cut-off >10 giving 4069 probes corresponding to 3844 DEGs (Data S1).

Clustering of gene expression profiles
The expression patterns of the identified DEGs in HL and LL were co-clustered with SPLINECLUSTER (Heard et al., 2005), using the mean expression profiles of the biological replicates generated from MAANOVA and a previous precision value of 0.001, as described previously (Bechtold et al., 2016;Windram et al., 2012).

GO analysis
GO annotation analysis was performed using DAVID (Huang et al., 2008) or AGRIGO (Du et al., 2010) with the GO Biological Process (BP) category (Ashburner et al., 2000).Overrepresented GO_BP categories were identified using a hypergeometric test with an FDR threshold of 0.05 compared against the whole annotated genome as the reference set.

Comparisons with published transcriptomics data
The 3844 HL DEGs were compared on a cluster-by-cluster basis with publicly available transcriptomics data.The references for each dataset can be found in the References.Each DEG list from published data was mapped to AGI codes when necessary, cleaned to obtain single AGI codes since in some microarray data, probes mapped to several genes or were listed as 'no_match' and were eliminated from the list.Overlaps within each cluster and their statistical significance were determined using a Hypergeometric Distribution Test [phyper function in R (v3.2.1)] in a custom R script, available upon request.When required, Venn diagrams of overlaps between datasets were plotted with Venny (http://bioinfogp.cnb.csic.es/tools/venny/index.html)and the significance of the overlaps calculated using the R phyper function.

VBSSM
A full description of VBSSM applied to this type of time-series transcriptomics data is provided in Bechtold et al. (2016).The individual expression data for each biological replicate (n = 4) for selected DEGs in HL was run through the VBSSM algorithm (Beal et al., 2005) on a local server at the University of Essex (Bechtold et al., 2016) to generate the GRNs as described in Results.The VBSSM output files were imported, mapped, and plotted with CYTOSCAPE (Shannon et al., 2003; http://www.cytoscape.org/).

Expression profiling by RNA-seq
Total RNA was extracted from mature leaves of each individual shoot giving four biological replicate samples per treatment and genotype.The RNA was quality controlled as previously described (Albihlal et al., 2018).Library construction after mRNA enrichment and double-stranded cDNA synthesis carried out using Illumina protocols by Novogene (UK) Ltd (Cambridge, UK; en.novogene.com/).Library sequencing was carried out on an Illumina HiSeq 4000 with a 150-bp end reads to a depth of 20 million.Extraction and quality control of data from raw fastq files were carried out using the program CASAVA (Hosseini et al., 2010).The mapping of reads to the TAIR10 Arabidopsis genome sequence, followed by sorting and indexing of BAM output files was carried out using default settings in the program HISAT2 (v2.0.5;Kim et al., 2015).Across all samples, >92.5% of bases read attained the Q30 score threshold.Transcript assembly and quantification was as fragments per kilobase of transcript sequence per million base pairs sequenced using HTseq (in union mode; Anders et al., 2015).Determination of differential expression between different genotypes and treatments was done using the program DESEQ2 (Love et al., 2014) after read count normalization and an adjusted P value threshold of <0.05 (negative binomial distribution P value model and Benjamini-Hochberg correction for multiple testing).Raw and processed data files were deposited in NCBI Gene Expression Omnibus (GSE158898).

Figure 1 .
Figure 1.Temporal patterns of gene expression in LL-and HL-exposed leaf 7. (a) Visual output of co-clustered expression values by SPLINECLUSTER.This was done for the 3844 genes already identified as differentially expressed in high light (HL) versus low light (LL) over the time of the experiment (see Results and Data S1).The values range from log 2 2.5 (red) to -log 2 2.0 (green).The 43 temporal clusters can be counted in the accompanying dendrogram.Time points are shown on the y-axes for the HL and LL gene expression.(b) Number of HL/LL differentially expressed (DE) probes first appearing at each time point.

Figure 2 .
Figure 2. Induction of acclimation by repeated daily exposure to high light (HL).(a) Plants were exposed daily to 4 h HL and Fq 0 /Fm 0 determined for mature leaves.After the HL, plants were dark adapted and imaged under increasing actinic photosynthetically active photon flux densities (PPFDs) from 200 to 1400 µmol m À2 sec À1 in 200 µmol m À2 sec À1 increments every 5 min.Data were collected as chlorophyll fluorescence images and processed digitally to collect values from mature leaves.Plants were treated in this way daily for 5 days: day 1 (blue), day 2 (red), day 3 (olive green), day 4 (purple), and day 5 (light blue).Data (mean AE SE) correspond to 38 plants at 24-28 days post-germination (dpg) over six experiments, and the asterisks show differences in chlorophyll fluorescence parameters between days 1 and 5 were significant (P ≤ 0.001; ANOVA and Tukey HSD).Full statistical data comparing all days of HL exposure are provided in Data S5.(b) Daily changes in Fq 0 /Fm 0 plotted from the data in A and Data S5 (right panel).Fq 0 /Fm 0 values are from the same plants over the daily HL exposures showing the increase in photosystem II (PSII) operating efficiency at 800 µmol m À2 sec À1 PPFD actinic light over the 5 days of the experiments.(c)Photosynthesis plotted as CO 2 assimilation rate (A) as a function of actinic PPFD in mature leaf 7 (mean AE SE; n = 8 plants for each treatment; 49 dpg).Measurements were taken the day after 1 (dashed lines) and 5 days (solid lines) of daily 4 h HL exposures (blue lines) along with the low light (LL) control plants (red lines) not subjected to this treatment.(d) Photosynthesis plotted as CO 2 assimilation rate (A) as function of leaf internal CO 2 concentration (C i ) in mature leaf 7 (mean AE SE; n = 8 plants for each treatment; 49 dpg).Measurements were taken the day after 5 days of daily 4 h HL exposures (blue line) along with the LL control (red line).A was determined by infra-red gas analysis (see Experimental procedures).Asterisks indicate significant differences (P < 0.02; covariant T and two-tailed F tests) between LL-and HLexposed plants.

©
2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384

Figure 3 .
Figure 3. Inferred high light (HL) gene regulatory network centred on BBX32.The network shown was generated from the time-series expression data for HL differentially expressed genes.Differentially expressed genes code for transcription (co-)factors that are also light-and/or PHYA/PHYB-regulated in de-etiolating seedlings.The network was generated using VBSSM (threshold z-score = 2.33; see Experimental procedures) and initially visualized using CYTOSCAPE (v3.3.2;Shannon et al., 2003) but redrawn manually to improve clarity.The network shown is from the second iteration of the modelling, which omitted expression data for LHY (First iteration; FigureS4a).Genes depicted in rectangular nodes were responsive to BBX32 overexpression in HL-and/or low light (LL)-exposed leaves by showing significantly (P adj < 0.05; negative binomial distribution probability model and Benjamini-Hochberg correction) higher (+) or lower (+) transcript abundance than Col-0 (see Figure5; Data S8).Locus codes for the network genes can be found in Experimental procedures.

Figure 4 .
Figure 4. Acclimation in BBX32 overexpression and bbx32-1 plants.Fq 0 /Fm 0 values determined from images of ≥4 mature leaves from eight plants (24-28 days post-germination) over two experiments (means AE SE), which had first been exposed to 4 h high light (HL) each day for 5 consecutive days (see Experimental procedures and legend for Figure 2).Chlorophyll fluorescence parameter values were collected at a range of actinic photosynthetically active photon flux densities (PPFDs) (as indicated) at the end of each daily HL exposure.(a) Fq 0 /Fm 0 values at day 1 (black lines) and day 5 (red lines) for mutant or overexpression plants (dashed line) and Col-0 (solid line) of the HL treatments for BBX32-10 and BBX32-12.Asterisks indicate difference between mutant genotype and Col-0 at day 5 (P < 0.01; ANOVA and Tukey HSD).(b) Daily Fq 0 /Fm 0 values at 800 µmol m À2 sec À1 PPFD actinic light of bbx32-1 compared with Col-0 showing differences that were significant (P < 0.01) only between days 2 and 4. (c) Photosynthesis plotted as CO 2 assimilation rate (A) as a function of incident PPFD in mature leaf 7 of low light-grown BBX32-10 (green line) and BBX32-12 (red line) compared with Col-0 (blue line) plants after 5 days of daily 4 h HL (see Experimental procedures).Data are the mean AE SE; n = 4 for each genotype at 49 days post-germination; Asterisk indicates significant differences (P < 0.02; covariant T and two-tailed F tests) between Col-0 and BBX32-10 and BBX32-12 at a given PPFD.Leaf A, as a function of PPFD, was determined by infra-red gas analysis (see Experimental procedures).

Figure 5 .Figure 6 .Figure 7 .
Figure 5. Partial validation of the BBX32-centric inferred gene regulatory network.Expression of 25 of the 47 transcription factor genes in the inferred network showing the effect of BBX32 overexpression.All the genes displayed significant differences (P adj < 0.05; negative binomial distribution probability model and Benjamini-Hochberg corrected) in transcript abundance in four replicate BBX32 overexpression plants compared with four Col-0 both under low light (LL; suffix 'a') and/or high light (HL; suffix 'b') conditions.Data are mean FPKM (n = 4 AE SE) from four plants per genotype and treatment.Tabulated FPKM data for these genes can be found in Data S8.Colour codes are brown and dark blue for Col-0 and BBX32-10 plants in LL respectively, salmon pink and light blue are Col-0 and BBX32-10 plants in HL.Cluster number for each gene is shown on each graph.

©
2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384

Figure 8 .
Figure 8. High light (HL) acclimation of photoreceptor and HY5 mutants.Plots show the photosystem II operating efficiencies (Fq 0 /Fm 0 ) determined from chlorophyll fluorescence images of ≥4 mature leaves from eight plants over two experiments (means AE SE).Plants had been exposed to 4 h HL each day for 5 consecutive days (see Experimental procedures and legend of Figure 2).Chlorophyll fluorescence parameter values were collected at a range of actinic photosynthetically active photon flux densities (PPFDs) at the end of days 1 and 5 of HL.Fq 0 /Fm 0 values at day 1 (black lines) and day 5 (red lines) for mutant plants (dashed line) and Col-0 (solid line) of the HL treatments for (a) cry1-304, (b) cry1-M32, (c) YHB, (d) hy5-2, and (e) hy5-215.Asterisks (panels a, b, d, e) indicate significant difference between mutant compared with Col-0 at day 5 (P < 0.01; ANOVA and Tukey HSD).Upward arrows (c) indicate significant difference between YHB and Col-0 at day 1 (P < 0.01; ANOVA and Tukey HSD).© 2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384 ; Data S8) and © 2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384

©
2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384

Figure 9 .
Figure 9. High light (HL) acclimation of photoreceptor signal transduction mutants.(a) Photosynthetic efficiency of the same single Col-0, cop1-4, and det1-1 plants after 1 and 5 days of daily 4 h HL exposure.Chlorophyll fluorescence (CF) images are of Fq 0 /Fm 0 (photosystem II operating efficiency) at a 400 µmol m À2 sec À1 actinic photosynthetically active photon flux densities (PPFDs).(b-e) Plots show the photosystem II operating efficiencies (Fq 0 /Fm 0 ) determined from CF images from eight plants (24-28 days post-germination) over two experiments (means AE SE).Plants had been exposed to 5 days of daily 4 h HL (see Experimental procedures and legend of Figure2).Note that because of the size of the cop1-4, pifq, and det1-1 plants, data were collected from whole rosettes rather than from mature leaves.CF parameter values were collected at a range of actinic PPFDs at the end of days 1 and 5 of HL.Fq 0 /Fm 0 values at day 1 (black lines) and day 5 (red lines) for mutant plants (dashed line) and Col-0 (solid line) of the HL treatments for (b) cop1-4, (c) spa1,2,3, (d) det1-1, and (e) pifq.Asterisks (e) indicate significant difference between mutant compared with Col-0 at day 5 (P < 0.01; ANOVA and Tukey HSD).Upward arrows (b,c) indicate significant difference between mutants and Col-0 at day 1 (P < 0.01; ANOVA and Tukey HSD).

©
2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384

©
2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384

Figure S1 .
Figure S1.Examples of temporal differentially expressed clusters.Figure S2.Comparison of plants exposed to daily HL for 5 days with their equivalent age LL controls and induction of HL acclimation in older plants.Figure S3.HL exposure used does not produce extensive photodamage to leaves.Figure S4.First draft inferred HL gene regulatory network and HL acclimation in lhy-21 plants.Figure S5.Temporal patterns of transcript abundance of the 12 most connected genes in the BBX32-centric GRN under LL and HL conditions Figure S6.HL acclimation of phyB-9, phyA-211, cry2-1, and pif4-2 plants compared with Col-0.Figure S7.Spectral properties of the growth lights and isolight used for HL exposure.Figure S8.Identification and characterization of cry1-M32.

Figure S2 .
Figure S1.Examples of temporal differentially expressed clusters.Figure S2.Comparison of plants exposed to daily HL for 5 days with their equivalent age LL controls and induction of HL acclimation in older plants.Figure S3.HL exposure used does not produce extensive photodamage to leaves.Figure S4.First draft inferred HL gene regulatory network and HL acclimation in lhy-21 plants.Figure S5.Temporal patterns of transcript abundance of the 12 most connected genes in the BBX32-centric GRN under LL and HL conditions Figure S6.HL acclimation of phyB-9, phyA-211, cry2-1, and pif4-2 plants compared with Col-0.Figure S7.Spectral properties of the growth lights and isolight used for HL exposure.Figure S8.Identification and characterization of cry1-M32.

Figure S3 .
Figure S1.Examples of temporal differentially expressed clusters.Figure S2.Comparison of plants exposed to daily HL for 5 days with their equivalent age LL controls and induction of HL acclimation in older plants.Figure S3.HL exposure used does not produce extensive photodamage to leaves.Figure S4.First draft inferred HL gene regulatory network and HL acclimation in lhy-21 plants.Figure S5.Temporal patterns of transcript abundance of the 12 most connected genes in the BBX32-centric GRN under LL and HL conditions Figure S6.HL acclimation of phyB-9, phyA-211, cry2-1, and pif4-2 plants compared with Col-0.Figure S7.Spectral properties of the growth lights and isolight used for HL exposure.Figure S8.Identification and characterization of cry1-M32.

Figure S4 .
Figure S1.Examples of temporal differentially expressed clusters.Figure S2.Comparison of plants exposed to daily HL for 5 days with their equivalent age LL controls and induction of HL acclimation in older plants.Figure S3.HL exposure used does not produce extensive photodamage to leaves.Figure S4.First draft inferred HL gene regulatory network and HL acclimation in lhy-21 plants.Figure S5.Temporal patterns of transcript abundance of the 12 most connected genes in the BBX32-centric GRN under LL and HL conditions Figure S6.HL acclimation of phyB-9, phyA-211, cry2-1, and pif4-2 plants compared with Col-0.Figure S7.Spectral properties of the growth lights and isolight used for HL exposure.Figure S8.Identification and characterization of cry1-M32.

Figure S7 .
Figure S1.Examples of temporal differentially expressed clusters.Figure S2.Comparison of plants exposed to daily HL for 5 days with their equivalent age LL controls and induction of HL acclimation in older plants.Figure S3.HL exposure used does not produce extensive photodamage to leaves.Figure S4.First draft inferred HL gene regulatory network and HL acclimation in lhy-21 plants.Figure S5.Temporal patterns of transcript abundance of the 12 most connected genes in the BBX32-centric GRN under LL and HL conditions Figure S6.HL acclimation of phyB-9, phyA-211, cry2-1, and pif4-2 plants compared with Col-0.Figure S7.Spectral properties of the growth lights and isolight used for HL exposure.Figure S8.Identification and characterization of cry1-M32.

Figure S8 .
Figure S1.Examples of temporal differentially expressed clusters.Figure S2.Comparison of plants exposed to daily HL for 5 days with their equivalent age LL controls and induction of HL acclimation in older plants.Figure S3.HL exposure used does not produce extensive photodamage to leaves.Figure S4.First draft inferred HL gene regulatory network and HL acclimation in lhy-21 plants.Figure S5.Temporal patterns of transcript abundance of the 12 most connected genes in the BBX32-centric GRN under LL and HL conditions Figure S6.HL acclimation of phyB-9, phyA-211, cry2-1, and pif4-2 plants compared with Col-0.Figure S7.Spectral properties of the growth lights and isolight used for HL exposure.Figure S8.Identification and characterization of cry1-M32.

Figure S9 .
Figure S9.Randomized loop design employed for loading samples on to the two-channel CATMA arrays.Data S1.The 3844 HL DEGs on a cluster-by-cluster basis.Data S2.Significantly enriched Biological Process GO Terms for the HL DEGs in each temporal cluster.Data S3.Differentially expressed genes from leaf 7 of Arabidopsis subjected to a 30-min temperature rise from 22 to 27°C under LL and HL conditions.Data S4.Significant overlaps, on a cluster-by-cluster basis, between publicly available gene expression datasets and transcriptomic meta-analyses for mutants and treatments that perturb © 2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384 . In some experiments, the diminished size of mutant plants rendered image processing problematic and © 2021 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2021), doi: 10.1111/tpj.15384 in such stated cases, whole rosette data were collected.The raw data were fed via Excel into a program in R to calculate, plot, and analyse statistically the CF parameters