Photoperiod-responsive changes in chromatin accessibility in phloem companion and epidermis cells of Arabidopsis leaves

Abstract

Photoperiod plays a key role in controlling the phase transition from vegetative to reproductive growth in flowering plants. Leaves are the major organs perceiving day-length signals, but how specific leaf cell types respond to photoperiod remains unknown. We integrated photoperiod-responsive chromatin accessibility and transcriptome data in leaf epidermis and vascular companion cells of Arabidopsis thaliana by combining isolation of nuclei tagged in specific cell/tissue types with assay for transposase-accessible chromatin using sequencing and RNA-sequencing. Despite a large overlap, vasculature and epidermis cells responded differently. Long-day predominantly induced accessible chromatin regions (ACRs); in the vasculature, more ACRs were induced and these were located at more distal gene regions, compared with the epidermis. Vascular ACRs induced by long days were highly enriched in binding sites for flowering-related transcription factors. Among the highly ranked genes (based on chromatin and expression signatures in the vasculature), we identified TREHALOSE-PHOSPHATASE/SYNTHASE 9 (TPS9) as a flowering activator, as shown by the late flowering phenotypes of T-DNA insertion mutants and transgenic lines with phloem-specific knockdown of TPS9. Our cell-type-specific analysis sheds light on how the long-day photoperiod stimulus impacts chromatin accessibility in a tissue-specific manner to regulate plant development.

Cell-type-specific changes in chromatin accessibility and transcriptomes occur after a shift from short- to long-day growth conditions.

Introduction

Flowering time is an important contributor to reproductive success and fitness of higher plants (Huijser and Schmid, 2011). Many different pathways including the photoperiod (day length), gibberellin acid, vernalization (a period of cold), thermo-sensory, autonomous, sugar, and aging pathways play critical roles in the regulation of flowering time in Arabidopsis thaliana (Arabidopsis) (Putterill et al., 2004; Turck et al., 2008; He, 2009; Wang et al., 2009; Wu et al., 2009; Andres and Coupland, 2012; Kumar et al., 2012; Wahl et al., 2013).

In the photoperiod pathway, transcription of CONSTANS (CO) is activated by GIGANTEA-mediated degradation of transcriptional repressors of the CYCLING DOF FACTOR (CDF) family (Imaizumi et al., 2005; Sawa et al., 2007; Fornara et al., 2009; Cheng et al., 2019). Under long days (LDs), CO, NUCLEAR FACTOR (NF)-YB, and NF-YC form a complex that directly activates transcription of FLOWERING LOCUS T (FT) (Wenkel et al., 2006; Gnesutta et al., 2017). As part of florigen, FT protein moves through the phloem from the distal leaf veins to the shoot apical meristem (SAM) to promote flowering (Corbesier et al., 2007; Tamaki et al., 2007; Turck et al., 2008). FT is mainly transcribed in the companion cells (CCs) of the leaf phloem (Kotake et al., 2003; Adrian et al., 2010), which suggests that these cell types alone take responsibility for providing flowering stimuli (Lopez-Salmeron et al., 2019). In contrast, it is still unclear, how many leaf-cell types perceive the photoperiod signal and whether cell-type-specific responses to photoperiod exist. An indication that tissue-specific responses occur was provided by the detection of an independent and robust circadian clock system in the leaf vasculature compared to mesophyll cells after physical separation of leaf tissue (Endo et al., 2014).

Several methods for purifying-specific cell types based on the expression of tissue marker genes have been applied in plants, such as fluorescence-activated cell sorting and isolation of nuclei tagged in specific cell types (INTACT) (Brady et al., 2007; Deal and Henikoff, 2011; Kawakatsu et al., 2016; Svozil et al., 2016). INTACT has been widely used in epigenetic studies due to its low cost and convenient operation (Deal and Henikoff, 2010). In brief, a cell/tissue-specific promoter is chosen to drive an engineered “nuclear targeting fusion” (NTF) protein that is composed of a nuclear envelope localization domain, a GREEN FLUORESCENT PROTEIN, and a biotin receptor peptide. After biotinylation of NTF that is catalyzed by a constitutively expressed bacterial BirA enzyme, nuclei of interest are purified using magnetic streptavidin-coated beads (Deal and Henikoff, 2010, 2011). INTACT has been used to isolate Arabidopsis root hair and nonroot hair nuclei to detect the cell-type-specific gene expression and histone modifications (Deal and Henikoff, 2010) and to explore histone modifications in the SAM and leaf phloem CCs (You et al., 2017, 2019).

We are interested in the specific response of cis-regulatory elements in different leaf cells in sensing the LD signal. Two genome-wide methods are available that allow both, detection of accessible chromatin and identification of putative transcription factor binding sites (TFBSs) through the occurrence of protected footprints within accessible chromatin regions (ACRs). While DNase-seq has been adapted to study the regulatory sequences in maize and Arabidopsis (Zhang et al., 2012; Oka et al., 2017; Alvarez et al., 2019; Yan et al., 2019), it is not well suited to be combined with low amounts of starting material. In contrast, assay for transposase-accessible chromatin using sequencing (ATAC-seq), making use of the transposase activity of the Tn5 enzyme coupled with oligonucleotide adapters that allow direct generation of high-throughput sequencing libraries, is well adapted to capture regulatory elements from limited starting material (Buenrostro et al., 2013; Lu et al., 2017; Maher et al., 2018; Potter et al., 2018; Reynoso et al., 2019). In plants, a combination of INTACT and ATAC-seq identified cell-specific ACRs in root hairs, shoot stem cells, and mesophyll cells (Bajic et al., 2018; Sijacic et al., 2018).

Our study investigated cell-type-specific and LD-responsive ACRs and their relative enrichment for TFBS. We separated ACRs with respect to their different genomic location at target genes. We combined INTACT, ATAC-, and RNA-seq to separate single cell types corresponding to phloem CCs and epidermis. In both cell types, chromatin accessibility was generally increased after perception of the LD signal; however, variable and specific response ACR patterns were detected by comparing both cell types. LD-induced ACRs (iACRs) pinpointed previously characterized flowering-related genes but also novel loci, such as TREHALOSE-PHOSPHATASE/SYNTHASE 9 (TPS9), for which we confirmed a positive role in flowering time regulation. iACRs of CCs were particularly enriched for binding events of flowering-related TFs. Our work highlights the importance of cell-type-specific studies to elucidate the function of noncoding elements in the regulation of gene regulatory networks.

Results

INTACT isolation yields highly purified nuclei from Arabidopsis leaf vasculature and epidermis

The leaf is the major organ that perceives photoperiodic signals. Several known floral regulators like FT and CO are mainly expressed in phloem CC, whereas others like the floral repressors FLOWERING LOCUS C (FLC) and SHORT VEGETATIVE PHASE (SVP) do not show a tissue-specific pattern of expression (Kotake et al., 2003; Lee et al., 2007; Castillejo and Pelaz, 2008; Li et al., 2008; Sawa and Kay, 2011). To dissect the specific function of the leaf vasculature in the regulation of flowering, we established an INTACT approach to enrich both phloem and epidermis nuclei.

In the background of Arabidopsis ACT2p:BirA expressing transgenic plants (Deal and Henikoff, 2010), we introduced Cucumis melo GALACTINOL SYNTHASE 1 (GAS1) (Haritatos et al., 2000) and Arabidopsis FIDDLEHEAD (FDH) (Pruitt et al., 2000) promoters to specifically drive the expression of NTF in CC and epidermis, respectively (transgenic lines named GAS1p:NTF and FDHp:NTF) (Figure 1, A and B). After sorting nuclei according to the INTACT procedure (see Materials and methods section and Deal and Henikoff, 2011), magnetic dynabeads specifically interacted with single nuclei in both GAS1p:NTF and FDHp:NTF samples, while there were almost no free nuclei in the sorted samples (Figure 1, C and D). To further validate the purity, we introduced a spike-in approach by adding Landsberg erecta (Ler) background material into the GAS1p:NTF or FDHp:NTF (Col-0 ecotype) samples before sorting (Moreno-Romero et al., 2017), and measured the Ler to Col-0 DNA ratio before and after INTACT. A purity of ˃98% was detected for both cell types (Figure 1E).

Figure 1.

Combination INTACT with ATAC- and RNA-seq to study chromatin accessibilities and gene expressions in vascular and epidermis cells. (A, B) Fluorescence microscopy images showing the cell-type-specific NTF signals (GREEN FLUORESCENT PROTEIN channel) in transgenic plant lines expressing GAS1p:NTF (A) and FDHp:NTF (B), respectively, in phloem CC and epidermis. Bar = 50 μm. (C, D) Fluorescence microscopy images showing streptavidin-coated magnetic beads after INTACT purification. White arrowheads indicate the beads bound by DAPI-stained NTF-labeled nuclei. Bar = 20 μm. (E) Purity of NTF-labeled nuclei isolated by INTACT from GAS1p:NTF;ACT2p:BirA and FDHp:NTF;ACT2p:BirA plants detected by comparing relative enrichment of Col-0 and spiked-in Ler molecular markers before and after purification by qPCR. (F) Flowering time of plants transiently shifted to LD as indicated after 14 days of growth in SD, controls were grown in constant SDs and LDs. Flowering time determined as total leaf number. Bar plots show mean±s.d. Significant differences were determined by t-test. *P < 0.05, ***P < 0.001, and ns indicates not significant. (G) Overview of the experimental design.

To understand chromatin accessibility and transcriptome variation in CC and epidermis in response to LD, we performed INTACT sorting after shifting 14-day-old short day (SD) grown seedlings for 3 days to LD; control plants remained in SD until harvest. GAS1p:NTF and FDHp:NTF plants were sampled before dawn in LD (ZT16) and SD (ZT8) conditions. In our greenhouse conditions, plants shifted for three LD flowered almost as early as plants that were continuously grown in LD (Figure 1F; Supplemental Figure S1), consistent with previous findings in Arabidopsis (Corbesier et al., 2007). As expected, expression of the flowering integrator gene FT was induced after three LD compared to SD controls (Supplemental Figure S2).

Quality control of ATAC-seq data and identification of ACRs

INTACT-isolated nuclei from leaf vascular CC (Va LD and Va SD) and epidermis cells (Ep LD and Ep SD) were further analyzed for their transcriptome and chromatin accessibility (Figure 1G; Supplemental Table S1). To evaluate the quality of our ATAC-seq data, we compared it with published DNase-seq (Zhang et al., 2012) and ATAC-seq (Lu et al., 2017; Maher et al., 2018; Potter et al., 2018) datasets. After removing organellar reads from the total filtered reads, on average 88% (ranging from 84% to 93%) nuclear genomic reads remained in our data set, which is comparable to the genomic reads in published data (Supplemental Figure S3). To identify ACRs, the enrichment of ATAC-seq reads was analyzed by using MACS2 (Model-based Analysis of ChIP-Seq 2) peak mapping software (Zhang et al., 2008b). Overlapped peaks called in both biological replicates of each cell type and treatment were considered as ACRs. We identified 16,847, 19,501, 15,064, and 17,330 ACRs in Va LD, Ep LD, Va SD, and Ep SD, respectively (Supplemental Figure S4A). Compared to published data, ˃80% of DNase I Hypersensitive Sites from wild-type leaves (Zhang et al., 2012) were reproduced in our INTACT-ATAC ACRs (Supplemental Figure S4B).

To estimate the signal to noise ratio of our ATAC-seq data, we analyzed the Signal Portion of Tags (SPOTs) value, which counts the percentage of nuclear reads enriched at ACRs compared to all mapped reads. We found average SPOT values ranging from 68% to 79%, which is comparable to SPOT average values of published data (Supplemental Figure S5). A genome browser overview of an exemplary 200-kb genomic region illustrates the occurrence of enriched signals as sharp peaks and a comparable low background in all four samples (Figure 2A). We performed a principal component analysis (PCA) comparing the enrichment values at all ACRs across all samples. Biological replicates clustered most closely together (Supplemental Figure S6). Taken together, 15,000–20,000 ACRs were identified in each leaf cell type grown in different photoperiods from a dataset of high quality.

Figure 2.

Characterization of the proximal and distal ACRs. (A) Genome browser showing ATAC-seq signals in a 200-kb region (Chr1:20,957,197-21,159,077) in four samples as indicated. Tracks show normalized values of one representative biological replicate for each sample. (B) The percentage of ACRs corresponding to categories Intergenic, TSS, Genebody, and TES for Ep SD, Ep LD, Va SD, and Va LD. TSS: ACRs located within −1,000 to +50 bp of a gene’s TSS; TES: ACRs located within −50 to +1,000 bp of a gene’s TES; Genebody: ACRs located within the genebody; Intergenic: all other ACRs. Number of ACRs indicated in bar plots. (C) Enrichment of conserved CNSs in distal (Intergenic) and proximal (all other ACRs) regions, control shows shuffled sequences from all ACRs. P-values were calculated based on Wilcoxon signed-rank tests. **P < 0.01. (D) Boxplots showing the number of potential TFBS in different ACR categories. 529 DAP-seq derived matrices for TFs were included in the analysis. P-values were calculated based on Wilcoxon signed-rank tests. ***P<0.001. (E) Boxplots showing expression levels of genes associated with TSS, Genebody, TES, and Intergenic ACRs. Boxplots show the median (horizontal line), second to third quartiles (box), and Tukey-style whiskers (beyond the box). Statistical significance was calculated based on pairwise Wilcoxon signed-rank tests.

Differential features of proximal and distal ACRs

We further analyzed the distribution of ACRs to different genomic regions in all four samples. The majority of ACRs (∼86%) localized at gene proximal regions [including ˂1-kb upstream of the transcriptional start site (TSS) ∼53%, genebody ∼22%, and ˂1-kb downstream of the transcription termination site (TES) ∼11%] and a minority (∼14%) of ACRs at gene distal or intergenic regions (Figure 2B). When overlapped with a previously determined set of conserved non-coding sequences (CNSs) (Van de Velde et al., 2014), the intergenic ACRs enriched a higher fraction of CNS than proximal ACRs and control regions (Figure 2C). Enrichment of putative Arabidopsis TFBS, which were determined using available DAP-seq peaks for 529 TFs (O'Malley et al., 2016) was detected for proximal and distal ACRs over control regions (Figure 2D). In addition, the expression level of genes associated only with TSS or genebody ACRs was significantly higher than those only with intergenic or TES ACRs, while genes associated with more than one ACR category expressed overall at higher levels than those associated with fewer (Figure 2E).

Taken together, ACRs overrepresented CNS and putative TFBS, the presence of a proximal and/or genebody ACRs correlated with higher expression of associated genes than their absence and a general additive trend of ACRs of different categories on gene expression levels became obvious.

The chromatin response to day length is tissue-specific

To further understand variation in chromatin accessibility during the shift from SD to LD in leaf-vasculature CC and epidermis cells, we identified responsive ACRs that were differentially enriched between samples (Supplemental Figure S7A and see Materials and methods section). If an ACR was identified in both biological replicates with MACS2 (FDR [false discovery rate] < 0.01) and showed increased or decreased tagment signals ˃1.5-fold after the shift from SD to LD in vasculature, it was considered as vasculature induced and repressed ACRs (i/rACRs); the same criteria were applied to identify epidermis i/rACRs (Supplemental Figure S7A and see Materials and methods section). Cell-type-specific responsive ACRs were further classified as vasculature-specific (Va), epidermis-specific (Ep), and overlapped (Va&Ep) i/rACRs (Supplemental Figure S7, B and C and Materials and methods section).

Overall, we obtained 6,196 iACRs, composed of 2,861 Va iACRs and 1,486 Ep iACRs as well as 1,849 Va&Ep iACRs (Figure 3A; Supplemental Figures S7 and S8; Supplemental Data Set 1). In general, LD predominantly leads to chromatin opening and phloem CCs appear to be more sensitive than cells of the epidermis, since the number of Va iACRs was ∼2 times higher than that of Ep iACRs. Overall, the number of rACRs was dramatically smaller than that of iACRs; 185 rACRs were counted as Va-specific, 722 as Ep-specific, and 138 as overlapping between the tissues (Figure 3B; Supplemental Figures S7 and S8; Supplemental Data Set 1). Most responsive ACRs showed quantitative changes of chromatin accessibility; however, in rare cases, ACRs were unique to a specific photoperiod condition (Supplemental Figure S9).

Figure 3.

Characterization of LD induced or repressed ACRs in isolated cell types. (A) and (B) Heatmaps showing ATAC-seq coverage from −1 to +1 kb around the center of LD-induced (A) and LD-repressed ACRs (B). Right panels show fold-changes in coverage between SD and LD samples. Va response ACRs; Ep response ACRs; Va&Ep, overlapped response ACRs between two tissues. (C) GO analysis showing enriched pathways of genes associated with LD-induced and LD-repressed ACRs per cell-type-specific category. (D) Bar plots showing percentage of responsive ACRs per ACR category. The number of all ACRs per category and tissue type is set to 100%. P-values were calculated based on Fisher’s exact tests. *P < 0.05, **P < 0.01, ***P < 0.001, and ns indicates not significant.

Genes associated with iACRs enriched several gene ontologies (Figure 3C; Supplemental Data Set 2). Importantly, large differences in enriched GO-terms were found between vasculature and epidermis cell types. Pathways related to circadian rhythm, regulation of gene expression, developmental processes, and hormone response enriched predominantly in the vasculature CC, whereas pathways related to defense response, stress response, and response to bacteria and fungi enriched in epidermis cells (Figure 3C). Genes associated with rACRs in the vasculature CC enriched GO-terms related to response to nitrogen compounds and oxidative stress, while rACR associated genes in epidermis cells enriched pathways related to photosynthesis, light harvesting, and light stimulus (Figure 3C). We evaluated if ACRs corresponding to different genomic categories showed a different response to photoperiod changes and whether such differences were tissue specific. A comparison of the fraction of responsive ACRs within each genomic category showed that distal iACRs were the most responsive category in the Va samples while a slightly higher proportion of proximal (TSS and genebody) ACRs were induced by LD in the epidermis (Figure 3D; Supplemental Figure S10).

Taken together, chromatin regions are more likely to open than close during a shift to LD in both leaf cell types and this trend is even stronger in vasculature CC than that in epidermis cells. In CC, the proportion of distal ACRs induced by LD is higher than that of proximal ACRs, while a trend toward induction of proximal ACRs is detected in the epidermis. Despite a large overlap in the response, distinct gene sets respond predominantly to the change in photoperiod in epidermis and vasculature.

Altered chromatin access at distal and proximal ACRs correlates with gene expression

Gene expression is strongly correlated with chromatin accessibility at TSS (Sullivan et al., 2014; Klemm et al., 2019), but a similar link could so far not be established between gene expression and distal ACRs in plants (Oka et al., 2017). INTACT-sorted nuclear RNA-seq has been widely used to analyze global gene expression in root hair, nonroot hair, mesophyll, phloem, and shoot meristem cells (Deal and Henikoff, 2010; You et al., 2017; Sijacic et al., 2018; You et al., 2019). It was previously shown for plants and animals that nuclear RNA-seq data can represent similar expression profiles as whole-cell RNA-seq data (Jacob et al., 2007; Zhang et al., 2008a). To evaluate how photoperiod-induced differential gene expression was related to proximal and distal ACRs in leaf vasculature and epidermis, we carried out RNA-seq with INTACT isolated nuclei (see Materials and methods section).

Pairwise comparisons were performed among the four samples; genes with greater than two-fold change and FDR < 0.05 were considered as differentially expressed genes (DEG). Generally, more vascular genes were activated in LD, whereas more epidermis genes were repressed. Among the LD-activated DEG, 408 were specific to the vasculature, 257 to the epidermis, and 288 detected in both cell types. Among the LD-repressed DEG, 145 were specific to the vasculature, 507 to the epidermis, and 330 detected in both cell types (Figure 4A; Supplemental Data Set 3). Tissue-specific marker genes such as ALTERED PHLOEM DEVELOPMENT (Bonke et al., 2003) and ARABIDOPSIS THALIANA ARABIDOPSIS H(+)-ATPASE 3 (DeWitt and Sussman, 1995) for vasculature, as well as ECERIFERUM 1 (Bourdenx et al., 2011) and PERMEABLE LEAVES 3 (Panikashvili et al., 2009) for epidermis were detected as differentially expressed in our RNA-seq dataset (Supplemental Figure S11A). GO-term enrichment analysis revealed that DEG in the vasculature CC related to hormone response and endogenous stimulus, and to pathways related to defense response and light stimulus in epidermis cells (Supplemental Figure S11B).

Figure 4.

Chromatin accessibility variation is correlated with differential gene expression. (A) Heatmap showing the Z-score of expression for each gene grouped by different DEG category. (B) Heatmap and average line plots showing Z-scores of expression variation for each gene associated with different ACR categories. For each tissue response type, the right panel shows the relative level of chromosome accessibility enrichment (black) and averages of gene expression scores per ACR categories: Intergenic (red), TSS (blue), Genebody (gray), and TES (yellow). (C) Association of responsive ACR categories [distal (Intergenic) and proximal (all other categories) per tissue type] with genes that show correlated differential expressed. Significant P-values were calculated with a permutation test (n = 100). **P < 0.01. (D) Genome browser showing ATAC-seq and RNA-seq signals around representative LD activated or repressed ACRs in vascular and epidermal cells. Note that changes in expression level of associated genes change in the same direction as chromatin access at ACRs. (E) Representative candidates of LD activated DEGs in phloem CCs that are associated with proximal (left panel) or distal iACRs (right panel). Raw read counts were normalized with DEseq2 for each tissue and treatment to compare gene expression in different samples.

Gene expression in LD clearly increased with increasing chromatin accessibility; likewise, reduced expression was linked to decreased chromatin access as roughly 70% of all dynamically changing proximal and distal ACRs showed a correlated variation in the expression of associated genes (Figure 4B). To test the significance of the observed relationship between differential chromatin accessibility and gene expression, we created random permutations of the DEG dataset and re-calculated how DEG related to i/rACRs after permutation. Across both tissues and irrespective of the genomic location of ACRs, the number of upregulated DEG related to iACRs and that of downregulated DEG related to rACRs was significantly higher than expected (Figure 4C). Representative examples for LD responsive i/rACRs associated with DEGs in vasculature CC and epidermis are shown in a genome browser (Figure 4D).

In sum, changes in chromatin access can be directly linked to gene expression not only for proximal but also for distal intergenic ACRs. Cell types carry a significant chromatin-to-expression-change signature, which may be linked to their selective response to the LD signal.

Identification of flowering regulators in the phloem based on chromatin-to-expression signature

We ranked our data sets according to the occurrence of distal and proximal iACRs related to transcriptional induction. Among the 14 top-ranked genes from the leaf CC dataset (Supplemental Data Set 4), genes with an established or putative link to flowering were prominent: ARABIDOPSIS RESPONSE REGULATORS (ARRs) 7, ARR4, and BBX17 clustered to the proximal ACR group; and FT and TPS9 clustered to distal ACR group (Figure 4E; Supplemental Figure S12).

We previously identified a distal enhancer, Block C, which participates in promoting FT expression in the leaf vasculature under LD (Adrian et al., 2010; Liu et al., 2014). The chromatin accessibility in the Block C region increases in response to LD, particularly in CCs, a result that could be confirmed by an independent ATAC-qPCR (ATAC followed by quantitative real-time PCR) assay (Figure 5A). As expected, RNA-seq data show that FT expression was significantly induced by LD in CC (Figure 5A). These results demonstrate that a characterized distal enhancer of FT carries the typical chromatin-to-expression signature that is observed in genome-wide data.

Figure 5.

FT and TPS9 associated with LD induced ACRs promote floral transition. (A and B) Genome browser tracks showing ATAC- and RNA-seq coverage around FT and TPS9 in vascular and epidermis cells. Orange and blue colors mark the LD and SD treatments, respectively. ACRs and primers are indicated as red boxes. Relative density of ATAC-seq signals and ATAC-qPCR enrichment for FT and TPS9 genes are shown in the panels on the right. For ATAC-qPCR, bar plots show mean±s.d. based on two biological replicates, and each biological replicate we included three PCR technical replicates. ATAC-seq P-value was calculated by Poisson test. *P < 0.05, ***P < 0.001, NS indicates not significant. ATAC-qPCR P-value was calculated by t test. ***P < 0.001, ns indicates not significant. (C) Representative image and flowering phenotype of wild-type Col-0, tps9-1, and tps9-2 plants. Bar plots show mean±s.d. of total leaf numbers in wild-type Col-0 (n = 18), tps9-1 (n = 21), tps9-2 (n = 17). P-value was calculated by t-test. *P < 0.05 and **P < 0.01. (D) Bioluminescent image of transgenic plant SUC2p:LUC. Bar = 100 μm. (E) TPS9 expression levels were measured by RT-qPCR in 12 independent transgenic lines of SUC2p:LUC, SUC2p:amiRNA-TPS9#1, and SUC2p:amiRNA-TPS9#2. Two mature leaves were collected from one transgenic plant harboring 7–8 true leaves. PP2A was used as internal control. Bar plots show mean±s.d. of three technical replicates. (F) The flowering-time phenotype of SUC2p:amiRNA-TPS9 and two controls of SUC2p:LUC and 5.7kbFTp:GUS. Basta positive plants were transferred to new soil pots in SDs for 3 weeks, and then shifted to LDs until bolting. The transgenic plants of SUC2p:LUC treated as vector control; and 5.7kbFTp:GUS in Col-0 background were treated as Basta positive control. Image of representative plants per genotype. (G) Number of total leaves in SUC2p:LUC (n = 86 independent T1 lines), SUC2p:amiRNA-TPS9#1 (n = 91 independent T1 lines), SUC2p:amiRNA-TPS9#2 (n = 95 independent T1 lines) and 5.7kbFTp:GUS (1 homozygous line, n = 23 individuals). Boxplots show the median (horizontal line), second to third quartiles (box) and s.d. (beyond the box). P-value was calculated by t test. ***P < 0.001, ns indicates not significant.

We selected TPS9 as a new representative for an analysis of gene function, which shows rapidly changing expression levels according to a light/dark diurnal cycle (Usadel et al., 2008). TPS1, a homolog of TPS9, accelerates the floral transition by both, activating FT expression in leaves and promoting the flowering program in SAM (Wahl et al., 2013). The promoter of TPS9 drives expression of the GUS reporter gene in the vasculature of leaves, roots, and shoot apices (Ramon et al., 2009); however, the biological function of TPS9 is not yet resolved.

Chromatin accessibility of ACR associated with TPS9 gene as well as RNA abundance was obviously higher in Va LD compared to Va SD (Figure 5B). In detail, we found 1.7-fold change (Va LD/Va SD) at iACR2 associated with the promoter of TPS9 (Figure 5B). ATAC-seq results were confirmed by ATAC-qPCR using primers to measure the relative fold changes at iACR2 of TPS9 (Figure 5B). To assess a potential function of TPS9 in the floral transition, we characterized Salk T-DNA insertion mutants. Two independent tps9 mutant lines (SALK_086992C and SALK_151275C, named tps9-1 and tps9-2, respectively) flowered significantly later than Col-0 control in LDs (Figure 5C; Supplemental Figure S13), while no obvious difference of flowering time was observed in SDs (Supplemental Figure S14) indicating that TPS9 contributes to promoting the floral transition in a photoperiod-dependent manner.

To further access the function of phloem-expressed TPS9 in flowering time regulation, we generated transgenic lines that knock-down TPS9 specifically in the phloem. Two independent artificial microRNAs (amiRNA), amiRNA-TPS9#1 and amiRNA-TPS9#2, which target exons 2 and 3 of TPS9, were driven by the phloem-specific SUCROSE-PROTON SYMPORTER 2 (SUC2) promoter (Truernit and Sauer, 1995) in the Col-0 background (Supplemental Figure S13). As control, a SUC2p:LUC (Luciferase gene) transgenic line showing a significant signal in leaf veins was used (Figure 5D). In addition, to remove the side effect on plant development by the herbicide Basta (glufosinate) that was used for selection of transgenic plants (Abdeen and Miki, 2009; Christ et al., 2017), we included the Basta-resistant homozygous line 5.7kbFTp:GUS/Col-0 (Adrian et al., 2010) as control. The expression levels of TPS9 were significantly lower in SUC2p:amiRNA-TPS9 than in SUC2p:LUC leaves (Figure 5E), while no significant expression difference between the genotypes was detected in shoot apices (Supplemental Figure S15). Like the tps9 T-DNA insertion mutants, SUC2p:amiRNA-TPS9 plants flowered significantly later than both control lines (Figure 5F). For statistical analysis of flowering time, we measured around 90 independent transgenic T1 lines for each vector, and found that there were about five more leaves in SUC2p:amiRNA-TPS9#1 and #2 lines compared to SUC2p:LUC controls (Figure 5G).

A signature of iACR and up-regulated expression in response to LD in leaf phloem CCs taken together, revealed a critical and tissue-specific role of TPS9 in the timing of flowering.

Transcription factor binding sites of floral regulators are enriched in vascular induced ACRs

The observed predominant increase of chromatin accessibility in response to LD (Figure 3), suggests that transcription factors (TFs) may be recruited to iACRs in LD (Jiang, 2015). We extracted flowering-related TFs from 306 flowering time genes that were collected in the Flowering-Interactive Database (FLOR-ID) (Bouche et al., 2016)), leading to the selection of 10 TF with available ChIP-seq and/or DAP-seq data. Included were six available ChIP-seq (PHYTOCHROME-INTERACTING FACTOR 4 [PIF4], SUPPRESSOR OF OVEREXPRESSION OF CO 1 [SOC1], FLOWERING LOCUS M [FLM], iFLMβ, FLC, and SVP) and six DAP-seq (AGAMOUS-LIKE 15 [AGL15], LATE ELONGATED HYPOCOTYL [LHY], CDF3, VERNALIZATION1 [VRN1], SQUAMOSA PROMOTER BINDING PROTEIN-LIKE 5 [SPL5], and SVP) data sets (Kaufmann et al., 2010; Deng et al., 2011; Immink et al., 2012; Pose et al., 2013; Pfeiffer et al., 2014; Mateos et al., 2015; O'Malley et al., 2016; Collani et al., 2019). We found enrichment for most potential (determined by DAP-seq analysis) and active TF-binding events (determined by ChIP-seq) at iACRs with slightly higher significance scores in categories that included vascular samples. In contrast, TF-binding events were more significantly enriched at rACRs that included Ep tissues (Figure 6A).

Figure 6.

Floral TFs are enriched to regulatory ACRs in LD vasculature. (A) Bubble plots showing enrichment of flowering-related TFs at genes associated with LD induced and LD repressed ACRs. Six ChIP-seq datasets (marked by asterisk) and 6 DAP-seq for 10 floral regulators was included in the analysis. (B) Boxplots showing the number of binding sites of 10 flowering-related TFs at associated ACRs. Boxplots show the median (horizontal line), second to third quartiles (box), and Tukey-style whiskers (beyond the box). P-values were calculated based on Wilcoxon signed-rank tests. (C) Network showing binding interactions of six flowering-related TFs (rhombus) with i/rACR regions of associated with genes that are related to flowering (circle). (D) Genome browser tracks showing ATAC-seq coverage and binding sites of flowering-related TFs for FT and TPS9. Va iACRs were marked by red boxes. Blocks A, C, and E were marked by grey boxes at FT locus.

To further test this overall trend, binding for all flowering-related TFs per ACR was normalized to length in kb and compared to regions of the same size randomly sampled from the Arabidopsis genome. Again, flowering-related TFs enriched preferentially at Va iACRs and Va&Ep iACRs compared to Ep iACR while more enrichment was detected at rACRs that did exclude the Va-specific category (Figure 6B). To evaluate to which extent the observed trend was related to the flowering-related bias in the selection of TFs, we analyzed all available DAP-seq data. Enrichment at responsive ACRs was detected for six TF families (TCP [TEOSINTE BRANCHED1/CYCLOIDEA/PROLIFERATING CELL FACTOR], MYB [myeloblastosis]-related, Homeobox/HB, FAR1 [FAR-RED IMPAIRED RESPONSE1], and CAMTA [calmodulin binding transcription activator]) mostly confirming the trend observed for flowering-related factors, with the notable exception of the CAMTA family, which showed the strongest enrichment for rACRs that included vascular samples and a slight preference for iACRs that included epidermis (Supplemental Figure S16; Supplemental Data Set 5).

To build cell-type-specific floral TF regulatory networks related to photoperiod change and flowering time regulation, we build a graphical map of interactions that included 6 flowering-related TFs and 16 flowering genes from the 306 FLOR-ID database that were associated with LD-responsive ACRs of different categories (Figure 6C). We also included TPS9 in the networks based on our validation of its phloem-specific role in flowering time regulation in LD. As expected from the previous analysis, a higher proportion of genes was connected to Va and Va&Ep iACRs than Ep iACRs, since only 2 out of 17 genes showed Ep-specific connectivity. Only three associations between flowering-related TFs and rACRs occurred (Figure 6C). LHY as a component of the circadian clock is connected to TPS9 through a Va iACRs located at the TSS. At FT, a LHY TFBS locates within the enhancer Block E located downstream of the gene, a second LHY-binding site is located with the distal enhancer Block C, which overlaps with an iACR (Figure 6, C and D). It should be noted that interactions between flowering TFs and flowering-related DEG were not restricted to ACRs or open chromatin regions, that is, CDF3, SVP, FLC, and VRN1 association outside of these regions was detected for FT and TPS9 (Figure 6D).

We also attempted to construct TF gene regulatory networks based on DEG. For flowering-related TFs, enrichment of binding sites was only informative to predict differential expression, if gene-associated ACRs were located at the TSS (Supplemental Figure S17). On the other hand, a tissue-specific network topology was clearly detected when the iACRs were connected to all available DAP-seq TF that was associated to DEG (Supplemental Figure S18). The topology suggests a clear separation of active gene regulatory modules acting in phloem CC and epidermis that does not predominantly involve TF and genes with a confirmed role in flowering time regulation.

Taken together, our results indicate that flowering-related TFs predominantly associate with LD-induced ACRs especially in samples that include phloem CCs, while they are more associated with rACRs in the epidermis. This trend is not restricted to TF not currently associated with a role in flowering time regulation, although the CAMTA TF family followed an opposite trend. Overall, for flowering-related TFs, differential gene expression was less informative than differential ACRs, unless these were located at the TSS.

Discussion

Cell-type-specific responsive ACRs during photoperiod change identified by INTACT-ATAC

Transcriptional regulation and changes in chromatin states are functionally interconnected, but the causality of their interactions is not yet fully understood. In this study, we focused on dissecting how plants differentially sense and translate a change in the day-length signal in two leaf tissues. We isolated vasculature CC and epidermis nuclei from leaves and analyzed chromatin accessibility and transcriptomes in response to LD. We observed a predominant increase in chromatin accessibility upon a change to LD photoperiod that was shared between cell types, but we also detected cell-specific dynamics, indicating that both cell types sense the change in photoperiod and that they engage different gene networks in their response to the signal (Figures 2 and 3).

Phloem CCs are specialized living cells in the leaf vasculature playing functions in ion transport, sugar translocation, and inter-organ communication including the photoperiodic response (Lopez-Salmeron et al., 2019). Several floral activators such as CO, FT, and repressors like TEM1 mainly express in phloem CCs (Castillejo and Pelaz, 2008; Turck et al., 2008; Adrian et al., 2010). Not unexpectedly, CC-specific changes in chromatin accessibility and associated changes in transcription affected genes involved in the regulation of flowering. In contrast, Ep responses were either related to biotic and abiotic stress or photosynthesis genes, for ACRs with, respectively, increased or decreased ATAC-seq signal. Interestingly, the location of responsive chromatin regions was distinct for both cell types: In CCs, an increase in chromatin accessibility is more likely to occur in gene-distal as in gene-proximal regions, while an increase is more often gene-proximal in the epidermis. These differences could be driven by the nature of the affected gene regulatory networks, their overall expression level, or other chromatin features. Indeed, it was recently shown that active and repressive histone marks H3K27me3 and H3K4me3 were mutually antagonistic in phloem cells during a response to photoperiodic change (You et al., 2019). Thus, there may be an extended correlation between the occurrence of distal ACRs and the prevalence of Polycomb Group and Trithorax Group-related epigenetic marks.

Open chromatin regions are preferred sites of active TF binding; in reverse, active transcription factor binding is likely to increase chromatin accessibility; furthermore, increased transcription per se may lead to open chromatin, in particular at proximal ACRs (Klemm et al., 2019). In our study, we found that many iACRs in CCs were bound by TFs related to flowering, based on the re-analysis of ChIP-seq datasets, or at least likely to contain flowering-related TFBS based on available DAP-seq data (Figure 6). This observation suggests that the increase in chromatin accessibility could be a consequence of the induction of flowering related TFs upon LD perception; however, it is also possible that single TFs in this group are responsible for a synchronized opening of ACRs at these genes, promoting the binding of other TFs. A previous study suggested that the binding of AP1 (APETALA1) and SEP3 (SEPALLATA3) MADS-box TFs precedes an increase in chromatin accessibility at their target regions during floral development (Pajoro et al., 2014). In the future, it will be interesting to differentiate between cause and consequence in the relationship between TF binding and chromatin access.

Cell-type-specific photoperiod responsive gene regulatory networks

To further investigate cell-type-specific interactions of TF and ACRs, we compared the association of TF families with cell-type-specific ACRs showing increased or decreased accessibility in response to LD using publicly available DAP-seq and ChIP-seq data (Figure 6; Supplemental Figures S16 and S18). Based on DAP-seq data, TFs of the TCP, MYB-related, Homeobox/HB, FAR1, and CAMTA families were significantly enriched at photoperiod-responsive ACRs (Supplemental Figure S16). Likewise, binding sites of a selection of flowering-related TFs were enriched at LD-responsive ACRs (Figure 6, A and B). Interestingly, most TFs and TF families were more significantly associated with iACRs in the vasculature, but with rACRs in the epidermis. The CAMTA TF family, which is involved in the coordination of stress responses and includes a Ca²⁺-regulated Calmodulin binding domain (Shen et al., 2015), represents an exception to this rule (Supplemental Figure S16). This association could be related to the observation that stress-responsive genes are overrepresented among both, differentially expressed genes and iACR-associated genes in the epidermis (Figure 3C;  Supplemental Figure S11B).

Tissue-specific TF–TF regulatory networks became obvious when transcriptionally upregulated and iACRs-associated TFs were connected to LD-responsive iACRs containing their respective binding sites (Supplemental Figure S18). Similarly, tissue-specific connectivity was visible in gene regulatory networks based on ChIP-seq data of flowering-related TFs, responsive ACRs, and flowering-related target genes (Figure 6C).

A tissue-specific chromatin-to-expression signature to identify novel flowering time regulators

Based on their strong CC-specific chromatin-to-expression signature we identified TPS9 as a novel candidate with a potential role in flowering time regulation and confirmed its role based on a late-flowering phenotype of T-DNA and phloem-specific knock-down mutants (Figures 4 and 5). TPS9 is a homolog of TPS1, which activates FT expression in leaves and promotes the flowering program in SAM (Wahl et al., 2013). A previous study showed that TPS9 is expressed in the leaf vasculature (Ramon et al., 2009). TPS9, which belongs to the clade II of the TPS family, is believed to encode a catalytically nonfunctional trehalose-6-P synthetase, due to a missing catalytic domain. It was suggested that TPS9 plays a role in trehalose monitoring (Ramon et al., 2009). Flowering-related TF bound to the iACR upstream of the TPS9 promoter (Figure 6). Our analysis exemplifies that gene regulatory networks constructed from cell-type-specific chromatin and expression signatures will help to further dissect flowering time regulation.

The FT enhancer Block C has been widely investigated to regulate FT transcription (Adrian et al., 2010; Cao et al., 2014; Liu et al., 2014; Zicola et al., 2019). Our study adds information on the level of cell-type-specific regulation of chromatin accessibility to further deepen our understanding of FT regulation. An increase in chromatin access in response to LD at Block C is observed in the vasculature (Figure 5), which is consistent with increased FT mRNA abundance in CCs upon the shift to LDs (Figures 4 and 5). Nevertheless, ACRs at Block C are detected in all four experimental conditions, including growth in SD, which corresponds to an absence of FT expression. This indicates that Block C is constitutively accessible but shows a further increase in chromatin access in correlation with gene expression. Since Block C is located at considerable distance from the TSS, an increase in chromatin accessibility is more likely a consequence of transcription factor binding than of increased transcription of the locus. Recently, a second enhancer, Block E, participating in LD induction of FT expression was identified downstream of the transcription cassette (Zicola et al., 2019), this enhancer is connected to LHY and CDF3 by an rACR overlapping with Block E (Figure 6). In the future, it will be interesting to investigate how chromatin chances at both enhancers are coordinated during FT induction.

Materials and methods

Plant growth conditions

Two wild-type accessions of A. thaliana, Col-0 and Ler and two INTACT transgenic lines GAS1p:NTF/ACT2p:BirA, and FDHp:NTF/ACT2p:BirA were used in this study. Plants were cultivated on soil in SDs with a photoperiod cycle of 8-h light/22°C and 16-h darkness/20°C in greenhouse. Then half of the 14 SD plants were moved to long-day-greenhouse (16-h light/22°C and 8-h darkness/20°C) and the other half were kept in SD condition. After extra 3LDs or 3SDs treatment, the plants were collected for INTACT nuclei isolation. Mutants of tps9-1 (SALK_086992C) and tps9-2 (SALK_151275C) ordered from NASC were all genotyped as T-DNA homozygous alleles

Plasmid construction

To generate INTACT transgenic line GAS1p:NTF/ACT2p:BirA, GAS1 promoter from entry clone (An et al., 2004) was introduced into GW::NTF-blue destination vector (Franziska Turck group, MPIPZ) by gateway LR reaction (Invitrogen, Carlsbad, CA, USA). GAS1p:NTF was then transformed into ACT2p:BirA Arabidopsis (from Roger Deal Lab, Emory University) mediated by GV3101 agrobacterium. For FDHp:NTF/ACT2p:BirA INTACT transgenic line, FDHp was first introduced into entry clone, and the following steps as same as generating GAS1p:NTF/ACT2p:BirA. For generation of phloem-specific TPS9 knock-down transgenic plants, TPS9-amiRNA#1 and #2 were designed by the amiRNA designer interface WMD3 (Web MicroRNA Designer 3) (http://wmd3.weigelworld.org/cgi-bin/webapp.cgi), integrated into the MIR319a backbone (Liang et al., 2012) and cloned into the binary vector SUC2-pGREEN, which was followed by GV3101 Agrobacterium-mediated transformation and Basta (BBI Life Science, no.77182-82-2) selection. The primers used in this section are listed in Supplemental Table 2.

INTACT nuclei isolation

INTACT nuclei isolation method is modified from previously described protocol (Deal and Henikoff, 2011; Lu et al., 2017). A 0.5-g plant material of INTACT line (Col-0 background) was mixed with 0.1 g Ler plants. The above ground part of seedlings cut by scissors is collected on an ice-cooled 60-mm glass plate with 0.5-mL ice-cooled NPB buffer. Chop the samples with a piece of razor blade to release the nuclei which are flitted by a microcloth. M-280 dyna-beads (Thermo Fisher, Waltham, MA, USA) were washed by NPBt buffer. As long as the crude nuclei are released and filtered, 15-μL Dyna-beads are added to mix well with nuclei, followed by adding 0.1% Triton X-100. After rotating the tubes for 45 min at 4°C, transfer the samples into 15-mL tube containing 13-mL NPBt. Enrich the beads conjured nuclei using 15-mL magnet rack and wash the beads conjured nuclei several times. The DAPI-dyed nuclei are tested under the microscope.

Purity test assay

According to the previous method (Moreno-Romero et al., 2017), Ler as Spike-in material, is mixed with INTACT lines (Col-0 background). Before nuclei purification, an aliquot of suspension was taken as a “Before Sample.” After the last washing step of purification, an aliquot of nuclei sample was taken as an “After Sample.” Use real-time qPCR to measure a Ler specific DNA region, Tα1-2, in “Before Sample” and “After Sample.” ACT8 is measured as the internal control. INTACT nuclei purity is further calculated.

RNA-seq

Total nuclei RNA was extracted from about 100,000 isolated nuclei using the RNeasy Micro Kit (Qiagen, Hilden, Germany; 74004). cDNA was amplified using the Single Cell Full Length mRNA Amplication Kit (Vazyme, Nanjing, China; N712) according to the user manual. The RNA-seq library of each sample was prepared using TruePrep DNA Library Prep Kit version 2 for Illumina (Vazyme; TD501). The library was purified by the VAHTS DNA Clean Beads (Vazyme; N411) and quantified using Qubit and Agilent 2100 before sequencing by the Illumina HiSeq XTen.

ATAC-seq library

Prepare 50,000 fresh isolated beads conjured nuclei for the consequent ATAC-seq tagmentation (Lu et al., 2017). Convert the buffer from NPBt into tagmentation reaction (Vazyme; TD501) using magnet rack. Incubate the beads conjured nuclei 37°C for 30 min (Bajic et al., 2018). Extract DNA by a CTAB (cetyltrimethylammonium bromide) -chloroform-based method and finally store the DNA water solution at −20°C. Follow the Vazyme TD501 manual to set up ATAC-seq library. Twelve cycles are applied to the PCR step. All these steps above, from material collection until DNA extraction must be executive continuously to ensure the chromatin fresh.

Sequencing information

Sequencing of ATAC- and RNA-seq was performed using the Illumina HiSeq XTen instrument at Novogene. ATAC- and RNA-seq were sequenced in paired-end 150 bp. Information on read counts and alignment statistics can be found in Supplemental Table S1.

ATAC-seq raw data processing and alignment

Raw reads were trimmed with Trimmomatic version 0.36 (Bolger et al., 2014). Reads were trimmed for NexteraPE with a maximum of two seed mismatches, palindrome clip threshold of 30, and simple clip threshold of 10. Reads shorter than 30 bp were discarded. Trimmed reads were aligned to the Arabidopsis TAIR10 reference genome using Bowtie2 version 2.11.1 (Langmead and Salzberg, 2012; Langmead et al., 2019) with the following parameters: “bowtie2 -X 1000.” Aligned reads with MAPQ > 10 were sorted using SAMtools version 1.3.1 (Li, 2011; Li et al., 2009 ) and clonal duplicates were removed using Picard version 2.16.0 (http://broadinstitute.github.io/picard/). PCA of ATAC-seq data using “PlotPCA” in DEseq2 (Love et al., 2014; Kim et al., 2019).

RNA-seq raw data processing, alignment, and expression quantification

Raw reads were trimmed with Trimmomatic version 0.36 (Bolger et al., 2014). Reads were trimmed for Trueseq3 barcodes with a maximum of two seed mismatches, palindrome clip threshold of 30, and simple clip threshold of 10. Reads shorter than 50 bp were discarded. The remaining reads were aligned to the same version reference genome with ATAC-seq using HISAT2 with the parameter “–dta.” Raw counts were normalized with DEseq2 to compare gene expression in different samples (Love et al., 2014; Kim et al., 2019). Gene expression values (normalized counts) were calculated by DEseq2. DEGs were also identified with DEseq2 with adjust P <0.05 and fold changes >2.

Identification of ACRs

ACRs for each tissue were identified with MACS2 (Zhang et al., 2008b). Peaks of each biological sample were called with the “–keep-dup all –extsize 200 –shift 100 -g 1.2e8” and overlapping peaks from both replicates were considered as ACRs. To deal with the variant SPOT in different samples, only reads within ACRs were used to normalize data. In general, ACRs from four cell types were merged and Tn5 integration counts in the merged ACRs (TCM) for each sample were calculated. ACR accessibilities in the eight samples were then normalized to 1 M based for TCMs per sample. ACRs were visualized in the Integrative Genomics Viewer version 2.8.2 (Robinson et al., 2011; Thorvaldsdottir et al., 2013).

Identification of differential i/rACRs

To call differential ACRs, the following filtering steps were performed: Generally, (1) differential ACRs were called with MACS2 (Zhang et al., 2008b) with the “–keep-dup all –extsize 200 –shift 100 -g 1.2e8” by pairwise comparison; (2) the differential peaks from both biological replicates were picked up and overlapped with the samples’ ACRs; (3) those differential peaks were further filtered by the Tn5 integration counts which were calculated and normalized based on the strategy detailed above (see “Identification of ACRs” in the above method part); (4) peaks with >1.5-fold changes and “length >50 bp” in pairwise comparison were considered as differential ACRs; and (5) cell-type-specific differential ACRs were then identified by comparing different cell types.

Differential ACRs with higher chromatin accessibilities after changing to LD conditions were labeled as induced (i) ACRs while those with lower accessibilities were labeled as repressed (r) ACRs.

Identification of ACR-targeted genes

The target genes of an ACR were identified based on the following strategy: (1) if an ACR was located within −1,000 to +50 bp of the gene’s TSS, it was labeled as the gene’s “TSS ACR”; (2) remaining ACRs located in the −50 to +1,000 bp of the gene’s transcriptional end site (TES), were labeled as “TES ACR”; (3) remaining ACRs located in genebodies were labeled as “gene body ACR”; and (4) remaining ACRs were labeled as “Intergenic ACRs” and associated with both, the gene located upstream and downstream.

TF–target gene interaction network

To analyze TF–target gene networks for LD-responsive ACRs, we identified TFBS in different categories of ACRs by finding the overlap between ACRs and published ChIP-seq/DAP-seq peaks. We also use DAP-seq datasets for family-wise enrichment at responsive ACRs of each type. Networks were further built after filtering interactions for iACRs and associated target genes by Cytoscape version 3.7.2 (Otasek et al., 2019). Interactions including known flowering-related TF and target genes were extracted to relate responsive ACRs and DEGs.

GO-term analysis

The panther from TAIR (The Arabidopsis Information Resource) was used for GO (PANTHER version 15.0 software) term enrichment (Mi et al., 2019). The cell-type-specific ACRs associated genes as input and generated representative pathways in biological process.

Identification of enriched motifs

To identify whether responsive ACRs were bound by TF, the overlapping between tissue-specific responsive ACRs and TF DAP-seq peaks (O'Malley et al., 2016) were determined by BEDTools intersect (Quinlan and Hall, 2010). For the analysis, random fragment sets from the non-ACRs regions were used as negative control and all the common ACRs from all four samples were used as positive control. To identify FLM, γFLMβ, and VRN1 floral TF motifs, the ChIP- or DAP-seq binding sites were used for analysis using Discriminative Regular Expression Motif Elicitation (Bailey et al., 2009). The rest floral TF were used DAP-seq or JASPAR dataset (http://jaspar.genereg.net/).

Accession numbers

ATAC- and RNA-seq data have been deposited at NCBI’s GEO database (accession number GSE142613). The following secure token has been created to allow review of record GSE142613.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Photoperiodic shift experiments confirm that 3LDs induce early flowering in SD grown Col-0.

Supplemental Figure S2. Tissue-specific marker gene expression in INTACT lines during photoperiodic change.

Supplemental Figure S3. Contamination of ATAC-seq samples with organellar reads.

Supplemental Figure S4. Identification of ACRs.

Supplemental Figure S5. Comparative SPOT analysis of ATAC-seq and DNase-seq datasets.

Supplemental Figure S6. PCA of ATAC-seq data.

Supplemental Figure S7. LD induced and repressed ACRs identified in two biological replicates.

Supplemental Figure S8. Response ACRs in vasculature and epidermis.

Supplemental Figure S9. Representative examples of responsive ACRs.

Supplemental Figure S10. Genomic distribution of responsive ACRs.

Supplemental Figure S11. Cell-type-specific DEGs in vasculature and epidermis.

Supplemental Figure S12. Chromatin accessibility and expression of representative genes showing LD-induction of chromatin access and expression in the vasculature.

Supplemental Figure S13. Identification of TPS9 mutants.

Supplemental Figure S14. Flowering time of tps9 mutants in SDs.

Supplemental Figure S15. TPS9 expression in shoot apices of amiRNA TPS9 knock-down lines.

Supplemental Figure S16. TF families enriched at LD responsive ACRs.

Supplemental Figure S17. The effect of floral TF enrichment at different categories and gene expression.

Supplemental Figure S18. Regulatory modules of TF-target TF in response to photoperiod.

Supplemental Table S1. Sequencing statistics.

Supplemental Table S2. Primer list.

Supplemental Data Set 1. Leaf phloem and epidermis differential ACRs in response to photoperiod.

Supplemental Data Set 2. GO enrichment in biological process of Differential ACRs.

Supplemental Data Set 3. Leaf phloem and epidermis differentially expressed genes in response to photoperiod.

Supplemental Data Set 4. LD-induced candidate genes in vascular phloem cell.

Supplemental Data Set 5. TF family enrichment based on differential ACRs.

 

L.Y.L. designed the experiments. Y.R.L., H.T., and C.W. performed most of the experiments such as ATAC-seq, RNA-seq, gene transformation, and phenotyping. Z.F.L. carried out most of the bioinformatic analysis and H.T. helped to prepare the figures. X.W.X., Y.J.ZH., J.Z., and Q.Z. helped to generate materials and carry out genotyping and phenotype analysis. L.Y.L., Z.F.L., and L.G.L. drafted the manuscript, and F.T. and Y.F.F. gave advice and revised the manuscript.

The author(s) responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plcell/pages/General-Instructions) is: Liangyu Liu (liangyu.liu@cnu.edu.cn).