Gene co-expression network analysis identifies BEH3 as a stabilizer of secondary vascular development in Arabidopsis

Abstract

In plants, vascular stem cells located in the cambium continuously undergo self-renewal and differentiation during secondary growth. Recent advancements in cell sorting techniques have enabled access to the transcriptional regulatory framework of cambial cells. However, mechanisms underlying the robust control of vascular stem cells remain unclear. Here, we identified a new cambium-related regulatory module through co-expression network analysis using multiple transcriptome datasets obtained from an ectopic vascular cell transdifferentiation system using Arabidopsis cotyledons, Vascular cell Induction culture System Using Arabidopsis Leaves (VISUAL). The cambium gene list included a gene encoding the transcription factor BES1/BZR1 Homolog 3 (BEH3), whose homolog BES1 negatively affects vascular stem cell maintenance. Interestingly, null beh3 mutant alleles showed a large variation in their vascular size, indicating that BEH3 functions as a stabilizer of vascular stem cells. Genetic analysis revealed that BEH3 and BES1 perform opposite functions in the regulation of vascular stem cells and the differentiation of vascular cells in the context of the VISUAL system. At the biochemical level, BEH3 showed weak transcriptional repressor activity and functioned antagonistically to other BES/BZR members by competing for binding to the brassinosteroid response element. Furthermore, mathematical modeling suggested that the competitive relationship between BES/BZR homologs leads to the robust regulation of vascular stem cells.

BEH3 contributes to the robust regulation of vascular stem cells via competition among BES/BZR homologs.

Introduction

Continuous stem cell activity relies on a strict balance between cell proliferation and cell differentiation. Plant vascular stem cells located in the cambium continuously undergo cell proliferation and give rise to conductive tissues, including xylem and phloem, as bifacial stem cells (De Rybel et al., 2016; Fischer et al., 2019; Smetana et al., 2019; Shi et al., 2019). Previous genetic studies showed that vascular stem cells are maintained by cell-to-cell communication via the peptide hormone tracheary element differentiation inhibitory factor (TDIF) and its specific receptor PHLOEM INTERCALATED WITH XYLEM (PXY), also named TDIF RECEPTOR (TDR; Ito et al., 2006; Fisher and Turner, 2007; Hirakawa et al., 2008). For the maintenance of vascular stem cells, the TDIF–PXY/TDR signaling module promotes cell proliferation via the transcription factors WUSCHEL-RELATED HOMEOBOX 4 (WOX4) and WOX14 (Hirakawa et al., 2010; Etchells and Turner, 2010; Etchells et al., 2013), and represses cell differentiation by inactivating the other transcription factor BRASSINOSTEROID INSENSITIVE1-EMS-SUPPRESSOR 1 (BES1) through its negative regulator GLYCOGEN SYNTHASE KINASE 3 (GSK3)-like proteins, including BRASSINOSTEROID-INSENSITIVE 2 (BIN2; (Kondo et al., 2014). In parallel, the TDIF–PXY signaling module inactivates another GSK3-like protein, BIN2-LIKE 1 (BIL1), which in turn promotes cambial cell division by repressing negative regulators of cytokinin signaling, including ARABIDOPSIS RESPONSE REGULATOR 7 (ARR7) and ARR15, through the suppression of MONOPTEROS (MP), also named AUXIN RESPONSE FACTOR 5 (Han et al., 2018). Moreover, the TDIF–PXY and EPIDERMAL PATTERNING FACTOR-LIKE (EPFL)–ERECTA (ER) signaling modules coordinate vascular tissue organization (Etchells et al., 2012; Uchida and Tasaka, 2013; Ikematsu et al., 2017). At the genetic level, the TDIF–PXY module interacts with ethylene signaling as a compensatory mechanism (Etchells et al., 2012). Thus, complicated regulatory networks comprising multiple signaling pathways and their crosstalk ensure the precise regulation of vascular stem cells (Kondo and Fukuda, 2015; Fischer et al., 2019).

Recently, omics-based approaches have been utilized to study vascular stem cells. The PXY-mediated transcriptional regulatory network, constructed based on a high-throughput enhanced yeast one hybrid assay, revealed a feed forward loop containing WOX14, TARGET OF MONOPTEROS6 (TMO6), and LOB DOMAIN-CONTAINING PROTEIN (LBD4; Smit et al., 2020). Transcriptome analysis in root cells using fluorescence-activated cell sorting (FACS) with the (pro)cambium marker ARR15pro:GFP, produced a list of cambium-expressed transcription factor-encoding genes (Zhang et al. 2019), including not only well-known cambium genes, such as AINTEGUMENTA (ANT; Randall et al., 2015; Smetana et al., 2019), KNOTTED-like from Arabidopsis thaliana 1/BREVIPEDICELLUS (KNAT1/BP; Venglat et al., 2002; Liebsch et al. 2014), and LBD4 (Smit et al., 2020), but also new cambium regulators, such as SHORT VEGETATIVE PHASE (SVP), PETAL LOSS (PTL), RESPONSE TO ABA AND SALT 1 (RAS1), and MYB DOMAIN 87 (MYB87). Furthermore, transcriptome analysis of the inflorescence stem using fluorescence-activated nuclear sorting (FANS) with tissue-specific promoters such as the PXY promoter (specific for proximal cambial cells) and the SUPPRESSOR OF MAX2 1-LIKE PROTEIN 5 (SMXL5) promoter (specific for distal cambial cells) provided informative vascular tissue-specific gene expression profiles (Shi et al., 2021). Nevertheless, the molecular mechanism underlying the robust control of vascular stem cells remains unclear.

The vascular system consists of various types of vascular cells at different developmental stages. To understand the complex vascular developmental program, we previously established a tissue culture system, Vascular Cell Induction Culture System Using Arabidopsis Leaves (VISUAL), which can synchronously and efficiently induce the transdifferentiation of mesophyll cells into xylem and phloem cells from vascular stem cells (Kondo et al., 2016). The loss-of-function mutant of the phloem regulator gene ALTERED PHLOEM DEVELOPMENT (APL) lacks differentiation of phloem sieve elements (Bonke et al. 2003). In addition to the microarray data of apl mutants (Kondo et al., 2016), the FACS datasets generated using a phloem marker gene, SIEVE ELEMENT OCCLUSION RELATED 1 (SEOR1; Froelich et al., 2011; Kondo et al. 2016), in VISUAL enabled the construction of a co-expression gene network covering the early-to-late phloem sieve element differentiation process (Kondo et al., 2016). Although such a network analysis is useful for a temporal dissection of the vascular differentiation process, this gene network is still only limited to phloem sieve element differentiation. Recently, we discovered that the loss-of-function bes1-1 mutant dramatically impairs the differentiation of xylem and phloem cells, resulting in the accumulation of vascular stem cells in VISUAL (Kondo et al., 2014, 2015; Saito et al., 2018).

In this study, we integrated the previously generated bes1-1 transcriptome dataset (Saito et al., 2018) with several time course transcriptome datasets covering 72 h of vascular differentiation induction generated here. We established a co-expression network of the VISUAL differentiation process utilizing these multiple transcriptome datasets, and identified gene modules corresponding not only to phloem but also xylem and cambial cells. We validated these classifications based on a comparison with the pre-existing vascular FACS and FANS datasets. Through network analysis, we identified BES1/BZR1 HOMOLOG 3 (BEH3) as a potential vascular stem cell regulator. The beh3-1 null mutant exhibited high variation in vascular size among individual samples, suggesting that BEH3 functions as a stabilizer of vascular stem cells. Further genetic and molecular analyses revealed that BEH3 antagonizes other BES/BZR homologs by competitively inhibiting their binding to the brassinosteroid (BR) response element (BRRE) in target gene promoters. Moreover, mathematical modeling analysis showed that the competitive relationship between BEH3 and other BES/BZR homologs enables the robust control of vascular stem cells.

Results

Co-expression network analysis using multiple VISUAL transcriptome data sets

To investigate the temporal progression of vascular cell differentiation with VISUAL (Kondo et al., 2015, 2016), we determined the expression profiles of vascular cell marker genes at 6-h intervals by reverse transcription quantitative polymerase chain reaction (RT-qPCR) and microarray analysis over a 72-h period from time of induction. The expression of the cambium marker genes TDR and HOMEOBOX GENE 8 (HB-8; Baima et al., 1995, 2001) was induced, followed by the phloem marker gene SEOR1 and the xylem marker genes VASCULAR-RELATED NAC-DOMAIN 5 (VND5; Zhou et al., 2014), VND6 (Kubo et al., 2005), and IRREGULAR XYLEM 3 (IRX3; Taylor et al., 1999; Supplemental Figure S1). Since these vasculature-related genes were highly expressed at 72 h, we selected 855 VISUAL-induced genes that are highly upregulated (more than eight-fold at 0 h versus 72 h after VISUAL induction) from the time course microarray data set (Supplemental Data Set S1). To identify new regulators of vascular stem cells, we conducted a gene network analysis using multiple VISUAL transcriptome datasets: (1) the 6-h interval time course dataset generated in this study (Supplemental Data Set S1); (2) the FACS dataset generated previously with the phloem marker SEOR1pro:SEOR1-YFP (YFP-positive rep1–3 and YFP-negative rep1–3; Kondo et al., 2016); (3) phloem-deficient apl mutants (wild-type [WT] rep1–3 and apl rep1–3; Kondo et al., 2016); and (4) bes1-1 mutants (WT 72-h rep1–3 and bes1-1 72-h rep1–3; Saito et al., 2018; Figure 1A). Since bes1-1 mutants accumulate cambial cells because of the inhibition of xylem and phloem differentiation in VISUAL (Saito et al., 2018), the transcriptome datasets of these mutants were also effective in dissecting cambial cells and differentiated xylem/phloem cells (Figure 1A). We performed co-expression network analysis using the weighted gene co-expression network analysis (WGCNA) R package (Langfelder and Horvath, 2008). In this network, 855 VISUAL-induced genes were divided into 4 distinct modules (modules I–IV) and 11 outgroup genes (others, O; Figure 1B).

Figure 1.

Classification of genes induced in the VISUAL, based on gene co-expression network analysis. A, Schematic representation of the multiple VISUAL transcriptome datasets used for gene co-expression network analysis, including the 6-h interval time course (this study), cell-sorting with the phloem marker SEOR1pro:SEOR1-YFP (SEOR1-positive; Kondo et al., 2016), the phloem-defective apl mutant (Kondo et al., 2016), and the bes1-1 mutant (Saito et al., 2018). + and – indicate the presence (+) or absence (–) of xylem cells (upper) and phloem cells (lower) in the WT (Col-0) and each mutant (apl and bes1-1) after VISUAL induction. B, Co-expression network of VISUAL-induced genes using the WGCNA package. Four distinct modules are highlighted with different colors. A node represents a gene. An edge indicates high correlation between nodes (topology overlap matrix [TOM] >0.02). C, Expression profiles of VISUAL-induced genes in the VISUAL 6-h interval time course microarray data. The induced genes showed more than eight-fold higher expression at 72 h than at 0 h. Expression is depicted as a heatmap, with the color scale representing the Z-score of expression values. Color bars on the right indicate modules corresponding to (B). D, Expression profiles of VISUAL-induced genes across the multiple VISUAL microarray data sets. VISUAL microarray data sets were obtained from comparisons between WT versus bes1-1 at 72 h (Saito et al., 2018), WT versus apl (Kondo et al. 2016), and SEOR1-negative versus SEOR1-positive cells (Kondo et al. 2016).

Modules III and IV represent xylem- and phloem-related genes, respectively

According to the 6-h interval time course dataset, genes from modules III to IV were expressed later than those of module I or II (Figure 1C). The expression of modules III and IV genes was downregulated in the bes1-1 mutant (Figure 1D), suggesting that genes in these modules are related to xylem and phloem cells, respectively. In addition, the expression of module-III genes was high in the SEOR1-positive FACS datasets and downregulated in apl mutants, whereas that of module-IV genes were unchanged in the FACS and apl mutant datasets (Figure 1D), suggesting that modules III and IV represent phloem- and xylem-related genes, respectively. Consistent with this idea, the known phloem-related genes RAB GTPASE HOMOLOG C2A (RABC2A; Kondo et al., 2016), BREVIS RADIX (BRX; Depuydt et al., 2013), GLUCAN SYNTHASE-LIKE 7 (GSL07; Barratt et al., 2011), and NAC DOMAIN-CONTAINING PROTEIN 45 (NAC045; Furuta et al., 2014) were part of module-III, while xylem-related genes such as IRX3, XYLEM CYSTEINE PEPTIDASE 1 (XCP1; Zhao et al., 2000; Zhong et al., 2010; Yamaguchi et al., 2011), VND5, and VND6 were grouped in module IV (Figure 1B;  Supplemental Data Set S1). To validate these results, we investigated the expression profiles of genes in each module using the root-FACS and stem-FANS datasets obtained with cell type-specific markers (Brady et al., 2007; Zhang et al., 2019; Shi et al., 2021). The expression of module III genes was quite high in the phloem-specific APLpro population from both FACS and FANS (Supplemental Figure S2, A and B; Brady et al., 2007; Shi et al., 2021), while module IV genes were highly expressed in the xylem-specific S18 (MYB46pro) population in FACS (Figure 1E;  Supplemental Figure S2, A and B; Brady et al., 2007) and VND7pro datasets in FANS (Supplemental Figure S3, A and B; Shi et al., 2021). Genes with high Z-scores (Supplemental Figure 2B, Z-score >1) in the phloem-specific APLpro and xylem-specific S18 populations were significantly enriched in modules III and IV, respectively (Figure 2A).

Figure 2.

Cross-validation of gene co-expression network analysis and expression profiles of cambium-related transcription factor genes. A, Enrichment analysis for high Z-score genes in the FACS datasets (Brady et al., 2007; Zhang et al., 2019). Among the 855 VISUAL-induced genes, the Z-scores of 714 genes could be calculated. The percentage of high Z-score genes (Z-score >1) to the total number of genes in each module and all VISUAL-induced genes was calculated and presented in the upper half of each block. The percentage was also visualized as a heatmap, from white to dark orange. The P-value of enrichment analysis (Fisher’s exact test) of the number of high Z-score genes is given in each block. Significantly enriched genes (P < 1 × 10⁻⁰⁵) are indicated in red. B, Heatmap representation of expression profiles of cambium-related genes (Module II) in the FANS dataset (Shi et al., 2021). Gene names are shown on the right side. Clades were defined by cluster analysis, based on gene expression profiles shown in Supplemental Figure 3C.

Module-II contains vascular cambium-related genes

Compared to modules III and IV genes, the expression of modules I and II genes was induced earlier and not reduced in the bes1-1 mutant (Figure 1, C and D), suggesting that modules I and II genes correspond to (pro)cambium-related genes. Genes in module I were relatively highly expressed in the procambium (regions ARRIII ± BAP), while the expression of module II genes was more specific to the developing cambium and cambium (regions ARRI ± BAP and ARRII ± BAP), based on the transcriptome analysis of ARR15-expressing cells purified by FACS (Supplemental Figure S2A; Zhang et al., 2019). Indeed, high Z-score genes (Supplemental Figure S2B, Z-score >1) in the procambium (region ARRIII + BAP) and cambium (region ARRI + BAP) were statistically enriched in modules I and II genes, respectively (Figure 2A). In addition, previously reported cambium-enriched genes, which were identified as genes enriched in ARR15-expressing cells (Zhang et al., 2019), were the most enriched in module II (Supplemental Figure S2C). These results suggest that modules I and II represent procambium- and cambium-related genes, respectively. Consistent with this idea, module I genes were expressed much earlier than module II genes (Figure 1C).

We then focused on module II genes to isolate new stem cell regulators. Module II consisted of 394 genes, including several known cambium-related genes such as PXY, SMXL5, ANT, HB-8, LBD3 (Smit et al., 2020), LBD4 (Zhang et al., 2019; Smit et al., 2020), SCARECROW-LIKE 28 (SCL28; Zhang et al., 2019), and MYB3R-4 (Zhang et al., 2019; Figure 1B;  Supplemental Data Set S1). Among these genes, we investigated 32 transcription factor-encoding genes in more detail (Figure 2B;  Supplemental Data Set S2). Of these 32 transcription factors, 13 overlapped with the transcription factors from the PXY-mediated transcriptional regulatory network (Smit et al., 2020). Previous studies have revealed that vascular stem cells are located in the overlapping expression zone of PXY and SMXL5, which mark the proximal/xylem and distal/phloem side of cambium, respectively (Shi et al. 2019). According to the expression pattern in the FANS transcriptome datasets (Shi et al., 2021), we classified these 32 transcription factor genes from module II into four clades (Figure 2B; Supplemental Figure S3C): 13 belonging to clades B and C were expressed mainly in PXY- and/or SMXL5-positive cells (Figure 2B). These genes included several known cambial regulators such as ANT, LBD3, LBD4, HB-8, and SCL28 (Figure 2B). Within clade B, we discovered BEH3 as one of the new cambial cell regulator candidates (Figure 2B;  Supplemental Data Set S2). Notably, BEH3 was highly expressed on the proximal/xylem side of cambium (PXY-positive cells) in the FANS dataset (Shi et al., 2021; Supplemental Figure S3D). Therefore, we selected BEH3 as a potential vascular stem cell-specific regulator for further analysis.

BEH3 promotes and stabilizes secondary vascular development

BEH3 encodes a member of the BES/BZR family of transcription factors (Yin et al., 2005). We recently showed that BES1 and its closest homolog BRASSINAZOLE RESISTANT1 (BZR1) redundantly promote vascular stem cell differentiation in VISUAL (Saito et al., 2018). The Arabidopsis genome encodes six BES/BZR homologs (Figure 3A). While BEH3 was highly induced in VISUAL, we detected expression of BES1, BZR1, and BEH4 to certain levels but with weaker induction than that of BEH3 (Figure 3B). In contrast, the expression of BEH1 and BEH2 rapidly decreased in VISUAL (Figure 3B). Consistently, the expression domain of BEH3pro:GUS gradually expanded over time in VISUAL (Figure 3C). In vivo expression analysis revealed that all BES/BZR family genes are expressed in the procambium and cambium (ARRI-III + BAP) in the FACS dataset, and in PXY-positive and/or SMXL5-positive cells in the FANS dataset (Brady et al., 2007; Zhang et al., 2019; Shi et al., 2021; Supplemental Figure S4). Indeed, the reporter construct BEH3pro:GUS showed preferential staining in the vascular tissues of leaves and hypocotyls (Figure 3D). In cross-sections of hypocotyls, we observed BEH3pro:GUS staining in the vascular area including cambium (Figure 3E). Together, these results suggest that all BES/BZR family genes are expressed in the vasculature, and that the BEH3 expression domain is much more specific to vascular development.

Figure 3.

BEH3 is preferentially expressed in vascular tissues. A, An unrooted phylogenetic tree of Arabidopsis BES/BZR family members. B, Time course expression analysis of BES/BZR family genes in VISUAL. Expression values at each time point based on 6-h interval time course microarray data are indicated. C–E, GUS activity in different tissues of BEH3pro:GUS transgenic lines: cotyledons subjected to VISUAL (C), shoots of a 20-day-old seedling (D), and hypocotyl cross-section of a 7-day-old seedling (E). Scale bars = 1 mm (C), 2 mm (D), and 100 µm (E).

Next, to investigate the function of BEH3 in vascular development, we analyzed the beh3-1 null mutant, which carries a T-DNA insertion in the 1st exon of the BEH3 locus (Lachowiec et al., 2018). We compared the secondary growth of vascular tissues between the WT (Col-0) and the beh3-1 mutant grown under a short-day photoperiod. The area within the cambium was slightly but significantly lower in the beh3-1 mutant than in the WT (Figure 4, A and B;  Supplemental Figure S5). In addition, the beh3-1 mutant appeared to exhibit greater variation in the area within the cambium than WT (Figure 4C). This observation was further verified by the values of the coefficient of variation (CV), which differed significantly between the beh3-1 mutant and WT, as shown by two statistical analyses: asymptotic test and modified signed-likelihood ratio test (Feltz and Miller, 1996; Krishnamoorthy and Lee, 2014; Figure 4D). These results suggest that BEH3 not only promotes but also stabilizes secondary vascular development.

Figure 4.

BEH3 regulates secondary vascular development. A, Hypocotyl cross-section of Col-0 and beh3-1 mutant plants grown under short-day conditions for 21 days. B, Box-and-whisker plot of relative areas within the cambium is shown in (A). Relative values were calculated in comparison to the mean value from Col-0. The upper and lower limits indicate the first and third interquartile ranges, respectively, and the middle bar indicates the median. Whiskers indicate 1.5× the interquartile ranges. The black dot indicates an outlier. Gray dots indicate each sample (n = 27 seedlings [Col-0] and 29 seedlings [beh3-1]). Significant differences were examined by Student’s t test (**P < 0.01). C, Variation in cambium size between Col-0 and beh3 plants. Gray dots indicate area within the cambium in each individual from (B) when ordered by increasing size. D, CV of areas within the cambium in Col-0 and beh3 plants shown in (B). Statistical differences between Col-0 and beh3 plants were examined using the asymptotic test and modified signed-likelihood ratio test (M-SLRT; **P < 0.01). E, Cross-sections of hypocotyl vasculature of 11-day-old Col-0 and bes1-D seedlings. Blue and green areas indicate xylem and phloem cells, respectively. F, Box-and-whisker plot of the number of (pro)cambial cell layers of the samples is shown in (E). Cell layers were calculated as the average number of minimal cell layers between xylem and phloem regions in two phloem poles. Gray dots indicate values in individual seedlings (n = 10 [Col-0] and 12 [bes1-D]). Significant differences were examined by Student’s t test (*P < 0.05, **P < 0.01). G, Cross-sections of the hypocotyls of transgenic plants expressing BEH3 under the control of the β-estradiol-inducible promoter. Transgenic plants were treated with (+) or without (–) 5-µM β-estradiol for 12 days. H, Box-and-whisker plot of the number of (pro)cambial cell layers in samples is shown in (G). The number of cell layers was calculated as the average number of minimal cell layers between xylem and phloem in two phloem poles. I, Low-magnification images of (G). J, Box-and-whisker plot of the area of the whole vascular region in samples is shown in (I). K, Box-and-whisker plot of areas within the cambium in samples is shown in (I). Relative values were calculated in comparison to the mean value of the β-estradiol-untreated samples. In (H, J, and K), gray dots indicate values from individual seedlings (n = 23). Significant differences were examined by Student’s t test (*P < 0.05, **P < 0.01). In (A, E, F, G, and I), scale bars = 100 µm.

BEH3 antagonizes BES1 during vascular development

Consistent with previous reports (Kondo et al., 2014; Saito et al., 2018), the bes1-D gain-of-function mutant, which harbors a mutation in the PEST sequence involved in protein degradation, exhibited a reduced number of (pro)cambial cell layers between the differentiated xylem and phloem tissues in hypocotyls relative to the WT (Figure 4, E and F). To compare the function of BEH3 and BES1, we analyzed the vascular phenotype of bes1-D. Since BEH3 does not possess an obvious PEST domain (Supplemental Figure S6), we generated a stable transgenic line harboring a β-estradiol-inducible BEH3 overexpressor (pER8:BEH3-cyan fluorescent protein [CFP]). In contrast to the bes1-D gain-of-function mutant (Figure 4, E and F), two independent BEH3 overexpression lines showed an increase in the number of (pro)cambium layers specifically in the presence of β-estradiol (Figure 4, G and H;  Supplemental Figure S7, A and B). In association with this result, the entire vascular area and area within the cambium also increased in these BEH3 inducible lines (Figure 4, I–K; Supplemental Figure S7C) in contrast to the beh3-1 mutant. These results suggest that BEH3 performs a function opposite to that of BES1 during secondary vascular development. BES/BZR homologs are also known as key transcription factors in BR signaling (Yin et al., 2005; Nolan et al., 2020). Although the loss-of-function mutant of the BR receptor bri1 exhibits a severe reduction in hypocotyl length in etiolated seedlings, the introduction of the gain-of-function mutant bes1-D rescues the bri1 phenotype (Yin et al., 2002). Overexpression of BEH3 upon β-estradiol treatment shortened hypocotyl length in the dark (Supplemental Figure S8), consistent with a recent report (Nguyen et al., 2021). These results suggest that the effects of BEH3 overexpression are opposite to those of BES1 gain-of-function mutation not only for vascular cambium regulation but also for BR signaling.

We showed that differentiation of xylem and phloem cells from vascular stem cells is disrupted in bes1-1 mutants in VISUAL (Figure 1A). To investigate the function of BEH3 and other BES/BZR homologs, we evaluated the contribution of each BES/BZR homolog to xylem differentiation with a semi-automatic calculation method using the xylem indicator BF-170 (Nurani et al., 2020) in VISUAL (Figure 5, A and B;  Supplemental Figure S9). Compared to the WT, single mutants including beh4-1 as well as bes1-1 and bzr1-1 showed a reduced xylem differentiation percentage in VISUAL (Figure 5, B–D; Supplemental Figure S10), indicating that BEH4 also promotes xylem differentiation. However, xylem differentiation in beh3-1, beh1-1, and beh2-5 single mutants was comparable to that in WT (Figure 5, B–D; Supplemental Figure S10). Next, we examined the consequence of inactivating a BES1 homolog in the bes1-1 mutant background for xylem differentiation. Only the bes1-1 beh3-1 double mutant showed high xylem differentiation, similar to that observed in the beh-3 single mutant and the WT; however, the reduced xylem differentiation seen in bes1-1 was not rescued in any of the other double mutants (Figure 5, B–D; Supplemental Figures S10 and S11). The differentiation from vascular stem cells into not only xylem but also phloem cells was disrupted in bes1 mutants, as shown previously (Saito et al., 2018). These phenotypic changes in bes1 mutants were accompanied by the downregulation of xylem-specific (IRX3) and phloem-specific (SEOR1) marker genes (Figure 5E); however, this reduction in gene expression recovered to WT levels in the bes1-1 beh3-1 double mutant (Figure 5E). RT-qPCR analysis also indicated that the single beh3-1 mutant shows slightly higher levels of IRX3 and SEOR1 transcripts than the WT at 48 h (Figure 5E;  Supplemental Table S1). The results of these genetic analyses suggest that BEH3 functions oppositely to BES1 in vivo and in VISUAL.

Figure 5.

BEH3 functions antagonistically to BES1 and BZR1 in VISUAL. A, Schematic representation of the transdifferentiation process in Col-0 and the bes1-1 mutant. The bes1-1 mutant failed to induce xylem and phloem differentiation in VISUAL, resulting in the accumulation of cambial cells. + and – indicate the presence (+) or absence (–) of xylem cells (upper) and phloem cells (lower) in the Col-0 and bes1-1 mutant after the VISUAL induction. B and C, Staining of ectopic xylem cells (B) and quantification of xylem cell differentiation (C) in bes/bzr single and double mutants. Bright blue signal in (B) indicates xylem cell walls stained with the lignin probe BF-170. Box-and-whisker plots in (C) show the quantification of the xylem differentiation percentage. Gray dots indicate the xylem differentiation percentage for each cotyledon (n = 7–15). Different lowercase letters indicate significant differences (P < 0.05; Tukey–Kramer test). D, Roles of BES/BZR family members in VISUAL vascular cell differentiation. Red, blue, and gray indicate “promotive”, “suppressive”, and “no effect”, respectively. E, Time course expression analysis of xylem-specific (IRX3) and phloem-specific (SEOR1) marker genes. Relative fold-changes were calculated in comparison to Col-0 at time 0 h. Data represent mean ± standard deviation (sd; n = 3; biological replicates). Results of two-way ANOVA are summarized in Supplemental Table S1. Different lowercase letters indicate significant differences at 48 h (P < 0.05; two-way ANOVA). F and G, Staining of ectopic xylem cells (F) and quantification of xylem differentiation (G) in bes/bzr triple mutants. Bright blue signal in (F) indicates xylem cell walls stained with BF-170. Box-and-whisker plots in (G) show the xylem differentiation percentage. Gray dots indicate the xylem differentiation percentage for each cotyledon (n = 8–21). Different lowercase letters indicate significant differences (P < 0.05; Tukey–Kramer test). H, Relative transcript levels of IRX3 and SEOR1 in bes/bzr triple mutants 48 h after induction. Relative fold-changes were calculated in comparison to Col-0 at time 0 h (n = 3; biological replicates). Different lowercase letters indicate significant differences (P < 0.05; Tukey–Kramer test). Scale bars = 1 cm (B and F).

BEH3 competes with other BES/BZR homologs

To unravel the molecular basis of the antagonistic function of BEH3 and BES1, we examined the functional difference between BEH3 and the other BES/BZR proteins. Amino acid sequence alignments suggested no obvious difference in known functional domains/motifs, such as the DNA-binding domain (DBD) or EAR motif (Yin et al., 2002; Wang et al., 2002; Oh et al., 2014; Supplemental Figure S6). Both BES1 and BZR1 are known to repress the transcription of the BR biosynthetic gene DWARF4 (DWF4) by directly binding to the BRRE and/or E-box motif in its promoter (He et al., 2005; Sun et al., 2010; Yu et al., 2011; Chung et al., 2011; Nosaki et al., 2018). Therefore, we tested the transcriptional control activity of BES/BZR members using a DWF4pro:Emerald Luciferase (ELUC) reporter construct transiently co-infiltrated with constructs expressing fusions between BES/BZR members and the rat glucocorticoid receptor (GR) in Nicotiana benthamiana leaves (Figure 6A). ELUC activity decreased gradually from 3 to 9 h after the activation of BES1, BZR1, and BEH4 by the addition of dexamethasone (DEX; Figure 6B, Supplemental Table S2; Aoyama et al., 1995), whereas BEH3 activation did not affect ELUC activity (Figure 6B;  Supplemental Table S2). Electrophoretic mobility shift assays (EMSAs) using the BRRE-core motif and E-box motif with the DBD of BES/BZR members indicated that BEH3 can bind to the BRRE and E-box (Figure 6C;  Supplemental Figure S12), as can BES1, BZR1, and BEH4 (Nosaki et al., 2018). These results suggest that BEH3 binds to the BRRE and E-box motif, but exhibits much lower transcriptional repressor activity than the other BES/BZR homologs. When BEH3 and BES1 were co-expressed in N. benthamiana leaves, the repressor activity of BES1 on DWF4pro:ELUC was weakened (Figure 6D;  Supplemental Table S3). In contrast, the co-expression of BZR1 with BES1 did not show such a combinatorial repressive effect (Figure 6D;  Supplemental Table S3), suggesting the possibility that BEH3 interferes with the function of the other BES/BZR homologs by competing for their target gene promoters.

Figure 6.

BEH3 competes with other BES/BZR homologs. A, Schematic representation of the constructs used to examine the transcriptional repressor activity of BES/BZR homologs. B, Transcriptional repressor activity of BES/BZR proteins in N. benthamiana transient expression assay. Luciferase (LUC) activity from N. benthamiana leaves infiltrated with the DWF4pro:ELUC construct was calculated as the average photon counts per second of six leaf disks and was normalized relative to the values of DEX-treated control samples without effector constructs. All values at 3 h were set to 1. The results of two replicates are shown as first and second. C, EMSA. The DNA-binding ability of BES/BZR proteins was tested against the G-box, BRRE-core, and mBRRE-core elements. Asterisks and arrowheads indicate the positions of free probe and DBD-containing protein-bound probe, respectively. D, Combinatory effects of BES/BZR homologs on transcriptional repressor activity in N. benthamiana transient expression assay. BZR1 or BEH3 was co-expressed with BES1 and the DWF4pro:ELUC reporter in N. benthamiana. LUC activity was calculated, normalized, and plotted as in (B). E, Relative transcript levels of BES/BZR-target genes in the estradiol-inducible BEH3 overexpression line. Eight-day-old seedlings were treated with 10-µM β-estradiol for 9 h. Relative expression levels were calculated relative to DMSO-treated (control) samples. Circles indicate each replicate (n = 3; biological replicates). Significant differences were examined by Student’s t test (**P < 0.01). F, Relative transcript levels of BES/BZR-target genes in 9-day-old bes1-D bzr1-D double mutant seedlings. Relative expression levels were calculated in comparison to Col-0. Circles indicate each replicate (n = 3; biological replicates). Significant differences were examined by Student’s t test (**P < 0.01).

In VISUAL, the bes1-1 mutant did not affect BEH3 expression, and neither did the beh3-1 mutant on BES1 expression (Supplemental Figure S13). To investigate the competitive genetic interaction between BEH3 and the other BES/BZR homologs, we analyzed several combinations of triple mutants in VISUAL. Although the beh3-1 mutant completely rescued the phenotype of the bes1-1 single mutant (Figure 5, B and C), restoration of xylem differentiation by the beh3-1 mutant was only partial in the bes1-1 bzr1-2 double mutant background (Figure 5, F and G). We did not detect a similar partial rescue in the bes1-1 bzr1-2 beh4-1 triple mutant (Figure 5, F and G). Consistent with the xylem differentiation percentage, the transcript levels of xylem- and phloem-specific marker genes in the bes1-1 bzr1-2 beh3-1 triple mutant were intermediate between those of the WT and the bes1-1 bzr1-2 double mutant (Figure 5H). Taken together, these results suggest that BEH3 acts antagonistically to other homologs in a competitive manner in VISUAL.

BEH3 passively inhibits the repressor activity of other BES/BZR homologs

Next, we investigated whether BEH3 affects the expression of known BES/BZR-target genes in planta using pER8:BEH3-CFP transgenic lines, in which expression of BEH3 was increased upon β-estradiol treatment (Figure 6E;  Supplemental Figure S14 and Supplemental Table S4). The expression of DWF4 and BRASSINOSTEROID-6-OXIDASE 2 (BR6ox2), both direct targets of BES1-induced repression (He et al., 2005; Sun et al., 2010; Yu et al., 2011), was remarkably repressed in the constitutively active double mutant bes1-D bzr1-D (Figure 6F) but was significantly increased in pER8:BEH3-CFP transgenic plants upon β-estradiol treatment (Figure 6E;  Supplemental Figure S14 and Supplemental Table S4). BES1 and BZR1 function as both transcriptional activators and repressors (Yin et al., 2005; He et al., 2005; Nolan et al., 2020). However, BEH3 overexpression did not reduce the mRNA levels of genes activated directly by BES1, including SMALL AUXIN UPREGULATED 15 (SAUR15) or INDOLE-3-ACETIC ACID INDUCIBLE 19 (IAA19; Yin et al., 2005; Sun et al., 2010; Oh et al., 2012; Figure 6E), whose expression was highly induced in the bes1-D bzr1-D double mutant (Figure 6F). These results suggest that the noncanonical BEH3 passively inhibits the transcriptional repressor activity of endogenous BES1 homologs by competing for their common binding motif.

BEH3 partially compensates for the loss of other BES homologs

In Arabidopsis, six BES/BZR transcription factors perform redundant functions in BR signaling (Yin et al., 2005; Lachowiec et al., 2018; Chen et al., 2019a, 2019b; Nolan et al., 2020). Previous studies showed that the bes1 hextuple mutant, carrying null alleles in all six BES/BZR genes, exhibits male sterility (Chen et al., 2019a). We hypothesized that if BEH3 constitutively competes with the other BES/BZR homologs, then the bes1 bzr1 beh1 beh2 beh4 quintuple mutant might show the same phenotype as the bes1 bzr1 beh1 beh2 beh3 beh4 hextuple mutant. To test this hypothesis, we generated a bes1 hextuple mutant (bes1-1 bzr1-2 beh1-2 beh2-5 beh3-3 beh4-1) and quintuple mutant (bes1-1 bzr1-2 beh1-2 beh2-5 beh4-1) using the CRISPR/Cas9 system (Supplemental Figure S15). Consistent with previous reports, the newly established bes1 hextuple mutant produced shorter siliques with no seed compared to the WT (Figure 7, A and B). However, the bes1 quintuple mutant successfully produced seeds in siliques (Figure 7, A and B). In addition, the primary root length of the bes1 hextuple mutant was shorter than that of the WT, whereas the primary root length of the bes1 quintuple mutant was comparable to that of the WT (Figure 7, C and D). We observed similar trends in secondary vascular development. Relative to the WT, the entire vascular area was greater in the bes1 hextuple mutant but comparable in the bes1 quintuple mutant (Figure 7, E and F). These results suggest that BEH3 partially shares functions with other BES/BZR homologs in the absence of other members.

Figure 7.

Genetic analysis of bes/bzr quintuple and hextuple mutants. A and B, Development of fruits (A) and seeds (B) in bes1 quintuple (bes1-q) and hextuple (bes1-h) mutants. Box-and-whisker plots in (B) indicate the number of seeds in each mature fruit. Gray dots indicate values of each fruit (n = 10 fruits per genotype). Different lowercase letters indicate significant differences (P < 0.05; Tukey–Kramer test). C and D, Primary root length of 11-day-old bes1-q and bes1-h mutant seedlings. To obtain homozygous bes1-h mutant plants, self-fertilized progenies of T₂ heterozygous bes1-q mutants, which were hemizygous for the beh3 mutation, were used. Gray dots in (D) indicate the values of each seedling (n = 10 [Col-0], 12 [bes1-q], and 13 [bes1-h]). E, Hypocotyl cross-sections of 11-day-old Col-0, bes1-q, and bes1-h mutant plants. F, Box-and-whisker plot of the entire vascular area of samples is shown in (E). Gray dots indicate values of each seedling (n = 9 [Col-0], 10 [bes1-q], and 6 [bes1-h]). Different lowercase letters indicate significant differences (P < 0.05; Tukey–Kramer test). Scale bars = 1 mm (A and C) and 100 µm (E).

BEH3 increases vascular robustness via competition with other homologs

The beh3-1 mutant exhibited higher variation in vascular size than the WT (Figure 4, C and D). Based on our results, we hypothesized that the competitive relationship between BEH3 and other BES/BZR homologs might contribute to stabilizing vascular size. To test this hypothesis on the basis of the molecular activity of BEH3, we constructed a mathematical model for competitive binding, with a focus on transcriptional repressor activity. Because BES1 functions to reduce vascular area, we postulated that BES/BZR homologs suppress the transcript levels of direct target genes, leading to the repression of cambial cell proliferation. Here, we assumed a total of three binding sites (n = 3) in the putative promoter and predicted the output as vascular area, when varying the promoter-binding rate of BEH3 (P_BEH3) and other homologs (P_O; Figure 8A). To set up the model, we first examined the effect of the binding of BEH3 ($\alpha$) and that of other homologs ($\beta$) on the output, when they occupied one binding site in the target gene promoter. When assuming the same transcriptional activity of BEH3 and other homologs ($\alpha =0.8$ and $\beta =0.8)$, the average predicted output value decreased gradually with the increase in P_BEH3 (Figure 8B), which was inconsistent with the enlargement of the vascular area upon BEH3 overexpression (Figure 4, G–K). Next, when assuming no transcriptional repressor activity of BEH3 compared to the other homologs ($\alpha =1$ and $\beta =0.8)$, the average output value gradually increased with the increase in P_BEH3 (Figure 8C), which is more in line with the BEH3 overexpression phenotype. However, this model could not account for the phenotype whereby the bes1 hextuple mutant exhibited a larger vascular area compared to the bes1 quintuple mutant (Figure 7, E and F). We thus tested a scenario in which BEH3 exhibits weaker transcriptional activity than the other homologs ($\alpha =0.95$ and $\beta =0.8$ [1 > $\alpha$ > $\beta$]). In this scenario, if P_O > 0.3, the average output value gradually increased with the increase in P_BEH3 (Figure 8D). However, if P_O was sufficiently low, the mean predicted output values increased as P_BEH3 decreased (Figure 8D), which possibly corresponds to the bes1 quintuple and hextuple mutant phenotypes. Furthermore, a reduction in P_BEH3 decreased the average output value, which corresponds to the smaller area within the cambium in the beh3-1 mutant when compared to the WT (Figure 4, A and B). We decided to use the parameters (n = 3, $\alpha =0.95$ and $\beta =0.8$ [1 > $\alpha$ > $\beta$]) for further modeling.

Figure 8.

Competitive relationship between BES/BZR family proteins enables the robust control of vascular stem cell activity. A, Schematic representation of the mathematical model that describes the competitive relationship between BEH3 and other BES1 homologs. Among n binding sites, BEH3 binds to i of n binding sites at a certain binding rate (P_BEH3). Then, other BES1 homologs bind to j of n–i binding sites at a certain binding rate (P_o). The signaling output (i, j) is defined as ${\alpha }^i{\beta }^j$, where $\alpha$ and $\beta$ represent coefficients of transcriptional activity (<1 in the case of a repressor) of BEH3 and other BES1 homologs, respectively. B–D, Simulated means of signaling output when verifying both P_BEH3 and P_o. The parameters were set as follows: n = 3, $\alpha$ = 0.8, $\beta$ = 0.8 (B); n = 3, $\alpha$ = 1, $\beta$ = 0.8 (C); and n = 3, $\alpha$ = 0.95, $\beta$ = 0.8 (D). Results using various P_BEH3 are indicated in different colors. E, Simulated CV (CV = standard deviation/mean) of output values when verifying both P_BEH3 and P_o (n = 3; $\alpha$ = 0.95; $\beta$ = 0.8). F, Model showing the competitive role of BES/BZR family members in the robust control of vascular stem cells.

To understand the significance of the competitive relationship between BEH3 and other BES/BZR members in terms of phenotypic variation, we simulated the CV by modeling. A reduction in P_BEH3 increased the CV of the output (Figure 8E), which mimics the vascular phenotype of the beh3-1 mutant. In addition, the CV also increased with a decrease in P_BEH3, even when changing the parameters n, $\alpha$, and $\beta$, while following the rule 1 > $\alpha$ > $\beta$ (Supplemental Figure S16). Collectively, our results indicate that cambium-expressed BEH3 competitively inhibits the transcriptional repressor activity of other BES/BZR homologs, thereby minimizing the fluctuation in vascular stem cell activity.

Discussion

In this study, we constructed a gene co-expression network during the vascular stem cell differentiation process in VISUAL (Figure 1). Our network analysis categorized vasculature-related genes into four distinct modules; procambium (module I), cambium (module II), phloem (module III), and xylem (module IV), which were validated by pre-existing FACS and FANS transcriptome datasets. Thus, a reconstitution approach with VISUAL was effective in dissecting the hierarchical process of vascular stem cell differentiation and enabled the identification of genes involved in this process. The FACS datasets (Zhang et al. 2019) showed a similar tendency as our network prediction (Figure 2;  Supplemental Figure S2). However, overlap between our module II genes and the cambium-enriched genes identified by the FACS dataset was statistically but weakly overrepresented (12.9%, 45 out 394 module II genes; Supplemental Figure S2C). These low overlap rates may result from the difference in the status of cambium/vascular stem cells between in vivo and in vitro conditions and/or from technical differences between cell sorting and whole tissue sampling. Therefore, a combinatorial approach involving bioinformatics and VISUAL will allow the accurate and efficient identification of genes of interest expressed in the vasculature, which is usually hidden by its deep-seated location. Indeed, we successfully identified a new cambium regulator, BEH3, in the gene list generated by combining module II genes with the FANS datasets (Shi et al., 2021). Moreover, several genes involved in xylem and phloem development remain unidentified. Therefore, our co-expression network will serve as a foundation for further investigation of vascular development.

Although previous studies indicated redundant roles of BES/BZR family members in BR signaling (Yin et al., 2005; Lachowiec et al., 2018; Chen et al., 2019a, 2019b; Nolan et al., 2020), little has been reported about their functional differences. Here, we discovered that BEH3 acts as a noncanonical member of the BES/BZR family. BEH3 is expressed preferentially in the vascular cambium, whereas BES1 and BZR1 are expressed ubiquitously (Figure 3; Supplemental Figure S4; Yin et al., 2002). During vascular development, the bes1-D gain-of-function mutant showed an accelerated rate of vascular stem cell differentiation and consequently fewer numbers of (pro)cambial cell layers, which eventually decreased vascular size (Figure 4; Kondo et al., 2014). Conversely, BEH3 overexpression increased the number of (pro)cambial cell layers, resulting in the enlargement of vascular size. Furthermore, the beh3-1 loss-of-function mutant showed a smaller vascular area than the WT. Because BEH3 showed a repressive effect on vascular cell differentiation in VISUAL (Figure 5), these results suggest that BEH3 plays a positive role in the maintenance of vascular stem cells, probably by inhibiting excess cell differentiation. In addition, the overexpression of BEH3 influenced hypocotyl elongation and BR-regulated gene expression in a manner opposite to that seen with other BES/BZR homologs (Figure 6E; Supplemental Figure S8; Yin et al., 2002; Nguyen et al., 2021). These results suggest the possibility that BEH3 performs important functions not only in vascular development but also in BR signaling. Moreover, it was recently reported that BEH3 is transcriptionally induced by osmotic stress and abscisic acid treatment and negatively regulates sensitivity to osmotic stress (Nguyen et al., 2021), suggesting a potential role for BEH3, as a noncanonical member of the BES/BZR family, in various aspects of plant growth and environmental adaptation.

Importantly, the beh3-1 loss-of-function mutant showed greater variation in vascular size than the WT. Molecular analyses suggested that BEH3 competes with other BES/BZR homologs for common DNA-binding motifs to inhibit their transcriptional repressor activity (Figure 6). Taking into account the phenotypic differences between bes1 quintuple and hextuple mutants, mathematical modeling showed that BEH3 exhibits weaker transcriptional activity than the other homologs (Figures 7 and 8). As a similar example in mammals, the transcription factor JunB, which exhibits weaker activity than its homolog c-Jun, can repress c-Jun-mediated transactivation (Deng and Karin, 1993). The results of our modeling and genetic analysis suggest that the presence of a weak-type member in a transcription factor family enables the fine-tuned control of signaling outputs. Taken together, we conclude that the competition between BES/BZR family members in the cambium potentially enables the stable and robust maintenance of vascular stem cells, thus ensuring continuous radial growth (Figure 8F). However, it remains unclear why only BEH3 partially lacks transcriptional repressor activity, and why BEH3 only competitively suppresses the repressor activity of its homologs. Further comparative interactome analyses of BEH3 and other BES1 homologs will help uncover the mechanism(s) underlying these functional differences between BES/BZR proteins.

Materials and methods

Plant materials

Seeds of Arabidopsis (Arabidopsis thaliana) bes1-1 (SALK_098634), beh1-1 (SAIL_40_D04), beh3-1 (SALK_017577), and beh4-1 (SAIL_750_F08) mutants were obtained from the Arabidopsis Biological Resource Center. All genotypes used in this study were in the Columbia-0 (Col-0) background. The bzr1-1 single mutant and bes1-1 bzr1-2 double mutant were generated previously using the CRISPR/Cas editing system (Saito et al., 2018), and bes1-1 beh1-1, bes1-1 beh3-1, bes1-1 beh4-1, bes1-1 bzr1-2 beh3-1, and bes1-1 bzr1-2 beh4-1 higher-order mutants were generated by crossing. Chen et al. reported that the bes1-1 mutant expresses one of the two BES1 splice variants (BES1-S); however, the other splice variant (BES1-L), which is functionally more important, is completely eliminated (Jiang et al., 2015; Chen et al., 2019b). The bes1-1 bzr1-2 beh2-5 beh4-1 quadruple mutant was generated by CRISPR/Cas9 editing using the pKAMA-ITACH vector (Tsutsui and Higashiyama, 2017). The single guide RNA (sgRNA; Supplemental Table S5) sequence corresponding to the target gene, BEH2, was cloned into the pKIR1.1 vector. The resulting construct was transformed into the bes1-1 bzr1-2 beh4-1 triple mutant. In the T₂ generation, Cas9-free plants homozygous for the beh2-5 mutation were selected as described previously (Tsutsui and Higashiyama, 2017). To isolate the beh2-5 single mutant and bes1-1 beh2-5 double mutant, the bes1 quadruple mutant was crossed with Col-0 and bes1-1. The bes1 quintuple mutant (bes1-1 bzr1-2 beh1-2 beh2-5 beh4-1) was also generated using CRISPR/Cas9 editing, which targeted the BEH1 locus in the bes1-1 bzr1-2 beh2-5 beh4-1 quadruple mutant background. To generate the bes1 hextuple mutant (bes1-1 bzr1-2 beh1-2 beh2-5 beh3-2 beh4-1), the BEH3 locus was edited using CRISPR/Cas9 in the bes1 quintuple mutant background. To generate a construct for BEH3 inducible expression under the control of the β-estradiol-inducible promoter, the coding sequence of BEH3 (At4g18890.1) was amplified from Col-0 cDNA by PCR (Supplemental Table S6) and then cloned into pENTR/D-TOPO (Life Technologies). Using LR clonase II (Life Technologies), the region between attL1 and attL2 in the entry vector was recombined into the destination vector, pER8-GW-CFP, which expresses the CFP-tagged proteins upon estrogen application (Ohashi-Ito et al., 2010). To generate the BEH3pro:GUS construct, a 1,997-bp promoter fragment upstream of the BEH3 translation start site was amplified from Col-0 genomic DNA by PCR (Supplemental Table S6) and cloned into pENTR/D-TOPO. The region between attL1 and attL2 in the entry vector was recombined into the destination vector, pGWB434. The constructs were transformed into Col-0 plants via Agrobacterium (Agrobacterium tumefaciens) strain GV3101 by the floral dip method (Clough and Bent, 1998). Plants were grown on conventional half-strength Murashige and Skoog (MS) medium, pH 5.7, at 22°C under 60–70 µmol m⁻² s⁻¹ continuous white light, unless stated otherwise.

VISUAL

Plants were analyzed using VISUAL as described previously (Kondo et al., 2016), with a slight modification. Briefly, cotyledons of 6-day-old seedlings grown in liquid half-strength MS medium were cultured under continuous light in MS-based VISUAL induction medium containing 0.25 mg·L^-1 2,4-D, 1.25 mg·L⁻¹ kinetin, and 20-µM bikinin. To observe ectopic xylem differentiation, the xylem indicator BF-170 (Sigma-Aldrich) was added to the induction medium (Nurani et al., 2020). After 4 days of induction, cotyledons were fixed in ethanol: acetic acid (3:1, v/v) and mounted onto microscope slides in a clearing solution (chloral hydrate: glycerol: water = 8:1:2 [w/w/v]). Images obtained on a fluorescent stereomicroscope (M165, Leica) were binarized using the “Threshold” function of the ImageJ software (Schneider et al., 2012). The xylem differentiation ratio was calculated based on the BF-170 positive area cut-off by suitable thresholding per whole cotyledon area.

RT-qPCR

Total RNA was extracted from cotyledons (6–10 cotyledons) subjected to VISUAL experiments or whole seedlings in other experiments using the RNeasy Plant Mini Kit (Qiagen). Quantitative PCR was performed using a LightCycler (Roche Diagnostics). UBQ14 (for VISUAL experiments) and ACT2 (for other experiments) were used as reference genes. Gene-specific primers and TaqMan Probe sets are listed in Supplemental Table S7.

Microarray experiments

To generate the VISUAL 6-h interval time course microarray datasets, VISUAL was performed as described previously (Kondo et al., 2016). After VISUAL induction, cultured cotyledons (6–10 cotyledons) were collected every 6 h. Sampling consisted of three independent biological replicates. Total RNA was extracted from cultured cotyledons the using RNeasy Plant Mini Kit (Qiagen). Transcript levels of marker genes in all replicate samples were calculated using RT-qPCR. All replicates of extracted RNA were mixed at each timepoint for microarray experiments. Microarray experiments were performed using the Arabidopsis Gene 1.0STArray (Affymetrix) according to the standard Affymetrix protocol.

Construction of the co-expression network

Based on the VISUAL 6-h interval time course microarray data described above, the VISUAL-induced genes were defined as more than eight-fold upregulated genes 72 h after VISUAL induction compared to time 0 h. Median normalized values in log₂ scale of the VISUAL-induced genes (855 genes) were obtained from multiple transcriptome datasets: the VISUAL 6-h interval time course (this study), WT versus bes1-1 (WT 72 h rep1–3 and bes1-1 72 h rep1–3 from Gene Expression Omnibus accession number GSE110199; Saito et al., 2018), WT versus apl mutant (WT rep1–3 and apl rep1–3 from GSE80026; Kondo et al., 2016), and SEOR1-YFP cell sorting data (YFP positive rep1–3 and YFP negative rep1–3 from GSE80027; Kondo et al., 2016). As described previously (Kondo et al., 2016), the VISUAL co-expression network was constructed using the WGCNA R package (Langfelder and Horvath, 2008). The adjacency matrix was calculated with soft thresholding power, which was chosen based on the criterion of scale-free topology (fit index = 0.8). For cross-validation, the microarray datasets of tissue-specific FACS in Arabidopsis roots (Brady et al., 2007 [GSE8934]; Zhang et al., 2019 [GSE125244]) and the RNA-seq dataset of tissue-specific FANS in Arabidopsis inflorescence stems ([GSE142034]; Shi et al., 2021) were used. Among VISUAL-induced genes, the genes that showed expression (values >0) in the normalized transcriptome datasets attached as supplemental datasets in Zhang et al. (2019) or Shi et al. (2021) were used for further studies. Z-score values of these genes were calculated based on the FACS and FANS datasets and were visualized as heatmaps and violin plots using R (version 3.5.1). For gene enrichment analysis, the number of high Z-score genes (Z-score >1) in each module was calculated relative to all VISUAL-induced genes. The significance of the enrichment was assessed using Fisher’s exact test.

Phylogenetic analysis

The phylogenetic tree of Arabidopsis BES/BZR homologs was generated by aligning full-length protein sequences using MEGA X (version 10.1, https://www.megasoftware.net/; Kumar et al., 2018) using the maximum likelihood algorithm. Bootstrap values (1,000 replicates) were used to assess confidence of each branch, whose frequency is indicated at each branch node, and the scale bar indicates the number of amino acid substitutions per site. Sequence alignment and machine-readable tree file of the phylogenetic analyses are provided as Supplemental File S1 and Supplemental File S2, respectively.

GUS staining

Cotyledons subjected to VISUAL experiments or whole seedlings were fixed in 90% acetone for 1 h at –20°C. GUS staining was performed according to the previously published protocol (Kondo et al., 2014).

Plant growth condition for vascular histological analysis

To minimize variation in growth conditions, Arabidopsis Col-0 (WT) and beh3-1 (SALK_017577) plants were cultured on peat pots (Jiffy-7; Jiffy Products International AS, Norway) at 22°C under continuous light. For vascular observation, freshly harvested seeds were sown on rockwool (Nitto Boseki, Tokyo, Japan) soaked with 0.5 g·L⁻¹ (w/v) liquid fertilizer (Hyponex, Japan) and grown at 22°C under short-day (8-h light and 16-h dark) conditions. To equalize culture conditions, 30 seeds of each genotype were divided into two rockwool pads (15 seeds each per genotype) on independent culture trays. After a few days, non- or late-germinating seeds were eliminated. After 21 days, hypocotyls were fixed in FAA solution containing 40% (v/v) ethanol, 2.5% (v/v) acetic acid, and 2.5% (v/v) formalin for histological analysis. For other vascular sectioning experiments, Arabidopsis seedlings were grown in half-strength liquid MS medium for 11 or 12 days under continuous light.

Histological analysis

Cross-sections of hypocotyls from seedlings were prepared as described previously (Kondo et al., 2014). Samples were fixed in FAA solution and gradually dehydrated using an ethanol series and then embedded in Technovit 7100 resin (Kulzer). Sections of 5-μm thickness were prepared on a RM2165 microtome (Leica). Sliced samples were stained with 0.1% (w/v) toluidine blue for 1 min. To evaluate the vascular phenotypes, “Cambium cell layer” and “whole vascular area” were defined as the average number of minimal cell layers between the xylem and phloem areas in two phloem poles and as the area inside the endodermis, respectively. For beh3-1 mutants, “area within the cambium”, defined as the area inside the middle of cambium layers, was measured because their vasculatures were too large to measure “whole vascular area” when grown under long-term culture.

Hypocotyl elongation assay under dark conditions

Seeds of two independent β-estradiol-inducible BEH3 overexpression lines were sown on half-strength MS medium containing DMSO or 10-µM β-estradiol. Plates were incubated in the dark after white light illumination to initiate germination. After 7 days, the plates were scanned using a flatbed scanner (CanoScan 9000F, Canon). Hypocotyl length was measured in ImageJ.

Confocal laser scanning microscopy

BF-170 signals from VISUAL samples were visualized using a FV1200 confocal laser scanning microscope (Olympus). Z-series fluorescence images were obtained at excitation and detection wavelengths of 473 and 490–540 nm, respectively. Z-projection images were created in ImageJ.

Production and purification of recombinant BES/BZR family proteins

Maltose-binding protein (MBP)-fused DBDs of BES1 (amino acids 20‒103) and BZR1 (amino acids 21‒104) were produced and purified as described previously (Nosaki et al., 2018). The coding regions for the DBDs from BEH3 (amino acids 4‒88) and BEH4 (amino acids 4‒89) were cloned into pMAL-c2X (New England Biolabs) and, subsequently, transformed into Escherichia coli Rosetta (DE3) cells (Novagen). MBP-fused DBDs of BEH3 and BEH4 were produced and purified as BES1 and BZR1.

Analysis of BES/BZR proteins by size exclusion chromatography

The purified proteins were loaded onto a Superdex 200 GL 10/300 (GE Healthcare) column and eluted with a buffer containing 20-mM Tris–HCl (pH 7.5), 250-mM NaCl, 1-mM DTT, and 5% glycerol. To estimate the multimerization state of BES/BZR family proteins, the following standards were used: thyroglobulin (Mr 670,000 Da), bovine globulin (Mr 158,000 Da), chicken ovalbumin (Mr 44,000 Da), equine myoglobin (Mr 17,000 Da), and vitamin B-12 (Mr 1,350 Da; Bio-Rad).

EMSAs

EMSAs of BES/BZR family proteins were performed as described previously (Nosaki et al., 2018).As a probe with BRRE-core motif, DNA fragment of the CONSTITUTIVE PHOTOMORPHOGENIC DWARF (CPD) promoter region (‒93 to ‒64) containing the BRRE-core motif was used. Probes with G-box (E-box) motif or mBRRE motif were designed by substitutions of the BRRE-core motif in the DNA fragment of the CPD promoter region. Probes were synthesized and labeled by carboxyfluorescein (Eurofins).

Luciferase assay

The coding sequences of BES1-L (At1g19350.3), BZR1 (At1g75080.1), BEH3 (At4g18890.1), and BEH4 (At1g78700.1) were cloned in-frame with the sequence of the ligand binding domain of the rat GR. An approximately 1,952-bp fragment upstream of the DWF4 transcription start site was cloned upstream of the enhanced green-emitting luciferase gene (ELUC) encoding ELUC with a PEST domain (Tamaki et al., 2020; Toyobo). The effector and reporter constructs, as well as a p19k suppressor construct, were introduced separately into Agrobacterium strain GV3101 MP90 and the agrobacteria mixed in equal volume and co-infiltrated into N. benthamiana leaves. After 2 days, leaf discs excised from the infiltrated leaves were incubated with 200 µM D-luciferin (Wako) for 2 h in white 24-well plates (PerkinElmer). Then, 10-µM DEX was added to each well, and luciferase activity was measured using the TriStar2 LB942 microplate reader (Berthold) system, as described previously (Tamaki et al., 2020).

Mathematical modeling

A mathematical model for the competitive binding between BEH3 and other homologs was constructed based on the following assumptions: (1) there are n binding sites at the promoter and (2) the weak-type BEH3 is competitively superior to the strong-type homologs in terms of binding to the promoter. The weak-type BEH3 occupies the vacant sites first, followed by the strong-type homologs, which occupy the remaining vacant sites. The binding probabilities of the weak-type BEH3 and strong-type homologs ($p_{\mathit{BEH}3}$ and $p_O$, respectively) are proportional to the transcript abundance of BEH3 and other homologs, respectively. Using $p_{\mathit{BEH}3}$ and $p_O$, the probability that i BEH3 and j homologs bind to the promoter is given as the product of two binomial distributions as follows:

$$f\left(i,j|p_{\mathit{BEH}3},p_O,n\right)=\left(\begin{array}{c}n\\ i\end{array}\right)p_{\mathit{BEH}3}^i{(1-p_{\mathit{BEH}3})}^{n-i}\times \left(\begin{array}{c}n-i\\ j\end{array}\right)p_O^j{(1-p_O)}^{n-i-j}.$$

Binding of BEH3 and other homologs to the binding sites suppresses the transcription level of direct target genes, leading to the repression of the proliferation of stem cells. The effect of the binding sites of BEH3 and other homologs on the cell proliferation rate of stem cells was designated as $\alpha$ and $\beta$, respectively. Because the effect of BEH3 is weaker than that of other homologs, $\alpha$ >$\mathrm{}\beta$ always holds. When i BEH3 and j homologs bind to the promoter, the cell proliferation rate is given as ${\alpha }^i{\beta }^j$. The cell division rate should be proportional to the area inside the cambium, which was measured in experiments because the area inside the cambium increases with the increase in the rate of cell division. When there are multiple individuals, the expected rate of cell division of stem cells is calculated as follows:

$$\begin{array}{c}E[\text{cell}\text{division}\text{rate}]=\sum _{i=0}^n\sum _{j=0}^{n-i}{\alpha }^i{\beta }^{j}f\left(i,j|p_{\mathit{BEH}3},p_O,n\right)\\ ={[\alpha p_{\mathit{BEH}3}+\left(1-p_{\mathit{BEH}3}\right)\left(\beta p_O+1-p_O\right)]}^n\end{array}$$ (1)

Using Equation 1, the variance of the cell division rate can be calculated as:

$$\begin{array}{c}V\left[\text{cell}\text{division}\text{rate}\right]={\left[{\alpha }^2{p}_{\mathit{BEH}3}+\left(1-p_{\mathit{BEH}3}\right)\left({\beta }^2{p}_O+1-p_O\right)\right]}^n\\ -{[\alpha p_{\mathit{BEH}3}+\left(1-p_{\mathit{BEH}3}\right)\left(\beta p_O+1-p_O\right)]}^{2n}\end{array}$$ (2)

The CV of the cell division rate was calculated as:

$$CV=\sqrt{V\left[\mathit{cell}\mathrm{}\mathit{division}\mathrm{}\mathit{rate}\right]}/E\left[\mathit{cell}\mathrm{}\mathit{division}\mathrm{}\mathit{rate}\right]$$

For calculations, the software Mathematica version 12 (Walfram) was used.

Accession numbers

Arabidopsis Genome Initiative locus identifiers for each gene are as follows: BEH3 (At4g18890), BES1 (At1g19350), BZR1 (At1g75080), CLE41 (At3g24770), PXY/TDR (At5g61480), WOX4 (At1g46480), WOX14 (At1g20700), BIN2 (At4g18710), BIL1 (At2g30980; AtSK23), ARR7 (At1g19050), ARR15 (At1g74890), MP/ARF5 (At1g19850), TMO6 (At5g60200), LBD4 (At1g31320), ANT (At4g37750), KNAT1/BP (At4g08150), SVP (At2g22540; AGL22), PTL (At5g03680), RAS1 (At1g09950), MYB87 (At4g37780), SMXL5 (At5g57130), APL (At1g79430), SEOR1 (At3g01680), HB-8 (At4g32880), VND5 (At1g62700; ANAC026), VND6 (At5g62380; ANAC101), IRX3 (At5g17420; CESA7), RABC2A (At5g03530), BRX (At1g31880), GSL07 (At1g06490), NAC045 (At3g03200), XCP1 (At4g35350), VND7 (At1g71930; ANAC030), LBD3 (At1g16530; ASL9), SCL28 (At1g63100), MYB3R-4 (At5g11510), BEH4 (At1g78700), BEH1 (At3g50750), BEH2 (At4g36780), BRI1 (At4g39400), DWF4 (At3g50660), BR6ox2 (At3g30180), SAUR15 (At4g38850; SAUR-AC1), IAA19 (At3g15540), CPD (At5g05690).

Microarray data are available at the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/). Accession numbers for the VISUAL 6-h interval time course microarray have been deposited at the Gene Expression Omnibus under series number GSE171548.

Supplemental data

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

Supplemental Figure S1 . Time course gene expression analysis of VISUAL-induced marker genes.

Supplemental Figure S2 . Comparative analysis of the VISUAL co-expression network and vasculature-related FACS data sets.

Supplemental Figure S3 . Comprehensive analysis of the VISUAL co-expression network and vasculature-related FANS datasets.

Supplemental Figure S4 . Expression profiles of BES/BZR homologs in vasculature-related transcriptome datasets.

Supplemental Figure S5 . Cross-section images showing the hypocotyl vasculature of the WT and beh3-1 mutant plants.

Supplemental Figure S6 . Amino acid sequence alignment of Arabidopsis BES/BZR homologs.

Supplemental Figure S7 . Hypocotyl vasculature phenotype of two independent β-estradiol-inducible BEH3-overexpressing transgenic lines.

Supplemental Figure S8 . Hypocotyl elongation in two independent pER8:BEH3-CFP transgenic lines under dark conditions.

Supplemental Figure S9 . Semi-automatic calculation of xylem differentiation ratio by the binarization method.

Supplemental Figure S10 . Ectopic xylem differentiation in VISUAL with beh2 and bes1 beh2 mutants.

Supplemental Figure S11 . Ectopic xylem cells in VISUAL observed under a confocal laser scanning microscope.

Supplemental Figure S12 . Preparation of MBP fusions of the DBDs of the BES/BZR family proteins for the EMSA.

Supplemental Figure S13 . Expression levels of BEH3 and BES1 in bes1-1 and beh3-1, respectively, at 72 h post-induction in VISUAL.

Supplemental Figure S14 . Time course expression analysis of BES/BZR target genes in the β-estradiol-inducible BEH3 overexpression line.

Supplemental Figure S15 . Construction of genome-edited lines for BES/BZR homologs.

Supplemental Figure S16 . Simulated CV of output values determined by the mathematical model of competitive binding between BEH3 and other homologs.

Supplemental Table S1 . Results of two-way ANOVA for Figure 5E.

Supplemental Table S2 . Results of Student’s t test for Figure 6B.

Supplemental Table S3 . Results of Student’s t test for Figure 6D.

Supplemental Table S4 . Results of two-way ANOVA for Supplemental Figure S14.

Supplemental Table S5 . Sequences of sgRNA for CRISPR/Cas9 system.

Supplemental Table S6 . List of cloning primers.

Supplemental Table S7 . List of RT-qPCR primers.

Supplemental Data Set S1 . List of VISUAL-induced genes.

Supplemental Data Set S2 . List of VISUAL-induced transcription factors.

Supplemental File S1 . Sequence alignment file for the phylogenetic analysis in Figure 3.

Supplemental File S2 . Machine-readable tree file for the phylogenetic analysis in Figure 3.

 

T.F., M.S., and Y.K. designed the experiments, coordinated the project, and wrote the manuscript; T.F, M.S., H.U., S.S., W.Y., S.N., T.M., and Y.K. performed the experiments; A.S. performed mathematical model simulations; M.T. and H.F. participated in discussions. All authors reviewed and edited the manuscript.

The authors 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) is: Yuki kondo (pkondo@tiger.kobe-u.ac.jp).