bHLH transcription factors cooperate with chromatin remodelers to regulate cell fate decisions during Arabidopsis stomatal development

Abstract

The development of multicellular organisms requires coordinated changes in gene expression that are often mediated by the interaction between transcription factors (TFs) and their corresponding cis-regulatory elements (CREs). During development and differentiation, the accessibility of CREs is dynamically modulated by the epigenome. How the epigenome, CREs, and TFs together exert control over cell fate commitment remains to be fully understood. In the Arabidopsis leaf epidermis, meristemoids undergo a series of stereotyped cell divisions, then switch fate to commit to stomatal differentiation. Newly created or reanalyzed scRNA-seq and ChIP-seq data confirm that stomatal development involves distinctive phases of transcriptional regulation and that differentially regulated genes are bound by the stomatal basic helix-loop-helix (bHLH) TFs. Targets of the bHLHs often reside in repressive chromatin before activation. MNase-seq evidence further suggests that the repressive state can be overcome and remodeled upon activation by specific stomatal bHLHs. We propose that chromatin remodeling is mediated through the recruitment of a set of physical interactors that we identified through proximity labeling-the ATPase-dependent chromatin remodeling SWI/SNF complex and the histone acetyltransferase HAC1. The bHLHs and chromatin remodelers localize to overlapping genomic regions in a hierarchical order. Furthermore, plants with stage-specific knockdown of the SWI/SNF components or HAC1 fail to activate specific bHLH targets and display stomatal development defects. Together, these data converge on a model for how stomatal TFs and epigenetic machinery cooperatively regulate transcription and chromatin remodeling during progressive fate specification.

Fig 1. Dynamic changes in transcriptional entropy and number of genes expressed during stomatal development.

(A) The principle of transcriptional entropy is illustrated by comparing the correlation of entropy and expressed genes for simulated cell populations with homogenous gene expression (same expression for all genes, blue line) and populations in which the top 10% of expressed genes comprise 40% (orange line) or 50% (grey line) of all transcripts. An increase in the number of expressed genes leads to a rise in entropy, whereas increased expression heterogeneity results in a drop in entropy. (B) Heatmap of entropy scores overlaid on a UMAP plot of scRNAseq data from epidermal cells of developing leaves. Data from [27]. (C, D) Boxplot of entropy scores (C) and number of genes expressed (D) in different epidermal cell types. Cell types are indicated in panel keys with M (meristemoids) being the most pluripotent cells and arrows below the plots indicating bidirectional paths to differentiation as guard cell (GC) or epidermal pavement cells (epi). The data underlying this figure can be found in S1 Data.

Fig 2. Differentially regulated genes reside in closed chromatin regions and undergo dynamic changes in chromatin signature during fate transitions.

(A) Openness of the chromatin determined by ATAC-seq [33] read enrichment at SPCH [28], MUTE [29], FAMA, and SCRM [30] targets and random intergenic intervals in meristemoids. Open regions are dark purple; closed regions are white. (B) Example of a shared bHLH target (SWEET5) that is nucleosome bound (MNase-seq [38]; top track, deep blue), heavily H3K27me3-methylated in early stomatal lineage cells (H3K27me3 ChIP-seq in FAMA^LGK cells [37]; second track, light blue) and shows a decrease in H3K27me3 in mature GCs (H3K27me3 ChIP-seq in GC [37]; third track, green); the bottom 5 tracks show binding of the stomatal lineage bHLHs and the chromatin remodeler BRM from ChIP-seq experiments. (C) Late stomatal lineage-specific expression pattern of SWEET5 in scRNA-seq. Violin plots show the expression pattern (cells with expression) of cell type markers (top 7) and SWEET5 (bottom) in the following cell clusters: alternative epidermal (pavement cells) fate (epi1-4), meristemoid (M and M/GMC), guard cell (GC1-3). (D) Genome-wide average fold enrichment of H3K27me3 at the gene body in WT GC (purple) and in FAMA^LGK cells (reprogrammed GCs with earlier lineage identity, orange) at SPCH, MUTE, FAMA, and SCRM targets and random genes. The data underlying this figure can be found in S1 Data. ATAC-seq, assay for transposase-accessible chromatin using sequencing; bHLH, basic helix–loop–helix; ChIP-seq, chromatin immunoprecipitation followed by deep sequencing; GC, guard cell; SCRM, SCREAM; scRNA-seq, single-cell RNA-sequencing; SPCH, SPEECHLESS; TES, target end sites; TSS, target start sites; WT, wild type.

Fig 3. The stomatal lineage bHLHs have the potential to bind nucleosomal DNA and induce chromatin remodeling.

(A) Nucleosomal density around the binding sites of the known pioneer factor LFY [18], the stomatal lineage bHLHs, and an unrelated bHLH, PIF4, in leaf tissue. Nucleosome density was determined by MNase-seq [38]. TF binding sites were determined from ChIP-seq data described in S1 Data. (B) Top nucleosome-associated ChIP-seq motifs identified at SCRM (top), MUTE (middle), and FAMA (bottom) peaks. (C) Average ChIP-seq fold enrichment around the center of Nuc or Non-nuc SCRM, MUTE, and FAMA peaks. MUTE and SCRM show a pattern similar to canonical animal pioneer factors. FAMA shows a noncanonical pattern that is different from animal pioneer factors. (D) Changes in nucleosome occupancy (MNase-seq signal) at nucleosomal MUTE peaks (left) and random peaks (right) after ectopic, EST-induced, MUTE expression. (E) Example of a decrease in nucleosome signal at a MUTE binding site upon MUTE induction. The top 2 traces represent the MNase-seq signal (nucleosome occupancy) after mock treatment (purple) and MUTE induction (EST, orange). The bottom traces show ChIP-seq binding of MUTE, FAMA, and SCRM at this locus. The data underlying this figure can be found in S1 Data. bHLH, basic helix–loop–helix; ChIP-seq, chromatin immunoprecipitation followed by deep sequencing; Non-nuc, nonnucleosomal; Nuc, nucleosomal; SCRM, SCREAM; TF, transcription factor.

Fig 4. The stomatal lineage bHLHs are associated with chromatin-related factors.

(A) Scatterplot of log2 fold enrichment of proteins in SCRM-TbID plants compared to wild-type samples as determined by unpaired 2-sided t test. Red circles are interaction candidates (proteins enriched vs. all controls). The positions of SCRM and its heterodimerization partners SPCH and FAMA are indicated in light red; positions of SWI/SNF components and HAC1 are indicated in dark red (filled circles). MUTE was not enriched due to the low abundance of the protein. (B) Y2H testing the interaction of SCRM (as AD-fusion) with the identified SWI/SNF complex components and HAC1 (as BD-fusion). Cells were spotted on SD-LT medium to confirm transformation with the constructs and SD-LTHA to test for interaction. Combinations with SWI3C, BRIP2, and HAC1 were grown on SD-LTHA with 5 mM 3-AT to suppress autoactivation (indicated by *). (C) Scheme of the spacing calculations between 2 DNA binding proteins used in D. Proteins that bind in a coordinated manner (blue line) show a narrow (constrained spacing) or wide (relaxed spacing) peak. Noncoordinated proteins (random, red dotted line) do not show a peak. (D) Density plot of the distance between SCRM and BRD1 (BRD1-SCRM), SCRM and BRM (BRM-SCRM), and SCRM and FAMA (FAMA-SCRM) shared ChIP-seq binding sites. Kolmogorov–Smirnov test: p-value for BRD1-SCRM vs. BRM-SCRM = 0.004787, other comparisons p-value < 2.2 × 10⁻¹⁶. (E) Plots of average ChIP-seq signal fold enrichment at shared (dark blue) and unique (light green) SCRM (left) and FAMA (right) peaks. In both cases, peaks shared by both SCRM and FAMA had higher signals than unique peaks. The data underlying this figure can be found in S1 Data. AD, activation domain; BD, binding domain; bHLH, basic helix–loop–helix; ChIP-seq, chromatin immunoprecipitation followed by deep sequencing; SCRM, SCREAM; SPCH, SPEECHLESS; SWI/SNF, SWITCH DEFECTIVE/SUCROSE NONFERMENTABLE; TbID, TurboID; Y2H, yeast two-hybrid.

Fig 5. The SWI/SNF complex and HAC1 are required for GC differentiation.

(A) Quantification of abnormal GC complexes in FAMAp::amiRNA lines of HAC1 (orange), BRM (red), and SWI3C (brown) compared to an amiRNAi control (scrambled sequence, grey) and to lines in which fama is rescued with a FAMA construct that persists to maturity (normal FAMA) and one with early-terminating expression (“short-lived” or short FAMA) in blue. The top panel reports the overall percentage of abnormal GCs in 21-day-old cotyledons. Each colored dot represents a plant with the mean and SE indicated as a black dot with whiskers. Statistical difference was tested by Kruskal–Wallis rank sum test (p-value = 0.0013), followed by a Conover post hoc test with holm correction (statistical difference between lines (adjusted p-value < 0.025) is shown as compact letter display above the data points). The bottom panel provides finer phenotypic dissection of abnormal GCs, with percentage of GCs in each of phenotypic classes represented by the size of the colored circles. (B-E) Confocal images of representative cells from cotyledon epidermis with cell outlines in magenta. (B) Reexpression of SPCH (left) and MUTE (right) translational reporters in GCs indicates reversion of GC cell identity to an early stomatal lineage state in FAMAp::amiHAC1. (C, D) Evidence that even morphologically normal FAMAp::amiHAC1 GCs have fate defects. Compared to WT GCs (C) with pore autofluorescence (yellow) and decreased ML1p-driven reporter activity (magenta plasma membrane signal), in FAMAp::amiHAC1 plants (D), autofluorescence is missing and ML1p remains active in older GCs. Arrows point to morphologically mature stomata. (E) Comparison of FAMA expression during late stages of GC development, showing early termination in FAMAp::amiHAC1 and FAMAp::amiBRM, similar to the “short-lived” FAMA line. The data underlying this figure can be found in S1 Data. BRM, BRAHMA; GC, guard cell; GMC, guard mother cell; M, meristemoid; SPCH, SPEECHLESS; SWI/SNF, SWITCH DEFECTIVE/SUCROSE NONFERMENTABLE; WT, wild-type.

Fig 6. A Model for chromatin reprogramming by the stomatal lineage bHLH heterodimers using transcriptional regulation of FAMA as an example.

(Top) The stomatal lineage bHLH TFs form heterodimers and can access targets located in repressive chromatin regions. They then recruit the SWI/SNF complex and HAC1 to evict nucleosomes and acetylate histone tails to induce chromatin reprogramming and maintain the expression of their targets at different stages of GC fate commitment. (Bottom) HAC1 and the SWI/SNF complex could impact FAMA and FAMA target gene expression at 2 stages. Experimental evidence suggests that FAMA itself is kept active by the action of HAC1 and the SWI/SNF complex (late FAMA), and this enables FAMA to promote terminal differentiation. Based on indirect evidence, early FAMA expression, which enables it to repress cell cycle progression, could also depend on HAC1 and SWI/SNF activity. bHLH, basic helix–loop–helix; GC, guard cell; SWI/SNF, SWITCH DEFECTIVE/SUCROSE NONFERMENTABLE; TF, transcription factor.