Skip to main content
PMC home page
eLife logoLink to eLife
. 2021 Sep 15;10:e69508. doi: 10.7554/eLife.69508

A novel bivalent chromatin associates with rapid induction of camalexin biosynthesis genes in response to a pathogen signal in Arabidopsis

Kangmei Zhao 1, Deze Kong 2, Benjamin Jin 1, Christina D Smolke 2,3, Seung Yon Rhee 1,✉
Editors: Nicola J Patron4, Jessica K Tyler5
PMCID: PMC8547951  PMID: 34523419

Abstract

Temporal dynamics of gene expression underpin responses to internal and environmental stimuli. In eukaryotes, regulation of gene induction includes changing chromatin states at target genes and recruiting the transcriptional machinery that includes transcription factors. As one of the most potent defense compounds in Arabidopsis thaliana, camalexin can be rapidly induced by bacterial and fungal infections. Though several transcription factors controlling camalexin biosynthesis genes have been characterized, how the rapid activation of genes in this pathway upon a pathogen signal is enabled remains unknown. By combining publicly available epigenomic data with in vivo chromatin modification mapping, we found that camalexin biosynthesis genes are marked with two epigenetic modifications with opposite effects on gene expression, trimethylation of lysine 27 of histone 3 (H3K27me3) (repression) and acetylation of lysine 18 of histone 3 (H3K18ac) (activation), to form a previously uncharacterized type of bivalent chromatin. Mutants with reduced H3K27me3 or H3K18ac suggested that both modifications were required to determine the timing of gene expression and metabolite accumulation at an early stage of the stress response. Our study indicates that the H3K27me3-H3K18ac bivalent chromatin, which we name as kairostat, plays an important role in controlling the timely induction of gene expression upon stress stimuli in plants.

Research organism: A. thaliana

eLife digest

In the fight against harmful fungi and bacteria, plants have an arsenal of chemicals at their disposal. For instance, species in the crucifer family – which includes mustard, cabbages and the model plant Arabidopsis thaliana – can defend themselves with camalexin, a compound produced soon after the plant receives signals from its attacker. What controls this precise timing, however, is still unclear.

For the genes that rule the production of camalexin to be ‘read’, interpreted, and ultimately converted into proteins, their DNA sequences first need to be physically accessible to the cell. This availability is controlled, in part, by adding or removing chemical groups onto histones, the spool-like structures which DNA wraps around. These precisely controlled modifications ultimately help to activate or repress a gene. Sometimes, activating and inhibiting chemical groups can be present in the same location, creating what is known as a bivalent chromatin domain.

Zhao et al. investigated whether histone modifications regulate when A. thaliana produces camalexin in response to an attack. A combination of bioinformatics and experimental approaches highlighted two chemical modifications (one repressive, the other activating) which were present on the same histone, forming a previously unknown bivalent chromatin domain. Mutant plants which did not carry these modifications could not produce camalexin at the right time. Further experiments showed that under normal conditions, both histone modifications were present. However, when the plant was under attack, the level of repressive and activating modifications respectively decreased and increased, leading to gene activation.

Together, the results by Zhao et al. suggest that both histone modifications are required for camalexin genes to respond appropriately to signals from a harmful agent. A deeper understanding of this new mechanism could, in turn, allow scientists to engineer crops that are better at resisting disease.

Introduction

Plants produce many metabolites that are essential for responding to environmental cues, the expression of which requires tight temporal control (Barco and Clay, 2019; Grotewold, 2005; Martin et al., 2010). One of the most prominent defense metabolites from Arabidopsis thaliana is camalexin. Camalexin is an indole alkaloid derived from tryptophan and can be induced rapidly by various biotic and abiotic stimuli in Arabidopsis (Stotz et al., 2011). Essential biosynthetic enzymes belong to the cytochrome P450 family, including CYP79B2/3, CYP71A12/13, and PHYTOALEXIN DEFICIENT 3 (PAD3) (Klein et al., 2013; Schuhegger et al., 2006). The expression of these enzymatic genes can be rapidly induced by various abiotic and biotic stresses, and pathogen-derived substances such as flagellin 22 (FLG22) and oligogalacturonides (Denoux et al., 2008). Previous studies revealed complex transcriptional and translational control of camalexin biosynthesis genes (Frerigmann et al., 2016; Stahl et al., 2016; Zhou et al., 2020; Birkenbihl et al., 2012; Frerigmann et al., 2015; Mao et al., 2011). At the transcriptional regulation level, transcription factors from the MYB family, including MYB34, MYB51, and MYB122, promote camalexin biosynthesis in response to Pseudomonas syringae infection (Stahl et al., 2016; Frerigmann et al., 2015). WRKY33 functions as an activator and directly binds to the promoters of camalexin biosynthesis genes (Birkenbihl et al., 2012). Besides transcription factors, CALCIUM-DEPENDENT PROTEIN KINASE (CPK)5/6 and MAPK3/6 can phosphorylate WRKY33 to enhance promoter binding and transactivation (Zhou et al., 2020; Mao et al., 2011). At the translational level, ribosome footprinting showed that genes involved in camalexin biosynthesis, CYP79B2 and CYP79B3, also increased translational efficiency under pattern triggered immunity (Xu et al., 2017; Yoo et al., 2020). Despite the rich knowledge of these upstream regulators of the camalexin biosynthetic pathway, it remains unknown how the rapid induction upon a pathogen signal is enabled.

The accessibility of target gene regions to transcription factors is determined by the dynamics of chromatin states in eukaryotic cells (Stricker et al., 2017). Chromatin states are controlled by epigenetic modifications that influence nucleosome accessibility (Vihervaara et al., 2018). Epigenetic modifications constitute various covalent decoration of chemical groups to histones and DNA, which are associated with promoting or repressing gene expression by altering chromatin accessibility to transcription factors (Stricker et al., 2017; Henikoff and Greally, 2016; Venkatesh and Workman, 2015). For example, trimethylation of lysine 27 of histone 3 (H3K27me3), established by the polycomb repressive complex 2, is associated with repressing gene expression (Carter et al., 2018). H3K27me3 represses gene expression by increasing chromatin condensation and limiting the recruitment of transcription factors and other components of the transcriptional machinery (Aranda et al., 2015). Trimethylation of lysine 4 of histone 3 (H3K4me3) is marked at actively transcribed genes (Zhang et al., 2009), which activates gene expression by promoting the recruitment of transcription initiation factors to promoters of target genes (Lauberth et al., 2013).

Epigenetic marks that play opposite roles on gene expression can co-localize at the same gene regions to form bivalent chromatins (Bernstein et al., 2006; Voigt et al., 2013). Bivalent chromatins were initially observed to be associated with genes involved in cell differentiation in human embryonic stem cells and the most reported variant is formed by H3K27me3 and H3K4me3 (Bernstein et al., 2006; Dattani et al., 2018; Sachs et al., 2013). Bivalent chromatins also exist in other mammalian cell types and can be formed by other epigenetic marks, including additional activating (H3K4me1, H3K4me2, H3K36me3, H3K9ac, H3K27ac, H3K14ac, H4K18ac) and repressive (H3K9me2, H3K9me3, H4K20me3, 5Mc) marks (van der Velde et al., 2021; Xu and Kidder, 2018; Charlet et al., 2016; Matsumura et al., 2015; Roy and Sridharan, 2017). In plants, the H3K27m3-H3K4me3 bivalent chromatin has been observed at FLOWERING LOCUS C (FLC) (Jiang et al., 2008) in A. thaliana and genes associated with cold stress in potato tuber (Zeng et al., 2019). Recently, a novel bivalent chromatin formed by H3K27me3 and H3K4me1 was identified in Brassica napus (Zhang et al., 2021). Therefore, bivalent chromatins can be formed by different pairs of epigenetic marks.

Despite many reports on the existence of bivalent chromatins, much less is known about their biological roles. The only function that has been proposed for a bivalent chromatin is to poise the expression of developmental genes in stem cells for rapid activation into differentiation (Bernstein et al., 2006). This hypothesis has been tested and supported by studies only on mammalian embryonic stem cells or cancer cells (Deng et al., 2013; Lima-Fernandes et al., 2019; Jia et al., 2012). The ‘poise genes for transcriptional activation’ role of bivalent chromatin has not yet been tested on a whole organism. And, to our knowledge, no other role has been implicated for bivalent chromatins.

Here, by integrating epigenomic profiles with a genome-wide metabolic network in Arabidopsis, we found that two epigenetic marks playing opposite roles on gene expression, H3K27me3 and H3K18ac, were enriched at camalexin biosynthesis and several other specialized metabolic pathways. Focusing on the camalexin biosynthesis pathway in this study, sequential chromatin immunoprecipitation (SeqChIP)-PCR confirmed that these two marks were co-localized in vivo and formed a previously uncharacterized type of bivalent chromatin. Mutants defective in H3K27m3 and H3K18ac suggested that both modifications were required to determine the timely induction of camalexin biosynthetic gene expression and camalexin accumulation under FLG22 treatment. This study revealed a new type of bivalent chromatin whose function may be to determine the timely induction of defense-compound biosynthesis in response to a pathogen signal.

Results

Camalexin biosynthesis pathway is one of the best characterized pathways associated with specialized metabolism in Arabidopsis (Stotz et al., 2011; Frerigmann et al., 2016). We identified epigenetic marks enriched in the camalexin biosynthesis pathway using an omics data-driven approach by integrating 16 publicly available epigenomic profiles with Arabidopsis genome-scale metabolic network annotation. High-resolution epigenomic profiles including histone variants, DNA methylation and histone modifications were obtained from seedlings grown under similar conditions (Wang et al., 2015; Luo et al., 2013; Roudier et al., 2011). To map these epigenetic marks on metabolic domains, pathways, and genes, we used genome-wide functional annotations of metabolism (Schläpfer et al., 2017). Enrichment analysis at the metabolic domain level revealed distinct patterns, including specialized metabolism being enriched with both a repression mark H3K27me3 and an activation mark H3K18ac (Figure 1A). Correlation of H3K27me3 and H3K18ac marks on the Arabidopsis genome had been reported previously (Luo et al., 2013), which we also observed using total metabolic genes (Pearson’s correlation coefficient 0.36, t-test p-value 4.6E-11) (Figure 1—figure supplement 1A). To determine if specialized metabolic genes drove the correlation of these two marks on metabolic genes, we removed specialized metabolic genes from the total metabolic genes and reanalyzed the correlation patterns. We observed decreased correlation between H3K27me3 and H3K18ac, indicating that specialized metabolic genes substantially contributed to the correlation of these two marks on metabolic genes (Pearson’s correlation coefficient 0.11, t-test p-value 3.5E-02) (Figure 1—figure supplement 1B). To determine how prevalent the H3K27me3-H3K18ac association was for specialized metabolic genes, we counted the number of genes marked by both modifications. We found that 37% (324 out of 887) of specialized metabolic genes were marked by both H3K27me3 and H3K81ac (Figure 1B). We then asked in which specialized metabolic pathways these two marks were significantly enriched by selecting the pathways that were annotated to specialized metabolism based on the Plant Metabolic Network database. To increase the statistical power, we only included the specialized metabolic pathways containing at least 10 genes, which resulted in 42 pathways. Among them, we found several pathways, including camalexin and glucosinolate biosynthesis pathways, to be significantly enriched with both epigenetic marks (Figure 1—figure supplement 2, Supplementary file 1). To decipher the nature and function of the H3K27me3-H3K18ac dual association on controlling gene expression, we focused our investigation on the camalexin biosynthetic genes.

Figure 1. Patterns of epigenetic modification across metabolism in Arabidopsis.

(A) Enrichment analysis shows different epigenetic modification patterns across metabolic domains. The heatmap represents log2 fold change (Log2FC) of enrichment or depletion of an epigenetic mark associated with each domain relative to total metabolic genes. Epigenetic modifications are color-coded based on their effect on gene expression; red represents activating mark and blue represents repressive mark. Genes only mapped to unique metabolic domains were included in this analysis. Significant enrichment or depletion is based on Fisher’s exact test, p-value < 0.05 and fold change >1.5. Gray cells represent no significant change. Black arrows indicate the patterns for trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac. (B) The co-occurrence of H3K27me3 and H3K18ac on total enzymatic genes in the genome and specialized metabolic genes.

Figure 1—source data 1. Source data used to generate Figure 1A, B.

Figure 1.

Figure 1—figure supplement 1. Correlation analysis of epigenetic marks based on their relative abundance on (A) total Arabidopsis metabolic genes, (B) non-specialized metabolic genes.

Figure 1—figure supplement 1.

Epigenetic modifications are color-coded based on their effect on gene expression; red represents activating mark and blue represents repressing mark. r represents Pearson’s correlation coefficient. The box highlighted by red represents Pearson’s correlation coefficient between H3K18ac and H3K27me3 . Gray cells represent insignificant correlation determined by the R package Hmisc.

Figure 1—figure supplement 2. Epigenetic marks associated with pathways involved in specialized metabolism including camalexin biosynthesis.

Figure 1—figure supplement 2.

The heatmap represents log2 fold change (Log2FC) of enrichment or depletion of a mark associated with each pathway relative to total metabolic genes based on Fisher’s exact test followed by a post hoc adjustment using false discovery rate (FDR). Gray cells represent no-significant enrichment or depletion at FDR of 0.01 and fold change of 1.5. The heatmap was generated using hierarchical clustering using the ggplot2 package version 3.1 in RStudio. Epigenetic modifications are color-coded based on their effect on gene expression; red represents activating marks and blue represents repressive marks.
Open in a new tab

We developed two mutually non-exclusive hypotheses about the role of H3K27me3 and H3K18ac on the expression of camalexin biosynthesis genes: (1) H3K27me3 and H3K18ac are co-localized on the same genes in the same cell and control their gene expression induction; or (2) H3K27me3 and H3K18ac are associated with the same genes but in different cells, which may contribute to determining cell-type specificity of this pathway. To distinguish between the two possibilities, we examined in vivo co-localization of the two modifications using SeqChIP-qPCR, which requires a two-step, serial chromatin pull-down with antibodies against these two modifications. Though more than 10 genes were annotated to camalexin biosynthesis pathway in the Plant Metabolic Network database, we focused on the three experimentally validated genes that are essential in synthesizing camalexin, CYP79B2, CYP71A13, and Phytoalexin Deficient 3 (PAD3) (Klein et al., 2013; Schuhegger et al., 2006Figure 2A). To assess the antibodies’ specificity and efficiency, we included several negative and positive controls, including a transcription factor-encoding gene called Golden-2-Like 1 (GLK) that was previously observed to be associated with both H3K27me3 and H3K18ac (Luo et al., 2013). All three camalexin biosynthesis genes showed significantly higher signals when pulled down with both antibodies against H3K27me3 and H3K18ac than without any antibody in the second pull-down (Figure 2B). We observed comparable pull-down efficiency for GLK between this study and the previous publication (Zeng et al., 2019). Altering the order of the two antibodies for H3K27me3 and H3K18ac in the pull-down showed similar patterns (Figure 2C). These results indicated that H3K27me3 and H3K18ac are co-localized at the camalexin biosynthesis genes in planta to form a bivalent chromatin.

Figure 2. H3K27me3 and H3K18ac are co-localized on camalexin biosynthesis genes in planta to form a bivalent chromatin.

Figure 2.

Open in a new tab

(A) Camalexin biosynthesis pathway map and the essential genes that can produce camalexin. (B-C) Sequential chromatin immunoprecipitation (ChIP)-qPCR confirms the co-localization of H3K27me3 and H3K18ac. Altering the order of antibodies against H3K27me3 or H3K18ac in two rounds of pull-downs shows similar results in sequential ChIP-qPCR. Negative control represents a genomic region that has low abundance of both H3K27me3 and H3K18ac. H3K27me3_only represents a genomic region that has high abundance of H3K27me3 and low abundance of H3K18ac. H3K18ac_only represents a genomic region that has high abundance of H3K18ac and low abundance of H3K27me3. NoAb represents no antibody in the second pull-down. Error bars represent standard deviation of data from six biological replicates. The experiments were performed twice with different plant samples and data were combined in this analysis.

Figure 2—source data 1. Source data used to generate Figure 2B,C.

The biological function of bivalent chromatins has long been hypothesized to poise gene expression for rapid activation upon signal perception. To date, this model has been tested only in mammalian stem cells and cancer cells (Zeng et al., 2019; Zhang et al., 2021; Deng et al., 2013), and no direct evidence is available in a whole organism context. As a first step to understand the role of the H3K27me3-H3K18ac bivalent chromatin on camalexin genes, we examined the transcriptional kinetics of camalexin biosynthesis genes under FLG22 induction in wild type and mutant lines with defective deposition of H3K27me3 or H3K18ac. We selected chromatin remodeler pickle (pkl-1) (Carter et al., 2018) and histone methyltransferase curly leaf 28 (clf28) (Doyle and Amasino, 2009) to study the function of H3K27me3 on gene expression. The protein complex involved in establishing H3K18ac has not been well characterized in Arabidopsis, which limits the genetic resources that can be utilized to study the function of this epigenetic modification. The available mutant lines associated with establishing H3K18ac are increased DNA methylation 1 and 2 (idm1 and idm2) (Qian et al., 2014; Qian et al., 2012). IDM1 is a histone acetyltransferase and IDM2 is a heat shock protein that functions in the same protein complex as IDM1 (Qian et al., 2014; Qian et al., 2012).

In wild-type plants, CYP71A13 and PAD3 were significantly induced within 30 min after FLG22 treatment and CYP79B2 was significantly induced within 1 hr after the treatment (Figure 3A-C, Supplementary file 3 to 5). In mutants with reduced deposition of H3K27me3 or H3K18ac, the induction patterns of the camalexin biosynthetic genes were disrupted. For pkl-1 and clf28 (reduced H3K27me3 marks), all three camalexin biosynthesis genes showed a significant induction of expression much faster than the wild type, within 5 min of FLG22 treatment, and the degree of induction was much higher than that in wild-type plants at 6 hr of the treatment (Figure 3A to C, Supplementary file 3 to 5). In contrast, in the lines with reduced H3K18ac marks, idm1 and idm2, genes were induced much later than the wild type. CYP71A13 and PAD3 were significantly induced within 1 hr after the treatment and CYP79B2 was induced within 3 hr after the treatment (Figure 3A to C, Supplementary file 3 to 5). To better understand the effect of genotype, FLG22 treatment, and duration of treatment on gene expression, we conducted a three-way ANOVA using raw transcripts of camalexin genes. Genotype, treatment, duration of treatment, and their interactions significantly contribute to the expression changes of CYP71A13 and PAD3 (Supplementary file 5). For CYP79B2, all experimental variables contributed significantly to its expression patterns except the genotype × treatment × duration of treatment (Supplementary file 5). These data generally support the idea that disrupting H3K27me3 and H3K18ac affects the temporal induction patterns of camalexin genes in response to FLG22 treatment.

Figure 3. H3K27me3-H3K18ac bivalent chromatin controls the timing of gene induction and camalexin accumulation upon a pathogen signal.

(A-C) Expression of the three essential genes in camalexin biosynthesis in response to flagellin 22 (FLG22) in the wild type (Col-0), H3K27me3-defective mutants clf28 and pkl-1, and H3K18ac-defective mutants idm1 and idm2. ACT2 was used as a reference gene. Relative gene expression at each time point was reported in Supplementary file 3 to 5. Fold change represents the induction level of camalexin biosynthesis genes in each genotype under FLG22 treatment relative to their corresponding untreated controls at each time point. (D) Camalexin accumulation in the wild type and mutants in response to FLG22. Camalexin content at each time point was reported in Supplementary file 6. Error bars represent standard deviation of data from six biological replicates of two experiments. (E-F) Abundance of epigenetic modifications at the genomic regions of camalexin biosynthesis genes with or without FLG22 treatment. Con-H3K18ac = chromatin was extracted from water-treated samples (Col-0) and pulled down by the antibody against H3K18ac. FLG-H3K18ac = chromatin was extracted from FLG22-treated samples (Col-0) and pulled down by the antibody against H3K18ac. Con-H3K27me3 = chromatin was extracted from water-treated samples (Col-0) and pulled down by the antibody against H3K27me3. FLG-H3K27me3 = chromatin was extracted from FLG22-treated samples (Col-0) and pulled down by the antibody against H3K27me3. Relative abundance of each modification was calculated by normalizing first to input, then to control plants with mock treatment. Two-week-old Arabidopsis seedlings of Col-0 were treated with 1 µM FLG22 or deionized water in these experiments. Stressed and control samples were collected 30 min after the treatment. Error bars represent standard deviation of data from six biological replicates of two experiments. * and ** represent p-values < 0.05 and 0.01, respectively, from Student’s two-tailed t-test. (G) A model of the H3K27me3-H3K18ac bivalent chromatin that acts as a kairostat to regulate the temporal pattern of gene expression in response to external stimuli. Upon a pathogen signal, bivalent chromatin controls the timing of induction for camalexin biosynthetic genes, with H3K18ac expediting and H3K27me3 attenuating expression to hit the presumed temporal ‘sweet spot’. Dashed lines represent the time of significant gene induction.

Figure 3—source data 1. Source data used to generate Figure 3A-F.

Figure 3.

Figure 3—figure supplement 1. Expression of CYP79B2 in the wild type and mutants in response to flagellin 22 (FLG22).

Figure 3—figure supplement 1.

Significance was tested via two-way ANOVA followed by post hoc Dunnett’s test (*p < 0.05, **p < 0.01, ***p < 0.0001) relative to mock-treated plants at each time point. Results represent the mean of two independent experiments with three technical replicates per sample per experiment.

Figure 3—figure supplement 2. Expression of CYP71A13 in the wild type and mutants in response to flagellin 22 (FLG22).

Figure 3—figure supplement 2.

Significance was tested via two-way ANOVA followed by post hoc Dunnett’s test (*p < 0.05, **p < 0.01, ***p < 0.0001) relative to mock-treated plants at each time point. Results represent the mean of two independent experiments with three technical replicates per sample per experiment.

Figure 3—figure supplement 3. Expression of PHYTOALEXIN DEFICIENT 3 (PAD3) in the wild type and mutants in response to flagellin 22 (FLG22).

Figure 3—figure supplement 3.

Significance was tested via two-way ANOVA followed by post hoc Dunnett’s test (*p < 0.05, **p < 0.01, ***p < 0.0001) relative to mock-treated plants at each time point. Results represent the mean of two independent experiments with three technical replicates per sample per experiment.

Figure 3—figure supplement 4. Camalexin accumulation in the wild type and mutants in response to flagellin 22 (FLG22).

Figure 3—figure supplement 4.

Significance was tested via two-way ANOVA by post hoc Dunnett’s test (*p < 0.05, **p < 0.01, ***p < 0.0001) relative to mock-treated plants at each time point. Results represent the mean of two independent experiments with three technical replicates per sample per experiment.
Open in a new tab

To test whether the changes in gene expression would lead to changes in metabolite accumulation, we measured camalexin content using liquid chromatography–tandem mass spectrometry (LC-MS/MS). For wild-type plants, camalexin accumulation significantly increased 3 hr after FLG22 treatment and continued to increase 6 hr after the treatment (Figure 3D, Supplementary file 6). Plants with reduced H3K27me3 marks, clf28 and pkl-1, started to accumulate more camalexin earlier, 1 and 3 hr after FLG22 treatment, and continued to increase at later time points. On the other hand, for mutants with reduced H3K18ac marks, idm1 and idm2, camalexin accumulated much later, at 6 hr after the treatment (Figure 3D, Supplementary file 6). We examined the effect of genotype, FLG22 treatment, and duration of the treatment on camalexin accumulation using three-way ANOVA (Supplementary file 6). Each of the three factors affected camalexin content significantly (p-values = 1.57E-05, 7.71E-09, 2.00E-16). However, the interaction between genotype and treatment and the three-way interaction, genotype × FLG22 × duration of treatment, were not significant (p-values = 6.20E-02 and 9.50E-01). One possibility is that the kinetics of metabolite changes might be different from transcripts since significant accumulation of camalexin was detected at later time points. Another possibility is metabolite data being noisier than gene expression data. In summary, the transcriptional kinetics analysis suggests that H3K27me3-H3K18ac bivalent chromatin might control the timing of camalexin production in response to FLG22, which is partially corroborated with metabolite accumulation patterns.

We next asked whether FLG22 treatment changed the relative amount of H3K27me3 and H3K18ac marks on the camalexin biosynthetic genes. To address this question, we quantified the abundance of H3K27me3 and H3K18ac on the camalexin biosynthetic genes with or without FLG22 treatment using ChIP-qPCR. For each gene tested, primers were designed to cover a genomic region that included 1 kb upstream and the entire transcribed region (Supplementary file 2). In plants treated with FLG22, the abundance of H3K18ac was significantly higher in the gene regions of CYP71A13, CYP79B2, and PAD3 compared to the mock-treated plants (Figure 3E). Moreover, the abundance of H3K27me3 modification was significantly lower in the regions of these genes in response to FLG22 (Figure 3F). These results suggest that the FLG22-induced changes of camalexin biosynthetic gene expression involve remodeling of the H3K27me3-H3K18ac bivalent chromatin. Taken together, this study demonstrates that a novel bivalent chromatin formed by H3K27me3 and H3K18ac controls the timing of induction of camalexin biosynthesis genes upon stress, with H3K18ac expediting and H3K27me3 attenuating expression to hit the presumed ‘sweet spot’ (Figure 3G). We have named this type of bivalent chromatin regulator a kairostat, inspired by the ancient Greek word ‘kairos’ meaning the right moment and ‘stat’ meaning regulating device.

Discussion

Plant specialized metabolites are integral for plant-environmental interactions and defense against herbivores and pathogens. Camalexin is a prominent phytoalexin in Arabidopsis and genes associated with its biosynthesis are induced rapidly upon pathogen signals such as FLG22 (Denoux et al., 2008; Stahl et al., 2016; Zipfel et al., 2004). In this study, we found that H3K27me3 and H3K18ac co-localized on camalexin biosynthesis genes forming a previously uncharacterized form of bivalent chromatin. We examined the role of H3K27me3-H3K18ac bivalent chromatin on the induction of camalexin biosynthesis genes by combining publicly available epigenomics data with genetics, molecular biology, and biochemistry. In wild-type (Col-0) plants, camalexin biosynthesis genes were activated at 30 min to 1 hr of FLG22 treatment and continued to increase their expression at later time points of the treatment. This is consistent with the previous study on induction kinetics of defense pathways, including camalexin, under FLG22 treatment (Denoux et al., 2008). In mutants with defective deposition of H3K27me3 or H3K18ac, the induction kinetics of camalexin biosynthesis genes and metabolite accumulation was accelerated or delayed, respectively.

Our results provide new evidence for how chemical defense mediated by camalexin may be regulated at the epigenetic level. However, we cannot rule out the possibility that other known mechanisms regulating camalexin genes may also affect transcriptional kinetics and metabolite accumulation. For example, H3K27me3 affects gene expression by altering chromatin accessibility to transcription factors (Aranda et al., 2015). Removing this repression mark may create a permissive environment and facilitate transcription factors, such as WRKY33, to bind to promoters of camalexin genes. At the translational level, camalexin biosynthesis genes can alter translational efficiency controlled by a highly enriched messenger RNA consensus sequence, R-motif, during pattern triggered immunity (Xu et al., 2017; Yoo et al., 2020). Additional studies are needed to unravel how different regulatory machineries work together to enable the rapid induction of camalexin genes upon stress signals.

In this study, we attempted to reveal the biological function of bivalent chromatin, which has long been hypothesized to poise developmental genes in embryonic stem cells for rapid activation upon a cell differentiation signal (Zhang et al., 2009). Only few studies directly tested this hypothesis and all of them focused on mammalian stem cells or cancer cells (Deng et al., 2013; Lima-Fernandes et al., 2019; Jia et al., 2012). For example, in embryonic stem cells, H3K27me3-H3K4me3 bivalent chromatin was associated with HoxB4, one of the key regulators controlling development. Inhibition of H3K4 methyltransferase, hSET1A, repressed the expression of HoxB4 and blocked the differentiation of blood cells (Deng et al., 2013). In addition, H3K27me3-H3K4me3 bivalent chromatin was observed in cancer-initiating cells. Disruption of the bivalent state through inhibition of the H3K27 methyltransferase EZH2 inhibited the self-renewal of cancer cells through de-repression of a key canonical marker of normal colonocyte differentiation, named Indian Hedgehog (Lima-Fernandes et al., 2019). Despite this knowledge, the role of bivalent chromatin on regulating metabolism remains unknown. In this study, we reported the co-localization of H3K27me3 and H3K18ac on camalexin biosynthesis genes and functionally examined the role of H3K27me3-H3K18ac bivalent chromatin in Arabidopsis. The results suggest that H3K27me3-H3K18ac bivalent chromatin maintains the proper timing of camalexin gene induction upon a pathogen signal.

To fully elucidate how bivalent chromatin integrates environmental cues to regulate gene expression, it would be critical to understand how bivalent chromatin is established and maintained and how different epigenetic marks function cooperatively to control gene expression. To achieve these goals, it is required to characterize all the proteins involved in establishing different epigenetic marks. Mechanisms that enable histone modification protein complexes to alter the abundance of each mark upon stress signals also remains to be elucidated. A recent study showed that NF-Y transcription factors can interact with histone modification protein REF6 to reduce H3K27me3 at the SOC1 locus (Hou et al., 2014). It would be interesting to examine whether transcription factors that directly bind to promoters of camalexin biosynthesis genes interact with proteins that can deposit or remove both epigenetic marks forming bivalent chromatin upon stress signals. The H3K27me3-H3K18ac bivalent chromatin and its role on the camalexin biosynthesis genes described here provide the molecular handle with which to open these lines of investigations in the future. With these knowledge gaps filled, kairostats can be engineered to function as a regulatory cassette to control the expression of target genes under specific tissues and conditions. This capability will provide novel routes to re-program biological processes involved in development and stress response, which have far-reaching implications in agriculture, engineering, and medicine.

Materials and methods

Annotation of metabolic genes, pathways, and domains

Arabidopsis metabolic enzymes and pathways were extracted from Plant Metabolic Network (http://www.plantcyc.org/). Enzymes were annotated based on the Ensemble Enzyme Prediction Pipeline (E2P2) v.3.0 (Schläpfer et al., 2017; Chae et al., 2014), metabolic pathways were predicted by Pathway Tools software v.18.5 (Karp et al., 2011), which were further validated by a process called Semi-Automated Validation and Integration (SAVI) v.3.02 (Zhang et al., 2010). In A. thaliana, 5451 genes were predicted to encode enzymes that catalyze 3585 reactions in small molecule metabolism, which were assigned to 627 pathways based on prediction and manual curation in the release of PMN 13.0 (Schläpfer et al., 2017). These pathways were further grouped into 13 metabolic domains based on the types of metabolites that they potentially synthesize (Schläpfer et al., 2017). The metabolic domain(s) assigned to a given pathway were transferred to the reactions, EC numbers, and proteins/genes associated with that pathway. To improve the statistical power, we included only those pathways that contained more than 10 genes when we performed epigenetic modification enrichment analysis on individual pathways in specialized metabolism (Sham and Purcell, 2014).

Epigenomic enrichment analysis

The 16 epigenomic profiles were downloaded from Wang et al., 2015, where previously published ChIP-seq and ChIP-chip data for all epigenetic marks were re-mapped to A. thaliana genome TAIR v.10 (Wang et al., 2015). Briefly, for each ChIP-seq profile, the signals were binned and ranked in the genome (Wang et al., 2015). We used the ranked epigenetic modification signals in the following analyses.

To identify the predominant epigenetic modifications associated with metabolic domains and pathways, we first mapped the ranked epigenetic modification signals to each gene, including the transcribed region, 1 kb upstream, and 500 bp downstream regions. Then, we calculated the average epigenetic modification density for each domain or pathway by taking the sum of the ranked modification signals observed within all the genes in that domain or pathway and normalizing it to the total length of all genes. To identify enriched marks within metabolic domains or pathways, we compared the density of each epigenetic modification within that domain or pathway to the background signal. We defined the background signal as the average density of each epigenetic modification by taking the sum of the epigenetic modification signals observed within total metabolic gene regions and normalizing it to the length of these genes.

The statistical significance of enrichment or depletion of epigenetic marks per domain or pathway was determined by fold change of at least 1.5 and Fisher’s exact test p-value, followed by a post hoc adjustment using false discovery rate (FDR) with the threshold of 0.01. The heatmaps were generated with gplots v.3 (Warnes, 2009) and ggplot2 v.3.1 (Wickham, 2016) packages in RStudio.

Plant materials and growth conditions

Seeds of A. thaliana accession Columbia (Col-0) (wild type) and previously characterized mutant lines clf28 (SALK_139371), pkl-1 (SALK_010693), idm1 (SALK_062999), and idm2 (SALK_138229) were used in this study. Seeds of these mutants were obtained from the Arabidopsis Biological Resource Center. In all the experiments described in this study, seeds were stratified at 4°C for 3 days before germination and grown in 0.5× Murashige and Skoog (MS) medium containing 1% sucrose and 0.8% agar at 23°C under long days (16 hr light/8 hr dark) with controlled light at 100 µmol photosynthetic photons/m2 s (PPFD). Two-week-old seedlings were harvested for the ChIP-qPCR, SeqChIP-PCR, transcriptional kinetics, and camalexin content measurement experiments.

ChIP-qPCR and SeqChIP-PCR

Two-week-old Arabidopsis seedlings (Col-0) were used in ChIP and SeqChIP (sequential ChIP) experiments. To obtain sufficient chromatin for immunoprecipitation, 200 seedlings (about 2 g) were pooled for each ChIP reaction. Plant tissues were fixed with crosslinking buffer containing 1% formaldehyde for 20 min by vacuum infiltration at room temperature and quenched with glycine (final concentration 100 mM) for 5 min. After grinding, the cross-linked chromatin was isolated using ice-cold nuclear lysis buffer (50 mM HEPES at pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% Na deoxycholate, 0.1% SDS, and protease inhibitor) and disrupted by sonication to yield DNA fragments of 200–600 bp (Zhu et al., 2016; Saleh et al., 2008).

To examine the abundance of H3K27me3 and H3K18ac 30 min after 1 µM FLG22 treatment, ChIP-qPCR was used to quantify their abundance throughout the genomic region and 1 kb upstream of the three essential camalexin biosynthesis genes. After chromatin crosslinking and shearing, DNA/protein complexes were equally divided and immunoprecipitated with antibodies against H3K27me3 (Millipore, Cat No 07–449), H3K18ac (Abcam, Cat No ab1191), or IgG as antibody control (Millipore, Cat No 12–370) by referring the ChIP protocol established using Arabidopsis tissues (Zhu et al., 2016; Saleh et al., 2008). DNA was purified using QIAquick PCR Purification Kit (Qiagen, Cat No 28104). Quantitative real-time PCR was performed using LightCycler 480 System (Roche) with the SensiFAST Sybr No-Rox mix (Bioline, Cat No BIO98020) from Bioline. Three biological replicates were used in each experiment. The ChIP experiments were performed twice with different Arabidopsis tissues and revealed similar results. Primers were designed to cover the 1 kb upstream of the transcription start site and the entire transcribed region (Supplementary file 2).

SeqChIP experiments were performed to examine in planta co-localization of H3K27me3 and H3K18ac on camalexin biosynthesis genes. Chromatin crosslinking and shearing followed the protocol described above for ChIP-qPCR. For each ChIP/re-ChIP assay, 20 µg DNA/protein complexes were immunoprecipitated with anti-H3K27me3 (Millipore, Cat No 07–449) and anti-H3K18ac (Abcam, Cat No ab1191) antibodies using the Re-ChIP-IT kit (Active Motif, Cat No 53016). Quantitative real-time PCR was performed using LightCycler 480 System (Roche) with the SensiFAST Sybr No-Rox mix (Bioline, Cat No BIO98020) from Bioline. To represent final results after two-step chromatin immunoprecipitation, the pull-down efficiency was calculated as the percent of first ChIP using the equation: 1st ChIP = 2^(Cq_2nd ChIP – Cq_1st ChIP). Three biological replicates were used in each experiment. The SeqChIP was performed twice using different Arabidopsis tissues and altering the order of antibodies in two rounds of pull-down showed similar results. Primer sequences used in this assay are provided in Supplementary file 3.

Transcriptional kinetics analysis

Two-week-old seedlings grown in MS agar media as described in the Plant materials and growth conditions section were used to study transcriptional kinetics. FLG22, a short peptide of 22 amino acids (sequence: QRLSTGSRINSAKDDAAGLQIA), was synthesized at Stanford Protein and Nucleic Acid (PAN) Facility. To induce the expression of camalexin biosynthesis genes, seedlings were sprayed with 1 µM FLG22 dissolved in deionized (DI) water. Control samples were sprayed with the same volume of DI water. FLG22 and water-treated tissues were sampled at 5 min, 30 min, 1 hr, 3 hr, and 6 hr after treatment and frozen immediately in liquid nitrogen.

After grinding in liquid nitrogen, RNA was extracted using RNeasy Plant Mini Kit (Qiagen, Cat No 74904) and 2 µg RNA was used for cDNA synthesis using SuperScript First-Strand Synthesis System (Thermo Fisher, Cat No 18080051). Quantitative real-time PCR was performed using LightCycler 480 System (Roche) with the SensiFAST Sybr No-Rox mix (Bioline, Cat No BIO98020) from Bioline. Primers used in this experiment are summarized in Supplementary file 4. A housekeeping gene, Actin 2 (AT3G18780), was used as the reference gene in the qPCR experiments. Three biological replicates were used in each experiment by polling 10 seedlings per replicate. The transcriptional kinetics assays were performed twice and showed similar results.

Camalexin content quantification using LC-MS

Two-week-old seedlings of Col-0, pkl-1, clf28, idm1, and idm2 plants were used to measure camalexin accumulation. Tissues were treated with FLG22 or DI water as described in the "Transcriptional kinetics analysis" section. Metabolites were extracted using a 75% methanol solution with 0.1% formic acid. For each sample, 150 mg of ground plant tissue was dissolved in 200 µL extraction solution. The extracted samples were vortexed at room temperature for 40 min and centrifuged at 32,000 g for 10 min to pellet plant debris. The supernatant was analyzed by an Agilent 1260 Infinity Binary HPLC paired with an Agilent 6420 Triple Quadrupole LC-MS/MS, with a reversed-phase column (Agilent EclipsePlus C18, 2.1 × 50 mm, 1.8 μm), water with 0.1% formic acid as solvent A and acetonitrile with 0.1% formic acid as solvent B, at a constant flow rate of 0.4 mL/min and an injection volume of 10 µL. The following gradient was used for metabolite separation: 0–0.5 min, 10–50% B; 0.5–5.5 min, 50–98% B; 5.5–6 min, 98% B; 6.00–6.01 min, 98–10% B; 2 min post-time equilibration with 10% B. The LC eluent was subjected to MS for 0.01–6.01 min with the ESI source in positive mode, gas temperature at 350°C, gas flow rate at 11 L/min, and nebulizer pressure at 40 psi. To detect camalexin, we set the multiple reaction monitoring (MRM) ion transition to be 201.26 → 59.2, fragmentor at 135 V, and collision energy at 40 V. The MRM ion transitions in this work were derived from product ion scan with precursor ion set at 201.26 and the most abundant product ion was chosen for MRM transition quantification. LC-MS/MS data files were analyzed using Agilent MassHunter Workstation software v.B.06.00. Metabolite content was quantified by integrating peak area under the ion count curve. The ion counts were calibrated to camalexin chemical standard (Cat Num Sigma SML1016) and converted to measurements of titer in molar concentrations (nM). Three biological replicates were used in each experiment by polling 50 seedlings per replicate. The experiments were conducted twice with different plant samples and showed similar results.

Acknowledgements

We thank Jennifer Brophy and Flavia Bossi for their critical feedback on this work. We acknowledge Jia-Ying Zhu and Yuchun Hsiao for help with ChIP-qPCR experiments and Hye-In Nam for help with growing plants. Funding: This work was supported in part by Carnegie Institution for Science Endowment and grants from the National Science Foundation (IOS-1546838, IOS-1026003), the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Genomic Science Program grant nos. DE-SC0018277, DE-SC0008769, and DE-SC0020366, and the National Institutes of Health (1U01GM110699-01A1). This work was done on the ancestral land of the Muwekma Ohlone Tribe, which was and continues to be of great importance to the Ohlone people.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Seung Yon Rhee, Email: srhee@CarnegieScience.edu.

Nicola J Patron, The Earlham Institute, United Kingdom.

Jessica K Tyler, Weill Cornell Medicine, United States.

Funding Information

This paper was supported by the following grants:

  • National Science Foundation IOS-1546838 and IOS-1026003 to Kangmei Zhao, Benjamin Jin, Seung Yon Rhee.

  • National Institute for Health Research 1U01GM110699-01A1 to Kangmei Zhao, Deze Kong, Christina D Smolke, Seung Yon Rhee.

  • U.S. Department of Energy DE-SC0018277 to Kangmei Zhao, Seung Yon Rhee.

  • Carnegie Institution for Science Endowment to Kangmei Zhao, Benjamin Jin, Seung Yon Rhee.

  • U.S. Department of Energy DE-SC0008769 to Seung Yon Rhee.

  • U.S. Department of Energy DE-SC0020366 to Seung Yon Rhee.

Additional information

Competing interests

No competing interests declared.

Author contributions

Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing - original draft, Writing - review and editing.

Formal analysis, Investigation, Methodology, Writing - original draft, Writing - review and editing.

Investigation.

Supervision, Writing - review and editing.

Funding acquisition, Supervision, Writing - original draft, Writing - review and editing.

Additional files

Supplementary file 1. Log2 fold change in the enrichment analysis to identify specialized metabolic pathways associated with both trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac.
elife-69508-supp1.xlsx (12.6KB, xlsx)
Supplementary file 2. Primers used to quantify the abundance of trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac in the genomic regions of camalexin biosynthesis genes using chromatin immunoprecipitation (ChIP)-qPCR in wild-type and mutant plants with or without flagellin 22 (FLG22) treatment.
elife-69508-supp2.docx (16.9KB, docx)
Supplementary file 3. Primers used in the sequential chromatin immunoprecipitation (ChIP)-qPCR experiment to examine the co-localization of trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac within camalexin biosynthesis genes.
elife-69508-supp3.docx (14.3KB, docx)
Supplementary file 4. Primers used to examine the expression of camalexin biosynthesis genes under flagellin 22 (FLG22) induction using qPCR.
elife-69508-supp4.docx (13.8KB, docx)
Supplementary file 5. The effect of genotype, flagellin 22 (FLG22) treatment, and time point on the expression change of camalexin genes.
elife-69508-supp5.docx (16.2KB, docx)
Supplementary file 6. The effect of genotype, flagellin 22 (FLG22) treatment, and time on camalexin accumulation.
elife-69508-supp6.docx (15.1KB, docx)
Transparent reporting form

Data availability

The data that support the findings of this study are available within the article and its supplementary information files. Arabidopsis metabolic genes and pathways are available at Plant Metabolic Network (https://plantcyc.org/).

References

  1. Aranda S, Mas G, Di Croce L. Regulation of gene transcription by Polycomb proteins. Science Advances. 2015;1:e1500737. doi: 10.1126/sciadv.1500737. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Barco B, Clay NK. Hierarchical and Dynamic Regulation of Defense-Responsive Specialized Metabolism by WRKY and MYB Transcription Factors. Frontiers in Plant Science. 2019;10:1775. doi: 10.3389/fpls.2019.01775. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bernstein BE, Mikkelsen TS, Xie X, Kamal M, Huebert DJ, Cuff J, Fry B, Meissner A, Wernig M, Plath K, Jaenisch R, Wagschal A, Feil R, Schreiber SL, Lander ES. A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell. 2006;125 doi: 10.1016/j.cell.2006.02.041. [DOI] [PubMed] [Google Scholar]
  4. Birkenbihl RP, Diezel C, Somssich IE. Arabidopsis WRKY33 is a key transcriptional regulator of hormonal and metabolic responses toward Botrytis cinerea infection. Plant Physiology. 2012;159:266–285. doi: 10.1104/pp.111.192641. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Carter B, Bishop B, Ho KK, Huang R, Jia W, Zhang H, Pascuzzi PE, Deal RB, Ogas J. The Chromatin Remodelers PKL and PIE1 Act in an Epigenetic Pathway That Determines H3K27me3 Homeostasis in Arabidopsis. The Plant Cell. 2018;30:1337–1352. doi: 10.1105/tpc.17.00867. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Chae L, Kim T, Nilo-Poyanco R, Rhee SY. Genomic signatures of specialized metabolism in plants. Science. 2014;344:510–513. doi: 10.1126/science.1252076. [DOI] [PubMed] [Google Scholar]
  7. Charlet J, Duymich CE, Lay FD, Mundbjerg K, Dalsgaard Sørensen K, Liang G, Jones PA. Bivalent Regions of Cytosine Methylation and H3K27 Acetylation Suggest an Active Role for DNA Methylation at Enhancers. Molecular Cell. 2016;62:422–431. doi: 10.1016/j.molcel.2016.03.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Dattani A, Kao D, Mihaylova Y, Abnave P, Hughes S, Lai A, Sahu S, Aboobaker AA. Epigenetic analyses of planarian stem cells demonstrate conservation of bivalent histone modifications in animal stem cells. Genome Research. 2018;28:1543–1554. doi: 10.1101/gr.239848.118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Deng C, Li Y, Liang S, Cui K, Salz T, Yang H, Tang Z, Gallagher PG, Qiu Y, Roeder R, Zhao K, Bungert J, Huang S. USF1 and HSET1A mediated epigenetic modifications regulate lineage differentiation and hoxb4 transcription. PLOS Genetics. 2013;9:e1003524. doi: 10.1371/journal.pgen.1003524. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Denoux C, Galletti R, Mammarella N, Gopalan S, Werck D, De Lorenzo G, Ferrari S, Ausubel FM, Dewdney J. Activation of defense response pathways by OGs and Flg22 elicitors in Arabidopsis seedlings. Molecular Plant. 2008;1 doi: 10.1093/mp/ssn019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Doyle MR, Amasino RM. A single amino acid change in the enhancer of zeste ortholog CURLY LEAF results in vernalization-independent, rapid flowering in Arabidopsis. Plant Physiology. 2009;151:1688–1697. doi: 10.1104/pp.109.145581. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Frerigmann H, Glawischnig E, Gigolashvili T. The role of MYB34, MYB51 and MYB122 in the regulation of camalexin biosynthesis in Arabidopsis thaliana. Frontiers in Plant Science. 2015;6:654. doi: 10.3389/fpls.2015.00654. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Frerigmann H, Piślewska-BednarekBednarek P, Sánchez-Vallet A, Molina A, Glawischnig E, Gigolashvili T. Regulation of Pathogen-Triggered Tryptophan Metabolism in Arabidopsis thaliana by MYB Transcription Factors and Indole Glucosinolate Conversion Products. Molecular Plant. 2016;9:682–695. doi: 10.1016/j.molp.2016.01.006. [DOI] [PubMed] [Google Scholar]
  14. Grotewold E. Plant metabolic diversity: a regulatory perspective. Trends in Plant Science. 2005;10:57–62. doi: 10.1016/j.tplants.2004.12.009. [DOI] [PubMed] [Google Scholar]
  15. Henikoff S, Greally JM. Epigenetics, cellular memory and gene regulation. Current Biology. 2016;26:R644–R648. doi: 10.1016/j.cub.2016.06.011. [DOI] [PubMed] [Google Scholar]
  16. Hou X, Zhou J, Liu C, Liu L, Shen L, Yu H. Nuclear factor Y-mediated H3K27me3 demethylation of the SOC1 locus orchestrates flowering responses of Arabidopsis. Nature Communications. 2014;5:4601. doi: 10.1038/ncomms5601. [DOI] [PubMed] [Google Scholar]
  17. Jia J, Zheng X, Hu G, Cui K, Zhang J, Zhang A, Jiang H, Lu B, Yates J, Liu C, Zhao K, Zheng Y. Regulation of pluripotency and self- renewal of ESCs through epigenetic-threshold modulation and mRNA pruning. Cell. 2012;151 doi: 10.1016/j.cell.2012.09.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Jiang D, Wang Y, Wang Y, He Y. Repression of flowering LOCUS C and flowering LOCUS T by the Arabidopsis Polycomb repressive complex 2 components. PLOS ONE. 2008;3:e3404. doi: 10.1371/journal.pone.0003404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Karp PD, Latendresse M, Caspi R. The pathway tools pathway prediction algorithm. Standards in Genomic Sciences. 2011;5:424–429. doi: 10.4056/sigs.1794338. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Klein AP, Anarat-Cappillino G, Sattely ES. Minimum set of cytochromes P450 for reconstituting the biosynthesis of camalexin, a major Arabidopsis antibiotic. Angewandte Chemie. 2013;52:13625–13628. doi: 10.1002/anie.201307454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lauberth SM, Nakayama T, Wu X, Ferris AL, Tang Z, Hughes SH, Roeder RG. H3K4me3 interactions with TAF3 regulate preinitiation complex assembly and selective gene activation. Cell. 2013;152 doi: 10.1016/j.cell.2013.01.052. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lima-Fernandes E, Murison A, da Silva Medina T, Wang Y, Ma A, Leung C, Luciani GM, Haynes J, Pollett A, Zeller C, Duan S, Kreso A, Barsyte-Lovejoy D, Wouters BG, Jin J, Carvalho DDD, Lupien M, Arrowsmith CH, O’Brien CA. Targeting bivalency de-represses Indian Hedgehog and inhibits self-renewal of colorectal cancer-initiating cells. Nature Communications. 2019;10:1436. doi: 10.1038/s41467-019-09309-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Luo C, Sidote DJ, Zhang Y, Kerstetter RA, Michael TP, Lam E. Integrative analysis of chromatin states in Arabidopsis identified potential regulatory mechanisms for natural antisense transcript production. The Plant Journal. 2013;73:77–90. doi: 10.1111/tpj.12017. [DOI] [PubMed] [Google Scholar]
  24. Mao G, Meng X, Liu Y, Zheng Z, Chen Z, Zhang S. Phosphorylation of a WRKY transcription factor by two pathogen-responsive MAPKs drives phytoalexin biosynthesis in Arabidopsis. The Plant Cell. 2011;23:1639–1653. doi: 10.1105/tpc.111.084996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Martin C, Ellis N, Rook F. Do transcription factors play special roles in adaptive variation. Plant Physiology. 2010;154:506–511. doi: 10.1104/pp.110.161331. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Matsumura Y, Nakaki R, Inagaki T, Yoshida A, Kano Y, Kimura H, Tanaka T, Tsutsumi S, Nakao M, Doi T, Fukami K, Osborne TF, Kodama T, Aburatani H, Sakai J. H3K4/H3K9me3 Bivalent Chromatin Domains Targeted by Lineage-Specific DNA Methylation Pauses Adipocyte Differentiation. Molecular Cell. 2015;60 doi: 10.1016/j.molcel.2015.10.025. [DOI] [PubMed] [Google Scholar]
  27. Qian W, Miki D, Zhang H, Liu Y, Zhang X, Tang K, Kan Y, La H, Li X, Li S, Zhu X, Shi X, Zhang K, Pontes O, Chen X, Liu R, Gong Z, Zhu JK. A histone acetyltransferase regulates active DNA demethylation in Arabidopsis. Science. 2012;336 doi: 10.1126/science.1219416. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Qian W, Miki D, Lei M, Zhu X, Zhang H, Liu Y, Li Y, Lang Z, Wang J, Tang K, Liu R, Zhu JK. Regulation of active DNA demethylation by an α-crystallin domain protein in Arabidopsis. Molecular Cell. 2014;55 doi: 10.1016/j.molcel.2014.06.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Roudier F, Ahmed I, Bérard C, Sarazin A, Mary-Huard T, Cortijo S, Bouyer D, Caillieux E, Duvernois-Berthet E, Al-Shikhley L, Giraut L, Després B, Drevensek S, Barneche F, Dèrozier S, Brunaud V, Aubourg S, Schnittger A, Bowler C, Martin-Magniette ML, Robin S, Caboche M, Colot V. Integrative epigenomic mapping defines four main chromatin states in Arabidopsis. The EMBO Journal. 2011;30 doi: 10.1038/emboj.2011.103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Roy S, Sridharan R. Chromatin module inference on cellular trajectories identifies key transition points and poised epigenetic states in diverse developmental processes. Genome Research. 2017;27:1250–1262. doi: 10.1101/gr.215004.116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Sachs M, Onodera C, Blaschke K, Ebata KT, Song JS, Ramalho-Santos M. Bivalent chromatin marks developmental regulatory genes in the mouse embryonic germline in vivo. Cell Reports. 2013;3:1777–1784. doi: 10.1016/j.celrep.2013.04.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Saleh A, Alvarez-Venegas R, Avramova Z. An efficient chromatin immunoprecipitation (ChIP) protocol for studying histone modifications in Arabidopsis plants. Nature Protocols. 2008;3:1018–1025. doi: 10.1038/nprot.2008.66. [DOI] [PubMed] [Google Scholar]
  33. Schläpfer P, Zhang P, Wang C, Kim T, Banf M, Chae L, Dreher K, Chavali AK, Nilo-Poyanco R, Bernard T, Kahn D, Rhee SY. Genome-Wide Prediction of Metabolic Enzymes, Pathways, and Gene Clusters in Plants. Plant Physiology. 2017;173:2041–2059. doi: 10.1104/pp.16.01942. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Schuhegger R, Nafisi M, Mansourova M, Petersen BL, Olsen CE, Svatos A, Halkier BA, Glawischnig E. CYP71B15 (PAD3) catalyzes the final step in camalexin biosynthesis. Plant Physiology. 2006;141:1248–1254. doi: 10.1104/pp.106.082024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sham PC, Purcell SM. Statistical power and significance testing in large-scale genetic studies. Nature Reviews. Genetics. 2014;15:335–346. doi: 10.1038/nrg3706. [DOI] [PubMed] [Google Scholar]
  36. Stahl E, Bellwon P, Huber S, Schlaeppi K, Bernsdorff F, Vallat-Michel A, Mauch F, Zeier J. Regulatory and Functional Aspects of Indolic Metabolism in Plant Systemic Acquired Resistance. Molecular Plant. 2016;9:662–681. doi: 10.1016/j.molp.2016.01.005. [DOI] [PubMed] [Google Scholar]
  37. Stotz HU, Sawada Y, Shimada Y, Hirai MY, Sasaki E, Krischke M, Brown PD, Saito K, Kamiya Y. Role of camalexin, indole glucosinolates, and side chain modification of glucosinolate-derived isothiocyanates in defense of Arabidopsis against Sclerotinia sclerotiorum. The Plant Journal. 2011;67:81–93. doi: 10.1111/j.1365-313X.2011.04578.x. [DOI] [PubMed] [Google Scholar]
  38. Stricker SH, Koferle A, Beck S. From profiles to function in epigenomics. Nature Reviews. Genetics. 2017;18:51–66. doi: 10.1038/nrg.2016.138. [DOI] [PubMed] [Google Scholar]
  39. van der Velde A, Fan K, Tsuji J, Moore JE, Purcaro MJ, Pratt HE, Weng Z. Annotation of chromatin states in 66 complete mouse epigenomes during development. Communications Biology. 2021;4:239. doi: 10.1038/s42003-021-01756-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Venkatesh S, Workman JL. Histone exchange, chromatin structure and the regulation of transcription. Nature Reviews. Molecular Cell Biology. 2015;16:178–189. doi: 10.1038/nrm3941. [DOI] [PubMed] [Google Scholar]
  41. Vihervaara A, Duarte FM, Lis JT. Molecular mechanisms driving transcriptional stress responses. Nature Reviews. Genetics. 2018;19:385–397. doi: 10.1038/s41576-018-0001-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Voigt P, Tee WW, Reinberg D. A double take on bivalent promoters. Genes & Development. 2013;27:1318–1338. doi: 10.1101/gad.219626.113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Wang C, Liu C, Roqueiro D, Grimm D, Schwab R, Becker C, Lanz C, Weigel D. Genome-wide analysis of local chromatin packing in Arabidopsis thaliana. Genome Research. 2015;25 doi: 10.1101/gr.170332.113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Warnes GR. Gplots: Various R programming tools for plotting data. R Package. 2009 https://www.scienceopen.com/document?vid=0e5d8e31-1fe4-492f-a3d8-8cd71b2b8ad9
  45. Wickham H. Ggplot2: Elegant Graphics for Data Analysis. Springer; 2016. [DOI] [Google Scholar]
  46. Xu G, Greene GH, Yoo H, Liu L, Marqués J, Motley J, Dong X. Global translational reprogramming is a fundamental layer of immune regulation in plants. Nature. 2017;545:487–490. doi: 10.1038/nature22371. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Xu J, Kidder BL. H4K20ME3 co-localizes with activating histone modifications at transcriptionally dynamic regions in embryonic stem cells. BMC Genomics. 2018;19:514. doi: 10.1186/s12864-018-4886-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Yoo H, Greene GH, Yuan M, Xu G, Burton D, Liu L, Marqués J, Dong X. Translational Regulation of Metabolic Dynamics during Effector-Triggered Immunity. Molecular Plant. 2020;13:88–98. doi: 10.1016/j.molp.2019.09.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Zeng Z, Zhang W, Marand AP, Zhu B, Buell CR, Jiang J. Cold stress induces enhanced chromatin accessibility and bivalent histone modifications H3K4me3 and H3K27me3 of active genes in potato. Genome Biology. 2019;20:123. doi: 10.1186/s13059-019-1731-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Zhang X, Bernatavichute YV, Cokus S, Pellegrini M, Jacobsen SE. Genome-wide analysis of mono-, di- and trimethylation of histone H3 lysine 4 in Arabidopsis thaliana. Genome Biology. 2009;10:R62. doi: 10.1186/gb-2009-10-6-r62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Zhang P, Dreher K, Karthikeyan A, Chi A, Pujar A, Caspi R, Karp P, Kirkup V, Latendresse M, Lee C, Mueller LA, Muller R, Rhee SY. Creation of a genome-wide metabolic pathway database for Populus trichocarpa using a new approach for reconstruction and curation of metabolic pathways for plants. Plant Physiology. 2010;153:1479–1491. doi: 10.1104/pp.110.157396. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Zhang Q, Guan P, Zhao L, Ma M, Xie L, Li Y, Zheng R, Ouyang W, Wang S, Li H, Zhang Y, Peng Y, Cao Z, Zhang W, Xiao Q, Xiao Y, Fu T, Li G, Li X, Shen J. Asymmetric Epigenome Maps of Subgenomes Reveal Imbalanced Transcription and Distinct Evolutionary Trends in Brassica napus. Molecular Plant. 2021;14:604–619. doi: 10.1016/j.molp.2020.12.020. [DOI] [PubMed] [Google Scholar]
  53. Zhou J, Wang X, He Y, Sang T, Wang P, Dai S, Zhang S, Meng X. Differential Phosphorylation of the Transcription Factor WRKY33 by the Protein Kinases CPK5/CPK6 and MPK3/MPK6 Cooperatively Regulates Camalexin Biosynthesis in Arabidopsis. The Plant Cell. 2020;32:2621–2638. doi: 10.1105/tpc.19.00971. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Zhu JY, Oh E, Wang T, Wang ZY. TOC1-PIF4 interaction mediates the circadian gating of thermoresponsive growth in Arabidopsis. Nature Communications. 2016;7:13692. doi: 10.1038/ncomms13692. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Zipfel C, Robatzek S, Navarro L, Oakeley EJ, Jones JDG, Felix G, Boller T. Bacterial disease resistance in Arabidopsis through flagellin perception. Nature. 2004;428 doi: 10.1038/nature02485. [DOI] [PubMed] [Google Scholar]

Decision letter

Editor: Nicola J Patron1

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Acceptance summary:

This manuscript provides evidence that chromatin with activating and repressing histone modifications at the same location enables the rapid activation of genes required for the biosynthesis of camalexin, a plant defence compound, upon detection of a pathogen signal. The data show that both H3K27me3 and H3K18ac are required to determine the timing of gene expression.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "A Bivalent Chromatin Controls Timing of Expression of Camalexin Biosynthesis Genes in Response to a Pathogen Signal in Arabidopsis" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Comments to the Authors:

We are sorry to say that, after consultation with the reviewers, we have decided that your work will not be considered further for publication by eLife.

In this manuscript, the authors present data showing that changes to H3K27me3/H3K18ac marks help regulate the main Arabidopsis phytoalexin. While the chromatin states of these specific genes have not previously been reported on directly, this form of regulation is known to occur in plants and H3K27me3 and H3K18ac have previously been shown to colocalize in a large number of genes in the Arabidopsis genome. Further, H3K27me3 marked genes are known to be targets of Polycomb Repressive Complex 2 (PRC2) and, therefore, the changes in the regulation of such genes in H3K27me3 defective mutants are as expected. The term bivalent chromatin has been misused in this manuscript to describe overlapping chromatin marks, which has led to the formation of confusing and incorrect hypotheses. In addition, concerns were raised about the pathway enrichment analysis and if this allows extension to specialized metabolism pathways in general. There is a potential for over-counting pathways when a metabolic pathway is inaccurately described by hierarchical database design. The phenomenon appears limited to fewer pathways than suggested and some statistical analysis are needed to support the broader claims. Finally, the changes to metabolite levels should be interpreted with reference to previously published data showing that post-translational mechanisms play a role in their regulation.

Reviewer #1 (Recommendations for the authors):

The focus of the manuscript is to investigate changes in chromatin state. The series of experiments to identify and examine changes in chromatin marks in response to flgg22 treatments investigate clearly defined hypotheses and the results are clear and accurately interpreted. Where the manuscript needs to be improved is the wider interpretation of the results in relation to changes in overall metabolite levels. The finding of bivalent chromatin and its role in gene regulation in response to biotic stress is interesting. However, the data is presented within a very narrow focus and interpretation of the findings within the wider background of literature about plant metabolic responses to pathogen infection is too limited. The manuscript would be improved by the inclusion of introductory sentences and a Discussion section focussing on how these new discoveries fit into the wider understanding of camalexin expression in response to pathogen infection. Though clearly chromatin state plays a role, it is one part of a multi-layered response that also includes changes in the rate of translation calcium-dependent phosphorylation of WRKY33 to enable to initiate transcription.

Reviewer #2 (Recommendations for the authors):

Technical suggestions.

In a number of cases, the pathway enrichment analysis gives the impression that there are a number of different pathways being linked to bivalent chromatin. However, it isn't clear if these are really different pathways or a construct of the categorization system. For example, in Figure 4B, Quercetin and Rutin biosynthesis each have two genes but Rutin is simply a Quercetin glycoside and so it is likely that these two genes are the same and this isn't two different pathways. There are other groupings like (Simple coumarins, Scopolin and phenylpropanoid; and Glucosinolate, indole glucosinolate and camalexin) where the separate genes are not really finding multiple different pathways but simply the same genes are finding different hierarchical categorizations. There is a similar complexity with at least six glucosinolate from Xmethionine treated as independent pathways as well as aliphatic glucosinolate, an overarching category encompassing these other terms. Yet these pathways are an artifact of glucosinolates involving a cyclic step that is determined by one enzyme (MAM1/2) adding a carbon and a second enzyme (CYP79F1/F2) having specificity for the chain length. Other than those two steps the rest of genes in these pathways. The complexity is that this cyclic structure can't be shown in a hierarchical construction. Thus to make the programming work they are broken into different "pathways". In general this makes it unclear if the specialized pathway enrichment analysis is being driven by an actual enrichment for all of specialized metabolism or did the identified genes just happen to track with these artificial pathways that can create the impression of all specialized metabolism. The key to this is to dissect how much the enrichment is driven by the same gene driving the identification of multiple pathways. Right now, it seems that this is not an enrichment with 23 independent specialized metabolism pathways but instead being driven by the issues of hierarchical programming and non-linear pathways in the original databases.

I should be clear that the above concern does not affect the specific camalexin molecular analysis, it simply suggests that the generalizations to all of specialized metabolism might be unsupported.

There are some missing statistical analysis. In Figure 3A-C, there should be some statistical analysis of the transcsript data as camalexin transcripts can easily have 3-4x biological variation so simply showing the fold-change doesn't provide information about what differences are meaningful in this pathway. Similarly, in Figure 3D, there should be some statistical analysis of the camalexin to test if there are time x genotype dependent effects on the metabolite. The standard error is not a statistical test. Similarly, the other replicates of this experiment should be combined into this analysis. Camalexin metabolites are fairly noisy and all the data should be combined and analyzed to show how reproducible the effects are.

I don't think the literature provides sufficient resolution to say that only CYP79B2 is involved in camalexin as is proposed in Figure 2. CYP79B2 KOs an induce camalexin suggesting that CYP79B3 is also involved. Similarly they CYP71A12/A13 specificity is still unresolved. I understand the focus when doing the CHIP but the genomic data should have information about these other genes and could be provided in the supplemental.

The metabolomics in response to FLG22 was intended to show that the gene expression changes were indicative of metabolite changes. Unfortunately, the MS finds largely unknown compounds which could be all related to camalexin synthesis or other CYP79B2 dependent metabolites. As such, it doesn't really support the broader argument about all specialized metabolites. There is also the chance that the enhanced response in clf28 could be due to lower basal levels. Right now with FC levels shown it isn't clear if this is an enhanced response or enhanced repression. It would help to show absolute levels and to conduct factorial analysis testing for genotype and genotype x treatment effects. This is a similar concern for the microarray analysis as well.

Editorial Suggestions

I'm not totally sure what the main point of the second paragraph of the discussion is conveying. This paragraph starts with ideas about bivalent chromatin in development and then ends with specific details about bivalent chromatin in camalexin. I feel there is a big point in this paragraph but as written I can't quite grasp it. eLife allows more space then is used and it would help the general reader to expand this section to let the big point come through clearer. I think it is about bivalent chromatin being a trigger switch when something must go from absolute off to absolute on and do so very quickly.

Related to the general idea of bivalent chromatin and camalexin, there are some complications around this pathway that aren't being integrated into the discussion. It doesn't change the results of this work but it does lay out a more complicated view then is being proposed. For example, the PAD mutants from Jane Glazebrook were isolated based on the accumulation of camalexin. A result of this work was the fact that the regulatory mutants (e.g. PAD4, etc) were highly specific to the stimulus being applied and often did not affect camalexin regulation in response to other stimuli. There is also work showing that the camalexin pathway has post-transcriptional regulation where the transcripts can accumulate to a high level with no metabolite accumulation. Finally, the FLG22 ability to induce camalexin is age dependent and is not seen in older soil grown plants based on the MAPK studies. Thus, the upstream and downstream regulatory processes around camalexin are significantly more complicated than the intro and discussion lays out by describing a couple of TFs. It would help to expand the discussion and introduction to convey this full complexity that is present in the established literature as it is critical to place the bivalent chromatin into the full context of camalexin regulation.

Reviewer #3 (Recommendations for the authors):

The genes being studied are targeted by the PRC2 complex and this is why they have H3K27me3. Past studies have discovered that H3K18ac and H2A.Z colocalize with H3K27me3 (Luo C, 2012). Yet, there is no mention of H2A.Z throughout this study? Would this now make this chromatin state 'trivalent' chromatin? This issue specifically highlights the incorrect premise of this study, that there is a "bivalent chromatin" at these metabolic genes. They, like many other H3K27me3 marked genes, are targets of PRC2 and silenced until proper environmental/development cues are perceived. This fits with the highly specialized need for these metabolic genes in unique environments.

There is no evidence presented that H3K18ac is an 'activating" mark. This should not be confused with other histone acetylation marks. Based on what was previously described by Luo C et al., 2012, it is a repressive mark given it's colocalization with H3K27me3.

How do you distinguish direct versus indirect effects when using pkl/clf/idm1 mutants? These genes are required for general gene regulation. For example, without clf, no H3K27me3 occurs leading to up-regulation of all PRC2 targeted genes?

I have major concerns that IDM1 is the HAT that acetylates H3K18 in gene bodies. IDM1 specifically binds to methylated DNA, which does not colocalize with H3K27me3 in Arabidopsis (see also Figure 1a). This would also explain the seemingly weak signals from all experiments using idm1. ChIP-seq data for IDM1 exists (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3575687/) and should be used to show it colocalizes specifically to the genes that have H3K27me3 versus other types of genes.

The term bivalent chromatin should be more carefully used in this study. Bivalent chromatin specifically refers to H3K4me3/H3K27me3. It should not be used to described H3K27me3/H3K18ac. There are numerous examples where bivalent chromatin is referring to the classic definition (H3K4me3/H3K27me3) to set up rationale for proposed experiments on H3K27me3/H3K18ac. This leads to incorrect hypotheses and is confusing to the reader.

eLife. 2021 Sep 15;10:e69508. doi: 10.7554/eLife.69508.sa2

Author response

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

Reviewer #1 (Recommendations for the authors):

The focus of the manuscript is to investigate changes in chromatin state. The series of experiments to identify and examine changes in chromatin marks in response to flgg22 treatments investigate clearly defined hypotheses and the results are clear and accurately interpreted. Where the manuscript needs to be improved is the wider interpretation of the results in relation to changes in overall metabolite levels. The finding of bivalent chromatin and its role in gene regulation in response to biotic stress is interesting. However, the data is presented within a very narrow focus and interpretation of the findings within the wider background of literature about plant metabolic responses to pathogen infection is too limited. The manuscript would be improved by the inclusion of introductory sentences and a Discussion section focussing on how these new discoveries fit into the wider understanding of camalexin expression in response to pathogen infection. Though clearly chromatin state plays a role, it is one part of a multi-layered response that also includes changes in the rate of translation calcium-dependent phosphorylation of WRKY33 to enable to initiate transcription.

We thank the reviewer for pointing this out and helping us improve the manuscript by providing a more holistic context of camalexin regulation. We revised Introduction and Discussion to include current understanding of post-transcriptional regulation on camalexin pathway and interpret the results in a more holistic view. Revisions in the Introduction:

“Previous studies revealed complex transcriptional and translational control of camalexin biosynthesis genes8, 9, 10, 11, 12, 13. At the transcriptional regulation level, transcription factors from the MYB family, including MYB34, MYB51 and MYB122 promote camalexin biosynthesis in response to P. syringae infection9, 12. WRKY33 functions as an activator and directly binds to the promoters of camalexin biosynthesis genes11. Besides transcription factors, CALCIUM-DEPENDENT PROTEIN KINASE (CPK)5/6 and MAPK3/6 can phosphorylate WRKY33 to enhance promoter binding and transactivation10, 13. At the translational level, ribosome footprinting showed that genes involved in camalexin biosynthesis, CYP79B2 and CYP79B3, also increased translational efficiency under pattern triggered immunity14, 15. Despite the rich knowledge of these upstream regulators of the camalexin biosynthetic pathway, it remains unknown how the rapid induction upon a pathogen signal is enabled”.

In Discussion, we added:

“Our results provide new evidence for how chemical defense mediated by camalexin may be regulated at the epigenetic level. However, we cannot rule out the possibility that other known mechanisms regulating camalexin genes may also affect the transcription kinetics and metabolite accumulation. For example, H3K27me3 affects gene expression by altering chromatin accessibility to transcription factors 21. Removing this repression mark may create a permissive environment and facilitate transcription factors, such as WRKY33, to bind to promoters of camalexin genes. At the translational level, camalexin biosynthesis genes can alter translational efficiency controlled by a highly enriched messenger RNA consensus sequence, R-motif, during pattern triggered immunity 14, 15. Additional studies are needed to unravel how different regulatory machineries work together to enable the rapid induction of camalexin genes upon stress signals.”

Reviewer #2 (Recommendations for the authors):

Technical suggestions.

In a number of cases, the pathway enrichment analysis gives the impression that there are a number of different pathways being linked to bivalent chromatin. However, it isn't clear if these are really different pathways or a construct of the categorization system. For example, in Figure 4B, Quercetin and Rutin biosynthesis each have two genes but Rutin is simply a Quercetin glycoside and so it is likely that these two genes are the same and this isn't two different pathways. There are other groupings like (Simple coumarins, Scopolin and phenylpropanoid; and Glucosinolate, indole glucosinolate and camalexin) where the separate genes are not really finding multiple different pathways but simply the same genes are finding different hierarchical categorizations. There is a similar complexity with at least six glucosinolate from Xmethionine treated as independent pathways as well as aliphatic glucosinolate, an overarching category encompassing these other terms. Yet these pathways are an artifact of glucosinolates involving a cyclic step that is determined by one enzyme (MAM1/2) adding a carbon and a second enzyme (CYP79F1/F2) having specificity for the chain length. Other than those two steps the rest of genes in these pathways. The complexity is that this cyclic structure can't be shown in a hierarchical construction. Thus to make the programming work they are broken into different "pathways". In general this makes it unclear if the specialized pathway enrichment analysis is being driven by an actual enrichment for all of specialized metabolism or did the identified genes just happen to track with these artificial pathways that can create the impression of all specialized metabolism. The key to this is to dissect how much the enrichment is driven by the same gene driving the identification of multiple pathways. Right now, it seems that this is not an enrichment with 23 independent specialized metabolism pathways but instead being driven by the issues of hierarchical programming and non-linear pathways in the original databases.

I should be clear that the above concern does not affect the specific camalexin molecular analysis, it simply suggests that the generalizations to all of specialized metabolism might be unsupported.

In general, we really appreciate the technical comments from reviewer #2 and believe our manuscript was improved by addressing their comments. This comment makes two points: (1) redundancy of reaction mapping to pathways in specialized metabolism and potential inflation of pathway enrichment in specialized metabolism and (2) the claim of the bivalent chromatin being generalized to specialized metabolism.

The reviewer brought up an important point regarding the gene redundancy in pathway enrichment analysis. This made us re-evaluate the pathway organization in Plant Metabolic Network databases. We are now exploring different strategies to organize pathways to address the shared reaction issue for enrichment analysis, which will require a substantial effort and we feel is out of the scope of this manuscript. Nonetheless, we feel this is an important caveat of the pathway enrichment analyses of metabolic pathway databases.

To address the comment more directly, we reported the number of metabolic genes marked by both marks by stating:

“To determine how prevalent the H3K27me3-H3K18ac association was for specialized metabolic genes, we counted the number of genes marked by both modifications. We found that 37% (324 out of 887) of specialized metabolic genes were marked by both H3K27me3 and H3K81ac (Figure 1B).”

Enrichment analyses show increased propensity for the occurrence of a phenomenon to a specific group (e.g. specialized metabolism vs total metabolism), but does not imply the occurrence to all members of the enriched group (Reimand et al., (2019) Nature Protocol). To see which additional pathways might be regulated similarly as camalexin pathway, we used RNA-seq and untargeted metabolomics to measure gene expression and metabolites change in different genotypes under FLG22 treatment. Several pathways were identified using enrichment analysis with metabolic genes showing similar gene expression patterns as camalexin genes. However, we feel that these are preliminary at best and divert attention away from the major discovery of a new type of bivalent chromatin and its potential biological role on camalexin biosynthesis pathway. Thus, we removed sections reporting these results in the manuscript.

There are some missing statistical analysis. In Figure 3A-C, there should be some statistical analysis of the transcsript data as camalexin transcripts can easily have 3-4x biological variation so simply showing the fold-change doesn't provide information about what differences are meaningful in this pathway. Similarly, in Figure 3D, there should be some statistical analysis of the camalexin to test if there are time x genotype dependent effects on the metabolite. The standard error is not a statistical test. Similarly, the other replicates of this experiment should be combined into this analysis. Camalexin metabolites are fairly noisy and all the data should be combined and analyzed to show how reproducible the effects are.

We wanted to show how the transcriptional kinetics of camalexin genes changed upon FLG22 treatment at different time points and we feel that fold change reflects the change of gene expression best. To better understand the effect of genotype, FLG22 treatment, and time points on gene expression, we conducted three-way ANOVA using raw transcripts of camalexin genes. Supplementary File 5 summarized whether each variable or interactions between variables significantly contributed to the variance among groups. The results showed that genotype, treatment, time point, and their pairwise interaction significantly contributes to the expression change of camalexin genes. The three-way interaction term Genotype: Treatment: Time was significant for CYP71A13 and PAD3 but not for CYP79B2.

To better reflect the effect of genotype and treatment on camalexin gene expression, we generated boxplots for these genes and compared the difference between mock and FLG22 treated samples using two-way ANOVA followed by Dunnett’s test. The results showed that camalexin genes were induced 5min after FLG22 treatment in mutant plants defective of H3K27me3. In mutant defective of H3K18ac, the significant induction of camalexin genes was detected at 1h after FLG22 treatment.

For camalexin accumulation, as the reviewer suggested, we examined the effect of genotype, FLG22 treatment, and time point on camalexin gene expression using three-way ANOVA by combining all technical and biological replicates from two independent experiments. In Supplementary file 6 are the results summarizing whether each variable or interactions between variables significantly contribute to the variance among groups.

The analysis showed that genotype, time, and FLG22 treatment affected camalexin gene expression significantly with p values lower than 0.05. The interaction between genotype and treatment was marginally significant with p value 0.062. This may be because metabolite data is nosier than the gene expression data. Another possibility is that the kinetics of metabolites might be different from transcripts since the significant accumulation of metabolite was detected at fairly late time points. We added theses caveats to the Result section reporting camalexin accumulation in response to FLG22 treatment. The interaction between treatment and time is statistically significant, which indicates the effect of FLG22 on camalexin gene induction varies based on the time of treatment. The variable Trial represents the two rounds of measurements of camalexin content in each sample. The p value for Trial is larger than 0.05, which suggests the rounds of experiments detected similar results. We further analyzed the difference of camalexin content between mock and FLG22 treated samples in different genotype by running two-way ANOVA with genotype and treatment as variables followed by Dunnett’s test.

I don't think the literature provides sufficient resolution to say that only CYP79B2 is involved in camalexin as is proposed in Figure 2. CYP79B2 KOs an induce camalexin suggesting that CYP79B3 is also involved. Similarly they CYP71A12/A13 specificity is still unresolved. I understand the focus when doing the CHIP but the genomic data should have information about these other genes and could be provided in the supplemental.

We revised Figure 2 to include CYP79B2, CYP79B3, CYP71A12, and CYP71A13 to the proposed pathway diagram. We selected CYP79B2, CYP71A13, and PAD3 in this study because these three genes are sufficient to produce camalexin in in vitro enzymatic assays (Klein et al., (2013) Angew Chem Int Ed Engl). Nonetheless, additional enzymes may play important roles in producing camalexin in vivo under particular conditions.

To respond to the reviewer’s comment, we plotted the expression of CYP79B3 and CYP91A12 and analyzed their gene expression change using one-way ANOVA in (Author response image 1). Post-hoc Tukey’s test shows that both genes showed significant induction in pkl-1 and Col-0, but not in idm1 plants, 30min after FLG22 treatment, which is consistent with the three essential genes we present in the manuscript.

Author response image 1.

Author response image 1.

Open in a new tab

The metabolomics in response to FLG22 was intended to show that the gene expression changes were indicative of metabolite changes. Unfortunately, the MS finds largely unknown compounds which could be all related to camalexin synthesis or other CYP79B2 dependent metabolites. As such, it doesn't really support the broader argument about all specialized metabolites. There is also the chance that the enhanced response in clf28 could be due to lower basal levels. Right now with FC levels shown it isn't clear if this is an enhanced response or enhanced repression. It would help to show absolute levels and to conduct factorial analysis testing for genotype and genotype x treatment effects. This is a similar concern for the microarray analysis as well.

We attempted to identify additional metabolic pathways that may be controlled by both H3K27me3-H3k18ac bivalent chromatin as camalexin genes in response to FLG22 signal. We used RNA-seq and untargeted metabolomics to measure gene expression and metabolites change in different genotypes treated under FLG22 treatment. Several pathways were identified using enrichment analysis with metabolic genes showing similar gene expression patterns as camalexin genes. We feel that the RNA-seq and metabolomics data provide valuable preliminary information for future studies but divert attention away from the major finding of this manuscript on camalexin biosynthesis pathway. Thus, we removed sections reporting these results in the manuscript.

Editorial Suggestions

I'm not totally sure what the main point of the second paragraph of the discussion is conveying. This paragraph starts with ideas about bivalent chromatin in development and then ends with specific details about bivalent chromatin in camalexin. I feel there is a big point in this paragraph but as written I can't quite grasp it. eLife allows more space then is used and it would help the general reader to expand this section to let the big point come through clearer. I think it is about bivalent chromatin being a trigger switch when something must go from absolute off to absolute on and do so very quickly.

We appreciate the reviewer’s comment about improving the section of Discussion on the function of bivalent chromatin. In the original submission, we discussed the observation of bivalent chromatin in the literature and its biological function discovered this study in the second paragraph of Discussion. To better highlight the knowledge gap filled in this study in understanding the function of bivalent chromatin, we rewrote this section to discuss the known knowledge on bivalent chromatin function and the main discovery of this manuscript. Here are the revised paragraphs:

“The function of bivalent chromatin has long been hypothesized to poise developmental genes in embryonic stem cells for rapid activation upon a cell differentiation signal22. Only few studies directly tested this hypothesis and all of them were focusing on stem cells or cancer cells47,48,49. For example, in embryonic stem cells, H3K27me3-H3K4me3 bivalent chromatin was associated with HoxB4, one of the key regulators controlling development. Inhibition of H3K4me methyltransferase, hSET1A, repressed the expression of HoxB4 and led to a block in the differentiation of blood cells47. In addition, H3K27me3-H3K4me3 bivalent chromatin was observed in cancer-initiating cells. Disruption of the bivalent state through inhibition of the H3K27 methyltransferase EZH2 inhibited the self-renewal of cancer cells through de-repression of a key canonical marker of normal colonocyte differentiation, named Indian Hedgehog48. Despite this knowledge, the role of bivalent chromatin on regulating metabolism remains unknown. In this study, we reported the colocalized of H3K27me3 and H3K18ac on camalexin genes and functionally examined the role of H3K27me3-H3K18ac bivalent chromatin in Arabidopsis. The results suggest that H3K27me3-H3K18ac bivalent chromatin maintains the proper timing of camalexin gene induction upon a pathogen signal.”

Related to the general idea of bivalent chromatin and camalexin, there are some complications around this pathway that aren't being integrated into the discussion. It doesn't change the results of this work but it does lay out a more complicated view then is being proposed. For example, the PAD mutants from Jane Glazebrook were isolated based on the accumulation of camalexin. A result of this work was the fact that the regulatory mutants (e.g. PAD4, etc) were highly specific to the stimulus being applied and often did not affect camalexin regulation in response to other stimuli. There is also work showing that the camalexin pathway has post-transcriptional regulation where the transcripts can accumulate to a high level with no metabolite accumulation. Finally, the FLG22 ability to induce camalexin is age dependent and is not seen in older soil grown plants based on the MAPK studies. Thus, the upstream and downstream regulatory processes around camalexin are significantly more complicated than the intro and discussion lays out by describing a couple of TFs. It would help to expand the discussion and introduction to convey this full complexity that is present in the established literature as it is critical to place the bivalent chromatin into the full context of camalexin regulation.

We thank the reviewer for pointing this out and helping us improve the manuscript by providing a more holistic context of camalexin regulation. We revised Introduction and Discussion to include current understanding of post-transcriptional regulation on camalexin pathway and interpret the results in a more holistic view. Revisions in the Introduction:

“Previous studies revealed complex transcriptional and translational control of camalexin biosynthesis genes8, 9, 10, 11, 12, 13. At the transcriptional regulation level, transcription factors from the MYB family, including MYB34, MYB51 and MYB122 promote camalexin biosynthesis in response to P. syringae infection9, 12. WRKY33 functions as an activator and directly binds to the promoters of camalexin biosynthesis genes11. Besides transcription factors, CALCIUM-DEPENDENT PROTEIN KINASE (CPK)5/6 and MAPK3/6 can phosphorylate WRKY33 to enhance promoter binding and transactivation10, 13. At the translational level, ribosome footprinting showed that genes involved in camalexin biosynthesis, CYP79B2 and CYP79B3, also increased translational efficiency under pattern triggered immunity14, 15. Despite the rich knowledge of these upstream regulators of the camalexin biosynthetic pathway, it remains unknown how the rapid induction upon a pathogen signal is enabled”

In Discussion, we added:

“Our results provide new evidence for how chemical defense mediated by camalexin may be regulated at the epigenetic level. However, we cannot rule out the possibility that other known mechanisms regulating camalexin genes may also affect the transcription kinetics and metabolite accumulation. For example, H3K27me3 affects gene expression by altering chromatin accessibility to transcription factors 21. Removing this repression mark may create a permissive environment and facilitate transcription factors, such as WRKY33, to bind to promoters of camalexin genes. At the translational level, camalexin biosynthesis genes can alter translational efficiency controlled by a highly enriched messenger RNA consensus sequence, R-motif, during pattern triggered immunity 14, 15. Additional studies are needed to unravel how different regulatory machineries work together to enable the rapid induction of camalexin genes upon stress signals.”

Reviewer #3 (Recommendations for the authors):

The genes being studied are targeted by the PRC2 complex and this is why they have H3K27me3. Past studies have discovered that H3K18ac and H2A.Z colocalize with H3K27me3 (Luo C, 2012). Yet, there is no mention of H2A.Z throughout this study? Would this now make this chromatin state 'trivalent' chromatin? This issue specifically highlights the incorrect premise of this study, that there is a "bivalent chromatin" at these metabolic genes. They, like many other H3K27me3 marked genes, are targets of PRC2 and silenced until proper environmental/development cues are perceived. This fits with the highly specialized need for these metabolic genes in unique environments.

Luo et al., (2012) The Plant Journal did not study H2A.Z and there is no information about H2A.Z colocalized with H3K27me3 in that publication. In addition, we respectively disagree with the reviewer that H2A.Z is known to colocalize with H3K27me3. In mammalian embryonic stem cells and neural progenitors, H2A.Z is deposited at promoters marked by H3K4me3-H3K27me3 bivalent chromatin and at active promoters solely marked by H3K4me3, but not the chromatin solely associated with only H3K27me3 (Ku et al., (2012) Genome Biology). In mouse embryonic stem cells, H3K27me3 enrichment correlates strongly with H2A.Z based on ChIP-seq profiles, which indicates the co-localization of H2A.Z and H3K27me3 (Wang et al. (2018) BMC Biology). In Arabidopsis, H2A.Z and H3K27me3 have overlapping targets, such as genes associated with thalianol biosynthesis, but the co-localized has not been experimentally examined (Sequeira-Mendes et al., (2014) Plant Cell, Carter et al., (2018) Plant Cell, Nützmann et al., (2015) New Phytologist). These results suggest that H3K27me3 and H1A.Z may have different relationships in different organisms and their colocalization in Arabidopsis has not been tested.

There is no evidence presented that H3K18ac is an 'activating" mark. This should not be confused with other histone acetylation marks. Based on what was previously described by Luo C et al., 2012, it is a repressive mark given it's colocalization with H3K27me3.

We disagree with the reviewer that H3K18ac was reported as a repression mark in Luo et al., (2012) The Plant Journal. We cited the sentence describing H3K18ac-H3K27me3 colocalization in this publication:

“Intriguingly, a strong correlation was detected between H3K18Ac and H3K27me3 in the Arabidopsis genome (Pearson r = 0.44, Figure 1b and Figure S4), although histone acetylation is not expected to co‐localize with a repressive mark such as H3K27me3.”

Indeed, H3K18ac was reported as an activation mark in Luo et al., (2012) The Plant Journal, which is cited here “Our analysis also showed that multiple active histone marks such as H3K4me3, H3K9Ac and H3K18Ac were significantly more depleted in gene bodies for genes in States 2, 3 and 4 compared to those in State 1 (Figure 5f–h and Figure S9b–d)”.

There are additional reports on H3K18ac functioning as an activation mark, such as Liu et al., (2013) The Plant Cell, “Among the seven known Lys residues available for acetylation in the H3 N-terminal tail, acetylated Lys-9, Lys-14, Lys-18, and Lys-23 are responsible for promoting transcription”. In addition, studies in other systems, such human cell lines and Plasmodium falciparum also revealed that H3K18ac promotes activation of target genes, such as Luo et al., (2019) Stem Cell Reports, Tang et al., (2020) Epigenetics and Chromatin, Wang et al., (2008) Nature Genetics, Roy et al., (2017) Genome Research.

How do you distinguish direct versus indirect effects when using pkl/clf/idm1 mutants? These genes are required for general gene regulation. For example, without clf, no H3K27me3 occurs leading to up-regulation of all PRC2 targeted genes?

The reviewer brought up a good point that it’s hard to distinguish the direct and indirect effects in the mutants with defective epigenetic modifications, which is also a challenge in the field. We used mutant lines in two independent genes for each mark to test the function of H3K27me3 (pkl, clf) and H3K18ac (idm1, idm2), which revealed similar results (Figure 3 A to D). In addition, the up-regulation of camalexin genes is associated with the alteration of these two marks (Figure 3 E to F). These results support that epigenetic modifications do play a role to regulate the expression of camalexin genes.

I have major concerns that IDM1 is the HAT that acetylates H3K18 in gene bodies. IDM1 specifically binds to methylated DNA, which does not colocalize with H3K27me3 in Arabidopsis (see also Figure 1a). This would also explain the seemingly weak signals from all experiments using idm1. ChIP-seq data for IDM1 exists (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3575687/) and should be used to show it colocalizes specifically to the genes that have H3K27me3 versus other types of genes.

We double checked the publication highlighted by the reviewer regarding ChIP-seq data in idm1 and this study only performed ChIP-PCR on target genes instead of ChIP-seq using idm1 mutant. So unfortunately, no ChIP-seq data from the paper could be obtained. In one of the follow-up studies from the same group that discovered IDM1 (Nie et al., (2019) PNAS), H3K18ac was profiled genome-wide using imd1 mutant in the background overexpressing SUC2. There, this dataset is also not appropriate for our study.

IDM1 is a histone acetyltransferase, which binds to DNA methylated region to initiate DNA demethylation process by recruiting SWR1 and ROS1 protein complexes (Qian et al., (2014) Molecular Cell, Nie et al., (2019) PNAS). We checked the DNA methylation pattern of camalexin genes using the methylome profiles generated by 1001 genomes project (http://neomorph.salk.edu/1001.aj.php) and the three genes we tested in this study have DNA methylation in the gene body. This suggests that camalexin genes can be targets of IDM1.

In addition, the enzyme complex required to establish H3K18ac is not well characterized, which limits the genetic resources we can use. We revised the text to reflect the challenge and limitations to dissect the function of H3K18ac on the induction of camalexin genes: “The protein complex involved in establishing H3K18ac has not been well characterized in Arabidopsis, which limits the mutant lines that can be utilized to study the function of this epigenetic modification. The functionally characterized mutant lines associated with establishing H3K18ac are increased DNA methylation 1 and 2 (idm1 and idm2)33, 34. IDM1 is the histone acetyltransferase and IDM2 is a heat shock protein that functions in the same protein complex as IDM1.”

The term bivalent chromatin should be more carefully used in this study. Bivalent chromatin specifically refers to H3K4me3/H3K27me3. It should not be used to described H3K27me3/H3K18ac. There are numerous examples where bivalent chromatin is referring to the classic definition (H3K4me3/H3K27me3) to set up rationale for proposed experiments on H3K27me3/H3K18ac. This leads to incorrect hypotheses and is confusing to the reader.

We disagree with the reviewer that bivalent chromatin specifically refers to H3K4me3 and H3K27me3 and we cited several references that support our definition used in the manuscript: “the bivalent domains (BDs) – that are distinguished for having two histone marks that are associated with both positive and negative transcriptional events”. Refs: Blanco et al., (2019) Trends in Genetics; Sneppen et al., (2019) Nature Communications; Zhang et al., (2020) Molecular Plant; Velde et al., (2021) Communications Biology, Jacquet et al. (2016) Molecular Cell. To avoid any confusion, we revised the title and the manuscript to clarify that we are reporting a novel type of bivalent chromatin instead of the well-studied variant formed by H3K4me3 and H3K27me3.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Figure 1—source data 1. Source data used to generate Figure 1A, B.
    elife-69508-fig1-data1.xlsx (85.2KB, xlsx)
    Figure 2—source data 1. Source data used to generate Figure 2B,C.
    elife-69508-fig2-data1.xlsx (11.3KB, xlsx)
    Figure 3—source data 1. Source data used to generate Figure 3A-F.
    elife-69508-fig3-data1.xlsx (14.3KB, xlsx)
    Supplementary file 1. Log2 fold change in the enrichment analysis to identify specialized metabolic pathways associated with both trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac.
    elife-69508-supp1.xlsx (12.6KB, xlsx)
    Supplementary file 2. Primers used to quantify the abundance of trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac in the genomic regions of camalexin biosynthesis genes using chromatin immunoprecipitation (ChIP)-qPCR in wild-type and mutant plants with or without flagellin 22 (FLG22) treatment.
    elife-69508-supp2.docx (16.9KB, docx)
    Supplementary file 3. Primers used in the sequential chromatin immunoprecipitation (ChIP)-qPCR experiment to examine the co-localization of trimethylation of lysine 27 of histone 3 (H3K27me3) and H3K18ac within camalexin biosynthesis genes.
    elife-69508-supp3.docx (14.3KB, docx)
    Supplementary file 4. Primers used to examine the expression of camalexin biosynthesis genes under flagellin 22 (FLG22) induction using qPCR.
    elife-69508-supp4.docx (13.8KB, docx)
    Supplementary file 5. The effect of genotype, flagellin 22 (FLG22) treatment, and time point on the expression change of camalexin genes.
    elife-69508-supp5.docx (16.2KB, docx)
    Supplementary file 6. The effect of genotype, flagellin 22 (FLG22) treatment, and time on camalexin accumulation.
    elife-69508-supp6.docx (15.1KB, docx)
    Transparent reporting form
    elife-69508-transrepform1.docx (246.4KB, docx)

    Data Availability Statement

    The data that support the findings of this study are available within the article and its supplementary information files. Arabidopsis metabolic genes and pathways are available at Plant Metabolic Network (https://plantcyc.org/).

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES