Introduction

Molecular mechanism underlying disease resistance in plants has been an active area of research. Although, the mechanism for resistance against biotrophic and necrotrophic pathogens may differ in detail, the basic outline of the mechanism is the same. It is now widely accepted that the R gene-mediated race-specific resistance is managed through two layers of immunity, namely pathogen associated molecular patterns (PAMP) triggered immunity (PTI) and effector triggered immunity (ETI). The gene R provides resistance, only when there is a corresponding avirulence (Avr) gene in the prevailing race of the pathogen. This has been described as gene-for-gene relationship (Flor 1942, 1971). In both PTI and ETI, the plant cells generally respond to the pathogen attack through nucleotide binding site-leucine rich repeats (NBS-LRR) (also called nod-like receptors = NLR), followed by downstream signaling pathways. These pathways form a network that is not fully understood, although some pathways including salicylic acid (SA), jasmonic acid (JA) and ethylene (Et) pathways are known to operate. More recently, formation of a resistosome with R gene encoded NBS-LRR protein has also been suggested (Wang et al. 2019a, b, c). Another recent study also demonstrated that on pathogen attack, the cell death is really caused due to NADase activity of the Toll/Interleukin receptor 1 (TIR) domain at the N-terminus of NBS-LRR receptor (Horsefield et al. 2019; Wan et al. 2019); NADase causes cleavage of the oxidized form of nicotinamide adenine dinucleotide (NAD+) leading to cell death and hypersensitive reaction (HR), thus providing protection against biotrophs.

In wheat, leaf rust is an important disease and causes major losses in yield (Draz et al. 2015). At least ~ 80 Lr genes for leaf rust resistance in wheat are already known (see Gill et al. 2019) and more genes are being regularly discovered. These genes include both the race-specific seedling resistance (Sr; also described as all stage resistance) genes and also the race-non-specific adult plant resistance (APR) genes. An important leaf rust resistance gene, which has been deployed for development of a number of resistant cultivars in India is Lr28. In order to understand the molecular mechanism underlying leaf rust resistance due to Lr28, a pair of NILs in the background of wheat cultivar HD2329 was developed at ICAR-Indian Agricultural Research Institute (ICAR-IARI), India. This pair of NILs has already been utilized in our laboratory for studies involving transcriptome, DNA methylation, histone modification and non-coding RNAs (miRNAs and lncRNAs). Using transcriptomics approach, we discovered a large number of downstream genes, which are either induced or repressed in the susceptible and resistant NILs following the attack by the pathogen (Dhariwal et al. 2011; Chandra et al. 2016; Singh et al. 2017; Sharma et al. 2018a).

During our earlier transcriptome studies involving the above pair of NILs, some evidence of epigenetic regulation of Lr28-mediated leaf rust resistance was available. This evidence included Lr28-mediated differential expression of genes encoding methyltransferases, non-coding RNAs and histone acetylases/methyltransferases. This prompted us to undertake epigenetic studies using the same experimental material. For instance, the role of histone acetylation was examined by us in controlling the expression of six genes involved in interaction between wheat and leaf rust pathogen (Sharma et al. 2018b). Studies are also available in rice, Arabidopsis and cowpea where the role of histone methylation was examined during abiotic/biotic stresses (Kim et al. 2008; Zong et al. 2013; Ayyappan et al. 2015). However, the role of histone methylation in regulation of leaf rust resistance in wheat has never been examined.

Histone acetylation is more generally correlated with open chromatin and therefore more active transcription (reviewed by Zentner and Henikoff 2013). By contrast, histone methylation can result in different transcriptional outcomes, depending on the specific individual amino acids that are modified and the degree of modification (reviewed by Li et al. 2007). Among histone methylation marks, two histone trimethylation marks (H3K4me3 and H3K27me3) have been widely used in understanding the epigenetic regulation of development, vernalization and abiotic stress tolerance, nitrogen use efficiency (NUE) in Arabidopsis, rice, wheat, Brachypodium distachyon, etc. (Kim et al. 2008; Diallo et al. 2012; Zong et al. 2013; Zhang et al. 2009; You et al. 2017; Huan et al. 2018; Wu et al. 2020). These two histone trimethylation marks have also been used in animal systems including mouse (Yang and Wilson 2018). H3K4me3 is widely known to be associated with promoter sequences and gene activation, while H3K27me3 is known to be associated with silenced genes or non-genic regions causing heterochromatinization and repression of gene expression through trans-regulation (Heintzman et al. 2007 and Zhou and Hu 2010). However, there are only few reports where the role of these two histone marks in the regulation of specific genes during biotic stress (disease resistance) has been examined. For example, the removal of repression mark H3K27me3 through Jumonji C domain protein JMJ705, which encodes H3K27me2/3 demethylase resulted into activation of defense related genes during infection by the bacterial pathogen Xanthomonas oryzae in rice (Li et al. 2013). However, no studies involving genome-wide histone profiling in relation to biotic stress have been conducted in cereals although a solitary genome-wide study for H3K9me2 and H4K12ac is available for rust (Uromyces appendiculatus) infected Phaseolus vulgaris (Ayyappan et al. 2015).

The present study was undertaken to examine the genome-wide occurrence of H3K4/K27 trimethylations and their effect on expression of downstream genes following inoculation by the most virulent race 77-5. A pair of NILs involving susceptible wheat cv. HD2329 and its resistant NIL HD2329 + Lr28 was used to study the effect of the presence/absence of the gene Lr28 on the occurrence of H3K4me3/K27me3 and resulting regulation of genome-wide expression of genes. A large number of genes were found to be differentially expressed due to binding of these two H3 trimethylation marks.

Materials and methods

Plant material and sample preparation

The details of the plant material used in the present study are available in an earlier report (Sharma et al. 2018a). In brief, the seedlings of susceptible cv. HD2329 and its resistant NIL (HD2329 + Lr28) were raised in a growth-chamber under controlled conditions of 16 h light (240 mmol m−2 s−1) with 25 °C and 8 h dark with 18 °C at the National Phytotron Facility, ICAR-IARI, New Delhi. The seedlings (14 days old) were inoculated by the most prevalent and virulent pathotype 77-5 (syn. 121R63-1) of Puccinia triticina Erikss. & E. Henn. (for details, see Sharma et al. 2018a), and leaf samples were collected as detailed below.

Eight leaf samples from seedlings included two biological replicates each for the following four treatments: (i) S0 (susceptible cv. HD2329 at 0 hbi), (ii) S96 (susceptible cv. HD2329 at 96 hai), (iii) R0 (resistant NIL HD2329 + Lr28 at 0 hbi) and (iv) R96 (resistant NIL HD2329 + Lr28 at 96 hai). Leaf samples were collected and fixed in formaldehyde buffer (1% formaldehyde, 0.4 M sucrose, 10 mM Tris–HCl, 1 mM EDTA, 1 mM PMSF) followed by vacuum infiltration.

Chromatin immunoprecipitation (ChIP) and preparation of ChIP-seq libraries

Antibodies specific for H3K4Me3 and H3K27Me3 marks were procured from Abcam, USA. The workflow for ChIP and preparation of ChIP-seq libraries included the following steps: (i) chromatin was isolated (using Digenode Plant ChIP-seq kit) from all the eight samples (as above); (ii) isolated chromatin was sheared and immuno-precipitated using each of the above two antibodies, thus making 16 samples; input DNA (total genomic DNA used as control) was also isolated from the above four samples (without biological replicates) following precipitation of chromatin using Digenode Plant ChIP-seq Kit; in this manner 20 samples were available. (iii) DNA was isolated from all the 20 ChIPed samples and was used for constructing ChIP-Seq libraries using Microplex Library Prep Kit V2; (iv) Quantity and quality check (QC) of each amplified library was done by Bioanalyzer2100 (Agilent Technology) using high sensitivity (HS) DNA chip based on Qubit concentration and mean peak size; (v) sequencing of 20 DNA libraries was outsourced to Xcelris Genomics Pvt. Ltd, Ahmedabad (India), who used Illumina platform for cluster generation and sequencing; paired-end (PE) sequencing (2 × 75 bp) was undertaken, which allowed template fragments to be sequenced in both forward and reverse directions. Sequencing data was generated with ~ 11–90 million reads/sample. Data in FASTQ files for all the 20 libraries was subjected to detailed bioinformatics analysis (for details of 20 libraries, see Table S1).

ChIP-seq data analysis

The filtered reads from each library were mapped to wheat reference genome-Ref-Seq v1.0 (IWGSC 2018; https://urgi.versailles.inra.fr/download/iwgsc/IWGSC_RefSeq_Assemblies/v1.0/) using Bowtie2.0 (Langmead and Salzberg 2012). Pearson’s correlation coefficients were calculated for all replicates/treatments to check reproducibility in the aligned reads between the replicates by deepTools (Ramírez et al. 2014). Scores for fraction of reads in peak (FRiP) were also calculated for each replicate for each treatment; > 1% FRiP was used as a cutoff. Peaks marked with H3K4me3 and H3K27me3 were identified using MACS2.0 (Model-based Analysis of ChIP-Seq by Zhang et al. 2008). The narrow/sharp peaks represented H3K4me3 and broad peaks represented H3K27me3 (Rintinch et al. 2014; Nakato and Shirahige 2017; Laczik et al. 2016). The peaks were localized into exon, intron, promoter (− 1 kb to + 100 bp), TTS—Transcription Termination Sites (− 100 bp to + 1 kb) and intergenic regions using HOMER (Hypergeometric Optimization of Motif EnRichment; Heinz et al. 2010). Differential binding sites (DBSs) for the two histone marks were extracted by using DiffBind package of Bioconductor (Stark and Brown 2011) at P value < 0.01. Perl-script was used to find known high confidence genes using IWGSC RefSeq v1.0 data; some genes were found overlapping with DBSs for both histone marks. In order to study their role in wheat-leaf rust interaction, the high confidence genes were categorized into the following 10 categories based on their functions in biotic stress response: (i) resistance, (ii) transcription factor (TF), (iii) oxidative stress response, (iv) transport, (v) photosynthesis, (vi) epigenetic regulation, (vii) metabolism, (viii) defense, (ix) plant hormone response, and (x) miscellaneous class.

The ChIP-Seq raw reads of susceptible genotype (HD2329) and resistant NIL (HD2329 + Lr28) at 0 hbi and 96 hai were deposited in Sequence Read Archive (SRA), National Center for Biotechnology Information (NCBI) and can be accessed through Bioproject Database under the accession number PRJNA588134.

Gene ontology analysis and identification of transposable elements (TEs)

Blast2GO (Conesa and Götz 2008) was used for functional annotation of genes associated with DBSs. Based on annotation, the genes were placed into the three well-known classes: cellular component, molecular function and biological process. The DBSs were also searched for TEs utilizing IWGSC RefSeq using Perl script.

Co-localized DBSs for H3K4me3 and H3K27me3

A DBS was considered co-localized for both the marks only if peak coordinates for both H3K4me3 and H3K27me3 had an overlap of at least one base. Genes were considered as co-localized, only if both marks occur within the gene, even if the marks do not overlap.

RNA-seq analysis and relationship of histone methylation with gene expression

In order to examine the relationship of histone methylation with gene expression, RNA-seq analysis was also conducted for the same four treatments (S0, S96, R0 and R96). For this purpose, RNA samples from two biological replicates of each treatment were pooled and the sequencing was outsourced to Xcelris Genomic Pvt. Ltd (pooling was done to reduce sequencing costs). The libraries were prepared with ~ 1 μg input total RNA using Illumina TruSeq Standard Total RNA Library Preparation Kit as per the manufacturer’s protocol. Briefly, the protocol involved the following steps (i) total RNA was Ribo depleted using plant rRNA removal mix and rRNA removal beads; (ii) RNA was subjected to purification, fragmentation and priming for cDNA synthesis; (iii) Ribo-depleted and fragmented RNA was converted into first-strand cDNA, followed by second-strand cDNA synthesis, A-tailing, adapter-index ligation; (iv) ligated RNA was amplified using recommended number of PCR cycles; (v) library quality and quantity was checked using Agilent DNA High Sensitivity Assay Kit. The pipeline used for bioinformatics analysis and identification of differentially expressed genes is presented in Fig. S1. For identification of differentially expressed genes (DEGs), log FC > 0 criterion was followed. The RNA-seq raw reads of susceptible genotype (HD2329) and resistant NIL (HD2329 + Lr28) at 0 hbi and 96 hai were deposited in Sequence Read Archive (SRA), National Center for Biotechnology Information (NCBI) and can be accessed through Bioproject Database under the accession number PRJNA588134.

In order to examine the relationship of histone methylation and gene expression, the genes associated with DBSs were compared with differentially expressed genes (DEGs) identified from RNA-seq data (as above). A DEG was considered overlapped with DBS only if it overlaps with peak coordinates at least by one base. This allowed us to identify genes, which were associated with modified histone marks and exhibited differential expression, assuming activation due to H3K4me3 and repression due to H3K27me3.

Relationship of histone methylation with DNA methylation

For examining the relationship of histone methylation with DNA methylation, differentially methylated genes (DMGs) identified using mathylated DNA immuno precipitation (MeDIP) analysis from our earlier study (Saripalli et al. 2020) were compared with the differentially histone modified (DH3M) genes. The MeDIP raw reads of susceptible (HD2329) and resistant (HD2329 + Lr28) NILs at 0 hbi and 96 hai were deposited in Sequence Read Archive (SRA), NCBI and can be accessed through Bioproject Database under the accession number PRJNA517568.

Chromatin states and genome annotation analysis

In order to study the chromatin states using epigenomic information (DNA methylation, H3K4me3, H3K27me3), ChromHMM (Ernst and Kellis 2012) was utilized. This exercise involved the following steps; (i) file for chromosome size was prepared and COORDS and ANCHORFILES for wheat were prepared utilizing IWGSC v1.1 annotation file; Perl scripts were used for this purpose; (ii) genome wise Cellmarkfiletable was prepared utilizing controls for histone marks; (iii) binarization was performed on BAM files by BinarizeBam separately for DNA methylation and histone modifications; (iv) binary files for DNA methylation and histone modifications were merged chromosome-wise; (v) model was learned using LearnModel for 15 chromatin states, which are biologically interpretable.

Validation by quantitative RT-PCR (qRT-PCR)

Identification of genes for qRT-PCR

Out of the genes associated with DBSs, 26 genes showing FDR < 0.05 or P < 0.01 with a FC > 2.5 or <  − 2.5 were selected for validation using qRT-PCR. These 26 genes included 17 genes for H3K4me3 and 9 genes for H3K27me3. Primers for the above genes were designed using Primer Express (ver. 3.0) software (Applied Biosystems, USA). Details of primers are available in Table S2.

RNA isolation, cDNA synthesis and qRT-PCR

For qRT-PCR, total RNA was isolated using TRIzol reagent (Ambion, USA) following manufacturer’s guidelines and treated with RNAse-free DNAse I (Invitrogen, USA) for 15 min to degrade any residual genomic DNA. First-strand cDNA was synthesized from this RNA using RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific, USA) according to manufacturer’s instructions. qRT-PCR was performed in 96 well plates with Step OnePlus Real Time PCR system using SYBR Green (both from Applied Biosystem, USA). All reactions were performed using three biological and three technical replicates. Target gene expression (fold-change between two treatments) was calculated using 2−∆∆CT method using actin gene for normalization (Livak and Schmittgen 2001). Heat maps were generated using online tool clustvis, and were used to depict the fold changes in differential binding/expression of genes following qRT-PCR/ChIP-Seq.

Results

ChIP sequencing and alignment with wheat genome

ChIP sequencing data from 20 libraries (including 16 IpDNA and 4 input DNA) had ~ 950 million paired end reads with an average of ~ 48 million reads/sample. An average of 89.88% reads from each library were mapped with wheat reference genome. The correlations between aligned reads of two replicates ranged from 0.54 to 0.85 (Table S1); the correlation between different treatments was also high (> 0.5); this was not unexpected, since the susceptible (S) and resistant (R) lines differed only for Lr28 gene and had almost similar genetic background (Fig. S2).

Differential binding sites (DBSs) with modified histone marks

Chromosomal distribution of DBSs

Differential binding sites (DBSs) for the two histone marks were distributed on all the 21 chromosomes (Fig. 1). Some DBSs (1061) could not be assigned to specific individual chromosomes and were labelled as ‘chromosome Un’. For H3K4me3, the number of DBSs ranged from 251 on the unidentified chromosome (Un) to 455 on chromosome 2A; for H3K27me3, this number ranged from 14 on chromosome 4D to 559 on the unidentified chromosome (Un). The number of DBSs for H3K4me3 in individual treatment pairs ranged from 296 (S96 vs R96) to 4428 (S0 vs. S96); for H3K27me3, it ranged from 295 (S96 vs. R96) to 733 (S0 vs. R0).

Fig. 1
figure 1

Distribution of DBSs in different treatment pairs on all the 21 chromosomes (1A to 7D) and the DBSs unassigned to any of the 21 chromosomes are indicated as Un. The DBSs indicated in each of the two histone marks, starting from outermost circle, (i) first circle indicates the treatment pair S0 vs S96, (ii) the second circle: S0 vs R0; (iii) the third circle S96 vs R96 and (iv) fourth circle R0 vs R96. Bars facing outwards indicate increased affinity and bars facing inwards indicate decreased affinity to H3K4me3 in outer four circles in red, and to H3K27me3 marks in inner four circles in blue. S0: HD2329 at 0 hbi; S96: HD2329 at 96 hai; R0: HD2329 + Lr28; R96: HD2329 + Lr28 at 96 hai

Full size image

DBSs in S (HD2329) and R (HD2329 + Lr28) lines

For H3K4me3, DBSs showing increased affinity at 96 hai (relative to 0 hbi) in each of the two NILs (S0 vs. S96 and R0 vs. R96) were abundant, relative to those between two NILs (S0 vs. R0 and S96 vs. R96). In the former two comparisons involving the same NILs at two time points, when comparisons were made between S and R lines, binding sites decreased at 96 hai in R NIL relative to S line with the passage of time (R0 vs. R96 relative to S0 vs. S96). For the H3K27me3, in S96 and R0, the binding sites showing decreased affinity were more abundant relative to S0, whereas in R96 the binding sites with increased affinity were more abundant relative to S96 and R0 (Figs. 1 and 2).

Fig. 2
figure 2

Binding sites showing increased and decreased affinity of histone marks: a H3K4me3 and b H3K27me3 in each of the four treatment comparison; in each comparison, second treatment is compared to first

Full size image

Distribution of DBSs in genomic regions

Relative frequencies of DBSs in five genomic regions (promoter, exon, intron, TTS and intergenic regions) for H3K4/K27me3 are depicted in Fig. 3. The DBSs of H3K27me3 were more frequently found in the intergenic region (84–88%) than in genic and promoter regions in comparison to H3K4me3 (18–23%). Binding of the H3K4me3 mark occurred with higher frequency in the exons relative to introns. In contrast, the binding of H3K27me3 in exons and the introns did not differ. Details of DBSs in exons, intron, promoter and TTS for both the marks are available in Tables S3–S10; whereas details of DBSs in intergenic regions for both the marks are available in Tables S11–S18. DBSs in the intergenic regions (associated with H3K27me3) were found to be involved in trans-regulation of genes that were located at > 50 kb away (upstream as well as downstream).

Fig. 3
figure 3

Proportions of different genomic regions, occupied by DBSs for two histone marks: a H3K4me3 and b H3K27me3 in each of the four treatment comparisons

Full size image

Differentially modified genes due to H3 histone marks (DH3M genes)

Differential binding sites (DBSs) other than those falling in intergenic regions were used for identification of differentially H3 modified (DH3M) genes; IWGSC reference sequence was used for this purpose. Many more DH3M genes (3869) were associated with H3K4me3 relative to those associated with H3K27me3 (186). Majority of DH3M genes in individual treatment pairs were unique; only a few genes occurred in more than one treatment pairs (Fig. 4; prepared based on protein domain information). For instance, if we consider the two important comparisons, namely S0 vs S96 and R0 vs R96, there were 2259 unique DH3M genes associated with H3K4me3 and 53 unique DH3M genes associated with H3K27me3 in S96 (relative to S0) whereas in R96 (relative to R0), 800 unique DH3M genes associated with H3K4me3 and 9 unique DH3M genes associated with H3K27me3. However, some of these unique DH3M genes shared common domains indicating their similar functions. Based on their domains, the DH3M genes were classified into 10 categories (Fig. 5). Some important categories of DH3M genes that may be used for further study of wheat-leaf rust interaction are listed in Table S3. Maximum DH3M genes associated with H3K4me3 belonged to the miscellaneous category in all the four treatment pairs; minimum number of DH3M genes belonged to growth hormone response category. However, in case of H3K27me3, many more genes belonged to defense response and metabolism categories in all the four treatment pairs.

Fig. 4
figure 4

Venn diagrams showing number of DH3M genes marked with a H3K4me3 and b H3K27me3. These venn diagrams have been prepared using the protein domains encoded by the genes. Therefore, the genes with the common domains in the same treatment pair have been omitted. The detailed list of gene IDs has been presented in Tables S3–S10

Full size image
Fig. 5
figure 5

Ten and nine different categories of genes identified in DBSs associated to a H3K4me3 and b H3K7me3 in four different treatment pairs, respectively

Full size image

Gene ontology analysis of DH3M genes

GO terms of DH3M genes for H3K4/K27me3 are shown in Fig. 6a and b. The DH3M genes can be broadly classified as follows: (i) DH3M genes for H3K4me3: in this category, the genes involved in binding activity were most abundant followed by catalytic activity, metabolic process and cellular process when compared to others. (ii) DH3M genes for H3K27me3: in this category, genes involved in binding activity, catalytic activity, metabolic processes, cellular processes and those belonging to different cellular components were predominant (for details, see Tables S19–S26).

Fig. 6
figure 6

GO analysis of DH3M genes associated with a H3K4me3 and b H3K27me3. The GO categories are distributed in three categories, namely, (i) cellular component (ii) molecular function and (iii) biological process

Full size image

Transposable elements (TE) in DBSs

TEs in DBSs were identified and the results are summarized in Table 1; as can be seen, TEs in each case could be placed in two categories, TEs with nested repeats (NR) and those with simple repeats (SR). It can also be seen that for H3K4me3, maximum number of TEs (775) were available in DBSs from the treatment S0 vs. S96 and minimum TEs (65) were available in DBSs of S96 vs. R96. The corresponding figures for H3K27m3 were 158 in S0 vs. R0 and 90 in S96 vs R96.

Table 1 Transposable elements in DBSs for H3K4me3 and H3K27me3
Full size table

Chromatin states associated with H3K4/K27me3 marks and DNA methylation

The results of analysis of chromatin states involving three epigenetic marks (DNA methylation, H3K4me3, H3K27me3) and four treatments (S0, S96, R0, R96) are depicted in Fig. 7. As many as 15 chromatin states were identified, which provided sufficient resolution to understand biologically meaningful patterns, across four different treatments. We used this model to produce five genome-wise chromatin state annotations, which are summarized in Table 2 and depicted as heat maps in Fig. 7. From this figure, following notable features of the analysis involving chromatin states were observed: (i) chromatin state 2, which represents repressed enhancers located 2 kb upstream of the TSS carrying H3K27me3, is predominant in the treatment R96; (ii) chromatin state 8, which represents transcribed genic region enriched with DNA methylation in TTS seems to be common in all the four treatments (S0, S96, R0 and R96), although the probability seems to be relatively low in R0; (iii) chromatin state 10, which represents repressed genic region with all the three epigenetic marks is common in all the four treatments; (iv) chromatin state 12 involving strong promoter enriched with H3K4me3 in TSS, seems to be common in S0, S96 and R0, but its probability is rather high in S0; (v) chromatin state 13 having the same attributes as state 12 is particularly common in R96; (vi) chromatin state 14 is particularly common in S0 suggesting the presence of weak promoters in many genes marked with H3K4me3; (vii) chromatin state 15 involving repetitive region enriched for DNA methylation and H3K4me3 seems to have high probability in R0.

Fig. 7
figure 7

a A heatmap of emission probabilities, where each row represents a different chromatin state, and each column represents a different epigenetic mark in individual treatments; the darker regions correspond to a greater probability of observing the mark in the state; b heat map of emission probablities displaying enrichment for six different external genomic annotations based on RefSeq and 15 chromatin states; c a heat map of transition parameters, where each row and column represents a chromatin state

Full size image
Table 2 Showing class of chromatin states with their mark and region of enrichment
Full size table

Differentially expressed genes identified using RNA-seq

RNA-seq analysis produced > 105 million reads for all the four treatments with maximum number of reads obtained for S96 (2,989,116). The mapping efficiency of these reads ranged from 89.19% (R96) to 93.49% (S0). The DEGs in the treatment pairs S0 vs R0 and R0 vs R96 were nearly similar but were relatively higher when compared with the remaining two treatment pairs, namely S0 vs S96 and S96 vs R96 (Table 3). A large number of genes were also expressed in only one of the two treatments of a pair in all the four treatment comparisons. The DEGs largely belonged to the following different classes of genes: (i) genes encoding receptor like kinases, TFs, and important enzymes involved in metabolic pathways, (ii) growth hormone responsive genes, defense response R genes, etc.

Table 3 Details of DEGs identified in all the four different treatment comparisons
Full size table

Relationship between histone trimethylation and gene expression

As many as 346 DH3M genes associated with H3K4me3 marks and 34 genes associated with H3K27me3 marks were also found to be differentially expressed during RNA-seq analysis. Summary of a comparison of number of DH3M genes during ChIP-seq analysis which were also found to be differentially expressed using RNA-seq is given in Table 4 (for details see Tables S27 and S28). It can be seen that among DH3M genes for H3K4me3, 232 DH3M genes with increased affinity showed high expression and 9 DH3M genes with reduced affinity showed lower expression. Similarly, for H3K27me3, two DH3M genes with increased affinity showed lower expression and 14 DH3M genes with reduced affinity showed higher expression. An opposite relationship was observed for the remaining 123 DH3M genes (105 associated with H3K4me3 and 18 associated with H3K27me3). Overall, only ~ 0.6% of the DEGs (identified using RNA-Seq) were found to be influenced due to histone modifications (both H3K4/K27me3).

Table 4 Comparison of binding affinity of DBSs for H3K4me3 and H3K27me3 with expression of genes
Full size table

Co-localized or bivalent DBSs for H3K4me3 and H3K27me3

As many as 14 DH3M genes associated with 125 DBSs exhibited association with both, H3K4me3 and H3K27me3; these were described as colocalized/bivalent DBSs (Table 5); the distribution of these 14 genes in treatment pairs was as follows: (i) 4 genes (having decreased affinity with both the marks) belonged to S0 vs R0; (ii) 6 genes (showing opposite affinities with the two marks) belonged to S0 vs S96; (iii) 3 genes (showing increased affinity to both the marks) belonged to R0 vs R96 and (iv) 2 genes (showing decreased affinity to both the marks) belonged to S96 vs R96.

Table 5 Co-localization of H3K4me3 & H3K27me3 within DBSs and DH3M genes
Full size table

Validation of identified genes by qRT-PCR

As mentioned earlier, a representative sample of 26 genes was used for qRT-PCR. Among 17 genes marked with H3K4me3 and used for qRT-PCR, expression of only 6 genes could be validated (Fig. 8a). Similarly, among 9 genes marked with H3K27me3 and used for qRT-PCR, expression of only 5 genes (Fig. 8b) could be validated. For H3K4me3, among six validated genes, two genes with increased affinity showed enhanced expression, while remaining four genes with reduced affinity showed reduced expression. Similarly, for the H3K27me3 three genes with increased affinity showed reduced expression while remaining two genes with reduced affinity showed enhanced expression. The functions of these genes validated using qRT-PCR are summarized in Table 6.

Fig. 8
figure 8

Heatmap showing relationship of histone modifications due to a H3K4me3 mark and b H3K27me3 (using ChIP-Seq) with gene expression (using qRT-PCR). Genes highlighted in yellow were validated using qRT-PCR analysis

Full size image
Table 6 Genes marked with H3K4/K27me3 for which qRT-PCR results could be validated along with their function
Full size table

Discussion

Molecular mechanism for disease resistance in wheat-leaf rust pathosystem is not fully understood; this is particularly true of the signal transduction pathways that follow the initial interaction between the protein encoded by an individual Lr gene and the corresponding effector produced by the pathogen. It is, however, known that an attack by the leaf rust fungus (P. triticina) is perceived by the host cells through cell-surface and intracellular receptors; the cell surface receptors have recently been shown to form a resistosome, which provides defense through cell death mediated hypersensitive reaction, the so called HR (Wang et al. 2019a, b, c). The perception signal also triggers differential activation/repression of a large number of downstream genes involved in more than one signal transduction pathways. Differential expression of these downstream genes has been a subject of intensive research. It has also been shown that the differential expression of genes is partly regulated by epigenetic modifications. The present study is a part of such studies, where role of two H3 histone modification marks (H3K4me3 and H3K27me3) in regulation of expression of these downstream genes of the host has been examined. For this purpose, we used a susceptible cv. HD2329 and its resistant NIL, the latter carrying the leaf rust resistance gene Lr28 that has been widely used for imparting resistance against leaf rust in many Indian cultivars. This gene, Lr28 which is located on chromosome arm 4AL was transferred from Aegilops speltoides and could not be cloned and characterized so far. However, in one of our own studies involving RNA-seq analysis using the same experimental material (Sharma et al. 2018a) and based on highly expressed genes identified on chromosome 4AL in resistant NIL, this gene was speculated to encode an ATP binding serine threonine protein kinase (receptor like kinase).

It is widely known that the epigenetic control of gene expression is broadly mediated by DNA methylation, ncRNAs and histone modifications. The present study is the first study, which involved identification of downstream genes, whose expression and crosstalk is regulated by the association of H3K4/27me3, primarily due to presence/absence of the gene Lr28. The study involved genome-wide ChIP sequencing to identify genomic regions, the so-called differential binding sites (DBS), associated with each of the two histone marks.

While studying association of modified histones, generally we do not know whether the histone marks are the cause or the result of activation or repression of relevant genes. Available information indicates the prevalence of both the possibilities. Alternatively, the histone marks may simply cause uncoiling or coiling of DNA in the form of nucleosomes, thus influencing expression of genes rather indirectly. For instance, available evidence indicates that H3K4me3 may not be the cause, but only the result of active transcription, since it has been shown that loss of H3K4me3 from genes involved in active transcription does not lead to reduction in transcription of differentially expressed genes (reviewed in Howe et al. 2017). In rice, it has also been shown that H3K4me3 is strongly associated with active transcription and removal of H3K4me3 by histone demethylase leads to gene repression (Chen et al. 2013).

In view of the above, we feel that caution should be exercised in interpreting the results of ChIP-seq and in comparing these results with the results of gene expression obtained from RNA-seq. During the present study also, the expected results of gene expression available from RNA-seq data do not always match with the results of ChIP-seq, if we assume that H3K4me3 is an activation mark and H3K27me3 is a repression mark. Such unexpected results were also reported earlier in common bean (Ayyappan et al. 2015), rice (Zong et al. 2013) and Arabidopsis (Brusslan et al. 2012).

DBSs for the two histone marks (H3K4/K27me3)

In the present study, a large number of DBSs in the genome were available for H3K4me3; the frequency of DBSs for H3K27me3 were fewer suggesting that relatively more genes are activated than repressed for infection to be successful. Also, there seem to be many more DBSs (4379) in S line relative to R line (2184), that are associated with H3K4me3 relative to those which are associated with H3K27me3, suggesting that in the S line, many more genes are activated relative to those which are repressed (Fig. 2). The results of H3K27me3 are largely in agreement with this conclusion, except that with the passage of time, there is many fold decrease in the number of DBSs for both the marks, suggesting that different sets of almost equal number of genes perhaps get downregulated and upregulated in the R line also. This also suggests complexity of the interacting genes following inoculation in both NILs.

Genomic distribution of DBSs

Distribution of DBSs in the different genomic regions (promoters, exons, introns, TTS and intergenic regions) also had some distinctive features including the following (Fig. 3; for details see Tables S3–S18).

  1. (i).

    DBSs in promoter regions The frequency of DBSs for H3K4me3 associated with promoter regions was much higher than those with H3K27me3. This is partly in agreement with the available results for yeast, fruit fly, Arabidopsis and mouse (Yang and Wilson 2018). This observation was further supported by chromatin state analysis (Fig. 7a). For instance, the chromatin state 12 mainly involving promoter regions associated with H3K4me3 in the TSS was common in S0, S96 and R96; probability being particularly high in S96. This suggested activation of a number of genes in susceptible cultivar due to binding of H3K4me3 in promoter region. High expression of such genes (with high affinity DBSs in promoter region) was also obtained in RNA-seq analysis; some of the proteins encoded by these genes include the following: ABC transporter family, polyubiquitin, sodium hydrogen exchanger, F-box protein, alpha/beta hydrolase, etc. While comparing these results with those of chromatin state analysis, it was observed that chromatin state 13 was exclusively enriched for strong promoter with H3K4me3 in R96 indicating the activation of some important genes perhaps involved in providing resistance. In contrast, the chromatin state 2 involving repressed enhancers located 2 kb upstream of TSS carrying H3K27me3 was predominant in S0, although it was present in other treatments also. This indicated repression of genes, which need to have low expression in susceptible genotype in the absence of the pathogen; these genes mainly encoded proteins for photosystem II, Ankyrin repeats, pentatricopeptide repeats, etc.

  2. (ii)

    DBSs in intergenic regions The frequency of DBSs located in the intergenic regions was much higher for H3K27me3 (84–88%) relative to those for H3K4me3 (28–46%); these results are not very different from those recently reported for mouse (Yang and Wilson 2018). Earlier in wheat also, similar situation was observed with these histone marks indicating that these marks may also be involved in trans-regulation of gene activity (Li et al. 2019). The abundance of the DBSs for H3K27me3 in the intergenic region may also explain the possible repression of transcription of the transposable elements (TEs), thus leading to genome stability (Walter et al. 2016). For some of DBSs in the intergenic regions, the frequencies were similar for both the epigenetic markers; these results are also in partial agreement with the results of chromatin state analysis, since chromatin state 4 showed high enrichment of both the marks in the intergenic regions. Later, we discuss the role of intergenic regions in trans-regulation of gene expression.

  3. (iii)

    DBSs in exons, introns and TTS The proportions of DBSs located in exons, introns and TTS for the two histone marks also differed with relative abundance of H3K4me3. Thus, overall, H3K4me3 mark is more spread out in the genome whereas H3K27me3 is mostly restricted to intergenic region. The distribution of the binding sites for activation (H3K4me3) and repression (H3K27me3) marks in the present study in wheat is in agreement with the distribution of the other activation (H4K12ac) and repression (H3K9me2) marks in the genome examined during rust infection in cowpea (Ayyappan et al. 2015). On individual chromosomes also, the DBSs for H3K4me3 are distributed all along the length of individual chromosomes whereas the DBSs for H3K27me3 had more clustered distribution (see Fig. 1).

Mechanism of action of H3K4me3 and H3K27me3

In view of the above, we believe that the mechanism of action of the two histone marks may differ. In fact, H3K27me3 and the associated intergenic sequences (> 50 kb from the DBS) may be involved in trans-regulators (for list of genes see Tables S11–S18). In contrast, H3K4me3 may facilitate the associated gene sequences in performing their function as cis-regulatory elements. The interactions between specific intergenic regions associated with H3K27me3 and distantly located genes, may actually be facilitated through at least following two different known mechanisms: (i) The folding of chromatin in space through an unknown mechanism may bring the DBSs in the intergenic region in close proximity to the genes to be trans-regulated; this may be further investigated through a study of 3D structure of chromatin conformations using the available techniques like 3C (Dekker et al., 2002; Miele and Dekker 2009), Hi-C (van Berkum et al. 2010) and Dip-C (Tan 2019). (ii) Trans-regulation through DBSs in the intergenic region carrying H3K27me3 may also involve lncRNAs, which are known to be derived from non-coding regions of the genome and act in trans by recruiting polycomb repressor complex 2 (PRC2), which mediates methylation of H3K27me3; similarly, PRC1 seems to limit the access of chromatin to transcription factors by binding with other nucleosome, thus affecting the expression of a distant gene (Lehmann et al. 2012; Brockdorfi 2013; Roberts et al. 2013). LncRNA mediated regulation of leaf rust disease resistance genes is also being investigated by us in a separate study.

Expression of DH3M genes due to H3K4/K27me3 in S and R lines

During the present study, as many as 2259 unique DH3M genes (out of a total of 2565 DH3M genes; based on the protein domains) associated with H3K24me3 were identified in the susceptible cultivar. In contrast, only 800 unique DH3M genes (out of total 1200 DH3M genes) were associated with H3K4me3 in resistant NIL. These DH3M genes must have been active, although the exact role of H3K4me3 following an attack by the pathogen cannot be inferred (Fig. 4). Further analysis also revealed that the number of DH3M genes that were associated with H3K4me3 differed in susceptible (S) and resistant (R) NILs. We like to discuss these results under four different categories of genes. (i) First, in S lines, the number of genes activated are almost three-fold the number of genes that are activated in the R line. That susceptibility accompanies activation of relatively large number of genes was observed in our earlier studies also, which involved either the transcriptome studies involving APR gene Lr48 or DNA methylation (MeDIP) studies involving Lr28; this activation may partly be attributed to association with H3K4me3, although loss of DNA methylation could be another factor, as reported by us in another study (Saripalli et al. 2020). In fact, a complex network of genes may be involved, which needs to be elucidated. (ii) Second, in R line at 96 hai, activation of some genes and repression of some other genes seem to play important roles in providing resistance. This was apparent from the results of differential binding of H3K4me3 and H3K27me3 in R96 line relative to those in either R0 or S96. This conclusion is based on abundance of sites showing either increased affinity or decreased affinity for H3K4me3 relative to those for H3K27me3. (iii) Third, S line and R line had 245 common DH3M genes associated with H3K4me3 (no such common genes were available for H3K27me3). These genes must be involved in important functions but may not be causal in determining resistance or susceptibility, unless trans-regulatory elements were involved, which could not be examined in the present study. The role of these genes shared by both NILs need to be subjected to further study. (iv) Fourth, in both S and R lines, there were many more unique DH3M genes associated with H3K4me3 relative to those associated with H3K27me3; these included 53 in S96 (relative to S0) and mere 9 in R96 (relative to R0)]. The implications are that relatively many more genes are activated than repressed in both the reactions (S0 vs S96 and R0 vs R96). The factors involved in activation of these genes will have to be elucidated, which at least in some cases may be due to histone H3 acetylation, as shown by us in our earlier study (Sharma et al. 2018b). However, some unique DH3M genes in the two opposite interactions perform different functions in the compatible and incompatible interactions, although these unique genes may contain one or more common domains.

The genes regulated by two histone modifications in S and R genotypes belonged to 10 different categories that are depicted in Fig. 5. The same 10 categories (but not necessarily the same genes) were also noticed for genes that were found to be differentially expressed in our earlier study on transcriptome analysis (Sharma et al. 2018a) and differentially methylated in our earlier study on MeDIP analysis (Saripalli et al. 2020). A more detailed study of the functions of DH3M genes (inferred through ChIP-Seq) and their assignment to three well known categories of functions (cellular, molecular and biological) suggested that majority of these genes were involved in catalytic or binding functions and metabolic (molecular functions) or cellular/biological processes (Fig. 6).

Relationship of histone modifications with gene expression (RNA-seq)

It is apparent from the data presented in Table 4 that a very small fraction of genes associated with the two histone marks (7% and 12%) exhibited the expected expression patterns in the transcriptome (RNA-seq) data. This is not surprising, keeping in view that a complex network of regulatory systems controls the expression of genes, and that the two histone marks used in the present study make only a tiny fraction of the overall regulatory system involving DNA methylation, siRNA and other histone marks (Takahashi et al. 2018; Sharma et al. 2018b). Similar comparisons were earlier made in rice and Arabidopsis; this match accounted for 24–26% genes in rice, under drought stress involving H3K4me3 mark (Zong et al. 2013) and 10–15% genes in Arabidopsis for leaf senescence involving H3K4/K27me3 marks (Brusslan et al. 2012). A number of studies in mammals also did not find a match between H3K4me3 binding and RNA-Seq data (for a review see Howe et al. 2017). According to some studies, H3K4me3 binding and transcription are two independent events; H3K4me3 seems to be involved in post-transcriptional events like splicing, transcription, termination, etc. (Howe et al. 2017). H3K4me3 has also been shown to be involved in other functions like DNA repair and targeting of double stranded breaks during meiotic recombination (Howe et al. 2017). In the present study also, a number of H3K4me3 associated genes were found to contain important domains involved in DNA repair (e.g. flap endonuclease) and targeting double stranded breaks. Thus, exclusive role of H3K4me3 in gene activation seems to be doubtful and needs detailed further investigation.

Co-localization of H3K4/27me3 marks (bivalent chromatin)

In a very small proportion of genes (14 out of 5328/305 genes; for complete list of DH3M genes and their gene IDs, refer to supplementary Tables S3 to S10); co-localization of the two histone marks (i.e. bivalent domains) was observed in coding/promoter regions (Table 5). Co-localization of histone H3 marks has also been reported in FLC gene for flowering, camalexin biosynthesis genes and a large number of other genes in Arabidopsis (Jiang et al. 2008; Zhang et al. 2009; Zhao and Rhee 2018); many more reports of this co-localization of the two histone H3 marks are available in mammalian genomes (Sachs et al. 2013; Dattani et al. 2018; Vestenhouw and Schier 2012). It has been also suggested that in these bivalent domains, the association of H3K4me3 with an hitherto inactive gene may help in subsequent activation. Some reports involving plant systems also suggest that transcription can take place even in the presence of repressive marks like H3K27me3, which is not sufficient for silencing of genes (Schubert et al. 2006; Adrian et al. 2010). In other cases, a hitherto active gene due to H3K4me3 may also be repressed due to subsequent association of H3K27me3 (Akkers et al. 2009). Together, the expression of genes in the presence of bivalent histone marks seems to be a complex process.

Even though the genes with co-localized histone marks in this study constitute only a tiny fraction of the total genes associated with DBSs, some of these genes also have a role in binding of colocalized histone marks. Such genes included genes like those encoding PHD domains (Qian et al. 2018); these genes showed increased affinity with H3K4me3 in the present study. Therefore, efforts have been made to study the mechanism involved in binding of two histone marks on the same site. For instance, it is known that certain proteins with domains described as ‘writer’, ‘eraser’ and ‘reader’ may be involved respectively in adding, removing and recognizing modified histones. In a recent study, Qian et al. (2018) demonstrated that a plant-specific histone ‘reader’ called SHORT LIFE (SHL) recognizes both H3K4me3 and H3K27me3, the former with the help of its domain PHD (plant homeodomain) and the latter with the help of its domain BAH (bromo-adjacent homology).

In the resistant NIL, colocalization of the two chromatin marks was observed for four genes encoding the following proteins: U-box protein (E3 ligase), ubiquitin carboxyl-terminal hydrolase-2 (UCH2), protein containing coiled-coil domain (e.g. NLR) involved in a variety of functions including plant immunity and poly[ADP-ribose] polymerase (PARP) involved in plant immunity (Table 5). Interestingly U-box protein is involved in ubiquitination, whereas UCH2 belonging to the deubiquitinase (DUB) superfamily has a role in deubiquitination (see Zhou and Zeng 2017). Genes encoding these two proteins were also found to exhibit high expression in RNA-seq data during expression of resistance to leaf rust. In rice, ubiquitination (involving some of the U-box proteins) has been shown to promote host immunity by degradation of avirulence effector AvrPiz-t of Magnaporthe oryzae, although ubiquitination is known to often attenuate PTI signalling through proteasome dependent degradation of pattern recognition receptor (for a review, see Zhou and Zeng 2017). Similarly, UCHs (causing deubiquitintion) have been shown to have a role in both the biotic and abiotic tolerance. For example, UCH60 was reported in seeds harvested from field grown disease infested wheat (Kamal et al. 2010). Similarly, the UCHs also seem to be involved in providing tolerance to heat stress in an important vegetable chieh-qua (Benincasa hispidsa) and drought stress in maize (Hu et al. 2015; Wang et al. 2019a, b, c). Further, the deubiquitinating enzymes such as AtUBP12 and NtUBP12 through their deubiquitinating activity have been implicated in suppression of plant cell death in Arabidopsis and tobacco (Ewans et al. 2011). Together these observations suggest a possible role of ubiquitination and deubiquitination in leaf rust resistance. The remaining two genes, namely, gene with coiled-coil domain and PARP gene are widely known to be involved in plant immunity (Hu et al. 2017; De Block et al. 2005; Amor et al. 1998). Therefore, we assume that these genes must also be involved in resistance, directly or indirectly.

In the susceptible cultivar, the bivalent chromatin associated genes encoding three different disease resistance-like proteins (R genes) and S-acyltransferase were identified. Two of these R-like genes containing TIR-NBS-LRR and disease resistance protein with RPM1 domains showed high expression during RNA-seq using the same treatments as used for ChIP-seq. The H3K4me3 mark also showed increased affinity with these DH3M genes, suggesting the possible role of the epigenetic regulation of these genes conferring susceptibility (for more details, see later).

Relationship between H3 lysine methylation and cytosine DNA methylation

A comparison of results of the present study with the results of our earlier study on the role of cytosine DNA methylation using MeDIP (Saripalli et al. 2020) allowed identification of 43 genes in S and R reactions which were perhaps regulated by differential histone methylation as well as DNA methylation (Table 7). Information on the expression of these genes in our RNA-seq data was not available and therefore, we could not correlate the patterns of histone/DNA methylation with the pattern of expression of these genes. However, in our present study, nearly equal number of genes showed antagonistic and synergistic relationship with differential DNA methylation. Results of H3K4me3 and H3K27me3 will be separately compared with DNA methylation results. (i) H3K4me3 mark was earlier shown to have a positive relationship with DNA hypomethylation in Arabidopsis (Zhang et al. 2009; Mathieu et al. 2005). This hypomethylation may be due to inhibition of MET1 (methyltransferase 1) activity by H3K4me3, since MET1 is involved in CG methylation in plants (Finnegan et al. 1996; Saze et al. 2003). (ii) The relationship of H3K27me3 mark with DNA methylation is still poorly understood and both positive and negative relationship are reported. The DNA methylation is a negative modulator of PRC2-chromatin interaction, so that hypomethylation occurs in the presence of H3K27me3 (Mendenhall et al. 2010; Meissner et al. 2008). However, a positive relationship between DNA methylation and histone modification has actually been reported in humans, where genes that are de novo methylated in cancer cells were found to be enriched with H3K27me3 mark (Schlesinger et al. 2007). The chromatin-state analysis also showed a distinct pattern in chromatin state 10, where repressed genic regions were enriched with all the three epigenetic marks (H3K4me3, H3K27me3 and DNA methylation) in TTS region, although the genes associated with the two different marks differed in the present study (Fig. 7).

Table 7 Common proteins encoded by DH3Ms genes with increased affinity (↑) and decreased affinity (↓) and differentially methylated genes (hyper- (↑) or hypo-methylated (↓) identified using ChIP-Seq and MeDIP, respectively
Full size table

Important genes involved in wheat-leaf rust interaction

The present study also allowed us to identify some important genes that are differentially expressed due to association with H3K4/K27s; these genes encode proteins with domains involved in biotic stress response and can be grouped into the following five classes. (i) Genes with increased association with H3K4me3 and showing high expression in both susceptible and resistant lines following inoculation. (ii) Genes showing decline in association with H3K27me3 and showing high gene expression in susceptible cultivar or resistant NIL. (iii) Genes showing decline in association with H3K4me3 and showing low expression in resistant NIL. (iv) Genes showing increased association with H3K4me3 and hypomethylated in susceptible cultivar. (v) Genes associate with two co-localized histone marks including genes showing increased association with H3K4me3 and decline in association with H3K27me3 in susceptible genotype or increased association with both the marks in resistant NIL (genes in this category discussed above). The role of some of these important genes will be discussed in relation to susceptibility and resistance to leaf rust.

Genes involved in susceptibility

Some of the genes belonging to four of the above five classes are known to be involved in sensitivity/susceptibility to biotic stresses in general (Lorang et al. 2007; Faris et al. 2010; Gupta et al. 2018; Pessina et al. 2016; Ravichandran et al. 2015) (See Supplementary Tables S27 and S28; Tables 5 and 7; Fig. 9). The most important of these genes that are identified during the present study and may be involved in susceptibility to leaf rust include three genes that encode proteins containing three different known domains, namely CC-NBS LRR, TIR-NBS-LRR and RPM1; these genes show an increase in affinity with H3K4me3 and a decline in affinity with H3K27me3. Two of these three genes (CC-NBS-LRR and RPM1 containing genes) also showed high expression in susceptible cultivar. This indicated that during the compatible interaction, these genes are perhaps induced due to pathogen attack only (due to their increased affinity with activation mark H3K4me3); otherwise these genes remain repressed. Thus, these genes might have an important role in susceptibility and hence may actually be susceptibility genes (S genes) with NBS-LRR domains (Gupta et al. 2018; Faris et al. 2010; Pessina et al. 2016). Such S genes have also been shown to confer susceptibility to Cochliobolus victoriae in Arabidopsis (Lorang et al. 2007), and to necrotrophic pathogens in wheat (Faris et al. 2010; Gupta et al. 2018). However, no such S-genes conferring susceptibility against leaf rust pathogen, have so far been reported in wheat, although several sensitivity gene including Tsn1 conferring susceptibility to spot blotch and Tsc1 against tan spot have been reported (McDonald et al. 2018). When Tsn1 interacts with ToxA gene in the pathogen, it leads to susceptibility of wheat to spot blotch. This ToxA-Tsn1 system represent an inverse-gene-to-gene relationship (Nawathe et al. 2020). We speculate that such a sensitivity gene for leaf rust in wheat may be discovered in future. Such genes encoding purple acid phosphatases, receptor like kinases, cytochrome P450, F-box, Mlo like protein, etc. were up-regulated and also exhibited increased affinity with H3K4me3 in susceptible cultivar (S96), indicating their possible role in susceptibility. Over-expression of purple acid phosphatase in Arabidopsis was earlier shown to inhibit expression of PR1, accumulation of salicylic acid and increased accumulation of H2O2 enhancing susceptibility to Pseudomonas syringeae (Ravichandran et al. 2015). The gene encoding (Mlo)-like protein is another important gene whose dominant allele acts as a susceptibility gene for powdery mildew infection in barley (Buschges et al. 1997) and in grapevine, the RNAi mediated loss of function mutations in MLO6 and MLO7 genes contributed to powdery mildew resistance (Pessina et al. 2016). Another gene encoding for phloem protein A10 was found to exhibit high expression in susceptible cultivar as revealed through qRT-PCR analysis. Phloem proteins are actually the lipid binding proteins involved in lipid mediated signalling in plants (Barbaglia et al. 2016). However, their role in increased susceptibility towards leaf rust infection or other biotic stresses is not known.

Fig. 9
figure 9

Modified from our earlier study, Sharma et al. (2018a)

A network of genes based on the common genes with DBSs that are differentially expressed or differentially methylated during compatible (S0 vs S96) and/or incompatible interactions (S96 vs. R96 or R0 vs. R96). Genes that are differentially expressed/DNA methylated/histone methylated in R lines (green boxes), S lines (yellow boxes), and in both the NILs (blue boxes). Genes highlighted in red font were associated with H3K4me3 mark, genes highlighted in purple were associated with H3K27me3 mark and genes highlighted in black font showed colocalization of both the H3K4/K27me3 marks.

Full size image

In addition to the above genes, there were also genes (carrying domains for bHLH TF, auxin response factor, F-box, etc.) which had high affinity with H3K4me3 in susceptible genotype and thus are perhaps involved in promoting leaf rust infection. Earlier, overexpression of bHLH in wheat and Arabidopsis was shown to exhibit enhanced susceptibility to P. striiformis (in wheat) and Pseudomonas syringae (in Arabidopsis) (Wang et al. 2015).

Genes involved in resistance

Out of a number of DH3M genes associated with resistance or defence responses (R0 vs R96) one of the genes encodes a protein for alpha–beta hydrolase (for details of genes refer to Supplementary Tables S27 and S28 and Fig. 9). These proteins provide core structure for phytohormone/ligand receptors for two signalling pathways, namely gibberellin and strigolactone signalling pathways that are known to play an important role in defense (Mindrebo et al. 2016). For instance, gibberellin is known to act as both positive and negative regulator of disease resistance. Transgenic rice over-expressing GA deactivating enzyme displayed enhanced resistance to Mo and Xoo pathogens suggesting the role of GA as negative regulator of disease resistance; in contrast, GA was shown to act as a positive regulator of disease resistance during infection with necrotrophic root pathogen like P. graminicola (for further details, refer to review by Bruyne et al. 2014). The role of strigolactones in defense response was examined in tomato, where the mutants deficient in strigolactone were found to be more susceptible to foliar fugal pathogens like Botrytis cinerea and Alternaria alternate. These tomato mutants also exhibited reduction in the content of defense hormones like jasmonic acid, salicylic acid and abscisic acid (Torres-Vera et al. 2014).

Another class of genes that were identified during the present study and may be involved in resistance encode proteins carrying domains for GDSL esterase/lipases that are known to be involved in providing resistance against Alternaria brassisicola in Arabidopsis (Oh et al. 2005). Resistance through GDSL esterase/lipases are mainly provided through signalling involving phytohormone like ethylene (Kwon et al. 2009); negative regulation through auxin signalling has also been reported (Lee et al. 2009).

Genes which showed low affinity with H3K4me3 and associated with reduced expression were also found to be associated with disease resistance (S96 vs R96). For instance, the genes encoding proteins carrying domains like TIR-NBS-LRR domain, C-type lectin domain, aspartyl protease domain perhaps serve as negative regulators of disease resistance. Similarly, WAK (wall associated kinase) and exocyst subunit 70 showing high affinity with H3K4me3 were repressed (perhaps due to their hypermethylation) in incompatible interaction (Table 7), suggesting that their repression is involved in providing leaf rust resistance. Some of these genes have actually been reported to provide resistance through their negative regulation. For instance, there are genes which encode proteins containing lectin domain (e.g., C-type lectin domain containing proteins) and function as receptor like kinases involved in perception and transduction of environmental stimuli (for a review see Singh and Zimmerli 2013). Such genes include Lec-RK V.5 (receptor kinase V.5) that were earlier reported to act as a negative regulator of Pseudomonas syringeae infection through stomatal closure in Arabidopsis, since plants lacking this gene were found to be resistant (Desclos-Theveniau et al. 2012). WAK is another receptor kinase, exemplified by OsWAK112d in rice, which has been shown to act as a negative regulator of blast disease, so that its down regulation enhances disease resistance (Delteil et al. 2016). Some other genes which could be validated using qRT-PCR seem to be involved in resistance, since, these were induced in resistant NIL (Table 6).

Another group of genes carrying domains for bHLH TF, auxin response factor, F-box, etc. (for details of genes see Table S26) which were associated with high affinity DBSs of H3K27me3 in resistant NIL perhaps act as negative regulators of leaf rust disease resistance assuming that H3K27me3 leads to repression of gene expression. Interestingly, the same set of genes were shown to promote leaf rust infection when associated with H3K4me3 (see above). This suggested that activation and repression of the same set of genes may be modulated by different histone marks during susceptible and resistant reaction. In addition, the genes carrying domains for ethylene response TF, F-box, transmembrane protein, etc. were activated in resistant NIL indicating their role in defense response. Earlier also, the overexpression of ethylene response TF was shown to provide enhanced tolerance against infection by Rhizoctonia cerealis (Zhu et al. 2014).

Genes involved in binding of H3K4/K27me3 to the chromatin

A number of DH3M genes associated with H3K4me3 are known to encode proteins that regulate the H3 histone lysine trimethylation. These DH3M genes mainly encode histone lysine methyltransferases/SET domain containing proteins, PHD (plant homeodomain) finger like proteins, S-adenosyl-l-methionine-dependent methyltransferase (SAM), etc. Histone lysine methyltransferase/SET domain containing proteins are preferentially recruited to the 5′ coding regions of transcriptionally active genes, where they catalyze trimethylation of H3K4 (Ng et al. 2003). Similarly, proteins carrying PHD (plant homeodomain) help in recognizing modified histones like H3K4me3 and trap K4me3 in an aromatic cage like binding pocket (Musselman and Kutateladze, 2018). This subsequently helps in normal gene expression with the help of other proteins (Sanchez and Zhou 2011). The SAM proteins mainly help in the transfer of methyl group from methionine of ATP to H3 lysine causing histone trimethylation.

Conclusions and summary

The present study on wheat leaf-rust pathosystem is the first of its kind, where we studied genome-wide gene regulation mediated by histone H3 methylation coupled with DNA methylation and chromatin state analysis. Our results showed that susceptibility to leaf rust infection seems to be controlled mainly through activation of a large number of genes, whereas leaf rust resistance/defense response seems to be controlled by activation and repression of relatively fewer genes as revealed by the association of H3K4/K27me3 marks. Our earlier findings also showed the involvement of histone acetylation and DNA methylation in regulating the expression of genes involved in wheat-leaf rust pathosystem. The results of differential histone modification, chromatin state analysis and DNA methylation also suggested that more than one epigenetic marks may be involved in wheat-leaf rust pathosystem. We also identified 43 DH3M genes, which also carried DNA methylation marks; there were also genes that are regulated uniquely by one of the two histone modifications. Although, we do not expect complete correspondence between the genes available in ChIP-seq and RNA-seq libraries, but the presence of merely 7% (H3K4me3) and 12% (H3K27me3) genes that were common to the RNA-seq data do indicate the possible role of other epigenetic mechanisms in the regulation of gene expression, which need to be examined. This small fraction of genes (and the encoded proteins) enriched by two histone methylation marks included the following: (i) RLKs and cell wall modifying enzymes, (ii) protein kinases (S/TPKs) and phosphatase (purple phosphatase), (iii) ENTH/VAS family protein involved in phospholipid signaling, (iv) TFs (scarecrow, wuschell, NAC, MYB), (v) genes involved in ROS homeostasis (cytB5, GST, cytP450), (vi) growth hormone responsive genes (alpha beta hydrolase, GDSL esterase/lipase), and (vii) R genes (NBS-LRR containing). These genes may be involved in PTI and ETI signalling in a sequential manner. Some of these genes (e.g. acid purple phosphatase, ROS homeostasis genes, etc.) are perhaps involved in promoting infection leading to susceptibility mostly by suppressing the activity of other defense response genes like PR1 or due to oxidative burst. Some other genes are involved in defense response either through negative regulation (e.g. lectin domain containing genes, RLKs like WAK) where the low expression leads to resistance or through positive regulation where high expression leads to resistance. Evidence of trans-regulation of genes due to intergenic DBSs of both these marks was also noticed. With a view to examine the role of the remaining epigenetic mechanisms, we are currently examining the role of ncRNA (miRNAs and lncRNAs) and DNA methylation (at the level of a single base) using whole genome bi-sulfite sequencing. Together, the genes identified during the present study and those to be identified by the future studies involved in wheat-leaf rust interaction will hopefully be used in future studies thus expanding the knowledge-base for plant immunity in general and wheat-leaf rust pathosystem in particular.