Features of DNA methylome in Capsicum spp. Authors: Ye Liu, Yingjie Mi, Manru Song, Hongji Yang, Lulu Liu, Qian Yin, Xueyan Yao, Guorong Yu, Ying Fu, Guochen Qin, Feng Liu, Xuan Huang, Yan Xue Published: 2025 ABSTRACT AbstractDNA methylation is a most heritable epigenetic modification. Being a major vegetable crop, pepper (Capsicum spp.) possesses an over 3 Gb genome populated with TEs. This indicates a rich reservoir of epigenetic regulatory mechanisms. However, this large and complex genome renders the study of DNA methylation unaffordable and technically challenging. In this study, we analyzed DNA methylome in Capsicum spp., with a focus on C. annuum ST-8. We found that the genomes of Capsicum spp. are heavily methylated, particularly in the non-CG contexts. This is true when comparing to wheat, whose genome is over 16 Gb, containing over 80% TEs and repeats. Interestingly, we observed genic non-CG methylation and found that it is likely maintained by the CMTs, instead of RdDM. Overall, there is a negative relationship between gene expression and H3K9me2, and a positive relationship between genic non-CG methylation and H3K9me2, despite that genes without genic CHH methylation also possess some H3K9me2. Finally, we performed salt stress treatment with and without priming, and profiled active chromatin features as well as transcriptomes. We found that regardless of the environmental stimuli and developmental stages, the overall negative relationship between transcription and H3K9me2 is stably maintained. Altogether, our study revealed features of DNA methylation in ST-8 and we suggest that these features are likely common in Capsicum spp.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13072-025-00647-w. Introduction DNA methylation is a most heritable epigenetic modification conserved in plants and mammals. It functions to regulate gene expression, repress transposons and maintain genome stability. In plants, DNA methylation occurs in three sequence contexts, CG, CHG and CHH, where H represents A, T or C. In Arabidopsis, DNA methylation is concentrated in pericentromeric heterochromatin where transposable elements (TEs) are densely populated [1]. The establishment of DNA methylation is dependent on RNA directed DNA methylation (RdDM) [1, 2]. In RdDM, RNA POLYMERASE IV (Pol IV) triggers the production of 24-nucleotide small RNAs (24-nt sRNAs) [3]. This is mediated by RNA DEPENDENT RNA POLYMERASE 2 (RDR2), which converts single stranded Pol IV RNA to double strand. The double stranded RNA was subsequently cleaved into 24-nt sRNA through DICER-LIKE PROTEIN 3 (DCL3) [4, 5]. The 24-nt sRNA was then incorporated into AGONAUTE 4 (AGO4) and guide the ribonucleoprotein to the scaffold RNA produced by RNA POLYMERASE V (Pol V) through sequence complementarity. Finally, through direct interaction with either Pol V or AGO4, DOMAINS REARRANGED METHYLASE 2 (DRM2) can be recruited to de novo methylate DNA in all sequence contexts [6, 7]. Once established, DNA methylation can be maintained by DNA methyltransferases. CG methylation is maintained by METHYLTRANSFERASE 1 (MET1) during DNA replication [8]. CHG is maintained mainly by CHROMOMETHYLASE 3 (CMT3), which is recruited to their target sites by H3K9me2 [9]. CHG methylation, in turn, recruits SUPPRESSOR OF VARIEGATION 3–9 HOMOLOGUE PROTEIN (SUVH) 4, 5 and 6 to deposit H3K9me2, thereby, forming a positive feedback loop [10, 11]. CHH methylation is maintained by DRM2 or CMT2 depending on the chromatin environment. DRM2 functions at RdDM sites, including the body of short TEs and edges of long TEs, while CMT2 functions at H1 containing constitutive heterochromatin, commonly distributed over the body of long TEs, which is inaccessible to RdDM [2, 12]. Similar to CMT3, CMT2 contains a BAH domain and a chromo domain for dual recognition of H3K9me2, forming a positive feedback loop [13]. Unlike most situations where DNA methylation occurs in all three sequence contexts, gene body methylation (gbM) occurs mainly in the CG context [14]. The function of gbM remains controversial. Studies showed that genes with gbM tend to be moderately or highly expressed. They tend to participate in house-keeping functions with less expression variance among tissues [15]. gbM has also been suggested to influence aberrant transcription initiation and splicing, while contradictory observations have been reported [15–18]. Besides, gbM is proposed to prevent TE insertion into genic regions, which could be detrimental to the function of a gene [19]. While the function of gbM remains elusive, its establishment has been revealed recently. In Eutrema salsugineum, CMT3 and gbM are both absent. Despite the seemingly irrelevance, CMT3 has been recently shown to occasionally methylate CG. CMT3 mediated CHG methylation triggers the deposition of H3K9me2, which is subsequently removed from active genes by INCREASED IN BONSAI METHYLATION1 (IBM1) [20, 21]. Erasure of H3K9me2 leads to the loss of non-CG methylation, while CG methylation is retained by MET1 [22, 23]. Pepper is one of the most widely grown vegetable crops worldwide, best known for its pungent flavor and importance in pharmaceuticals. The 3 Gb genome of pepper is highly complex and populated with TEs and repeats (~ 80%), hampering the understanding of its functional genome. However, the large and complex genome also suggests a rich reservoir of epigenetic regulatory mechanisms. Recently, significant progresses have been made in understanding the functional genome of pepper, shedding light on its domestication, evolution and unique metabolic pathways [24–29]. In addition, systemic profiling of the epigenome landscape has been achieved, leading to the identification of cis-regulatory elements and unique epigenetic features such as intergenic H3K4me1-H3K27me3 bivalent regions [29, 30]. However, owing to the genome size and intrinsic complexity, few studies have focused on the methylome of pepper, despite its significance especially in maintaining the stability of such complex genomes. In this study, we systemically analyzed the methylome of Capsicum annuum ST-8 in leaf tissues of 30-day-old seedlings and noticed several interesting features. First, we observed high levels of DNA methylation in pepper, even in comparison to the aerial part of 14-day-old wheat seedlings, whose genome is over 16 Gb with over 80% TEs [31–33]. This is particularly true in the non-CG sequence contexts. Besides, we observed non-CG methylation over genic regions and found that this is likely dependent on CMT2 and CMT3, instead of RdDM. Overall, we observed a negative relationship between gene expression and H3K9me2, and a positive relationship between genic CHH methylation and H3K9me2, despite that some genes without genic CHH methylation also possess H3K9me2. To reveal the potential protein components associated with heterochromatin, we performed cross linked immunoprecipitation mass-spectrometry (CLIPMS) against H3K9me2 and validated the nuclear localization of several candidate proteins. Finally, we performed salt stress treatment with and without priming, and profiled active chromatin features as well as transcriptomes. We found that regardless of the environmental stimuli or developmental stages, the overall negative relationship between transcription and H3K9me2 is stably maintained. In conclusion, our study revealed unexpected features of DNA methylation in ST-8 and we suggest that these features are likely common in Capsicum spp. Results Pepper genome possesses high levels of non-CG methylation We have previously profiled the epigenome of Capsicum annuum ST-8 using leaf tissues of 30-day-old seedlings (6-leaf stage) [30], and found that DNA methylation levels are significantly higher than those in the rosette leaves of 5-week-old Arabidopsis in all sequence contexts (Fig. 1a). To test whether this is specific to ST-8, we compared our results to C. chinense Choco and C. frutescens Frut4 published by Jaiswal et al. [34]. It should be noted that leaf samples from 60-day-old plants were profiled in their study. Despite of different developmental stages, similar results were obtained. Specifically, genome-wide CG and CHG methylation in both species are above 80% and 70%, respectively. CHH methylation levels are over 20% [34]. DNA methylation in Choco is slightly higher than that in Frut4, indicating the presence of species variations [34]. We then processed their hybrid F1 methylation data using our pipeline and obtained similar results, validating our methods (Fig. 1a). Therefore, heavy DNA methylation in all sequence contexts is likely common in Capsicum spp. To reveal the extent of variations in different tissues, we compared leaf DNA methylation to that in root, stem and floral bud from ST-8, as well as leaf and floral bud from Arabidopsis [30, 35, 36]. We found that in Arabidopsis, floral bud displays slightly higher DNA methylation in all sequence contexts compared to leaf, while in ST-8, DNA methylation is comparable between these two tissues (Fig. 1a; Fig. S1). In ST-8, root and stem possess slightly lower level of DNA methylation relative to leaf and bud (Fig. S1). It should be noted floral buds were collected after flowering, while all other ST-8 tissues were collected from 30-day-old seedlings. DNA methylation levels in Pepper. a DNA methylation levels in C. annuum ST-8 and hybrid F1 of C. frutense and C. chinense, Arabidopsis, wheat, tomato and maize in CG, CHG and CHH sequence contexts. b Metaplots showing the DNA methylation levels over all annotated genes and regions 10 Kb upstream and downstream in ST-8, F1 hybrid, wheat, maize and tomato Arabidopsis genome contains low TE content (~ 23% in Col-0) [37] while ~ 80% of pepper genomes are occupied by TEs [26]. Thus, the dramatic discrepancy in DNA methylation levels may be a consequence of different TE contents. Therefore, we further examined methylomes in wheat and maize, over 80% of whose genomes are occupied by TEs [38, 39]. We also included tomato, a close relative of pepper whose genome contains about 70% TEs [40]. The WGBS datasets from wheat and maize were acquired from the aerial part of 12-day-old and 14-day-old seedlings, respectively, consisting mainly of young leaf tissues [41, 42]. DNA methylome from tomato was obtained from mature leaves [43]. As a result, wheat and tomato possess similar levels of genome-wide CG methylation compared to pepper, while non-CG methylation levels are significantly lower. Despite similar TE content, DNA methylation is lower in all sequence contexts in maize compared to wheat and pepper (Fig. 1a). To reveal the distribution of DNA methylation in the functional genome, we examined genic regions of all annotated genes and included 5 Kb extensions before transcription start sites (TSS) and after transcription end sites (TES) to represent intergenic regions. As a result, we observed higher DNA methylation in both genic and intergenic regions in pepper, compare to Arabidopsis, maize, wheat and tomato (Fig. 1b). This is also true in ST-8 root, whose DNA methylation is the lowest among the four tissues examined. Consistent with previous observations, non-CG DNA methylation in the pepper F1 hybrid is higher than those in ST-8 leaf and root (Fig. 1b, middle and right panels). This difference exists in both genic and intergenic regions. However, between ST-8 leaf and root, difference in non-CG methylation is mainly contributed by the intergenic regions. Regardless of these tissue specific differences, DNA methylation in pepper is overall higher than that in the other plants tested. Therefore, we conclude that pepper genome contains relatively high levels of DNA methylation, particularly in the non-CG sequence contexts. We suggest that this phenomenon is likely common in Capsicum spp. It should be noted that the methylation datasets adopted for this comparison might be acquired from different tissues and different growth stages. DNA methylation changes might be expected during plant growth and development. Presence of non-CG methylation over genic regions in pepper The observation of non-CG methylation over genes is surprising, as genic DNA methylation usually occurs only in the CG context. To test whether this phenomenon is ST-8 specific, we further included C. chinense Choco, C. frutescens Frut4 as well as their F1 hybrid for comparison [34]. As expected, CG methylation is enriched over genic regions and depleted at TSS and TES, a pattern characteristic of gbM (Fig. 2a, left panels). Consistently, we observed enrichment of CHG and CHH methylation over genic regions in all these four genetic backgrounds, with a bias towards the TSS (Fig. 2a, middle and right panels). The presence of genic non-CG methylation in ST-8 leaf was further validated through visual inspection (Fig. 2b). Presence of non-CG DNA methylation over genic regions in pepper. a Enrichment of DNA methylation in C. annuum ST-8 (top panels), C. Chinense Choco (second panels), C. Frutescens Frut4 (third panels) and their hybrid F1 (forth panels) over all annotated pepper genes. b Screenshot showing the presence of non-CG methylation over Caz02g19410 as an example. c Enrichment of DNA methylaiton in Arabidopsis over all annotated genes For comparison, we plotted DNA methylation levels over genic regions in Arabidopsis leaf and bud, ST-8 root, wheat, maize and tomato. As expected, genic CG methylation is present in all these samples, while non-CG methylation is depleted in Arabidopsis, wheat and maize (Fig. 2c; Fig. S2a-c). Similar to leaf, ST-8 root showed slight enrichment of non-CG methylation over genic regions (Fig. S2d). To our surprise, we also observed a slight enrichment of genic non-CG methylation in tomato, despite of being much weaker (Fig. S2e). In conclusion, we report the presence of non-CG DNA methylation over genic regions in pepper, and that this unusual phenomenon is likely common in Capsicum spp. Characteristics of genic non-CG methylation in pepper To further characterize genic non-CG methylation, we calculated DNA methylation levels over all annotated genes individually using ST-8 leaf methylome. Arabidopsis leaf was used as a control considering its well characterized gene structures. We found that most Arabidopsis genes possess lower than 20% genic CG methylation, which exceeds 40% over most ST-8 genes (Fig. 3a). Consistent with previous studies, CHG and CHH methylation are barely detectable in Arabidopsis genes [23], while ST-8 genes contain up to about 40% CHG and 15% CHH methylation (Fig. 3a). Next, we asked whether the genic non-CG methylation is contributed by intronic TEs which are subsequently spliced out during mRNA processing. Accordingly, we calculated DNA methylation levels separately over exons and introns. Consistent with previous reports, CG methylation is more enriched over exons in Arabidopsis (Fig. 3b) [14]. Similar observation was found in ST-8 leaf (Fig. 3b). As expected, non-CG methylation remains depleted over introns and exons in Arabidopsis (Fig. 3b). Intriguingly, in ST-8 leaf, a subset of genes possesses non-CG methylation at relatively comparable levels between introns and exons (Fig. 3b). Characterization of non-CG methylation over genes. a DNA methylation levels in CG, CHG and CHH sequence context in all annotated genes in Arabidopsis and pepper. b DNA methylation levels in exons and introns of all annotated genes in Arabidopsis and pepper. c Expression levels of genes ranked by CHH methylation levels over genic regions. d Screenshot of an expressed gene with high levels of genic non-CG methylation. e GO biological process enrichment of genes ranked by CHH methylation levels over genic regions DNA methylation usually represses gene expression while gbM is known to associate with moderately or highly expressed house-keeping genes [23]. Therefore, we asked whether and how genic non-CG methylation might be correlated with gene expression. To answer this question, we separated all annotated genes into four categories according to their non-CG methylation levels and assessed transcription in each category. RNA-seq data from leaf tissues of 30-day-old ST-8 seedling were adopted for the analysis [30]. We used CHH methylation as a proxy because CHG and CHH methylation levels are positively correlated (Fig. S3a, b). As a result, we found that for genes with genic CHH methylation, there is an overall negative correlation between methylation level and expression (Fig. 3c), suggesting that genic non-CG methylation might repress expression. However, we did find genes that are heavily non-CG methylated, and yet are highly expressed (Fig. 3d). Interestingly, genes without CHH methylation are expressed only slightly higher than the heavily methylated genes. Finally, we performed GO term analysis to reveal the function of genes in each methylation category. We found that genes with no CHH methylations are enriched in mRNA splicing, translation and photosynthesis (Fig. 3e). Genes carrying moderate or excessive non-CG methylation (> 0.1) are enriched in ATP/ADP transport, purine nucleotide transport, positive regulation of protein metabolism, ribonucleotide biosynthesis, etc. (Fig. 3e). Therefore, genes with different levels of non-CG methylation might have different biological functions in pepper. In conclusion, we found that in pepper, non-CG DNA methylation is present over introns as well as exons of a subset of genes and that excessive methylation in non-CG contexts may repress transcription. Genic non-CG methylation is associated with H3K9me2 modification Non-CG methylation can be established through RdDM or maintained by CMT2 and CMT3, depending on the chromatin context. Studies in Arabidopsis revealed a division of labor between RdDM and CMTs. RdDM is responsible for CHH methylation over the body of short TEs and the edges of long TEs, characterized by the production of 24-nt sRNAs. In contrast, CMT2 and CMT3 are responsible for the body of long TEs, characterized by dense H3K9me2 and largely independent of sRNAs [12] (Fig. S4a-d). To reveal whether the genic non-CG methylation is a consequence of RdDM or the CMTs, we performed sRNA seq and assessed the correlation between genic non-CG methylation and the enrichment of H3K9me2 and 24-nt sRNAs. Again, we adopted CHH methylation for the analysis as it is highly correlated with CHG methylation (Fig. S2a, b). As a result, we found that genic CHH methylation is better correlated with H3K9me2, in comparison to the abundance of 24-nt sRNAs (Fig. 4a). We then examined global enrichment of H3K9me2 and 24-nt sRNA over all annotated genes in ST-8. As expected, we observed depletion of H3K9me2 over genic regions in comparison to intergenic regions. However, the center of the genes displays slight enrichment of H3K9me2 in comparison to the TSS and TES (Fig. 4b). This enrichment is biased towards the TSS, a pattern very similar to the genic non-CG methylation (Fig. 2a). However, 24-nt sRNA is evenly depleted throughout the gene body (Fig. 4b). To validate, we again categorized all genes according to their genic CHH methylation levels. As expected, genes with the highest genic CHH methylation (> 0.2) possess the highest levels of H3K9me2, which is more abundant than the intergenic regions (Fig. 4c). Genes showing the highest expression (methylation levels 0-0.1) are significantly depleted of H3K9me2. Interestingly, the other two groups (methylation levels 0 and 0.1–0.2) possess H3K9me2 levels similar to the intergenic regions. However, the group with higher expression (methylation levels 0.1–0.2) shows clear depletion of H3K9me2 at TSS and TES, potentially reflecting the presence of the transcription machinery (Fig. 4c). In contrast, 24-nt sRNA remains depleted in all groups, suggesting the absence of RdDM activity (Fig. 4d). Association of genic non-CG methylation with other heterochromatin features. a Scatter plots showing the association of genic CHH methylation with H3K9me2 levels (left panel) and the 24-nt siRNA abundance (right panel). b Metaplots showing the enrichment of H3K9me2 (left panel) and 24-nt siRNA (right panel) over all annotated genes in pepper. c Metaplot showing the enrichment of H3K9me2 over the genes categorized by their genic CHH methylation levels. d Metaplot showing the enrichment of 24nt siRNA over the genes categorized by their genic CHH methylation levels Altogether, our results suggested that genic non-CG methylation is likely dependent on the maintenance pathways mediated by the CMTs and H3K9me2, instead of RdDM. Proteins associated with H3K9me2 marked heterochromatin To reveal the potential composition of heterochromatin in pepper, we performed crosslinked immunoprecipitation couple with mass spectrometry (CLIPMS) with H3K9me2. It should be noted that our aim is to provide a list of potential proteins that associate with H3K9me2 marked heterochromatin. The functions of the identified proteins are inferred solely based on sequence homology. Therefore, further investigations are required before the function of these proteins can be assigned or validated. As expected, we identified CMT2 among the most enriched proteins (Table 1). We also detected RdDM components including NRPE subunits. RdDM is known to be at the edges of long TEs and throughout short TEs, while H3K9me2 is highly enriched at the body of long TEs [10, 44, 45]. Therefore, detection of RdDM components may suggest that the resolution of our method is not high enough to delineate the boundaries between RdDM and H3K9me2 marked constitutive heterochromatin. Meanwhile, RdDM loci also possess low levels of H3K9me2, deposited by SUVH4/5/6, for the recruitment and initiation of Pol IV mediated 24-nt sRNA production [11, 46–48]. Table 1Proteins associated with H3K9me2 marked heterochromatin in pepperProtein IDAnnotationCTH3K9me2Arabidopsis homologuerep1rep2Gene IDAnnotationPHT95849.1DNA (cytosine-5)-methyltransferase CMT201411AT4G19020CMT2PHT87914.1DNA-directed RNA polymerases IV and V subunit 201110AT3G23780NRPE2PHT74769.1DNA-directed RNA polymerase V subunit 1099AT2G40030NRPE1PHT62805.1Adenosylhomocysteinase01112AT4G13940SAHH1PHT62682.1hypothetical protein21918AT1G31280AGO2PHT62921.1U5 small nuclear ribonucleoprotein kDa helicase13230AT1G20960BRR2PHT74276.1Elongation factor 20220AT1G56070LOS1PHT75776.1Pre-mRNA-processing-splicing factor 832934AT1G80070SUS2PHT84342.1ATPase family AAA domain-containing protein0117AT1G05910BRAT1PHT89826.1Nuclear pore complex protein01212AT1G14850NUP155PHT95063.1SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily A member 5, partial0171AT5G18620CHR17PHT65965.1Hypothetical protein12124AT1G64790ILAPHT70493.1Hypothetical protein01010AT1G67680SRP72PHT71304.1Hypothetical protein02718AT1G65440SPT6LPHT86904.1Hypothetical protein01818AT5G61970SRP68PHT69388.1Hypothetical protein01113AT3G22440FRIGIDA-like protein 4aPHT81189.1Hypothetical protein01114AT5G48385FRIGIDA-like protein 3PHT81217.1Hypothetical protein11716AT2G03150RSA1PHT88542.1Hypothetical protein21721AT5G07350AtTudor1 Proteins associated with H3K9me2 marked heterochromatin in pepper As expected, we detected a list of hypothetical proteins, reflecting a substantial gap in the molecular and functional characterizations of genes in pepper. To reveal the potential functions of these hypothetical proteins, we searched for their closest homologues in Arabidopsis. Among the homologues, we identified AGO2, further supporting the close association between RdDM and H3K9me2. Besides, we identified chromatin remodelers and factors associated with transcription machinery. However, we do not know whether these transcription factors are directly associated with the genic H3K9me2 or in close proximity with constitutive heterochromatin. Additionally, their molecular functions await further experimental characterization. Finally, we examined the subcellular localization of several genes on the list using agrobacterium mediated transient expression in ST-8 leaf, as well as in Nicotinana benthamiana. As expected, CMT2 and FRI were localized to the nuclei but seemingly excluded from the nucleoli in both pepper and N. benthamiana (Fig. 5a-b). The NRPE2 homologue formed punctate signals in both ST-8 and N. benthamiana (Fig. 5c). However, using the transient expression system, the protein is likely over-expressed, potentially leading to artificial phase separation. The SWI/SNF homologue was also confirmed to be a nuclear protein. It was even distributed in the nucleoplasm in ST-8 and more concentrated in the nucleoli in N. benthamiana (Fig. 5d). Whether this is an over-expression artefact remains to be tested. Nuclear localizaiton of proteins identified by H3K9me2 CLNIP. a Nuclear localization of PHT95063.1/CMT2-GFP transiently expressed in ST-8 (upper) and N. benthamiana leaf (lower panels). b Nuclear localization of PHT81189.1/FRI-GFP transiently expressed in ST-8 (upper) and N. benthamiana leaf (lower panels). c Nuclear localization of PHT87914.1/NRPE-GFP transiently expressed in ST-8 (upper) and N. benthamiana leaf (lower panels). d Nuclear localization of PHT86904.1/SWI/SNF-GFP transiently expressed in ST-8 (upper) and N. benthamiana leaf (lower panels) In summary, we revealed a list of potential H3K9me2 associated heterochromatic proteins using CLIPMS and validated the nuclear localization of a subset of proteins on the list. Genic non-CG methylation is associated with H3K9me2 and represses transcription To reveal the relationship between genic non-CG methylation and transcription, we treated ST-8 with salt stress with and without priming (Fig. 6a). Leaf tissues were collected at indicated timepoints and used for RNA seq (Fig. 6a). Differentially expressed genes (DEGs) were identified (FC > 2, p value < 0.1). We defined DEGs between T1 and T1’ as priming responsive genes; T3’ and T2’ as salt stress responsive genes; T3 and T3’ as primed salt stress responsive genes. GO term analysis of the differentially expressed genes (DEGs) revealed that translation and photosynthesis were upregulated after priming treatment (Fig. 6b, left panels). As expected, high salt treatment triggered the activation of stress responses and protein folding while photosynthesis and metabolism are repressed (Fig. 6b, middle panels). Interestingly, photosynthesis and metabolism were both activated, or alternatively, less repressed, compared to high salt treatment without priming, while defense responses were downregulated (Fig. 6b, right panels). In summary, the GO analysis suggests that our treatments and analysis were reliable. Association of genic non-CG methylation with transcription reprogramming. a Treatment conditions of primed salt stress and salt stress. b GO term biological function enrichment of DEGs upregulated (left panels) and downregulated (right panels) in different timepoints. c Expression of genes ranked by genic CHH methylation levels at T1’. d Metaplots showing the intensity of active histone marks and chromatin accessibility over genes ranked by genic CHH methylation levels. e Gene length ranked by genic CHH methylation levels. f Exon length ranked by genic CHH methylation levels We then analyzed the transcriptome at T1’ with 6-week-old leaf tissues without salt treatment. Similar to the analysis before, we grouped all genes into four categories according to their genic CHH methylation levels. Despite different growth stages, the same expression trend was observed (Fig. 6c). Specifically, genes with genic CHH methylation overall showed a negative relationship between expression and DNA methylation. Genes without methylation were expressed slightly higher than those that are heavily methylated. Similar patterns were observed with all timepoints and treatments included in this study (Fig. S5a-f). In addition to the published repressive histone marks, we profiled chromatin features of active transcription. Specifically, we performed ChIP-seq against H3K4me3, H3K27ac and H3K4me1 as well as ATAC-seq at all the timepoints. Consistently, we found that genes with moderate levels of genic CHH methylation display the most prominent active chromatin features (Fig. 6d). Interestingly, while genes with 0 CHH methylation show slightly higher expression in comparison to the genes with heavy CHH methylation (> 0.2), they possess the least active chromatin features. Consistent with the depletion of the repressive mark H3K9me2 over TSS of genes with heavy CHH methylation (> 0.2), we observed a clear enrichment of active chromatin features over TSS in this category. These features are consistently observed in all the treatments regardless of the treatment or growth stages included in this study (Fig. S6a-c). Therefore, we conclude that genic non-CG methylation likely represses gene expression through H3K9me2, regardless of growth and environmental stimuli. Discussion Epigenetic regulation is crucial for genome stability and gene expression. Intensive studies have been carried out in the model plant Arabidopsis, leading to significant breakthroughs, for example, the identification and comprehensive mechanistic elucidation of RdDM [1, 49–51]. However, as a model organism, Arabidopsis carries a small and simple genome, representing an intrinsic disadvantage in studying epigenetic regulations. For example, mCHH islands are commonly found in plants with large genomes including wheat, maize and pepper [30, 41, 52–54]. They are present right in front of the TSS and tend to associate with highly expressed genes. However, likely due the low contents of TEs and repeats, mCHH islands are absent in Arabidopsis. Therefore, to fully appreciate the complexity and biological significance of epigenetic regulation, it is necessary to examine plants with more complex genomes. In this study, we examined DNA methylation in Capsicum annuum ST-8. Its over 3 Gb genome is populated with TEs and repeats. We report that genome-wide high level of DNA methylation is likely common in Capsicum spp., particularly in the non-CG sequence contexts. It is noteworthy that DNA methylation can be dynamic throughout development and may vary in different tissues. For example, Arabidopsis floral tissue possesses slightly higher levels of DNA methylation than leaf, and pepper root contains the lowest DNA methylation among leaf, stem and floral tissues. Therefore, we focused our comparisons on leaf methylomes from different plants. One caveat is that the leaf samples from different plants were collected at different ages, which might potentially introduce developmental variations into the comparison. Despite these variations, pepper overall contains higher DNA methylation levels in comparison to the other plants examined here. In addition to wheat, maize, tomato and Arabidopsis methylome, which we re-processed using our pipeline, we also searched for published DNA methylation results in other plants for comparison. In soybean cotyledon, DNA methylation in CG, CHG and CHH sequence contexts are about 66%, 45% and 11% respectively [55]. Sugarcane methylation patterns have been profiled in leaf, root, rind and pith, among which the pith contains the highest levels of genome wide methylation, at about 62% (CG), 40% (CHG) and 5% (CHH) [56]. In a study aimed to improve gray leaf spot resistance, tomato scions were grafted onto pepper or eggplant rootstocks, and the scion methylome were profiled. Interestingly, DNA methylation levels were significantly higher in tomato scions when grafted onto pepper than those grafted onto eggplant or tomato itself [57]. This suggests that DNA methylation levels are highly dynamic. Despite the wide range of DNA methylation levels and highly dynamic nature, genomes of Capsicum spp. are relatively heavily methylated among the plants we examined. Interestingly, we observed genic non-CG methylation covering both exons and introns of some genes, which has been rarely reported to our knowledge. Some studies showed that gymnosperms, including P. taeda, P. glauca and Norway spruce contain non-CG methylation over genes [58, 59]. While the cause and biological significance of genic non-CG methylation remain elusive, we showed that it does not seem to be a consequence of high TE content. Wheat, which carries a much larger genome and over 80% TEs, lacks genic non-CG methylation. In contrast, tomato, another Solanaceae species, showed slight enrichment of genic non-CG methylation, despite of a much smaller genome and lower TE content. Several potential reasons might contribute this feature. First, high quality genome has only recently been assembly for Capsicum spp. and the functional genome is under studied. Therefore, mis-annotation of gene and TE structures might exist. Second, variations of the methylation pathway components might exist. It would be interesting to functionally characterize these components in plants with and without genic non-CG methylation. For example, whether the genic methylation remover IBM1 is functionally diverged. However, searching for the methylation components and validating their functions remain technically challenging especially in pepper. While the biological significance of genic non-CG methylation remains elusive, we found that it is likely deposited by the CMTs and associated with the heterochromatin mark H3K9me2. Consistently, we found that genes with heavy genic non-CG methylation tends to be lowly expressed. However, we also observed genes with heavy non-CG methylation while highly expressed. One potential explanation to this seemingly contradictory result is that we performed sequencing with bulked samples, meaning that DNA methylation and transcription may be from different cells. Single cell methylome analysis coupled with RNA seq would be necessary to answer the question. It should also be noted that gene structure in pepper is largely based on bioinformatic prediction and mis-annotation might be present. In summary, we observed high levels of DNA methylation in Capsicum spp., particularly in the non-CG sequence contexts. In addition, we report the presence of genic non-CG methylation and proposed that this is likely caused by CMTs instead of RdDM. We suggest that genic non-CG methylation is likely associated with heterochromatinization and transcription repression. Our study revealed the necessity and value of studying plants with large and complex genomes. Materials and methods Plant materials and growth conditions C. annuum “Zhangshugang” (ST-8) was kindly provided by Dr. Feng Liu. Plants were kept in a greenhouse under long day condition (25℃, 16 h light/8 hr dark). For sRNA seq experiment, leaf samples from 4-week-old seedling were collected. For salt stress treatment, seeds were germinated at 28℃ in the dark and transferred to ½ Hoagland solution (Phygene Biotechnology, China). Hoagland solution was changed every three days. Four-week-old seedlings were subjected to ½ Hoagland solution with or without 50 mM NaCl as control and priming, respectively, for 7 days. This is followed by another 7-day of recovery in ½ Hoagland solution. Finally, salt stress treatment was performed with ½ Hoagland solution supplemented with 200 mM NaCl. After 7-day of stress treatment, leaf tissues were harvested for RNA seq, ChIP seq and ATAC seq. sRNA seq Leaves of 4-week-old soil-grown seedlings were used for sRNA seq. Total RNA was first extracted using FreeZol Reagent (Vazyme, Najing, China). sRNA from around 10 to 30 nucleotide long were excised and recovered after gel electrophoresis. sRNA libraries were constructed using VAHTS Small RNA Library Prep Kit for IIlumina V2, coupled with VAHTS Small RNA Index Primer Kit for Illumina (Vazyme, Najing, China) according to the manufacturer’s instructions. Sequencing was performed with Illumina NovaSeq 6000 platform. RNA seq Total RNA was extracted using FreeZol Reagent according to the manufacturer’s instruction (Vazyme). Libraries were made with VAHTS Unive V8 RNA-seq Library Prep Kit for Ilumina according to the manufacturer’s instruction (Vazyme). Sequencing was performed with the Illumina NovaSeq 6000 system. ChIP seq Two grams of tissues pooled from at least three plants were used for each replicate and two biological replicates were performed for each ChIP. All samples were crosslinked in vitro with Nuclei Isolation Buffer (50 mM HEPES, 1 M sucrose, 5 mM KCl, 5 mM MgCl2, 0.6% Triton X-100, 0.4mM PMSF, 5 mM benzamidine, cOmplete EDTA-free Protease Inhibitor Cocktail (Roche)) supplemented with 1% formaldehyde for 12 min with rotation. Glycine was immediately added to stop the crosslinking. Lysate was filtered through Miracloth and centrifuged for 20 min at 2880 g at 4°C. The pellet was resuspended in 1 ml of extraction buffer 2 (0.25 M sucrose, 10 mM Tris-HCl pH 8, 10 mM MgCl2, 1% Triton X-100, 5 mM BME, 0.1 mM PMSF, 5 mM Benzamidine, and 1x protease inhibitor cocktail tablet) and centrifuged at 12,000 g for 10 min at 4°C. The pellet was resuspended in 500 µL extraction buffer 3 (1.7 M sucrose, 10 mM Tris-HCl pH 8, 2 mM MgCl2, 0.15% Triton X-100, 5 mM BME, 0.1 mM PMSF, 5 mM Benzamidine, 1x protease inhibitor cocktail tablet) and layered on top of 500 µL extraction buffer 3 (1.7 M sucrose, 10 mM Tris-HCl pH 8, 2 mM MgCl2, 0.15% Triton X-100, 5 mM BME, 0.1 mM PMSF, 5 mM Benzamidine, 1x protease inhibitor cocktail tablet) and centrifuged at 12,000 g for 1 h at 4°C. The pellet was lysed with 400 µL nuclei lysis buffer (50 mM Tris pH 8, 10 mM EDTA, 1% SDS, 0.1 mM PMSF, 5 mM Benzamidine, 1x protease inhibitor cocktail tablet). A total of 1.7 ml of ChIP dilution buffer (1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris pH 8, 167 mM NaCl, 0.1 mM PMSF, 5 mM Benzamidine, 1x protease inhibitor cocktail tablet) was added to the lysed nuclei. Chromatin was sheared with Bioruptor Plus (Diagenode) for 30 min with 30 s on and 30 s off and incubated with the corresponding antibodies overnight at 4°C (anti H3K4me1 antibody (Abcam), anti H3K4me3 antibody (Abcam), anti H3K27ac antibody (Abcam)). Chromatin-bound proteins were immunoprecipitated with Protein A and Protein G magnetic Dynabeads (Invitrogen) for 2 h at 4°C. Dynabeads were washed twice with low salt buffer (150 mM NaCl, 0.2% SDS, 0.5% Triton X-100, 2 mM EDTA, 20 mM Tris pH 8), once with high salt buffer (200 mM NaCl, 0.2% SDS, 0.5% Triton X- 100, 2 mM EDTA, 20 mM Tris pH 8), once with LiCl buffer (250 mM LiCl, 1% Igepal, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris pH 8), and once with TE buffer (10 mM Tris pH 8, 1 mM EDTA). Elution was performed with 250 ul elution buffer (1% SDS, 10 mM EDTA, 0.1 M NaHCO3) by incubating at 65°C with shaking twice. Four hundred microliters of eluted complexes were reverse crosslinked by incubation at 65°C overnight with the addition of 20 ul of 5 M NaCl followed by protease K treatment (20 ug in 10 mM EDTA and 40 mM Tris pH 8) at 45°C for 1 h. DNA fragments were precipitated with EtOH overnight at − 20°C. Libraries were prepared with Ovation Ultra Low System V2 kits following the manufacturer’s instructions. All libraries were sequenced at a length of 150 bps pair-end with the NovaSeq 6000 platform (Illumina). ATAC seq ATAC seq was performed with two biological replicates. Each biological replicate includes pooled leaf samples from at least three plants. ATAC seq libraries were prepared with the help of Biomarker Technologies (Biomarker). Briefly, 50,000 nuclei were treated with Tn5 transposases for 30 min at 37◦C. DNA was purified with the MinElute PCR purification kit (Qiagen) according to the manufacturer’s instructions. Libraries were prepared using TruePrep DNA library prep kit V2 for illumina (Vazyme). Sequencing was performed with the Illumina 6000 platform (Illumina). sRNA seq analysis Raw sequencing data were processed using Trim Galore (v0.6.10) [60] with the parameters --quality 20 -a AGATCGGAAGAGC --length 18 --max_length 30 to remove 3’ adapters and filter low-quality reads (Q < 20), retaining fragments of 18–30 nt in length. Quality-filtered reads were aligned to the Rfam database using Bowtie (v1.1.2) [61] under stringent parameters (-v 1 -k 1 --un) to remove non-coding RNAs (including rRNAs and tRNAs). The remaining reads were then mapped to the pepper reference genome (Zhangshugang version) available at: http://ted.bti.cornell.edu/ftp/pepper/genome/Zhangshugang/Zhangshugang_genome.fa.gz. using ShortStack (v4.1.1) [62] with the parameters --dicermin 20 --dicermax 24 for specific analysis of 20–24 nt small RNAs. RNA seq analysis Raw sequencing reads were adapter-trimmed and quality-filtered using Fastp (v0.23.4) [63] to generate clean data. The clean reads were then aligned to the Zhangshugang reference genome using HISAT2 (v2.2.1). Read counts for each gene were quantified using the featureCountsfunction from the Subread package. Differential gene expression analysis was performed with the edgeR package in R, applying a significance threshold of adjusted Pvalue ≤ 0.01 and an absolute log2-fold change ≥ 1. Gene Ontology (GO) enrichment analysis for the differentially expressed genes was conducted using the clusterProfiler package. Crosslinked IPMS, LC-MS/MS and data analysis Crosslinked IPMS, LC-MS/MS and data analysis were performed according to Yang et al. [30] using anti-H3K9me2 antibody (Abcam). Differentially detected proteins were identified using edgeR [64] with significance thresholds set at p ≤ 0.05 and fold change ≥ 1.5. DNA methylation analysis Sequencing data of pepper hybrid (NCBI SRR16307080) [34], Capsicum chinense (CNCB CRR1250606-07) [34], Arabidopsis thaliana (SRR6410887-88) [65], Triticum aestivum (SRR8126631) [41], Zea mays (SRR29015388) [42], and Solanum lycopersicum (SRR5901103) [43] were adopted from previous publications. Raw reads were processed using Fastp (v0.23.4) [63] for adapter trimming and quality filtering. Clean reads were aligned to their respective reference genomes using Bismark (v0.24.2) [66] with Bowtie2 [61] as the alignment engine, followed by PCR duplicate removal using deduplicate_bismark. Regional methylation levels were calculated using BatMeth2 (v2.1.0) [67] with the following parameters: calmeth -Q 20 --remove_dup --coverage 4 -nC 1 --Regions 600 --step 50,000. A custom R script was developed to integrate CG, CHG, and CHH methylation levels across genomic features (genes, exons, and introns). Following the calculation of CHH methylation levels across genic and promoter regions using BatMeth2’s methyGff(coverage ≥ 4, step size 0.01), genes were classified into four categories (0, 0–0.1, 0.1–0.2, > 0.2). All subsequent comparative analyses and visualizations, including boxplots, were conducted using ggplot2 in R. Histone modification analysis Datasets for H3K9me2 and H3K27me3 in pepper were obtained from CNGBdb CNSA (accessions CNR0515149-52 and CNR0515157-60) [29], Arabidopsis thaliana (SRR15853573-74, SRR15853581-86) [65], Triticum aestivum (SRR10015383, SRR8126623-26) [41], Zea mays (SRR26146587-88, SRR26146591-92, SRR13198954-57) [68, 69], and Solanum lycopersicum (SRR26146505-06, SRR26146509-10, SRR24709427, SRR24709430-31) [69]. Raw sequencing data were quality-filtered using Fastp (v0.23.4) [63] and aligned to their respective reference genomes using Bowtie2 (v2.5.3) [61]. Enriched regions (peaks) were identified using MACS2 (v2.2.7.1) [70], and histone modification profiles were visualized using Deeptools (v3.5.1) [71] to generate distribution profiles and heatmaps. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1. Acknowledgements Lead contact: Yan Xue. Author contributions Y.L., M.S., H.Y, L.L, Q.Y. X.Y and G.Y. performed the experiments. Y.M. performed the bioinformatic analysis. Y.F. and G.C. performed LC MS/MS. Y.X., X.H. and F.L. designed the experiments and analysis. Y.X. wrote the manuscript. All authors read and approved the manuscript. Funding This work was supported by Natural Science Foundation of Shaanxi Province (2025JC-YBMS-241 to X.H.), the Shandong Provincial Natural Science Fund for Excellent Young Scientists Fund Program (2023HWYQ-111 to Y.X.) and Shandong Provincial Natural Science Foundation (ZR2024QC239 to L.L.). Data availability The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in National Genomics Data Center (Nucleic Acids Res 2022), China National Center for Bioinformation / Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA026607) that are publicly accessible at [https://ngdc.cncb.ac.cn/gsa] (https://ngdc.cncb.ac.cn/gsa). Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Ye Liu, Yingjie Mi, Manru Song and Hongji Yang have contributed equally to this work. Contributor Information Feng Liu, Email: jwszjx@hunau.edu.cn. Xuan Huang, Email: xuanhuang@nwu.edu.cn. Yan Xue, Email: yan.xue@pku-iaas.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in National Genomics Data Center (Nucleic Acids Res 2022), China National Center for Bioinformation / Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA026607) that are publicly accessible at [https://ngdc.cncb.ac.cn/gsa] (https://ngdc.cncb.ac.cn/gsa).