The chromatin landscape of the moss Physcomitrella patens and its dynamics during development and drought stress



The moss Physcomitrella patens is an important model organism for evo-devo studies. Here, we determined the genome-wide chromatin landscape of five important histone three (H3) modifications (H3K4me3, H3K27me3, H3K27Ac, H3K9Ac and H3K9me2) and describe the changes to these histone marks in two contrasted situations, developmental transition and abiotic (drought) stress. Integrative analysis of these histone H3 modifications revealed their preferential association into 15 chromatin states (CS) in genic regions of the P. patens genome. Synergistic relationships that influence expression levels were revealed for the three activating marks H3K4me3, H3K27Ac and H3K9Ac, while an antagonistic relationship was found between CS containing the H3K27me3 and H3K27Ac marks, suggesting that H3K27 is a key indexing residue regarding transcriptional output. Concerning the alteration of histone marks in response to developmental transition (juvenile to adult) and drought stress, the three activating marks H3K4me3, H3K27Ac and H3K9Ac show significant changes in both situations. However, changes to H3K27me3 are central only for genes differentially expressed during development. Interestingly, genes induced during drought stress show significant histone mark toggling during developmental transition. This situation suggests that drought induced adult (gametophore expressed) genes are primed to respond to this stress during the juvenile to adult transition.


Chromatin is a central player in organizing the genome at both structural and functional levels, and takes part in establishing the so-called epigenome (Wolffe and Hayes, 1999; Van Steensel, 2011; Luger et al., 2012). Histones, which form the core components of the chromatin, undergo post-translational modifications that have been associated with a range of processes such as DNA replication and repair, cellular differentiation, chromosome condensation and transcriptional activity (Berger, 2007; Groth et al., 2007; Kouzarides, 2007; Li et al., 2007). The discovery of enzymes that can either deposit (‘writers’) or remove (‘erasers’) histone modifications, as well as chromatin-binding proteins (‘readers’), has led to the view that histone modifications can significantly extend the information potential of the genetic code (Strahl and Allis, 2000; Lee et al., 2010; Rando, 2012). Although some differences in the establishment and maintenance of histone modifications have been uncovered across different kingdoms, common themes have emerged (Berger, 2007; Köhler and Villar, 2008; Hennig and Derkacheva, 2009; Feng et al., 2010). For instance, histone H3 and histone H4 acetylation are, in general, associated with active transcription, while methylation of lysine residues has been associated with either transcriptional activation or repression, depending on which lysine residue is methylated and how many methyl groups are added.

In plants, comparatively few genome-wide histone mark studies were performed so far and have been carried out mainly by chromatin immunoprecipitation followed by microarray analysis (ChIP-chip). For example, in the model flowering plant Arabidopsis thaliana eight histone marks were analysed and four chromatin states (CS) were found to be representative of 90% of the 120 Mbp genome (Roudier et al., 2011). Other ChIP-chip-based epigenomic analyses have been undertaken in Arabidopsis, to study the dynamics of chromatin changes during development or in response to environmental cues, but mainly focusing on a few histone modifications or on a single response (Charron et al., 2009; Lafos et al., 2011; Bourbousse et al., 2012). More recently, CS in A. thaliana have been analysed using ChIP followed by deep sequencing (ChIP-seq) data of nine histone modifications (Luo et al., 2013) and three to four H3 modifications were analysed in rice and maize (Wang et al., 2009; He et al., 2010). Only a few ChIP-seq studies that dealt with dynamic changes of histone modifications have been carried out in plants so far, namely H3K4 methylation in response to dehydration (van Dijk et al., 2010) in A. thaliana; H3K4me3/K3K27me3 changes in response to leaf senescence (Brusslan et al., 2012) in A. thaliana; H3K9Ac in response to cold in maize (Hu et al., 2012) and H3K4me3 under drought in rice (Zong et al., 2013). Very recently, epigenomic changes on four histone marks, induced by hyperosmotic priming, were analysed in A. thaliana seedlings (Sani et al., 2013). Genome-wide histone mark data are therefore restricted to a few model flowering plants, mainly focusing on A. thaliana with its atypically small genome. Epigenomic changes induced by abiotic stress have been analysed in the two crop plants, maize and rice, as well. We therefore decided to carry out a genome-wide ChIP-seq study that involved a developmental progression as well as an abiotic stress condition, to contrast the two situations and their associated epigenetic changes.

We selected the non-flowering model plant, Physcomitrella patens (Rensing et al., 2008) as in contrast with seed plants and metazoans it has a haploid-dominant life cycle. Moreover, P. patens is an important model for evo-devo studies (Prigge and Bezanilla, 2010) due to its informative phylogenetic position, its comparatively simple morphology and easily conductible reverse genetics approaches through gene targeting employing homologous recombination (Kamisugi et al., 2006). The P. patens v1.6 genome annotation is spread over close to 2000 scaffolds and harbours 32 275 protein coding genes (Zimmer et al., 2013), about half of the approximately 500 Mbp genome is derived from transposons (Rensing et al., 2008). The haploid gametophytic generation comprises two distinct developmental stages that are: (i) filamentous protonemata representing the juvenile stage; and (ii) leafy gametophores that represent the adult asexual stage (Cove, 2005). These two principal developmental stages display relatively simple morphologies (for the most part, a single cell layer thick) with fewer cell fates as compared with flowering plants (Cove, 2005; Prigge and Bezanilla, 2010). As in animals and seed plants, the Polycomb repressive complex 2 (PRC2) of P. patens is critically important for the normal development of moss (Mosquna et al., 2009; Okano et al., 2009) in that P. patens mutants impaired for PRC2 members undergo protonema development but lack gametophores. As the PRC2 has been shown to catalyse trimethylation of lysine 27 on H3 (H3K27me3) in different organisms [for review see (Köhler and Villar, 2008; Butenko and Ohad, 2011)], the phenotype observed in P. patens PRC2 mutants suggests strongly that H3K27me3 plays an important role in this developmental transition in moss (Mosquna et al., 2009; Okano et al., 2009). Nevertheless, genes that are targeted by H3K27me3 modification remain unknown in P. patens.

In this study we report the genome-wide histone modification patterns of P. patens, integrated into functional, structural and dynamic frameworks. Genome-scale epigenetic signatures of five commonly studied histone modifications (H3K4me3, H3K27me3, H3K27Ac, H3K9Ac and H3K9me2) were generated using ChIP-seq. In addition to providing epigenomic profiles of histone modifications in a haploid plant, we investigated switches of these chromatin marks under two contrasting situations: (i) developmental transition from protonema to gametophores; and (ii) dehydration stress on gametophores. Our study provides comprehensive insight into the moss epigenome and its dynamics, which reveals distinctive behaviors of histone marks in response to development and abiotic stress.


Characterizing the Physcomitrella patens chromatin landscape

Genome-wide maps of five histone modifications (H3K4me3, H3K27me3, H3K27Ac, H3K9Ac and H3K9me2) were generated using chromatin immunoprecipitation followed by high-throughput sequencing on the SOLiD platform. These histone modifications were chosen for two main reasons: (i) they fall into the main chromatin types/states recently defined in Drosophila melanogaster and Arabidopsis thaliana (Filion et al., 2010; Roudier et al., 2011); and (ii) they are involved in altering lysine residues that can be modified by contrasted post-translational modifications that can produce opposite functional output (methylation versus acetylation), thus representing flexibility in chromatin messages from the same modification site. Furthermore, to assess how the epigenome changes in response to developmental transition and environmental stress, epigenomic profiling was performed in three conditions: (i) 8-day-old protonemata representing the juvenile stage of moss development (filamentous tissue extending by tip growth); (ii) 6-week-old leafy gametophores representing the adult, asexual developmental stage; and (iii) such gametophores slowly dehydrated to 50% water loss (see Experimental Procedures). Thus, together with the controls, 24 libraries were sequenced on the SOLiD4 platform and processed reads were mapped to the Physcomitrella patens V1.6 genome (cf. Data S1 and Figure S1). All the epigenomic profiles generated are available publicly through the genome browser at; an example of an epigenomic profile is represented by a genome browser snapshot in Figure S2. In addition, to determine the relation between the epigenomic profiles and gene expression patterns, transcriptomic data were obtained via microarray profiling and incorporated into our functional analysis.

Global organization of histone H3 modifications in P. patens

Genomic regions decorated significantly with histone H3 modifications (‘peaks’) were identified for all conditions using the MACS (model-based analysis of ChIP-Seq) software (Zhang et al., 2008) (Table S1, for details see Methods S1, Tables S2, S3 and S4). Regardless of the condition, H3K4me3 and H3K9Ac modifications presented the highest number of regions detected (about 22 000 and 25 000, respectively), while H3K27me3 was detected at about 7000 regions. Comparing gametophore and protonema developmental stages, we observed that they share 40 and 60% of regions modified with H3K27Ac and H3K9me2, respectively (Figure S3 and Figure 1). The distribution of identified regions revealed a strong enrichment of H3K9me2 at genomic regions carrying transposable elements (TEs) (Figure 1a and Figure S3b,d). In contrast, the four other histone modifications (H3K4me3, H3K27me3, H3K27Ac and H3K9Ac) showed enrichment in genic regions (Figure 1a,b and Figure S3; see Figure S4 for promoter definition). ChIP-qPCR on selected genes and one TE using two independent samples validated the ChIP-seq data (Figure S5e–j and Table S5). To gain insight into the structural and functional organization of histone modifications over genes, we used the ANCORP (ANchored CORrelative Pattern) pipeline that allows visualization and correlation of genome-wide datasets (Luo and Lam, 2010; Luo et al., 2013). Expression levels (Figure 1c) and gene length (Figure 1d) were used as anchors to sort genes. Histone modification occupancy was represented across aligned transcript units from 1 kbp upstream of the transcription start site (TSS) to 5 kbp downstream. We observed that different histone modifications associated with different expression levels: (i) H3K4me3, H3K27Ac and H3K9Ac coincide with high expression level (q-value = 9.99e–36, Fisher's exact test with subsequent false discovery rate adjustment (Benjamini and Hochberg, 1995) – applied for correction for multiple testing throughout), whereas (ii) H3K27me3 coincides with low gene expression level (q-value = 9.99e–36, Fisher's exact test; Figure 1c and Figure S6a,c). In addition to these functional properties, structural characteristics are also revealed, namely a relative enrichment of histone marks toward the 5′ end of the genes compared with their 3′ end (Figure 1b,d and Figure S6b,d). Preferential association of histone marks with distinct ranges of gene lengths is evident. While H3K27me3 is associated with short genes (as defined in the Experimental Procedures section; q-value = 1.3e–5, Fisher's exact test), H3K4me3, H3K27Ac and H3K9Ac tend to be associated with long genes (q-value ≤ 1.66e-35, Fisher's exact test) (Figure 1d and Figure S6b,d). The general absence of H3K9me2 over genes is also observed, as indicated previously (Figure 1 and Figure S6).

Figure 1.

Genome-wide distribution of histone modifications.

(a) Distribution of significantly enriched histone modification regions in genic, intergenic and transposable elements (TE) regions in protonema.

(b) Distribution of histone modification regions within genic annotations in protonema.

(c) Histone modification occupancy density of annotated genes sorted by expression level (for protonema). All genes were aligned to the transcription start site (TSS, indicated by the white dashed line) and ranked according to their expression level. Each line represents a single gene with 1 kbp upstream and 5 kbp downstream of the TSS. Histone modification occupancy is indicated as a heat map with high (yellow) and low (blue) values.

(d) Histone modification occupancy density for protonema of annotated genes sorted by gene length. Legend identical to (c), however regions downstream of annotated transcripts are masked with a grey colour. For analogous representations of gametophores and dehydrated gametophores see Figures S3 and S6.

Chromatin marks are preferentially arranged into 15 chromatin states

To study the combinatorial association of histone modifications, we defined and visualized CS for each condition using the ANCORP pipeline (Luo et al., 2013). Briefly, a gene is considered to be decorated by a chromatin modification if there is an overlap of an identified peak of the modification with the gene (from transcription start to transcription stop). The resulting CS of each gene was represented by a binary code: presence or absence of any detected histone modification. The CS for all P. patens v1.6 annotated genes (Zimmer et al., 2013) were organized into clusters, based on their binary code (chromatin content) to reveal: (i) groups of genes with identical CS; and (ii) the association among individual modifications as illustrated by a dendrogram (Figure 2a and Figure S7). Out of the 32 CS theoretically possible, 27 combinations were observed at least once in all three conditions by this approach (Table S6 and Figure 2a). To check whether the detected CS are significantly biased as compared with random expectation, the combinatorial frequency of the individual histone marks was compared with the observed number of CS (Table S6). This method demonstrates that 15 out of the 27 observed CS are significantly biased as compared with random expectation in all three conditions (q-value < 0.05, see Experimental Procedures for details), denoting preferential association of particular chromatin modifications. Out of these 15 CS, five are less frequent than expected (namely, the five individual marks), while the other 10 are more abundant than expected by chance (Table 1). Among those CS, the CS #9 H3K4me3-H3K9Ac is the only one that switches abundance in that it is more frequent than random in gametophores and less frequent than random in protonemata. Moreover, only six CS represent between 70–80% of the annotated genes (CS #5 H3K4me3-H3K9Ac-H3K27Ac; #6 H3K4me3-H3K9Ac-H3K27me3; #9 H3K4me3-H3K9Ac; #13 H3K9Ac, #15 H3K27me3 and #16 no modification), regardless of the experimental condition.

Table 1. Summary of the 15 statistically significant CS and the ‘no modification’ state. The third column shows the abundance of the CS compared with random expectation (green for higher and red for lower; all but CS #9 show the same tendency throughout). Columns 4–8 mark the contribution of individual chromatin marks to the respective CS (grey for absence and brown for presence). Column 9 contains the observed number of genes per CS in gametophores, while columns 10 and 11 denote whether the observed number of genes in the other conditions are significantly different (red for less and green for more) or not (white). Columns 12 and 13 denote the tendency of CS to mark genes with respect to expression levels and gene length (expression levels and length categories as defined in Experimental Procedures; n.a. = no association)Thumbnail image of
Figure 2.

Global organization of histone modifications into chromatin states in Physcomitrella patens.

(a) Organization of annotated P. patens genes based on the binary chromatin state of individual genes (gametophore). Each row in the figure represents the digitized pattern of chromatin modifications for a single gene: yellow depicts the presence and black the absence of histone modification. The number on the left defines the 15 chromatin state (CS) significantly different from random expectation (see Tables 1 and S6). The dendrogram at the top indicates clusters of individual modifications. Gene expression level (b) and gene length (c) were plotted as correlative patterns of (a). Green lines in (c) indicate transcripts. For analogous representations for protonema and dehydrated gametophores see Figure S7.

H3K27 is a key indexing residue

We observed a higher number of regions marked with H3K27Ac in protonemata as compared with gametophores (Figure 1 and Figure S3a). At the level of CS, this increased number of protonemal H3K27Ac-decorated regions results in an increased proportion of genes belonging to CS #5 (H3K4me3-H3K9Ac-H3K27Ac) in protonema (37.9%) as compared with gametophore (14.0%) (Table 1). At the same time, fewer genes were found in CS #9 (H3K4me3-H3K9Ac) in protonema (9.8%) as compared with gametophores (35.2%) (Table 1; also compare Figure 2a with Figure S7a). In total, 7419 CS #5 genes lose potentially activating H3K27 acetylation during the developmental transition from juvenile (protonemata) to adult (gametophore) growth form. Gene Ontology bias analyses revealed that those genes are enriched for RNA processing, RNA and protein metabolism and translation, while they are depleted for response to different stimuli (Figure S8). Protonemata are considered to represent an, albeit transient, housekeeping stage, and they are in much closer contact with the substrate than gametophores. The removal of a synergistically activating mark (see below) might render this large subset of genes (that might typically play a less important role in gametophores as in protonemata) more ready for deactivation by (tri)methylation at K27 of histone H3, respectively might be resulting in toning down gene expression strength.

To determine the relations between CS, expression levels and gene length, we used ANCORP, employing the gene order resulting from the CS clusters as an anchor to plot the expression levels and gene length as correlative patterns (Figure 2 and Figure S7). As observed at the level of individual histone marks, particular CS also associate with distinct expression levels and gene length. Regarding gene length, CS that contain H3K27me3 tend to mark short and shorter than average genes as compared with H3K4me3-containing CS (q-value ≤ 2.21e–6, q-value ≤ 0.03, respectively, Fisher's exact test) (Table 1 and Figure 2a,c) and Figure S7). Regarding expression levels, the CS analysis revealed a positive synergistic effect of the combination of the three marks H3K4me3, H3K9Ac and H3K27Ac on transcription (Table 1; Figure 2a,b and Figure S7). Indeed, expression levels are highest when these three marks are combined (CS #5 H3K4me3-H3K9Ac-H3K27Ac; Figure 2 and Figure S7 and Table 1). Interestingly, H3K27Ac is rarely found alone. For instance in protonema, 87% of H3K27Ac marked genes are also marked with both H3K4me3 and H3K9Ac. This preferential association of H3K27Ac with both H3K4me3 and H3K9Ac, together with high expression levels, suggests that H3K27Ac is important to reinforce transcriptional activity. This suggestion is also supported by expression levels that are much higher for CS #5 (H3K4me3–H3K9Ac–H3K27Ac; on average 112e3, 87e3 and 61e3 in gametophores, dehydrated gametophores and protonema, respectively) than for CS #9 (H3K4me3–H3K9Ac; 57e3, 46e3 and 27e3). In contrast, H3K27 residues that are trimethylated (H3K27me3) rather than acetylated (H3K27Ac), are associated with lower expression levels. For example, CS #6 (H3K4me3–H3K9Ac–H3K27me3) coincides with low expression levels (q-value = 9.99e–36, Fisher's exact test) while the CS differing by absence of H3K27me3, CS #9 (H3K4me3–H3K9Ac) coincides with high expression levels (q-value = 9.99e–36, Fisher's exact test) (Table 1). This difference in expression levels is even more pronounced when comparing the two CS that differ by different marking of the H3K27 residue (H3K27Ac versus. H3K27me3): CS #5 (H3K4me3–H3K9Ac–H3K27Ac) coincides with high expression levels (q-value = 9.99e–36, Fisher's exact test; expression levels see above) whereas CS #6 (H3K4me3–H3K9Ac–H3K27me3) coincides with low expression levels (see above; average expression levels 23e3, 25e3, 12e3). As the CS were obtained using whole tissues and cannot be resolved into specific cell types resolution, these observations need to be carefully interpreted (see Discussion). Interestingly though, we found that the 206 genes that toggle from potentially activating CS #5 (H3K4me3–H3K9Ac–H3K27Ac) in protonema to potentially repressing CS #6 (H3K4me3–H3K9Ac–H3K27me3) in gametophores are enriched for Gene Ontology (GO) terms involved in organ/reproductive development (Figure S9). However, none of these 206 genes is transcriptionally down-regulated in gametophores. Rather, the average expression of these 206 genes is already low in protonemata (43e3, median 13e3; while 61e3 is the average for CS #5 in protonema) and remains low in gametophores (average 46e3, median 13e3; 23e3 average for CS #6 in gametophores).

Taken together, our data reveal a strong interplay between the marking of the H3K27 residue and transcriptional output, thus underlining an important role of the H3K27 residue in chromatin indexing. The data emphasize that there are explicit combinations of histone marks in P. patens that are preferentially associated with tissue/developmental stage, gene length or expression levels (Table 1).

Dynamic regulation of histone modification during developmental transition and dehydration stress

We investigated histone modification toggling (as revealed from gain or loss of particular chromatin marks) in two situations: (i) the two primary gametophytic tissues, protonemata and gametophores (representing a developmental transition from juvenile to adult plant); and (ii) an abiotic stress condition on gametophores, namely dehydration to 50% water loss. RT-qPCRs on selected differentially expressed genes (DEG) were performed in order to validate this transcriptomic data set (Figure S5a–d and Table S5) and showed good congruence, as previously described for this platform (Wolf et al., 2010). Both responses are accompanied by almost exclusive up-regulation of gene expression: (i) regarding the developmental transition, 248 genes were found to be up-regulated in gametophores as compared with protonemata and only three genes were down-regulated; and (ii) regarding dehydration stress, 205 genes were found to be up-regulated in dehydrated gametophores as compared with control, whereas 11 genes were found to be down-regulated (Table S7; all q-value < 0.05, Cyber-t-test). These numbers of DEG are defined in a strict way to exclude false positives, and are in line with previous analyses (e.g. Wolf et al., 2010). Gene Ontology bias analyses were performed on these two DEG sets (Figure 3a,b). As expected, the set of DEG up-regulated under dehydration is most significantly enriched for GO terms associated with water deprivation, whereas the developmental DEG set is enriched for GO terms related to general functions (such as CO2 transport, response to light, primary cell wall biogenesis) and depleted of genes involved in gene expression. From these two DEG sets, CS of genes were retrieved for each developmental stage and/or condition (protonemata, gametophores and gametophores under drought). Figure 4(a–d) represent proportion of genes associated with a given CS for both DEG sets as compared with all genes marked by this CS. Under all conditions the CS containing no histone H3 modification (#16) is detected for a smaller proportion of genes in the DEG sets as compared with the genome-wide situation, suggesting that DEG are more susceptible to be decorated with histone modifications than are unregulated genes. In addition, some CS reveal significant changes with regard to gene proportion within the DEG sets as compared with all genes marked by the corresponding CS. In particular, CS within the developmental DEG set show a higher proportion of H3K27me3-containing CS in protonema (CS #11: H3K9Ac–H3K27me3; #8: H3K4me3–H3K27me3, #6 H3K4me3–H3K9Ac–H3K27me3 and #15 H3K27me3), whereas CS within the dehydrated DEG set show a significant higher proportion of the CS containing the three activating marks H3K4me3, H3K9Ac and H3K27Ac in dehydrated gametophores (CS #5 H3K4me3–H3K9Ac–H3K27Ac).

Figure 3.

Gene Ontology (GO) bias of developmental transition and stress response.

Gene Ontology bias analysis was performed on the developmental DEG set (a) and the dehydration DEG set (b); red colour stands for under-represented GO terms whereas green marks over-represented GO terms. (c) and (d) represent unique GO terms significantly biased within the H3K4me3-H3K9Ac chromatin state (CS) (CS #9) for genes belonging to the dehydration DEG in (c) gametophores and (d) dehydrated gametophores (terms common to gametophores and dehydrates gametophores were removed to represent the unique set). The same colour code was used as in (a) and (b).

Figure 4.

Proportion of genes associated with a given chromatin state (CS) for both DEG sets as compared with all genes that possess this CS.

Gene proportion within a given CS found in the respective DEG set were compared with gene proportion found in all genes marked with the same CS for: (a) developmental DEG in protonema and (b) in gametophores; (c) dehydration DEG in gametophores; and (d) in dehydrated gametophores. Asterisks (*) denote significant differences in gene proportion between DEG and all genes marked with the respective CS (q-value < 0.05).

In order to understand which functional gene categories are present in different CS, GO bias analyses of DEGs versus genome-wide were conducted and are summarized in Table S8. Interestingly, these GO analyses revealed substantial differences, even for some CS that showed no significant change with regard to gene proportion: Figure 3(c,d) illustrate one of these examples for CS H3K4me3–H3K9Ac in the gametophore and dehydrated gametophore condition, respectively. Here, the DEGs that carry this CS are enriched for response to water deprivation (Figure 3d), while this situation is not the case for unstressed gametophores. These results demonstrate histone mark dynamics within the DEG sets during development and stress responses. In order to evaluate and visualize these changes in chromatin modification, we illustrated the differentially modified histone marks for the developmental transition (Figure 5a) and for dehydration stress (Figure 5d) within the up-regulated genes for each comparison. Different histone dynamics were observed when comparing the two sets (Figure 5a,d). With regard to developmental transition (Figure 5a), increase in gene expression is associated with a massive loss of H3K27me3 in gametophores, and a significant trend of gaining H3K4me3, H3K9Ac and H3K27Ac marks in comparison with genome-wide histone modification changes (Fisher's exact test, P-value < 0.01 see Figure S10; genome-wide histone modification changes for both situations are represented in Figure S11). With regard to drought, transcriptional induction does not significantly involve H3K27me3 marks, but coincides again with a significant gain of the three activating marks H3K4me3, H3K9Ac and H3K27Ac in dehydrated gametophores as compared with control gametophores (Figure 5d; Fisher's exact test, P-value < 0.01). Figure 5(e) and (f) illustrate histone modification changes on selected genes as displayed in the genome browser. Because the gene Pp1s267_21V6.1 (coding for a late embryogenesis abundant (LEA) protein) belongs to both the developmental and dehydrated DEG sets, we chose this gene to check its chromatin behavior using two independent samples. ChIP-qPCR assays validated the changes found by ChIP-seq on this gene (Figure S5). Taken together, these results suggest that genes up-regulated in response to development (8-day-old protonemata versus 6-week-old gametophores) and to dehydration stress (8 h of dehydration) have similar histone mark behavior regarding the gain of activating marks (H3K4me3, H3K9Ac and H3K27Ac), but differ with regard to H3K27me3, the loss of which being specific to the developmental transition. Of note, 33.8% (85 genes) of the developmental DEG and 39.4% (86 genes) of the stress DEG exhibit no change of CS.

Figure 5.

Dynamic regulation of histone modification in genes up-regulated during developmental transition and stress response.

Up-regulated genes during developmental transition (a, b) and under dehydration stress (c, d) were ordered according to their fold change (most up-regulated on top). Histone modification switches within these two gene sets were represented for developmental transition (a) and (c) (= changes in gametophore as compared with protonema) and for stress response (b) and (d) (= changes in dehydrated gametophores as compared with gametophores): yellow marks a gain of histone mark, blue a loss, whereas no change is represented in black. Significance of histone mark changes (gain and loss) were determined by Fisher's exact test (P-value < 0.01) in comparison to the genome-wide histone modification changes (+ denotes significantly more abundant than genome-wide; −, significantly less abundant; NS, not significantly different). (e) and (f) represent genome browser snapshots of histone modifications for two selected genes (black boxes represent exons, lines represent introns) that respond to developmental transition (Pp1s7_65V6.1, a GDSL domain-containing zinc finger protein) and dehydration stress (Pp1s155_102V6.1, a PI-PLC-X domain-containing protein), respectively.

Dehydration responsive genes undergo significant histone H3 modification switches during developmental transition, but not vice versa

As expected, the developmental DEG set (up-regulated genes) does not show any significant change in chromatin marks during dehydration stress, as revealed from gain or loss of particular chromatin marks (Figure 5b). Surprisingly, however, when assessing the histone modification changes of stress-induced genes during the developmental transition (Figure 5c) we found significant alterations in the level of particular histone marks. Here, significant gain of H3K4me3 and H3K9Ac marks, significant loss of H3K27me3 and less removal of H3K27Ac than observed genome-wide were found for genes up-regulated in response to dehydration stress (Figure 5c: Fisher's exact test, P-value < 0.01, see Figure S8). This unexpected observation led us to investigate the transcriptional behavior of the 205 drought up-regulated genes during the transition from protonemata to gametophores. While 42% (86 genes) of these genes show transcript fold-changes above 1.2 in gametophores as compared with protonema, only 18 (8.8%) were found to be significantly up-regulated (q-value < 0.05, Cyber-t-test) during the developmental transition (Figure S12, outliers in the boxplot). Therefore, the 205 dehydration up-regulated genes are not significantly altered in their expression during developmental transition (except for 18 genes), but nevertheless they show a significant change of histone modifications during development. Given the P. patens life cycle, gametophores are probably more likely to face dehydration stress than would juvenile protonemata, as the spores germinate on wet soil and therefore initial conditions do not present the juvenile plant with dehydration stress (see Discussion). We therefore hypothesize that the uncoupling between gene expression and histone modification in this specific situation may reflect the fact that the histone modification changes observed during the developmental transition might serve to prime genes to respond more rapidly to dehydration stress in the gametophore phase. Indeed, among the 125 dehydration induced genes that change their CS during development are genes annotated as LEA (six genes), ERD (early response to dehydration, two genes), dehydrin/dehydration/desiccation related (five genes). Five more are involved in the response to ROS (reactive oxygen species), and one each in cell wall restructuring and ABA (abscisic acid) synthesis (Table S9).

To assess the specificity of this priming, we examined chromatin changes during the developmental transition for 406 genes that were found to be up-regulated in gametophores under UV-B irradiation (Wolf et al., 2010). Interestingly, the only significant histone mark switches observed in the UV-regulated genes, as compared with the genome-wide dynamic, are reduced loss of H3K9Ac and loss of H3K27Ac marks (P-value < 0.05, Fisher's exact test), but no significant alteration of H3K4me3 and H3K27me3. We also assessed the histone mark changes of predicted ABA responsive genes (Timmerhaus et al., 2011), as ABA is involved in several abiotic stress pathways, including drought and UV-B. Both ABA responsive and UV-B regulated genes share a significant loss of H3K27Ac (P-value < 0.05, Fisher's exact test) as compared with genome-wide changes, plus they exhibited a higher gain of H3K27me3 marks. In summary, all observed differences of UV-B and ABA regulated genes during the developmental transition point in a different direction than those shown by the genes up-regulated by dehydration, supporting the specificity of the epigenetic dehydration priming during the protonema-to-gametophore switch.


Conservation of chromatin organization in land plants

Here, we investigate the chromatin landscape of the model moss P. patens. In addition to being important for evo-devo studies, its life cycle makes P. patens a useful model to study epigenetic processes during the haploid gametophytic generation.

Recent genome-wide profiling of chromatin components in different organisms (metazoans and flowering plants) have identified a low combinatorial complexity of chromatin components into chromatin types/states that have distinct properties (Ernst and Kellis, 2010; Filion et al., 2010; Luo and Lam, 2010; Kharchenko et al., 2011; Riddle et al., 2011; Roudier et al., 2011; Dunham et al., 2012; Luo et al., 2013). Here, we find that most histone marks are organized into 15 abundant CS (Table 1) that differ with regard to their association with tissue, transcriptional output and gene length (Table 1). The histone modifications investigated fit with key features attributed to major CS in other organisms. For instance, the significant H3K9me2 enrichment in TEs found in P. patens is a typical attribute also found in CS in Drosophila cells and in the seed plant Arabidopsis thaliana (Filion et al., 2010; Riddle et al., 2011; Roudier et al., 2011; Luo et al., 2013). The three marks H3K4me3, H3K9Ac and H3K27Ac are predominantly located within genic regions and coincide with high expression levels (Figures 1 and 2), two features that are similar to animals and plants (reviewed in Van Steensel, 2011). Interestingly however, combinations of these histone modifications are necessary to associate the CS with high expression levels, as the CS representing either of the single marks alone are not associated with strong expression (Table 1). Concerning H3K27me3, our data show that this histone mark is significantly associated with repressed genes in P. patens (Figures 1 and 2). This finding is consistent with numerous previous studies that demonstrated that H3K27me3 is a signature of the conserved Polycomb repressive complex 2 (PRC2), involved in transcriptional gene silencing in many organisms (Hennig and Derkacheva, 2009; Simon and Kingston, 2009; Morey and Helin, 2010), including P. patens (Mosquna et al., 2009; Okano et al., 2009). Taken together, the P. patens genome-wide chromatin landscape presented here illustrates that typical chromatin features known from diploid organisms (i.e. segregation of histone modifications into ‘repressive’/’active’ marks and into genes/TEs) (Roudier et al., 2009; Van Steensel, 2011; Dunham et al., 2012) hold true for the haploid-dominant moss as well. As reported for other plant studies (Roudier et al., 2009; Luo and Lam, 2010), the apparent ‘spreading’ of the H3K27me3 mark in the gene body also distinguishes it from the three gene activation marks that are more spatially localized near the TSS of genes.

H3K27 represents a key indexing residue

Regarding combinatorial association of chromatin marks, we found that the H3K27 residue performs a key role in chromatin indexing dependent on its post-translational modification: acetylation versus trimethylation. The H3K27Ac mark is associated preferentially with both H3K4me3 and H3K9Ac marks and the presence of these three marks on genes coincides with the highest expression levels (Figure 2 and Figure S7 and Table 1). Similarly, preferential association of H3K27Ac with H3K9Ac, coupled with high gene expression level, has also been found in Arabidopsis (Charron et al., 2009). This observation suggests that H3K27Ac deposition may require the presence of both H3K4me3 and H3K9Ac marks, although other scenarios are possible including the one in which deposition of these three activating marks occurs independently in different cell types (see below). This situation could be similar to those found in a different context in mammalian stem cells and Drosophila embryos. The emerging picture from studies with these organisms suggests that cis-regulatory enhancer elements undergo high histone modification dynamics: H3K4me1 deposition precedes that of H3K27Ac to activate enhancers, an effect that is counterbalanced by the repressive H3K27me3 mark (Creyghton et al., 2010; Hawkins et al., 2011; Rada-Iglesias et al., 2011; Bogdanovic et al., 2012; Bonn et al., 2012). As in the moss we also found that genic CS differing only in acetylation versus trimethylation of the H3K27 residue have opposite transcriptional outputs, it is likely that H3K27 represents a key residue in the chromatin language of many organisms. Nonetheless, the significance of CS needs to be interpreted with caution, as many studies simultaneously sample multiple tissues and cell types that might represent a mixture of distinct CS. In the present study, chromatin profiling was performed on tissues as well. Even if P. patens has the advantage to have a simple anatomy and low complexity in tissue organization, the gametophores and protonemata sampled contain distinct cell types that perform specific functions (e.g. plant growth, photosynthesis, nutrient assimilation). Thus, the observed CS could be due to different chromatin states occurring in particular cell types. Additional work at the cell-specific level would provide a deeper understanding of the patterns and significance of chromatin organization in P. patens (Deal and Henikoff, 2010; Steiner et al., 2012). However, the H3K27me3 mark has a peculiar behavior, as this histone mark showed toggling in response to developmental transition and not under drought stress. Similar to the situation in other organisms, K27 trimethylation of histone H3 apparently marks genes involved in development. From intensive characterization of proteins that mediate H3K27me3 deposition (PRC2 proteins), the involvement of H3K27me3 in developmental processes has been well documented in both animals and plants (Köhler and Villar, 2008; Butenko and Ohad, 2011; Lafos et al., 2011; Bemer and Grossniklaus, 2012). PRC2 has been shown to be required for maintenance of stem cells in animals as well as for meristematic activity in plants (Reyes, 2006; Hennig and Derkacheva, 2009). In addition, these proteins are key players for cell fate determination in animals and developmental transition in plants (such as the transition from embryo to seedling, vegetative to reproductive phase, flower organ formation) (Hennig and Derkacheva, 2009; Schuettengruber and Cavalli, 2009; Bouyer et al., 2011). In P. patens, loss of function in PRC2 components prevents formation of gametophores (Mosquna et al., 2009; Okano et al., 2009). This finding is in agreement with the H3K37me3 genome-wide dynamic shown here, that emphasizes an important role of H3K27me3 during the protonema-to-gametophore transition.

Stress-induced genes in gametophore are primed during the protonema-to-gametophore transition for future response to drought stress

During drought stress of gametophores, the three activating marks H3K4me3, H3K27Ac and H3K9Ac are highly dynamic (Figure 5d). Focusing on a restricted number of drought stress-responsive genes, Kim and co-workers (Kim et al., 2008) showed that these three marks are also associated with genes that are up-regulated during drought stress in Arabidopsis. Genome-wide data in response to dehydration stress are limited, but recent studies focused on H3K4 methylation show that H3K4me3 patterns respond dynamically to dehydration stress in Arabidopsis and rice (van Dijk et al., 2010; Zong et al., 2013), a situation also found in this study with Physcomitrella (Figure 5d).

We observed significant histone mark changes during the protonema-to-gametophore transition for genes that are transcriptionally induced in gametophores during dehydration stress (Figure 5c). We hypothesize that the histone modification changes observed during the developmental transition may prime genes to respond more rapidly to dehydration stress later on. Looking at the P. patens life cycle, protonemata arise from spores that germinate under wet conditions, typically in mud or moist ground. Within a few days after germination, protonemata start to bud and to develop gametophores. Protonema is therefore not expected to suffer overly from dehydration. Gametophores, however, are more likely to be subjected to repeated rounds of dehydration and rehydration, as they are not in direct contact with the soil and their development continues for several weeks or months. Thus, preparing genes during the protonema-to-gametophore transition to respond to forthcoming drought stress could be an adaptive response that would allow the plant to better cope with fluctuating water availability in the gametophore stage. Interestingly, this CS priming is not found in the case of UV-B induced genes (Wolf et al., 2010), perhaps because UV-B stress is more likely to affect moss at both the protonema and gametophore developmental stages. Also, ABA regulated genes (Timmerhaus et al., 2011) do not follow the trend seen for the genes induced by dehydration.

The idea of a role for histone modification changes in preparing genes to respond to later signals has emerged in embryonic stem cells of animals, playing a role at the developmental level (cell fate choice) rather than being involved in stress response (for review see (Pietersen and van Lohuizen, 2008; Vastenhouw and Schier, 2012). In a recent flowering plant study (Sani et al., 2013), hyperosmotic treatment was used to acclimate Arabidopsis seedlings. In their study, genes were found to be primed epigenetically to respond to this stress by shortening and fragmentation of H3K27me3 islands. In Physcomitrella, we find that those genes that respond to drought are significantly decreased in their H3K27 trimethylation during development, suggesting that in plants developmental progression might be used to prime genes for activation under certain stress conditions later on. The differences in CS toggling observed between dehydration versus UV-B stress and genes regulated by the stress hormone ABA might be founded in the poikilohydric lifestyle of P. patens–changing water availability is an expected environmental cue.

In summary, we describe a genome-wide, integrated epigenomic map for P. patens and provide a dynamic portrait of key histone modifications in response to the principal gametophytic developmental transition and to subsequent dehydration stress. We expect this survey to serve as a resource for the community to further explore and compare chromatin-based regulation of gene activity in Physcomitrella and in cross-species approaches, but also to help developmental biologists and physiologists to decipher the relationship and interactions between chromatin and specific developmental and physiological states.

Experimental Procedures

Plant material and growth conditions for ChIP experiments

The Gransden 2004 strain of Physcomitrella patens was used in this study (Rensing et al., 2008). Cultures were grown in 9-cm Petri dishes on PpNH4 medium which corresponds to the minimal medium described by (Ashton et al., 1979), supplemented with 5 mM NH4-tartrate. Medium was solidified with 0.7% Agar (Sigma-Aldrich A7002, and overlaid with a cellophane disk for protonema culture. Cultures were grown under the following environmental conditions: 16 h/8 h light/dark cycle, 80 μmol sec−1 m−2 photosynthetically active radiation, 22/19°C day/night temperature. Eight-day-old protonemata and 6-week-old gametophores were harvested. Drought stress was performed by transferring 6-week-old gametophores from agar plates to the plastic base of 9-cm Petri dishes. Dishes were placed without lids in desiccators that contained saturated NaCl to provide a 75% relative humidity atmosphere for 8 h at 22°C. These conditions allowed gametophore colonies to reach 50% fresh weight (water) loss in a reproducible way (see Figure S13 for stress-induced phenotype).

Chromatin immunoprecipitation (ChIP) and sequencing

The detailed ChIP protocol is available in Methods S1. Briefly, antibody cross-reactivity was checked using MODified Histone Peptide Array and The MODified Array Labeling kit (Active Motif, (Figure S14). The validated antibodies used are listed in Table S10. The ChIPed DNA from two identical reactions was pooled to construct Applied Biosystems SOLiD 4 System libraries following the manufacturer's protocol. Sequencing was carried out by CeGaT (Center for Genomics and Transcriptomics, using the SOLiD 4 system, resulting in 50-nt long single ended reads.

Microarray and differentially expressed genes

Plants were grown under long day conditions (16 h white light/8 h dark) on Knop medium as described previously (Wolf et al., 2010). Therefore, growth conditions were very similar to those for the ChIP samples except for NH4 supplementation. Dehydration was performed as described above until the plants had lost 50% fresh weight. Isolation of RNA, generation of microarray raw data and data processing was also carried out as described in Wolf et al. (2010). Briefly, DEGs for the developmental transition (protonema on solid medium versus asexual gametophores) and dehydration stress (asexual gametophores versus gametophores dehydrated to 50% water loss) were detected between treatment and control (three biological replicates for each condition) using the Cyber-t-test (Cho and Walbot, 2001) with subsequent false discovery rate adjustment (Benjamini and Hochberg, 1995), selecting only genes with a q-value < 0.05 (Table S7). All expression data were normalized by median scaling to 10 000. As the microarray platform is based on v1.2 gene models, those were mapped onto the v1.6 gene models and all subsequent analyses conducted on the latter.

See Data S1 for ChIP, quantitative polymerase chain reaction (PCR) analysis, and all computational and statistical analysis.

Data access

The ChIP-seq data from this study have been submitted to the NCBI Sequence Read Archive (, the accession numbers are listed in Table S11. ChIP-seq peaks and gene annotation tracks are additionally available via for browsing and visualization.

The microarray data are available in the Array Express ( public repository under the accession numbers E-MTAB-914 (dehydration) and E-MTAB-917 (protonemata and gametophores). In addition, these data are available via as microarray datasets (protonema samples are called: Media_Gransden_WT9_LKM pH 5.8_ptn_rep_1, 2 and 3; gametophore samples are called: Media_Gransden-WT9_SKM pH 5.8_ptn_rep1, 2 and 3 and dehydrated gametophores are called: Light/dehydr._Gransden-WT9_dehyr. 50%FW_jgam_rep1, 2 and 3).


We thank Dr Mark Diamond (Rutgers University) for his advice on sample sonication; Dr Rong Di (Rutgers University) for her technical support and Drs Daniel Lang (Freiburg University) and Nir Ohad (Tel Aviv University) for discussions on the manuscript. Funding for this work was provided by Marie Curie Actions (European Commission EMBOCOFUND2010, GA-2010-267146) and EMBO Long-Term Fellowships to T.W. (ALTF 1166-2011) and by the German Ministry of Education and Research to S.A.R. (BMBF FRISYS). Funding for chromatin work done in E.L.s and M.L.s laboratories was provided by support from the School of Environmental and Biological Sciences at Rutgers University. T.W. expresses his thanks to Dr Jerzy Paszkowski (University of Geneva) for his support.

Conflicts of Interest

The authors declare no competing financial interests or other conflicts of interest.