Abstract
In eukaryotic cells, linker histone H1 anchors in and out ends of nucleosome DNA to promote chromatin to fold into the 30 nm fiber. However, if H1 plays a role in coordinating the three-dimensional (3D) chromatin architecture, DNA methylation, and transcriptional regulation is not clear. We engineered H1 knockout mutants in Arabidopsis thaliana which shows pleiotropic phenotypes. Using High-throughput Chromosome Conformation Capture (Hi-C), we found that H1 complete depletion dampens inter- and intra-chromosomal interactions, as well as intra- and inter-chromosomal arm interactions. MNase accessibility assays followed by sequencing (MNase-seq) showed that the nucleosome density decreases in centromeric regions and increases in chromosome arms. In contrast, DNA methylation level in CHG and CHH contexts increases in centromeric regions and decreases in chromosome arms as revealed by whole genome bisulfite sequencing (WGBS) in h1 mutant. Importantly, the functional link between DNA methylation and gene transcription is defected, and the extensive switches between chromatin compartment A and B are uncoupled from genome-wide DNA methylation and most of gene transcriptions upon H1 depletion. These results suggested that linker histone H1 works as linkers among chromatin compartmentalization, DNA methylation and transcription.
Genomic DNA in eukaryotic cells is densely packaged into chromatin to fit within the small volume of nucleus. The eukaryotes achieve DNA package depending on two types of major proteins, the core histones (H2A, H2B, H3 and H4) and the linker histone H1. Two copies of each of the four core histones wrapped with ∼147 bp of DNA termed as nucleosome core particle together with an additional variable length of linker DNA constitute nucleosome, which is the repeating structure unit of chromatin (Kornberg 1974; Luger et al. 1997). In addition, H1 binds to nucleosome with ∼10 bp of DNA at both the entry and exit sites of nucleosome core particle to facilitate the folding of chromatin into a 30 nm fiber (Bednar et al. 1998; Song et al. 2014; Thoma et al. 1979). H1 is the most variable histone, with multiple H1 family members are present in different organisms. For example, mice and human contain H1-coding genes up to eleven, consisting of seven somatic subtypes and four germ line-specific subtypes. Deletion of any one or two H1 subtypes did not noticeably affect mouse development due to the compensatory effect of other subtypes (Drabent et al. 2000; Fan et al. 2001; Fantz et al. 2001; Lin et al. 2000; Sirotkin et al. 1995). Elimination of H1c, H1d and H1e led to a decrease in the ratio of H1 to nucleosome core particle up to 50 % and embryonic lethality. In addition, deletion of the three H1 subtypes induced the decrease of global nucleosome spacing, reduction of local chromatin compaction, specific effect on transcriptions of imprinted or X chromosome genes, massive epigenetic changes and alteration of topological organization at the most active chromatin domains (Fan et al. 2005; Maclean et al. 2011; Yang et al. 2013).
H1s in plants are divided into two groups: the main H1s which are ubiquitously and stably expressed, and the minor H1s which are stress-induced and evolutionarily conserved from monocotyledonous to dicotyledonous plants. The drought stress and abscisic acid (ABA)-inducible H1 was initially discovered in wild tomato Solanum pennellii, followed by its identification in different organisms, including Arabidopsis and Nicotiana tabacum (Cohen and Bray 1990; Cohen et al. 1996). Being different from mammalians, Arabidopsis H1 is encoded by only three genes: a stress-inducible minor variant H1.3 (AT2G18050) and two main variants H1.1 (AT1G06760) and H1.2 (AT2G30620) (Ascenzi and Gantt 1997; Gantt and Lenvik 1991; Przewloka et al. 2002). H1.1 and H1.2 are distributed in all vegetative tissues and organs, including leaves, roots, hypocotyls and meristems, and H1.3 is detected in constitutive guard cell-specific tissues and facultative environmentally regulated tissues (Rutowicz et al. 2015).
Plant DNA methylation occurs in CG, CHG and CHH sequence contexts which are primarily catalyzed by DNA methyltransferase 1 (MET1), chromomethylase 3 (CMT3) and the de novo methyltransferase (domains rearranged methylases, DRMs) (Law and Jacobsen 2010). The majority of DNA methylation occurs in transposable elements (TEs) in CG, CHG and CHH contexts and is essential to the inhibition of TE activity. Substantial methylation is also found in the bodies of active genes, in which generally occurs in CG context (Law and Jacobsen 2010). In Arabidopsis, H1 was known to be involved in DNA methylation and gene transcription. Knockdown of H1 in Arabidopsis causes stochastic changes in both hypo- and hyper-DNA methylation in a variety of gene contexts (Wierzbicki and Jerzmanowski 2005). The effect of H1 on DNA methylation has been explained by that DNA METHYLATION 1 (DDM1) plays a role in removing H1 from chromatin to facilitate the access of DNA-methylation machinery (Lyons and Zilberman 2017; Rea et al. 2012; Wierzbicki and Jerzmanowski 2005; Zemach et al. 2013). In addition, H1 and DNA methylation jointly repress TEs and aberrant intragenic transcripts (Choi et al. 2020). The mutation of H1 was known to affect nucleosome density (Choi et al. 2020; Rutowicz et al. 2019; Willcockson et al. 2020). Genome-wide nucleosome maps in yeast, animals and plants revealed that the nucleosome distribution patterns correlate with high-order chromatin organization and transcription levels (Lee et al. 2004; Li et al. 2014; Parnell et al. 2008; Sala et al. 2011; Weiner et al. 2010; Yuan et al. 2005). However, the function of H1 in maintaining high-order chromatin organization is poorly understood in plant.
Hi-C is widely used to map chromatin organization of architectures (Feng et al. 2014; Liu et al. 2016; Wang et al. 2015), and revealed that the chromatin is partitioned into different domains based on different scales and compaction levels, including A/B compartments (Meaburn and Misteli 2007; Wang et al. 2015), TADs (Dixon et al. 2012; Lieberman-Aiden et al. 2009; Nora et al. 2012), and chromatin loops (Jin et al. 2013; Rao et al. 2014). The compartment A for euchromatin with active transcription and compartment B for heterochromatin with repressed transcription can be classified in chromatin regions based on the interaction pattern. There are no obvious TADs, but TAD-like domains in Arabidopsis (Dong et al. 2017; Liu et al. 2017). The smaller structure feature is chromatin loops which appear at 10 kb to 1 Mb within TADs (Phillips and Corces 2009) and play roles in transcription, recombination and replication (Mukherjee and Mukherjea 1988). The studies of Arabidopsis three-dimensional genome revealed that genome doubling modulates the transcription genome-widely by changed chromatin interactions (Zhang et al. 2019); 3D chromatin organization rearrangement correlates with heat stress-induced transposon activation (Sun et al. 2020); theme of genome structure is the formation of structural units correspond to gene bodies (Liu et al. 2016); KNOT, a structure similar to flamenco locus of Drosophila, is present in Arabidopsis (Grob et al. 2014); and chromatin interactions are related to various epigenetic marks in active or inactive chromatin (Feng et al. 2014).
Recently, partial (about 50 %) depletion of H1 in mouse embryonic stem cells was found to cause no significant change in the overall 3D genome (Geeven et al. 2015); The 3D genome organization is correlated with chromatin compaction and the epigenetic landscape in mouse partial h1 mutants including h1c/h1d/h1e triple (Willcockson et al. 2020) and h1c/h1e double mutant (Yusufova et al. 2020). In this study, we generated Arabidopsis H1 null mutants, and applied Hi-C and genome-wide approaches to revealing the roles of H1 in regulating 3D genomic organization, nucleosome distribution, DNA methylation and transcription. We showed that H1 depletion impairs the functional links among 3D chromatin interactions, DNA methylation or transcription. We further revealed that H1 modulates the nucleosome density and distribution along the chromosomes, which correlate with the changes of 3D chromatin structure, DNA methylation and transcription.
Results
The phenotypes and transcriptome of Arabidopsis linker histone h1 null mutant
To analyze the role of H1 in chromatin organization, we first generated an h1.1-1h1.2-1h1.3-1 triple T-DNA insertion line which shows no obvious developmental defects (Supplemental Fig. S1A, B) and has only slightly decondensed chromocenters (Supplemental Fig. S1C) as the transcription of H1.2 in this mutant with T-DNA inserted in the promoter of H1.2 is significantly induced (Supplemental Fig. S1D). We then constructed h1 null mutants by crossing h1.1-1 h1.3-1 double mutant with two CRISPR/Cas9-edited (Feng et al. 2013; Yan et al. 2015) independent h1.2 mutants, h1.2-2 and h1.2-3, which have pre-mature stop codons due to a nucleotide insertion of “G” or “A” in the second exon of H1.2, 949 bp and 906 bp from transcription start site (TSS), respectively (Supplemental Fig. S1E). To eliminate the potential off-target mutations in the CRRISPR experiments, we compared the phenotypes between h1.1-1 h1.2-2 h1.3-1 (h1-1) and h1.1-1 h1.2-3 h1.3-1 (h1-2) with each of them harboring an independent CRISPR/Cas9-edited h1.2 mutation, we found that these two lines show similar phenotypes including decreased growth and serrated leaves (Fig. 1A, B). For further studies, we used h1-1 (named as h1 hereafter) in which the complete deletion of H1 was confirmed by western blots, compared to the partial depletion of H1 in h1.1-1 h1.2-1 h1.3-1 mutant (Supplemental Fig. S1F). The primary roots of h1 seedlings are shorter than those of wild type (Col) (Fig. 1C), and enlarged nuclei in h1 leaves were indicated by guard cell nuclei without endoreduplication (Supplemental Fig. S1G, H). At the subnuclear level, we observed that H1 is highly enriched in DAPI-dense heterochromatic chromocenters in wild type (Supplemental Fig. S2A), and depletion of H1 causes dramatic decondensation of chromocenters (Fig. 1D). We then compared the subnuclear distributions of H1.1, H1.2 and H1.3, and found that H1.2 co-localizes with H1.1 or H1.3 in nuclear foci by transiently co-expressing H1.1-YFP/H1.2-mCherry or H1.3-YFP/H1.2-mCherry (Supplemental Fig. S2B).
To evaluate the impact of H1 on transcription, we profiled the transcript levels in h1 by RNA-seq. Among 1974 genes significantly regulated in h1, 1268 are up-regulated and 706 down-regulated (Fig. 1E; Supplemental Fig. S3A; Supplemental Table S1). Gene ontology (GO) analysis revealed that these differentially expressed genes (DEGs) are involved in a variety of response processes (Supplemental Fig. S3B; Supplemental Table S2), indicating the functions of H1 in responses to environmental and developmental cues. To verify the results of RNA-seq, we analyzed the transcript levels of several response-related genes by qRT-PCRs. The results were consistent with those in RNA-seq (Supplemental Fig. S3C; Supplemental Table S1). In addition, we found that the differentially up-regulated and down-regulated genes distribute in all five chromosomes (Fig. 1F).
H1 depletion causes the dampened inter- and intra-chromosomal interactions and extensive A/B compartment switches
To dissect genome-wide chromatin architecture, we performed Hi-C experiments of h1 mutant (Supplemental Table S3) and Col with each of them having a high reproducibility between two biological replicates (Supplemental Fig. S4A, B). The Hi-C data of Col were obtained from our previous report (Zhang et al. 2019), and used as the control because the seedlings of h1 null mutant and Col were grown and sampled in parallel at the same time and conditions. Compared to the typical Hi-C heatmap of Col, a clearly homogenous pattern of Hi-C heatmap was observed for h1 mutant (Fig. 2A). To address the chromosome clustering traits, we calculated the interaction difference matrix from the reads with the same sequencing depth between Col and h1. The genome-wide chromatin interaction difference matrix between Col and h1 revealed that h1 knockout results in dampened inter- and intra-chromosomal interactions (Fig. 2A-C), decreased inter- or intra-arm interaction (Fig. 2D, E), and inter-pericentromeric or telemetric interaction frequencies (Fig. 2F, G).
To quantitatively assess the chromatin contacts, we calculated interaction decay exponents (IDEs), which characterize chromatin packing as the slopes of a linear fit of average interaction intensities detected at a given range of genomic distance (Grob et al. 2014). The results displayed that IDEs of intra-chromosome arms (Supplemental Fig. S5A-F), pericentromeres (Supplemental Fig. S6A-F) and telomeres (Supplemental Fig. S7A-F) in h1 are all lower than those in Col.
Next, we defined the active (A) and inactive (B) chromatin compartments in h1 and Col by Pearson Correlation (PC1) values (Fransz et al. 2000; Grob et al. 2013). We compared the compartments A/B through the first principal component at a 50 kb resolution between Col and h1 mutant. We found that 31.0% and 27.5% of the genome showed conserved compartment A and B, respectively (Fig. 3A-C). We observed that H1 depletion induces extensive switches between chromatin compartment A and B (Fig. 3A-C; Supplemental Fig. S8A-D; Supplemental Fig. S9A-D), with 18.5% converted from compartment A to B, and 22.9% converted from compartment B to A (Fig. 3C). In addition, we found that the compartment transition from A to B occurs more than B to A on chromosome 1 (Chr1), Chr2, Chr3 and Chr5, and B to A more than A to B only on Chr4 (Fig. 3D).
The nucleosome density decreases in centromeric regions and increases in chromosome arms in h1 mutant
To study the potential relationship between the observed weakened chromatin interactions in h1 and nucleosome density, we first examined micrococcal nuclease (MNase) accessibility which reflects the nucleosome occupancy and gene accessibility by digesting naked DNA through MNase (Li et al. 2014). The results showed that the chromatin in h1 is more resistant to MNase digestion than Col (Fig. 4A), and all of the mononucleosome ratios were significantly decreased under different MNase levels (Fig. 4B). In addition, western blots indicated that the total H3 and H4 protein levels increase in h1 compared to Col (Fig. 4C, D). These results indicated that the complete depletion of H1 results in an elevated average density of nucleosomes in chromatin.
To examine the chromosomal distribution of nucleosomes in h1 null mutant, we performed MNase accessibility assays followed by sequencing (MNase-seq) (Supplemental Table S4) in Col and h1 mutant. The results showed that significant alteration of nucleosome density was presented along all the five chromosomes (Supplemental Fig. S10). To refine the changes of nucleosome packaging in different chromosome regions, we analyzed the nucleosome density on centromeres, pericentromeres and chromosome arms, respectively. We found that the nucleosome density dramatically decreases on centromeres, while increases in both right- and left-chromosome arms. In contrast, nucleosome density in the right- and left-pericentromeres showed no obvious change (Supplemental Fig. S11A-E).
We then analyzed the nucleosome distribution in different types of genes, and found that the overall patterns of nucleosome distribution in protein-coding genes (PCGs), pseudogenes and transposable elements (TEs) in h1 mutant were similar to those in Col. The evenly spaced nucleosome distribution is observed in gene bodies of PCGs, but not in promoters of PCGs (Supplemental Fig. S12A-C). However, compared to Col, PCGs display a higher nucleosome density in their promoter regions (Supplemental Fig. S12A); pseudogenes and TEs show a slight decrease of nucleosome density (Supplemental Fig. S12B, C).
We further analyzed the nucleosome distribution in up-regulated, down-regulated and unregulated genes in h1 mutant. We observed that the evenly spaced nucleosome distribution is absent in the bodies of DEGs, but kept in the bodies of non-DEGs upon H1 depletion (Supplemental Fig. S12D-F). In addition, the nucleosome density in h1 mutant is elevated in promoters, but not in gene bodies, irrelevant of DEGs or non-DEGs (Supplemental Fig. S12G, H).
The levels of CHG and CHH DNA methylations increase in centromeric regions and decrease in chromosome arms in h1 mutant
Loss of H1 was reported to cause the elevation of total genomic DNA methylation (Rea et al. 2005). Given that the nucleosome density shows variation among different chromosomal regions, we investigated the potential variations of DNA methylation along chromosomes upon complete depletion of H1. Compared to Col, we found that the DNA methylation level (Supplemental Table S5) in h1 null mutant increases most obviously over centromeric regions (Supplemental Fig. S13). In addition, DNA methylation levels in CG, CHG and CHH sequence contexts show different distribution patterns along chromosomes (Supplemental Fig. S13B), with the hypermethylation and hypomethylation of CG more evenly distributed in the genome. In contrast, the hypermethylation sites of CHG and CHH locate mainly around centromeres, while the hypomethylation sites of CHG and CHH distribute on chromosome arms with the hypomethylation sites of CHH enriched in the chromosome regions adjacent to centromeres (Supplemental Fig. S13C). Compared to that the nucleosome density decreases in centromeric regions and increases in chromosome arms in h1 mutant, the reverse correlation between DNA methylation and nucleosome density in centromeric regions and chromosome arms was thus indicated.
Next, we analyzed the effect of H1 on DNA methylation in various DNA elements, and found that the DNA methylation level of repeat sequences increased significantly in h1 (Supplemental Fig. S14A). In addition, we found that hypermethylation and hypomethylation of gene bodies have higher ratios for longer genes, and hypermethylation of TEs have a higher ratio for longer TEs, while hypomethylation of TEs is not related to their sizes upon h1 mutation (Supplemental Fig. S14B).
To examine potential genes that might regulate DNA methylation in h1 mutant, we analyzed the transcription levels of DNA methyltransferases (MET1, DRM2, DRM3, CMT2 and CMT3), DNA demethylase (DME, ROS1, DML2 and DML3), histone methyltransferases (SUVH4, SUVH5 and SUVH6), histone demethylase (IBM1) and the chromatin remodeling factor (DDM1) involved in DNA methylation directly or indirectly in RNA-seq data. We found that the transcription levels of these DNA methylation-related genes do not change obviously in the absence of histone H1 (Supplemental Fig. S15), implying that the change of DNA methylation might not result from the altered levels of DNA methylation-related genes in h1 mutant.
DNA methylation is uncoupled from the transcriptional regulation in h1 mutant
We then analyzed the relationship between transcription and DNA methylation which generally serves as a repressive marker of transcription (Zemach et al. 2010). To this end, we first analyzed the normalized RNA-seq FPKM of all differential hyper- and hypo-methylated genes, and found that there are no obvious differences (Fig. 5A). We then analyzed the normalized RNA-seq FPKM of DEGs intersected with hyper- or hypo-methylation genes. We found that FPKMs of both hyper- and hypo-DNA methylation genes in h1 were higher than those in Col (Fig. 5B), in contrast to expectation that the FPKM of hyper-methylation genes decreases and FPKM of hypo-methylation genes increases.
We found that only a small portion of DEGs (233/1974, about 11.29%) localizes in the differential DNA methylation regions (Fig. 5C). Among these 233 genes, 81 genes are RNA-up-regulated and DNA methylation-down-regulated, and 32 genes are RNA-down-regulated and DNA methylation-up-regulated as indicated in Venn diagram which includes RNA-seq data of 1974 DEGs and WGBS data of 4139 differentially methylated genes (Fig. 5C; Supplemental Table S6). Given that DNA methylation can also serve as an active marker in some circumstances (Harris et al. 2018), we then clarified the overall relationship between transcription and DNA methylation by bootstrapping randomized analysis by randomly selecting 5000 group of equal number (1974) of non-regulated genes to determine the percentage of those genes overlapped with DNA methylation-related genes. The result showed that the percentage of DEGs localized in the percentage of randomly selected genes (Fig. 5D), indicating that DNA methylation is no longer a regulatory factor of gene transcription upon H1 depletion.
DNA methylation is uncoupled from the chromatin compartmentalization in h1 mutant
DNA methylation was known to correlate with chromatin compartmentalization, and used to reconstruct compartments A/B in Hi-C analysis (Fortin and Hansen 2015). To gain insight into the correlation between DNA methylation and chromatin structure, we analyzed DNA methylation in the chromatin regions with converted compartments. We found that only 20% differentially methylated genes (830/4139, about 20%) overlapped with chromatin compartment A/B conversions (Fig. 6A). In addition, we found the Cs, CG, CHG, and CHH methylation levels of both A to B and B to A chromatin compartment transition-related genes in h1 are lower than those in Col (Supplemental Fig. S16A-D), in contrast to the general expectation that the methylation level in A to B regions increases, while that in B to A regions decreases.
We then analyzed the DNA methylation levels of genes intersected between DNA methylation and A/B switches, and found that the Cs, CG, CHG, and CHH methylation levels of both compartments A to B or B to A switch-related genes in h1 are lower than those in Col (Fig. 6B-E). We further analyzed the promoter regions and TSSs of genes intersected between DNA methylation and compartment A/B switches (Supplemental Table S7). The results indicated that the Cs, CG, CHG, and CHH methylation levels of both compartments A to B or B to A switch-related genes in h1 are lower than those in Col (Supplemental Fig. S16E-L). Together, we concluded that there is no correlation between DNA methylation and chromatin compartment A/B conversions in h1 mutant.
The switches of compartment A/B are largely uncorrelated with gene transcription
To reveal the role of chromatin folding in regulating gene expression upon H1 depletion, we analyzed the correlation between chromatin compartment switches and transcription. We found that most of DEGs (1568/1974, about 79.4%) show no overlap with chromatin compartment switches between A and B (Fig. 6F) with only a small portion of DEGs (406/1974, about 20.6%) overlapped with chromatin compartment A/B switches (Fig. 6F). In addition, although chromatin compartment switches account for 41.4% (18.5%+22.9%, Fig. 3C) of the Arabidopsis genome, only 7.7% (1974/25498) (Arabidopsis Genome Initiative 2000) of genes are DEGs in h1 mutant, supporting that most of the A/B switches do not affect transcription.
To further evaluate the relevance between A/B switches and transcription, we analyzed the normalized RNA-seq FPKM of genes in compartment A/B switches, and found that genes in compartments transited from B to A show only slightly higher expression levels than those in compartments transited from A to B in h1 mutant (Fig. 6G). We also noticed that most of DEGs (1568 genes) show no overlap with A/B switches (Fig. 6H). We further found that only 10.9 % (139/1268) up-regulated genes localized in compartment B to A switches (Fig. 6I), and 10.3% (73/706) down-regulated genes localized in compartments A to B switches (Fig. 6K). To know the confidence of these results, we performed bootstrapping randomized analysis. We randomly selected 5000 group of equal number (139) of up-regulated genes to determine the percentage of those genes overlapped with compartment A/B switches-related genes. The results showed that the top 5 percentile of randomly selected genes (2.6%) are only slightly lower than the percentage of up-regulated genes localized in compartment B to A switches (139/2690≈5.2%) (Fig. 6J). In addition, we randomly selected 5000 group of equal number (73) of down-regulated genes to determine the percentage of those genes overlapped with compartment A/B switches-related genes, and found that the top 5 percentile of randomly selected genes (1.2%) are only slightly lower than the percentage of up-regulated genes localized in compartments A to B switches (73/25220 ≈2.9%) (Fig. 6L). We concluded that chromatin compartment switches are predominantly not related to the gene regulation in h1 knockout mutant.
Discussion
The linker histone H1 binds and locks the adjacent nucleosomes to maintain chromatin structure (Fan et al. 2005). In the study, we found that unlocking in and out ends of nucleosome DNA by H1 complete depletion dampens the chromatin interactions, alters nucleosome density and distribution, changes the level and distribution of DNA methylation and affects the expressions of a subset of genes. Notably, loss of H1 disrupts the connections among chromatin interactions, DNA methylation and transcription, in which the H1 depletion-induced changes of nucleosome density and distribution might play an important role.
The role of linker histone H1 in regulating plant growth and development
Arabidopsis, which harbors only three genes encoding linker histone H1, is a good model organism to study the functions of H1. The partial mutants were reported, such as RNAi triple knockdown mutant (Wierzbicki and Jerzmanowski 2005), or T-DNA insertion triple knockdown mutant h1.1-1 h1.2-1 h1.3-1 with subtle phenotypes at developmental transitions (Rea et al. 2012; Rutowicz et al. 2019). In this study, we generated knockout mutants of histone H1s by T-DNA insertions and CRISPR/Cas9-based gene editing, which show pleiotropic phenotypes and are fertile. RNAi lines of H1 showed pleiotropic developmental abnormalities including changed size, serrated, small or elongated leaves, and reduced apical dominance, and plants with aberrant phenotypes have a considerably greater reduction of H1 expression than plants with no visible changes (Wierzbicki and Jerzmanowski 2005). We observed the h1 null mutant phenotypes of small plant size, short primary roots, serrated, small and elongated leaves, indicating the roles of H1 in plant growth and development by regulating a subset of genes.
There are 11 H1-coding genes in human and mouse, including replication dependent H1.1-H1.5, non-replication dependent H1.0 and H1.X, sperm specific H1t, H1T2 and H1LS1, and egg specific H1oo (Godde and Ura 2009; Kowalski and Palyga 2012). Knockouts of H1.3, H1.4 and H1.5 in mouse resulted in the death of embryos with altered distribution of nucleosomes and without change of the nuclear size (Fan et al. 2003). However, Tetrahymena strains that lacked either macronuclear or micronuclear histone H1 proteins or both were fully viable, showing similar growth to the wild type strains, implicating that H1 is more important for higher organisms. The increasing number of H1 genes during the evolution of higher organisms might be due to the needs for precise and specific regulation of gene expressions or responses to environmental stimuli as the tissue-specific expression of H1 was observed in mammalian and the expressions of response genes were induced by H1 depletion as shown in this study. H1 can also affect the phenotypes of organisms by modulating the post-translational modifications of histones (Ausio 1992; Herrera et al. 2000).
The role of H1 in functional connections among chromatin interactions, DNA methylation and transcription
At the cellular level, H1 depletion results in the increased nuclear size and decondensation of heterochromatic chromocenters. Hi-C analysis showed that the complete depletion of H1 in Arabidopsis dampens the chromatin interactions, supporting the role of linker H1 in the high-order chromatin structure in addition to its role in facilitating the folding of chromatin into 30 nm fiber. Moreover, significant alterations of nucleosome density and distribution were induced along all chromosomes. The nucleosome repeat length of H1-rich genes was observed to decrease substantially in Arabidopsis h1 knockdown mutants (Choi et al. 2020). The increase of nucleosome density and decrease of nucleosome repeat length might prevent chromatin DNA from getting closer to each other, leading to increased nuclear size and decreased chromatin interactions which might affect the nucleating of heterochromatin in chromocenters.
Further, we found the extensive switches between chromatin compartment A and B upon H1 depletion, indicating the important function of H1 in defining the chromatin domain identity. A/B compartments were known to be related to DNA methylation (Fortin and Hansen 2015). However, our results showed that both the changes of DNA methylation and expressions of most DEGs are uncorrelated with the chromatin compartment switches in h1 mutant, indicating that the linker histone also serve as a linker between chromatin compartment identity and DNA methylation directly or indirectly. Our results from the Arabidopsis H1 knockout line are different from those in mammalian or human H1 knockdown cells. In embryonic stem cells with about 50 % depletion of H1, the 3D genome showed no significant change and the alterations in TAD configuration coincide with epigenetic landscape changes but not with transcriptional output changes (Geeven et al. 2015). In addition, most A/B genomic compartments and TADs are unchanged upon deletion of H1C, H1D and H1E in mammals (Willcockson et al. 2020). In h1c/h1d/h1e triple and h1c/h1e double mutants, the chromatin architectural and epigenetic changes underlie the transcriptional alterations (Willcockson et al. 2020; Yusufova et al. 2020). How H1 maintains the relationship between the 3D structure of chromatin and DNA methylation is worthy of further study.
Interestingly, the compartment switches from A to B are more than B to A on chromosome 1 (Chr1), Chr2, Chr3 and Chr5, while B to A are more than A to B only on Chr4 (Fig. 3D), which might be related to the specific chromatin status of Chr4 with nucleolar associated chromatin domain (NAD) and nucleolar organizer region 4 (NOR4) (Pontvianne et al. 2016; Rabanal et al. 2017).
The correlation between the nucleosome density and DNA methylation
In eukaryotes, DNA methylation plays important roles in chromatin structure (Zhang et al. 2017). In ascomycete fungi, loss of H1 leads to genome-wide hypermethylation (Barra et al. 2000; Seymour et al. 2016). In mouse, the mutation of H1 reduces the DNA methylation levels in specific loci (Yang et al. 2013). In Arabidopsis, knockdown of H1 increases DNA methylation levels of heterochromatic elements and decreases DNA methylation levels of euchromatic TEs in all sequence contexts (Choi et al. 2020; Rutowicz et al. 2015; Zemach et al. 2013). In animals, CpG methylation induces tight wrapping of DNA around the histone core accompanied by a topologic change, and the changes in physical properties of nucleosomes induced by CpG methylation may contribute to the formation of repressive chromatin architectures (Lee and Lee 2012). The non-CG methylations (CHG and CHH) in plant are mainly distributed in heterochromatin region (Du et al. 2015). Our results revealed that the levels of CHG and CHH DNA methylations increase in centromeric regions and decrease in chromosome arms, while the nucleosome density dramatically decreases in centromeres and increases in chromosome arms in h1 mutant. One possible reason might be that the more nucleosome facilities stronger chromatin condensation, which prevents DNA methyltransferases or DNA demethylases to access to target sites as the transcription levels of these DNA methyltransferase or demethylase genes do not change obviously. The study of DDM1 supported this idea, showing DDM1 remove H1 from chromatin in order to facilitate the access of DNA methylation factors (Lyons and Zilberman 2017). Other studies also pointed out that nucleosomes are substantial obstacles to DNA methylation (Baubec et al. 2015; Huff and Zilberman 2014). However, the precise mechanism of H1 involved in DNA methylation needs to be intensely investigated, and the functions of HMGA proteins, which compete with histone H1 to bind to linker DNA (Catez and Hock 2010; Charbonnel et al. 2014), also need to be examined upon H1 complete depletion.
The role of H1-defined nucleosome distribution in the gene expression
Only a subset of genes are specifically regulated upon depletion of H1, however, the underlying mechanism about the specific gene regulation through H1 is not clear. Upon unlocking in and out ends of nucleosome DNA by H1 complete depletion, the nucleosome density in promoters elevates, but not in gene bodies (Supplemental Fig. S12G, H), and the phased distribution pattern of nucleosome in gene bodies of regulated genes is lost, but not in gene bodies of unregulated genes (Fig. 5D, F), indicating the role of the specific nucleosome deposition defined by H1 in the regulation of gene expression. Alternatively, transcription may also affect the differential nucleosome distribution between promoter and gene bodies. It was reported that H1 inhibits RNA polymerases from binding to chromatin (Krishnakumar et al. 1995), and regulates neuronal activation through modulating immediate early gene (IEG) expression, in which H1 is replaced by PARP on IEG promoters after neuronal stimulation (Azad et al. 2018). Therefore, H1 may also modulate the specific nucleosome density to regulate accessibility of transcription machinery to chromatin and affect gene expression.
This study unveiled for the first time the effect of H1 on the relationships among the 3D chromatin structure, DNA methylation and gene expression. In animal cells, no obvious change in the three-dimensional chromatin structure was observed upon partial depletion of linker histones (Geeven et al. 2015; Willcockson et al. 2020; Yusufova et al. 2020). In Arabidopsis, we observed enlarged nuclei and obvious decondensation of heterochromatic chromocenters, which correlates with a more homogenous pattern of chromatin interaction heatmap and dramatic changes of chromatin compartments upon H1 complete depletion. Our results shed light on the roles of linker histones in the maintenance of proper genome folding and coordinating chromatin compartmentalization, DNA methylation and gene expression.
Methods
Plant materials and growth conditions
Wild type Arabidopsis thaliana (Col-0 ecotype) and mutant plants were grown under 16 h light/8 h dark at 22 ℃. T-DNA insertion lines of h1.1-1 (SALK_128430C), h1.2-1 (SALK_002142) and h1.3-1 (SALK_025209) were obtained from Arabidopsis Biological Resource Center and confirmed by PCR-based genotyping. Primers used were listed in Supplemental Table S8.
The knockout mutants of h1-1 and h1-2 through CRISPR/Cas9 -mediated editing were generated as previously reported (Feng et al. 2013; Yan et al. 2015). The sgRNAs were listed in Supplemental Table S8.
Constructs and transient expression
The cDNAs of H1.1, H1.2, and H1.3 were amplified by PCR from Col cDNAs using primers listed in Supplemental Table S8, digested by EcoRI/SalI, and subcloned into EcoRI/SalI-treated vectors of pCambia1300-35S-N1-YFP and pCambia1300-35S-N1-mCherry (Fang and Spector 2007). The constructs were confirmed by sequencing and introduced into Agrobacterium tumefaciens (GV3101) by electroporation.
Arabidopsis plants were transformed by the floral dip method (Clough and Bent 1998). Transient expression and colocalization analysis were performed as described (Fang and Spector 2010).
Western blot
Total proteins were extracted from 10-day-old seedlings. 100 mg materials were collected and ground in liquid nitrogen. The powders were suspended in 3×SDS buffer (sample: buffer 1:3), then boiled at 100 ℃ for 10 min and centrifuged for 2 min at 12,000 rpm, the supernatants were separated on 15 % SDS-PAGE. Antibodies to H3 (Sigma H9289; 1:3000), H4 (Active Motif 61300; 1:3000), and H1 (Agrisera AS111801; 1:3000) were used.
DAPI staining
Leaves from 4-week-old plants were used. After washes in PBS, nuclei were counterstained in Fluoromount mounting media with DAPI (4′, 6-diamidino-2-phenylindole, YEASEN 36308ES11). Images of nuclei were acquired as described (Shi et al. 2011).
Hi-C library preparation
Hi-C experiments were performed essentially as described (Grob et al. 2014) with some modifications. Two biological replicates of Col we reported previously (Zhang et al. 2019) (GSE114950) and two biological replicates of h1 null mutant were sampled in parallel with Col under same growth condition. Briefly, 2.5 g aerial parts of the seedlings were fixed and ground into powder in liquid nitrogen. 600 U HindIII restriction enzyme were used to digest the extracted nuclei by incubating overnight at 37℃, then the digested chromatin was blunt-ended with 1μl 10mM dATP, dTTP, dGTP and 25μl 0.4mM biotin-14-dCTP and 100 U Klenow fragment for 45 min at 37℃. The ligation reaction was performed in 10 time volume of ligation buffer under constant shaking with 745μl 10× ligation buffer, 10% Triton X-100, 80μl 10 mg/ml BSA and ATP, 100 Weiss U T4 DNA ligase, at 16℃ for 6 h. The nuclei were then reverse-crosslinked with proteinase K at 65℃overnight. Subsequently, the extracted chromatin was fragmented into a mean size of 300 bp using a sonicator (Covaris S220). Hi-C libraries were constructed with NEB Next Multiplex Oligos kit and KAPA Hyper Prep Kit. The final library was subject to sequencing on an Illumina HiSeq 2000 instrument with 2 × 150-bp reads.
Hi-C sequencing data processing
Hic-pro (Servant et al. 2015) and Bowtie2 (Langmead and Salzberg 2012) were used for Hi-C read mapping. The clean Hi-C reads of Col and h1 mutant were aligned to Arabidopsis reference genome (TAIR10) after removing the adapter. Following with HiC-Pro and Juicer software (Durand et al. 2016), valid pairs of Col and h1 mutant were used to create interaction matrixes with bin size 50 kb for further analysis. The interaction matrixes were normalized with KR method from Juicer (Durand et al. 2016). The reproducibility of two biological replicates was tested with Pearson correlation coefficient from the interaction matrixes (Lin et al. 2018). After excluding the pericentromeres as reported (Grob et al. 2014), the first principal component was used to identify compartments A and B with Juicer.
Calculation of chromatin interaction and interaction decay exponents
The normalized interaction matrix from Col was divided by the normalized interaction matrix from h1 mutant, with all the zeros in matrixes replaced with 1% quintiles from the non-zero elements in each matrix, which were used to analyze the difference of interaction matrixes between Col and h1 mutant. We used Log2 transformation and median normalization to standardize the difference matrix. Interaction decay exponents (IDEs) were calculated (Grob et al. 2014) for chromosomes, pericentromeres and telomeres to study the interaction frequency changes dependent on the genome distance.
Bootstrapping analysis
In the bootstrapping strategies (Zhang et al. 2019; Buonaccorsi et al. 2018), we randomly selected 5000 groups (n = 5000 times) of the relevant genes of equal number, which were subjected to the same analysis to determine the percentage of those groups fallen in differential interaction bins or differential methylation bins. The percentile of the test sample lie above the top 5 percentile of the control distribution was considered confidently.
Micrococcal nuclease treatment assay
MNase assay was performed using protocol described (Zhang et al. 2018) with slight modification. For each MNase assay, 2 g of 10-day-old seedlings were used for nuclear extraction. The materials were ground in liquid nitrogen and shaked in lysis buffer containing 50 mM HEPES (pH 7.5), 10 % glycerol, 1 % Triton X-100, 1mM EDTA (pH 7.5), 150 mM NaCl, 5 mM β-mercaptoethanol and protease inhibitor cocktail (Roche) for 1 h at 4℃. The lysis mixture was filtered through a 40 μm cell strainer (BD) into fresh 50 ml tubes, nuclei were collected by centrifugation at 4℃ for 20 min at 4000 g. Nuclei were then washed twice with buffer A containing 25 mM Tris–HCl (pH 7.5), 0.44 M sucrose, 10 mM MgCl2, 0.1% Triton-X and 10 mM β-mercaptoethanol, and washed once with buffer B containing 20 mM Tris–HCl (pH 7.5), 352 mM sucrose, 8 mM MgCl2, 0.08% Triton-X, 8 mM β-mercaptoethanol and 20% glycerol, then resuspended in 100 μl of buffer B and flash frozen in liquid nitrogen. The nuclei obtained were divided into four portions for MNase treatments by incubating with 0.5 U, 1 U, 4 U and 8 U of MNase (TaKaRa 2910A) in MNase digestion buffer containing 20 mM Tris–HCl (pH 8.0), 5 mM NaCl and 2.5 mM CaCl2 for 20 min at 37℃. Reactions were stopped by addition of 500 mM EDTA. Digested DNA was purified and separated on 2 % agarose gel. For MNase-seq, all the nuclei were treatment with 4 U of MNase and mononucleosome-sized fragments were gel purified for sequencing library generation.
MNase-seq analysis
Approximately 1 μg of purified mononucleosome-sized DNA fragment was used for Illumina library generation per manufacturer’s instruction. Library construction and sequencing were performed by Genergy Biotechnology Co. Ltd. (Shanghai, China). Sequencing was carried out as single-end 50 bp reads on Illumina HiSeq-2000. Data analysis was carried out as previously described (Li et al. 2014). Briefly, the quality of raw reads was examined by FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). All clean reads were mapped to the TAIR 10 genome with the BOWTIE (http://bowtie.cbcb.umd.edu). Nucleosome positions were identified using the TEMPLATE FILTER software (http://compbio.cs.huji.ac.il/NucPosition/TemplateFiltering/Home.html).
RNA-seq analysis
Total RNAs were extracted from 10-day-old seedlings using the RNeasy plant mini kit (Qiagen). Yield and RNA integrity were detected by using an Agilent 2100 Bioanalyzer, and RNA purity was determined by using a Nanodrop ND-1000 spectrophotometer. cDNA library construction and sequencing were performed by Beijing Genomics Institute (BGI) using BGISEQ-500 platform for 50 bp single-end sequencing as previously described (Huang et al. 2018). At least 20 M clean reads of sequencing depth were obtained for each sample. Three independent biological replicates were performed. The raw reads were trimmed and quality controlled by Trimmomatic with default parameters (http://www.usadellab.org/cms/uploads/supplementary/Trimmomatic). Then clean reads were separately aligned to Arabidopsis thaliana genome from TAIR 10 with orientation mode using tophat software (http://tophat.cbcb.umd.edu/). The level for each transcript was calculated using the fragments per kilobase of exon per million mapped reads (FPKM) method. Cuffdiff (http://cufflinks.cbcb.umd.edu/) was used for differential expression analysis. GO functional enrichment analysis was carried out by Goatools (https://github.com/tanghaibao/Goatools). The differential expression analysis was run using the classical normalization method DESeq2 R package (Love Huber and Anders 2014) with a 0.05 p-value, 0.05 false discovery rate, and cutoff of 1 log-fold change. The hypergeometric test was performed as previously described (Wollmann et al. 2017).
DNA methylation analysis
For whole genome bisulfite sequencing (WGBS), genomic DNA (gDNA) was extracted from 10-day-old seedlings with the DNeasy plant mini kit (Qiagen) per manufacturer’s introduction. Library construction and sequencing were performed by Beijing Genomics Institute (BGI) using Illumina HiSeq-2000 for 100 bp paired-end sequencing. The raw paired end reads were trimmed and quality controlled by SeqPrep (https://github.com/jstjohn/SeqPrep) and Sickle (https://github.com/najoshi/sickle) with default parameters. All clean reads were mapped to the TAIR 10 genome with the BSMAP aligner (http://code.google.com/p/bsmap/) allowing up to 2 mismatches. Uniquely mapped reads were used to determine the cytosine methylation levels as previously stated (Lister, O’Malley, Tonti-Filippini, Gregory, Berry, Millar and Ecker 2008).
Data access
The Hi-C, WGBS and RNA-seq datasets have been submitted to NCBI (PRJNA680865), and MNase-seq datasets have been submitted to NCBI (PRJNA695028).
Competing interest statement
The authors declare no competing interests.
Author contributions
Z.S. and Y.F. designed and performed experiments, analyzed data, and wrote the manuscript. M.L. generated h1 mutants, acquired DNA methylation and RNA-seq data. Y.W. helped data analysis. H. Z. performed Hi-C experiments.
Acknowledgments
This work was supported by National Natural Science Foundation of China (31871230 to Y.F. and 31971334 to Z.S.).