Normal Patterns of Histone H3K27 Methylation Require the Histone Variant H2A.Z in Neurospora crassa

Abstract

Neurospora crassa contains a minimal Polycomb repression system, which provides rich opportunities to explore Polycomb-mediated repression across eukaryotes and enables genetic studies that can be difficult in plant and animal systems. Polycomb Repressive Complex 2 is a multi-subunit complex that deposits mono-, di-, and trimethyl groups on lysine 27 of histone H3, and trimethyl H3K27 is a molecular marker of transcriptionally repressed facultative heterochromatin. In mouse embryonic stem cells and multiple plant species, H2A.Z has been found to be colocalized with H3K27 methylation. H2A.Z is required for normal H3K27 methylation in these experimental systems, though the regulatory mechanisms are not well understood. We report here that Neurospora crassa mutants lacking H2A.Z or SWR-1, the ATP-dependent histone variant exchanger, exhibit a striking reduction in levels of H3K27 methylation. RNA-sequencing revealed downregulation of eed, encoding a subunit of PRC2, in an hH2Az mutant compared to wild type, and overexpression of EED in a ΔhH2Az;Δeed background restored most H3K27 methylation. Reduced eed expression leads to region-specific losses of H3K27 methylation, suggesting that differential dependence on EED concentration is critical for normal H3K27 methylation at certain regions in the genome.

IN eukaryotes, DNA-dependent processes in the nucleus are regulated by chromatin-based mechanisms (Luger 2003). One heavily studied group of proteins that are particularly important for maintaining stable gene repression are the polycomb group (PcG) proteins. In plants and animal cells, PcG proteins assemble into polycomb repressive complexes 1 and 2 (PRC1 and PRC2), which play key roles in repression of developmental genes (as reviewed in Müller (1995), Hennig and Derkacheva (2009), Simon and Kingston (2009), Schuettengruber et al. (2017), Kuroda et al. (2020)). PRC2 is a multi-subunit complex that deposits mono-, di-, and trimethyl groups on lysine 27 of histone H3, and trimethyl H3K27 is a molecular marker of transcriptionally repressed facultative heterochromatin (Cao et al. 2002; Czermin et al. 2002; Kuzmichev et al. 2002; Müller et al. 2002). PcG proteins are absent from the model yeasts, Saccharomyces cerevisiae and Schizosaccharomyces pombe, but core PRC2 components have been identified and characterized in several fungi, including Neurospora crassa, Fusarium graminearum, Cryptococcus neoformans, Epichloë festucae, and Fusarium fujikuroi (Veerappan et al. 2008; Aramayo and Selker 2013; Connolly et al. 2013; Jamieson et al. 2013; Chujo and Scott 2014; Schotanus et al. 2015; Studt et al. 2016). In these fungi, PRC2 is required for repression of key fungal genes, suggesting that this enzyme complex is functionally conserved between fungi, plants, and animals (Connolly et al. 2013; Jamieson et al. 2013; Dumesic et al. 2015).

In N. crassa, the catalytic subunit of PRC2 is SET-7, a protein with homology to EZH1/EZH2 in humans and curly leaf (CLF), medea (MEA), or swinger (SWN) in Arabidopsis (Jones and Gelbart 1993; Abel et al. 1996; Chen et al. 1996; Goodrich et al. 1997; Grossniklaus et al. 1998; Czermin et al. 2002; Kuzmichev et al. 2002; Chanvivattana et al. 2004; Schwartz and Pirrotta 2007). Neurospora EED is essential for catalysis and is a homolog of mammalian enhanced ectoderm development (EED), Drosophila extra sex combs (Esc) and Arabidopsis fertilization independent endosperm (FIE) (Schumacher et al. 1998; Ng et al. 2000; Chanvivattana et al. 2004). SUZ-12 is the third essential component of PRC2 in Neurospora and is named SUZ12 in humans, su(z)12 in Drosophila and embryonic flower 2 (EMF2), vernalization 2 (VER2), or fertilization-independent seed 2 (FIS2) in Arabidopsis (Birve et al. 2001; Chanvivattana et al. 2004). N. crassa CAC-3/NPF is an accessory subunit homologous to mammalian retinoblastoma binding protein 46/48 (RBAP46/68) in humans, and multicopy suppressor of IRA1-5 (MSI1-5) in Arabidopsis (Huang et al. 1991; Qian et al. 1993; Derkacheva et al. 2013). In contrast to PRC2, PRC1 components appear to be absent from the fungal kingdom (Jamieson et al. 2013; Lewis 2017).

The presence of a minimal polycomb repressive system in well studied fungi such as N. crassa provides an opportunity to explore the diversity of polycomb-mediated repression across eukaryotes and enables genetic studies that can be difficult in plant and animal systems. Indeed, genetic studies have provided insights into PRC2 control in Neurospora. Deletion of cac-3/npf causes region-specific losses of H3K27me3 at telomere-proximal domains, and telomere repeat sequences are sufficient to nucleate a new domain of H3K27me3-enriched chromatin (Jamieson et al. 2013, 2018). In constitutive heterochromatin domains, heterochromatin protein-1 (HP1) prevents accumulation of H3K27me3 (Basenko et al. 2015; Jamieson et al. 2016). Thus, regulation of H3K27 methylation occurs at multiple levels. Despite recent advances, the mechanisms that regulate PRC2 in fungal systems and eukaryotes in general is poorly understood.

In addition to the core histones (H2A, H2B, H3, and H4), eukaryotes also encode nonallelic histone variants. One of the most conserved and extensively studied histone variants is H2A.Z, which is enriched proximal to transcription start sites (TSS) and in vertebrate enhancers (Guillemette et al. 2005; Barski et al. 2007; Creyghton et al. 2008; Bargaje et al. 2012; Weber et al. 2014; Latorre et al. 2015; Dai et al. 2017; Gómez-Zambrano et al. 2019). Functional studies of H2A.Z have linked presence of this variant in nucleosomes to gene activation, gene repression, maintaining chromatin accessibility, and a multitude of other functions (Adam et al. 2001; Meneghini et al. 2003; Rangasamy et al. 2004; Bruce et al. 2005; Guillemette et al. 2005; Dhillon et al. 2006; Xu et al. 2012; Hu et al. 2013; Neves et al. 2017). Notably, H2A.Z has been implicated in the direct regulation of H3K27 methylation in mouse embryonic stem cells (mESCs) and in plants (Surface et al. 2016; Zhang et al. 2017; Carter et al. 2018). In mESCs, there is a strong correlation between the activity of PRC2, enrichment of H3K27me3, and the presence of H2A.Z (Wang et al. 2018). Colocalization of SUZ12, a subunit of PRC2, and H2A.Z has been found in mESCs at developmentally important genes, such as HOX clusters (Creyghton et al. 2008). In addition, H2A.Z is differentially modified at its N- and C- terminal tails at bivalent domains that are “poised” for activation or repression upon differentiation (Ku et al. 2012; Surface et al. 2016). N-terminal acetylation (acH2A.Z) or C-terminal ubiquitylation (H2A.Zub) repress or stimulate the action of PRC2 through interactions with the transcriptional activator BRD2 or the PcG protein complex PRC1 (Surface et al. 2016). It is important to note that functional studies of H2A.Z are challenging because this histone variant is essential for viability in most organisms, including Drosophila, Tetrahymena, mouse, and Xenopus (van Daal and Elgin 1992; Iouzalen et al. 1996; Liu et al. 1996; Clarkson et al. 1999; Faast et al. 2001; Ridgway et al. 2004).

In Arabidopsis thaliana, a genetic interaction between PICKLE (PKL), a chromatin remodeler that promotes H3K27me3, and PIE-1 (homolog to SWR-1), the remodeler which deposits H2A.Z, was recently reported (Carter et al. 2018). PKL has been found by ChIP-seq at loci enriched for H3K27me3, and is proposed to determine levels of H3K27me3 at repressed genes in Arabidopsis (Zhang et al. 2012). In rice callus and seedlings, H2A.Z is found at the 5′ and 3′ ends of genes that are highly expressed. In repressed genes, H2A.Z is found along the gene body, and this pattern closely mimics the presence of H3K27me3 (Zhang et al. 2017). This is a notable difference between plants and other eukaryotes.

We investigated the relationship between H2A.Z and PRC2 in the filamentous ascomycete Neurospora crassa, and report that H2A.Z is required for normal enrichment of H3K27me2/3 across the genome. Our findings show that loss of H2A.Z leads to region-specific losses of H3K27me2/3 in N. crassa. Expression levels of eed, encoding a PRC2 subunit, are reduced in the absence of H2A.Z and ectopic expression of eed can restore H3K27me2/3 in an H2A.Z-deficient strain. Together, these data suggest that H2A.Z regulates facultative heterochromatin through transcriptional regulation of the PRC2 component EED and points to differential requirements for EED at discrete PRC2-target domains.

Materials and Methods

Strains and growth media

Strains used in this study are listed in Supplemental Material, Table S1. Strains were grown at 32° in Vogel’s minimal medium (VMM) with 1.5% sucrose or glucose for DNA-based protocols and RNA-based protocols, respectively (Davis and de Serres 1970). Liquid cultures were shaken at 180 rpm. Crosses were performed on synthetic crossing (SC) medium in the dark at room temperature (Davis and de Serres 1970). Ascospores were collected 14 days after fertilization. To isolate cross progeny, spores were spread on solid VMM plates containing FGS (1X Vogel’s salts, 2% sorbose, 0.1% glucose, 0.1% fructose, and 1.5% agar) and incubated at 65° for 1 hr as previously described (Davis and de Serres 1970), after which spores were picked using a sterile inoculating needle and transferred to agar slants with appropriate medium (typically VMM). To test for sensitivity to DNA damaging agents, 5 µl of a conidial suspension was spotted on VMM containing FGS plates containing concentrations of methyl methanesulfonate (cat. # 129925-5g; Sigma–Aldrich) between 0.010% and 0.030% (w/v).

To construct the N-terminal FLAG-tagged eed allele, we amplified the eed region with primers, MK #51: GGCGGAGGCGGCGCGATGCAAATTTGTCGGGACCG and MK #52: TTAATTAATGGCGCGTTACTTCCCCCACCGCTGAA (Table S5), from wild-type genomic DNA (FGSC 4200). The amplified fragment was cloned into the AscI site of pBM61::CCGp-N-3xFLAG (Honda and Selker 2009) by InFusion cloning (cat. # 639648; Takara). The new plasmid was then digested with DraI and transformed into a his-3;mus-52::bar strain. Primary transformants were selected on VMM plates, and then back-crossed to wild type to isolate homokaryons (his-3::Pccg-1-3xflag-eed). We next crossed the homokaryon (his-3::Pccg-1-3xflag-eed) to Δeed::hph (FGSC 14852) to obtain Δeed;his-3::Pccg-1-3xflag-eed. 3xFLAG-EED expression and deletion of eed deletion were confirmed by western blots probed with anti-FLAG antibody (cat. # F1804; Sigma Aldrich) and genotyped by PCR with primers, LL #155: TCGCCTCGCTCCAGTCAATGACC and LL #466: TGTGGGCGATTTGAGCGTGC, respectively. The Δeed;his-3::Pccg-1-3xflag-eed strain was then crossed to the ΔhH2Az::hph (FGSC 12088) strain to obtain ΔhH2Az;Δeed;his-3::Pccg-1-3xflag-eed. 3xFLAG-EED expression and deletion of eed were confirmed by western blots with anti-FLAG antibody (cat. # F1804; Sigma Aldrich) and genotyping with eed deletion primers (see above). Deletion of hH2Az was confirmed by PCR with primers AC #24: GAACAAGCCGATTGCTGTCC and AC #23: TGTATAGAACGCTGCCAAGGA.

For the H2AZ-GFP gene replacement construct, a 1-kb segment including the end of the hH2Az coding region was amplified by PCR with primers #1577: CGGAAAGGGCAAGTCGTCTG and #1578: CCTCCGCCTCCGCCTCCGCCGCCTCCGCCAGCCTCCTGAGCCTTGGCCT and a 500-bp segment of the 3′ flanking region was amplified with primers #1579: TGCTATACGAAGTTATGGATCCGAGCTCGCTGCACCGAAAAACTCGACG and #1580: GTGACGAGGGGAGATTGCTC. The cassettes containing the GFP segment and the hph gene were amplified using M13 forward and reverse primers from pGFP::hph::loxP (Honda and Selker 2009). The three fragments were mixed and then assembled by overlapping PCR with primers #1577 and #1580 above. The cassette was transformed into the Δmus-52 strain (FGSC 15968) by electroporation.

Transformation and complementation assays

Transformations were performed as previously described (Margolin et al. 1997). To carry out ectopic complementation of the ΔhH2Az::hph strain, two linear gene fragments were electroporated into the mutant strain. Specifically, the bar (confers Basta resistance) was amplified with primers LL #148 CCGTCGACAGAAGATGATATTGAAGGAGC and LL #149 AATTAACCCTCACTAAAGGGAACAAAAGC (Avalos et al. 1989), and the wild-type hH2Az gene fragment including its native promoter (genomic coordinate 1390154–1393398 of GCA_000182925.2 assembly accession) was amplified with primers AC #27 CCCAATCCTAGAATCCCGTCG and AC #21 TAAAAGAGCTGCTGTCGCACG, and fragments were co-integrated into the ΔhH2Az::hph strain, followed by selection of transformants on Basta-containing plates (VMM with 2% sorbose, 0.1% glucose, 0.1% fructose, 1.5% agar, and 200 μg/mL Basta). Transformants were transferred to agar slants and then screened by PCR, and Southern blots with the North2South Biotin Random Prime Labeling and Detection Kit (cat. #17175; ThermoFisher) and the wild-type hH2Az gene fragment generated from primers AC#27: CCCAATCCTAGAATCCCGTCG and AC#21: TAAAAGAGCTGCTGTCGCACG was used as a probe. Genomic DNA was extracted from wild type, ΔhH2Az, and the two complemented strains (ACt9-3, ACt12-1). A 500 ng aliquot of DNA from all strains was subjected to double restriction digests with EcoRI-HF (cat # R3101S; NEB) and EcoRV-HF (cat # R3195S; NEB) in CutSmart Buffer. Digests were incubated at 37° overnight and then subjected to heat inactivation for 20 min at 65°. Digests were run on a 0.7% agarose gel and transferred overnight to a nylon membrane which was then UV crosslinked. Hybridization and detection followed the manufacturer’s instructions.

Race tube assay

Race tubes were prepared with 15 ml of VMM plus 1.5% sucrose and 1.5% agar. Strains were grown on VMM plates with 1.5% sucrose and 1.5% agar for 16 hr before using a 6 mm cork borer to extract mycelial agar plugs from the edge of growing hyphae. This plug was used for inoculating each tube at one end. Strains were inoculated in triplicate. Measurements were taken at 9, 23, 47, and 60 hr to determine linear growth rates.

Protein extraction and western blotting

Strains were grown at 32° shaken in 18 × 150 mm glass test tubes at 180 rpm in 5 ml VMM with 1.5% sucrose. After 16 hr, tissue was harvested using filtration, washed once in phosphate buffered saline (PBS), and suspended in 1 ml of ice-cold protein extraction buffer [50 mM HEPES pH 7.5, 150 mM NaCl, 0.02% NP-40, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (PMSF; P7626; Sigma), one tablet Roche cOmplete mini EDTA-free Protease Inhibitor Cocktail (cat. # 11836170001; Roche)]. Tissue was subjected to sonication by Diagenode Bioruptor UCD-200 to deliver 22.5 30 sec pulses at 4°. After two rounds of centrifugation at 13,200 rpm for 10 min, supernatant was mixed with 2× Laemmli buffer and boiled for 5 min. Samples for FLAG and H2A.Z western blots were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene difluoride (PVDF) membranes in Tris-glycine transfer buffer (25 mM Tris, 200 mM glycine) or CAPS transfer buffer (10 mM, pH 11) containing 20% methanol at constant 100 V for 1 hr at 4°, respectively. Membranes were blocked with Tris-buffered saline (TBS; 10 mM Tris, pH 7.5, 150 mM NaCl) including 3% milk powder or phosphate-buffered saline including Tween 20 (PBS-T; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, KH₂PO₄, 0.01% Tween-20) including 30% milk powder for 1 hr, and incubated overnight with anti-FLAG antibody (cat. # F1804; Sigma Aldrich) in TBS plus 3% milk or H2A.Z antibody in PBS-T plus 3% milk. Detection was performed with horseradish peroxidase-conjugated secondary antibodies and SuperSignal West Femto chemiluminescent substrate (cat. # 34094; ThermoFisher).

Chromatin immunoprecipitation

To carry out chromatin immunoprecipitation (ChIP), conidia were inoculated in 5 ml of liquid VMM plus 1.5% sucrose and grown for 18 hr for wild type and other strains with typical growth rates. Slow growing ΔhH2Az::hph strains were grown for 24 hr to isolate cultures at a similar developmental stage. ChIP was performed as described previously (Sasaki et al. 2014; Seymour et al. 2016; Ferraro and Lewis 2018). In brief, mycelia were harvested using filtration and were washed once in PBS prior to cross-linking for 10 min in PBS containing 1% formaldehyde on a rotating platform at room temperature. After 10 min, the reaction was quenched using 125 mM glycine and placed back on the rotating platform for 5 min. Mycelia were harvested again using filtration, washed once with PBS, then resuspended in 600 µl of ChIP lysis buffer [50 mM HEPES, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, one tablet Roche cOmplete mini EDTA-free Protease Inhibitor Cocktail (cat. # 11836170001; Roche)] in 15 ml conical tubes. Chromatin was sheared by sonication after lysing cell walls with the QSONICA Misonix S-4000 ultrasonic processor (amplitude 10, 30 sec processing, 1 sec on, 1 sec off), using the Diagenode Bioruptor UCD-200 (Intensity level: Medium, three rounds of 15 min (30 sec on, 30 sec off) to deliver 22.5 30 sec pulses at 4°. Water temperature was kept at a constant 4° by using a Bio-Rad cooling module (cat. # 170-3654) with variable speed pump to circulate 4° water while processing samples. Lysates were centrifuged at 13,000 rpm in an Eppendorf 5415D microcentrifuge for 5 min at 4°. For ChIP reactions with antibodies against N. crassa H2A.Z, 1, 2.5, or 5 µl of antibody was used [antibody supplied by Qun He, China Agricultural University (Liu et al. 2017; Dong et al. 2018)]. For detection of H3K27 di- and trimethylation (H3K27me2me3; Active Motif 39535), and GFP-tagged H2A.Z (GFP; Rockland 600-301-215), 1 µl of the relevant antibody was used. Protein A/G beads (20 µl) (cat. # sc-2003; Santa Cruz) were added to each sample. Following overnight incubation, beads were washed twice with 1 ml lysis buffer without protease inhibitors, once with lysis buffer containing 500 mM NaCl, once with lysis buffer containing 50 mM LiCl, and finally with TE (10 mM Tris-HCl, 1 mM EDTA). Bound chromatin was eluted in TES (50 mM Tris pH 8.0, 10 mM EDTA, 10% SDS) at 65° for 10 min. Chromatin was de-crosslinked overnight at 65°. The DNA was treated with RNase A for 2 hr at 50°, then with proteinase K for 2 hr at 50°, and extracted using phenol-chloroform-isoamyl alcohol (25:24:1) followed by chloroform extraction. After final chloroform extraction, DNA was precipitated using ethanol precipitation with 2 vol of ethanol, 1/10 vol 3 M NaOAc, pH 5.2, and 0.025 mg/mL glycogen overnight at −20°. DNA pellets were washed with 70% ethanol and resuspended in TE buffer. Samples were then prepared for Illumina sequencing.

RNA extraction

Conidia were inoculated into 100 × 15 mm plates containing 25 ml of VMM + 1.5% glucose and grown for 36–48 hr to generate mycelial mats. Using a 9 mm cork borer, five to seven disks were cut out of the mycelial mat and transferred to 125 ml flasks with 50 ml of VMM + 1.5% glucose and allowed to grow for 12 hr at 29° in constant light while agitating at ∼90–100 rpm. Disks were harvested using filtration and flash-frozen with liquid nitrogen. Frozen tissue was transferred to 1.5 ml RNase-free tubes with 100 µl sterile RNase-free glass beads and vortexed to lyse tissue in phenol:chloroform (5:1) pH 4.5. Three sequential acid phenol:chloroform extractions were performed followed by ethanol precipitation using two volumes of ethanol and 1/10 vol of 3 M NaOAc pH 5.2, incubated overnight at −20°. Samples were centrifuged at 13,200 rpm in 4° for 30 min and pellets were then washed in RNase-free 70% ethanol, and resuspended in RNase-free water. Samples were quantified using the Invitrogen Qubit 2.0 fluorometer (cat. # Q32866) and RNA quality was checked on a denaturing agarose gel. After quality was verified 10 µg of RNA for each sample was subjected to Turbo DNase treatment (cat. # AM2238; Invitrogen) at 37° for 30 min and then another acid phenol:chloroform extraction was performed to inactivate enzyme and purify the RNA. Samples were subjected to another ethanol precipitation as described above, this time with the addition of 1 µl of RNase-free glycogen (5 mg/ml). Samples were centrifuged at 13,200 rpm in 4° for 30 min and the pellets were washed with RNase-free 70% ethanol, then resuspended in RNase-free water. Quality and quantity were again checked with denaturing gel and with the Invitrogen Qubit 2.0 fluorometer. Samples were then prepared for Illumina sequencing.

ChIP library preparation

Libraries were constructed as described (Sasaki et al. 2014; Seymour et al. 2016; Ferraro and Lewis 2018). In brief, the NEBNext Ultra II End Repair/dA-tailing Module (cat. # E7546S), NEBNext Ultra II Ligation Module (cat. # E7546) were used to clean and A-tail DNA after which Illumina adapters were ligated. The ligation products were amplified to generate dual-indexed libraries using NEBNext Ultra II Q5 Hot Start HiFi PCR Master Mix (cat. # M0543S). Size selection with magnetic beads was performed after the adapter ligation and PCR steps with Sera-Mag SpeedBeads (cat. # 65152105050250) suspended in a solution of 20 mM PEG 8000, 1 mM NaCl, 10 mM Tris-HCl, 1 mM EDTA (Rohland and Reich 2012).

RNA library preparation

Libraries were prepared according to the Illumina TruSeq mRNA stranded Library Kit (cat. # RS-122-2101). In brief, mRNA selection via polyA tails was performed using RNA purification beads and washed with bead washing buffer. Fragmentation and cleanup were performed enzymatically using the Fragment, Prime, Finish Mix and incubated at 94° for 8 min. First-strand synthesis using the SuperScript II RT enzyme and First Strand Synthesis Act D Mix was incubated as described, and second-strand synthesis used the Second Strand Marking Mix with resuspension buffer was incubated for 1 hr to generate cDNA. The final steps in the library preparation are the same as the above ChIP-seq library preparation with exception of two extra bead cleanup steps: one prior to A-tailing and adapter ligation, two after adapter ligation.

Libraries were pooled and sequenced on a NextSeq500 instrument at the Georgia Genomics and Bioinformatics Core to generate single or paired-end reads.

Data analysis

For ChIP-seq data, short reads (<20 bp) and adaptor sequences were removed using TrimGalore (version 0.4.4), cutadapt version 1.14 (Martin 2011), and Python 2.7.8, with fastqc command (version 0.11.3). Trimmed Illumina reads were aligned to the current N. crassa NC12 genome assembly available from NCBI (accession # GCA_000182925.2) using the BWA (version 0.7.15) (Li and Durbin 2009), mem algorithm, which randomly assign multi-mapped reads to a single location. Files were sorted and indexed using SAMtools (version 1.9) (Li et al. 2009). To plot the relative distribution of mapped reads, read counts were determined for each 50 bp window across the genome using DeepTools to generate bigwigs (version 3.3.1) (Ramirez et al. 2016) with the parameters –normalizeUsing CPM (counts per million), and data were displayed using the Integrated Genome Viewer (Thorvaldsdottir et al. 2013). The Hypergeometric Optimization of Motif EnRichment (HOMER) software package (version 4.8) (Heinz et al. 2010) was used to identify H3K27me3 peaks in wild type and ΔhH2Az against input using “findPeaks.pl” with the following parameters: -style histone. Bedtools (version 2.27.1) “intersect” (version 2.26.0) was used to determine the number of peaks that intersect with other peak or gff files, for determining PRC2-target genes the parameter of -f 0.70 was used. Heatmaps, Spearman correlation matrix (Figure S7) and line plots were constructed with DeepTools (version 3.3.1) (Ramirez et al. 2016).

For RNA-seq data, short reads (<20 bp) and adaptor sequences were removed using TrimGalore (version 0.4.4), cutadapt version 1.14 (Martin 2011), and Python 2.7.8, with fastqc command (version 0.11.3). Trimmed Illumina single-end reads were mapped to the current N. crassa NC12 genome assembly using the Hierarchical Indexing for Spliced Alignment of Transcripts 2 (HISAT2: version 2.1.0) (Kim et al. 2019) with parameters –RNA-strandness R then sorted and indexed using SAMtools (version 1.9) (Li et al. 2009). FeatureCounts from Subread (version 1.6.2) (Liao et al. 2013) was used to generate gene level counts for all RNA bam files. Raw counts were imported into R and differential gene expression analysis was conducted using Bioconductor: DeSeq2 (Love et al. 2014). Volcano plot and box plots were generated in R using DeSeq2 and ggplot2 (Wickham 2009).

Data deposition

The authors state that all data necessary for confirming the conclusions presented in the article are represented fully within the article. Raw sequence data associated with this paper are available through the NCBI GEO database (accession # GSE146611). Supplemental data have been uploaded to figshare. Supplemental material available at figshare: https://doi.org/10.25386/genetics.12486482.

Results

Normal patterns of H3K27me2/3 enrichment require the presence of H2A.Z or SWR-1

Normal H3K27me2/3 patterns in plants and in mESCs depend on the histone variant H2A.Z (Creyghton et al. 2008; Carter et al. 2018), but the underlying mechanism is ill-understood. To determine if H2A.Z also plays a role in polycomb group repression in N. crassa, we performed ChIP-seq to examine H3K27me2/3 enrichment in an H2A.Z deletion strain (ΔhH2Az::hph, hereafter ΔhH2Az) and compared this to wild type and Δset-7. Inspection of the data in the Integrative Genomics Viewer (IGV) genome browser (Thorvaldsdottir et al. 2013) revealed that the ΔhH2Az mutant displayed a significant reduction in H3K27me2/3 (Figure 1A). This reduction was apparent on all chromosomes (Figure S1) but H3K27me2/3 was not completely abolished, as observed in the Δset-7 strain, which lacks the catalytic subunit of PRC2 (Figure 1A). To quantify the change in H3K27me2/3 patterns, we called peaks of H3K27me2/3 enrichment using HOMER (version 4.8; Heinz et al. 2010). We identified 325 peaks of H3K27me2/3 in wild type, hereafter referred to as PRC2-target domains (Table S2). Consistent with previous studies, these peaks comprised ∼6% of the N. crassa genome (Jamieson et al. 2013; Basenko et al. 2015). These regions are typically larger than single genes, ranging in size from 500 bp to 108 kb, with an average size of 7.7 kb. We next plotted H3K27me2/3 levels across the 5′ end of all 325 domains for wild type and ΔhH2Az (Figure 1B). Inspection of heatmaps and the genome browser revealed that H3K27me2/3 levels were reduced in many, but not all, PRC2-target domains in ΔhH2Az. Using HOMER software to identify PRC2-target domains in ΔhH2Az revealed 239 peaks (Table S3). These were slightly smaller, with an average size of 5.5 kb, and comprised only 3% of the N. crassa genome. To determine if the peaks observed in the ΔhH2Az strain are in wild-type locations, we compared peaks only from assembled contigs. Using bedtools intersect, we found that all peaks in ΔhH2Az overlap with wild-type peaks, indicating that ΔhH2Az exhibits significant loss of H3K27me2/3 from normal domains, but does not gain H3K27me2/3 in new locations (Table S4).

Figure 1.

H2A.Z is required for normal patterns of H3K27 methylation. (A) Genome browser images illustrate H3K27me2/3 enrichment on N. crassa linkage group (“chromosome”) III for wild type, ΔhH2Az, and Δset-7. A segment of chromosome III is displayed at higher resolution to illustrate depletion of internal H3K27me2/3 domains. (B) H3K27me2/3 in the ΔhH2Az strain exhibits striking depletion of most H3K27me2/3 domains, with overall lower enrichment of this modification. Heatmaps display 325 PRC2-target domains (rows) ordered by wild type enrichment for wild type, ΔhH2Az, and Δset-7 strains centered on the 5′ end of each domain + or –1000 bp for a total window size of 2000 bp. (C) Genome browser images illustrate H3K27me2/3 enrichment on chromosome III for two ectopic complemented strains of ΔhH2Az+hH2Az^wt and the Δswr-1 strain. The segment of chromosome III is displayed at higher resolution to illustrate rescue by complementation and depletion of H3K27me2/3 in Δswr-1 background. (D) Heatmaps of H3K27me2/3 rescue in complemented strains [ΔhH2Az+hH2Az^wt (ACt9-3 and ACt12-1)] and depletion in the Δswr-1 strain. The heatmaps are ordered as in (B) and depict the domain boundary + or –1000 bp for a total window size of 2000 bp.

Since H2A.Z is required for maintaining genome stability in yeast and animals, our findings raised the possibility that a second site mutation could be responsible for the observed phenotype (Krogan et al. 2004; Rangasamy et al. 2004; Dhillon et al. 2006; Greaves et al. 2007). To confirm that loss of H3K27me2/3 was due to the absence of H2A.Z, we first backcrossed the original deletion strain (FGSC 12088) to wild type (Colot et al. 2006). Four independent ΔhH2Az progeny all displayed similar reduction in H3K27me2/3 levels (Figure S2). In addition, the backcrossed ΔhH2Az strain displayed slow and variable growth (Figure S3) and was hypersensitive to the DNA-damaging agent methyl methanesulfonate (MMS). This is consistent with previous studies that have demonstrated poor growth of ΔhH2Az in S. cerevisiae and in N. crassa (Jackson and Gorovsky 2000; Liu et al. 2017).

We next introduced a wild-type copy of the hH2Az gene with its native promoter into ΔhH2Az (Figure S4, A and B). This complemented defects in growth and MMS-sensitivity, and fully restored H3K27 methylation, suggesting that loss of H2A.Z was responsible for all observed phenotypes in the deletion mutant (Figure 1, C and D and Figure S3). Using HOMER software to identify PRC2-target domains in the complemented strains (n = 2) revealed an average of 449 peaks with an average size of 6.6 kb, comprising ∼7% of the N. crassa genome. To determine if the peaks observed in the complemented strains are in wild-type locations, we compared peaks only from assembled contigs. Using bedtools intersect, we found that most peaks in overlap with wild-type peaks. The six peaks that were reported as not intersecting were viewable in the genome browser in the same location as wild type, but of a slightly different amplitude, which affects peak calling and is likely due to ChIP efficiency from multiple different experiments. Because a specific chromatin remodeling complex, SWR1, exchanges H2A.Z for H2A in plants, yeast, and animals (Kobor et al. 2004; Mizuguchi et al. 2004; Deal et al. 2007; Wong et al. 2007; Luk et al. 2010), we next examined H3K27me2/3 in a deletion strain lacking the N. crassa homolog of the SWR1 ATPase (Δswr-1 also known as Δcrf1-1; Borkovich et al. 2004). The swr-1 mutant displayed a similar reduction in H3K27me2/3 (Figure 1, C and D). Together, these data demonstrate that H2A.Z is required for normal H3K27me2/3 in N. crassa.

Deletion of hH2Az results in region-specific loss of H3K27me2/3

Visual inspection of the ChIP-seq data revealed losses of H3K27me2/3 from PRC2-target domains located at internal (i.e., nonsubtelomeric regions >200 kb from the telomere repeats) chromosome sites, but not at telomere-proximal sites (i.e., <200 kb from the telomere repeats) (Figure 2A). To quantify this, we inspected ChIP-seq results for H3K27me2/3 for both classes and found retention of H3K27 methylation in telomere-proximal regions with progressive loss in domains farther from chromosome ends. Previously published work showed that a cac-3/npf deficient strain has H3K27me2/3 loss, which was observed primarily in the telomere-proximal regions (Jamieson et al. 2013); cac-3/npf encodes an accessory subunit of PRC2 in N. crassa homologous to the conserved PRC2 components Msl1-5, NURF55, Rpbp46/48, found in plants, Drosophila, and humans, respectively.

Figure 2.

Deletion of hH2Az results in region-specific loss of H3K27me2/3. (A) Genome browser images illustrate H3K27me2/3 enrichment on linkage group (“chromosome”) III for wild type, ΔhH2Az, Δcac-3/npf, and Δset-7. The two segments of chromosome III are displayed at higher resolution to visualize region-specific loss. Left panel displays the end of the chromosome to ∼300 kb (left panel) and from ∼400 to 600 kb (right panel). The telomere-proximal H3K27me2/3 regions are only moderately affected by the deletion of hH2Az, whereas internal domains show a more dramatic loss of H3K27me2/3. (B) Heatmaps of H3K27me2/3 enrichment for wild type, ΔhH2Az, Δcac-3/npf, and Δset-7 across PRC2-target domains organized by their proximity to the telomere. The top section is restricted to domains that are <200 kb away from the chromosome ends (“telomere-proximal domains”), plotted from largest to smallest (123 domains). The bottom of the heatmaps contain the domains that are >200 kb away from chromosome ends (“internal domains”), also plotted from largest to smallest (186 domains). Heatmaps are centered on the 5′ edge of PRC2-target domains + or −1000 bp for a total window size of 2000 bp. The ΔhH2Az strain retains most telomere-proximal H3K27me2/3, as opposed to the Δcac-3/npf strain where almost all H3K27me2/3 enrichment is lost from telomere-proximal regions.

To better visualize which regions of the genome in the Δcac-3/npf or ΔhH2Az strains lose enrichment of H3K27me2/3, we again divided all 325 H3K27me2/3 peaks in the wild-type strain into telomere-proximal sites (123 peaks, average size 8261 bp) (Figure 2B, top) and internal sites (186 peaks, average size 7509 bp) (Figure 2B, bottom). The loss was again most dramatic at the internal regions in the hH2Az deletion strain, where most PRC2-target domains showed significant reduction of H3K27me2/3 levels. Specifically, we found that telomere-proximal regions show normal levels of H3K27me2/3.

Previous work demonstrated that the placement of repetitive telomere repeat sequences (5′-TTAGGG-3′) in a euchromatic locus can induce de novo H3K27 methylation across large regions (Jamieson et al. 2018). Together, these data demonstrate that the absence of H2A.Z is more detrimental for the establishment and/or maintenance of internal domains of H3K27me2/3 in N. crassa.

Neurospora H2A.Z localizes to promoter regions but not to PRC2-target domains

We next asked if H2A.Z colocalizes with H3K27 methylation, as has been reported for plants and mESCs (Creyghton et al. 2008; Zhang et al. 2017; Carter et al. 2018). We used a strain expressing a C-terminal H2A.Z-GFP fusion protein to perform ChIP-seq with antibodies against H3K27me2/3 and GFP. We confirmed the H2A.Z-GFP was functional by testing the growth rate with race tubes, as disruption of H2A.Z function leads to significantly impaired growth in Neurospora (Figure S3 and Figure S5).Visual inspection of the enrichment profiles in a genome browser revealed a mostly mutually exclusive localization pattern (Figure 3A). There are some small H2A.Z peaks in PRC2-target domains, such as in Figure 3A; however, these were rare (Figure 3B). The genomic locations with the highest enrichment for H2A.Z-GFP are the regions immediately before and after the TSS of most genes, with low enrichment in gene bodies and 3′ ends (Figure 3C). On average, we find little enrichment of H2A.Z-GFP in the promoters and gene bodies of H3K27me2/3 enriched genes or at the center of H3K27me2/3 peaks, confirming that H3K27me2/3 and H2A.Z are largely mutually exclusive (Figure 3, D and E).

Figure 3.

H3K27me2/3 and H2A.Z are not colocalized in N. crassa mycelium. (A) Genome browser images of ChIP-seq for H2A.Z-GFP (green) and H3K27me2/3 (blue) enrichment across Linkage Group (“chromosome”) VII. A segment of chromosome VII is displayed at higher resolution to visualize the distinct patterns of each modification. Distinct peaks are located at the start of many genes in the genome yet few H2A.Z peaks are present within transcriptionally silent PRC2-target domains. (B) Heatmaps of H3K27me2/3 (blue) and H2A.Z-GFP (green) enrichment ordered by H3K27me2/3 enrichment. Heatmaps are centered on the transcription start site (TSS) for all 10,397 genes, + or –1000 bp for a full window size of 2000 bp. (C) Gene profile of H2A.Z-GFP (green line) and H3K27me2/3 (blue line) enrichment for all 10,397 genes (fit to 1000 bp for gene body length) in the genome –1000 bp upstream of TSS, and +1000 bp downstream of TES. (D) Gene profile of H2A.Z-GFP (green line) and H3K27me2/3 (blue line) enrichment for only H3K27me2/3 genes (589 genes) enriched in the genome –1000 bp upstream of TSS and + 1000 bp downstream of TES. (E) Line plot centered on 325 PRC2-target domains displays very low enrichment for H2A.Z-GFP (green line) and high enrichment for H3K27me2/3 (blue line).

To validate H2A.Z enrichment, we also performed ChIP-seq on wild type using an antibody raised against the native N. crassa H2A.Z protein (Liu et al. 2017). We performed a western blot using the H2A.Z antibody on both the chromatin and soluble fractions of wild type, H2A.Z-GFP, and ΔhH2Az and identified a 15 kDa band in the chromatin fraction only in wild type, confirming our ability to use this antibody to recognize H2A.Z (Figure S6C). These H2A.Z ChIP-seq experiments show the same localization as the H2A.Z-GFP ChIP-seq experiments (Figure S6, A and B). Using HOMER, we called peaks and found that there were far fewer peaks identified, which is likely due to the higher signal-to-noise ratio when using the H2A.Z antibody. Overall, 695 peaks were identified, as compared to 3104 in the H2A.Z-GFP strain; however, when viewed in the genome browser, the peaks clearly overlap (Figure S6A), and, when we intersected the peaks to determine if they were in the same location, we find that 80% of the peaks were in common with the GFP peaks (Figure S6D). The localization of H2A.Z-GFP at the TSS of 5704 genes (over half of all genes) is similar to findings in multiple other organisms (Guillemette et al. 2005; Barski et al. 2007; Creyghton et al. 2008; Bargaje et al. 2012; Weber et al. 2014; Latorre et al. 2015; Dai et al. 2017; Gómez-Zambrano et al. 2019).

H2A.Z is crucial for proper regulation of 11% of the genes in N. crassa, including eed

Previous studies have implicated H2A.Z in multiple roles related to transcription including gene activation and repression (Dhillon et al. 2006; Creyghton et al. 2008; Valdes-Mora et al. 2012; Hu et al. 2013; Kim et al. 2013; Surface et al. 2016). We therefore asked if H2A.Z regulates H3K27me2/3 by regulating expression of one or more PRC2 components. We performed RNA sequencing of wild type, ΔhH2Az, Δset-7, and the double mutant ΔhH2Az;Δset-7 to determine which genes exhibit differential expression in the absence of H2A.Z. Deletion of histone variant H2A.Z causes both positive and negative misregulation of a large number of genes (Figure 4A). After Benjamini–Hochberg correction (Yoav Benjamini and Hochberg 1995), there are 1066 genes with differential transcription (adjusted P value < 0.05 and fold change ≥ 1.5). Of these 1066 genes, there are similar numbers of genes up- and downregulated in the absence of H2A.Z (559 genes upregulated and 507 downregulated) (Figure 4A and Table S6).

Figure 4.

H2A.Z is important for the proper regulation of a large number of genes in N. crassa, including eed. (A) Volcano plot of differentially expressed genes in ΔhH2Az. Deletion of hH2Az misregulates a large number of genes in both directions; however, there are slightly more genes that are upregulated upon deletion of hH2Az. It is likely that H2A.Z is necessary for the proper expression of a large percentage of genes in N. crassa. EED is labeled and highlighted in green on the plot. Genes enriched for H3K27me2/3 are in gray (FDR corrected P-value > 0.05) or pink (FDR corrected P-value < 0.05) corresponding to their significance values. The eed gene is significantly downregulated in the deletion strain. (B) Genome browser images of each gene and its corresponding H2A.Z enrichment, for all PRC2 components, there is enrichment of H2A.Z near the TSS. Boxplots of normalized counts for all subunits of PRC2 [eed (light blue), set-7 (purple), suz-12 (pink) cac-3/npf (dark blue)] in wild type, ΔhH2Az, Δset-7, and ΔhH2Az;Δset-7 backgrounds. Downregulation of eed is dependent on hH2Az deletion. (C) Heatmap showing the difference of H2A.Z-GFP enrichment for all transcripts that are misregulated in the ΔhH2Az strain. Heatmap is ordered by enrichment,: top panel transcripts that are upregulated in ∆hH2Az (559), bottom panel: transcripts that are downregulated in ΔhH2Az (507). (D) Volcano plot of 548 differentially expressed PRC2-target genes divided by their proximity to the end of the chromosome. Genes that are not significant are in black, significant telomere-proximal genes are in orange, significant internal genes are in purple, and other genes are in gray (other are genes that do not overlap 70% or more with H3K27me2/3 domains). More internal genes show upregulation (27) than downregulation, and the telomere-proximal genes show a similar number of up and downregulated genes.

H2A.Z distribution flanks the TSS throughout much of the genome, and we wanted to know if this distribution was different for the transcripts that were misregulated in the ΔhH2Az strain. There is higher enrichment at the TSS for transcripts that are downregulated in the absence of H2A.Z than there is for transcripts that are upregulated in the absence of H2A.Z (Figure 4C). These data suggest that, in Neurospora, H2A.Z may be more important for activation of genes than for repression.

We looked more closely at the PRC2-target genes and their differential expression status. Using the previously generated domains of internal and telomere-proximal H3K27me2/3 domains, we found 506 internal and 262 telomere-proximal genes. Out of these, 78 PRC2-target genes are misregulated (FDR corrected P value < 0.05 and fold change ≥ 1.5) in the ΔhH2Az strain. Of these, 31 are located in internal domains and 41 are located in telomere-proximal domains. Due to the stringency of the intersections between domains and genes, there are six that are denoted as “other” (six genes are not covered by 70% or more H3K27me2/3). There are approximately even numbers of genes that are up (46) and downregulated (32); however, we find a large number of the internal genes are upregulated (27) in comparison to the telomere-proximal genes, which are almost evenly up and downregulated (Figure 4D).

We next examined expression levels of genes encoding individual PRC2 components (Figure 4B). We found that expression of eed is significantly reduced in ΔhH2Az by more than ninefold (FDR-corrected P value = 8.96 × 10⁻⁰⁷), whereas cac-3/npf, suz-12, and set-7 were expressed at similar levels in both wild type and ΔhH2Az (Figure 4, A and B).

The eed gene showed the most dramatic change in expression compared to wild type in either ΔhH2Az or ΔhH2Az;Δset-7, but is expressed normally in the single mutant Δset-7. This indicated that deletion of H2A.Z is likely responsible for its downregulation. As an essential component of PRC2, EED is required for catalytic activity. EED is also important for recognition of the H3K27me2/3 mark and has been implicated in maintenance and/or spreading of H3K27me3 from nucleation sites (Hansen et al. 2008; Xu et al. 2010). Since H2A.Z is localized proximal to the promoters of a little over half the genes (5704) in the N. crassa genome, we examined the H2A.Z localization at the eed gene. There is a large peak of H2A.Z enrichment at the promoter of eed (Figure 4B), which appears to be crucial for normal eed expression. Promoters of other PRC2 components are also enriched for H2A.Z, but apparently are not dependent on H2A.Z for their expression. Together, these data suggest that H2A.Z is required for the proper expression of eed.

Overexpression of EED rescues H3K27 methylation levels in the absence of H2A.Z

To determine if downregulation of eed is responsible for the depletion of H3K27me2/3 observed in ΔhH2Az, we constructed a strain that lacks both eed and hH2Az, and we introduced N-terminal tagged 3xflag-eed into the his-3 locus driven by the strong constitutive clock controlled gene-1/glucose-repressible gene-1 (ccg-1/grg-1) promoter (his-3::Pccg1-3xflag-eed). We calculated expected expression levels of this construct using native ccg-1 levels observed in our RNA-seq experiment, and we expect eed to be expressed at ∼100 times the native level. To confirm this construct was being expressed at the same level in both the Δeed and Δeed;ΔhH2Az backgrounds, we performed an anti-FLAG western blot (Figure S4C). Our results confirm that the deletion of H2A.Z does not alter 3xFLAG-EED expression driven by the ccg-1 promoter. After performing H3K27me2/3 ChIP-seq in this strain, we find that the majority of H3K27me2/3 peaks are recovered in the genome (Figure 5A), but the growth phenotype of the Δeed;ΔhH2Az strain is only partially rescued (Figure S4D). Strains with deletions of either set-7 or eed, which lack any H3K27me2/3, exhibit a wild-type-like growth rate (Basenko et al. 2015; Jamieson et al. 2016). Therefore, we conclude that the defective growth of ΔhH2Az strains is likely due other functions of H2A.Z and not the loss of H3K27me2/3. There are some qualitative differences in peak shape and not all peaks are fully restored (Figure 5B), which could indicate that H2A.Z contributes to normal H3K27me2/3 via additional mechanisms. To quantify the restoration of peaks in the overexpression strain, we called peaks of H3K27me2/3 using HOMER. We identified a total of 445 peaks, which comprised ∼6% of the genome, analogous to wild type. The restored domains tended to be smaller than wild type, ranging from 500 bp to 59 kb, with an average size of 6.1 kb. To determine if the peaks observed in this strain are in wild-type locations, we again compared peaks only from assembled contigs. Using bedtools intersect, we found 301 peaks in common (out of 308 wild-type peaks on assembled contigs) indicating that the peaks in the overexpression strain were smaller but located in the same regions as in wild type. Nevertheless, the significant restoration of H3K27me2/3 suggests that reduced eed expression is the major contributor to the loss of H3K27me2/3 in the ΔhH2Az strain.

Figure 5.

Overexpression of eed rescues H3K27 methylation levels in the absence of H2A.Z. (A) Partial restoration of H3K27 methylation throughout the genome in a ΔhH2Az;Δeed strain containing his-3::Pccg-1-3xFlag-eed and overexpressing eed at ∼100× the native level. Most H3K27 methylation is restored, though there are some qualitative differences in the peak patterns between the overexpression strain and wild type. (B) Heatmaps of H3K27me3 enrichment across PRC2-target domains sorted by first by telomere-proximal (123 domains) or internal domains (186 domains) and then by enrichment, centered on each domain + or –1000 bp for a full window size of 2000 bp for wild type, ΔhH2Az, Δeed;his-3::Pccg-1-3xflag-eed, and Δeed;ΔhH2Az;his-3::Pccg-1-3xflag-eed. Not all domains are fully rescued to wild-type levels.

Discussion

H2A.Z is a highly conserved histone variant that has been linked to gene activation and repression, and control of H3K27 methylation. We report here that N. crassa H2A.Z is required for normal methylation of H3K27 in facultative heterochromatin domains. In contrast to the situation in plants and animals, we find that N. crassa H2A.Z does not colocalize with H3K27me2/3. In undifferentiated mammalian cells and in plant cells, H2A.Z colocalized with PRC2 components, H3K27me3, SUZ12, or both (Creyghton et al. 2008; Ku et al. 2012; Surface et al. 2016; Zhang et al. 2017; Carter et al. 2018; Wang et al. 2018). In mESCs, H2A.Z is found at developmentally important loci where SUZ12 is also enriched (Creyghton et al. 2008; Ku et al. 2012). In addition, this histone variant is proposed to regulate lineage commitment by functioning as a “molecular rheostat” to drive either activation or repression of genes (Subramanian et al. 2015; Surface et al. 2016; Zhang et al. 2017). This colocalization of PRC2 and H2A.Z is not seen in differentiated murine cells, and ubiquitylated residues on the C-terminal tail of H2A.Z have been hypothesized as integral for cells to maintain undifferentiated status (Ku et al. 2012; Surface et al. 2016). In plants, H2A.Z displays significant colocalization with H3K27me3 in the gene bodies of PcG-repressed genes even in differentiated tissues (Zhang et al. 2017; Carter et al. 2018). Our work highlights an important structural difference between facultative heterochromatin in plants and filamentous fungi. In both plants and animals, PRC1 and PRC2 work together to maintain stable gene repression in a number of ways including, the classic hierarchical model (L. Wang et al. 2004), ubiquitylation of H2A proteins (H. Wang et al. 2004), and inhibition of machinery involved in transcription (King et al. 2002). PRC1 monoubiquitylates H2A.Z in these organisms, which is important for Polycomb binding and repression (Surface et al. 2016; Gómez-Zambrano et al. 2019). H2A.Z may not be required for H3K27me2/3 in fungi due to the absence of PRC1. Although we did not observe colocalization of H2A.Z and H3K27me2/3 in N. crassa, it remains possible that these two chromatin features overlap in specific developmental cell types (e.g., during sexual development or meiosis). Future work is needed to test this possibility.

In N. crassa, we generally find histone H2A.Z at the promoters of a large number of genes in the genome. When viewing the localization using a metaplot, which averages the enrichment of all H2A.Z marked nucleosomes, it appears that H2A.Z flanks the TSS. Genome-wide localization of H2A.Z has been performed in a variety of organisms including Arabidopsis, Caenorhabditis elegans, S. cerevisiae, mouse, and Drosophila. H2A.Z is generally found in the promoters of active and inactive genes, as well as in vertebrate enhancers (Guillemette et al. 2005; Barski et al. 2007; Creyghton et al. 2008; Bargaje et al. 2012; Weber et al. 2014; Latorre et al. 2015; Gómez-Zambrano et al. 2019). The +1 nucleosome—the first nucleosome after the TSS, containing H2A.Z—has been postulated as a lower energy barrier to transcription elongation in Drosophila and Arabidopsis (Weber et al. 2014; Dai et al. 2017). Our data are consistent with an important promoter-specific role for N. crassa H2A.Z.

Indeed, in N. crassa we find that the eed gene contains a large peak of H2A.Z in the +1 nucleosome, and we find that H2A.Z is required for the proper expression of eed. To our knowledge, this is the first report of H2A.Z specifically regulating the eed gene. Previous studies in mESCs demonstrate that appropriate binding of multiple factors to the eed promoter are required for the normal expression of eed (Ura et al. 2008, 2011). It is also possible that there are transcription factors that can bind to the sequence associated with the H2A.Z-containing nucleosome. Nucleosomes that contain H2A.Z protect ∼120 bp of DNA from MNase digestion as opposed to nucleosomes with canonical H2A that protect 147 bp (Tolstorukov et al. 2009). This may leave more sequence available for transcription factor binding between H2A.Z-containing nucleosomes.

We observed that reduced eed expression levels leads to region-specific losses of H3K27me2/3, rather than a more general, or global, reduction. In contrast to our work, reduced Eed is reported to cause a global decrease in H3K27me3 in mESCs. In these cells, reduced expression of Eed was observed following downregulation of Oct3/4, which, in turn, led to a global reduction of H3K27me3 (Ura et al. 2008, 2011). In N. crassa, repetitive sequences (e.g., the canonical telomere repeats) are sufficient to induce an artificial H3K27me3 domain when inserted into a locus normally devoid of H3K27me3 (Jamieson et al. 2018). It is interesting that, even though we also observe the loss of H3K27 methylation throughout much of the genome, regions proximal to the telomeres retained H3K27me2/3. This suggests that PRC2 is being recruited to the telomeric region and the downregulation of eed causes a defect in the propagation of the H3K27me2/3 modification into topologically associated, nearby regions, and we will address this possibility in future Hi-C studies. Another possibility is that the internal domains have a special requirement for EED in spreading, or for the maintenance of H3K27 methylation following DNA replication. Alternatively, EED may interact directly with transcription factors that control assembly of facultative heterochromatin at certain internal domains, while other PRC2-associated proteins may be more important for targeting PRC2 to telomeres. Future studies are needed to distinguish between these possible working models.