Regional control of histone H3 lysine 27 methylation in Neurospora

Abstract

Trimethylated lysine 27 on histone H3 (H3K27me3) is present in Drosophila, Arabidopsis, worms, and mammals, but is absent from yeasts that have been examined. We identified and analyzed H3K27me3 in the filamentous fungus Neurospora crassa and in other Neurospora species. H3K27me3 covers 6.8% of the N. crassa genome, encompassing 223 domains, including 774 genes, all of which are transcriptionally silent. N. crassa H3K27me3-marked genes are less conserved than unmarked genes and only ∼35% of genes marked by H3K27me3 in N. crassa are also H3K27me3-marked in Neurospora discreta and Neurospora tetrasperma. We found that three components of the Neurospora Polycomb repressive complex 2 (PRC2)—[Su-(var)3–9; E(z); Trithorax] (SET)-7, embryonic ectoderm development (EED), and SU(Z)12 (suppressor of zeste12)—are required for H3K27me3, whereas the fourth component, Neurospora protein 55 (an N. crassa homolog of p55/RbAp48), is critical for H3K27me3 only at subtelomeric domains. Loss of H3K27me3, caused by deletion of the gene encoding the catalytic PRC2 subunit, set-7, resulted in up-regulation of 130 genes, including genes in both H3K27me3-marked and unmarked regions.

Polycomb group proteins form multimeric complexes to establish, maintain, and recognize the trimethylation of histone H3K27 (H3K27me3) (1, 2). Polycomb repressive complex 2 (PRC2), which was first described in Drosophila and consists of four core proteins: enhancer of zeste [E(Z)], extra sex combs (ESC), suppressor of zeste12 [SU(Z)12], and p55, is directly responsible for methylation of H3K27 (1, 2). The SET [Su(var)3–9; E(z); Trithorax] domain protein E(Z) is the catalytic subunit of the complex (3). SU(Z)12 and p55 each appear to facilitate nucleosome binding, whereas ESC apparently boosts the enzymatic activity of E(Z) and modestly contributes to nucleosome binding (4). Embryonic ectoderm development (EED), the mammalian homolog of ESC, was found to bind to H3K27me3, raising the possibility that it plays a role in the propagation of this histone mark (5). PRC2 has been conserved throughout evolution, with core subunits present in metazoans, plants, and even protists (6, 7).

In both animals and plants, H3K27me3 is commonly associated with transcriptionally silenced genes involved in development (8). Deletion of a PRC2 subunit increases the expression of some H3K27me3 genes (9–15), but the mechanism for gene repression by H3K27me3 is largely unknown (16, 17). The distribution of H3K27me3 varies among organisms; both Drosophila and mammals typically exhibit broad H3K27me3 domains of up to several hundred kilobases, including both transcribed and regulatory regions (18, 19). In contrast, H3K27me3 regions are rather short in Arabidopsis, with most less than 1 kb, and are largely restricted to the transcribed regions of single genes (20). This difference in H3K27me3 distribution suggests the possibility of distinct mechanisms for the control of this modification in metazoans and plants.

H3K27me3, like H3K9me3, appears to be absent from some simple model organisms, such as Saccharomyces cerevisiae. Fission yeast Schizosaccharomyces pombe sports H3K9 methylation but lacks H3K27 methylation (21). In Neurospora crassa, H3K9me3 directs DNA methylation and marks centromeric and interstitial segments of heterochromatin, which largely comprise inactivated transposons (22, 23). While studying gene silencing at the telomeres (24), we discovered that N. crassa also sports H3K27me3, allowing us to exploit this organism to study the control and function of this histone modification.

Considering the lack of information on H3K27me3 in fungi, we analyzed the distribution and function of H3K27me3 in N. crassa and two other species of Neurospora. Sizable H3K27me3 domains were found concentrated near the telomeres on all seven linkage groups (LGs) of N. crassa, and the distribution has been partially conserved in the genus. H3K27me3 covers a substantial number of specialized silent genes. The PRC2 complex, but not the PRC1 complex, is conserved in N. crassa. We found that three members of the PRC2 complex are required for H3K27me3 [SET-7, EED and SU(Z)12]; the fourth, Neurospora protein 55 (NPF; a homolog of p55), is required for H3K27me3 on just a subset of targets.

Results

Distribution of H3K27me3 in Neurospora.

We used ChIP-sequencing (Seq) to generate a high-resolution map of H3K27me3 distribution throughout the genome of N. crassa (Fig. 1A) and identified 223 H3K27me3 domains, ranging from 0.5 to 107 kb (average 12.5 kb), together occupying 2.8 Mb of the 41-Mb genome (SI Appendix, Table S1; Dataset S1). This fraction of the genome (6.8%) is similar to the H3K27me3 occupancy found in Arabidopsis, Drosophila, and mammals (25–27). Interestingly, the H3K27me3 domains of N. crassa are found predominantly near telomeres (Fig. 1A). We identified 774 predicted genes that are completely included within these domains and an additional 165 predicted genes partially covered by H3K27me3 (“border genes”) (Dataset S2). The multigene domain arrangement of H3K27me3 in N. crassa is reminiscent of animal systems (19, 28) and contrasts the situation observed in Arabidopsis, in which this mark is associated with individual genes (20).

Fig. 1.

Genome-wide H3K27me3 ChIP-Seq analysis of wild-type, Δset-7, and Δnpf strains. (A) The distribution of H3K27me3 in the WT strain (blue) grown in Bird’s medium, displayed above the predicted genes (green ticks), is represented to scale on the seven LGs of N. crassa. Below the genes, the absence of H3K27me3 enrichment in a Δset-7 strain (gray) and the regionally affected H3K27me3 distribution in a Δnpf strain (inverted black traces) are displayed. LG I is divided at the right end of its centromere into IL and IR. Red arrows indicate the locations of primer sets used for qChIP experiments. (B) Part of the right arm of LG V near the telomere (dotted line) is expanded to detail mutually exclusive H3K27me3 and H3K9me3 domains. H3K9me3 data for the whole genome are presented in SI Appendix, Fig. S2. (C) Relative H3K27me3 enrichment was determined by qChIP at the telomeres of LG I (IL and IR) and at two genic regions sporting H3K27me3 in LG VII for WT and for strains deleted for genes encoding the PRC2 subunits (SI Appendix, Table S5). (D) H3K27me3 ChIP-Seq read densities for WT, Δset-7, and Δnpf for the regions assayed by qChIP (primers indicated by red arrows).

We verified the H3K27me3 distribution determined by ChIP-Seq in three ways. First, we carried out ChIP-microarray experiments for LG VII. Equivalent results were obtained with ChIP-Seq and ChIP-microarray methods (SI Appendix, Fig. S1A). Second, we used ChIP-Seq to assess the distribution of H3K27me3 in different Neurospora culture media [Vogel’s (29) and Bird media (30)]; virtually identical distributions of H3K27me3 (SI Appendix, Fig. S1B; Datasets S1 and S3) were observed. Third, we used ChIP followed by real-time quantitative PCR to verify H3K27me3 enrichment at LG I telomeres and at two genes on LG VII (qChIP; Fig. 1C).

Our study on telomere silencing in N. crassa provided early evidence of both H3K9me3 and H3K27me3 in several telomeric regions (24). The conventional ChIP experiments did not provide information on whether these two marks truly colocalize in N. crassa, however. To address this possibility, we performed ChIP-Seq for H3K9me3 and compared the distributions of these two marks (SI Appendix, Fig. S2). Interestingly, we found that H3K27me3 often neighbors H3K9me3, but each mark forms distinct domains with little or no overlap (Fig. 1B; SI Appendix, Fig. S2). As found in more limited surveys (22, 23), H3K9me3, which mirrors the distribution of DNA methylation in N. crassa (23), was almost exclusively associated with gene-depleted, A:T-rich sequences altered by repeat-induced point mutation (RIP). In contrast, H3K27me3 domains include numerous predicted genes and the base composition of these regions is not skewed in any obvious way.

Neurospora H3K27me3-Marked Genes Are Distinctive.

As a first step to explore the possible function of the H3K27me3 mark in Neurospora, we surveyed the underlying genes for their evolutionary conservation and predicted functions. It became obvious that the H3K27me3-marked genes are not representative of the overall genome. The average predicted size of proteins encoded by H3K27me3 genes is smaller than that of genes not marked by H3K27me3 (373 vs. 513 amino acids; SI Appendix, Fig. S3). Furthermore, an unusually high fraction of the 774 H3K27me3-marked genes have no predicted function (71.4%; compared with 38.2% of genes in the genome overall; SI Appendix, Fig. S4). Although most of the H3K27me3-marked genes are unannotated, the annotated set does contain representatives of a full spectrum of categories (e.g., metabolism, cellular transport; SI Appendix, Fig. S4).

The high level of novelty among genes marked by H3K27me3 prompted us to investigate their relative conservation. We found that H3K27me3-marked genes are substantially less conserved than genes not marked by this modification. Seventy-nine percent of N. crassa H3K27me3-marked genes have orthologs found only in fungi, compared with 49% for non–H3K27me3-marked genes (Fig. 2A). Moreover, unlike N. crassa genes generally, a high proportion of H3K27me3-marked genes are limited to the Neurospora genus or to closely related genera in the Sordariomycetes class; 30% are Neurospora-specific (compared with 9% for non– H3K27me3-marked genes) and an additional 26% are limited to the Sordariomycetes (compared with 8% for non–H3K27me3-marked genes) (Fig. 2A).

Fig. 2.

Conservation of H3K27me3 genes in Neurospora species. (A) The phylogenetic tree depicts the classification of Neurospora species and their relationship to common model organisms. The pie charts illustrate conservation of N. crassa orthologs in H3K27me3 domains (H3K27me3 genes) or outside H3K27me3 domains (non-H3K27me3 genes). (B) Conservation and H3K27me3 status of N. crassa genes with or without H3K27me3 relative to N. tetrasperma and N. discreta. Full species names of the organisms indicated in A are: Aspergillus nidulans, Arabidopsis thaliana, Coprinopsis cinerea, Homo sapiens, Podospora anserina, and Sordaria macrospora.

Conservation of H3K27me3 in Neurospora Species.

Our observation that N. crassa H3K27me3-marked genes show a strong fungal-specific bias raised two potentially related questions: (i) How conserved are H3K27me3-marked genes within the Neurospora genus? and (ii) To what extent is the mark itself conserved? To address these questions, we determined the distribution of H3K27me3 in two other Neurospora species, N. tetrasperma and N. discreta. Our ChIP-Seq analyses demonstrated that H3K27me3 covers a similar fraction of each of the three genomes and that all three species have a similar number of H3K27me3 domains (SI Appendix, Table S2; Datasets S4 and S5). The N. tetrasperma and N. discreta genomes are not yet fully assembled, so it is not certain that the H3K27me3 domains are preferentially near the ends of chromosomes as in N. crassa.

Although all three species show comparable fractions of their genomes associated with this mark, we found striking evidence of dynamics. Among N. crassa H3K27me3-marked genes, only ∼35% are marked in both N. tetrasperma and N. discreta, ∼12% are unmethylated in both comparative species, and nearly 25% are methylated in one comparative species but not the other (Fig. 2B; SI Appendix, Table S3). Conversely, homologs of 2.5% of unmethylated N. crassa genes are marked with H3K27me3 in N. tetrasperma and/or N. discreta (Fig. 2B; SI Appendix, Table S3). Moreover, compared with non–H3K27me3-marked genes, a high fraction of N. crassa genes associated with the H3K27me3 mark are absent in one or both of the comparative species (∼14% and ∼9%, respectively, compared with ∼6% and ∼3% for unmethylated N. crassa genes; Fig. 2B). Thus, N. crassa genes that are not found in one or both of the sister species are marked by H3K27me3 more frequently than those found in all three species (Fig. 2B; SI Appendix, Table S3). In sum, we found partial conservation of the H3K27me3 mark among three closely related species of Neurospora.

PRC2 Complex Is Conserved in N. crassa.

Pioneering work in Drosophila demonstrated that the PRC2 complex is responsible for methylation of H3K27. This complex consists of four core subunits in Drosophila—E(Z), ESC, SU(Z)12, and p55 —which are highly conserved in plants and animals, although some subunits are duplicated in higher eukaryotes (6). The genome of N. crassa contains one homolog for each of the PRC2 subunits, with their predicted functional domains largely intact (SI Appendix, Figs. S5 and S6). The N. crassa homolog of the gene for the catalytic subunit, E(Z), is set-7 (31); the homolog of the p55 gene, which we named npf, was previously named chromatin assembly complex 3 (cac-3) because it was potentially a component of the putative Neurospora chromatin assembly factor 1 (CAF-1) complex (31).

To determine whether the four putative PRC2 subunits form a complex in N. crassa, we fused a 3XFLAG epitope tag to the amino terminus of the EED homolog expressed under the control of the qa-2 promoter. We purified tagged EED using an anti-FLAG affinity gel and identified EED and associated proteins by mass spectrometry. In addition to EED, we found strong coverage for the other three putative PRC2 subunits, SET-7, SU(Z)12, and NPF, implying that, indeed, a PRC2-like complex forms in N. crassa (SI Appendix, Table S4).

To explore the function of the N. crassa PRC2 complex, we obtained knockout strains for the corresponding genes (32). In contrast to their essential role in developmental processes of higher eukaryotes (6), we found that none of the four PRC2 homologs is essential in N. crassa. Indeed, strains with knockouts of set-7, eed, or su(z)12 displayed no growth defects under standard growth conditions (SI Appendix, Fig. S7A). However, deletion of npf resulted in a slow-growth phenotype; its linear extension rate is ∼84% of that of WT (SI Appendix, Fig. S7 A and B). In other systems, besides its role in PRC2, NPF (called RbAp46/RbAp48 in mouse) has been shown to be a histone-binding protein and a component of ATP-dependent nuclear remodeling complexes (33). Considering that growth was not retarded in knockouts for the other three PRC2 subunits, it seems likely the slow-growth of the npf strain is due to a role of this protein in complexes other than PRC2. We also found that crosses homozygous for a deletion of set-7 were fruitful, indicating that H3K27me3 is not essential for the sexual cycle.

NPF Is Differentially Required for H3K27me3.

We initially used both immunoblotting and ChIP to test mutants lacking components of the N. crassa PRC2 complex for H3K27me3, but found that available antibodies were most reliable for ChIP experiments. We used qChIP to access the level of H3K27me3 enrichment near both telomeres on LG I and at two genic regions on LG VII in each of the PRC2 knockout strains (Fig. 1C). H3K27me3 enrichment was completely lost from the LG I telomeres in all four PRC2 mutants. Similarly, H3K27me3 was eliminated at the two genic regions in the Δset-7, Δeed, and Δsu(z)12 strains. Surprisingly, there was only a partial reduction of H3K27me3 at these genic regions in the Δnpf strain. To explore this further, we analyzed the distribution of H3K27me3 across the whole genome by ChIP-Seq in Δset-7 and Δnpf strains. Consistent with the qChIP results, we did not observe enrichment for H3K27me3 in the Δset-7 strain, implying that the histone methyltransferase catalytic subunit of the PRC2 complex is absolutely required (Fig. 1). Consistent with the initial qChIP results, we found a differential loss of H3K27me3 in the Δnpf strain (Fig. 1). Both the size and number of H3K27me3 domains was reduced in the Δnpf strain compared with WT (SI Appendix, Table S1; Datasets S6 and S7). Although H3K27me3 enrichment was reduced across most areas of the genome, the mark was specifically absent near the telomeres of all of the chromosomes. We conclude that SET-7, EED, and SU(Z)12, but not NFP, are absolutely required for H3K27me3 in N. crassa.

H3K27me3 Domains Are Transcriptionally Quiescent.

To determine whether H3K27me3 represents a repressive mark in N. crassa, we analyzed gene expression in a WT strain by RNA-Seq (Fig. 3A). As in other model systems, we found few or no transcripts produced from the H3K27me3-marked genes. To illustrate the negative correlation between H3K27me3-marked genes and expression, we plotted transcript abundance versus H3K27me3 level across the genome (Fig. 3C). The H3K27me3-marked genes (blue) and the genes falling on the domain borders (black) showed extremely low transcript levels; the vast majority of transcripts were produced by non-H3K27me3 genes (green).

Fig. 3.

Deletion of set-7 de-represses a subset of Neurospora genes. (A) RNA-Seq read densities for WT (black) and Δset-7 (green) are displayed below the genes (green ticks) for LG V; H3K27me3 enrichment (blue) is included for reference. (B) Two genes (NCU08907 and NCU11087) within an H3K27me3 domain are expanded along with the corresponding RNA-Seq reads. Northern confirmation of increased NCU08907 expression in the Δset-7 mutant; 18S rRNA stained with methylene blue is shown as a loading control. (C) Transcript abundance, expressed as fragments per kilobase of exon per million fragments mapped (FPKM), plotted vs. H3K27me3 level (reads) for genes contained within H3K27me3 domains (blue circles), genes partially contained in H3K27me3 domains (H3K27me3 “border” genes; black diamonds), and for genes outside of H3K27me3 domains (non-H3K27me3 genes; green crosses). The Δset-7 up-regulated genes that were verified by Northern blots are indicated by red dots.

We next asked whether the absence of H3K27me3 in a Δset-7 strain is sufficient to increase expression of H3K27me3-marked genes. Using a stringent threshold for the change of expression (∼7.5-fold), we found 130 genes with increased expression in the Δset-7 strain relative to the WT strain (Dataset S8). De-repression of four H3K27me3-marked genes was confirmed by Northern blot analyses of total RNA (Fig. 3 A and B; SI Appendix, Fig. S8). In addition, five genes showed lower transcript levels in a Δset-7 strain (Dataset S9). Overall, the functional classification of the up-regulated genes is similar to that of the total H3K27me3-marked genes, consisting of primarily unannotated genes (SI Appendix, Figs. S4 and S9). Interestingly, of the 130 de-repressed genes, only 21 fell completely within the H3K27me3 domains identified in the WT strain. This result is similar to what has been observed in Arabidopsis (9). Thus, although loss of H3K27me3 may be necessary, it is not sufficient to increase expression of the majority of H3K27me3-marked genes under the conditions of our experiment.

Discussion

Elements of the Polycomb repression system, originally uncovered in Drosophila, have been found in a variety of higher animals and plants, but not in yeast species that have been examined (S. cerevisiae and S. pombe). Previously we found that H3K27me3, a hallmark of the Polycomb system, is represented in the model filamentous fungus N. crassa (24, 34). Here, we present a genome-wide analysis of the distribution of this chromatin mark, characterize the underlying machinery, and start to explore its function and evolutionary dynamics. A sizable fraction of the N. crassa genome (6.8%) is marked by H3K27me3, covering 774 genes in 223 domains. These domains, some of which are hundreds of kilobases long, are found preferentially near the ends of the chromosomes (Fig. 1). Unlike the case in Arabidopsis, in which H3K27me3 covers single genes in domains of less than 1 kb (20), the broad domains in N. crassa (12.5 kb average) are reminiscent of Drosophila and mammals, which average 70 and 43 kb, respectively (18, 35).

H3K27me3 and H3K9me3 are both regarded as repressive marks (1, 8, 15, 16, 36) but are distributed differently. In N. crassa, as in higher eukaryotes, H3K9me3 is a feature of constitutive heterochromatin and is found principally associated with centromeric heterochromatin, which is characterized by H3K9me3, DNA methylation, a paucity of genes, and an abundance of repeats that show evidence of RIP (23, 37). H3K9me3 is also found in N. crassa associated with numerous small islands of sequences mutated by RIP and near telomeres (24), adjacent to where we found H3K27me3. Unlike H3K9me3, we show that H3K27me3 is in gene-rich regions. Notably, the H3K27me3 and H3K9me3 regions do not appear to overlap (Fig. 1 and SI Appendix, Fig. S2). This is consistent with reports from plant and animal systems that describe mostly mutually exclusive H3K27me3 and H3K9me3 distributions (18, 26–28, 38–40). It will be interesting to learn whether the machinery responsible for methylating H3K9 and H3K27 are inherently incompatible.

As a step to investigate the mechanism of H3K27me3 in N. crassa, we identified and tested homologs of PRC2 components identified in other organisms (SI Appendix, Fig. S5). We found that H3K27me3 absolutely depends on three of the PRC2 components: SET-7 (equivalent to EZH2), EED, and SU(Z)12. Thus, unlike the situation in Drosophila and other animals, H3K27me3 is not essential in N. crassa. Interestingly, we found that the fourth component of the presumptive N. crassa PRC2 complex, NPF (Neurospora homolog of Drosophila P55 and mammalian P48), is not required for all H3K27me3. In particular, NPF is only required for H3K27me3 domains near telomeres; domains farther from telomeres are somewhat affected, shrinking in the absence of NPF. We conclude that NPF is not required for the methyltransferase activity of PRC2, unlike SET-7, EED, and SUZ12. Perhaps NPF and its homologs in other organisms, which have been reported to bind histone H4 (41, 42), help tether the PRC2 complex to nucleosomes via its six WD40 domains. This is consistent with the observation that a trimeric Esc-E(z)-Su(z)12 complex trimethylates H3K27 in vitro, but is unable to bind nucleosomes (4). It is interesting that various genomic regions are differentially dependent on NPF. Perhaps regions that do not lose H3K27me3 in the Δnpf strain rely on another WD40 domain–containing protein. We are unaware of direct evidence from other organisms of a comparable, genome-wide influence of NPF homologs on H3K27me3 or other histone modifications, but there are clues that the effect is not limited to N. crassa; the combined findings of two studies with Arabidopsis revealed that H3K27me3 marks approximately half of the genes that become de-repressed in a mutant for the NPF homolog (20, 43). Although Arabidopsis does not contain the broad H3K27me3 domains observed in N. crassa, Drosophila, and mammals, selective reduction of H3K27me3 resulting from loss of msi1 could be responsible for the de-repression of a subset of H3K27me3-marked genes and their presumptive indirect targets (20, 43).

As in other organisms, we found that N. crassa genes marked by H3K27me3 produce little or no transcripts. The silence of these genes is not a simple consequence of this mark, however, because elimination of H3K27me3 by mutation of set-7 did not derepress the bulk of the genes. The Δset-7 strain showed up-regulation of 130 genes, but only 21 of these fell within H3K27me3 domains, representing 2.7% of the 774 genes marked by H3K27me3 in the WT strain. The 109 genes that showed increased expression but are located outside of H3K27me3 domains represent 1.1% of the total genes not marked by H3K27me3. Thus, the increased gene expression observed in the set-7 mutant is modestly skewed toward H3K27me3-marked genes. That only a small subset of genes within H3K27me3 domains was up-regulated suggests that, in addition to loss of the repressive mark, activating signals may be required to express genes in H3K27me3 domains. There are also indications in human, Drosophila, and Arabidopsis that depletion of Polycomb group genes is only sufficient to activate a subset of H3K27me3-marked genes, leading researchers to postulate a secondary layer of regulation (9, 11, 14, 44–47). In Arabidopsis, the up-regulation of genes not marked by H3K27me3 in an H3K27me3-deficient background is thought to result from induction of transcriptional regulators (9). Because the majority of H3K27me3-marked genes in N. crassa are unannotated, it is not yet clear if this is the case in Neurospora; it will be interesting to learn whether the up-regulated non–H3K27me3-marked genes are controlled by any of the up-regulated H3K27me3-marked genes.

Although H3K27me3 preferentially marks developmental genes in animals, it is not yet clear if this is the case in N. crassa, because the vast majority of the marked genes have not been characterized. Similarly, although some developmental genes are marked by H3K27me3 in Arabidopsis, a high proportion of H3K27me3-marked genes are also functionally unknown genes in this organism (20). Lack of gene annotation often reflects a lack of characterized orthologs in other species. Indeed, we found that N. crassa H3K27me3-marked genes show a striking lack of conservation; most are confined to the fungal kingdom, with the largest fractions confined to the class Sordariomycete and genus Neurospora (Fig. 2A). Thus, H3K27me3 seems to preferentially mark poorly conserved or “new” genes. Conceivably, as new genes are incorporated into a genome, H3K27me3 could serve as a “safety” mechanism by silencing them.

To investigate the evolutionary dynamics of H3K27me3 and of the associated genes, we analyzed the distribution of H3K27me3 in two closely related Neurospora species. We found that H3K27me3 is present in both N. tetrasperma and N. discreta and that the number and size of the H3K27me3 domains in both species is similar to N. crassa (SI Appendix, Table S2). Although these closely related species predominantly share a common set of genes, we found that those that were not shared (i.e., those that are unique to N. crassa genes, or those only found in only one of the other Neurospora species examined) are approximately threefold more frequently marked with H3K27me3 than are genes in the overall genome (SI Appendix, Fig. S10). Interestingly, among orthologs common to the three species, 35% of the N. crassa H3K27me3-marked genes are also marked by H3K27me3 in both N. tetrasperma and N. discreta. A sizable number of genes that are H3K27me3-marked in N. crassa are not marked in at least one of the other species (24.6% “MU” in Fig. 2B) or do not currently have the gene in one (14.1% “M−” in Fig. 2B) or both (9.0% “–” in Fig. 2B) other species. Thus, although the fraction of the genome that is marked by H3K27me3 in the three genomes is equivalent, the distribution of the mark and the associated genes are not highly conserved (Fig. 2B).

The presence of H3K27me3 and PRC2 components in N. crassa and some other lower eukaryotes (e.g., Chlamydomonas reinhardtii (7)) and absence of this mark and the associated machinery from some other lower eukaryotes [e.g., S. cerevisiae and S. pombe (21, 24)] suggests that this system has not been retained throughout evolution (7), and is consistent with its nonessential role in N. crassa. Previous studies suggested PRC2 arose before PRC1 and showed that it is the more conserved of the two Polycomb complexes (2). N. crassa appears to lack PRC1 homologs, raising the question of how the H3K27me3 mark is “read” in this organism.

Materials and Methods

Neurospora Strains and Methods.

Neurospora strains used in this study (SI Appendix, Table S1) were grown and crossed following standard procedures (48). The Δset-7, Δeed, Δsuz12, and Δnpf strains were generated by the Neurospora gene knockout project (32) and obtained from the Fungal Genetic Stock Center (FGSC; www.fgsc.net). RNA isolation and Northern blotting was done as described (49), except the mycelium, grown 16 h at 32 °C, was disrupted using a minibead beater (Biospec).

ChIP.

ChIP was performed as previously described (37) using anti-H3K27me3 (Active Motif 39535; ChIP-Seq and qChIP), anti-H3K27me3 (Upstate 07-449; ChIP-chip), anti-H3K9me3 (Active Motif 39161; ChIP-Seq), and anti-H3K4me2 (Active Motif 39141; qChIP). ChIP-chip procedures, including microarray design, sample labeling, microarray hybridization, and data analysis, were conducted as described (23). For qChIP, real-time PCR experiments were performed three times using FAST SYBR Green master mix (Kapa) with the listed primers (Table S2) and analyzed using a Step One Plus Real Time PCR System (Life Technologies). Relative enrichment of H3K27me3 at representative telomeres and genic regions was determined versus input and then standardized to relative enrichment of H3K4me2 at hH4.

Sequencing.

Neurospora strains used for ChIP-Seq and RNA-Seq were grown in liquid media as described in the figures. A detailed description of cDNA preparation, preparation of ChIP-enriched DNA and double-stranded cDNA for sequencing, and description of sequence analysis is available in SI Appendix, SI Materials and Methods. Sequencing reads can be downloaded from NCBI (accession no. SRA0688854). ChIP-Seq reads and ChIP-chip data were mapped to the N. crassa OR74A reference genome (50) (www.broadinstitute.org/annotation/genome/neurospora/MultiDownloads.html), N. crassa OR74A v10 genome assembly, N. tetrasperma FGSC 2508 mat A v2.0 reference genome (51), or the N. discreta FGSC 8579 mat A (US Department of Energy Joint Genome Institute).

Bioinformatic Analysis.

Orthologs were identified by aligning the three Neurospora genomes with Mercator (52) and identifying orthologs as genes found in the same position between the genomes. Phylogenetic clades were identified by clustering genes into orthologous groups using OrthoMCL (53), which first links genes by similarity with the BLASTP program followed by the MCL graph algorithm, which identifies groups through a Markov Clustering procedure (54). Comparison of orthology relationships and H3K27me3-marked and -unmarked genes was completed with custom Perl scripts (https://github.com/hyphaltip/H3K27) written with BioPerl (55). Functional classification of genes/proteins was conducted using MIPS FunCat (http://mips.helmholtz-muenchen.de/genre/proj/ncrassa/Search/Catalogs/searchCatfirstFun.html). Domain prediction used RSEG software (http://smithlab.usc.edu/histone/rseg/) with a bin-size of 500 bp. The RSEG difference program was run to determine domain differences between H3K27me3 domains predicted from independent ChIP-Seq experiments.