H4K20me3 controls Ash1-mediated H3K36me3 and transcriptional silencing in facultative heterochromatin

Facultative heterochromatin controls development and differentiation in many eukaryotes. In metazoans, plants, and many filamentous fungi, facultative heterochromatin is characterized by transcriptional repression and enrichment with nucleosomes that are trimethylated at histone H3 lysine 27 (H3K27me3). While loss of H3K27me3 results in derepression of transcriptional gene silencing in many species, additional up- and downstream layers of regulation are necessary to mediate control of transcription in chromosome regions enriched with H3K27me3. Here, we investigated the effects of one histone mark on histone H4, namely H4K20me3, in the fungus Zymoseptoria tritici, a globally important pathogen of wheat. Deletion of kmt5, the gene encoding the sole methyltransferase responsible for H4K20 methylation, resulted in global derepression of transcription, especially in regions of facultative heterochromatin. Reversal of silencing in the absence of H4K20me3 not only affected genes but also a large number of novel, previously undetected, non-coding transcripts generated from regions of facultative heterochromatin on accessory chromosomes. Transcriptional activation in kmt5 deletion strains was accompanied by a complete loss of Ash1-mediated H3K36me3 and chromatin reorganization affecting H3K27me3 and H3K4me2 distribution in regions of facultative heterochromatin. Strains with a H4K20M mutation in the single histone H4 gene of Z. tritici recapitulated these chromatin changes, suggesting that H4K20me3 is essential for Ash1-mediated H3K36me3. The Δkmt5 mutants we obtained are more sensitive to genotoxic stressors and both, Δkmt5 and Δash1, showed greatly increased rates of accessory chromosome loss. Taken together, our results provide insights into a novel, and unsuspected, mechanism controlling the assembly and maintenance of facultative heterochromatin. Significance Facultative heterochromatin contains genes important for specific developmental or life cycle stages. Transcriptional regulation of these genes is influenced by chromatin structure. Here, we report that a little studied histone modification, trimethylation of lysine 20 on histone H4 (H4K20me3), is enriched in facultative heterochromatin and important for transcriptional repression in these regions in an important agricultural pathogen. Furthermore, normal levels of H4K20me3 are essential for deposition of another repressive histone mark, Ash1-mediated H3K36me3, and affect the distribution of other marks including H3K27me3. We conducted the first genome-wide assessment of H4K20 methylation levels in a fungus, and our discoveries reveal that multiple chromatin modifications are required to establish transcriptional silencing, providing the framework to understand epistasis relationships among these histone marks.


122
To determine the function of kmt5 and ash1, we deleted the respective genes in the wild-type 123 isolate Zt09 (37) by replacement with a gene encoding hygromycin phosphotransferase (hph), 124 conferring resistance to hygromycin B. Mutant strains were confirmed by PCR and Southern 125 assays ( Figure S3). Both deletion mutants exhibit phenotypic differences compared to the wild-126 type strain. The ∆kmt5 mutant showed increased sensitivity to genotoxic stress and altered 127 morphology at higher temperatures (28°C versus the usual 18°C; Figure 1D). Deletion of ash1 128 ( Figures 1D, S4).

131
Loss of Kmt5 and Ash1 result in increased chromosome loss 132 Previously, we measured accessory chromosome loss as an indicator for overall genome stability 133 (32). A four-week growth experiment including five replicate cultures of wild-type, ∆kmt5, and 134 ∆ash1 strains revealed an increase in accessory chromosome loss in both mutants (Table 1).

135
Increased chromosome loss rates did not affect all accessory chromosomes equally;

141
ChIP-seq uncovers different heterochromatin states in Z. tritici 142 To determine the effects on chromatin in both ∆kmt5 and ∆ash1 mutants, we first performed

177
Facultative heterochromatin marked by H4K20me3, H3K36me3, and H3K27me3 shows 178 strong interactions 179 To further characterize chromatin states and their influence on spatial interactions in the nucleus, 180 we performed Hi-C experiments to find regions in the genome that physically interact. We

196
Kmt5 is essential for Ash1-mediated catalysis of H3K36me3 and normal H3K27me3 197 distribution 198 Deletion of kmt5 resulted in complete loss of H4K20me3 ( Figures 1B and 4)

215
We found that most changes in chromatin states in ∆kmt5 and ∆ash1 mutants occurred in regions 216 that we characterized as facultative heterochromatin (cluster 1), i.e., regions that show 217 enrichment for H3K27me3, H3K36me3, and H4K20me3, but to a much lesser extent in regions 218 outside of facultative heterochromatin ( Figure S7). Looking specifically at enrichment of the three 219 marks in those regions in wild type and mutants ( Figure 4B), we found that, as expected, 220 H4K20me3 is absent in ∆kmt5 but only slightly reduced in ∆ash1. H3K27me3 was reduced in 221 many regions in ∆kmt5 but slightly increased in ∆ash1, and H3K36me3 was greatly reduced in 222 ∆ash1 and only slightly less so in ∆kmt5 when compared to enrichment in wild type. Taken 223 together, we showed that Kmt5 is important for facultative heterochromatin assembly and 224 maintenance and that it is essential for Ash1 activity.

226
H4K20M mutation alters abundance and distribution of H4K20me3 227 To test whether the presence of the Kmt5 protein or its catalytic activity, i.e., methylation of 228 H4K20, is important for H3K36me3, we constructed histone H4K20 mutants, replacing lysine 229 residue 20 with a methionine. We generated two different types of mutants, one a clean

236
we observed striking differences in H4K20me3 distribution in the strain with the ectopic hH4 K20M 237 integration (H4 K20/K20M ). In this mutant, H4K20me3 enrichment was retained only in specific 238 regions, co-localizing with Ash1-mediated H3K36me3 and H3K27me3 but lacking from the rest 239 of the genome. Analysis of the regions enriched with H4K20me3 in the hH4 K20/K20M mutant 240 revealed that H3K36me3 was still slightly reduced compared to wild type, but levels were higher 241 than in both ∆kmt5 and hH4 K20M mutants, while H3K27me3 enrichment was comparable to wild 242 type ( Figure 5A-C). Presence of H4K20me3 in these regions suggests that these regions are 243 preferential targets of Kmt5. Based on data collected with mutant histone H4 alleles, presence 244 of H4K20me3 appears to be necessary to deposit H3K36me3 and H3K27me3 and thus for 245 maintenance of a facultative heterochromatin state.

247
Tagging of Kmt5 impairs protein function and limits H4K20me3 spreading 248 In order to characterize Kmt5 in more detail, we generated strains harboring N-terminally (V5-249 gfp-kmt5) or C-terminally (kmt5-gfp-3xFLAG) tagged Kmt5 alleles by either replacing the 250 endogenous kmt5 in wild type or complementing the ∆kmt5 mutant at the original locus. Based 251 on ChIP-seq analyses, none of these mutants retained or restored wild-type levels of H4K20me3 252 enrichment. H4K20me3 levels were similar to those observed in the H4 K20/K20M strain in specific 253 locations co-localizing with Ash1-mediated H3K36me3 and H3K27me3 but mostly absent from the rest of the genome ( Figure 5D). The C-terminally tagged Kmt5 mutant maintained some H4K20me3 outside of facultative heterochromatin but not at levels comparable to those in wild 256 type.

257
Notably, integration of the C-terminally tagged Kmt5 in either wild-type or ∆kmt5 backgrounds 258 led to similar defects in H4K20me3 distribution. This suggests the difference in H4K20me3 is not 259 a result of impaired de novo deposition of H4K20me3 after complete loss of the modification 260 but rather a defect in H4K20me3 genome-wide deposition or maintenance associated with 261 impaired function of the tagged Kmt5 protein ( Figure S8).

263
Eaf3 is essential for Ash1 activity 264 To further investigate the connection between H4K20me3 and H3K36me3, we searched the 265 genome for predicted proteins that, in other organisms, have been shown to bind to H4K20me3    H4K20me3 in ∆ash1 mutants, it is not absence of H3K36me3 but a mechanism upstream of Ash1 289 recruitment that influences H4K20me3 levels in ∆eaf3.

291
A Rad23 homolog may act as a histone H4K20me3 demethylase 292 A recent study found that two human hHR23 proteins, homologs of S. cerevisiae Rad23, act as 293 demethylases of H4K20me1/2/3 in human cells (48). We found one potential homolog of yeast 294 Rad23 in Z. tritici (previously not annotated, on chr 2: 3,656,371-3,657,715) and deleted the 295 putative rad23 gene to determine whether Rad23 is involved in H4K20me3 demethylation in Z. 9 tritici. We found increased levels of H4K20me3 in facultative heterochromatin in ∆rad23 strains 297 and minor differences in H4K20me3 levels outside of these regions indicating that Rad23 may 298 be involved in demethylation of H4K20, particularly in facultative heterochromatin regions 299 ( Figure 6B and C, S7, S8).

301
H4K20me is required for transcriptional repression in facultative heterochromatin 302 To investigate whether the observed changes in chromatin structure correlate to changes in gene 303 expression, we sequenced mRNA from three replicates of wild type, ∆kmt5, ∆ash1, ∆eaf3, 304 hH4 K20M , and hH4 K20/K20M strains and determined which genes were differentially expressed 305 ( Figures 7A, S10). As a control for the H4 K20/K20M strain, we also included a strain that carries an 306 ectopic wild-type H4 copy. We detected few (only 10) differentially expressed genes between 307 wild type and this strain, concluding that an additional wild-type H4 copy has negligible impact 308 under these conditions (Table S1). We showed previously (32) that loss of H3K27me3 alone has 309 surprisingly minor effects on transcription, as only a few genes, mostly located on accessory  considerably lower (26% in ∆ash1, 19% in hH4 K20/K20M , and 15% in ∆eaf3). H4K20me3 levels in 319 these mutants were largely unchanged or even increased (∆eaf3) suggesting that presence of 320 H4K20me3 may limit derepression. We found that more than 55% of upregulated genes were 321 shared in ∆kmt5 and ∆ash1 mutants and almost 90% of upregulated genes in the hH4 K20M mutant 322 were also upregulated in ∆kmt5, ∆ash1, or both. In general, a large proportion of upregulated 323 genes are shared in the mutants we sequenced, indicating that they affect a common silencing 324 pathway ( Figure S10). Most of the upregulated genes (~94%), as well as genes in facultative 325 heterochromatin in general, lack functional annotations (only 57 out of 976 [~6%] genes have 326 been previously annotated in these regions), making it difficult to assess enrichment for putative 327 functions among those genes. A list including all genes, expression levels, and predicted function 328 can be found in Table S1.

329
We noticed that, especially on accessory chromosomes, many regions without previously

422
To understand the biochemistry of a putative Kmt5 complex, we tagged the protein with GFP 423 and short tags for protein purification. We showed that both N-and C-terminal tags greatly

437
How then does H4K20 trimethylation govern facultative heterochromatin de novo assembly and 438 maintenance? Our current working model (Figure 8) proposes that in maintenance mode either 439 Kmt5 alone, or in a complex, methylates H4K20 in regions that are already enriched with 440 methylated H4K20. For de novo methylation, some "signal" must be present. When Kmt5 is 441 reintroduced into a ∆kmt5 mutant, the protein, alone or in a complex, is able to detect such 442 signals and methylate H4K20, as observed in the tagged Kmt5 strains we generated. These

448
Independently from how Kmt5 first finds its targets, it seems likely that H4K20me3 is bound by 449 Eaf3 which then stimulates activity of Ash1, as shown for MRG15 in flies (47, 63); this is supported 450 by finding increased H4K20me3 levels in the absence of Eaf3 and the requirement of Eaf3 for 451 Ash1-mediated H3K36me3. Whether Eaf3 is part of an Rpd3S-like complex in filamentous fungi, 452 or whether it may directly recruit Ash1 to regions of H4K20me3 remains unknown. We propose 453 that Ash-1-mediated H3K36 methylation results in a switch of Eaf3 binding from H4K20me3 to 454 H3K36me3, though it is also possible that it may bind both modifications at the same time.

455
Little is known about proteins that interact with Kmt5 in fungi except for S. pombe Pdp1 (42).

456
Here we showed that there is no true ortholog for this protein in filamentous fungi, and that the

517
Generation of plasmids for fungal transformation 518 Plasmids were generated by Gibson Assembly (66)

551
ChIP experiments were carried out as described previously (70)

677
The authors declare that no competing interests exist.

679
Major Impact of Ash1 on Genome Stability.