The ACF chromatin remodeling complex is essential for Polycomb repression

Establishing and maintaining appropriate gene repression is critical for the health and development of multicellular organisms. Histone H3 lysine 27 (H3K27) methylation is a chromatin modification associated with repressed facultative heterochromatin, but the mechanism of this repression remains unclear. We used a forward genetic approach to identify genes involved in transcriptional silencing of H3K27-methylated chromatin in the filamentous fungus Neurospora crassa. We found that the N. crassa homologs of ISWI (NCU03875) and ACF (NCU00164) are required for repression of a subset of H3K27-methylated genes and that they form an ACF chromatin remodeling complex. This N. crassa ACF complex interacts with chromatin throughout the genome, yet association with facultative heterochromatin is specifically promoted by the H3K27 methyltransferase, SET-7. H3K27-methylated genes that are upregulated when iswi or acf1 are deleted show a downstream shift of the +1 nucleosome, suggesting that proper nucleosome positioning is critical for repression of facultative heterochromatin. Our findings support a direct role for the ACF complex in Polycomb repression.


ABSTRACT 23
Establishing and maintaining appropriate gene repression is critical for the health and 24 development of multicellular organisms. Histone H3 lysine 27 (H3K27) methylation is a 25 chromatin modification associated with repressed facultative heterochromatin, but the 26 mechanism of this repression remains unclear. We used a forward genetic approach to 27 identify genes involved in transcriptional silencing of H3K27-methylated chromatin in the 28 filamentous fungus Neurospora crassa. We found that the N. crassa homologs of ISWI 29 (NCU03875) and ACF (NCU00164) are required for repression of a subset of H3K27-30 methylated genes and that they form an ACF chromatin remodeling complex. This N. octamer of histone proteins (Kornberg 1974;Luger et al. 1997). Nucleosomes can be 57 precisely positioned on DNA by ATP-dependent chromatin remodeling proteins to 58 produce a chromatin landscape that modulates accessibility for DNA transactions, such 59 as transcription (Lai and Pugh 2017). In particular, the precise positioning of the +1 60 nucleosome, the first nucleosome downstream of the transcription start site (TSS), is 61 SNP mapping for one mutant identified a region on linkage group VI that 107 contained essentially 100% Oak Ridge SNPs, indicating the likely position of the critical 108 mutation (Figure 1figure supplement 1A). Within this region we found a point mutation 109 (CTT -> CCT) in iswi (NCU03875) predicted to cause a leucine to proline substitution at 110 a conserved position (L430P; Figure 1B). This same approach was used on a second 111 strain to map and identify a two base pair insertion (GTC -> TTGTC) on linkage group III 112 To test if deletion of these two identified genes would also cause derepression of 115 our H3K27 methylation mutant selection gene, we created strains with the 116 NCU07152::nat-1 replacement and either ∆iswi or ∆acf1 alleles. Deletion of either iswi 117 or acf1 resulted in nourseothricin-resistance, equivalent to the original mutants identified 118 in our selection ( Figure 1C). In addition, we showed that introduction of an ectopic, wild-119 type copy of iswi or acf1 into the corresponding original mutant strain largely restored 120 silencing of both H3K27 methylation mutant selection genes ( Figure 1D). We noticed 121 that disruption of these two genes resulted in an early conidiation phenotype that was 122 even evident in the spot tests ( Figure 1C,D) but this was not accompanied by an 123 increased linear growth rate. In fact, the ∆iswi strain showed a decreased growth rate 124 relative to wild type; the ∆acf1 strain grew comparably to wild type (Figure 1figure  125 supplement 1C). Together, these data confirm that iswi and acf1 are required to 126 maintain the drug sensitivity of strains containing the H3K27 methylation mutant 127 selection genes and are thus good candidates for genes involved in repression of 128 H3K27-methylated chromatin. 129 iswi and acf1 are required for repression of a subset of H3K27-methylated genes 130 We performed mRNA-seq on ∆iswi and ∆acf1 strains to determine if loss of these genes 131 had specific effects on transcription within H3K27-methylated domains, or if they were 132 general transcriptional regulators. We found that while the majority of gene expression 133 changes observed upon loss of ISWI or ACF1 occurred outside of H3K27-methylated 134 domains ( Figure 2figure supplement 1A,B), genes marked by H3K27 methylation 135 were significantly enriched in the upregulated gene sets for ∆iswi and ∆acf1 strains 136 ( Figure 2A, B). Furthermore, nearly all (92%) of the H3K27-methylated genes that were 137 upregulated in ∆acf1 were also upregulated in ∆iswi, showing significant (P < 8.036E-138 54) overlap between these two gene sets ( Figure 2C). This demonstrates that ISWI and 139 ACF1 are not simply involved in the repression of the two H3K27-methylated genes we 140 used in our initial selection (NCU05173 and NCU07152), but also necessary for the 141 repression of a large overlapping set of H3K27-methylated genes. 142 Because we previously showed that the H3K27 methyltransferase SET-7 and the 143 H3K36 methyltransferase ASH1 are involved in repression of H3K27-methylated genes, 144 we compared the gene expression profiles from ∆iswi and ∆acf1 strains to previously 145 generated mRNA-seq data from ∆set-7 or catalytic null ASH1 (ash1 Y888F ) strains (  a considerably greater increase in transcripts from H3K27-methylated genes than did 148 either deletion of set-7 or mutation of ash1 ( Figure 2D). Deletion of acf1 strains showed 149 a smaller increase in total mRNA abundance, even compared to the set-7 or ash1 150 mutants ( Figure 2D). Increases in total transcript abundance can possibly result from a 151 strong upregulation of few genes or a more modest upregulation of many genes. To 152 investigate these possibilities, we created heatmaps that contained H3K27-methylated 153 genes ordered from most upregulated to most repressed in ∆iswi strains and compared 154 this rank-ordered list to gene expression changes in ∆acf1, ∆set-7, and ash1 Y888F strains 155 ( Figure 2E). Consistent with our previous results, ISWI and ACF1 were found to be 156 involved in the repression of a similar set of H3K27-methylated genes; however, this 157 analysis revealed that deletion of iswi led to greater gene derepression. We also found 158 that loss of ASH1 function derepresses a similar gene set as loss of either ISWI or 159 crassa, we affinity-purified over-expressed 3xFLAG-ISWI from N. crassa cellular 170 extracts. Immunopurified samples were digested down to peptides and analyzed by 171 mass spectrometry (MS) to identify potential interacting proteins. We focused on 172 proteins whose counts comprised greater than 0.4% of the total spectrum counts. We 173 found that ISWI co-purifies with ACF1 as well as with CRF4-3 (NCU02684), a homolog 174 of Ioc4 and member of the Saccharomyces cerevisiae Isw1b complex (Vary et al. 2003 were also identified in a recent independent analysis of ISWI-interacting proteins in N. 182 crassa and named ISWI accessory factor 1 and 2 (IAF-1 and IAF-2), respectively 183 (Kamei et al. 2021). CRF4-3 was not previously identified as an ISWI-interacting 184 partner, (Kamei et al. 2021) but for consistency, we will adopt the new nomenclature 185 and refer to this protein as ISWI accessory factor 3 (IAF-3). 186 To confirm these interactions and to gain more information on the possible 187 formation of ISWI-containing subcomplexes, we engineered a C-terminal HA tag at the 188 endogenous locus of each of the four most prominent putative ISWI-interacting 189 partners: ACF1, IAF-3, IAF-1 and IAF-2. These proteins were purified by 190 immunoprecipitation and subjected to MS to identify interacting partners. Interactions 191 between ISWI and all four proteins was confirmed, with each HA-tagged protein pull-192 down yielding high unique peptide counts for ISWI. Additional interactions, with lower 193 unique peptide counts and lack of reciprocal pull-downs were also found ( Figure 3A). 194 These data suggest that ISWI forms multiple distinct protein complexes and, 195 importantly, that ISWI and ACF1, two proteins identified in our selection for factors 196 involved in repression of H3K27 methylated genes, interact. The ACF1-HA pull-down 197 identified two histone fold proteins, NCU03073 (HFP-1) and NCU06623 (HFP-2) (Kamei 198 et al. 2021), as interacting partners ( Figure 3A). These proteins were also identified in 199 our ISWI pull-downs, albeit below our threshold, and suggest the presence of an N.  To determine if any of the identified ISWI-interacting proteins, beyond ACF1, 205 were involved in H3K27-methylated gene silencing, we first examined whether they 206 were required for silencing the NCU07152::nat-1 selection marker. As previously 207 shown, deletion of iswi or acf1 results in robust growth on nourseothricin indicating 208 strong derepression of the nat-1 gene. Deletion of iaf-3 or hfp-1 also derepressed the 209 nat-1 marker. Strains with deletion of iaf-1 and iaf-2 showed more modest growth on 210 nourseothricin ( Figure 3B). To determine the extent to which the other three ISWI-211 interacting partners may contribute to H3K27 methylation-mediated repression, we 212 performed mRNA-seq on strains with deletions of iaf-3, iaf-1, or iaf-2. We found that 213 H3K27-methylated genes were modestly enriched in the ∆iaf-3 and ∆iaf-2 upregulated 214 gene sets ( Figure 3C,E), but not enriched in the ∆iaf-1 gene set ( Figure 3D). Taken 215 together, these findings suggest that ISWI forms multiple protein complexes in N. crassa 216 and that the ACF complex makes the greatest contribution to repression of H3K27 217 methylated genes. 218 219 iswi and acf1 are required for H3K27 methylation at a subset of genes 220 To investigate if derepression of H3K27-methylated genes in ∆iswi and ∆acf1 strains is 221 due to loss of H3K27 methylation in these regions, we performed H3K27me2/3 ChIP-222 qPCR at the two H3K27 methyl-marked genes (NCU05173 and NCU07152) used for 223 our initial mutant selection. Both of these genes retained H3K27me2/3 ( Figure 4A). To 224 evaluate H3K27me2/3 distribution genome-wide, we performed ChIP-seq. We 225 compared H3K27me2/3 at each gene and found, using two-fold cutoffs, that ∆iswi 226 strains showed increased H3K27me2/3 for 218 genes and decreased H3K27me2/3 for 227 373 genes ( Figure 4B). In contrast, ∆acf1 strains showed only 3 genes with 228 H3K27me2/3 gains and 80 genes with losses ( Figure 4C). By comparing H3K27me2/3 229 changes between the two genotypes, we found that the vast majority of genes that 230 changed H3K27me2/3 level in ∆acf1 changed in the same direction in ∆iswi : 2 of 3 231 H3K27me2/3 gains and 77 of 80 H3K27me2/3 losses in ∆acf1 were also changed in 232 ∆iswi. By examining H3K27me2/3 genome-wide, we found that much of the distribution 233 in ∆iswi and ∆acf1 strains mirrored that of wild type ( Figure 4D     one quarter of the genes that lost H3K27me2/3 were upregulated (100/373), while 16% 243 (35/218) of genes that gained H3K27me2/3 were downregulated. In ∆acf1 strains, about 244 one third of genes (26/80) that lost H3K27me2/3 were also upregulated and one of the 245 three genes that gained H3K27me2/3 was downregulated. These data show that loss of 246 ISWI has a greater effect on H3K27 methylation than loss of ACF1, and that differences 247 in H3K27 methylation do not account for all changes in gene expression in these 248 deletion strains. (DamID-seq) (Zhou 2012). We found that ACF1 localization is not restricted to one part 255 of the genome, but rather appears to interact with chromatin indiscriminately genome-256 wide ( Figure 5A, B). However, when set-7 was deleted, eliminating H3K27 methylation, 257 ACF1 localization to H3K27-methylated regions was reduced relative to wild type, 258 suggesting that H3K27 methylation, or SET-7 presence, promotes ACF1 interactions 259 specifically with these genomic regions ( Figure 5A Dam), we found that both constructs localized to non-H3K27-methylated genes at 266 similar levels, and this was independent of set-7 presence ( Figure 5C). In contrast, 267 ACF1-Dam localized to H3K27-methylated genes more than Free-Dam and this 268 increased localization was partially dependent on set-7 ( Figure 5D). These data suggest 269 that ACF1 association with facultative heterochromatin is promoted by, but not fully 270 dependent on, an intact PRC2 complex and/or H3K27 methylation. ∆iswi, ∆acf1, ∆iaf-3, ∆iaf-1, ∆iaf-2 or ∆set-7). We found that when all SD genes were 295 considered, deletion of iswi or acf1 was more likely to result in a downstream shift (>30 296 bp) of the +1 nucleosome than when iaf-3, iaf-1, iaf-2 or set-7 were deleted ( Figure 6 -297 figure supplement 2A). This trend held when only H3K27-methylated SD genes (n=358) 298 were considered ( Figure 6A). Importantly, a significant portion of the H3K27-methylated 299 genes with a shifted nucleosome is shared between iswi and acf1 (P < 9.91E-13) 300 ( Figure 6B). These data suggest that ISWI and ACF1 may work in concert to position 301 the +1 nucleosome at a subset of genes, including those in H3K27-methylated regions. 302 Analysis of nucleosomes positions at all SD genes in wild-type and mutant 303 strains (∆iswi, ∆acf1, ∆iaf-3, ∆iaf-1, ∆iaf-2 and ∆set-7) revealed some differences in precluded a strict analysis of direct ACF1 targets but considering that our data support a 311 functional role in transcriptional repression at H3K27-methylated genes, we restricted 312 our analysis to these regions (H3K27-methylated SD genes; n=358). The MNase signal 313 plots revealed that the +1 nucleosome shifted downstream in the absence of iswi or 314 acf1 ( Figure 6C, D); in contrast, no shift was seen when other ISWI-interacting partners 315 (∆iaf-3, ∆iaf-1, ∆iaf-2) were deleted ( Figure 6E-G). There was also no shift observed in 316 ∆set-7 strains when all H3K27-methylated SD 317 genes were considered ( Figure 6H), nor when we limited our analysis to H3K27-318 methylated SD genes that were upregulated upon deletion of set-7 ( Figure 6figure  319 supplement 2C). These findings suggest that ISWI and ACF1 act to position the +1 320 nucleosome at a substantial subset of H3K27-methylated genes. Furthermore, SET-7 321 and hence H3K27 methylation are not required for nucleosome positioning by 322

ISWI/ACF1 and transcriptional upregulation is not sufficient to shift nucleosomes. 323
To test if the downstream nucleosome shift at H3K27-methylated genes in ∆iswi 324 or ∆acf1 strains correlated with increased gene expression, we further focused our 325 analysis to look at nucleosome positions in H3K27-methylated SD genes that were 326 upregulated when iswi or acf1 was deleted. We found that the +1 nucleosome shifted 50 327 bp downstream on average at H3K27-methylated SD genes that were upregulated 328 (FDR < 0.05) in ∆acf1 strains ( Figure 6I) whereas no such shift was seen in the +1 329 nucleosome of H3K27-methylated SD genes that were not upregulated in ∆acf1 strains 330 ( Figure 6J). Similarly, H3K27-methylated genes that were upregulated (FDR < 0.05) in 331 ∆iswi display a more prominent downstream shift of the +1 nucleosome than those 332 genes that were not upregulated ( Figure 6K,L). Taken together, these data suggest that 333 positioning of the +1 nucleosome by ISWI and ACF1 at a subset of H3K27-methylated 334 genes contributes to transcriptional repression.

Selection for mutants defective in Polycomb silencing 463
The selection was carried out as previously described (Wiles et al. 2020

MNase-seq data analysis 553
Paired-end sequence reads were aligned to the Neurospora crassa genome (OR74A) 554 using Bowtie2 (version 2.3.3) with the option "-q -p 4 -X 250 --no-discordant --no-mixed 555 --no-unal." Paired-end alignment reads with maximum 250bp distance gap between 556 them were used in subsequent analysis. This length corresponds to mononucleosomes. 557 Only correctly aligned paired-end alignment reads were filtered using samtools (version 558 1.5) commands "samtools view -hf 0x2 input.bam | grep -v "XS:i:" Dyad Coverage was 559 The average score of dyad coverage vector for every 182 bp using the region -100 bp to 576 +1000 bp from TSS was calculated for each gene. The closest peak from TSS was 577 defined as +1 nucleosome position.

Conflict of interest 605
The authors declare that they have no conflicts of interest. 606  ChIP-qPCR data for H3K27me2/3 at the two genes used for the initial mutant selection 868 (NCU05173 and NCU07152) in the indicated strains. Filled bars represent the mean of 869 technical triplicates and error bars show standard deviation (** for P < 0.01, * for P < 870 0.05, and ns for not significant; all relative to wild type by unpaired t-test). Data are from 871 one representative experiment that was performed three times. B, Scatter plot showing 872 correlation of H3K27me2/3 at genes in wild type and ∆iswi based on biological 873

REFERENCES
replicates of ChIP-seq data. Green points represent genes with increased H3K27me2/3 874 levels (at least 2-fold over wild type) and red points represent genes with decreased 875 H3K27me2/3 levels (at least 2-fold relative to wild type) when iswi was deleted. C, 876 Average nucleosome signal at all H3K27-methylated SD genes plotted from MNase-seq 935 data for the indicated mutants and wild type. The three colored lines represent biological 936 replicate 1, biological replicate 2, and the average of the replicates for the strains 937 indicated in the key. Arrows in (C) and (D) indicate the shifted +1 nucleosome. I, 938 Average nucleosome signal at SD genes that are upregulated (FDR < 0.05) and marked 939 by H3K27 methylation in ∆acf1 strains. The three colored lines represent biological 940 replicate 1, biological replicate 2, and the average of the replicates. The boxed, shaded 941 region is enlarged in the lower panel. J, Same as panel (I) but for H3K27-methylated SD 942 genes that are not upregulated in ∆acf1 strains. K, Average nucleosome signal at SD 943 genes that are upregulated (FDR < 0.05) and marked by H3K27 methylation in ∆iswi 944 strains. The three colored lines represent biological replicate 1, biological replicate 2, 945 and the average of the replicates. Arrow indicates the shifted +1 nucleosome. L, Same 946 as panel (K) but for H3K27-methylated SD genes that are not upregulated in ∆iswi 947 strains. 948