PhpC^NF-Y transcription factor infiltrates heterochromatin to generate cryptic intron-containing transcripts crucial for small RNA production

Abstract

The assembly of repressive heterochromatin in eukaryotic genomes is crucial for silencing lineage-inappropriate genes and repetitive DNA elements. Paradoxically, transcription of repetitive elements within constitutive heterochromatin domains is required for RNA-based mechanisms, such as the RNAi pathway, to target heterochromatin assembly proteins. However, the mechanism by which heterochromatic repeats are transcribed has been unclear. Using fission yeast, we show that the conserved trimeric transcription factor (TF) PhpC^NF-Y complex can infiltrate constitutive heterochromatin via its histone-fold domains to transcribe repeat elements. PhpC^NF-Y collaborates with a Zn-finger containing TF to bind repeat promoter regions with CCAAT boxes. Mutating either the TFs or the CCAAT binding site disrupts the transcription of heterochromatic repeats. Although repeat elements are transcribed from both strands, PhpC^NF-Y-dependent transcripts originate from only one strand. These TF-driven transcripts contain multiple cryptic introns which are required for the generation of small interfering RNAs (siRNAs) via a mechanism involving the spliceosome and RNAi machinery. Our analyses show that siRNA production by this TF-mediated transcription pathway is critical for heterochromatin nucleation at target repeat loci. This study reveals a mechanism by which heterochromatic repeats are transcribed, initiating their own silencing by triggering a primary cascade that produces siRNAs necessary for heterochromatin nucleation.

Introduction

Eukaryotic genomes are organized into “open” and “closed” chromatin domains referred to as euchromatin and heterochromatin^1,2. Heterochromatin represses gene expression to ensure proper cellular differentiation and safeguards genome stability by silencing repetitive DNA elements^3,4. The assembly of heterochromatin domains involves posttranslational modification of histones, which creates regions with characteristic hypoacetylation of histones and high levels of histone H3 lysine-9 methylation (H3K9me)^2,5. Heterochromatin is nucleated at specific sites, known as nucleation sites, and then spread to surrounding sequences². Evidence suggests that transcripts from target loci, including those derived from repetitive DNA elements, play an important role in guiding heterochromatin assembly proteins to nucleation sites^1,2,6. However, despite the need for transcription within repressive heterochromatin domains, the mechanisms involved are not fully understood.

The fission yeast Schizosaccharomyces pombe is a valuable model system for studying heterochromatin assembly, as it shares conserved factors and pathways with higher eukaryotes. The Clr4 histone methyltransferase, like its mammalian counterparts SUV39H1 and SUV39H2, methylates histone H3K9^7,8, which in turn recruits HP1 chromodomain proteins^7,9,10,11,12. In addition to methylating H3K9, Clr4^Suv39h can recognize preexisting H3K9me3 through its chromodomain, and its ability to both “read” and “write” H3K9me is crucial for spreading and epigenetic inheritance of heterochromatin^2,13. Spreading of H3K9me and its associated HP1 proteins allows heterochromatin-associated effectors, such as histone deacetylases (HDACs) involved in transcriptional gene silencing, to exert their influence across extended chromosomal domains².

Genome-wide mapping studies have discovered heterochromatin domains widely distributed across the S. pombe genome. While small heterochromatin islands are found at various regulated genes, the major targets of heterochromatin assembly are the pericentromeric repeat regions, sub-telomeres, and the silent mating-type (mat) region^14,15. These constitutive heterochromatin domains are coated with H3K9me and HP1 proteins and each of them contains a specific class of repeat elements, referred to as dg and dh repeats, which serve as RNAi-dependent heterochromatin nucleation centers^16,17,18. The silent mat region contains a single copy of the dg- and dh-like element, referred to as cenH, which serves as an RNAi-dependent heterochromatin nucleation center^16,19. Heterochromatin targeted to cenH and a nearby site spreads across the ~20-kb domain surrounded by inverted repeat boundary elements (IR-R and IR-L) through the Clr4^Suv39h read-write mechanism^13,16,20.

RNAi-mediated heterochromatin assembly necessitates the transcription of repeat elements by RNA polymerase II (RNAPII)^21,22. During the S-phase, dg/dh repeats are transcribed bidirectionally^23,24, producing transcripts that are converted into siRNAs by the RNAi factors Argonaute (Ago1), Dicer (Dcr1), and RNA-dependent RNA polymerase (Rdp1)^{14,25,26,27,28}. The siRNAs are incorporated into the RNA-induced transcriptional silencing (RITS) complex, comprising Ago1, Tas3, and Chp1²⁸, which is then guided by the siRNA to nascent repeat transcripts to recruit Clr4^Suv39h, thereby methylating H3K9^13,29,30.

A noteworthy aspect of RNAi-mediated heterochromatin assembly is a self-reinforcing loop: siRNAs generated by RNAi facilitate heterochromatin assembly, promoting the stable association of RNAi pathway proteins with repeat loci and ensuring efficient transcript processing^26,31. Together, H3K9me and Swi6^HP1 play an important role in this process by engaging the RNA-dependent RNA polymerase complex (RDRC) to convert repeat transcripts into double-stranded RNAs for siRNA production by Dcr1^{11,26,31,32,33}. However, substantial levels of siRNAs can still be produced in the absence of heterochromatin modifications^32,34. This H3K9me- and Swi6^HP1-independent processing of dg/dh transcripts by RNAi is considered essential for the initial heterochromatin nucleation. The precise mechanism by which RNAi machinery initiates siRNA accumulation independently of heterochromatin remains to be fully elucidated.

Here, we focused on identifying factors driving the transcription of repeat elements within repressive heterochromatin and discovered how the resulting transcripts are processed by the RNAi machinery to produce siRNAs. We identify a transcription factor (TF) complex, PhpC, analogous to the human NF-Y trimeric TF complex, that can infiltrate repressive heterochromatin via its histone-fold domain and promotes transcription of repeat elements. PhpC, which recognizes CCAAT box motifs, acts in a cooperative manner with another TF, named Moc3, that we show also localizes to cenH and dh repeats, regardless of the heterochromatic state or cell cycle. Our work further reveals that TF-mediated transcription of heterochromatic repeats produces transcripts that contain inefficiently spliced introns, termed cryptic introns. Splicing machinery is required to engage RNAi machinery to these cryptic intron-containing transcripts to generate siRNAs and trigger heterochromatin nucleation. This study advances our understanding of the mechanisms that allow heterochromatic repeats to be transcribed to generate siRNAs for RNAi-mediated heterochromatin nucleation.

Results

TFs bind to repeat elements within heterochromatin domains

Although heterochromatin is generally inaccessible to transcriptional machinery, the transcription of heterochromatic repeat elements is essential for RNAi-mediated targeting of heterochromatin^21,22,23,24. This raises a long-standing question: how are these elements transcribed?

To answer this question, we mapped the distribution of TFs, including those that show genetic interactions with the RNAi machinery and components of the Clr4^Suv39h complex³⁵. We focused on detecting binding to heterochromatic repeat elements, such as those at centromeres and the silent mat region, by performing chromatin immunoprecipitation-sequencing (ChIP-seq) analysis of TFs tagged with GFP. Atf1 and Pcr1, ATF-CREB family proteins previously known to bind specific sites at the silent mat region³⁶, were included as controls. As expected, our analyses detected TFs localizing to gene promoters, lncRNAs, and other ncRNAs (tRNAs and rRNAs) (Supplementary Fig. 1a and Supplementary Data 1). Notably, the histone fold domain-containing TF Php5^37,38 and a Zn finger type DNA binding protein Moc3³⁹ also showed significant enrichment at the cenH element within the silent mat region and the dh centromeric repeat elements adjacent to siRNA hotspots (Fig. 1a, b and Supplementary Fig. 1b). The binding peak of Php5 overlapped with that of Moc3 but not with those of Atf1 and Pcr1 that localize to the previously described CAS (Clr3-attracting sequence) element (Fig. 1c and Supplementary Fig. 1c)⁴⁰.

Fig. 1: The transcription factors Php5 and Moc3 bind to heterochromatic repeats.

a, b ChIP-seq analysis of TF distribution at the silent mat locus (a) and centromere 1 (b) in a wild-type background strain. The blue and small orange lines represent transcripts and small RNAs, respectively. Annotations: cenH refers to homology to centromeric dg/dh; IR signifies an Inverted Repeat. c The enlarged section from (a) shows the distribution of the specified TFs. CAS denotes Clr3-attracting sequence. d ChIP-qPCR analysis of the distribution of the indicated GFP-tagged transcription factors at cenH in wild-type cells. The data is expressed as the relative fold enrichment compared to the leu1 control locus and is normalized to the untagged strain. Significant enrichments are defined as those ≥ 2-fold (dotted line). Data from 3 independent biological experiments are presented as the mean ± SD. The positions of qPCR primers for dh and cenH are indicated by the black lines and arrows in (b) and (c), respectively. Source data are provided as a Source Data file.

Since cenH shares approximately 96% homology with centromeric dg and dh repeats¹⁹, ChIP-seq analysis cannot conclusively determine if TFs localize to both or only one of the locations. In fact, the observed cenH peak contains a mix of reads that map uniquely to cenH in addition to some reads that map equally well to pericentromeric dg/dh. Therefore, to further validate our ChIP-seq results, we performed site-specific qPCR using primers specific to each of the heterochromatic locations adjacent to the siRNA hotspots. By using this method, we confirmed that Php5 and Moc3 were indeed enriched at both cenH and dh loci (Fig. 1d and Supplementary Fig. 1b). These results indicate that TFs can access repeat elements located within repressive heterochromatin domains.

Heterochromatic TFs exhibit partial colocalization across the genome

The co-localization of Php5 and Moc3 at cenH and dh heterochromatic repeats suggested a broader collaboration between these factors. To further explore this, we performed a comparative analysis of the genome-wide distribution patterns of these TFs alongside others. A hierarchical clustering based on the signal strength of the TFs at 2,273 binding sites identified six distinct clusters (Supplementary Fig. 1d). Atf1 and Pcr1, known for their role in stress response⁴¹, were grouped within one cluster (Supplementary Fig. 1d). Despite their colocalization at heterochromatin repeat elements, Php5 and Moc3 were found in separate clusters. Only ~24% (180/762) of Moc3-occupied sites, including several ncRNAs, also contained Php5, while approximately 46% (180/389) of Php5-occupied peaks also had Moc3 (Supplementary Fig. 1e, f and Supplementary Data 2). Gene Ontology (GO) analysis revealed both Moc3 and Php5 bind to the promoters of target genes that are linked to various pathways involved in energy production, such as NADH metabolic processes, ATP generation, and glycolytic pathways (Supplementary Fig. 1g and Supplementary Data 2). Overall, these findings suggest that although Php5 and Moc3 colocalize at sites within constitutive heterochromatin domains, their colocalization in other genomic regions is only partial. This indicates potential shared as well as distinct functions of these factors.

Php5 forms a trimeric complex that localizes to heterochromatic repeats

Unlike Moc3, which is present only in Schizosaccharomycetes, Php5 is highly conserved from budding yeast (Hap5) and pathogenic yeast Candida albicans (Hap5) to humans (NF-YC). In other organisms, Php5 orthologs are known to associate with two other proteins, HsNF-YA/ScHap2/CaHap2 and HsNF-YB/ScHap3/CaHap3, to form a trimeric complex^{38,42,43,44,45}. Genetic evidence indicates that a trimeric complex likely also exists in S. pombe, but this has not been biochemically confirmed. We therefore focused on identifying factors that are associated with Php5. For this, we performed immuno-affinity purification of GFP-tagged Php5 (Fig. 2a) and analyzed the purified fraction by mass spectroscopy. In addition to Php5, these analyses identified NF-YA and NF-YB orthologs, Php2 and Php3, respectively (Fig. 2b). Therefore, Php5 associates with Php3^NF-YB and Php2^NF-YA to form the Php complex (PhpC), as observed in other yeasts and humans. Interestingly, our analyses also identified Moc3, as well as Atf1 and Pcr1, in the Php5 purified fraction (Fig. 2b). Moreover, several other proteins, including the components of RSC and Ino80 remodeler complexes, also co-purified with Php5 (Supplementary Fig. 2a).

Fig. 2: Php5 forms a trimeric complex that localizes to heterochromatic repeats.

a Php5 immunopurified fractions were prepared from untagged and Php5-GFP expressing cells and were probed with anti-GFP antibodies. b Immunopurified fractions from untagged and Php5-GFP expressing cells were subjected to mass spectrometry analysis. The percentage of total peptide coverage for the indicated immunopurified proteins is displayed. Results from one biological replicate are shown; see also Supplementary Fig. 2a. c, d ChIP-qPCR analysis of the indicated GFP-tagged TFs was performed at cenH (c) and dh (d) in wild-type cells. Data from 3 independent biological experiments are presented as the mean ± SD of the relative fold enrichment compared to the leu1 control locus. The p-values were calculated using a two-tailed paired t-test. e, f ChIP-seq analysis to determine the localization of the indicated GFP-tagged TFs at the silent mat locus (e) and centromere 1 (f) in wild-type cells. Note that the Php5-GFP ChIP-seq data is the same as in Fig. 1a, b, as the experiments were performed simultaneously. g, h Euler diagrams represent the number of loci bound by Php5, Php3, and Php2 and contain a high (g) or low (h) abundance of CCAAT boxes. i, j Heat maps displaying ChIP-seq enrichments of PhpC subunits at loci containing high (i) or low (j) abundance of CCAAT boxes. Source Data are provided as a Source data file.

We next investigated whether Php2 and Php3 bind at the heterochromatic regions where Php5 is found. ChIP-seq analyses revealed that both proteins co-localized with Php5 at the cenH and dh repeats adjacent to siRNA hotspots, and this result was confirmed by ChIP-qPCR (Fig. 2c–f). Notably, the enrichment level of Php2 at both locations was lower than that of the Php3 and Php5 subunits. The reason for this enrichment difference is unclear, as live-cell imaging indicated similar expression levels for all PhpC subunits (Supplementary Fig. 2b).

The trimeric NF-Y complex and its counterparts recognize a pentanucleotide motif called the CCAAT box^42,43,44. This led us to ask if the PhpC subunits are also associated with chromosomal regions containing this motif. To address this, we counted the CCAAT motifs on both strands within -/+ 500 bp of the peak center, and as previously proposed, motifs separated by ≤25 bp were counted as single ones⁴⁶. A cumulative total of 920 genomic binding peaks were found for Php2, Php3, and Php5. Notably, approximately 73% (673/920) of Php2, Php3, and Php5 peaks correspond to genomic regions containing more than one CCAAT box, including those at the cenH and dh loci (Fig. 2g, h and Supplementary Fig. 2c). The levels of Php proteins were relatively higher at loci where the number of CCAAT boxes was greater (Fig. 2i, j and Supplementary Data 3). In particular, CCAAT boxes were especially enriched at loci simultaneously occupied by all three PhpC subunits (Fig. 2g). For example, PhpC subunits localize to the CCAAT box-containing locus sme2, a lncRNA essential for sexual development⁴⁷, and to the promoter regions of known target genes such as cyc1⁺ and pcl1⁺ (Supplementary Fig. 2d, e)^38,48. Taken together, these results indicate that Php2^NF-YA, Php3^NF-YB, and Php5^NF-YC form a trimeric complex that localizes to numerous loci containing the CCAAT box motif, including the cenH and dh heterochromatic repeat elements.

TF binding to heterochromatic repeats is interdependent

We next asked if PhpC and Moc3 independently localize to heterochromatic regions or if their localization is mutually dependent. We performed ChIP-seq and ChIP-qPCR analyses of PhpC subunits and Moc3 in strains where one of the other factors was deleted. Consistent with the fact that Php2/3/5 forms a complex, loss of Php3 localization at cenH and dh elements was observed in the absence of php2 or php5 (Fig. 3a, b, and Supplementary Fig. 3a). Likewise, the localization of Php2 and Php5 at these sites decreased upon the deletion of any other PhpC subunit (Supplementary Fig. 3b, c). Interestingly, we also observed that Moc3 occupancy at both cenH and dh heterochromatic regions was contingent upon PhpC (Fig. 3a, c and Supplementary Fig. 3d). Moreover, the localization of PhpC subunits was compromised in moc3∆ cells (Fig. 3a, b and Supplementary Fig. 3b, c). The nuclear localization of PhpC and Moc3 persisted upon loss of any of the other factors (Supplementary Fig. 3e), consistent with the fact that these proteins still independently localize to a significant fraction of euchromatic locations (Supplementary Fig. 3f–h and Supplementary Data 4). These findings suggest that the TFs act in a cooperative manner to localize to heterochromatic repeat elements.

Fig. 3: The binding of PhpC and Moc3 to heterochromatin is interdependent and requires CCAAT boxes.

a ChIP-seq analysis of GFP-tagged Php3 and Moc3 at the silent mat locus and centromere 1 in the indicated strains. b, c Fold enrichments of GFP-tagged Php3 and Moc3 were determined by ChIP-qPCR analysis at cenH and dh. d ChIP-seq analysis of GFP-tagged Php3 and Moc3 expressed in wild-type or CCAAT^mut strains. CCAAT boxes are indicated by vertical yellow lines in the schematics above, and the asterisk represents the location of two mutated CCAAT boxes. e, f Fold enrichments of GFP-tagged Php3 and Moc3 were determined by ChIP-qPCR analysis at cenH and dh. Data from 3 independent biological experiments are presented as the mean ± SD of the relative fold enrichment compared to the leu1 control locus. The p-values were calculated using a two-tailed paired t-test (b, c, e, f). WT Moc3-GFP and Php3-GFP ChIP-Seq data are replotted from Figs. 1a, b and  2e, f(a, d). Source data are provided as a Source Data file.

The CCAAT box is required for PhpC heterochromatin localization

The regions within cenH and dh elements where TFs bind contain two CCAAT boxes that are separated by approximately ~60 bp. To investigate the significance of these elements for the heterochromatin association of TFs, we modified both CCAAT boxes in cenH to ATGAC (Supplementary Fig. 3i, hereafter referred to as CCAAT^mut). ChIP-seq and ChIP-qPCR analyses revealed a drastic decrease in the binding of PhpC subunits at cenH in CCAAT^mut cells (Fig. 3d, e and Supplementary Fig. 3j), whereas the localization of PhpC at the centromeric dh elements (Fig. 3d, e and Supplementary Fig. 3j) and euchromatic locations (Supplementary Fig. 3k, and Supplementary Data 4), where CCAAT boxes remained unchanged, were unaffected. Consistent with our results showing that PhpC is required for Moc3 localization at heterochromatic repeats, CCAAT^mut cells were defective in Moc3 binding at cenH (Fig. 3d, f). These results emphasize the importance of CCAAT boxes in guiding TFs to specific sequences within a heterochromatin domain at the silent mat region.

Heterochromatic localization of TFs is independent of the cell cycle

Previous studies have shown that transcription from heterochromatic repeats occurs predominantly during S-phase^23,24, suggesting that transcription factors may access repeat loci specifically during this cell cycle phase. However, our ChIP analyses were conducted in asynchronously growing cells, making it difficult to determine if only the cells in the S-phase show TF enrichment at heterochromatic repeats. The human counterpart of PhpC, NF-Y, remains bound to chromatin during mitosis⁴⁹. To address whether PhpC and Moc3 remain similarly bound to chromatin, we performed ChIP-seq analyses in cells that carry a cold-sensitive mutation in the β-tubulin gene (nda3-KM311), which causes mitotic prophase arrest at low-temperature⁵⁰. Php3 and Moc3 continued to associate with heterochromatic repeat elements even when cells were shifted to low temperature (Fig. 4a–d and Supplementary Fig. 4a, b). Furthermore, the genome-wide distribution of Php3 and Moc3 remained largely unchanged in mitotic-arrested cells (Supplementary Data 4), indicating that PhpC and Moc3 maintain their association with heterochromatin outside of S-phase.

Fig. 4: The histone fold domain of PhpC is pivotal for heterochromatin localization.

a, b ChIP-seq analysis of Php3 and Moc3 localization at the silent mat locus (a) and centromere 1 (b) in wild type, clr4∆, and nda3-KM311 strains. WT Moc3-GFP and Php3-GFP ChIP-seq data are replotted from Figs. 1a, b and 2e, f. Vertical yellow lines designate CCAAT boxes. Note that the highlighted CCAAT-enriched region is accessible to Php3 in clr4∆ cells. The small black line denotes the dg-annealing primer used in Supplementary Fig. 4c. c, d ChIP-qPCR analysis of GFP-tagged Php3 or Moc3 was performed at cenH (c) and dh (d) in the indicated strains. e, f ChIP-seq analysis of Php3 localization at the silent mat locus (e) and centromere 1 (f) in wild type and php3^L12R strains. g ChIP-qPCR analysis of GFP-tagged Php3 was performed at cenH and dh in indicated strains. h Representative live-cell images of strains expressing GFP-tagged Php3 or Php3^L12R from two independent experiments with similar results. Data from 3 independent biological experiments are presented as the mean ± SD. The p-values were calculated using a two-tailed paired t-test (c, d, g). Source data are provided as a Source Data file.

We further investigated the impact of heterochromatin on TF occupancy across heterochromatic regions. PhpC and Moc3 peaks at cenH and dh elements were mostly unchanged in cells lacking Clr4^Suv39h (Fig. 4a–d and Supplementary Fig. 4a, b). Interestingly, the absence of Clr4^Suv39h resulted in the emergence of additional minor peaks of PhpC, and its binding extended across dh and part of the dg repeat at centromeres (Fig. 4b and Supplementary Fig. 4c). However, global Php3 and Moc3 binding at gene promoters, lncRNAs, and other ncRNAs remained unchanged in clr4∆ cells (Supplementary Data 4). Therefore, heterochromatin disruption allows the binding of transcription factors to other sites within heterochromatic regions.

TF can localize to heterochromatin without HATs and chromatin remodelers

Histone acetyltransferases (HATs) and chromatin remodelers play a key role in regulating chromatin accessibility and facilitating the binding of TFs to target gene loci⁵¹. To determine if HATs are required for TF localization at heterochromatic loci, we investigated the distribution of Php3 and Moc3 in cells devoid of both Gcn5, a subunit of the SAGA complex, and Mst2, a MYST acetyltransferase. Interestingly, we found that in those cells, both TFs can still localize to cenH and dh heterochromatic repeat elements, albeit with a slight reduction (Supplementary Fig. 4d, e). This result suggests that HAT activity has a minimal impact on TF localization to heterochromatin regions.

Since we identified components of the RSC and Ino80 chromatin remodelers in purified Php5 fractions, we also studied the possible roles of remodelers in TF heterochromatic binding (Supplementary Fig. 2a). Specifically, we assessed TF localization in cells lacking the RSC subunit Rsc1, the SWI/SNF ATP-dependent chromatin remodeler Snf22 and/or bearing a mutation in the ATPase domain of Ino80 (i.e., ino80^K873A). In all cases, we observed that Php3 and Moc3 binding to cenH and dh regions were only marginally reduced but nevertheless not abolished (Supplementary Fig. 4f–i). These findings suggest that PhpC and Moc3 can access heterochromatic regions independently of the HATs and chromatin remodeling factors examined.

Histone-fold domains in PhpC subunits are vital for accessing heterochromatin

The PhpC subunits Php3 and Php5 contain H2B- and H2A-like histone-fold domains (HFDs), respectively (Supplementary Fig. 5a). We constructed strains bearing mutations of conserved residues of Php3 (L12R) and Php5 (A112D, R113D) HFDs to investigate the role of these domains in TF localization to heterochromatic loci. Remarkably, we observed that in both mutants, the binding of PhpC subunits Php3 and Php5 at cenH and dh heterochromatic repeats was drastically reduced (Fig. 4e–g, Supplementary Fig. 5b), while the nuclear localization of both TFs was preserved (Fig. 4h, Supplementary Fig. 5c). Indeed, site-specific ChIP-qPCR analysis revealed a ~ 90% reduction in Php3 and a ~ 80% reduction in Php5 in HFD mutants compared to WT (Fig. 4g, Supplementary Fig. 5b). On the other hand, the global binding of Php3 at gene promoters, lncRNAs, and other ncRNAs, was reduced but not abolished in HFD mutant cells (Supplementary Fig. 5d, e, Supplementary Data 4). Therefore, HFDs of PhpC subunits are necessary to facilitate TF accessibility to heterochromatic regions.

TFs promote RNAPII transcription of cenH heterochromatic repeats

As mentioned above, RNAPII transcription of heterochromatic repeat elements is crucial for heterochromatin nucleation. Since PhpC and Moc3 localize to heterochromatic repeats, we investigated if these TFs promote transcription at these sites. To avoid complications associated with the analysis of multicopy repeat elements, we exploited previously described sequence polymorphisms¹⁹ to focus primarily on transcripts derived from the cenH element. In wild-type (WT) cells, cenH transcripts are challenging to detect. To overcome this, we utilized cells lacking Clr4^Suv39h, wherein transcription is enhanced and transcripts are easier to detect. As expected, we observed increased cenH transcript levels in clr4∆²⁶ (Fig. 5a, b and Supplementary Fig. 6a, b). Interestingly, the lack of PhpC significantly decreased cenH transcript levels in clr4∆ (Fig. 5a, b). Similar changes were observed in moc3∆ (Fig. 5a and Supplementary Fig. 6b)⁵². Moreover, introducing CCAAT^mut into clr4∆ cells led to a notable reduction in cenH transcript levels (Fig. 5a and Supplementary Fig. 6a). These results were confirmed by RNA-seq analyses. Notably, the loss of PhpC and Moc3, as well as the CCAAT^mut mutation in Clr4-deficient cells, caused a substantial reduction in transcripts originating from the bottom strand of the cenH element, while the levels of transcripts from the top strand remained largely unchanged (Fig. 5c, d). These findings indicate that TFs play a significant role in driving transcription of the bottom strand of the cenH element, suggesting that their binding defines an active promoter responsible for transcribing downstream cenH sequences.

Fig. 5: Cooperative function of PhpC and Moc3 promotes cenH transcription.

a RT-PCR was performed to analyze cenH transcripts in the indicated strains, with leu1 serving as the control for +RT and -RT reactions. The primer binding site at cenH is indicated by the black line. b RT-qPCR analysis of cenH transcripts in the indicated strains. Relative expression was compared to the leu1 control locus. c RT-PCR was performed to analyze cenH top and bottom strand transcripts in the indicated strains, with leu1 used as the control for +RT and -RT reactions. The top and bottom strand-specific primer binding site at cenH is indicated by the red and yellow lines, respectively. Representative results from two independent experiments are shown. d RNA-seq expression profile of the cenH top and bottom strand in the indicated strains. Reads were uniquely mapped to cenH. e, f RNA polymerase II (8WG16) occupancy at cenH was determined by ChIP-seq (e) and ChIP-qPCR (f) analyses in the indicated strains. Reads were uniquely mapped to cenH. For f, ChIP data is presented as relative fold enrichment compared to the tRNA control locus. Data from 3 independent biological experiments are presented as the mean ± SD. The p-values were calculated using a two-tailed paired t-test (b, f). Source data are provided as a Source Data file.

We then investigated the role of TFs in RNAPII occupancy at cenH. ChIP-seq analysis of the Rpb1 subunit of RNAPII was performed, and the results were confirmed by site-specific ChIP-qPCR. RNAPII occupancy increased at cenH in clr4∆ (Fig. 5e, f and Supplementary Fig. 6c). Removal of PhpC or Moc3, or the introduction of CCAAT^mut mutation in clr4∆ cells, resulted in a considerable reduction in RNAPII levels at cenH (Fig. 5e, f). Similar results were obtained using an antibody against RNAPII CTD Ser-2P (Supplementary Fig. 6c). As a control, we also examined the impact of php2∆ at a known Php target and confirmed reduced transcription and loss of RNAPII occupancy independent of Clr4^Suv39h (Supplementary Fig. 6d, e). Overall, these findings demonstrate the importance of TFs in promoting cenH transcription by RNAPII.

TF-mediated cenH transcription supports Swi6^HP1-independent siRNA production

Since transcription of heterochromatic repeats by RNAPII is necessary for the generation of siRNAs by the RNAi machinery, we investigated whether TF binding to cenH is also crucial for siRNA production. As described above, two distinct mechanisms engage RNAi machinery to process repeat transcripts. In addition to H3K9me and Swi6^HP1 recruiting RITS and RDRC to facilitate the conversion of repeat transcripts into siRNAs by Dcr1^26,31,32,33, RNAi machinery can also be targeted to repeat-derived RNAs via a Swi6^HP1-independent mechanism³².

To determine the impact of php3∆, moc3∆, or CCAAT^mut on siRNA production from cenH, we conducted small RNA sequencing analysis in the presence or absence of Swi6^HP1. To rule out the contribution of centromeric siRNA due to the homology between cenH and dg/dh repeats, we analyzed only the reads that uniquely map to cenH. Single mutant cells carrying php3∆, moc3∆, or CCAAT^mut showed little or no change in cenH siRNAs compared to WT cells (Fig. 6a). On the other hand, loss of Swi6^HP1, which causes an increase in heterochromatic repeat transcription (Supplementary Fig. 7a), showed a corresponding increase in cenH siRNAs. However, combining swi6∆ with either php3∆, moc3∆, or CCAAT^mut led to a severe reduction or loss of siRNAs mapping to cenH (Fig. 6a). Together, these results suggest that while Swi6 can mediate siRNA generation from the top strand, TF-driven transcription of the cenH bottom-strand is required to produce Swi6^HP1-independent siRNAs.

Fig. 6: PhpC- and Moc3-mediated cenH transcription is crucial for producing Swi6^HP1-independent siRNAs.

a, b siRNA-seq profiles at the cenH region in the indicated strains. Note that the reads were uniquely mapped to cenH. c, d ChIP-qPCR analysis of H3K9me3 enrichment at the silent mat locus in the indicated strains. Data is presented as the mean of 2 independent biological experiments. The numbered lines in the schematic denote the location of the primers. Source data are provided as a Source data file.

Processing of TF-generated transcripts into siRNAs requires the spliceosome and cryptic intron splice sites

The observation that PhpC-mediated transcription of the bottom cenH strand is required for Swi6^HP1-independent production of siRNAs led us to investigate the unique features of this transcript strand that may help recruit the RNAi machinery. Closer examination revealed that a region of cenH sharing homology to dh repeats contains multiple cryptic introns, which have been implicated in RNAi-mediated heterochromatin formation^53,54,55. RNAi machinery interacts with the splicing factors that are necessary for proper heterochromatin formation⁵⁶. Notably, these cryptic introns are predominantly found in the bottom transcript strand (Supplementary Fig. 7b).

We wondered whether the presence of multiple cryptic introns in these TF-driven transcripts engages splicing factors implicated in siRNA production by the RNAi machinery. To investigate this, we tested if splicing machinery is required for Swi6^HP1-independent production of siRNAs, similar to the TFs involved in bottom-strand transcription. A temperature-sensitive mutation in the conserved spliceosome component Cwf10^EFTUD2 (cwf10-1)⁵⁶ alone caused only a modest defect in the production of siRNAs that map to cenH (Supplementary Fig. 7c). On the other hand, when cwf10-1 was combined with swi6∆, the double mutant cells showed a dramatic reduction in cenH siRNAs (Supplementary Fig. 7c).

We next mutated the splice sites of all five cryptic introns found in the cenH bottom strand (Supplementary Fig. 7b, hereafter referred to as crypticSS^mut). Whereas mutating the cryptic intron splice sites alone did not cause major changes in the levels of siRNAs mapping to cenH, combining crypticSS^mut with swi6∆ resulted in a severe reduction of cenH siRNAs (Fig. 6b). Together, these results suggest that cryptic introns in the TF-driven cenH transcripts engage RNAi machinery via splicing machinery to generate Swi6^HP1-independent siRNAs.

H3K9me targeting to cenH requires TF-driven transcription

The siRNAs are believed to provide specificity for RNAi-mediated targeting of Clr4^Suv39h to methylate H3K9². At the silent mat region, heterochromatin targeted to cenH by RNAi spreads in cis across the entire silenced domain via a mechanism involving Swi6^HP1 and the read-write activity of Clr4^{Suv39h 13,16}. In cells defective in heterochromatin spreading, such as the swi6 mutant, H3K9me is mainly restricted to the cenH region and fails to spread to surrounding sequences¹⁶.

We explored the contribution of the TF-driven transcription of cenH in heterochromatin nucleation by mapping H3K9me3 throughout the silent mat region in the swi6∆ background. As expected, the loss of Swi6^HP1 effectively limited H3K9me3 to the cenH region (Fig. 6c). Remarkably, targeting of H3K9me3 to cenH was severely affected when swi6∆ was combined with CCAAT^mut or crypticSS^mut (Fig. 6c, d). Moreover, loss of any subunit of PhpC or Moc3 in swi6∆ cells yielded similar results for both H3K9me3 and H3K9me2 (Fig. 6c and Supplementary Fig. 7d). Consistent with our results showing that TFs are dispensable for Swi6^HP1-dependent production of siRNAs mapping to cenH (Fig. 6a, b and Supplementary Fig. 7c), loss of these factors, CCAAT^mut or crypticSS^mut alone had no major impact on heterochromatin assembly in cells expressing functional Swi6^HP1 (Fig. 6c, d and Supplementary Fig. 7d).

To further validate our findings, we assessed the impact of combining CCAAT^mut with swi6∆ on the localization of the Clr4^Suv39h complex. Raf2, a subunit of the Clr4^Suv39h complex, is normally distributed throughout the silent mat region¹³. However, its localization was restricted to the cenH nucleation site in swi6∆ cells (Supplementary Fig. 7e). Consistent with the results observed for H3K9me3, loss of TFs or mutating CCAAT in swi6∆ cells drastically reduced the Raf2 loading at cenH (Supplementary Fig. 7e). Together, these results indicate that TF-mediated cenH transcription plays an important role in cryptic intron-dependent generation of siRNA and heterochromatin nucleation, which is unveiled upon loss of Swi6^HP1.

TFs promote de novo establishment of a repressive heterochromatin domain

Swi6^HP1-independent production of siRNAs may play a role in the de novo targeting of H3K9me and heterochromatin assembly. To investigate this, we studied whether PhpC binding to cenH is necessary for the initial formation of heterochromatin at the silent mat region using a sensitized reporter system. In mat1-M cells lacking a local silencer element (REII), the disruption of heterochromatin assembly leads to derepression of the mat2-P locus. This results in abnormal sporulation (haploid meiosis) due to the simultaneous expression of M (mat1M) and P (mat2P) mating-type information in haploid cells. Cells undergoing haploid meiosis exhibit dark staining upon exposure to iodine vapor, while WT colonies stain yellow. Additionally, the level of expression of ura4⁺ inserted near mat2P (mat2P::ura4⁺) offers a further assessment of heterochromatic silencing⁵⁷.

To conduct the heterochromatin establishment assay, we cultured WT and CCAAT^mut cells, both carrying the REII∆ mat2P::ura4⁺ reporter, in the presence of the histone deacetylase inhibitor trichostatin A (TSA). TSA is known to erase H3K9me marks and relieve gene silencing³⁶. Following the removal of TSA, WT cells successfully reestablished silencing at the silent mat region (Fig. 7a). However, CCAAT^mut cells were deficient in reestablishing heterochromatic silencing, as indicated by the expression of mat2P::ura4⁺ and the dark iodine staining of colonies (Fig. 7a). This is consistent with lower levels of H3K9me3 in CCAAT^mut cells compared to WT cells (Fig. 7b). Conversely, both WT and CCAAT^mut cells cultured without TSA maintained reporter gene silencing, forming colonies that stained yellow when exposed to iodine vapor (Fig. 7a), consistent with similar levels of H3K9me3 (Fig. 6c).

Fig. 7: PhpC contributes to de novo heterochromatin establishment.

a 10-fold serial dilution assay of wild-type (WT) and CCAAT^mut strains on the indicated medium before and after trichostatin A (TSA) treatment and washout. Iodine staining was performed on non-selective (N/S) plates. b ChIP-qPCR analysis of H3K9me3 enrichment at the silent mat locus in the indicated TSA-treated strains. Data is presented as the mean of 3 independent biological experiments. Numbered lines in the schematic indicate the location of the primers. c Representative live-cell images of WT and CCAAT^mut strains harboring the mat2P::GFP reporter at the indicated time points (generations) after TSA treatment and washout. d Quantification of the cell fraction in the “ON” state (e.g., GFP-positive) at the indicated time points is shown. N = 260–651 cells for each data point. Source data are provided as a Source Data file.

Next, we utilized a fluorescence-based single-cell silencing assay to investigate the kinetics of de novo heterochromatin establishment in CCAAT^mut cells that were defective in TF binding at cenH. Replacing the ura4⁺ reporter with a GFP gene enables accurate measurement of the fraction of GFP-positive (“ON”) cells after TSA washout. As expected, TSA treatment induced GFP expression in most cells in both WT and CCAAT^mut strains (Fig. 7c). Following TSA washout, WT cells reestablished GFP reporter silencing more rapidly than the CCAAT^mut mutants, which showed compromised reestablishment of heterochromatic silencing, as indicated by the significant number of cells that continued to express GFP, even after more than 20 generations (Fig. 7c, d). Similar results were observed in CCAAT^mut moc3∆ cells (Supplementary Fig. 8a, b). It is noteworthy, however, that the proportion of CCAAT^mut cells with a silenced mat region increased with each generation in a stochastic manner, suggesting that heterochromatin eventually forms at this locus, likely through alternative nucleation mechanisms^36,58.

Together, these results emphasize the important role of TFs in promoting the establishment of heterochromatin. We find that TFs drive the transcription of cryptic intron-containing cenH transcripts, which are necessary for the Swi6^HP1-independent production of siRNAs involved in targeting H3K9me and heterochromatin assembly.

Discussion

Heterochromatin is generally repressive, preventing the underlying DNA sequences from being accessed by transcriptional machinery. Nonetheless, RNAPII transcription of heterochromatic repeat elements occurs and is necessary for recruiting heterochromatin assembly proteins to nucleation sites, such as through RNAi machinery^{2,21,22,23,24,59}. Despite advances in understanding heterochromatin assembly pathways, exactly how heterochromatic repeats are transcribed remains to be fully elucidated. In this study, we demonstrate that two distinct transcription factors co-bind to specific sites within repeat elements in constitutive heterochromatin domains. Our analyses reveal that the binding of these transcription factors is crucial for transcribing RNAs that contain multiple cryptic introns, engaging RNAi machinery through spliceosome components, and triggering siRNA production for de novo heterochromatin assembly.

We show that in addition to a Zn-finger domain-containing protein Moc3, a trimeric transcription factor complex PhpC, related to the mammalian NF-Y^44,45, drives transcription of one of two strands of its target heterochromatic repeat elements. A recent study implicated Moc3 as a positive regulator of dh element transcription but could not detect its binding at the target repeats in wild-type cells⁵². This study suggested that TF binding is blocked once heterochromatin is assembled. Our results contrast with these previous findings. We show that both Moc3 and PhpC can access cenH and dh repeat elements, irrespective of the heterochromatin state. Indeed, comparable levels of both TFs are detected at these loci in WT and clr4∆ cells. Furthermore, we show that both TFs can infiltrate heterochromatin even in the absence of major HATs and chromatin remodelers. It is clear from these results that constitutive heterochromatin is not completely inaccessible to TFs that are required to promote the transcription of repeat elements.

How do TFs gain access to sequences within heterochromatin regions? Interestingly, the localization of PhpC and Moc3 to the heterochromatin region occurs in an interdependent manner, while this is not the case for binding to euchromatin regions. Loss of PhpC or mutating its binding site (CCAAT) dramatically impairs Moc3 binding to the cenH element. Similarly, we find that moc3∆ cells are defective in PhpC localization at heterochromatic repeats. These findings suggest that the collaborative action of these TFs is crucial for accessing heterochromatic regions. Interestingly, the PhpC subunits, Php3 and Php5, each possess an H2B/H2A-like HFD involved in histone-histone and histone-DNA interactions⁴⁶. Considering that the human counterpart of PhpC, NF-Y, forms a nucleosome-like structure to gain chromatin accessibility^60,61,62, it is plausible that PhpC uses HFDs to displace nucleosomes to access the CCAAT box⁶³. In this regard, we show that mutating conserved residues in the HFDs of Php3 or Php5 disrupts their localization to heterochromatin. TF heterochromatin accessibility is likely further facilitated by binding of Moc3 to DNA via its Zn-finger domain. In this sense, PhpC and Moc3 function as “pioneer transcription factors” to displace nucleosomes to access underlying DNA sequences⁶⁴. Regardless of their binding mode to heterochromatic sequences, our analyses show that TFs act directly to promote repeat transcription, as opposed to causing indirect effects resulting from misregulation of their target gene expression. Indeed, we show that mutating the CCAAT box, which affects both PhpC and Moc3 binding, impairs cenH transcription similarly to what is observed in the absence of these TFs.

Previous studies have shown that RNAPII transcription of the heterochromatic repeat elements cenH and dh occurs preferentially during the S-phase of the cell cycle^23,24. This S-phase requirement for repeat transcription is dictated by the heterochromatin machinery involved in transcriptional silencing². In cells defective in heterochromatin assembly, cenH and dh elements are transcribed throughout the cell cycle²³. The restriction of heterochromatic repeat transcription to the S-phase contrasts with our findings that both PhpC and Moc3 can bind to heterochromatic repeats outside of the S-phase in wild-type cells. This suggests that heterochromatin obstructs additional steps necessary for RNAPII transcription beyond TF binding. Heterochromatin-associated HDACs, which are believed to prevent opening of chromatin by restricting access of chromatin remodelers to heterochromatin^65,66 and counteracting an anti-silencing factor Epe1⁶⁷, may prevent the assembly of active RNAPII transcriptional complexes, thereby inhibiting transcription initiation and/or elongation. In this regard, the perturbation of heterochromatin during the S-phase, such as during DNA replication, may provide a “window of opportunity” for TFs to recruit RNAPII complexes and promote transcription of heterochromatic repeats.

Our findings reveal that the generation of cenH transcripts via TFs is crucial for producing siRNAs independent of Swi6^HP1 and for the de novo targeting of H3K9me by Clr4^Suv39h. In cells lacking Swi6^HP1, loss of PhpC, Moc3, or mutation of the CCAAT box, which impair the transcription of the cenH bottom strand, significantly impede siRNA production. The bottom strand cenH transcript contains multiple cryptic introns, and their processing into siRNAs by RNAi factors requires splicing machinery. Importantly, we show that mutating cryptic intron splice sites in cells lacking Swi6^HP1 results in drastic reduction in siRNA production. These observations, together with previous studies², suggest a model for the initial generation of siRNAs that trigger de novo heterochromatin assembly at target repeat loci. The spliceosome, bound to TF-driven cryptic intron-containing transcripts, likely provides specificity for recruiting RNAi factors to repeat transcripts for initial siRNA production (Fig. 8). In this regard, it has been shown that stalling of the spliceosome at cryptic introns recruits RDRC^56,68, likely in collaboration with a homolog of mammalian Ars2 (named Pir2 in S. pombe)⁵⁴ to generate double-stranded RNA substrates for Dcr1. The siRNAs produced through this process are loaded onto Ago1-containing RITS, which facilitates the recruitment of Clr4^Suv39h to methylate H3K9. Once heterochromatin is established, TFs remain associated with repeat elements for continuous production of siRNAs. However, Swi6^HP1 bound to H3K9me can independently recruit RDRC to further promote the RNAi-mediated processing of bidirectional repeat transcripts and amplify siRNA production^31,32,33. This model explains why mutating TFs or their recruitment sites at cenH in cells with functional Swi6^HP1 does not significantly impact siRNA production and heterochromatin assembly.

Fig. 8: Model showing the role of the histone fold domain-containing PhpC^NF-Y in heterochromatic repeat transcription and siRNA generation.

The top panel illustrates how the histone fold domain (HFD) containing the trimeric TF complex PhpC, which recognizes underlying CCAAT boxes at heterochromatic regions such as cenH, binds to heterochromatic repeats in cooperation with Moc3. The bottom panel depicts how PhpC facilitates the transcription of the cenH bottom strand containing multiple cryptic introns (depicted as red rectangles). These introns likely stall the spliceosome that then recruits the RNAi machinery, particularly the RNA-dependent RNA polymerase Rdp1-containing protein complex RDRC, initiating siRNA production. The siRNAs are loaded onto the Ago1-containing RITS complex. Together with the protein Stc1, RITS recruits the Clr4^Suv39h methyltransferase, promoting H3K9me and leading to de novo establishment of heterochromatin, independently of Swi6^HP1. Once H3K9me is established, Swi6^HP1 binds and also recruits the RNAi machinery via Ers1, maintaining a pool of siRNAs that further facilitate heterochromatin assembly. Consequently, the simultaneous disruption of both pathways results in the loss of siRNAs and H3K9me.

Our ability to specifically block cenH bottom-strand transcription has provided crucial insight into how cells initiate siRNA production to nucleate heterochromatin via mechanisms involving cryptic introns and the splicing machinery. Other mechanisms, such as the generation of Dcr1-independent but Ago1-dependent small RNAs, referred to as primal RNAs, might also contribute⁶⁹. However, considering the studies showing that H3K9me levels are similar in ago1∆ and dcr1∆ cells, and even in ago1∆ dcr1∆ double mutant cells^70,71, the role of primal RNAs is likely only minor.

The RNA-based targeting of heterochromatin is widely conserved across various systems. RNAi, an ancient defense and regulatory mechanism, plays a crucial role in forming heterochromatin at repetitive DNA elements and other loci in various organisms, including C. elegans and plants^72,73,74,75. This suggests the potential existence of factors analogous to the TFs described here, which may infiltrate repressive chromatin domains to mediate transcription at RNAi target loci, thus maintaining a steady pool of siRNAs via mechanisms that also require splicing factors^68,76,77,78. Indeed, several transcription factors have been identified as binding to heterochromatic repeats in higher eukaryotes^79,80,81,82. Although these TFs generally repress transcription, it is plausible that their binding promotes transcription of heterochromatic repeats under specific conditions to facilitate RNA-mediated heterochromatin formation. In this regard, we note that multiple pathways utilize RNAPII transcription to nucleate heterochromatin^{2,83,84,85,86}. Besides the RNAi machinery, evolutionarily conserved nuclear RNA processing factors, including Enhancer of Rudimentary homolog (ERH), are required for H3K9me3 and silencing of both repetitive DNA elements and lineage-specific genes^{2,15,83,84,87}.

Future studies will investigate whether TFs capable of infiltrating repressive chromatin domains, including “pioneer transcription factors”, contribute to maintaining heterochromatin domains through RNA-based mechanisms.

Methods

Strains and growth conditions

Strains were created through genetic crosses or a PCR-based method. The addition of epitope tags or gene deletions was confirmed using gene-specific PCR primers. The expression and nuclear localization of GFP-tagged transcription factors were confirmed through live-cell imaging.

The construction of the strain harboring GFP at the XbaI site next to mat2-P(Bg-Bs) was previously described⁶⁵. The CCAAT^mut and crypticSS^mut strains were generated using CRISPR, and derived strains were generated by a genetic cross. The strains were cultured in rich yeast extract plus adenine (YEA) media using standard protocols unless otherwise specified. Oligonucleotides and strains used in this study are listed in Supplementary Data 5.

Chromatin immunoprecipitation (ChIP)

The strains with epitope tags were grown until OD₆₀₀ 0.5. The cultures were then shifted to 18 °C for 2 hours and cross-linked with 3% paraformaldehyde (Sigma, Cat# P6148) for 30 minutes at 18 °C. Cells were washed with 1x PBS and crosslinked with dimethyl adipimidate dihydrochloride (DMA; Sigma Cat#285625) for 45 minutes at room temperature. Subsequently, the cells were washed again with 1x PBS and lysed with zirconia beads plus lysis buffer (50 mM HEPES/KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% deoxycholate, and protease inhibitors, Roche, Cat#183617001). The cell lysates were sheared using a Bioruptor-300 (Diagenode) to an approximate size of 300–600 base pairs. 3 μg of anti-GFP antibody (Abcam, Cat#ab290) was used for GFP immunoprecipitation. For RNA polymerase II immunoprecipitation, 25 μl of anti-Pol II antibody (8WG16, Santa Cruz Biotechnology, sc56767) was used. For Raf2-Myc immunoprecipitation, 25 μl of c-Myc antibody (Santa Cruz Biotechnology, Cat#9E10) was used. For H3K9me immunoprecipitation, 2 μl of anti-H3K9me2 (Abcam, Cat#ab115159) and anti-H3K9me3 (Abcam, Cat#ab8898) antibodies were used. Chromatin extracts were incubated with antibodies overnight at 4 °C, and an equal volume (25 μl) of Protein A plus (Pierce, Ref#22811) or Protein G (Invitrogen, Ref#15920010) agarose beads were added for four hours at 4 °C. For FLAG immunoprecipitation, 50 μl of anti-FLAG M2 affinity gel (Sigma, Cat#A2220) was directly incubated with chromatin extract for 4 hours at 4 °C. The beads were then washed twice with lysis buffer, once with lysis buffer containing 0.5 M NaCl, once with wash buffer (10 mM Tris-HCl pH8, 0.25 M LiCl, 1 mM EDTA, 0.5% NP-40, 0.5% DOC), and once in TE buffer, pH 8. The chromatin–antibody complex was eluted using TES buffer (50 mM Tris-HCl pH 8, 10 mM EDTA, 1% SDS) and, along with the whole-cell extract (WCE) input, was de-crosslinked by heating to 65 °C overnight and purified using MinElute Spin Columns (Qiagen Cat# 28004). For qPCR analyses, iTaq Universal SYBR Green Supermix (Bio-Rad, Cat# 1725120) was used. Delta-delta Ct normalization used leu1 as the reference gene in all cases unless indicated otherwise. The fold enrichment was further normalized to tagged/untagged in all cases unless indicated otherwise. Oligonucleotides used for ChIP-qPCR are listed in Supplementary Data 5.

Chromatin immunoprecipitation-sequencing (ChIP-seq)

Sequencing libraries were generated using NEBNext Ultra II DNA library prep kit for Illumina (Illumina, Cat# E7645) according to the manufacturer’s protocol. The library size was analyzed using an Agilent 4200 Tape Station system (Agilent). Samples were multiplexed, and single-end reads were sequenced on the Illumina MiSeq platform using the MiSeq reagent kit V3 according to the manufacturer’s protocol.

Protein purification and mass spectrometry

GFP-tagged Php5 was purified from 2 liters of cell culture. Cells expressing Php5-GFP or the untagged control were cultured at 30 °C. The cells were washed with water and then flash-frozen in liquid nitrogen. The cell pellets were thawed on ice and ground with glass beads in a Pulverisette 6 system (Labsynergy) in lysis buffer (20 mM HEPES–KOH pH 7.6, 20% glycerol, 500 mM NaCl, 2 mM MgCl2, 150 mM KCl, 0.1% IGEPAL, 1 mM dithiothreitol (DTT), 1 mM EDTA, Roche complete protease inhibitors cocktail and 1 mM PMSF). The lysate was then cleared by centrifugation at 27,000 g for 1 hour, and the supernatant was incubated with anti-GFP agarose beads (GFP-Trap Chromotek, Cat#AB_2631357) for 2 hours at 4 °C. The beads were thoroughly washed in BAC150 (20 mM HEPES–KOH pH 7.6, 20% glycerol, 2 mM MgCl2, 1 mM EDTA, 150 mM KCl, 0.1% IGEPAL, Roche complete protease inhibitors cocktail, and 1 mM PMSF) and then eluted with 200 μl of 0.2 M glycine (pH-2.5). 1/10 of the eluted proteins were probed with an anti-GFP antibody (Roche) to confirm the efficiency of the pull-down (Uncropped blots are shown in the Source Data file), and the rest of the eluted proteins were precipitated with 10% TCA and resuspended in the sample buffer. The proteins were resolved on a 4%–12% Bis-Tris Gel (Invitrogen, Cat#NP0321BOX) and visualized using SimplyBlue SafeStain (Invitrogen, Cat#465034). Protein bands were excised from the gel and subjected to mass spectrometry. All purification and mass spectrometry experiments were repeated twice. The excised gel bands were subjected to in-gel trypsin digestion at a 20 ng/μl concentration for effective peptide extraction. The samples were desalted using Pierce C18 spin columns (Thermo Fisher Scientific), dried, and resuspended in 0.1% trifluoroacetic acid before loading onto an Acclaim PepMap 100 C18 LC column (Thermo Fisher Scientific). This process was carried out with a Thermo Easy nLC 1000 LC system connected to a Q Exactive HF mass spectrometer (Thermo Fisher Scientific). Peptide elution occurred through a 5% to 36% acetonitrile gradient with 0.1% formic acid over 56 minutes at a 300 nL/min flow rate. The QEHF performed MS1 scans in the orbitrap at a resolution of 60,000, with a maximum injection time of 120 ms and an AGC target of 1e6. For MS2 scans, a normalized collision energy of 27 was applied at a resolution of 15,000, with a maximum injection time of 50 ms and an AGC target of 2e5. The raw MS data was analyzed with Proteome Discoverer 2.2 and SEQUEST HT software, targeting the UniProt S. pombe proteome database from the European Bioinformatics Institute (https://www.uniprot.org/proteomes/UP000002485). The parent ion mass tolerance was 10 ppm, while the fragment ion mass tolerance was set to 0.02 Da. The minimum peptide length was established at 6 amino acids, allowing for a maximum of two missed cleavages.

RNA isolation and cDNA preparation

Cells were cultured in a YEA medium under normal growth conditions and then transferred to 18 °C for 2 hours before total RNA isolation. The RNA was isolated by incubating yeast cells in hot phenol (pH 5.2) at 65 °C for 15 minutes, followed by three additional phenol-chloroform extractions⁸⁸. The RNA was precipitated using sodium acetate-ethanol and quantified using NanoDrop (Thermo). Approximately 50 μg of total RNA was treated with Turbo DNase I (Invitrogen, Cat# AM2238), and about 1 µg of total RNA was used for cDNA preparation using Oligo (dT) and hexamer primers following the manufacturer’s protocol.

qPCR and RT-PCR

Primers specific to cenH were used to assess expression (see Supplemental Data 5). We performed qPCR using iTaq Universal SYBR Green Supermix (Bio-Rad Cat# 1725120) on the QuantStudio3 platform (Thermo Fisher Scientific). For RT-PCR, we used Taq DNA polymerase (NEB, Cat#M0273L) and followed the manufacturer’s instructions. The leu1 gene served as the endogenous control for normalizing the transcript levels. Uncropped gels are shown in the Source Data file.

RNA-seq library preparation

Cells were cultured in YEA medium and transferred to 18 °C for 2 h before isolating total RNA using the hot phenol method. The RNA quality was checked (RIN > 8), and approximately 350 ng of total RNA was used for rRNA removal with the TruSeq Standard Total RNA LP kit (Illumina, Ref#15032615). Libraries were prepared using the NEBNext Ultra Directional RNA Library Prep kit (NEB, Cat#E7760) and sequenced on the NextSeq500 platform.

Small RNA-seq library preparation

Small RNA was purified from 4 OD₅₉₅ units of log phase cells using the MasterPure Yeast RNA Purification Kit (Lucigen Cat# MPY03100). Small RNAs (21-25nt) were excised after denaturing electrophoresis, and then ethanol precipitated and resuspended in DEPC-treated water. Adaptor ligation and PCR amplification were performed using the NEBNext Small RNA Library Prep Set for Illumina (NEB, Cat# E7300S). The final library was sequenced on the MiSeq platform.

TSA treatment and dilution assay

Cells were initially cultured in YEA media and subsequently treated with TSA (MedChemExpress, Cat#HY-15144, used at a final concentration of 35 μg/ml) for 60 hours. After washing out TSA, the cells were grown in PMG-ura media for 24 hours, then diluted to ~0.005 OD₆₀₀ into YEA media and incubated again for another 60 hours. Both untreated and TSA-treated cells were serially diluted and spotted on appropriate PMG plates. H3K9me3 ChIP-qPCR was performed as previously described. Data from three independent biological replicates are shown in the figures.

Microscopy analysis of TSA-treated cells

Cells were initially grown in YEA media and then treated with TSA for 24 hours (approximately six generations) using the same conditions as above. After treatment, cells were washed with water and cultured in YEA liquid media at 26 °C for 4 days. Daily dilution to 0.025OD₆₀₀ in fresh YEA media was performed to prevent the cells from entering the stationary phase. Under these conditions, the doubling time was approximately 3 hours. For imaging, the cells were mounted in 2% agarose. Twenty 0.35-μm z-sections were captured for GFP fluorescence using a Delta Vision Elite microscope (Leica) with a 100×1.35 NA oil lens (Olympus). Images were deconvolved and projected into 2D maximum-intensity images using SoftWorx (Leica). Analysis was done using ImageJ-based Fiji software.

Multiple sequence alignments

Protein sequences were downloaded from the respective database and aligned using Jalview software.

Cryptic-intron identification at cenH

cenH shares 96% homology with the cen2 dh element (SPRPTCENB.4). cen2 dh has six cryptic introns, which are detectable in wild-type (two cryptic introns) and rrp6Δ ago1Δ double mutant cells (four cryptic introns). We carefully searched for these intron splice sites at cenH and identified five putative intronic sequences. We then mutated the 5’ and 3’ splice sites, including the branch point; additionally, we also mutated a few potential splice sites at cenH loci. These mutations were confirmed by Sanger sequencing.

Statistics and reproducibility

The statistical significance of qPCR results was assessed using two-tailed paired t-tests. RT-PCR experiments were performed at least twice, yielding similar results, as shown in Fig. 5a and c. All fluorescence images in Figs. 4h, 7c, and Supplementary Figs. 2b, 3e, 5c, and 8a are from over 200 yeast cells.

Data processing and computational analysis

ChIP-seq data analysis

Single-end short reads were quality trimmed with fastp⁸⁹ and aligned using the BWA aligner⁹⁰ to the S. pombe V2 reference sequence⁹¹ in which the Pombase provided 20 kb mat contig was replaced with the sequence of the full 40 kb mat region (MAT). In addition, both the mat region on chromosome 2 (chrII 2,109,748–2,138,781), and the mat1M locus on the 40 kb MAT contig (MAT 4,489–5,615) were masked from alignment. Bedgraphs of ChIP enrichment were produced using the MACS2⁹² ‘callpeaks’ function to make broad calls with options ‘–nomodel–extsize 147’, followed by the MACS2 ‘bdgcmp’ function to compute fold enrichment (FE). Bedgraphs of FE of the IP over input divided by the FE of the untagged control over untagged input, referred to hereafter as ‘FE bedgraphs’, were used in downstream analysis. These bedgraphs were converted to a more convenient fixed-interval format by taking the mean signal over non-overlapping 10 bp windows.

TF site identification

A library of TF binding sites was taken from the ‘summits’ BED files generated for ChIPs of 16 TFs in WT backgrounds by MACS2 when run as described above. Each site in this collective set was then expanded by 25 bp symmetrically to produce a library of sites of uniform 50 bp width. Overlapping sites were subsequently merged to yield the final set of 2273 unique transcription factor binding sites that were used in the downstream analysis.

After site identification, the signal strength for each TF at each site was determined from the FE bedgraphs. Because both the width and central tendency of site occupancy varied considerably among the 16 TFs, the TF signal strength per site was taken as the maximum signal over an extended interval of 1 kb straddling the site center. Signal strengths per site for TFs in mutant backgrounds were calculated similarly.

RNA-seq data analysis

Single-ended short reads from RNA-seq experiments were quality-trimmed using fastp⁸⁹ and aligned using the STAR aligner⁹³. Variable interval bedgraphs, including both uniquely and multi-mapping reads and normalized to counts per million mapped reads, were generated by STAR and were further processed to produce 10 bp fixed interval versions in the manner described above for ChIP-Seq.

Visualization of genomic alignments

The Integrative Genomics Viewer⁹⁴ (v.2.13.1) was used to visualize bedgraphs of NGS alignments. The Integrated Genome Browser⁹⁵ was used to visualize small RNA SGR alignments.

Site-centered heat maps

ChIP signals for each of the 16 TFs in WT backgrounds were plotted across the subset of the 2273 sites originally assigned to the TF in the respective MACS ‘summits’ BED file. Signals were taken from FE bedgraphs and signal density was plotted across sites using a combination of the computeMatrix and plotHeatmap programs of the Deeptools suite⁹⁶.

Hierarchical clustering of the 16 TFs on the basis of the vector of signal strength at each of the 2273 TF sites was performed by applying the Ward algorithm to the Euclidian distances between vectors as implemented in the Orange data mining suite⁹⁷. Euler diagrams were produced using the ‘eulerr’ R package (https://cran.rproject.org/web/packages/eulerr/index.html), and the significance of site set overlap was assessed using the hypergeometric distribution as implemented in the ‘SuperExactTest’ R package (https://www.r-project.org). Heatmaps were produced using the ‘ComplexHeatmap’ R package⁹⁸. Gene enrichments were computed using the Orange bioinformatics package.

Small RNA-seq data analysis

Single-ended short reads were quality trimmed with fastp⁸⁹ and aligned to the S. pombe reference using Novoalign (www.novocraft.com/products/novoalign). SGR files, whose format allows the simultaneous display of read counts from both strands, were subsequently created from the Novoalign-generated BAM files using custom scripts (https://doi.org/10.5281/zenodo.14285969). SGRs were either filtered to remove multi-mapping reads or left unfiltered to include multi-mapping reads and subsequently normalized to counts per 10 million mapped reads (CP10M).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.