PIWI proximity proteome reveals Set1-mediated piRNA biogenesis for transposon silencing in telomere

16 Silencing complexes formed by PIWI -clade Argonaute (Ago) proteins and 17 PIWI-interacting RNAs (piRNAs) are essential guardians of genome integrity, 18 controlling the deleterious activities of transposable elements (TEs) in animal germline. 19 However, our understanding of PIWI-piRNA-directed TE silencing remains incomplete. 20 Here, we systemically characterize the proximity proteome of PIWI members, Piwi, 21 Aubergine (Aub), and Ago3 in the germline of Drosophila ovaries. Functional screening 22 identifies previously uncharacterized factors involved in TE silencing, including 23 H3K4me3 writer and transcriptional coactivator Set1. Transcriptome analysis reveals 24 that Set1 acts as an indispensable repressor for TEs , particularly those forming 25 telomeres. The involvement of Set1 in Piwi pathway is further supported by its critical 26 role in the production of antisense, TE-targeting piRNAs. Notably, catalytic activity of 27 Set1 is dispensable for TE silencing . Genome-wide chromatin binding analysis using 28 CUT&Tag demonstrates that Set1 preferentially associates with TE sequences and 29 localizes to subtelomeric piRNA cluster loci , suggesting a role in promoting piRNA 30 precursor transcription through direct binding . Collectively, these findings uncover a 31 noncanonical function of Set1 in Piwi-mediated TE silencing and telomere control in 32 germline nuclei. 33 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

34 Transposable elements (TEs) are mobile DNA sequences constituting a substantial part 35 of genomes in most organisms (Wells & Feschotte, 2020). Although TEs act as drivers 36 of genomic evolution and species diversification, the harmful potential as mutagens 37 requires the hosts to establish the control mechanisms. The control is particularly 38 important in germline cells where activated TEs accumulate damages on the inheritable 39 genomes and impair the production of functional gametes. In animals, TE control in 40 germline cells relies on silencing mediated by complexes of PIWI clade Argonaute 41 (Ago) family proteins and the sequence-specific guide, PIWI -interacting (pi)RNAs 42 (Wang et al, 2023b). 43 PIWI-piRNA-directed TE silencing has been studied in different animals 44 including the invertebrate model, Drosophila melanogaster. D. melanogaster possesses 45 three PIWI members including Piwi, Aubergine (Aub), and Ago3 , all of which are 46 expressed in germline cells. Piwi differs from the other two in that it functions in the 47 nucleus. Guided by piRNAs, Piwi complexes recognize the nascent transcripts from 48 target loci, mediating the (co -)transcriptional silencing and heterochromatin formation 49 (Le Thomas et al, 2013; Yu et al, 2015b; Sienski et al, 2015a; Mugat et al, 2020; Ariura et 50 al, 2024). In contrast, Aub and Ago3 are cytoplasmic proteins, localizing to nuage, a 51 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint perinuclear membraneless organelle formed in germline cells (Suyama & Kai, 2025; 52 Kawaguchi et al, 2025). In nuage, Aub and Ago3 mediate the reciprocal cleavage of 53 targets containing sense and antisense TE sequences. 3’ cleavage products generated by 54 either of the two proteins are passed to the other (i.e., from Aub to Ago3, or from Ago3 55 to Aub), then processed into a piRNA. Newly generated piRNAs direct the next round 56 of target cleavage and piRNA production. These cycles are called ping -pong, coupling 57 TE degradation and piRNA amplification in germline cells. Because target cleavage 58 occurs between the nucleotides complementary to 10th and 11 th nucleotides of guide 59 piRNA, and new piRNA is produced from the 5’ end of cleavage product, germline 60 piRNAs exhibit 10-nt overlap signature (ping-pong signature). In addition to ping-pong, 61 precursor fragments incorporated to Aub can be processed into multiple piRNAs on 62 mitochondrial outer membrane. These phasing/trailer piRNAs are mainly loaded onto 63 Piwi, translocating the complexes to the nucleus (Ge et al, 2019). Ping-pong cycles 64 process diverse transcripts including intact TE mRNAs and long non -coding RNAs 65 derived by noncanonical transcription of heterochromatinized, large intergenic regions 66 accumulating truncated TE sequences, defined as piRNA cluster loci (Brennecke et al, 67 2007a). 68 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint piRNA-directed TE silencing is achieved through interactions between PIWI 69 proteins and other components . These include nuclear factors involved in 70 Piwi-piRNA-directed target heterochromatinization (e.g., Panx, Nxf2, SetDB1/Eggless) 71 (Zhao et al, 2019; Murano et al, 2019; Fabry et al, 2019; Batki et al, 2019; Yu et al, 72 2015b; Sienski et al , 2015a). In cytoplasm, Tudor domain proteins (Krimper, Tejas, 73 Tapas, Qin/Kumo) and RNA helicases (Spn-E, Vas) are enriched in nuage and support 74 the ping-pong cycles mediated by Aub and Ago3 (Webster et al, 2015; Lin et al, 2023; 75 Handler et al, 2011; Zhang et al, 2011; Anand & Kai, 2012; Qi et al, 2011; Saito et al, 76 2010; Lim & Kai, 2007; Kawaguchi et al , 2025) . Nonetheless, characterizing the 77 interactions between PIWI members and other factors are often challenging because of 78 their dynamics in piRNA biogenesis and TE silencing. Hence, the repertoire of piRNA 79 pathway components is heretofore unelucidated, and ou r understanding of 80 piRNA-directed TE silencing remains incomplete. 81 Proximity-dependent biotin labelling (BioID/TurboID) is a powerful tool for 82 characterizing the physical interactions and proximity relationships of proteins in vivo 83 (Roux et al, 2012; Branon et al, 2018; Choi‐Rhee et al, 2004). These techniques rely on 84 the promiscuous biotin ligase activity exhibited by BirA derived from Escherichia coli. 85 TurboID utilizes an efficient and compact BirA variant called mini(m)Turbo established 86 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint through directed evolution (Branon et al, 2018), which we previously introduced in male 87 germline cells to study testis -specific factors (Iki et al, 2023; Kai et al, 2025; Iki et al, 88 2020). In order to deepen the understanding of piRNA-directed TE silencing, this study 89 characterized the proteins in the physical proximity of individual PIWI proteins in 90 female germline cells where TE silencing has been extensively studied . The obtained 91 proximity proteome of germline PIWI members not only confirmed reported 92 interactions but also revealed the unrecognized links. Moreover, PIWI proximity 93 proteome contained a list of factors which have not been characterized in piRNA 94 pathway thus far. Of those, this study further characterizes Set1, a histone 95 methyltransferase and transcriptional coactivator conserved across eukaryotes. Our data 96 highlight the hitherto unappreciated importance of Set1 in Piwi-piRNA-directed 97 silencing of TEs forming telomeres in Drosophila genome. 98 99 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

100 Labelling of PIWI proximity factors and purification from germline cells 101 To characterize the proximity proteome of PIWI proteins using TurboID, we generated a 102 series of transgene s encoding mTurbo -GFP-fused Piwi, Aub, or Ago3 (Figure 1A). 103 These transgenes were expressed in ovarian germline cells under the control of nos 104 promoter activity (UASp/Gal4 binary system using NGT 40;nos-Gal4 (NN) driver) 105 (Rørth, 1998; Grieder et al , 2000) . The biotinylation activities of individual fusion 106 proteins were confirmed by detecting the streptavidin-HRP-dependent signals (Figure 107 S1A). However, pulldown using streptavidin was inefficient and biotinylated proteins 108 largely remained in the flow through , possibly due to interfer ing factors accumulated 109 during differentiation. To circumvent this technical challenge, we depleted one of the 110 differentiation factor s, bag of marbles (bam) (McKearin & Spradling, 1990) , by 111 germline knockdown (GLKD) , and collected germline stem cell (GSC)-like cells 112 (GSCLCs) impaired in differentiation (Figure 1A, S1BC). Although TurboID in 113 GSCLCs showed weaker biotinylation signals in the input, the pulldown efficiency was 114 dramatically improved (Figure S1A). 115 In GSCLCs, mTurbo-GFP alone displayed a broad distribution across both the 116 nucleus and cytoplasm (Figure 1B). In contrast, mTurbo -GFP-PIWI proteins exhibited 117 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint subcellular localization patterns consistent with those of their endogenous counterparts; 118 Piwi accumulated in the nucleus, while Aub and Ago3 were enriched in the perinuclear 119 nuage. The concordance suggests that N -terminal mTurbo -GFP tagging does not 120 substantially disrupt the interaction networks of individual PIWI proteins. Supporting 121 this notion, N-terminal mTurbo-FLAG fusion preserved Aub functions in male germline 122 (Iki et al, 2023). Notably, the steady-state levels of fusion proteins were below those of 123 endogenous counterparts (Figure S1D) . mTurbo-GFP-Ago3 was the least stable and 124 barely detectable in the input (Figure 1C , anti -GFP), accounting for the weak 125 streptavidin-HRP signals in Ago3 -TurboID condition (Figure 1 C). Nevertheless, the 126 streptavidin pulldown displayed distinct signal patterns, suggesting 127 biotinylation-dependent enrichment of specific factors in each PIWI-TurboID condition. 128 129 Characterization of germline PIWI proximity proteomes 130 Mass spectrometry analysis and the label-free quantification (LFQ) of streptavidin 131 pulldown samples identified the proximity factors of individual PIWI proteins (average 132 abundance ratios >1 compared to the control GFP-TurboID conditions in biological 133 duplicates) (Table S1). These included 37, 33, and 49 factors for Piwi, Aub, and Ago3, 134 respectively (Figure 2A and 2B, S2). Notably, the identified proximity factors contained 135 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint known piRNA pathway components (Figure 2A, highlighted in yellow, Table S2), and 136 the significant enrichment was confirmed by Gene Ontology (GO) analysis (Figure 2C). 137 These results suggest that the proteome reflects the physical and functional interactions 138 between PIWI members and other factors in germline cells. 139 140 Screening of PIWI proximity factors involved in piRNA biogenesis and TE 141 silencing 142 Most factors identified in the proximity of PIWI members have not been functionally 143 characterized in piRNA pathway or TE silencing. Hence, w e investigated the possible 144 involvement of 54 factors, for which short hairpin-based RNAi lines were available, by 145 germline knockdown (GLKD) screening using NN driver. Given that impaired piRNA 146 biogenesis can be associated with nuage collapse and mislocalization of its components, 147 we examined the localization of Krimp, a key scaffold of nuage (Lim & Kai, 2007; Patil 148 & Kai, 2010). In parallel, we observed the accumulation of Gag protein encoded by a 149 non-LTR retroelement HeT-A, as a readout of TE derepression (Shpiz et al, 2011). We 150 confirmed that GLKD of aub leads to the accumulation of HeT-A Gag protein around 151 the oocytes of developing egg chambers (Figure 3A) , and the loss of Krimp from 152 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint perinuclear nuage and the formation of abnormal foci in the cytoplasm (Figure 3B , 153 S3B). 154 GLKD screening identified several factors including Hrb27C and Set1 , the 155 proximity factor of Aub and Piwi, respectively (Table S3). Both factors are conserved 156 across eukaryotes , and their molecular and physiological functions have been 157 characterized. Hrb27C (also known as Hrp48) is a n hnRNP family member required in 158 GSC for the maintenance (Yan et al, 2014; Ables et al, 2016; Finger et al, 2023). Set1 is 159 an enzyme acting inside COMPASS complex responsible for di- and tri-methylation of 160 histone H3 at lysine 4 (H3K4me2/3), a modification associated with transcription start 161 sites (TSSs) for RNA polymerase (pol)II activation (Ardehali et al, 2011; Mohan et al, 162 2011; Wang et al , 2023a) . We observed distinct phenotypes in hrb27C-GLKD and 163 set1-GLKD. In hrb27C-GLKD, abnormal Krimp foci appeared in the cytoplasm, similar 164 to aub-GLKD (Figure 3B). However, HeT-A Gag protein was under detectable level 165 (Figure 3A). In set1-GLKD, though abnormal Krimper foci formation was not clear, 166 HeT-A Gag protein accumulated around the oocyte. In both hrb27c- and set1-GLKD 167 conditions, perinuclear Krimp signals could be weaker but still observed (Figure 3B). 168 Consistently, Aub and Piwi did not show drastic changes in their localization (Figure 3C, 169 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint S3C). Of note, RT -qPCR confirmed target downregulation in individual GLKD 170 conditions (Figure S3D). 171 172 set1 and hrb27C functionally interact with piwi and aub, respectively, to control 173 TEs 174 To investigate the global effect of Hrb27C or Set1 depletion in the germline on TE 175 expression, we compared the transcriptome from hrb27C- or set1-GLKD ovaries with 176 that from gfp-GLKD control ovaries (Figure 4 A, S4 A). In parallel, aub- and 177 piwi-GLKD ovary transcriptomes were generated as references for TE derepression. 178 Our analysis included germline stem cell -like cell (GSCLC) condition in which PIWI 179 proximity proteome was characterized (Figure 4B, S4B). First, comparison of 180 gfp-GLKD transcriptomes between non-GSCLC and GSCLC contexts revealed that TE 181 transcript levels are generally elevated in GSCLCs (Figure S4C). This baseline increase 182 may account for the relatively modest TE derepression observed for aub- or 183 piwi-GLKD in GSCLC setting (Figure 4AB). 184 Hrb27C was identified as the proximity factor of Aub (Figure 2). H owever, 185 unlike aub-GLKD, hrb27C-GLKD did not result in clear TE derepression in 186 non-GSCLC ovaries (Figure 4A, S4A) . In contrast , TE derepression was evident in 187 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint GSCLC ovaries (Figure 4B, S4B) , and the patterns were positively correlated between 188 hrb27C- and aub-GLKD conditions (r = 0.57, Figure S4D). These suggest a functional 189 link between Hrb27C and Aub in GSCs. 190 For Set1 identified as the proximity factor of Piwi (Figure 2), its GLKD caused 191 remarkable derepression of a subset of TEs in non-GSCLC ovaries (Figure 4A, S4A). 192 These set1-sensitive TEs included HeT-A, consistent with its Gag protein accumulation 193 (Figure 3) , as well as TART family members , which together with HeT-A maintain 194 telomeres as HTT array (Villasante et al, 2007). TAHRE, another HTT component, was 195 not analyzed due to the lack of annotated insertions. In addition to HTT, several LTR 196 retrotransposons including HMS-Beagle, Max-element, diver, 3S18, and gypsy12 197 exhibited robust derepression (Figure 4A) . This result was corroborated by analyses 198 based on mapping reads to consensus sequences of individual TE families (Figure 4C). 199 Overall, the derepression profile of set1-GLKD showed a strong positive correlation 200 with that of piwi-GLKD (r = 0. 73), but not with aub-GLKD (r = 0.02, Figure 4 C). 201 Consistent with this, re-analysis of publicly available dataset (GSE103582) indicated 202 that HTT members are more sensitive to the loss of piwi than that of aub (Figure 4D, 203 S4E) (Teixeira et al , 2017a). Moreover, similar to set1-GLKD, HTT members were 204 strongly derepressed upon loss of Panoramix (Panx), an essential cofactor of Piwi 205 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint (GSE71374) (Sienski et al, 2015b; Yu et al, 2015a). In contrast to the pronounced effect 206 observed in germline, knockdown of set1 in ovarian follicle cells using traffic jam-Gal4 207 did not derepress zam and gypsy, which are silenced by piwi in this somatic lineage 208 (Figure S4F). Taken together, these results suggest that Set1 function is associated with 209 Piwi-mediated TE silencing in germline cells. 210 211 Catalytic activity-independent role of Set1 in TE silencing 212 The C-terminal SET domain of Set1 is responsible for H3K4me3 modification (Mohan 213 et al , 2011; Ardehali et al , 2011; Wilson et al , 2002) . To investigate whether this 214 methyltransferase activity is required for TE silencing , we performed rescue 215 experiments by expressing RNAi-resistant transgene encoding GFP -tagged wild -type 216 Set1 (Set1 WT) or a catalytically inactive variant (Set1E1613K) in germline cells depleted 217 of endogenous Set1 (set1-GLKD) (Figure 5A) (Vidaurre et al, 2024; Hallson et al, 2012). 218 Both GFP-Set1WT and GFP -Set1E1613K proteins accumulated in the germline nuclei 219 (Figure 5B). Weaker GFP signals observed for E1613K variant suggest reduced protein 220 stability. 221 Transcriptome analysis revealed that HTT and other TE families were fully 222 re-repressed by GFP -Set1WT introduced in the set1-GLKD background, indicating the 223 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint depletion of set1 is causal for their derepression (Figure 5C). Remarkably, although less 224 effective than GFP-Set1WT, expression of GFP-Set1E1613K led to substantial 225 re-repression of TEs. This effect was most pronounced for HTT, with TART-A1 as an 226 exception (Figure 5C). Together, these results suggest that Set1 plays a noncanonical, 227 catalytic activity-independent role in TE silencing. 228 To further support the above notion , we analyzed H3K4me3 signals in 229 set1-GLKD and the transgene-rescue conditions. Ovary immunotaining showed that 230 H3K4me3 signals were depleted from germline nuclei in set1-GLKD (Figure 5 B). 231 Germline H3K4me3 signals were fully recovered by GFP -Set1WT, while no recovery 232 was observed with GFP-Set1E1613K. We further employed CUT&Tag to genome-widely 233 characterize H3K4me3 signals. H3K4me3 peaks were identified around the 234 transcription start site (TSS) of genes including ago3 and aub (Figure 5D). Consistent 235 with the immunostaining data, H3K4me3 signals on ago3 and aub were markedly 236 reduced in set1-GLKD ovaries, and fully restored by GFP-Set1WT, but not by 237 GFP-Set1E1613K (Figure 5D). These results confirm the catalytic inactivity of E1613K 238 variant and further support the catalytic activity -independent role in TE silencing . 239 Notably, despite these changes in H3K4me3 signals, steady-state transcript levels of 240 piRNA pathway components were only modestly affected in set1-GLKD ovaries 241 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint (Figure S5 A). Moreover, RT -qPCR analysis showed that depletion of COMPASS 242 subunits led to milder derepression of HeT-A compared with that of set1 (Figure S5B). 243 Taken together, these results argue against the notion that Set1 functions solely as an 244 H3K4me3 writer and represses TEs simply by promoting the expression of piRNA 245 pathway components. 246 247 Loss of TE-targeting piRNAs in set1-depleted ovaries 248 TE derepression pattern suggests the functional link between Set1 and Piwi (Figure 4, 249 S4). Moreover, consistent with the ir close proximity (Figure 2), Set1 and Piwi interact 250 physically, as Piwi co-precipitated with GFP-Set1WT and GFP -Set1E1613K expressed in 251 germline cells (Figure 6A). Notably, the interaction appears transient or weak, given the 252 dependency on crosslinking (Figure S6A) . These physical and functional links 253 prompted us to examine whether piRNA expression is affected by set1 depletion. 254 Deep-sequencing of ovarian small RNAs showed that overall abundance of piRNAs 255 (23~29-nt fragments and those derived from TE sequences) was largely unchanged in 256 set1-GLKD (Figure 6B). However, analysis at the level of individual TE families 257 revealed a striking loss of piRNAs mapping to telomeric HTT members (Figure 6C). 258 Notably, antisense piRNAs, which directly target cognate TE transcripts, were reduced 259 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint more severely than sense piRNAs. A prominent example is HeT-A, for which sense 260 piRNAs were largely unaffected whereas antisense species were nearly lost (Figure 261 6CD). Supporting the link to Piwi pathway, re -analysis of published small RNA -seq 262 data (GSE71374) revealed a similar antisense-biased reduction of HTT-mapping 263 piRNAs in ovaries lacking Panx, a key cofactor of Piwi (Figure 6E) (Yu et al, 2015a; 264 Sienski et al, 2015b). An antisense-biased reduction of piRNAs was also observed for 265 other Set1-controlled TEs, including 3S18, gypsy12, Max-element, diver, and 266 HMS-Beagle (Figure 6C). Despite this loss, the ping-pong signature ( 10-nt overlap 267 frequency) of piRNAs was preserved (Figure S 6). This is consistent with the 268 maintenance of nuage structure in set1-GLKD ovaries (Figure 3B) . Together, t hese 269 findings indicate that Set1 is required to express TE-targeting functional piRNAs, 270 supporting its crucial role in Piwi-mediated TE silencing. 271 272 Set1 binds TE sequences and localizes to subtelomeric piRNA cluster loci 273 How does Set1 contribute to the expression of antisense piRNAs? To address this 274 question, we analyzed the chromatin binding by GFP-Set1 expressed in set1-depleted 275 germline cells, using CUT&Tag with anti-GFP antibodies. In parallel with GFP-Set1WT, 276 GFP-Set1E1613K was included in the analysis , given the functionality in TE silencing 277 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint (Figure 5) . Comparison between GFP -Set1 and GFP control conditions identified 278 Set1-binding peaks ( p<0.01, Figure S7A ). Notably, genes and TEs associated with 279 peaks from Set1WT or Set1 E1613K showed limited overlap (Figure 7A) . However, this 280 does not indicate their targets are distinct; rather, it reflects differential affinity between 281 Set1WT and Set1 E1613K to shared targets (r=0.85, Figure 7B). Consistent with its role as 282 H3K4me3 writer, targets identified from Set1WT peaks were predominantly endogenous 283 genes (541 of 548, Figure 7A). These bindnig peaks appeared around transcription start 284 site ( TSS) and coincided with set1-dependent H3K4me3 signals, as exemplified with 285 aub (Figure 7C). In contrast, targets identified from Set1E1613K peaks were enriched with 286 TEs (Figure 7A, 84 of 306) , including HeT-A, TART, diver, HMS-Beagle, and 287 Max-element, regulated by Set1 (Figure 4A, 5C). At the level of consensus sequences , 288 two major peaks were typically observed: one spanning the 5’UTR-ORF1 region and 289 another within the 3’UTR (Figure 7 D, S7CD). Collectively, these results revealed the 290 affinity of Set1 to TE sequences. 291 We mention that H3K4me3 signals on TEs behaved differently from those at 292 gene TSS: they increased upon set1 depletion and decreased upon expression of both 293 Set1WT and Set1 E1613K, indicating a lack of correlation with Set1’s catalytic activity 294 (Figure 7D, S7CD). Instead, TE-associated H3K4me3 signals were positively correlated 295 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint with the transcript abundance (Figure 5) , suggesting that these marks are deposited 296 either by residual Set1 in the set1-depleted germline cells or through Set1-independent 297 mechanisms (Ardehali et al, 2011). 298 3’UTRs of HTT members possess promoter and TSS for bidirectional 299 transctiption (Figure 7D) (Danilevskaya & Arkhipova, 1997; Radion et al , 2017; 300 Maxwell et al, 2006). Intriguingly, HTT 3’UTRs interacting with Set1 are maintained as 301 truncated insertions in subtelomeric piRNA cluster loci. HeT-A{}6278 in cluster 3 is one 302 of those 3’UTR fragments. Alignment to the consensus sequence suggests that 303 HeT-A{}6278 maintains promoter and antisense TSS but not sense TSS (Figure S7E). 304 piRNA hot spot starting from th is 3’UTR fragment implies the involvement in piRNA 305 precursor transcription (Figure 7E). Set1 showed affinity to other sites on cluster 3 , 306 which is contrastive to non-telomeric clusters lacking, if not any, Set1-binding signals 307 (Figure S7F). Insertions from set1-sensitive LTR retrotransposons are rare within major 308 piRNA clusters. Max{}2206 is one of exceptions found in 42AB cluster, but it did not 309 exhibit Set1 binding (Figure S7F). Hence, Set1 binding to cluster loci is biased toward 310 subtelomeric regions, whose piRNA production is extremely set1-dependent (Figure 311 7E). Set1 binding to LTR retrotransposons may instead reflect its affinity for potentially 312 mobile copies located outside piRNA clusters (Wang et al, 2018). Collectively, t hese 313 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

support a model in which Set1 binding to TE sequences promotes precursor 314 transcription for piRNA biogenesis in germline cells (Figure 7F). 315 316 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

317 To deepen the understanding of piRNA -directed TE silencing, this study explored the 318 proximity factors of individual PIWI proteins expressed in female germline cells. The 319 proteome provided additional clues supporting the known interactions, and, more 320 importantly, revealed otherwise unrecognized links between factors in piRNA 321 biogenesis. First, we point out that Zucchini (Zuc), an endonuclease on mitochondrial 322 outer membrane, was identified in the proximity of both Aub and Piwi . These three 323 proteins are together involved in phased piRNA production (Ge et al, 2019), but their 324 physical links have not been reported in ovaries. It should be noted, a study conducting 325 TurboID using Zuc as bait did not find Aub or Piwi in its proximity (Nguyen et al, 2023). 326 This might be due to the difference of drivers (matalfa -Gal4 vs nos -Gal4), stages of 327 germline cells (differentiating cells vs stem -cell like cells), or bait proteins (Zuc vs 328 PIWI proteins) . In addition to Zuc, our data also showed that Nxf3 and Bootlegger 329 (Boot) mediating the export of piRNA precursors are in the proximity of Ago3, but not 330 of Aub (ElMaghraby et al, 2019; Kneuss et al, 2019). The selective proximity supports 331 the model whereby precursor transcripts delivered by Nxf3 and Boot from nucleus to 332 nuage in cytoplasm are targeted by Ago3-piRNA complexes (Wang et al, 2023b). 333 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Hrb27C emerged as a functionally relevant partner of Aub (Figure 2-4). As a 334 member of hnRNP family, Hrb27C regulates the transport, localization, and translation 335 of interacting mRNAs during oogenesis (Yano et al, 2004; Goodrich et al, 2004; Huynh 336 et al, 2004). Both Hrb27C and Aub are intrinsic factors required for GSC maintenance 337 (Ma et al, 2017; Rojas‐Ríos et al, 2017; Rojas-Ríos et al, 2024; Finger et al, 2023). In this 338 study, we provided evidence supporting the involvement of Hrb27C in Aub -mediated 339 TE silencing in GSCLCs (GSC-like cells induced by depletion of differentiation factor, 340 bam) (Figure 4). The cooperation between Hrb27C and Aub in GSCs warrants further 341 detailed investigation. 342 H3K4 methyltransferase Set1 was identified and characterized as a key factor 343 in Piwi-mediated TE silencing in female germline (Figure 2-7). We initially as sumed 344 that, despite the proximity to Piwi, TE silencing by Set1 could be attributable to its 345 general coactivator function (Ardehali et al, 2011; Mohan et al, 2011). Set1 acting as 346 H3K4me3 writer inside COMPASS potentially underlies the expression of germline 347 genes including piRNA pathway components. However, our results argue against this 348 notion but instead shed light on its specialized role in Piwi pathway. First, set1 and piwi 349 but not aub showed similar TE derepression pattern upon depletion from germline cells 350 (Figure 4) . Second, set1 depletion caused severe loss of piRNAs derived from HTT, 351 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint which we found ocurring in panx-deficient ovaries disrupting Piwi pathway (Figure 6). 352 Moreover, w e demonstrated that Set1 can control TEs , HTT members in particular, 353 without its catalytic activity (Figure 5). Last, CUT&Tag revealed the affinity of Set1 to 354 TE sequences, which is associated with piRNA production (Figure 7). 355 How Set1 recognizes TE sequences has remained elusive. Set1 interacts with 356 phosphorylated C -terminal domain ( CTD) of RNA polII in elongation phase, which 357 allows targeted recruitment and H3K4me3 deposition on actively transcribed genes (Ng 358 et al, 2003). In o ur CUT&Tag analysis, H3K4me3 signal s within TE sequences were 359 largely overlap ped with Set1 -binding regions. Although our results did not support a 360 role for Set1 in depositing these TE-associated H3K4me3 (Figure 7D), RNA polII on 361 transcriptionally active TEs could contribute to Set1 recruitment. However, HeT-A 362 showed only mild derepression by downregulation of a COMPASS component, Wdr82 363 that is the orthologue of yeast Swd2 linking Set1 and RNA polII (Figure S5A) (Bae et al, 364 2020). Hence, RNA polII-dependent recruitment of Set1 as part of COMPASS may be 365 insufficient, and additional mechanisms may contribute. The e levated TE binding 366 observed for the catalytically inactive Set1 variant, compared with the wild-type protein, 367 support a possible alternative mode of recruitment (Figure 7B). Given the physical and 368 function links, involvement of Piwi machinery in recruiting Set1 to TEs is a plausible 369 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint possibility. Our proximity proteome did not identify SetDB1/Eggless, the H3K9me3 370 writer responsible for heterochromatin formation at Piwi-piRNA target loci (Rangan et 371 al, 2011; Osumi et al , 2019; Akkouche et al , 2017) . This absence suggests that our 372 dataset may be biased toward a specific stage at which Piwi cooperates with Set1 . 373 Further characterization of additional factors identified by our Piwi proxiome should 374 provide deeper insight into the mode of actions of Set1 in TE silencing. 375 Most piRNAs are derived from defined genomic regions called piRNA c luster 376 loci (Brennecke et al , 2007b) . The heterochromati c structure of these loci requires 377 noncanonical RNA polII recruitment for piRNA precursor transcription, mediated by 378 H3K9me3 reader Rhino and the cofactors (Klattenhoff et al, 2009; Mohn et al, 2014; 379 Andersen et al , 2017) . Recent studies uncovered cluster-specific molecular 380 underpinnings for piRNA precursor transcription. These include DNA-binding protein, 381 Kipferl, and enhancer of zeste, E(z), the writer of H3K27me3 , each required at distinct 382 cluster loci (Akkouche et al, 2025; Baumgartner et al, 2022). In gonadal soma, Traffic 383 jam (Tj), a fly orthologue of large Maf transcription factors , activates soma-restricted 384 cluster, flamenco (Rivera et al, 2025; Alizada et al, 2025). Subtelomeric piRNA cluster 385 loci involve Set1 (this study) and NSL complex (Iyer et al , 2023). Together, these 386 findings indicate that precursor transcription and piRNA production are governed by 387 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint diverse chromatin modifiers and transcriptional regulators . Such diversity may be 388 advantageous, enabling spaciotemporal control of the piRNA repert oire to support 389 multitasks of a common set of PIWI proteins. By localizing to subtelomeric cluster loci, 390 Set1 would exert its role in telomere control with Piwi . Characterization of telomere 391 phenotypes in set1-depleted germline cells will be an important direction for future 392 studies. Beyond piRNA clusters, Set1 also exhibited affinity for several TE families 393 including HMS-Beagle and Max-element, both of which maintains mobility in female 394 germline and transmits new copies to next generations through oocyte targeting (Yang et 395 al, 2023; Wang et al , 2018). The role of Set1 in silencing of such active TEs also 396 represent a future avenue. 397 Fly Set1 can silence TEs forming telomeres regardless of its methyltransferase 398 activity (Figure 5). This finding is reminiscent of molecular features reported for yeast 399 Set1, which was originally identified as a transcriptional repressor in the telomere and 400 silent mating loci (Nislow et al, 1997). Notably, its repressive functions on genes and 401 TEs are often associated with H3K4 methylation-independent mechanisms (Jezek et al, 402 2023; Lee et al, 2018; Lorenz et al, 2012). However, t he m echanisms underlying TE 403 silencing are different between the two organisms, as yeast lacks PIWI clade proteins 404 and does not produce piRNAs. Nonetheless, Set1-dependent antisense non-coding 405 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint RNAs have been characterized as repressors of retrotransposons in yeast (Berretta et al, 406 2008). The role of Set1 in generating antisense , silencing-competent transcripts for 407 genome defense may be conserved across broader contexts. 408 409 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

410 Fly stocks and culture 411 All stocks and crosses were raised at 25 ˚C on a molasses/yeast medium [5% (w/v) dry 412 yeast, 5% (w/v) corn flower, 2% (w/v) rice bran, 10% (w/v) glucose, 0.7% (w/v) agar, 413 0.2% (v/v) propionic acid, and 0.05% (v/v) p -hydroxy butyl benzoic acid]. Fly stocks 414 used in this study are listed in Table S4. 415 416 Plasmid construction and generation of transgenic flies 417 All the primers used for plasmid construction are listed in Table S 5. To generate 418 UASp-mTurbo-GFP, UASp -mTurbo-GFP-Piwi, UASp -mTurbo-GFP-Aub and 419 UASp-mTurbo-GFP-Ago3, mTurbo fragment was amplified by PCR using 420 attB_miniTurbo_Fw_general and miniTurbo_linker_Rv_general as primers and 421 3xHA-miniTurboNLS_pCDNA3 (addgene #107172) as template. GFP fragment was 422 amplified by PCR using 2 -1vLinker>GFP5_F and Ti876 -GFP-G5-R as primers. Piwi, 423 Aub or Ago3 fragment was amplified by PCR using primer pairs of Linker_to_Piwi and 424 Piwi_to_Vector, Ti893-G5-Aub-CDS-F and Ti894 -Aub-UASp-R1, or Linker_to_Ago3 425 and Ago3 _to_Vector, respectively, with cDNA from yw ovaries as template . PCR 426 fragments were introduced into the XbaI site in pUASp-K10-attB vector (Koch et al, 427 2009) using In -Fusion HD Cloning (Takara). UASp-mTurbo-GFP and 428 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint UASp-mTurbo-GFP-Ago3 were integrated to attP40 , while UASp -mTurbo-GFP-Piwi 429 and UASp-mTurbo-GFP-Aub were integrated to attP2, and attP-3B sites, respectively. 430 431 Proximity-dependent biotin pulldown in ovarian germline cells 432 After eclosion, female s expressing mini(m)Turbo -GFP-PIWI proteins were reared at 433 25˚C for 2 -3 ( non-GSCLC ovaries) or 5 -7 days (GSCLC ovaries) in the modified 434 molasses/yeast medium supplemented with 100 μM biotin (Nacalai). For individual 435 conditions, ~ 300 ovaries from 150 progenies were collected . Ovaries were 436 homogenized in 150 μL of PI -lysis buffer [50 mM t ris-HCl (pH 7.5), 500 mM NaCl, 2 437 mM EDTA, 2 mM dithiothreitol (DTT), 0.4% (w/v) SDS, and cOmplete Protease 438 Inhibitor Cocktail Tablet (Roche)] , using Bioruptor (Diagenode) for 30s (power H) for 439 six times with 30s intervals. Triton X -100 was then added to sa mples at a final 440 concentration of 2% (v/v), and homogenization was further performed for 30s for three 441 times with 30s intervals. After centrifugation (20,000 g, 10 min, 4 ˚C), the supernatant 442 was diluted with equal amount of 50 mM Tris -HCl (pH 7.5) buffer and incubated with 443 pre-equilibrated 15-μL slurry volume of Dynabeads MyOne Streptavidin C1 (Thermo 444 Fisher Scientific), overnight at 4 ˚C with gentle rotation. Next day, beads were washed 445 twice with W1 buffer [50 mM tris -HCl (pH 7.5), 250 mM NaCl, 0.2% (w/v) SDS, 1 446 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint mM EDTA, and 1 mM DTT], twice with W2 buffer [50 mM H EPES-KOH (pH 7.4), 447 500 mM NaCl, 0.1% (w/v) deoxycholate (Nacalai), 1% (v/v) Triton X -100, and 1 mM 448 EDTA], twice with W3 buffer [10 mM tris -HCl (pH 8.0), 250 mM LiCl, 0.5% (w/v) 449 deoxycholate, 0.5% (v/v) NP-40 (Nacalai), and 1 mM EDTA], and four times with W4 450 buffer [50 mM tris -HCl (pH 7.5) and 50 mM NaCl]. The purified proteins bound to 451 beads were stored at -80˚C until analysis. 452 453 Immunoblotting and biotinylated protein detection 454 Protein s amples wer e denatured at 95°C for 5 min in 2× protein loading buffer [4% 455 (w/v) SDS, 200 mM DTT, 0.1% (v/v) bromophenol blue (BPB), and 20% (v/v) 456 glycerol] saturated with biotin, resolved by SDS –polyacrylamide gel electrophoresis 457 (SDS-PAGE) and transferred to 0.2 -μm polyvinylidene difluoride membrane (Wako) 458 using the semi-dry system (Trans-blot Turbo, Bio-Rad). The membrane was blocked in 459 4% (w/v) skim milk (Nacalai) in 1× phosphate -buffered saline (PBS) supplemented 460 with 0.1% (v/v) Tween 20 and further incubated with primary antibodies: rabbit 461 anti-GFP (1:1000, Clonetech), mouse anti -Piwi (1:100) (Saito et al, 2006), guinea pig 462 anti-Aub (1:1000) and rat anti -Ago3 (1:200 ) (Lim et al, 2022), mouse anti -Tubulin 463 (1:3000, Santa Cruz) . Secondary antibodies were anti-guinea pig 464 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint immunoglobulins-HRP (1:1000; Dako), anti -rabbit immunoglobulin G (IgG) –HRP 465 (1:3000; Bio-Rad), anti-mouse IgG-HRP (1:3000; Bio -Rad), and anti-rat IgG (1:1000; 466 DAKO). HRP-conjugated Streptavidin (Clonetech) was used for detecting biotinylated 467 Proteins. The chemiluminescent signals generated with Chemi-Lumi One (Nacalai) 468 were detected by Chemidoc MP Imaging system (Bio-Rad). The images were processed 469 with Fiji. 470 471 Proteomic Sample Preparation and Label -Free Quantification using Orbitrap 472 Eclipse Tribrid Mass Spectrometer 473 Biotinylated proteins bound to magnetic beads were reduced with 1 mM dithiothreitol 474 (DTT) and alkylated with 5.5 mM iodoacetamide. Proteins were digested overnight at 475 37°C with Trypsin Gold, Mass Spectrometry Grade (Promega, Madison, WI). The 476 resulting peptides were captured and desalted using ZipTip C18 (Millipore, Billerica, 477 MA). Shotgun proteomic analysis was performed using an Orbitrap Eclipse Tribrid 478 mass spectrometer equipped with a FAIMS Pro interface (Thermo Fisher Sci entific, 479 Waltham, MA), which was coupled to a Vanquish Neo UHPLC system (Thermo Fisher 480 Scientific, Waltham, MA). Peptides were separated on a reversed -phase column using a 481 linear gradient of 2 –24% acetonitrile in 0.1% formic acid at a flow rate of 300 nL/m in. 482 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Full MS scans were acquired in the Orbitrap at a resolution of 120,000, followed by 483 MS/MS scans in the ion trap using higher -energy collisional dissociation (HCD) with a 484 normalized collision energy of 35% and a maximum injection time of 10 ms. Label-free 485 quantification (LFQ) was performed using Proteome Discoverer version 2.5 (Thermo 486 Fisher Scientific, Waltham, MA) with the Sequest HT search engine. MS/MS spectra 487 were searched against the UniProt Drosophila melanogaster reference proteome 488 (UP000000803). Protein identifications were filtered at a false discovery rate (FDR) of 489 <1%. Peptide intensities were extracted using the Minora Feature Detector node and 490 used for protein quantification. GO analysis was conducted using MetaScape. (Zhou et 491 al, 2019) 492 493 Croslinking and immunoprecipitation 494 Ovaries were dissected from 50 females expressing GFP, GFP -Set1WT, or 495 GFP-Set1E1613K in phosphate -buffered saline (PBS) and immediately fixed with 0.1% 496 (w/v) paraformaldehyde (Electron Microscopy Sciences) for 10 min at room 497 temperature (RT). Fixation was quenched with 125 mM glycine. Samples were then 498 snap-frozen in liquid nitrogen and stored at -80°C until use . Ovary s amples were 499 homogenized with a pestle in lysis buffer [20 mM Tris -HCl (pH 7.5), 135 mM NaCl , 500 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint 1.5 mM MgCl 2, 10% (v/v) glycerol, 0.2% (v/v) TritonX-100] supplemented with 501 cOmplete Protease Inhibitor Cocktail (Roche). Pre-equilibrated a nti-GFP 502 antibody-conjugated magnetic beads (MBL, D153-11) were added to ovarian lysate and 503 incubated for 1h at 4°C. After three times of washing with lysis buffer, bead-bound 504 proteins were extracted by boiling in SDS-PAGE loading buffer [50 mM Tris–HCl (pH 505 6.8), 2% (w/v) SDS, 100 mM 1,4-dithiothreitol (DTT), 10% (v/v) glycerol, 0.05%(w/v) 506 bromophenol blue]. 507 508 Histochemistry and image acquisition 509 Ovaries were dissected from adult females in 1x PBS buffer supplemented with 0.4% 510 (w/v) bovine serum albumin (BSA; Wako) and fixed in 5.3% (v/v) paraformaldehyde 511 (Nacalai) in 0.67x PBS buffer for 10 min. To observe DNA, ovaries were incubated 512 with 1 μM 40,6-diamidino-2-phenylindole (DAPI) in PBX buffer [1x PBS containing 513 0.2% (v/v) Triton X -100]. For immunostaining, fixed ovaries were washed with PBX 514 and incubated with Image -iT™ FX Signal Enhancer (Invitrogen) for 30 min and PBX 515 containing 2% (w/v) BSA for 30 min for blocking. The primary antibody incubation 516 was performed overnight at 4 ˚C, and ovaries were washed with PBX at RT for 1h. The 517 secondary antibody incubation was then performed at RT for 2 h, and then ovaries were 518 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint washed with PBX at RT for 1 h. The antibodies used for immunostaining were rabbit 519 anti-HeT-A Gag (1:2000),(Lin et al, 2023) guinea pig anti-Krimp (1:2000),(Lim & Kai, 520 2007) mouse anti-Piwi (1:50, Santa Cruz, sc-390946), guinea pig anti-Aub (1:500),(Lim 521 et al, 2022) mouse anti-GFP (1:50, Invitrogen, A11120) . Secondary antibodies were as 522 follows: Alexa Fluor 488 - and 555 -conjugated goat antibodies at 1:500 (Molecular 523 Probes) and CF®633 goat antibodies at 1:500 (Biotium) . Antibodies were diluted in 524 0.4% (w/v) BSA containing PBX as the working solution. Images were taken by ZEISS 525 LSM 900 using C-Apochromat 40x/1.20 W Korr objective lens and processed with 526 ZEISS ZEN 3.0 and Fiji. 527 528 Reverse transcription and qPCR analysis 529 Total RNA was extracted from ovaries using TRIzol™ LS (Invitrogen) following the 530 manufacturer’s instructions. Using DNase I (NEB) –treated RNA, cDNA was 531 synthesized with 2.5 μM oligo(dT) adaptor using SuperS cript III reverse transcriptase 532 (Thermo Fisher Scientific). Quantitative reverse transcription PCR (qPCR) reaction was 533 performed using SYBR™ Green qPCR Master Mix (Thermo Fisher Scientific) and 534 gene-specific primers (Table S5) in QuantStudio 5 Real-Time PCR system (ABI). 535 536 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Transcriptome analysis 537 Total RNA was extracted from 100 ovaries (non-GSCLC ovaries or GSCLC ovaries) 538 using TRIzol™ LS (Invitrogen). To obtain GSCLC ovaries expressing shRNA for aub, 539 hrb27C, set1 or gfp, shRNA lines were combined ( bam shRNA; aub shRNA, bam 540 shRNA; hrb27C shRNA, bam shRNA; set1 shRNA, bam shRNA; gfp shRNA) and 541 crossed with NN. DNase I-treated RNA was sent to Rhelixa (Japan) for library 542 construction and deep -sequencing. Polyadenylated RNA selection was performed with 543 NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB). Libraries were 544 constructed with the NEBNext Ultra II Directional RNA Library Prep Kit (NEB) and 545 sequenced by using NovaSeq 6000 (Illumina). Trimming was performed by Cutadapt 546 (v1.18, -j 12 -a AGATCGGAAGAGCA CACGTCTGAACTCCAGTCA -A 547 AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT, -m 20). Trimmed paired-end 548 reads were mapped to the genome (BDGP6.46, dm6) using STAR ( v2.7.10b, 549 --outFilterMultimapNmax 100 --outSAMmultNmax 1 --outMultimapperOrder Random). 550 Read counting was performed with featureCounts (-M -p --countReadPairs -t exon -g 551 gene_id) using Drosophila_melanogaster.BDGP6.46.112.gtf (ensembl.org) . Using the 552 sum of RPK ( read per kilobase) for individual genes and TEs, TPM ( transcript per 553 million) was calculated. Trimmed paired -end reads were also mapped to consensus 554 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint sequences of individual TE families (transposon_sequence_set.embl.v.9.41, flybase ) 555 with STAR using the same options . Mappers on individual TE families were counted 556 using pileup.sh (BBMap). TPM were calculated using the sum of RPK for genes and 557 TEs obtained in genome mapping . Published small RNA data ( GSE103582 and 558 GSE71374) were processed in the same way. 559 560 Small RNA sequencing and analysis 561 Biological duplicate dataset for Set1 -present condition was provided by gfp-GLKD 562 ovaries (control RNAi condition), and set1-GLKD ovaries expressing GFP -Set1WT 563 (rescue condition). Dataset for Set1 -depleted condition was provided by set1-GLKD 564 ovaries, and th e siblings of rescue conditions (CyO). T otal RNA was extracted using 565 TRIzol™ LS (Invitrogen) from 60 to 100 ovaries of ~3 days old adult females. After the 566 addition of chloroform and centrifugation (12,000g 15min, 4˚C), short (< ~200 nt) RNA 567 in the aqueous phase was purified using RNA Clean & Concentrator 5 (R1015 Zymo ). 568 To deplete 2S ribosomal RNA (rRNA) , 2 pmol of complementary oligo DNA 569 (5’-AGTCTTACAACCCTCAACCATATGTAGTCCAAGCAGCACT-3’) was added 570 per 1 μg of RNA. The mixture was heated (95 ˚C, 2min), gradually cooled down to form 571 DNA/RNA hybrid, and then treated with RNase H (NEB) at 37 ˚C for 30min. RNase H 572 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint was inactivated by heating at 65 ˚C for 20 min. The 2S rRNA -depleted RNA mixture 573 was loaded onto 8 M urea -polyacrylamide gel (12%) and size-separated in parallel with 574 visible RNA ladder (Dynamarker DM253, BioDynamics Laboratory Inc.) in 0.5x 575 tris-borate EDTA buffer. Gel within the range from 20 to 30nt was excised and passed 576 through a small hole opened by needle (23Gx1, TERUMO) at the bottom of 0.5 mL 577 DNA LoBind tubes (Eppendorf) by centrifugation (15,000g). The fine particles were 578 recovered in 2.0 mL DNA LoBind tubes, and RNAs were eluted overnight at 4 ˚C with 579 gentle rotation in the presence of 300 mM NaOAc (pH 5.2). After removing the 580 particles using cell ulose 0.22 µm membrane filter Spin -X (Costar, 8160), RNA was 581 precipitated in the presence of 80% (v/v) ethanol and glycogen (40 μg/mL) (Nacalai) for 582 overnight at -20˚C. After centrifugation (20,000g, 20min, 4˚C), RNA pellet rinsed twice 583 with 80% (v/v) ethanol was resuspended in RNase-free water. Small RNA libraries were 584 prepared using NEBNext® Multiplex Small RNA Library Prep Set for Illumina (NEB, 585 E7300S) following manufacture’s procedure. After 15 cycles of PCR amplification, the 586 libraries were purifi ed using MagMAX™ Pure Bind Beads (Applied Biosystems), 587 size-separated in 3% (w/v) low melting agarose (HydraGene) in the presence of SYBR 588 gold (Thermo Fisher). ~150 -bp library fragments containing 20~30 -bp inserts were 589 purified using QIAquick gel extractio n kit (Qiagen). Libraries were sequenced by 590 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint RIMD NGS core facility ( The University of Osaka ) using the NovaSeqX Plus platform 591 (Illumina). Adaptor sequences (AGATCGGAAGAGC) were removed by Trim Galore 592 (v0.6.2, --length 15) with quality filter cut off (Phred ≥20). Adaptor-trimmed reads were 593 mapped to rRNA, tRNA, snoRNA, and snRNA sequences using bowtie (v1.3.1), and 594 the unmappers were collected (samtools view -f 4). From the unmappers, 18~29 -nt 595 reads were selected as small RNAs using seqkit (v2.4.0 ) and mapped to miRNA 596 precursors using bowtie ( -v 0) to collect “miRNA” populations. The size -selected 597 (23~29-nt) unmappers were considered as “piRNA”. These piRNA fragments were 598 mapped to TE consensus sequences (transposon_sequence_set.embl.v.9.41, flybase ) 599 using bowtie allowing up to 3 mismatches and taking one from multi mappers (-v 3 -M 600 1 --best --strata). Fw and Rv mappers were counted for individual TEs using pileup.sh. 601 Read counts were normalized using the sum of 18~29 -nt reads and reads per million 602 (RPM) was obtained. For visualization, Fw and Rv mappers were separated using 603 splitsam.sh. Bedgraph was generated using bedtools (v2.26.0, genomecov -bga -split), 604 values normalized using scale factors given by the sum of 18~29-nt reads. Mean values 605 of biological replicates were obtained using bedtools unionbedg. Track view was 606 generated in IGV (v2.16.0). piRNA overlap scores (z) were measured using signature.py 607 (Antoniewski, 2014) . piRNA fragments were also mapped to Drosophila genome 608 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint (BDGP6.46, dm6) using bowtie without allowing mismatch (-v 0 -M 1 --best --strata). 609 Bedgraph generation, normalization, averaging, and visualization follow the processes 610 in the analysis using TE consensus sequences. Published small RNA data (GSE71374) 611 were processed in the same way. 612 613 CUT&Tag sequencing and analysis 614 We referred (Anderson et al, 2023) for performing whole ovary C UT&Tag. Twenty 615 ovaries from 10 adult females at ~ 3 days old were dissected in PBS buffer 616 supplemented with cOmplete Protease Inhibitor (Roche). Ovaries were permeabilized in 617 PBX buffer (PBX containing 0.2%(v/v) Triton-X100) for 30 min at RT. After removing 618 PBX, ovaries were washed once with Wash+ buffer ( 20mM HEPES [pH 7.4], 150 mM 619 NaCl, 0.5 mM spermidine[Nacalai], 2 mM EDTA, 1%[w/v] BSA, cOmplete Protease 620 Inhibitor) and incubated overnight at 4 ˚C with primary antibodies diluted by 1:50 with 621 Wash+ buffer. Anti-H3K4me3 (#61979, mouse, Active Motif), or anti-GFP (#598, rabbit, 622 MBL) antibodies were used . Next day, ovaries were washed three times with Wash+ 623 buffer, and incubated with secondary antibodies diluted by 1:50 with Wash+ buffer. 624 Anti-mouse IgG (#52885L, Cell Signalling), or anti -rabbit IgG (#35401S, Cell 625 Signalling) were used. After removing the secondary antibody solution, ovaries were 626 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint washed three times with 300Wash+ buffer (20 mM HEPES [pH 7.4], 300 mM NaCl, 0.5 627 mM spermidine, cOmplete Protease Inhibitor ) and incubated for 1h at RT with loaded 628 pAG-Tn5 ( #79561S, Cell Signalling) diluted by 1:25 with 300Wash+ buffer. After 629 removing the Tn5 solution, ovaries were washed three times with 300Wash+ buffer. 630 Tagmentation was performed by incubating ovaries in 300Wash+ buffer supplemented 631 with 10 mM MgCl2 for 1h at 37˚C. After removing the supernatant, ovaries were treated 632 with collagenase (2 mg/mL, Sigma C9407) in HEPESCA buffer (50 mM HEPES [pH 633 7.4], 360 µM CaCl 2) for 1h at 37 ˚C. Then, the reaction mixture was added with SDS, 634 EDTA, and proteinase K (Nacalai) (final concentration 0.2%[w/v], 16 mM, and 0.3 635 mg/mL, respectively), and further incubated for 1h at 58 ˚C. After the reaction, DNA 636 was purified by using Quick -DNA Microprep kit (D30 20, ZYMO). Using the purified 637 DNA as template, libraries were synthesized by PCR (72 ˚C 5min, 98 ˚C 30s, 14 cycles 638 of [98 ˚C 10s, 63 ˚C 15s], 65 ˚C 5min) using Ultra II Q5 Master Mix (NEB) and 639 dual-indexed primers (Table S 5). Generated libraries were purified using x1.3 volume 640 of AmpureXP (beckman). The concentration of libraries was measured using Qubit 641 dsDNA HS assay kit (Thermo). Libraries were sequenced with NovaSeqX (Illumina) 642 and 150+ 150bp paired -end reads were obtaine d (H3K4me3 libraries sequenced by 643 RIMD in The University of Osaka , GFP libraries sequenced by Rhelixa ). Adaptor 644 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint trimming was performed by Cutadapt ( -a 645 CTGTCTCTTATACACATCTCCGAGCCCACGAGAC -A 646 CTGTCTCTTATACACATCTGACGCTGCCGACGA, -m 20). Trimmed reads were 647 mapped to D. melanogaster genome (BDGP6.46, dm6) using Bowtie 2 -p 12 --no-unal 648 --end-to-end --very-sensitive --no-mixed --no-discordant -q --phred33 -I 10 -X 700 ). 649 Read counts were normalized using deeptools (bamCoverage --normalizeUsing CPM 650 -of bigwig --binSize=1). Average bigwig was generated from biological duplicate 651 bigwig data using wiggletools and ucsc-wigtobigwig (wigToBigWig). Track view was 652 generated in IGV (v2.16.0). Counting reads on gene TSS±500bp (H3K4me3) or exon 653 (GFP) was performed using featureCounts (v2. 1.1, -M -p --countReadPairs -O) with 654 Drosophila_melanogaster.BDGP6.46.112.gtf (Ensembl). RPM (read per million) was 655 calculated using the number of aligned reads in Bowtie2 mapping . Trimmed reads were 656 also mapped to TE consensus sequences (transposon_sequence_set.embl.v.9.41, 657 flybase). Bigwig was generated using deeptools (bamCoverage), normalizing the values 658 with scale factors given by the sum of aligned reads on the genome . Mean values of 659 biological replicates were obtained using wiggletools and wigToBigWig 660 (ucsc-wigtobigwig). Track view was generated in IGV (v2.16.0). 661 662 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

663 We thank Dr. Prof. Xin Chen for providing fly stocks. We thank Chisato Yanagisawa, 664 Minako Moriguchi for their helps with maintenance of fly lines. We are also grateful to 665 Bloomington Drosophila Stock Center for providing fly stocks. We also thank all 666 members in our lab for their insightful discussion and suggestions. This study is 667 supported by JSPS Grant -in-Aid for Scientific Research C (22K06081) for TI, Life 668 Science Foundation of Japan (J231503018) for TI, Takeda Foundation (J241503007) for 669 TI, Daiichi Sankyo Foundation of Life Science (J241503010) for TI, Naito Foundation 670 (J241503011) for TI, JSPS Grant -in-Aid for Scientific Research B (21H02401) for TK, 671 JSPS Grant -in-Aid for Transformative Research Areas A (21H05275) for TK, 672 Grant-in-Aid for JSPS Fellows (23KJ1521) for WI, and Grant from Open and 673 Transdisciplinary Research Initiatives (OTRI) RNA Frontier Science Division for TI. 674 The founders had no role in study design, data collection and analysis, decision to 675 publish, or preparation of the manuscript. 676 677 Author contributions 678 Conceptualization: TI and TK 679 Methodology: TI, TK, and WI 680 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Investigation: TI, WI, HK-H and MO 681 Supervisions: TI and TK 682 Writing-original draft: WI and TI 683 Writing-review and editing: TI, WI and TK 684 685 Competing interest statement 686 The authors declare no competing interests. 687 688 Data availability 689 Newly generated transcriptome data are available with BioProject accession ID: 690 PRJNA1345876; Drosophila ovary polyA transcriptome sequencing, PRJNA1345540; 691 Drosophila ovary small RNA sequencing, PRJNA1345531; Drosophila ovary 692 CUT&Tag sequencing 693 694 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Figure legends 695 Figure 1. PIWI proximity biotin labelling in germline stem cell-like cells (GSCLCs) 696 (A) Experimental design for PIWI proximity biotin labelling in female germline cells . 697 mTurbo-GFP-Piwi, -Aub, or -Ago3 is expressed in germline cells using UASp/Gal4 698 system with NGT40; nos-Gal4 (NN) drive r. mTurbo-GFP serves as negative contr ol. 699 Depletion of bam using short hairpin (sh)RNA enables TurboID in GSCLC s. (B) 700 Subcellular localization of mTurbo-GFP fusion proteins in GSCLC ovary. Upper panels; 701 fluorescent signals from GFP (green). Lower panels; immunostaining signals from 702 endogenous Piwi, Aub, or Ago3 (green). DAPI (blue) for nuclei. Scale bar = 50 μm or 5 703 μm in the insets. (C) Blotting images for p roteins extracted from GSCLC ovaries 704 expressing mTurbo fusion proteins (Input), and the streptavidin-bound fraction 705 (Pulldown). Upper panels; immunoblotting with anti-GFP antibody. Lower panels; 706 blotting with streptavidin-HRP. White and black arrowheads indicate mTurbo -GFP and 707 mTurbo-GFP-PIWI proteins, respectively. 708 709 Figure 2. Proximity proteome of PIWI proteins in germline cells 710 (A) Network diagram summarizing proximity factors of individual PIWI proteins (Piwi, 711 Aub, and Ago3) identified by mass spectrometry ( nanoLC-MS/MS) followed by 712 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint label-free quantification. Yellow nodes: factors known to be involved in piRNA pathway. 713 Size of c ircle reflects abundance ratio (mTurbo -GFP-PIWI/mTurbo-GFP) based on 714 biological duplicate TurboID data. (B) Venn diagram show s the number of proximity 715 proteins shared by or unique to individual PIWI members. (C) GO enrichment analys is 716 of PIWI proximity factors. Hierarchical clustering and heatmap show significantly 717 enriched GO terms. Color intensity represents the significance of enrichment ( -log10 718 p-value). 719 720 Figure 3. Germline knockdown (GLKD) screening of PIWI proximity factors 721 (A) Immunofluorescence signals from HeT-A Gag protein (top panels) in egg chambers 722 of control (y w) or GLKD of indicated genes. DNA stained with DAPI (bottom panels). 723 Arrowheads; Gag proteins accumulating around oocyte. Scale bars = 20 μm. (B) 724 Immunofluorescence signals of Krimp (green) in the egg chambers of control ( y w) or 725 of indicated GLKD conditions . DNA stained with DAPI (magenta). Fluorescence 726 signals along yellow arro ws are plotted in the bottom panels (signals normalized by 727 setting 1 for the highest value of DAPI). Scale bars = 10 μm. (C) Immunofluorescence 728 signals of Aub or Piwi (green) in the egg chambers of indicated conditions. 729 730 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Figure 4. Germline knockdown (GLKD) effect on ovary transcriptome 731 (A-B) Transcriptome analysis and comparison in non-GSCLC ovaries (A) or in GSCLC 732 ovaries (B). Transcript levels (TPM) of genes and TE insertions in set1-, hrb27C-, aub-, 733 or piwi-GLKD condition were compared with those in gfp-GLKD control. Dots in the 734 volcano plots; genes (gray) and TEs (gray with red outline) included in the differential 735 expression analysis (EdgeR). ( C) Scatter plot c omparing the effect of GLKD on the 736 transcript levels of i ndividual TE families (consensus sequence mapping). Left : 737 set1-GLKD compared with aub-GLKD. Right: set1-GLKD compared with piwi-GLKD. 738 Fold change from control gfp-GLKD (TPM/TPM) were shown. r = pearson correlation 739 coefficient. (D) Transcript levels (TPM) for HeT-A and TAHRE in different 740 transcriptome datasets. GSE103582 including aub and piwi knockouts, and GSE71374 741 including panx knockout were analyzed together(Yu et al, 2015a; Teixeira et al, 2017b). 742 Error bars indicate ±standard deviation (SD) of biological duplicate data. 743 744 Figure 5. Effect of set1 transgene expression on H3K4me3 modification and TE 745 expression in set1-GLKD ovaries 746 (A) Schemetic of GFP-Set1 transgenes (Vidaurre et al, 2024). Nucleotide substitution to 747 replace 1613 th glutamate ( E) of Set1 polypeptide to lysine ( K) abolishes catalytic 748 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint activity. Nonsense substitutions render resistance to siRNA targeting. Both placed 749 downstream of UASp and expressed by germline driver NN. (B) GFP fluorescence (top) 750 and immunofluorescence staining for H3K4me3 and Vas (bottom) in egg chambers. Vas 751 marks germline cells. Scale bar = 10μm. (C) Transcript levels (TPM) of TE families in 752 gfp-GLKD, set1-GLKD, and transgene expression conditions ( GFP-Set1WT or 753 GFP-Set1E1613K expressed in set1-GLKD). Error bars indicate ±SD from biological 754 duplicates. (D) CUT&Tag analysis of H3K4me3 in ovaries of indicated conditions. 755 Left: track view at the ago3 and aub gene loci. Bar graph on the right shows H3K4me3 756 signals (mean CPM) around TSS (±500bp) of ago3, aub and piwi . Error bars indicate 757 ±SD from biological duplicates. 758 759 Figure 6. Co-precipitation of Piwi with Set1, and e ffect of set1-GLKD on piRNA 760 accumulation in ovaries 761 (A) Immunoprecipitation of GFP or GFP -Set1 proteins from ovaries using anti -GFP 762 antibody (GFP-IP) after crosslinking treatment. Tubulin serves as loading control. (B) 763 Proportion (%) of miRNA and piRNA in ovarian small RNAs. miRNA: reads mapping 764 to miRNA precursor s. piRNA: 23~29-nt reads (excluding rRNA, tRNA, snoRNA, 765 snRNA, and miRNA mappers), or 23~29-nt reads mapping to TE consensus sequences. 766 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Control: mean values given by single replicate of gfp-GLKD and GFP-Set1WT-rescue 767 conditions. set1-GLKD: mean values given by biological duplicate of set1-GLKD 768 condition. Error bars indicate ±SD of biological duplicates. (C) Fold changes (FC) of 769 piRNA levels (RPM) in set1-GLKD compared to control condition. piRNA : 23~29-nt 770 mappers on individual TE families . TEs having few mappers ( RPM <5) are excluded. 771 Sense or antisense piRNA : mappers having TE sense strand or the reverse 772 complementary sequences. (D) Track view for 23~29-nt mappers on consensus 773 sequences of telomeric TE families. (E) Fold change (FC) of the levels (RPM) of 774 piRNA mapping to telomeric TE families . Small RNA libraries for panx KO and the 775 heterozygous control conditions (GSE71374) are included (Yu et al, 2015a). 776 777 Figure 7. Genome-wide Set1-binding analysis, and a model for Set1 function in TE 778 silencing 779 (A) Venn diagram for genes and TEs having GFP-Set1WT- or GFP-Set1E1613K-binding 780 peaks. Peak calling by MACS2 (p<0.01). (B) Scatter plot comparing fold differences of 781 CUT&Tag signals (GFP-Set1E1613K - GFP GFP -Set1WT - GFP) on genes and TEs 782 identified by peak calling . r = pearson correlation coefficient. (C, D) Track view for 783 CUT&Tag signals of GFP proteins (green) and H3K4me3 (black). aub (C) and HeT-A 784 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint (D) represent genes or TEs, respectively. TSS: transcription starting site. (E) Track view 785 for CUT&Tag signals of GFP proteins (green) and piRNA levels on subtelomeric 786 piRNA cluster 3. piRNAs having genomic plus strand sequence (Blue) or minus strand 787 sequences (Red) are separately shown . (F) A model for Set1 -mediated production of 788 antisense-biased piRNAs on subtelomeric cluster loci and on individual TE insertions 789 outside of cluster loci. Subtelomere regions maintain fragments of 3’UTR derived from 790 HTT, which are bound by Set1 to initiate antisense transcription. Catalytic activity of 791 Set1 is dispensable. Similar mechanism is applied for antisense piRNA production from 792 potentially mobile TE insertions outside of piRNA cluster loci . Antisense piRNAs are 793 loaded onto Piwi to reinforce transcriptional silencing. 794 795 Figure S1. Optimization of germline PIWI proximity labelling and purification 796 (A) Comparison of biotinylation efficiency and biotinylated protein purification 797 between non-GSCLC and GSCLC ovaries. Co omassie brilliant blue (CBB) staining 798 serving as protein loading control. Ovaries expressing mTurbo -GFP or 799 mTurbo-GFP-Aub were analyzed with negative control (without mTurbo protein 800 expression). Asterisks : non-specific signals. (B) Schematic illustration f or germline 801 differentiation and the function of Bam. Left; the germarium in wild -type, non-GSCLC 802 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint ovaries. A germline stem cell (GSC, blue) divides to produce a daughter cell (red) which 803 expresses Bam and starts the differentiation. Two-to-three GSCs are normally 804 maintained in a germarium. Right: by bam-GLKD, daughter cells cannot differentiate 805 and thus accumulate as GSCLCs. (C) A pair of wild -type ovaries (y w, left) and that of 806 bam-GLKD ovaries (right). Scale bar = 1 mm. (D) Immunoblotting images com paring 807 the expression level of mTurbo -GFP-tagged Piwi or Aub with that of endogenous Piwi 808 or Aub in GSCLC ovaries. 809 810 Figure S2. Quality assessment of PIWI proximity proteome data 811 V olcano plots for the proximity factors of Piwi, Aub, or Ago3 in individual biological 812 replicates. Red dots highlight the factors showing significant enrichment (abundance 813 ratio > 1, adjusted p < 0.05). Statistics relies on background-based t-test. 814 815 Figure S3. Germline knockdown (GLKD) efficiency and the effect on protein 816 subcellular localization 817 (A-C) Plotting the fluorescence signals from egg chambers of indicated genotypes, 818 related to Figure 3B. (A) Krimp (green) and DAPI (magenta) , (B) Aub (green) and 819 DAPI (magenta), (C) Piwi (green) and DAPI (magenta) . (D) Transcript levels of aub, 820 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint hrb27C and set1 in individual GLKD conditions (ΔΔCt method using rp49 as reference). 821 Error bars indicate standard deviation (SD) from three biological replicates. p values 822 from unpaired Student's t-test. 823 824 Figure S4. GLKD effect on ovary transcriptome, and somatic KD effect on TEs 825 (A-B) Transcriptome analysis and comparison in non-GSCLC ovaries (A) or in GSCLC 826 ovaries (B). Transcript levels (TPM) of genes and TE insertions in set1-, hrb27C-, aub-, 827 or piwi-GLKD conditions were compared with those in the control gfp-GLKD. Dots in 828 the MA plots; genes (gray) and TEs (gray with red outline) included in the differential 829 expression analysis (EdgeR). (C) Transcript levels (TPM) of i ndividual TE families in 830 gfp-GLKD (consensus sequence mapping) compared between non-GSCLC and GSCLC 831 ovaries. (D) Scatter plot c omparing the effect of GLKD on the transcript levels of 832 individual TE families (consensus sequence mapping). Left : hrb27C-GLKD compared 833 with aub-GLKD. Right : hrb27C-GLKD compared with piwi-GLKD. F old change s 834 (TPM/TPM) from control gfp-GLKD are shown. r = pearson correlation coefficient. (E) 835 Transcript levels (TPM) for TART-A1, TART-B, and TART-C, in different transcriptome 836 datasets. GSE103582 including aub and piwi knockout conditions, and GSE71374 837 including panx knockout condition were analyzed. Error bars indicate ± SD of 838 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint biological duplicates. (F) RT-qPCR measurement of TE transcript levels in ovaries upon 839 somatic KD of gfp, piwi or set1 using tj-Gal4 driver. ΔΔCt method using rp49 as 840 reference. Error bars indicate ±SD from three biological replicates. 841 842 Figure S5. Effect of COMPASS subunit GLKD on HeT-A expression, and effect of 843 set1-GLKD on the expression of piRNA pathway factors 844 (A) RT-qPCR measurement of HeT-A transcript levels in COMPASS subunit GLKD 845 ovaries. ΔΔCt method using rp49 as a reference. Error bars indicate ±SD from three 846 biological replicates. p value from unpaired Student's t -test is indicated. Asterisk; p < 847 0.05. (B) Bar graph shows the transcript levels (TPM) of piRNA pathway components 848 in gfp-GLKD and set1-GLKD ovaries. Error bar s indicate ± SD of biological 849 duplicates. 850 851 Figure S6. Immunoprecipitation of GFP-Set1 without crosslinking, and e ffect of 852 set1-GLKD on ping-pong signature of piRNAs 853 (A) Immunoprecipitation of GFP or GFP -Set1 proteins from ovaries using anti -GFP 854 antibody (GFP -IP) without crosslinking treatment. Tubulin serves as loading control. 855 (B) Ping-pong signature analysis on piRNAs mapping to telomeric TE families . Z10 856 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint scores (reflecting 10-nt overlap frequenc y within piRNA groups) were highlighted . 857 Error bars indicate ±SD of biological duplicates. 858 859 Figure S7. Genome-wide Set1-binding analysis 860 (A) Venn diagram shows overlap of genes and TEs between replicate 1 and 2 of 861 GFP-Set1WT CUT&Tag (peak calling using GFP control) . (B) Venn diagram shows 862 overlap of genes and TEs between replicate 1 and 2 of GFP -Set1E1613K CUT&Tag (peak 863 calling using GFP control) . (C) Track view for CUT&Tag signals of GFP proteins 864 (green) and H3K4me3 (black) on TAHRE (left) and TART-C (right). (D) Track view for 865 CUT&Tag signals of GFP proteins (green) and H3K4me3 (black) on HMS-Beagle (left) 866 and Max-element (right). (E) Alignment of HeT-A{}6278 to HeT-A consensus sequence. 867 (F) Track view for CUT&Tag signals of GFP proteins (green) and piRNA levels on 868 42AB locus. Blue and Red indicate piRNAs containing genomic plus and minus strand 869 sequences, respectively. 870 871 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint

872 Ables ET, Hwang GH, Finger DS, Hinnant TD & Drummond-Barbosa D (2016) A Genetic Mosaic 873 Screen Reveals Ecdysone-Responsive Genes Regulating Drosophila Oogenesis. G3 874 Genes|Genomes|Genetics 6: 2629–2642 875 Akkouche A, Kneuss E, Bornelöv S, Renaud Y, Eastwood EL, van Lopik J, Gueguen N, Jiang 876 M, Creixell P, Maupetit-Mehouas S, et al (2025) Binding of heterochromatin 877 protein Rhino to a subset of piRNA clusters depends on a combination of two histone 878 marks. Nat Struct Mol Biol 32: 1517–1527 879 Akkouche A, Mugat B, Barckmann B, Varela-Chavez C, Li B, Raffel R, Pélisson A & 880 Chambeyron S (2017) Piwi Is Required during Drosophila Embryogenesis to License 881 Dual-Strand piRNA Clusters for Transposon Repression in Adult Ovaries. Mol Cell 882 66: 411-419.e4 883 Alizada A, Martins A, Mouniée N, Rodriguez Suarez J V., Bertin B, Gueguen N, Mirouse 884 V, Papameletiou A-M, Rivera AJ, Lau NC, et al (2025) The transcription factor 885 Traffic jam orchestrates the somatic piRNA pathway in Drosophila ovaries. Cell 886 Rep 44: 115453 887 Anand A & Kai T (2012) The tudor domain protein Kumo is required to assemble the nuage 888 and to generate germline piRNAs in Drosophila. EMBO J 31: 870–882 889 Andersen PR, Tirian L, Vunjak M & Brennecke J (2017) A heterochromatin-dependent 890 transcription machinery drives piRNA expression. Nature 549: 54–59 891 Anderson JT, Henikoff S & Ahmad K (2023) Chromosome-specific maturation of the 892 epigenome in the Drosophila male germline. Elife 12 893 Antoniewski C (2014) Computing siRNA and piRNA Overlap Signatures. In Methods in 894 Molecular Biology pp 135–146. Humana Press Inc. 895 Ardehali MB, Mei A, Zobeck KL, Caron M, Lis JT & Kusch T (2011) Drosophila Set1 is 896 the major histone H3 lysine 4 trimethyltransferase with role in transcription. 897 EMBO J 30: 2817–2828 898 Ariura M, Solberg T, Ishizu H, Takahashi H, Carninci P, Siomi H & Iwasaki YW (2024) 899 Drosophila Piwi distinguishes transposons from mRNAs by piRNA complementarity 900 and abundance. Cell Rep 43: 115020 901 Bae HJ, Dubarry M, Jeon J, Soares LM, Dargemont C, Kim J, Geli V & Buratowski S (2020) 902 The Set1 N-terminal domain and Swd2 interact with RNA polymerase II CTD to recruit 903 COMPASS. Nat Commun 11: 2181 904 Batki J, Schnabl J, Wang J, Handler D, Andreev VI, Stieger CE, Novatchkova M, 905 Lampersberger L, Kauneckaite K, Xie W, et al (2019) The nascent RNA binding 906 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint complex SFiNX licenses piRNA-guided heterochromatin formation. Nat Struct Mol 907 Biol 26: 720–731 908 Baumgartner L, Handler D, Platzer SW, Yu C, Duchek P & Brennecke J (2022) The Drosophila 909 ZAD zinc finger protein Kipferl guides Rhino to piRNA clusters. Elife 11 910 Berretta J, Pinskaya M & Morillon A (2008) A cryptic unstable transcript mediates 911 transcriptional trans -silencing of the Ty1 retrotransposon in S. cerevisiae. 912 Genes Dev 22: 615–626 913 Branon TC, Bosch JA, Sanchez AD, Udeshi ND, Svinkina T, Carr SA, Feldman JL, Perrimon 914 N & Ting AY (2018) Efficient proximity labeling in living cells and organisms 915 with TurboID. Nat Biotechnol 36: 880–887 916 Brennecke J, Aravin AA, Stark A, Dus M, Kellis M, Sachidanandam R & Hannon GJ (2007a) 917 Discrete Small RNA-Generating Loci as Master Regulators of Transposon Activity 918 in Drosophila. Cell 128: 1089–1103 919 Brennecke J, Aravin AA, Stark A, Dus M, Kellis M, Sachidanandam R & Hannon GJ (2007b) 920 Discrete Small RNA-Generating Loci as Master Regulators of Transposon Activity 921 in Drosophila. Cell 128: 1089–1103 922 Choi‐Rhee E, Schulman H & Cronan JE (2004) Promiscuous protein biotinylation by 923 Escherichia coli biotin protein ligase. Protein Science 13: 3043–3050 924 Danilevskaya ON & Arkhipova IR (1997) Promoting in Tandem: The Promoter for Telomere 925 Transposon HeT-A and Implications for the Evolution of Retroviral LTRs 926 ElMaghraby MF, Andersen PR, Pühringer F, Hohmann U, Meixner K, Lendl T, Tirian L & 927 Brennecke J (2019) A Heterochromatin-Specific RNA Export Pathway Facilitates 928 piRNA Production. Cell 178: 964-979.e20 929 Fabry MH, Ciabrelli F, Munafò M, Eastwood EL, Kneuss E, Falciatori I, Falconio FA, 930 Hannon GJ & Czech B (2019) piRNA-guided co-transcriptional silencing coopts 931 nuclear export factors. Elife 8 932 Finger DS, Williams AE, Holt V V & Ables ET (2023) Novel roles for RNA binding proteins 933 squid , hephaesteus , and Hrb27C in Drosophila oogenesis. Developmental Dynamics 934 252: 415–428 935 Ge DT, Wang W, Tipping C, Gainetdinov I, Weng Z & Zamore PD (2019) The RNA-Binding 936 ATPase, Armitage, Couples piRNA Amplification in Nuage to Phased piRNA Production 937 on Mitochondria. Mol Cell 74: 982-995.e6 938 Goodrich JS, Clouse KN & Schüpbach T (2004) Hrb27C, Sqd and Otu cooperatively regulate 939 gurken RNA localization and mediate nurse cell chromosome dispersion in 940 Drosophila oogenesis. Development 131: 1949–1958 941 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Grieder NC, Cuevas M de & Spradling AC (2000) The fusome organizes the microtubule 942 network during oocyte differentiation in Drosophila. Development 127: 4253–4264 943 Hallson G, Hollebakken RE, Li T, Syrzycka M, Kim I, Cotsworth S, Fitzpatrick KA, 944 Sinclair DAR & Honda BM (2012) dSet1 Is the Main H3K4 Di- and 945 Tri-Methyltransferase Throughout Drosophila Development. Genetics 190: 91–100 946 Handler D, Olivieri D, Novatchkova M, Gruber FS, Meixner K, Mechtler K, Stark A, 947 Sachidanandam R & Brennecke J (2011) A systematic analysis of Drosophila TUDOR 948 domain-containing proteins identifies Vreteno and the Tdrd12 family as essential 949 primary piRNA pathway factors. EMBO J 30: 3977–3993 950 Huynh J-R, Munro TP, Smith-Litière K, Lepesant J-A & Johnston DS (2004) The Drosophila 951 hnRNPA/B Homolog, Hrp48, Is Specifically Required for a Distinct Step in osk mRNA 952 Localization. Dev Cell 6: 625–635 953 Iki T, Kawaguchi S & Kai T (2023) miRNA/siRNA-directed pathway to produce noncoding 954 piRNAs from endogenous protein-coding regions ensures Drosophila 955 spermatogenesis. Sci Adv 9 956 Iki T, Takami M & Kai T (2020) Modulation of Ago2 Loading by Cyclophilin 40 Endows 957 a Unique Repertoire of Functional miRNAs during Sperm Maturation in Drosophila. 958 Cell Rep 33: 108380 959 Iyer SS, Sun Y, Seyfferth J, Manjunath V, Samata M, Alexiadis A, Kulkarni T, Gutierrez 960 N, Georgiev P, Shvedunova M, et al (2023) The NSL complex is required for piRNA 961 production from telomeric clusters. Life Sci Alliance 6: e202302194 962 Jezek M, Sun W, Negesse MY, Smith ZM, Orosz A & Green EM (2023) Set1 regulates telomere 963 function via H3K4 methylation–dependent and -independent pathways and calibrates 964 the abundance of telomere maintenance factors. Mol Biol Cell 34 965 Kai T, Zheng J & Iki T (2025) Paralog‐Dependent Specialization of Paf1C 966 Subunit, Ctr9, for Sex Chromosome Gene Regulation and Male Germline 967 Differentiation in Drosophila. Genes to Cells 30: e70040 968 Kawaguchi S, Isshiki W & Kai T (2025) Factories without walls: The molecular 969 architecture and functions of non-membrane organelles in small RNA-guided genome 970 protection. Biochimica et Biophysica Acta (BBA) - General Subjects 1869: 130811 971 Klattenhoff C, Xi H, Li C, Lee S, Xu J, Khurana JS, Zhang F, Schultz N, Koppetsch 972 BS, Nowosielska A, et al (2009) The Drosophila HP1 Homolog Rhino Is Required for 973 Transposon Silencing and piRNA Production by Dual-Strand Clusters. Cell 138: 974 1137–1149 975 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Kneuss E, Munafò M, Eastwood EL, Deumer U-S, Preall JB, Hannon GJ & Czech B (2019) 976 Specialization of the Drosophila nuclear export family protein Nxf3 for piRNA 977 precursor export. Genes Dev 33: 1208–1220 978 Koch R, Ledermann R, Urwyler O, Heller M & Suter B (2009) Systematic Functional 979 Analysis of Bicaudal-D Serine Phosphorylation and Intragenic Suppression of a 980 Female Sterile Allele of BicD. PLoS One 4: e4552 981 Lee KY, Chen Z, Jiang R & Meneghini MD (2018) H3K4 Methylation Dependent and 982 Independent Chromatin Regulation by JHD2 and SET1 in Budding Yeast. G3 983 Genes|Genomes|Genetics 8: 1829–1839 984 Lim AK & Kai T (2007) Unique germ-line organelle, nuage, functions to repress selfish 985 genetic elements in Drosophila melanogaster. Proceedings of the National Academy 986 of Sciences 104: 6714–6719 987 Lim L-X, Isshiki W, Iki T, Kawaguchi S & Kai T (2022) The Tudor Domain-Containing 988 Protein, Kotsubu (CG9925), Localizes to the Nuage and Functions in piRNA 989 Biogenesis in D. melanogaster. Front Mol Biosci 9: 1–15 990 Lin Y, Suyama R, Kawaguchi S, Iki T & Kai T (2023) Tejas functions as a core component 991 in nuage assembly and precursor processing in Drosophila piRNA biogenesis. 992 Journal of Cell Biology 222 993 Lorenz DR, Mikheyeva I V., Johansen P, Meyer L, Berg A, Grewal SIS & Cam HP (2012) 994 CENP-B Cooperates with Set1 in Bidirectional Transcriptional Silencing and 995 Genome Organization of Retrotransposons. Mol Cell Biol 32: 4215–4225 996 Ma X, Zhu X, Han Y, Story B, Do T, Song X, Wang S, Zhang Y, Blanchette M, Gogol M, 997 et al (2017) Aubergine Controls Germline Stem Cell Self-Renewal and Progeny 998 Differentiation via Distinct Mechanisms. Dev Cell 41: 157-169.e5 999 Maxwell PH, Belote JM & Levis RW (2006) Identification of multiple transcription 1000 initiation, polyadenylation, and splice sites in the Drosophila melanogaster 1001 TART family of telomeric retrotransposons. Nucleic Acids Res 34: 5498–5507 1002 McKearin DM & Spradling AC (1990) bag-of-marbles: a Drosophila gene required to 1003 initiate both male and female gametogenesis. Genes Dev 4: 2242–2251 1004 Mohan M, Herz H-M, Smith ER, Zhang Y, Jackson J, Washburn MP, Florens L, Eissenberg 1005 JC & Shilatifard A (2011) The COMPASS Family of H3K4 Methylases in Drosophila. 1006 Mol Cell Biol 31: 4310–4318 1007 Mohn F, Sienski G, Handler D & Brennecke J (2014) The Rhino-Deadlock-Cutoff complex 1008 licenses noncanonical transcription of dual-strand piRNA clusters in Drosophila. 1009 Cell 157: 1364–1379 1010 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Mugat B, Nicot S, Varela-Chavez C, Jourdan C, Sato K, Basyuk E, Juge F, Siomi MC, 1011 Pélisson A & Chambeyron S (2020) The Mi-2 nucleosome remodeler and the Rpd3 1012 histone deacetylase are involved in piRNA-guided heterochromatin formation. Nat 1013 Commun 11: 2818 1014 Murano K, Iwasaki YW, Ishizu H, Mashiko A, Shibuya A, Kondo S, Adachi S, Suzuki S, 1015 Saito K, Natsume T, et al (2019) Nuclear RNA export factor variant initiates 1016 piRNA‐guided co‐transcriptional silencing. EMBO J 38 1017 Ng HH, Robert F, Young RA & Struhl K (2003) Targeted Recruitment of Set1 Histone 1018 Methylase by Elongating Pol II Provides a Localized Mark and Memory of Recent 1019 Transcriptional Activity. Mol Cell 11: 709–719 1020 Nguyen TTM, Munkhzul C, Kim J, Kyoung Y, Vianney M, Shin S, Ju S, Pham-Bui H-A, Kim 1021 J, Kim J-S, et al (2023) In vivo profiling of the Zucchini proximal proteome in 1022 the Drosophila ovary. Development 150 1023 Nislow C, Ray E & Pillus L (1997) SET1 , A Yeast Member of the Trithorax Family, 1024 Functions in Transcriptional Silencing and Diverse Cellular Processes. Mol Biol 1025 Cell 8: 2421–2436 1026 Osumi K, Sato K, Murano K, Siomi H & Siomi MC (2019) Essential roles of Windei and 1027 nuclear monoubiquitination of Eggless/ SETDB 1 in transposon 1028 silencing. EMBO Rep 20 1029 Patil VS & Kai T (2010) Repression of Retroelements in Drosophila Germline via piRNA 1030 Pathway by the Tudor Domain Protein Tejas. Current Biology 20: 724–730 1031 Qi H, Watanabe T, Ku H-Y, Liu N, Zhong M & Lin H (2011) The Yb Body, a Major Site 1032 for Piwi-associated RNA Biogenesis and a Gateway for Piwi Expression and 1033 Transport to the Nucleus in Somatic Cells. Journal of Biological Chemistry 286: 1034 3789–3797 1035 Radion E, Ryazansky S, Akulenko N, Rozovsky Y, Kwon D, Morgunova V, Olovnikov I & 1036 Kalmykova A (2017) Telomeric Retrotransposon HeT-A Contains a Bidirectional 1037 Promoter that Initiates Divergent Transcription of piRNA Precursors in 1038 Drosophila Germline. J Mol Biol 429: 3280–3289 1039 Rangan P, Malone CD, Navarro C, Newbold SP, Hayes PS, Sachidanandam R, Hannon GJ & 1040 Lehmann R (2011) piRNA Production Requires Heterochromatin Formation in 1041 Drosophila. Current Biology 21: 1373–1379 1042 Rivera AJ, Lee J-HR, Gupta S, Yang L, Goel RK, Zaia J & Lau NC (2025) Traffic Jam 1043 activates the Flamenco piRNA cluster locus and the Piwi pathway to ensure 1044 transposon silencing and Drosophila fertility. Cell Rep 44: 115354 1045 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Rojas-Ríos P, Chartier A, Enjolras C, Cremaschi J, Garret C, Boughlita A, Ramat A 1046 & Simonelig M (2024) piRNAs are regulators of metabolic reprogramming in stem 1047 cells. Nat Commun 15: 8405 1048 Rojas‐Ríos P, Chartier A, Pierson S & Simonelig M (2017) Aubergine and pi 1049 RNA s promote germline stem cell self‐renewal by repressing the 1050 proto‐oncogene Cbl. EMBO J 36: 3194–3211 1051 Rørth P (1998) Gal4 in the Drosophila female germline. Mech Dev 78: 113–118 1052 Roux KJ, Kim DI, Raida M & Burke B (2012) A promiscuous biotin ligase fusion protein 1053 identifies proximal and interacting proteins in mammalian cells. Journal of Cell 1054 Biology 196: 801–810 1055 Saito K, Ishizu H, Komai M, Kotani H, Kawamura Y, Nishida KM, Siomi H & Siomi MC (2010) 1056 Roles for the Yb body components Armitage and Yb in primary piRNA biogenesis in 1057 Drosophila. Genes Dev 24: 2493–2498 1058 Saito K, Nishida KM, Mori T, Kawamura Y, Miyoshi K, Nagami T, Siomi H & Siomi MC (2006) 1059 Specific association of Piwi with rasiRNAs derived from retrotransposon and 1060 heterochromatic regions in the Drosophila genome. Genes Dev 20: 2214–2222 1061 Shpiz S, Olovnikov I, Sergeeva A, Lavrov S, Abramov Y, Savitsky M & Kalmykova A (2011) 1062 Mechanism of the piRNA-mediated silencing of Drosophila telomeric 1063 retrotransposons. Nucleic Acids Res 39: 8703–8711 1064 Sienski G, Batki J, Senti K-A, Dönertas D, Tirian L, Meixner K & Brennecke J (2015a) 1065 Silencio/CG9754 connects the Piwi–piRNA complex to the cellular heterochromatin 1066 machinery. Genes Dev 29: 2258–2271 1067 Sienski G, Batki J, Senti KA, Dönertas D, Tirian L, Meixner K & Brennecke J (2015b) 1068 Silencio/CG9754 connects the piwi-piRNA complex to the cellular heterochromatin 1069 machinery. Genes Dev 29: 2258–2271 1070 Suyama R & Kai T (2025) pi RNA processing within non‐membrane structures 1071 is governed by constituent proteins and their functional motifs. FEBS J 292: 2715–1072 2736 1073 Teixeira FK, Okuniewska M, Malone CD, Coux R-X, Rio DC & Lehmann R (2017a) 1074 piRNA-mediated regulation of transposon alternative splicing in the soma and germ 1075 line. Nature 552: 268–272 1076 Teixeira FK, Okuniewska M, Malone CD, Coux RX, Rio DC & Lehmann R (2017b) 1077 PiRNA-mediated regulation of transposon alternative splicing in the soma and germ 1078 line. Nature 552: 268–272 1079 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Le Thomas A, Rogers AK, Webster A, Marinov GK, Liao SE, Perkins EM, Hur JK, Aravin 1080 AA & Tóth KF (2013) Piwi induces piRNA-guided transcriptional silencing and 1081 establishment of a repressive chromatin state. Genes Dev 27: 390–399 1082 Vidaurre V, Song A, Li T, Ku WL, Zhao K, Qian J & Chen X (2024) The Drosophila histone 1083 methyltransferase SET1 coordinates multiple signaling pathways in regulating 1084 male germline stem cell maintenance and differentiation. Development 151 1085 Villasante A, Abad JP, Planelló R, Méndez-Lago M, Celniker SE & de Pablos B (2007) 1086 Drosophila telomeric retrotransposons derived from an ancestral element that was 1087 recruited to replace telomerase. Genome Res 17: 1909–1918 1088 Wang H, Fan Z, Shliaha P V., Miele M, Hendrickson RC, Jiang X & Helin K (2023a) H3K4me3 1089 regulates RNA polymerase II promoter-proximal pause-release. Nature 615: 339–1090 348 1091 Wang L, Dou K, Moon S, Tan FJ & Zhang ZZ (2018) Hijacking Oogenesis Enables Massive 1092 Propagation of LINE and Retroviral Transposons. Cell 174: 1082-1094.e12 1093 Wang X, Ramat A, Simonelig M & Liu M-F (2023b) Emerging roles and functional mechanisms 1094 of PIWI-interacting RNAs. Nat Rev Mol Cell Biol 24: 123–141 1095 Webster A, Li S, Hur JK, Wachsmuth M, Bois JS, Perkins EM, Patel DJ & Aravin AA (2015) 1096 Aub and Ago3 Are Recruited to Nuage through Two Mechanisms to Form a Ping-Pong 1097 Complex Assembled by Krimper. Mol Cell 59: 564–575 1098 Wells JN & Feschotte C (2020) A Field Guide to Eukaryotic Transposable Elements. Annu 1099 Rev Genet 54: 539–561 1100 Wilson JR, Jing C, Walker PA, Martin SR, Howell SA, Blackburn GM, Gamblin SJ & Xiao 1101 B (2002) Crystal Structure and Functional Analysis of the Histone 1102 Methyltransferase SET7/9. Cell 111: 105–115 1103 Yan D, Neumüller RA, Buckner M, Ayers K, Li H, Hu Y, Yang-Zhou D, Pan L, Wang X, Kelley 1104 C, et al (2014) A regulatory network of Drosophila germline stem cell self-renewal. 1105 Dev Cell 28: 459–473 1106 Yang F, Su W, Chung OW, Tracy L, Wang L, Ramsden DA & Zhang ZZ (2023) Retrotransposons 1107 hijack alt-EJ for DNA replication and eccDNA biogenesis. Nature 620: 218–225 1108 Yano T, de Quinto SL, Matsui Y, Shevchenko A, Shevchenko A & Ephrussi A (2004) Hrp48, 1109 a Drosophila hnRNPA/B Homolog, Binds and Regulates Translation of oskar mRNA. 1110 Dev Cell 6: 637–648 1111 Yu Y, Gu J, Jin Y, Luo Y, Preall JB, Ma J, Czech B & Hannon GJ (2015a) Panoramix enforces 1112 piRNA-dependent cotranscriptional silencing. Science (1979) 350: 339–342 1113 Yu Y, Gu J, Jin Y, Luo Y, Preall JB, Ma J, Czech B & Hannon GJ (2015b) Panoramix enforces 1114 piRNA-dependent cotranscriptional silencing. Science (1979) 350: 339–342 1115 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Zhang Z, Xu J, Koppetsch BS, Wang J, Tipping C, Ma S, Weng Z, Theurkauf WE & Zamore 1116 PD (2011) Heterotypic piRNA Ping-Pong Requires Qin, a Protein with Both E3 Ligase 1117 and Tudor Domains. Mol Cell 44: 572–584 1118 Zhao K, Cheng S, Miao N, Xu P, Lu X, Zhang Y, Wang M, Ouyang X, Yuan X, Liu W, et 1119 al (2019) A Pandas complex adapted for piRNA-guided transcriptional silencing 1120 and heterochromatin formation. Nat Cell Biol 21: 1261–1272 1121 Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, Benner C & Chanda 1122 SK (2019) Metascape provides a biologist-oriented resource for the analysis of 1123 systems-level datasets. Nat Commun 10: 1523 1124 1125 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Figure 1 A mTurbo GFP PIWI UASp (Piwi, Aub or Ago3) Gal4 B GSCLC ovaries mTurbo-GFP-PiwimTurbo-GFP mTurbo-GFP-Aub mTurbo-GFP-Ago3 GFP DAPI GSCLC ovaries (bam-GLKD) expressing: anti-Piwi DAPI anti-Aub DAPI anti-Ago3 DAPIbam shRNAUASp Gal4 ♀ Gal4 nos promoter Input Streptavidin pulldown 180 140 100 75 60 anti-GFP 180 140 100 75 60 45 35 Streptavidin-HRP C (kDa) mTurbo-GFP mTurbo-GFP-Piwi mTurbo-GFP-Aub mTurbo-GFP-Ago3 mTurbo-GFP mTurbo-GFP-Piwi mTurbo-GFP-Aub mTurbo-GFP-Ago3 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Figure 2 A B C .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint 0 5 0 0. 5 1 0 5 Figure 3 aub-GLKDy w hrb27C-GLKD set1-GLKD Krimp HeT-A Gag A B aub-GLKDy w hrb27C-GLKD set1-GLKD 0 5 0 5 0 1 2 0 5 0 5 0 5 Normalized signal Distance (µm) 0 5 DAPIAubDAPI Normalized signal DAPI Normalized signal PiwiDAPI Distance (µm) Distance (µm) Distance (µm) 0 1 0 5 10 0 5 10 0 5 100 5 10 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint GSE71374 gfp-GLKD set1-GLKD aub-GLKD piwi-GLKD panx het panx KO 0 10 00 20 00 30 00 0 20 0 40 0 60 0 B 80 40 0 120 160 200 log2FC (hrb27C-GLKD / gfp-GLKD) log2FC (aub-GLKD / gfp-GLKD)log2FC (set1-GLKD / gfp-GLKD) HeT-A HMS-Beagle TART-B HeT-A gypsy12 log2FC (hrb27C-GLKD / gfp-GLKD) Figure 4 A C Non-GSCLC ovary (Genes and individual TE insertions) D GSCLC ovary (Genes and individual TE insertions) -log 10 FDR -log 10 FDR log2FC (set1-GLKD / gfp-GLKD) Transcript level (TPM) r = -0.02 log2FC (aub-GLKD / gfp-GLKD) log 2FC ( set1 -GL KD /gfp -GLKD ) TEs included in EdgeR output Genes and TEs included in EdgeR output TAHRE Non-GSCLC ovary (Individual TE families) log2FC (piwi-GLKD / gfp-GLKD) log2FC (piwi-GLKD / gfp-GLKD) log2FC (aub-GLKD / gfp-GLKD) 80 40 0 100 120 140 60 20 -10 -5 0 5 10 15 -10 -5 0 5 10 15 -10 -5 0 5 10 15 TEs included in EdgeR output Genes and TEs included in EdgeR output -10 -5 0 5 10 15 -10 -5 0 5 10 15 -10 -5 0 5 10 15 -10 -5 0 5 10 15 -10 -5 0 5 10 15 log2FC (piwi-GLKD / gfp-GLKD) mdg3 DM88 copia Circe mdg3HMS-Beagle flea Max-element diver flea HMS-Beagle diver 4 0 -4 6 8 10 2 -2 HeT-A gypsy12 TAHRE TART-A1 TART-B TART-C Max-element 40-4 6 8 102-2 diver HMS-Beagle flea mdg3 4 0 -4 6 8 10 2 -2 40-4 6 8 102-2 r = 0.73 HeT-A TAHRE gypsy12 TART-B TART-A1 TART-C Max-element HMS- Beagle diver HeT-A HeT-A mdg3 invader3 HMS-Beagle TART-A Max-element diver gypsy12 TART-C 3S18 3S18 3S18 TART-B TART-A rover aub het aub KO piwihet piwiKO This study GSE103582 3S18 3S18 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Figure 5 A B GFPGFP-Set1WT SET RNAi-resistant nonsense substitutions GFPGFP-Set1E1613K SET E1613K Transcript level (mean TPM) GLKD GFP set1 set1 set1 GFP-set1WT GFP-set1E1613K Transgene400 300 200 100 0 400 300 200 100 0 3000 2000 1000 0 HeT-A TAHRE TART-A1 TART-B TART-C diver HMS- Beagle 3S18 Max- element Telomeric TE families C D [0-50] [0-50] [0-50] [0-50] GLKD GFP set1 set1 set1 GFP-set1WT GFP-set1E1613K Transgene ago3 ago3 aub piwi 200 160 120 80 40 0 H3K4me3 signal (mean CPM) aub [0-15] [0-15] [0-15] [0-15] TSS RRM RRM GFPH3K4me3Vas set1-GLKDgfp-GLKD GFP-Set1WT in set1-GLKD GFP-Set1E1613K in set1-GLKD H3K4me3 (CUT&Tag) TSS .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint HMS-Beagle (0.51) diver (0.59) 0 60 120 Figure 6 D log2FC 23-29nt mapper (set1-GLKD / control) C B Control set1 GLKD 6,063 bp HeT-A 10,463bp TAHRE 13,424bp TART-A1 0 1 2 -1 -2 -3 0 60 120 Individual TE families ordered by FC value Proportion (%) Control set1-GLKD 60 50 40 30 20 10 0 0 60 120 Sense Antisense Sense+Antisense TART-C (0.16) TART-A1 (0.24) TART-B (0.35) TAHRE (0.49) HeT-A (1.10) TART-C (0.18) TART-A1 (0.19) TAHRE (0.20) HeT-A (0.23) TART-B (0.28) TART-C (0.16) TART-A1 (0.20) TART-B (0.29) TAHRE (0.31) HeT-A (0.51) Max-element (1.36) diver (1.60) HMS-Beagle (1.86) gypsy12 (0.72) gypsy12, 3S18 (0.52) Max-element (0.71) gypsy12 (0.66) Max-element (0.84) HMS-Beagle (0.99) diver (1.01) A E set1-GLKD / Control Sense Antisense log2FC 23-29nt mapper Sense Antisense HeT-A TAHRE TART-A1 TART-B TART-C 1 0 -1 -2 -3 1 0 -1 -2 -3 HeT-A TAHRE TART-A1 TART-B TART-C HeT-A TAHRE TART-A1 TART-B TART-C HeT-A TAHRE TART-A1 TART-B TART-C 3S18 (2.04) 3S18 (1.13) CPM [0-250] [0-250] CPM [0-250] [0-250] 10,176bp TART-B Sense Antisense Sense Antisense panx KO / Het (GSE71374) 11,124bp TART-C [0-250] [0-250] [0-250] [0-250] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] [0-50] WT EKGFP GFP-Set1 Piw i GFP-Set1 GFP Tubulin Input WT EKGFP GFP-Set1 GFP-IP (Crosslinking) 245 35 100 60 piRNA .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint Subtelomeric HTT cluster set1-GLKD Figure 7 A B TE 84 (HeT-A, TART, diver, HMS-Beagle, Max, etc) Gene 222 TE 7 Gene 541 (aub, etc) TE 1 Gene 91 Peak call (p<0.01) GFP-Set1 (WT or E1613K) / GFP control -4 -2 0 2 4 6 8 10 12 -4 -2 0 2 4 6 8 10 12 log2 fold difference (GFPSet1WT - GFP) TEs with peaks Genes and TEs with peaks C D GFP GFP-Set1-WT in set1-GLKD H3K4me3 gfp-GLKD [0-15] [0-15] [0-15] [0-15] [0-50] [0-50] [0-50] ORF1 TSSHeT-A (consensus) TSS aub [0-500] [0-500] [0-500] [0-500] [0-50] [0-50] [0-50] 5’ 3’ GFP-Set1-EK in set1-GLKD GFP-Set1-WT in set1-GLKD GFP-Set1-EK in set1-GLKD GFP E CUT&Tag Cluster 3 (chr4:1,267,100 -1,348,250) TE Gene track GFP GFP-Set1-WT in set1-GLKD [0-15] [0-15] [0-15] GFP-Set1-EK in set1-GLKD GFP(CUT&Tag ) Control [0-10] plus st rand (antisense) [0-10] minus strand (sense) [0-10] plus st rand (antisense) [0-10] minus strand (sense) set1-GLKD piRNA 6083 F Precursors piRNA (Sense < Antisense) 3’UTR HTT Piwi TE Individual TE insertions Telomere Precursor Antisense piRNA Piwi Set1 log2 fold difference (GFPSet1E1613K - GFP) r = 0.85 Set1Set1 Set1 Set1 Set1 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 30, 2026. ; https://doi.org/10.64898/2026.03.30.715253doi: bioRxiv preprint