The Drosophila histone methyltransferase SET1 coordinates multiple signaling pathways in regulating male germline stem cell maintenance and differentiation

In multicellular organisms, homeostasis and regeneration of many tissues largely depend on adult stem cells. These endogenous stem cells often undergo asymmetric cell divisions to allow both the stem cell to maintain its own population through self-renewal and the differentiation process to replace cells lost under physiological and pathological conditions (Kahney et al., 2019; Knoblich, 2008; Morrison and Spradling, 2008; Venkei and Yamashita, 2018). Disruption of these processes can potentially result in the misregulation of stem cell activity, leading to cancer or tissue degeneration (Clevers, 2005; Knoblich, 2010; Morrison and Kimble, 2006; Zion et al., 2020). There are two main modes of dysregulation in adult stem cell lineages that can result in the unrestrained cell proliferation that underlies tumorigenesis: constraints of normal stem cell expansion may be compromised or inactivated. This situation could occur if the dependence of stem cells on the niche is disrupted, resulting in a niche-independent overpopulation of stem cells or enhancement of stem cell activities. Alternatively, the transit-amplifying cells could fail to exit proliferation and enter the terminal differentiation program (Clarke and Fuller, 2006; Mukherjee et al., 2015; Sell, 2010; Zhang and Hsu, 2017).

Drosophila spermatogenesis is an excellent model system in which to study stem cell proliferation and differentiation (Davies and Fuller, 2008; Fuller, 1993; Gleason et al., 2018). Spermatogenesis is initiated with asymmetric division of the germline stem cell (GSC) to produce a self-renewed GSC and a gonialblast (Fig. 1A). The gonialblast undergoes four rounds of transit-amplifying mitotic divisions as spermatogonial cells. After mitosis, the 16 spermatogonial cells enter meiosis with a prolonged G2 phase as the primary spermatocytes. In addition to the GSC-derived germline lineage, the testis also has at least two somatic cell populations: the hub cells and the cyst cells (de Cuevas and Matunis, 2011). The hub cells are post-mitotic and they support the GSCs and the cyst stem cells (CySCs) (Kiger et al., 2001; Leatherman, 2013; Leatherman and Dinardo, 2010; Losick et al., 2011; Tran et al., 2000; Yamashita et al., 2003). The CySCs give rise to the cyst cells that support the germ cells throughout spermatogenesis (Leatherman and Dinardo, 2008; Lim and Fuller, 2012). The CySC also undergoes an asymmetric cell division to produce a self-renewed CySC and a cyst cell, which never divides again (Cheng et al., 2011). Two cyst cells encapsulate the germ cells as they divide and differentiate (Fig. 1A) (Fuller and Spradling, 2007; Matunis et al., 2012; Spradling et al., 2011).

Two important signaling pathways involved in the maintenance of the GSC niche are the JAnus Kinase Signal Transducer and Activator of Transcription (JAK-STAT) and Bone Morphogenetic Protein (BMP) signaling pathways (Amoyel et al., 2014; Herrera and Bach, 2019; Inaba et al., 2015; Kiger et al., 2001; Leatherman and Dinardo, 2008, 2010; Shivdasani and Ingham, 2003; Tulina and Matunis, 2001). Furthermore, epigenetic mechanisms, such as chromatin remodeling and histone modifications, have been shown to play important roles in the male germline lineage (Cherry and Matunis, 2010; Eun et al., 2017; Feng et al., 2018; Tarayrah and Chen, 2013; Tarayrah et al., 2015). In particular, multiple examples have shown the interplay between extrinsic signaling pathways and intrinsic epigenetic mechanisms in the Drosophila adult testes (Feng et al., 2017; Gleason and Chen, 2023; Vidaurre and Chen, 2021). For example, previous studies have shown that the H3K27me3 histone demethylase, Ubiquitously transcribed Tetratricopeptiderepeat gene on the X chromosome (UTX), acts as a negative regulator of JAK-STAT signaling by maintaining the transcription of Socs36E, proper expression and function of which maintain the balance between GSCs and CySCs (Amoyel et al., 2016; Issigonis et al., 2009; Singh et al., 2010; Tarayrah et al., 2013). Moreover, genes of the Epidermal Growth Factor (EGF) signaling pathway may be directly regulated by the H3K27me3 methyltransferase Enhancer of zestes [E(z)] in cyst cells, to ensure proper germ cell identity (Eun et al., 2014). Additionally, the Drosophila demethylase for H3K4me3, Little imaginal disc (Lid; Lysine demethylase 5, Kdm5), regulates JAK-STAT signaling in the male germline to maintain GSC activity and proliferation (Tarayrah et al., 2015). However, unlike the methyltransferase and demethylase for H3K27me3, the methyltransferases for H3K4me3 have not been studied in the Drosophila male germline.

SET domain containing 1 (Set1) is the main H3K4me3 methyltransferase in Drosophila, and its function is conserved from yeast to mammals (Lee et al., 2007a; Simonet et al., 2007). The post-translational modification of H3K4me3 has previously been shown to correlate with active transcription and is enriched at the promoter and the 5′ coding regions of genes (Bernstein et al., 2002; Heintzman et al., 2007; Pokholok et al., 2005; Santos-Rosa et al., 2002). The role of H3K4me3 in activating transcription is not fully understood, although it is thought that H3K4me3 acts as a docking scaffold for the transcription pre-initiation complex and chromatin remodeling complexes (Vermeulen et al., 2007; Wysocka et al., 2006). In yeast, Set1 catalyzes all methylation forms of H3K4 (i.e. H3K4me1/2/3) and its catalytic activity is modulated via a multi-subunit protein complex known as the Complex of proteins associated with Set1 (COMPASS) (Dehe et al., 2005; Miller et al., 2001; Nagy et al., 2002; Roguev et al., 2001). The Drosophila Set1 also interacts with the COMPASS complex to catalyze H3K4 di- and tri-methylation (H3K4me2/3) at the promoter proximal regions (Ardehali et al., 2011; Hallson et al., 2012; Mohan et al., 2011). In the Drosophila female germline, it has been shown that Set1 regulates GSC maintenance and differentiation; however, the mechanism has not been fully determined (Xuan et al., 2013; Yan et al., 2014).

Here, we use a series of molecular genetics and cell biology tools to study the roles of Set1 in the Drosophila testis. We found that Set1 is required for the normal function of early-stage male germ cells by regulating several key signaling pathways in a methyl-transferase-dependent manner. This study provides an example of how the intrinsic epigenetic mechanisms cooperate with the extrinsic signaling pathways to determine and maintain stem cell fate.

To explore the function of Set1 in the Drosophila male germline, a short hairpin RNA (shRNA or RNAi) specifically targeting the coding sequence of the Set1 gene driven by the early-stage germline driver nanos-Gal4 (nos-Gal4) (Van Doren et al., 1998) was employed for cell type-specific knockdown (KD) experiments. The Set1 null allele is lethal at the pupal stage and conventional mosaic analysis cannot be performed owing to the genomic location of the Set1 gene, which is very close to the centromere of chromosome 3. Therefore, the KD strategy is required to interrogate its roles in vivo (Ardehali et al., 2011; Hallson et al., 2012). In nos>set1 RNAi (Set1 KD) testes, H3K4me3 signal was greatly reduced in the early-stage germ cells, from GSCs to spermatogonial cells (Fig. S1B), compared with germ cells at the comparable stages in the nos>mCherry RNAi testes (nos>Control RNAi or Ctrl KD; Fig. S1A). In addition, in both Set1 KD and Ctrl KD testes, H3K4me3 signals were present in the somatic gonadal cells, indicating germline-specific and efficient inactivation of the methyltransferase function of Set1 (Fig. S1).

To investigate potential defects in the testis, a time course experiment was performed whereby Set1 KD and Ctrl KD flies were grown at 18°C until eclosion and then shifted to 25°C for 0, 1, 3, 5 or 7 days (Fig. 1B). We also examined an earlier development stage at the third instar larvae (L3), when male GSCs and their niche have already been established (Le Bras and Van Doren, 2006). At L3, 100% of Set1 KD and Ctrl KD testes displayed normal morphology (Fig. 1C,D). At day 0 and day 1 post-eclosion, 98% and 56%, respectively, of the Set1 KD testes, showed a germ cell loss phenotype compared with none of the Ctrl KD testes (Fig. 1C,D). At day 3 post-eclosion, although 45% of Set1 KD testes had germline loss, 55% had an early germ cell overpopulation phenotype, whereby the germ cells surrounding the hub formed a large, disorganized cluster with very few intercalating cyst cells and were often devoid of or had very few late-stage spermatocytes (Fig. 1C,D). By days 5 and 7, 95% and 100%, respectively, of Set1 KD testes exhibited this early-stage germ cell overpopulation phenotype (Fig. 1C,D). Here, we used a membrane marker (anti-Armadillo) to identify cyst cells that encapsulate germ cells, as shown previously (Feng et al., 2017). Any cyst with more than 16 germ cells was considered overpopulated. To test further whether these overpopulated germ cells resulted from over-proliferation, immunostaining using an antibody against a mitosis-enriched marker, H3S10P (phosphorylation at Serine 10 of H3) (Ranjan et al., 2022; Xie et al., 2015), revealed increased H3S10P-positive early-stage germ cells in the Set1 KD testes at days 3, 5 and 7, compared with earlier time points (days 1 and 3) as well as the Ctrl KD testes throughout this time course (Fig. S2).

In addition to the germline phenotypes observed in the Set1 KD testes, two other phenotypes were detected in the somatic cell lineages of the testis. First, the hub area was significantly increased compared with the Ctrl KD testes throughout this time-course experiment, with a higher degree of difference toward the later time points (Fig. 1F). Second, the number of cyst cells, positively stained with both a cyst cell marker, Traffic jam (Tj) (Li et al., 2003), and an early-stage cyst cell marker, Zinc-finger homeodomain protein 1 (Zfh1) (Eun et al., 2014; Issigonis et al., 2009; Leatherman and Dinardo, 2008), was significantly increased in the Set1 KD testes compared with the Ctrl KD testes at the later time points in the time-course experiments (Fig. 1G, Fig. S3). These changes in cyst cell number coincided with the changes in the germline phenotypes over the duration of the time course in the germline Set1 KD testes, indicating that Set1 acts in the germ cells to regulate somatic gonadal cells in a non-cell-autonomous manner.

Because the Gal4/UAS system may still be functional at 18°C, knockdown of Set1 at earlier stages of development could contribute to those phenotypes detected in adulthood (Brand and Perrimon, 1993). In Set1 KD and Ctrl KD at L3, the overall early germline morphology in Set1 KD testes resembled that of the Ctrl KD testes with no early-stage germ cell proliferation and less severe germ cell loss compared with adult Set1 KD testes (Fig. 1C,D). Phenotypes such as germline stem cell number and hub size showed much less or no significant difference between the Set1 KD and Ctrl KD testes, respectively (Fig. 1E,F), whereas the cyst cell number continued to increase in the Set1 KD testes compared with the Ctrl KD testes from L3 to day 7 (Fig. 1G). Therefore, to probe further the requirement of Set1 in adulthood, we knocked down Set1 in the early germline using the temperature-sensitive Gal80 controlled by the tubulin promoter (tub-Gal80^ts). At the permissive temperature (18°C), functional Gal80 protein inhibits Gal4 from associating with the UAS sequences, thus turning on the RNAi expression, but at the restrictive temperature (29°C) Gal80 is inactivated and Gal4 can associate with UAS and activate RNAi (McGuire et al., 2003). The tub-Gal80^ts, nos>set1 RNAi (Set1 ts-KD) and tub-Gal80^ts, nos>Control RNAi (Ctrl ts-KD) flies were grown at 18°C until eclosion, when they were shifted to 29°C for 0, 7, 14, 21 and 28 days, respectively, for another time-course experiment (Fig. 2A). At day 0, the Set1 ts-KD and Ctrl ts-KD testes had comparable germline morphology, indicating that active Gal80 at 18°C effectively prevents Set1 knockdown (Fig. 2B,C, day 0). However, 7 days after shifting to 29°C, 73% of the Set1 KD testes showed germline loss, and 27% exhibited the early germline overpopulation phenotype (Fig. 2B,C, day 7). These results indicate that the early germline phenotypes can be induced upon knocking down Set1 in adulthood. In addition to the germline phenotypes detected in the Set1 KD time course (Fig. 1C-E), in Set1 ts-KD testes at later time points, a new phenotype classified as ‘none’ was detected (no Vasa-positive germ cells could be detected at day 21 and day 28; Fig. 2B), whereby the testes are almost completely devoid of germ cells (Fig. 2B,C: 48.5% at day 14, 78% at day 21 and 65% at day 28). Throughout this time course, GSC number was consistently reduced at every time point in Set1 ts-KD testes compared with the Ctrl ts-KD testes (Fig. 2D, days 7, 14, 21 and 28). These results demonstrate that Set1 is required intrinsically for GSC maintenance in adulthood. Furthermore, the hub area was significantly increased at the later time points in the Set1 ts-KD testes (Fig. 2E, days 14, 21 and 28), whereas the cyst cell number was significantly reduced in Set1 ts-KD testes at the earlier time points (Fig. 2F, day 7 and day 14). The variation in somatic gonadal cell phenotypes between Set1 KD and Set1 ts-KD could be due to the non-cell-autonomous effects induced by the strength of germline Set1 KD and/or the different temperature shift regimes, as it has been shown that temperature contributes to germ cell differentiation in Drosophila (Gandara and Drummond-Barbosa, 2022, 2023). Together, both strategies to knock down Set1 in the germline shown in Figs 1 and 2 demonstrate that Set1 is likely required cell-autonomously for adult GSC maintenance and proper differentiation.

To determine whether the detected GSC loss phenotype (Figs 1E and 2D) is specific to Set1 knockdown in the early germline, we first used cell type-specific strategies to knock down Set1 in the CySC lineage using the tj-Gal4 driver (Li et al., 2003). In tj>set1 RNAi testes, H3K4me3 signal is reduced in the CySCs and cyst cells compared with the signals in other cell types in the adult testis (Fig. S4A). At 7 days post eclosion, no obvious germline phenotype could be detected in the tj>set1 RNAi testes compared with the tj>Control RNAi testes (Fig. S5A). The number of GSCs was actually increased in the tj>set1 RNAi testes (Fig. S5D), contrary to what was detected in the nos>set1 RNAi testes (Figs 1E and 2D). The hub area exhibited no significant difference between the tj>set1 RNAi testes and the tj>Control RNAi testes (Fig. S4A′), but the number of cyst cells was significantly increased in the tj>set1 RNAi testes (Fig. S4A″). In addition, to determine the stage specificity of the GSC loss phenotype in nos>set1 RNAi testes, we used the stage-specific bam-Gal4 (bag of marbles) driver, which turns on target gene expression specifically in 4- to 16-cell spermatogonial cysts (Chen and McKearin, 2003). In bam>set1 RNAi testes, H3K4me3 was diminished in the germline past the 4-cell spermatogonial cyst stage but still present in the GSCs and very early-stage germ cells (Fig. S4B). At 7 days post-eclosion, there were no obvious morphological changes in the germline of bam>set1 RNAi testes compared with bam>Control RNAi testes (Fig. S5B). In addition, the number of GSCs and the hub area in bam>set1 RNAi testes and bam>Control RNAi testes were not significantly different from each other (Fig. S5E), even though the bam>set1 RNAi testes had significantly more cyst cells (Fig. S4B-B″). In summary, these results indicate that knockdown of Set1 in early-stage germ cells is responsible for the detected GSC loss phenotypes shown in Figs 1 and 2.

To confirm these findings, we used nos-Gal4ΔVP16, bam-Gal80 to drive Set1 RNAi only in the GSCs, gonialblasts, and 2-cell spermatogonial cysts as Gal80 protein driven by the bam promoter prevents RNAi expression in 4- to 16-cell spermatogonial cysts (Eliason et al., 2018). In the nos-Gal4ΔVP16, bam-Gal80>set1 RNAi testes, 95% had a reduction in H3K4me3 specifically in GSCs, gonialblasts and early-stage spermatogonial cells (Fig. S4C). At 7 days post-eclosion, the hub size and number of cyst cells in the nos-Gal4ΔVP16, bam-Gal80>set1 RNAi testes were comparable to those in nos-Gal4ΔVP16, bam-Gal80>Control RNAi testes (Fig. S4C′,C″). However, the number of GSCs in the nos-Gal4ΔVP16, bam-Gal80>set1 RNAi testes was significantly reduced compared with the nos-Gal4ΔVP16, bam-Gal80>Control RNAi testes (Fig. S5F). Together, these data demonstrate that knockdown of Set1 in the early-stage germ cells can partially account for the GSC loss phenotype seen in the nos>set1 RNAi testes. The differences in other phenotypes, especially the non-cell-autonomous phenotypes, are likely due to the cell type and stage specificities, as well as the strength of the KD effects, using different drivers. We focused on using nos>set1 RNAi testes for the following experiments to understand the mechanisms underlying the Set1 KD germline phenotypes.

To confirm that the phenotypic effects seen in the nos>set1 RNAi testes (Fig. 3A, adapted from Fig. 1D) are due to the knockdown of the Set1 gene, but not off-targets effects of RNAi, a rescue experiment was performed. In nos>set1 RNAi testes, a GFP-tagged Set1 cDNA transgene was expressed using the same nos-Gal4 driver. Furthermore, because the Set1 RNAi target sequence is within the coding region of Set1, silent mutations were made within the target region to prevent KD of the transgene, namely a GFP-tagged Set1 cDNA transgene with the RNAi silent mutations (WT Rescue; Fig. 3B, Fig. S6A). As controls, three additional transgenes were generated and crossed individually to the same Set1 RNAi background to perform rescuing experiments: the WT Rescue transgene with RNAi silent mutations and an E-to-K mutation in the catalytic SET domain (Mut Rescue; Fig. 3C, Fig. S6B), a GFP-tagged Set1 cDNA transgene without the RNAi silent mutations (WT; Fig. 3D, Fig. S6C), and a GFP transgene with a nuclear localization signal (GFP-NLS; Fig. 3E, Fig. S6D). At day 1 and day 5 post eclosion, 93% and 77% of the nos>WT Rescue, set1 RNAi testes showed full rescue of the Set1 knockdown phenotypes (Fig. 3B′,B″, Fig. S6A′). By contrast, in the nos>Mut Rescue, set1 RNAi testes, only 2% showed full rescue, whereas 82% and 38% exhibited the germline loss phenotype, as well as 16% and 60% displaying the early germ cell proliferation phenotype at day 1 and day 5 post eclosion, respectively (Fig. 3C′,C″, Fig. S6B′), indicating that the catalytic SET domain is indispensable for the normal function of Set1 in the early-stage male germline. Both the nos>WT, set1 RNAi testes and nos>GFP-NLS, set1 RNAi testes show similar results compared with the nos>Mut Rescue, set1 RNAi testes (Fig. 3D′,D″,E′,E″, Fig. S6C′,D′), suggesting that adding back Set1 without changing the RNAi recognition sequences or a non-specific GFP-NLS transgene does not rescue the Set1 knockdown phenotypes in the male germline.

Furthermore, the GFP signals from both the nos>WT Rescue and the nos>Mut Rescue GFP-tagged fusion proteins are exclusively detected in the early germline nuclei, indicating that the silent mutations make these transgenes resistant to the RNAi knockdown, allowing for their proper expression and localization (Fig. 3B′,C′, Fig. S6A′,B′). In contrast, no GFP signal could be detected in the nos>WT, set1 RNAi testes without the RNAi silent mutations, indicating effective knockdown. In summary, these results support that the germline phenotypes observed in nos>set1 RNAi testes are due to the methyltransferase activity of the Set1 protein.

To understand the molecular mechanisms underlying the function of Set1 in the germline, RNA sequencing (RNA-seq) was performed with three independent biological replicates to compare the transcriptomes between nos>Control RNAi (Ctrl KD) and nos>set1 RNAi (Set1 KD) testes at 0, 1, 3 and 5 days post-eclosion, respectively (Fig. 4, Fig. S7). To obtain a global picture of the transcriptome changes between Ctrl KD and Set1 KD testes at different time points, principal component analyses were performed for all 24 samples (Fig. S7A). Among the 24 samples, the Ctrl KD samples cluster together and the Set1 KD samples cluster together (Fig. S7A). In addition, the biological replicates for each Set1 KD sample time points cluster more closely with one another than the replicates of the Ctrl KD sample time points (Fig. S7A). This can be explained as all 12 Ctrl KD samples should have similar gene expression patterns to one another given that the age difference between each consecutive time points is only 1-2 days. In contrast, the 12 Set1 KD samples do not, as visible changes could be detected at each time points in the Set1 KD testes. To analyze in more detail the transcriptome changes between the Ctrl KD and Set1 KD samples at every time point, a heatmap of all the differentially expressed genes was created (Fig. S7B). The 16,610 genes included in the heatmap, according to the Drosophila Release 6 reference genome (dos Santos et al., 2015), could be classified into four groups. Group 1 contains genes that are highly expressed at the earlier time points of the Set1 KD samples (days 0 and 1), whereas Group 2 has the genes that are highly expressed at the later time points of the Set1 KD samples (days 3 and 5). Additionally, Group 3 consists of the genes that are downregulated in the Set1 KD samples. Finally, Group 4 comprises the genes that had variable expression profiles across all the time points and thus did not fit the clear expression patterns seen for the genes in groups 1-3 (Fig. S7B).

After these initial analyses on all 24 samples, we focused on day 3 of the Set1 KD samples for two reasons. First, it is the time point when the germline mass phenotype became predominant; second, it had the most differentially expressed genes even with a stringent cutoff (≥log₂1.3=2.46-fold and P<0.05; Fig. 4A, Fig. S7C). At this time point, 1216 genes were significantly upregulated and 764 genes were significantly downregulated in the Set1 KD samples compared with the Ctrl KD samples (Fig. 4A, Fig. S7C). Gene ontology (GO) analysis of the 1216 upregulated genes showed enriched functions, with a significant category being ‘negative regulation of transcription’ (Fig. 4B).

Interestingly, components of multiple signaling pathways, such as JAK-STAT, BMP, EGF, Notch, Hedgehog, Wnt and Hippo signaling pathways, were found to be upregulated in the Set1 KD compared with the Ctrl KD samples at day 3 (Fig. S7D). For the EGF, Hedgehog and Wnt signaling pathways, multiple core components were significantly upregulated, whereas for the Notch and Hippo pathways some components were significantly upregulated and others were significantly downregulated. Among all these, only the JAK-STAT, BMP and EGF signaling pathways had at least two core components that made the significant cut-offs (Fig. 4C, Fig. S7D). For example, the JAK-STAT pathway genes Stat92E and upd3, and the BMP pathway genes dpp and wit, were found to be significantly upregulated (Fig. 4C). Given the importance of both signaling pathways in the testis (Inaba et al., 2015; Kawase et al., 2004; Kiger et al., 2001; Leatherman and Dinardo, 2008, 2010; Schulz et al., 2004; Singh et al., 2010; Tulina and Matunis, 2001), these results indicate that the early-stage germ cell overpopulation phenotype could be due to ectopic expression of both JAK-STAT and BMP signaling pathway genes.

To investigate further the functional relevance of both JAK-STAT and BMP signaling pathways with the Set1 KD phenotypes, we examined two key downstream genes of JAK-STAT and BMP pathways, Stat92E and Mad, respectively. Both genes were identified to be upregulated in the Set1 KD testes compared with the Ctrl KD testes based on the RNA-seq results (Fig. 4D). However, given that Set1 generates H3K4me3 to activate transcription, the direct target genes of Set1 are likely downregulated in Set1 KD testes. Nevertheless, we reason that the ectopic expression of Stat92E or Mad could contribute to the early-germline overpopulation phenotypes in the Set1 KD testes. Next, we tested potential genetic interactions of them with Set1.

In heterozygous loss-of-function mad¹² testes (Raftery et al., 1995) or Ctrl KD testes, pMad was detectable in the GSCs (Fig. 5A,D, mad¹²/+; Fig. S8A) and 100% showed normal morphology (Fig. 5E). In contrast, in the Set1 KD testes, pMad was detected in GSCs and ectopically in the spermatogonial cells (Fig. 5A,D, nos>set1 RNAi). Overall, 85% of the Set1 KD testes exhibited the early germ cell overpopulation phenotype (Fig. 5E). Even though both pMad patterns were detectable in Set1 KD testes with one copy of the mad¹² allele (Fig. 5A,D, mad¹²/+; nos>set1 RNAi), the addition of one mad¹² allele to the Set1 KD background reduced the occurrence of the early germ cell overpopulation phenotype from 85% to 65% (P<0.05; Fig. 5E). Furthermore, we tested thickveins (tkv), which encodes an upstream receptor of the BMP pathway and displays increased expression in the Set1 KD testes compared with the Ctrl KD testes (Fig. S7D). The loss-of-function tkv⁴ allele (Lecuit and Cohen, 1997; Terracol and Lengyel, 1994) also suppressed the early germ cell overpopulation phenotype from 85% to 64% (P<0.05; Fig. 5E), as well as the GSC loss phenotype (P<0.05; Fig. 5F). Together, these results suggest that Set1 normally represses the BMP signaling pathway. Without this repression, ectopic activity can occur, leading to abnormal germline morphology and composition.

Additionally, in heterozygous testes with the loss-of-function stat92E⁰⁶³⁴⁶ allele (Hou et al., 1996) or Ctrl KD testes, Stat92E expression was detected in the GSCs (white outline in Fig. 5B,C, stat92E⁰⁶³⁴⁶/+; Fig. S8B), but in Set1 KD testes Stat92E was not only detectable in GSCs but also in hub cells (yellow outline in Fig. 5B, nos>set1 RNAi) and in spermatogonial cells (cyan outline in Fig. 5B, nos>set1 RNAi). Interestingly, upon compromising Stat92E with stat92E⁰⁶³⁴⁶ in the Set1 KD background, the expression of Stat92E was no longer found in hub cells (yellow outline in Fig. 5B, nos>set1 RNAi; stat92E⁰⁶³⁴⁶/+) or in the spermatogonial cells. Furthermore, when the level of Stat was reduced in nos>set1 RNAi; stat92E⁰⁶³⁴⁶/+ testes, the incidence of the early-stage germ cell overpopulation phenotype decreased from 85% to 11% (P<10⁻⁴; Fig. 5E). These results suggest that enhanced Stat92E expression upon Set1 inactivation significantly contributes to the early germ cell overpopulation phenotype in Set1 KD testes. Moreover, compromising the BMP pathway by mad¹²/+ fully suppresses the GSC loss phenotypes in Set1 KD testes, whereas compromising the BMP pathway by tkv⁴ can also partially suppress the GSC loss phenotype in Set1 KD testes (P<0.05; Fig. 5F). Halving the level of Stat92E could partially suppress the GSC loss phenotype in Set1 KD testes (P<0.05; Fig. 5F). However, none of the conditions could fully restore the loss of some spermatogonial cells in the Set1 KD testes (Fig. 5E), suggesting that other downstream factors of these two signaling pathways or components from other potentially involved signaling pathways (Fig. S7D) could contribute to the loss of early-stage germ cells at the Set1 KD background. Together, these genetic interaction results support the functional relationships between Set1 and BMP as well as JAK-STAT pathways, with the primary role of BMP signaling relating to the GSC loss phenotype and a predominant contribution of the JAK-STAT pathway to the early-stage germline overpopulation phenotype, consistent with the transcriptome results shown in Fig. 4.

This study explores the in vivo roles of Drosophila Set1, a histone methyltransferase responsible for ‘writing’ the H3K4me3 histone modification, in the Drosophila testis. Through a series of cell type- and stage-specific RNAi knockdown via a time course regime, our findings unveil a fascinating progression of germline defects (Fig. 6A-D). These Set1 loss-of-function phenotypes have both cellular and molecular specificities. First, the phenotypes are predominantly linked to the loss of Set1 function in early-stage germ cells, as compromising Set1 in late-stage germ cells or somatic gonadal cells fails to replicate these defects. Moreover, we conducted rescuing experiments utilizing transgenes encoding both the wild-type and catalytic inactive forms of Set1. Only the wild-type Set1 effectively rescues the knockdown phenotypes, underscoring the specificity of these effects to the methyltransferase activity of Set1. To further understand the underlying molecular mechanisms, we performed RNA-seq experiments to identify genes with changed expression upon knocking down Set1 gene. Through this assay, we identified several crucial signaling pathway genes, such as Stat92E and Mad, the downstream components of the JAK-STAT and BMP signaling pathways, respectively, that have increased expression when Set1 is knocked down. Genetic interaction analyses further suggest a functional relationship between Set1 and these signaling pathways, as mutations of the Stat92E and Mad genes suppress the Set1 knockdown phenotypes. Collectively, our investigations shed light on the molecular mechanisms at play, enhancing our comprehension of the pivotal decision-making process governing proliferation versus differentiation within adult stem cell lineages. This insight bears significant implications for both stem cell biology and the broader field of cancer biology.

Here, our RNA-seq results reveal that upon knockdown of Set1 in the Drosophila male germline, a key downstream factor of the JAK-STAT signaling pathway, Stat92E, is upregulated (Fig. 4D). This is unexpected for the direct targets of Set1, given that Set1 is the methyltransferase for the transcriptional activating mark H3K4me3. In addition, previous ChIP-seq results using progenitor germ cell-enriched bam testes (McKearin and Spradling, 1990) indicate that H3K4me3 is not enriched at the promoter region of the Stat92E gene (Fig. S9A) (Gan et al., 2010). These results suggest that Stat92E is unlikely to be a direct target of Set1. However, given the suppression of the early germ cell overpopulation phenotype when adding a single copy of a Stat92E mutation in the Set1 KD background, they likely interact in vivo. We hypothesize that Set1 acts in the Drosophila male germline by regulating the transcriptional activity of a JAK-STAT inhibitor. When Set1 is compromised, the expression of this inhibitor is possibly downregulated, allowing for the ectopic expression of Stat92E (Fig. 6E). One candidate JAK-STAT inhibitor is protein tyrosine phosphatase 61F (Ptp61F). Ptp61F is a known inhibitor of the JAK-STAT pathway that has been previously identified as a target of Lid, the H3K4 demethylase, in the germline (Baeg et al., 2005; Tarayrah et al., 2015). Additionally, Ptp61F is slightly downregulated in the RNA-seq data (Fig. S9G) and is enriched with the H3K4me3 mark at its promoter regions (Fig. S9C).

In addition to Stat92E, the downstream factor of the BMP signaling pathway Mad is also upregulated in the Set1 KD background (Fig. 4D). Unlike Stat92E, H3K4me3 is enriched at the promoter region of the Mad gene in progenitor germ cells (Fig. S9B) (Gan et al., 2010). However, given that Mad is slightly upregulated (Fig. 4D) and suppression of the early germ cell overpopulation phenotype was detected when a single copy of Mad mutation is combined with the Set1 KD condition (Fig. 5A,E), we hypothesize that BMP inhibitor-encoding genes could be direct targets of Set1, downregulation of which upon inactivation of Set1 could lead to the ultimate upregulation of Mad. When examining the candidate BMP inhibitors Cul-2, MAN1 and Ube3a (Cul-2 is a BMP inhibitor in the Drosophila female germline, whereas both MAN1 and Ub3a are BMP inhibitors in other Drosophila tissues, as shown previously; Ayyub et al., 2015; Duan et al., 2020; Laugks et al., 2017; Li et al., 2016; Wagner et al., 2010), all three BMP inhibitors have downregulated expression to varying degrees in the Set1 KD testes (Fig. S9G) and H3K4me3 enrichment can be detected at their corresponding promoter regions (Fig. S9D-F) (Gan et al., 2010). Additional experiments are needed to determine whether these inhibitors of JAK-STAT and BMP signaling are bona fide direct target genes and have a functional relationship with Set1 in the Drosophila male germline.

Notch signaling has been shown to regulate the transit-amplification stage (Ng et al., 2019). Removal of Notch in the cyst cells or Delta in the germline result in spermatogonial cell death. Although Delta is not differentially expressed, Notch is upregulated whereas two other Notch pathway components, Suppressor of Hairless [Su(H)] and Serrate (Ser), have decreased expression in Set1 KD testes (Fig. S7D). Increased expression of Wnt signaling components, such as arm, is detected using RNA-seq (Fig. S7D), which could result in abnormal clusters enriched with undifferentiated germ cell as reported (Pancratov et al., 2013). Finally, the EGF signaling is important for cyst cell differentiation and germline maintenance and differentiation (Feng et al., 2017), which regulates encapsulation of germ cells by the cyst cells (Amoyel et al., 2016; Schulz et al., 2004; Tran et al., 2000). The increased expression of multiple EGF signaling components in Set1 KD testes (Fig. S7D) could act in conjunction with JAK-STAT and BMP signaling pathways. Because our RNA-seq analyses used the entire tissue, further studies are needed to address all these possibilities.

In Drosophila, the lid gene encodes the histone demethylase responsible for ‘erasing’ the H3K4 methylation (Eissenberg et al., 2007; Lee et al., 2007b), enzymatic activity of which directly antagonizes the function of Set1. It has been shown that in the male germline Lid is required for the proper levels of Stat92E. In the lid knockdown or mutant cells, Stat92E expression is reduced (Tarayrah et al., 2015). Therefore, the two histone-modifying enzymes with opposite activities, Set1 and Lid, display antagonistic regulation of Stat92E expression. Therefore, this work introduces a new epigenetic regulator within the complex framework of the JAK-STAT signaling pathway in the Drosophila testis niche. Together, these studies highlight the delicate balance maintained by a histone methyltransferase and a histone demethylase in orchestrating a major signaling pathway.

As the primary H3K4 methyltransferase in Drosophila, Set1 has also been demonstrated to play crucial roles in the female germline. For example, it has been shown that Set1 acts with the E3 ubiquitin ligase Bre1 to produce the majority of H3K4me3 modifications in regulating the maintenance and differentiation of female GSCs. Knockdown of Set1 in the early-stage germline results in a significant loss of GSCs in the germaria, due to decreased expression of BMP signaling components, such as Mad or Dad (Bray et al., 2005; Xuan et al., 2013). This phenotype resembles the male GSC loss phenotype observed in this study. However, in the male germline, inactivation of Set1 predominantly affects the JAK-STAT pathway over the BMP pathway in our study. This effect is observed to progress over time, resulting in upregulated Stat92E levels. Consequently, we observe an overpopulation of early-stage germ cells during the late phase of our time-course experiments. Based on an extensive RNAi screen that identified hundreds of candidate genes with potential roles in the female germline, Set1 stands out as a crucial factor required for female GSC maintenance and proper differentiation. When Set1 is knocked down in the female germline, ovaries exhibit a spectrum of phenotypes, including empty ovarioles likely due to female GSC loss as well as so-called ‘pseudoegg’ chambers filled with undifferentiated cells (Yan et al., 2014). The latter phenotype resembles the early-germline overpopulation exhibited in the Set1 male germline knockdown effects shown in this work (Fig. 1C,D). Moreover, for the non-cell-autonomous roles of Set1 in the Drosophila ovary, Set1 acts in somatic gonadal cells to impact the female germline. For example, compromising Set1 in different somatic gonadal cells in the ovary results in either GSC loss or overpopulated GSC-like cells in the germaria (Xuan et al., 2013). By contrast, compromising Set1 in the somatic gonadal cells in the testis does not seem to cause any obvious phenotypes (Fig. S4A). However, in the Drosophila testis Set1 acts in the germline non-cell autonomously to affect niche architecture and cyst cell number (Fig. 1F,G).

In the somatic stem cell lineages, two types of neuroblast (NBs), type I NBs (NB-I) and type II NBs (NB-II), differ in number (e.g. 184 NB-Is and 16 NB-IIs) and lineage progression to ultimately produce the diverse range of neurons and glia in the brain (Bello et al., 2008; Boone and Doe, 2008). Interestingly, Set1 seems to be dispensable for the NB-II differentiation, which is distinct from the profound functions of Set1 in both ovary and testis for proper GSC differentiation (Yan et al., 2014), suggesting different requirements of histone modifications and their modifying enzymes in distinct stem cell systems.

Many short-lived cell types in the body are produced from adult stem cells, either continuously or in response to physiological signals or trauma. During these processes, genetic lesions or epigenetic misregulations can drive cells to divide continuously instead of undergoing proper differentiation. This shift in cellular properties can initiate diseases such as cancer.

The COMPASS complex has been shown to be involved in pathogenesis such as cancer (Shilatifard, 2012). Here, our results demonstrate significant phenotypic changes resulting from germline knockdown of Set1 in fly testes, beginning with germ cell loss and progressing to the development of germline tumors within a relatively short timeframe (Fig. 1B-D). This delayed appearance raises an intriguing possibility: a resilient subset of ‘survivor cells’, which evade early cell death, initiates a phase of uncontrolled and rapid over-proliferation of undifferentiated cells, reminiscent of the phenomenon observed for ‘cancer stem cells’ (Beck and Blanpain, 2013). Given that various histone methyltransferases are targeted for therapeutic treatment in cancers, our discovery of initial cell loss followed by over-proliferation in response to knockdown of a key histone methyltransferase raises concerns about the potential use of these enzyme inhibitors in cancer therapy (Morera et al., 2016).

Flies were raised under standard yeast/molasses medium at 25°C unless stated otherwise. The following flies were used: nos-Gal4 (with VP16)/Cyo (Van Doren et al., 1998), nos-Gal4 (without VP16 or ΔVP16) on the second chromosome (from Yukiko Yamashita, University of Michigan, USA), UAS-set1 RNAi (Bloomington Drosophila Stock Center, BL33704), UAS-mCherry RNAi (Bloomington Drosophila Stock Center, BL35785), tj-Gal4/Cyo (Tanentzapf et al., 2007), bam-Gal4/TM6B (Chen and McKearin, 2003), bam-Gal80 on the third chromosome (from Juliette Mathieu and Jean-René Huynh, Collège de France, France), UAS-GFP.nls (Bloomington Drosophila Stock Center, BL4775), UASp-FRT-EGFP-set1 (WT)-PolyA-FRT (this study), UASp-FRT-EGFP-set1 (WT rescue)-PolyA-FRT (this study), UASp-FRT-EGFP-set1 (Mutant rescue)-PolyA-FRT (this study), mad¹², FRT40A/Cyo (Bloomington Drosophila Stock Center, BL58785), stat92E⁰⁶³⁴⁶/TM3, Sb (Bloomington Drosophila Stock Center, BL11681), tkv⁴ FRT40A/Cyo (Bloomington Drosophila Stock Center, BL58786), vasa-GFP knock-in (from Dr Tatjana Trcek, Johns Hopkins University, Baltimore, MD, USA).

To study the function of Set1 in the Drosophila testis, two RNAi lines, UAS-set1 RNAi and UAS-mCherry RNAi, were crossed with different drivers, nos-Gal4/Cyo, bam-Gal4/TM6B and tj-Gal4/Cyo, at 18°C. Flies at the larval stage L3 were collected and newly eclosed progenies were transferred to new vials at 25°C and aged for 0, 1, 3, 5 or 7 days before dissection.

For Set1 function study in the germline stem cells and gonialblast cells, the two RNAi lines UAS-set1 RNAi and UAS-mCherry RNAi were crossed with nos-Gal4ΔVP16/Cyo; bam-Gal80/MKRS at 25°C. Newly eclosed progeny were transferred to new vials and maintained at 25°C for 7 days before dissection.

To determine the role of Set1 in the Drosophila early germline at adulthood, two RNAi lines, vasa-GFP knock-in/CyO; UAS-set1 RNAi and vasa-GFP knock-in/CyO; UAS-mCherry RNAi, were crossed with nos-Gal4/Cyo; tub-Gal80^ts at 18°C. Newly eclosed progenies were transferred to new vials at 29°C and aged for 0, 7, 14, 21 or 28 days before dissection.

To determine whether expression of EGFP-set1 cDNA in the early germline is sufficient to rescue nos>set1 RNAi germ cell phenotypes, flies with the following genotypes were grown at 25°C and dissected at 1 and 5 days post eclosion: nos-Gal4/UASp-FRT-EGFP-set1 (WT rescue)-PolyA-FRT; UAS-set1 RNAi/+, nos-Gal4/UASp-FRT-EGFP-set1 (Mutant rescue)-PolyA-FRT; UAS-set1 RNAi/+, nos-Gal4/UASp-FRT-EGFP-set1 (WT)-PolyA-FRT; UAS-set1 RNAi/+, nos-Gal4/UAS- GFP.nls; UAS-set1 RNAi/+.

To identify whether Set1 genetically interacts with Mad, tkv and Stat92E, the alleles mad¹², tkv⁴ and stat92E⁰⁶³⁴⁶ were used. Flies with the following genotypes were grown at 25°C and dissected at 5 days post eclosion: nos-Gal4/mad¹²; UAS-set1 RNAi/+, nos-Gal4/mad¹², nos-Gal4/tkv⁴; UAS-set1 RNAi/+, nos-Gal4/tkv⁴, nos-Gal4/+; UAS set1 RNAi/stat92E⁰⁶³⁴⁶, UAS set1 RNAi/stat92E⁰⁶³⁴⁶; nos-Gal4/+; UAS set1 RNAi/+.

For the lines UASp-FRT-EGFP-set1 (WT rescue)-PolyA-FRT and UASp-FRT-EGFP-set1 (Mutant rescue)-PolyA-FRT, the RNAi recognition sequence was altered. Because the RNAi recognition sequence is in the coding region of the Set1 gene (AAGGTGCAGAGTATAAGAGTA), the third base for each codon of this sequence was changed to make a silent mutation that would allow for the protein to be translated properly but for the RNAi to not recognize the mRNA. For the line UASp-FRT-EGFP-set1 (Mutant rescue)-PolyA-FRT, a mutation in the SET domain at G4713A to make the E1613K amino acid replacement was made.

Testes were dissected in Schneider's insect media and then fixed in 4% formaldehyde in 1× PBST (1× PBS with 0.1% Triton X-100) for 8 min at room temperature (RT). The testes were rinsed three times in 1×PBST and washed three times for 5 min each time using 1× PBST at RT. Testes were incubated with primary antibodies in 1× PBST+3% bovine serum albumin at 4°C for at least one night. Samples were then rinsed three times in 1× PBST and washed three times for 5 min each time in 1× PBST and then incubated in a 1:1000 dilution of Alexa Fluor-conjugated secondary antibody in PBST+5% normal goat serum for 2 h at RT or rotating for a minimum of 24 h at 4°C. Samples were rinsed three times in 1× PBST and washed three times, 5 min each in 1× PBST and then mounted for microscopy in Vectashield antifade mounting medium (H-1400, Vector Laboratories) with or without DAPI. Samples were imaged using a Leica SP8 or Stellaris 5 confocal microscope with a 63× oil immersion objective. Images were analyzed using ImageJ software. Primary antibodies used were: Vasa (rabbit, 1:5000; from Ruth Lehmann, Skirball Institute of Biomolecular Medicine, NY, USA), Fas III (mouse, 1:50; DSHB, 7G10), Armadillo (mouse, 1:50; DSHB, N2 7A1), α-Spec (mouse, 1:50; DHSB, 3A9), 1B1 (mouse, 1:50; DSHB, 1B1-s), H3S10P (mouse, 1:2000; Abcam, ab14955), Zfh1 (rabbit, 1:5000; from Ruth Lehmann), TJ (guinea pig, 1:1000; from M. Van Doren, Johns Hopkins University, USA), H3K4me3 (rabbit, 1:400, Cell Signaling Technology, 9751S), GFP (chicken, 1:1000; Abcam, ab13970), Stat92E (rabbit, 1:200; gift from Denise Montell, University of Santa Barbara, CA, USA) and pMad (rabbit, 1:100; Abcam, ab52903). All secondary antibodies were Alexa Fluor Conjugated secondaries from Thermo Fisher Scientific used at 1:1000 dilution: goat anti mouse 405 A31553, 488 A23723, 568 A11004, 633 A21050, 647 A21235; goat anti rabbit 488 ab150077, 568 A11011, 633 A21202; goat anti guinea pig 488 A11073, 568 SAB4600080, 633 A21105; goat anti chicken 488 ab150169.

nos-Gal4; UAS-mCherry RNAi and nos-Gal4; UAS-set1 RNAi flies were collected as newly eclosed males and aged for 0, 1, 3 and 5 days at 25°C after shift from 18°C. Approximately 15 pairs of testes for each genotype were dissected in Schneider's media+10% fetal bovine serum as one replicate. Three replicates were generated for each time point and genotype. The testes were then desheathed in 500 μl of lysis buffer (trypsin LE+2 mg/ml collagenase). The samples were incubated in lysis buffer for 10 min in a 37°C water bath with gentle vortex mixing every 2 min. The samples were then filtered through a 40 μm tissue culture filter followed by a 10 min centrifugation at 1200 rpm (106 g). The cells were washed with 200 μl of PBS and pelleted again for 5 min at 1200 rpm (106 g). Total RNA was purified using the Quick-RNA Microprep Kit (R1050, Zymo Research Corporation) following the manufacturer's instructions. The libraries were generated using the reagents provided in NEBNext Ultra II Directional RNA library Prep Kit for Illumina (E7760S, New England Biolabs Inc.) and NEB Next Poly(A) mRNA Magnetic Isolation Module (E7490, New England Biolabs Inc.). The Illumina compatible libraries were sequenced with Illumina Novaseq6000 sequencer at the National Institutes of Health sequencing facility.

The sequencing reads were examined using FastQC quality software (Galaxy Version 0.73+galaxy0) (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads that passed the quality filter were then mapped to the Drosophila genome (D. melanogaster August 2014; BDGP Release 6+ISO11 MT/dm6) using Bowtie 2 version 2.5.1 (Langmead and Salzberg, 2012). For gene mapping, the gene model (Drosophila_melanogaster.BDGP6.87.gtf) was utilized, and we followed an RNA-seq data analysis tutorial using Galaxy (https://training.galaxyproject.org/training-material/topics/transcriptomics/tutorials/ref-based/tutorial.html), as detailed in the work (Batut et al., 2018; Hiltemann et al., 2023). Aligned reads were summarized to genes using featureCounts (Liao et al., 2014) with gene features obtained from NCBI RefSeq dm6 assembly. Raw counts were normalized to transcripts per million reads (TPM) counts. Principal component analysis was performed on 16,610 genes with detectable expression in at least one sample, based on which sample distance matrix was computed. To assess differential gene expression between Set1 knockdown and controls, DESeq2 (Love et al., 2014) was applied on samples from each day, using the Benjamini and Hochberg method for multiple testing correction with default DESeq2 parameters. A threshold of log₂ fold change of 1.3 and adjusted P<0.05 was applied to ascertain differentially expressed genes, for which EnrichR (Chen et al., 2013) was applied to assess enrichment of molecular pathways in each set. All statistical analysis was conducted using R software version 4.2.0.

All phenotypic quantification was carried out in Fiji (ImageJ). GSC number was determined by counting every cell that was either Vasa positive or Tj negative and directly next to the hub. Early cyst cell number was quantified by counting every Tj-positive cell in a whole testis z-stack. Hub area quantification was determined by drawing a line around the z-slice with the largest hub size based on Armadillo or Fasciclin III immunostaining and using the area measurement in Fiji. For the early germ cell proliferation phenotype, Armadillo was used as the marker to delineate the germline cysts (Feng et al., 2017). Any cyst with more than 16 cells was considered overpopulated.

Data were subjected to the Shapiro–Wilk test to determine whether the data were normally distributed or skewed. For normally distributed data, an unpaired, two-sample t-test was used to compare two individual datasets with each other. For skewed data, the Wilcoxon signed rank test was used to compare two individual datasets with each other. A χ² test was used to compare distributions of datasets containing data from different categories. Two-way ANOVA with interaction and Šidák multiple comparison test were used to compare two individual datasets with each other that had two independent variables. One-way ANOVA and Dunnett's T3 multiple comparison test were used when more than two individual datasets were compared with each other that had one independent variable. Data are presented with error bars representing the s.e.m. Significant differences based on these statistical analyses are noted by asterisks (*P<0.05, **P<10⁻², ***P<10⁻³, ****P<10⁻⁴).

We thank Drs S. Myong, D. Drummond-Barbosa, C. Wu and Y. Kim for critical comments. We thank Chen lab members for insightful suggestions. We thank Yingying Li for generating the transgenes. We thank Johns Hopkins Integrated Imaging Center for confocal imaging.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202729.reviewer-comments.pdf