Sequential changes at differentiation gene promoters as they become active in a stem cell lineage

In stem cell lineages, proliferating precursors give rise to specific differentiating cell types. A major question is how the correct differentiation genes, kept silent in precursor cells, are selectively activated upon the switch to differentiation. Reversal of repression by the Polycomb group (PcG) epigenetic transcriptional silencing machinery has been implicated in the switch from proliferating precursor cells to differentiation in stem cell lineages. In embryonic stem cells (ESCs), PcG components are often associated with transcriptionally silent genes that will be turned on later during lineage specification (Boyer et al., 2006; Chamberlain et al., 2008; Lee et al., 2006). However, PcG action is unlikely to be the only mechanism that keeps the differentiation genes silent, as PcG proteins are not necessary for ESC self-renewal. Instead, PcG action appears to be required for the faithful execution of lineage-appropriate gene expression during ESC differentiation (Surface et al., 2010). In adult stem cell lineages, lack of the mammalian PcG protein Bmi1 results in loss of long-term self-renewing capacity in hematopoietic and neural stem cells, whereas overexpression of Bmi1 is associated with leukemia (Lessard and Sauvageau, 2003; Molofsky et al., 2003; Park et al., 2003). Consistent with the oncogenic features of Bmi1, several genes repressed by Bmi1 act to restrict the multipotency of progenitors in the hematopoietic lineage (Akala et al., 2008). Likewise, reversal of PcG-mediated epigenetic silencing by specific histone demethylases is required for keratinocyte differentiation in epidermis (Sen et al., 2008).

The PcG proteins act in at least two distinct but cooperating protein complexes: Polycomb repressive complex 1 (PRC1) and PRC2 (for a review, see Schwartz and Pirrotta, 2007). PRC2 initiates repression by methylating histone H3 at Lys27 (Cao et al., 2002; Czermin et al., 2002; Kuzmichev et al., 2002; Muller et al., 2002). Recruitment of PRC1 may be initiated by the binding of Polycomb via its chromodomain to the trimethylated histone H3 Lys27 (H3K27me3) mark (Fischle et al., 2003; Min et al., 2003). PRC1 is found at silenced gene promoters even in the presence of the RNA polymerase II (Pol II) transcription machinery, suggesting that Polycomb might not block the formation of the preinitiation complex (Dellino et al., 2004). In the preinitiation complex, the C-terminal domain (CTD) of Pol II is in an unphosphorylated state. The CTD becomes phosphorylated at Ser5 as the polymerase initiates transcription of the first few nucleotides, and phosphorylation on Ser2 of the CTD is associated with Pol II fully engaged in transcript elongation (reviewed by Phatnani and Greenleaf, 2006). Thus, antibodies specific for the hypophosphorylated CTD provide a way to detect Pol II that is still at the promoter. In addition to the H3K27me3 mark placed by PRC2 activity, the PRC1 component dRing (Sce – FlyBase) acts as an E3 ubiquitin ligase to mono-ubiquitylate histone H2A at Lys119 (H2AK119ub) (Wang et al., 2004a), which blocks efficient transcriptional elongation (Stock et al., 2007). In addition to silencing gene expression by modifying specific histones, PcG complexes may repress transcription by compacting the chromatin and excluding transcriptional activators (Francis et al., 2004).

In stem cell lineages, if silencing by PcG action blocks the expression of differentiation genes, then the normal developmental program must reverse this epigenetic silencing in a gene-specific and lineage-appropriate fashion to allow differentiation of the proper cell type(s). Failure to do so might lead to abnormalities in development and underlie early pathological progression toward cancer.

To understand the mechanisms that reverse silencing by PcG during differentiation in stem cell lineages, we investigated the sequence of events at the promoters of three terminal differentiation genes as cells progress from proliferating precursors to active differentiation in the Drosophila male germline. Male germ cells differentiate from adult stem cell precursors, first undergoing transit-amplifying mitotic divisions as spermatogonia, followed by switching to the spermatocyte program of cell growth and meiosis. In spermatocytes, a dramatic, cell type-specific transcription program is initiated in preparation for subsequent spermatid differentiation. This transcription program requires the cooperative activity of spermatocyte-specific homologs of components of the core transcription factor TF_IID complex (the tTAFs) (Hiller et al., 2004; Hiller et al., 2001; Lin et al., 1996), as well as the testis-specific meiotic arrest (tMAC) complex, which resembles the mammalian MIP/dREAM (Drosophila RB, E2F and Myb) and the C. elegans SynMuv complexes (Ayyar et al., 2003; Beall et al., 2007; Jiang et al., 2007; Jiang and White-Cooper, 2003; Perezgasga et al., 2004; White-Cooper et al., 2000; White-Cooper et al., 1998). Expression of the tTAFs and testis-specific components of the tMAC complex, such as always early (aly), is turned on for the first time in spermatocytes, early in the G2 phase of meiosis I (White-Cooper et al., 2000; Hiller et al., 2001; Hiller et al., 2004; Chen et al., 2005). The tTAFs appear to activate the expression of terminal differentiation genes in part by counteracting repression by the PcG transcriptional silencing machinery. As a result of tTAF action, Polycomb is removed from differentiation gene promoters and sequestered to spermatocyte nucleoli, along with several other PRC1 components (Chen et al., 2005).

Here we show that a choreographed series of events under the control of a cell type- and stage-specific developmental program converts terminal differentiation genes from a transcriptionally silent state in precursor cells to a fully active state in spermatocytes. Our results in the germline stem cell (GSC) lineage provide a paradigm for how epigenetic silencing can be reversed in a gene-selective and stage-specific manner to allow expression of terminal differentiation genes and highlight the role of cell type-specific transcription machinery in this process.

Flies were raised on standard cornmeal molasses agar medium at 25°C unless stated otherwise. Drosophila strains are described previously (Ayyar et al., 2003; Gonczy et al., 1997; Hiller et al., 2004; Hiller et al., 2001; Jiang et al., 2007; Jiang and White-Cooper, 2003; Lin et al., 1996; McKearin and Ohlstein, 1995; McKearin and Spradling, 1990; Perezgasga et al., 2004; White-Cooper et al., 2000; White-Cooper et al., 1998). The allelic combinations for mutants were: bam¹/bam^Δ86, aly²/aly^5P, can¹², y,w, comm^Z1340, tomb^GS12862, topi^{143-68/157-89}, achi/vis^Z3922, mip40^1a/4a. The E(z)^−/− clones were induced by FLP recombinase that was exclusively expressed in germ cells using the germline-specific nos-GAL4 driver with a UAS-FLP transgene in males of genotype nos-GAL4/Y; UAS-FLP; E(z)⁷³¹ FRT2A/GFP3-13-7 FRT2A.

Immunofluorescence staining using anti-Sa, anti-Can or anti-Fib was performed using the methanol fixation procedure as described (Chen et al., 2005). Immunofluorescence staining using anti-E(z), anti-Su(z)12 or anti-H3K27me3 was performed using the formaldehyde fixation procedure as described (Hime et al., 1996).

Primary antibodies were anti-Can (1:1000), anti-Sa (1:100), anti-Fib (undiluted), anti-E(z) (1:100), anti-Su(z)12 (1:100) and anti-H3K27me3 (1:200; Millipore 07-449). Secondary antibodies were from the Alexa Fluor-conjugated series (1: 200; Molecular Probes).

Chromatin was visualized by DAPI (4′,6-diamidino-2-phenylindole) staining in VECTASHIELD medium (Vector Labs H-1200). Images were taken using the Leica TCS SP2 AOBS confocal system and processed using Adobe Photoshop.

The nht¹/nht² and nht¹/nht²; E(z)⁶¹/E(z)⁷³¹ flies were isolated at the day of eclosion at 25°C (permissive temperature) and shifted to 29°C (restrictive temperature) for 7 days. Twenty pairs of testes were dissected from each genotype and total protein was extracted, separated by 4-20% PAGE (Invitrogen, EC6025BOX) and transferred to a membrane (GE Healthcare RPM303F). The blot was probed with rabbit anti-H3K27me3 (1:400; Millipore 07-449) followed by HRP-conjugated anti-rabbit (1:5000, GE Healthcare NIF824). ECL reagents (GE Healthcare RPN2108) were used for detection. The blot was stripped for 30 minutes at 50°C with stripping buffer (100 mM 2-mercaptoethanol, 2% SDS, 62.5 mM Tris-HCl pH 6.7) and then re-probed with anti-H3 (1:5000; Abcam ab1791), followed by the same anti-rabbit secondary antibody (1:5000) and ECL detection.

Twenty pairs of testes from yw (wt), nht¹/nht² (nht) and nht¹/nht²; E(z)⁶¹/E(z)⁷³¹ [nht; E(z)] males (grown under the same conditions as described above) were dissected and used for total RNA extraction with TRIzol (Invitrogen 15596-018). cDNAs were synthesized with M-MLB reverse transcriptase (Promega M1701). Real-time PCR analyses were performed using TaqMan probes with Universal PCR Master Mix (Applied Biosystems 4304437) in an ABI 7300 machine. The TaqMan probes for dj, fzo, Mst87F and Rpl32 are Dm02361628_s1, Dm02150476_s1, Dm02362808_s1 and Dm02151827_g1, respectively (Applied Biosystems). Each PCR reaction was performed in duplicate and the Ct numbers for each reaction were collected and averaged. The relative expression level of each target gene was calculated from 2^{Ct(Rpl32)–Ct(target gene)}, using Rpl32 as an internal control.

ChIP experiments using dissected testes were performed as described (Cao et al., 2002; Chen et al., 2005), except that Protein A Dynabeads (Invitrogen 100.01D) were used instead of Protein A Agarose beads. Each ChIP experiment used: 5 μl anti-Sa, 5 μl anti-Polycomb [from R. Kingston and R. Jones (Chen et al., 2005)], 5 μl anti-Pol II (ascites, 8WG16, Covance MMS-126R), 4 μl anti-Pol II (1 μg/μl, 4H8, Abcam ab5408), 5 μl anti-H3K4me3 (0.4 μg/μl, Abcam ab8580), 2 μl anti-H3K27me3 (1 μg/μl, Millipore 07-449), 1 μg anti-H3 (Abcam ab1791). All ChIP experiments were repeated at least three times in independent biological replicates.

Input DNA, mock precipitated DNA and ChIP DNA were analyzed using gene-specific primers (see below) with the Universal PCR Master Mix (Applied Biosystems 4304437) in an ABI 7300 machine. Quantitative PCR analyses of the ChIP experiments were performed as described (Chen et al., 2005). The ChIP and mock DNA were normalized to the input DNA and are plotted as a percentage of the input DNA (Input %) in Fig. S3 in the supplementary material. Further normalization was performed to convert the raw percentage input data to fold change relative to the constitutively expressed CycA gene from the same sample. For a given antibody, relative enrichment compared with CycA in the same sample is plotted in Fig. 3, with CycA set to 1 in order to compare across the different genotypes. Primers used were (5′ to 3′):

Total RNA from ~200 pairs of fly testes (bam¹/bam^Δ86, aly²/aly^5P, sa¹/sa², y,w) was extracted using TRIzol. The genomic DNA was degraded using 2 units DNase I (Fermentas EN0521, Glen Burnie, MD, USA) at 37°C for 20 minutes. RNA integrity was checked by gel electrophoresis (1% agarose). Approximately 4 μg total RNA from each biological replicate was used to generate labeling probes to hybridize with the Affymetrix GeneChip Drosophila Genome 2.0 Array according to the Affymetrix protocol. Three biological replicates were performed for each genotype. Microarray hybridization was processed at the Core Facility at the Stanford University School of Medicine and the raw data were exported from the Affymetrix Microarray Suite (MAS). The CEL files were used for signal normalization with RMA as part of the limma package (Bioconductor). The accession number for microarray data is GSE 28728 (Gene Expression Omnibus).

Consistent with the model that terminal differentiation genes may be acted upon by the PcG transcriptional silencing machinery in precursor cells, immunofluorescence staining revealed abundant PRC2 proteins E(z) and Su(z)12 in GSCs, gonialblasts, spermatogonia and very early spermatocytes (Fig. 1A,B). Strikingly, E(z) and Su(z)12 protein levels abruptly dropped in spermatocytes in early G2 of meiotic prophase, immediately prior to the initiation of tTAF protein expression (arrowheads in Fig. 1A-A″,B-B″). E(z) and Su(z)12 are likely to be downregulated at least in part at the level of mRNA expression or stability, as in situ hybridization revealed a decrease in mRNA levels in the spermatocyte region compared with the apical region of the testis, where stem and precursor cells are located (see Fig. S1 in the supplementary material). The downregulation of PRC2 components was not due to the action of either the tTAFs or the tMAC component Aly, as Su(z)12 protein levels still dropped abruptly in early spermatocytes in testes from tTAF [e.g. ryan express (rye, TAF12L)] or aly mutant males (see Fig. S2A-A″,B-B″ in the supplementary material). Conversely, loss of PRC2 did not allow precocious tTAF expression in precursor cells, as the tTAF Cannonball (Can, TAF5L) was not turned on in E(z) mutant spermatogonia (see Fig. S2C-C″ in the supplementary material). At least some of the E(z) mutant germ cells differentiated into round or elongating spermatids (see Fig. S2D-D″ in the supplementary material), indicating that E(z) function was not required for germ cell differentiation from spermatocyte to spermatid.

The abrupt decline in Su(z)12 and E(z) proteins in differentiating male germ cells was dependent on the switch to the spermatocyte program. The Su(z)12 and E(z) proteins remained high in the spermatogonial cysts that overproliferate in bag of marbles (bam) mutant testes (Fig. 1C,C′), in which the transition from spermatogonia to spermatocytes is entirely abolished (Gonczy et al., 1997; McKearin and Spradling, 1990).

Immunofluorescence staining revealed that the histone modification H3K27me3, which is made by the PRC2 component E(z), was abundant in early male germ cells and persisted into spermatocytes, even after the abrupt downregulation of E(z) and Su(z)12 proteins (Fig. 1D, arrowheads; Fig. 1E, cells on the left side). The fact that spermatocytes have already completed pre-meiotic S phase, which is the last DNA replication event of this lineage, prior to the downregulation of E(z) protein might contribute to the perdurance of the H3K27me3 modification in later-stage germ cells, including spermatocytes. Notably, marked clones of germ cells that were null mutant for E(z) had very much reduced H3K27me3, even as spermatocytes (Fig. 1D,E, dotted circles), indicating that E(z) encodes the sole or predominant histone methyltransferase in Drosophila male germ cells that generates the H3K27me3 epigenetic mark recognized by the antibody.

Although two PRC2 components, E(z) and Su(z)12, are downregulated after male germ cells switch from the spermatogonial mitotic proliferation to the spermatocyte cell growth and differentiation program, the PRC1 components Polycomb, Polyhomeotic and dRing remain expressed in spermatocytes (Chen et al., 2005), indicating that the expression of PRC1 components and of certain PRC2 components are regulated differentially in the male GSC lineage.

The state of the promoters of three terminal differentiation genes in precursor cells was assessed by chromatin immunoprecipitation (ChIP) using bam mutant testes, which are enriched with precursor cells but lack spermatocytes and spermatids. The terminal differentiation genes Mst87F, don juan (dj) and fuzzy onions (fzo) are not expressed in spermatogonia but are first transcribed in spermatocytes, in which they require wild-type function of both tMAC and tTAFs for normal expression (Hiller et al., 2004; Hiller et al., 2001; White-Cooper et al., 1998). Microarray analysis indicated that these differentiation genes are upregulated ~100- to 200-fold in wild-type compared with bam mutant testes (Fig. 2). fzo and dj were still essentially silent in testes from males mutant for the tMAC subunit aly. Although low levels of Mst87F transcript were detected by microarray in aly mutant testes (Fig. 2), expression was so low that Mst87F mRNA is not detected by northern in aly mutant testes although abundant transcript is detected in wild type (White-Cooper et al., 1998). The three differentiation transcripts were measurably expressed in tTAF [spermatocyte arrest (sa, TAF8L)] mutant testes, higher than in aly mutant testes but much lower than in wild-type testes, based on northern blot, in situ hybridization (White-Cooper et al., 1998; Hiller et al., 2001) and microarray analysis (Fig. 2).

To compare ChIP results among different genotypes, which by necessity had to be tested in separate reactions, enrichment of the promoter regions of the terminal differentiation genes was compared with enrichment of the Cyclin A (CycA) gene in the same sample (Fig. 3). Similar trends held when the ChIP data were plotted as a percentage of the input DNA (Input %) without normalizing to CycA (see Fig. S3 in the supplementary material); however, in this case it is harder to compare absolute levels among different genotypes. CycA was chosen for normalization for two reasons. First, our quantitative analysis of CycA mRNA levels by microarray (Fig. 2D) showed that levels of its transcript were similar in bam, aly, sa mutant and wild-type testes. In situ analysis of CycA mRNA levels also indicates its presence in both precursor cells and spermatocytes, as well as in tTAF and aly mutant spermatocytes (White-Cooper et al., 1998). Second, previous studies have shown that CycA protein is expressed in both spermatogonia and spermatocytes in wild-type, tTAF and aly mutant testes but not in spermatids (Lin et al., 1996), and that the level of CycA protein is equivalent between wild-type and either tTAF or aly mutant testes (White-Cooper et al., 1998). These features make CycA a good control on a per cell basis for the germ cells in which most transcription is taking place. Thus, normalizing for CycA across genotypes should correct for different numbers of spermatogonia and spermatocytes per testis in wild-type versus mutant genotypes, as well as for the presence of spermatids (which are mainly transcriptionally inactive but maintain stabilized mRNAs for many differentiation genes) in wild-type testes but not in mutant testes.

Reflecting the transcriptionally silent state of the terminal differentiation genes in proliferating spermatogonia, when antibodies against H3K4me3, a marker of transcriptionally active chromatin (Barski et al., 2007; Dou et al., 2006; Milne et al., 2002), were used in ChIP experiments with bam mutant testes, enrichment for the promoter regions of Mst87F, dj and fzo was much lower than at the promoter region of CycA, which is actively transcribed in spermatogonia (Fig. 3A and see Fig. S3A in the supplementary material). Likewise, ChIP from bam mutant testes with antibodies against the unphosphorylated form of Pol II (8WG16), which is the form in the preinitiation complex, revealed that Pol II occupancy at the promoter region of Mst87F, dj and fzo was near background and significantly lower than Pol II occupancy near the promoter of the CycA gene (P<0.01 in Fig. 3B; see Fig. S3B in the supplementary material). Similar results were obtained with antibodies (4H8) that recognize both unphosphorylated and Ser5 phosphorylated forms of Pol II (Guenther et al., 2007; Stock et al., 2007) (P<0.05 in Fig. 3C; see Fig. S3C in the supplementary material), suggesting that the differentiation genes do not have RNA polymerase at their promoters in precursor cells.

By contrast, ChIP with anti-H3K27me3 substantially enriched for Mst87F, dj and fzo compared with CycA in bam mutant testes (Fig. 3D and see Fig. S3D in the supplementary material), consistent with action of the PRC2 component E(z) on terminal differentiation genes in precursor cells. Likewise, ChIP with anti-Polycomb also enriched for Mst87F, dj and fzo compared with CycA in bam mutant testes (Fig. 3E and see Fig. S3E in the supplementary material), suggesting occupancy of these terminal differentiation genes by the PRC1 machinery. Together, the ChIP results suggest that the terminal differentiation genes tested lack the active markers Pol II and H3K4me3, but show markers of action by PcG components in precursor cells, consistent with ChIP-seq results (Gan et al., 2010a) (see Fig. S4 in the supplementary material).

As male germ cells proceed along the differentiation pathway, a Bam-dependent switch allows exit from the transit-amplifying spermatogonial stage and entry into the spermatocyte differentiation program (Insco et al., 2009). Once this switch occurs, the cells turn on expression of a number of spermatocyte-specific transcripts, including several testis-specific components of the tMAC complex and the five tTAFs. In a second transcriptional wave, tMAC and tTAFs then activate the expression of many spermatid differentiation genes, including Mst87F, dj and fzo (White-Cooper et al., 1998).

The switch from spermatogonial proliferation to the spermatocyte differentiation program appears to trigger the recruitment of Pol II to terminal differentiation gene promoters, perhaps poising these genes for robust expression. Mst87F, dj and fzo were still largely transcriptionally silent in aly mutant testes, which have large numbers of spermatocytes (Fig. 2). ChIP analysis indicated that H3K4me3 levels near the Mst87F, dj and fzo promoters were extremely low in aly mutant testes (Fig. 3A and see Fig. S3A in the supplementary material), comparable to the level in bam mutant testes. However, the relative enrichment of the Mst87F, dj and fzo promoter regions by ChIP with antibodies specific for the unphosphorylated form of Pol II (8WG16) increased in aly mutant compared with bam mutant testes (Fig. 3B and see Fig. S3B in the supplementary material). Indeed, in aly mutant testes, the relative enrichment of Mst87F, dj and fzo by ChIP with antibodies against the unphosphorylated form of Pol II was similar to the enrichment of CycA, which is actively expressed (Fig. 3B and see Fig. S3B in the supplementary material).

The unphosphorylated form of Pol II was also detected as associated with the differentiation gene promoters by ChIP from tTAF mutant testes, in which Mst87F, dj and fzo transcripts are detectable but still very low (Fig. 2). Upon ChIP from can mutant testes with antibodies against the unphosphorylated form of Pol II, enrichment of Mst87F was slightly higher than that of CycA (P<0.05, Fig. 3B) and enrichment of dj and fzo was similar to that of CycA (Fig. 3B). Enrichment of the differentiation gene promoters by ChIP with anti-Pol II was higher in wild-type than in can testes (Fig. 3B,C). By contrast, in bam mutant testes, enrichment of the three terminal differentiation genes by ChIP with anti-unphosphorylated Pol II (8WG16) was substantially lower than that of CycA (P<0.01, Fig. 3B). ChIP using the antibody (4H8) that recognizes both unphosphorylated and Ser5 phosphorylated forms of Pol II gave similar results. Enrichment of the three terminal differentiation genes by ChIP with the 4H8 anti-Pol II antibody in aly or can mutant testes was similar to that of the active CycA gene promoter (Fig. 3C), with the exception that enrichment of Mst87F in aly mutant testes was slightly lower than that of CycA. By contrast, enrichment of the differentiation gene promoters by ChIP with the 4H8 anti-Pol II antibody was significantly lower than that of CycA in bam mutant testes (P<0.01 for Mst87F and dj, P<0.05 for fzo, Fig. 3C). The presence of unphosphorylated Pol II (recognized by the 8WG16 antibody) or Ser5 phosphorylated Pol II (recognized by the 4H8 antibody) at the differentiation genes in aly mutant testes, in which the genes are still largely silent, is reminiscent of the stalled Pol II observed at many transcriptionally quiescent genes in ESCs (Guenther et al., 2007; Stock et al., 2007), as well as in Drosophila embryos (Muse et al., 2007; Zeitlinger et al., 2007). Enrichment of the differentiation gene promoters by ChIP with anti-Pol II was still higher in wild-type than in can mutant testes (Fig. 3B,C).

Presence of the PRC1 complex near the promoter region of the terminal differentiation genes might contribute to their transcriptionally silent state even in the presence of Pol II in aly or tTAF mutant testes. ChIP with antibodies against the PRC1 component Polycomb enriched for the promoter regions of the three terminal differentiation genes 3- to 4-fold relative to the CycA promoter in bam testes and by 2- to 3-fold relative to CycA in aly or can mutant testes. By contrast, in wild-type testes, where these genes are fully expressed, enrichment for the three differentiation genes was comparable to that of CycA (Fig. 3E and see Fig. S3E in the supplementary material). The level of Polycomb at the differentiation genes in spermatocytes might be even lower than that indicated by the ChIP results using entire wild-type testes (Fig. 3E and see Fig. S3E in the supplementary material), as wild-type testes also contain precursor cells, in which the genes are highly occupied by Polycomb (see bam data in Fig. 3E and Fig. S3E in the supplementary material). However, because spermatocytes significantly outnumber precursor cells in wild-type testes, we can still gain information on the chromatin state of differentiation genes using whole testes, especially for changes in which a particular chromatin mark or protein occupancy appears or turns on as cells switch from precursor to spermatocyte. The relative enrichment of Polycomb at the differentiation genes in this study (~3- to 4-fold enrichment of Mst87F in can mutant compared with wild-type testes) was comparable to that reported in a previous study (Chen et al., 2005), when the results were plotted in the same way (see Fig. S5 in the supplementary material).

The wild-type function of tTAFs is required to reduce the occupancy of Polycomb at the differentiation gene promoters (Fig. 3E) and to turn on robust expression of the differentiation genes (Fig. 2) (Chen et al., 2005). Analysis of a double-mutant strain generated by combining the tTAF mutant no hitter (nht, TAF4L) with an E(z) conditional allele, E(z)⁶¹, which was used previously to inactivate E(z) activity at restrictive temperature (Wang et al., 2004b), indicated that the action of tTAFs is required for more than just abolishing silencing by PcG. When nht; E(z)⁶¹ double-mutant flies were grown to adulthood at permissive temperature and then shifted to 29°C (the restrictive temperature) for 7 days, little E(z) function remained, as H3K27me3 was almost undetectable in the double-mutant testes extracts, as compared with the robust levels in testes from sibling control nht mutant males that were raised under exactly the same conditions (Fig. 4A). However, no significant change in transcript levels for Mst87F, dj or fzo was detected by quantitative RT-PCR analyses in nht; E(z) double-mutant testes as compared with nht single-mutant sibling controls (Fig. 4B). In both nht and nht; E(z) mutant testes, transcript levels for the differentiation genes were very much reduced compared with those of wild-type testes shifted through the same temperature regimen in parallel, suggesting that inactivation of E(z) is insufficient to turn on differentiation gene expression without tTAF function.

A number of spermatid differentiation genes require wild-type function of both tMAC and the tTAFs for normal expression in spermatocytes (White-Cooper et al., 1998). Results from ChIP analysis of testis extracts with antibodies directed against the TAF8L Sa indicated that the proper function of spermatocyte-specific components of tMAC is required for the tTAFs to properly associate with the promoter regions of these differentiation genes. ChIP of wild-type testes with antibodies against Sa enriched for Mst87F, dj and fzo ~3- to 6-fold over the level of the control gene CycA, which is transcribed in spermatocytes independently of tTAFs (Fig. 5A). By contrast, although Sa protein was present in aly mutant spermatocytes (Fig. 5C′), ChIP of aly mutant testes with anti-Sa did not significantly enrich for any of these three target genes compared with CycA (Fig. 5A).

The wild-type function of tMAC is also required for proper subnuclear localization of tTAFs and PRC1 proteins in spermatocytes. In wild-type spermatocytes, immunofluorescence analysis revealed colocalization of tTAF proteins (e.g. Sa and Can) with Polycomb and other PRC1 components to a subdomain within the nucleolus, interdigitated with the nucleolar marker Fibrillarin (Fib) (Fig. 5B-B″) (Chen et al., 2005). By contrast, in aly mutant spermatocytes, the tTAFs were concentrated in lobe-shaped domains outside of, but next to, the Fib-enriched nucleolar region (Fig. 5C-C″,D-D″). Similar results were observed for the TAF4L Nht (data not shown) and for the TAF6L Meiosis I arrest (Mia) (Metcalf and Wassarman, 2007). The perinucleolar lobes containing mislocalized tTAFs were also apparent in freshly squashed live samples viewed by phase contrast microscopy, in which they appeared as phase-dark structures extending from the nucleolar periphery (Fig. 5F-F″, compare with the wild-type spermatocyte in Fig. 5E-E″). A substantial amount of the Polycomb protein in aly mutant spermatocytes was also localized to the same perinucleolar domains (Fig. 5H-H″, Fig. 4I-I″). Some Polycomb still retained association with condensing chromatin in aly mutant spermatocytes (Fig. 5H,H′,I,I′), as in wild-type spermatocytes (Fig. 5G,G′).

The localization of tTAFs to the perinucleolar lobes in aly mutant spermatocytes was not due to total disruption of the nucleolus, as phase contrast microscopy (Fig. 5F″) and staining for the nucleolar marker Fib (Fig. 5C″,D″) indicated that the nucleolus was present. It is possible that the wild-type function of Aly is required to maintain proper nucleolar architecture in spermatocytes (Metcalf and Wassarman, 2007), and that collapse of nucleolar architecture in aly mutant spermatocytes squeezed the tTAF-containing subcompartment to the perinuclear lobes.

Similar mislocalization of tTAFs to the lobed perinucleolar domains was also observed in spermatocytes mutant for other testis-specific components of the tMAC complex, including cookie monster (comr), tombola (tomb), matotopetli (topi) and achintya/vismay (achi/vis; achi – FlyBase) (Fig. 5J-M). However, even though Myb-interacting protein 40 (Mip40) has been identified biochemically as a component of tMAC (Beall et al., 2007), tTAFs were localized within the nucleolus, interdigitated with the Fib-positive subcompartment, in mip40 mutant spermatocytes, similar to in the wild type (Fig. 5N), suggesting that Mip40 has different role(s) than the other tMAC components in spermatocytes.

Our results suggest a stepwise series of developmentally programmed events as terminal differentiation genes convert from a transcriptionally silent state in precursor cells to full expression in differentiating spermatocytes (Fig. 6). In precursor cells, differentiation genes are repressed and associated with background levels of hypophosphorylated Pol II and H3K4me3. These genes also display elevated levels of H3K27me3 and Polycomb at the promoter region, suggesting that they are acted upon by the PcG transcriptional silencing machinery. Notably, the differentiation genes studied in precursor cells here did not show the hallmark bivalent chromatin domains enriched for both the repressive H3K27me3 mark and the active H3K4me3 mark that have been characterized for a cohort of differentiation genes in mammalian ESCs (Bernstein et al., 2006).

The cell fate switch from proliferating spermatogonia to the spermatocyte differentiation program initiates both global and local changes in the transcriptional regulatory landscape, starting a cell type-specific gene expression cascade that eventually leads to robust transcription of the terminal differentiation genes. Globally, soon after the switch from spermatogonia to spermatocytes, core subunits of the PRC2 complex are downregulated, including E(z), the enzyme that generates the H3K27me3 mark. Locally, after male germ cells become spermatocytes, Pol II accumulates at the terminal differentiation gene promoters, although these genes still remain transcriptionally silent, with low H3K4me3 and high Polycomb protein levels near their promoters.

The next step awaits the expression of spermatocyte-specific forms of core transcription machinery and chromatin-associated regulators, including homologs of subunits of both the general transcription factor TF_IID (tTAFs) and the MIP/dREAM complex (Aly and other testis-specific components of tMAC) (Fig. 6B). The tMAC complex acts either locally or globally, perhaps at the level of chromatin or directly through interaction with tTAFs, to allow recruitment of tTAFs to promoters of target terminal differentiation genes. The action of tTAFs then allows full and robust transcription of the terminal differentiation genes, partly by displacing Polycomb from their promoters.

Strikingly, the two major PcG protein complexes appear to be regulated differently by the germ cell developmental program: whereas the PRC2 components E(z) and Su(z)12 are downregulated, the PRC1 components Polycomb, Polyhomeotic and dRing continue to be expressed in spermatocytes. The global downregulation of the epigenetic ‘writer’ E(z) in spermatocytes might facilitate displacement of the epigenetic ‘reader’, the PRC1 complex, from the differentiation genes, with the local action of tTAFs at promoters serving to select which genes are relieved of PRC1. In addition, the tTAFs act at a second level to regulate Polycomb by recruiting and accompanying Polycomb and several other PRC1 components to a particular subnucleolar domain in spermatocytes (Fig. 6C) (Chen et al., 2005). It is not yet known whether sequestering of PRC1 to the nucleolus by tTAFs plays a role in the activation of terminal differentiation genes, perhaps by lowering the level of PRC1 that is available to exchange back on to differentiation gene promoters. Conversely, recruitment of PRC1 to the nucleolar region might have a separate function, such as in chromatin silencing in the XY body as observed in mammalian spermatocytes (Baarends et al., 2005; Takada et al., 2007).

Our findings indicate that, upon the switch from spermatogonia to spermatocytes, the terminal differentiation genes go through a poised state, marked by presence of both active Pol II and repressive Polycomb, before the genes are actively transcribed. Stalled Pol II and abortive transcript initiation are emerging as a common feature in stem/progenitor cells. This mechanism may prime genes to rapidly respond to developmental cues or environmental stimuli (Muse et al., 2007; Zeitlinger et al., 2007). Stalled Pol II could represent transcription events that have initiated elongation but then pause and await further signals, as in the regulation of gene expression by the androgen receptor (Zhao et al., 2008) or by heat shock (Lis, 1998; Rasmussen and Lis, 1995; Rougvie and Lis, 1988). Alternatively, Pol II might be trapped at a nascent preinitiation complex, without melting open the DNA, as found in some instances of transcriptional repression by Polycomb (Dellino et al., 2004). Although our ChIP analyses did not have the resolution to distinguish whether Pol II was stalled at the promoter or had already initiated a short transcript, the results with antibodies specific for unphosphorylated Pol II suggest that Pol II is trapped in a nascent preinitiation complex. The PRC1 component dRing has been shown to monoubiquitylate histone H2A on Lys119 near or just downstream of the transcription start site (Wang et al., 2004a). We propose that in early spermatocytes, before expression of the tTAFs and tMAC, the local action of PRC1 in causing H2AK119ub at the terminal differentiation gene promoters might block efficient clearing of Pol II from the preinitiation complex and prevent transcription elongation.

Removal of PRC1 from the promoter and full expression of the terminal differentiation genes in spermatocytes require the expression and action of tMAC and tTAFs. Cell type-specific homologs of TF_IID subunits have been shown to act gene-selectively to control developmentally programmed gene expression. For example, incorporation of one subunit of the mammalian TAF4b variant into TF_IID strongly influences transcriptional activation at selected promoters, directing a generally expressed transcriptional activator to turn on tissue-specific gene expression (Liu et al., 2008).

The local action of the tTAFs to relieve repression by Polycomb at target gene promoters provides a mechanism that is both cell type specific and gene selective, allowing expression of some Polycomb-repressed genes while keeping others silent. Similar developmentally programmed mechanisms may also reverse PcG-mediated epigenetic silencing in other stem cell systems. Indeed, striking parallels between our findings and recent results from mammalian epidermis (Ezhkova et al., 2009; Sen et al., 2008) suggest that molecular strategies are conserved from flies to mammals. In mouse epidermis, the mammalian E(z) homolog Ezh2 is expressed in stem/precursor cells at the basal layer of the skin. Strikingly, as we observed for E(z) and Su(z)12 in the Drosophila male GSC lineage, the Ezh2 level declines sharply as cells cease DNA replication and the epidermal differentiation program is turned on. Overexpression of Ezh2 in epidermal precursor cells delays the onset of terminal differentiation gene expression (Ezhkova et al., 2009), and removal of the Ezh2-generated H3K27me3 mark by the Jmjd3 (Kdm6b) demethylase is required for epidermal differentiation (Sen et al., 2008).

In particular, our results suggest a possible explanation for the conundrum that, although PcG components are bound at many transcriptionally silent differentiation genes in mammalian ESCs, loss of function of PcG components does not cause loss of pluripotency but instead causes defects during early embryonic differentiation (Boyer et al., 2006; Chamberlain et al., 2008; Surface et al., 2010). In Drosophila male germ cells, events during the switch from precursor cell proliferation to differentiation are required to recruit Pol II to the promoters of differentiation genes. Without this differentiation-dependent recruitment of Pol II, loss of Polycomb is not sufficient to precociously turn on terminal differentiation genes in precursor cells. Rather, Polycomb that is pre-bound at the differentiation gene promoters might serve to delay the onset of their transcription after the mitosis-to-differentiation switch. Robust transcription must await the expression of cell type- and stage-specific components of the transcription machinery. These might in turn guide gene-selective reversal of Polycomb repression to facilitate appropriate differentiation gene expression in specific cell types.

We thank P. Dimario and M. Pollard for anti-Fibrillarin; J. Muller for E(z)⁷³¹ FRT2A flies and anti-Su(z)12 antibody; R. Jones for anti-E(z) antibody; R. Paro for Pc-GFP flies and anti-Pc; R. Kingston and A. Saurin for anti-Pc; H. White Cooper for tMAC mutant strains and Q. Gan for analyzing and depositing microarray data. C.L. was supported in part by a Stanford Dean's Fellowship and J.R.M.P. by an NSF Predoctoral Fellowship. This work was supported by NIH K99/R00 HD055052 Pathway to Independence Award and Research Grant No. 05-FY09-88 from the March of Dimes Foundation to X.C. and by NIH 3 R01 GM061986 to M.T.F. Deposited in PMC for release after 12 months.