Chromatin and Transcription Transitions of Mammalian Adult Germline Stem Cells and Spermatogenesis

Transcription

1 Resource Chromatin and Transcription Transitions of Mammalian Adult Germline Stem Cells and Spermatogenesis Saher Sue Hammoud,,4 Diana H.P. Low,, Chongil Yi, Douglas T. Carrell, 4 Ernesto Guccione,,, * and Bradley R. Cairns, * Howard Hughes Medical Institute, Department of Oncological Sciences, and Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City, UT 84, USA Division of Cancer Genetics and Therapeutics, Institute of Molecular and Cell Biology, A*STAR (Agency for Science, Technology and Research), Singapore 974, Singapore Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 974, Singapore 4 Department of Surgery, University of Utah School of Medicine, Salt Lake City, UT 84, USA *Correspondence: eguccione@imcb.a-star.edu.sg (E.G.), brad.cairns@hci.utah.edu (B.R.C.) SUMMARY Adult germline stem cells (AGSCs) self-renew (Thy + enriched) or commit to gametogenesis (Kit + enriched). To better understand how chromatin regulates AGSC biology and gametogenesis, we derived stage-specific high-resolution profiles of DNA methylation, 5hmC, histone modifications/variants, and RNA-seq in AGSCs and during spermatogenesis. First, we define striking signaling and transcriptional differences between AGSC types, involving key self-renewal and proliferation pathways. Second, key pluripotency factors (e.g., Nanog) are silent in AGSCs and bear particular chromatin/ attributes that may poise them for reactivation after fertilization. Third, AGSCs display chromatin poising/bivalency of enhancers and promoters for embryonic transcription factors. Remarkably, gametogenesis occurs without significant changes in and instead involves transcription of DNA-methylated promoters bearing high RNAPol, HK9ac, HK4me, low CG content, and (often) 5hmC. Furthermore, key findings were confirmed in human sperm. Here, we reveal AGSC signaling asymmetries and chromatin/ strategies in AGSCs to poise key transcription factors and to activate DNA-methylated promoters during gametogenesis. INTRODUCTION Cellular epigenomes are complex, reflecting the cell s current transcriptional program, its developmental history, and the poising of genes for future expression. Viewed this way, male adult germline stem cells (AGSCs) represent a particularly interesting node in development. First, they are the final germline stem cell, originating from primordial germ cells (PGCs). Second, AGSCs sit at the beginning of a unipotent developmental pathway involving meiosis and, subsequently, spermiogenesis. A major question of AGSC biology is whether and how DNA methylation () and chromatin patterns play major roles in AGSC identity, unipotency, self-renewal, and their preparation for and commitment to spermatogenesis. Likewise, once committed to spermatogenesis, do and chromatin dynamics help guide the developmental stages while still retaining competency for genome pluripotency after fertilization? This work aims to address these questions through extensive and chromatin profiling of AGSCs and key stages of spermatogenesis in the mouse, and through comparisons to PGCs and embryonic stem cells (s). Consideration of AGSCs is informed by studies on their precursors: zygotes, inner cell mass (ICM)/s, and PGCs. Zygotes and PGCs are the two mammalian cell types that undergo extensive genomewide changes in their epigenomes, termed reprogramming (Guibert et al., ; Hajkova et al., 8; Reik et al., ; Sasaki and Matsui, 8; Smith et al., ). Following fertilization, the paternal genome undergoes chromatin repackaging, and both the paternal and maternal genomes experience active and passive lowering of (Guibert et al., ; Hajkova et al., 8; Reik et al., ; Sasaki and Matsui, 8; Smith et al., ). The second phase of reprogramming occurs in PGCs during their migration and early residence within the gonad, and it involves nearly complete removal of, including the erasure of parental imprints (Dawlaty et al., ; Hackett et al.,, ; Seisenberger et al., ; Seki et al., 5). Following this phase, levels in PGCs/gonocytes are slowly restored (embryonic day [E].5 E6.5), and sex-specific imprints are established in the male gonocytes prior to birth, generating spermatogonial stem cells (SSCs) in the neonate and AGSCs in the adult. Here, we provide a genomic understanding of two key AGSC states: the self-renewing and transplantable state (Thy + ) versus the state committed to gametogenesis (Kit + ). We also provide an in-depth genomic examination of spermatogenesis. SSCs/AGSCs either self-renew to form two single unpaired cells (A s ) or instead form a paired cell connected by an intracellular bridge (A pr ), from which subsequent divisions give rise to aligned chains of A al and B spermatogonia committed to meiosis. Meiosis then creates tetraploid primary spermatocytes, diploid secondary spermatocytes, haploid round Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. 9

2 spermatids, and, ultimately, spermatozoa (Hess and Renato de Franca, 8). These transitions have been characterized by global changes in chromatin (Delaval et al., 7; Khalil et al., 4; Oakes et al., 7; Soumillon et al., ; Turner, 7). However the gene-specific locations of these modifications and their relationships to and transcription have largely not been characterized at a genomewide scale. Interestingly, in humans and mice, mature germ cells retain nucleosomes (often modified at HK4me and HK7me) at the promoters of hundreds of genes of developmental importance in the embryo, including Hox-, Fox-, Sox- and Gata-family genes, particular micrornas (mirnas), and long-noncoding RNAs (Brykczynska et al., ; Hammoud et al., 9). The coincidence of HK4me (correlated with gene activation) with HK7me (correlated with silencing) is termed bivalency a chromatin property discovered at developmental genes in s that underlies gene poising (Bernstein et al., 6). Notably, these bivalent promoters are almost invariably DNA hypomethylated in both s and sperm. Recently, bivalent nucleosomes have been localized at developmental genes in spermatocytes and spermatids in mice (Erkek et al., ; Lesch et al., ), and also at E6.5 PGCs (Ng et al., ), but AGSCs have not been tested. Here, our work uses extensive transcriptome and chromatin profiling to address several outstanding issues, including: () the transcriptome and chromatin/dna differences between self-renewing versus committed/differentiating AGSCs, () whether bivalency and DNA hypomethylation resides at genes for embryo development in AGSCs, () whether key genes for pluripotency (e.g., Nanog and Sox) are bivalent/ poised in AGSCs, and (4) how chromatin and are used to regulate genes during germline development. RESULTS Genomic Profiling of Thy + or Kit + AGSCs and Staged Spermatogenesis in Mice Thy + /Kit AGSCs are self-renewing and transplantable, whereas Thy /Kit + cells are poorly transplantable and largely committed to gametogenesis (Oatley et al., 9). Here, we isolated these two populations from the adult (8 weeks, detailed later). In addition, we isolated cells from the three main stages of gametogenesis: spermatocytes, spermatids, and mature sperm (Figure A; for all isolations and purity/validation, see Figure SA available online). Below, we provide extensive profiling of these cell types. analyses involved whole-genome shotgun bisulfite sequencing (conversion efficiency > 99%) and typically > genome coverage (Table SA). For comparisons, we reprocessed data from s and PGCs (Seisenberger et al., ; Stadler et al., ). Our 5hmC and histone modification data sets in germ cells were compared to prior s and/ or PGC data sets (Mikkelsen et al., 7; Ng et al., ). Transcriptional profiling of long and various small RNAs involved strand-specific RNA sequencing (RNA-seq). Thy + and Kit + AGSCs Display Distinctive Gene Expression Signatures We first examined the known molecular signatures/markers that distinguish Thy + versus Kit + AGSCs, using stringent criteria (p <.5 and is greater than or equal to 4-fold change). First, we find Thy or Kit selectively expressed, in alignment with our enrichment goals (Figures B and C). In Thy + cells, we find colony-stimulating factor receptor (Csfr) and integrin B (Itgb) specifically expressed (Oatley et al., 9) and Gfra (the receptor for the GDNF ligand) expressed at somewhat higher relative levels a pathway of known importance for AGSC self renewal in neonates. In Kit + cells, Stra8 and multiple aldehyde dehydrogenases (e.g., Aldh) are specifically expressed, consistent with known roles for retinoic acid in committing Kit + cells to meiosis (Figures B, C, and SB). Notably, Aldh, Stra8, and Spo are the only promoters that lose HK7me (Figure SB; data not shown), suggesting their bivalent-to-active chromatin transition. Thus, our data sets verify the known distinctive differences between Thy + and Kit + cells and reveal very rare instances of bivalency resolution in germline development. Thy + and Kit + AGSCs Display Distinctive Signaling Pathways and Chromatin Transitions We then identified factors with strong expression selectively in Thy + versus Kit + cells. Notably, Thy + cells selectively upregulate particular receptors (Fgfr and Tnfr), cell cycle regulators (Ccne), proto-oncogenes (Fyn and Lyn), growth factors (Tgfb, Gdf5, and Pdgf), and tumor suppressors (Dab and Aim)(Figures B and C). Likewise, we find many signaling factors/pathways specifically expressed/upregulated in Kit + cells, including multiple receptors (Egfr, Fgfr, Lifr, Pdgfra, Esr) and several nuclear hormone orphan receptors, signaling factors (Bmp4), and cell cycle factors (Ccnd) (Figures B and C; data not shown). Notably, several JNK pathway target genes (Atf, Jun, p5, and Rac) are markedly upregulated in Kit + AGSCs (compared to Thy + ), consistent with their proliferative state. It is interesting that we found H-Ras pathway components highly expressed in both AGSC types. Furthermore, several key transcription factors are expressed specifically in Kit + cells, including c-myc, Fos, and Solhlh/. Here, roles for Solhlh/ in promoting spermatogonial differentiation and meiosis are known (Ballow et al., 6; Hao et al., 8), whereas functional roles for c-myc or Fos remain to be explored. Taken together, our data sets should prove highly useful in optimizing cell markers for transplantation studies, for studying signaling within the testicular niche, and in revealing new factors that may promote AGSC renewal or differentiation (Figure D). Rare Differences between Thy + and Kit + AGSCs We find global patterns of Thy + and Kit + AGSCs remarkably similar (r =.96; Figure E; Figure SC) and differences at promoters extremely rare; only seven promoters exhibit >% change in fraction : Syce, Tex, Hormad (important for meiosis), Ankrd7, and Ccin lose, whereas Kcng4 and Kcnmb (potassium channels) gain. However, moderate increases in both bulk and promoter levels are observed on transitioning to the Kit + stage (Figures A and B, clusters and 4; Table SA [note: PGCs/gonocytes are recovering from genomewide reprogramming]) but are proportional to the slight increase in genome average rather than a conversion from unmethylated to methylated. Thus, the transition between self-renewal and commitment to gametogenesis does not generally involve changes in promoter. 4 Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc.

3 A Scytes Stids AGSC Sperm Thy+ Total Germ cells small RNAs long RNAs 5hmC HK4me HK7me HK7ac HK4me HK9ac HA.Z Pol Pol BisSeq HK4me BisSeq B Log (fpkm (Thy+ / Kit +)) 5 5 Thy Tgfb Csfr Ccne Itgb Dab AIm Pdgf Aldhs Lifr Pdgfr Ccnd Egfr Stra8 Itgb Sohlh Myc C-Kit red = p-value <.5 e+ e+ e+ e+4 e+5 Mean of Normalized Counts C RNAseq E6.5 Thy+ FPKM m Chr9 5 5 Thy 4,85, 4,855, Chr8 Csfr 6,7, 6,9, 4 4 Chr Itgb 77,, Tgfb Chr7 6,47, 6,49, RNAseq E6.5 Thy+ FPKM m Aim Chr 7,5, 7,4, Chr8 7,54, Gdf Lifr Chr5 7,, Pdgfr Chr5 75,55, RNAseq E6.5 Thy+ FPKM m Chr5 4 Kit 76,, Myc Chr5 6,8, 6,8, Chr8 Itgb,, Chr,4, Egfr 6,7, 6,8, D E Intracellular: Ccne ; Fynn; Lyn; Dab; Aim Surface: CsfR; Itgb; Fgfr; Gfra; Ilr Tnfra; Notch/ Thy+ Active pathways: FGF, RAS, JAK/STAT RA, EGF, JNK, LIF Secreted: Tgfb; Gdf5; Pdgf Secreted: Bmp4 Intracellular: Ccnd; Esr; Orphan NHR Solhl/; Myc; Fos Stra8; Aldh Surface: Itgb; Egfr; Fgfr; Lifr; Pdgfr; Notch Fraction Methylation r = Thy+ Fraction Methylation Figure. Expression of Key Stem Cell and Self-Renewal Factors Differ between Thy + and Kit + AGSCs (A) Graphical summary of data sets generated (gray boxes)., spermatocytes;, spermatids. (B) MA plot of Thy + /Kit + expression data sets. Red dots indicate p <.5. (C) Transcriptional dynamics of stem cell factors in s and germ cells. Browser snapshots of Itgb/, Myc, Klf4, Csfr, Thy, Kit, Lifr, Tgfb, Aim, Gdf5, Egfr, and Pdgfra. y axis: RNA-seq (FPKM). Note: mouse (m) RNA-seq data are from Klattenhoff et al. (); PGC E6.5 RNA-seq data are from Seisenberger et al. (). (D) Graphical summary of genes and pathways differentially expressed between Thy + and Kit + cells. (E) Scatterplots comparing (fraction CG methylation) at promoters (TSS ± kb) in Thy + versus Kit +. See also Figure S and Tables SA SC and SF. Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. 4

4 A Fraction DNA Methylation (mcg/cg) C Fraction (mcg/total CG) HK7me Qval FDR HK4me Qval FDR D kb E6.5 Thy+ CGI 4 E.5 5 Thy+ 4 E.5 5 Thy+ Chr E.5 Thy+ HK4me HK7me TSS E6.5 Thy+ +5kb B, RefSeq Promoters Cluster Dev. TFs (Hox/Gata) Cluster Immune Processes Cluster Cell-cell Signaling, Ecto/Mesoderm dev. Cluster4 RNA Processing DNA Repair Proliferation E6.5 Thy+ HoxD Cluster Fraction DNA Methylation..67. AGSCs, s, and PGCs Share DNA Demethylation and Bivalency of Developmental Genes Next, we compared chromatin/ patterns of germline and s. At promoters, profiles of s, E6.5 PGCs, and AGSCs (Thy + and Kit + ) are highly similar (r >.85 for all pairwise comparisons; Figure SC); evident at promoter (class average) maps of (Figure A), promoter clustering (Figure B, clusters and ), and when viewing along the physical map (HoxD locus; Figure C). As expected, DNA hypomethylation aligns with CpG island strength (Figure C; Table SB). Notably, both Thy + and Kit + AGSCs display regions of bivalent chromatin with underlying DNA hypomethylation, which localized to the promoters of most transcription factors utilized during embryo development, including Hox-, Fox-, Sox-, Tbx-, and Gata-family transcription factors (e.g., HoxD locus; Figures C and D, cluster ; Table SA), but not other gene classes (e.g., housekeeping genes, Figure D, cluster 4; Table SB) properties shared with PGCs and s (Farthing et al., 8). It is interesting that, when considering bivalent genes, we found a 9% or 7% overlap when comparing s to AGSCs or AGSCs to PGCs, respectively (p <.). However, AGSCs have many more bivalent genes than s, including zinc finger clusters, protocadherins, Wnt pathway, germline specification, -5kb +5kb Fraction methlyation 74,5, 74,55, 74,6, 74,65, Evx Hoxd-as 4 Cluster Cluster Cluster Cluster4 E6.5 Thy+ MTX E F G z-score H PC Fraction CG Methylation HK4me FPKM HK7me FPKM Thy+ Meiosis-Specific and DNA repair and AGSC Genome Average E6.5 Genome Average E6.5 Thy+ Meiosis-Specific and DNA repair E.5 Thy+ E.5 Thy+ E.5 Thy+ E6.5 m 5 5 PC 4 Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. Stra8 Sycp Spo Spata Aldh Figure. and Chromatin Dynamics between Germ Stem Cells and s (A) Class average maps of (x axis, TSS ± 5 kb; y axis, fraction CG methylation) for all cell types. (B) K-means clustering (n = 4) of (mean fraction CG methylation) at TSS regions (± kb). data set from Stadler et al. (); E6.5 from Seisenberger et al. (). (C) A snapshot of the Hoxd cluster and neighboring genes, depicting (mean fraction CG methylation), CGIs, and HK4me or HK7me (q value [Qval] FDR) across all cell types. (D) K-means clustering of active and repressive histone modifications at all TSS (± kb), with enriched GO terms in the middle panel and class average plots on the right. Dev., developmental. (E) Average promoter of meiotic and DNA repair genes in s, PGCs, and AGSCs. (F) Boxplots representing the average HK4me or HK7me levels at meiotic and DNA repair genes. (G) RNA-seq heatmap (Z score) for selected meiotic genes (note: peak expression of Sycp, Spata, and Spo is in spermatocytes; data not shown). (H) PCA plot depicting transcriptome relationships between s, AGSCs, and PGCs. See also Figure S and Tables SA SC. PRDM family, toll-like receptors, and solute carrier genes. Finally, developmental transcription factors display relatively wide promoter and gene DNA hypomethylation (Figure D [compare clusters and 4, right panel]). Thus, key genes for embryo development are selectively packaged/poised in large regions of bivalent chromatin with accompanying DNA hypomethylation in the germline cycle. Profiles of AGSCs and s Differ at Key Loci for Germline Development Although is similar overall, comparison of AGSCs (either Thy + or Kit + ) to s yield differentially methylated promoters (with >% change in fraction methylation; Figure B, cluster ; Table SC). Genes DNA methylated in s, but demethylated in Thy + and Kit + AGSCs included Dppa, Stra8, Piwil, Mov, Scp-family, Tex-family, Rnf7, Sohlh, Tdrd, Catsper-family, Nobox, Adad, and Spag and GO analyses enriched gene categories of germline reproductive development (meiosis, Piwi-interacting RNAs [pirnas], chromatoid bodies), consistent with silencing of these genes in and their utilization in AGSC (Table SC). Conversely, genes specifically methylated in AGSCs included particular small nucleolar RNAs [snornas], and notably, key genes involved in self-renewal. AGSCs and PGCs Differ at Genes for Olfactory Receptors, Proliferation, and Meiosis We observed 5 gene promoters that acquire methylation between E6.5 PGCs/gonocytes and Kit + AGSCs (Figure B,

5 A RNAseq FPKM 5-hmC Qval FDR Fraction DNA Methylation HK7me Qval FDR HK4me Qval FDR C SOX m E6.5 Thy+ S Cytes S Tids S Cytes S Tids m E6.5 Thy+ S Cytes S Tids m E.5 Thy+ S Cytes S Tids m E.5 Thy+ 5 5 S Cytes S Tids Nanog Chr6,65,,66, 5hmC HK4me HK7me OCT4 KLF4 PGC Chr7 Oct4 Chr4 5,54, 5,55, Chr Sox 5,64, 5,645, 4,54, 4,554, Klf4 (E6.5) AGSC D Fraction kb NANOG kb kb ,,,, Chr OCT4.5kb Prdm kb B Chr OCT4/SOX Lefty 8,865, 8,87,.5kb NANOG LEFTY PRDM4 PLZF Fraction kb SOX.5kb kb KLF4.5kb m E6.5 Thy+ S Cytes STids Figure. AGSCs and s Differ in Chromatin Packaging and Utilization of Pluripotency Factors (A) Browser snapshots depicting genomic features of key pluripotency genes (Oct4, Nanog, Sox, Prdm4, Klf4) in mouse (m) and stages of spermatogenesis. A black bar denotes promoter regions, and a gold bar denotes the enhancer regions. S Cytes, spermatocytes; S Tids, spermatids. (B) Same as in (A) for the self-renewal factor Lefty. (C) Summary schematic for key pluripotency factors in s, PGCs (E6.5), and AGSCs. (D) Class average maps of (fraction CG methylation) across binding site regions (±.5 kb) for pluripotency factors determined in s. See also Figure S and Tables SD, SE, and SA SE. cluster 4; Table SC). Notably, the majority (9) are olfactory receptors, which may be needed for PGC homing/migration to the gonad but are silenced after residence is established. The remainder include genes linked to cell proliferation or metastasis (i.e., Cecams, Prl), proinflammatory cytokines (Il, Ifi7la, Irf7), and three mirnas (mir-87, mir-467b, and mir-48) with roles in apoptosis, angiogenesis, and cancer, respectively. Notably, we found the promoters of many key meiosis-specific/promoting genes (e.g., Stra8, Tex/4/5/9/4, Spo, Spata, Sycp//, DMRTc) partially DNA methylated in E6.5 PGCs (comparable to the PGC genome average) but fully hypomethylated in both AGSC populations (Figures E and SB; and data not shown). Curiously, meiotic gene promoters in Thy + cells are marked with HK4me, but not transcribed, and lack HK7me suggesting alternative modes of repression (Figures F and G). Thus, between E6.5 and AGSCs, many key genes for meiosis lose (likely to prepare for activation during spermatogenesis), whereas key genes linked to proliferation or cell migration become methylated and silenced. AGSCs, PGCs at E6.5, and s Differ in Stem/Germ Cell Markers and Pluripotency Factor Utilization Principal-component analysis of RNA profiling reveals similarity between AGSC types but striking differences between AGSCs and s or PGCs (Figure H), especially regarding pluripotency. Notably, AGSCs express high Klf4 and low Oct4, but in contrast to s, AGSCs entirely lack expression of Nanog, Sox, Lefty, and Prdm4 (Figure A), extending on recent data from E6.5 PGCs/gonocytes (Seisenberger et al., ). Furthermore, AGSCs share additional factors expressed in s (Klf, Etv5, Pin, Aldh, Myc, Msh, and Rex) (Figure SA). Although AGSCs lack certain self-renewal factors present in s, they Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. 4

6 A Refseq Genes B lncrnas Genes C NATs D pirna Clusters Cluster Cluster Cluster Expression z-score - Cluster Cluster Cluster Cluster Cluster Cluster4 Cluster Cluster4 E F HK4me LogRatio 4 5. r = Fraction Cluster5 5 7 Cluster5 Cluster Cluster4 Cluster5 Refseq Genes lncrna Genes NATs pirna Clusters. r = Fraction Spermatocytes Fraction Spermatids r = r = -.6 r = Fraction Cluster Fraction of Gene Promoters with >6% G Refseq Gene Promoters Key: All panels Typical; Promoter <6% Atypical; Promoter >6% 9 Peak Exp. in Peak Exp. in Peak Exp. in Peak Exp. in Peak Exp.in 4 4 # promoters CG obs/exp CG obs/exp CG obs/exp CG obs/exp CG obs/exp H Refseq Gene Promoter Attributes in Round Spermatid Reads per million.5 HK4me read density Promoter region.4 RNAPol read density HK7ac read density 5hmC read density HA.Z read density HK9ac read density Promoter region Promoter region Promoter region Promoter region Promoter region (legend on next page) 44 Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc.

7 express additional factors (absent in s) of known importance in the self-renewal of adult stem cells (e.g., Bmi, Lrig; Figure SA) or neonatal SSCs (e.g., Plzf, Dazl, Taf4b, and Gfra; Figure SA; data not shown). Taken together, AGSCs utilize a very different self-renewal network than s. Furthermore, certain factors of known importance in neonatal SSC self-renewal and proliferation (e.g., Bcl6b, Lhx, Sox, Neurog; Figure SA; data not shown) are not expressed in our isolated AGSCs, suggesting considerable regulatory differences in neonatal (and/or cultured) SSCs versus AGSCs. We then explored how chromatin might underlie the regulation of these key self-renewal genes in AGSCs. It is interesting that we found the transcriptional start site (TSS) regions of Nanog and Lefty DNA methylated in AGSCs (Figures A and B) but not in PGCs/gonocytes (E6.5), revealing acquisition between PGCs (E6.5) and AGSCs. Remarkably, for both Nanog and Lefty, we found the proximal enhancer DNA hypomethylated and bivalent (note that this enhancer is intragenic for Lefty), and we also observed the Lefty promoter highly enriched with 5hmC (Figures A and B). Furthermore, Sox and Prdm4 are also silenced in AGSCs but DNA hypomethylated and clearly marked with promoter HK7me (Figure A). Taken together, AGSCs (Kit + and Thy + ) share with PGCs/gonocytes (E6.5) a lack of Nanog, Lefty, and Sox expression but utilize different modes of chromatin repression than PGCs (E6.5) (Figures A C) and differ from PGCs in their lack of Prdm4 expression. Notably, for these four pluripotency/self-renewal genes, we observed either promoter or enhancer chromatin/ modifications (HK4/K7me or 5hmC) in mature sperm that may enable their rapid transition to the active state in the embryo. Also, we found that the binding sites for NANOG and SOX (and SOX/OCT4 compound sites determined in s) are DNA methylated in AGSCs (Thy + and Kit + ), whereas OCT4- only sites or KLF4 sites are DNA hypomethylated (Figure D) (Wang et al., 6). Thus, the loss of SOX and NANOG proteins in AGSCs is strongly reflected in the status of their genomic binding sites. Notably, 5hmC is enriched at 5% of NANOG-binding sites in all germ cell stages (Figure SB), which again may facilitate active removal in the embryo. We next identified active and poised germline enhancers (via generating HK4me, HK7ac, and HK7me profiles), defined their status, and then determined candidate transcription factor binding sites within them. Notably, sites for particular transcription factors were enriched at enhancers for developmental genes (ELK, ELK4, NFAT, ARIDa, and NRF; via regulatory sequence analysis tool, false discovery rate [FDR] <.; Figure SA, cluster ; Table SD). Taken together, we identified candidate enhancer regions in AGSCs and associated transcription factor binding sites. Neither the Commitment of AGSCs to Gametogenesis nor Gametogenesis Itself Involves Significant Changes in Next, we extend our analyses to gametogenesis, examining Kit + AGSCs onward, and begin with. It is somewhat surprising that gametogenesis does not involve appreciable changes in promoter : no promoters pass threshold criteria (i.e., >% change in fraction methylation), nor are there persistent changes in intergenic/enhancer linked to transcriptional changes during gametogenesis (Figure SB; data not shown; note promoter chromatin-transcription relationships of the aggregate data in Figure SC). Notably, we find the enhancers neighboring embryonic developmental transcription factors (e.g., Hox, Fox, Evx, Hand, Irx, and Lxh genes, and Nanog itself) DNA hypomethylated in all germline stages and marked with HK7me, consistent with poising (Tables SA SE; Figure SD, cluster 4). Developing Germ Cells Express a Large Repertoire of Stage-Specific RNAs Our transcriptome analyses involved examination of four main gene types: () known protein-coding RNAs (RefSeq), () long noncoding RNAs (lncrnas) in isolation, () new, nonannotated transcription units (NATs; see Experimental Procedures for characterization), and (4) known pirna clusters (Figures 4A 4D). For each gene type, we identified gene sets with peak expression (RNA accumulation) during a particular stage, using normalization and clustering programs. (Note: for mature sperm, the peak gene set represents residual RNAs, not peak transcription.) For RefSeq genes, comparisons of Kit + AGSCs to stages of gametogenesis revealed 9, differentially expressed genes, which were then clustered (Figure 4A) and subjected to gene ontology (GO) analyses (Tables S4A S4E). First, our data confirm known/predicted stage-specific transcription of particular RefSeq protein-coding genes during gametogenesis (Tables S4A S4E). Notably, for lncrnas, we found the majority of the known/annotated lncrna repertoire expressed during spermatogenesis, with striking stage specificity and interesting GO categories (FDR <.5; Figure 4B; Table S4F). Furthermore, we observed a large number of NATs (95; fragments per kilobase of exon, per million mapped reads [FPKM] >.4; Figure 4C), that peak in spermatids, in agreement with recent data (Soumillon et al., ). Of these, 7 are large intergenic and independently transcribed units (inats), validated by clear independent HK4me peaks (Figures 4C and S4A). To investigate inat coding potential, we applied CPAT, which calculates coding probability (Figure S4B; Tables S4G and S4H). For pirnas, previous work in staged postnatal mouse testes defined two phases of pirna production: prepachytene (primarily MILI bound) and Figure 4. Association of the Gametogenesis Program with Low-CG Content and DNA-Methylated Promoters (A D) RNA-seq hierarchical clustering of four gene classes (as separate panels) in s and across germ cell stages (note: AGSCs are Kit + ). For each gene class, the clusters representing peak expression (RNA accumulation) in each stage are examined for their promoter status directly below, (F)., spermatocytes;, spermatids. (E) Density scatterplots and correlations of HK4me and in s, AGSCs (Kit + ), spermatocytes, and spermatids. (x axis, fraction CG methylation; y axis HK4me, log ratio). (F) Transcribed gene promoters, defined in (A) through (D) are partitioned into Typical (lacking ) or Atypical (bearing ). (G) Histogram depicting the number and CG frequency (Observed/expected (Obs/exp)) of transcribed Typical and Atypical promoters, in the Refseq class. (H) Class average maps of the chromatin attributes of Typical and Atypical promoters (±.5 Kb) in the round spermatid, in the Refseq class. See also Figures S, S4, and S5 and Tables S4 and S6. Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. 45

8 A 5hmC HK7ac HK9ac HA.Z HK7me HK4me HK4me BisSeq RNA Pol BisSeq BisSeq Amplicon RNA Pol RNAseq (-) 5 Chr8 Typical,8,,,,,,4, Crem 5hmC HK7ac HK9ac 5 HK7me HK4me HK4me BisSeq RNA Pol BisSeq BisSeq Amplicon RNA Pol,78,,79,,8, Chr6 RNAseq (-) Atypical Socs Tnp Prm Prm 5hmC HK7ac HK9ac HAz HK7me HK4me HK4me BisSeq RNA Pol BisSeq BisSeq Amplicon RNA Pol RNAseq (+) pirna Cluster Chr8,75,,745,,755, 5hmC HK7ac HK9ac HK7me HK4me HK4me BisSeq RNA Pol BisSeq BisSeq Amplicon RNA Pol RNAseq (+) RNAseq (-) Chr8 5 pirna 67,, 67,, B RNAPol Read Counts RNAPol Read Counts 8K 6K Crem 45K 5K 4K 5K 5K K 5K CG Fraction Methylation 5 5 Typical pirna Cluster CG Fraction Methylation HK4me Read Counts HK4me Read Counts RNAPol Read Counts 4K Prm 4K K K K 8K K 6K 4K K K CG Fraction Methylation RNAPol BisSeq Reads HK4me BisSeq Reads RNAPol Read Counts 6K 4K K Atypical pirna Cluster CG Fraction Methylation 4K K HK4me Read Counts 6K HK4me Read Counts D % -cell C Negative A and B AGSC S Cytes S Tids Embryo Oocyte 8-6 ICM Blast * mir-a 6 paternally provided mirnas log(fpkm)>.5 4 hours * All Sponges % Morulas Negative A and B * 7 hours ** Log FPKM All Sponges 5 5 mir-59 mir-66 mir-5 mir-7 mir-876 mir-69 mir-6a mirna # Survived Sponges Injections % cell % Morula % Blastocyst Neg A and/or B 7 6 (88.5%) 59 (84.%) 46 (65.7%) All Sponges 9 59 (65.6%) (6.66%) 7 (%) % Blastocysts Negative A and B 96 hours ** All Sponges * ** p-value =.8 p-value =. (legend on next page) 46 Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc.

9 pachytene (primarily MIWI bound) (Li et al., ). Our AGSC pirnas overlap nearly completely (9%) with previously designated prepachytene pirnas (Figure 4D, cluster ), whereas our spermatocytes and spermatids pirnas data sets overlap nearly completely (95%) with previously designated pachytene pirnas (Figure 4D, cluster ). Thus, our data in the adult mice aligns well with postnatal mice regarding pirna phases and cell origins. Taken together, we provide deep stage-specific data sets of protein coding, lncrna, inat/intergenic units, and pirnas for future study. Common Transcription of Atypical Promoters with and HK4me in the Germline In somatic cells, is typically anticorrelated with both transcription and HK4me. Surprisingly, we found and transcription essentially uncorrelated in spermatocytes or spermatids, and relationships of with HK4me either weak or absent (Figure 4E), suggesting major differences from observations in s or somatic cells. To characterize active genes bearing, we examined the status of gene promoters (lncrnas ± 5 base pairs [bp], Refseq ± kb flanking the TSS) transcribed at each stage (Figure 4F). Remarkably, for RefSeq, lncrnas, and inats, a high fraction (% 45%) of the genes with peak transcription in spermatids bear at their TSS (Figure 4F examines the status of the peak/red clusters within Figures 4A 4D). This fraction of genes with was still quite high (% %) in spermatocytes and Kit + AGSCs, compared to the extremely low levels in s (%) (Figure 4F). Thus, we observed two very different classes of germline promoters: () the majority class, showing Typical relationships, and () a very sizeable minority with atypical relationships active, yet bearing correlated and HK4me termed hereinafter Atypical promoters. In keeping, GO analyses of Atypical promoters enrich almost exclusively categories of male gametogenesis (Tables S5A and S5B). For pirnas, we found virtually all promoters driving prepachytene/kit + AGSC pirnas unmethylated, in keeping with their intragenic derivation. In contrast, % (6/5) of unique pachytene pirna clusters are driven from promoters bearing, and these methylated promoters are almost all intergenic (8/6) (Figure 4F). Thus, resides at the promoters of independently transcribed/intergenic pachytene pirna promoters but not intragenic prepachytene/ Kit + AGSC promoters. Taken together, a marked feature of gametogenesis especially spermatid-specific genes and pirnas is the preponderance of active Atypical promoters (% 45% of total). In principle, the activity of Atypical promoters might be enabled by low CG density; mcg-dependent repression involves methyl-cg-binding domain (MBD) proteins complexed to histone deacetylases (HDACs). If Atypical promoters were to uniformly display low CG content, then they might fall below a critical MBD-HDAC threshold. It is striking that we found Atypical promoters uniformly of low CG content, whereas Typical promoters display standard CG content distributions a relationship true for each of the four gene types (Figure 4G; Figures S5A S5C). In contrast, promoters are almost all Typical, and very few display low CG content (Figure 4G; Figures S5A S5C). However, low CG content itself cannot be sufficient for expression, as these genes are silent in s and somatic cells. To define chromatin attributes that might further distinguish Typical from Atypical promoters, we profiled the HA histone variant HA.Z (typically anticorrelated with ), 5hmC (a modification that deters the binding of most repressive MBD proteins [MeCP excepted]), and both HK7ac and HK9ac (present at unmethylated active genes/enhancers) (Refseq and noncoding genes; Figure 4H; Figures S5D and S5E). Notably, HK7ac and HK9ac were enriched at both Typical and Atypical promoters (Figure 4H; Figures S5D and S5E). In contrast, HA.Z was present at high levels only at Typical promoters (Figure 4H), as well as at low levels at bivalent developmental genes (data not shown). Remarkably, in round spermatids, 5hmC was selectively present (FDR < ) at 55% of active Atypical RefSeq genes and 7% of active Atypical pirna promoters, but not at Typical promoters (i.e., Crem; Figure 4H). Remarkably, additional chromatin immunoprecipitation sequencing (ChIP-seq) profiling of RNA polymerase II (RNAPol) revealed more RNA- Pol at Atypical promoters than Typical promoters (Figure 4H; Figures S5D and S5E), with examples provided in Figures 5 and S6A. Thus, a large fraction of the gametogenesis and pachytene pirna program involves transcription from low CG content promoters that bear, HK4me, 5hmC, and high acetylation levels but unlike most active genes in somatic cells lack HA.Z. ChIP-BisSeq Verifies High HK4me and RNAPol on DNA-Methylated Promoters To directly examine coincidence of, HK4me, and RNAPol, we performed ChIP bisulfite sequencing (ChIP- BisSeq) using multiplex sequencing of candidate TSS regions of multiple genes (both Typical and Atypical) in round spermatids. It is interesting that ChIP-BisSeq with HK4me revealed at the TSS of Atypical, but not Typical, promoters (Figure 5A). To determine whether these regions are truly transcribed in the presence of or instead transcribed while DNA is demethylated, and then subsequently remethylated we performed ChIP-BisSeq with RNAPol from total germ cells and again observed underlying only Atypical promoters Figure 5. Chromatin and Relationships of Atypical Mouse Promoters and Small Noncoding RNA Repertoire (A) Representative browser snapshots of two Typical (Crem or pirna, at left) and two Atypical (protamine or pirna cluster, at right) promoters in round spermatids. Note: RNAPol BisSeq from total germ cells. Note that, for Atypical genes, the HA.Z track was omitted, as it lacked peaks. (B) RNAPol BisSeq (blue line) and HK4me BisSeq (red line) read distribution of fraction CG methylation. (C) H-clustering of mirna from all stages of spermatogenesis, mature oocytes, cleavage, and blastocyst (oocyte and embryo small RNAs obtained from Ohnishi et al. ). S Cytes, spermatocytes; S Tids, spermatids. (D) Injection of pooled mirna sponges targeting paternally provided mirnas into pronuclear mouse embryos reduces blastocyst frequency. *p =.8; **p =.. See also Figure S6. Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. 47

10 (Figure 5A). Furthermore, we examined the distribution of sequencing reads with respect to their status; here, the vast majority of the reads (>95%) bore mcg at the vast majority of the CGs within that read (Figure 5B; Figure S6B). In contrast, for Typical genes (e.g., Crem or certain pirna clusters), the vast majority of the reads were mostly hypomethylated CGs (<. mcg at >95% of reads; Figure 5B), demonstrating that Atypical genes are transcribed with a methylated TSS in spermatids. Coincidence of HK4me, RNAPol, and was also observed at both lncrnas and pirna clusters, which also show stage-specific expression and stage-specific occupancy by RNAPol. Here, ChIP-BisSeq again revealed underlying HK4me and RNAPol at the analyzed Atypical pirna promoters (Figure 5A; right panel) but not those with Typical promoters (Figure 5A, left panel). Sperm-Specific mirnas Are Associated with EMT, Proliferation, and Pluripotency We profiled mouse mirnas from AGSCs and gametogenesis stages, combined with previous data from oocyte, morula, ICM, and blastocysts (Ohnishi et al., ), yielding 5 total mirnas (log FPKM >.5) (Figure 5C; Table S4I). Clustering analyses revealed three major types: () mirnas present in male germ cells including mature sperm (expressed at varying levels) but absent in oocytes (termed misp High ; 6 total), () mirnas expressed in Kit + AGSCs and downregulated in differentiated germ cells and absent in early embryos (termed miagsc High ; 5 total; unknown function), and () mirnas provided both maternally and paternally and maintained during embryogenesis (5 total). Regarding the misp High type, our results confirm mir-4c as the most abundant mirna in mature sperm (log FPKM = ) (Liu et al., ) and show further its high expression in Kit + AGSCs and depletion in the oocyte. It is interesting that mir- 4c affects the first cleavage division (Liu et al., ). Notably, when considering the top (log FPKM > ) misp High members, most have roles in cancer, epithelial-to-mesenchymal transition (EMT), cell proliferation, or translational repression of the pluripotency network (Sox, Klf4, and Oct4 mrnas): mir- a/b, mir-a, mir-96a/b, mir-4, mir-45, mir-4, mir- 8-5p, mir-9, mir-45, mir-4, mir-96, and mir-47. To explore possible functional roles for mirnas retained in mature sperm in embryos, we selected eight mirnas (from misp High ) known to target MAPK, ERB, or TGFb signaling pathways. Here, we injected into the male pronucleus an equimolar concentration of a pool of eight mirna sponges, which resulted in a significant reduction in blastulation rate (65% versus 5%), as compared to the negative control (scrambled sponge pool; Figure 5D; p <.5). Taken together, we have defined the mir- NAs of Kit + AGSCs and gametogenic stages and provide initial preliminary data that paternally provided mirnas may affect embryogenesis. Conservation in Human Sperm of Enhancer Poising, Atypical Promoters, and Small RNAs We then tested for conservation of key concepts in mature human sperm by profiling, chromatin, and multiple RNA species (Figure 6A). Enhancer profiling revealed poised enhancers (HK4me+HK7me, but lacking H.); as in the mouse, these localize (by GO analysis) near developmental factors of importance in the embryo (Figure S7A; Tables SF and SG). Regarding chromatin poising of pluripotency network factors, clear human-mouse similarity is observed at SOX and PRDM4 (which bear bivalent promoters) and also LEFTY (with a 5hmC-enriched and a poised internal enhancer). However, human OCT4 and NANOG promoters are DNA methylated and lack a clear poised enhancer (Figure S7B) in humans. Notably, regions with HK4me slightly correlated with DNA methylation (Figure 6B), whereas HK7me strongly anticorrelated (p <.). To find and examine Atypical human promoters, we extended the MNase ChIP-BisSeq protocol with HK4me to a genomewide format (Figure 6C; note that mature human sperm is transcriptionally silent but retains HK4me at previously transcribed promoters). Notably, read distributions (Figure 6C) reveal two types of loci: () those bearing high HK4me and (Atypical), which map almost solely to low CG-content promoters (Figure 6D), and () loci with high HK4me lacking (Typical), which map largely to intermediate-high CG-content promoters (Figure 6D). Promoters of genes with high HK4me and are genes highly transcribed during gametogenesis including protamine genes, transition protein genes, and 5% of pirna clusters (for pirnas, promoters were defined by coincident RNA-seq read starts and HK4me peaks) (Figure 6E). In contrast, Typical genes (e.g., Crem) lack underlying their promoter HK4me peaks (Figure 6E). Notably, we observed a 45% overlap between conserved Atypical promoters identified in the mouse and those in humans (Tables S5C and S5D). Taken together, the data are consistent with Atypical promoter utilization as a conserved property of mouse and human gametogenesis. RNA Profiles in Human Sperm Regarding RNAs in mature sperm, we performed RNA-seq on two donors (D and D), deriving separately long RNA (> bases) and small RNA (< bases). As data sets from D and D were highly correlated (r =.85, p <.), they were combined (> million reads). Long RNA was of low abundance, and those detected were those RNAs most highly transcribed during spermatogenesis or spermiogenesis and are therefore likely residual (Table S6A). For mirnas, we found a large number of mature mirnas (5; FPKM > ) shared between the two donors (8% overlap), greatly increasing previous estimates (5 mirnas), with complete agreement/overlap (Figure S7B; Tables S6B and S6C). It is interesting that the vast majority (85%) of mirnas common between D and D in mature sperm overlapped with those in ES cells (H cells; Figure S7C; Table S6C). Furthermore, of the top misp High members in mice, 5 are within the top mirnas in human sperm. Furthermore, for those mirnas with shared mouse/ human nomenclature (8), we observed a 59% overlap (9/ 8) of mirnas found in mouse and human mature sperm (Figure S7D; Table S6D). Finally, we reveal a large number (9,5; FPKM > ) of pirnas shared between D and D data sets (79% overlap; Figure S7E; Table S6E), although with almost no overlap with previous data sets (Krawetz et al., ). Thus, we greatly increase the known repertoire of retained mirnas and pirnas in sperm. 48 Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc.

11 A Human Donor Human Donor D+D small RNAs long RNAs HK4me HK7ac H. HK4ac HK4me BisSeq B HK4me Log Ratio HK4me Log Ratio r= r= Fraction Methylation 7 6 r= D + D Fraction Methylation C HK4me-BisSeq Read Counts 6e+5 4e+5 e+5 HK4me BisSeq e D + D BisSeq Fraction CG Methylation CG Obs/Exp D Counts 9 6 Typical; HK4me HypoMeth DNA Atypical; HK4me HyperMeth DNA E Typical Atypical HK4me HK4me BisSeq RNAseq (+) Chr CREM 5,4, 5,48, 5,54, HK4me HK4me BisSeq RNAseq (-) Chr6 75,7,,8, TNP PRM PRM PRM HK4me HK4me BisSeq RNAseq (+) 5 pirna cluster HK4me HK4me BisSeq RNAseq (+) 5 pirna cluster Chr 45,6, 45,68, 45,74, Chr5 9,9, 9,94, 9,96, RNAseq (-) 5 RNAseq (-) 5 Figure 6. Co-Occupancy of HK4me and at Low-CG Content Regions in Human Sperm (A) Graphical summary of human sperm data sets generated (gray boxes). (B) Density scatterplots and correlations of HK4me and DNA methylation in (top) and mature sperm (bottom). (x axis, fraction CG methylation, range, ; y axis, HK4me log ratio). (C) The distribution of fraction CG methylation for the reads from the HK4me BisSeq data set. The reads were partitioned into DNA hypomethylated (blue bars) or hypermethylated (red bars) for use in (D). (D) CG frequency (Obs/Exp, observed/expected) of the DNA hypomethylated or hypermethylated BisSeq reads from (C). (E) Browser snapshots of representative Typical and Atypical promoters in human sperm. Depicted are (D+D, fraction CG methylation), D+D HK4me ChIP-seq (Qval FDR), and HK4me-BisSeq (fraction CG methylation). Arrows in the pirna clusters are used to depict the predicted TSS of the transcribed cluster, based on HK4me peaks. See also Figures S6 and S7. DISCUSSION Our work uses extensive genomic profiling to better understand AGSC self-renewal, commitment to gametogenesis, gametogenesis stages, and germline genome pluripotency. Foremost, we aimed to understand how chromatin packaging and transcription networks help regulate these processes and poise/prepare genes for each subsequent stage. Our work provides new insights into AGSC biology and reveals several remarkable features of chromatin utilization during gametogenesis. Distinctive Signaling and Transcription Pathway Signatures Define Thy + versus Kit + AGSCs First, regarding self renewal, our transcriptional profiling reveals clear differences between the Thy + versus Kit + AGSCs, including the selective presence or absence of particular signaling factors, tumor suppressors, proto-oncogenes, and transcription factors. It is interesting that only Thy + cells express Csfr, Ilr, and Tnfr, providing receptors for a set of cytokines secreted by cells (macrophages, monocytes, or dendritic cells) residing in the interstitium adjacent to the wall of the seminiferous tubule but absent in Kit + cells (which migrate centrally). Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc. 49

12 AGSCs Spermatocytes PGCs Thy + --> Stem cell markers and RA synthesis Demethylation Enhancer dynamics Methylation Distinctive Chromatin poising Transcription pre-pachytene meiotic gene demethylation chromatin / Bivalency in gametogenesis dynamics Widespread NATs pachytene pirnas Pluripotency network Lefty Sox Nanog Prdm4 Klf4 Oct4??? No changes / intergenic transcription transition Histone-protamine Round spermatids Elongated spermatids Chromatin poising/bivalency at enhancers and promoters of genes utilized in embryo development HK4me + HK7me Transcription in gametogenesis HK4me + HK9ac + HK7ac HA.Z Pol HK4me + HK9ac + HK7ac Pol (eg. Hox, Sox, Fox, Tbx, Gata-family TFs) Typical promoters - High CG (eg. Crem) Typical promoters - Low CG Atypical promoters - Low CG (eg. pirnas) ICM ES cells sperm = = 5hmC Zygote Figure 7. Summary Schematic of Transcription and Chromatin Transitions within the Germline Cycle See Discussion for details. Furthermore, we reveal an interesting example of cytokine/ receptor asymmetry (Pdgf in Thy +, Pdgfra in Kit + ) that raises the possibility of signaling between different AGSC cell types. RAS signaling promotes AGSC/SSC proliferation in vitro, and we find H-Ras highly expressed in both Thy + and Kit + AGSCs in vivo. Notably, Harvey rat sarcoma viral oncogene homolog (H-RAS) pathway activation elicits SSC self-renewal through transcriptional regulation of the cyclins Ccne or Ccnd. It is interesting that we found Ccne expressed only in Thy + cells and Ccnd expressed only in Kit + cells. Thus, RAS signaling appears involved in both Thy + and Kit + AGSCs, but certain regulators and targets are different. Furthermore, transplantation of cultured SSCs bearing constitutively active H-RAS into the testis gives rise to germ cell tumors, emphasizing the importance of RAS regulation in vivo (Lei et al., ). Notably, we find Dab (a known modulator of the RAS pathway in other cell types) expression in Thy + cells, which raises the question of whether DAB may be modulating RAS signaling and preventing uncontrolled proliferation of AGSC. In D. melanogaster, SSC selfrenewal is also regulated by NFkB, and only in Thy + cell do we observe high Aim, which controls cell proliferation by suppressing NFkB. Our data sets also verify the known involvement of retinoic acid in the Thy + -to-kit + transition and here reveal two key promoters (Aldh and Stra8) as bivalent in Thy + cells but only HK4me (and transcribed) in Kit + AGSCs. Notably, Aldh and Stra8 provide rare examples in our data sets of a bivalent gene losing HK7me and then becoming active. Beyond these rare exceptions, germline developmental transitions almost never involve gene activation by resolving bivalency (removing HK7me), whereas HK7me removal commonly accompanies gene activation during differentiation. Overall, our results reveal new signaling pathways, factors, and processes for study in AGSCs and their niche and may inform their proper culturing in vitro. AGSCs Poise the Factors Needed for Self-Renewal and Embryo Development We found the expression of pluripotency factors in AGSCs more similar to PGCs than s. For example, AGSCs and PGCs (E6.5) lack expression of Nanog, Sox, and Lefty (and, additionally, Prdm4 in AGSCs) but may utilize different repression mechanisms; Nanog and Lefty promoters have high in AGSCs but not in E6.5 PGCs (Seisenberger et al., ) (Figure 7). Remarkably, in mature sperm (human and mouse), each of these genes possesses (or maintains from AGSCs) bivalency and/or 5hmC at their TSS or proximal enhancer, providing alternative modes of poising for their rapid activation in the embryo. Notably, these candidate poising features appear conserved in human sperm at SOX, PRDM4, and LEFTY but not at OCT4 and NANOG; their methylation in sperm necessitates prompt demethylation in the embryo, a process that might be facilitated by the slightly later maternal-zygotic transition in humans than mice. AGSCs appear to blend the use of factors present in PGCs with those typically associated with adult stem cells. For example, AGSCs resemble other adult stem cells in silencing Nanog and Lefty yet expressing Myc, Aldh, Lrig, Olfm, and Bmi (Sangiorgi and Capecchi, 8). Also, AGSCs share with neural stem cells the GFRA/GDNF signaling axis (a pathway critical for AGSC self-renewal) and also Foxo expression (Lei et al., ; Ro et al., ). Furthermore, prior genetic work has implicated PLZF (Zinc Finger and BTB Domain Containing 6) and ETV5 (Ets Variant 5) in AGSC self-renewal (Costoya et al., 4; Dovere et al., ; Hobbs et al., ), results supported here by high transcript levels in AGSCs. Thus, a unique blend of these factors help AGSCs to poise the like pluripotency program for future utilization in the embryo while simultaneously enforcing a unipotent developmental program (gametogenesis) within the testicular niche. Notably, 5 Cell Stem Cell 5, 9 5, August 7, 4 ª4 Elsevier Inc.