sign in sign up
<<
Home , ACIN1, CALM3, FOXD3, KLF12, KLF15, MAPK phosphatase

Oncogene (2008) 27, 7180–7191 & 2008 Macmillan Publishers Limited All rights reserved 0950-9232/08 $32.00 www.nature.com/onc ORIGINAL ARTICLE Androgen-induced programs for prostate epithelial growth and invasion arise in embryogenesis and are reactivated in cancer

EM Schaeffer1,2,3,5, L Marchionni3,5, Z Huang1,2, B Simons1, A Blackman3,WYu3, G Parmigiani1,3,4 and DM Berman1,2,3

1Department of Pathology, Johns Hopkins University School of Medicine, Baltimore, MD, USA; 2Department of Urology, Johns Hopkins University School of Medicine, Baltimore, MD, USA; 3Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, MD, USA and 4Department of Biostatistics, Johns Hopkins University School of Medicine, Baltimore, MD, USA

Cancer cells differentiate along specific lineages that most significant links to the development and cancer, we largelydetermine their clinical and biologic behavior. highlight coordinate induction of the factor Distinct cancer phenotypes from different cells and organs Sox9 and suppression of the proapoptotic phospholipid- likelyresult from unique expression repertoires binding A1 that link earlyprostate established in the embryo and maintained after malignant development to earlyprostate carcinogenesis. These transformation. We used comprehensive results credential earlyprostate development as a reliable analysis to examine this concept in the prostate, an organ and valid model system for the investigation of and with a tractable developmental program and a high pathways that drive . propensityfor cancer. We focused on gene expression in Oncogene (2008) 27, 7180–7191; doi:10.1038/onc.2008.327; the murine prostate rudiment at three time points during published online 15 September 2008 the first 48 h of exposure to androgen, which initiates proliferation and invasion of prostate epithelial buds into Keywords: prostate organogenesis; androgen signaling; surrounding urogenital sinus mesenchyme. Here, we show prostate cancer; microarray that androgen exposure regulates genes previouslyim- plicated in prostate carcinogenesis comprising pathways for the and tensin homolog (PTEN), fibroblast (FGF)/-activated (MAPK), and Wnt signaling along with cellular Introduction programs regulating such ‘hallmarks’ of cancer as angiogenesis, , migration and proliferation. We The discovery of proto-oncogenes in the 1970’s (Stehelin found statisticallysignificant evidence for novel androgen- et al., 1976) and tumor suppressor genes a decade later induced gene regulation events that establish and/or (Friend et al., 1986) launched an effort to identify a maintain prostate cell fate. These include modulation of common genetic basis for all cancers. This approach has gene expression through microRNAs, expression of yielded significant insights into common molecular specific transcription factors, and regulation of their machinery regulating tumor initiation and growth predicted targets. Byquerying public gene expression (Hanahan and Weinberg, 2000), but fails to account databases from other tissues, we found that rather than for dramatic differences in the behaviors of tumors generallycharacterizing androgen exposure or epithelial arising in different sites. As revealed by recent genomic budding, the earlyprostate development program more approaches, the site of origin is the dominant influence closelyresembles the program for prostate cancer. on gene expression in cancers (Ramaswamy et al., 2001) Most importantly, early androgen-regulated genes and and specifies particular oncogenic mutations (Garraway functional themes associated with prostate development and Sellers, 2006). Tissue-specific behaviors of cancers were highlyenriched in contrasts between increasingly almost certainly reflect lineage-specific epigenetic pro- lethal forms of prostate cancer, confirming a ‘reactivation’ grams that operate during embryogenesis in cells and of embryonic pathways for proliferation and invasion in tissues from which cancers arise. Indeed, reawakening of prostate cancer progression. Among the genes with the embryonic programs has long been posited for cancer (Bailey and Cushing, 1925), and could underlie tissue Correspondence: Associate Professor DM Berman, Department of specific modes of regulating critical aspects of the Pathology, Johns Hopkins University School of Medicine and Sidney malignant phenotype such as survival, angiogenesis, Kimmel Cancer Center, CRB2 Room 545, 1550 Orleans street, invasion and migration. The advent of comprehensive Baltimore, MD, 21231 USA. genomic profiling techniques now permits this hypo- E-mail: [email protected] thesis to be tested, and to link cancer to embryogenesis 5These authors contributed equally to this work. Received 31 March 2008; revised 18 July 2008; accepted 1 August 2008; through cellular pathways that define specific lineages published online 15 September 2008 and organs (Garraway and Sellers, 2006). Identification Embryonic gene expression in prostate cancer EM Schaeffer et al 7181 of these pathways and their functions should identify urogenital sinus (UGS), an embryonic rudiment present targets for more specific and less toxic cancer therapies. in both sexes. These interactions lead to outgrowth of The prostate gland represents a prime target for such buds from the UGS epithelium (UGE) that proliferate an analysis as the genetic causes of prostate cancer are and invade surrounding urogenital sinus mesenchyme. poorly understood and because there are excellent AR expression has been described in both UGE and models of prostate development and . While urogenital sinus mesenchyme (Drews et al., 2001; B other cancers have canonical cancer initiating mutations Simons, EM Schaeffer and DM Berman, unpublished (for example, K-ras for pancreatic carcinoma and observations), although epithelial AR is dispensable for Adenomatosis Polyposis Coli for colon carcinoma), induction of prostate development (Cunha et al., 2004). the driving force for prostate cancer initiation in remains uncertain. Answers may lie in cellular programs activated by androgen (AR) signal- Results ing, which strictly controls prostate epithelial cell fate in embryogenesis and in cancer. Hormonal manipulation of mouse prostate development Upon binding androgen, AR signals through genomic In mouse, epithelial expression of the androgen-sensitive and nongenomic modes (Manolagas et al., 2002). family member Nkx3. 1 begins by embryonic Genomic responses entail nuclear translocation of day 16 (e16), and is the earliest reported marker of liganded AR and activation of transcription at regula- prostate development (Bhatia-Gaur et al., 1999). We tory regions containing AR-binding sites. Nongenomic comprehensively profiled gene expression in the UGS signaling, in contrast, comprises protein–protein inter- between e16 and e17.5, when prostate buds first emerge. actions in the cytoplasm and is exceedingly rapid, with The developmental fate of the UGS is bipotential in measurable responses within minutes of androgen both sexes and depends soley on androgens to drive exposure (Kousteni et al., 2001). AR signaling, through prostate formation in males (Jost, 1968). The absence of genomic and/or nongenomic routes, is necessary and androgens leads to vaginal/urethral formation in sufficient for prostate organogenesis. In response to females (Wilson et al., 1980). Taking advantage of this circulating and its local conversion to the dynamic, we induced prostate development in the more potent androgen, dihydrotestosterone, AR signal- androgen-naı¨ ve yet androgen-sensitive female UGS with ing induces epithelial-mesenchymal interactions in the precisely timed 6 and 12 h intrauterine exposure

15.5–16 d.p.c. 17.5 d.p.c. Mouse public domain Molecular Histologic Prostate degeneration/regeneration (GSE5901; Wang et al., 2007) Mesenchyme Epithelium Lung development (GSE1423; Lu et al., 2004)

UGS Androgen-dependent gene expression in salivary gland (GSE3995: Treister et al., 2006) Androgen 48 h Endogenous (testis) Human public domain data Macrodissected prostate cancer profiling (Lapointe et al., 2004) Exogenous (injection) 12 h

Androgen 6 h LCM prostate cancer profiling (GSE6099: Tomlins et al., 2007) Androgen start

Differential gene expression analysis * Gene expression measurements: preprocessing, quality control evaluation * Statistical analysis of gene expression (Linear model analysis and empirical Bayes approach) * Gene cross-referencing with gene and homologene identifiers for cross platform and organismal analysis

Analysis of functional annotation (AFA) * Enrichment evaluation using: (GO), KEGG pathways and TFBS targets in prostate development and human cancer progression by Wilcoxon rank-sum test * Enrichment evaluation for mouse expression signatures in human prostate tumors by Wilcoxon rank-sum test

Figure 1 Flowchart of data acquisition and analysis. (a) Schematic of early prostate development. The embryonic prostate rudiment, the urogenital sinus (UGS). Mesenchyme (light blue) surrounds epithelium (darker green). In the mouse, prostate-specific gene expression begins by embryonic day 16 (e16) followed by prostate epithelial budding at e17.5. Prostate development proceeds spontaneously in males in response to endogenous androgens or can be engineered in females in response to exogenous androgens. We comprehensively profiled androgen-induced gene expression changes in pharmacologically virilized female UGS at 6 and 12 h after injection with a potent androgen (dihidrotestosterone, 50 mg/kg) and in physiologic prostate development at e17.5, B48 h after the onset of androgen-induced transcriptional changes. (b) List of data sets from the public domain used in integrative analysis. (c) Linear models and Bayesian approaches were used to identify differentially expressed genes. (d) Significantly enriched themes were identified through functional annotation enrichment analysis (see Supplementary methods for analytic protocols). See online version for color figure. LCM, laser capture microdissection.

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7182 to pharmacological levels of dihydrotestosterone Nkx3.1 was highly induced (adj. Po10À4) among the (Figure 1a). For the latest time point, approximating transcripts confirmed by reverse transcriptase–PCR. We 48 h of physiologic androgen exposure, we compared further validated these results with separate biologic UGS gene expression in unmanipulated male and female replicates and with RNA from embryos treated with e17.5 littermates. In all cases, we isolated high quality pharmacologic AR blockade (Supplementary Figure 1). RNA, prepared labeled cRNA pools and competitively Using the 48 h time point as a standard, we performed hybridized androgen-exposed pools against androgen- ‘correspondence at the top (CAT)’ analysis (Irizarry naı¨ ve pools on Agilent 44 000 probe whole genome et al., 2005) to compare identities of top ranking (most mouse microarrays. High quality expression data was differentially expressed) genes in lists from different time confirmed as described in Supplementary (S) methods. points. CAT analysis showed that both pharmacologic and physiologic androgen exposure induced highly Distinct and prostate-specific phases of gene expression in related gene expression programs in the UGS, early prostate development (Figure 2d). This was particularly true when comparing The genomic response to androgen varied dramatically gene expression at 12 h and 48 h; gene lists from these with the duration of exposure to the hormone. After 6, two time points showed 40% identity for top ranking 12 and 48 h of androgen exposure, respectively, 693, 177 genes. By connecting physiologic and pharmacologic and 829 genes were differentially expressed (adj. androgen responses in the UGS, these results validated Po0.03) (Figure 2a). Complete lists are provided in our experimental approach. the Supplementary material (Supplementary Tables 1– Several genes were consistently androgen regulated; 3). Pairwise comparisons between any two time points either induced (Aspn, Klf9, Synpr, Gadd45 g, Sox9, revealed an overlap between 10–42% of differentially Adamdec1 and Tle1) or suppressed (Prrx1 and Inhba) at expressed genes. Despite some diversity in the most all three time points (Figure 2b). This cohort may highly differentially expressed genes across different represent particularly important genes in establishing time points, we successfully validated the androgen- and maintaining prostate lineage. mediated developmental program using a variety of methods. Starting with the latest time point, reverse The genomic response to androgen is dynamic transcriptase–PCR analysis of 20 transcripts confirmed The overall pattern of genomic response to androgen differential expression that was concordant with the was suppression, mild induction and robust induction of array results (Supplementary Figure 1). As expected, gene expression at 6, 12, and 48 h, respectively

0.6 12v48 h Induced 12 h Aspn 6 h 0.4 ] Pros. vs Pros. 11 Klf9 606 ] Pros. vs lung 90 Synpr 10 Gadd45g 0.2 66 66 Sox9 Adamdec1

Tle1 Proportion in Common 0 687 0 1000 2000 Repressed Number of genes compared Prrx1 48 h Inhba 0.6 12v48 h Ratio suppressed/induced genes 0.4 ] Pros. vs Pros. 6 h 74%

12 h 31% ] Pros. vs salivary 0.2 48 h 46%

050100% Proportion in common 0 Suppressed 01000 2000 Induced Number of genes compared Figure 2 Androgen-induced gene expression in early prostate development is dynamic and organ specific. (a) Distinct and overlapping genomic responses to androgen at 6, 12 or 48 h of exposure (see text). Values represent differential expression at adj. Po0.05. (b) Nine genes showed concordant up- or downregulation at all time points. (c) Chart shows ratio of differentially expressed genes either suppressed (black) or induced (gray) by androgen at indicated time point. (d and e) Similarity of gene lists was determined by pairwise correspondence at the top plot analyses of statistically top ranking genes in (d) branching morphogenesis of prostate compared to lung (Lu et al., 2004) or (e) adult salivary gland (Treister et al., 2005).Y-axis represents the proportion of identical genes between two array sets, whereas X-axis represents the number of genes compared. Note there is a correspondence between all prostate comparisons (Pros vs Pros) with particularly high concordance (arrow) between pharmacologically regulated genes (12 h) and physiologically regulated genes (48 h).

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7183 (Figure 2c). To better understand this pattern of Discrete and overlapping features of prostate development regulation, we analyzed functional annotations of and prostate homeostasis differentially expressed genes (Supplementary Methods). Prostate homeostasis in adults can be studied by At 6 h, an overwhelming majority (76%) were sup- hormonal manipulation of the mature gland and pressed (Figure 2c), and most (58%) of the suppressed consists of three phases: active cell death upon androgen genes contained one or more predicted binding site for withdrawal (regression), castrate steady state and miRNAs, far exceeding the number expected by chance androgen induced growth (regeneration; English et al., (Po1 Â 10À5; w2-test). Considering previously described 1987). Cross platform analysis of the 1536 and 2197 roles for miRNAs in , and genes differentially expressed (adj. Po0.001) in devel- specific functions in targeting complementary tran- opment or regeneration (Wang et al., 2007b), respec- scripts for destruction (Stefani and Slack, 2008), our tively, revealed only a limited number of differentially results suggest a mechanism for rapid gene suppression expressed genes enriched in both processes (333 genes, in response to androgen. After longer exposure to 9%) (Supplementary Figure 2). Annotation analysis androgen, the majority of genes were induced, with 71 reported previously for prostate regeneration noted (69%) and 403 (54%) genes showing increased expres- significant enrichment of adhesion, proliferation and sion at 12 and 48 h, respectively. metabolic themes (Wang et al., 2007b). We hypothesized At equal statistical stringency (log odd>1.1, adjusted that prostate organogenesis would involve a wider Po0.03), we found more differentially expressed genes variety of themes that govern additional processes, such at the 6 and 48 h time points than at 12 h. A portion of as lineage differentiation, extracellular matrix remodel- this effect stems from additional biologic replicates ing and migration. available for 6 and 48 h (Supplementary Methods). However, biological factors likely explain most of the variability in numbers of differentially expressed genes. Androgen-activated embryonic gene expression invokes These include expression levels passing through neutral the ‘hallmarks’of cancer as they change from repressed at 6 h to induced at 48 h, By analysis of functional annotation, we identified and an amplification effect of significant themes in early prostate development expression. Accordingly, 50% of genes with enrichment (Table 1, Supplementary Tables 4–12). A self renewing at both 6 and 48 h were repressed at 6 h and induced at population of stem cells has recently been described in 48 h. Several transcription factors were induced at 12 h the adult murine prostate (Lawson et al., 2007), and and their predicted target genes were differentially stem cell-associated genes were enriched (Po0.05) expressed at 48 h (Supplementary Tables 2, 12). Exam- among androgen response genes (Supplementary Table ples include induction of Sox9 followed by differential 19). Sca1 expression, which in part characterize stem expression of 33 of its predicted targets (Supplementary cells in the adult prostate (Xin et al., 2005) was highly Table 12), including fibroblast growth factor R3 induced at the 48 h time point and may signal the (FgfR3), Hoxb9 and Tle1 (see below). Altogether, these emergence of an androgen-regulated progenitor popula- data indicate that androgen exposure reorganizes the tion at this stage. All three time points exhibited genomic repertoire of the UGS toward prostatic growth enrichment (adj. Po0.001) of AR targets as predicted and differentiation–first by suppressing a bipotential by transcription factor-binding site analysis. We also gene expression program, and then by inducing tran- identified significant (adj. Po0.05) enrichment of genes scription factors that initiate and carry out prostate- associated with ‘hallmark’ characteristic activities of specific gene expression. cancer (Hanahan and Weinberg, 2000) including angio- genesis, apoptosis, migration, motility and proliferation (Table 1). Similarly, and despite the dominant role of Prostate-specific responses to androgen in the embryo post-translational modifications in their regulation, To further define the specificity of expression programs in several extracellular and intracellular signaling path- embryonic prostate, we used publicly available gene ways showed significant androgen-induced responses. expression data to compare early prostate development These included oncogenic pathways, such as FGF/ to embryonic and androgen responsive gene expression in mitogen-activated protein kinase (MAPK), phosphatase other tissues. As the prostate utilizes programs of and tensin (PTEN)-PI3K-mTOR (mammalian target branching morphogenesis during development, we em- of rapamaycin) and Wnt (Table 1). These findings ployed CAT analysis to compare top ranking genes in indicate that cancers may reactivate cellular programs prostate development with top ranking genes associated initiated in development, a concept we test more with embryonic branching in the lung (Lu et al., 2004). formally below. Similarly, we compared gene expression in the developing We also captured evidence of transcription factor prostate and another androgen-responsive organ, the adult networks acting downstream of the AR. Transcription salivary gland (Treister et al., 2005). We observed distinct factors including AR, Foxf2 and Sox9 were induced clustering of top ranking genes by organ (Figures 2d and within 12 h of androgen exposure and associated with e). Rather than generic programs for branching morpho- coordinate expression of their predicted target genes genesis and androgen response, these analyses highlight a (Table 2, Supplementary Tables 11,12). This observa- distinct developmentally and hormonally restricted lineage tion not only identifies new androgen-regulated acti- program governing prostate development. vities of these transcription factors, but also cross

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7184 Table 1 Functional annotation enrichment during early androgen induced prostate development Androgen exposure 6 h 12 h 48 h

Induced Supressed Induced Suppressed Induced Suppressed

GO metaclass Adhesion, Adamts2 Aspn Timp2 Anxa9 Aspn Bcl6 Sox9 Col8a1 Spon2 Aspn Mmp7 Mmp16 Mmp3 cell–cell Sparcl1 Calca Cldn7 Robo1 Stat5a Vcam1 Calca Pcdh8 Smoc1 Fgfr3 Pten Robo2 cell–ECM Egfr Vcam1 Sox9 Tgfbi Sox9 Sparcl1 interactions Angiogenesis Figf Pten Bmp4 Id1 Notch1 Fgf10 Fgfr2 Shh Cxcl12 Cxcr4 Fgfr2 Id1 Hif1a Pten Runx1 Nr2f Rhob Pdgfa Shh Smad5 Sox17 Apoptosis Casp12 Casp4 Bcl2l2 Inhba Kras Bcl6 Gadd45 g Inhba Mtch1 Bcl2 Gadd45 g Dcc Egln3 Gas1 Ebag9 Pdcd10 Notch1 Rhob Sox9 Stat5a Trp63 Sgk Sox9 Stat5a Inhba AnnexinA1 Rnf7 Cellular motility, Egfr Pten Sdcbp Ablim1 Btg1 Cfl1 Bcl6 Lama5 Shh Cxcl12 Stmn1 Etv4 Nck2 Cxcr4 Dcc Epha7 Ephb1 migration Epha7 Nav1 Rras2 Shh Ephb2 Nrp2 Development Adamts2 Fgf18 Notch1 Pfn1 Zeb2 Fgf10 Shh Fgfr2 Cxcl12 Cxcr4 Foxa1 Robo2 Slit2 Mycn and differentiation Bmp4 Zeb2 Lama5 Lama5 Shh Wnt4 Fgf18 Pthlh Proliferation/ Ebag9 Htra1 Btg1 Crim1 Myocd Alox12 Esm1 Ar Esr1 Rerg Sertad2 Sertad3 Gas6 Igfbp2 Igfbp6 Socs1 Tgfb3 Epigenetics Tle1 Mbd1 Dnmt3a Ncor1 Bcl6 Ncoa1 Tle1 and chromatin Dnmt1 Hist1h1b Trp63 Bnip3 remodeling

KEGG Steroid Hmgcr metabolism TGF-b-signaling Rps6kb1 Smurf2 Bmp4 Crebbp Id1 Fst Id1 Id2 Id3 Amhr2 Inhba pathway Id2 Id3 Inhba Tfdp1 Smad5 Tgfb3 Tgfbr2 MAPK signaling Fgf10 Fgfr2 Stmn1 Fgf12 Fgfr2 Fgfr3 Dusp2 Dusp4 Dusp7 pathway Gadd45 g Map3k6 Mapk13 Pdgfa Fgf18 Jund1 Pla2g4a Phosphatidyl- Pik3r3 Prkcb1 Calm3 Pip5k3 Plcb1 Calm1 Plcd1 Pik3r2 Pip5k2b inositol signaling Pten Prkcb1 Pten system Cell cycle Orc6 l Cdc27 Cdk2 Cdk4 Gadd45g Gadd45g Tgfb3 Gadd45 g Trp53 mTOR signaling Figf Pik3r3 Rheb Eif4b Pten* Hif1a Igf1 pathway Rps6kb1 Pten* Pik3r2 Wnt signaling Dvl1 Prkcb1Tle1 Ctbp1 Lef1 Trp53 Fzd7 Plcb1 Sfrp2 Sfrp1 Wnt5b Mmp7 Sfrp2 Wif1 Dkk2 Sox17 pathway Vangl2 Wnt6 Wif1 Tle1 Rspo1 Wnt4 Tle1 Rspo1Wnt7a Wnt9a VEGF signaling Pik3r3 Prkcb1 Kras Pla2g4a

TFBS FREAC2 Bcl9 Inhba Lmo4 Efemp1 Inhba Nsg2 Tcf15 Pdgfc Fgf12 Dusp4 Grem1 Inhba Klf3 Nfix Tiam1 Pdgfc Tle1 Trim2 Hhip Hoxa10 Klf12 Socs1 GATA3 Nrgn Syncrip Bcl2l1 Bmp4 Efemp1 Filip1 Amhr2 Elavl4 Edn1 Foxa1 Penk1 Amhr2 Esr1 Igf1 Hoxa10 Prrx1 Syn1 Penk1 Trp63 Prrx1 Pias1 Xpr1 Prrx1 Sh3kbp1 NKX3.1 Nedd4 Notch1 Efemp1 Loxl2 Nnat Efemp1 Kit Ptn Cntnap Prox1 Rnf111 Tlk1 Tnrc15 Trp63 Rnf128 Tacstd2 Slitrk1 Smad5 Wnt7a SOX9 Lhx6 Tle1 Hoxa10 Inhba Fgfr2 Maob Basp1 Elavl4 Dusp6 Fgfr2 Fgfr3 Inhba Meis1 Nfix Nav1 Sox4 Ets1 Tle1Fgf10 Trim2 Olfm1 Inhba Foxa1 Hoxa10 Cdca7 Tcf4 Sfrp1 Androgen Atrnl1 Rheb Inhba Lrp1 Epha7 Bcl6 Klf15 Lrrk1 Amph Olfm1 Id1 Id2 Sgk Smox Wnt7a Inhba receptor Syncrip Gsr Id1 Shh Smoc1 Inhba Gap43 Foxa1 Map4k5 Wnt9a Nnat Epha7 LEF1 Rnf11Arrb2 Fap Nav1 Dnmt3a Pla2g4a Bcl6 Prrx Crmp1 Pla2g4a Fgfr3 Shh Wnt7a Wnt9a Jund1 Gbe1 Ktn1 Dusp7 Gata2 Fgf10 Trp63 Elavl3 Elavl4 Foxa1 Tmprss2 Dusp4 Ephb2 Notch1 Foxf2 Hmga2

Table includes gene ontology (GO), Kegg pathways and consensus transcription factor-binding sites (TFBS) for predicted promoter regions of genes regulated by androgen in the developing prostate. Selected genes driving each enrichment are listed here. GO metaclasses are made up of several enriched categories listed in Supplementary Tables 4–12, along with a complete list of differentially expressed genes responsible for enrichment. Bold type indicates genes discussed in the text. *PTEN—asterisk indicates manual addition to gene set. PTEN has been shown to be a component of the mTOR signaling pathway, but is not yet annotated as such in GO terms (Yeager et al., 2008). validates the effects of androgens across different time androgen response in prostate development and points in our analysis. Altogether, this work, including highlights androgen-mediated programs for regulated detailed annotated gene lists (see Supplementary invasion, proliferation, survival and other cellular material), provides the first comprehensive map of the behaviors that become aberrantly regulated in cancer.

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7185 Table 2 Embryonic genes enriched in cancer transitions GO category Gene list

Programmed cell death Serinc3 Dapk1Acin1Rybp Cadm1Gas1Eif2ak2 Sgk Sox9 Trp53bp2 Bclaf1, Transcription Cited2 Drap1 Elf4 Ell2 Ncoa6 Rybp Wwtr1 Rbpsuh Jun Smad5 Mtf1 Sox9 Tcf7l2 Nrip1 Tcf3 Tbpl1 Angiogenesis Canx Cited2 Pdpn Tissue and Postn Col3a1 Col12a1 Dag1 Dpt Fbln1 Fbn2 Lgals3 Lum Mfap4 Mmp16 Ogn Bgn Sparc Timp3 Adamts1 extracellular Dag1Grem1Mgp Wwtr1 matrix remodeling Cell Adhesion Efs Postn Col12a1 Dpt Alcam Cadm1 Itha3 Otgav Mfap4 Neo1 Sympk Nrp2 Cd34 Cell movement Pdpn Akt2 Cell differentiation Acin1 Wwtr1 Id4 Notch3 Mapk1 Wnt Wif1 Csnk1a1 Sfrp1 Tcf7l2 Tle1Tle3 Tcf3 BMPsignaling Smad5 Serine/ Cited2 Smad5 kinase signaling Epha5 Ephb1 Ephb2 Ntrk2 signaling

Table lists enriched (adj. Po0.05) Gene ontology (GO) functional annotation categories and corresponding genes differentially expressed (adj. Po0.01) in both prostate development (*>* 1 time point) and prostate cancer progression (*>* 1 contrast). For a complete list of genes enriched in development and cancer, see Supplementary Tables 13–18.

Signaling in prostate development: FGF/MAPK ized. PTEN function suppresses a downstream signaling Secreted fibroblast growth factor (FGF) ligands signal cascade whose effectors PI3 K, AKT (a proto-onco- through their receptors to activate intracellular path- gene), mTOR and (S6K) promote a ways controlling proliferation. FGFs have multiple wide variety of cellular functions, including growth, known roles in development and constitutive activation survival, migration and angiogenesis. PTEN loss is of the pathway in mouse prostate leads to cancer common in advanced prostate cancer (McMenamin (Acevedo et al., 2007; Memarzadeh et al., 2007). In et al., 1999; Wang et al., 2003) and PTEN loss is development, Fgf7 and Fgf10 have been intensively oncogenic in the mouse prostate, perhaps due to its studied as candidate paracrine signals that facilitate promotion of epithelial cell growth and survival in the epithelial budding in response to androgen (Donjacour face of androgen withdrawal (Shen and Abate-Shen, et al., 2003). However, despite dramatic effects of these 2007). Intriguingly, androgen rapidly induced several peptides on prostate ductal outgrowth, Fgf7 and Fgf10 components of this pathway, including Pik3r, a PI3K demonstrate largely equivalent expression patterns in regulatory subunit Rheb, a small GTPase Ras homolog male and female UGS tissue. Thus, other factors must required for mTOR activity (Long et al., 2005), and be required to induce prostate development, possibly by Rps6kb1, a subunit of the mTOR target, ribosomal modulating responsiveness to Fgfs (Thomson, 2001; protein S6 kinase. By the 48-h time point, PTEN Donjacour et al., 2003). In the developing prostate, we transcripts were suppressed, along with the Pik3r2 identified sexually dimorphic expression of Fgf receptors regulatory subunit of PI3K. These results indicate and ligands as part of the MAPK pathway through dynamic androgen-responsive regulation of the Pten which they signal (Table 1). Fgf18, which binds to signaling pathway. Future studies in prostate develop- FgfR3, (reviewed in (Mohammadi et al., 2005) is ment may therefore elucidate normal Pten/PI3K/mTOR induced at 6 h and then is highly suppressed later in regulation, its functional roles and more effective prostate development. Fgf12 was induced in males, strategies for pathway measurement and manipulation. although it reportedly does not bind FgfRs (Olsen et al., 2003). Two Fgf receptors showed increased expression in androgen-induced tissue; FgfR3, which can bind a Signaling in prostate development: the Wnt pathway wide range of Fgf ligands, and Fgfr2 which encodes the Extending an observation made previously by Serial preferred receptor for Fgfs 7 and 10 (Mohammadi et al., Analysis of Gene Expression in the urogenital sinus 2005). These results implicate Fgf12 and Fgf18 in UGS (Zhang et al., 2006), analyses of functional annotation differentiation and suggest that previously observed revealed Wnt signaling as one of the most androgen sexually dimorphic responses to Fgf7 and Fgf10 could regulated pathways in early prostate development (adj. stem from male-specific expression of FgfR2. Po0.001), with regulation of multiple Wnt ligands and soluble inhibitors (Table 1). Wnt ligands are highly conserved secreted molecules that play critical but Signaling in prostate development: PTEN/PI3K/mTOR pleiotropic roles in cell- during embryogen- signaling esis (Nusse, 2005). Studies of this pathway in prostate Aberrant signaling through PTEN and downstream cancer show up- and downregulation of Wnt ligands, effectors is highly implicated in prostate cancer, but its both stimulatory and inhibitory, with little or no expression in early prostate development is uncharacter- agreement as to their net effect (Verras and Sun,

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7186 2006). Likewise, nearly all Wnts are expressed in the that characterize transitions to more invasive lesions developing prostate (not shown) along with Wnt (that is, normal vs locally invasive cancers; lower grade inhibitory factor (Wif-1), and secreted frizzled , vs higher grade locally invasive cancers and localized vs which can modulate UGS development in culture metastatic cancers). Comparing these lists to lists of (Joesting et al., 2005; Table 1). Androgen-responsive differentially expressed genes at each of the develop- expression was also observed for Wnt ligand coactiva- mental time points revealed significant (adj Po0.001) tor, R-spondin1 (Rspo1; Nam et al., 2007), and several enrichment of developmental genes in invasive transi- intracellular components of the pathway, including tions, especially in more aggressive prostate cancers disheveled 1 (Dvl1), lymphoid enhancer-binding factor (Figure 3a, top). In contrast, gene expression in adult 1 (lef1) and -like enhancer of split 1 (Tle1; prostate regeneration showed little enrichment in cancer Table 1). Of special interest, the Wnt target gene matrix (Figure 3a, bottom). The latter result was somewhat metalloproteinase 7 (Mmp7; Crawford et al., 1999) has a surprising, given the previously reported enrichment of well-established role in cancer invasion (Ii et al., 2006), cell cycle, cell adhesion and metabolic genes in the and is induced during prostate budding (Table 1). These prostate regeneration study (Wang et al., 2007b). results suggest a general role for Wnt signal modulation Using laser capture microdissection to isolate and in prostate development that may underlie the me- profile pure epithelial cell populations from the prostate, chanics of epithelial invasion in prostate development Tomlins et al. (2007) have assigned much of the and prostate cancer. variation in previously published prostate cancer gene expression studies to varying ratios of cancer epithelia to benign stroma. We therefore performed a second Prostate carcinogenesis and progression reactivates analysis, looking for enrichment of developmental genes androgen-induced embryonic gene expression programs in the cancer contrasts from Tomlins et al. (2007). Having demonstrated that androgen-responsive em- Surprisingly, these analyses (Figure 3b) showed in- bryonic gene expression programs regulate processes creased enrichment of embryonic genes compared with that are fundamental to cancer, we examined the macrodissected bulk cancer samples (Figure 3a). En- expression of embryonic prostate genes in human richment was significant for all time points (adj prostate cancer tissues. Like prostate organogenesis, Po0.001), and included transitions for categories initiation and progression of prostate cancer depends on indicating malignant transformation (normal vs PIN), AR signaling (Scher and Sawyers, 2005). Neoplastic invasion (PIN vs cancer) and aggressiveness (Gleason transformation in the prostate is first recognized as high grade; Figure 3b). The genes driving these processes are grade prostatic intraepithelial neoplasia (PIN), a non- listed in Supplementary Tables 13–18. Thus, invasive invasive lesion that precedes invasive cancer (Bostwick prostate cancer adopts embryonic programs particularly and Qian, 2004). Invasive prostate cancer can be during neoplastic transformation and invasion (PIN vs indolent, particularly in lower grade tumors (Gleason cancer) as indicated by enrichment across all three 6 or less). Higher grade tumors (Gleason 8–10) more developmental time points (boxes 2, 5 and 6 in often metastasize and cause death. Using public domain Figure 3b). In addition, genes differentially expressed gene expression data from a large and well-characterized in prostate cancer progression overlap most significantly series (Lapointe et al., 2004), we assembled lists of genes with genes differentially expressed at the earliest (6 h)

Normal vsCancer cancer Localgrade vs(low metastatic vs high) Nml vs PINPIN vs Cancercancer gradeLocal (lowvs metastatic vs high) 6 6 h 1234 Duration Duration 12 Androgen 12 h 5

Dev Androgen 48 48 h 6 Regressing P < 0.00001 P < 0.00001 Castrate P <- 0.006 P < 0.001 Regenerating Regen P = 0.08 P = 0.05 Figure 3 Embryonic gene expression in human prostate cancer. (a) Genes differentially expressed at each time point (6, 12, or 48 h of androgen exposure; Y-axis) in early prostate development (top half of heat map) are identical to genes differentially expressed at different stages of prostate cancer progression (X-axis), whereas prostate regeneration (bottom half) shows little relationship to cancer. In this data set, based on macrodissected cancers (Lapointe et al., 2004), Gleason grade 6 tumors are labeled ‘low grade’ and ‘high grade’ are Gleason grades 8–10. Degree of shading indicates statistical significance in comparisons between two gene sets (b) Differentially expressed genes in early prostate development are also enriched in a similar prostate cancer progression study (Tomlins et al., 2007) using microdissected epithelial and cancer cells. Cancer comparisons include normal epithelium vs high-grade prostatic interaepithelial neoplasia (Nml vs PIN), PIN vs invasive cancer, cancer grade, (low vs high) and local vs metastatic tumors. Developmental genes enriched in cancer transitions (boxes labeled 1–6) are listed in Supplementary Tables 13–18.

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7187 time point in prostate development, when prostate identity is first becoming established. Although most of these genes are repressed at the 6-h time GO categories point in development (Figure 2c), the same genes P < 0.00001 show a more equal balance between induction and Development P = 0.001 repression (Supplementary Tables 13–18) in cancer. P = 0.05 These observations identify androgen-responsive gene 6 h 12 h 48 h Nml vs PINPIN vs CancercancerLocal grade vs (low metastatic vs high) networks that establish prostate identity and operate in Programmed cell death prostate cancer. Comparing and contrasting normal and Apoptosis malignant operation of these networks will likely yield Regulation of transcription useful insights into the molecular basis of prostate Angiogenesis cancer. ECM remodeling Receptor kinase signaling Wnt signaling BMP signaling Embryonic genes drive neoplastic transformation and Cell migration invasion Microtubule regulation Several shared functional themes from early prostate Tyrosine kinase signaling development also operate in human prostate cancer Adaptor Proteins in signaling tissues, although at distinct stages of disease progression (Figure 4). Selected differentially expressed genes Transcription factor binding sites (P 0.05) and themes enriched in development and FoxD3 o CCT1 cancer are listed in Table 2. Complete lists are provided Lef1 in Supplementary Tables 13–18. The transition between Sox9 normal prostate and PIN is characterized by enrichment Lef4 in pathways regulating transcription and apoptosis MicroRNAs (Figure 4a). The latter observation is in keeping with MiR-17-5p striking suppression of programmed cell death pre- MiR-21 viously observed in PIN (Zeng and Kyprianou, 2005). In contrast, the transition from noninvasive PIN to Figure 4 Transitions to increasingly invasive cancers are char- invasive cancer is enriched for extracellular matrix acterized by the activation of distinct pathways, transcription factors and microRNA target genes. Analysis of functional remodeling, cell motility and angiogenesis, as well as annotation in development and cancer transition reveals distinct signaling pathways including Wnt and bone morphoge- (a) Gene ontology (GO) categories, (b) predicted transcription netic proteins (Figure 4a). These observations outline a factor binding sites in differentially expressed genes and (c) program for prostate invasion that serves prostate predicted targets of specific miRNAs at each cancer transition development and cancer. (for each category listed, there is enrichment in at least one of the three developmental time points and at least one of the four cancer Interestingly, other investigators have reported that transitions). Contrasts include benign vs prostatic intraepithelial metastatic potential can pre-exist in localized tumors neoplasia (Nml v PIN), PIN vs cancer, cancer grade (low vs high), rather than being induced by the external environment and localized vs metastatic tumors. (Bernards and Weinberg, 2002; Kang et al., 2003). This concept may explain why the transition from localized to metastatic lesions did not represent themes expected Reduced Anxa1 expression characterizes 91% percent of to underlie , such as cell migration and PIN lesions/cases, and is maintained in 94% of prostate adhesion. Instead, enriched processes included micro- cancers (Kang et al., 2002). Anxa1 expression has not tubule dynamics, mitotic spindle regulation and en- been described in prostate development. In contrast to hanced intracellular signaling (tyrosine kinase activity its loss in cancer, Anxa1 transcript expression was and adaptor proteins). Proteins in the latter class make strongly induced at the 48-h time point (Supplementary excellent drug targets (Zwick et al., 2002), and to the Table 3). Immunohistochemical staining of developing extent that these genes might facilitate prostate cancer prostates confirmed increased expression of Anxa1 in metastasis, they warrant further investigation. male tissue; however, there was a striking reduction of expression in emerging epithelial buds (Figures 5a–c). Prostate budding recapitulates the aspects of neoplastic This pattern of expression is consistent with a role for transformation Anxa1 repression in growth, survival, and/or invasion of As a complement to statistical analyses and reverse prostate epithelium. Future studies involving genetic transcriptase–PCR confirmation, we performed immu- and/or pharmacologic manipulation of the developing nohistochemical analysis of early prostate development. prostate should further elucidate this role. Here, we show examples of two genes, Anxa1 and Sox9, Sox9 is a member of the SOX (Sry-related high which are coordinately suppressed and activated, mobility group) family of transcription factors that respectively in prostate epithelial buds, mirroring their binds a DNA consensus sequence and enhances steroid regulation in prostate carcinogenesis. binding. Functionally, it has been Annexin A (Anxa1) is a potent inducer of cell death in linked to a number of themes shared by development prostate cancer cell lines (Ornstein and Tyson, 2006). and cancer (Table 2). In developing prostate, Sox9 and

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7188

Female Male

α-AnxA1 α-AnxA1 α-AnxA1

Sox9 p63

PIN PIN

Sox9 p63 UGS

Figure 5 Annexin and Sox9 in epithelial invasion. (a and b) Immunohistochemical localization of Annexin A1 (AnxA1) in UGS tissue from female (a) and male (b) e17.5 littermates. (c) Immunohistochemical localization of Annexin A1 in male e18 UGS showing decreased. Annexin A1 at tips of invading prostate epithelial buds (arrow). (d) Immunofluorsescent localization of Sox9 protein (green) at tips of invading prostate epithelial buds at e18. Antibodies against p63 (red) show nearly ubiquitous expression in UGS epithelium. Nuclei appear blue (DAPI stain). (e) Hematoxylin and eosin stain of PIN tissue microarray. Higher power inset demonstrates prominent nucleoli characteristic of PIN. (f) Adjacent tissue section showing immunofluorsescent localization of Sox9 protein (green) in predominantly luminal cells of the same PIN lesion shown in panel e. Antibodies against p63 (red) label basal cells. Nuclei appear blue (DAPI stain). Higher power inset demonstrates localization of Sox9 protein (green) in luminal epithelial cells in contrast with basal expression of p63 (red).

its predicted transcriptional targets are differentially with a trend toward increased expression with increasing expressed during each of the three time points after grade (P ¼ 0.1; Supplementary Table 20). Nuclear Sox9 androgen exposure (Supplementary Tables 1–3, 10–12). was also noted in similar numbers (17/31, 55%) of Immunohistochemical analysis revealed that Sox9 is advanced (lymph node metastasis) prostate cancer cases. expressed in early prostate epithelium and is particularly In sum, alterations in the geographic and temporal concentrated in the tips of invading prostate buds location of Sox9 expression parallel the earliest events in (Figure 5d). This expression pattern contrasts to that of the neoplastic transformation of prostate epithelial cells. Anxa1 with its reduced staining in bud tips and differs These results suggest that Sox9 may play a critical role from p63, which is diffusely expressed in UGE at this in the early, initiating phase of prostate carcinogenesis stage of development. Altogether, this suggests Sox9 is and contribute aspects of the basal/stem cell phenotype critically located to facilitate prostate outgrowth, a to prostate cancer. This pattern differs from recent hypothesis supported by recent work demonstrating its work, which identified increased Sox9 in metastatic essential role in prostate development (Thomsen et al., rather than localized lesions in a murine prostate cancer 2008; Z Huang, B Simons, EM Schaeffer and DM model (Acevedo et al., 2007). It is likely that Sox9 has Berman, unpublished observations). We noted that Sox9 several context-dependent functions in prostate epithe- transcripts and predicted targets were differentially lium and prostate cancer. The list of predicted Sox9 expressed across prostate cancer progression (Supple- target genes that participate in prostate development mentary Tables 14–18), and others have linked Sox9 to (Supplementary Table 21) represents a useful starting invasion, cell growth and metastasis in prostate cancer point for further investigation. models. We further characterized SOX9 expression in a panel of 219 human prostate cases, using immunohistochem- istry and Immunofluorescence. In benign prostate tissue, Discussion Sox9 is expressed predominantly in basal epithelial cells, the stem cell compartment for prostate epithelia. (Xu Previous studies have linked primitive embryonic gene et al., 2005; Wang et al., 2007a). We discovered aberrant expression profiles to aggressive subsets of brain and Sox9 expression in neoplastic luminal epithelial cells in lung cancers (Kho et al., 2004), thereby supporting 51 of 53 (96%) PIN lesions (Supplementary Table 20), balances between differentiation state and growth indicating an association with neoplastic transformation potential that operate similarly in organogenesis and (Figures 5e and f). In prostate carcinomas (n ¼ 105), tumorigenesis. Here, we identified prostate-specific basal cell gene expression is lost; however Sox9 programs for growth, survival, angiogenesis, and inva- expression is maintained in cancer cells (n ¼ 59, 56%) sion that originate in organogenesis and are reactivated

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7189 at specific steps in cancer progression. This curated list Probe synthesis, hybridization to oligonucleotide arrays and of androgen-regulated programs in prostate develop- detection ment and carcinogenesis provides a roadmap for Amplification, labeling, hybridization, and detection of 250 ng understanding prostate growth and invasion. Prostate samples were carried out according to the manufacturer’s development in the mouse becomes a tractable experi- directions (Agilent). mental system in which to investigate the specific functions of these genes. This model bypasses the Statistical analysis difficulty of probing gene regulation in human prostate Data were processed without background subtraction with cancer cell lines, which represent rare and possibly packages from R/Bioconductor (www.bioconductor.org/) skewed exceptions to the rule that human prostate (Ihaka and Gentleman, 1996; Gentleman et al., 2004). With- cancers do not adapt to growth in the laboratory. in-array ‘loess’ and between-arrays ‘scale’ normalization Unlike the extant models of prostate cancer involving methods were applied to log2 expression ratios. Moderated t-statistics (by empirical Bayes shrinkage of standard errors transgenic mice, prostate development is unbiased by a (Smyth, 2004)) log odds ratios of differential expression and preselected genetic lesion. Development is reproducible, adjusted P-values (Benjamini et al., 2001) were obtained from genetically (Xin et al., 2003) and pharmacologically a linear model accounting for biological replicates’ correlation, (Berman et al., 2004) tractable, and shown here to be and dye/group effects. reflected the entire spectrum of human prostate cancer Functional themes were obtained from Gene ontology progression. (Ashburner et al., 2003), KEGG (Kanehisa et al., 2004), A unique feature of these studies was the ability to MsigDb(Subramanian et al., 2005). Enrichment analysis was induce prostatic lineage commitment and growth by performed by one-sided Wilcoxon test and multiple testing controlled induction of signaling by AR, a gene with correction performed separately (Benjamini et al., 2001). lineage-specific oncogenic properties (Garraway and Detailed statistical methods, public microarray data set identification and MIAME (Minimal Information about Sellers, 2006). Evidence indicates the operation of other Microarray Experiments) information are provided in Supple- lineage-specific oncogenes in melanocytes (Garraway mentary materials. Data will be hosted in the Gene Expression and Sellers, 2006), lung (NKX2-1; Weir et al., 2007) and Omnibus database. elsewhere in the body (Garraway and Sellers, 2006), suggesting that gene expression programs relevant to other types of cancers can be identified, manipulated Immunohistochemistry Detection was performed as described in Berman et al.(2004) and modeled in the embryo. using Sox9 (Chemicon, Billerica, MA, USA) and Annexin A1 (Invitrogen, Carlsbad, CA, USA) antibodies.

Materials and methods Double immunofluorescence Mice and tissues Detection was performed with anti-Sox9 and anti-p63 C57/Bl6J (The Jackson Laboratories, Bar Harbor, ME, USA) (Santacruz, Santa Cruz, CA, USA) antibodies as described pregnancies were timed according to scheduled 4 h pairings. in supplement (Bivalacqua et al., 2007). Paired pregnant females were injected intraperitoneally with dihydrotestosterone (Sigma-Aldrich, St Louis, MO, USA) at Tissue microarrays 50 mg/kg or triolein vehicle at e16.0. 6 or 12 h later, embryos Human tissues (Summarized in Supplementary Table 20) were were sexed and female UGS tissue dissected. For the 48-h time stained with Sox9 antibody and scored as described in point tissue from unmanipulated male and female littermates Supplementary Methods. was harvested at e17.5.

RNA isolation Acknowledgements Frozen tissue was homogenized and total RNA purified using the RNeasy system (Qiagen, Hilden, Germany) and analyzed We thank W Matsui and N Watkins for comments on the using a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, manuscript, A DeMarzo and J Epstein for TMAs and YQ CA, USA). Each time point yielded pools of ‘androgen Chen for prostate regeneration data. These studies were exposed’ and ‘androgen naı¨ ve’ RNA. There were a total of funded by the Evensen Family, Passano and Patrick C Walsh 8,6 and 10 pools of RNA at the 6-, 12- and 48-h time points, Prostate Cancer Foundations, and NIH5K08DK059375 (DB) respectively. Each pool was obtained from 3–4 individual NIHK08 DK081019 (EMS), NIH5P30CA06973-39 (GP) and UGS. NSF034211 (GPand LM).

References

Acevedo VD, Gangula RD, Freeman KW, Li R, Zhang Y, Wang F Bailey P, Cushing H. (1925). Medulloblastoma cerbelli: a common type of et al. (2007). Inducible FGFR-1 activation leads to irreversible midcerebellar glioma of childhood. Arch Neurol Psychiatry 14: 192–223. prostate adenocarcinoma and an epithelial-to-mesenchymal transi- Benjamini Y, Drai D, Elmer G, Kafkafi N, Golani I. (2001). tion. Cancer Cell 12: 559–571. Controlling the false discovery rate in behavior genetics research. Ashburner M, Mungall CJ, Lewis SE. (2003). Ontologies Behav Brain Res 125: 279–284. for biologists: a community model for the annotation of genomic Berman DM, Desai N, Wang X, Karhadkar SS, Reynon M, data. Cold Spring Harb Symp Quant Biol 68: 227–235. Abate-Shen C et al. (2004). Roles for Hedgehog signaling in

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7190 androgen production and prostate ductal morphogenesis. Dev Biol estrogen or androgen receptors: dissociation from transcriptional 267: 387–398. activity. Cell 104: 719–730. Bernards R, Weinberg RA. (2002). A progression puzzle. Nature 418: Lapointe J, Li C, Higgins JP, van de Rijn M, Bair E, Montgomery K 823. et al. (2004). Gene expression profiling identifies clinically relevant Bhatia-Gaur R, Donjacour AA, Sciavolino PJ, Kim M, Desai N, subtypes of prostate cancer. Proc Natl Acad Sci USA 101: 811–816. Young P et al. (1999). Roles for Nkx3.1 in prostate development and Lawson DA, Xin L, Lukacs RU, Cheng D, Witte ON. (2007). cancer. Genes Dev 13: 966–977. Isolation and functional characterization of murine prostate stem Bivalacqua TJ, Deng W, Kendirci M, Usta MF, Robinson C, Taylor cells. Proc Natl Acad Sci USA 104: 181–186. BK et al. (2007). Mesenchymal stem cells alone or ex vivo gene Long X, Lin Y, Ortiz-Vega S, Yonezawa K, Avruch J. (2005). Rheb modified with endothelial nitric oxide synthase reverse age- binds and regulates the mTOR kinase. Curr Biol 15: 702–713. associated erectile dysfunction. Am J Physiol Heart Circ Physiol Lu J, Qian J, Izvolsky KI, Cardoso WV. (2004). Global analysis of 292: H1278–H1290. genes differentially expressed in branching and non-branching Bostwick DG, Qian J. (2004). High-grade prostatic intraepithelial regions of the mouse embryonic lung. Dev Biol 273: 418–435. neoplasia. Mod Pathol 17: 360–379. Manolagas SC, Kousteni S, Jilka RL. (2002). Sex steroids and bone. Crawford HC, Fingleton BM, Rudolph-Owen LA, Goss KJ, Recent Prog Horm Res 57: 385–409. Rubinfeld B, Polakis P et al. (1999). The metalloproteinase McMenamin ME, Soung P, Perera S, Kaplan I, Loda M, Sellers WR. matrilysin is a target of beta-catenin transactivation in intestinal (1999). Loss of PTEN expression in paraffin-embedded primary tumors. Oncogene 18: 2883–2891. prostate cancer correlates with high Gleason score and advanced Cunha GR, Cooke PS, Kurita T. (2004). Role of stromal-epithelial stage. Cancer Res 59: 4291–4296. interactions in hormonal responses. Arch Histol Cytol 67: 417–434. Memarzadeh S, Xin L, Mulholland DJ, Mansukhani A, Wu H, Teitell Donjacour AA, Thomson AA, Cunha GR. (2003). FGF-10 plays an MA et al. (2007). Enhanced paracrine FGF10 expression promotes essential role in the growth of the fetal prostate. Dev Biol 261: 39–54. formation of multifocal prostate adenocarcinoma and an increase in Drews U, Sulak O, Oppitz M. (2001). Immunohistochemical localisa- epithelial . Cancer Cell 12: 572–585. tion of androgen receptor during sex-specific morphogenesis in the Mohammadi M, Olsen SK, Ibrahimi OA. (2005). Structural basis for fetal mouse. Histochem Cell Biol 116: 427–439. fibroblast activation. Cytokine Growth Factor English HF, Santen RJ, Isaacs JT. (1987). Response of glandular Rev 16: 107–137. versus basal rat ventral prostatic epithelial cells to androgen Nam JS, Park E, Turcotte TJ, Palencia S, Zhan X, Lee J et al. (2007). withdrawal and replacement. Prostate 11: 229–242. Mouse R-spondin2 is required for apical ectodermal ridge main- Friend SH, Bernards R, Rogelj S, Weinberg RA, Rapaport JM, Albert tenance in the hindlimb. Dev Biol 311: 124–135. DM et al. (1986). A human DNA segment with properties of the Nusse R. (2005). Wnt signaling in disease and in development. Cell Res gene that predisposes to retinoblastoma and osteosarcoma. Nature 15: 28–32. 323: 643–646. Olsen SK, Garbi M, Zampieri N, Eliseenkova AV, Ornitz DM, Garraway LA, Sellers WR. (2006). Lineage dependency and lineage- Goldfarb M et al. (2003). (FGF) survival oncogenes in human cancer. Nat Rev Cancer 6: 593–602. homologous factors share structural but not functional homology Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S with FGFs. J Biol Chem 278: 34226–34236. et al. (2004). Bioconductor: open software development for Ornstein DK, Tyson DR. (2006). Proteomics for the identification of computational biology and bioinformatics. Genome Biol 5: R80. new prostate cancer biomarkers. Urol Oncol 24: 231–236. Hanahan D, Weinberg RA. (2000). The hallmarks of cancer. Cell 100: Ramaswamy S, Tamayo P, Rifkin R, Mukherjee S, Yeang CH, 57–70. Angelo M et al. (2001). Multiclass cancer diagnosis using tumor Ihaka R, Gentleman R. (1996). R: A language for data analysis and gene expression signatures. Proc Natl Acad Sci USA 98: graphics. J Comput Graph Stat 5: 299–314. 15149–15154. Ii M, Yamamoto H, Adachi Y, Maruyama Y, Shinomura Y. (2006). Scher HI, Sawyers CL. (2005). Biology of progressive, castration- Role of matrix metalloproteinase-7 (matrilysin) in human cancer resistant prostate cancer: directed therapies targeting the androgen- invasion, apoptosis, growth, and angiogenesis. Exp Biol Med receptor signaling axis. J Clin Oncol 23: 8253–8261. (Maywood) 231: 20–27. Shen MM, Abate-Shen C. (2007). Pten inactivation and the emergence of Irizarry RA, Warren D, Spencer F, Kim IF, Biswal S, Frank BC et al. androgen-independent prostate cancer. Cancer Res 67: 6535–6538. (2005). Multiple-laboratory comparison of microarray platforms. Smyth GK. (2004). Linear models and empirical Bayes methods for Nat Methods 2: 345–350. assessing differential expression in microarray experiments. Stat Joesting MS, Perrin S, Elenbaas B, Fawell SE, Rubin JS, Franco OE Appl Genet Mol Biol 3: article3 (e-pub 2004 Feb 12). et al. (2005). Identification of SFRP1 as a candidate mediator of Stefani G, Slack FJ. (2008). Small non-coding RNAs in animal stromal-to-epithelial signaling in prostate cancer. Cancer Res 65: development. Nat Rev Mol Cell Biol 9: 219. 10423–10430. Stehelin D, Varmus HE, Bishop JM, Vogt PK. (1976). DNA related to Jost A. (1968). Full or partial maturation of fetal endocrine systems the transforming gene(s) of avian sarcoma viruses is present in under pituitary control. Perspect Biol Med 11: 371. normal avian DNA. Nature 260: 170–173. Kanehisa M, Goto S, Kawashima S, Okuno Y, Hattori M. (2004). The Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, KEGG resource for deciphering the genome. Nucleic Acids Res 32: Gillette MA et al. (2005). Gene set enrichment analysis: a knowl- D277–D280. edge-based approach for interpreting genome-wide expression Kang JS, Calvo BF, Maygarden SJ, Caskey LS, Mohler JL, Ornstein profiles. Proc Natl Acad Sci USA 102: 15545–15550. DK. (2002). Dysregulation of annexin I protein expression in high- Thomsen MK, Butler CM, Shen MM, Swain A. (2008). Sox9 is grade prostatic intraepithelial neoplasia and prostate cancer. Clin required for prostate development. Dev Biol 316: 302–311. Cancer Res 8: 117–123. Thomson AA. (2001). Role of androgens and fibroblast growth factors Kang Y, Siegel PM, Shu W, Drobnjak M, Kakonen SM, Cordon- in prostatic development. Reproduction 121: 187–195. Cardo C et al. (2003). A multigenic program mediating breast cancer Tomlins SA, Mehra R, Rhodes DR, Cao X, Wang L, Dhanasekaran metastasis to bone. Cancer Cell 3: 537–549. SM et al. (2007). Integrative molecular concept modeling of prostate Kho AT, Zhao Q, Cai Z, Butte AJ, Kim JY, Pomeroy SL et al. (2004). cancer progression. Nat Genet 39: 41–51. Conserved mechanisms across development and tumorigenesis Treister NS, Richards SM, Lombardi MJ, Rowley P, Jensen RV, revealed by a mouse development perspective of human cancers. Sullivan DA. (2005). Sex-related differences in gene expression in Genes Dev 18: 629–640. salivary glands of BALB/c mice. J Dent Res 84: 160–165. Kousteni S, Bellido T, Plotkin LI, O’Brien CA, Bodenner DL, Han L Verras M, Sun Z. (2006). Roles and regulation of Wnt signaling and et al. (2001). Nongenotropic, sex-nonspecific signaling through the beta-catenin in prostate cancer. Cancer Lett 237: 22–32.

Oncogene Embryonic gene expression in prostate cancer EM Schaeffer et al 7191 Wang H, McKnight NC, Zhang T, Lu ML, Balk SP, Yuan X. (2007a). Xin L, Lawson DA, Witte ON. (2005). The Sca-1 cell surface marker SOX9 is expressed in normal prostate basal cells and regulates androgen enriches for a prostate-regenerating cell subpopulation that can initiate receptor expression in prostate cancer cells. Cancer Res 67: 528–536. prostate tumorigenesis. Proc Natl Acad Sci USA 102: 6942–6947. Wang S, Gao J, Lei Q, Rozengurt N, Pritchard C, Jiao J et al. Xu J, Lowey J, Wiklund F, Sun J, Lindmark F, Hsu FC et al. (2005). (2003). Prostate-specific deletion of the murine Pten tumor The interaction of four genes in the inflammation pathway suppressor gene leads to metastatic prostate cancer. Cancer Cell 4: significantly predicts prostate cancer risk. Cancer Epidemiol 209–221. Biomarkers Prev 14: 2563–2568. Wang XD, Wang BE, Soriano R, Zha J, Zhang Z, Modrusan Z et al. Yeager N, Brewer C, Cai KQ, Xu XX, Di Cristofano A. (2008). (2007b). Expression profiling of the mouse prostate after castration Mammalian target of rapamycin is the key effector of phosphati- and hormone replacement: implication of H-cadherin in prostate dylinositol-3-OH-initiated proliferative signals in the thyroid folli- tumorigenesis. Differentiation 75: 219–234. cular epithelium. Cancer Res 68: 444. Weir BA, Woo MS, Getz G, Perner S, Ding L, Beroukhim R et al. Zeng L, Kyprianou N. (2005). Apoptotic regulators in prostatic (2007). Characterizing the cancer genome in lung adenocarcinoma. intraepithelial neoplasia (PIN): value in prostate cancer detection Nature 450: 893–898. and prevention. Prostate Cancer Prostatic Dis 8: 7–13. Wilson JD, Griffin JE, George FW. (1980). Sexual differentiation: Zhang TJ, Hoffman BG, Ruiz de Algara T, Helgason CD. (2006). early hormone synthesis and action. Biol Reprod 22:9. SAGE reveals expression of Wnt signalling pathway Xin L, Ide H, Kim Y, Dubey P, Witte ON. (2003). In vivo regeneration members during mouse prostate development. Gene Expr Patterns of murine prostate from dissociated cell populations of postnatal 6: 310–324. epithelia and urogenital sinus mesenchyme. Proc Natl Acad Sci USA Zwick E, Bange J, Ullrich A. (2002). Receptor tyrosine kinases as 100(Suppl 1): 11896–11903. targets for anticancer drugs. Trends Mol Med 8: 17–23.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene