Abstract: Genetically engineered mice are being used increasingly for delineating the molecular mechanisms of prostate cancer development. Epithelium-stroma interactions play a critical role in prostate development and tumorigenesis. To better understand gene expression patterns in the normal sexually mature mouse prostate, epithelium and stroma were laser-capture microdissected from ventral, dorsolateral, and anterior prostate lobes. Genome-wide expression was measured by DNA microarrays. Our analysis indicated that the gene expression pattern in the mouse dorsolateral lobe was closest to that of the human prostate peripheral zone, supporting the hypothesis that these prostate compartments are functionally equivalent. Stroma from a given lobe had closer gene expression patterns with stroma from other lobes than epithelium from the same lobe. Stroma appeared to have higher expression complexity than epithelium. Specifically, stromal cells had higher expression levels of genes implicated in cell adhesion, muscle development, and contraction, in structural constituents of cytoskeleton and actin binding, and in components such as sarcomere and extracellular matrix collagen. Among the genes that were enriched in the epithelium were secretory proteins, including seminal vesicle protein secretion 2 and 5. Surprisingly, prostate stroma expressed many osteogenic molecules, as confirmed by immunohistochemistry. A "bone-like" environment in the prostate may predispose prostate cells for survival in the bone. Chemokine Cxcl12 but not its receptor, Cxcr4, was expressed in normal prostate. In prostate tumors, interestingly, Cxcl12 was up-regulated in epithelial cells with a concomitant expression of Cxcr4. Expression of both the receptor and ligand may provide an autocrine mechanism for tumor cell migration and invasion. Genetically engineered mice are being used increasingly for delineating the molecular mechanisms of prostate cancer development. Epithelium-stroma interactions play a critical role in prostate development and tumorigenesis. To better understand gene expression patterns in the normal sexually mature mouse prostate, epithelium and stroma were laser-capture microdissected from ventral, dorsolateral, and anterior prostate lobes. Genome-wide expression was measured by DNA microarrays. Our analysis indicated that the gene expression pattern in the mouse dorsolateral lobe was closest to that of the human prostate peripheral zone, supporting the hypothesis that these prostate compartments are functionally equivalent. Stroma from a given lobe had closer gene expression patterns with stroma from other lobes than epithelium from the same lobe. Stroma appeared to have higher expression complexity than epithelium. Specifically, stromal cells had higher expression levels of genes implicated in cell adhesion, muscle development, and contraction, in structural constituents of cytoskeleton and actin binding, and in components such as sarcomere and extracellular matrix collagen. Among the genes that were enriched in the epithelium were secretory proteins, including seminal vesicle protein secretion 2 and 5. Surprisingly, prostate stroma expressed many osteogenic molecules, as confirmed by immunohistochemistry. A "bone-like" environment in the prostate may predispose prostate cells for survival in the bone. Chemokine Cxcl12 but not its receptor, Cxcr4, was expressed in normal prostate. In prostate tumors, interestingly, Cxcl12 was up-regulated in epithelial cells with a concomitant expression of Cxcr4. Expression of both the receptor and ligand may provide an autocrine mechanism for tumor cell migration and invasion. Prostate cancer is a major health issue in the United States. Genetically engineered mice are an important model for studying the molecular mechanisms of prostate tumorigenesis (1Kasper S. Smith Jr., J.A. J. Urol. 2004; 172: 12-19Crossref PubMed Scopus (43) Google Scholar). The usefulness of mice to model human prostate cancer is well accepted (1Kasper S. Smith Jr., J.A. J. Urol. 2004; 172: 12-19Crossref PubMed Scopus (43) Google Scholar, 2Abate-Shen C. Shen M.M. Trends Genet. 2002; 18: S1-S5Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar). However, gene expression in mouse prostate has not been well characterized. More importantly, expression differences between the mouse prostate epithelium and stroma are largely unknown, despite the critical role of stromal-epithelial interactions in the development of normal prostate and prostate cancer. The mouse prostate consists of three distinct lobes (anterior, dorsolateral, and ventral), which differs anatomically from the human prostate (a single lobe with peripheral, transition, and central zones (3Shappell S.B. Thomas G.V. Roberts R.L. Herbert R. Ittmann M.M. Rubin M.A. Humphrey P.A. Sundberg J.P. Rozengurt N. Barrios R. Ward J.M. Cardiff R.D. Cancer Res. 2004; 64: 2270-2305Crossref PubMed Scopus (488) Google Scholar)). Gene expression profiling allows for global assessment of gene expression patterns and identification of differentially expressed genes. In human prostate cancer cells, differentially expressed genes have been analyzed by SAGE (serial analysis of gene expression) (4Waghray A. Schober M. Feroze F. Yao F. Virgin J. Chen Y.Q. Cancer Res. 2001; 61: 4283-4286PubMed Google Scholar), expressed sequence tag library analysis (5Nelson P.S. Clegg N. Eroglu B. Hawkins V. Bumgarner R. Smith T. Hood L. Nucleic Acids Res. 2000; 28: 212-213Crossref PubMed Scopus (19) Google Scholar), and DNA microarrays (6Bubendorf L. Kolmer M. Kononen J. Koivisto P. Mousses S. Chen Y. Mahlamaki E. Schraml P. Moch H. Willi N. Elkahloun A.G. Pretlow T.G. Gasser T.C. Mihatsch M.J. Sauter G. Kallioniemi O.P. J. Natl. Cancer Inst. 1999; 91: 1758-1764Crossref PubMed Scopus (324) Google Scholar, 7Dhanasekaran S.M. Barrette T.R. Ghosh D. Shah R. Varambally S. Kurachi K. Pienta K.J. Rubin M.A. Chinnaiyan A.M. Nature. 2001; 412: 822-826Crossref PubMed Scopus (1430) Google Scholar, 8Luo J. Duggan D.J. Chen Y. Sauvageot J. Ewing C.M. Bittner M.L. Trent J.M. Isaacs W.B. Cancer Res. 2001; 61: 4683-4688PubMed Google Scholar, 9Magee J.A. Araki T. Patil S. Ehrig T. True L. Humphrey P.A. Catalona W.J. Watson M.A. Milbrandt J. Cancer Res. 2001; 61: 5692-5696PubMed Google Scholar, 10Welsh J.B. Sapinoso L.M. Su A.I. Kern S.G. Wang-Rodriguez J. Moskaluk C.A. Frierson Jr., H.F. Hampton G.M. Cancer Res. 2001; 61: 5974-5978PubMed Google Scholar, 11Singh D. Febbo P.G. Ross K. Jackson D.G. Manola J. Ladd C. Tamayo P. Renshaw A.A. D'Amico A.V. Richie J.P. Lander E.S. Loda M. Kantoff P.W. Golub T.R. Sellers W.R. Cancer Cell. 2002; 1: 203-209Abstract Full Text Full Text PDF PubMed Scopus (2035) Google Scholar). Some of the identified genes have been further characterized (12Waghray A. Keppler D. Sloane B.F. Schuger L. Chen Y.Q. J. Biol. Chem. 2002; 277: 11533-11538Abstract Full Text Full Text PDF PubMed Scopus (62) Google Scholar, 13Takayama T.K. McMullen B.A. Nelson P.S. Matsumura M. Fujikawa K. Biochemistry. 2001; 40: 15341-15348Crossref PubMed Scopus (156) Google Scholar, 14Lin B. Ferguson C. White J.T. Wang S. Vessella R. True L.D. Hood L. Nelson P.S. Cancer Res. 1999; 59: 4180-4184PubMed Google Scholar, 15Klezovitch O. Chevillet J. Mirosevich J. Roberts R.L. Matusik R.J. Vasioukhin V. Cancer Cell. 2004; 6: 185-195Abstract Full Text Full Text PDF PubMed Scopus (232) Google Scholar). Overall, this approach has identified many candidate biomarkers for prostate cancer and provided a wealth of information to further delineate molecular mechanisms of prostate cancer development. In the present study, we have determined mRNA expression profiles of the ventral, dorsolateral, and anterior prostate lobes in C57BL/6 and 129/SvEv mice, two of the most widely used strains in transgenic studies. Gene expression patterns were compared between epithelium and stroma, between different prostate lobes, and between mouse and human prostate. Differential expression of selected genes was validated by immunohistochemistry in normal and tumor mouse prostate. Importantly, our analysis provided, for the first time, molecular evidence establishing a functional equivalence between mouse dorsolateral lobe and human prostate peripheral zone, where the majority of human prostate cancers originate. Mouse Strains—Four- to five-week-old male C57BL/N6Tac and 129S6/SvEvTac mice were purchased from Taconic (Germantown, NY). Mice were housed in a pathogen-free facility for at least 1 week to reduce potential shipping stress. Prostate-specific Pten knock-out mice were generated by crossing Ptenloxp/loxp mice (16Lesche R. Groszer M. Gao J. Wang Y. Messing A. Sun H. Liu X. Wu H. Genesis. 2002; 32: 148-149Crossref PubMed Scopus (316) Google Scholar) with the ARR2Probasin-Cre transgenic line, PB-Cre4, where the Cre recombinase is under the control of a modified rat prostate-specific probasin promoter (17Wu X. Wu J. Huang J. Powell W.C. Zhang J. Matusik R.J. Sangiorgi F.O. Maxson R.E. Sucov H.M. Roy-Burman P. Mech. Dev. 2001; 101: 61-69Crossref PubMed Scopus (293) Google Scholar) (obtained from the NCI, National Institutes of Health mouse repository, mouse.ncifcrf.gov/). Prostate Lobe Dissection and Laser-capture Microdissection—The anterior (AP), 2The abbreviations used are:APanterior prostateVPventral prostateMBEImodel-based expression indexesDLdorsolateralAPEanterior prostate epitheliumAPSanterior prostate stroma dorsolateral (DL), and ventral (VP) prostate lobes from 6-week-old C57BL/N6Tac and 129S6/SvEvTac mice were dissected in RNase-free phosphate-buffered saline and snap-frozen in liquid nitrogen. Ten lobes of each type from the same mouse strain were pooled and embedded in tissue freezing media (catalog no. TFM-5, Triangle Bio-medical Sciences, Inc.) and kept at -80 °C. Five-μm thick sections were cut with an RNase-free blade in a cryostat at -20 °C and stained with the HistoGene LCM frozen section staining kit (Arcturus, Mountain View, CA). Epithelium and stroma were separated by laser-capture microdissection with a Leica LMD microscope (Leica Microsystems, Inc., Bannockburn, IL) (Fig. 1). anterior prostate ventral prostate model-based expression indexes dorsolateral anterior prostate epithelium anterior prostate stroma Microarray Analysis—Total RNA was isolated from microdissected tissues using the PicoPure RNA isolation kit (Arcturus). RNA quality and quantity were measured with Bioanalyzer pico chips (Agilent Technologies, Palo Alto, CA) (Fig. 1) and spectrophotometry. Approximately 200 ng of RNA from each sample were used for target labeling using the two-cycle cDNA synthesis protocol (Affymetrix, Santa Clara, CA). Fifteen μg of cRNA was hybridized to mouse expression 430 chip sets. Two to three biological replicas were performed for each data point. A total of 26 MOE430A and 26 MOE430B chips were used. Raw Data Processing—cel files were processed with the dCHIP software (18Li C. Hung Wong W. Genome Biol. 2001; 2 (RESEARCH0032)Google Scholar, 19Li C. Wong W.H. Proc. Natl. Acad. Sci. U. S. A. 2001; 98: 31-36Crossref PubMed Scopus (2713) Google Scholar). All arrays were normalized to a base-line array with a median probe intensity of 100. Because all normalization plots showed a correlation value with the base-line array r > 0.7, arrays were normalized and modeled in a single group. The perfect match-only model was used to obtain model-based expression indexes (MBEI). Identification of Differentially Expressed Genes—For epithelium-stroma comparisons, the anterior prostate epithelial (APE) samples were compared with the anterior prostate stromal (APS) samples (replica arrays of AP epithelium and stroma, respectively, from 129/SvEv and C57BL) with settings of APE/APS or APS/APE >2 (i.e. increase or decrease >2-fold) and APE-APS or APS-APE >100 (i.e. MBEI differences >100). Similarly, DL epithelium was compared with DL stroma and VP epithelium to VP stroma. Analysis of variance was performed among all samples. To identify differentially expressed genes between prostate lobes, replica arrays for stromal and epithelial samples of both mouse strains were combined for each lobe. Comparisons were performed between AP and DL, between AP and VP, and between DL and VP with the same criteria as above. Analysis of variance was performed among AP, DL, and VP. To identify differentially expressed genes between the prostate of C57BL and 129/SvEv, the C57 sample was compared with the 129 sample (replica arrays of APE, APS, DL epithelium, DL stroma, VP epithelium, and VP stroma from C57BL and 129/SvEv, respectively). Expression of Ortholog Genes in the Human Prostate Peripheral Zone—Raw data of 10 HG-U95Av2 arrays from normal human prostate tissues originating from the peripheral zone (11Singh D. Febbo P.G. Ross K. Jackson D.G. Manola J. Ladd C. Tamayo P. Renshaw A.A. D'Amico A.V. Richie J.P. Lander E.S. Loda M. Kantoff P.W. Golub T.R. Sellers W.R. Cancer Cell. 2002; 1: 203-209Abstract Full Text Full Text PDF PubMed Scopus (2035) Google Scholar) (N01, N36, N38, N39, N40, N58, N59, N60, N61, N62) were obtained from the Harvard microarray data base and processed using dCHIP. Arrays were normalized to a base-line array with a median probe intensity of 100. MBEI was calculated using the perfect match-only model. The list of differentially expressed genes between the mouse prostate lobes was queried using the NetAffx Analysis Center server (Affymetrix), and a list of ortholog genes on the HG-U95Av2 chip was obtained. The average MBEI of these ortholog genes was then obtained from the above mentioned ten human prostate samples. "High-level" Analysis—Hierarchical clustering and principal component analysis were performed with differentially expressed genes using the Genesis software (20Sturn A. Quackenbush J. Trajanoski Z. Bioinformatics (Oxf.). 2002; 18: 207-208Crossref PubMed Scopus (1484) Google Scholar). Comparative analysis of gene sets in the gene ontology space was performed using the GoSurfer software with a setting node of ≥5 genes (a node in the gene ontology classification tree must have at least five genes with the same function from the input list) and a p value <0.01 (a significantly larger proportion of genes associated with the node from the input list compared with the list of total genes available on the Affymetrix arrays). Immunohistochemistry—Prostate lobes of Pten+/+, Pten+/-, and Pten-/- mice were dissected from 8-week-old F2 generation male mice, fixed with in formaldehyde, and embedded paraffin. Wild-type Pten+/+ and heterozygous (Pten+/-) prostates were histologically normal, whereas prostates of homozygous knock-out (Pten-/-) mice all developed carcinoma. Tissue sections were deparaffinized and rehydrated, and antigens were retrieved by incubating sections in antigen unmasking solution (catalog no. H-3300, Vector Laboratories Inc., Burlingame, CA) at 95-96 °C for 1 h. Immunostaining was performed using the DAKO EnVion + system peroxidase (DAB) kit (catalog no. K4010, DakoCytomation Inc., Carpenteria, CA) according to the manufacturer's instructions. All antibodies were used at 5 ng/μl concentration. Normal rabbit or goat IgG was used as a control. Rabbit anti-Ctgf (sc-25440), rabbit anti-Sparc (sc-25574), rabbit anti-Cxcr4 (sc-9046), goat anti-Cxcl12 (sc-6193), and normal rabbit (sc-2027) and goat IgG (sc-2028) were purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). Rabbit anti-goat IgG (catalog no. 305-005-0030) was purchased from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA). DNA Microarray Analysis—Mouse anterior, dorsolateral, and ventral prostate lobes were obtained from 6-week-old male 129S6/SvEvTac and C57BL/N6Tac mice. Epithelium and stroma were separated by laser-capture microdissection (Fig. 1). Global gene expression of the mouse prostate was measured using Affymetrix DNA oligo microarrays (26 MOE430A and 26 MOE430B chips). Two to three biological replicas were performed for each data point, i.e. stroma and epithelium of the anterior, dorsolateral, and ventral prostate from 129/SvEv and C57BL mice. Raw data were processed using the dCHIP software. All arrays showed good quality with low levels of array outliers (0.07-3%) and single outliers (0.06-1.5%). Expression of Known Prostate Markers—The expression pattern of some known prostate markers was checked on our data set (Fig. 2). Serine protease inhibitor p12, spermine-binding protein p25, Hoxb13, and Foxa1, known to be expressed in the ventral prostate (21Mills J.S. Needham M. Thompson T.C. Parker M.G. Mol. Cell. Endocrinol. 1987; 53: 111-118Crossref PubMed Scopus (29) Google Scholar, 22Chen L.Y. Lin Y.H. Lai M.L. Chen Y.H. Biol. Reprod. 1998; 59: 1498-1505Crossref PubMed Scopus (46) Google Scholar, 23Mills J.S. Needham M. Parker M.G. Nucleic Acids Res. 1987; 15: 7709-7724Crossref PubMed Scopus (36) Google Scholar, 24Economides K.D. Capecchi M.R. Development (Camb.). 2003; 130: 2061-2069Crossref PubMed Scopus (132) Google Scholar, 25Mirosevich J. Gao N. Matusik R.J. Prostate. 2005; 62: 339-352Crossref PubMed Scopus (74) Google Scholar), were detected at highest levels in the ventral samples. Likewise, the β-microseminoprotein gene was highly expressed in ventral and dorsolateral prostate (26Thota A. Karajgikar M. Duan W. Gabril M.Y. Chan F.L. Wong Y.C. Sakai H. Chin J.L. Moussa M. Xuan J.W. J. Cell. Biochem. 2003; 88: 999-1011Crossref PubMed Scopus (15) Google Scholar). The seminal vesicle protein secretion 2 (27Kwong J. Lui K. Chan P.S. Ho S.M. Wong Y.C. Xuan J.W. Chan F.L. Prostate. 2003; 56: 81-97Crossref PubMed Scopus (12) Google Scholar) and 5 (28McDonald C.J. Eliopoulos E. Higgins S.J. EMBO J. 1984; 3: 2517-2521Crossref PubMed Scopus (20) Google Scholar) genes were expressed in dorsolateral prostate with the highest level in epithelium, consistent with the known expression patterns. The tranglutaminase 4 (29Ho K.C. Quarmby V.E. French F.S. Wilson E.M. J. Biol. Chem. 1992; 267: 12660-12667Abstract Full Text PDF PubMed Google Scholar) gene had high expression in anterior and dorsolateral lobes. The vimentin and α-actin genes are known to be expressed in the stroma, whereas the E-cadherin and cytokeratin 8 genes are expressed in the secretory epithelium; these gene products were highest in the corresponding tissue type. p63 and the cytokeratin 5 and 14 genes, which are expressed in basal epithelial cells, were generally higher in the epithelium. The neuroendocrine cell markers chromogranin A and synaptophysin (3Shappell S.B. Thomas G.V. Roberts R.L. Herbert R. Ittmann M.M. Rubin M.A. Humphrey P.A. Sundberg J.P. Rozengurt N. Barrios R. Ward J.M. Cardiff R.D. Cancer Res. 2004; 64: 2270-2305Crossref PubMed Scopus (488) Google Scholar) were expressed at higher levels in the stroma. Comparison between Epithelium and Stroma—A list of differentially expressed genes was identified by comparing the expression profile of epithelium to stroma of the ventral, dorsolateral, or anterior prostate (supplemental Table I). Epithelium and stroma had distinctive expression patterns as illustrated by hierarchical clustering and principal component analysis (Fig. 3). In general, similar tissues from both the C57BL and 129/SvEv mouse strains clustered together. Epithelial tissues of the same lobe from different strains of mice had the closest gene expression pattern. Dorsolateral epithelium showed an expression pattern distinct from anterior and ventral prostate epithelium, with a greater number of genes expressed at higher levels in dorsolateral epithelium. Stroma tissues of the same lobe from different mouse strains also had a similar gene expression pattern. Anterior and ventral prostate stroma had a closer expression pattern than dorsolateral stroma as well. Of the differentially expressed genes, a vast majority was expressed at higher levels in the stroma (TABLE ONE). Gene ontology mapping analysis indicated that these genes are implicated in biological processes such as cell adhesion, muscle development, and contraction, in biological functions such as structural constituent of cytoskeleton, actin binding, and extracellular matrix constituents, and in biological components such as sarcomere and extracellular matrix collagen (Fig. 4). These categories reflect products of the two major cell types in prostate stroma, i.e. fibroblast and smooth muscle cells. Among the epithelial samples, the dorsolateral lobe had the greatest number of transcripts expressed at high levels.TABLE ONEDifferential gene expression between epithelium and stroma in mouse prostate glandsTissueAPDLVPEaE, epithelium.SbS, stroma.ESESNumber of higher expressed transcriptscE/S or S/E>2;E-S or S-E>100.3223361940115Gene ontology noded≥5 genes in each node. Cell adhesion✓✓✓ Muscle contraction✓✓ Muscle development✓✓ Extracellular matrix structural constituent conferring tensile strength activity✓✓✓ Structural constituent of cytoskeleton✓✓✓ Calcium ion binding✓✓✓ Actin binding✓✓ Sarcomere✓✓ Actin cytoskeleton✓✓ Collagen✓✓✓a E, epithelium.b S, stroma.c E/S or S/E>2;E-S or S-E>100.d ≥5 genes in each node. Open table in a new tab Comparison among Anterior, Dorsolateral, and Ventral Lobes—Differentially expressed genes were identified by comparing expression profiles of the ventral, dorsolateral, and anterior prostate after combining data from stroma and epithelia (supplemental Table II). The AP had a higher expression of enzymes with hydrolase activity (e.g. cathepsin C, complement component 1 r and s, pancreatic lipase-related protein 1 and 2, matrix metalloproteinase 7 genes, RIKEN cDNA 4921509F24). The DLP presented a higher level of molecules related to the microtubule-based process and microtubule cytoskeleton (e.g. α-tubulin 1 and 2, β-tubulin 3 and 5, kinesin 2 and 5c, and microtubule-associated protein tau genes). The VP showed an elevated level of oxidoreductase activity (e.g. pyrroline-5-carboxylate synthetase, aldehyde dehydrogenase family 1 (subfamily A1 and A7), carbonyl reductase 2 and 3, cytochrome P450 2f2, and hypothetical protein MGC11688 genes) as well as enzyme inhibitor activity (e.g. tissue inhibitor of metalloproteinase 4, stefin A1, procollagen type VI α 3, and annexin A1 genes). The ventral, dorsolateral, and anterior prostate had distinctive expression patterns as illustrated by hierarchical clustering (Fig. 5A). The same lobe from different mouse strains was most similar, and the AP and DLP shared a closer expression pattern compared with the VP (Fig. 5A). Comparison between Mouse Prostate Lobes and Human Prostate Peripheral Zone—196 ortholog genes were identified from the mouse 430 chip set and the human U95Av2 chip (Affymetrix). Model-based expression indexes were normalized within a given sample, and expression patterns were analyzed by hierarchical clustering and principal component analysis. Gene expression pattern in the peripheral zone of the human prostate appeared to be closest to that of the mouse dorsolateral lobe (Fig. 5B). This supports the hypothesis that the mouse dorsolateral lobe is equivalent to the human prostate peripheral zone (33Price D. Natl. Cancer Inst. Monogr. 1963; 12: 1-27PubMed Google Scholar). Expression of Osteogenic Genes in Prostate—Many of the genes that were differentially expressed between stroma and epithelium are known to play important roles in osteogenesis (TABLE TWO). Noticeably, all of these genes were expressed at higher levels in the stroma. The Cxcl12 chemokine and its receptor Cxcr4, connective tissue growth factor Ctgf, and Sparc (osteonectin) were further studied by immunohistochemistry in normal and tumor prostate of the Pten knock-out mice. Tumor suppressor gene Pten inactivation occurs frequently in human prostate tumor cells, and prostate-specific deletion of the Pten gene in mice leads to hyperplasia, intraepithelial neoplasia, invasive carcinoma, and metastasis, recapitulating the steps of prostate cancer development in humans (30Wang S. Gao J. Lei Q. Rozengurt N. Pritchard C. Jiao J. Thomas G.V. Li G. Roy-Burman P. Nelson P.S. Liu X. Wu H. Cancer Cell. 2003; 4: 209-221Abstract Full Text Full Text PDF PubMed Scopus (873) Google Scholar). Anterior, dorsolateral, and ventral prostate lobes were dissected from 8-week-old Pten+/+, Pten+/- and Pten-/- mice, embedded in paraffin, and used for immunohistochemistry. Prostates of the Pten+/+ and Pten+/- mice were histologically normal, whereas prostates of the Pten-/- mice all developed carcinoma. In normal prostate, consistent with the microarray result, higher expression levels of Ctgf, Sparc (osteonectin), and Cxcl12 were seen in the stroma than in the epithelium (Fig. 6). Ctgf and Sparc were highly expressed in stromal cells (Fig. 6A), whereas strong immunoreactivity for Cxcl12 was found throughout the stroma, including the extracellular matrix (Fig. 6B). Sparc and Cxcl12 were also detectable in epithelial cells with an apical staining (Fig. 6). Normal prostate glands from Pten+/+ and Pten+/- had identical expression patterns (data not shown). Interestingly, in tumor prostate tissues, stromal Ctgf, Sparc, and Cxcl12 seemed to be reduced or lost, whereas they appeared to be increased in epithelial cells (Fig. 6). Expression of Cxcr4, the receptor for Cxcl12, was not detectable by DNA microarray or by immunohistochemistry in normal prostate glands. Strikingly, 60-70% of anterior and dorsolateral glands of tumor prostate expressed Cxcr4 (Fig. 6B). Ventral prostate glands also expressed Cxcr4, although to a variable extent among individual mice.TABLE TWOA list of "bone-related" molecules expressed in mouse prostate stroma cellsProbe setGeneTitleAccession no.p valueAPaFold higher expression in stroma than in epithelium.DLaFold higher expression in stroma than in epithelium.VPaFold higher expression in stroma than in epithelium.FunctionRef.1437889_x_atBgnBiglycanAI9318626.709E-102.913.281.78KO leads to osteoporosis1448323_a_atBC0195029.855E-122.913.311.83401416405_atBC0195024.072E-092.312.921448823_atCxcl12Chemokine (C-X-C motif) ligand 12BC0066400.00042793.861.982.65KO leads to bone marrow myelopoiesis411416953_atCtgfConnective tissue growth factorNM_0102172.518E-163.653.652.59Osteogenesis421424007_atGdf10Growth differentiation factor 10 (bone morphogenetic protein 3B)L421140.00204573.433.451.99Osteogenesis431448416_atMglapMatrix γ-carboxyglutamate (Gla) proteinNM_0085970.00067862.972.092.15Bone mineralization441419663_atOgnOsteoglycin (osteoinductive factor)BB5420510.0007097.254.521.95Bone formation451423606_atPostnPeriostin (osteoblast specific factor)BI1105653.365E-089.973.542.9Bone formation461448392_atSparcSecreted acidic cysteine-rich (glycoprotein (osteonectin)NM_0092420.00040555.152.412.59Bone calcification471416589_at8.022E-063.943.151.8a Fold higher expression in stroma than in epithelium. Open table in a new tab FIGURE 6Immunohistochemistry of osteogenic molecules in the normal and tumor prostate. Normal prostate tissues from Pten+/+ and Pten+/- and carcinoma tissues from Pten-/- mice were used for immunostaining of Ctgf, Sparc, Cxcl12, and Cxcr4. Tissues from Pten+/+ and Pten+/- prostate had similar expression patterns (data not shown). Bars indicate the scale. A, expression of Ctgf and Sparc proteins in normal (Pten+/+) and tumor (Pten-/-) prostate. B, expression of Cxcl12 and Cxcr4 proteins in normal (Pten+/+) and tumor (Pten-/-) prostate. Cxcr4-positive glands were counted under a microscope by two individuals. Insets, close-up views.View Large Image Figure ViewerDownload Hi-res image Download (PPT) Gene Expression Pattern in Different Mouse Prostate Lobes and Relation to Human Prostate—The human prostate is architecturally different from the mouse prostate, being composed of central, transition, and peripheral zones aggregated in a compact manner (3Shappell S.B. Thomas G.V. Roberts R.L. Herbert R. Ittmann M.M. Rubin M.A. Humphrey P.A. Sundberg J.P. Rozengurt N. Barrios R. Ward J.M. Cardiff R.D. Cancer Res. 2004; 64: 2270-2305Crossref PubMed Scopus (488) Google Scholar). Over 80% of human prostate carcinomas arise from the peripheral zone (31McNeal J.E. Cancer. 1969; 23: 24-34Crossref PubMed Scopus (280) Google Scholar, 32Sakr W.A. Grignon D.J. Anal. Quant. Cytol. Histol. 1998; 20: 417-423PubMed Google Scholar). Different transgenic mice have been used to model human prostate cancer in which tumors predominantly form in different lobes (1Kasper S. Smith Jr., J.A. J. Urol. 2004; 172: 12-19Crossref PubMed Scopus (43) Google Scholar, 2Abate-Shen C. Shen M.M. Trends Genet. 2002; 18: S1-S5Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar). Based on anatomic studies, Price (33Price D. Natl. Cancer Inst. Monogr. 1963; 12: 1-27PubMed Google Scholar) had proposed in 1963 that the mouse dorsolateral lobe was equivalent to the human prostate peripheral zone. Our observation that the gene expression pattern in the peripheral zone of the human prostate was closest to the mouse dorsolateral lobe constitutes the first molecular evidence in support of this hypothesis. Separate stroma and epithelium data were not available for the human prostate in the Harvard array set (11Singh D. Febbo P.G. Ross K. Jackson D.G. Manola J. Ladd C. Tamayo P. Renshaw A.A. D'Amico A.V. Richie J.P. Lander E.S. Loda M. Kantoff P.W. Golub T.R. Sellers W.R. Cancer Cell. 2002; 1: 203-209Abstract Full Text Full Text PDF PubMed Scopus (2035) Google Scholar). Thus, for comparison with the human peripheral zone, mouse prostate arrays were digitally combined by lobe. Therefore, our analysis of gene expression patterns between the mouse dorsolateral lobe and the human peripheral zone does not distinguish the relative contributi