Background
Prostate stromal mesenchyme fibromuscular cells provide a regulatory extracellular matrix and direct epithelial differentiation and development through growth factors and androgen stimulation [
1]. The critical role stromal cells play in prostate development has been demonstrated by co-implantation in animals of stem cells and stromal cells to achieve functional glandular development [
2‐
7]. Although prostate cancer is epithelial in origin, there is a growing body of evidence suggesting that the stromal microenvironment plays a significant role in the cancer process [
8‐
12]. Prostate tumor-associated or 'reactive' stroma is characterized by a decrease in smooth muscle cell differentiation and an increase in myofibroblasts and fibroblasts, with characteristics of a wound repair stroma [
13]. Defining the gene expression changes in the stroma of prostate cancer has been the focus of several recent studies and is an important step in defining the underlying mechanisms of stromal-epithelial interaction in cancer. Previous studies have characterized gene expression profiles of tumor-associated stromal cells isolated by laser-capture microdissection (LCM) [
14,
15] and cultures established from histologically confirmed cancer tissues [
16]. These studies have identified genes that are potentially involved in processes such as proliferation and angiogenesis. Current thought is that the tumor-associated stroma always co-exists with cancer [
17], and that it may contribute to the gain of metastatic potential by tumor cells and the progression towards androgen-independence [
12].
In this study, we sought to identify genes specific to prostate stromal cells that might function in organ specific stromal induction of epithelial development, and to isolate viable stromal cell populations associated with cancer by magnetic cell sorting (MACS) for gene expression analysis and comparison between these cells and their normal counterpart [
18‐
20]. CD90/THY1 is a cell surface molecule expressed in a wide variety of cells including stem and progenitor cells [
21‐
27]. It is thought to be involved in cell recognition, adhesion, and lymphocyte activation [
26]. Elevated expression of CD90 has been found in the stromal cells of primary prostate cancer [
28]. Previously, CD90
hi cells isolated from primary stromal cell cultures of prostate cancer were shown to differentially express several genes associated with tumor-promotion [
16]. Here, we used differential expression of CD90 to isolate viable CD90-expressing stromal cells directly from prostate cancer specimens for gene expression profiling and comparison to normal tissue stromal cells. We also used differential expression of CD13 in the bladder stroma to isolate viable CD13-expressing stromal cells from bladder cancer specimens for further comparison. To date there has not been an established marker comparable to CD90 in prostate stroma that differentiates bladder tumor-associated stroma from normal. A population of CD13
+ cells in the so-called superficial lamina propria was regarded here as the prostate-equivalent bladder stromal cells because of its proximity to the urothelium [
20]. These profiles provide important cell-type specific gene expression data for future
in vitro differentiation and development studies to compare cancer-associated and normal tissue stromal cells. We used cell sorting rather than LCM because cell sorting results in a viable population that could subsequently be grown in cell culture whereas LCM cannot. Identification and isolation of a viable, sufficiently pure, cancer-associated stromal cell population from tumor specimens will provide an essential research tool for the study of prostate carcinogenesis.
Methods
Tissue specimens
The methods of tissue collection, expression data generation and analysis used in this study have been published previously [
19,
29,
30]. The tissue samples consisted of prostate tissue specimens obtained from 13 patients undergoing radical prostatectomy under approval by the University of Washington Institutional Review Board. The same approach was used for both cancer-free and cancer-enriched (where at least 85% of the cells in the corresponding frozen section were of cancer) samples. Upon receipt of a radical specimen, 3-mm thick transverse sections were made of the prostate after inking the exterior surface. Cancer-free samples, weighing between 2 and 10 g, were harvested primarily from the anterior aspect of the prostate (transition zone) as described [
19,
20,
31]. Corresponding frozen sections were histologically assessed to confirm the specimens were free of cancer. Cancer-enriched samples, weighing at least 0.1 g, were dissected from the opposing aspect of the non-fixed section adjacent to the block of tissue that had been sectioned. To minimize possible RNA degradation, organs resected in the operating room were immediately submerged in ice-cold saline solution. The pathology characteristics of the tumors from which stromal cells were obtained were as follows - 08-021: Gleason 5+4, T3a, 4.5 cc tumor volume; 08-028: Gleason 3+4, T2c, 2.5 cc; 08-032: Gleason 4+4, T3b, 27 cc. Whole tissue transcriptomes were generated from the following tumor samples - 05A: Gleason 3+4, T3a, 6 cc; 05B: Gleason 3+4, T2c, 3.4 cc; 05C: Gleason 4+5, T3a, 3 cc; 05D: Gleason 4+5, T3aN1, >5 cc; 05E: Gleason 3+4, T2a, 2.5 cc; and their matched non-cancer samples.
To obtain bladder stromal cells for analysis, tissue specimens were obtained from cystoprostatectomy specimens. Transverse sections of intact urethra and both ureters were taken to assess the surgical margins. After palpating the bladder externally to locate any masses and inking grossly concerning lateral surgical margin(s), the bladder was opened, avoiding transecting the tumor. Regions of bladder mucosae and wall that appeared grossly normal both visually and by palpation were identified as "normal" urinary bladder. An approximately 2 × 2 cm portion of bladder wall with minimal perivesicle fat was excised and cut into several pieces. Two pieces were histologically characterized to verify that the tissue collected was free of either invasive or in situ carcinoma. The others were used for cell sorting. The pathology characteristics of the tumor from which CB stromal cells were obtained was as follows: specimen 07A contained was a high-grade urothelial carcinoma, stage pT2N0.
For cell sorting, the collected specimens were processed within hours. The tissue was rinsed with Hanks balanced salt solution (HBSS) and minced for enzymatic digestion overnight at room temperature with 0.2% collagenase type I (Invitrogen, Carlsbad, CA) in RPMI-1640 media supplemented with 5% FBS and 10
-8 M dihydrotestosterone or serum-free media on a magnetic stirrer. The resultant cell suspension was filtered with 70-μm Falcon cell strainer, diluted with an equal volume of HBSS, and aspirated with 18-gauge needle. The single cell preparation was partitioned into stromal and epithelial fractions on a discontinuous Percoll density gradient (Amersham Pharmacia, Piscataway, NJ) as described previously [
32,
33].
Western blot analysis of digestion media
HBSS-diluted tissue digestion media was centrifuged and the supernatant was collected. Protein concentrations were measured using Bradford Assay (BioRad, Hercules, CA). Sample buffer and 0.1 M DTT were added to amount of media containing 60 μg protein. The samples were heated to 70° for 10 min, electrophoresed on 4-20% gradient SDS-polyacrylamide gel (BioRad), and electrotransferred to PVDF membrane (Hybond-P, Amersham). The membrane was immersed in 5% nonfat dry milk in PBS-Tween for 30 min, and probed with tissue inhibitor of metalloproteinase 1 (TIMP1) antibody (1:500; MAB3300, Millipore, Temecula, CA) or CD90 antibody (1:500; 5E10, BD-PharMingen, San Diego, CA) for 60 min, followed by horseradish peroxidase conjugated anti-mouse IgG. After washing, the membrane was incubated with Luminol (Santa Cruz Biotechnology, Santa Cruz, CA) and immunoreactive bands were visualized using Biomax MR light film (Kodak, Rochester, NY). Prostate specific antigen (PSA) antibody (1:1000; A67-B/E3, Santa Cruz) was used for loading control.
MACS cell isolation
Cell types were sorted using monoclonal antibodies specific for tumor-associated prostate stromal cells (CD90), tumor-associated bladder stromal cells (CD13) and normal bladder stromal cells (CD13) with MACS in the same manner as was previously used to generate the normal prostate stromal cell (CD49a) transcriptome [
19]. The gradient-purified stromal cell fraction was resuspended in 100 μl 0.1% bovine serum albumin (BSA)-HBSS, and CD90-Phycoerythrin (PE) mouse monoclonal antibody (1:20, 5E10, BD-PharMingen) or CD13-PE (1:20, WM15, BD-PharMingen) added for 15 min at room temperature in the dark. The reaction was stopped by 1 ml 0.1% BSA-HBSS and centrifugation. The labeled cells were resuspended and 15 μl paramagnetic microbead conjugated anti-PE antibody (Miltenyi Biotec, Auburn, CA) was added for 15 min. After incubation the positive and negative cells were separated by AutoMACS cell sorter (Miltenyi Biotec) using double positive sort program.
FACS analysis of sorted cells
Aliquots of positive and negative cell fractions were analyzed by fluorescence activated cell sorting (FACS) (Becton Dickinson, Mountain View, CA) to monitor the sort efficiency; only >85% CD90
+ and CD13
+fractions were used for microarray experiments. The purity level was chosen based on our own observations (unpublished data) and studies by Szaniszlo
et al., that showed that the transcriptome of a 75% pure sorted cell population is largely identical to that of a 100% pure population [
34]. The sorted stromal cells were pelleted by centrifugation and lysed immediately in RLT (Qiagen, Valencia, CA). Total RNA was extracted for gene expression analysis using RNeasy (Qiagen).
Gene expression profiling on Affymetrix DNA microarrays
Quality and concentration of RNA were determined using Agilent 2100 Bioanalyzer and RNA Nano or Pico Labchip (Agilent Technologies, Santa Clara, CA). Only RNA samples that were of sufficient concentration and showed no degradation as evidenced by distinct ribosomal bands at 18S and 28S were used for microarray experiments using the Human Genome U133 Plus 2.0 GeneChips (Affymetrix, Santa Clara, CA). The U133 Plus 2.0 array contained probesets representing 54,675 genes, splice variants, and ESTs. The GeneChips were prepared, hybridized, and scanned according to the protocols provided by Affymetrix (P/N 702232 Rev. 2). Briefly, 200 ng of RNA was reverse transcribed with poly (dT) primer/T7 promoter, and the cDNA was made double-stranded.
In vitro transcription was performed to produce unlabeled cRNA. Next, first-strand cDNA was produced with random primers, and the cDNA was made double-stranded.
In vitro transcription was performed with biotinylated ribonucleotides, and the biotin-labeled cRNA was hybridized to the GeneChips. The chips were washed and stained with streptavidin-PE using Affymetrix FS-450 fluidics station. Data was collected with Affymetrix GeneChip Scanner 3000. A total of 15 array datasets were obtained from the following sample types obtained from 10 patients: two CD90
+ prostate tumor-associated stromal, two CD13
+ normal bladder stromal, one CD13
+ bladder tumor-associated stromal, and five whole tissue prostate cancer and five normal tissue from matched pairs. Additionally, 8 array datasets were obtained from the following sample types obtained from 7 patients: five CD49a
+ normal prostate stromal (published previously [
19]), and three CD26
+ prostate cancer (with one sample analyzed twice, Pascal
et al., submitted). All datasets have been deposited in GEO with the following accession number: GSE17906.
Microarray data analysis
Differential gene expression was determined by HTself, a self-self based statistical method for individual microarrays [
35]. All possible combinations of pair-wise comparisons among experiments were taken to create sets of ratios. The test used virtual self-self experiments to derive intensity-dependent cutoffs. Accordingly, a probeset was considered significantly differentially expressed if all its log-ratio combinations were outside the 99.9% credibility cutoff. The computational analysis results were verified by dataset query of known differentially expressed genes. Pathway analysis of selected genes was done with KEGG. Functional and ontology enrichment analysis was performed using the DAVID web-based tool [
36].
Gene expression validation
Reverse transcriptase polymerase chain reaction (RT-PCR) and quantitative real-time PCR (qPCR) were used to validate expression obtained by DNA arrays. For each cell/tissue sample, 1 μg RNA was reverse transcribed with Superscript II RT (Invitrogen) at 50° for 50 min followed by 10 min at 70°. Gene-specific primers for PCR (Additional file
1) were designed to produce amplicons of 100-650 bp in size. PCR was carried out at 95°/30 s, 55°/30 s, 72°/1 min for 35 cycles. PCR products were resolved on 2% agarose gels. Smooth muscle actin (ACTA2), ribosomal protein RPL0 and GAPDH served as internal references for the various PCR experiments. Results were derived from specimens different from those used for transcriptome analysis.
Discussion
In prostate cancer, both the epithelial and stromal compartments can be distinguished by CD immunostaining from their normal/benign counterpart cell types [
28]. CD immunostaining can also distinguish stromal cells of the prostate and the bladder [
20]. Our data show that the stromal cell type in cancer differs from that in normal/benign in the expression of a significant number of genes, both down-regulated and up-regulated. A recently reported genome-wide analysis showed no somatic DNA changes in breast and ovarian tumor-associated stromal cells [
48] so that DNA mutation could not likely contribute to the expression alteration. With expression alteration, the influence of CP stromal cells on epithelial differentiation and function is consequently different from that of NP stromal cells. The down-regulation of markers associated with SMC differentiation and of organ-specific genes suggests a loss of normal stromal signaling. We recently developed a co-culture system in which NP stromal cells were shown to induce differentiation of an embryonal carcinoma stem-cell line into a cell type with stromal cell gene expression, including the up-regulation of CNTN1, as well as that of epithelial markers [
49]. This system could allow us to further explore the role of genes identified here in cell-cell interaction. Bladder stromal genes could be similarly tested. The involvement of stromal cells in cancer is a feature of the tissue organization theory of carcinogenesis [
50], which equates cancer to inborn errors of development. Cancer is characterized by disruption of reciprocal intercellular signaling that maintains tissue organization in repair and turnover. For example, irradiation of breast stroma caused tumor formation from implanted non-irradiated mammary epithelial cells while non-irradiated stroma did not [
51]. Of the down-regulated genes, PENK is a hormone known to function in development [
52]. CNTN1 is known to mediate cell surface interactions in the development of the nervous system by signaling between axons and myelinating glial cells [
53] and has been shown to promote cellular adhesion and invasion of lung cancer cells [
54]. Also, CNTN1 has been shown to promote differentiation in CD90+ bone marrow stromal cells [
55]. Both PENK and CNTN1 have been detected in the prostate in other studies [
52,
56], and cancer down-regulation of PENK was evident in published datasets [
52]. SPOCK3 encodes a secreted protein involved in diverse steps of neurogenesis [
57] and has been shown to inhibit tumor invasion [
57]. MAOB catalyzes oxidative deamination of biogenic and xenobiotic amines with important roles in the metabolism of neuroactive and vasoactive amines in the central nervous system and peripheral tissues [
58]. CNTN1 was detected here all NP samples analyzed, down-regulated in CP, and not detected in metastasis. A similar pattern is shown by MAOB and SPOCK3 to a lesser degree. MAOB was found to be expressed in lymph node metastases. The down-regulation of organ-specific genes appears to be a feature also of bladder cancer-associated stromal cells. Future analyses such as transcriptome determination of the various cell types in the bladder will be required to more fully understand the role of stromal-epithelial interaction in bladder carcinogenesis. TRPA1 functions in signal transduction and growth control, HSD17B2 may have a role in bone development, and SALL1 a role in kidney development. Interestingly, the bladder stromal genes SALL1 and IL24 appear to be expressed in CP stromal cells.
In addition, these cancer-associated stromal cells could also be the source of potential cancer biomarkers consequential to their production of secreted/extracellular protein products. For each cancer epithelial cell, there are about 5-10 associated CP stromal cells (True
et al., submitted); as a result, a higher concentration of stromal-derived markers such as those identified here would be expected. The cell surface-anchored CD90 was detected in tissue digestion media [
37] as well as in media of cultured stromal cells [
59]. The increase in CD90 protein in cancer tissue was measured by quantitative proteomics [
37] and was detectable by Western blot analysis. Therefore, secreted proteins like SFRP4 with a 12-fold increase in expression in tumor-associated stroma could potentially serve as markers for the presence of cancer. With appropriate antibodies, ELISA-type of assays could be developed for these marker proteins. Alternatively, multiplex-type of assays could be designed to detect these genes. The abnormal expression of Wnt pathway member SFRP4 and WT1 in stromal cells and that of other members in the epithelial cancer cells (Pascal
et al., submitted) suggest a significant role of Wnt signaling in prostate carcinogenesis.
Two published datasets of tumor-associated
vs. normal tissue stromal cells were available for comparison to ours. Genes scored as up-regulated (no down-regulated ones were found) by Richardson
et al. [
14] did not match those by Zhao
et al. [
16]. Expression of only a subset of these genes was scored as up-regulated in our datasets. The method of cell isolation could be a factor. Richardson
et al. used laser-capture microdissection, but without immunostaining (e.g., with CD90) it is difficult to avoid capturing any NP stromal since the cancer-associated stroma does not extend beyond 10 cells from the tumor glands (True
et al., submitted). Zhao
et al. used cultured CP and NP stromal cells. Cell culturing is known to alter gene expression (e.g., expression of epithelial CD markers [
31]). The datasets of Joesting
et al. [
60] were also generated from cultured cells but did not match well with those of Zhao
et al. In addition, incorrect array probesets and non-agreement between array signals and protein expression determined by immunostaining could compromise data quality [
18]. Our finding that genes involved in SMC differentiation and organ specificity were down-regulated in CP stromal cells is in line with the hypothesis that stromal cell differentiation in tumor is abnormal. The up-regulation of CD90 is notable because CD90 is a stem cell marker, and CD90
+ fibroblasts are considered to be a primitive cell type. Attenuated smooth muscle differentiation and down-regulation of developmental genes such as CNTN1, SPOCK3 and MAOB are an indication of such a possibility. Furthermore, the possibility exists that these tumor-associated stromal cells are similar to the normal stromal cells adjacent to the basal epithelium. Future comparison of the tumor-associated stromal cells to CD90
+ normal stromal cells may provide even greater insight into the role of stromal-epithelial interaction and how it might be disrupted in prostate carcinogenesis.
The cancer stroma contains not only fibromuscular cells but also CD45
+ white blood cells, CD31
+ endothelial cells of blood vessels, and nerve elements. These cell types could conceivably have a function in the cancer process as well [
61,
62]. Similar experimental analyses can be used to characterize and compare these cell types isolated from tumor and normal tissue.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
LEP and YAG conceived and carried out experiments, analyzed data, and drafted the manuscript. RZNV analyzed microarray data and helped write the manuscript. TKT assisted with study design, and helped write the manuscript. LDT provided pathology analysis, assisted with study design and helped write the manuscript. LSP performed flow cytometry and cell culture. AAC performed Western blot experiments. AYL conceived experiments, study design and coordination, and revised the manuscript. All authors read and approved the final manuscript.