Estrogen receptor α in cancer-associated fibroblasts suppresses prostate cancer invasion via modulation of thrombospondin 2 and matrix metalloproteinase 3

Abstract

The prostate cancer (PCa) microenvironment contains active stromal cells known as cancer-associated fibroblasts (CAF) that may play important roles in influencing tumor progression. Here we studied the role of CAF estrogen receptor alpha (ERα) and found that it could protect against PCa invasion. Immunohistochemistry on prostatectomy specimens showed that PCa patients with ERα-positive stroma had a significantly lower risk for biochemical recurrence. In vitro invasion assays further confirmed that the stromal ERα was able to reduce PCa cell invasion. Dissection of the molecular mechanism revealed that the CAF ERα could function through a CAF–epithelial interaction via selectively upregulating thrombospondin 2 (Thbs2) and downregulating matrix metalloproteinase 3 (MMP3) at the protein and messenger RNA levels. Chromatin immunoprecipitation assays further showed that ERα could bind to an estrogen response element on the promoter of Thbs2. Importantly, knockdown of Thbs2 led to increased MMP3 expression and interruption of the ERα mediated invasion suppression, providing further evidence of an ERα–Thbs2–MMP3 axis in CAF. In vivo studies using athymic nude mice injected with CWR22Rv1 (22Rv1) PCa epithelial cells and CAF cells ± ERα also confirmed that mice coimplanted with PCa cells and CAF ERα+ cells had less tumor foci in the pelvic lymph nodes, less metastases, and tumors showed less angiogenesis, MMP3, and MMP9 (an MMP3 downstream target) positive staining. Together, these data suggest that CAF ERα could play protective roles in suppressing PCa metastasis. Our results may lead to developing new and alternative therapeutic approaches to battle PCa via controlling ERα signaling in CAF.

Introduction

Although the majority of prostate cancers (PCas) are epithelial in nature, it is becoming more and more apparent that the tumor microenvironment (TME) is vital for the transformation of normal epithelial cells to cancer cells (1). Cancer-associated fibroblast (CAF) cells are stromal cells which have been transformed either via the TME or injury (2). These cells have been shown to induce normal prostate cells to grow tumors in a tissue recombination model (3) as well as having a role in extracellular matrix (ECM) disruption and immune cell infiltration (4,5).

Estrogen has been studied extensively in PCa initiation and progression (6,7). Typically, estrogen actions in the prostate can be summarized as either growth inhibiting through estrogen receptor beta activation (8) and suppression of testosterone via the pituitary axis (9), or growth stimulating through estrogen receptor alpha (ERα) (10). Previous studies have looked at the stromal and epithelial roles of ERα in normal prostate development and in squamous metaplasia (11–13), however, the role of CAF ERα in PCa invasion has yet to be fully elucidated.

Estrogen sources are varied. In males, testosterone can be converted to estrogens. Adipose tissue can be a de novo source of estrogen (14,15), which is important as aging males are observed to have an increase in fat deposition. In addition to de novo synthesis by fatty tissue, environmental estrogens and phytoestrogens such as Bisphenol A (16), genestein (17) or polyfluorinated iodine alkanes (18) have all been shown to be active estrogenic compounds. With the increasing knowledge and prevalence of estrogenic compounds in the human body, there is a pressing need to elucidate the effects that estrogens may have in the PCa microenvironment. In the current study we determine that the presence of CAF ERα in the stroma is linked to a better clinical outcome. Cell line studies indicate that CAF ERα may exert its protective effects via modulation of an anti-angiogenic factor, thrombospondin 2 (Thbs2) and an ECM remodeling factor and matrix metalloproteinase family activator, matrix metalloproteinase 3 (MMP3). We examined these molecules in an in vivo model using orthotopic implantation of CAF cells infected with either vector or ERα mixed with PCa cells. We present here for the first time, to our knowledge, evidence that CAF ERα could play a protective role in cancer progression.

Materials and methods

Cell lines

PCa cell lines, 22Rv1, PC3 and TRAMP C1 were purchased from the American Type Culture Collection, Rockville, MD. C4-2 cells were a gift from Dr Chung (19). CAF cells were primary cultures from transgenic adenocarcinoma of the prostate mice in our lab. All cells were maintained in RPMI media with 10% fetal bovine serum and 1% Penicillin/streptomycin.

Lentiviral ERα transduction of CAF cells and firefly luciferase infection of 22Rv1

The ERα complementary DNA (cDNA) was cloned into the PmeI site of pWPI lentiviral vector to generate a lentivirus ERα expression construct. The 293T packaging cells were transiently transfected with pMD2.G and psPAX2 with pWPI or pWPI–ERα to produce lentiviral particles. The supernatant containing lentiviral particles, was collected 48h post-transfection of 293T cells. Polybrene was added to the culture media of CAF cells at a final concentration of 4mg/l. The lentiviral supernatant was filtered and added to culture media to transduce CAF cells for 48h. The viral vector or ERα transduced CAF cells were then subjected to antibiotic selection using 5mg/µl blasticidin (Invitrogen, Grand Island, NY) for 10 days. The firefly luciferase cDNA was constructed with pCDNA backbone with G418 antibiotics and was used to transduce 22Rv1 cells in a similar manner. The same procedure was followed for the construction of pWPI–MMP3, except the selection marker was neomycin (50 µg/ml), and BamH1 was used as the restriction enzyme.

Invasion assay

We first cultured CAF ± ERα cells (2×10⁶) in 10cm² plates. After cells adhered, we changed to fresh 5% charcoal dextran fetal bovine serum media. Then, 10nM 17ß estradiol (E2) was added and conditioned media (CM) were collected after 24h incubation. Before seeding PCa cells for invasion assay, we precoated transwells with matrigel (the concentration of matrigel was 0.2mg/ml, 100 μl into the transwell insert) and air dried overnight. We added 600 μl CM into each well of 24 well plates, then moved transwells (coated with matrigel) into each well. We seeded PCa cells (22Rv1, C4-2, or PC3) at 5×10⁴/150 μl per well in blank media. After 24h incubation, we washed, fixed and stained transwells. The number of cells per representative fields (~10 fields at ×100 magnification) were counted and averaged. The mean of three transwell inserts per group was compared.

Zymography

In 10cm² plates, 2×10⁶ CAF vector or ERα cells were seeded in blank media and 5ml of CM was collected after 24h. The protein was concentrated 25× and 40 µl CM mixed with 4× non-denaturing sample buffer before loading on an 8% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel containing casein (0.1%). The gel was run for 2h at 100 volts. The gel was renatured for 60min, developed for 24h and then stained with coomassie blue for 30min. After destaining, the gel was visualized in a Chemidoc XRS (Bio-Rad, Hercules, CA).

Chromatin immunoprecipitation assay

CAF vector or CAF ERα cells were treated with E2 for 24h and then fixed in 1% formaldehyde. The cells were lysed and sonicated to shear DNA to 200–1000bp in length. Lysates were diluted 10-fold into chromatin immunoprecipitation dilution buffer. Sepharose beads were used to clear the cell extracts in order to reduce the non-specific background. Mouse ERα antibody MC-20 (Santa Cruz Biotechnology, Dallas, TX) was added and allowed to precipitate overnight. Gamma binding G sepharose beads (Pharmacia Biotech) were used to collect the antibody–protein complex. The beads were washed and the beads transferred to a new tube where salt was added to reverse the protein–DNA crosslinks. Ethylenediaminetetraacetic acid and proteinase K were added to degrade the protein. Chromosomal DNA was eluted using a Gel Elution Kit. Using Dragon ERE Finder version 3 (http://datam.i2r.a-star.edu.sg/ereV3/) a putative estrogen response element (ERE) site was found ~4kb upstream of the transcription initiation site. Primers were designed flanking this ERE (GA-GATCA-CAC-TGACC-TC). Quantitative-PCR analysis was performed using the primers; sense primer: 5′-GCGTCTGATGCCTATGTTCTGAAAG-3′; antisense primer: 5′-GCAAGAGGTCCTGAGAGGATGAG-3′.

Thbs2 exogenous protein collection

Human embryonic kidney 293 cells were infected with lentiviral pWPI-puromycin-Thbs2 plasmid or empty vector. After 24h of lentivirus infection the media was collected and used to reinfect human embryonic kidney 293 cells. After confirmation of protein expression via green fluorescent protein reporter expression, media was changed to blank media for 24h and then collected. Media were concentrated (~10-fold), total protein was measured and 30 µg of total protein was used for Thbs2 western blotting. The remaining protein was treated equally to parental CAF cells.

Human tissue samples and immunohistochemistry

We retrieved 75 PCa specimens obtained by radical prostatectomy performed at the University of Rochester Medical Center. Appropriate approval from the Institutional Review Board at our institution was obtained prior to construction and use of the tissue microarray. The mean age of the patients at presentation was 59.9 years (range: 47–71 years) and the mean follow up after the surgery was 21.3 months (range: 1–38 months). None of the patients had received therapy with hormonal reagents, radiation or other anticancer drugs pre- or postoperatively prior to clinical or biochemical recurrence. Biochemical recurrence was defined as a single prostate-specific antigen level of ≥0.2ng/ml. Immunohistochemical staining was performed on the automated staining system, as described previously (20). Briefly, the sections (4 μm thick) were immunohistochemically labeled, using the primary antibody to ERα (E115 clone, 1:100 dilution, Epitomics, Burlingame, CA). All these stains were manually scored, separately in non-neoplastic epithelial cells, carcinoma cells and stromal cells (primarily fibroblastic cells) adjacent to the tumors, by one pathologist (H.M.) blinded to patient identity. German Immunoreactive Score was calculated by multiplying the percentage of immunoreactive cells (0% = 0; 1–10% = 1; 11–50% = 2; 51–80% = 3; 81–100% = 4) by staining intensity (negative = 0; weak = 1; moderate = 2; strong = 3). Cores with the immunoreactive score of 0 or 1 were considered negative (0) and those with the immunoreactive scores of 2–4, 6–8 and 9–12 were considered weakly positive (1+), moderately positive (2+) and strongly positive (3+), respectively.

Immunohistochemistry in mouse tissues

Immunohistochemistry staining was carried out as described previously (21,22). Sections were incubated with the following primary antibodies: anti-MMP9, anti-MMP3, anti-CD31 (AbCam, Cambridge, MA), and antifirefly (Santa Cruz Biotechnology) in 3% bovine serum albumin in phosphate-buffered saline overnight at 4°C followed by respective secondary antibodies. The stained slides were mounted and visualized by a bright-field microscope. Cells were quantified using 10 or more ×400 magnification images from anterior prostates from six or more mice/group. The total number of positive cells was divided by the total number of cells in the image and then multiplied by 100 to get a percentage of positive cells.

Western blotting

Cells were lysed in radioimmunoprecipitation assay buffer + 10% proteinase inhibitor, centrifuged and the soluble protein removed and quantified. 50 µg of protein was mixed with 4× loading dye and loaded onto a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis. After running, the gel was transferred to a polyvinylidene difluoride membrane overnight. Primary antibodies used were anti-MMP3 (AbCam, 1:1000), anti-Thbs2 (Santa Cruz Biotechnology, 1:100), anti-glyceraldehyde 3-phosphate dehydrogenase (Genetex 1:2000) and anti-ERα (Santa Cruz Biotechnology, 1:500). Blots were incubated for 1h at room temperature in primary antibody. After washing, the secondary antibody was applied for 1h, followed by exposure with an enhanced chemiluminescence reagent.

Protein array and densitometry analysis

CM from CAF vector or CAF ERα cells was collected 24h after incubation with E2. CM was used for protein array analysis as per the manufacturer’s instructions (R&D systems). Briefly, the membrane was blocked using the provided blocking buffer. Samples were prepared with the conjugating buffer and exposed to the membrane at 4°C overnight. The horseradish peroxidase buffer was added after washing and finally developed using Femto enhanced chemiluminescence reagent for varying time points. The time point used was just before the image showed saturation of pixels as determined by the Image Lab software (Bio-Rad). Densitometry analysis was carried out using Image J to quantify the pixel density of each spot and then normalize those values to the six control spots.

RNA extraction and quantitative real-time PCR analysis

Total RNA was extracted by Trizol reagent (Invitrogen, CA) according to the manufacturer’s instructions. RNAs (1 μg) were subjected to reverse transcription using Superscript III transcriptase (Invitrogen). The obtained cDNAs were used for qPCR using a Bio-Rad CFX96 system with SYBR green. Primers used were as follows: Thbs2 Forward 5'-GGGGACACTTTGGACCTCAAC-3' and Reverse 5'-GCAGCCCACATACAGGCTA-3'. MMP3 Forward 5'-TTAAA GACAGGCACTTTTGGCG-3', and Reverse 5'-CCTCGTATAGCCCAGAA CT-3' Serpin F1 Forward 5'-TCTGGGAAAGGGTTCACTTTACC-3', and Reverse 5'-GACACGCCATAGGGAGAGAAG-3'. Expression levels were normalized to the expression of glyceraldehyde 3-phosphate dehydrogenase RNA.

Orthotopic implantation

We mixed 22Rv1 cells that had been transduced with firefly cDNA and CAF±ERα cells (ratio of 22Rv1: CAF is 9:1). After anesthesia, the abdomens of 8-week-old male athymic nude mice were surgically opened in a sterile environment. 9 x 10⁵ 22Rv1 cells were mixed with 1 x 10⁵ CAF ERα+/- cells in Matrigel. The ratio of cells to matrigel was 1:1. Twenty microlitre of this mixture were injected into each lobe of the anterior prostate using a 25 gauge needle. The abdomens were then closed and sutured using 5-0 adsorbable sutures (Ethicon). Twelve animals were used per group. Mice were monitored by live-image system for tracking tumor growth and metastasis every other week. Tumors were harvested 12 weeks after injection. Tumor sizes, weights and metastasis to pelvic lymph node were compared. All mice experiments were under a protocol approved by the Institutional Animal Care and Use Committee of the University of Rochester Medical Center.

Statistical analysis

Values were expressed as mean ± standard deviation. Student’s t-test was used to calculate P-values. P-values were two sided and considered statistically significant when α < 0.05. Chi squared one-tailed test was used to evaluate metastatic load between vector and ERα orthotopically injected mice. Survival rates in patients were calculated by the Kaplan–Meier method and comparison was made by log-rank test. P-values <0.05 were considered significant.

Results

Patients with ERα-positive stroma show better outcomes

We immunohistochemically stained for ERα in a tissue microarray with 75 radical prostatectomy specimens and evaluated its expression in PCa and stromal cells. Three cases were inadequate for evaluation as there were no or small amounts of stromal cells. 42 out of 72 (37 cases, 1+; 5 cases, 2+) cases were positive for ERα in the tumor stroma. And 55 out of 72 (all 1+) and 31 out of 75 (all 1+) patients were positive for ERα in carcinoma cells and benign epithelial cells, respectively, yet its expression in tumor cells did not correlate with tumor recurrence (P = 0.952). There were no statistically significant correlations between ERα expression in PCa or stromal compartment and histopathologic features of the tumors including Gleason score, pathologic stage (pT) and lymph node metastasis. Nonetheless, Kaplan–Meier analysis coupled with log-rank test revealed ERα negativity in stromal cells strongly (P = 0.002) correlated with the risk of biochemical recurrence (Figure 1). None of the patients with ERα-positive stromal cells developed recurrence. These results suggest that CAF ERα may play a protective role in PCa progression.

Fig. 1.

Immunoreactivity to ERα in stromal cells correlates with patients’ outcomes. ERα expression was immunohistochemically assessed in a PCa tissue microarray consisting of 75 radical prostatectomy specimens. Kaplan–Meier analysis was performed according to the status of ERα expression in the stromal compartment, and comparison was made by log-rank test. Biochemical recurrence was defined as a single prostate-specific antigen level of ≥0.2ng/ml (P = 0.0020, log-rank test).

CAF ERα inhibits PCa cell invasion

Our data show that stromal ERα may play a protective role in PCa progression so we decided to determine the CAF ERα role in this phenotype. We first established a CAF cell line from the prostate stromal region of transgenic adenocarcinoma of the prostate mice. These cells were then stably transfected with ERα-cDNA (or its empty plasmid vector as control) (Figure 2A). We then applied the transwell system (Figure 2B) with the CM from CAF cells to study its influence on PCa cell invasion. As shown in Figure 2C and Supplementary Figure 1 (available at Carcinogenesis Online), CM from CAF cells were able to enhance human PCa PC3 cell invasion, and addition of ERα in CAF cells led to suppression of PC3 cell invasion. This invasion phenotype was not estrogen dependent as shown in Supplementary Figure 1 (available at Carcinogenesis Online). Similar results were obtained when we replaced PC3 cells with other PCa epithelial cells including human, androgen independent, C4-2 cells and mouse TRAMP C1 cells (Figure 2C). Quantification of those invasive cell numbers (Figure 2D–F) concluded that CAF ERα could play a protective role to suppress PCa invasion.

Fig. 2.

CAF ERα is able to suppress PCa invasion in an in vitro model. (A) CAF cells were created that contained either an ERα or empty vector plasmid. Western blot and quantitative-PCR analysis verify ERα expression in CAF cells (qPCR on right side panel). (B) In vitro model of transwell invasion assay. Transwell inserts were coated with a thin layer of matrigel compound and allowed to dry to create a barrier that PCa cells must cross in order to invade. CM from CAF cells with or without ERα was collected and placed into the lower chamber of the transwell plates. PCa cells were put into the top chamber and allowed to invade for a period of time. (C, D) PCa cell invasion is decreased when treated with CAF ERα CM. The transwell inserts were fixed and stained before light microscopy visualization. In those three PCa cell lines, invasion was significantly decreased when CAF ERα CM was used as the chemoattractant compared with PCa cells where CAF vector CM was used.

Thbs2 and MMP3 are CAF ERα targets

To determine the mechanism by which CAF ERα was able to exert its effects, we applied a protein array using CM from CAF vector and CAF ERα cells. There were 53 different cytokine/chemokine antibodies mounted on the array membrane (Supplementary Table 1, available at Carcinogenesis Online). After incubation, washing and application of the substrate, densitometry analysis results revealed that, among others, the protein levels of Thbs2 and MMP3 were changed in the CM collected from CAF ERα cells (Figure 3A). Addition of E2 in the culture medium was able to enhance the expression levels of Thbs2 and MMP3. However, this gene change was also seen when CAF cells were cultured in charcoal stripped fetal bovine serum medium, indicating that the genotype is not estrogen exclusive (Supplementary Figure 1, available at Carcinogenesis Online). We arbitrarily picked Serpin F1 as control since densitometry (Figure 3B) showed no change in protein level. Using qPCR analysis we further confirmed the messenger RNA (mRNA) of Thbs2 and MMP3 are selectively changed in CAF ERα cells (Figure 3C), suggesting these 2 proteins are ERα downstream targets in prostate CAF cells. Other targets showed inconsistent mRNA levels, were not related to our study or did not have enough literature support to justify further study (Supplementary Figure 2, available at Carcinogenesis Online). Thbs2 is a potent antiangiogenic factor (23) as well as a regulator to modulate the expression of MMP2 (24), a key player for PCa invasion (25). MMP3 may activate MMP9 (26), which is also a well-known player in PCa invasion. Both Thbs2 and MMP3 were selected as the downstream effectors to further investigate the CAF ERα roles in prostate TME.

Fig. 3.

Mechanism of CAF ERα suppression of PCa invasion via upregulation of Thbs2, downregulation of MMP3. (A) Protein array analysis shows that Thbs2 and MMP3 are downstream ERα targets. Protein array was performed using CM from CAF vector and ERα cells as the input. (B) Densitometric analysis, performed by normalizing protein levels to the control spots, revealed that Thbs2 was selectively upregulated in CAF ERα CM. Conversely, CAF ERα CM showed a marked decrease in secreted MMP3 levels. Serpin F1 was used as a negative control. (C) The quantitative PCR confirms protein expression. To see if the increase was due to either a change in import/export signaling or was a change at the transcriptional level, qPCR analysis was run. Not only was the secreted amounts of protein forms of both MMP3 and Thbs2 changed, but the transcriptional regulation of these genes was also altered consistent with the protein level. Serpin F1 was used as quantitative PCR negative control.

Disruption of the CAF ERα-mediated suppression of PCa cell invasion by Thbs2–shRNA

To test whether upregulation of Thbs2 was essential for the CAF ERα mediated invasive phenotype of PCa cells, shRNA of Thbs2 was infected into CAF ERα cells (Figure 4A). CM collected from these cells was introduced to PC3 PCa cells. Consistently, CM from CAF ERα+ cells was able to reduce PC3 invasion significantly. Supportively, knockdown of Thbs2 with Thbs2–shRNA led to a reversal of the CAF ERα protective roles to suppress PC3 cell invasion (Figure 4B and D).

Fig. 4.

Disruption of Thbs2 pathway results in rescue of ERα invasion phenotype. (A) CAF ERα cells were infected with lentivirus containing shThbs2. Verification of Thbs2 knockdown at the mRNA level was achieved using qPCR. (B, D) Knockdown of Thbs2 created much more invasive PCa cells. After verification of Thbs2 knockdown (B), CM was collected from vector+shLuc, ERα+shLuc, vector+shThbs2, ERα+shThbs 2 CAF cells (shLuc is used as control for shThbs2) CAF cells. The media was used as a chemoattractant in the invasion assay as described in Materials and methods. (C) Analysis of ERα binding to the promoter of Thbs2 using chromatin immunoprecipitation assay. Chromatin immunoprecipitation assay was performed in CAF vector or CAF ERα cells treated with 10nM E2 for 24h. Protein–DNA complexes were pulled down using the ERα antibody (MC-20). The 3rd and 4th lanes showed the equal input of genomic DNA prior to be pulled down by IgG or MC-20. The eluted DNA fragments were used to amplify the ERα protein bound sites. Quantification of the resultant PCR product with quantitative PCR is shown.

Further mechanism dissection found ERα could modulate Thbs2 at the transcriptional level. As shown in Figure 4C, ERα could bind to a putative ERE (GATCACACTGACC compared with canonical ERE GGTCAnnnTGACC) located ~4kb upstream of the Thbs2 transcription start site as shown via a chromatin immunoprecipitation in vitro binding assay.

CAF ERα cells reduce MMP3 expression

We then applied a casein sodium dodecyl sulfate–polyacrylamide gel electrophoresis zymography assay with CM from CAF vector or CAF ERα cells (CM collected from CAF vector or ERα cells transfected with an MMP3 overexpression plasmid was used as positive control), and results showed an obvious difference in MMP3 expression levels in CM collected from CAF vector compared with CAF ERα cells (Figure 5A). Additionally, western blot data also reveal a substantial decrease in MMP3 protein levels (Figure 5B). These data suggest that MMP3 protein was downregulated in an ERα dependent manner.

Fig. 5.

Mechanistic study of the ERα regulation of MMP3 and Thbs2. (A) Casein zymograpy results of CAF vector, ERα, vector + MMP3, and ERα + MMP3 CM. CM from different CAF cells was collected as before. We used 40 µl of media to run 8% casein zymography. After development, fixing, staining and destaining, bands were visualized using a Chemidoc. Densitometry analysis was performed using CAF vector CM as a control. (B) Western blot of MMP3 levels. CAF cells infected with ERα or ERα + MMP3, or an empty vector, were harvested and protein extracted. Fifty microgram of protein were run per lane and the membrane hybridized with MMP3, or glyceraldehyde 3-phosphate dehydrogenase antibodies, appropriate secondary antibodies, and then developed. (C) The reduced MMP3 expression is reversed in CAF ERα/shThbs2 cells. mRNA from CAF ERα/shLuc control and CAF ERα+/shThbs2 cells were collected as described in Materials and methods. Quantitative-PCR analysis was performed to detect mRNA levels of MMP3. (D) Thbs2 expression levels in human embryonic kidney 293 cell media. Eighty microgram of concentrated human embryonic kidney 293 +/− Thbs2 CM were loaded onto an 8% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel and immunoblotted using a Thbs2 antibody. (E,F) Protein and mRNA levels of MMP3 post Thbs2 treatment. Cell lysates from parental CAF cells that had been treated with or without Thbs2 were used to collect cellular protein, or underwent reverse transcription. These cells were used to determine the MMP3 mRNA or protein amount using western blot. (G) CAF cells partially lose their ability to attract PCa cells. Post Thbs2 treatment, CAF cell CM was unable to attract as many PC3 cells to invade compared with CAF cells which had not been treated with Thbs2. This difference was found to be significant (n = 3, *P < 0.05). (H) MMP3 overexpression leads to increased PCa cell invasion. CM from CAF Vector/ ERα or Vector/ ERα+MMP3 was used as a chemoattractant for the PCa invasion assay as described previously. PC3 cells were allowed to invade through a transwell insert coated with matrigel.

CAF ERα downregulated MMPs are modulated via a Thbs2 axis

To test the functional linkage of these two ERα downstream target genes, we examined whether knocked-down Thbs2 via shRNA–Thbs2 in ERα positive CAF cells would influence MMP3 expression. Supportively, our data showed that ERα downregulated MMP3 around ~50%, and knocking-down Thbs2 by shRNA could reverse the ERα effect and resulted in a significant induction of MMP3 (Figure 5C).

To determine if the converse was true, that expression of Thbs2 could modulate MMP3 levels, we produced exogenous Thbs2 protein by transfecting the Thbs2 plasmid into human embryonic kidney 293 cells (Figure 5D), collecting the CM and then treating parental CAF cells with CM from either vector or Thbs2 producing human embryonic kidney 293 cells. Concurrent with ERα expression, Thbs2 treated CAF cells showed a decrease in MMP3 expression levels at both the mRNA level and protein level (Figure 5E and F). The CAF cells treated with Thbs2 CM showed a significant decrease in their ability to affect the invasion of PCa PC3 cells compared with vector-CM-treated CAF cells (Figure 5G).

Together, our data indicate that there is a functional linkage of the ERα/Thbs2/MMP3 pathway in CAF cells.

Reversal of CAF ERα downregulation of MMP3 is able to reverse the invasion phenotype

We overexpressed MMP3 into CAF ERα cells (Figure 5B) in order to test the MMP3 effect on the invasion inhibition created by ERα. We used CM from these cells to induce PC3 cells to invade. We show here that MMP3 overexpression is able to significantly reverse the ERα inhibited PCa invasion (Figure 5H). This provides further evidence of MMP3’s important role in CAF ERα inhibited PCa invasion.

Mice orthotopically implanted with CAF ERα/PCa cells show a less invasive phenotype compared with CAF vector/PCa implants

Eight-week-old athymic nude mice were implanted with a mixture of CAF cells +/− ERα and 22Rv1 cells. The invasion phenotype was characterized by pelvic lymph node metastasis. Mice injected with 22Rv1/CAF vector cells showed larger lymph nodes with more metastatic foci (Figure 6A) than those in 22Rv1/CAF ERα cells. Overall examination of metastatic incidence in these mice also confirmed CAF ERα mice showed much less metastases (Figure 6B). To ensure those metastatic tumors were from 22Rv1 PCa cells, we also tagged injected 22Rv1 cells with a luciferase plasmid and results indicated that a significant reduction in lymph node luciferase staining with less tumor foci was seen in 22Rv1/CAF ERα implanted mice (Figure 6C and D). Although 22Rv1/CAF ERα tumors showed no difference in tumor size at time of harvested (Supplementary Figure 4, available at Carcinogenesis Online), earlier time points showed a marked increase in tumor growth (data not shown).

Fig. 6.

Orthotopic implantation of nude mice with CAF and PCa cells. (A) Histological analysis of mouse metastatic tumor tissue. Pelvic lymph nodes were removed from mice and sectioned for H + E staining. The metastatic foci are labeled yellow, orange star on vector ×40 picture denotes cutaway area of ×400 picture. (B) Quantification of metastatic load. Among mice that developed tumors the number of mice that had metastases was quantified. CAF ERα+ 22Rv1 injected mice had less metastasis than 22Rv1 cells mixed with CAF vector although this result was not found to be statistically significant (P = 0.14). (C, D) Less lymph node invasion in 22Rv1/CAF ERα+ implants by luciferase visualization. 22Rv1 PCa cells used for orthotopic injection were labeled with a luciferase construct. Postsacrifice luciferase signal was visualized using immunofluorescence. The amount of luciferase signal/cell number/area of lymph node was quantified and found to be significantly reduced (P < 0.05). (E) Reduced MMP9 and angiogenesis in 22Rv 1/CAF ERα+ coimplant tumors. Immunohistochemistry staining of MMP9 shows that in CAF ERα+ mice less intense MMP9 staining is seen and this staining is localized to the stromal cells. In CAF ERα- mice more intense MMP9 staining is observed and this MMP9 staining is more diffuse, showing in several epithelial cells as well. CD31 Immunohistochemistry was performed to stain endothelial cells. Reduced CD31 immunostaining signals indicated less endothelial cells and reduced angiogenesis in the 22Rv1 coinjected with CAF ERα+ cells than coimplant with CAF ERα−. Insets are 200% magnification (*, P < 0.05; **, P < 0.01).

Thbs2 and MMP3 are regulated by CAF ERα in in vivo mouse models

To further verify the obtained in vitro cell line data in our in vivo mouse model, we examined the expression of Thbs2 and MMP3 downstream targets CD31 and MMP9, respectively, as well as MMP3 expression. In PCa and CAF vector mice MMP9 and MMP3 are seen in a greater quantity and with stronger intensity than PCa and CAF ERα+ mice (Figure 6E, quantification: lower panel). As Thbs2 is an anti- angiogenesis molecule, we stained CD31 to determine the amount of new vasculature in the prostate tumors of PCa CAF±ERα mice. CD31 has been applied as an angiogenesis marker (27) previously. Our data showed a dramatic decrease of CD31 staining in 22Rv1 PCa coimplanted with CAF ERα+ cells (Figure 6E, quantification: lower panel).

Discussion

PCa survival, proliferation and migration can be increased by stromal cells (28,29). Activated fibroblasts, or CAF cells (30), are characterized as stroma with a loss in smooth muscle cells, decrease in normal tissue structure, increased angiogenesis, nuclear atypia and genetic alterations (31). Two markers of CAF are transforming growth factor beta 3 and CD90 (32,33). Consistently, our CAF cells were found to have lower transforming growth factor beta 3 and higher CD90 expression than the normal prostate fibroblasts we established (data not shown). Here we present for the first time a novel mechanism whereby stromal ERα can play a protective role to suppress PCa invasion. We showed that this protection is potentially imbued via lation of the anti-angiogenic factor Thbs2 and downregulation of the ECM degrader and matrix metalloproteinase family activator, MMP3. These results were confirmed in an in vivo mouse model that recapitulated the TME during PCa progression. Additionally, we were able to see that downstream functional targets of the two genes were altered, there were less endothelial cells and decreased MMP9 staining in the tumor site of CAF ERα/22Rv1 injected mice compared with CAF vector/22Rv1 mice.

Early studies suggested that epithelial-to-mesenchymal transformation can be a driving force of cancer invasion (34,35). We assayed several epithelial-to-mesenchymal transformation markers in PCa cells treated with CM from CAF vector versus CAF ERα cells for 2 weeks. Our results showed several epithelial-to-mesenchymal transformation markers were downregulated in CAF-ERα-CM-treated PCa cells compared with vector CM-treated PCa cells (Supplementary Figure 3, available at Carcinogenesis Online). These data indicate that in addition to regulating PCa invasion through targets such as Thbs2 and MMP3 in the CAF cells, there are mechanisms by which CAF ERα executes its protective roles in PCa invasion, for example, by preventing the epithelial-to-mesenchymal transformation in PCa epithelial cells.

Although not a main focus of our study, it should be noted that we were able to show that the Thbs2/MMP3 pathway was estrogen independent (Supplementary Figure 1, available at Carcinogenesis Online). Estrogen can strongly activate ERs, and indeed stimulation of Thbs2 and inhibition of MMP3, respectively, in the E2 stimulated group is magnitudes higher. However, in the absence of E2, there is still a substantial and significant induction of the Thbs2/MMP3 pathway. This is possible as even in the absence of E2, ERα can be activated via post-translational modifications or growth factors (36–38). Determining which growth factors or post-translational modifications modify ERα could lead to potential therapeutic benefit via a more cell specific response. This is also of importance as E2 stimulation in other studies using normal, WPMY-1 human fibroblasts has shown an ‘increase’ in PCa invasion (39). Interestingly, in that study, the pathway was transforming growth factor beta 1 mediated; however, in our CAF studies, we did not see transforming growth factor beta 1 significantly upregulated (data not shown). More research into the potential differences between normal and CAF cells could reveal differential therapeutically relevant pathways.

Thbs2 is a member of a family of five glycoproteins that are involved in cell adhesion, migration and invasion in normal and cancerous cells (23,40,41). Although much is known about the physiological role of Thbs2, less information exists about the Thbs2 role in PCa. One study has shown that E2 has been able to influence Thbs2 levels in Bos Taurus (cow), but this study did not address receptor specificity (42) so the specific ERα modulation of this gene in selective cells has yet to be researched. Here we show that Thbs2 in PCa CAF is an ERα target gene and is able to prevent PCa invasion potentially through downregulation of MMP9 and MMP3 bioactivity. Additionally, 22Rv1/CAF ERα orthotopically injected mice displayed a reduced amount of endothelial cells, which was probably caused by Thbs2 upregulation.

Although both of these are novel ERα targets in PCa as far as we know, MMP3 has recently been shown to be positively regulated by ERα in a bone study (43). MMP3 is a master controller of the matrix metalloproteinase family, able to activate several of the other family members as well as having ECM remodeling capabilities itself. Here we show that activation of ERα is able to heavily restrict the ability of MMP3 enzymatic activity, transcription and activation of other MMPs, such as MMP9. Although we show in mice that Thbs2 is a direct ERα target, an initial drawback is the lack of an ERE site on the human Thbs2 gene 4kb upstream of the transcription start site (data not shown). However, an analysis of the 5000bp upstream region of human Thbs2 revealed several possible indirect pathways through which ERα could regulate transcription. The study found consensus sequences for the following transcription factors; nuclear factor-kappaB, nuclear factor-Y, p53, myc common factor 1, activator protein 1, Sp1, CF-1 and GATA (44). Of these eight transcription factors, nuclear factor-kappaB (45), nuclear factor-Y (46), Sp1 (47), activator protein 1 (48) and p53 (49) have all been shown to bind ERα. With so many potential ERα cofactors, it is quite possible that ERα is still able to exert an influence on the human version of this gene. Both MMP3 and Thsbs2 make good therapeutic targets as each has a dual function. Thbs2 is able to inhibit MMP2 activity through protein–protein interaction as well as inhibit angiogenesis which is commonly thought of as an important step in the progression of many cancers (50). MMP3 on the other hand is able to activate other MMPs, notably MMP9 through cleavage of the pro form of the enzyme to form the active form. MMP3 also contains ECM degradation abilities, allowing it to degrade collagen types II, III, IV, IX and X and proteoglycans, fibronectin, laminin and elastin (51). A way to modulate both of these molecules at once could be highly advantageous in the treatment of PCa. Currently, we show that ERα is able to upregulate Thbs2 and downregulate MMP3 that can explain why stromal ERα positive PCa patients had better clinical outcomes. Further studies could elucidate the Thbs2 to MMP3 regulation as well as possible therapies to activate CAF ERα in order to better treat PCa.

It is a commonly held belief that larger tumors lead to poorer prognosis and increased metastasis. Interestingly, we found no correlation between tumor size and the invasiveness of the tumors. In fact, earlier time points revealed that CAF ERα may increase the tumor growth. This result leads us to believe that CAF ERα may play dual roles in the TME depending on the tumor stage. Together, our data provide multiple evidences that CAF ERα can regulate Thbs2 to decrease invasive capability, potentially through decreased tumor angiogenesis. Although the crosstalk between Thbs2 and MMP3 needs to be further elucidated, we provide compelling evidence that regulation does exist between ERα, Thbs2 and MMP3 in CAF cells to regulate the tumor invasive potential.

Supplementary material

Supplementary Table 1 and Supplementary Figures 1–4 can be found at http://carcin.oxfordjournals.org/

Funding

National Institutes of Health (CA137474).

Glossary

Abbreviations:

CAF

cancer-associated fibroblasts

cDNA

complementary DNA

CM

conditioned media

E2

17ß-estradiol

ERα

estrogen receptor alpha

ERE

estrogen response element

ECM

extracellular matrix

MMP3

matrix metalloproteinase 3

PCa

prostate cancer

Thbs2

thrombospondin 2

TME

tumor microenvironment.