Matrix Metalloproteinases Contribute Distinct Roles in Neuroendocrine Prostate Carcinogenesis, Metastasis, and Angiogenesis Progression

In men, prostate cancer is the most diagnosed cancer and is second only to lung cancer in cancer deaths in the United States (1). Most human prostate carcinomas respond to hormonal therapy initially but often become more aggressive and hormone refractory. Most adenocarcinomas contain foci marked by a characteristic neuroendocrine-like differentiation (NED) pattern (2–4). These foci do not express androgen receptor and are considered androgen insensitive (5). As prostate tumors progress, their histopathology may exhibit increased NED with associated androgen independence (6). Increased levels of NED biomarkers correlate with increased tumor progression, refractoriness to hormone-based regimens, and poor prognosis (4, 7). However, little is known about the factors regulating growth of these NED-positive tumors.

CR2-TAg transgenic mice express SV40 tumor antigen (SV40 TAg), under the control of transcriptional regulatory elements from the cryptidin-2 gene, in a subset of neuroendocrine cells present in the prostate (8, 9). By 8 weeks of age, prostatic intraepithelial neoplasia (PIN) lesions are evident. Focal invasion begins at 10 to 12 weeks, and visible tumors are formed by around 16 weeks. One hundred percent of these mice develop prostate tumors by 24 weeks, with the majority exhibiting metastases to peripheral tissues, including lung, liver, lymph nodes, and bone. The tumor is characterized by cells with a high nuclear to cytoplasmic volume ratio and abundant rosette formation characteristic of a neuroendocrine tumor. The neoplastic cells in CR2-TAg mice express neuroendocrine markers, are androgen receptor-negative, and show androgen independence.

Matrix metalloproteinases (MMP) are a family of enzymes that cleave a broad range of components of ECM, basement membrane, growth factors, and cell surface receptors (10, 11). MMPs are upregulated in cancer progression, can act as oncogenes, and promote invasion and metastasis in virtually all solid tumors (10, 11). These enzymes play a role not only in tumor initiation and invasion but also in angiogenesis, metastasis, and releasing of other tumor-promoting factors. Stromal and inflammatory cells, rather than tumor cells, typically synthesize MMPs, which can then act on the stroma and regulate the tumor microenvironment as well as acting on tumor cells themselves (10, 11).

Several MMPs are overexpressed in prostate cancer progression, and androgen ablation or castration increases levels of MMPs (12–15). Indeed, increased expression of MMP-2 in cancer cells is an independent predictor of decreased prostate cancer disease-free survival (16). Moreover, a synthetic inhibitor of MMPs decreases tumor growth and metastases in a rat prostate cancer model (17), and reduced expression of MMP-9 in prostatic carcinoma cells results in reduced lung metastases but does not affect the tumor growth rate (18).

The high penetrance of tumor progression makes the CR2-TAg mouse an attractive model for characterizing the contribution of MMPs to aggressive neuroendocrine tumor progression. Therefore, in this study we used both pharmacologic and genetic approaches to determine the effect of MMP-2, MMP-7, and MMP-9 on the progression of carcinogenesis, metastasis, and angiogenesis.

Mice homozygous for null alleles of the MMP-2 (19), MMP-7 (20), MMP-9 (21), and CR2-TAg transgene (8, 9) were genotyped using published protocols. All were backcrossed to FVB/N in University of California-San Francisco (UCSF) mouse tumor model core. Mice were maintained under pathogen-free conditions in the UCSF barrier facility. All animal protocols were reviewed and approved by the UCSF Institution Animal Care and Use Committee.

Two nanomoles of MMPSense molecular imaging agent (VisEn Medical, Inc.) were injected into the tail vein of mice 12 to 18 h before imaging (22, 23). Mice used in Fig. 1 were CR2-TAg;MMP-7^+/− or MMP-7^+/+. All tomographic studies were performed using VisEn's fluorescence molecular tomography (FMT1) system. Fluorochrome quantification was determined by establishing three-dimensional regions of interest (ROI). Fluorochrome concentration was calculated automatically from the reconstructed images using FMT1 Imaging Software in the context of preacquired calibrations. Data are expressed as pmol fluorescence per ROI.

Gelatin zymography was performed as described previously (24). For MMP-7 Western blots, we loaded 30 μg of protein per lane, as determined by Bradford protein assay (Bio-Rad), on NuPAGE 4% to 12% Bis-Tris gradient gels with MES running buffer (Invitrogen) and used goat anti-mouse MMP-7 (R&D Systems) at a dilution of 1:500 in 3% milk prepared in PBS with 0.1% Tween. Chemiluminescence was detected using a Fujifilm ImageQuant LAS-4000, and relative intensity of bands was quantified using Photoshop to calculate intensities relative to actin levels.

The prostate lobes were removed, weighed, fixed in 4% paraformaldehyde overnight, and stored embedded in paraffin or frozen in OCT until sectioning. Lung, liver, lymph node, and bone were collected and processed similarly to the prostate tissue. Bones were decalcified in EDTA following fixation.

All antibodies and conditions used for immunostaining of tissues in this study are listed in Supplementary Table S1. We used the following antibodies: anti-MMP-2 (Cell Signaling), anti-MMP-2 (Chemicon), anti-MMP-7 (R&D Systems; ref. 25) and MMP-9 (26), anti-phosphorylated histone H3 (Cell Signaling), anti-cleaved caspase-3 (Cell Signaling), anti-CD31 (Pharmingen), anti-α-smooth muscle actin (SMA) Clone 1A4 Cy3 conjugate (Sigma), anti-F4/80 (Caltag), anti-CD45 (BD Pharmingen), anti-synaptophysin (DAKO), rabbit anti-SV40-TAg (gift of Dr. Doug Hanahan), biotinylated antirabbit and antirat IgG (Vector), Alexa 488, 568, 594 antirat or antirabbit secondary antibodies (Molecular Probes), and peroxidase-conjugated antirat IgG (Jackson ImmunoResearch).

We cut 20-μm sections for vasculature staining from frozen blocks and usually 5-μm sections from paraffin-embedded blocks for H&E staining and immunocytochemistry. Images were captured using a Leica DMR microscope.

For tumor samples harvested from 24-wk-old animals, tissue sections were stained by H&E and given a score of 1 to 3, wherein 1 = “PIN” (>60% of tissue is PIN and normal glands), 2 = “Mixed” (40–60% PIN, 40–60% Carcinoma), or 3 = “Carcinoma” (>60% of tissue is carcinoma). Mouse prostate pathology terms used follow the Bar Harbor classification system (27). For 16-wk prostate tissue samples, H&E-stained sections were scored blinded by a trained pathologist for the glands containing invasive foci as a percentage of total glands.

AG3340/Prinomastat (28) was provided by David Shalinsky (Agouron Pharmaceuticals). The inhibitor was prepared at 3 mg/mL in acidified drinking water (pH 2.3) and filter sterilized. Mice received the drug p.o. ad libitum beginning at 6 wk of age, changed every 2 wk, and continued until tissues were harvested at 24 wk. Control mice received the same volume of acidified drinking water without inhibitor.

Harvested lung and liver were immediately scored for visible surface-associated macrometastatic (macromet) nodules and then processed. H&E-stained sections were scored for micrometastases (micromet). For each mouse, one section was analyzed initially. If metastasis was not detected, then two to six more sections were analyzed per mouse.

For vessel density, vasculature was counted manually from sections stained for CD31 and SMA. We defined a pericyte-covered vessel as a CD31⁺ vessel with at least one surrounding SMA⁺ cell. Vasculature area was quantified using Image J 1.38× (NIH). Average vessel size was calculated by dividing the vasculature area by vessel density. Tumor cell invasion was scored for SV40 Tag-positive stromal cells associated with vasculature (perivascular invasion) or cells not associated with vasculature (infiltrative invasion) as a percentage of total invasive cells (i.e., number of perivascular + infiltrative invasive cells).

Kaplan-Meier survival curves were analyzed by log-rank tests to determine the significance of observed differences between the curves. Metastases were analyzed by Fisher's exact test. Median tumor weights and invasive status (perivascular or infiltrative) were compared using Mann-Whitney tests. Statistical analysis other than for metastases used Prism 4 software (GraphPad Software, Inc.). All cohorts composed of three or more genotypes were compared using one-way ANOVA and using unpaired Student's t tests for comparing two cohorts. In all experiments, P < 0.05 was considered statistically significant.

We first analyzed MMP activity in the CR2-TAg model using MMP substrate-based tomography. CR2-TAg transgenic mice and nontransgenic controls were injected with a pan-MMP probe, MMPSense, that becomes fluorescent following cleavage by active MMPs, including MMP-2, MMP-3, MMP-9, and MMP-13 (22, 23). Noninvasive whole body optical tomography indicated that CR2-TAg mice had significantly increased MMP activity in the region of their prostate compared with nontransgenic littermates (P = 0.004; Student's t test; Fig. 1A).

By gelatin zymography, we found no significant difference in MMP-2 activity at 16 and 24 weeks of age in nontransgenic littermate prostate samples (Fig. 1B). At 16 weeks (local microinvasion stage), dorsolateral prostate samples from CR2-TAg mice had increased levels of active MMP-2, pro-MMP-9, and active MMP-9. By 24 weeks, when 100% of the CR2-TAg mice have tumors and the majority have liver metastases (9), active MMP-2 and MMP-9 levels were significantly increased in the prostate, seminal vesicles (marker of local prostate cancer invasion) and liver (frequent site of metastasis).

Western blot analyses revealed increased levels of active MMP-7 by 12 weeks in CR2-TAg prostates, with variable activity at 24 weeks (Fig. 1B and E).

Prostatic stromal cells may be involved in progression of prostate adenocarcinomas (29, 30). These previous studies suggest that increased MMP-2 and MMP-9 levels reflect stromal cell, rather than tumor epithelial cell, contributions (10).

MMP-2 localized to the stroma within and surrounding the CR2-TAg tumors and colocalized with F4/80 (macrophage marker) and α-SMA (a marker of stromal carcinoma-associated fibroblasts, pericytes, and smooth muscle cells) at 24 weeks (Fig. 1C and D). MMP-9 localized to the stroma surrounding the tumor and in stromal regions within the tumor itself (Fig. 1D; Supplementary Fig. S1A and B), often adjacent to clusters of malignant cells and in regions enriched for blood vessels and stroma. MMP-9 was prominent in macrophages, detected by costaining with anti-F4/80 (Fig. 1C). Therefore, MMP-2 and MMP-9 reside in stromal rather than epithelial cells.

The prostate contains three epithelial cell types: luminal secretory cells, basal cells found between the luminal cells and basement membrane, and scattered neuroendocrine cells. We observed MMP-7 staining at the apical surface of infrequent luminal cells located in normal prostate as well as in abnormal neoplastic glands from 16-week-old CR2-TAg mice (Supplementary Fig. S1C and D). MMP-7 did not colocalize with the neuroendocrine cell marker synaptophysin by immunofluorescence, indicating that MMP-7–positive cells are not likely from the neuroendocrine lineage (Supplementary Fig. S1E). By 24 weeks of age, MMP-7 was visible in nonneoplastic prostatic glandular tissue, but not within the tumor itself (Fig. 1D), similar to its localization in normal mouse and rat prostate and in human tumors of the breast and colon (25, 31).

In a pilot experiment, CR2-TAg mice were treated with the oral metalloproteinase inhibitor AG3340 that potently inhibits MMP-2, MMP-3, MMP-9, MMP-13, and MMP-14 (picomolar IC₅₀ values) and inhibits other MMPs less well (e.g., MMP-1, MMP-7; refs. 22, 28). Mice were given AG3340 (n = 5) or placebo control (n = 4) in their drinking water beginning at 6 weeks after birth until 24 weeks. Two of the control mice died before 24 weeks, giving a median survival of 164 days (data not shown). In contrast, all AG3340-treated CR2-TAg mice survived to 24 weeks.

AG3340-treated transgenic mice had significantly reduced tumor burdens at 24 weeks compared with the control vehicle-treated mice (P = 0.02, Mann-Whitney test) or with wild-type untreated CR2-TAg mice (P = 0.0004, Mann-Whitney test; Fig. 2A). AG3340-treated mice had no macromets to the lung or liver (data not shown), whereas 38% of the untreated mice had macromets to the liver and 14% had macromets to the lung (Supplementary Table S5). Prostate samples from AG3340-treated mice (n = 5) were less invasive with a more glandular morphology, whereas control prostates predominantly had invasive carcinoma (Fig. 2C; Supplementary Table S2). Treated tissues had reduced levels of stromal CD45+ cells (Supplementary Fig. S1). AG3340-treated mice still developed small invasive foci with characteristic neuroendocrine rosette patterning and had small carcinomas. We concluded that AG3340 treatment limits, but does not prevent, neuroendocrine carcinoma progression.

We bred CR2-TAg animals to mice deficient for MMP genes to determine the contributions of specific MMPs to tumor progression. MMP-2–deficient CR2-TAg mice (CR2-TAg;MMP-2^−/−) had increased survival with a median survival of 201 days (n = 39) compared with 188 days for transgenic mice homozygous for the wild-type MMP-2 allele (CR2-TAg;MMP-2^+/+; n = 52; P = 0.0007; log-rank test) or 188 days for heterozygotes (CR2-TAg;MMP-2^+/−; n = 71; P = 0.001; Fig. 2D; Supplementary Table S2).

The lack of MMP-7 or MMP-9 did not affect survival (Supplementary Fig. S2). CR2-TAg;MMP-7^−/− animals had a median survival of 173 days (n = 27), similar to CR2-TAg;MMP-7^+/− cohort mice (169 days; n = 14; P = 0.81). CR2-TAg;MMP-9^−/− mice had a median survival of 170 days (n = 26), which was similar to 171 days for CR2-TAg;MMP-9^+/− mice (n = 52; P = 0.75).

The increased survival in CR2-TAg;MMP-2^−/− mice was related to reduced tumorigenicity (Fig. 2A; Supplementary Table S4). Median tumor weights in CR2-TAg;MMP-2^−/− mice were significantly lower (0.33 g) than in CR2-TAg;MMP-2^+/+ mice (2.19 g; P < 0.0001; Mann-Whitney test) or CR2-TAg;MMP-2^+/− mice (1.11 g; P = 0.002). The median tumor weights in CR2-TAg;MMP-2^+/− mice were also significantly lower than in the CR2-TAg;MMP-2^+/+ mice (1.11 g versus 2.19 g, respectively; P = 0.04), suggesting that MMP-2 dosage may contribute to tumor burden. In contrast, CR2-TAg;MMP-7^−/− and CR2-TAg;MMP-9^−/− mice did not show significantly reduced tumor burdens compared with their heterozygous controls (P = 0.73 and P = 0.15, respectively; Supplementary Fig. S2).

We next characterized a cohort of CR2-TAg;MMP-9^−/− and CR2-TAg;MMP-9^+/− mice at 16 weeks of age using morphometric analysis of prostate sections. CR2-TAg;MMP-9^−/− mice (n = 6) had a significant increase in the percentage of glands with invasive foci compared with the CR2-TAg;MMP-9^+/− controls (n = 5; P = 0.01; Student's t test; Fig. 2D). Therefore, in the absence of MMP-9, invasive carcinoma induced by CR2-Tag is accelerated. However, this difference did not alter the survival curve (Supplementary Fig. S2).

CR2-TAg;MMP-7^−/− and CR2-TAg;MMP-9^−/− tumors were predominantly invasive carcinomas, resembling advanced neuroendocrine tumors with little or no normal glandular tissue (Supplementary Table S1; Fig. 2C).

Compared with the MMP-7– and MMP-9–deficient mice, the CR2-TAg;MMP-2^−/− mice had less invasive carcinoma than CR2-TAg;MMP-2^+/+ mice (Fig. 2C). Similar to the AG3340-treated transgenic mice, the CR2-TAg;MMP-2^−/− mice had numerous glands containing high-grade PIN. In contrast, MMP-7– and MMP-9–deficient transgenic mice had predominantly carcinoma and invasive phenotypes at 24 weeks, similar to their heterozygous MMP control cohorts (Supplementary Table S1).

We analyzed lung and liver tissues for metastasic nodule formation at 24 weeks and scored for visible nodules (macromets). H&E-stained tissue sections were also analyzed by microscopy for smaller neoplastic nodules (micromets). The percentage of CR2-TAg;MMP-2^−/− mice with lung metastases was significantly reduced compared with the MMP-2^+/+ transgenic animals (P = 0.01; Fisher's exact test), with a trend toward reduction in percentage of mice with predominantly micromets (P = 0.05; Fig. 3; Supplementary Table S5).

Although MMP-9 and MMP-7 deficiencies decrease melanoma and lung metastases (32), CR2-TAg;MMP-7^−/− mice had no statistically significant difference in metastasis compared with CR2-TAg;MMP-7^+/− controls (for lung, P = 0.64; for liver, P = 1). CR2-TAg;MMP-9^−/− mice also showed no difference in the percentage of mice with metastasis compared with CR2-TAg;MMP-9^+/− mice (for lung, P = 1; for liver, P = 0.54).

To determine if proliferation or apoptosis contributed to reducing tumor burden in the MMP-2^−/− mice, we stained tumor sections from 24-week-old animals with anti-phosphorylated histone H3 (mitotic proliferation marker) and anti-cleaved caspase-3 (apoptosis marker) and counted positively stained cells per carcinoma region per mouse (Supplementary Fig. S4). There was no significant difference in phosphorylated histone H3 staining across the different genetic backgrounds, suggesting that there is no difference in proliferation rate within the tumor region (P = 0.54; one-way ANOVA).

We detected no significant difference in cleaved caspase-3 staining except in AG3340-treated 24-week-old CR2-TAg animals, which had significantly reduced cleaved caspase-3 staining compared with the untreated transgenic control (P = 0.04; Student's t test) but not compared with the wild-type control (P = 0.09).

Because MMPs are known regulators of vascular density, permeability, and angiogenesis, we investigated the changes in vasculature. In CR2-TAg transgenic mice, blood vessels, as shown by CD31-immunostaining, were present primarily in the peripheral stroma up until 12 weeks of age, but at 16 and 24 weeks tumors were well vascularized (data not shown; Supplementary Fig. S5).

We analyzed the vasculature in MMP-deficient mice to determine MMP contribution. At 24 weeks of age, tumors in CR2-TAg; MMP-2^−/− mice had reduced numbers of CD31⁺ blood vessels (P = 0.01, Student's t test) but not in the number of mature (CD31⁺SMA⁺) vessels (P = 0.16; Fig. 4).

By comparison, tumors of CR2-TAg; MMP-7^−/− mice at 24 weeks of age had decreased blood vessel size (P = 0.02) and reduced endothelial area (P = 0.046) but similar vessel density compared with control heterozygous mouse tumors (Fig. 5; Supplementary Fig. S2). The tumor vessels in the control cohort varied in size and included hyperdilated, irregularly shaped vessels, whereas MMP-7–deficient vessels were more regularly shaped and smaller (Fig. 5).

Previous studies in glioblastoma show that MMP-9–deficient tumor cells are highly invasive and invade perivascularly (33). At 24 weeks of age, CR2-TAg;MMP-9^−/− tumor cells showed tumor stromal invasion that had an increased representation of perivascular invasion and a decrease in infiltrative invasion. Heterozygous control tumors had comparatively more vasculature-independent stromal invasion of tumor cells (Fig. 6).

At 24 weeks of age, CR2-TAg;MMP-9^−/− mice had similar numbers of vessels within the tumor but decreased vessel size compared with heterozygous controls (Fig. 6 and data not shown). Tumors from the heterozygous control mice had hyperdilated, irregularly shaped vessels, whereas the vasculature within the MMP-9–deficient tumors was more elongated, thin, and regularly shaped. These tissues also showed a reduction in stromal CD45⁺ cells (Supplementary Fig. S1).

Taken together these data suggest that MMP-2 promotes neovascularization of immature blood vessels and MMP-7 increases blood vessel size without an effect on tumor burden. MMP-9 also increases vessel size, reduces the rate of focal invasion of glandular epithelium, and promotes vasculature-independent infiltrative invasion.

Here we used optical imaging, MMP inhibition, and genetic approaches to determine the role of three MMPs in an aggressive, highly penetrant metastatic tumor model of neuroendocrine origin (Supplementary Table S6). Differential localization and microenvironmental cues contribute to activation and substrate selection of MMPs and other proteases. Clearly, MMP-2, MMP-7, and MMP-9 have different functions in tumorigenesis in the CR2-TAg neuroendocrine prostate cancer model.

MMP-2 has been implicated in survival, angiogenesis, and metastasis in studies using several tumor models (34–36). Here we show that MMP-2 is a strong contributor to prostate carcinogenesis. We found that MMP-2 deficiency results in a reduction of immature blood vessel number, which likely contributes to the reduced tumor burden in these mice, because neovascularization is necessary to support tumor growth. Previous studies showed MMP-2 has a role in angiogenesis: MMP-2 deficiency decreased corneal angiogenesis (37). Also, in a glioblastoma model, MMP-2^−/− had increased survival, increased vessel density, and increased invasive phenotype, migrating along blood vessels in the brain parenchyma (34). However, this vasculature was not functional due to decreased pericyte coverage.

We observed that MMP-7 and MMP-9 contribute to tumor progression, although they did not affect survival or metastasis in CR2-TAg mice. Previous studies found different roles for MMP-7 in angiogenesis and invasion. Corneal deficiency in MMP-7 leads to increased endothelial area coverage (38). Also overexpressed MMP-7 in prostate cells leads to increased invasion of neighboring tissue (31).

Whereas MMP-9 contributed to vascular remodeling in our model, these aggressive tumors still were able to adapt to its absence and the resulting reduced angiogenesis by increasing perivascular invasion along a more normal vasculature, as we observed previously in glioma (33). Ultimately the MMP-9–deficient tumors developed similar survival and both tumor and metastatic burdens compared with controls. Thus, the same MMP has quite different outcomes in different tumors. How is this regulated? In a number of tumor models, MMP-9 deficiency causes a reduction in tumor progression and metastasis (10). In this study, by 24 weeks, MMP-9 deficiency led to an increased dependence on perivascular invasion and a reduction in vasculature-independent invasion. Whether MMP-9 has protumorigenic or antitumorigenic effects depends on the environment and stage in tumor progression. For example, in neuroblastoma and orthotopically injected pancreatic human cancer cells, loss of MMP-9 results in decreased microvessel size and decreased vascularization (39, 40), and in a skin cancer model, MMP-9^−/− reduces the number of tumors but increases the malignancy of the tumors formed (41).

The role of MMP-9 may be determined by the balance between the recruitment of myeloid cells, release of angiogenic factors, and promotion of invasion in a given microenvironment. For example, MMP-9 triggers the release of matrix-bound vascular endothelial growth factor in the RIP-TAg model (42). Also, recruitment of bone marrow–derived cells to the tumor stroma and vasculature is MMP-9 dependent in neuroblastoma and other tumor models (43, 44). Similarly we saw a reduction in CD45⁺ cells found in the stroma surrounding the tumor. Loss of MMP-9 can delay the angiogenic switch and transition to malignancy, as is the case when macrophage infiltration is prevented in the MMTV-PyMT breast cancer (45). Future studies will be necessary to determine the role, if any, of myeloid cells to these phenotypes.

The response of individual MMPs in the CR2-TAg microenvironment is a balancing act between the substrates and cofactors unique to that tissue. Overlapping substrates could account for a certain amount of MMP redundancy. For example, MMP-2 may be able to compensate for MMP-9^−/−. We detected increased expression of MMP-2, MMP-7, and MMP-9 following treatment with AG3340, as has been seen previously using other MMP inhibitors (Supplementary Fig. S3; refs. 46–48). Interestingly, in glioblastoma, MMP-2 and MMP-9 deficiency stimulates perivascular invasion, similar to our observations seen here following MMP-9 deficiency (33, 34). Also, strain-dependent variances account for some variation between genotypes in breast cancer studies and may play a role in humans (49).

We exploited the CR2-TAg animal model for characterizing MMP genes and identifying potential therapeutic treatments of aggressive neuroendocrine malignancies. We found that treatment with the MMP inhibitor AG3340 decreased tumor burden and slowed tumor progression. Similarly, AG3340 treatment of a neuroblastoma tumor progression model increased survival and reduced tumor burden, lung metastasis, and microvessel size (39). Because AG3340 is a strong inhibitor of tumorigenesis when given to the mouse early in tumor progression, an MMP inhibitor, such as AG3340, may have therapeutic efficacy in certain forms of prostate cancer, although we predict it would be insufficient to result in complete tumor ablation, because here the tumor adapted to the microenvironment and became invasive.

Based on our study, therapeutic use of broad spectrum MMP inhibitors may have unpredictable results, possibly even a more invasive tumor pathology, which has been seen in some human studies (50). Targeting MMPs does not seem to stop tumor progression of late, highly aggressive tumors. Cleverly, the tumors progress aggressively through alternative routes. Thus, treatment of the most aggressive tumors will require similarly aggressive multihit therapies. Combination therapies, including selective MMP inhibition, may be of therapeutic benefit for the most aggressive prostate malignancies with neuroendocrine features or other types of neuroendocrine tumors.

K. Williams is an employee of VisEn Medical, Inc. The other authors disclosed no conflicts of interest.

We thank David Shalinsky (Agouron) for the MMP inhibitor AG3340, Doug Hanahan for the SV40-TAg antibody, Lynn Matrisian for the MMP-7 knockout mice and antibody, Elena Atamaniuc and Bernard Thompson (T.W.) for mouse care, Helen Capili for tissue sectioning, and the Werb laboratory for helpful discussions.

Grant Support: CA072006 (Z. Werb), University of California-San Francisco Prostate Specialized Programs of Research Excellence CA089520 (Z. Werb), American Cancer Society Postdoctoral Fellowship (L.E. Littlepage), Ruth L. Kirschstein National Research Service Award CA103534 (L.E. Littlepage), Career Development award from Bay Area Breast Specialized Programs of Research Excellence CA58207 (M.D. Sternlicht), and K08 Award NS063456 (J. Phillips).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.