An amino-bisphosphonate targets MMP-9–expressing macrophages and angiogenesis to impair cervical carcinogenesis (original) (raw)
Cervical carcinogenesis and the angiogenic phenotype in HPV/E2 mice. We evaluated angiogenesis in the distinctive neoplastic stages that arise in the HPV/E2 transgenic mouse model of cervical cancer. The histopathological progression to cervical malignancy through low-grade dysplasia (CIN-1/2), to high-grade dysplasia (CIN-3), and then to cervical carcinoma (SCC), is shown in Figure 1A, in comparison with nontransgenic estrogen-treated female (N/E2) control mice; at the defined end point of 7 months of age, when approximately 90% of the HPV/E2 females have invasive cancer, there are no macroscopic LN or distant metastases (not shown). The cervical vasculature was visualized in these lesional stages by CD-31 immunostaining (Figure 1B), revealing modest neovascularization below the stromal/epithelial junction in CIN-1/2 lesions, followed by intense angiogenesis in CIN-3 lesions and cervical carcinomas, as previously observed in human cervical carcinogenesis (3). Given that VEGF-A and its principal signaling receptor, VEGF-R2/flk-1, are commonly involved in angiogenic switching and tumor growth (13, 14) and are implicated in human cervical neoplasia, we assessed their possible involvement in the angiogenic phenotype. Using double-labeled immunofluorescence with a mAb that specifically recognizes VEGF in complex with VEGF-R2 (GVM39) (15), along with the Meca-32 Ab to visualize endothelial cells (ECs), we observed a modest increase in the VEGF/VEGF-R2 complex in vessels below the stromal/epithelial junction of CIN-1/2 lesions and a substantial increase in the high-grade CIN-3 lesions and invasive cervical tumors as compared with the normal cervix (not shown) and N/E2 mice (Figure 1C).
Stages of cervical carcinogenesis and angiogenic profile of HPV/E2 mice. (A) H&E staining shows the steps of tumor progression. (B) Blood vessels (red) were detected with anti–CD-31 Ab; nuclei were visualized by DAPI (blue). The number of vessels per field were: N/E2, 8.2 ± 1.2; CIN-1/2, 14.8 ± 1.1; CIN-3, 32.2 ± 1.7; SCC, 35.2 ± 1.8. Significant differences were observed between CIN-3/SCC and CIN-1/2 (P < 0.01), CIN-3/SCC and N/E2 (P < 0.01), and CIN-1/2 and N/E2 (P < 0.01), but not between CIN-3 and SCC (P = 0.309). (C) VEGF-A/VEGF-R2 complex was detected with GVM39 Ab (green) and ECs with anti–Meca-32 Ab (red). VEGF/VEGF-R2 complex colocalizes with most vessels in CIN-3 and SCC and with a subset in CIN-1/2 (arrows). Arrowheads in CIN-1/2 indicate vessels that do not bind GVM39 Ab. Minimal GVM39 staining was detected in N/E2 stroma. The percentage of VEGF-A/VEGF-R2–labeled vessels was quantitated: N/E2, 8.6 ± 1.2; CIN-1/2, 44.1 ± 2.9; CIN-3, 90.2 ± 3.1; SCC, 91.4 ± 3.4. Significant differences were observed between CIN-3/SCC and CIN-1/2 (P < 0.01), CIN-3/SCC and N/E2 (P < 0.01), and CIN-1/2 and N/E2 (P < 0.01). CIN-3 and SCC were not significantly different (P = 0.690). Values are mean ± SEM. P values were calculated using the Wilcoxon test. Scale bars in A and C: 50 μm (N/E2); 25 μm (CIN-1/2, CIN-3, and SCC). Scale bars in B: 50 μm. E, normal or dysplastic cervical epithelia; S, stroma; T, tumor cells.
MMP-9 and macrophages are associated with the angiogenic vasculature in dysplasias and tumors. Motivated by previous studies in mouse models functionally implicating MMP-9/gelatinase B as a regulator of VEGF signaling and angiogenesis in other organs (13, 16–19), as well as in broader cancer phenotypes (20, 21), and by reports that MMP-9 is upregulated in human cervical neoplasias (7, 8), we evaluated expression of MMP-9 and related MMPs during cervical carcinogenesis. RT-PCR analysis of nine different MMPs revealed that only MMP-9 was markedly upregulated during tumor progression in the cervix of HPV/E2 mice (Figure 2A). By contrast, a related protease, MMP-2, was constitutively expressed in all stages (Figure 2A); mRNAs for seven other MMPs were analyzed, and we observed unchanged or undetectable levels of expression (see Supplemental Figure 1, available at http://www.jci.org/cgi/content/full/114/5/623/DC1).
MMP-9 expression and activity is upregulated in macrophages during tumor progression. (A) RT-PCR analysis revealed increased MMP-9 expression in CIN-3 and SCC as compared with CIN-1/2. No MMP-9 expression was detected in normal cervix (not shown) or in N/E2. MMP-2 was equally expressed at all stages. (B) Immunohistochemical analysis using an anti–MMP-9 Ab revealed no MMP-9 in N/E2 cervix and minimal expression in CIN-1/2 lesions (arrows); in contrast, MMP-9 was detected in the stroma proximal to CIN-3 lesions and tumors. (C) Zymography showing gelatinase activity in tissue lysates of different stages. Both pro–MMP-9 (inactive form, 105 kDa) and active MMP-9 (95 kDa) were upregulated in CIN lesions and tumors as compared with controls. pro–MMP-2 (72 kDa) was slightly increased during progression, but no active MMP-2 (62 kDa) was detected. (D) Gelatinase activity was measured using a fluorescin-gelatin assay in the absence or presence of the MMP inhibitor 1,10 Phe (4 mM). Statistically significant increases in gelatinase activity were observed in CIN lesions and tumors compared with controls: CIN-3/SCC versus CIN-1/2 (P < 0.01); CIN-3/SCC versus N/E2 (P < 0.01); CIN-1/2 versus N/E2 (P < 0.01); CIN-3 and SCC were not significantly different (P = 0.102). Values are mean ± SEM. P values were calculated using the Wilcoxon test. (E) Colocalization of MMP-9 (red) and macrophages (CD-68, green) was observed in the stroma underlying CIN-3 and surrounding SCC. Few MMP-9–expressing macrophages were detected in stroma adjacent to CIN-1/2 lesions (arrows). No MMP-9 expression was observed in macrophages in N/E2 mice (arrowheads). Scale bars: 50 μm (N/E2); 25 μm (CIN-1/2, CIN-3, and SCC).
Using immunohistochemistry (IHC), we confirmed the RNA analysis, detecting MMP-9 expression in scattered cells, both in the stroma adjacent to the epithelial interface of CIN lesions and in tumors (Figure 2B); no such cells were detected in cervix of control or N/E2 mice (Figure 2B, and data not shown). To ascertain whether the appearance of MMP-9 RNA and protein immunoreactivity reflected proteolytically active enzyme, we performed gelatin zymography, which detects both latent pro- and active forms of the enzyme. We observed, consistent with RNA expression, an increase in both forms of MMP-9, particularly in the high-grade dysplasias (CIN-3) and carcinomas, compared with controls (Figure 2C). In contrast, we detected only a slight increase in abundance of the MMP-2 pro form and no evident upregulation of active MMP-2 (Figure 2C). We have substantiated and further quantified the profile of protease activity in these stages using a biochemical assay for gelatinase activity, which showed statistically significant upregulation of total gelatinolytic activity in CIN-3 lesions and SCC as compared with controls (Figure 2D). Consistent with the zymogram analysis (Figure 2C), no significant difference in activity was observed between CIN-3 and SCC (Figure 2D).
It has previously been demonstrated in the related K14-HPV16 skin cancer model, that MMP-9 is supplied by bone marrow–derived cells rather than by transformed epithelial cells (17). To identify the MMP-9–expressing cells in the cervix, we performed immunostaining for markers of various inflammatory cell types, comparing the different stages of tumor progression in HPV/E2 mice. We observed a substantive increase in the number of infiltrating macrophages, as assessed by tissue immunostaining with Ab’s to F4/80 (data not shown) and CD68 (Supplemental Figure 2); by contrast, the abundance of other leucocytes was not appreciably altered (Supplemental Figure 2). Using double-labeled immunofluorescence, we detected MMP-9 expression in macrophages infiltrating the stroma underlying CIN-3 lesions and in tumors (Figure 2E). Notably, the increase in MMP-9–expressing macrophages was concordant with the intense angiogenesis seen in CIN-3 as compared with CIN-1/2 (Figure 2E). Only slight and sporadic MMP-9 immunoreactivity was detected in ECs of the angiogenic vasculature or in stromal fibroblasts associated with the tumors (data not shown). Thus, these particular infiltrating macrophages seem unable to induce MMP-9 in other stromal cell types, but rather themselves serve as the major source of MMP-9.
The nitrogen-containing (amino-) bisphosphonate zoledronic acid impairs angiogenesis, progression, and growth of cervical carcinomas. Molecular and histological evaluation of cervical cancer progression in humans and in HPV/E2 mice presents striking parallels between the two pathways (5, 6, 11, 12). The HPV/E2 mouse model thus presents a platform for preclinical testing of drugs designed to interrupt critical pathways implicated in angiogenesis and lesional progression of cervical carcinoma. Given the above data implicating MMP-9 in angiogenesis and progression, we became interested in assessing the effects of MMP inhibitors (MMPIs) on tumor growth and angiogenesis. Therapeutic trials in a transgenic mouse model of neuroendocrine cancer had previously demonstrated the capability of a prototypical MMPI, BB-94, to inhibit angiogenesis and tumor growth when treatment was initiated during premalignant stages (13, 22). Therefore, we conducted analogous early-stage trials targeting CIN-2/3 lesions with BB94, which produced reductions in the density of the angiogenic neovasculature (data not shown) and in tumor incidence and tumor growth/burden (Figure 3, A and B).
BB94 reduces tumor incidence and tumor volume in HPV/E2 mice. Preclinical trials using the broad-spectrum MMPI BB94 (batimastat) in HPV/E2 mice. (A) Decrease of tumor incidence (51% reduction) in 7-month-old BB94-treated (n = 15) mice compared with control mice (n = 20). (B) Reduction of tumor volume in BB94-treated animals (63% reduction; **P < 0.01) compared with controls. Values are mean ± SEM. P values were calculated using the Wilcoxon test. Drug treatment, tumor volume, and histological scores were determined as described in Methods.
While encouraged, we were aware that human clinical trials in various disease indications using BB94 and other active site–targeted MMPI had failed due to side effects and toxicity as well as poor efficacy against end-stage tumors (23). We were thus led to consider a much different agent, zoledronic acid (ZA; Zometa), a nitrogen-containing (amino-) bisphosphonate (N-BP) that is FDA approved to reduce skeletal complications of bone metastasis in patients with multiple myeloma and several solid tumor types with minimal side effects (24, 25). Recent studies in cell culture and xenotransplant tumor models have variously suggested that ZA and other BPs have antiangiogenic and MMP-inhibitory activity (26–31). These data led us to test ZA on cervical neoplasias in HPV/E2 mice.
We treated female mice bearing CIN-3 lesions with ZA (100 μg/kg daily, subcutaneously) or with vehicle alone for 6 weeks, asking whether ZA could prevent tumor formation and/or affect angiogenesis (in a prevention trial [PT]). Although ZA is inoculated intravenously on a monthly regimen in humans, a number of preclinical studies suggested that it was best supplied in a daily regimen in mice, reflecting much different pharmacokinetics (26, 27). Fluorescin-lectin perfusion and immunostaining with Meca-32 Ab revealed a decrease of 56% in the number of blood vessels proximal to the basement membrane in dysplastic lesions and cervical carcinomas from ZA-treated mice compared with analogous lesions in controls (Figure 4, A and B). At the defined end point (5 months of age), the control mice had a tumor incidence of 85% (Figure 4C). Remarkably, ZA was able to limit the tumor incidence to 30% (Figure 4C) and reduce tumor volume 61% (Figure 4D).
ZA inhibits angiogenesis and reduces tumor incidence and growth. (A) Perfusion of HPV/E2 control mice and ZA-treated mice (6 weeks of treatment; PT) with fluorescin-lectin revealed dramatic changes in the 3-dimensional organization of the vasculature proximal to CIN-3 lesions of treated mice at 5 months of age. (B) Vessels’ density, as assessed by Meca-32 immunostaining, was significantly reduced in ZA-treated mice as compared with controls (56% reduction). Results are mean ± SEM of five fields per mouse from a total of eight mice. (C) CIN-2/3 lesion-bearing mice (_T_0, beginning of treatment; 3.5 months old) treated with ZA or vehicle for 6 weeks (PT) showed a 55% reduction in the tumor incidence at the end of the treatment (_T_1; 5 months old) (n = 16 control, n = 10 ZA-treated). (D) Tumor volume of ZA-treated mice was reduced by 61% compared with untreated controls (PT; n = 16 control, n = 10 ZA-treated). (E) Mice bearing SCC (T0; 5 months old) treated with ZA for 1 month (_T_1) in a RT showed a 57% decrease in tumor volume (n = 15 control, n = 10 ZA-treated). Values are mean ± SEM. **P < 0.01; #P < 0.001. P values were calculated using the Wilcoxon test. Tumor volume and histological scores were determined as described in Methods. Scale bar: 50 μm.
Next, to assess the effects of ZA on the growth of established cervical carcinomas, we performed a regression trial (RT), treating the mice with ZA for 1 month starting at 5 months of age, when 85% had malignant tumors and the remainder had CIN-3 lesions. One month later the control mice had a tumor incidence of 90% and increased tumor volume (Figure 4E). In contrast, the ZA regimen impaired growth of the invasive carcinomas, reducing tumor volume 57% from that at the start (Figure 4E). Thus, ZA was able to impact both premalignant and malignant stages of cervical carcinogenesis, reducing angiogenesis both in high-grade dysplasias and cancers, interfering with progression to and subsequent growth of invasive tumors.
Having demonstrated clear and convincing efficacy, we sought to investigate the mechanism by which ZA impaired angiogenesis and tumor incidence and growth, first by examining the frequency of apoptosis in ZA-treated mice. We used caspase-3, an effector of the apoptotic program, as a biomarker for apoptosis, monitoring the presence of its activated form by IHC in ZA-treated mice; we observed a statistically significant increase in cells expressing active caspase-3 in treated versus control tumors (Figure 5, A–C). ZA treatment also increased the incidence of apoptosis in CIN-3 lesions, whereas fewer apoptotic cells were detected in CIN-3 of control or vehicle-treated mice (Figure 5, D–F). Similar results were obtained using the TUNEL assay to detect DNA fragmentation (data not shown). We further investigated the apoptotic rate in ECs in neoplastic lesions of mice treated with ZA. Double-label immunofluorescence using Meca-32 and caspase-3 Ab’s showed an increase in the number of apoptotic ECs within vessels (6 ± 1/field, P < 0.001) of the stroma proximal to CIN-3 lesions (Figure 5, G–I) and in cervical carcinomas (5 ± 1/field, P < 0.001; data not shown). No apoptotic ECs were detected in analogous lesions from vehicle-treated control mice (Figure 5, J–L).
ZA induces apoptosis in epithelial cells and ECs in tumors and CIN-3 lesions. (A–C) A significant increase in apoptosis in cervical carcinomas was observed in ZA-treated mice compared with controls (A) after 6 weeks of treatment (PT) as revealed by caspase-3 immunostaining (#P < 0.001, Wilcoxon test). Representative sections of tumors from treated and untreated mice are shown in B and C. (D–F) Similar increase in apoptosis was detected in CIN-3 epithelium of ZA-treated mice as compared with controls (D); treated versus untreated lesions are exemplified in E and F. Five fields per mouse were counted to assess the caspase-3–positive cells. Values are mean ± SEM. Increased apoptosis was observed in vessels in CIN-3 lesions of ZA-treated mice (G–I) compared with controls (J–L) as detected by colocalization of Meca-32 (green) with caspase-3 (red). Arrows indicate apoptotic ECs. Scale bars: 50 μm (B and E); 25 μm (C, F, and G–I).
The squamous epithelium of the cervix in HPV/E2 mice shows a progressive increase in proliferation rates in the successive stages of carcinogenesis (11, 12), motivating us to ask whether ZA was also impacting the cell division cycle. We did not detect significant differences in the rates of proliferation of neoplastic cervical epithelial cells in CIN-3 lesions or cervical carcinomas nor in their adjacent stroma, as compared with untreated controls (see Supplemental Figure 3). Taken together, these data suggest that the biological effects of ZA result from increased apoptosis in epithelial and endothelial cells of the lesions. By contrast, ZA does not appear to act as an antimitotic drug, since it does not reduce proliferation of the transformed cervical epithelial cells.
ZA inhibits MMP-9 expression by macrophages and activity in CIN-3 and cervical carcinomas. The ZA trial design was motivated by the above data showing that macrophages infiltrating the stroma of CIN-3 lesions and tumors produced MMP-9, a pro-angiogenic protease, and by the emerging appreciation of ZA as an inhibitor of angiogenesis and of MMP activity. Moreover, N-BPs have been shown to impair the function of macrophages in vitro (32, 33). We asked, therefore, whether ZA therapy was affecting the abundance of infiltrating macrophages, their expression of MMP-9, or the conversion of pro–MMP-9 into active MMP-9. We assessed the abundance of both F4/80+ (data not shown) and CD68+ macrophages (Figure 6) and of MMP-9–expressing cells in the stroma of the ZA-treated cervix, compared with vehicle controls following a 6-week PT. Using double-labeled immunofluorescence, we detected only a slight decrease in the number of infiltrating macrophages in ZA-treated mice (10% reduction, data not shown). In contrast, we observed a 71% reduction of MMP-9 expression in macrophages in the stroma proximal to CIN-3 lesions (Figure 6, A and B) and a 73% reduction of MMP-9 in macrophages associated with tumors relative to vehicle-treated controls (data not shown). We detected a similar inhibition of MMP-9 expression by macrophages (68% reduction) in the RT targeting preexisting cervical carcinomas (Figure 6B). Since ZA did not completely inhibit MMP-9 expression, we measured the total gelatinolytic activity in tissues of ZA-treated compared with control mice by gelatinase assay. Consistent with the immunofluorescence results (Figure 6, A and B), we observed a strong inhibition of the total gelatinolytic activity in ZA-treated mice in both PTs and RTs compared with controls (Figure 6C). We also performed gelatin zymography to reveal the pro form and active forms of MMP-9. In agreement with gelatinolytic biochemical assay, after 6 weeks of treatment (in a PT), ZA reduced the abundance of the MMP-9 pro form 70% (reflecting the reduced number of MMP-9–expressing macrophages) and almost completely inhibited the activation of latent pro–MMP-9, in that the smaller active form of the protease was barely detectable (Figure 6D). We observed a slight decrease in the abundance of the MMP-2 pro form, although the significance is unclear, since we did not detect appreciable MMP-2 activity in either the untreated control or the ZA-treated neoplastic cervix (Figure 6D). These data suggest a dual effect of ZA: it suppressed the expression of MMP-9 by infiltrating macrophages, and it inhibited the proteolytic activity of the remaining (perhaps preexisting) MMP-9 in the stroma associated with CIN-3 lesions and cervical carcinomas.
ZA inhibits MMP-9 expression and activation in macrophages. (A) Reduced MMP-9 expression was detected in macrophages in the stroma adjacent to CIN-3 in ZA-treated mice compared with controls as revealed by colocalization of MMP-9 (green) and CD-68 (red). Arrows show MMP-9–expressing macrophages; arrowheads indicate macrophages that do not express MMP-9. (B) Quantification of double-labeled MMP-9+/CD-68+ cells (5 fields per mouse) revealed a 71% reduction in the PT (n = 8 control, n = 6 ZA-treated) and a 68% reduction in the RT (n = 10 control, n = 8 ZA-treated). Similar results were obtained with double-labeled MMP-9+/F4/80+ cells (not shown). #P < 0.001. (C) Gelatinase activity in tissue extracts was measured by incubation with fluorescin-conjugated gelatin in the absence or presence of the MMP inhibitor 1,10 Phe (4 mM). Gelatinase activity was lower in ZA-treated compared with control cervixes in both PT and RT. **P < 0.01 versus control. N/E2 indicates estrogen-treated normal cervix. (D) Zymography showing the pro- and active forms of gelatinases in tissue lysates from both control and ZA-treated mice. Pro- and active forms of MMP-9 and MMP-2 are indicated by arrows. Scale bar: 25 μm. Values are mean ± SEM. P values were calculated using the Wilcoxon test.
Gene KO mice reveal the functional importance of MMP-9 for cervical carcinogenesis. The dual MMP-9 inhibitory effects of ZA treatment clearly suggested that MMP-9 expressed by macrophages is critical for tumor angiogenesis and cancer progression in the cervix. Most MMP inhibitors, while possibly selective for a particular protease, typically inhibit multiple MMPs (23, 30, 31). To assess whether MMP-9 was functionally involved in regulation of angiogenesis and tumor growth, we used a genetic approach involving HPV16 mice carrying gene KOs of either MMP-2 or MMP-9, afforded by the fact that both protease KO mice are viable (34, 35). To functionally evaluate the roles of MMP-9 and MMP-2 in cervical cancer progression, we analyzed the cervical cancer phenotype in HPV/E2 mice that had been intercrossed to MMP-9–null or MMP-2–null mice (all maintained in the FVB/n background) to render them deficient in one or the other protease. The tumor phenotype was evaluated 6 months later and compared with HPV/E2 control mice at the same age. HPV/E2–MMP-2–null mice (n = 15) exhibited the same tumor incidence as HPV/E2 control mice (n = 40) and had an inconsequential 16% reduction in tumor volume (P = 0.728) compared with controls (Figure 7, A and B). In contrast HPV/E2-MMP-9–null mice (n = 25) evidenced a 66% reduction in tumor incidence and a 76% reduction in tumor volume (P < 0.001) (Figure 7, A and B). Immunostaining with Meca-32 Ab revealed reduced vessel density in HPV/E2-MMP-9–/– mice, whereas no difference in vascularity was observed in HPV/E2-MMP-2–/– mice, compared with controls (data not shown). Therefore, genetic ablation of MMP-9 demonstrated that MMP-9 activity is functionally significant for tumor progression in the cervix. Moreover, the phenotypic effects of the gene KO were remarkably similar to the responses seen in the ZA trails, consistent with the hypothesis that the actions of ZA involve MMP-9 as a principal target.
ZA inhibits the formation of VEGF/VEGF-R2 complexes on the neoplastic vasculature, phenocopying a MMP-9 gene KO. (A) Decrease in tumor incidence in 7-month-old HPV/E2-MMP-9–/– (66% reduction; n = 25) compared with HPV/E2-MMP-2–/– (n = 15) and control mice (n = 40). (B) Reduction of tumor volume in 7-month-old HPV/E2-MMP-9–/– (76% reduction; n = 25) compared with HPV/E2-MMP-2–/– and control mice. Tumor volume and histological scores were determined as described in Methods. (C–G) VEGF/VEGF-R2 complex was detected by immunohistochemistry using GVM39 Ab (green), while the ECs were visualized with the anti–Meca-32 Ab (red). Significant reduction in the VEGF/VEGF-R2 complex was observed on vessels in HPV/E2-MMP-9–/– (77% reduction compared with controls; n = 15) and ZA-treated (73% reduction compared with controls; n = 8) mice compared with HPV/E2-MMP-2–/– (n = 10) and controls (n = 20) (C). Representative analyses of tumors are shown in D–G. Arrows indicate the VEGF/VEGF-R2 complex associated to the vessels. Scale bar: 25 μm. Values are mean ± SEM. #P < 0.001. P values were calculated using the Wilcoxon test.
ZA impairs VEGF mobilization to its receptor by MMP-9. VEGF-A associates with its principal signaling receptor VEGF-R2 in the highly angiogenic lesions of the cervical carcinogenesis pathway (Figure 1C). Motivated by our previous observation that MMP-9 could mobilize VEGF and increase its association with VEGF-R2 (13), we sought to determine whether MMP-9 was also serving such a function in cervical neoplasias, and, if so, was ZA limiting VEGF mobilization? We analyzed the association of VEGF with VEGF-R2 in CIN-3 lesions and in tumors in ZA-treated HPV/E2 mice, in MMP-2 and MMP-9 gene KO HPV/E2 mice, and in control HPV/E2 mice. As expected from unaltered tumor incidence and volume, VEGF ligand-receptor complexes were readily detected in the vasculature of MMP-2–deficient tumors (Figure 7, C and E) and CIN-3 lesions (data not shown), similar to vessels of control mice (Figure 7, C and D). In contrast, the vessels in tumors or CIN-3 lesions of MMP-9–null mice showed little or no VEGF/VEGF-R2 complex on ECs (Figure 7, C and F; the data for CIN-3 lesions are analogous but not shown). Notably, the immunostaining of vessels with GVM39 Ab in ZA-treated mice (in both prevention and regression trials) revealed a similarly reduced association of VEGF with VEGF-R2 in tumors (Figure 7, C and G) and in CIN-3 lesions (data not shown). A comparable reduction of GVM39 immunostaining (74% reduction compared with controls) was also observed in BB94-treated cervical carcinomas (see Supplemental Figure 4, A–C) and CIN-3 lesions (data not shown), supporting the interpretation that ZA is inhibiting angiogenesis by interfering with MMP-9–mediated mobilization of VEGF.