This study has presented novel applications for an amino-bisphosphonate, ZA, currently used clinically to ameliorate bone metastasis. The strategy forthcoming involves treatment of either CIN or established cervical cancer so as to elicit stasis or regression rather than progression and growth. We demonstrate that ZA inhibits angiogenesis in premalignant lesions (CIN-3) and in cervical tumors by targeting both macrophage expression and proteolytic activity of an MMP, MMP-9. Therapeutic regimes using ZA impaired progression of premalignant lesions to invasive carcinoma and antagonized the growth of preexisting cervical tumors. These results are likely to be relevant to human cervical carcinogenesis. MMP-9 has been shown to be upregulated in CIN-3 lesions and cervical cancers in humans (refs. 7
and our unpublished observations), and elevated expression and activity correlate with poor prognosis for subsequent metastasis and tumor recurrence (37
). Notably, MMP-9 and the related MMP-2 have been detected in both epithelial and stromal compartments in human cervical lesions (8
). We found MMP-9 expression exclusively in the stroma in the mouse model in infiltrating macrophages. MMP-9 is similarly expressed by macrophages (and other inflammatory cells), infiltrating both stroma and epithelia of human CIN-2/3 and cervical cancers (E. Giraudo and D. Hanahan, unpublished data); the reasons that MMP-9–expressing macrophages do not infiltrate the neoplastic epithelial compartments in the mouse model are presently unclear. We performed a genetic test to assess the importance of this and a related protease by analyzing the cervical cancer phenotype in HPV/E2
mice that carried a gene KO of either MMP-9 or MMP-2. The HPV/E2
-MMP-9–null mice had a markedly lower tumor incidence and a statistically significant reduction of tumor volume as well as reduced vascularity compared with control HPV/E2
mice, whereas HPV/E2
-MMP-2–null mice were largely unaffected, revealing MMP-9 as the key gelatinolytic MMP regulating cervical cancer progression in this mouse model.
These results are consistent with previous data showing in the related K14-HPV16 skin model that MMP-9 but not MMP-2 activation coincided with the angiogenic switch in premalignant lesions (18
) and that MMP-9, supplied by bone marrow–derived cells, was sufficient for skin carcinogenesis in mice otherwise deficient in MMP-9 (17
). Furthermore, MMP-9 but not MMP-2 has been implicated in triggering the angiogenic switch during pancreatic neuroendocrine carcinogenesis in RIP1-Tag2 transgenic mice, in part by increasing the bioavailability of VEGF-A to its receptor (13
). Thus, although MMP-9 and MMP-2 have historically been considered very similar in substrate specificity, there is increasing evidence that they have distinctive targets as well as different patterns of zymogen activation. It is notable that the active form of MMP-2 was not increased in abundance, in marked contrast to the significant increase of MMP-9 activity during cervical carcinogenesis. Consistent with this difference between MMP-2 and MMP-9, RT-PCR analysis showed increased expression only of MMP-9 transcripts during carcinogenesis, whereas levels of MMP-2 and other MMPs variously implicated in malignant phenotypes, such as MMP-3, MMP-11, MMP-12, MMP-14, were unchanged, while others, MMP-8, MMP-10 and MMP-13, were not detectably expressed. Notably, the effects of genetically ablating MMP-9 were similar to those resulting from ZA therapy, consistent with the proposition that ZA’s targeting of MMP-9 is indeed a key part of its biological activity in this context. Taken together, analyses of this mouse model and the cognate human cancer predict that MMP-9 functionally contributes to human cervical carcinogenesis resulting from an infiltration of neoplastic lesions by MMP-9–expressing macrophages.
ZA is an unconventional metalloprotease inhibitor with multiple modes of action.
Our results complement and extend a series of recent publications indicating that ZA and other BPs do more than just inhibit osteoclast functions in the bone. Several reports have documented antiangiogenic activity of ZA in cell culture and in vivo angiogenesis bioassays (26
) or showed that other BPs antagonize both angiogenesis and tumor growth in traditional subcutaneous xenotransplant tumor models (28
). ZA and other N-BPs also have specific effects on intracellular signaling pathways. For example, N-BPs have been shown to interfere with RAS signaling in cultured tumor cells, apparently by inhibiting the mevalonate biosynthetic pathway, which limits protein prenylation of Ras and other small GTP-binding molecules and thereby disrupts cellular functions (38
), including expression of VEGF (28
). Similar intracellular effects are evident in macrophages and osteoclasts in vitro following treatment with BPs (33
). Interestingly, distinct BPs may differentially affect macrophages in different contexts: a recent study of macrophages isolated from human peripheral blood treated with two BPs, clodronate and pamidronate, revealed a complex modulation of MMP-9 expression and secretion (41
), but not the suppression we observe in vivo with ZA. It is possible, therefore, that ZA has differential capability to repress MMP-9 gene expression and activity and/or that macrophages infiltrating tissues such as the cervix may respond differentially to BPs than peripheral blood monocytes; these possibilities deserve future investigation. In light of our results, it is reasonable to postulate that ZA is suppressing signaling pathways regulating MMP-9 biosynthesis and secretion by macrophages infiltrating the cervix, since we see only minimal reductions in the abundance of macrophages associated with CIN-3 and cervical cancers despite an almost complete loss of MMP-9 immunoreactivity (by IHC) and protease activity (by gelatin assay and zymography). We infer that the susceptibility of macrophages to ZA reflects their lineage relationship to osteoclasts (42
), which selectively take up BPs (32
). The biochemical mechanisms by which ZA inhibits MMP-9 gelatinolytic activity in this tissue context remains to be defined, as does the molecular basis by which it suppresses biosynthesis in macrophages. MMP-9 activity is known to be regulated at many levels (45
), including gene transcription, secretion of the latent pro form, activation of the latent form by proteolytic cleavage (45
), and enzymatic efficiency in the face of endogenous protease inhibitors. We suspect that ZA is interfering both with conversion of latent to active MMP-9 and with enzyme function, the latter due to ZA’s recognized ability to chelate zinc (30
), a necessary cofactor for MMP-9 activity. Finally, it is possible that ZA is affecting the interaction of MMP-9 with tissue inhibitors of metalloproteases, now known to upregulate and downregulate MMP activity through complex interactions (48
Seeking to investigate the mechanistic basis for the observed histological responses in the neoplastic lesions, we assessed the frequency of apoptosis and found substantially increased frequencies of apoptosis in both transformed cervical epithelia and ECs, much as we and others have seen in the course of analyzing the effects on MMP-9 gene KOs and angiogenesis inhibitor therapies in other mouse models of cancer (13
). The data are consistent with the hypothesis that the antineoplastic responses are due to impaired angiogenesis and reduced vascularity, which results in large part from the targeting of MMP-9. Consistent with this hypothesis, we found that MMP-9–deficient HPV/E2
and ZA-treated HPV/E2
mice have similarly reduced association of VEGF with VEGF-R2 in both CIN-3 lesions and cervical carcinomas. It is conceivable that ZA is, in addition, directly affecting the angiogenic ECs, as has been reported for cultured ECs (26
). In regard to ZA’s ability to inhibit MMP-9, it is pertinent to note that MMP-9 is involved in mobilization of circulating endothelial progenitors cells (CEPs) and hematopoietic stem cells (HSCs) from the bone marrow (50
). While ZA has no effect on the abundance of infiltrating macrophages in cervical neoplasias, it is possible that mobilization of CEPs and HSCs is affected, which warrants future consideration.
In regard to other potential targets of ZA action, we cannot exclude the possibility that ZA also has a direct, proapoptotic effect on the neoplastic epithelia, and, indeed, there are reports of such activity against tumor cells in vitro (38
) and in vivo (39
). Our data show that ZA is not a traditional antimitotic chemotherapeutic drug in the cervix in that it does not reduce the proliferation rates of transformed cervical epithelial cells in CIN-3 lesions or cervical tumors. Nevertheless, we leave open the possibility that ZA is affecting the HPV16-expressing keratinocytes both indirectly (by inhibiting both MMP-9 activation and expression by macrophages and thereby impairing angiogenesis), as well as directly. Finally, we cannot rule out the possibility that MMP-9 is also antagonizing angiogenesis by generating an antiangiogenic fragment of the collagen IVα3
chain, called tumstatin, as Hamano et al. have recently shown (54
); our data indicate that in balance MMP-9 is pro-angiogenic in the cervix, as shown both by ZA treatment and in HPV/E2
-MMP-9–null mice. The cervical carcinomas are analyzed at defined end points (6–7 months of age), when the tumors are relatively small compared with the large tumors where Hamano et al. observed their effects. MMP-9 (and its inhibition by ZA) could have more complex effects in the context of large tumor burden, particularly if tumstatin mobilization by MMP-9 proves to be a factor in cervical cancer pathways. Notably, in another mouse model, that of pancreatic islet carcinoma, several classical MMP inhibitors have been found to be efficacious against early-stage but not late-stage tumors when used as monotherapy (perhaps because of the “tumstatin effect”); by contrast, a combinatorial regimen involving an MMPI along with “metronomic chemotherapy” (55
) produced regression and survival benefit against large end-stage pancreatic tumors (56
). This experience leads us to further suggest that ZA should also be considered for inclusion in combinatorial regimens when targeting large tumors and/or late-stage disease to circumvent a potential tumstatin effect.
We have analyzed the angiogenic phenotype and assessed a novel antiangiogenic therapy in a mouse model of cervical carcinogenesis that is driven by the etiologic viral oncogenes, faithfully recapitulating the lesional stages of the human disease. Having identified MMP-9 and its macrophage source as functional contributors to angiogenesis and tumor progression, we evaluated a targeted therapy involving a clinically approved drug, the amino-bisphosphonate ZA, which demonstrated promising efficacy against both premalignant lesions and cancers of the cervix. In the course of a PT, ZA inhibited angiogenesis and limited the progression of premalignant precursors to invasive carcinomas of the cervix, suggesting that ZA (or better yet an orally bioavailable derivative) might be considered for adjuvant treatment following excision of CIN-3 lesions for women at high risk for recurrence and progression. We have also shown that ZA is able to interfere with the growth of established cervical carcinomas, producing partial responses as a monotherapy. Perhaps ZA could be added to the standard of care for cervical cancer (57
) to assess its potential for improving efficacy by inhibiting angiogenesis. Furthermore, given that we have identified molecular and cellular targets for ZA, thereby increasing the body of evidence (20
) that MMP-9 and tumor-enhancing macrophages are functionally important for angiogenesis and carcinogenesis in a variety of organs, it may be reasonable to consider incorporating ZA into conventional therapeutic regimens for additional tumor types and, indeed, other diseases with an angiogenic component where these biomarkers are present. In that regard, it may be pertinent to consider more frequent dosing than the current monthly schedule, which was based on ameliorating pain associated with bone metastasis and not on dose-limiting toxicity or clinical antineoplastic efficacy; instead, perhaps suppression of MMP-9 expression in tissue-infiltrating macrophages (or imaging of its protease activity in the target tissue) may be an appropriate biomarker for optimizing an efficacious antiangiogenic dosing regimen.