Experimental studies performed prior to 1990 led to the widely held belief that matrix metalloproteinases (MMPs) produced by cancer cells are of critical importance in tumor invasion and metastasis. Based on this evidence, the pharmaceutical industry produced several well tolerated, orally active MMP inhibitors (MMPIs) which demonstrated efficacy in mouse cancer models. Phase III clinical trials initiated in 199798 using marimastat, prinomastat (AG3340), and BAY 12-9566 alone or in combination with standard chemotherapy in patients with advanced cancers (lung, prostate, pancreas, brain, GI tract) have recently been reported; no clinical efficacy was demonstrated. Bayer and Agouron have discontinued their ongoing Phase III drug trials of MMPIs in advanced cancer. In retrospect, the failure of MMPIs to alter disease progression in metastatic cancer might have been anticipated since MMPs appear to be important in early aspects of cancer progression (local invasion and micrometastasis) and may no longer be required once metastases have been established. Our understanding of MMP pathophysiology in cancer has expanded considerably in the past 10 years. Current views indicate that: (1) most MMPs in tumors are made by stromal cells, not carcinoma cells; (2) cancer cells induce stromal cells to synthesize MMPs using extracellular matrix metalloproteinase inducer (EMMPRIN) and cytokine stimulatory mechanisms; and (3) MMPs promote cell migration and the release of growth factors sequestered in the extracellular matrix. MMPs have a dual function in tumor angiogenesis: MMP-2 and MT1-MMP are required in breaking down basement membrane barriers in the early stage of angiogenesis, while other MMPs are involved in the generation of an angiogenic inhibitor, angiostatin. In spite of considerable recent progress in identifying multiple roles of MMPs in disease, our understanding of MMP function in cancer is far from complete (see Table 1). Based on accumulated data, it is recommended that future MMPI trials focus on: (1) patients with early stage cancer; (2) the use of MMPIs along with chemotherapy; (3) the measurement of MMPs in tumor tissue and blood as a means of identifying patients who are more likely to respond to MMPI therapy; and (4) identification of biomarkers that reflect activation or inhibition of MMPs in vivo.
Background leading to the development of matrix metalloproteinase inhibitors in cancer treatment
During the past 40 years, the pharmaceutical industry has focused the vast amount of its resources in the oncology field on developing drugs that kill cancer cells. Although considerable progress has been made in the treatment of selective types of cancers (leukemias, lymphomas, testicular cancer), the more common forms of human cancers (lung, gastrointestinal, and prostate cancer) remain relatively resistant to standard cytotoxic chemotherapy. Frustration with this avenue of cancer treatment has been a stimulus to develop other approaches.
The inability to control metastasis is the leading cause of death in patients with cancer. Control of metastasis, therefore, represents an important therapeutic target. Based on a sound understanding of the biochemistry of matrix metalloproteinases (MMPs) and the accumulation of considerable experimental evidence implicating MMPs in cancer dissemination, the pharmaceutical industry has invested heavily in developing effective MMP inhibitors (MMPIs) for the treatment of cancer. Since MMPs also play an important role in tumor angiogenesis (Fang et al., 2000; Stetler-Stevenson, 1999), MMPIs may have a dual role in the treatment of cancer.
Several recent review articles have presented an optimistic view of the use of MMPIs in cancer treatment; much of this optimism was based on preclinical studies in experimental animals and model systems (Belotti et al., 1999; Curran and Murray, 1999; DeClerck and Imren, 1994; Johnson et al., 1998; Nelson et al., 2000; Parsons et al., 1997; Wojtowicz-Praga et al., 1997). The results of randomized Phase 3 clinical trials using three different MMPIs have recently been reported; none of these trials have demonstrated clinical effectiveness of MMPIs in advanced cancer. In this report, we will emphasize that the use of MMPIs in patients with pre-existing metastasis is like locking the barn door after the horse has bolted out of the stable. The scientific and industrial communities have learned an important lesson, which should be useful in planning new protocols for use of MMPIs at earlier stages of cancer.
Before discussing the details of clinical trials, a review of the basic biology, chemistry, cell biology and pathophysiology of MMPs will be presented.
Structure and function of MMPs
Although serine, cysteine, and aspartic proteases have been implicated in various aspects of cancer progression, most of the recent emphasis has been on MMPs. The family of MMPs now includes more than 20 related enzymes that degrade a variety of extracellular matrix components (Birkedal-Hansen, 1995; Nagase and Woessner, 1999; Parks and Mecham, 1998; Stetler-Stevenson, 1999). Both a descriptive name, typically based on a preferred substrate, and an MMP numbering system have been used; the latter will be used in this review except for the membrane type-MMP sub family (MT-MMP).
All MMPs share several highly conserved domains, including an activation locus [PRCGXPD] in the amino-terminal pro domain and a zinc atom binding domain [VAAHExGHxxGxxH] in the active site (catalytic domain), with three histidines coordinating the zinc (Birkedal-Hansen et al., 1993). A pre region (signal peptide) targets the proteins for secretion or insertion into the plasma membrane. The cysteine residue in the N-terminal pro-domain maintains the enzyme in an inactive state by coordinating the active-site zinc atom in the catalytic domain. This link must be broken by proteolytic cleavage or conformational modification to permit function of the active enzyme (cysteine switch mechanism). The majority of MMPs have additional domains including a hemopexin-like C-terminal domain (shaped like a beta propeller fold with pseudo-fourfold symmetry) which is important in substrate recognition and in inhibitor binding, and a hinge region which links the hemopexin and catalytic domains. MMP-2 and MMP-9 also contain a fibronectin-like region.
Three MMPs that attack interstitial collagens have been identified: MMP-1, MMP-8 and MMP-13. These collagenases cleave the triple helical collagen molecule at a single site. Two MMPs (MMP-2 and MMP-9) are capable of cleaving basement membrane type IV collagen, but their enzymatic activity is far greater against gelatins; hence they are often referred to as gelatinases. MMP-2 and MMP-9 are also responsible for further degradation of the large 3/4 and 1/4 collagen fragments and other proteins including fibronectin, laminin, and elastin. Two MMPs, MMP-3 and MMP-10, have been identified as stromelysins and degrade various proteoglycan components of the extracellular matrix (ECM), as well as fibronectin, laminin, and gelatin. MMP-11 (stromelysin-3) does not degrade extracellular matrix proteins efficiently, but is effective in degrading the serine proteinase inhibitor α-1 anti-proteinase inhibitor (anti-trypsin) and thus may potentiate the action of serine proteinases. MMP-7 (matrilysin) is a short, truncated proteinase which degrades nonfibrillar collagen, fibronectin, and laminin. MMP-12 (metalloelastase) is capable of degrading native elastin.
A subset of six membrane type MMPs (MT-MMPs) contain a transmembrane domain of approximately 20 amino acids which attaches the enzymes to the cell surface and a short cytoplasmic domain (Sato et al., 1994) which is involved in intracellular trafficking (Lehti et al., 2000; Nakahara et al., 1998). MT-MMPs activate proMMP-2 at the cell surface, leading to enhanced cellular invasion in vitro (Sato et al., 1994). A stoichiometric concentration of tissue inhibitor of metalloproteinases (TIMP-2), functioning as a connecting molecule between MT-MMP and proMMP-2, is required for proMMP-2 activation, but an excess of TIMP-2 is inhibitory. MT1-MMP is also capable of hydrolyzing fibrillar collagen and digesting vitronectin, laminin, fibronectin, enactin (d'Ortho et al., 1997; Ohuchi et al., 1997; Pei and Weiss, 1996) fibrinogen, and fibrin (Hiraoka et al., 1998).
MMP activity is tightly regulated both intracellularly at the level of gene expression and following secretion by the action of activators of proenzymes and inhibitors. MMPs are generally produced by cells at very low levels, however cellular expression is rapidly induced at times of tissue remodeling. MMPs are transcribed and secreted by the constitutive secretory pathway, except in the case of neutrophils, macrophages (Birkedal-Hansen et al., 1993), and Paneth cells (Lopez-Boado et al., 2000); these cells store MMPs in secretory granules. Cytokines, chemokines, and growth factors play a key role in modulation of most MMP secretion, especially during inflammation, wound healing, and cancer (Crawford and Matrisian, 1996; Saren et al., 1996). MMP-2 is constitutively produced (Long et al., 1998). MMPs are also sequestered as inactive zymogens in the ECM after secretion, thereby providing a reservoir of latent enzyme (Yu and Woessner, 2000).
Activation of MMPs is achieved by removal of the N-terminal prosequence of approximately 80 amino acids to yield mature enzyme (Birkedal-Hansen et al., 1993). Although the physiologic activation of MMPs is a matter of speculation, the initial cleavage event can be carried out in vitro by a variety of serine proteinases including plasmin, trypsin, and neutrophil elastase. Some activated MMPs can further activate other proMMPs. MMPs containing a furin-like recognition domain (RXKR) in their propeptides, MMP-11, MMP-23 (Velasco et al., 1999), and MT-MMPs (Sato et al., 1994; Strongin et al., 1995) are activated intracellularly in the trans-Golgi network by a group of calcium-dependent transmembrane serine proteinases (furin/PACE/kex-2).
Activated MMPs are modulated by endogenous proteinase inhibitors including TIMP-1, -2, -3, -4 which specifically regulate MMPs in tissues and α2-macroglobulin (α2-M), which is a broad-spectrum protease inhibitor prominently displayed in plasma (Birkedal-Hansen et al., 1993; Stetler-Stevenson, 1999). All of the TIMPs are capable of inhibiting all of the MMPs following formation of tight non-covalent 1 : 1 complexes. The exception to the rule is that TIMP-1 is a poor inhibitor of MT-MMPs. In addition to binding to the active site of MMPs, the C-terminal domains of TIMP-1 and TIMP-2 form complexes with the C-terminal domains of proMMP-9 and proMMP-2, respectively. These complexes control the activity of the bound MMPs.
Controlled regulation of MMP activity is associated with normal physiologic processes including ovulation, trophoblast invasion, mammary involution and embryonic development. It has been proposed that inappropriate overexpression of MMPs or under expression of TIMPs constitute part of the pathogenic mechanism in cancer and other diseases demonstrating tissue injury, such as rheumatoid arthritis, multiple sclerosis (Kieseier et al., 1999), aortic aneurysms, arterial restenosis lesions (Libby, 1995), and bullous skin disorders (Liu et al., 1998).
Role of MMPs in cancer dissemination
A positive correlation between tumor progression and the expression of multiple MMP family members (MMP-1, MMP-2, MMP-7, MMP-9, MMP-11, MT1-MMP) in tumor tissues has been demonstrated in numerous human and animal studies (Ellerbroek and Stack, 1999; Stetler-Stevenson et al., 1996). The ratio of activated to total MMP levels, especially MMP-2, has also been correlated with tumor aggressiveness (Davies et al., 1993). Based on numerous preclinical studies, it was proposed that targeting MMP activity may provide a mechanism to prevent cancer dissemination (Stetler-Stevenson et al., 1996). Further support for the role of MMPs in cancer dissemination came from the demonstration that TIMPs were able to interfere with experimental metastasis (DeClerck and Imren, 1994; Montgomery et al., 1994).
Atypical aspects of MMP/TIMP function in cancer
A considerable amount of information has gradually accumulated to suggest that the role of MMPs in cancer is far more complicated than initially presumed. For example, increased TIMP-1 levels in human cancer tissues have been associated with poor prognosis (Lu et al., 1991). It is uncertain whether this reflects the growth potentiating properties of TIMPs (Guedez et al., 1998; Hayakawa et al., 1994). Other experimental studies employing cancer cells transfected with TIMP-1 cDNA demonstrated that MMPs may act primarily to alter the extracellular environment to allow sustained growth in an ectopic site as opposed to having a specific role in allowing the cells to extravasate from the blood stream (Cameron et al., 2000; Chambers and Matrisian, 1997; Nelson et al., 2000). In some experimental tumor systems, however, increased MMP production did not correlate with increased metastasis (Zucker et al., 1992). One potential explanation is that excess proteolysis may degrade matrix signals and receptors, thereby disrupting cell matrix interactions and inhibiting migration (Parks and Mecham, 1998).
Cleavage of matrix components release polypeptide fragments with new biological properties, as well as releasing signaling components embedded within the matrix (Figure 1). For example, cleavage of laminin-5 (a component of the ECM) by MT1-MMP and MMP-2 promotes migration of cells (Giannelli et al., 1997; Koshikawa et al., 2000). A number of soluble growth factors are secreted and then stored in an inactive form bound to extracellular matrix molecules. During enhanced proteolysis, these factors are then freed to act on their target receptors. MMP-3 has been shown to cleave the matrix molecule decorin, resulting in the release of transforming growth factor-β (TGF-β) in its more biologically active form (Imai et al., 1997). Furthermore, receptors for growth factors are targeted for proteolysis by MMPs including FGF type I receptor, which mediates the effects of fibroblast growth factor (see McCawley and Matrisian, 2000). IGF binding protein-1 has been identified as a potential physiologic substrate for MMP-11. Although IGF binding proteins can confer latency on IGF-I and IGF-II, their degradation can restore the activity of these growth factors, thereby affecting tumor growth (Noel et al., 1997).
Although it seems counter intuitive, other aspects of MMP function are potentially beneficial in slowing tumor progression. In such cases, MMPIs may be harmful if they interfere with the body's natural mechanism to inhibit cancer (Figure 1). For example, the synthesis of MMP-7 and MMP-9 in endothelial cells has been reported to be increased in integrin α1-knockout mice. Based on the capacity of these MMPs to cleave plasminogen and produce the angiogenesis inhibitor, angiostatin, mice exhibiting high tumor levels of these MMPs and high plasma levels of angiostatin, had decreased tumor vascularity and slower tumor growth compared to their wild-type counterparts (Pozzi et al., 2000). While the data from these experiments can be otherwise interpreted (i.e. angiostatin is generated by a non-MMP mechanism), the inverse correlation between MMP levels and tumor size seems indisputable. It remains to be determined whether a similar scenario occurs in human cancers.
Production of MMPs by tumor stromal cells
As a result of numerous studies of oncogene transformed cells, cancer cell lines, and experimental tumor models, it had long been assumed that cancer cells were responsible for producing the MMPs in human tumors (Liotta, 1992; Nicolson, 1991). This concept came under attack in 1990 when Basset et al. (1990) reported that stromal fibroblasts surrounding tumor cells, not the tumor cells themselves, were responsible for producing MMP-11 in human breast cancer. Other investigators have similarly employed in situ hybridization to demonstrate the localization of MMP-1, MMP-2, MMP-3 and MT1-MMP mRNA primarily in stromal fibroblasts, especially in proximity to invading cancer cells, but not in the carcinoma cells in human breast, colorectal, lung, prostate, and ovarian cancers (Nelson et al., 2000). MMP-9 has been localized to inflammatory cells (macrophages and neutrophils), rather than fibroblasts or tumor cells in colorectal cancer tissue. Immunolocalization studies which identify MMPs using specific antibodies, rather than mRNA expression, however, have generally identified MMP-2 and MMP-9 protein in cancer cells. This data reinforces the concept that tumor cells have docking sites which bind stromal cell secreted MMPs, i.e. MMP-9 binding to collagen type IV components on the cell surface (Olson et al., 1998), and to CD-44 (Yu and Stamenkovic, 2000) and MMP-2 binding to αvβ3 integrin (Brooks et al., 1996). Hence, tumor cells may function as a receptacle for stromal MMPs.
An explanation for the production of MMPs by reactive stromal cells in a tumor came from the discovery of Extracellular Matrix MetalloPRoteinase INducer, EMMPRIN, by Biswas (Biswas et al., 1995). EMMPRIN is an intrinsic plasma membrane glycoprotein produced in high amounts by cancer cells, which stimulates local fibroblasts to synthesize MMP-1, MMP-2 and MMP-3. Tumor cell interactions with fibroblasts via EMMPRIN leads to fibroblast-induced local degradation of basement membrane and ECM components, thus facilitating tumor cell invasion (Figure 2). On examination of human lung and breast cancer tissue, EMMPRIN expression in cancer cells far exceeded that of normal epithelial cells (Polette et al., 1997). Recent studies have demonstrated that MMP-1 binds to EMMPRIN on the tumor cell surface, thus indicating that following EMMPRIN stimulation of MMP synthesis in fibroblasts, a surface localized MMP-1/EMMPRIN complex arms the cancer cell for degradation of the ECM (Guo et al., 2000). The possibility of developing EMMPRIN inhibitors for treatment of cancer is a future consideration.
MMP-7 represents an exception to the generalization that fibroblasts and inflammatory cells produce the MMPs in a tumor; MMP-7 is produced in neoplastic epithelial cells in human cancer tissue and not in stromal cells (Nelson et al., 2000). Recent data suggests that MMP-7 cleavage of Fas ligand is an important mediator of epithelial cell apoptosis (Powell et al., 1999).
Role of MMPs in tumor angiogenesis
A role for MMPs, especially MMP-2, in tumor neoangiogenesis has been elegantly demonstrated in experimental models (Fang et al., 2000). MMPs are required for vascular endothelial cells to penetrate their underlying basement membrane in order to produce capillary sprouts (Haas and Madri, 1999; Stetler-Stevenson, 1999). MT1-MMP appears to be important in selected aspects of neonatal angiogenesis (Holmbeck et al., 1999; Zhou et al., 2000), whereas both MMP-2 (Itoh et al., 1998) and MT1-MMP are probably involved in tumor angiogenesis. Vascular endothelial growth factor (VEGF) can indirectly activate MMP-2 in endothelial cells (in the presence of coagulation factors) by inducing tissue factor synthesis leading to thrombin-induced activation of pro-MMP-2 on the cell surface (Zucker et al., 1998) (Figure 2).
Development of MMP inhibitors as novel anti-cancer agents
Based on the logical but unproved concept that cancer metastasis in humans requires cancer cells to release MMPs that digest surrounding connective tissues (Liotta and Stetler-Stevenson, 1991; Stetler-Stevenson et al., 1996), the pharmaceutical industry invested considerable effort over the past 10 years to develop safe and effective MMPIs for use in patients with cancer. More recently, it has been recognized that the pathogenetic function of MMPs in cancer is far more complicated than initially conceived. An attractive aspect of inhibiting MMP activity in cancer is that the target of the anti-cancer therapy will include components produced by non-malignant cells in the tumor stroma; these cells presumably will not mutate and develop drug resistance (McCawley and Matrisian, 2000).
Initially, drugs were targeted to the chemical functional group that chelates the active site zinc(II) ion which is a ubiquitous feature of MMPs. Peptide and peptide-like compounds have been designed which combine backbone features (P1, P1′, P2′, p53′ regions) which favorably interact with the enzyme subsites (S1, S1′, S2′, S3′ pockets) and functionality capable of binding zinc (chelator) in the catalytic site (Gomis-Ruth et al., 1997; Grams et al., 1995). These drugs essentially mimic the collagen substrate of MMPs, and thereby work as competitive, potent, but reversible inhibitors of enzyme activity (Brown and Whitaker, 1999). Both broad spectrum and partial selective inhibitors of MMPs have been developed. Selective inhibitors provide greater specificity and theoretically, more safety than broad-spectrum MMP inhibitors. On the other hand, experimental studies have revealed that several of the MMP family members are coexpressed in different types of cancer, thereby making it difficult to identify a single MMP as being critical to the disease process.
MMPIs that featured a hydroxamic acid zinc binding group were initially identified as highly potent compounds. The majority of inhibitors currently in clinical testing have been designed with this functional group (Grams et al., 1995). Early examples of these compounds included the pseudopeptide hydroxamic acid MMP inhibitor, batimastat® (British Biotech) which showed broad specificity in inhibition of members of the MMP family, while displaying minimal activity against other classes of metalloproteinases (Brown and Whittaker, 1999). Specific enzyme recognition and selectivity were obtained by structural modification of the P1′ (primary) and P2′/P3′ region(s), taking advantage of distinctions in the depth and composition of the S1′, S2′ and S3′ pockets of individual MMPs (Beckett and Whittaker, 1998). Orally active, broad spectrum compounds were first developed by British Biotech (Brown and Whittaker, 1999). Workers at Agouron used a structure-based inhibitor design program that employed high resolution X-Ray crystallography to discover potent inhibitors that are selective for the deep-pocket (MMP-2, MMP-3, MMP-8, MMP-9, MMP-13 and MT1-MMP) over the shallow-pocket enzymes (MMP-1 and MMP-7). Orally absorbed prinomastat was developed using this approach with the expectation that sparing MMP-1, which was suspected to be responsible for arthralgia side effects, would increase the therapeutic index of MMPIs (Shalinsky et al., 1999).
Preclinical studies of MMPIs
Numerous studies using MMPIs as drugs in cancer models have demonstrated their ability to delay primary tumor growth and block experimental metastasis. Both syngeneic and xenograft tumor models have been employed. Batimastat and the orally effective marimastat have IC50 values for MMP-1, MMP-2, MMP-7, MMP-9 and MMP-12 ranging between 316 nM. Despite marked inhibition of metastasis with MMPIs, the animals eventually developed local and distant metastases (Brown and Whittaker, 1999). Initiation of batimastat (or prinomastat) when tumor burden was minimal has a more profound effect on tumor growth inhibition than initiation of treatment at the time of large tumor bulk (Eccles et al., 1996; Shalinsky et al., 2000). One excellent study described the use of batimastat in a mouse model of pancreatic islet cell angiogenesis and multistage tumor progression. This study included: (1) a prevention trial in which batimastat was given before tumor nodules had progressed beyond the stage of carcinoma in situ, (2) an intervention trial in which treatment was given when small tumors were present, and (3) a regression trial in which treatment was delayed until large tumors had developed. Batimastat produced a 49 per cent reduction in angiogenesis in the prevention trial, an 83 per cent reduction in tumor burden in the intervention trial, but had no effect on large, invasive tumors (Bergers et al., 1999). This latter observation is of importance in examining clinical trial results of MMPIs employed in advanced human malignancies. The use of concurrent versus sequential administration of synthetic MMPIs with cytotoxic agents in early-stage tumors has also been addressed. Survival after implantation of human tumor xenografts in nude mice was extended several fold in mice treated sequentially with chemotherapy followed by batimastat, whereas survival was considerably less in mice treated with each single agent (Giovazzi et al., 1998).
Workers at Agouron have reported extensive testing of the effects of prinomastat in various tumor models. The lack of permanent suppression of tumor growth across animal models and the recognition that other factors are involved in neoplastic progression, supported the case for use of prinomastat in combination with cytotoxic chemotherapy (Shalinsky et al., 1999). Prinomastat alone and in combination with cytotoxic chemotherapy, was active against many human tumors in immunodeficient mice and in syngeneic metastasis models (Shalinsky et al., 2000). The antitumor efficacy of prinomastat in mice was associated with maintaining minimum effective plasma concentrations of prinomastat (Shalinsky et al., 1998, 1999). Additionally, resistance to prinomastat did not develop after extended treatment in vivo.
BAY 12-9566, a novel, non-peptide biphenyl MMPI with a Zn-binding carboxyl group, was developed by Bayer Corporation with selective for MMP-2, MMP-9, MMP-11, MMP-13 and MT1-MMP (IC50 613 nM). BAY 12-9566 has a very long terminal plasma half-life (90100 h) related to its extremely high plasma protein binding fraction (>99.99%) (Gatto et al., 1999). After removal of a primary human breast cancer in nude mice, BAY 12-9566 resulted in inhibition of tumor regrowth and reduction in pulmonary metastasis (see Nelson et al., 2000).
Design of clinical trials of MMPIs in patients with cancer
Guidelines for development of human trials for testing MMPIs in cancer were discussed in 1994 (Zucker, 1994). These included: (1) selecting cancer types which have a high probability of metastasis in spite of standard local treatment (surgery/radiotherapy) directed at the primary tumor; (2) selecting cancer types with high levels of MMPs; (3) treating patients with the best standard treatment modality (chemotherapy/radiotherapy/surgery) prior to or simultaneously with an MMPI; and (4) initiating treatment prior to the development of metastases and continuing treatment until the likelihood of metastasis is diminished. Unfortunately, it was not feasible to meet all of these trial design objectives in the initial Phase III trials.
Clinical pharmaceutical trials of MMPIs in patients with cancer were begun in 1997. Results employing marimastat, prinomastat, and BAY 12-9566 have been recently presented. Phase III trials with MMPIs determined the optimal biological dose rather than the maximum tolerable dose; this represented an important conceptual advance in cancer therapy. Given the limitation of secondary end points of response to MMPIs, clinical development of MMPIs rapidly proceeded to phase III trial design, with the primary end points of survival or progression-free survival.
Batimastat, which is poorly soluble and consequently has poor bioavailability when administered orally or parenterally, was the first MMPI tested in phase I studies (Rasmussen, 1999). This compound was superseded by marimastat, an orally active agent. The initial pharmacokinetic work done in healthy volunteers demonstrated a linear doseplasma concentration relationship with a mean half-life of 810 h. Plasma concentrations at all dose levels studied were well in excess of required inhibitory concentrations, indicating that oral administration of the drug produces pharmacological active drug levels (Rasmussen, 1999).
Three different approaches have been taken in phase IIIII MMPI trials (Nelson et al., 2000). The first involved a direct comparison between a MMPI and standard chemotherapy; these trials include either marimastat or BAY 12-9566 versus gemcitabine in pancreatic cancer. The second strategy involved concomitant administration of a MMPI with chemotherapy compared with chemotherapy alone. These trials include: (1) marimastat plus gemcitabine in pancreatic cancer; (2) prinomastat in combination with paclitaxel and carboplatin in non small cell lung cancer; (3) prinomastat in combination with cisplatin and gemcitabine in non small cell lung cancer; and (4) prinomastat in combination with mitoxantrone and prednisone in hormone refractory prostate cancer. The third strategy compares a MMPI with placebo in patients with relatively low-volume disease or no overt clinical evidence of disease after standard chemotherapy. These trials include: (1) marimastat in pancreatic cancer; (2) marimastat in unresectable glioblastoma; (3) marimastat in advanced gastric cancer; (4) marimastat in small cell lung cancer; (5) marimastat in non small cell lung cancer; (6) marimastat in breast cancer; (7) BAY 12-9566 in small cell lung cancer; (8) BAY 12-9566 in non small cell lung cancer; and (9) BAY 12-9566 in ovarian cancer. It should be emphasized that the likelihood of metastasis at the time of entry into the trial in each of these disease settings is very high; hence the poor prognosis of each of these patient categories.
Results of early phase IIIII trials of MMPIs
Marimastat has been tested in phase II dose ranging studies in more than 400 patients with far advanced pancreatic, ovarian, colorectal, and prostate cancer. Cancer specific antigens were used as surrogate markers for biologic activity (CA 19/9 in pancreatic cancer, CA 125 in ovarian cancer, CEA in colorectal cancer, and PSA in prostate cancer). Analysis of these studies by British Biotech using non-standard methodology suggested that marimastat treatment significantly reduced the rates of rise of all four cancer-specific antigens in a dose-dependent fashion (Rasmussen, 1999).
Phase III trials of marimastat in patients with unresectable pancreatic cancer and post-chemotherapy progressing gastric cancer have been presented. Therapeutic superiority of marimastat over gemcitabine in pancreatic cancer was not demonstrated. A recent trial of marimastat in glioblastomas also revealed a lack of efficacy of the drug. In a large randomized double-blind placebo controlled study of patients with inoperable gastric adenocarcinoma following achievement of stable disease with chemotherapy, patients were randomized to oral marimastat versus placebo. Median survival was 167 days for marimastat versus 135 days for placebo (P=0.07). Exclusion of patients who did not receive any study treatment accentuated the differences (P=0.046). Progression-free survival was also significantly improved by marimastat. The authors concluded that this is the first definitive clinical data supporting the use of MMPIs in cancer (Fielding et al., 2000).
Bayer has reported the results of several clinical trials. In a large pancreatic cancer trial, gemcitabine alone resulted in significantly prolonged overall survival (6.4 mo.) and progression-free survival as compared to BAY 12-9566 alone (3.2 mo.) (Moore et al., 2000). Interim analysis of a drug trial of BAY 12-9566 versus placebo in patients with stable disease following cytoreductive chemotherapy for advanced small cell lung cancer resulted in shorter survival in drug-treated patients. Based on these results, all clinical trials of BAY 12-9566 were suspended in September 1999. Additional preclinical models are in progress to investigate the potential reasons for this unexpected result.
In August 2000, Agouron reported an interim analysis of the use of prinomastat in combination with chemotherapy in the treatment of patients with advanced prostate cancer after failure of hormonal therapy and of patients with advanced non small cell lung cancer. In both disease settings, prinomastat did not demonstrate clinical efficiency. As a result of these findings, Agouron has prematurely ended these clinical trials. Future trials of prinomastat will focus on early stage cancer (Shalinsky et al., 2000).
In terms of safety, the only clear-cut drug related toxicity identified so far with hydroxamate-derived MMPIs is a characteristic musculoskeletal syndrome consisting of tendonitis manifested by joint pain, stiffness, edema, reduced mobility, and skin discoloration (Giovazzi et al., 1998). BAY 12-9566 has a different side-effect profile which included asymptomatic elevation of hepatic enzymes and mild thrombocytopenia.
In terms of the therapeutic use of MMPIs in other diseases, Roche has recently discontinued trials of Ro 32-3555 in rheumatoid arthritis because of lack of efficacy in a well controlled clinical study. Needless to say, the failure of the initial clinical MMPI trials appears to have put a damper on initiation of new pharmaceutical trials of MMPIs in cancer.
It is impossible for us to resist the editorial temptation of being a Monday morning quarterback. The probability of failure of MMPIs in late stage human cancer was predictable from preclinical studies. What do these results tell us about the potential usefulness of MMPIs in early stage cancer? Actually, the negative reports to date have not been very instructive. However, a positive trend for benefit was suggested by marimastat in earlier-stage gastric cancer. Additionally, one MMPI, BAY 12-9566, is contraindicated in selected cancers. In contrast, marimastat and prinomastat have not produced adverse effects in cancer patients. Clinical drug trials of MMPIs in selected early stage cancers, with procurement of tumor tissue for examination of potential tumor response markers, seems like a next logical step. The problem with this proposal is the competition with more established anti-cancer drug trials for entry of patients with early disease into clinical studies. Furthermore, the high cost of performing these types of drug trials leads pharmaceutical companies to demand strong preliminary data predicting success of the study. In that regard, the lack of success of the recent clinical trials of MMPIs has been a major setback for the industry. Hopefully, a better understanding of the role of MMPs in cancer progression will soon be at hand, thus providing more insight in developing the drug protocols of tomorrow. In the future, the identification of surrogate markers to identify response to MMPIs is a high priority for advancing the use of MMPIs in disease (Greenwald et al., 1999).
The authors would like to thank Dr David Shalinsky for helpful advice in preparation of this manuscript and Michelle Hymowitz for production of the figures.