Bisphosphonates were initially used either as corrosion inhibitors or as complexing agents in the textile, fertilizer or oil industries 11. It was shown in the 1960s that pyrophosphate (Fig. 1) inhibits calcification of tissues by binding to hydroxapatite but that orally administered pyrophosphates are hydrolysed in the gut. Bisphosphonates are resistant to hydrolysis and so could be administered orally 12. The first bisphosphonate used for human treatment was etidronate for myositis ossification 13. It was discovered that bisphosphonates inhibited osteoclastic-mediated bone resorption 1415, and this led to their use as bone protective agents.

All bisphosphonates have the same generic structure (Fig. 1). The hydrolysable oxygen atom that separates the two phosphate groups in pyrophosphates is replaced with a more stable carbon atom. The P–C–P structure is responsible for giving bisphosphonates their high affinity for bone, which can further be enhanced by the substitution of a hydroxyl group for the R1 side chain 16. The R2 side chain is important for conferring the antiresorptive potency to the bisphosphonates. Extensive modifications of the R2 side chain showed that a basic primary nitrogen group attached to an alkyl chain, such as in pamidronate and alendronate (Fig. 2), produced more potent antiresorptive agents (10-fold to 1000-fold) than the earlier generation bisphosphonates, such as etidronate, with a single methyl group side chain.

When the nitrogen atom was combined as a tertiary amine in the R2 side chain, such as in ibandronate and olpadronate, the bisphosphonates were even more potent. However, the most potent bisphosphonates to date are those that contain the nitrogen within a cyclic structure, such as in risedronate and zoledronate. These cyclic nitrogen structures are up to 10,000-fold more active than etidronate. The structures of the side chains of the bisphosphonates used for human therapy are shown in Figure. 2 along with their relative potency at inhibiting bone resorption in rats as compared with etidronate 17.

Bisphosphonates were initially thought to affect bone resorption by a physical process that directly inhibited the dissolution of mineralisation 11. However, the effective levels of some newer bisphosphonates to inhibit bone resorption were well below the levels needed to have a physical effect. Bisphosphonates predominantly affect bone resorption by altering the cellular metabolism, although their ability to bind to the calcified matrix is important to localise them to bone.

Bisphosphonates target the osteoclast directly either by increasing apoptosis or by affecting metabolic activity 11. The circulating levels of bisphosphonates are extremely low as they are rapidly absorbed by bone, suggesting that circulating levels are not relevant to function. This is further supported by the fact that a single large dose of bisphosphonates can have a sustained effect on bone resorption 11. To avoid the potential gastrointestinal problems associated with the newer nitrogen-containing bisphosphonates, both risedronate and ibandronate are undergoing clinical trails to determine whether monthly injection is a suitable delivery method for these bisphosphonates 11.

During bone resorption, the osteoclasts demineralise the extracellular matrix of the bone, and bisphosphonates are released from the bone surface and absorbed by osteoclasts 1819. Cellular uptake of bisphosphonates leads to the loss of the ruffled border between the osteoclast and the bone surface 2021, to disruption of the cytoskeleton 1922 and to loss of function. Both alendronate and tildronate can inhibit several protein tyrosine phosphatases 232425 and can also affect small GTPases such as Rho and Rac, explaining some of the cytoskeletal effects. Bisphosphonates also inhibit the osteoclast proton pumping H⁺ ATPase 262728 and lysosomal enzymes, which contributes to the loss of function.

Bisphosphonates can cause osteoclasts 29, macrophages 3031 and myeloma cell lines 3233 to undergo apoptosis. The induced osteoclast apoptosis may be caused by a general disruption that pushes the cells towards apoptosis, which may account in part for the inhibition of bone resorption. Bisphosphonates also inhibit osteoclast differentiation, the formation of osteoclast-like cells in culture 34 and the generation of mature osteoclasts in culture 353637; these effects will contribute to the prevention of bone resorption.

The more potent nitrogen-containing bisphosphonates inhibit the farnesyl diphosphate synthase enzyme of the mevalonate pathway 38 that is responsible for producing cholesterol and isoprenoid lipids. Two of the isoprenoid lipids, farnesyldiphosphate and geranyldiphosphate, are required for the normal prenylation of the small GTPases such as Ras, Rho and Rac; a process essential for the correct functioning of these enzymes 39 (see Fig. 3). These small GTPases control the osteoclast cell morphology, the cytoskeletal arrangement, membrane ruffling, the trafficking of vesicles and apoptosis 394041. It is believed that the more potent nitrogen-containing bisphosphonates inhibit osteoclast function by inhibiting these small GTPases. This effect on protein prenylation in osteoclasts has been demonstrated in vivo for aldronate 42.

Bisphosphonates that have a structure similar to pyrophosphate (e.g. chlodronate and etidronate) become incorporated into nonhydrolysable analogues of ATP 4344, which accumulate within the osteoclast leading to impaired function. Chlodronate, etidronate and tiludronate can all be metabolised in mammalian cells 4245, via the cytoplasmic aminoacyl-tRNA enzymes. ATP analogues accumulate within the cytoplasm, where they interfere with numerous biological processes, eventually causing both osteoclast and macrophage apoptosis 42. This appears to have been confirmed when the nonhydrolysable ATP analogue metabolite of chlodronate produced identical effects to that seen for chlodronate alone 4246. Encapsulated chlodronate works in an identical manner to cause apoptosis in macrophages in vivo by a buildup of nonhydrolysable ATP products in the cytoplasm 42. The more potent bisphosphonates that contain a nitrogen in the side chain are not metabolised in this way 152546.

Bisphosphonates inhibit calcification by binding to the surface of solid calcium phosphate crystals and acting as crystal poisons affecting both crystal growth and dissolution 47. There is a positive correlation between the binding effects of the various bisphosphonates and their ability to inhibit crystallisation 48, further supporting a physical mechanism.

Bisphosphonates are excellent inhibitors of bone resorption, with their potency varying according to the structure of the side chains. Treatment with bisphosphonates reduces the steady-state level of resorption dependent upon the administered dose 4950. Many different osteoporosis models have been investigated 515253545556. Bisphosphonates are also effective in decreasing bone loss and increasing mineral density in postmenopausal osteoporosis 575859606162 and corticosteroid-induced bone loss 63. Bisphosphonates improve the biomechanical properties of bone in both normal animals and models of osteoporosis 5164656667 and, along with hormone replacement therapy, calcium and vitamin D supplementation, have led to a significant improvement in the management of osteoporosis. It has also been demonstrated that, in humans, bisphosphonates inhibit tumour-induced bone resorption, correct hypercalcaemia, reduce pain, prevent the development of new osteolytic lesions, prevent fractures and, consequently, improve the quality of life for the patients 476869707172.

Considerable progress has been made in the design of new and effective bisphosphonates. The original assumption that the mechanism of action of these compounds involved a strong physical interaction with the mineral phase only partially explains their action. It is now recognised that many of the effects result from interfering with essential cellular functions of osteoclasts. Some actions of the bisphosphonates can be separated, with different roles for the backbone and side chains of the molecule. In the future, it is probable that specific bisphosphonates will be produced that can target individual metabolic pathways within the cell to produce more bone-specific actions with less action on neighbouring cell types, reducing the occurrence of side effects.

The first synthetic MMP inhibitors bound tightly at the active site and therefore blocked enzyme activity. Effective inhibitors mimicked the peptide sequences around the cleavage site in the substrate, and the scissile bond was replaced by a chelating group, such as hydroxamate, that bound to the active site zinc (Fig. 5). Other chelating groups such as carboxylic acid, thiol and phosphorus have been used and large numbers of inhibitors have been made 109. Early inhibitors (e.g. batimastat, BB-94) showed broad-spectrum inhibition for many MMPs but showed poor oral availability. Marimastat (BB-2516; British Biotech, Oxford, UK), a chemically modified form of batimastat, shows similar broad-spectrum MMP inhibition and is also orally active. Initial designs focused on broad-range inhibitors as it was thought that many MMPs were involved in cancer. Many inhibitors were produced using conventional pharmaceutical screening processes before crystal structures were available, and the detailed chemical design of MMP inhibitors is reviewed in 109110.

As the crystal structures of the catalytic domains of MMP-1, MMP-2, MMP-3, MMP-7, MMP-8, MMP-13, MMP-14 and MMP-16 became available, new nonpeptide inhibitors with increased specificity for individual MMPs were made. At least some of the variation in substrate specificity among MMPs can be explained by differences in the six specificity subsites in the active site cleft and surrounding sequences (Fig. 5). The first subsite on the carboxy-terminal side of the substrate scissile bond, the S' pocket, is particularly important; for example, it is deeper in MMP-3 and larger in MMP-8 than it is in MMP-1. This, and differences out to the S' 4 position, offer possibilities for designing specificity in synthetic inhibitors. Two examples of nonpeptidyl hydroxamate inhibitors are prinomastat (AG-3340; Agouron, San Diego, USA) inhibitor, which is selective for gelatinases over collagenases, and Ro32-3555 (Roche, Basel, Switzerland), which is an effective inhibitor of collagenases but has less potency against MMP-2 and MMP-3.

Many early studies were involved with the treatment of cancer. For example, marimastat (BB-2516) (an orally administered hydroxamate inhibitor of MMPs with limited ability to inhibit sheddase activity) has been in clinical development since 1994 111. A recent study showed that marimastat significantly improved the survival of patients with advanced gastric cancer 112.

CellTech (Slough, Berks, UK) developed highly specific gelatinase inhibitors and proposed that these could be effective for the treatment of cancer and bone resorption 113. Agouron developed prinomastat (AG3340) 114, a MMP inhibitor with a K_i value in the picomolar range for the inhibition of MMP-2, MMP-9, MMP-13 and MMP-14, with lower activity against MMP-1 and MMP-7 115. Chiroscience (Cambridge, UK) developed D2163, now licensed to Bristol-Myers Squibb (BMS-275291; New York, USA), for the potential treatment of cancer. This compound inhibits a broad range of MMPs associated with cancer but does not inhibit shedding events, and it gives a 10-fold increase in systemic exposure for a given dose when compared with marimastat. BMS-275291 does not exhibit any deleterious side effects with tendons and joints. The compound prevents angiogenesis in a mouse model, and phase I studies in healthy volunteers have been completed showing that good plasma levels were achieved and that the compound is well tolerated. BMS-275291 has now entered phase II studies, and the results will be reported in 2003.

Novartis (formerly Ciba-Geigy; Basel, Switzerland) published information regarding an orally active hydroxamate MMP inhibitor, CGS 27023A. This is a broad-spectrum inhibitor with a nanomolar K_i against MMP-1, MMP-2, MMP-3, MMP-9, MMP-12 and MMP-13, and is chondroprotective in both the rabbit menisectomy model of OA and the guinea pig model of spontaneous OA 115.

Another inhibitor, tanomastat (BAY 12–9566; Bayer Corporation, West Haven, CT, USA), targets MMP-3, MMP-2, MMP-8, MMP-9 and MMP-13, with low activity against MMP-1, and was proposed for the treatment of OA. It is effective in guinea pig and canine models of OA 116, and human trials of BAY 12–9566 given to 300 OA patients for 3 months reported no musculoskeletal side effects. The drug was detectable in human cartilage of treated patients undergoing joint replacement 117. However, BAY 12–9566 was withdrawn from an 1800-patient phase III trial in OA following negative results in a separate cancer trial of the same drug 115 (see Safety of MMP inhibitors).

Trocade (Ro 32–3555; Roche, Welwyn Garden City, UK), a selective collagenase inhibitor, was used in phase III trials for the treatment of RA. It has a low nanomolar K_i against MMP-1, MMP-8 and MMP-13, with approximately 10-fold to 100-fold lower potency against MMP-2, MMP-3 and MMP-9. It blocks IL-1-induced collagen release from cartilage explants and, in vivo, has prevented cartilage degradation in a rat granuloma model, in a Propionibacterium acnes-induced rat arthritis model and in an OA model using the SRT/ORT mouse 117. Clear evidence of protection to bone and cartilage was apparent even where active inflammation was present. Trocade had no effect on acute inflammation in rodent models, so presumably did not inhibit TNFα converting enzyme at these concentrations 118.

Large-scale trials of trocade in RA patients, however, were terminated because of a lack of efficacy. Future data will show whether adequate concentration of this compound within the joint was achieved and whether the dosing schedule used was appropriate. This was the first large-scale trial in RA, and its failure to complete does leave the future of therapies targeted at the collagenases in jeopardy 119. The clinical evaluation of these drugs is difficult as long trials have to be conducted with radiographs being the most reliable measure of joint damage. While some progress has been made with the use of magnetic resonance imaging, this technology is not yet been proven and routine centres do not have access to a validated method for quantitation.

Several studies have shown that the antibiotic tetracycline and its derivatives inhibit MMPs. Micromolar concentrations of tetracycline are sufficient to inhibit collagenase activity by 50%; greater activity is seen with some modified tetracyclins 120. Doxycycline hyclate (Periostat^®; CollaGenex Pharmaceuticals, Newtown, PA, USA), at a subantimicrobial dose, is the only MMP inhibitor approved by the US Food and Drug Administration as an adjunct therapy in adult periodontitis 120. The results of patient studies for the use of tetracyclins to treat patients with rheumatic disease are equivocal 121, although there were positive trends. There is evidence that, in combination with nonsteroidal anti-inflammatory drugs, these compounds can be effective, and further studies are planned with the more potent derivatives 120. One advantage of the tetracyclins, if they prove to be effective, is that they are already in clinical use with a known side-effect profile.

Bisphosphonates may directly inhibit the activity of several MMPs as tiludronate inhibited MMP-1 and MMP-3 but did not affect MMP production in periodontal ligament cells 122. Chlodronate was able to inhibit MMP-8 in peri-implant sulcus fluid 123.

When any new class of drugs is used for the first time it will raise issues concerning safety. MMPs are involved in a large variety of physiological processes so the rate of wound healing, growth and foetal development could all be affected. There is a balance between matrix synthesis and matrix breakdown in connective tissues, so inhibition of MMPs could cause deposition of a matrix, leading to fibrosis, although dose ranging studies should avoid such complications.

The most advanced safety data available is for marimastat, where musculoskeletal pain and tendonitis are identified as reversible side effects in treated patients 124. These effects commence in the small joints of the hand and spread to the arms, shoulders and other joints if the treatment is continued. The symptoms are time and dose dependent and could be reversible. These symptoms are also seen with the Roche compound Ro 31–9790, and this led to its development as an arthritis treatment being stopped. Some suggest that these side effects are caused by inhibition of MMP-1, but the effect is reproducible in rodents, which only express MMP-1 during embryogenesis 105, and no such events were noted with trocade, which inhibits MMP-1 very effectively. It is probable that inhibition of an uncharacterised sheddase enzyme, similar to TNFα converting enzyme, contributes possibly via effects on inflammation. All new compounds can be very effectively screened in rodent models for these musculoskeletal events and those which cause side effects discarded.

The Bayer compound BAY 12–9566 was withdrawn as it was associated with increased tumour growth and poor survival times in small cell lung cancer, but no other cases of these effects have been reported 125. It is not logical to assume that an effect seen with one member of this class of compounds will automatically be seen by all and that there are significant differences in chemical structure and metabolism of individual inhibitors.

Inhibition of cartilage collagen destruction still remains a viable target to prevent joint destruction in arthritic disease. However, the trials of MMP inhibitors have been extremely disappointing. New agents are still under development and these may overcome some of the problems of both delivery and side effects. A key to future success is to identify the specific MMPs that are responsible for arthritic tissue destruction in the different diseases. This will allow highly specific inhibitors that target individual enzymes and potentially reduce side effects. It should be possible to screen all new inhibitors across the MMP family to ensure specificity. A variety of explanations have been offered to explain why the metalloproteinase inhibitors have been unsuccessful in clinical trials in patients with joint diseases. In addition to a wealth of in vitro and animal in vivo data, much of the evidence supporting their role in human disease has shown that they are 'at the scene of the crime'. There is no doubt that MMPs are present and active in joint diseases. However, it could be that the MMPs or collagenases, in spite of their presence within the joint, are the wrong target in the arthritides, and that other enzyme systems such as cathepsin K predominate 126127.

If MMPs are responsible for cartilage damage, however, then it is possible that some compounds may not penetrate the cartilage/bone synovial interface and are therefore ineffective. Compounds may be altered in body compartments so that their specificity for individual MMPs is altered. An unrecognised MMP may be the major player in collagen turnover but is not inhibited by the available inhibitors, which were screened against a limited set of available MMPs. Altering the balance between active MMPs and TIMPs with an artificial inhibitor may upregulate the synthesis of more MMP and so promote destruction. It could be that activation is the major control point, and inhibition of the enzyme systems responsible for activation is the key to success. Alternatively, MMPs are involved in many cellular functions and the side-effect profiles of compounds that would prevent joint destruction have meant that effective compounds are excluded during development.

Although synthetic inhibitors of MMPs have so far proved to be ineffective, there are several alternative approaches that can be considered. Much research is being directed towards the role of cytokines and signalling pathways in the regulation of the MMPs in disease. Blocking of certain cytokines or inhibiting the inflammatory cell signalling pathways may produce an alternative approach to inhibiting MMP production and activity. AntiTNFα therapy has proved successful for treating RA patients, and the levels of MMPs are reduced. Gene therapy approaches in which chondroprotective cytokines are overexpressed in affected joints show that the overexpression of IL-4 and IL-13 in experimental models of arthritis prevents MMP-induced cartilage destruction. It may also be possible to increase the expression of the TIMPs. For example, calcium pentosan polysulphate stimulates the production of TIMP-3 in human synovial fibroblasts and rheumatoid synovium without affecting MMP production 128.

As more detailed information about the structure of MMPs and their interaction with substrates becomes available, it may be possible to design inhibitors that target areas of the enzyme other than the active site. For example, the C-terminal haemopexin-like domain of collagenases has long been known to be required for collagenolysis, presumably because of interactions with the substrate 129. The activation of the proenzyme is also a valid target, again requiring a detailed knowledge of the underlying biology. An understanding of the regulation of expression of both MMPs and TIMPs at the molecular level may allow us to modulate the levels of both enzymes and inhibitors expressed by cells during disease. This will require detailed analysis of the signalling pathways involved in transforming a cytokine signal at the cell surface to induce the expression of proteinase or inhibitor. Considerable effort is being expended in preventing the production of MMPs by interfering with the cytokine signalling pathways 130. Finally, there is interest in the synthesis of modified TIMPs that are specifically targeted to inhibit specific enzymes 131132133.

It is interesting that, when patients are treated with the new anticytokine therapies, a greater protection of the joint is obtained when this is combined with more conventional treatments. It is probable that blocking of MMPs will be more effective if combined with treatments that target earlier steps in inflammation. Furthermore, as noted earlier, MMPs are not alone in being implicated in joint disease. Serine proteinases are believed to be involved in MMP activation, and cysteine proteinases have been shown to degrade collagen. It may be necessary to combine proteinase inhibitors, either in sequence or with other agents that hit other specific steps in the pathogenesis, before the chronic cycle of joint destruction found in these diseases can be broken.