Fistulising Crohn’s disease: MMPs gone awry

Fistulae are a common complication of Crohn’s disease (CD), and their most common perianal manifestation is present in 14–38% of CD patients in referral populations.¹ Despite advances in conservative treatment, fistulae rarely heal, while surgical resection is effective but frequently does not prevent local recurrence or fistulising disease at other sites.¹ Remodelling of the extracellular matrix (ECM) is a key event in chronic bowel inflammation,^2,3 especially in CD which is characterised by both active fibrogenesis (that is, ECM production and deposition), leading, for example, to stricture formation, and by enhanced fibrolyis (that is, breakdown and removal of ECM), such as occurs in fistula formation. While fibroblasts and myofibroblasts, and to a minor degree endothelial and epithelial cells, produce the intestinal ECM, many more cell types are involved in ECM breakdown by releasing a broad spectrum of enzymes that can degrade essentially every ECM component, such as the various collagens, non-collagenous (glyco-) proteins, glycosaminoglycans, and proteoglycans. The most important of these enzymes are the matrix metalloproteinases (MMPs), a structurally related class of at least 20 zinc dependent proteases.^4,5 MMPs are classified according to their primary specificities and structural features into interstitial collagenases (MMP-1, MMP-8, MMP-13, MMP-18), gelatinases (degrading denatured and basement membrane collagens: MMP-2, MMP-9), stromelysins (degrading a broad spectrum of ECM substrates: MMP-3, MMP-7, MMP-10, MMP-11), elastases (MMP-12), and membrane-type (transmembrane) MMPs (with broad substrate specificities: MMP-14, MMP-15, MMP-16, MMP-17, MMP-24, MMP-25).

As uncontrolled MMP activity would virtually lead to dissolution of organs, MMP activity is strictly controlled, not only at the transcriptional and translational levels, but also by the requirement of proteolytic activation of the zymogens and by several protease inhibitors. Proteolytic activation occurs by the plasminogen activator-plasmin cascade and by certain active MMPs themselves, such as MMP-3, MMP-10, MMP-14, and MMP-24. The most important physiological inhibitors are the tissue inhibitors of metalloproteinases (TIMP), mainly TIMP-1, which blocks almost all MMPs, and TIMP-2, which is specific for MMP-2. A fifth level of control is localisation of secreted MMPs either at cell membrane compartments that are involved in local ECM proteolysis (for example, by binding to ECM receptors that mediate cellular locomotion through the matrix^6,7) or on certain ECM components, such as collagens. Interestingly, it is preferentially the inactive proforms of MMPs that bind to collagen which become released from the ECM on proteolytic activation.^8,9

This illustrates that both homeostasis and proteolytic remodelling of the ECM are highly sophisticated and tightly regulated processes, and that simply measuring upregulation or downregulation of MMPs and their inhibitors will give a very incomplete picture of ECM remodelling at best. In this line, no characteristic difference is found between procollagen mRNA levels as a measure of fibrogenesis² and MMP and TIMP mRNA levels as a measure of fibrolysis^10–12 in inflamed colonic specimens from patients with CD and ulcerative colitis. This was initially unexpected as CD often leads to stenosis due to excess ECM deposition while fibrosis is uncommon in ulcerative colitis. But it comes as no surprise when one considers the complex regulation of fibrolysis in the transmural inflammation of CD.

The paper by Kirkegaard and colleagues¹³ in this issue of Gut brings our understanding of the role of MMPs in CD a step further [see page 701]. Using complementary techniques, they identified and investigated MMP-3 and MMP-9 as key players in fistulae of CD and other causes. At first glance the occurrence of fistulae, a manifestation of enhanced fibrolytic activity, in a fibrostenotic intestinal disease may appear paradoxical. However, severe inflammation which always leads to release and activation of MMPs by cells of the inflammatory infiltrate can drive both processes: fistulae will form when there is no rapid compensatory fibrogenic response to fill up the defect, perhaps favoured by a too quick re-epithelialisation or re-endothelialisation; fibrosis will develop when the fibrogenic response is strong and quick, leading to scar formation. Both processes can obviously coexist in close proximity. What are the factors determining these divergent pathways? Recent studies demonstrated that at least early activation of MMPs is necessary to allow for a fibrogenic response,^14,15 either by destroying basement membranes which have a role in maintaining cellular quiescence and differentiation in the gut¹⁶ with subsequent mesenchymal activation, or by allowing mesenchymal cell migration and proliferation by removal of constraining ECM.

Thus the cellular source and temporospatial pattern of MMP release and activation probably explain these divergent pathologies. As mesenchymal cells (fibroblasts, myofibroblasts, and to a lesser extent endothelial cells) and inflammatory cells (mainly macrophages, monocytes, and neutrophils) are the major sources of MMPs and TIMPs, there must be a differently tuned interplay of these cells driving either fistula or fibrosis. Kirkegaard et al observed that acute fistulising inflammation in CD and other aetiologies is characterised by high expression of MMP-3 and MMP-9 coupled with high activity of MMP-2 and MMP-9 in inflammatory cells, while in chronic fistulae, MMP-9 almost disappeared but MMP-3 expression was maintained, with a shift to mesenchymal, mainly myofibroblastic, cells.¹³ The finding that TIMPs remained low in all fistula specimens is of interest as TIMP-1 in particular is a major determinant of fibrogenesis, as exemplified in liver fibrosis.¹⁷ Low expression of TIMP-1 is also characteristic of osteoarthritis, the hallmark of which is unopposed MMP activation in the joint leading to destruction of cartilage and bone, somewhat reminiscent of fistula formation in the gut.¹⁸ Also, it is osteoarthritis that has led to a novel concept of chronic fibrolytic inflammation that is driven by altered synovial fibroblasts, the equivalent of which could be myofibroblasts in chronic fistulising diseases. These fibroblasts/myofibroblasts can secrete atypical chemokines attracting fibrolytic inflammatory cells on the one hand, and have a fibrolytic and migratory phenotype themselves on the other.¹⁹ MMP-3, which has a broad substrate specificity and which acts as a proactivator (in addition to plasmin), at least of MMP-1, MMP-2, and MMP-9, has been identified as a particularly aggressive enzyme in intestinal inflammation³ which favours invasiveness.²⁰ This does not exclude an important contribution by MMP-12, MMP-13, MMP-14, and MMP-19 which are expressed in inflamed intestine²¹ but were not considered in this study.

Can we derive therapeutic concepts from this and related studies on intestinal inflammation? A drug that effectively prevents emerging or eliminates chronic fistulae would be most welcome. Anti-inflammatory, antibiotic, and immunosuppressive approaches can lead to long term fistula control or even closure in 25–50% of patients, roughly doubling the placebo response which is unsatisfactory.¹ Anti-tumour necrosis factor α (TNF-α) strategies can lead to good short term and occasionally long term results in cases that are refractory to these therapies but treatment is costly and may have grave side effects.¹ None the less, this approach is rational as TNF-α is a potent inducer of MMP-3 in intestinal myofibroblasts.³

Based on this and other studies, four novel strategies aimed at inhibiting intestinal inflammation and remodelling in inflammatory bowel disease in general, or in fistulising CD in particular, merit further exploration. These are the use of: (1) inhibitors of growth factors or cytokines, such as epidermal growth factor, hepatocyte growth factor, insulin-like growth factor (IGF-I and IGF-II), fibroblast growth factor (aFGF, bFGF), interleukin 1, and stem cell factor that induce MMPs in epithelial and mesenchymal cells²²; (2) synthetic inhibitors of kinases that trigger MMP production, especially p38 kinase²³; (3) synthetic MMP inhibitors that were shown to ameliorate experimental colitis in mice²⁴; and (4) inhibitors of integrins (that is, cellular receptors for ECM proteins that can bind to and/or induce certain MMPs).^6,25–28 Inhibition of growth factors/cytokines and kinases always carries the risk of side effects because of their importance in normal tissue regeneration and of their ubiquitous expression in other cells and organs. Inhibition of MMPs appears more promising, especially as the first generation of broad spectrum inhibitors which have multiple side effects are becoming replaced by orally available subtype specific agents.¹⁸ However, the necessity for cell-type specific targeting (for example, to MMP-3 expressing myofibroblasts of fistulae) remains an unsolved problem. Therefore, it is attractive to target integrins that promote MMP-expression as many of these receptors are cell-type specific. Thus MMP-1, MMP-2, and MMP-3, MMP-14, MMP-1 and MMP-3, and MMP-9 are induced and activated by integrins α2β1, α4β1, α6β1, and αvβ3, respectively, which are predominantly found on activated (myo-) fibroblasts.^6,25–28 Such inhibitors have (in part) been developed, and orally available agents are in phase I and II clinical studies (for example, for metastatic tumour disease).²⁹ It remains to combine the most promising of these agents, alone or in combination, with established therapies in preclinical and clinical studies, for our patients with complicated inflammatory bowel disease.