Fell-Muir Lecture: Metalloproteinases: from demolition squad to master regulators

Abstract

Two families of the Metzincin clan of metalloproteinases, the matrix metalloproteinases and the disintegrin metalloproteinases have attracted much attention as important effectors of cellular interactions with their environment. They appear to play significant roles in the modulation of components of the extracellular matrix, matrix and cell receptors, as well as the cytokines and growth factors and their receptors. Such functions at the ‘cutting edge’ of cell biology puts these enzymes in pivotal roles in the orchestration of the rapid response of cells to their environment, acting as key switches between different signalling pathways. Inevitably such enzymes should be regarded as suitable targets for therapeutic approaches to many diseases where such pathways become dysregulated. A major challenge to the development of direct inhibitors of catalysis has been the broad structural similarity of the Metzincin catalytic site. More detailed knowledge of active site structures has helped to some extent to resolve the development of more specific chemical inhibitors and selected enzymes are now being targeted. An alternative strategy is the consideration of the role of the extracatalytic domains that are determinants of specificity at a variety of levels. Dissecting the relationships between structure and function of these interaction sites is allowing the development of new approaches to inhibition of enzyme function. Antibodies are proving useful tools in this respect and may pave the way to a novel biologics approach to disease therapy.

Introduction

Of the 566 proteases identified in the human genome, 273 are classed as ‘extracellular’ (Overall & Blobel 2007) and may be viewed as having potential roles in modulating the way that cells interact with their environment. In reality, the 16 intramembrane and 277 ‘intracellular’ enzymes can in some situations be functional in the extracellular environment too. This suggests that proteolysis is a very valuable tool for cell-environment interactions, but an extremely challenging task for biologists to untangle the potential webs of interacting activities that can occur. Division of the proteases into mechanistic classes and into clans and superfamilies of related structures has been the basis of the methodical approach to the proteases that has superceded the old ‘grind/culture and find’ approach of previous times. I was fortunate enough to be involved in the early stages of the analysis of some of the proteinases of the Metzincin clan, initially from the perspective of their structure and function in the biology of the extracellular matrix and broadening out into their more general capabilities in extracellular biology. The field is becoming ever more complex, with many contradictions and enigmas. Whatever, all aspects of the modern metalloproteinase story have been excellently reviewed in recent times. Here, I will focus on some recent examples of studies of the matrix metalloproteinases (MMPs) and the disintegrin metalloproteinases (ADAMs) that illustrate the challenges that we face, particularly when considering these enzymes as potential therapeutic targets. Alternative strategies for the development of specific inhibitors of individual enzymes, targeting extracatalytic sites, could be a way forward for the design of potential therapeutics in well defined diseases, and these will be discussed.

Matrix metalloproteinases

Proteolytic degradation of the extracellular matrix (ECM) has long been regarded as a key activity of the resident cells within tissues, orchestrating development and subsequent remodelling and repair processes. Such activities are also a major feature of disease processes, when cells of the innate immune system and incoming cells from other tissue compartments also contribute to the modulation of the ECM. The MMPs, a family of secreted proteinases, play a significant part in ECM proteolysis, orchestrating remodelling, as well as limited clipping to generate signalling neo-epitopes, or the release of sequestered growth factors, to impact on cell behaviour (McCawley & Matrisian 2001; Visse & Nagase 2003; Morrison et al. 2009).

The human genome has 24 MMP genes (MMP23 is duplicated) and are members of the clan of metalloendopeptidases (proteinases) termed Metzincins. Much knowledge has accumulated on the ECM degrading capabilities of the MMPs over the five decades since Gross and Lapiere discovered the first collagenolytic MMP in resorbing tadpole tails (Gross & Lapiere 1962), but important new substrates are still being discovered. Novel MMP substrate data have been established by the use of screening technologies including biochemical assessments, cell based assays and whole organism studies (Overall & Blobel 2007; Butler et al. 2009) and have indicated the potential for MMP regulation of chemokines, cytokines and growth factors, as well as cell adhesion molecules and other membrane associated proteins. Hence, MMPs can act as activators or regulators of many aspects of cell biology and may not be predominantly acting as ECM modulators (Murphy et al.2008,Morrison et al. 2009; Fanjul-Fernandez et al. 2010). Such studies have also served to emphasise the complexity of MMP functions, dependent on the precise cell type and situation. We also need to bear in mind that there are many other proteases that are active in the chemokine and signalling peptide arena; amino and carboxypeptidases and proteinases of other mechanistic classes have key regulatory contributions in some cell types.

In wound healing, the processing of chemokines provides an interesting example of the complexities of MMP modulation (detailed fully in Gill & Parks 2008). MMP-1 and MMP-3 cleave the CCL cytokines, MCPs 1–4 to generate soluble receptor antagonists and MMP-2 enhances the activity of CXCL5. MMP-8 increases the activity of CXCL5 and 8, but MMP-9 degrades CXCL1, 4 and 7. There are also species differences between mouse and human to consider; human CXCL8 is activated by MMP-8 and MMP-9 proteolysis, but the mouse orthologues are not processed by these enzymes. Hence, although the use of MMP gene ablated mice is a useful strategy for determining possible functions, it may not always give a clear picture of the role in human diseases.

The mild phenotypes elicited on ablation of individual Mmp genes in mice have been documented, but it should be noted that with pathological challenge more definitive phenotypes are emerging (Fanjul-Fernandez et al. 2010). This suggests that MMPs are significantly regulated in pathological conditions and may participate in disease progression positively or negatively. In view of this the precise roles of MMPs will take substantial further work to clarify. One of the most profound constitutive phenotypes is seen with the ablation of MT1-MMP (MMP-14). This membrane associated enzyme is thought to orchestrate collagen and fibrin proteolysis in relation to cell migration through 3D matrices (Hotary et al. 2002). However, MT1-MMP is also able to cleave cell adhesion molecules including CD44, integrins and ICAM and it activates proMMP-2 and proMMP-13. MT1-MMP ablation in vivo causes craniofacial dysmorphism, arthritis, osteopenia, dwarfism, and fibrosis of soft tissues, probably because of the loss of collagen remodelling in developing connective tissues (Itoh & Seiki 2006).

Disintegrin metalloproteinases

Besides the MMPs, families of Metzincin proteinases that contribute to the extracellular activities of cells are the ADAMs/adamalysins, the astacins and the pappalysins (Gomis-Rüth 2009; MEROPS http://merops.sanger.ac.uk/). The ADAMs family appear to be complementary to the MMPs in terms of extracellular function. The membrane anchored ADAMs number 25 in human, but they do not all have all the features of a functional Metzincin catalytic domain and are unlikely to be proteolytically active. Thirteen ADAMs are thought to be functional proteinases and play important roles in the regulation of cytokines and growth factors by the modification of both ligands and receptors, as well as cell adhesion molecules and other membrane proteins (Moss & Bartsch 2004; Blobel 2005; Garton et al. 2006; Reiss et al. 2006; Edwards et al. 2008; Gerg et al. 2008; Murphy et al. 2008; Reiss & Saftig 2009). The ADAMTS; ADAMs with one or more thrombospondin type 1 repeats, number 19 and are not transmembrane proteins, although they may be closely allied with cell surface molecules (Apte 2009). They are involved in the maturation of procollagens, the processing of von Willebrand Factor and the turnover of proteoglycans. The importance of ADAMs and ADAMTSs in development, in normal physiological and in pathological processes has been documented (Porter et al. 2005; Edwards et al. 2008; Apte 2009). Astacins also have extracellular substrates in some cases but are rather fewer in number (Sterchi et al. 2008). ADAMTS and Astacins will not be further discussed here, although their potential for overlapping roles with MMPs and ADAMs is considerable and requires detailed analysis.

A major function of the active metalloproteinase forms of the ADAMs appears to be the proteolysis of the entire ectodomain of transmembrane proteins, a process termed ‘shedding’, that releases soluble and partially functional intact ectodomain forms of the membrane associated protein (Edwards et al. 2008). ADAM17 (tumour necrosis factor-alpha convertase, TACE) is the best studied of this family of enzymes and is the principal protease involved in the activation of pro-tumour necrosis factor α (TNFα). ADAM17 sheddase functions cover a broad range of other cell surface molecules, notably the active forms of epidermal growth factor receptor (EGFR) ligands, amphiregulin, transforming growth factor α, heparin binding EGF and epiregulin (Horiuchi et al. 2007). The solubilisation of growth factors could potentially orchestrate different modes of signalling: autocrine, paracrine, juxtacrine and even exocrine in the case of TNFα (Blobel 2005). Confusingly, ADAM17 also appears to regulate the levels of cytokine receptors too, and such activities can down-modulate signalling or generate decoy receptors. Bell et al. reported the release of soluble TNFα, TNFRI, and TNFRII release by neutrophils and macrophages upon stimulation with various microbial antigens (Bell et al. 2007), providing the first direct in vivo evidence that ADAM17 is a primary and non-redundant sheddase. The role of ADAM17 in inflammation requires more detailed study as ADAM17-mediated TNFα shedding should result in enhanced TNFα signalling, but as ADAM17 also cleaves TNFRI and TNFRII, the shedding of these receptors could attenuate the initial enhancement of systemic TNFα activity. Another notable example is the role of ADAM17 in the regulation of interleukin 6 (IL6) function by the generation of a soluble form of the IL6 receptor. The formation of a complex of IL6 with its soluble receptor allows signalling through the co-receptor gp130, in a process termed ‘trans signalling’ (Rose-John et al. 2006). The function of vascular endothelial growth factor-A (VEGF-A) is also regulated by ADAMs as ADAM17 sheds VEGF receptor 2. VEGF-A stimulates endothelial cell ADAM17 activity with effects on VEGFR2 and other important angiogenesis modulators such as Tie-2, TNFα and EGFR ligands (Swendeman et al. 2008). The type I transforming growth factor β receptor 1 has also been shown to be shed by ADAM17 activity, such that ADAM17 ablation in cancer cells enhanced TGFβ responsiveness, including epithelial-mesenchymal transition (Liu et al. 2009). Garton et al. (2006) have proposed that shedding of cell surface inflammatory proteins provides a mechanism to rapidly regulate leucocyte adhesion to endothelial cells, de-adhesion and coordinate transitions between the different steps involved in leucocyte recruitment. Hence, L-selectin and VCAM-1 ectodomains have been shown to be released by ADAM17 proteolysis. The significant feature of proteolytic mechanisms for the regulation of such protein-protein interactions is the speed with which they can act (we are limited by the sensitivity of our assays but activity can be seen within minutes in vitro). This means that an understanding of the regulation of ADAM17 and other ADAMs is a key goal of many studies.

Several other ADAMs have important sheddase functions in particular tissue contexts. Another major family member, ADAM10, is a principal player in signalling via the Notch and Eph/ephrin pathways (Janes et al. 2005; Tousseyn et al. 2009). For a growing number of substrates, foremost among them being Notch, cleavage by ADAM sheddases is essential for their subsequent ‘regulated intramembrane proteolysis’ (RIP). Intramembrane presenilin complexes (γ-secretase) can generate cleaved intracellular domains from the remaining ‘stubs’ of transmembrane proteins left after ADAM cleavage and these translocate to the nucleus and regulate gene transcription. In other cases, RIPping seems to release the transmembrane ‘stubs’ for degradation by the proteasome (Wolfe 2006; Hass et al. 2009). Hence, ADAM activities are key to many signalling pathways from the cell to its environment and vice versa, as well as in general cellular housekeeping (Edwards et al. 2008). Confusion arises as the same substrate can be shed by different ADAMs in different cells, meaning that careful analysis is required for each individual situation. Studies of Adam9 deficient mice showed that pathological neovascularization in an oxygen-induced retinopathy (OIR) model was reduced compared to wild type mice. In cell-based assays, the overexpression of ADAM9 enhanced the ectodomain shedding of EphB4, Tie-2, Flk-1, CD40, VCAM, and VE-cadherin, so the enhanced expression of ADAM9 could potentially affect pathological neovascularization by increasing the shedding of these and other membrane proteins from endothelial cells (Guaiquil et al. 2009). The same authors found an increase in retinal re-vascularization, but fewer neovascular tufts in the OIR model in Adam8 deficient mice relative to wild-type. In cell-based assays, overexpression of ADAM8 increased the ectodomain shedding of several membrane proteins, including some that were ADAM9 substrates: CD31, Tie-2, Flk-1, Flt-1, EphrinB2, EphB4, VE-cadherin, Kit ligand, E-selectin, and neuregulin-1β2 (Guaiquil et al. 2010).

To add to this, different stimuli can activate different ADAM proteolytic pathways and MMPs too can be invoked in the same cell type. A few examples will be documented here.

Defining the functional metalloproteinases: MMPs vs. ADAMs

Although MMPs are likely to be the major contributors to ECM turnover amongst the Metzincins, there have been some descriptions of ADAM activities against specific matrix proteins (Seals & Courtneidge 2003). Such studies have largely been conducted in vitro, hence the importance of ruling out contamination with other proteases. The major overlap in MMP and ADAM functions comes from studies of the proteolysis of cell surface proteins including cell adhesion molecules, growth factors and their receptors. For instance, in a rodent model that mimics the osteoblastic and osteolytic changes associated with human metastatic prostate cancer, microarray analysis identified Mmp-7, as being upregulated in osteoclasts at the tumour-bone interface. MMP-7 was capable of processing TRANCE/RANKL (TNF family member receptor activator of nuclear κB ligand) to a soluble form that promoted osteoclast activation. Mmp-7 deficient mice demonstrated reduced RANKL processing and prostate tumour-induced osteolysis (Lynch et al. 2005). However, in another study, soluble RANKL produced by isolated osteoblasts from Mt1-mmp deficient mice was markedly reduced, membrane RANKL rose and their intrinsic osteoclastogenic activity was promoted, consistent with the findings of increased osteoclastogenesis in vivo in these mice (Hikita et al. 2006). Numerous other proteolytic activities at the cell surface have been variously ascribed to MMP-7, MT1-MMP or ADAMs. Although MMP-7 is a soluble enzyme it is thought to be associated with cell membrane and ECM heparan sulphate proteoglycans (Yu & Woessner 2000), whilst MT1-MMP and the ADAMs are membrane associated proteins, allowing for efficient access to cell membrane substrates. HSPGs, including syndecans are intimately linked with chemokine bioavailability and probably regulate the activity of a plethora of other signalling and regulatory proteins such as integrins (Morgan et al. 2007). MMP-7 has been shown to cleave syndecan-1 in wounded lung epithelia in vivo, releasing bound chemokines (Li et al. 2002) and enhancing cell migration and re-epithelialisation by the modification of α2β1 integrin binding to collagen. This effect was lost in Mmp-7-/- mice (Chen et al. 2009). However, Pruessmeyer et al. (2010) found that syndecan-1 released into the bronchoalveolar fluid of mice could be reduced by the administration of a relatively specific ADAM17 inhibitor. Isolated lungs perfused with TNFα and interferon γ increased ADAM17 activity and syndecan release. Both constitutive and induced syndecan shedding was prevented by the ADAM17 inhibitor. Other studies, albeit only based on cell studies, have implicated MMP-9 or MT1-MMP in syndecan-1 shedding (Endo et al. 2003; Brule et al. 2006). The cleavage of fibroblast syndecan-1 by MT1-MMP promotes cell proliferation in co-cultured tumour cells (Su et al. 2008).

MMP-7 has been shown to play a role in FasL shedding, potentiating epithelial cell apoptosis in the involuting prostate (Powell et al. 1999). Mmp-7-/- mice showed a reduction in the apoptotic index in the involuting prostate compared with wild-type animals. On the other hand, using loss and gain of function studies in murine embryonic fibroblasts (MEFs), Schulte et al. showed that the ADAM10 is critically involved in the shedding of FasL (Schulte et al. 2007). In primary human T cells, FasL shedding was significantly reduced after inhibition of ADAM10. The resulting elevated membrane associated FasL expression was associated with an increase in T cells undergoing activation induced cell death. O’Reilly et al. have shown that membrane bound FasL is essential for Fas induced apoptosis in T cells (O’Reilly et al. 2009), hence the solubilisation of FasL by metalloproteinases could protect against cell death in such situations.

Metalloproteinases as therapeutic targets

The MMPs and, more recently, the ADAMs and ADAMTSs have been considered as therapeutic targets and there have been extensive attempts to develop low molecular weight inhibitors targeting the active site (Hu et al. 2007; Nuti et al. 2007; Fosang & Little 2008; DasGupta et al. 2009).

The upregulation of MMPs has been reported in cancer, arthritis, vascular pathologies and numerous inflammatory diseases (Hu et al. 2007; Murphy 2008; Murphy & Nagase 2008; Pruessmeyer & Ludwig 2009). As regulation can occur at the level of gene transcription, cellular trafficking, proenzyme activation and endogenous inhibition, there are many opportunities for perturbation and a loss of coordination of their function (Lopez-Otin & Overall 2002). The tetracyclines are weak inhibitors of MMP activity but also prevent their synthesis in some cases and have been trialled is some degradative diseases (Yao et al. 2004). Targeting regulatory cytokines and growth factors, gene transcription or signal transduction pathways are all strategies that have been considered for the development of therapeutics (Overall & Lopez-Otin 2002; Folgueras et al. 2004). ADAMs and ADAMTSs can also show enhanced levels in similar pathological situations, and could be susceptible to similar approaches, although this is still a developing field and not fully explored (Porter et al. 2005; Edwards et al. 2008; Boutet et al. 2009; Murphy 2009).

Initially, the zinc ion at the heart of the hydrolytic process in metalloproteinase action was a key element of inhibitory drug design, with many chelating compounds being synthesised and assessed as inhibitors of MMPs, ADAMs or ADAMTS. The problems associated with general zinc chelators, compounded by the similarities in the structures of the active site of the Metzincins, led to a lack of specificity and side effects. More recent efforts at synthesising inhibitors have often moved away from the use of a chelating moiety and the advent of crystal structures or good models have led to an improvement in the ability to target individual MPs (Yiotakis & Dive 2008). Nevertheless, the knowledge of their precise activities in vivo is key to the exploitation of such inhibitors and the field still has a long way to go in that respect.

Alternative strategies for the development of MP inhibitors are now being evaluated (Sela-Passwell et al. 2010). Of particular interest, as discussed in more detail below, is the role of extra catalytic domains of these enzymes in determining their level of activity, specificity, or site of action. The definition of such exo sites is a painstaking process of individual evaluations, but this could be fruitful for the most prominent MPs in the near future. Other possibilities for the design of inhibitors of MMPs’ activities include the down regulation of gene expression by modifying the intracellular signalling pathways or their downstream transcription factors. Blockade of mitogen activated protein kinase (MAPK) pathways, nuclear factor (NF)-κB or activator protein (AP)-1 have been demonstrated to be effective in partially modulating MMP activities, or the direct blockade of inflammatory cytokines (Mix et al. 2004). Such approaches may also be used to target some ADAM/ADAMTS activities, but their regulation is not so well understood at this stage.

Novel approaches to metalloproteinase inhibition

The Metzincins are categorised as having a proteolytic domain (Figure 1) containing the zinc binding motif HEXXHXXGXXH that contains the three histidines binding the catalytic zinc and a glutamic acid residue that acts as a general acid/base during peptide bond hydrolysis. The active site of the Metzincins contains a conserved methionine providing hydrophobicity to the zinc environment. The remarkable overall similarity of the topology of the active site cleft of the Metzincins has meant that early generations of inhibitors chelating the active site zinc and interacting with residues within the active site had inhibitory activities across the different families, leading to confusing data in biological studies and apparent side effects when used as therapeutic entities (Overall & Lopez-Otin 2002; Georgiadis & Yiotakis 2008).

Figure 1.

Domain structure of the Metzincin families: MMPs, ADAMs and ADAMTS. The archetypal catalytic domain harbouring the active site zinc ion and Metzincin loop is preceded by a propeptide domain that is necessary for folding and latency of the active site. C-terminal domains are extremely varied and determine substrate and inhibitor binding properties, as well as interactions with other extracellular molecules. Elucidating the function of the extracatalytic domains and assessing them as targets for the design of novel proteinase inhibitors is providing exciting new insights into structure-function relationships and overall cellular roles of these enzymes.

The emerging complexity of Metzincin functions alluded to in the above discussion and in the many reviews of the field have led to the concept that a greater understanding of the expression profile and biological role of the individual enzymes is needed, coupled with intensive efforts to produce more specific inhibitors. Because of the nature of the catalytic sites, approaches will necessitate the use of sophisticated theoretical and experimental techniques to identify the specific structural features that can be exploited to attain the necessary specificity. Alternative approaches to metalloproteinase inhibition are to define exosites that determine catalytic efficiency and substrate or location specificity, including the contribution of other domains. Some examples of the role of extra-catalytic site involvement are discussed below.

Sagi and colleagues have recently considered the potential role of exosite or allosteric protein regions domain distinct from the catalytic cleft, in determining MMP activity. They raised the possibility that MMP enzymatic and non-enzymatic activities may be modified via antagonist molecules targeted to such sites or to alternative enzyme domains and discussed potential enzyme domains as targets for designing highly selective MMP inhibitors (Sela-Passwell et al. 2010).

Probing extra-catalytic domains: MMPs

Besides the archetypal secretory signal sequence, a regulatory propeptide and catalytic domain the MMPs have a C-terminal hemopexin-like domain, with the exception of MMP-7, MMP-23 and MMP-26 (Figure 1). MMP-2 and MMP-9 also have three repeats of the fibronectin type II motif inserted into the catalytic domain. Many of the MMPs are ostensibly soluble secreted proteins, but their association with ECM components and other proteins, including integrins in some cases, has been described. There are six membrane anchored MMPs, termed ‘membrane-type’ (MT), four are transmembrane proteins with small cytoplasmic domains (MT-1, MT-2, MT-3 and MT-5 MMP) and two with glycophosphatidyl inositol anchors (MT-4 and MT-6 MMP). There is considerable interest in the roles of the extracatalytic domains of the MMPs. The N-terminal propeptide acts to maintain the latency of the MMPs by the presentation of a cysteine residue into the active site as a zinc coordinator. The activation mechanism (known as the ‘cysteine switch) involves proteolytic cleavages of the propeptide causing a destabilisation of the cysteine-zinc interaction (Ra & Parks 2007). There are numerous examples of MMP domain contributions to macromolecular substrate specificity; the fibronectin type II domains of MMP-2 and MMP-9 are important for the cleavage of denatured collagens, type IV collagen and elastin, but they do not influence the hydrolysis of small peptides (Murphy et al. 1994; Shipley et al. 1996). The hemopexin domain of MMP-2 was found to bind chemokines such as monocyte chemoattractant protein-3 and hence facilitated its cleavage (McQuibban et al. 2000). The hemopexin domains of the collagenolytic MMPs (MMP-1, MMP-2, MMP-8, MMP-13 and MT1-MMP) are important for their ability to bind and cleave native triple helical collagen monomers. It has been shown that collagenases locally unwind the triple helix to allow access of the individual alpha chains to the active site centre (Chung et al. 2004). Hence, the collagen binding site is composed of elements of both the catalytic and the hemopexin domains. It is also clear that inter domain flexibility is key for collagenases’ and possibly other MMPs’ activity, as discussed by Sela-Passwell et al. (2010). In particular the long flexible ‘linker’ sequences between the catalytic and the hemopexin domains confer specific properties on individual MMPs. An interesting example is proMMP-9 that has a 64 amino acid linker between the catalytic and hemopexin domains that is proline-rich and heavily O-glycosylated and termed the O domain. Structural analysis to characterize a full-length structural model of pro-MMP-9 and the molecular character of its OG linker domain showed an elongated protein with the OG domain acting as a flexible 30 Å long linker between the two terminal domains. The degree of flexibility of the OG domain plays a potential role in the regulation of binding and processing of substrates, ligands, and receptors required for MMP-9 activities. Blocking such domain flexibility by antagonists could thus constitute a way of regulating the pathological activities of this enzyme (Rosenblum et al. 2007).

MT1-MMP has been shown to require the hemopexin domain for cell surface clustering, and a role in the formation of functional oligomers as part of the mechanism for proMMP-2 activation has been proposed. The hemopexin domain also mediated MT1-MMP interactions with CD44 (Suenaga et al. 2005). It is possible that the role of MMP domains in the interaction with ECM or cell surface proteins are critical for their function, for example, the fibronectin-like domain of MMP-2 and MMP-9 mediated binding to collagen. Active MMP-7 auto-activation is regulated by binding to highly sulfated glycosaminoglycans, such as heparin, chondroitin-4,6-sulfate (CS-E), and dermatan sulfate (Ra et al. 2009). Keratinocyte proMMP-1 was found to bind to collagen bound α2β1 integrin during cell migration. Competition studies suggested that the integrin α2 I domain interacts with the linker and hemopexin domains of pro-MMP-1, but not with the pro-domain (Dumin et al. 2001). Recent structural studies have shown that the prodomain of MMP-1 interacts with both the catalytic and hemopexin domains (Jozic et al. 2005), resulting in a ‘closed’ configuration as the pro-domain partially blocks the putative collagen binding site. The removal of the pro-domain upon activation opens the groove (‘open’ form) and allows MMP-1 to bind to collagen (Iyer et al. 2006; Li et al. 1995). It is possible therefore that integrin binding is specific to proMMP-1 because of the specific conformation of the proform. How might all these activities be investigated? One possibility is the use of fragment antibody libraries or monoclonal antibodies to generate reagents that interfere with specific activities and then to utilise mapping techniques to determine the epitopes involved. Conformation specific antibodies have proved immensely useful in the understanding of integrin function and similar success could be envisaged for the protease field (Askari et al. 2009).

ProMMP-2 activation by MT1-MMP on the cell surface, requires the formation of a quaternary molecular complex involving TIMP-2 (Itoh & Seiki 2006). ProMMP-2 hemopexin domain binds tightly to the non-inhibitory C-terminal domain of the inhibitor (Butler et al. 1998). This complex binds to active MT1-MMP on the cell surface through the free N-terminal MMP inhibitory domain of TIMP-2. This presumably orients the propeptide of proMMP-2 to an adjacent active MT1-MMP, and the specific reaction of two MT1-MMP molecules is driven by the hemopexin domains. TIMP-3 and TIMP-4 form a complex with proMMP-2 in a similar manner to TIMP-2 and the proMMP-2-TIMP-4 complex interacts with MT1-MMP, but it is non-productive in terms of MMP-2 activation (Bigg et al. 2001). The biological significance of the proMMP-2-TIMP-3 complex is not known. The hemopexin domain of proMMP-9 can also bind to TIMP-1 and TIMP-3 through their C-terminal domains (O’Connell et al. 1994). ProMMP-9 in neutrophils partially binds to neutrophil gelatinase associated lipocalin-like molecule (NGAL) through an intermolecular disulfide bond (Kjeldsen et al. 1993). However, the biological significance of these complexes is not clear, except that proMMP-9-TIMP complexes inhibit metalloproteinases and activation of proMMP-9 in the complex by MMPs is restricted (Ogata et al. 1995).

Exploring the extracatalytic domains: ADAMs

The 13 human ADAMs with functional metalloproteinase domains are comprised of a large propeptide domain, a disulphide bonded catalytic domain linked to a series of extensively disulphide bonded structures, namely a disintegrin domain, a cysteine rich domain and, in some cases an EGF-like domain that may act as a rigid spacer connecting the ectodomain to a transmembrane region. This is then linked to a cytoplasmic domain with significant potential for regulatory and signalling properties (Figure 1) (Takeda 2008).

The ADAM propeptides appear to have a role in initial protein folding and are similar to MMPs in that they possess a short consensus sequence toward their C-terminus including a cysteine residue. ADAM prodomains mediate the latency of the immature form and proteolytic processing to the mature, active form frequently occurs in the trans-Golgi network, mediated by proprotein convertases. The cysteine switch mechanism, whereby the cysteine of the propeptide coordinates the catalytic zinc ion does not always hold for the ADAMs, for example, ADAM10 and ADAM17. Nevertheless, the propeptide is a reasonable inhibitor of catalytic activity, by insertion within the catalytic domain. In the case of ADAM17 the potency of the prodomain as an inhibitor depends on the presence of the cysteine-rich/disintegrin domain: the prodomain is much less capable of inhibiting this form relative to the catalytic domain in isolation (Gonzales et al. 2004, 2008). This indicates the possible role of the non-catalytic domains in catalytic function. There have been several studies assessing the ability of the isolated propeptide, or derived sequences, to act as ADAM inhibitors (Moss et al. 2007; Gonzales et al. 2008). Further information about the nature of the interactions between the propeptide and the mature ectodomain will be invaluable in pinpointing exosite motifs that modify catalytic function.

The metalloproteinase domain is remarkably similar in structure throughout the ADAM and ADAMTS families, being ellipsoidal with a deep catalytic cleft harbouring the functional zinc ion. The accessibility and structure of the juxtamembrane ‘stalk’ region of the ADAM substrates seems to be a major specificity determinant and there are few specific peptide sequence criteria for ADAM cleavage. Solomon et al. used stopped-flow x-ray spectroscopy methods together with transient kinetic analyses to demonstrate that the catalytic zinc ion undergoes dynamic charge transitions before substrate binding to the metal ion. This suggests that the enzyme catalytic core is immensely flexible and interacts with distant exosites that are key to function (Solomon et al. 2007). The catalytic domain is linked to the disintegrin and cysteine rich domains. It has been proposed that flexibility between these domains may be important for fine-tuning substrate recognition and may allow adjustments to the spatial alignment of the catalytic site and exosite (Takeda et al. 2007). From structural data for the closely related snake venom haemorrhagins (Takeda 2008) and for isolated domains of ADAM10 (Janes et al. 2005) it appears that the disintegrin and cysteine rich domains form a disulphide bonded entity exposing a C-terminal hypervariable region in the latter which is likely to be a major interaction site with other molecules, including substrates. Many studies have shown integrin binding activity of the ADAMs that appear to be implicated in ADAM biology was attributed to the disintegrin domain, but the haemorrhagin structures show that the putative disintegrin loop is packed against a sub-domain of the cysteine rich region and would be inaccessible for protein binding, unless significant conformational changes can occur which are not represented in the crystal structure (Takeda et al. 2007). The crystal structures indicate that the ADAM ectodomains are C-shaped with the catalytic site and the hypervariable site juxtaposed. Preliminary studies using chimeras of different domains from different ADAMs do imply that these extra-catalytic regions might confer specificity to an individual ADAM and could form the basis for studies of more targeted inhibitors than has been achieved with catalytic domain interactants (Reddy et al. 2000). In the case of ADAM17, studies with the inhibitory propeptide and the natural inhibitor TIMP-3 have shown that active site binding kinetics are significantly modified by the presence of the disintegrin and cysteine rich domains (Lee et al. 2002; Gonzales et al. 2004).

Based on ELISA binding studies with novel fragment antibodies to ADAM17 we found that isomerisation of the disulphide bonds in ADAM17 and the subsequent conformational changes form the basis for the modulation of ADAM17 activity (C. Tape, S. Willems & G. Murphy, unpublished). The shuffling of disulphide bond patterns in ADAMs has been suggested by a number of recent adamalysin crystal structures, with distinct disulphide bond patterns altering the relative orientations of the domains, as well as the overall shape of the enzyme (Guan et al. 2010). The potential for the disintegrin and cysteine rich domains to bind to integrins, syndecans and other cell surface proteins could have implications for the modulation of ADAM functions, both as proteinases and as adhesion molecules. This field is in its infancy, but the prospects for a better understanding of ADAM structure-function relationships are good and highly promising for novel approaches to inhibitor development.