Matrix metalloproteinases at a glance

Matrix metalloproteinases (MMPs) belong to the metzinicin superfamily of metalloproteinases, which also includes the ‘a zinc and distintegrin metalloproteinase’ (ADAM) family (Primakoff and Myles, 2000). MMPs are zinc-dependent endopeptidases, with their catalytic activity depending on the binding of a zinc ion to a histidine-containing motif in their active site, which enables the cleavage of internal peptide bonds in their substrates (Stocker et al., 1995). Currently, 28 MMP family members have been described, of which 23 are expressed in human cells (Kohrmann et al., 2009; Nagase and Woessner, 1999; Pendas et al., 1996; Sternlicht and Werb, 2001) (see poster). Generally, MMPs are widely expressed in both animals and plants, ranging from Drosophila (Llano et al., 2000) to Arabidopsis (Maidment et al., 1999) and Chlamydomonas (Kinoshita et al., 1992). Here, we focus mostly on findings relating to human or murine MMPs. We present an overview of the current knowledge on MMP domain structure, activation and inhibition, their respective target proteins, and the intra- and extra-cellular trafficking pathways that enable local availability of MMPs, as well as the involvement of MMPs in health and disease. However, considering the wealth of information gained by decades of research on MMPs, not all relevant aspects can be addressed in this Cell Science at a Glance article. We therefore also refer the reader to comprehensive reviews covering the entire range of MMP activity and regulation, in both physiological and pathological processes (Cabral-Pacheco et al., 2020; Chang, 2023; de Almeida et al., 2022; Hey et al., 2022; Malemud, 2006; Moracho et al., 2022; Niland and Eble, 2020; Serra, 2020; Strouhalova et al., 2023).

MMPs can be sorted into six groups, based on substrate specificity and sequence similarity. These groups comprise collagenases (MMP-1, -8, -13 and -18), gelatinases (MMP-2 and -9), stromelysins (MMP-3, -10 and -11), matrilysins (MMP-7 and -26), enamelysins (MMP-20), a further diverse group (MMP-19, -21, -23A, -23B, -27 and -28), as well as membrane type (MT) MMPs (MMP-14, -15, -16, -17, -24 and -25). Vertebrate MMPs are numbered sequentially; however, for historical reasons, there is no MMP-4, MMP-5 or MMP-6, as the respective gene products had already been given a different designation (MMP-2 or MMP-3) (Parks and Shapiro, 2001).

All MMP polypeptides contain a N-terminal signal sequence or pre-domain, which is removed upon synthesis in the endoplasmic reticulum (ER) (Nagase et al., 2006). Mature MMP proteins are characterized by their typical domain structure; they contain an N-terminal pro-domain that is involved in autoinhibition by ligating the zinc ion in the catalytic domain. Removal of this pro-domain is achieved by convertases, such as furin, other MMPs or by autoactivation (Kinoshita et al., 1998; Strongin et al., 1995), and is a necessary step in MMP activation. All MMPs feature a 160–170 amino acid long zinc-dependent catalytic domain with a zinc ion arranged by three histidine residues at the center. The catalytic domain confers substrate specificity through binding of residues of the substrate that are directly adjacent to the target peptide bond, as well as by binding secondary substrate sites by motifs outside of the active site itself (Overall, 2001). This N-terminal domain is followed by a hinge region that further connects to a hemopexin- or vitronectin-like domain. This domain is involved in binding tissue inhibitors of metalloproteinases (TIMPs) (Ugarte-Berzal et al., 2016), recognition of certain substrates, such a collagen fibrils (Sternlicht and Werb, 2001), oligomerization (Cha et al., 2002) and the regulation of MMP intracellular trafficking (Dufour et al., 2008) (see poster).

Of note, not all of the MMP domains are present in every isoform. For example, MMP-7 (matrilysin) and MMP-26 (endometase) consist of only the pro- and catalytic domains, thus exhibiting the minimal set of domains required for an activatable proteinase (Piskor et al., 2020). Moreover, additional regions can also be present, such as in the gelatinases MMP-2 and MMP-9, which contain three fibronectin type II repeats within their catalytic domain (Bauvois, 2012). In addition, MMP-23, the only group II membrane-type (MT-)MMP member, contains an N-terminal type II signal anchor that enables association with the cell surface, and a C-terminal domain containing a cysteine-rich and a proline-rich region (Velasco et al., 1999).

Six MMP isoforms are attached to the cell surface by membrane-associated domains and are thus termed as ‘membrane-type MMPs’ (MT-MMPs). This subset is referred to either by their MT-MMP numerical nomenclature, i.e. MT(1-6)-MMP, or by their systematic MMP number. For example, MT1-MMP is another name for MMP14 (see poster for the full list of alternative MMP names).

The group consisting of MMP-14, -15, -16 and -24 contains a transmembrane domain that is followed by a short cytoplasmic tail involved in local enrichment at the plasma membrane, cellular signaling, endocytosis and recycling (El Azzouzi et al., 2016; Strouhalova et al., 2023; Uekita et al., 2001). A second group consisting of MMP-17 and -25 contains a C-terminal glycophosphatidyl inositol (GPI) anchor that allows association with the cell surface by insertion of the fatty acid chains of inositol phospholipids into the outer leaflet of the plasma membrane (Itoh et al., 1999; Kojima et al., 2000), without any intracellular region.

Anchoring to the cell surface restricts the area of potential proteolytic activity of MT-MMPs to sites on or directly adjacent to the plasma membrane. This is in contrast to what is seen for soluble MMPs, which, once secreted from the cell, can diffuse into the extracellular space. Moreover, MMPs can also be released as cargo of extracellular vesicles (discussed below) or can be shed from the cell surface; therefore, MT-MMPs can also act at a distance from the cell of origin (see poster). Vice versa, some soluble MMPs, such as MMP-7 and MMP-12 can interact with lipid bilayers through electrostatic interactions or bind to heparin sulfate proteoglycans, and they are thus able to associate with the cell surface, despite the absence of any specific membrane-binding domains (Van Doren et al., 2017).

MMP activity is spatiotemporally regulated at different levels, such as through removal of the pro-domain and association with inhibitors, as well as via intra- and extra-cellular trafficking. Although the N-terminal pre-domain is cleaved off during translation, the presence of the pro-domain ensures the protein is in an inactive state through autoinhibitory interaction with the catalytic site. Removal of the pro-domain either by convertases, such as furin (Kang et al., 2002; Yana and Weiss, 2000), serine proteinases, such as plasmin, thrombin or trypsin (Ra and Parks, 2007), autoactivation (Strongin et al., 1995) or other MMPs (Itoh and Seiki, 2004), leads to an active enzyme. Of note, activation of MT1-MMP by furin can occur intracellularly, prior to its embedding in the plasma membrane, as shown in A375 cells (Mazzone et al., 2004). However, it is unclear whether this reflects a general principle in MMP activation (see poster).

Another level of activity regulation comprises the interaction of MMPs with TIMPs, which bind to MMPs and block their activity. Currently, four TIMPs have been described (TIMP-1, -2, -3 and -4); all of them are capable of inhibiting MMPs. For details of inhibition of specific MMPs by specific TIMPs, see Cabral-Pacheco et al. (2020). An example for the complex interaction between TIMPs and MMPs is the in vitro finding that TIMP-2 competes with TIMP-4 for binding of MMP-2 and can thus inhibit MMP-2 activation (Bigg et al., 1997). However, in some cases, TIMP binding is also a prerequisite for MMP activation. For instance, TIMP-2 has been shown to form a complex with an MT1-MMP dimer and pro-MMP-2, with subsequent activation of MMP-2 through the removal of its pro-domain by one of the MT1-MMP molecules (see poster) (Jackson et al., 2017).

MMPs can proteolytically cleave a wide variety of substrates, including components of the extracellular matrix, cell-surface-associated proteins and intracellular proteins (de Almeida et al., 2022). This can result in the degradation of the target (e.g. ECM components), but also in MMP activation through removal of inhibitory domains as is the case for other MMPs or ADAM proteases. Although individual MMPs exhibit a certain level of substrate specificity, they also share target proteins (see poster). One of the most prominent targets of MMPs are collagens, which are mostly cleaved by soluble MMPs, such MMP-2 (Aimes and Quigley, 1995) and MMP-9 (Rosenblum et al., 2010), but are also targeted by MT1-MMP (Tam et al., 2002) and others. Collagenases, such as MMP-2 and MMP-9, can also cleave gelatin (i.e. denatured collagen) (Nikolov and Popovski, 2021; Toth and Fridman, 2001). Other ECM components cleaved by MMPs include elastin (by MMP-9, -10, -11 and -12), laminin (by MMP-3, -7 and -15), fibronectin (by MMP-3, -7, -10, -11, -12, -13, -14, -15, -25 and -26), tenascin (by MMP-1 and -15) and proteoglycans (e.g. aggrecan by MMP-1, -7, -8, -9, -11, -13, -14, -19 and -20) (Morrison et al., 2009) (see poster).

Important MMP targets at the cell surface include cell adhesion proteins, such as CD44 (Werny et al., 2023) and integrins (Gomez et al., 2012), as well as ADAMs (Nakamura et al., 2004). Furthermore, MT1-MMP can form a complex with TIMP-2 and CD44 at the surface of HeLa cells; this results in cleavage of CD44, which negatively influences cell–cell and cell–ECM adhesion (Wohner et al., 2023). Integrins are central regulators of cell–ECM adhesion (Bachmann et al., 2019). Accordingly, cleavage and shedding of the of integrin β1 ectodomain by MMP-2, which reduces cell–substrate adhesion, promotes invasion of human colorectal tumor cells (Kryczka et al., 2012). Of note, interactions between MMPs and integrins are neither one-sided nor restricted to proteolysis. For example, integrin β1-mediated signaling can regulate phosphorylation of MT1-MMP at the Tyr567 residue, which promotes MT1-MMP endocytosis and recycling to sites of invadopodia formation, resulting in increased cell invasion, as shown for MDA-MB-231 human breast carcinoma cells (Grafinger et al., 2020). Upon binding to ECM, integrins can also upregulate the expression of MMPs (Yue et al., 2012). For further reading on the multifaceted relationship between integrins and MMPs, we refer the reader to Niland and Eble (2020) and Yue et al. (2012).

As noted above, the cell-surface-anchored ADAM proteinases are further interactors of MMPs. For example, MMP-7 has been shown to cleave off the pro-domain of ADAM28, leading to its activation (Mochizuki et al., 2004). Moreover, pro-forms of MMPs can be activated through cleavage by other MMPs; this includes cleavage of MMP-2 by MT1-MMP (Deryugina et al., 2001) and of pro-MMP-9 by MMP-3 (Ogata et al., 1992). These findings also indicate that care should be taken before attributing an observed effect to a specific MMP.

Finally, a prominent intracellular MMP target is the centrosomal protein pericentrin (Golubkov et al., 2005). Interestingly, human pericentrin-2 has been shown to be cleaved by MT1-MMP, whereas murine pericentrin is resistant to MT1-MMP activity (Golubkov et al., 2005), pointing to the potential relevance of species-specific effects in MMP activity and regulation. For details on further intracellular targets of MMPs, see Cauwe and Opdenakker (2010).

Of note, MMPs play a variety of roles in both physiological and pathological scenarios (see Box 1). Many of these roles are connected to their ability to remodel the extracellular matrix. A prominent consequence of this remodeling is the local invasion of cells. As mentioned above, ECM-directed activity of MMPs is often localized at podosomes or invadopodia, the main adhesion and invasion structures of many cell types (Linder et al., 2023) (see poster). Respective important MMP isoforms include MT1-MMP, MMP-2 and MMP-9, which have been detected at podosomes of macrophages (El Azzouzi et al., 2016), endothelial cells (Osiak et al., 2005; Tatin et al., 2006) and smooth muscle cells (Thatcher et al., 2017), among others, as well as in invadopodia of various cell types (Lagarrigue et al., 2010; Monsky et al., 1993; Redondo-Munoz et al., 2008).

Interaction of MMPs with other proteins can be complex and does not necessarily involve proteolysis. For example, although CD44 can be cleaved by MT1-MMP, thereby promoting invasion of human carcinoma and osteosarcoma cells (Kajita et al., 2001), non-proteolytic association of CD44 with the hemopexin domain of MT1-MMP can also lead to its localization to lamellipodia of HT-1080 human epithelial cells (Mori et al., 2002). Of note, in particular, the cytoplasmic tail of MT1-MMP has been shown to exert structural and regulatory roles that are independent of its proteolytic activity (Strouhalova et al., 2023). For example, association of MT1-MMP with TIMP-2 can lead to activation of the Ras-Raf-ERK signaling cascade, which requires the ⁵⁷³YCQR⁵⁷⁶ motif of the cytoplasmic tail of MT1-MMP and is independent of its proteolytic activity (D'Alessio et al., 2008; Strongin, 2010). Furthermore, binding of the MT1-MMP C-terminus to β1 integrin regulates endothelial cell invasion and branching morphogenesis (Mori et al., 2013), whereas the F-actin-binding ⁵⁷¹LLY⁵⁷³ motif within the MT1-MMP tail (Uekita et al., 2001) assists in its recruitment to invadopodia (Yu et al., 2012) and podosomes (El Azzouzi et al., 2016). Moreover, this motif enables the MT1-MMP C-terminus to facilitate the reformation of podosomes at sites of previously disassembled structures, thus providing structural memory and supporting directional cell migration (El Azzouzi et al., 2016) (see poster). The reformation of podosomes is likely initiated by the binding of short cytoplasmic actin filaments to the MT1-MMP C-terminus, which could act to locally concentrate Arp2/3 complexes and actin-associated adaptor proteins, such as Tks5 (also known as SH3PXD2A) or cortactin, facilitating the local nucleation of branched actin networks and recruitment of further podosome components (El Azzouzi et al., 2016).

Another important aspect of MMP regulation is their limited spatiotemporal dwelling time at specific intracellular compartments or at the cell surface. Initially, MMPs are synthesized in the ER as zymogens (Seiki and Yana, 2003) and are then transported to the Golgi for further maturation, such as glycosylation (Yana and Weiss, 2000). Intra-Golgi trafficking is regulated by proteins such as nucleobindin-1 (NUCB1), as shown for MMP-2 and MT1-MMP in MDA-MB-231 breast cancer cells (Pacheco-Fernandez et al., 2020). Moreover, MMPs can also regulate gene transcription, either indirectly, by inducing respective signaling cascades, as shown for MMP-2, MMP-7 and MT1-MMP in MCF-7 cells (Eisenach et al., 2010), or directly, by being transported to the nucleus where they act as transcription factors, as shown for MMP-12 driving IκBα transcription in murine macrophages (Marchant et al., 2014).

In the cytoplasm, soluble MMPs are transported within vesicles, whereas MT-MMPs are embedded in the vesicle membrane, with any C-terminal extensions facing the cytoplasm (see poster). As microtubules constitute the long-range transport system of cells, MMP transport is mainly microtubule based, with motor proteins such as kinesins and dynein providing kinetic energy (Hey et al., 2022). Microtubules are polar structures, with their plus-ends mostly pointing to the cell periphery. Accordingly, the plus-end-directed motors kinesin-1 and kinesin-2 have been shown to be important for exocytic trafficking of MT1-MMP in primary macrophages, with the minus-end-directed motor dynein acting as a counterplayer by mediating transport towards the cell interior (Wiesner et al., 2010). All three motors have also been shown to regulate MT1-MMP trafficking in MDA-MB-231 cells (Marchesin et al., 2015), pointing to the general involvement of these specific motors in MT1-MMP transport. Furthermore, kinesin KIF13A has been shown to regulate MT1-MMP exocytic trafficking in HT-1080 fibrosarcoma cells. Of note, this exocytic trafficking is inhibited by the kinesin KIF9 through a yet unknown mechanism (Gifford et al., 2022). Moreover, the kinesin KIF16B is the major motor driving recycling of re-endocytosed MT1-MMP back to the cell surface of primary macrophages (Hey et al., 2023).

Generally, endocytosis of membrane-associated MMPs is followed by either storage in endolysosomal vesicles, degradation in mature lysosomes or recycling back to the cell surface. Endocytosis can proceed by clathrin- or caveolin-dependent pathways, as shown for MT1-MMP in HT1080 fibrosarcoma cells and endothelial cells, respectively (Galvez et al., 2004; Remacle et al., 2003). In addition, uptake by clathrin-independent carriers or GPI-anchored protein enriched compartments (CLICs and GEECs) has been demonstrated for MT4-MMP in MDA-MB-231 cells (Truong et al., 2016).

MMP transport is further modified by regulatory proteins, including c-Jun terminal kinase interacting proteins (JIP)-3 and JIP-4 (also known as MAPK8IP3 and SPAG9, respectively), which have been shown to coordinate the activities of kinesin-1 and dynein activity during MT1-MMP transport in MDA-MB-231 cells (Marchesin et al., 2015). Furthermore, Rab GTPases also regulate vesicular transport of MMPs. For example, Rab8 regulates MT1-MMP exocytosis in MDA-MB-231 cells (Bravo-Cordero et al., 2007), and Rab8 isoform a (Rab8a) modulates MT1-MMP exocytosis in primary human macrophages (Wiesner et al., 2013). Work in macrophages has further shown that Rab5a regulates MT1-MMP re-endocytosis, whereas Rab14 and Rab22a, respectively, are involved in fast and slow recycling of the protease back to the cell surface (Wiesner et al., 2013) (see poster). MT1-MMP recycling involves trafficking through Rab7-positive endolysosomes, but this can also result in the degradation of part of the MT1-MMP pool (Planchon et al., 2018; Wiesner et al., 2013). By contrast, MT1-MMP recycling in MDA-MB-231 cells has been shown to depend on Rab4 proteins, pointing to respective differences in MMP trafficking between primary and transformed cells (Linder and Scita, 2015).

Soluble NSF attachment protein receptor (SNARE) proteins are involved in membrane fusion and are categorized into vesicle-localized (v-)SNAREs and target membrane-localized (t-)SNAREs (Wang et al., 2017). Their interaction mediates the transfer of (MMP-containing) vesicles from one cellular compartment to another. For example, the v-SNARE Bet1 and the t-SNARE syntaxin-4 mediate transport of MT1-MMP from the ER to invadopodia in MDA-MB-231 cells (Miyagawa et al., 2019). Further important SNAREs include the v-SNAREs VAMP7 and VAMP8, and the t-SNARE SNAP23, all of which have been shown to regulate MT1-MMP transport in various cells (Rohl et al., 2019; Steffen et al., 2008; Williams and Coppolino, 2011).

MMP exocytosis often takes place at specific sites of the plasma membrane, most notably at cell–matrix contacts, such as invadosomes, an umbrella term for podosomes (El Azzouzi et al., 2016) and invadopodia (Sakurai-Yageta et al., 2008), as well as at focal adhesions (Wang and McNiven, 2012). Of note, a preferential delivery of MMPs to these sites has not only been reported in 2D systems, but also in 3D contexts. Indeed, MT1-MMP has been shown to localize to 3D podosomes of macrophages, where it colocalizes with sites of matrix degradation (Wiesner et al., 2013), and to areas where migration of tumor cells and neutrophils through collagen networks is restricted owing to a dense ECM (Ferrari et al., 2019; Wolf et al., 2013).

MMPs are released into the extracellular space by secretion in case of soluble isoforms or by shedding of the ectodomains of cell-surface attached MT-MMPs (Elkington et al., 2009). In addition, both soluble and membrane-bound MMPs can be released as cargo of extracellular vesicles (EVs). EVs range in size from 30–100 nm; they include ectosomes that bud off the plasma membrane, and exosomes that are released from multivesicular bodies (MVBs) (El Andaloussi et al., 2013). The processes that lead to sorting of MMPs into EVs are not well understood. Nevertheless, it has been shown that the v-SNARE VAMP3 and the tetraspanin CD9 regulate sorting of MT1-MMP into EVs in LOX melanoma cells (Clancy et al., 2015). Moreover, release of exosomes containing MT1-MMP, MMP-2 and MMP-9 from SCC61 head and neck squamous carcinoma cells has been shown to be regulated by cortactin, Rab27a and coronin 1B, by stabilizing cortical actin-rich docking sites for multivesicular late endosomes (Sinha et al., 2016). Similarly, coronin 1C has been shown to be important for the release of MT1-MMP positive EVs (Tagliatela et al., 2020).

MMP-containing EVs can have a variety of effects, both on the surrounding ECM and on recipient cells. For example, knockdown of VAMP3 results in decreased sorting of MT1-MMP into EVs and, subsequently, in reduced cell invasion (Clancy et al., 2015). Interestingly, MT1-MMP-positive EVs in MDA-MB-231 cells have been shown to be preferentially released close to invadopodia, which is followed by gelatin degradation in this region (Beghein et al., 2018). Furthermore, the uptake of MMP-3-positive EVs from murine lung metastatic LuM1 cells in an allograft mouse model increased the invasiveness of recipient cells from multiple organs (Okusha et al., 2020). This has been shown to be based on translocation of MMP-3 to their nucleus, resulting in increased expression of connective tissue growth factor (CTGF) (Okusha et al., 2020). Finally, EVs can also induce the release of MMPs by a recipient cell; EVs from AsPC1 cells have been shown to lead to increased MMP-9 secretion in THP-1 cell-derived macrophages after their initial uptake (Linton et al., 2018). For further reading on MMP trafficking, release and recycling, we refer the reader to Hey et al. (2022).

Following their initial discovery (Gross and Lapiere, 1962), decades of research have demonstrated the fundamental importance of MMPs in a plethora of pathological and physiological functions, revealing many fascinating facets of these multifunctional proteins. Still, despite the wealth of data on MMP regulation, many important questions still need to be addressed. For instance, it is currently unclear whether soluble and membrane-associated MMPs travel in the same vesicle populations or are sorted to discrete vesicles. It is also not clear whether active and inactive forms of MMPs, either with or without any associated proteins, such as TIMPs, are transported together (see poster). Of note, MMPs are currently being rediscovered as viable targets for therapies for the treatment of cancer, as well as infectious, inflammatory and neurological diseases. In this regard, the development not only of isoform- but also of context-specific inhibitors for therapeutic applications constitute future challenges. Moreover, it is now evident that MMPs are not part of strictly hierarchical activation cascades, but are integral members of a highly flexible proteinase network within cells, and we anticipate that more of their fascinating properties are revealed, the more we will find out about the intricate roles MMPs have in cells and tissues.

We apologize to all authors whose work was not mentioned due to space limitations. We thank Martin Aepfelbacher for continuous support.

A high-resolution version of the poster and individual poster panels are available for downloading at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.261898#supplementary-data.