Breaching the basement membrane: who, when and how?

Breaching the basement membrane: who, when and how?

Review Feature Review Breaching the basement membrane: who, when and how? R. Grant Rowe and Stephen J. Weiss Division of Molecular Medicine and Gene...

1MB Sizes 0 Downloads 24 Views

Review

Feature Review

Breaching the basement membrane: who, when and how? R. Grant Rowe and Stephen J. Weiss Division of Molecular Medicine and Genetics, Department of Internal Medicine, The Life Sciences Institute, University of Michigan, 210 Washtenaw, Ann Arbor, MI 48109–2216, USA

The basement membrane (BM), a specialized network of extracellular matrix macromolecules, surrounds epithelial, endothelial, muscle, fat and nerve cells. During development, immune surveillance and disease states ranging from cancer to fibrosis, host cells penetrate the BM by engaging tissue-invasive programs, the identity of which remain largely undefined. Although it is commonly assumed that all cells employ similar mechanisms to cross BM barriers, accumulating evidence indicates that cells might selectively mobilize protease-dependent or -independent invasion programs. New data indicate that protease-dependent transmigration is largely reliant on a group of membrane-anchored metalloenzymes, termed the membrane-type matrix metalloproteinases, which irreversibly remodel BM structure. By contrast, mechanisms that enable protease-independent transmigration remain undefined and potentially involve the reversible disassembly of the BM network. Further characterization of the molecular mechanisms underlying BM transmigration should provide important insights into pathophysiologic tissue remodeling events and also enable the development of novel therapeutics. Introduction Early in development, animals ranging from flies to humans direct the embryonic epithelium to orchestrate the organization of an extracellular, supramolecular network of proteins, glycoproteins and proteoglycans, termed the basement membrane (BM) [1–3]. This conglomerate of structural macromolecules coalesces to form a dense, 100– 300 nm-thick lamina that underlies all epithelia and, in higher organisms, ensheaths endothelial cells, nerves, smooth muscle cells and adipocytes [1,4,5]. The assembled BM provides adherent cells with structural support and functional cues by virtue of its biomechanical properties, display of adhesion receptor ligands and repertoire of matrix-bound growth factors [4,6]. With a pore size in the order of 50 nm, only small molecules are able to passively diffuse across this thin, but structurally rugged, barrier [6–8]. Nonetheless, normal cells are able to traffic freely and rapidly across BMs by activating tissue-invasive programs during morphogenesis and immune surveillance [9–13]. Furthermore, in a repetitive theme

familiar to biologists, cell populations participating in pathologic events such as cancer can inappropriately coopt ‘normal’ BM-transmigration programs to dire and, most often, lethal consequences by driving the metastatic process [14].

Glossary Anchor cell: a specialized cell in C. elegans that invades the vulval epithelium and crosses the underlying BM during normal worm development. Basement membrane: a specialized form of extracellular matrix comprised of an interwoven mixture of type IV collagen, laminins, nidogen and sulfated proteoglycans that lies beneath epithelial and endothelial cells and surrounds muscles, nerves, adipocytes and smooth muscle cells. Branching morphogenesis: during development and angiogenesis, the formation of a branched network of epithelial (or endothelial)-lined tubules from a cell aggregate. This process occurs during development of organs such as lung and kidney and the vasculature. Cathepsins: a family of acidophilic cysteine proteinases that reside primarily within the lysosomal network, but that might traffic extracellularly when the enzymes have been implicated in turnover of extracellular matrix molecules. Epithelial–mesenchymal transition: the process by which a differentiated epithelial cell reprograms its gene expression profile to downregulate epithelial characteristics, such as homotypic adhesion and basolateral polarity, and activate mesenchymal characteristics, such as increased motility and invasive activity. Extracellular matrix: polymeric network of proteins, glycoproteins, proteoglycans and glycosaminoglycans that supports and compartmentalizes tissues while regulating cell fate and function. Fibrosis: pathologic accumulation of fibroblasts and extracellular matrix molecules in interstitial compartments of tissues, thought to be initiated by EMT processes. Hydroxyproline: a hydroxylated derivative of the amino acid proline that is generally restricted to collagen family members and which promotes helical structure and thermal stability. Imaginal disc: primitive structure in Drosophila larva that moves to the outside of the adult body and gives rise to external anatomical structures (e.g. legs, wings, eyes and antennae). Invadopodium: specialized actin-rich cytoplasmic protrusion containing adhesion molecules, signal transduction machinery and proteinases that is used to invade extracellular matrix barriers. Lathyrogen: a chemical inhibitor of lysyl oxidase, an enzyme crucial for oxidizing the e-amino group of lysine or hydroxylysine to an aldehyde derivative that participates in the formation of intermolecular and intermolecular collagen crosslinks. Matrix metalloproteinases: a family of approximately 25 zinc-dependent, neutral pH-optima, mammalian metalloenzymes that are synthesized as secreted or membrane-tethered zymogens. After processing to their active forms, the proteinases (as a family) are able to cleave all known ECM components. Primitive streak: in the two-layered mammalian embryo, the site at which epiblast cells invade through the underlying BM to differentiate into the mesoderm. Type IV collagen: a group of six collagenous protein chains [a1(IV)–a6(IV)] that belong to the 28-member collagen superfamily. Restricted largely to BMs, the chains assemble to form three distinct heterotrimers of a1a1a2, a3a4a5 or a5a5a6. The a1a1a2 heterotrimer is present in the BM of all tissues with the other chain composites displaying a restricted tissue distribution.

Corresponding author: Weiss, S.J. ([email protected]).

560

0962-8924/$ – see front matter ß 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tcb.2008.08.007 Available online 9 October 2008

Review The transmigration of cells across the BM has an unquestionably important role in normal and neoplastic events and has been the subject of thousands of reports – not to mention innumerable reviews – in the literature [1– 3,6,14]. Nevertheless, it probably comes as a surprise that the mechanisms that enable cells to cross this structural barrier remain largely unknown and the subject of considerable debate. Whereas the ability of a migrating cell to perforate the BM has been almost uniformly ascribed to proteolytic events, >500 proteinases are encoded within the mammalian genome, thus complicating efforts to identify a subset of crucial, matrix-degrading enzymes [15]. As such, using various model constructs designed to recapitulate BM structure in vitro, most investigators have, over the past 20 years, accepted a largely circumstantial premise that secreted proteinases directly regulate transmigration in both physiologic and pathologic states [16–18]. Very recently, however, evidence has begun to accumulate indicating that accepted dogma might now be ripe for revisiting [1,19–21]. New insights into BM assembly and structure have raised serious concerns regarding the use of most of the in vitro models used for analyzing invasion [22–24]. Furthermore, in vivo studies of mice harboring inactivating mutations of proteinases commonly implicated as the ‘usual suspects’ in BM transmigration seldom seem to exhibit frank defects in BM invasion-associated events [25–28]. Instead, recent studies support the contention that a small subset of membraneanchored metalloproteases assumes a previously unrecognized role in this process [21]. Although we would be pleased to inform readers that solutions to all queries regarding BM invasion programs lie within this perspective, this is far from the case and continued efforts are required to build a definitive model of BM transmigration. Rather, it is our intent to highlight existing conundrums and caveats in the field, to pose possible solutions as to the means by which normal and neoplastic cells traverse BM barriers and to outline experimental systems in which these hypotheses might be evaluated and tested in rigorous fashion. BM structure Comprising >50 distinct macromolecules, the predominant components of BMs are intertwined meshworks of polymeric laminin and type IV collagen (see Glossary) [2– 4,6]. Distinct from all other BM constituents, only laminin and type IV collagen are able to self-assemble into polymers [4]. Until recently, most models of BM organization assumed that type IV collagen serves as the major scaffolding upon which the laminin network is deposited. Newer studies indicate, however, that laminin polymers function as the initial template for BM assembly [29,30]. Laminin isoforms are a family of 16 heterotrimeric glycoproteins composed typically of a 400-kD a chain, a 200-kD b chain and a 200-kD g chain [31]. The coalescence of a, b and g chains leads to the formation of a cruciform trimer with three long arms and one short arm stabilized by disulfide bonds [31]. Supramolecular organization is triggered when laminins are concentrated at the plasma membrane of BMassociated cells via binding of long-arm globular domains to cellular receptors that range, in a tissue-specific man-

Trends in Cell Biology

Vol.18 No.11

ner, from sulfated glycolipids to a-dystroglycan [29,30]. In turn, stable interactions between the short-arm globular domains of adjacent laminin molecules enable the organization of a single layer of obliquely oriented laminin heterotrimers [29,32]. Like the laminins, type IV collagens also assemble into a supramolecular network at the surface of epithelial and endothelial cells, adipocytes, muscle cells and nerves through processes either mediated directly by integrins (e.g. a2b1 and a1b1) or indirectly via type-IV-collagenbinding interactions with laminin and other BM-associated macromolecules [2,4]. There are six distinct type IV collagen chains [a1–6(IV)] that display tissue-specific distribution patterns [2,3]. In the trans-Golgi network

Figure 1. Type IV collagen network structure. Type IV collagen protomers are triplehelical assemblies that contain N-terminal 7S domains that oligomerize to form 7S tetramers covalently stabilized by disulfide and (hydroxyl)lysine-derived crosslinks. The globular C-terminal NC1 domains of two type IV collagen protomers also interact to form dimers composed of six individual collagen chains [termed NC1 hexamers (Hex)] that are held in association by strong non-covalent forces and (hydroxyl)lysine-methionine thioether crosslinks. Each individual collagen chain is synthesized with an N-terminal 7S domain that is decorated with a N-linked oligosaccharide (black, g-shaped symbol). Formation of 7S domain tetramers and NC1 dimers support the assembly of a crosslinked, stabilized meshwork within the basement membrane that is reinforced by supramolecular twisting and lateral associations between the triple-helical collagenous domains (arrowheads). Figure adapted, with permission, from Ref. [2].

561

Review (TGN), a(IV) chains form heterotrimeric protomers that contain a central discontinuous triple-helical domain rich in Gly-X-Y motifs, in which X and Y are often proline and hydroxyproline residues, and an N-terminal 7S domain and a C-terminal globular non-collagenous 1 (NC1) domain [2]. After protomer secretion, dimeric associations between two NC1 trimers and tetrameric interactions between 7S domains promote the assembly of a polymeric meshwork [2,4] (Figure 1). Remarkably, the six genetically distinct a chains assemble to form only three heterotrimers (i.e. a1a1a2, a3a4a5 or a5a5a6) [2,3]. The a1(IV) and a2(IV) chains, the so-called ‘classic’ chains, are present in the BM of all tissues, whereas the other two heterotrimers display a more restricted pattern of distribution [2,3]. Unlike the laminins, type IV collagen polymers are stabilized by a network of covalent crosslinks [2]. Indeed, this structural characteristic probably provides the structural basis for the long-appreciated fact that type IV collagen cannot be extracted from BMs assembled in vivo unless the animals are fed a lathyrogen (i.e. a compound such as b-aminopropionitrile that prevents collagen crosslinking by inhibiting lysyl oxidase activity) and the tissues treated with a strong reducing agent [33]. These observations support the contention that disulfide bonds and lysyl oxidase-catalyzed aldimines dominate intermolecular crosslinks generated within the 7S domain [33,34]. Further, more recent analyses of the type IV collagen NC1 domain have led to

Trends in Cell Biology Vol.18 No.11

the novel demonstration of a new, and heretofore unprecedented, type of intermolecular thioether crosslink, which is formed between specific methionine and hydroxylysine or lysine residues residing within apposing NC1 trimers [22,24,35]. Hence, the highly ordered and crosslinked nature of the type IV collagen network confers BMs with their structural integrity [18] and most likely presents migrating cells with their most formidable barrier during transmigration. Independent of the self-polymerizing properties of the laminins and type IV collagens, BM networks are further bridged by non-covalent interactions with nidogens 1 and 2, which in turn also bind the BM heparin sulfate proteoglycan, perlecan [4,6]. Perlecan and other BM-associated proteoglycans, including collagens XV, XVIII and agrin, probably confer charge-dependent selective filtration properties to BMs and serve as reservoirs for heparin-binding growth factors including fibroblast growth factor 2 (FGF2), vascular endothelial growth factor (VEGF) and plateletderived growth factor (PDGF) [36,37]. When do cells cross BMs? Multiple cell types traverse BM barriers in the course of developmental, inflammatory, fibrotic and neoplastic processes (Figure 2). During mammalian development, the early embryo consists of two apposed layers of epithelial cells. With the inception of gastrulation, genesis of the

Figure 2. BM transmigration. (a) During gastrulation, epiblast cells invade at the primitive streak, penetrating the epithelial BM to populate the space beneath the germ layer to produce the mesoderm. (b) At the initial phase of tumor invasion, tumor cells acquire the ability to breach epithelial BM barriers and migrate into the stromal matrix. After traversing the stroma, tumor cells cross the vascular BM both at the primary tumor site (termed intravasation) and at the site of metastasis (i.e. extravasation) to colonize distant organs. (c) During normal immune surveillance and the inflammatory response, leukocyte populations, including polymorphonuclear neutrophils (PMNs), cross the vascular BM barrier to reach the interstitial compartment.

562

Review third germ layer (the mesoderm) occurs when epithelial cells at the primitive streak adopt mesenchymal (i.e. fibroblast-like) characteristics and penetrate the underlying BM to populate the intervening space [11,12,38] (Figure 2a). This theme is conserved throughout development; for example, imaginal-disc eversion and anchor-cell invasion that occur during Drosophila melanogaster and Caenorhabditis elegans morphogenesis, respectively, likewise require cell penetration of the BM [1,19,20]. With the exception of the extracellular matrix remodeling events associated with reproduction in adult mammals (e.g. follicular rupture, blastocyst implantation, uterine resorption and mammary gland involution), it is not clear that proteolytic BM remodeling or epithelial invasion programs are necessarily re-engaged in the adult, save for pathologic processes such as cancer [1,19]. During carcinoma progression, non-invasive, dysplastic epithelial cells proliferate locally as a lesion, termed carcinoma in situ. Unchecked, the localized lesion will acquire the capacity to perforate the subjacent epithelial BM barrier, becoming an invasive carcinoma [14,39] (Figure 2b). It should be stressed, however, that epithelial cells need not adopt a neoplastic status to display BM-invasive characteristics. For example, recent studies demonstrate that, in pathologic fibrotic disease states (e.g. most commonly affecting the lungs or kidneys), epithelial cells can assume a mesenchymal phenotype as they traverse the underlying BM, infiltrate underlying stromal tissues and deposit an excess of interstitial matrix components [40,41]. Although analyses of BM invasion in the current literature most commonly focus on matrix remodeling events associated with development, fertilization and disease, it is important to note that this structural barrier is traversed daily by billions of cells in healthy adult tissues in the course of normal immune surveillance (Figure 2c). Heterogeneous populations of white blood cells (i.e. leukocytes) continuously traffic through BMs as they patrol host tissues in search of potential microbial pathogens [9,10]. Given the fact that leukocytes, as well as normal or neoplastic epithelial cells, can transmigrate BMs, it has long been assumed that each of these cell populations would engage similar mechanisms to cross this structural barrier [17,18,42]. Yet, evidence to support this contention remains far from clear. Mechanisms of BM transmigration: an introduction To probe the mechanisms underlying BM transmigration, it must first be asked: what are the properties of BMs and the associated invasive cell populations that influence the transmigration process? From a structural perspective, the migrating cell will be confronted with a semi-permeable, type-IV-collagen-rich barrier, the pore size of which is dictated by both extracellular matrix (ECM) density and crosslinking [5,21,43]. In turn, the degree to which the cell can deform its cytoplasm and nucleus to traverse a structural pore will determine whether proteolytic remodeling is a required step during BM transmigration [44–46]. Because migrating cells are unable to efficiently negotiate pores with diameters <2.0 mm in size [47,48], a spatial limitation that exceeds BM pore size by a factor of 40-fold [7,8], it would seem that BM organization must be modu-

Trends in Cell Biology

Vol.18 No.11

lated to accommodate cell traffic. Larger BM discontinuities have been identified within distinct sites, such as the pulmonary bed, where circumscribed areas devoid of BM macromolecules have been detected, but their frequency and tissue-specific localization seem inconsistent with a more general role in providing unimpeded passageways for transmigrating cells [23,49]. How might BMs be remodeled to support transmigration? Conceptually, BM pore size might be altered by either proteolytic or non-proteolytic processes. Transmigration associated with proteolysis would involve the localized dissolution of the BM at sites of cell invasion. By contrast, non-proteolytic BM remodeling would necessitate the reversible disassembly of BM superstructure so as to temporarily increase pore diameter to a caliber permissive for transmigration. Currently, data exist to support both of these operating paradigms. Breaking down the door: protease-dependent BM transmigration Evidence for proteolytic BM destruction during transmigration is supported by the observation that tissue-invasive events associated with developmental and disease states are characterized frequently by irreversible changes in BM structure [1,12,13,21,38,39]. During development in worms, flies and mammals, BM effacement is commonly observed at sites of invasion (e.g. within the vulva in C. elegans and the primitive streak in mammals) [1,19,20]. Similarly, BM discontinuities have been identified at sites of carcinoma invasion in vivo [39,50,51,173]. Together, these observations have lent considerable support to the proposition that protease-dependent BM degradation is an event essential to transmigration [1,14,52]. Accordingly, the expression of ECM-degradative proteases has been documented in association with the tissue-invasive phenotype of both normal and neoplastic epithelial cells [1,14,19,52]. Nevertheless, the fact that the mammalian degradome is comprised of hundreds of proteases has led many to speculate whether the identification of those enzymes necessary and/or sufficient for BM transmigration might prove to be an impossible task [53,54]. A historical perspective To date, in vivo models have implicated a confusing – and sometimes contradictory – array of serine proteases [55], cathepsins [56] and metalloproteases [53,57] in tissueinvasive processes. Studies designed to identify the role of specific proteinases in ECM remodeling events associated with invasion in vivo must, however, be interpreted cautiously. In the in vivo setting, transmigration across an intact, cross-linked BM devoid of any preformed passageways is not only associated with the dissolution of the intervening barrier, but also the generation of multiple chemokines and mitogens [19,54,58]. Because proteolytic remodeling of the ECM can effect the generation of novel chemotactic molecules, the inactivation of chemo-repulsive agents, the release or activation of matrix-bound growth factors and the unmasking of cryptic epitopes that can modify cell genotype and or phenotype [19,54,58], the mechanistic interpretation of an ‘invasive’ phenotype in the in vivo setting is challenging, if not problematic. 563

Review Indeed, given the inherent complexities associated with whole animal studies, it might realistically be asked: (i) is it possible to use in vivo models to identify the specific proteases involved directly in the BM remodeling program that specifically underlie invasion; and (ii) if not, how do we proceed? From a reductionist perspective, an ‘ideal’ model might rely on an ex vivo system, wherein the ability of a normal or neoplastic cell type to traverse a BM barrier could be monitored directly. Studies initiated in the late 1970 s used just such an approach with tissues isolated from donor organisms [16,59]. A popular source of tissue was the human placenta, wherein the amniotic BM is sandwiched between an overlying layer of epithelial cells and a dense, underlying layer of interstitial collagen (dominated by an interwoven mat of fibrillar type I collagen) [59,60]. After removal of the amniotic epithelial layer, a cell type of interest could be cultured atop the ‘naked’ matrix and its invasion through the BM tracked [16,61]. Interestingly, such studies lent early support to the contention that BM invasion requires the mobilization of a subset of metalloenzymes encoded by the matrix metalloproteinase (MMP) gene family [16,59,61], a conclusion that might well hold 30 years later. The type IV ‘collagenase’ conundrum: are the ‘usual’ suspects the key players? With close to 20 000 publications in the literature dedicated to the MMP family, it is perhaps not surprising that these proteases have been associated with tissue-invasive and remodeling events in a variety of developmental and pathophysiologic states ranging across the evolutionary spectrum [1,19,20,62]. In overview, MMPs are a class of 25 enzymes that are synthesized as latent zymogens [19,53]. When productive conformational changes occur between the autoinhibitory MMP prodomain and the catalytic domain, proteolytic activity is unmasked [63]. MMPs are conveniently divided into two general classes; the secreted MMPs and membrane-type MMPs (MT-MMPs), with several subclass denominations based on domain composition [19,53]. With specific regard to BM remodeling, attention has long been focused on the so-called type IV collagenases, MMP-2 (gelatinase A) and MMP-9 (gelatinase B) [6,18,52]. Because the covalently crosslinked, polymeric network of type IV collagen provides BMs with the bulk of their structural integrity [2,43], the early classification of MMP-2 and MMP-9 as type IV collagenases drew considerable attention to these enzymes – all the more so given that these metalloproteinases were found to be widely expressed during the remodeling of normal and neoplastic tissues [52,62,64,65]. This series of observations, although establishing a strong correlation between MMP-2 or MMP9 expression and events associated with BM remodeling, requires cautious interpretation because landmark studies assigning type IV collagenolytic activity to these enzymes have been called into question. Whereas the cross-linked type IV collagen network of the native BM is thermally stable at 378C and resistant to many forms of proteolytic attack [66–68], the solution-phase a(IV) chains used as substrate in early studies are more sensitive to thermal 564

Trends in Cell Biology Vol.18 No.11

denaturation and, hence, proteolytically labile (i.e. as the helical domains of type IV collagen unwind as a function of increasing temperature, the individual chains display an increased number of cryptic sites susceptible to proteolytic attack) [68–70]. Accordingly, although MMP-2 or MMP-9 can degrade type IV collagen in solution at 378C, the ability of the proteinases to cleave the insoluble, polymeric substrate is limited at lower temperatures at which triplehelical character is maintained [67–70]. These observations are consistent with other reports, although largely overlooked, documenting the resistance of crosslinked type IV collagen to MMP-2- or MMP-9-dependent attack at physiologic temperatures despite the large number of non-helical domains found within its structure [67–70]. Hydroxyproline-rich peptides (presumably arising from type IV collagen degradation products) can be released from isolated BMs incubated with either MMP-2 or MMP-9 [62,71], but such tissues can also contain non-helical a(IV) chains that would prove susceptible to attack by these proteases, thus, confounding interpretation of these results [72]. Similarly, although BMs isolated from model organisms (e.g. sea urchins) have been used as substrates for studies of mammalian cell invasion [73], the low hydroxyproline content of their type IV collagen renders them sensitive to thermal denaturation at 378C [74]. MMP-2 and MMP-9: in vivo studies These issues notwithstanding, the roles of MMP-2 and MMP-9 in BM remodeling and transmigration can be further questioned from the perspective of targeted gene inactivation and mutation in mice and humans, respectively. In animal studies, mice harboring inactivating mutations in Mmp2, Mmp9 or both Mmp2 and Mmp9 do not exhibit gross defects in a variety of developmental events associated with BM remodeling and are viable and fertile [25,28,75]. Indeed, during events ranging from branching morphogenesis and inflammation to postnatal angiogenesis (a program of new blood vessel development that is initiated by the focal dissolution of the surrounding vascular BM), matrix remodeling proceeds in an unabated fashion in MMP-2 and MMP-9 double-null mice [25,27,28]. Citation of these specific reports should not be construed to indicate that MMP-2 or MMP-9 do not assume major roles in matrix remodeling events. In disease models, for example, tumor growth, metastasis and angiogenesis – pathologic states linked frequently to BM remodeling – can be modulated in either MMP-2- or MMP-9-null mice [64,65,76,77]. It is difficult, however, to divine a consistent role for these proteases in pathologic BM remodeling. Depending on the specific model, MMP-2-null status might, or might not, affect tumor growth [64,76,78]. Similarly, MMP-9-deficiency can slow, or actually accelerate, cancer-cell proliferation and angiogenesis [26,64,79,80]. Adding to these complexities, loss of MMP-2 can, paradoxically, accelerate bone loss and induce joint-erosive disease in mice and humans [81–83]. Given that the type IV collagenolytic activities of MMP2 and MMP-9 remain the subject of debate, and that the double-null mice remain viable and fertile, it seems likely that the crucial proteolytic targets for these MMPs lie outside the BM itself. Indeed, context-specific proteolytic

Review cleavage of both ECM and non-ECM substrates can regulate both cell invasive and migratory processes through mechanisms distinct from the removal of structural barriers in the complex in vivo environment. For example, MMP-dependent proteolysis of adhesion molecules such as b-dystroglycan [84,85], integrins [86] or cadherins [19] can affect cell motility. Further, chemotactic factors can either be activated or inactivated following MMP-2 or MMP-9-dependent proteolysis, and more recent proteomic studies have identified hundreds of new substrates for these enzymes [19,87]. Hence, MMP-2 and MMP-9 can potentially modulate cell invasion through multiple mechanisms, emphasizing the limits of in vivo models as a means of identifying the mechanisms by which these proteases impact BM transmigration. Revisiting the 1970s: modeling BM transmigration ex vivo Given the complexities and uncertainties associated with attempts to generate mechanistic insights from in vivo studies alone, it might be argued that the time is ripe for returning to more ‘simple’ models that enable direct analyses of cell–BM interactions ex vivo. Although the amniotic BM first championed >30 years ago provided a powerful tool for analysis, several caveats limit its use for studies of transmigration per se. First, early studies demonstrated that only very small numbers of cells successfully traversed the amniotic BM, a finding perhaps consistent with the fact that this structural barrier does not normally support cell traffic of any type in vivo [16,88]. Second, the amniotic BM is peppered with structural breaches that are native to its inherent architecture, which might permit protease-independent trafficking until the migrating cell reaches the dense, underlying layer of interstitial collagen [60]. Attempts to circumvent many of these limitations led investigators to adopt the use of a BM-mimic derived from harvested extracts of Engelbreth-Holm-Swarm (EHS) sarcoma cells grown in vivo [89]. Commonly termed Matrigel, this viscous but soluble mixture of tumor-derived type IV collagen, laminin, heparin sulfate, proteoglycans and nidogens can be stored in liquid form at 48C [89]. After warming to 378C, the extract polymerizes to form a hydrogel that has been used extensively as a BM-like substratum for transmigration studies [89]. In terms of dissecting tissue-invasive programs, however, it is frequently overlooked that Matrigel is comprised primarily of a laminin isoform found rarely in adult tissues [90] and lacks the complex mix of covalent crosslinks that characterize type IV collagen polymers in vivo [91,92]. Indeed, in contrast to the insolubility of BMassociated type IV collagen deposited in the in vivo setting, Matrigel hydrogels are readily solubilized by chaotropic solutions [91]. Hence, in the absence of the major structural barrier that distinguishes the type IV collagen scaffolding of most BMs assembled in vivo, it is unclear whether Matrigel can be used to define the proteolytic machinery required for transmigration in situ [21]. Although proteinases are nonetheless implicated frequently in Matrigel ‘invasion’ assays [93,94], the ability of MMPs to alternatively drive a motile response by

Trends in Cell Biology

Vol.18 No.11

cleaving cell-adhesion molecules, generating bioactive molecules by hydrolyzing Matrigel-associated macromolecules or even activating proteases embedded within the BM extract itself cannot be discounted readily [95–97]. Given the caveats associated with analyses of tissueinvasive behavior in vivo or with BM mimics in vitro, what alternatives exist? From our perspective, the preferred substratum might be characterized as: (i) an authentic BM assembled in vivo; (ii) a structural barrier known to support normal and neoplastic cell trafficking in vivo; and (iii) a tissue amenable to ex vivo culture. Given that the peritoneal BM fulfills each of these criteria [51,98–100], we opted recently to use this sheet-like tissue as a (patho)physiologically relevant model for characterizing the invasive machinery necessary to support BM transmigration [21]. Strikingly, despite the apparent complexity of the proteinase genome with all the complicating features of protease redundancy, compensation and adaptive development, these studies seem to narrow the cast of suspects to a small subclass of proteases that function as crucial determinants of the BM-transmigration program [19– 21,101]. A novel BM-transmigration program Given a consensus that carcinoma cells and their normal counterparts mobilize similar machinery to traverse BMs (i.e. both cell populations undergo an epithelial–mesenchymal cell transition as they downregulate cell–cell and cell–BM interactions, rearrange cytoskeletal elements and adopt a fibroblast-like motile phenotype [102]), ex vivo studies of pro-invasive activity were initiated to define the cancer cell–BM-transmigration program [21]. Using denuded peritoneal BM explants, multiple carcinoma cell lines (e.g. breast, prostate, pancreatic and squamous cell) exhibited invasive activity either ‘spontaneously’ in response to motogenic signals generated by aberrant transcription factor expression or after the addition of proinvasive growth factors [21,102]. In each case, carcinoma cells initiated invasion by first inserting pseudopod-like membrane extensions through the BM barrier in a manner that morphologically recapitulates cancer cell invasion in vivo (Figure 3a,b). Consistent with a protease-dependent mechanism, type IV collagen degradation products could also be identified in the perforated BM [21]. Although earlier studies had provided indirect evidence supporting a role for serine, cysteine or aspartyl proteinases during BM invasion programs [55,56,103], validated inhibitors of these respective proteinase families were unable to suppress invasive activity in this ex vivo model [21]. In marked contrast, inhibitors directed against MMP-family members not only ablated the cancer-cell-mediated degradation of BM-associated type IV collagen, but transmigration as well [21]. As proteolysis and invasion were unaffected by tissue inhibitor of metalloproteinases-1 (TIMP-1), an endogenous inhibitor specific for most secreted MMPs as well as the glycosylphosphatidylinositol (GPI)-anchored MMPs (i.e. MT4-MMP and MT6-MMP) [104], these data provided indirect support for the contention that BM remodeling might be mediated by one or more of the TIMP-1-insensitive MT-MMPs (i.e. MT1-MMP, MT2MMP, MT3-MMP and MT5-MMP) [21], which are 565

Review

Trends in Cell Biology Vol.18 No.11

Figure 3. Proteolytic invasion of BM barriers. (a) Transmission electron micrograph of a carcinoma cell extending an invadopodium-like process across a BM in vivo. Arrowheads mark the BM. (b) Recapitulation of the in vivo BM-transmigration program in vitro by culturing MDA-MB-231 breast carcinoma cells atop a peritoneal explant. The invading cell extends an invadopodial process (*) through the intact peritoneal BM (arrows) into the underlying interstitial matrix (I). (c) When engineered to express MT1-MMP, invasion-null COS-1 cells activate a BM-transmigration program. Inset magnifies the discontinuous BM (arrows) associated with the COS-1 cell invadopodium. (d,e) Scanning electron micrographs of peritoneal BM (en face view) before (0 days) and after an eight-day culture period with MT1-MMP-expressing cancer cells. After detergent lysis, the BM is decorated with corroded pits that mark sites of invasion. Part (a) adapted, with permission, from Ref. [173]. Parts (b–d) adapted, with permission, from Ref. [21].

characterized by a type I transmembrane domain that tethers the MMP to the cell surface. Indeed, BM degradation and transmigration by cancer cells of epithelial, mesenchymal or neural crest origin were inhibited effectively by TIMP-2, a second member of the TIMP family that targets all of the secreted and the MT-MMP family members [21,105]. Most striking, perhaps, is the fact that cancer cells of multiple origins were unable to mobilize alternate proteolytic systems that supported BM invasion – even after eight or more days in culture – when MMP activities were inhibited [21]. In an effort to identify those MMPs that directly confer host cells with the ability to remodel or transmigrate BM barriers, a series of secreted or membrane-anchored MMPs were subsequently expressed in COS cells (i.e. an epithelial cell type that displays no BM degradative or invasive activity) [21]. Independently of their secretion as latent or fully active enzymes, neither MMP-2, 3, 7, 9, 11 nor 13 endowed COS cells with BM remodeling activity [21]. Although the possibility that these proteinases hydrolyzed 566

BM components at levels below the limits of morphologic or biochemical resolution used in these studies cannot be ruled out, none of the secreted MMPs tested conferred COS cells with BM pro-invasive potential [21]. In marked contrast, however, three MT-MMP family members (i.e. MT1-MMP, MT2-MMP or MT3-MMP) conferred COS cells with the ability to both degrade and transmigrate the subjacent BM [21] (Figure 3c). BM remodeling activity did not, however, extend to MT4-MMP, MT5-MMP or MT6-MMP. Importantly, although it might be argued that a requirement for MT1, 2, 3-MMPs during BM transmigration arose as a consequence of their ability to indirectly generate type IV collagenolytic activity by converting the MMP-2 zymogen to its active form [86], MT-MMP-dependent BM remodeling and invasion proceeded in an unabated fashion in the absence of active MMP-2 [21]. Because these findings are limited to an experimental system wherein the implicated MT-MMPs were overexpressed in recipient COS cells, further studies were performed, which demonstrated that specific MT-MMP silencing by small inhibitory

Review

Trends in Cell Biology

Vol.18 No.11

Figure 4. Invadopodium structure and function. At sites of proteolytic BM disruption, integrin activation promotes F-actin polymerization and cytoskeletal tension to induce cytoplasmic protrusion toward matrix targets. MT-MMP zymogens are processed to their active form within the TGN by proprotein convertase family members such as furin. Activation of RhoA and Cdc42 signals IQGAP1 and the exocyst complex to deliver Rab8-coated vesicles containing activated MT-MMPs from the TGN to ECM contact sites at the developing invadopodium. After VAMP-7-mediated vesicle fusion, surface MT-MMP activity generates a pericellular zone of BM and type I collagen degradation products associated with low pH generated by a vacuolar-type H+-ATPase and/or Na+/H+ exchanger isoform 1 (V-ATPase or NHE1). Proteolysis-dependent translocation of MT-MMPs, continued F-actin polymerization, integrin-mediated adhesion and bioactive matrix degradation products collaborate to promote invadopodium extension into the interstitial matrix, triggering BM transmigration.

RNAs (siRNAs) abrogated the ability of human cancer cells to traverse BMs isolated from either mouse, rat, canine or human sources [21]. A required role for MT-MMPs must, of course, be validated in an in vivo system. This work is ongoing, but our preliminary data using the developing chick embryo as a host organ indicate that our in vitro findings might well prove applicable to the in vivo setting. Nevertheless, the specific BM components cleaved by the MT-MMPs in situ remain to be characterized, as does their role in promoting BM transmigration. MT-MMPs function as pro-invasive motors that drive BM transmigration The ability of any single proteinase to effectively transform an invasion-null cell into a BM-invasive phenotype is, to say the least, surprising. Analyses of the structural and functional relationships that underlie the ability of this subset of MT-MMPs to mediate BM remodeling and transmigration have yielded several insights into this process: First, MT1-MMP, MT2-MMP and MT3-MMP, although synthesized as latent enzymes, undergo intracellular activation before their display at the cell surface via the proteolytic removal of their respective prodomains by members of the ubiquitously expressed proprotein convertase family [86,106,107]. Second, in the antiproteinase-rich extracellular environment of plasma, the active proteinases only display BM-degradative and pro-invasive activities when tethered to cell surface – a characteristic that might imbue these enzymes with unique activities [21,108]. For example, whereas soluble MMPs depend on diffusion to locate target substrates, the localization of MT-MMP catalytic activity to the cell surface provides for the direct apposition of

enzyme and substrate at sites directed by the repertoire of receptors expressed on the cell membrane at the invasive front [109,110]. Additionally, MT-MMPs could be further concentrated at sites of active BM remodeling by trafficking to an actin-rich, adhesive cell protrusion termed, the invadopodium [109,110] (Figure 4). Although the precise mechanisms underlying MT-MMP trafficking to zones of BM effacement remain unclear, the MT1-MMP cytoplasmic domain itself is dispensable, because cells expressing an MT1-MMP cytosolic tail-deletion mutant retain invasive activity while continuing to focus proteolytic activity to invadopodia [21,108]. More recent studies indicate that the localization of MT1-MMP to the invasive front of migrating cells might be regulated by a cohort of signaling molecules including Rab8, the exocyst complex and membrane fusion machinery [111–113] (Figure 4). Third, despite the fact that the C-terminal MT1-MMP hemopexin-homology domain has been postulated to endow the protease with the ability to bind and unwind triple-helical collagen domains, thereby enabling the catalytic domain successive access to the individual collagen chains during hydrolysis [108,114–116], hemopexindeletion mutants of MT1-MMP fully support BM remodeling [21,108]. Nonetheless, a crucial role for the MT-MMP catalytic domain is underlined by the fact that chimeric MT1-MMP constructs harboring the catalytic domain of the interstitial collagenase, MMP-1, do not support BM degradative activity [21]. Likewise, neither BM remodeling nor transmigration are elicited when COS cells are engineered to express membrane-tethered, active forms of either MMP-2 or MMP-9 [21]. The possibility that triplehelicase activity resides within the MT1-MMP catalytic domain or the proline-rich hinge region that bridges the 567

Review catalytic and hemopexin domains remains to be determined [108]. Fourth, a recent elucidation of the biophysical mechanics underlying collagen degradation indicates that MT-MMPs could promote BM transmigration not only by dissolving ECM barriers, but also by acting as pro-invasive motors [117]. In this model, MMPs display a proteolysisdependent diffusion bias wherein the MMP affinity for native substrates exceeds that for degraded products, preventing retrograde diffusion of MMPs into sites depleted of targets in accordance with a ‘burnt bridge’ model [117]. Hence, MT-MMPs operating at the cell surface–BM interface might use a proteolysis-dependent diffusion mechanism to ‘pull’ the cell through the native ECM concurrently with matrix degradation, thus, potentially integrating proteolysis and invasion into a single concerted process. Of course, attempts to assign specific or direct-acting properties to the MT-MMPs are complicated by the fact that matrix proteolysis itself can exert global effects on cell function. In the course of BM proteolysis, the MT-MMPexpressing cell would be bathed in a host of bioactive degradation products in association with potential fluctuations in local pH [6,19,54,58,118] (Figure 4). Further, as new ligands are exposed (e.g. as the leading edge of the invading cell traverses the perforated BM, contact would be made with the underlying interstitial tissue), integrin usage would switch with attendant changes in downstream signaling and the application of cell-mediated mechanical forces [119,120]. Finally, changes in the biomechanical properties of the degraded BM itself would be expected to affect both cell shape and, consequently, gene expression [121–123]. Independent of the impact of these various signaling pathways on cellular behavior, MT1–3MMPs seem unique relative to all other proteinases examined to date with regard to their ability to consolidate focal proteolysis with BM transmigration. Doubtless, other proteinases as well as glycosidases and sulfatases – derived from normal and neoplastic cells alike – can hydrolyze multiple BM components [71,124–129], but their ability to directly promote invasive activity has not been demonstrated. Opening the door: non-proteolytic BM transmigration Clearly, proteinases, especially MT-MMPs, can function as crucial regulators of transmigration by degrading BMassociated structural barriers and generating a conduit through which normal and malignant epithelial cells can pass. By contrast, are there any BM-transmigration events wherein proteolysis is dispensable? Recently, non-proteolytic means of cell invasion have been reported using Matrigel as a BM surrogate [103,130,131]. Carcinoma cells are capable of invading Matrigel matrices using a protease-independent mode of migration termed ‘amoeboid’ movement, which has been proposed as a compensatory mode of invasion that is activated when proteolysis is absent or inhibited [103,130,131]. By contrast, using native BMs as a structural barrier, carcinoma cells are unable to invade the crosslinked meshwork in the absence of MT-MMP activity [21,101]. Because the network of BM components found in Matrigel is stabilized almost entirely by non-covalent 568

Trends in Cell Biology Vol.18 No.11

forces, the pathophysiologic relevance of amoeboid-type invasion in this medium remains to be defined. Studies of protease-independent migration in Matrigel or other non-crosslinked ECM constructs are not without value, given that paradigms derived from these studies might apply to modes of migration operative within specialized matrices. For example, migration through Matrigel could recapitulate cell migration through non-crosslinked BMs that potentially define fetal tissues (i.e. although commonly overlooked, the composition of the fetal or early postnatal ECM is distinct from that in mature animals [132–134]) or select tissues, such as the lens capsule, in which BM crosslinks can remain purposefully unassembled [135,136]. Leukocyte BM transmigration Whereas the importance of non-proteolytic BM transmigration by carcinoma cells requires further study in other experimental models, are there other cell types that might traverse native, crosslinked BMs in the absence of proteolysis? In vivo, enormous numbers of leukocytes traffic daily across the BM underlying the vascular endothelium – and in contrast to the remodeling events associated with developmental processes or cancer – without leaving detectable perforations in their migratory wake [9]. Likewise, neutrophils (a leukocyte sub-population that mediates host defense and acute inflammation), rapidly traverse the subendothelial BM in vitro via a process that not only leaves the BM intact morphologically, but that also proves resistant to broad-spectrum protease inhibitors – including those directed against MMPs [9]. Indeed, unlike carcinoma cells, neutrophils have not been reported to express MT1MMP, MT2-MMP or MT3-MMP. Although neutrophils do synthesize MT6-MMP [137,138], BM invasive activity has not been assigned to this enzyme [21]. Interestingly, these results are consistent with the fact that neutrophil trafficking through BMs in vivo is unaffected by gene-targeting strategies that ablate each of the major serine proteinases or secreted MMPs known to be expressed in these cells [139–142]. These observations should not be interpreted as indicating that neutrophil-derived proteinases are incapable of degrading BM components. To the contrary, the neutrophil armamentarium is well equipped with the means to proteolytically dissolve BMs, but bulk degradation alone does not seem to support the tightly regulated process of transmigration [9,143]. Though other studies have reported that leukocyte-derived proteinases can modulate tissue-invasive responses in vivo, the observed effects have either been limited to the use of non-specific, synthetic inhibitors or probably occur as an indirect consequence of the ability of the proteinases to generate chemoattractants in a feed-forward fashion [144–148]. For example, neutrophil recruitment to inflammatory sites in vivo is attenuated in mice harboring inactivating mutations in MMP-8 (i.e. the major secreted collagenase expressed by neutrophils) or dipeptidyl peptidase I (i.e. a lysosomal enzyme required for the activation of the major neutrophil serine proteinases, elastase, cathepsin G and proteinase 3) [144–147]. In each case, however, defects in leukocyte trafficking were not ascribed to an impediment in BM transmigration per se. Rather, neutrophil-derived

Review proteinases regulated the magnitude and kinetics of the leukocyte influx indirectly by locally generating ECMdegradation products and inducing the expression or processing of chemokines, which each function as powerful chemoattractants that sustain the inflammatory response [144–147]. Further distinguishing leukocyte transmigration from other modes of invasion is the fact that in vitro studies have demonstrated that neutrophils only traverse BM barriers (including the peritoneum) in the presence of an overlying endothelium that is in direct contact with the BM– a requirement unshared by carcinoma cells [9] (R.R. and S.W., unpublished) (Figure 5a). How might these contrasting observations be reconciled and might they establish a precedent for a new model of BM transmigration?

Trends in Cell Biology

Vol.18 No.11

Cooperative interactions between leukocytes and endothelial cells: a mechanism for controlling BM architecture? Endothelial cells actively participating in leukocyte trafficking are known to display increased cytoskeletal tension relative to the resting endothelium, owing to specific signals from adherent leukocytes [149–151]. A putative mechanism underlying reversible BM remodeling could rely on a precedent wherein the endothelial cell functions as the gatekeeper of transmigration by exerting a mechanical ‘pull’ on the underlying BM [149]. That is, in response to signals generated during leukocyte-endothelial cell adhesive interactions, exogenous tension is generated, which might regulate ECM structure and organization as a consequence of the inherent physicochemical charac-

Figure 5. BM transmigration by neutrophils. (a) In the upper panel, a transmission electron micrograph shows two neutrophils (PMN) that have crossed an endothelial cell monolayer in response to chemotactic signals and now reside temporarily atop the underlying BM. In the lower panel, a PMN begins to insert a cytoplasmic process across the endothelial cell BM (arrowheads). (b) Endothelial cells in association with the vascular BM. The type IV collagen network of the vascular BM underlying the endothelial cells is stabilized by disulfide and (hydroxyl)lysine-derived crosslinks in the 7S tetramer and (hydroxy)lysine–methionine thioether bonds in the NC1 dimer (inset). (c) During neutrophil (PMN) transmigration, endothelial cell tractional forces and/or surface-associated enzymatic activity, including protein disulfide isomerases, might promote disruption of type IV collagen network crosslinks and local BM disassembly, increasing local pore size so as to permit neutrophil transmigration via a proteinaseindependent process.

569

Review teristics of the BM. Indeed, BMs have been proposed to display thixotropic properties: an increase in applied force generates a change in BM viscosity, possibly via solubilization of protein network components, thus, altering BM permeability to macromolecules and perhaps even cells [152]. How might these observations be resolved on a molecular level? Currently, we hypothesize that a reversible, endothelial cell-mediated disassembly of quaternary interactions or covalent crosslinks within the type IV collagen network might generate a BM microenvironment wherein nonproteolytic modes of migration might function (Figure 5). From a teleologic perspective, a reversible process would enable the rapid and constitutive trafficking of leukocytes across intact BMs (i.e. it is difficult to envision a proteolytic process that would constantly require the repair of a neverending stream of BM defects). Plasticity of supramolecular interactions in BMs is supported by the recent observation that cryptic epitopes embedded within crosslinked a(IV) NC1 dimer domains can be exposed in vivo, indicating that hydroxylysine or lysine–methionine crosslinks and the associated intermolecular quaternary interactions are not necessarily static, irreversible structures [35,153]. Additionally, lysyl oxidase-generated aldimine groups found in the 7S domains of type IV collagen are potentially labile in the presence of tractional forces [154], indicating that contractile forces activated during leukocyte transmigration [88,102] might reverse BM crosslinks. Furthermore, a(IV) disassembly and repair in vascular BMs could be facilitated by the action of cell-surface-associated enzymes, such as protein disulfide isomerases, that are potentially capable of catalyzing thiol-disulfide exchange reactions between intra- and inter- molecular bonds within the type IV collagen network [155]. Hence, cooperation between tractional forces and the activity of cell surface enzymes on either the endothelial cell or leukocyte itself theoretically would enable the reversible opening and closing of BM crosslinks to promote local a(IV) disassembly in tandem with a required increase in BM pore size (Figure 5b,c). A requirement for ‘opening and closing’ type-IV-collagen-associated crosslinks is predicated on the assumption that all BMs are constructed similarly. Because BMs can vary in terms of their content of covalently crosslinked NC1 dimers [2,3,136], it remains possible – although unproven – that the vascular BM contains regions characterized by lower degrees of NC1 dimerization. In such a scenario, reversible disruptions of type IV collagen quaternary interactions might be possible, thus, permitting non-proteolytic transmigration to occur without changing crosslink status [119,148,153]. Along these lines, it is interesting to note that new forms of synthetic rubbers – extensible polymers that are usually stabilized by covalent crosslinks – have recently been developed from networks of hydrogen-bonded materials alone [156]. In contrast to conventional rubbers that contain covalent crosslinks, these new synthetics can be cut and then induced to ‘self-heal’ by simply joining the fractured edges together at room temperature, whereby the physical properties of the native material are restored [156]. Intriguingly, the self-healing rubber is comprised of only two materials – one 570

Trends in Cell Biology Vol.18 No.11

that is able to directionally associate with its identical partners to form linear chains and a second that is able to non-covalently crosslink the chains together [156]. In many ways, BMs follow a similar structural design wherein self-associating polymers of type IV collagen and laminin are non-covalently held in association with one another by the nidogens [4,6]. Though type IV collagen networks can also be stabilized by covalent forces, perhaps to better deal with mechanical stresses that act upon motile organisms, reversible association of these crosslinks could recapitulate the structure of a self-healing rubber. In this manner, a reversible system might be envisioned wherein leukocyte trafficking could be accommodated by a self-repairing structure. Indeed, it has been proposed that the vascular BM displays a unique supramolecular organization at the level of post-capillary venule, rendering it permissive for transmigration [148]. Of course, leukocyte traffic is not restricted to the vascular bed and, in response to pro-inflammatory stimuli, immune cells are able to cross BM barriers throughout the body. Applying Occam’s razor, we are hesitant to invoke complex mechanisms whereby different leukocyte populations – each displaying their own unique repertoire of proteolytic enzymes – mobilize distinct systems to cross variably structured BMs. Although functional and morphologic evidence available to date indicates that non-proteolytic mechanisms might be activated during leukocyte BM transmigration, rigorous proofs of reversible, cell-mediated changes in BM structure have yet to be established. Tissue invasion without BM transmigration? Finally, although our perspective has focused on BM invasion, it is important to recall that organized patterns of cell movement through complex tissues can be observed under conditions in which the BM might never be crossed. During normal branching morphogenesis – in both prenatal and postnatal states – epithelial-cell-lined tubules extend through the dense interstitial matrices without displaying BM gaps [157–159]. In this specialized case, a continuous BM expands laterally because new matrix is apparently spliced into the existing BM (perhaps by depositing newly synthesized BM macromolecules into a substratum comprised of, as yet, uncrosslinked components) [157–159]. This organized structure then moves forward and infiltrates stromal tissue that undergoes structural reorganization in response to uncharacterized mesenchymal-cell activities [157–159]. Should neoplastic cells prove able to access similar ‘invasion’ programs, normal tissues could well be infiltrated by differentiated tumor-cell populations that retain intact BM structure. Beyond the BM: the 3D frontier After local BM transmigration, invasive cells previously situated atop a 2D BM substratum encounter the interstitial matrix, a 3D network of fibrillar collagens, glycoproteins and proteoglycans [160]. To proceed in this new environment, epithelial cells adapt to the 3D milieu via acquisition of a mesenchymal cell phenotype, alteration of adhesion receptor complement and redirection of proteolytic enzymes to support invasion through the stroma. Interestingly, the ability of normal and solid tumor cells

Review to infiltrate, proliferate and differentiate within the interstitial compartment has also been linked to MT1-MMP activity [108,161–164]. The recent characterization of cellular pathways that trigger the epithelial–mesenchymal transition (EMT) has begun to define this process [102], but the means by which cells coordinate proteolysis, EMT, and BM transmigration into a concerted 2D-to-3D transition process is far from understood. Concluding remarks and future perspectives Despite the obvious strengths of in vitro models, the full repertoire of proteolytic and non-proteolytic systems available to normal or neoplastic cells might only be accessible in vivo. As such, additional insights into the role of MTMMPs in BM transmigration will be gleaned from further analyses of gene-targeted mice. Whereas many developmental BM-invasive events are intact in both Mmp14 and Mmp16 (MT3-MMP) mutant mice [165,166], current studies indicate that the expression of either MT1-MMP, MT2-MMP or MT3-MMP alone is sufficient for BM invasion [21]. As such, functional redundancy for the MTMMPs in developmental BM transmigration is likely. A mouse model of Mmp15 (MT2-MMP) disruption has yet to be characterized. However, because MT2-MMP has been reported to be the dominant MT-MMP expressed in epithelial cells [167], the study of Mmp15 deficiency alone or in combination with other MT-MMP-null mutants should augment our understanding of proteinase-dependent BM transmigration. Likewise, in vivo studies could help address the largely unknown variability in tissue levels of type IV collagen cross-linking and might dictate the final balance between deploying proteolytic versus non-proteolytic mechanisms for BM invasion. Nevertheless, although the ability to extrapolate insights developed in model organisms to mammalian species can sometimes prove limiting, it is interesting to note that Drosophila and C. elegans MMPs assume major roles in controlling BM remodeling and invasion by normal and neoplastic cells [1,19,20]. A pre-eminent role for MT1–3-MMPs in BMtransmigration programs does, however, stand in contrast to the often-cited failure of synthetic MMP inhibitors to exert beneficial therapeutic effects in cancer patients [168]. Given the complex substrate repertoire of MMPs and the absence of MT-MMP-specific inhibitors, this outcome is not necessarily unexpected [54]. More importantly – and perhaps, simply – it remains unclear whether the levels of MMP inhibitors reached in human patients have been sufficient to provide a long-term blockade of powerful proteases localized specifically to the cell-matrix interface [169,170]. Consistent with this caveat, preliminary studies indicate that currently available inhibitors would have to be used at levels far exceeding those reached in clinical trials to block invasion through the BM or the underlying interstitium [21], thereby possibly necessitating the development of new targeted therapies. Should further studies reinforce a crucial role for MT-MMPs in BM invasion programs, the value of developing such inhibitors would probably be worthy of reconsideration. Do protease-independent mechanisms of BM transmigration also warrant further study? Clearly, careful examination of a potential, non-proteolytic leukocyte BM

Trends in Cell Biology

Vol.18 No.11

invasion program would elucidate the novel mechanisms that underlie immune cell trafficking across native BMs and the molecular processes that might regulate BM disassembly and repair. Indeed, given the proposition that tumor cells might cop-opt myeloid stratagems for crossing BM structures by fusing with leukocytes in vivo to generate a unique brand of ‘hybrid vehicle’ [171,172], an analysis of proteinase-dependent and independent mechanisms of BM invasion could provide new and important insights into physiologic and pathologic invasion programs in the years to come. References 1 Sherwood, D.R. (2006) Cell invasion through basement membranes: an anchor of understanding. Trends Cell Biol. 16, 250–256 2 Khoshnoodi, J. et al. (2008) Mammalian collagen IV. Microsc. Res. Tech. 71, 357–370 3 Hudson, B.G. et al. (2003) Alport’s syndrome, Goodpasture’s syndrome, and type IV collagen. N. Engl. J. Med. 348, 2543–2556 4 Yurchenco, P.D. et al. (2004) Basement membrane assembly, stability and activities observed through a developmental lens. Matrix Biol. 22, 521–538 5 Candiello, J. et al. (2007) Biomechanical properties of native basement membranes. FEBS J. 274, 2897–2908 6 Kalluri, R. (2003) Basement membranes: structure, assembly and role in tumour angiogenesis. Nat. Rev. Cancer 3, 422–433 7 Abrams, G.A. et al. (2000) Nanoscale topography of the corneal epithelial basement membrane and Descemet’s membrane of the human. Cornea 19, 57–64 8 Abrams, G.A. et al. (2000) Nanoscale topography of the basement membrane underlying the corneal epithelium of the rhesus macaque. Cell Tissue Res. 299, 39–46 9 Huber, A.R. and Weiss, S.J. (1989) Disruption of the subendothelial basement membrane during neutrophil diapedesis in an in vitro construct of a blood vessel wall. J. Clin. Invest. 83, 1122–1136 10 Ley, K. et al. (2007) Getting to the site of inflammation: the leukocyte adhesion cascade updated. Nat. Rev. Immunol. 7, 678–689 11 Fujiwara, H. et al. (2007) Regulation of mesodermal differentiation of mouse embryonic stem cells by basement membranes. J. Biol. Chem. 282, 29701–29711 12 Viebahn, C. et al. (1995) Morphology of incipient mesoderm formation in the rabbit embryo: a light- and retrospective electron-microscopic study. Acta Anat. (Basel) 154, 99–110 13 Viebahn, C. (1995) Epithelio-mesenchymal transformation during formation of the mesoderm in the mammalian embryo. Acta Anat. (Basel) 154, 79–97 14 Christofori, G. (2006) New signals from the invasive front. Nature 441, 444–450 15 Puente, X.S. et al. (2003) Human and mouse proteases: a comparative genomic approach. Nat. Rev. Genet. 4, 544–558 16 Mignatti, P. et al. (1986) Tumor invasion through the human amniotic membrane: requirement for a proteinase cascade. Cell 47, 487–498 17 Liotta, L.A. et al. (1980) Metastatic potential correlates with enzymatic degradation of basement membrane collagen. Nature 284, 67–68 18 Hu, J. et al. (2007) Matrix metalloproteinase inhibitors as therapy for inflammatory and vascular diseases. Nat. Rev. Drug Discov. 6, 480–498 19 Page-McCaw, A. et al. (2007) Matrix metalloproteinases and the regulation of tissue remodelling. Nat. Rev. Mol. Cell Biol. 8, 221–233 20 Srivastava, A. et al. (2007) Basement membrane remodeling is essential for Drosophila disc eversion and tumor invasion. Proc. Natl. Acad. Sci. U. S. A. 104, 2721–2726 21 Hotary, K. et al. (2006) A cancer cell metalloprotease triad regulates the basement membrane transmigration program. Genes Dev. 20, 2673–2686 22 Than, M.E. et al. (2002) The 1.9-A crystal structure of the noncollagenous (NC1) domain of human placenta collagen IV shows stabilization via a novel type of covalent Met-Lys cross-link. Proc. Natl. Acad. Sci. U. S. A. 99, 6607–6612 571

Review 23 Howat, W.J. et al. (2001) Basement membrane pores in human bronchial epithelium: a conduit for infiltrating cells? Am. J. Pathol. 158, 673–680 24 Vanacore, R.M. et al. (2005) Identification of S-hydroxylysylmethionine as the covalent cross-link of the noncollagenous (NC1) hexamer of the a1a1a2 collagen IV network: a role for the posttranslational modification of lysine 211 to hydroxylysine 211 in hexamer assembly. J. Biol. Chem. 280, 29300–29310 25 Perez, S.E. et al. (2005) Matrix metalloproteinases 2 and 9 are dispensable for pancreatic islet formation and function in vivo. Diabetes 54, 694–701 26 Hamano, Y. et al. (2003) Physiological levels of tumstatin, a fragment of collagen IV a3 chain, are generated by MMP-9 proteolysis and suppress angiogenesis via aV b3 integrin. Cancer Cell 3, 589–601 27 Baluk, P. et al. (2004) Matrix metalloproteinase-2 and -9 expression increases in Mycoplasma-infected airways but is not required for microvascular remodeling. Am. J. Physiol. Lung Cell. Mol. Physiol. 287, L307–L317 28 Itoh, T. et al. (2002) The role of matrix metalloproteinase-2 and matrix metalloproteinase-9 in antibody-induced arthritis. J. Immunol. 169, 2643–2647 29 McKee, K.K. et al. (2007) Role of laminin terminal globular domains in basement membrane assembly. J. Biol. Chem. 282, 21437–21447 30 Li, S. et al. (2005) Laminin-sulfatide binding initiates basement membrane assembly and enables receptor signaling in Schwann cells and fibroblasts. J. Cell Biol. 169, 179–189 31 Marinkovich, M.P. (2007) Tumour microenvironment: laminin 332 in squamous-cell carcinoma. Nat. Rev. Cancer 7, 370–380 32 McMillan, J.R. et al. (2003) Ultrastructural orientation of laminin 5 in the epidermal basement membrane: an updated model for basement membrane organization. J. Histochem. Cytochem. 51, 1299–1306 33 Kleinman, H.K. et al. (1982) Isolation and characterization of type IV procollagen, laminin, and heparan sulfate proteoglycan from the EHS sarcoma. Biochemistry 21, 6188–6193 34 Reiser, K. et al. (1992) Enzymatic and nonenzymatic cross-linking of collagen and elastin. FASEB J. 6, 2439–2449 35 Vanacore, R.M. et al. (2008) A role for collagen IV cross-links in conferring immune privilege to the goodpasture autoantigen: structural basis for the crypticity of b cell epitopes. J. Biol. Chem. 283, 22737–22748 36 Iozzo, R.V. (2005) Basement membrane proteoglycans: from cellar to ceiling. Nat. Rev. Mol. Cell Biol. 6, 646–656 37 Yang, B.G. et al. (2007) Binding of lymphoid chemokines to collagen IV that accumulates in the basal lamina of high endothelial venules: its implications in lymphocyte trafficking. J. Immunol. 179, 4376–4382 38 Nakaya, Y. et al. (2008) RhoA and microtubule dynamics control cellbasement membrane interaction in EMT during gastrulation. Nat. Cell Biol. 10, 765–775 39 Spaderna, S. et al. (2006) A transient, EMT-linked loss of basement membranes indicates metastasis and poor survival in colorectal cancer. Gastroenterology 131, 830–840 40 Venkov, C.D. et al. (2007) A proximal activator of transcription in epithelial-mesenchymal transition. J. Clin. Invest. 117, 482–491 41 Higgins, D.F. et al. (2007) Hypoxia promotes fibrogenesis in vivo via HIF-1 stimulation of epithelial-to-mesenchymal transition. J. Clin. Invest. 117, 3810–3820 42 Wu, B. et al. (2007) Wnt signaling induces matrix metalloproteinase expression and regulates T cell transmigration. Immunity 26, 227– 239 43 Poschl, E. et al. (2004) Collagen IV is essential for basement membrane stability but dispensable for initiation of its assembly during early development. Development 131, 1619–1628 44 Pajerowski, J.D. et al. (2007) Physical plasticity of the nucleus in stem cell differentiation. Proc. Natl. Acad. Sci. U. S. A. 104, 15619–15624 45 Wolf, K. et al. (2007) Multi-step pericellular proteolysis controls the transition from individual to collective cancer cell invasion. Nat. Cell Biol. 9, 893–904 46 Nakayama, M. et al. (2005) Rho-kinase and myosin II activities are required for cell type and environment specific migration. Genes Cells 10, 107–117 47 Kuntz, R.M. and Saltzman, W.M. (1997) Neutrophil motility in extracellular matrix gels: mesh size and adhesion affect speed of migration. Biophys. J. 72, 1472–1480

572

Trends in Cell Biology Vol.18 No.11 48 Tucker, S.P. et al. (1992) Migration of polarized epithelial cells through permeable membrane substrates of defined pore size. Eur. J. Cell Biol. 58, 280–290 49 Warfel, K.A. and Hull, M.T. (1988) Basal lamina fenestrations in the human colon: transmission and scanning electron microscope study. Anat. Rec. 220, 68–75 50 Frei, J.V. (1962) The fine structure of the basement membrane in epidermal tumors. J. Cell Biol. 15, 335–342 51 Birbeck, M.S. and Wheatley, D.N. (1965) An electron microscopic study of the invasion of ascites tumor cells into the abdominal wall. Cancer Res. 25, 490–497 52 Liotta, L.A. and Kohn, E.C. (2001) The microenvironment of the tumour-host interface. Nature 411, 375–379 53 Overall, C.M. and Kleifeld, O. (2006) Tumour microenvironment opinion: validating matrix metalloproteinases as drug targets and anti-targets for cancer therapy. Nat. Rev. Cancer 6, 227–239 54 Overall, C.M. and Blobel, C.P. (2007) In search of partners: linking extracellular proteases to substrates. Nat. Rev. Mol. Cell Biol. 8, 245– 257 55 Pillay, V. et al. (2007) The urokinase plasminogen activator receptor as a gene therapy target for cancer. Trends Biotechnol. 25, 33–39 56 Mohamed, M.M. and Sloane, B.F. (2006) Cysteine cathepsins: multifunctional enzymes in cancer. Nat. Rev. Cancer 6, 764–775 57 Martin, M.D. and Matrisian, L.M. (2007) The other side of MMPs: protective roles in tumor progression. Cancer Metastasis Rev. 26, 717– 724 58 Schenk, S. and Quaranta, V. (2003) Tales from the crypt[ic] sites of the extracellular matrix. Trends Cell Biol. 13, 366–375 59 Liotta, L.A. et al. (1977) Degradation of basement membrane by murine tumor cells. J. Natl. Cancer Inst. 58, 1427–1431 60 Aplin, J.D. et al. (1985) The extracellular matrix of human amniotic epithelium: ultrastructure, composition and deposition. J. Cell Sci. 79, 119–136 61 Persky, B. et al. (1986) Inhibition of proteolytic enzymes in the in vitro amnion model for basement membrane invasion. Cancer Res. 46, 4129–4134 62 Zeisberg, M. et al. (2006) Stage-specific action of matrix metalloproteinases influences progressive hereditary kidney disease. PLoS Med. 3, e100 63 Fu, X. et al. (2008) Activation and silencing of matrix metalloproteinases. Semin. Cell Dev. Biol. 19, 2–13 64 Bergers, G. et al. (2000) Matrix metalloproteinase-9 triggers the angiogenic switch during carcinogenesis. Nat. Cell Biol. 2, 737–744 65 Masson, V. et al. (2005) Contribution of host MMP-2 and MMP-9 to promote tumor vascularization and invasion of malignant keratinocytes. FASEB J. 19, 234–236 66 Linsenmayer, T.F. et al. (1984) Thermal stability of the helical structure of type IV collagen within basement membranes in situ: determination with a conformation-dependent monoclonal antibody. J. Cell Biol. 99, 1405–1409 67 Eble, J.A. et al. (1996) The recognition sites of the integrins a1b1 and a2b1 within collagen IV are protected against gelatinase A attack in the native protein. J. Biol. Chem. 271, 30964–30970 68 Wheatcroft, A.C. et al. (1999) Evidence of in situ stability of the type IV collagen triple helix in human inflammatory bowel disease using a denaturation specific epitope antibody. Matrix Biol. 18, 361–372 69 Mackay, A.R. et al. (1990) Studies on the ability of 65-kDa and 92-kDa tumor cell gelatinases to degrade type IV collagen. J. Biol. Chem. 265, 21929–21934 70 Mackay, A.R. et al. (1990) Basement membrane type IV collagen degradation: evidence for the involvement of a proteolytic cascade independent of metalloproteinases. Cancer Res. 50, 5997–6001 71 Mott, J.D. et al. (1997) Nonenzymatic glycation of type IV collagen and matrix metalloproteinase susceptibility. Kidney Int. 52, 1302–1312 72 Kajimura, D. et al. (2004) Non-helical type IV collagen polypeptides in human placenta. Biochem. Biophys. Res. Commun. 314, 11–16 73 Zeng, Z.Z. et al. (2006) Role of focal adhesion kinase and phosphatidylinositol 30 -kinase in integrin fibronectin receptormediated, matrix metalloproteinase-1-dependent invasion by metastatic prostate cancer cells. Cancer Res. 66, 8091–8099 74 Mayne, J. and Robinson, J.J. (2002) Comparative analysis of the structure and thermal stability of sea urchin peristome and rat tail tendon collagen. J. Cell. Biochem. 84, 567–574

Review 75 Vu, T.H. et al. (1998) MMP-9/gelatinase B is a key regulator of growth plate angiogenesis and apoptosis of hypertrophic chondrocytes. Cell 93, 411–422 76 Itoh, T. et al. (1998) Reduced angiogenesis and tumor progression in gelatinase A-deficient mice. Cancer Res. 58, 1048–1051 77 Ratzinger, G. et al. (2002) Matrix metalloproteinases 9 and 2 are necessary for the migration of Langerhans cells and dermal dendritic cells from human and murine skin. J. Immunol. 168, 4361– 4371 78 Du, R. et al. (2008) Matrix metalloproteinase-2 regulates vascular patterning and growth affecting tumor cell survival and invasion in GBM. Neuro-oncol. 10, 254–264 79 Bendrik, C. et al. (2008) Gene transfer of matrix metalloproteinase-9 induces tumor regression of breast cancer in vivo. Cancer Res. 68, 3405–3412 80 Acuff, H.B. et al. (2006) Matrix metalloproteinase-9 from bone marrow-derived cells contributes to survival but not growth of tumor cells in the lung microenvironment. Cancer Res. 66, 259–266 81 Martignetti, J.A. et al. (2001) Mutation of the matrix metalloproteinase 2 gene (MMP2) causes a multicentric osteolysis and arthritis syndrome. Nat. Genet. 28, 261–265 82 Inoue, K. et al. (2006) A crucial role for matrix metalloproteinase 2 in osteocytic canalicular formation and bone metabolism. J. Biol. Chem. 281, 33814–33824 83 Mosig, R.A. et al. (2007) Loss of MMP-2 disrupts skeletal and craniofacial development and results in decreased bone mineralization, joint erosion and defects in osteoblast and osteoclast growth. Hum. Mol. Genet. 16, 1113–1123 84 Yamada, H. et al. (2001) Processing of b-dystroglycan by matrix metalloproteinase disrupts the link between the extracellular matrix and cell membrane via the dystroglycan complex. Hum. Mol. Genet. 10, 1563–1569 85 Agrawal, S. et al. (2006) Dystroglycan is selectively cleaved at the parenchymal basement membrane at sites of leukocyte extravasation in experimental autoimmune encephalomyelitis. J. Exp. Med. 203, 1007–1019 86 Itoh, Y. and Seiki, M. (2006) MT1-MMP: a potent modifier of pericellular microenvironment. J. Cell. Physiol. 206, 1–8 87 Dean, R.A. and Overall, C.M. (2007) Proteomics discovery of metalloproteinase substrates in the cellular context by iTRAQ labeling reveals a diverse MMP-2 substrate degradome. Mol. Cell. Proteomics 6, 611–623 88 Azzarelli, B. and Lafuze, J. (1987) Amniotic basement membrane: a barrier to neutrophil invasion. Am. J. Obstet. Gynecol. 156, 1130–1136 89 Kleinman, H.K. and Martin, G.R. (2005) Matrigel: basement membrane matrix with biological activity. Semin. Cancer Biol. 15, 378–386 90 Virtanen, I. et al. (2000) Laminin a1-chain shows a restricted distribution in epithelial basement membranes of fetal and adult human tissues. Exp. Cell Res. 257, 298–309 91 Kleinman, H.K. et al. (1986) Basement membrane complexes with biological activity. Biochemistry 25, 312–318 92 Even-Ram, S. and Yamada, K.M. (2005) Cell migration in 3D matrix. Curr. Opin. Cell Biol. 17, 524–532 93 Ueda, J. et al. (2003) Sequence-specific silencing of MT1-MMP expression suppresses tumor cell migration and invasion: importance of MT1-MMP as a therapeutic target for invasive tumors. Oncogene 22, 8716–8722 94 Rizki, A. et al. (2008) A human breast cell model of preinvasive to invasive transition. Cancer Res. 68, 1378–1387 95 Koshikawa, N. et al. (2005) Membrane-type matrix metalloproteinase-1 (MT1-MMP) is a processing enzyme for human laminin g 2 chain. J. Biol. Chem. 280, 88–93 96 Mackay, A.R. et al. (1993) Identification of the 72-kDa (MMP-2) and 92-kDa (MMP-9) gelatinase/type IV collagenase in preparations of laminin and Matrigel. Biotechniques 15, 1048–1051 97 Farina, A.R. et al. (1996) Identification of plasminogen in Matrigel and its activation by reconstitution of this basement membrane extract. Biotechniques 21, 904–909 98 Easty, D.M. and Easty, G.C. (1974) Measurement of the ability of cells to infiltrate normal tissues in vitro. Br. J. Cancer 29, 36–49 99 Buck, R.C. (1973) Walker 256 tumor implantation in normal and injured peritoneum studied by electron microscopy, scanning electron microscopy, and autoradiography. Cancer Res. 33, 3181–3188

Trends in Cell Biology

Vol.18 No.11

100 Witz, C.A. et al. (2001) Composition of the extracellular matrix of the peritoneum. J. Soc. Gynecol. Investig. 8, 299–304 101 Margulis, A. et al. (2005) E-cadherin suppression accelerates squamous cell carcinoma progression in three-dimensional, human tissue constructs. Cancer Res. 65, 1783–1791 102 Peinado, H. et al. (2007) Snail, Zeb and bHLH factors in tumour progression: an alliance against the epithelial phenotype? Nat. Rev. Cancer 7, 415–428 103 Friedl, P. and Wolf, K. (2003) Tumour-cell invasion and migration: diversity and escape mechanisms. Nat. Rev. Cancer 3, 362–374 104 English, J.L. et al. (2006) Individual Timp deficiencies differentially impact pro-MMP-2 activation. J. Biol. Chem. 281, 10337–10346 105 Zhao, H. et al. (2004) Differential inhibition of membrane type 3 (MT3)-matrix metalloproteinase (MMP) and MT1-MMP by tissue inhibitor of metalloproteinase (TIMP)-2 and TIMP-3 regulates proMMP-2 activation. J. Biol. Chem. 279, 8592–8601 106 Yana, I. and Weiss, S.J. (2000) Regulation of membrane type-1 matrix metalloproteinase activation by proprotein convertases. Mol. Biol. Cell 11, 2387–2401 107 Wu, Y.I. et al. (2007) Activation-coupled membrane-type 1 matrix metalloproteinase membrane trafficking. Biochem. J. 407, 171–177 108 Li, X-Y. et al. (2008) Molecular dissection of the structural machinery underlying the tissue-invasive activity of MT1-MMP. Mol. Biol. Cell 19, 3221–3233 109 Artym, V.V. et al. (2006) Dynamic interactions of cortactin and membrane type 1 matrix metalloproteinase at invadopodia: defining the stages of invadopodia formation and function. Cancer Res. 66, 3034–3043 110 Linder, S. (2007) The matrix corroded: podosomes and invadopodia in extracellular matrix degradation. Trends Cell Biol. 17, 107–117 111 Bravo-Cordero, J.J. et al. (2007) MT1-MMP proinvasive activity is regulated by a novel Rab8-dependent exocytic pathway. EMBO J. 26, 1499–1510 112 Sakurai-Yageta, M. et al. (2008) The interaction of IQGAP1 with the exocyst complex is required for tumor cell invasion downstream of Cdc42 and RhoA. J. Cell Biol. 181, 985–998 113 Steffen, A. et al. (2008) MT1-MMP-dependent invasion is regulated by TI-VAMP/VAMP7. Curr. Biol. 18, 926–931 114 Itoh, Y. et al. (2006) Cell surface collagenolysis requires homodimerization of the membrane-bound collagenase MT1-MMP. Mol. Biol. Cell 17, 5390–5399 115 Lafleur, M.A. et al. (2006) Type I collagen abrogates the clathrinmediated internalization of membrane type 1 matrix metalloproteinase (MT1-MMP) via the MT1-MMP hemopexin domain. J. Biol. Chem. 281, 6826–6840 116 Tam, E.M. et al. (2004) Characterization of the distinct collagen binding, helicase and cleavage mechanisms of matrix metalloproteinase 2 and 14 (gelatinase A and MT1-MMP): the differential roles of the MMP hemopexin c domains and the MMP2 fibronectin type II modules in collagen triple helicase activities. J. Biol. Chem. 279, 43336–43344 117 Saffarian, S. et al. (2004) Interstitial collagenase is a Brownian ratchet driven by proteolysis of collagen. Science 306, 108–111 118 Cardone, R.A. et al. (2005) The role of disturbed pH dynamics and the Na+/H+ exchanger in metastasis. Nat. Rev. Cancer 5, 786–795 119 Shintani, Y. et al. (2008) Collagen I-mediated up-regulation of Ncadherin requires cooperative signals from integrins and discoidin domain receptor 1. J. Cell Biol. 180, 1277–1289 120 Gaggioli, C. et al. (2007) Fibroblast-led collective invasion of carcinoma cells with differing roles for RhoGTPases in leading and following cells. Nat. Cell Biol. 9, 1392–1400 121 Chun, T.H. et al. (2006) A pericellular collagenase directs the 3dimensional development of white adipose tissue. Cell 125, 577–591 122 Somogyi, K. and Rorth, P. (2004) Evidence for tension-based regulation of Drosophila MAL and SRF during invasive cell migration. Dev. Cell 7, 85–93 123 Discher, D.E. et al. (2005) Tissue cells feel and respond to the stiffness of their substrate. Science 310, 1139–1143 124 Felbor, U. et al. (2000) Secreted cathepsin L generates endostatin from collagen XVIII. EMBO J. 19, 1187–1194 125 Xu, J. et al. (2001) Proteolytic exposure of a cryptic site within collagen type IV is required for angiogenesis and tumor growth in vivo. J. Cell Biol. 154, 1069–1079

573

Review 126 Heljasvaara, R. et al. (2005) Generation of biologically active endostatin fragments from human collagen XVIII by distinct matrix metalloproteases. Exp. Cell Res. 307, 292–304 127 Cheng, S. et al. (2006) Matrix metalloproteinase 2 and basement membrane integrity: a unifying mechanism for progressive renal injury. FASEB J. 20, 1898–1900 128 Nawroth, R. et al. (2007) Extracellular sulfatases, elements of the Wnt signaling pathway, positively regulate growth and tumorigenicity of human pancreatic cancer cells. PLoS One 2, e392 129 Patel, V.N. et al. (2007) Heparanase cleavage of perlecan heparan sulfate modulates FGF10 activity during ex vivo submandibular gland branching morphogenesis. Development 134, 4177–4186 130 Gadea, G. et al. (2007) Loss of p53 promotes RhoA-ROCK-dependent cell migration and invasion in 3D matrices. J. Cell Biol. 178, 23–30 131 Sahai, E. and Marshall, C.J. (2003) Differing modes of tumour cell invasion have distinct requirements for Rho/ROCK signalling and extracellular proteolysis. Nat. Cell Biol. 5, 711–719 132 Kelleher, C.M. et al. (2004) Vascular extracellular matrix and aortic development. Curr. Top. Dev. Biol. 62, 153–188 133 Niederreither, K. et al. (1995) Coordinate patterns of expression of type I and III collagens during mouse development. Matrix Biol. 14, 705–713 134 Rooney, P. and Kumar, S. (1993) Inverse relationship between hyaluronan and collagens in development and angiogenesis. Differentiation 54, 1–9 135 Kelley, P.B. et al. (2002) Collagen IV in the developing lens capsule. Matrix Biol. 21, 415–423 136 Vanacore, R.M. et al. (2004) The a1.a2 network of collagen IV. Reinforced stabilization of the noncollagenous domain-1 by noncovalent forces and the absence of Met-Lys cross-links. J. Biol. Chem. 279, 44723–44730 137 Kang, T. et al. (2001) Subcellular distribution and cytokine- and chemokine-regulated secretion of leukolysin/MT6-MMP/MMP-25 in neutrophils. J. Biol. Chem. 276, 21960–21968 138 Matsuda, A. et al. (2003) Clusterin, an abundant serum factor, is a possible negative regulator of MT6-MMP/MMP-25 produced by neutrophils. J. Biol. Chem. 278, 36350–36357 139 Levesque, J.P. et al. (2004) Characterization of hematopoietic progenitor mobilization in protease-deficient mice. Blood 104, 65–72 140 Betsuyaku, T. et al. (1999) Neutrophil emigration in the lungs, peritoneum, and skin does not require gelatinase B. Am. J. Respir. Cell Mol. Biol. 20, 1303–1309 141 Balbin, M. et al. (2003) Loss of collagenase-2 confers increased skin tumor susceptibility to male mice. Nat. Genet. 35, 252–257 142 Pham, C.T. (2006) Neutrophil serine proteases: specific regulators of inflammation. Nat. Rev. Immunol. 6, 541–550 143 Weiss, S.J. (1989) Tissue destruction by neutrophils. N. Engl. J. Med. 320, 365–376 144 Gutierrez-Fernandez, A. et al. (2007) Increased inflammation delays wound healing in mice deficient in collagenase-2 (MMP-8). FASEB J. 21, 2580–2591 145 Pagano, M.B. et al. (2007) Critical role of dipeptidyl peptidase I in neutrophil recruitment during the development of experimental abdominal aortic aneurysms. Proc. Natl. Acad. Sci. U. S. A. 104, 2855–2860 146 Lin, M. et al. (2008) Matrix metalloproteinase-8 facilitates neutrophil migration through the corneal stromal matrix by collagen degradation and production of the chemotactic peptide Pro-GlyPro. Am. J. Pathol. 173, 144–153 147 Tester, A.M. et al. (2007) LPS responsiveness and neutrophil chemotaxis in vivo require PMN MMP-8 activity. PLoS One 2, e312 148 Wang, S. et al. (2006) Venular basement membranes contain specific matrix protein low expression regions that act as exit points for emigrating neutrophils. J. Exp. Med. 203, 1519–1532 149 Rabodzey, A. et al. (2008) Mechanical forces induced by the transendothelial migration of human neutrophils. Biophys. J. 95, 1428–1438

574

Trends in Cell Biology Vol.18 No.11 150 Saito, H. et al. (1998) Endothelial myosin light chain kinase regulates neutrophil migration across human umbilical vein endothelial cell monolayer. J. Immunol. 161, 1533–1540 151 Breslin, J.W. et al. (2006) Involvement of ROCK-mediated endothelial tension development in neutrophil-stimulated microvascular leakage. Am. J. Physiol. Heart Circ. Physiol. 290, H741–H750 152 Simpson, L.O. (1980) Basement membranes and biological thixotropy: a new hypothesis. Pathology 12, 377–389 153 Borza, D.B. et al. (2005) Goodpasture autoantibodies unmask cryptic epitopes by selectively dissociating autoantigen complexes lacking structural reinforcement: novel mechanisms for immune privilege and autoimmune pathogenesis. J. Biol. Chem. 280, 27147– 27154 154 Davison, P.F. (1989) The contribution of labile crosslinks to the tensile behavior of tendons. Connect. Tissue Res. 18, 293–305 155 Essex, D.W. et al. (2001) Protein disulfide isomerase and sulfhydryldependent pathways in platelet activation. Biochemistry 40, 6070– 6075 156 Cordier, P. et al. (2008) Self-healing and thermoreversible rubber from supramolecular assembly. Nature 451, 977–980 157 Abrahamson, D.R. and Perry, E.W. (1986) Evidence for splicing new basement membrane into old during glomerular development in newborn rat kidneys. J. Cell Biol. 103, 2489–2498 158 Hinck, L. and Silberstein, G.B. (2005) Key stages in mammary gland development: the mammary end bud as a motile organ. Breast Cancer Res. 7, 245–251 159 Williams, J.M. and Daniel, C.W. (1983) Mammary ductal elongation: differentiation of myoepithelium and basal lamina during branching morphogenesis. Dev. Biol. 97, 274–290 160 Yamada, K.M. and Cukierman, E. (2007) Modeling tissue morphogenesis and cancer in 3D. Cell 130, 601–610 161 Sabeh, F. et al. (2004) Tumor cell traffic through the extracellular matrix is controlled by the membrane-anchored collagenase MT1MMP. J. Cell Biol. 167, 769–781 162 Filippov, S. et al. (2005) MT1-matrix metalloproteinase directs arterial wall invasion and neointima formation by vascular smooth muscle cells. J. Exp. Med. 202, 663–671 163 Chun, T.H. et al. (2004) MT1-MMP-dependent neovessel formation within the confines of the three-dimensional extracellular matrix. J. Cell Biol. 167, 757–767 164 Hotary, K.B. et al. (2003) Membrane type I matrix metalloproteinase usurps tumor growth control imposed by the three-dimensional extracellular matrix. Cell 114, 33–45 165 Shi, J. et al. (2008) Membrane-type MMPs enable extracellular matrix permissiveness and mesenchymal cell proliferation during embryogenesis. Dev. Biol. 313, 196–209 166 Holmbeck, K. et al. (1999) MT1-MMP-deficient mice develop dwarfism, osteopenia, arthritis, and connective tissue disease due to inadequate collagen turnover. Cell 99, 81–92 167 Szabova, L. et al. (2005) Expression pattern of four membrane-type matrix metalloproteinases in the normal and diseased mouse mammary gland. J. Cell. Physiol. 205, 123–132 168 Coussens, L.M. et al. (2002) Matrix metalloproteinase inhibitors and cancer: trials and tribulations. Science 295, 2387–2392 169 Peterson, J.T. (2006) The importance of estimating the therapeutic index in the development of matrix metalloproteinase inhibitors. Cardiovasc. Res. 69, 677–687 170 Renkiewicz, R. et al. (2003) Broad-spectrum matrix metalloproteinase inhibitor marimastat-induced musculoskeletal side effects in rats. Arthritis Rheum. 48, 1742–1749 171 Vignery, A. (2005) Macrophage fusion: are somatic and cancer cells possible partners? Trends Cell Biol. 15, 188–193 172 Pawelek, J.M. (2005) Tumour-cell fusion as a source of myeloid traits in cancer. Lancet Oncol. 6, 988–993 173 Kinjo, M. (1978) Lodgement and extravasation of tumour cells in blood-borne metastasis: an electron microscope study. Br. J. Cancer 38, 293–301