Background
Cancer cell invasion of tissue matrices is a fundamental aspect of metastasis. Extracellular matrices (ECM) are generally considered to be of two types, basement membrane and stromal/interstitial. Basement membrane matrices are normally deposited beneath epithelia, and its components characteristically include collagen IV, laminin, perlecan and nidogen, which interact to form a thin, dense, cross-linked polymeric network with high tensile strength. Stromal/interstitial matrices form the majority of the body connective tissue and are composed primarily of fibrillar collagen I that is cross-linked into a stable meshwork to impart 3D structural support. As both basement membrane and stromal matrices present a steric barrier to cell transmigration, matrix remodelling is a necessary and critical contributor to metastasis. Tumour cells acquire the ability to surmount ECM barriers by expressing a range of proteases [
1], particularly members of the matrix metalloprotease (MMP) family [
2‐
4]. MMPs are vital for the degradation of both basement membrane and stromal matrices: the gelatinases MMP-2 and MMP-9, and transmembrane MMPs are critical mediators of basement membrane remodelling [
5,
6], whereas the cleavage of stromal fibrillar collagen I networks is limited to MMPs-1, -8, -13 and the transmembrane MMPs [
2].
In vitro assays are valuable for evaluating the potential role of candidate modulators of invasive behaviour, particularly in the present era of high throughput proteomic and genomic screens which are identifying large numbers of possible therapeutic targets. Cancer cell invasion is typically assessed
in vitro using the transwell Matrigel invasion assay. Matrigel, an extract derived from mice harbouring Engelbreth-Holm Swarm (EHS) tumours, is rich in laminin and collagen IV and is therefore used as a surrogate basement membrane for investigating a variety of cell behaviours, including cancer cell invasion [
5,
7]. For invasion assays, a thin layer of Matrigel is coated onto a porous membrane in Boyden or Transwell chambers and cell penetration is assessed. As such, the assay is considered to be a reliable and valuable test to evaluate cancer cell invasiveness [
5,
8‐
11]. In an assay similar to the Matrigel chemoinvasion assay, transwell membranes can be coated with collagen I to reflect cellular invasion through the confines of stromal/interstitial matrices.
For cancers such as ovarian, gastric and colon, which metastasise within the peritoneal cavity, it is paramount that the
in vitro models adequately reflect the processes that occur during peritoneal dissemination. Epithelial ovarian cancers (EOC) are the most deadly of the gynaecological cancers and are the fifth leading cause of cancer-related deaths in North American women [
12]. The majority of EOCs metastasize locally in a manner that does not involve haematological transport. Ovarian tumour cells exfoliate and are carried via peritoneal fluid to secondary sites in the abdominal cavity where their attachment, invasion of the submesothelial connective tissue, and proliferation form peritoneal deposits. An inflammatory response typically accompanies disease progression and alters the peritoneal membrane in a manner that renders it prone to cancer cell adhesion [
13,
14]. This further facilitates tumour dissemination such that a self-promoting vicious cycle of metastasis ensues and inevitably leads to impaired functioning of abdominal organs: the obstruction and malfunctioning of the gastrointestinal tract are a frequent cause of morbidity from ovarian cancer [
15,
16]. Devising effective strategies to prevent further metastatic spread is instrumental for improving survival of patients diagnosed with advanced disease. Important cell behaviours that contribute to ovarian cancer disease progression include adhesion (cell-cell and cell-matrix), migration, and invasion [
17].
The surfaces of the peritoneal cavity are covered by a layer of mesothelial cells that function in an antiadhesive manner to promote gliding of the abdominal viscera [
18]. The mesothelial layer also protects against cancer cell attachment [
19], as it conceals the underlying connective tissue matrix to which tumour cells preferentially attach [
19,
20]. Collagen I is present in abundance beneath the peritoneal mesothelium [
21]. The collagen-binding integrins α2β1 and α3β1 mediate
in vivo peritoneal metastasis of gastric tumour cells [
22,
23], and their importance in this process has been inferred for ovarian cancer cells [
24‐
27]. Furthermore, collagen I is the preferred substrate for ovarian cancer cell attachment [
28], and stimulation of motile [
25,
29] and invasive [
24] behaviour.
In these studies we compared the performance of Matrigel and collagen I substrata as in vitro invasion matrices, both in 2D (planar) and 3D contexts. In particular, we sought to determine whether cell penetration of these matrices required MMP activity, reflecting the mechanisms needed for cancer cell invasion in vivo. We show that in contrast to the invasion of collagen I matrices, MMP-mediated proteolysis is not required for cell penetration of Matrigel for ovarian cancer cells in either 2D transwell or 3D spheroid cell invasion systems. This contrasts with the requirement of MMP activity for the invasion of basement membranes in vivo, indicating the limitations of Matrigel for the evaluation of cancer cell invasion.
Discussion
Establishing
in vitro invasion assays that accurately reflect the circumstances
in vivo is paramount to revealing mechanisms relevant to cancer metastasis. During peritoneal metastasis, cancer cells preferentially adhere to the submesothelial ECM rather than to the mesothelial cells [
19,
20]. The submesothelial connective tissue is periodically exposed at milky spots (lymphatic portals), which are prevalent on the surface of the omentum and the subdiaphragmatic peritoneum. In the early stages of peritoneal dissemination, cancer cells preferentially adhere to the milky spots [
38] likely due to exposure of the underlying ECM at these sites [
20,
39]. Rather than effectively eliminating tumour cells, immune cells, including those at the milky spots, may promote tumour dissemination through their release of inflammatory mediators [
40]. With cancer progression, the mesothelium becomes compromised [
41] due to the action of inflammatory cytokines, including tumour necrosis factor α and interleukin 1β, which cause mesothelial cell retraction [
20,
42]. The widespread exposure of the underlying ECM causes a corresponding shift in the pattern of tumour cell attachment such that it is no longer limited to peritoneal surfaces with milky spots [
20]. Therefore investigating cancer cell interaction with, and invasion of, the submesothelial ECM is vital. Unlike the archetypal basement membrane underlying true epithelia, consisting of a distinct collagen IV and laminin-rich sheet, the submesothelial ECM is not a typical basement membrane. [
21]. Instead, collagen I and FN abut the mesothelial monolayers, co-localizing with thin deposits of collagen IV and laminin [
21].
Based on its composition, Matrigel is believed to resemble basement membrane matrices, such as that of the endothelium, which cancer cells penetrate during haematological metastasis frequently used for
in vitro invasion studies [
9,
10]. The attraction of the Matrigel invasion assay is that it provides a rapid, simple method to evaluate invasion within hours [
5] as compared to the multiple days required for tumour cell invasion of intact basement membranes isolated from tissues. Despite the unique metastatic process and critical role of collagen I in peritoneal metastasis, the
in vitro invasion of ovarian, gastric and colon cancer cells is also routinely assessed using Matrigel matrices. However, collagen I is comparable to Matrigel in terms of being commercially available and easy to prepare as a transwell coating or thick (3D) gel. Collectively, our experiments show that matrices reconstituted from Matrigel do not adequately reflect the barrier function of basement membranes, in either transwell or 3D invasion systems, whereas the barrier function of stromal matrices can be mimicked using polymerised collagen I. We show that ovarian cancer cells can migrate through Matrigel matrices, in the absence of MMP activity, in either thin 2D transwell or thick 3D spheroid invasion contexts. In contrast, ovarian cancer cell penetration through a collagen I barrier required MMP activity.
Matrigel differs from authentic basement membranes in terms of the relative abundance of and interactions between the components [
43], and this may underly its inability to mimic the barrier function of intact basement membranes. Collagen IV is essential for basement membrane tensile strength and stability, forming a cross-linked network with which the laminin network interacts. Whereas the components of Matrigel are chemically and immunologically similar to the major components of basement membranes [
44], the relative abundance of and interactions between the components differ. Most notably, Matrigel matrices are substantially less cross-linked than basement membranes [
43]. As it is the cross-linked structure that imparts strength and integrity to intact basement membrane matrices, it follows that Matrigel matrices would have lower resistance to cell penetration and be unable to reflect the in vivo situation.
Collagen I is arguably the most important ECM component with which ovarian cancer cells interact during peritoneal dissemination. It is the preferred substrate for adhesion and migration of ovarian cancer cells and also stimulates their invasive behaviour. An abundant constituent of the peritoneal stromal matrix that is present directly beneath the mesothelium, collagen I is exposed to the peritoneal cavity at milky spots and as a result of mesothelial retraction. The importance of tumour cell interaction with collagen I in peritoneal metastasis is supported by the fundamental role of its receptors α2β1 integrin and α3β1 integrin [
23,
45,
46]. Altogether, this provides strong rationale for the use of collagen I matrices in investigations pertaining to invasive behaviour by ovarian and other cancers that undergo peritoneal metastasis.
In 3D culture, MMP-mediated proteolysis was required for invasion of matrices formed from acid-extracted collagen I but not of matrices formed from pepsin-digested collagen I (Vitrogen/Purecol). Differences in the requirement for MMP activity for ovarian cancer cells to invade collagen I matrices formed from acid-extracted rat tail collagen versus pepsin-solubilized collagen may be attributed to differences in the structural integrity of the reconstituted collagen. Pepsin cleaves within the telopeptide regions of collagen I molecules (creating atelocollagen), whereas these domains remain intact when acid-extraction is used for solubilization. In addition to promoting collagen I assembly [
47], intact telopeptide regions are critical to the strength and stability of fibrillar collagen I matrices through the provision of lysine residues required for intermolecular cross-link formation [
36]. These covalent cross-links prevent collagen molecules from sliding past one another and are the basis of the tensile strength of collagen fibre systems. The collagen cross-links re-establish during reconstitution of the acid-extracted collagen I [
37], but not pepsin-digested collagen I (atelocollagen) preparations. Consequently, it is conceivable that cancer cells should be able to migrate through the more compliant atelocollagen matrices without the requirement for its degradation. Interestingly, although MMP activity was not required for cell penetration of 3D gels of atelocollagen, it was required when the atelocollagen was dried onto transwells. This affixing of collagen to the transwell membrane may compensate for the lack of cross-linking, effectively preventing fibrils from sliding past one another as would occur in a 3D gel.
The barrier function of Matrigel was implied but not verified in the original description of the Matrigel chemoinvasion assay [
10]. However, cell penetration of Matrigel does not always correlate with invasiveness. Studies have shown that normal fibroblasts are capable of passing through Matrigel, whereas many invasive epithelial cancer cell lines are not, and correlation between invasive potential and capacity for Matrigel penetration was found to be lacking [
48‐
50]. Rather, cells of mesenchymal lineage had a superior capacity for Matrigel penetration compared to those of epithelial origin, regardless of whether they were malignant or had invasive potential
in vivo. Moreover, extensive migration of the cells into Matrigel occurred in the absence of matrix degradation [
48], which is consistent with our findings. Therefore, the use of Matrigel as an invasion matrix is inconsistent with the current dogma that MMPs are important for cancer cell invasion. Conversely, we have found that the
in vitro collagen I invasion capacity of a panel of ovarian cancer cell lines reflected the
in vivo capacity for peritoneal tumour formation reported by others [
31].
The function of MMPs in matrix proteolysis is well established, yet this family of proteases also has critical roles in various additional physiological functions that include the cleavage of cell adhesion molecules and growth factors [
51]. The results obtained for HOC-7 cells emphasize that MMP activity can influence Matrigel transmigration (in a cell line specific manner) for reasons unrelated to alleviation of a matrix barrier. The apparent marked GM6001-mediated inhibition of HOC-7 Matrigel "invasion" was paralleled by a comparable reduction in migration through uncoated transwells, yet HOC-7 scratch wound migration was unaffected by MMP inhibition. It is plausible that MMP inhibition blocks the cell-cell detachment required for these cells to migrate through the narrow (8 μm) transwell pores because in contrast to the fibroblast-like HEY and ES-2 cells, the HOC-7 cells have an epithelial morphology, grow in tight clusters, express E-cadherin [
31], and migrate as a sheet. This interpretation is supported by evidence that E-cadherin cleavage is mediated by MMPs [
1,
32,
33].
Matrigel contains numerous growth factors (TGFβ, EGF, FGF, PDGF, IGF [
7]) that may be activated or released from the matrix by proteolytic cleavage, including by some MMPs. The chemotactic response to various agents differs between cells [
52], and could contribute to observed differences in sensitivity to MMP inhibitors between cell lines (eg. HEY versus ES-2 in this study). This might explain why MMP inhibition has had, at best, only a modest influence on Matrigel penetration as compared to collagen I invasion.
The importance of considering auxiliary factors involved in the transmigration process that are unrelated to matrix penetration has been emphasized [
52], yet most studies evaluate treatment effects for Matrigel penetration in the absence of migration controls. Thus, without careful examination, the requirement of MMPs for cell dissociation/migration in some cancer cells lines may have been misinterpreted as a requirement for proteolytic degradation of matrices. In exceptional studies where migration control experiments have been performed, treatments tended to affect Matrigel "invasion" and migration similarly (for example [
53‐
55]). Thus, the impact on Matrigel penetration should be considered secondary to an altered motility unless alterations in matrix proteolysis are demonstrable.
Our results do not preclude the possibility that cells utilize alternative proteolytic systems for Matrigel penetration. For example, cathepsins have been implicated in transwell Matrigel penetration [
54,
56,
57] by cancer cells, including ovarian [
58]. However, marked reductions in cell motility have also been documented in response to cathepsin protease inhibition [
54]. Therefore, it is unclear whether Matrigel degradation by cathepsin proteases is a requirement for cell penetration
per se, or whether the observed reductions in Matrigel penetration merely reflect reduced cell motility. Although other proteolytic systems may influence invasion, it is widely accepted that MMP activity is a critical aspect of tumour penetration of basement membranes
in vivo.
Recent studies suggest cancer cells can circumvent the requirement for matrix proteolysis and migrate through tissues by adopting an amoeboid form of movement [
59‐
62]. It is notable that these studies were performed in 3D atelocollagen or Matrigel gels, not in acid-extracted collagen I. Whereas channels lined with collagen I degradation products were generated during cancer cell invasion of 3D acid-extracted collagen I matrices, such channels were not observed when pepsin-digested collagen was used [
63]. Thus, the phenomenon of protease-independent tumour cell invasion may be limited to those matrices with low levels of cross-linking.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
KLS performed all studies and drafted the manuscript. TJB performed statistical analysis. TJB and MJR revised the manuscript. All authors were involved in the conception of the study and data interpretation. All authors have read and approved the final manuscript.