Tumor Angiogenesis: MMP-Mediated Induction of Intravasation- and Metastasis-Sustaining Neovasculature

Abstract

Metastasis is a distinct stage of cancer progression that requires the development of angiogenic blood vessels serving as conduits for tumor cell dissemination. An accumulated body of evidence indicates that metastasis-supporting neovasculature should possess certain structural characteristics allowing for the process of tumor cell intravasation, an active entry of cancer cells into the vessel interior. It appears that the development of tumor vessels with lumens of a distinctive size and their structure supported by a discontinuous pericyte coverage, together constitute critical microarchitectural requirements to: (a) provide accessible points for vessel wall penetration by primary tumor cells; (b) provide enough lumen space for a tumor cell or cell aggregate upon intravasation; and (c) allow for sufficient rate of blood flow to carry away intravasated cells from the primary tumor to the next, proximal or distal site. This review will primarily focus on the functional roles of matrix metalloproteinases (MMPs), which catalytically trigger the development of an intravasation-sustaining neovasculature at the early stages of tumor growth and are also required for the maintenance of a metastasis-supporting state of blood vessels at later stages of cancer progression.

1. Introduction

The establishment and progression of metastases is a complex multi-step process involving dynamic interactions between cancer cells and their microenvironment (Egeblad et al., 2010; Hanahan and Weinberg, 2011; Valastyan and Weinberg, 2011). Within the primary tumor, these reciprocal interactions initiate angiogenic vessel assembly and culminate in the formation of vascular networks contributing to both tumor growth and cancer cell spread to secondary sites (Weis and Cheresh, 2011). This metastasis-supporting vasculature is shaped in part through continuous proteolytic modifications of the tissue extracellular matrix (ECM) and various cell surface molecules, particularly by the matrix metalloproteinases (MMPs).

The notion that tumor metastasis occurs “due to the conveyance of cells from the primary tumor to some other part of the body, by the blood or lymphatic vessels, where they develop into similar growths” has been acknowledged as early as 1897 (Lovett and Councilman, 1897). However, it took almost 75 years to suggest that cancer progression could be halted by therapeutic controlling of tumor angiogenesis (Folkman, 1971), and another 15 years to conceptualize that the induction of tumor angiogenesis is a critical step in the early development of a solid tumor required for its transition from avascular premalignant to vascular neoplasic stage (Folkman et al., 1989). By 2000, the induction of tumor angiogenesis by an “angiogenic switch” has become regarded as a hallmark of cancer (Hanahan and Folkman, 1996; Hanahan and Weinberg, 2000). During that time it has been also established that this angiogenic switch requires a critical angiogenic factor, VEGF, and its proteolytic release from tumor matrix by MMPs, predominantly by the MMP-9 delivered into the tumor microenvironment by tumor-infiltrating leukocytes (Bergers et al., 2000). Nevertheless, the precise mechanisms whereby MMP-9 and other MMPs induce the development and also sustain those distinct angiogenic vessels capable of supporting tumor cell intravasation and metastatic dissemination of aggressive cancer cells, are still in the spotlight of cancer research.

Although functionally MMPs have been linked to tissue neovascularization in the early 1990s and multifaceted roles of MMPs in tumor angiogenesis since then are regarded as well established, the literature on how MMPs are mechanistically involved in angiogenic vessel development and angiogenesis-dependent metastasis is still expanding. Thus, when limited to “MMP/Tumor Angiogenesis/Metastasis” as key word criteria, more than 6,300 publications are indicated in the PubMed database since 2011 to the present versus approximately 3,000 publications in the whole decade from 2001 to 2010 and only 30 publications within the 5 years from 1991 to 1995. However, the majority of these “key-words-filtered” MMP publications is centered either on general involvement of MMPs in the neovascularization process or on various aspects of MMP-mediated activation and migration of endothelial cells and not on the unique roles of MMPs in angiogenesis-dependent metastasis. Therefore, we will review almost exclusively the original in vivo studies, which illuminate specific mechanisms underlying the MMP-mediated induction and development of a tumor angiogenic vasculature that is functionally and structurally capable of sustaining intravasation and dissemination of cancer cells.

2. Tumor angiogenic vasculature and tumor cell intravasation

2.1. Dysfunctionality of angiogenic vasculature versus its sustainability to support tumor cell intravasation and metastasis

According to a generally-accepted notion, angiogenic vessels developing in the primary tumor are structurally abnormal and functionally immature (Carmeliet and Jain, 2011). The tumor vasculature is described as chaotic and torturous, irregular in lumen diameters, dilated and highly permeable, deficient in pericyte coverage and abnormal in endothelial lining (De Bock et al., 2011; Garcia-Roman and Zentella-Dehesa, 2013; Goel et al., 2011). Nevertheless, this seemingly impaired vasculature is functional enough to provide not only the nutrients for a growing tumor, but also the conduits for metastatic dissemination of the escaping tumor cells. Whereas the leakiness and enhanced vessel permeability are compatible with the ability of angiogenic vasculature to supply nutrients to the primary tumor, the overall immature state and dysfunctionality of angiogenic vessels appear to be at odds with the capacity of the same vascular networks to sustain active intravasation of tumor cells and their dissemination to secondary sites. Exemplifying such apparent contradiction, the diminishment of pericyte recruitment to primary tumors developing in mice genetically devoid of MMP-9, a critical angiogenic enzyme, has been associated with collapsed morphology of tumor vasculature and substantially inhibited metastasis (Chantrain et al., 2006). Therefore, the pericyte-mediated vessel stabilization (Raza et al., 2010) and architectural support are essential for the functionality of tumor angiogenic vasculature and its ability to sustain tumor cell dissemination.

2.2. Tumor cell intravasation is supported by lumen-containing angiogenic vasculature

While establishing molecular pathways whereby MMPs regulate the process of tumor cell dissemination, we have noticed that high levels of tumor cell intravasation and metastasis frequently correlated with the development and enhanced density within primary tumors of lumen-containing, perfusable blood vessels. The issue of lumen-containing vessels became an important characteristic of intratumoral vasculature, particularly in the settings where tumor cell intravasation and metastasis were significantly diminished, but the overall density of microvessels or total volume of vasculature, two parameters frequently employed in cancer studies, were not affected (Bekes et al., 2011b; Juncker-Jensen et al., 2013). From a mechanistic point, lumen space and blood flow represent obvious requirements for an intravasating tumor cell to enter the angiogenic vessel and be carried away to a secondary site. Therefore, the development of a network of anastomosed angiogenic vessels containing circulating blood and possessing a certain level of blood pressure would likely be prerequisites for the process of metastatic dissemination. In this regard, the vessel-covering pericytes would structurally support the newly-developing tumor vessels and prevent their collapse, allowing for both adequate lumen space to accommodate the size and volume of an intravasating tumor cell or tumor cell aggregates, and sufficient blood flow to carry away the intravasated cells. Reflecting functional importance of these vascular characteristics, the improved pericyte coverage and blood flow are among the “normalization” responses of tumor-associated vasculature to antiangiogenic therapies (Carmeliet and Jain, 2011; Goel et al., 2011; Jain, 2005), which can lead to accelerated tumor progression and enhanced distant metastasis (Ebos et al., 2009; Loges et al., 2009; Paez-Ribes et al., 2009; Sennino et al., 2012; Stockmann et al., 2008).

3. Functional interplay of MMPs and VEGF in the induction of angiogenic vessels and their maintenance during tumor development

3.1. VEGF as a critical molecule governing the development and microarchitecture of angiogenic vessel networks

Numerous molecular pathways and systems have directly or indirectly been implicated in the induction of angiogenesis at the early stages of tumor development and in the maintenance of metastasis-supporting vascular networks during late stages of cancer progression. However, the VEGF-A (VEGF) molecule represents a critical factor that regulates practically all aspects of tumor-induced angiogenesis, including endothelial cell sprouting and assembly, lumen formation, vessel dilation and permeability, and the microarchitecture and patterning of vascular networks (Chung and Ferrara, 2011; Ellis and Hicklin, 2008; Ferrara et al., 2003; Goel and Mercurio, 2013; Welti et al., 2013). Within the tumor microenvironment, the cancer cells appear to serve as the major source of angiogenesis-inducing VEGF in vivo (Hoeben et al., 2004), although different types of stromal cells, including cancer-activated fibroblasts (De Francesco et al., 2013; Ito et al., 2007) and alternatively activated dendritic cells (Riboldi et al., 2005) and infiltrating leukocytes such as neutrophils (Jablonska et al., 2010; Scapini et al., 2000; Scapini et al., 2004; Schruefer et al., 2005), macrophages (Coffelt et al., 2010b; Kiriakidis et al., 2003) and T lymphocytes (Owen et al., 2003), can also supply VEGF.

Complex roles of VEGF in tumor angiogenesis and, by implication, in angiogenesis-dependent metastasis involve both positive and negative regulations of blood vessel development by the VEGF molecule. Whereas VEGF levels within the tumor environment appear to directly correlate with the overall microvessel density (Huss et al., 2001; Takahashi et al., 1998; Takahashi et al., 1995), VEGF also disrupts interactions between vascular pericytes and activated endothelial cells, causing incomplete pericyte coverage of angiogenic vessels (Greenberg et al., 2008). In human xenograft-mouse models, specific inhibition or trapping of VEGF produced by cancer cells results in significant diminishment of tumor angiogenesis, concomitant with reduced tumor growth and, consequently, inhibited metastasis (Byrne et al., 2003; Crawford and Ferrara, 2008; Huang et al., 2003; Kanai et al., 1998; Wang et al., 2008; Warren et al., 1995). However, in syngeneic cancer models, specific depletion of VEGF in tumor-infiltrating myeloid cells can dramatically accelerate tumor progression and normalize pericyte coverage of tumor-associated vasculature (Stockmann et al., 2008). Furthermore, the lack of substantial benefits of anti-VEGF therapies attributed to resistance, adaptation and normalization of tumor vasculature in cancer patients (Bergers and Hanahan, 2008; Claes et al., 2008; Goel et al., 2011; Mancuso et al., 2006; Shojaei et al., 2007; Welti et al., 2013), has clearly indicated that the functional activities of VEGF molecule are more broad and complex than its direct effects on the endothelium or indirect influencing of vascular pericytes and might overlap with multiple molecular networks and processes (Dawson et al., 2009; Garmy-Susini et al., 2010; Mazzone et al., 2009; Mitra and Schlaepfer, 2006; O’Connell et al., 2011).

3.2. Functional links between VEGF and MMPs during tumor angiogenesis

Specific molecular pathways governing VEGF-mediated tumor angiogenesis involve cross-talk and complex links between VEGF and MMP systems. Moreover, VEGF and certain MMPs can be regulated in a mutually coordinated manner on a transcriptional level. Thus, expression of both VEGF and MMP-9 can be induced simultaneously by a common master regulator, a hypoxia-inducible factor HIF-1 (Hoeben et al., 2004; Liao and Johnson, 2007). Counterbalancing such a mutual induction of VEGF and MMP-9 expression, secreted protein acidic and rich in cysteine (SPARC), was shown to negatively and concurrently regulate both VEGF and MMP-9 in a medulloblastoma model, resulting in development of smaller tumors with fewer blood vessels (Bhoopathi et al., 2010). In a similar manner, endogenous SPARC was shown to inhibit expression of both VEGF and MMP-7 in gastric cancer cells and therefore, SPARC silencing elevated angiogenesis, whereas SPARC overexpression diminished angiogenesis in gastric tumor xenografts in an MMP-7/VEGF-dependent manner (Zhang et al., 2012). Interestingly, protein expression of VEGF was correlated with the expression of MMP-2 and MMP-9, and expressions of both MMPs and VEGF were closely linked to metastasis and angiogenesis in gastric cancer patients (Zheng et al., 2006).

In addition to mutual regulation, MMPs and VEGF were shown to exert seemingly unilateral effects on each other’s expression and production. Tumor-produced VEGF can regulate production of distinct MMPs in stromal cells and in activated endothelial cells. Thus, VEGF produced by ovarian cancer cells enhances the expression of host MMP-9 in the ovaries of xenograft-bearing mice, in part due to the induction of neutrophil influx and delivery of neutrophil MMP-9 (Belotti et al., 2008). Furthermore, VEGF produced by distant primary tumors facilitates lung metastasis through induction of MMP-9 production in VEGF receptor (VEGFR)-positive endothelial cells and macrophages involved in the formation of premetastatic niches (Hiratsuka et al., 2002). On the other hand, via STAT activation and VEGFR2 signaling, VEGF was shown to significantly reduce MMP-9 production in B-cell leukemia cells (Ugarte-Berzal et al., 2010), indicating the complexity of VEGF-mediated regulation of MMP-9 expression.

While VEGF is involved in regulation of some MMPs and MMP-9 in particular, certain MMPs of tumor and stromal cell origin can regulate VEGF expression. Thus, the overexpression of MMP-14 in cancer cells was shown to increase VEGF production and angiogenesis in glioblastomas (Deryugina et al., 2002) and breast carcinomas (Sounni et al., 2002; Sounni et al., 2004). MMP-2 also was shown to induce VEGF expression in A549 cells through the binding to αvβ3 and subsequent integrin signaling (Chetty et al., 2010). Furthermore, cancer cell-produced MMP-13 (collagenase-3) has recently been shown to promote secretion of VEGF by fibroblasts and endothelial cells and to induce tumor angiogenesis in vivo (Kudo et al., 2012). MMP-mediated regulation of VEGF-induced tumor vascularization was demonstrated in a xenograft model in vivo, in which MMP blockade almost completely inhibited VEGF production and significantly reduced the volume of angiogenic vasculature (Woenne et al., 2010). Therefore, in multiple ways, MMPs directly and indirectly can influence the VEGF-mediated development of an angiogenic vasculature.

3.3. MMP-mediated release of matrix-sequestered VEGF in vivo

Since the majority of produced VEGF is sequestered in the ECM deposited by tumor and stromal cells, the proteolytic release of angiogenic factors from tissue matrix is regarded as a prerequisite for in vivo induced angiogenesis. Different isoforms of VEGF have been shown to bind with different affinities to heparin-binding matrix glycoproteins, with a trend that high mol. wt. forms are more tightly bound than low mol. wt. forms and the most common 165 kDa isoform, VEGF1₆₅, exhibiting an intermediate level of binding (Houck et al., 1992). In vitro, a number of recombinant MMPs, including MMP-1, 3,-7, −9, −16 and −19, were shown to cleave full-length recombinant VEGF1₆₅, generating lower mol. wt. fragments (Lee et al., 2005). Then, recombinantly-produced VEGF fragments mimicking in molecular weight those generated by MMP-3 cleavage in vitro, were demonstrated to induce either dilation or branching of angiogenic vessels in vivo, prompting a notion that MMPs release matrix-bound VEGF by direct intramolecular processing (Lee et al., 2005). However, MMP-dependent release of specific soluble fragments from VEGF that was actually sequestered in tumor matrix was not demonstrated in this study, neither for in vitro nor for in vivo conditions. Furthermore, the full-length VEGF1₆₅ was shown to be completely resistant to several MMPs, including MMP-1,-3,-7,-9, and −13 (Hashimoto et al., 2002; Hawinkels et al., 2008), therefore not supporting the intramolecular cleavage as a mechanism of VEGF release.

In contrast, the release of VEGF via MMP-mediated digestion of matrix proteins sequestering VEGF molecules is supported by several biochemical studies. In vitro, matrix-bound VEGF1₆₅, actually embedded into the ECM deposited by tumor cells, was shown to be releasable by recombinant MMP-2 and MMP-9 (Ebrahem et al., 2010; Hawinkels et al., 2008). In addition, MMP-3 (stromelysin-1) and MMP-7 (matrilysin) were capable of releasing in vitro the intact VEGF₁₆₅ from its complex with connective tissue growth factor (CTGF), and this release was accompanied by digesting CTGF without any generation of low mol. wt. VEGF fragments (Hashimoto et al., 2002). Furthermore, quantitative modeling of proteolytic release of matrix-sequestered VEGF not only indicated that soluble VEGF was essentially insensitive to the cell-mediated proteolysis in vitro, but also suggested that VEGF₁₆₅ conversion to lower mol. wt. form would not occur in vivo (Vempati et al., 2010). In agreement with the mechanism, whereby VEGF would be released from the matrix as intact molecule and not cleaved fragments, the full-length VEGF₁₆₅ was shown in vivo to regulate different aspects of tumor angiogenesis depending on the concentration: when used at relatively low concentrations, VEGF₁₆₅ increased the density of angiogenic vessels, but at a 3-fold higher concentrations, the same VEGF₁₆₅ phenotypically switched normal-looking vessels to abnormal-dilated vessels typical of tumor vasculature (Parsons-Wingerter et al., 2006). Therefore, the proteolytic digestion of matrix proteins and proteoglycans by serine proteases (Houck et al., 1992) or MMPs (Hawinkels et al., 2008) appears to constitute a mechanism that is substantially supported by experimental data explaining how matrix-sequestered VEGF becomes bioavailable in vivo

MMP-mediated release of VEGF capable of inducing development of angiogenic vasculature was shown in several in vivo angiogenesis models. Importantly, VEGF-induced angiogenesis was not correlated with the levels of total VEGF present in tissue matrix, but specifically with the levels of soluble VEGF extracted from the matrix without detergents. Thus, corneal neovascularization in a micropocket model that requires the release of exogenously supplied VEGF, was stimulated by MMP-2 and MMP-9 in a VEGF-dependent manner (Ebrahem et al., 2010). Further indicating an importance of MMPs in VEGF-dependent angiogenesis, the MMP-9 produced by ovarian cancer cells released biologically active VEGF that in turn induced endothelial cell motility, enhanced endothelium permeability, and facilitated ascites formation (Belotti et al., 2003). In breast cancer xenografts engineered to express the stromal-bound form of MMP-9, the released VEGF promoted tumor angiogenesis via enhanced complexing of VEGF with endothelial cell VEGFR, the association that was dependent on MMP-9 activity (Mira et al., 2004). Stromal MMP-13 expressed by cancer-associated fibroblasts was shown to be essential for VEGF release and VEGF-dependent maintenance of angiogenesis in a skin squamous cell carcinoma model (Lederle et al., 2010).

However, as it will be shown below, the functional contribution of MMP-executed release of VEGF that triggers the angiogenic switch in vivo has been attributed almost exclusively to the proteolytic activity of host MMP-9 delivered mainly by tumor-infiltrating leukocytes.

4. Tumor-associated leukocytes and their MMP-9 in VEGF-induced angiogenesis

4.1. The role of inflammatory leukocytes in the generation of angiogenesis-inducing tumor microenvironment

The induction of tumor angiogenesis by tumor-infiltrating leukocytes involves a multi-step sequence of events, originating at the level of dysregulated tumor cells expressing and secreting various proangiogenic molecules and inflammatory cytokines, followed by a specific response of activated endothelial cells to assemble into angiogenic blood vessels, and culminating in the generation of a functional angiogenic vasculature partially covered with pericytes. The inflammatory cell influx into the tumor microenvironment is regulated by production of distinct leukocyte chemoattractants for neutrophils, circulating monocytes and monocytic precursors of macrophages, and different types of lymphocytes (Egeblad et al., 2010; Erez and Coussens, 2011; Grivennikov et al., 2010; Quail and Joyce, 2013). The relative production of lineage-specific cytokines can be tumor type-specific and/or tumor stage-specific, which would determine the preferential type of leukocytes found in a given tumor infiltrate (Coffelt et al., 2010a; Curry et al., 2014; Dumitru et al., 2012; Galdiero et al., 2013; Tazzyman et al., 2009; Tazzyman et al., 2013). In addition, different types of inflammatory leukocytes could compensate for each other and adopt angiogenesis-inducing functions of the absent leukocyte partner (Pahler et al., 2008; Sawanobori et al., 2008; Tan et al., 2013). Furthermore, depending on specific localization in the tumor area, tumor-associated leucocytes, e.g. macrophages and neutrophils, were shown to acquire either protumoral and proangiogenic characteristics or anti-tumorigenic traits (Casazza et al., 2013; Fridlender et al., 2009; Galdiero et al., 2013; Lawrence and Natoli, 2011; Mantovani et al., 2011; Piccard et al., 2011; Qian and Pollard, 2010) and to exhibit unique gene expression profiles and functions across different cancer types (Dalton et al., 2014; Elpek et al., 2014; Qian et al., 2009; Smith and Kang, 2013).

Tumor-associated leukocytes were credited with the expression of various MMPs, including MMP-1, −2, −3, −9, −10, −11, −14, −15, −16, −17, −19, −24, −26, −27, −28, mainly through the studies which demonstrated the corresponding MMP mRNA transcripts in normal leukocytes or immortal cell lines representing particular leukocyte types (Bar-Or et al., 2003; Sithu et al., 2007). Based mostly on immunohistochemical stainings, different populations of tumor-associated leukocytes have been liberally ascribed with the ability to produce and deliver angiogenesis-inducing MMP-9. However, the actual production on a protein level or functional importance in tumor development and angiogenesis was not demonstrated for the majority of these leukocyte-associated MMPs, with the exception of MMP-9. As described below, elegant experimental approaches applied to spontaneous metastasis models allowed to prove that MMP-9 produced by inflammatory leukocytes is a crucial molecule regulating tumor angiogenesis and angiogenesis-dependent tumor cell dissemination.

4.2. Tumor-infiltrating leukocytes deliver angiogenesis-inducing MMP-9

Genetic ablation of MMP-9 in tumor recipients, resulting in reduced microvessel density in developing tumors and even preventing the angiogenic switch during cancer progression, provided original evidence for the functional involvement of host MMP-9 in tumor angiogenesis (Bergers et al., 2000). Nevertheless, it was the reconstitution of tumor-bearing Mmp9-knockout (KO) mice by transplantation of wild-type, MMP-9-competent hematopoietic cells that provided a direct proof that tumor angiogenesis-inducing MMP-9 was delivered by tumor-infiltrating myeloid cells (Bergers et al., 2000; Chantrain et al., 2006; Jodele et al., 2005). Depending on the prevalence of a specific type of MMP-9-positive leukocytes in a particular tumor infiltrate, different tumor-associated leukocytes have been ascribed with the capacity of MMP-9 delivery. In general, the “reappearance” of MMP-9-positive tumor-associated macrophages (TAMs) concomitant with the “restoration” of tumor angiogenesis and metastasis in genetically-deficient tumor bearers transplanted with wild-type hematopoietic cells led to a prevailing conclusion that TAMs are critical for MMP-9-mediated tumor angiogenesis (Huang et al., 2002), even though other types of MMP-9-positive myeloid cells were not rigorously screened or investigated. Thus, when tumors genetically lacking TAMs but manifesting high levels of MMP-9 were searched for the cell source of host MMP-9, an important compensatory mechanism evoking neutrophil infiltration was established (Pahler et al., 2008), highlighting the functional difference between the MMP-9-expressing macrophages versus MMP-9-delivering tumor-infiltrating neutrophils. Further complicating the mechanisms underlying the functions of leukocyte-produced MMP-9 in tumor angiogenesis, genetic ablation of MMP-9 not only decreased microvessel density in ovarian carcinomas in vivo, but also decreased macrophage infiltration into primary tumors (Huang et al., 2002). Similar dependence of infiltrating macrophages on MMP-9 and its proteolytic activity is manifested in plasminogen-knockout mice, where supplementation of active MMP-9 was shown to restore impaired macrophage influx (Gong et al., 2008). In support of the notion that intratumoral MMP-9 can locally regulate influx and presentation of leukocytes in developing tumors, it was demonstrated that the adenovirus-mediated induction of MMP-9 in breast cancer xenografts dramatically altered their cytokine profile and enhanced production of neutrophil chemoattractants, KC and MIP-2, leading to a massive neutrophil infiltration (Leifler et al., 2013).

Pharmacological or genetic depletion of a specific hematopoietic lineage in tumor-bearing mice allowed for more precise attributing of MMP-9 delivery and MMP-9-dependent angiogenesis to a particular category of MMP-9-positive tumor-associated leukocytes. By using mast cell-deficient transgenic mice, the infiltration by mast cells delivering MMP-9-activating proteases was linked to the MMP-9-mediated angiogenic switch and resulting tumor angiogenesis in the skin carcinogenesis model (Coussens et al., 1999). Specific ablation of MMP-9-positive TAMs with zoledronic acid resulted in reduced tumor angiogenesis, leading to a conclusion that TAMs deliver angiogenesis-inducing MMP-9 (Giraudo et al., 2004; Rogers and Holen, 2011). It appears that tumor infiltration by inflammatory leukocytes belonging to a particular hematopoietic lineage can depend on specific tumor microenvironments, which also can change with the stage of tumor development.

Several independent studies repeatedly demonstrated that the elimination of MMP-9-bearing neutrophils with a granulocyte-specific Gr-1 antibody (Fleming et al., 1993) or prevention of neutrophil infiltration with anti-inflammatory agents consistently caused inhibition of tissue neovascularization (Shaw et al., 2003) and tumor angiogenesis concomitant with a significant reduction of intravasation and metastasis (Bekes et al., 2011b; Jablonska et al., 2010; Pekarek et al., 1995; Tazawa et al., 2003). Neutrophil-depletion studies also allowed to link tumor-infiltrating neutrophils specifically with the initial angiogenic switch in the genetic multistage model of pancreatic β-cell cancer (Nozawa et al., 2006; Shojaei et al., 2008). In addition to lineage depletion studies, abundant immunohistochemical and protein array data clearly link infiltration of primary tumors with MMP-9-bearing neutrophils with angiogenic progression, metastatic disease, poor prognosis and survival in patients with several cancer types (Dumitru et al., 2012; Gregory and McGarry Houghton, 2011; Houghton, 2010; Kuang et al., 2011; Tazzyman et al., 2009; Tazzyman et al., 2013). Furthermore, a comparative analysis of TAMs and tumor-associated neutrophils (TANs) isolated from syngeneic mouse tumors demonstrated that only TANs showed high levels of MMP-9 mRNA expression and intracellular MMP-9 protein (Sawanobori et al., 2008). That inflammatory neutrophils constitute the major tumor-associated leukocyte type expressing abundant MMP-9, has also been demonstrated in our recent study employing human xenografts or syngeneic murine tumors (Deryugina et al., 2014). Specifically, when tumors or isolated TAMs and TANs were double-stained for MMP-9 and respective macrophage- or neutrophil-specific antigens, only TANs were emitting strong immunofluorescent signal for MMP-9 (Figure 1, A and B). Together, the majority of functional and morphological data point to tumor-infiltrating neutrophils as a critical tumor-associated leukocyte type that is responsible for delivery of angiogenesis-inducing MMP-9.

Figure 1. TANs constitute the major cellular source of angiogenic proMMP-9.

(A) LLC tumor was stained for MMP-9 (red) and either macrophage-specific antigen F4/80 or neutrophil-specific antigen Ly6G (green). Note the MMP-9-positive fibrillar matrix surrounding the MMP-9/Ly6G–positive neutrophils that appear yellow due to the overlap of red and green immunofluorescent signals. Bars, 20 µm.

(B) Isolated TAMs and TANs were stained for common cell myeloid marker CD11b (green) and MMP-9 (red). Only isolated TANs are positively stained for MMP-9, producing yellow signal indicating the overlap of MMP-9 (red) and lineage-specific (green) signals. Bars, 10 µm.

(C) Western blot analysis of MMP-9 and TIMP-1 produced by isolated TAMs and TANs. Whereas both cell types do not produce detectable TIMP-1, only TANs release substantial quantities of proMMP-9.

(D) Angiogenic potential of isolated TAMs and TANs and their respective products, TAMs’ conditioned medium and TANs’ releasate, as measured in the in vivo angiogenesis assay in live chick embryos. Note that TANs and TANs releasate induce higher levels of angiogenesis compared to low angiogenic TAMs and their conditioned medium.

4.3. Dependence of tumor angiogenesis on the bioavailability of VEGF released by inflammatory cell MMP-9

The actual functional link between the MMP-9 delivered by tumor-infiltrating leukocytes and endothelial cells responding to angiogenic induction has been established in studies demonstrating that the levels of tumor angiogenesis are proportional to the levels of VEGF liberated by the activity of MMP-9. The initial evidence was provided in the transgenic model of pancreatic cancer, where MMP-9-mediated liberation of matrix-sequestered VEGF induced the angiogenic switching in avascular pre-malignant tumors (Bergers et al., 2000). The overall involvement of bone marrow-derived myeloid cells in the delivery of VEGF-releasing MMP-9 and initiation of VEGF-dependent tumor angiogenesis was also shown in a glioblastoma model (Du et al., 2008a). In a cervical cancer model, reduced tumor angiogenesis due to diminished levels of mobilized VEGF was associated with MMP-9 produced by TAMs (Giraudo et al., 2004). In prostate cancer bone xenografts, where MMP-9 expression was immunohistochemically linked to the osteoclast cell population, significantly fewer and smaller blood vessels were observed in MMP-9 null mice, concurrent with lower levels of soluble VEGF produced by MMP-9-negative osteoclasts in vitro (Bruni-Cardoso et al., 2010). However, although many macrophage-depleting and macrophage-inducing studies have emphasized an overall importance of tumor-infiltrating monocytes and TAMs in tumor angiogenesis through VEGF-involving mechanisms (Green et al., 2009; Halin et al., 2009; Zaynagetdinov et al., 2011; Zeisberger et al., 2006; Zhang et al., 2010), these TAM-centric studies did not directly or conclusively implicate MMP-9 produced by TAMs neither in MMP-9-mediated VEGF mobilization in vivo, nor in the induction of metastasis-supporting tumor vasculature. In contrast, MMP-9 produced by tumor-infiltrating neutrophils and TANs has been repeatedly shown to release matrix-bound VEGF during in vivo angiogenesis. Thus, VEGF-induced angiogenesis in the mouse brain specifically required neutrophil infiltration and delivery of neutrophil MMP-9 to produce functionally bioavailable VEGF (Hao et al., 2007). In the genetic model of pancreatic cancer progression, neutrophil MMP-9 was demonstrated to act as a critical contributor to the bioavailability of VEGF and induction of the angiogenic switch (Nozawa et al., 2006). By cleaving heparan sulphates in the matrix deposited by colon carcinoma cells, neutrophil MMP-9 was specifically implicated in a colorectal cancer model in liberation of actual matrix-sequestered VEGF and induction of VEGF-dependent angiogenesis (Hawinkels et al., 2008). Recently, neutrophils and their TIMP-free MMP-9 have also been shown to increase VEGF bioavailability during inflammatory lymphangiogenesis (Tan et al., 2013). In addition to release of VEGF, our studies on physiological and tumor-induced angiogenesis highlighted the importance of neutrophil MMP-9-mediated release of another angiogenic factor, basic fibroblast growth factor, or FGF-2 (Ardi et al., 2009).

4.4. Possible VEGF-independent functions of neutrophil MMP-9

Neutrophil-derived MMP-9 has also been ascribed with direct, VEGF-independent functions during tumor angiogenesis. In a 3-dimensional in vitro assay, sprouting from endothelial cell spheroids could be 2-fold enhanced by either exogenous VEGF or neutrophil MMP-9, but VEGF-induced angiogenesis was insensitive to MMP-9 inhibition and vice versa, MMP-9-induced angiogenesis was insensitive to VEGF blockade (Bausch et al., 2011). Surprisingly, the combined use of VEGF and MMP-9 produced a synergistic, 6-fold enhancement and full ablation of this enhancement required simultaneous inhibition of VEGF and MMP-9. Based on these in vitro data, the in vivo findings demonstrating that the inhibition of MMP-9 activity causing a reduction of vascular density in tumors was not accompanied by significant changes in total VEGF, were interpreted as evidence against MMP-9-mediated release of VEGF and for VEGF-independent functions of MMP-9 in the angiogenic switch (Bausch et al., 2011). However, since the total, detergent-extracted VEGF rather than soluble VEGF was measured in this study, it is still possible that MMP-9 activity was involved in liberation of matrix-sequestered VEGF in the tumor. It is also possible that neutrophil MMP-9 is capable of catalytically activating the angiogenic process independent of VEGF release, e.g. by a proteolytic digest of various ECM proteins, including collagenase-cleaved collagen fibrils, resulting in less stiff matrices more amenable for endothelial cell invasion.

5. Specific roles of neutrophils and neutrophil MMP-9 in the induction of intravasation-and metastasis-sustaining vasculature

5.1. Differential ability of tumor variants to intravasate depends on inflammatory neutrophil influx

To investigate the MMP-involving mechanisms governing the process of tumor cell intravasation, we have employed a number of congenic tumor cell variants selected in vivo from human cancer cell lines, including HT-1080 fibrosarcoma, PC-3 prostate carcinoma and HEp-3 epidermoid carcinoma, for low and high levels of dissemination (i.e., lo/diss and hi/diss variants). The respective pairs generate primary tumors of comparable sizes, but hi/diss and lo/diss variants display up to 100-fold differentials in the levels of vascular intravasation in the live chick embryo model and up to 25-fold difference in spontaneous lung metastasis in various mouse models, including orthotopic implantation models (Bekes et al., 2011a; Deryugina et al., 2005; Juncker-Jensen et al., 2013; Partridge et al., 2007). Furthermore, the HT-hi/diss and PC3-hi/diss cells are approximately 2–3-fold more angiogenic than their respective lo/diss counterparts in the chick embryo onplant and mouse angiotube models (Bekes et al., 2011b), suggesting that different levels of angiogenesis in the primary tumors might determine different levels of intravasation. Histological and immunohistochemical analyses of primary tumors, however, indicated that intravasation differential between hi/diss and lo/diss variants directly correlated with the density of lumen-containing vessels rather than with the overall density of the tumor-associated angiogenic vasculature (Bekes et al., 2011b; Deryugina et al., 2014; Juncker-Jensen et al., 2013). Therefore, the distinct ability to induce and sustain the development of angiogenic vessels possessing specific-sized lumens has been linked for the first time to the capacity of tumor cells to complete the process of vascular intravasation.

The intravasation differential of our tumor variants as it relates to the intratumoral networks of lumen-containing angiogenic blood vessels and tumor cell dissemination has, in turn, been linked to the differential ability of these tumor variants to induce inflammatory cell influx. Specifically, the density of TANs and the rates of neutrophil influx were substantially different between the hi/diss and lo/diss pairs of fibrosarcoma and prostate carcinoma variants whereas the levels of TAMs were similar (Bekes et al., 2011b). Furthermore, the close association between the levels of neutrophil infiltration and tumor cell intravasation was linked directly to the levels of neutrophil proMMP-9 delivered by tumor-infiltrating neutrophils into the primary tumor site. Thus, inhibition of neutrophil infiltration into primary HT-hi/diss and PC-hi/diss tumors by functionally blocking tumor cell-produced IL-8, a potent and specific neutrophil chemoattractant, resulted in a substantially inhibited development of a lumen-containing intratumoral vasculature concomitantly with significantly reduced levels of intravasation. Importantly, these effects of inflammatory neutrophil deficit were reversed by the supplementation of developing tumors with purified neutrophil proMMP-9, which in the HT-hi/diss in vivo metastasis model restored in a coordinated fashion the levels of intravasation and the density of lumen-containing vasculature. Further illuminating the functional role of neutrophil proMMP-9 in tumor-induced angiogenesis, supplementation of HT-lo/diss and PC-lo/diss cells with purified neutrophil proMMP-9, increased their low angiogenic potentials in the collagen onplant model, bringing angiogenesis to hi/diss levels. Reciprocally, the use of an anti-inflammatory drug, ibuprofen, inhibited hi/diss-induced angiogenesis, and these ibuprofen-mediated effects were also reversed and brought back to control levels by exogenous delivery of purified neutrophil MMP-9 (Bekes et al., 2011b).

5.2. Unique angiogenesis-inducing potency of TIMP-free MMP-9 produced by neutrophils and neutrophil-like MMP-9 produced M2-polarized macrophages

Compared to tumor-infiltrating monocytes or TAMs and other types of tumor-associated inflammatory leukocytes, inflammatory neutrophils appear to be much better equipped to deliver angiogenesis-inducing quantities of MMP-9, capable of liberating VEGF and bFGF and executing other matrix-modifying activities (Owen and Campbell, 1999). Our initial studies have demonstrated that neutrophil MMP-9 is exceptionally powerful in the induction of tumor angiogenesis, functioning as a pro-angiogenic factor at low nanomolar concentrations, comparable with the angiogenesis-inducing concentrations of VEGF and bFGF (Ardi et al., 2007; Ardi et al., 2009). This unusual angiogenic potency of neutrophil MMP-9 was linked first, to the unique form of the zymogen produced by neutrophils as an inactive proenzyme unencumbered by the natural MMP-9 inhibitor, TIMP-1, and second, to the catalytic action of an activated enzyme. In other cell types, secreted proMMP-9 is bound to TIMP-1, which not only significantly slows the activation of the MMP-9 zymogen, but also can immediately inhibit the proteolytic activity of the once activated enzyme (Ogata et al., 1995; Okada et al., 1992). Therefore, a TIMP-1-free form of proMMP-9 has a much higher probability for in vivo activation and execution of its enzymatic activity before either of these biochemical processes would be inhibited by endogenous TIMP-1 or TIMP-2, the latter a potent and abundant natural inhibitor of many activated MMPs.

In contrast to neutrophils, monocytic cell lines, like most cells, produce their proMMP-9 in a conventional stoichiometric 1:1 complex with TIMP-1 rendering it with low angiogenesis-inducing activity (Ardi et al., 2007), suggesting that monocyte-derived macrophages would also produce proMMP-9 complexed with TIMP-1. The production of proMMP-9/TIMP-1 complex was indeed demonstrated for normal human and mouse monocytes differentiated in vitro from either peripheral blood or bone marrow precursors in response to M-CSF (Zajac et al., 2013). Similarly, a TIMP-1-encumbered status of MMP-9 was also characteristic of proMMP-9 produced by mouse and human macrophages polarized into “classic” M1 phenotype. Remarkably, when in vitro-generated macrophages were polarized towards an “alternative” M2 phenotype, they shut down their TIMP-1 expression and initiated the production of a “neutrophil-like”, TIMP-1-free proMMP-9 (Zajac et al., 2013). Furthermore, we have recently demonstrated that TAMs isolated from murine tumors and manifesting specific characteristics of M2-skewed macrophages, such as production of arginase and the lack of iNOS production, also did not express TIMP-1 and therefore produced TIMP-1-free proMMP-9 (Deryugina et al., 2014).

5.3. Inflammatory neutrophils and TANs constitute the major source of angiogenesis-inducing MMP-9

Despite compelling evidence for the delivery of angiogenesis-inducing MMP-9 to the tumor microenvironment by bone marrow-derived myeloid cells, the majority of the studies were not accompanied by a quantitative analysis of MMP-9 production or comparison of the angiogenic potency of the MMP-9 produced by various tumor-associated leukocytes. In attempt to fill this apparent gap in the biology of MMP-9-mediated tumor angiogenesis and to clarify the controversy in the accumulated evidence regarding the leukocyte delivery of MMP-9, we recently have conducted a quantitative investigation and compared MMP-9-producing capacities of isolated neutrophils and monocyte-derived macrophages of human and murine origin as well as TANs and TAMs isolated from syngeneic murine tumors (Deryugina et al., 2014; Zajac et al., 2013). In addition, a special focus was placed on quantification of MMP-9 produced by monocytes and different types of macrophages, namely mature macrophages versus macrophages polarized towards M1 or M2 phenotypes. Consistent with its TIMP-free status, MMP-9 purified from M2-macrophages and TAMs, has been shown to be as angiogenic as proMMP-9 from neutrophils or TANs, when compared on a molar basis as purified zymogens. However, on a cell-to-cell level, murine macrophages and TAMs were shown to produce 40–50 times less proMMP-9 in 48 hours compared to the intracellular MMP-9 load of blood-circulating neutrophils and TANs, which can release their cargo on demand. Our quantification indicated that while 1×10⁶ neutrophils and TANs can release approximately 100–200 ng proMMP-9 within 1–2 hr of incubation, 1×10⁶ macrophages and TAMs would require at least 10 weeks to synthesize and secrete the same amount of proMMP-9 [Figure 1C and (Deryugina et al., 2014)]. Thus, it appears that inflammatory neutrophils constitute the major source of MMP-9 capable of matrix remodeling, liberation of VEGF, and induction of angiogenesis in primary tumors. Consistent with this notion, the isolated TANs and their released cellular contents but not TAMs and their conditioned medium were capable of inducing high levels of in vivo angiogenesis (Figure 1D).

5.4. Microarchitecture and pericyte coverage of intravasation-sustaining vasculature depends on neutrophil influx and delivery of angiogenic MMP-9

MMP-9-dependent tumor angiogenesis especially at the early stages of primary tumor development appear to be determined specifically by the rates of neutrophil influx (Bekes et al., 2011b; Deryugina et al., 2014; Nozawa et al., 2006; Shojaei et al., 2008; Zajac et al., 2013; Zijlstra et al., 2006), which, in turn, is regulated by the levels of neutrophil attractants expressed by tumor cells. Thus, in several cancer models enhanced IL-8 production has specifically been linked to high levels of tumor infiltration by neutrophils and tumor angiogenesis (Benelli et al., 2002; Brat et al., 2005; De Larco et al., 2004; Haqqani et al., 2000; Kim et al., 2001; Sparmann and Bar-Sagi, 2004; Waugh and Wilson, 2008; Yao et al., 2007). In a positive feed-back loop, neutrophil MMP-9 also has been shown to significantly potentiate the functional activity of IL-8 (Van den Steen et al., 2000). By using our hi/diss variants in a spontaneous metastasis avian model, in which neutrophil infiltration and therefore the delivery of neutrophil MMP-9 were inhibited by IL-8 blockade, we functionally linked the levels of intravasation and metastasis with the development of an angiogenic, lumen-containing vasculature (Bekes et al., 2011b). Histological examinations of angiogenic vasculature in syngeneic tumors developing in wild-type versus MMP-9-KO mice also indicates that the lumen-containing vasculature was fully developed only when tumors were infiltrated with MMP-9-positive leukocytes (Chantrain et al., 2004). Our recent detailed analyses of syngeneic tumors developing in MMP-9 null and wild-type mice confirmed that the majority of influxing MMP-9-positive leukocytes were mature neutrophils and that the lack of neutrophil MMP-9 delivery in Mmp9-KO hosts resulted in the development of collapsed, lumen-devoid vasculature severely deficient in pericyte coverage [Figure 2A and (Deryugina et al., 2014)]. Thus, while in a MMP-9-competent microenvironment almost 60% of tumor vessels developed lumens of >11 µm in diameter, less than 20% of all tumor vessels presented with lumens of this size in Mmp9-KO hosts (Figure 2B). Furthermore, 4-times more pericytes were associated with blood vessel lumens in WT tumors compared with MMP-9-deficient tumors (Figure 2B). It would appear that in wild type animals, the distinctive albeit discontinuous pericyte coverage apparently allows for enough structural support of lumen-containing vessels, but also for their sufficient penetrability by intravasating tumor cells. Therefore, neutrophil MMP-9 constitutes a critical microenvironmental contributor to the development of an architecturally sound intratumoral vasculature capable of sustaining maximal tumor cell intravasation and metastatic dissemination.

Figure 2. Development of intratumoral lumen-containing angiogenic vessels and their partial coverage with pericytes depend on the host MMP-9.

(A) LLC tumors grown in wild-type (WT) or MMP-9 knockout (KO) mice were immunostained for the endothelial cell marker CD31 (green) and pericyte marker NG2 (red). Blood vessels in WT tumors exhibit enlarged lumens (outlined) associated with more pericytes than the vessels in MMP-9-deficient tumors. Bars, 20 µm.

(B) Lumen-containing vessels (Y axis on the left) and average number of lumen-associated pericytes (Y axis on the right) were quantified in WT tumors versus MMP-9-deficient LLC tumors. Note a 4-fold differential in both parameters in favor of angiogenic vessels in MMP-9-competent tumors.

In addition to MMP-9 delivered to the tumor microenvironment by neutrophils, other MMPs and non-neutrophilic cells have also been implicated in pericyte coverage of angiogenic vasculature. Thus, complete genetic ablation of MMP-2 in a glioblastoma model resulted in the development of impaired and dysfunctional tumor vasculature that surprisingly, provided a attractive scaffold for tumor cell invasion along these poorly perfused vessels (Du et al., 2008b). Recently, pericyte coverage of tumor-associated vasculature and its metastasis-supporting ability has also been linked to a heterogeneous population of CD13-positive bone marrow-derived myeloid cells, represented by tumor-associated monocytes, macrophages, granulocytes, dendritic cells and T lymphocytes, implicated in production of a variety of proangiogenic factors, including MMP-9 (Dondossola et al., 2013). However, it remains unclear how much of the MMP-9 is produced by different types of CD13-positive leukocytes and whether this specific MMP-9 regulates pericyte coverage of tumor vasculature.

6. EGFR-mediated regulation of tumor cell intravasation through an MMP-9/VEGF-involving mechanism

Some recent findings suggest that MMP-9-dependent regulation of tumor angiogenesis can also be governed by specific tumor cell receptors, which regulate MMP-9 expression at the transcriptional level. Thus, activation of the pathways downstream of epidermal growth factor receptor (EGFR) increases MMP-9 expression and activity in many cancer cell types, including breast carcinomas, lung carcinomas, head and neck squamous carcinomas and glioblastomas (Choe et al., 2002; Ellerbroek et al., 2001; Kim et al., 2009; O-Charoenrat et al., 2000; Zhao et al., 2010). Reciprocally, anti-EGFR agents, which can directly effect EGFR-overexpressing tumor and endothelial cells, also inhibit expression of MMP-9 and reduce tumor growth indirectly by blocking VEGF production (Normanno and Gullick, 2006; Pore et al., 2006; Ratushny et al., 2009). Whereas EGF/EGFR-regulated MMP-9 functions have mostly been attributed to well-established roles of tumor cell-produced MMP-9 in matrix remodeling and tumor cell stromal invasion (Hwang et al., 2011; Zhao et al., 2010), downregulation of VEGF production in tumor xenografts treated with an EGFR inhibitor was linked to normalization of morphology and permeability of tumor-associated vessels (Cerniglia et al., 2009).

Providing novel insights to the role of tumor cell-produced MMP-9 to tumor angiogenesis and angiogenesis-dependent tumor cell dissemination, we have recently linked the expression and activity of tumor cell EGFR to development of an intravasation-supporting vasculature (P. Minder and EID, unpublished data). First evidence has come from the finding that all our hi/diss variants, independent of tissue origin, i.e. fibrosarcoma, prostate carcinoma and epidermoid carcinoma, express significantly higher levels of EGFR compared to their lo/diss counterparts. Interestingly, increased expression of EGFR in breast cancer cells has also been linked to enhanced intravasation and metastasis in a breast cancer mouse model, but these higher rates of cell dissemination were linked to increased cell motility of EGFR-overexpressing tumor cells (Xue et al., 2006). In contrast, we found no correlation between the motility of tumor cells in our HT-1080 and HEp-3 lo/diss and hi/diss variants (Deryugina et al., 2005; Juncker-Jensen et al., 2013), indicating that additional mechanisms or changes in tumor microenvironment other than overall EGFR levels could indirectly regulate EGFR-mediated tumor cell intravasation.

To investigate mechanistic links between EGFR and EGFR-dependent tumor cell dissemination, we employed our recently-established microtumor model, allowing for simultaneous measurements of tumor cell intravasation, tumor angiogenesis, vascular permeability and vascular microarchitecture of the intratumoral vasculature (Deryugina, in press). As an initial approach, the levels of EGFR were downregulated in HT-hi/diss cells by siRNA interference. It is important that in our model, down-regulation of EGFR significantly reduced HT-hi/diss intravasation independent of EGFR effects on tumor growth and overall levels of angiogenesis (Figure 3A), prompting the analysis of specific EGFR-dependent attributes of the intratumoral vasculature. Thus, EGFR silencing was associated with a severe underrepresentation of lumen-containing vessels, reduction in the density of TANs and bioavailable VEGF (Figure 3B). Similar effects were observed when EGFR function was inhibited with the EGFR-specific tyrosine kinase inhibitor, erlotinib, implicating EGFR activity and signaling in the development of intravasation-sustaining blood vessels. However, supplementation of purified neutrophil MMP-9 to EGFR-silenced tumors, fully restored the levels of intravasation along with the restoration of 15–30 µm-sized vessels and bioavailable VEGF (Figure 3B), thereby confirming functional regulation of intravasation-sustaining vascular development by EGFR via the MMP-9/VEGF pathway,

Figure 3. Tumor cell EGFR regulates the development of angiogenic vessels in the HT-hi/diss microtumors in a neutrophil MMP-9/VEGF-dependent manner.

(A) GFP-tagged HT-hi/diss cells were treated with control siRNA (siCtrl) or EGFR siRNA (siEGFR) and grafted on the CAM of live chick embryos. After 5 days, intratumoral blood vessels within developed microtumors were highlighted in vivo with red-fluorescent lectin. Despite the similar size of HT-hi/diss microtumors (upper panels; bars, 100 µm), EGFR silencing resulted in the development of thin and collapsed blood vessels with reduced size lumens compared to control (lower panels; bars, 50 µm). Note that treatment of siEGFR tumors with purified neutrophil proMMP-9 restored the appearance of intratumoral blood vessels.

(B) Quantification of intratumoral angiogenic vessels with 15–30 µm-lumens, the density of intratumoral MMP-9-positive TANs, the levels of intravasation and the levels of bioavailable VEGF in tumors. Downregulation of EGFR resulted in a significant underrepresentation of angiogenic vessels with the lumens of specific 15–30 µm size, concomitant with the reduction of neutrophil influx and tumor cell intravasation. Both blood vessel development and intravasation rates were restored by supplementation of siEGFR-silenced tumors with purified neutrophil MMP-9 (nMMP-9), which also fully restored the levels of bioavailable VEGF.

7. VEGF-induced tumor angiogenesis mediated by MMP-9 expressed by cancer cells and cancer-associated fibroblasts

The role of tumor cell MMP-9 in the overall process of tumor metastasis has been associated mainly with the regulation of tumor growth through the mechanisms involving matrix remodeling and stromal invasion by tumor and endothelial cells. In these processes, tumor cell-produced MMP-9 was shown to function through cooperation with complementary molecules such as uPA, uPAR, and cathepsin B (Bjorklund and Koivunen, 2005; Deryugina and Quigley, 2006; Farina and Mackay, 2014). Since the expression of MMP-9 and VEGF in tumor cells appears to be coordinately regulated on the transcriptional level, this particular mutual connection has been implicated in tumor angiogenesis and metastasis models where MMP-9 and VEGF were either enhanced or downregulated in a tandem fashion. Thus, the simultaneously enhanced production of MMP-9 and VEGF in A549 lung carcinoma cells in response to cytochrome P450 omega-hydroxylase resulted in a 2-fold increase in microvessel density in primary tumors and a similar 2-fold increase in metastasis from lung tumors to distant organs (Yu et al., 2011). Reciprocally, downregulation of MT1-MMP (MMP-14) in human ovarian cancer OVCAR-4 cells by RNA interference resulted in a significant diminishment of MMP-9 and VEGF production, concomitant with a dramatic reduction in tumor cell-induced angiogenesis in matrigel plugs (Kaimal et al., 2013).

Some agents appear to regulate levels of MMP-9 expression independently of the VEGF gene. Thus, late SV40 factor (LSF), which directly targets the MMP-9 gene and induces MMP-9 production, was shown to enhance tumor angiogenesis in an MMP-9-dependent manner (Santhekadur et al., 2012). In an orthotopic implantation model, MMP-9 produced by breast cancer cells regulated spontaneous lung metastasis via facilitating the motility-dependent crossing of the endothelial cell barrier by cancer cells (Kim et al., 2011). A recent study also re-confirmed overall dependence of tumor angiogenesis on the stromal MMP-9 produced by cancer-associated fibroblasts activated by specific interactions with tumor cells (Taguchi et al., 2014).

8. Tumor angiogenesis and metastasis regulated by tumor cell-produced MMP-1 inducing an intravasation-supporting vasculature

A critical example of a tumor cell-produced MMP regulating the development of intravasation-sustaining vasculature is represented by secreted MMP-1 in the head and neck cancer model described in our recent publication (Juncker-Jensen et al., 2013). By using our HEp-3 dissemination variants in a microtumor model, we have linked MMP-1 secreted by HEp3-hi/diss carcinoma cells to their high levels of intravasation into CAM vasculature and metastasis to the liver. Similar to EGFR-dependent intravasation-sustaining blood vessels, MMP-1-dependent intravasation also required the presence of lumens of 15–30 µm and increased levels of vascular permeability as measured by high mol. wt. Dextran exudation. However, in the case of tumor-produced MMP-1 it appeared to be a proteolytic activation of the endothelial cell membrane-tethered receptor, PAR1, which ultimately facilitated HEp3-hi/diss intravasation (Juncker-Jensen et al., 2013). These in vivo data corroborate a MMP-1/PAR1-involving mechanism previously demonstrated in vitro for endothelial cells and for tumor cell-endothelial cell settings (Boire et al., 2005; Brinckerhoff et al., 2000; Tressel et al., 2011; Yang et al., 2009). Furthermore, it has recently been demonstrated that MMP-1/PAR1-pathway is involved in the up-regulation of VEGFR in the endothelial cells (Mazor et al., 2013).

An additional mechanism whereby tumor cell-produced MMP-1 could regulate the development of metastasis-supporting vasculature was underscored in our recently-established orthotopic mouse model of head and neck cancer (Anna Juncker-Jensen, EID and JPQ, unpublished data). Similar to the chick embryo model, the tumor-associated vasculature in MMP-1-competent primary tumors developing in the buccal mucosa of immunodeficient mice manifesting high rates of spontaneous lung metastasis has been characterized by the presence of lumens of ∼11–15 µm. Furthermore, MMP-1 deficiency in HEp3-hi/diss tumors caused the development of dense but lumen-deficient intratumoral vasculature, reduced neutrophil influx and significant inhibition of spontaneous metastasis to the lungs. This MMP-1 repression in HEp-hi/diss cells also resulted in downregulation of IL-8 production in primary tumors, suggesting that MMP-1-dependent angiogenesis involves IL-8-induced neutrophil inflammatory response and a delivery of neutrophil MMP-9 into the tumor microenvironment. This putative tumor MMP-1/tumor IL-8/neutrophil MMP-9 axis might represent yet another example of complex and diversified pathways whereby MMPs of tumor and host origin directly, via matrix degradation, or indirectly, via angiogenic factor release and activation, regulate the quality of metastasis-sustaining angiogenic vasculature.

9. Conclusions

The overall dependence of tumor angiogenesis on the expression of various MMPs is still in the spotlight of experimental and clinical cancer research, deeply focusing on unraveling the precise interrelations between different regulatory molecules and proteolytic activities of specific MMPs (Al Rawashdeh et al., 2014; Barillari et al., 2014; Depner et al., 2014; Gong et al., 2014; Kahlert et al., 2014; Zhang et al., 2014). In this review, however, we attempted to summarize the data on the mechanisms whereby various MMPs, characterized by different substrate specificities, e.g. collagenases and gelatinases, and originating from different cell types, e.g. tumor cells, inflammatory leukocytes and cancer-associated fibroblasts, induce within primary tumors the development of a distinct angiogenic vasculature capable of sustaining tumor cell intravasation and intravasation-dependent metastasis. As indicated by in vivo imaging and quantitative measurements in our studies, this intravasation- and metastasis-sustaining vasculature should possess specific microarchitectural characteristics such as partial but distinctive pericyte coverage, vessel lumens of ∼10–30 µm in diameter, and elevated vascular permeability, all of which are required for optimal support of active tumor cell intravasation and dissemination of metastatic cells via vascular routes. It would appear that the major biochemical mechanisms that operate in MMP-mediated development of the intravasation-sustaining vasculature are proteolytic remodeling of the tumor matrix and the linked release of angiogenic factors like VEGF and bFGF. In this regard, the majority of accumulated evidence indicates that the proMMP-9 delivered by tumor-influxing neutrophils possesses the highest biochemical and physiological potentials, such as rapid release, high local concentration, TIMP-free status, high rates of activation and efficiency of substrate catalysis, to digest tumor matrix and release matrix-sequestered VEGF and bFGF into the tumor microenvironment. In turn, the released and functionally induced VEGF and bFGF directly regulate the microarchitecture and functions of the intratumoral vasculature, including its ability to sustain tumor cell intravasation and metastasis.

Highlights.

• MMPs are involved in angiogenesis-dependent intravasation and metastasis

• Inflammatory cell MMP-9 triggers the onset of tumor neovascularization

• IL-8-responding neutrophils are the major source of angiogenesis-inducing MMP-9

• Neutrophil MMP-9 catalytically releases angiogenesis factor VEGF from tumor matrix

• MMP-9/VEGF axis regulates intravasation- and metastasis-sustaining neovasculature