Tumor MMP-1 Activates Endothelial PAR1 to Facilitate Vascular Intravasation and Metastatic Dissemination Free

Deaths from cancer occur mainly because tumor cells metastasize from the primary tumor site to secondary vital tissues and organs likely by using the new angiogenic vasculature of the tumor as conduits. Metastasis, which is a complex multistep process, is still one of the major unsolved problems in cancer biology. Intravasation, the step in metastasis whereby escaping primary tumor cells enter the vasculature, is the least studied process in the entire metastatic cascade. Intravasation is experimentally understudied because there are very few if any in vivo models that accurately recapitulate the entry of tumor cells into the vasculature and also allow for quantification of the intravasation events. Furthermore, real-time imaging of escaping primary tumor cells and in vivo microscopic analysis of the structure and functionality of tumor-associated vasculature remain problematic for most laboratories. Because of these modeling and methodologic issues, no clear signature molecules that directly contribute to the intravasation event have been identified. However, several mechanisms have been linked to the processes and events leading up to the intravasation step, such as primary tumor cell escape and migration and protease-mediated tumor cell invasion. In regard to the latter, proteolytic degradation of the basement membrane and stromal matrix by specific members of the matrix metalloproteinase (MMP) family of enzymes could provide functional molecular links to tumor cell escape, transendothelial migration, and possibly to tumor cell-mediated active entry into the vasculature. The MMPs comprise a family of zinc-dependent endopeptidases that proteolytically modify the extracellular matrix in the primary tumors and metastatic sites as well as cleave distinct molecules on the surface of tumor and stromal cells (1–3).

A number of MMP genes have been linked to development and progression of squamous cell carcinomas (SCC), which constitute 90% of head and neck cancers, the fifth leading cause of cancer-related deaths (4). The MMP genes that have been linked to SCC progression, include Mmp1, 2, 3, 7, 9, 10, 11, 12, and 13, most of which are overexpressed in the SCC tissue compared with normal mucosa (4, 5). However, a remarkable, more than 70-fold differential between normal oral mucosa and oral SCCs was shown for the Mmp1 gene, which was found to be third best predictor among 25 signature genes (5), suggesting a critical role of MMP-1 protein in SCC progression. Furthermore, while the expression of many MMPs in primary SCCs is associated with stromal or inflammatory cells rather than carcinoma cells, MMP-1 protein expression has been attributed to cancer cells at least in oral SCCs (5). In addition, MMP-1 has shown up as one of the signature genes for the metastatic phenotype for human breast cancers (6–8) and has also been validated as part of a set of 63 genes associated with the progression and metastasis of advanced cervical carcinomas (9). All these considerations clearly warrant mechanistic study of the functional contribution of tumor-produced MMP-1 to metastasis of SCCs.

To functionally analyze the role of MMP-1 in overall metastatic dissemination and specifically the intravasation step of SCCs, we used the human epidermoid carcinoma cell line, HEp3, which is highly metastatic in both mice and chick embryos (10, 11). A distinctive feature of the chick embryo model, which is based on the grafting of human tumor cells on the chorioallantoic membrane (CAM), is that it uniquely allows for quantitative monitoring of intravasation into the CAM vasculature during spontaneous metastasis. With regard of intravasation, the HEp3 cells, when grafted onto the CAM at early passages, give rise to primary tumors and also disseminate to internal organs through the process of intravasation. These early passage-selected HEp3 cells have been referred to as the highly disseminating variant, HEp3-hi/diss. After 25 to 70 days in culture, the HEp3-hi/diss cells still maintain full tumorigenic capacity, but substantially reduce their metastatic potential and become low-disseminating cells, denoted herein as HEp3-lo/diss. The lost metastatic potential of the HEp3-lo/diss cells can be recovered by in vivo passaging on the CAM, allowing for a continuous source of aggressive HEp3-hi/diss cells. These 2 congenic HEp3 variants, presenting a distinct differential in their metastatic behavior, provide a suitable in vivo model system for identifying molecular factors that functionally contribute to the intravasation step of carcinoma cells.

By using the low and high metastatic variants of HEp3 carcinoma, we have shown in this study that tumor-produced MMP-1 functionally contributes to SCC vascular dissemination. However, in contrast to the established collagenous matrix remodeling and cell invasion-promoting functions of MMP-1, we have illuminated the mechanism whereby tumor-produced MMP-1 regulates development of an intravasation-sustaining intratumoral vasculature and assists carcinoma cell intravasation by inducing vascular permeability through the activation of endothelial PAR1. By placing MMP-1–producing carcinoma cells into an in vivo microenvironment where tumor cells confront intratumoral PAR1-positive angiogenic vasculature, our study for the first time presents evidence of how aggressive carcinoma cells, via secreted MMP-1, partially disrupt the endothelial barrier making it more amenable for the active entry of escaping cancer cells into the circulation.

Human endothelial EA.hy926 cells (CRL-2922) were obtained from the American Type Culture Collection (ATCC). Before this study, the cells were verified for expression of human CD44, CD31, CD105, and factor VIII–related antigen by immunocytochemistry. The cells were also tested and proved negative for Mycoplasma contamination. Human head and neck carcinoma cell line Detroit 562 was obtained from ATCC and used within 6 month after purchasing. Unless specified, the cells were maintained as monolayer cultures in a humidified incubator at 5% CO₂ and 37°C. Cell passaging was carried out by brief exposure of cell monolayer to minimal volume of trypsin–EDTA solution followed by plating of detached cells in fresh Dulbecco's Modified Eagle Medium (DMEM)–10% FBS.

Human epidermoid carcinoma HEp3-hi/diss cells were originally isolated from a neck lymph node metastasis of a patient diagnosed with head and neck carcinoma (12). The HEp3 cells were not deposited to any tissue or cell bank. The original HEp3 cell line was thoroughly characterized in animal models by Ossowski and Reich (13, 14) and became widely used in cancer research. In our laboratory, the parental HEp3 cells are maintained since 1986 (15) and have been repeatedly used in various cancer research studies (11, 16–19). The ability of HEp3 cells to form metastatic tumors after orthotopic implantations into the tongue and buccal mucosa of nude mice was verified in our laboratory in 2012.

The HEp3-hi/diss cells were generated from primary tumors developing from parental HEp3 cells grafted on the CAM of chick embryos as described previously (20). Briefly, CAM tumors were dissected from the CAM, minced, and incubated with dispase (1.25 U/mL) to generate single-cell suspension. After washing, the cells were placed into DMEM supplemented with 10% FBS (HyClone), sodium pyruvate (Invitrogen), and 10 μg/mL gentamicin (Gibco; D-10). Aliquots of primary cells were frozen at day 5 to provide a supply of HEp3 cells at low passage numbers. The cells that were maintained in culture for 25 days or less exhibited high levels of spontaneous dissemination and were designated as HEp3-hi/diss. The cells that were cultured for more than 75 days lost dissemination capacity and were designated as HEp3-lo/diss. The parental HEp3, HEp3-hi/diss, and HEp3-lo/diss cells were tested for Mycoplasma contamination and confirmed to be negative.

Chick embryo spontaneous metastasis assays were conducted as described previously (21). Briefly, 4 × 10⁵ HEp3 cells were grafted on the top of the CAM. Where indicated, the developing tumors were treated topically on day 1 and 3 with 30 to 100 ng of recombinant MMP-1 (provided as a mixture of zymogen and activated species by Abcam) or with MMP inhibitor GM6001 (Calbiochem-EMD Chemicals). After 5 days, the primary tumors were excised, weighed and fixed in 10% Zn-formalin for histologic examination. Portions of the distal CAM and the liver were examined by quantitative Alu-PCR to determine numbers of human tumor cells within chicken tissue background.

Embryos were inoculated with 5 × 10⁴ tumor cells. At 2 hours (cell arrest) or 5 days (tissue colonization), the portions of the CAM and liver were analyzed by Alu-qPCR to determine numbers of human cells. CAM colonization was also examined by fluorescence microscopy as described previously (22). For this analysis, the cells were prelabeled with 5 μmol/L CellTracker Green, whereas the CAM vasculature was highlighted with Rhodamine-conjugated Lens culinaris agglutinin (LCA; Vector Labs).

HEp3-hi/diss cells treated with control and MMP-1 siRNA constructs were labeled with 5 μmol/L CellTracker Green and resuspended at 2 × 10⁶/mL. Five to 7 small boluses of tumor cells (3–5 μL) were injected directly into the CAM mesoderm of day 10 chick embryos incubated ex ovo as described previously (23). Five days after cell injections, embryos were inoculated with Rhodamine-conjugated LCA to highlight the vasculature. Portions of the CAM with microtumors were imaged using a Carl Zeiss AxioImager microscope. Quantification of tumor cell escape and invasion was conducted using ImageJ software (NIH, Bethesda, MD). The mean of invasion distances from the microtumor–CAM border was determined for individual microtumors. A total of 11 to 13 individual microtumors from 6 to 8 embryos were analyzed for each variable in 2 to 3 independent experiments.

HEp3-hi/diss cells were treated with control or MMP-1 siRNA constructs, prelabeled with CellTracker Green and mixed with neutralized type I collagen (Becton Dickinson) at 1 × 10⁷ cells per mL and 6 10-μL aliquots were placed separately on the top of the intact CAM of embryos developing for 9 or 10 days ex ovo to generate microtumors. After 6-day incubation, Rhodamine-conjugated LCA was inoculated intravenously to highlight the CAM vasculature. The portions of the CAM containing microtumors were visualized in a fluorescent Olympus microscope. The images were analyzed for the overall structure of intratumoral vessels and their diameter.

To measure intratumoral vasculature permeability, CAM microtumors were generated from unlabeled HEp3-hi/diss cells. The developing microtumors were treated topically with 20 μL of vehicle (1% DMSO in PBS) or solutions of PAR1 antagonists, SCH79797 or RWJ56110 (Tocris Bioscience) applied at 5 to 50 μmol/L concentrations. The treatments were carried out on days 1, 3, and 5 or days 2 and 4. On day 6, embryos were inoculated first with the permeable, TRITC-conjugated dextran with molecular weight of 155 kDa. Nonpermeable fluorescein isothiocyanate (FITC)–conjugated dextran with molecular weight of 2,000 kDa was inoculated 45 to 60 minutes later, displacing the 155-kDa dextran from the vasculature. Images of microtumors and intratumoral vasculature were acquired in red and green fluorescence channels at the fixed exposure time, and individual microtumors were excised and lysed in 0.3 mL modified radioimmunoprecipitation assay (RIPA) buffer. Levels of red fluorescence, reflecting amounts of tissue-retained low molecular weight TRITC-conjugated dextran, were measured in lysates of individual tumors at 576 nm (excitation at 557 nm). Levels of green fluorescence, reflecting the volume of perfusable vasculature, were measured at 516 nm (excitation at 492 nm).

Data processing and statistical analyses were conducted using GraphPad Prism Software (GraphPad Software Inc.). For the chick embryo assays, percentage changes were calculated from the pooled differences determined as the ratios of numerical values for individual embryos over a mean of the control in the corresponding experiment. Statistical significance was evaluated using two-tailed unpaired Student t test for P < 0.05.

Quantitative real-time PCR (qRT-PCR), Alu-qPCR, MMP-1 silencing, cell maintenance, cell adhesion, chemotactic invasion in Transwell, endothelial permeability and transendothelial migration of HEp3-hi/diss cells, immunohistochemistry, Western blotting, and gelatin zymography are described in the Supplementary Data.

Comparative analysis of high- and low-disseminating HEp3 variants was conducted in spontaneous versus experimental metastasis models to identify those temporal steps of the dissemination cascade where nonmetastatic and metastatic carcinoma cells manifest most of their functional differential. In a spontaneous metastasis model, HEp3-hi/diss and HEp3-lo/diss cells gave rise to primary tumors with almost equal weight (approximately 166 ± 6 mg), indicating similar tumorigenic potentials of HEp3 variants. However, only HEp3-hi/diss cells exhibited high levels of dissemination compared with their late-passage counterparts, reflected in a 25-fold differential in CAM intravasation and 10-fold differential in liver metastasis (P < 0.0001; Fig. 1A).

Whether overall metastatic potential of HEp3 variants depended on their ability to complete the late steps of cell dissemination was tested in an experimental metastasis model involving tumor cell inoculations directly into the blood circulation and thus bypassing the early events of metastasis, including primary tumor formation and tumor cell intravasation. The HEp3-lo/diss and HEp3-hi/diss cells did not differ in their 2-hour arrest in the CAM vasculature, indicating their similar ability to interact with the luminal surface of vascular endothelium. Furthermore, HEp3 variants differed only by 3- to 4-fold in their ability to colonize the CAM and liver tissue as measured at day 5 after cell inoculations (Fig. 1B). When compared with the 10- to 25-fold differentials in spontaneous metastasis, it seems that the dissemination disparity between the HEp3 variants lies primarily in their ability to complete early steps of the metastatic cascade, preceding and including the process of intravasation into the vasculature.

Because MMPs have been strongly implicated in progression of different cancer types, we thought to identify those distinct MMP(s), the expression of which would correlate with the dissemination potential of the HEp3 carcinoma variants. We first evaluated HEp3-hi/diss cells for overall expression of several candidate MMP genes, including Mmp1, 2, 3, 7, 8, 9, 10, 13, and 14. Although overall mRNA levels of Mmp3, 7, 9, 10, and 13 were relatively low (C_t values between 24 and 30; ΔC_t values between 11 and 17), the expression levels of Mmp1, 2, 8, and 14 were relatively high (C_t values between 14 and 23; ΔC_t values between 3 and 10). Comparative analysis showed that levels of the highly expressed Mmp2 and 14 were similar between the HEp3-lo/diss and HEp3-hi/diss variants, whereas the Mmp8 expression was higher in HEp3-lo/diss cells (Fig. 1C), the latter consistent with reports documenting reciprocal correlation between MMP-8 levels and tumor aggressiveness (24–26). In contrast, the levels of Mmp1 expression were consistently and significantly higher in HEp3-hi/diss cells (P < 0.01), exceeding by 3-fold the levels observed in the HEp3-lo/diss variant (Fig. 1C) and suggesting that this collagenase may contribute to metastatic spread of aggressive HEp3 cells.

The relative MMP-1 production by HEp3-hi/diss and HEp3-lo/diss cells was analyzed by Western blotting of conditioned medium (Fig. 1D). Comparison with the position of the residual zymogen and activated forms of recombinant proMMP-1 allows for a conclusion that MMP-1 is secreted by HEp3 cells mainly as a approximately 52-kDa proenzyme. Characteristically, the production of proMMP-1 gradually decreased with the time HEp3-hi/diss cells spent in culture, concomitantly with the loss of intravasation and metastatic capacity manifested in the resulting HEp3-lo/diss cells (>75 days in culture). Thus, the overexpressed MMP-1 gene and its protein product may determine in part the dissemination phenotype and metastatic fate of epidermoid carcinoma cells.

To examine the specific contribution of MMP-1 to intravasation of carcinomas, HEp3-hi/diss cells were transfected with MMP-1 short hairpin RNA (shRNA) construct, generating a cell line that stably expressed 90% less MMP-1 protein as compared with control shRNA-treated cells (Fig. 2A, left). In addition, MMP-1 silencing was achieved by transient transfection of HEp3-hi/diss cells with siRNA constructs, resulting in a profound (95%–99%) and relatively long-lasting downregulation of MMP-1 protein production (Fig. 2A, right). The lack of major off-target effects of MMP-1 silencing on other MMPs is indicated by similar levels of secreted MMP-2 in HEp3-hi/diss cells treated either with control or MMP-1 shRNA (Fig. 2A, zymograms). Furthermore, the unchanged levels of MMP-2 forms (namely, 68-kDa proenzyme, 64-kDa activation intermediate, and 62-kDa fully activated enzyme; Fig. 2A), are a valid indicator that the activity levels of MMP-14 also have not been affected by downregulation of MMP-1 as MMP-14 is the only naturally expressed MMP responsible for MMP-2 processing and activation (27, 28).

Functional significance of MMP-1 expression for epidermoid carcinoma dissemination was evaluated by using the 2 types of MMP-1–silenced HEp3-hi/diss cells in both spontaneous and experimental metastasis models. MMP-1 silencing with either shRNA or siRNA constructs did not affect the tumorigenic capacity of HEp3-hi/diss cells (Fig. 2B, top). In agreement, control and MMP-1–silenced HEp3-hi/diss transfectants showed no significant difference in proliferation rates in 2-dimensional (2D) cultures and, more importantly, in 3-dimensional (3D) collagen gels (Supplementary Fig. S1). However, the deficiency in MMP-1 protein in HEp3-hi/diss cells caused either by shRNA or siRNA treatments was associated with a substantial 80% to 97% decrease in both CAM intravasation and liver metastasis (Fig. 2B, middle and bottom).

Immunohistologic analysis showed the HEp3-hi/diss tumors originating from control and MMP-1–silenced cells exhibited overall similar morphology. Although MMP-1–deficient tumors manifested a more intense staining for tumor cells at the tumor–stroma border, no significant differences were observed between invasive fronts of control and MMP-1 siRNA tumors (Fig. 2C and D). Interestingly, control CAM tumors, developing from the MMP-1–competent cells, displayed more blood vessels containing intravascular tumor cells as compared with primary tumors developed from HEp3-hi/diss cells treated with MMP-1 siRNA (Fig. 2D), further indicating a difference in tumor cell intravasation capacity. In agreement with Alu-qPCR data on intravasation, quantitative image analysis indeed indicated higher proportion of blood vessels harboring intravascular tumor cells in control, MMP-1–competent tumors (60.4% ± 5.3%) as compared with tumors originating from MMP-1–deficient cells (31.8% ± 2%; P = 0.0008).

We next verified whether specific downregulation of MMP-1 would interfere with extravasation and colonization abilities of HEp3-hi/diss cells in the experimental metastasis model. MMP-1 deficiency due to shRNA or siRNA silencing did not affect either 2-hour vascular arrest or levels of CAM and liver colonization measured at 5 days after cell inoculations (Supplementary Fig. S2A). When analyzed by epifluorescence microscopy, siRNA-treated HEp3-hi/diss cells also showed no difference in the morphology of HEp3-hi/diss colonies in the CAM tissue (Supplementary Fig. S2B). These experimental metastasis findings strongly suggest that MMP-1 is functionally involved in early events in the metastatic cascade.

To show functional contribution of MMP-1 to early steps of metastasis, we examined whether CAM intravasation and liver dissemination of HEp3-hi/diss cells, impaired by siRNA-induced MMP-1 protein deficiency, could be rescued by exogenously supplied human recombinant MMP-1. Sustained MMP-1 deficiency in the HEp3-hi/diss cells was confirmed by Western blot analysis (Fig. 3A), but while it did not affect significantly their ability to form primary tumors, it caused a considerable 60% to 70% decrease in the levels of CAM intravasation and liver metastasis as compared with the cells transfected with control siRNA (P < 0.0001; Fig. 3B). This significant impairment of MMP-1 siRNA-transfected cells in intravasation and liver metastasis was fully rescued by the supplementation of developing tumors with 5–15 nmol of recombinant MMP-1, strengthening the notion that MMP-1 may be a functional contributor to dissemination of SCCs. These in vivo rescuing effects of exogenous MMP-1, which was supplied as a mixture of proenzyme and enzyme species (Fig. 3A, first lane on the left), were completely abolished by a potent MMP inhibitor, GM6001 (P < 0.05), indicating that the functional contribution of MMP-1 requires the proteolytic activity of the enzyme (Fig. 3B). Moreover, supplementation of MMP-1 protein to primary tumors also significantly increased intravasation levels of tumor cells escaping from HEp3-lo/diss primary tumors (Fig. 3C), supporting that MMP-1 deficiency was one of the contributing factors in low dissemination capacity of HEp3-lo/diss cells.

We next analyzed whether MMP-1 produced by HEp3-hi/diss cells functionally contributes to the escape from primary tumors and stromal invasion as these processes are likely prerequisites of tumor cell intravasation. Putative involvement of MMP-1 in stromal invasion was examined in an in vivo stromal invasion model, in which human tumor cells are injected directly into the mesoderm of chick embryos developing ex ovo and allowed to form intramesodermal microtumors (22). Cell escape from these microtumors and invasion of the escaped cells within mesodermal stroma is then visualized by epifluorescence microscopy. In this assay, HEp3-hi/diss cells treated with MMP-1 siRNA constructs showed a diminished escape from microtumors and also showed approximately 45% decrease in the migration distance compared with siRNA control (P < 0.0001; Fig. 4A). The most plausible mechanism for the observed effects would involve diminished collagen degradation mediated by MMP-1, a potent collagenase, which could modify the collagen-rich stroma of the CAM. However, this conclusion is not supported by a complete lack of any inhibitory effects of MMP-1 silencing on the overall adhesive, migratory, and invasive capacities of HEp3-hi/diss cells tested in vitro (Supplementary Fig. S3). Therefore, although the inhibited invasion in vivo might indicate less escape and/or lower motility of MMP-1–depleted HEp3-hi/diss cells escaping from primary tumors, the relatively modest diminishment in stroma invasion is unlikely to account for the substantial 80% to 97% reduction of tumor cell intravasation caused by stable (shRNA) or transient (siRNA) downregulation of MMP-1 (Fig. 2B).

To investigate whether the effects of MMP-1 on tumor cell invasion and intravasation in vivo may be indirect and independent from the putative collagenolytic activity of MMP-1 on matrix degradation, we examined the involvement of MMP-1 in the induction of tumor angiogenesis, an MMP-dependent process that ultimately provides aggressive cancer cells with conduits for vascular spreading. To address this possibility, we evaluated the extent of angiogenic network development within intramesodermal microtumors originating from HEp3-hi/diss cells treated with either control or MMP-1 siRNA constructs. Both tumor types exhibited comparable numbers of CAM blood vessels coalescing toward the border of intramesodermal microtumors (Fig. 4B). However, intratumoral angiogenic vessels networks seem distinctly different between control and MMP-1–silenced microtumors. Thus, more than 70% of control microtumors contained well-developed, perfusable intratumoral angiogenic blood vessels. In contrast, almost 90% of MMP-1–deficient HEp3-hi/diss microtumors failed to develop robust intratumoral angiogenic networks and therefore these intramesodermal microtumors presented with rather thin, poorly perfused blood capillaries (Fig. 4C and D). Thus, the average diameter of 5 most prominent intratumoral blood vessels was 25 μm within individual control microtumors, but reached only 12 μm in microtumors developing from MMP-1–silenced cells (Fig. 4E). More detailed distribution analysis indicated that almost 60% of blood vessels in control HEp3-hi/diss microtumors were of 20 μm or more in diameter, whereas an 80% majority of vessels in MMP-1–deficient microtumors had a lumen diameter less than 15 μm and only 7% of blood vessels were 20 μm or more in diameter (Fig. 4F). Together, these findings suggested that under in vivo conditions, the MMP-1 produced by aggressive HEp3 cells is involved in the regulation of intratumoral angiogenesis and the overall structure of the tumor neovasculature.

The above notion was further investigated in a modification of the spontaneous metastasis assay, which involves the establishment of topical CAM microtumors in ex ovo–incubated embryos by grafting 5 to 6 collagen droplets each containing 1 × 10⁵ cells (Fig. 5A). Importantly, this newly established assay fully reproduced dissemination differentials, manifested by control and MMP-1–deficient HEp3-hi/diss cells in the standard in ovo model, and showed more than 90% reduction in intravasation and liver metastasis by silencing of MMP-1 (Fig. 5B). Notably, this inhibition of tumor dissemination occurred in the absence of significant cell invasion into the surrounding stroma (compare Fig. 5C versus Fig. 4A), further suggesting the involvement of mechanisms other than MMP-1–mediated collagen-remodeling. To probe for this notion, we analyzed possible correlations between the levels of MMP-1–mediated tumor cell dissemination and the structure of intratumoral vessels visualized by epifluorescence microscopy of topical CAM tumors. Both control and MMP-1–silenced microtumors triggered similar robust vascular responses, attracting large and medium size CAM blood vessels, which under low magnification seem to converge onto topically developing microtumors (Fig. 5C). However, quantitative analysis of images acquired at higher magnification deeper into the tumors indicated that almost 90% of control microtumors had well-developed intratumoral angiogenic networks as opposed to less than 40% of siMMP-1 microtumors. Western blot analysis of CAM microtumors excised from individual embryos confirmed that MMP-1 siRNA silencing in HEp3-hi/diss cells was still sustained in microtumor tissue at the end of the 6-day long experiment (Fig. 5D). This differential in MMP-1 expression closely correlated with the pattern and extent of intratumoral vasculature development manifested in a higher density of well-defined, lumen-containing angiogenic vessels within control microtumors (Fig. 5E, top). In contrast, MMP-1–deficient microtumors exhibited a substantially reduced density of lumen-containing vessels, most of which appeared poorly perfused and less interconnected (Fig. 5E, bottom). Furthermore, the angiogenic vessels within the 2 types of microtumors differed significantly in their mean lumen diameter (12.1 ± 0.9 μm in control tumors vs. 6.9 ± 0.3 μm in MMP-1–deficient tumors; P < 0.001; Fig. 5F, bars on the left). In addition, diameter distribution analysis indicated that 25% to 30% of blood vessels in control HEp3-hi/diss microtumors contained lumens with diameter of more than 15 μm, whereas microtumors developing from siMMP-1–treated cells were almost completely devoid of this size of blood vessels (28.3% ± 4.2% vs. 0.8% ± 0.8%; P < 0.0001; Fig. 5F, bars on the right). Importantly, this significant difference in the size and overall quality of the intratumoral angiogenesis is highly comparable with the intravasation and metastasis differentials displayed by the 2 cell types as measured in the same model system (Fig. 5B).

Together, our angiogenesis analyses conducted on live microtumors indicate that MMP-1 may facilitate tumor cell intravasation by inducing development of a dilated intratumoral angiogenic vasculature capable of sustaining tumor cell intravasation, whereas MMP-1 silencing seems to severely compromise vascular microarchitecture by narrowing vessel diameters, which in turn, would prevent tumor cell intravasation.

Tumor cell penetration into vessels via an MMP-1–mediated disruption of the endothelial barrier can constitute an additional mechanism whereby carcinoma cell–produced MMP-1 would facilitate intravasation independent of the collagenolytic activity of MMP-1 and its putative effects on stromal invasion. Because MMP-1–silencing did not affect the overall adhesive, migratory, and invasive capacities of HEp3-hi/diss cells (Supplementary Fig. S3), we further explored the possibility that the ability of tumor cells to cross over an endothelial barrier may not depend on collagenolytic activity of MMP-1. This was investigated in a Transwell assay, which combines measurement of endothelium permeability by Evans blue exudation with coordinated quantification of tumor cell transendothelial migration in an environment free of interstitial collagen (Fig. 6). Control, MMP-1–competent HEp3-hi/diss cells, but not the MMP-1–silenced cells (Fig. 6A), were able to induce substantial Evans blue diffusion (Fig. 6D), strongly implicating carcinoma cell-produced MMP-1 in the disruption of endothelium integrity. This notion was further affirmed when plasmin-activated recombinant MMP-1 (Fig. 6B) was able to rescue dampened levels of endothelial permeability induced in the presence of MMP-1–silenced HEp3-hi/diss cells (Fig. 6D, left). Furthermore, control HEp3-hi/diss cells exhibited approximately 3- to 4-fold higher rates of transendothelial migration as compared with their MMP-1–silenced counterparts, thereby establishing a strong positive link between MMP-1 production by carcinoma cells and their ability to cross endothelial barriers. This link is further reenforced by more than 2-fold increase of transendothelial migration of MMP-1–silenced cells by addition of activated exogenous MMP-1 (Fig. 6D, right).

To verify whether carcinoma-produced MMP-1 disrupted vascular integrity through processes involving endothelial PAR1 activation (29), we used a specific PAR1 antagonist, SCH79797, which was added at 5 μmol/L into Transwell. The PAR1 antagonist significantly and concomitantly diminished the ability of MMP-1–competent HEp3-hi/diss cells to induce endothelial permeability and cross over the transendothelial barrier (Fig. 6D, black bars). Correspondingly, these 2 diminished capacities of the MMP-1–deficient cells were rescued by the addition of exogenous active MMP-1 (Fig. 6D, horizontally striped bars). Because gene expression of PAR1 in the targeted endothelial cells is 10-fold higher than in HEp3-hi/diss cells and has not been affected by siRNA downregulation of MMP-1 (Fig. 6C), these in vitro findings suggest that a tumor MMP-1/endothelial PAR1 axis is likely involved in altering vascular integrity and rendering epidermoid carcinoma cells with the ability to efficiently penetrate through endothelial barriers.

To validate that MMP-1 functionally contributes to the development of an intratumoral vasculature manifesting high levels of endothelial permeability, we compared in vivo the exudation levels of permeable dextran within the topical HEp3-hi/diss microtumors. Thus, tumor-bearing embryos were first, inoculated with TRITC-labeled dextran (155 kDa) and then, after 1-hour incubation, perfused with nonpermeable FITC-conjugated dextran (2,000 kDa). Microtumors were evaluated in a fluorescent microscope for the levels of low molecular weight dextran exudation (red fluorescence) and perfusion of intratumoral vasculature (green fluorescence). The leakage of permeable dextran and the volume of perfusable vasculature in individual microtumors were then quantified fluorometrically.

No exudation of low molecular weight dextran was observed in the CAM tissue distal to the tumors (Fig. 7A, left). However, the permeable dextran was detected at significant levels within control MMP-1–competent microtumors as evidenced by diffuse red fluorescence (Fig. 7A, middle). In contrast, the intensity of red fluorescence was significantly reduced by MMP-1 silencing, (Fig. 7A, right), which could be appreciated more readily in a monochromatic tone (Fig. 7A, bottom). Permeable dextran exudation was quantified in the lysates of individual microtumors either relative to the volume of perfusable intratumoral vasculature or as an independent parameter. This analyses confirmed that specific vascular permeability in MMP-1–deficient tumors was diminished 2.3-fold (P < 0.05) compared with control microtumors (Fig. 7B, top). Furthermore, the comparison of absolute levels of dextran exudation independent of the volume of perfusable vasculature in the individual tumors indicated approximately a 10-fold differential in the overall vascular permeability between control and MMP-1–silenced microtumors (Fig. 7B, bottom).

The PAR1 antagonists, SCH79797 and RWJ56110, previously shown to target endothelial PAR1 in chick embryos (30), were used to substantiate in vivo the link between tumor-produced MMP-1 and PAR1-mediated signaling within intratumoral vasculature. The HEp3-hi/diss cells were grafted on the CAM of the ex vivo chick embryos and the developing microtumors were treated with either of PAR1 antagonists or vehicle. Similar to MMP-1 silencing, SCH79797 caused a significant reduction in the levels of permeable dextran exudation compared with the vehicle-control tumors as depicted visually or quantified fluorometrically (Fig. 7C and D). When related to perfusable vascular volume, specific vascular permeability in SCH79797-treated tumors was diminished 2.4-fold (P < 0.05), whereas the overall levels of dextran exudation indicated a more that 5-fold decrease caused by this specific inhibitor of PAR1 signaling (P < 0.005). Comparable results were obtained for the PAR1 antagonist RWJ56110 (Supplementary Fig. S4), further strengthening the observed inhibitory effects of SCH79797. In addition to significant diminishment of vascular permeability, the treatment with SCH79797 also reduced the levels of HEp3-hi/diss CAM intravasation (Supplementary Fig. S5). Together, these findings are consistent with the notion that tumor-produced MMP-1 functionally regulates endothelial integrity in the intratumoral vasculature via a PAR1 signaling axis.

To corroborate our findings generated with HEp3-hi/diss cells, we have screened a number of human head and neck carcinomas, including FaDu, Detroit 562, SCC-9, SCC-15, SCC-25, and A-253, obtained from the ATCC. From 6 tested cell lines, we chose Detroit 562 carcinoma as these cells produce MMP-1, form primary CAM tumors, and intravasate into the CAM vasculature. To examine the effects of MMP-1 on vascular integrity and cell intravasation of Detroit 562 cells, we downregulated MMP-1 by RNA interference and used the control and MMP-1–depleted cells in our CAM microtumor assay (Supplementary Fig. S6). Complementing our major findings in HEp3-hi/diss cells, downregulation of MMP-1 by siRNA in Detroit 562 cells also resulted in a concomitant decrease of vascular permeability and intravasation, thereby corroborating our notion that carcinoma-produced MMP-1 is involved in the regulation of vascular integrity and tumor cell dissemination.

MMP-1 is one of the most upregulated proteinases in a variety of carcinomas (31–33), but the precise role of this interstitial collagenase in carcinoma progression and metastasis remains unclear. Thus, although MMP-1 efficiently cleaves native triple-helical collagen in vitro, its contribution to matrix degradation and stromal invasion in vivo has been overshadowed by the well-documented role of another collagenase, namely MT1-MMP (MMP-14; refs. 34, 35). In the present study, we have shown that MMP-1 expression correlated positively with the overall metastatic ability of HEp3-hi/diss cells, a highly metastatic variant of the human head and neck SCC, HEp3. In our search for MMPs that could in part be responsible for the substantial dissemination differential between HEp3 variants, we profiled the expression of several MMP genes and identified Mmp1 as the most differentially overexpressed in HEp3-hi/diss cells compared with HEp3-lo/diss cells. Although expression of the Mmp1 gene has been found to be the third best oral SCC tumor predictor (5), we have shown for the first time the direct involvement of MMP-1 enzyme in SCC metastasis. Thus, the ability of HEp3-hi/diss cells to complete intravasation and metastatic dissemination from the primary tumor to secondary organs depended on MMP-1 expression, both at gene and protein levels, strongly suggesting that MMP-1 could be a critical contributor to vascular intravasation in SCCs. Because the intravasation step is the least studied step in the entire metastatic cascade, the clear identification of a contributing molecule in a live animal model is a significant finding.

The functional contribution of the MMP-1 protein to spontaneous dissemination of HEp3-hi/diss cells was confirmed by downregulation of MMP-1 by RNA interference. Importantly, the inhibitory effects of MMP-1 deficiency on metastatic dissemination of MMP-1–silenced cells were reversed by recombinant MMP-1 exogenously supplemented to developing HEp3-hi/diss primary tumors at low nanomolar levels. Furthermore, exogenously supplemented purified MMP-1 substantially increased the intrinsically low-disseminating potential of HEp3-lo/diss cells. Together, our findings validate MMP-1 as a critical proteinase that functionally regulates in vivo the intravasation process of epidermoid carcinoma cells.

Despite the fact that specific silencing of MMP-1 in HEp3-hi/diss cells by 2 RNAi approaches substantially, up to 97%, reduced intravasation and metastatic dissemination, stromal invasion in vivo was reduced only partially. Moreover, when the MMP-1–silenced HEp3-hi/diss cells and their control counterparts were examined for ability to invade across native collagen and basement membrane barriers, no significant reduction in invasion was observed for the MMP-1–silenced cells. These results are in apparent discrepancy with recently published data on the proteolytic role of murine MMP-1a (homolog of human MMP-1) in collagen and Matrigel invasion of murine lung cancer and melanoma cells (36), but are in agreement with the lack of MMP-1–mediated effects on matrix invasion of human tumor cells showed previously by the Weiss laboratory (34, 35, 37). Our findings suggested that other functional catalytic properties of MMP-1, aside from the putative collagen degradation and extracellular matrix remodeling normally associated with an interstitial collagenase, might be involved in the indicated contribution of MMP-1 to vascular intravasation.

Immunohistochemical analysis of primary tumors showed a substantial diminishment in the number of intravascular HEp3-hi/diss cells within tumor-associated blood vessels when carcinoma cell MMP-1 was specifically downregulated, suggesting that MMP-1 activity might be involved directly in the actual blood vessel entry process. This notion, along with the lack of substantial effects of MMP-1 silencing on tumor growth and stromal invasion in the in vivo settings, prompted us to examine the mechanistic involvement of MMP-1 in modulating interactions between tumor cells and intratumoral vasculature. By using our new intramesodermal microtumor model, we found that MMP-1–silenced HEp3-hi/diss microtumors had a thin and poorly perfused blood vessel network. More detailed evaluation by epifluorescent microscopy showed a dramatic difference in the microarchitecture of intratumoral vessels between control, MMP-1–competent microtumors and MMP-1–silenced HEp3-hi/diss microtumors. Thus, MMP-1 downregulation resulted in development of weak and immature tumor vascular networks represented mainly by vessels of 7 to 12 μm in diameter, that is, with a lumen size that apparently would not be amenable for efficient intravasation of the larger carcinoma cells. Importantly, the presence of an altered vascular structure in MMP-1–deficient tumors closely coordinated with the reduced rates and numbers of intravascular tumor cells and the substantially diminished levels of tumor cell intravasation and metastasis. These results further substantiate the functional significance of tumor-derived MMP-1 in regulation of intratumoral angiogenesis, vascular structure, and integrity, and ultimately, the tumor cell capacity for vascular penetration. Overall, our findings show that development of an intravasation-supporting intratumoral vasculature depends on specific mechanisms involving tumor-produced MMP-1, likely independent of its collagenolytic activity. That MMP-1 can be a critical contributing factor to the ability of carcinoma cells to regulate the development of tumor vasculature sustaining intravasation and metastasis, was further corroborated by the inhibitory effects of MMP-1 downregulation in another human SCC, Detroit 562.

Having observed MMP-1–dependent changes in vascular structure in our in vivo metastatic models, we investigated the effects of MMP-1–modulation on endothelial barrier integrity, namely on endothelial permeability during tumor cell transendothelial migration. Control MMP-1–competent and MMP-1–silenced HEp3-hi/diss cells were compared in our dual Transwell assay that uniquely combines quantification of both tumor cell–induced permeability of endothelial barriers and tumor cell transendothelial migration. We have shown for the first time that tumor-produced MMP-1 can regulate the permeability of the endothelial barrier and make it more penetrable for transmigrating tumor cells. It has previously been shown in a nontumor setting that exogenously added recombinant MMP-1 enzyme can induce vascular permeability and endothelial barrier disruption in mice undergoing endotoxin-induced sepsis (29). By using the EA.hy926 endothelial cell line in vitro, the authors delineated a putative MMP-1–involving mechanism and showed that endotoxin-mediated endothelial barrier disruption was mediated by overall PAR1 activation via an induction of Rho-GTP activity (29). A study by Goerge and colleagues also reported that the PAR1 pathway can be activated by conditioned medium containing MMP-1 secreted by tumor cells (38). Furthermore, in a number of recent clinical studies MMP-1 and PAR1 coexpression or colocalization in tumor tissues positively correlated with poor prognosis or unfavorable outcome for 3 different types of human carcinomas (39–41).

The findings of our study unify MMP-1–induced PAR1 activation showed selectively in either tumor cells (42–44) or endothelial cells (29, 45) and also complements PAR1 activation induced in immortalized endothelial cells by conditioned medium harvested from MMP-1–producing tumor cells (38). By direct introduction of MMP-1–producing, but, importantly, PAR1-deficient HEp3-hi/diss cells, to PAR1-positive endothelial cells and measuring endothelial barrier permeability in the presence or absence of a specific PAR1 antagonist, our study for the first time provides direct evidence that the MMP-1/PAR1 axis indeed plays a functional role during tumor cell–endothelial cell interactions in vitro. More importantly, our in vivo studies also indicate that HEp3-hi/diss–produced MMP-1 might mediate the complex effects on intratumoral vascular microarchitecture and permeability via endothelial PAR1.

There are several suggested mechanisms by which PAR1-activation might be involved in increasing vascular permeability. Activation of PAR1 on vascular endothelial cells induces their activation and promotes exocytosis of Weibel–Palade bodies containing the chemokine interleukin 8 (IL)-8 (38). In turn, the released IL-8, via binding to the receptor CXCR2 expressed on tumor cells, can contribute to the disruption of the endothelial barrier by enhancing tumor cell force generation and cytoskeletal remodeling dynamics in tumor cells directly confronting the endothelium (46). In addition, cleaved and activated PAR1 is a strong activator of G_12/13–driven Rho-GTPases, which, in turn, induce actin-dependent contraction of endothelial cells, resulting in a persistently increased endothelial permeability (47, 48). Because platelet MMP-1 can cleave PAR1 at a distinct site that strongly activates Rho-GTP pathways (49), it is plausible that MMP-1 produced by tumor cells can activate PAR1 on confronting vascular endothelial cells, leading to the development of an intravasation-sustaining vasculature that facilitates active entry of the aggressive carcinoma cells that produced the MMP-1.

It remains unknown whether or not the overall process of cancer cell intravasation and metastatic dissemination is directly linked to PAR1-modulated endothelial permeability by tumor-produced MMP-1, but our study suggests that a tumor MMP-1/endothelial PAR1 signaling axis could be a valid molecular target to functionally compromise the intratumoral vasculature during antimetastatic therapy for patients with aggressive epidermoid carcinomas. This contention could be especially important in view that it might not be the surrounding vasculature adjacent to the primary tumor, but rather the intratumoral vasculature that provides major conduits for primary tumor dissemination.

No potential conflicts of interest were disclosed.

Conception and design: A. Juncker-Jensen, E.I. Deryugina, J.P. Quigley

Development of methodology: A. Juncker-Jensen, E.I. Deryugina, L.H. Engelholm, J.P. Quigley

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Juncker-Jensen, E.I. Deryugina, I. Rimann, E. Zajac, J.P. Quigley

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Juncker-Jensen, E.I. Deryugina, I. Rimann, T.A. Kupriyanova, L.H. Engelholm, J.P. Quigley

Writing, review, and/or revision of the manuscript: A. Juncker-Jensen, E.I. Deryugina, J.P. Quigley

Study supervision: E.I. Deryugina, L.H. Engelholm, J.P. Quigley

The authors thank Chenxing Li for her excellent technical assistance. This is manuscript number 21996 from The Scripps Research Institute.

This work was supported by postdoctoral fellowships from the Lundbeck and Willum Kann Rasmussen Foundations (A. Juncker-Jensen); NIH grants R01CA105412 and R01CA129484 (J.P. Quigley); a postdoctoral research fellowship from Nachwuchsförderungskredit/Stiefel-Zangger Foundation, University of Zurich (I. Rimann); and Human Diversity and Re-Entry Award from NIH/NCI (E. Zajac).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.