Tumor cell-produced matrix metalloproteinase 9 (MMP-9) drives malignant progression and metastasis of basal-like triple negative breast cancer

Matrix metalloproteinases (MMPs) have been implicated in diverse roles in breast cancer development and progression. While many of the different MMPs expressed in breast cancer are produced by stromal cells MMP-9 is produced mainly by the tumor cells themselves. To date, the functional role of tumor cell-produced MMP-9 has remained unclear. Here, we show that human breast cancer cell-produced MMP-9 is specifically required for invasion in cell culture and for pulmonary metastasis in a mouse orthotopic model of basal-like breast cancer. We also find that tumor cell-produced MMP-9 promotes tumor vascularization with only modest impact on primary tumor growth, and that silencing of MMP-9 expression in tumor cells leads to an altered transcriptional program consistent with reversion to a less malignant phenotype. MMP-9 is most highly expressed in human basal-like and triple negative tumors, where our data suggest that it contributes to metastatic progression. Our results suggest that MMP9 may offer a target for anti-metastatic therapies for basal-like triple negative breast cancers, a poor prognosis subtype with few available molecularly targeted therapeutic options.


INTRODUCTION
Breast cancer is the most frequently diagnosed cancer in women in the United States, and accounts for the second highest number of cancer deaths [1]. Breast cancers can be classified into subtypes based on expression of histopathological markers or on gene expression profiles; using either approach, different subtypes have markedly different prognostic outlooks [2][3][4][5]. Major differences in survival among breast cancer subtypes derive in part from the availability of appropriate molecularly targeted adjuvant therapies and in part from the intrinsic aggressiveness of the cancer subtypes. While the introduction of tamoxifen and trastuzumab have improved outcomes for patients with hormone receptor-positive and HER2-positive cancers, respectively [6,7], no equivalent targeted agents have yet been identified for breast cancers of the "triple negative" histological subtype (negative for estrogen receptor (ER), progesterone receptor (PR), and HER2) and the molecularly defined basal-like subtype [8,9]. Furthermore, triple negative or basal-like cancers are more often highly invasive, spreading to lymph nodes while primary tumors are still small, and leading to early relapse with distant metastasis [9][10][11]. Outlooks for these high risk groups can be improved by unraveling the mechanisms that enable uncontrolled cancer spread, and by identifying new points of therapeutic intervention particularly for those breast cancer subtypes that are refractory to currently available molecularly targeted therapies.
The matrix metalloproteinases (MMPs) are www.impactjournals.com/oncotarget a family of zinc-dependent endopeptidases with important functions in extracellular matrix remodeling during development and in inflammation and wound repair processes [12,13]. As regulators of the tumor microenvironment, they are also important contributors to cancer initiation, development, and progression via multiple mechanisms [14,15]. Of particular relevance to metastasis, MMPs degrade basement membranes and expose cryptic peptide epitopes in the extracellular matrix, stimulating cellular invasion [14,15]. They also directly modify integrins and other cancer cell adhesion molecules, and proteolytically activate powerful cytokines such as TGF-β, leading to induction of epithelial-mesenchymal transition, a comprehensive phenotypic alteration characterized by enhanced cell motility [14][15][16]. MMPs at the primary tumor site can also release soluble factors into circulation that can help to establish a metastatic niche in distant organs, aiding subsequent colonization by tumor cells [14]. Interestingly, MMPs are produced both by tumor cells themselves and by complicit stromal cells in the tumor microenvironment, in response to reciprocal paracrine stimulatory interactions [17]. Matrix metalloproteinase 9 (MMP9), also known as gelatinase B, is an MMP that is strongly associated with aggressive and metastatic breast cancer. It is one of 70 genes in the Rosetta poor prognosis signature for breast cancer patients [3], and also correlates with poor prognosis in other breast cancer DNA microarray datasets [18,19]. The relevant source of MMP9 as a poor prognosis marker in human cancer remains unclear. Experimental metastasis studies in mouse models in which lung carcinoma or melanoma cells were injected into MMP9-deficient mice showed that host MMP9 promotes metastatic colonization of the lung [20,21]. Metastatic colonization was enhanced in MMP9-knockout mice implanted with MMP9-expressing bone marrow [20], and MMP9 also promotes lung metastasis in the MMTV-PyVT multistage mammary tumor model, potentially deriving from the pro-angiogenic contribution of neutrophil MMP9 in the lungs [22]. Such studies support the concept that MMP9 can promote metastasis when expressed by stromal cells, at least in mouse models. However, several large studies examining MMP9 expression in human tumors by immunohistochemistry have demonstrated that MMP9 is most widely produced by tumor cells, with lower incidence of expression in fibroblasts and immune cells [23][24][25][26]. While MMP9 expression in stromal cells has important prognostic implications in human breast cancer [24,25], the significance of the more substantial expression of MMP9 expression by tumor cells has not been clarified. While in some studies high expression of MMP9 by tumor cells was associated with higher rates of lymph node metastasis [27,28], distant metastasis [25], and poorer relapse-free survival [23,25], other studies have reported associations of tumor cell MMP9 expression with better overall survival or recurrence-free survival [24,26].
Here, by manipulating tumor cell MMP9 expression in breast cancer cell lines and in an orthotopic mouse model of metastatic human breast cancer, we dissect the functional significance of tumor cell-produced MMP9 for invasion and metastasis. Our results specifically implicate tumor cell-produced MMP9 in invasion and metastatic progression in models of basal-like triple negative breast cancer, and suggest MMP9 as a potential therapeutic target to combat metastasis in this poor prognosis breast cancer subtype for which few current molecularly targeted therapies are effective.

Tumor cell-produced MMP9 expression drives invasiveness of basal-like triple negative breast cancer cell lines
We have previously found that tumors derived from MDA-MB-231 human breast cancer cells orthotopically implanted in mice show evidence of gelatinase activity, and that treatment of MDA-MB-231 cells with tissue inhibitor of metalloproteinases-1 (TIMP-1) can reduce cellular invasiveness [29]. Because transcripts for gelatinases MMP2 and MMP9 are both expressed at low but detectable levels in MDA-MB-231 cells, we evaluated the impact of individually silencing MMP2 or MMP9 expression using targeted lentiviral shRNA constructs, to identify the specific gelatinase(s) responsible for promoting invasion (Fig. 1). Both MMP2 and MMP9 expression could be effectively diminished by shRNA knockdown (Fig. 1A,B). MMP2 knockdown reduced cellular invasiveness in Matrigel transwell assays by about 50%, while MMP9 knockdown had a much more substantial effect, reducing invasion by 90% (Fig. 1C), suggesting MMP9 as a major effector of invasive behavior in MDA-MB-231 cells.
To extend these initial observations, we carried out MMP9 silencing experiments using multiple MMP9targeted lentiviral shRNA constructs and additional cell lines. MDA-MB-231 cells are triple negative, and represent an accepted model of the basal-like breast cancer subtype [30,31]. Triple negative breast cancers are often aggressive and have a high risk of early metastatic relapse [8,9]. Hypothesizing that MMP9 may act as a general driver of the invasive/metastatic propensities of triple negative breast cancers, we evaluated the impact of MMP9 knockdown on basal-like, triple negative breast cancer cell lines BT-549 and SUM159PT. Knockdown of MMP9 expression in MDA-MB-231 cells (Fig 2A), BT-549 cells (Fig. 2B), and SUM159PT cells (Fig. 2C) led to consistent suppression of invasiveness in Matrigel transwell assays (Fig. 2D,E, F, respectively). www.impactjournals.com/oncotarget  In this model, primary tumor growth could be followed in vivo by bioluminescence imaging (Fig. 3A), which detected evidence of metastasis by 11 weeks after tumor cell implantation (Fig. 3B). Ex vivo bioluminescence imaging (Fig. 3C) and immunohistochemistry (Fig. 3D) confirmed the presence of pulmonary metastases in mice for which metastasis was detected in vivo.
Mice were implanted with 1×10 5 MDA-MB-231-luc2 cells stably transfected either with a nontarget control virus (NT) or with an MMP9-targeted lentiviral shRNA construct (KD). Mice were monitored regularly by in vivo bioluminescence imaging and euthanized at 11 weeks post-implantation. Ex vivo imaging showed that all of the control mice had evidence of pulmonary metastasis, whereas none of the mice bearing tumors in which MMP9 was silenced showed any bioluminescence signal in the lungs (Fig. 3E); this difference in outcomes was highly significant (p=0.0079; Fisher exact test). The extent of pulmonary metastasis in each mouse was quantified by bioluminescence flux (Fig. 3F) and by histopathological and histochemical analysis of a single section through all lung lobes (Fig. 3G), each of which likewise showed significant differences between groups.

Tumor cell-produced MMP9 promotes vessel formation in an orthotopic xenograft model of basal-like triple negative breast cancer
Excised primary tumors from the above experiment were also examined for effects of MMP9 suppression. Tumor sections stained for MMP9 protein expression showed a trend of lower overall stain intensity in the MMP9 KD group, although there was considerable variation within each group (Fig. 4A,B), suggesting that the knockdown was not uniformly maintained throughout the latter part of the 11 week experimental time course.
MMP9 has been implicated previously as a critical mediator in the processes of tumor angiogenesis and vasculogenesis [32,33], and so we also assessed the extent of tumor vascularization by staining the tumors for endothelial cell marker CD31. The tumors of the MMP9 KD group had significantly reduced staining compared to the control group (  Furthermore, the CD31 staining pattern showed that the vascular morphology also differed between the two groups; whereas control tumors possessed properly formed blood vessels with structured lumina, the MMP9 KD tumor vasculature was more disorganized and often lacked lumina (Fig. 4D). Surprisingly, the reduction in blood vessel density did not appear to correlate with a reduction in tumor growth, as the average tumor weight (Fig. 4E) and ex vivo bioluminescence signal (Fig. 4F) were only slightly lower in the group with tumor cell MMP9 KD (not significant).

MMP9 is associated with a tumorigenic expression profile in MDA-MB-231 cells
To investigate how the tumor cell-produced MMP9 drives the invasive/metastatic phenotype of triple negative/ basal breast cancer cells, we performed transcriptional profiling of the MDA-MB-231 cells transduced with nontarget and MMP9 KD lentivirus, and found substantial alterations associated with MMP9 KD (Fig. 5A). We found that 1423 transcripts were differentially regulated (p<0.05, FC>2, annotated expression data in Supplemental Table 1). We subjected the list of differentially expressed genes to a NextBio meta-analysis [34], and found significant overlap with datasets comparing metastatic vs. nonmetastatic breast cancers (  Table 6). These results suggest that the MMP9 that is expressed in these triple negative breast cancer cells activates a wide variety of pro-tumorigenic responses. Consistent with these results, Ingenuity Pathway Analysis (IPA) of the differentially expressed genes (Supplemental Table 7) identified a top-ranked interaction network that included a prominent nexus of genes associated with MMPs and stromal molecules (Fig. 5G; Supplemental Table 7). Using qRT/PCR, we validated transcriptional alterations in three of the most highly regulated genes in this network: the forkhead transcription factor FOXQ1 (Fig. 5H), the urokinase-type plasminogen activator (PLAU, Fig. 5I), and the BRCA1 interacting protein C-terminal helicase 1 (BRIP1, Fig. 5J).

MMP9 is most highly expressed in Basal-like and ER-negative breast cancers
Given our results supporting a role for tumor cell-produced MMP9 in the invasion and metastasis of cell lines derived from human basal-like triple negative breast tumors, we hypothesized that MMP9 may be particularly highly expressed in these types of cancers. By interrogating a large meta-analysis of published breast cancer microarray datasets [35], we found that MMP9 expression was most highly elevated in basal-like breast cancers, as classified according to the intrinsic subtypes of Hu et al. [36] (Fig. 6A) or the PAM50 subtypes [37] (Fig. 6B), when compared to other molecular subtypes (p<0.00001). This analysis also revealed that estrogen receptor (ER) negative tumors express a significantly higher level of MMP9 than do ER positive tumors (p<0.00001) (Fig. 6C). In assessing a potential association between MMP9 and tumor grade, we found that MMP9 expression was associated with high grade tumors (p= <0.00001) (Fig. 6D). These results, in combination with our findings using cell culture and mouse models to functionally implicate tumor cell-produced MMP9 in invasion and metastasis, suggest that the aggressive phenotype and poor outcome generally associated with basal-like triple negative breast cancers [8,10] are in part driven by MMP9. Taken together, our results suggest MMP9 as a potential target for antimetastatic therapies for this particular patient subset.

DISCUSSION
MMPs as a group are often considered to derive mainly from the stroma, although immunohistochemical and in situ hybridization studies show considerable diversity in staining patterns among the MMPs. In breast cancer, multiple large IHC studies have found MMP9 to be most consistently expressed by tumor cells, whereas MMP9 staining of stromal cells is rarer [23][24][25][26]. However, the functional implications for progression and metastasis of tumor cell-produced MMP9 have remained obscure, as previous experimental studies measuring metastasis as an endpoint have employed transgenic models best suited for probing the role of stromal cell-produced MMP9. In this study, we have identified MMP9 expression by tumor cells themselves as a critical factor mediating the invasion and metastasis of basal-like triple negative breast cancers. In our in vivo model, the impact of MMP9 knockdown in tumor cells on spontaneous metastasis was striking, resulting in a complete blockade of pulmonary metastasis. Consistent with prior studies implicating MMP9 in angiogenesis and vasculogenesis, we also found that tumor cell-produced MMP9 was a significant contributor to the vascularization of the orthotopic tumors, although the impact of MMP9 knockdown on primary tumor growth was surprisingly modest. It may be that the fewer blood vessels present in the MMP9 knockdown tumors were still functional and sufficient to support tumor growth.
The source of MMP9 found in human breast tumors has been seen to have significant bearing on its prognostic interpretation. Studies that evaluated the prognostic value of stromal MMP9 found significant associations with shorter recurrence-free survival [24,25], while the prognostic interpretation of tumor cell MMP9 has remained more ambiguous. Some studies have found such expression to be a negative prognostic factor [23,25,27,38], while others have reported associations between MMP9 expression and more favorable outcomes in certain patient subgroups [24,26]. Our present findings suggest triple negative breast cancer specifically as a focus for future clinical studies examining the prognostic significance of tumor cell MMP9, as it may be that the pathways by which tumor MMP9 drives progression and metastasis are most activated or most necessary for this breast cancer subtype.
Why should the cell from which MMP9 is derived be important for its role in cancer progression or for its prognostic significance? One possibility is that tumor cell MMP9 may carry out specialized functions made possible by specific localization, through generation of high local   [35], and included 1,881 patients from 11 distinct studies. For each plot, MMP9 expression differences between groups is highly significant (p<0.00001).
concentrations at a key point of action. Furthermore, although a soluble secreted enzyme, MMP9 possesses protein-protein binding domains that can serve to tether it to the cancer cell surface, facilitating spatially discrete and directional activities [39]. MMP9 can be localized to the surface of murine mammary carcinoma cells via the hyaluronan receptor CD44 [40], as a component of tumor cell invadopodia structures [41], where it can facilitate invasion and angiogenesis by proteolytic activation of latent TGF-β [42]. CD44-bound MMP9 may also possess migration-promoting signaling activities that are distinct from its proteolytic activity [43]. In other types of tumor cells, binding of MMP9 to specific integrins can promote cell migration [44,45]. Further studies will be needed to unravel the mechanisms by which tumor cell-produced MMP9 drives invasion and metastatic progression specifically in models of basal-like triple negative breast cancer.
As a first step toward elucidating the mechanisms by which basal-like triple negative cancer cell MMP9 is involved in invasion and metastasis, we carried out transcriptional profiling to compare MDA-MB-231 cells expressing endogenous levels of MMP9 versus cells in which MMP9 expression was silenced by RNA interference. MMP9 silencing resulted in broad and comprehensive transcriptional alterations, suggesting that the blockade of metastasis results from a complex transformation in cellular phenotype from a more malignant to a less malignant state. MMP9 has primarily been viewed as a direct effector of cellular motility, both through cleavage of ECM molecules and of cellcell and cell-ECM attachments. Our results suggest that the autologous effects of MMP9 on the intrinsic breast cancer cell phenotype may be even more far-reaching, that MMP9 may act upstream of many genes as a master regulator of the malignant phenotype. We identified and validated several known metastasis mediators as regulated by MMP9 (Fig. 5), including Foxq1 [46] and PLAU [47], as well as the suspected breast cancer susceptibility gene BRIP1 [48]. The mechanisms by which MMP9 orchestrates this comprehensive phenotypic switch are likely to involve outside-in cell surface signal transduction, where the role of MMP9 has been increasingly appreciated in recent years [49].
Breast cancer prognosis varies widely among subtypes, and while outcomes have improved in recent years for estrogen receptor (ER)-expressing and HER2overexpressing tumors, similar gains have not been achieved for patients with triple negative breast cancers lacking these molecular targets [9], most of which also belong to the poor prognosis basal-like intrinsic subtype [10,50]. For patients with triple negative disease, the majority will have residual disease after initial treatment, a high risk of metastatic relapse within the first few years, and a median survival of only 1 year after relapse [8,9]. Treatment options for these patients are limited following progression after standard chemotherapy regimens [8,9], and new molecular targets and targeted therapies are urgently needed. Our data implicating tumor cell-produced MMP9 as a molecular driver of metastatic progression in models of basal-like triple negative breast cancer, in combination with the meta-analyses showing that MMP9 is most highly upregulated in these types of tumors, suggests MMP9 as a druggable target that may prove useful in these patients for whom established targeted therapies are of minimal benefit.
MMPs are not new as therapeutic targets, and past clinical trials of early MMP inhibitors proved disappointing [51,52]. The broad-spectrum MMP inhibitors previously trialed in breast cancer produced serious dose-limiting musculoskeletal toxicity, failed to reach therapeutic plasma levels, and did not extend survival [53][54][55], likely due in part to the inhibitors' inability to distinguish among MMPs [56,57]. This was a critical problem, because some MMPs serve a primarily protective function, and indiscriminate inhibition of all MMPs can lead to poorer outcomes [58][59][60][61]. In the decade that has passed since the abandonment of early generation MMP inhibitors as experimental cancer drugs, many novel and innovative approaches to more selective MMP9 inhibition have emerged. These include a new family of mechanism-based small molecule inhibitors of gelatinases [62,63], as well as peptides and small molecules that inhibit subsets of MMP9 substrate or ligand interactions by blocking exosites on the collagen-binding domain [45,64] or hemopexin domain [18,45,65,66]. Candidates for development of therapeutic proteins include neutralizing monoclonal antibodies targeting the MMP9 catalytic domain [67,68], and recombinant engineered versions of natural tissue inhibitors of metalloproteinases (TIMPs) [29,69]. Recent comparative structural analyses of MMPs and TIMPs highlight new concepts underlying molecular specificity and offer promise for future development of yet more selective biologic inhibitors [70][71][72][73][74]. As future mechanistic studies reveal in greater detail the specific targets or interactions of tumor cell MMP9 that are essential for driving metastasis in basal-like triple negative cancers, it is anticipated that an optimal inhibitory approach can be selected and carried forward into clinical development, for the potential benefit of patients suffering from this challenging disease.
Conditioned media containing infective lentivirus particles were produced using HEK 293FT cells following supplier protocols. For lentiviral transduction, MDA-MB-231-luc2, SUM159PT, or BT-549 cancer cells were seeded in 6 well plates, and then 0.6 mL regular medium, 0.4 ml of lentiviral medium and 0.6 µL of polybrene (10 mg/ mL) were added. The medium was changed after 24 h and after another 24 h cultures were subjected to puromycin selection. Pooled transduced cells were maintained under puromycin selection for up to a week before use in experiments.

RNA extraction, cDNA synthesis, and quantitative real-time PCR
RNA was isolated from cultured cells using TRIZol reagent (Invitrogen) according to manufacturer protocols. cDNA was synthesized according to kit specifications using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Quantitative real-time PCR was performed using TaqMan gene expression assays (Applied Biosystems) on an Applied Biosystems 7900HT Fast Real-Time PCR System according to manufacturer protocols. TaqMan assays employed included: GAPDH, Hs99999905_m1; MMP9, Hs00957555_m1; MMP2, Hs00234422_m1; FOXQ1, Hs00536425_s1; PLAU, Hs00170182_m1; and BRIP1, Hs00230743_m1. Data were analyzed using SDS RQ Manager Software (Applied Biosystems), using GAPDH as an endogenous control for normalization.

Matrigel transwell invasion assays
Cellular invasion assays were similar to our previously published protocols [29,75] with minor modifications. Briefly, BD Falcon 24-well cell culture inserts (8.0 µm) were coated with 50 µg Matrigel basement membrane matrix in 100 µL of serum free medium appropriate for the cell type (EMEM for MDA-MB-231-luc2 cells; Ham F-12 media for SUM159PT cells; RPMI 1640 for BT-549 cells) and placed at 37 °C for 4 h, and then residual medium was aspirated and replaced with cells (2.5×10 4 MDA-MB-231-luc2 or BT-549 cells or 5×10 4 SUM159PT cells per well) suspended in 300 μL of the appropriate serum free medium supplemented with 0.1% BSA. The lower invasion chambers contained 750 µL/well of NIH/3T3 cell-conditioned serum free medium (DMEM supplemented with 50 µg/mL ascorbic acid) as chemo-attractant. Assays were incubated 18 hours at 37°C in 5% CO 2 . Non-invading cells were removed from the insert by scrubbing with a cotton swab, and then cells on the lower surface of the filter were fixed with methanol, stained with crystal violet, and counted using Image-Pro 6.3 software (Media Cybernetics) as previously described [75].

Mouse orthotopic tumor model and bioluminescence imaging
Animal Studies were conducted in accordance with recommended guidelines and approval of the Mayo Clinic Institutional Animal Care and Use Committee (IACUC) (protocol A12409). Prepared MDA-MB-231-luc2 cells (1×10 5 cells per animal), in a total volume of 100 µl of 50% Matrigel/50% medium per animal, were kept on ice until surgical implantation. Six week old female Nod/Scid mice were anesthetized and a small ventral incision was made to visualize the left 4 th mammary fad pat. Cells were implanted at the blood vessel bifurcation near the lymph node. Tumor growth and metastasis were monitored each week over the time course by intraperitoneal injection with 150 mg/kg of D-luciferin and bioluminescence imaging (IVIS Spectrum 3D imaging system, Caliper Life Sciences) essentially as described previously [75]. At 11 weeks mice were injected intraperitoneally with luciferin and euthanized by CO 2 asphyxiation, and then the tumors and lungs were excised, imaged ex vivo for bioluminescence signal, and formalin fixed and paraffin embedded for immunohistochemistry staining.

Transcriptional microarray and NextBio analysis
Isolated total RNA was assessed with Affymetrix U130A gene expression chips which were processed and normalized via GCRMA. Duplicate technical replicates were averaged and then analyzed using Genespring GX via t-tests using previously described methods [76,77]. Thus, results presented are averages of two separate experiments performed in duplicate. Nextbio (www. nextbio.com) meta-analysis [34] was performed to identify similarities between our dataset and published datasets as previously described [78]. Ingenuity Pathway Analysis (IPA) was performed using the web interface (www. ingenuity.com) to build gene interactions associated with expression of MMP9.

Meta-analysis of human breast cancer expression data
The online platform Gene expression-based Outcome for Breast cancer Online (GOBO), including data from 1,881 patients and 11 studies employing Affimetrix U133A microarrays, was used for metaanalysis of associations between MMP9 expression and other histopathological and molecular classifications (http://co.bmc.lu.se/gobo/) [