TIMP-2 Interaction with MT1-MMP Activates the AKT Pathway and Protects Tumor Cells from Apoptosis

Membrane-type 1 matrix metalloproteinase (MT1-MMP), a transmembrane proteinase with an extracellular catalytic domain and a short cytoplasmic tail, degrades a variety of extracellular matrix (ECM) components. In addition, MT1-MMP activates intracellular signaling through proteolysis-dependent and independent mechanisms. We have previously shown that binding of tissue inhibitor of metalloproteinases-2 (TIMP-2) to MT1-MMP controls cell proliferation and migration, as well as tumor growth in vivo by activating the Ras—extracellular signal regulated kinase-1 and -2 (ERK1/2) pathway through a mechanism that requires the cytoplasmic but not the proteolytic domain of MT1-MMP. Here we show that in MT1-MMP expressing cells TIMP-2 also induces rapid and sustained activation of AKT in a dose- and time-dependent manner and by a mechanism independent of the proteolytic activity of MT1-MMP. Fibroblast growth factor receptor-1 mediates TIMP-2 induction of ERK1/2 but not of AKT activation; however, Ras activation is necessary to transduce the TIMP-2-activated signal to both the ERK1/2 and AKT pathways. ERK1/2 and AKT activation by TIMP-2 binding to MT1-MMP protects tumor cells from apoptosis induced by serum starvation. Conversely, TIMP-2 upregulates apoptosis induced by three-dimensional type I collagen in epithelial cancer cells. Thus, TIMP-2 interaction with MT1-MMP provides tumor cells with either pro- or anti-apoptotic signaling depending on the extracellular environment and apoptotic stimulus.


Introduction
Membrane-type 1 matrix metalloproteinase (MT1-MMP, MMP-14), a transmembrane proteinase with an extracellular catalytic site and a 20-amino acid cytoplasmic domain, degrades a variety of extracellular matrix (ECM) components and activates the proenzyme forms of MMP-2 and MMP-13 [1]. Based on these features MT1-MMP has been implicated as a central component of the proteolytic mechanisms of a variety of physiological and pathological

Cells and Culture Media
Human MCF-7 breast adenocarcinoma cells and MCF-7 cells stably transfected with MT1-MMP cDNA under control by the tetracycline resistance transactivator in the Tet-Off conformation have been described [37]. Human MDA-MB-435 melanoma cells were from American Type Culture Collection (ATCC; HTB-129). MCF-7 and MDA-MB-435 cells were grown in DMEM supplemented with 10% fetal calf serum (FCS), 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. Apoptosis Induction MCF-7 cell transfectants were grown in medium containing 10% FCS; after 24 h (day 0) the medium was replaced with serum-free medium supplemented with TIMP-2 with or without the indicated additions. TIMP-2, U0126 or LY294002 were added every other day without changing the medium. The cells were collected at day 7 for Western blotting analysis of PARP degradation. 3D type I collagen gels (2 mg/ml) were made from a neutralized solution of bovine type I collagen (BD Biosciences, San Jose, CA, USA). Cells suspended in medium containing 10% FCS were added to the collagen mix prior to gelling, and gels were cast in 12-well plates. TIMP-2 was added every other day. The cells were recovered from 3D cultures at day 7 by dissolving the gels in 2 mg/ml bacterial collagenase (Sigma-Aldrich). The reaction was stopped with FCS, and the cells were immediately used for further analysis. MDA-MB-435 cells grown on glass coverslips were transiently transfected with MT1-MMP siRNA (as described below) 24 h and 48 h after seeding. Eight hours after the second transfection the culture medium was changed to serum-free medium with or without addition of TIMP-2 (100 ng/ ml). TIMP-2 or an equivalent volume of control medium was subsequently added every 24 h in the following 2 days. The cells were then fixed and stained for characterization of apoptotic nuclei as described below.

Chromatin Morphology Assay
MDA-MB-435 cells on glass coverslips were serum-starved to induce apoptosis as described above. Apoptosis was characterized by staining the nuclei with Hoechst 33342 (Sigma; 10 mg/ ml) to detect the condensed or fragmented chromatin pattern characteristic of apoptotic cells. For each experiment, 300 to 400 nuclei from 5 random fields of each coverslip were examined at high magnification (400X) with a Zeiss Axioskope 2 photomicroscope. The results are expressed as mean number of apoptotic nuclei ± sd / 10 X field, determined in three separate experiments.

Transient Transfection
The plasmids encoding wt and mutant MT1-MMP have been described [37]. The dominant negative mutant of Ras (HRASN17) was a kind gift of Dr. Mark R. Philips (NYU School of Medicine, New York, USA). The constructs (4 μg) were transiently transfected into sub-confluent cells in 6-well plates using 10 μl of Lipofectamine (Invitrogen, Grand Island, NY, USA) according to the manufacturer's instructions. Twenty-four hours after transfection, the cells were incubated with medium containing 0.5% FCS for additional 24 h, and immediately used for the experiments. MT1-MMP siRNA (siGENOME SMART) and control siRNA pools (Dharmacon GE Life Sciences, Piscataway, NJ) were transiently transfected into subconfluent MDA-MB-435 cells in 6-well plates using 5 μl of Lipofectamine according to the manufacturer's instructions. For the AKT activation experiments, twenty-four hours after transfection the cells were incubated with medium containing 0.5% FBS for additional 24 h before being treated with TIMP-2. For the apoptosis experiments the cells were treated as described above (Apoptosis induction).

Densitometry
Quantitative analysis of Western blot bands was performed with ImageJ 10.2 software (National Institutes of Health). Data are shown as mean ± SE of the ratio between the reading of the sample and that of the corresponding loading control.

Statistical Analysis
The data were analyzed using the Students' t test. A p value 0.05 was considered as significant.

TIMP-2 Interaction with MT1-MMP Induces AKT Activation
We have previously shown that TIMP-2 binding to MT1-MMP activates the Ras-ERK1/2 pathway by a proteolysis-independent mechanism [37]. To study the potential role of MT1-MMP and TIMP-2 in AKT activation we used human MCF-7 mammary carcinoma cells stably transfected with MT1-MMP under control by the tetracycline resistance transactivator (Tet-Off) [37]. In the presence of doxycycline (DOX; 1 μg/ml) in the culture medium these cells express virtually undetectable levels of MT1-MMP, like the parental non-transfected cells; removal of DOX induces expression of high levels of MT1-MMP [37,38] (Fig 1A and 1B). Our previous work had shown that ERK1/2 activation in MT1-MMP expressing cells becomes detectable after 5 min and peaks at 15 min of TIMP-2 (100 ng/ml) treatment [37]. Therefore, we incubated our MT1-MMP Tet-Off transfectants with TIMP-2 (100 ng/ml) for 15 min, and characterized AKT activation by Western blotting with antibody to phosphorylated AKT. TIMP-2 treatment of cells devoid of MT1-MMP (grown in the presence of DOX), or MT1-MMP expression in the absence of TIMP-2 resulted in modest AKT activation. However, TIMP-2 treatment of MT1-MMP expressing cells strongly upregulated AKT activation. Similarly, consistent with our previous findings [37], TIMP-2 induced ERK1/2 activation in MT1-MMP expressing cells but had no such effect on cells devoid of MT1-MMP (Fig 1A and 1B). These findings indicated that TIMP-2 interaction with MT1-MMP activates AKT as well as ERK1/2.
To confirm these results we transiently transfected MCF-7 cells with MT1-MMP cDNA or with the empty vector as a negative control. Consistent with our previous results, addition of TIMP-2 to the culture medium strongly induced AKT activation in the MT1-MMP transfectants but had no such effects on the control cells (Fig 1C and 1D). To investigate whether TIMP-2 activation of AKT is a unique feature of MCF-7 cells or results from the high levels of MT1-MMP expressed by the transfected cells, we characterized the effect of TIMP-2 on AKT activation in human MDA-MB-435 melanoma cells, which constitutively express MT1-MMP. To analyze the requirement for MT1-MMP we transfected these cells with MT1-MMP siRNA or control, scrambled siRNA. Forty-eight hours later the cells were treated with TIMP-2, and analyzed for MT1-MMP expression and AKT activation. MT1-MMP expression was reduced by approximately 80% in the cells transfected with MT1-MMP siRNA relative to the control siRNA transfectants (Fig 2). TIMP-2 addition to the culture medium of control siRNA transfectants resulted in increased level of MT1-MMP, an effect that reflects the stabilization and accumulation of the relatively low amount of active MT1-MMP on the cell surface [39]. Consistent with our previous results, addition of TIMP-2 to the culture medium of control siRNA-transfected cells induced rapid (15 min) activation of AKT. Conversely, TIMP-2 had no such effect in MT1-MMP siRNA transfectants, which expressed almost undetectable levels of MT1-MMP (Fig 2).
AKT activation in MT1-MMP expressing cells occurred within 15 min of TIMP-2 addition and high levels of active AKT persisted for at least 2 h (Fig 3), showing that TIMP-2 interaction  [40][41][42][43] and to the K D of TIMP-2 for MT1-MMP (0.77-2.54 nM) [44,45], indicating that the effect of TIMP-2 on AKT activation can occur under physiological conditions. TIMP-2 Activation of AKT Is Independent of the Proteolytic Activity of MT1-MMP TIMP-2 activation of the Ras-ERK1/2 pathway is mediated by MT1-MMP through a proteolysis-independent mechanism [37]. To investigate whether the effect of TIMP-2 on AKT activation is also independent of the proteolytic activity of MT1-MMP, we tested the effect of Ilomastat (GM6001) on AKT activation in MT1-MMP expressing cells. Ilomastat, a broadspectrum MMP inhibitor, inhibits MT1-MMP activity as efficiently as TIMP-2. We reasoned that if the effect of TIMP-2 on activation of AKT results from inhibition of MT1-MMP activity, Ilomastat would also activate AKT in MT1-MMP expressing cells. The results (Fig 4A and 4B), showed that Ilomastat (50 μM) did not induce AKT activation, indicating that TIMP-2induced activation of AKT does not require inhibition of MT1-MMP activity. Moreover, Ilomastat completely blocked AKT activation by TIMP-2, a result consistent with our previous finding that Ilomastat inhibits TIMP-2 activation of ERK1/2 by competing with TIMP-2 for binding to MT1-MMP [37].
To further investigate the involvement of MT1-MMP activity in AKT activation by TIMP-2 we transiently transfected MCF-7 cells with an MT1-MMP mutant devoid of proteolytic activity (E240A). Cells transfected with wt MT1-MMP or the empty vector were used as positive and negative controls, respectively. TIMP-2 induced AKT activation in cells transfected with MT1-MMP E240A as efficiently as in wt MT1-MMP transfected cells but not in the control, empty vector transfectants (Fig 4C and 4D). Thus, the results of these two experiments showed that AKT activation by TIMP-2 requires TIMP-2 binding to MT1-MMP and is independent of the proteolytic activity of MT1-MMP.

TIMP-2 Activation of AKT Is Mediated by Ras
Our previous studies have shown that TIMP-2 binding to MT1-MMP induces Ras activation [37]. Because AKT can also be activated by Ras [52,53], we hypothesized that TIMP-2 activation of Ras results in the downstream activation of both ERK1/2 and AKT. To investigate this hypothesis we transiently transfected MT1-MMP expressing MCF-7 cells with RasN17, a dominant negative mutant of Ras that blocks activation of endogenous Ras [53]. Overexpression of RasN17 inhibited TIMP-2 induction of both AKT and ERK1/2, showing that AKT activation by TIMP-2 is mediated by Ras (Fig 6).

TIMP-2 -MT1-MMP Interaction Induces Prosurvival Signaling through Both the ERK1/2 and AKT Pathways
To investigate the biological role of TIMP-2 -MT1-MMP activation of AKT we characterized its effect on apoptosis. AKT mediates prosurvival signaling in a variety of cell types, and is of particular importance in tumor cells [54]. Therefore, we hypothesized that TIMP-2 interaction with MT1-MMP generates prosurvival signaling. To investigate this hypothesis we serum starved MT1-MMP-expressing MCF-7 cells in the presence or absence of TIMP-2, and characterized apoptosis by Western blotting analysis of PARP degradation, a marker of apoptosis. As shown in Fig 7A and 7B), TIMP-2 dramatically reduced PARP degradation, indicating that TIMP-2 interaction with MT1-MMP generates prosurvival signaling.
To confirm that the effect of TIMP-2 on apoptosis is not a unique feature of our MCF-7 cell transfectants we used the MDA-MB-435 cells transfected with MT1-MMP siRNA described above. As shown in Fig 7C, TIMP-2 strongly inhibited PARP degradation in cells transfected with control siRNA, which constitutively express MT1-MMP, but had no such effect on MT1-MMP siRNA-transfected cells, which express very low levels of MT1-MMP (Fig 2).
In addition to PARP degradation, we also characterized the effect of TIMP-2 on the chromatin condensation and nuclear fragmentation typical of apoptosis. For this purpose we transfected MDA-MB-435 cells with MT1-MMP siRNA or control siRNA, serum-starved them in the presence or absence of TIMP-2, and measured apoptotic nuclei by DAPI staining as  (Fig 7D) were consistent with those obtained by the analysis of PARP degradation. TIMP-2 dramatically decreased the number of apoptotic nuclei in cells transfected with control siRNA but had no such effect on MT1-MMP siRNAtransfected cells, which express virtually no MT1-MMP (Fig 2). Therefore, these results showed that TIMP-2 interaction with MT1-MMP protects tumor cells from apoptosis.
To investigate the prosurvival signaling activated by TIMP-2 -MT1-MMP interaction we characterized the effect of inhibition of the ERK1/2 and AKT signaling pathways on apoptosis. We reasoned that blocking Ras activation, the common upstream activator of the two pathways, as well as the selective inhibition of either the downstream ERK1/2 or AKT pathway should abrogate the anti-apoptotic effect of TIMP-2. In a first set of experiments, cells transiently transfected with RasN17 or control empty vector were serum starved in the presence or absence of TIMP-2. As expected, TIMP-2 strongly reduced apoptosis in the control transfectants (Fig 8A and 8B). However, although after serum starvation apoptosis was downregulated in RasN17-transfected cells, treatment with TIMP-2 did not further reduce apoptosis, indicating that TIMP-2-activated survival signaling requires Ras activation. In a second set of experiments (Fig 8C and 8D), we serum starved the cells in the presence or absence of LY294002, a synthetic PI3K inhibitor that blocks AKT activation, and UO126, a MEK inhibitor that blocks ERK1/2 activation. As expected, both inhibitors upregulated apoptosis, but this effect was non-significantly reversed by addition of TIMP-2, indicating that the prosurvival signaling activated by TIMP-2 / MT1-MMP interaction is mediated by both the ERK1/2 and AKT pathways.

TIMP-2 Interaction with MT1-MMP Activates Pro-or Anti-Apoptotic Signaling Depending on Context
Our observation that TIMP-2 induces prosurvival signaling contrasts with a previous finding that the proteolytic activity of MT1-MMP protects MCF-7 cells from collagen I-induced apoptosis in three-dimensional (3D) culture [55]. MT1-MMP degradation of collagen I abrogates apoptotic signaling generated by collagen I in epithelial cells. Because TIMP-2 inhibits MT1-MMP activity, it is expected to abrogate its prosurvival effect. However, our findings of the prosurvival signaling of TIMP-2 in 2D culture raised the question whether in3D culture the anti-apoptotic or anti-survival effect of TIMP-2 would be prevalent. Therefore, in parallel experiments we induced apoptosis either by growing cells in 3D collagen gel or by serum starvation. Cells expressing or devoid of MT1-MMP were grown in the presence or absence of TIMP-2 either as monolayers in plastic culture dishes in serum-free medium or in suspension in 3D collagen gel in complete medium. Consistent with previous findings [55], growth in 3D collagen induced apoptosis (Fig 9A and 9C) in cells devoid of MT1-MMP, and this effect was  strongly reduced by MT1-MMP expression. Addition of TIMP-2 to the culture medium neutralized the anti-apoptotic effect of MT1-MMP, and increased apoptosis to a level comparable to that of cells devoid of MT1-MMP. In contrast, TIMP-2 dramatically reduced apoptosis, and virtually suppressed it in serum-starved MT1-MMP expressing cells grown on plastic (Fig 9B  and 9D), but had no such effect on cells devoid of MT1-MMP. Therefore, these results showed that TIMP-2 and MT1-MMP can have both pro-and anti-apoptotic signaling depending on the environment and apoptotic stimulus.

Discussion
We have previously shown that TIMP-2 binding to MT1-MMP activates the Ras-ERK1/2 pathway through a proteolysis-independent mechanism. Here we show that TIMP-2 interaction with MT1-MMP activates AKT by a similar mechanism and provides tumor cells with prosurvival signaling.
The data presented show several aspects of the mechanism of AKT activation by TIMP-2. Like ERK1/2, AKT is activated by TIMP-2 concentrations in the range of those found in tissues or biological fluids (10-100 ng/ml, or 0.4-4.0 nM) [40][41][42][43] and similar to the K D of TIMP-2 for MT1-MMP (0.77-2.54 nM) [37,44,56]. TIMP-2-induced activation of AKT is not mediated by inhibition of MT1-MMP proteolytic activity but requires TIMP-2 binding to MT1-MMP. Two findings support this conclusion. Ilomastat, a broad-spectrum MMP, does not induce AKT activation although it inhibits MT1-MMP activity as efficiently as TIMP-2, showing that MT1-MMP activation of AKT is not mediated by inhibition of MT1-MMP activity. Moreover, Ilomastat prevents AKT activation by TIMP-2. This observation is consistent with the wellestablished concept that Ilomastat, as well as other hydroxamic acid-based MMP inhibitors, competes with TIMP-2 for binding to MT1-MMP, and thus prevents TIMP-2 from binding to MT1-MMP [45]. Our previous studies have shown that Ilomastat dose-dependently inhibits TIMP-2 binding to MT1-MMP and blocks TIMP-2 induced activation of ERK1/2 [37]. In addition, TIMP-2 activation of AKT is mediated by proteolytically inactive MT1-MMP (MT1-MMP E240A). We have previously shown that TIMP-2 binds to MT1-MMP E240A mutant as well as to wt MT1-MMP, and that mutant TIMP-2 devoid of MMP inhibitory activity (Ala-TIMP-2) binds to MT1-MMP and activates ERK1/2 as efficiently as wt TIMP-2 [37]. Thus, AKT activation is mediated by TIMP-2 interaction with the MT1-MMP catalytic domain and independent of the proteolytic activity of MT1-MMP.
Our previous studies have shown that the TIMP-2 concentration that activates Ras-ERK1/2 signaling also inhibits MMP-2 activation by the MT1-MMP Tet-Off transfectants we used in the present manuscript [37]. However, these cells do not express MMP-2, and TIMP-2 induction of Ras-ERK1/2 activation occurs in the absence of MMP-2 [37,38], indicating that MMP-2 is not involved in MT1-MMP control of AKT signaling.
In addition to similarities between ERK1/2 and AKT activation by TIMP-2 / MT1-MMP, we also found differences. Although both ERK1/2 and AKT activation are mediated by Ras, TIMP-2 activation of ERK1/2 is inhibited by PD173074, a specific inhibitor of FGFR1, whereas AKT activation is not inhibited. The immediate upstream activator of AKT is phosphoinositide 3-kinase (PI3K), a heterodimeric enzyme consisting of a p85 regulatory subunit that harbors an SH2 domain, and a catalytic p110 subunit. Unlike ERK1/2, PI3K can be activated through different independent mechanisms, either triggered by recruitment of the p85 subunit to a tyrosine kinase receptor and consequent activation of the catalytic p110 subunit or, alternatively, p110 activation can be mediated by Ras independently of p85 [52,53,57]. It is not clear whether the mechanism of election is determined by specific conditions or linked to a specific biological response. We speculate that AKT activation by TIMP-2 / MT1-MMP is mediated by a different, proteolysis-independent mechanism that may result in Ras activation at a location on the plasma membrane different from that of FGFR. This hypothesis implies that MT1-MMP can activate intracellular signaling through multiple mechanisms, and is supported by the wellestablished notion that MT1-MMP interacts with a variety of cell surface and transmembrane proteins that can mediate activation of intracellular signaling [9,46]. Thus, it is possible that AKT activation is mediated by TIMP-2 / MT1-MMP interaction with a transmembrane protein (e.g. growth factor receptor, integrin) other than FGFR1. This hypothesis warrants further investigation.
TIMP-2 binding to MT1-MMP controls several important cell functions by a proteolysisindependent mechanism. Our previous work has shown that MT1-MMP -TIMP-2 interaction upregulates cell proliferation and migration [37]. Our present data show TIMP-2 binding to MT1-MMP induces anti-apoptotic signaling through the proteolysis-independent activation of Ras and both the downstream ERK1/2 and AKT pathways. We also found that inhibition of Ras activation protects the cells from apoptosis induced by serum starvation, indicating that Ras can relay both pro-and anti-apoptotic signaling. This observation may seem inconsistent with the well-established notion that constitutive activation of Ras promotes oncogenesis by stimulating cell proliferation and survival. However, a number of independent studies have shown that ectopic expression or endogenous activation of Ras is a critical step for the initiation of a death program by tumor cells in response to pharmacological or environmental insults such as growth factor (serum) deprivation. The mechanism(s) that determine the switch between the pro-and anti-apoptotic effect of Ras activation remain unclear, and may depend on the balance between Ras isoforms and interaction between different signaling pathways (reviewed in [52]).
Like Ras, TIMP-2 can also have both pro-and anti-apoptotic effects, depending on the cellular microenvironment and apoptotic stimulus. Consistent with previous reports [55], we found that MT1-MMP protects tumor epithelial cells from collagen I induced apoptosis. In the presence of collagen I TIMP-2 upregulates apoptosis as it blocks the protective effect of MT1-MMP. However, in the absence of collagen I (cells grown on plastic) TIMP-2 protects the cells from starvation-induced apoptosis by activating prosurvival signaling through MT1-MMP. Although both experimental conditions are artificial, the growth of cells in 3D collagen gels mimics the setting of invasive carcinoma cells in a stroma, where they become exposed to a pro-apoptotic extracellular matrix. Conversely, cells grown on plastic secrete and adhere to the extracellular matrix they produce, mimicking the condition of epithelial cells adherent to their own basement membrane. In this setting TIMP-2 interaction with MT1-MMP protects the cells from apoptosis by activating prosurvival pathways. This finding can explain the paradoxal observation that, while high levels of MT1-MMP are associated with aggressiveness in a variety of human malignancies [26], [27], high levels of TIMP-2 also correlate with a poor prognosis. Indeed, in some tumors a negative outcome correlates more closely with TIMP-2 than with MT1-MMP levels [28][29][30][31][32][33][34][35], and high TIMP-2 levels in primary breast carcinomas are associated with the development of distant metastases [30,36].
In conclusion, the data reported show a novel biological function of MT1-MMP and TIMP-2 that can have an important role in tumor biology. The pharmacological inhibition of TIMP-2 -MT1-MMP activation of pro-survival signaling can provide a novel approach to the treatment of tumor development and progression.