Functional relationship between matrix metalloproteinase-11 and matrix metalloproteinase-14

MMP-11 is a key factor in physiopathological tissue remodeling. As an active form is secreted, its activity must be tightly regulated to avoid detrimental effects. Although TIMP-1 and TIMP-2 reversibly inhibit MMP-11, another more drastic scenario, presumably via hydrolysis, could be hypothesized. In this context, we have investigated the possible implication of MMP-14, since it exhibits a spatiotemporal localization similar to MMP-11. Using native HFL1-produced MMP-11 and HT-1080-produced MMP-14 as well as recombinant proteins, we show that MMP-11 is a MMP-14 substrate. MMP-14 cleaves MMP-11 catalytic domain at the PGG(P1)-I(P1′)LA and V/IQH(P1)-L(P1′)YG scissile bonds, two new cleavage sites. Interestingly, a functional test showed a dramatical reduction in MMP-11 enzymatic activity when incubated with active MMP-14, whereas inactive point-mutated MMP-14 had no effect. This function is conserved between human and mouse. Thus, in addition to the canonical reversible TIMP-dependent inhibitory system, irreversible MMP proteolytic inactivation might occur by cleavage of the catalytic domain in a MMP-dependent manner. Since MMP-14 is produced by HT-1080 cancer cells, whereas MMP-11 is secreted by HFL1 stromal cells, our findings support the emerging importance of tumor-stroma interaction/cross-talk. Moreover, they highlight a Janus-faced MMP-14 function in the MMP cascade, favoring activation of several pro-MMPs, but limiting MMP-11 activity. Finally, both MMPs are active at the cell periphery. Since MMP-14 is present at the cell membrane, whereas MMP-11 is soluble into the cellular microenvironment, this MMP-14 function might represent one critical regulatory mechanism to control the extent of pericellular MMP-11 bioavailability and protect cells from excessive/inappropriate MMP-11 function.

MMP-11 exhibits a sharply defined spatial and temporal expression pattern, suggesting a local and rapid function [9]. MMP-11 possesses a cleavage site for intracellular furin-type convertases, between its prodomain and catalytic domain. Thus, an active form of MMP-11 is secreted, while several other MMPs need to be extracellularly processed to be functional [15,16]. Therefore, MMP-11 enzymatic activity should be drastically downregulated once the enzyme is no longer needed. Indeed, its inactivation might represent a critical step in the regulation of its function. It is generally accepted that MMPs are negatively regulated by a family of four endogenous proteins called TIMP-1 to TIMP-4 [17,18]. Indeed, TIMP-1 and TIMP-2 inhibit MMP-11 function. However, TIMPs are not specific and TIMP-mediated inhibition is reversible. A more permanent scenario may exist for tightly downregulating MMP-11 bioactivity. Interestingly, an MMP-dependent MMP-11 processing has been reported, but the nature of the MMP involved remains unknown [19]. MMP-14 might be a good candidate. Similar to MMP-11, MMP-14 plays an essential role during normal tissue remodeling processes as well as during benign and malignant diseases [20][21][22].
In the present study, we investigated the possible functional relationship between MMP-11 and MMP-14. We demonstrate that MMP-11 is a MMP-14 substrate and identify two cleavage sites on MMP-11. Interestingly, cleavage leads to the loss of MMP-11 enzymatic activity, highlighting a way to irreversibly inactivate MMP-11 and a novel negative regulator function for MMP-14.
We also checked the ability of wild type or point-mutated recombinant cat-MMP-14 to activate endogenous MMP-2 produced by U87MG cell line (American Type Culture Collection), using zymography. Cell culture medium incubated  with either cat-MMP-14 or cat-MMP-14 m (4 h, 37°C) where loaded onto a 10% polyacrylamide gel containing 0.1% gelatin (Invitrogen Corp., Carlsbad, CA) and separated by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE). After electrophoresis, the gel was renatured in 2.5% Triton X-100 solution at room temperature for 30 min with gentle agitation, and incubated in developing buffer (50 mmol/L Tris-HCl, pH 7.6; 200 mmol/L NaCl, 5 mmol/L CaCl 2 ) at room temperature overnight with gentle agitation. Transparent bands of gelatinolytic activity were visualized by staining with 0.5% Coomassie Blue R250.

Mass spectrometry
Protein bands were in-gel reduced, alkylated, and trypsin digested. The extracted peptides were dried (SpeedVac 5301 Concentrator Eppendorf, Hambourg, Germany), resuspended (3 lL 0.1% HCOOH), injected in the nano liquid chromatography, and analyzed on a linear ion trap LTQ XL ETD with nanoESI source interfaced to a nanoLC system (Ultimate 3000; Thermo Fisher Scientific, Germering, Germany). Samples were desalted and concentrated (reverse phase precolumn 300 lm i.d. 9 5 lm, 3 min, 20 lL/min), and separated (Pepmap C18, 15 cm 9 75 lm i.d). MS/MS spectra were recorded in the data-dependent mode on the five most intense ions observed in the mass range m/z 350-1600. A micro zoom scan was acquired to determine the peptide charge. Parameters for acquiring collision induced dissociation MS/MS spectra: activation time = 30 msec, activation Q = 0.25, relative collision energy = 35%, and an isolation width of 3 m/z. ETD MS/ MS spectra were acquired using: supplemental activation enabled, reaction time = 100 msec, reagent automatic gain control target = 1 E5 ion counts, full AGC target = 3 E4, isolation width of 3 m/z. Proteome Discoverer 1.3 (Thermo Fisher Scientific) with Sequest â search engine was used for spectra preprocessing protein ID using the Mouse or the Human Swissprot database (version 15.7). Sequest â (San Jos e, CA) results were filtered with Xcorr versus charge state 1.5-1, 2.5-2, 3-3, 3.2-4.

Results and statistical analyses
For each study, at least three independent experiments were performed. All of them gave similar results. Moreover, MMP-11 and MMP-14 enzymatic activity analyses were performed in triplicate and statistical differences were evaluated using student's t test. P values lower than 0.05 were considered as significant. *P < 0.05; **P < 0.01; ***P < 0.001.

MMP-11 and MMP-14 exhibit similar spatiotemporal expression at the tumor invasive front
We hypothesized that a functional relationship might exist between MMP-11 and MMP-14. This could be very important in malignant conditions since MMP-11 was shown to play an essential role during the early steps of cancer cell invasion of adjacent connective tissues [1,5,6]. MMP-14 has been reported to be expressed by both stromal and cancer cells [22,23]. Thus, we checked for the presence of MMP-14 in the MMP-11-expressing area. MMP-11 and MMP-14 immunohistochemistry was performed on serial sections of human breast cancer biopsies. As expected from previous data [6], peritumoral fibroblasts and cancer-associated adipocytes (CAAs) that are adipocytes located at the cancer cell contact [31], expressed MMP-11 at the tumor invasive front. Interestingly, MMP-14 staining gave similar images (Fig. 1A). This was observed in all samples examined (more than 20) indicating that this is a common event during local invasion.
HT-1080 and HFL1 cell coculture leads to MMP-11 decrease Since MMP-11/MMP-14 interaction is quite impossible to study in vivo, we presumed that similar dynamic events should occur in vitro. We used HT-1080 fibrosarcoma and HFL1 nontumoral stromal cell lines grown in serum-free conditions (Fig. 1B). As expected, immunofluorescence experiments showed MMP-14 at the plasma membrane of HT-1080 cells and MMP-11 in the secretory pathway of HFL1 cells (Fig. 1C). HFL1 cells were boosted for MMP-11 expression by TPA treatment [16] (Fig. 1D). Western blot analyses of HFL1 cell culture medium (CM; containing secreted MMP-11) using the 5ST4A9 antibody directed against human MMP-11 showed two bands (55 kDa and 47 kDa) corresponding to the pro-and mature forms of MMP-11, respectively [16]. Interestingly, coculture of the two cell lines led to a dramatic decrease in MMP-11 compared with the levels observed when HFL1 cells were grown alone, indicating the occurrence of MMP-11 proteolysis (Fig. 1D). Then, HT-1080 whole cell lysate (CL; containing membrane-bound MMP-14) and HFL1 CM from cells grown alone were incubated together for varying times. Western blot analysis with the 3H3 antibody recognizing MMP-14 showed no decrease in MMP-14, indicating that HFL1 CM did not contain enzyme(s) affecting MMP-14. However, more than 50% decrease in pro-MMP-11 and MMP-11 was seen in the presence of HT-1080 CL. This effect was already complete after 1 h of incubation (Fig. 1E).
Altogether, these data identify MMP-11 as a substrate of MMP-14.
Recombinant Cat-MMP-14 cleaves HFL1produced MMP-11 To further study the MMP-14 cleavage, we produced the human MMP-14 recombinant catalytic domain (Cat-MMP- 14), and test its activity against HFL1-produced MMP-11. A constant quantity of HFL1 CM was first treated with vehicle alone or with a constant Cat-MMP-14 quantity for various times. Western blotting analysis of the resulting mixtures showed that the pro-MMP-11 and MMP-11 bands progressively disappeared with increasing incubation times. The pro-form was completely cleaved after 120 min of reaction. Concomittantly a lower band recognized by the 5ST4A9 antibody appeared at 32 kDa whose intensity increased and became optimal after 120 min. Half of its production was shown after 45 min (Fig. 3A).

MMP-11 contains two cleavage sites for MMP-14
The cleavage sites were investigated using recombinant proteins. Silver nitrate staining showed one main band at 47 kDa for FL-MMP-11, and at 22 kDa for Cat-MMP-14, as reported [24][25][26][27]. FL-MMP-11 incubation with Cat-MMP-14 in a stoichiometric manner for increasing times led to three main bands numbered F1-F3 at 36, 29, and 19 kDa, respectively (Fig. 4A). They already appeared at 2 min incubation and their intensity increased with incubation time. F1 and F2 bands were obviously MMP-11related (sizes higher than Cat-MMP-14). F3 analysis by nano liquid chromatography-mass spectrometry using a LTQ XL ETD mass spectrometer (Thermo Fisher Scientific, San José, CA) showed that it corresponded to MMP-11 fragment (data not shown) confirming that MMP-14 cleaves MMP-11 but the reverse is not true. Consistently, western blot analysis of sample mixtures showed a decrease in the FL-MMP-11 band (Fig. 4B). Then, a con- ratio. Setting the band intensity obtained for the highest Cat-MMP-14 concentration at 100%, half of the F1-F3 bands were produced for 1:2 to 1:1 MMP-11:MMP-14 molar ratio (Fig. 4C). Cat-MMP-14 m had no effect ( Fig.  4B and D). Thus, MMP-11 cleavage by MMP-14 leads to three major protein fragments indicating the presence of two cleavage sites.

MMP-14 cleaves MMP-11 at its catalytic domain
Cleavage site motif for a MMP involves residues both Nand C-terminal to the scissile bond. Human F1-F3 fragments were subjected to Edman sequencing analysis to identify their N-terminus sequences (N-ter). The N-ter of F1, F2, and F3 was ILALAFFRL, LXGQPXPTVT, and MFVLSG, respectively (Fig. 5A). These data allowed to map the two MMP-14 cleavage sites along the MMP-11 sequence, between glycine 175 and isoleucine 176, and between histidine 255 and leucine 256. Interestingly, both PGG(P1)-I(P1 0 )LA and VQH(P1)-L(P1 0 )YG sites are located in the catalytic domain of MMP-11. Similar mouse MMP-11 F1-F3 fragments were generated by Cat-MMP-14 cleavage (data not shown). Sequence analysis and Edman determination showed that MMP-14 cleaved the mouse MMP-11 between glycine 179 and isoleucine 180, and between histidine 259 and leucine 260 revealing  Virtually all MMPs form tight 1:1 complexes with the TIMPs. The structure of the catalytic domain of MMP-14 in complex with TIMP-2, a natural MMP-14 inhibitor, is known [32]. We used it to guide a rigid body positioning of MMP-11 [33] inside the MMP-14 active site. Interestingly, molecular modeling showed the superposition of residues of the cleavable MMP-11 PGG-ILA sequence and TIMP-2 (Fig. 5B). Thus, TIMP-2 residues CSCSP placed in the active site of MMP-14 are shown to superpose to MMP-11 residues GILAH (179-183). It also indicated that conformational flexibility of MMP-11, particularly on the N-terminal side of the cleavage site, is a prerequisite to the cleavage of MMP-11 by MMP-14. Conformational flexibility of MMP's has been documented [34,35]. Thus, an nuclear magnetic resonance study of human stromelysin-1 [34] indicated that in the absence of inhibitor, the protein is highly flexible. Interestingly, the structure of the catalytic domain of MMP-11 indicates that  this cleavable site is situated in the b strand IV, immediately C-terminal to the S-shaped loop [33] and the normal mode calculations indicate that this region displays important conformational flexibility (Fig. 5C). These structural insights prompted us to further investigate the importance of the G and I residues located in P1 and P1 0 of the PGG-ILA cleavage site for cleavage efficiency. We produced recombinant FL-MMP-11 mutant where we replaced GI by AV (FL-MMP-11/AV) and tested the activity of Cat-MMP-14 on this mutant. Interestingly, western blot analysis showed that FL-MMP-11/AV cleavage was dramatically reduced compared to FL-MMP-11 (Fig. 5D). Together, these results showed that the GI sequence is essential for MMP-11 cleavage by MMP-14.
Thus, MMP-14 cleaves MMP-11 within its catalytic domain, and this function is conserved between human and mouse.

MMP-14 inactivates MMP-11 enzymatic activity
We first checked if cleavage by MMP-14 affects human MMP-11 enzymatic function using the a1-PI quantitative colorimetric assay. The Cat-MMP-11 and Cat-MMP-14 catalytic domains were used alone or after preincubation (Fig. 6A). Cat-MMP-11 alone was greatly more efficient to cleave a1-PI than Cat-MMP-14 when tested alone as previously reported [24]. When preincubated together, Cat-MMP-11 activity was dramatically reduced. More than 50% was decreased after 1 h of preincubation (P < 0.001), and it was completely abolished after 3 h (P < 0.001). On the contrary, when Cat-MMP-11 and Cat-MMP-14 were tested together but without preincubation (0 h), the level of a1-PI cleavage remained constant indicating that Cat-MMP-11 was still active (Fig. 6A). Inactive Cat-MMP-14 m had no effect. The human MMP-11 presents a characteristic that is the replacement of a canonical Proline, found in almost all catalytic domain of MMPs including mouse MMP-11, by an Alanine (Ala-235) [1,33]. To test the impact of this difference, similar experiments were conducted with mouse Cat-MMP-11. Similar results were obtained (Fig. 6B) indicating that this residue is not important for MMP-14 site recognition/function.
Finally, to study the impact of the hemopexin domain, we tested the MMP-14 activity on the full-length form of MMP-11 using the mouse FL-MMP-11 ( Fig. 6C and D). Using same quantity of FL-MMP-11 and Cat-MMP-14, when added without preincubation, mixture was as efficient as FL-MMP-11 alone. However, Cat-MMP-14 preincubation led to a significant 20% decrease in FL-MMP-11 activity (P < 0.01). We then used twofold more Cat-MMP-14 than FL-MMP-11. At this concentration, Cat-MMP-14 activity alone was about half of that of FL-MMP-11 when tested alone (Fig. 6D). When added without preincubation, mixture efficiency was increased in comparison with FL-MMP-11 alone, indicating that both MMPs were still active, and their enzyme activities were cumulated (P < 0.01). By contrast, after preincubation, the FL-MMP-11 inactivation was more pronounced (40%; P < 0.001). It should be noticed that this value is greatly underestimated since it includes the intrinsic Cat-MMP-14 activity. Cat-MMP-14 m had no activity. Thus, MMP-14 also inactivates the full-length MMP-11 activity, indicating that the hemopexin domain does not greatly altered MMP-14 site recognition/function. Collectively, these data indicate that MMP-14 inactivates enzymatic activity of the full-length or C-terminally truncated MMP-11, regardless their human or mouse origin.

Discussion
In this study, we demonstrate that MMP-11 is a substrate of MMP-14. MMP-14 cleaves human and mouse MMP-11 between the G and I residues of the PGG-ILA, and the H and L of V/IQH-LYG conserved sequences, both located in the catalytic domain. This cleavage leads to the MMP-11 inactivation. Since MMP-14 is produced by HT-1080 cancer cells, whereas MMP-11 is secreted by HFL1 stromal cells, our findings support the emerging importance of tumor-stroma interaction/cross-talk. These data are of interest since MMP-11 and MMP-14 are potent factors of the pericellular microenvironment in a broad spectrum of biological events, including cancer cell invasive steps.
That MMP-14 cleaves other MMPs is not new. Many reports have shown that MMP-14 processes several pro-MMPs leading to their activation: pro-MMP-8 [36], pro-MMP-13 [37], and most notably pro-MMP-2 [18,21,38,39]. MMP-14 is therefore involved in multiple zymogen activation pathways. By contrast, we now show that MMP-14 hydrolyzes an active MMP and negatively regulates its enzymatic activity through direct cleavage of the catalytic domain. Accordingly, autocatalytic MMP-14 cleavage on the cell surface leading to an inactive membrane-tethered species has been reported [40][41][42]. Thus, MMP-14-dependent hydrolysis of MMP might be another way to control the extent of one MMP activity, independently of the canonical reversible TIMP-dependent inhibitory process. Interestingly, it was reported recently that MT1-MMP proteolytically inactivates a disintegrin and metalloproteinase 9 (ADAM9), a member of another type of metalloproteinases [43].
In addition to pro-MMPs, other substrates including the interstitial collagen, adhesive glycoproteins and proteoglycans, latent enzymes and growth factors, chemokines, and more recently intracellular substrates have been reported for MMP-14 [12]. A survey of known cleavage sites showed that MMP-14 recognizes various sequences. PxG/P (P1)-L(P 0 1) and PL(G/P/A)(P1)-L(P1 0 )(R/M) consensus sequences were defined using bacteriophage peptide display libraries [44]. Although original, the PGG(P1)-I(P 0 1) LA and V/IQH(P1)-L(P1 0 )YG cleavage sites are partially conserved, like the P and G residues in P3 and P1 for the first site, and P and L residues in P3 and P1 0 for the second. A search of the Blast protein sequence database shows that PGG-ILA sequence is only present in MMP-11, whereas V/IQH-LYG is shared with MMP-26. Whether MMP-14 cleaves MMP-26 remains to be studied.
The crystallographic structure of the mouse MMP-11 catalytic domain complexed to RXPO3, a phosphinic inhibitor mimicking the transition-state, shows MMP-11 to have five-stranded beta-sheets and three alpha helices [33]. It is a negatively charged globular protein containing several pockets. The S1 0 subsite, a well defined hydrophobic pocket of variable depth depending on the MMP, is crucial for the specificity of the enzymatic activity of MMPs. The S1 0 MMP-11 pocket is larger than those of the other MMPs due to the presence of an E at position 215 [33]. Interestingly, we observed that the G(P1) identified in this study is one of the residues involved in Ca ++ binding and which contributes to the correct structure of the S1 0 MMP-11 pocket. This suggests that part of the catalytic site of MMP-11 serves as cleavage site for MMP-14.
MMP-11 cleavage by MMP-14 generates a 36 kDa MMP-11 fragment consistent with the size expected for the MMP-11 F1 fragment from the PGG-ILA cleavage. Interestingly, Mari et al. [19] have previously reported, in a tumor/stroma cell coculture assay, that the released active form of MMP-11 is processed via an unidentified MMP-dependent mechanism to a 35 kDa protein lacking enzymatic activity. The authors identified several sites of cleavage located C-terminal to the PGG-ILA sequence. It is therefore tempting to speculate that the 35 kDa MMP-11 form reported by these authors might result from PGG-ILA MMP-14-mediated cleavage.
MMP-14 is a transmembrane protein while MMP-11 is soluble into the cellular microenvironment. By hydrolyzing MMP-11, membranous MMP-14 might provide spatial restriction of MMP-11 proteinase activity and locally protect specific and vital pericellular ECM and cells from the inappropriate function of MMP-11.
That MMP-14 negatively regulates MMP-11 activity is consistent with the spatial expression pattern that both proteins share. In addition, a number of in vivo functional observations show that MMP-14 acts often in an opposing manner to MMP-11. For example, it has been shown that MMP-14 coordinates adipocyte differentiation by acting as an inducer of adipogenesis [45], whereas MMP-11 induces adipocyte dedifferentiation and acts as a negative regulator of adipogenesis [6,46]. Therefore, one of the main questions is to identify which MMP-14 function depends on MMP-11 inactivation. While pro-MMP activation or collagen cleavage by MMP-14 seems to be MMP-11-independent, this is less evident during adipogenesis, where MMP-14 might balance MMP-11 function. Moreover, since MMP-14 has numerous substrates, dynamic competitive processes might exist between them and MMP-11. The relative level of in vivo interaction between these two proteins remains to be determined.
In conclusion, the present work brings new insight into the critical regulatory mechanisms of MMP-11 function. Our data also highlight the continuously increasing complexity of the MMP network. To date, MMP-14 was known as an activator of MMP function. Our results indicate that MMP-14 is also a negative regulator of MMP function. This further reinforced the concept of MMP-14 as a central player in the MMP cascade.