The FASEB Journal • Research Communication

Regulation of matrix metalloproteinase-2 (MMP-2)activity by phosphorylation

Meltem Sariahmetoglu,*,‡ Bryan D. Crawford,* Hernando Leon,† Jolanta Sawicka,*Laiji Li,§ Barbara J. Ballermann,§ Charles Holmes,‡ Luc G. Berthiaume,� Andrew Holt,*Grzegorz Sawicki,*,1 and Richard Schulz*,†,2

Departments of Pharmacology,* Pediatrics,† Biochemistry,‡ Medicine,§ and Cell Biology,�

Cardiovascular Research Group, University of Alberta, Edmonton, Alberta, Canada

ABSTRACT The regulation of matrix metalloprotein-ases (MMP) has been studied extensively due to thefundamental roles these zinc-endopeptidases play indiverse physiological and pathological processes. How-ever, phosphorylation has not previously been consid-ered as a potential modulator of MMP activity. Theubiquitously expressed MMP-2 contains 29 potentialphosphorylation sites. Mass spectrometry reveals that atleast five of these sites are phosphorylated in hrMMP-2expressed in mammalian cells. Treatment of HT1080cells with an activator of protein kinase C results in achange in MMP-2 immunoreactivity on 2D immuno-blots consistent with phosphorylation, and purifiedMMP-2 is phosphorylated by protein kinase C in vitro.Furthermore, MMP-2 from HT1080 cell-conditionedmedium is immunoreactive with antibodies directedagainst phosphothreonine and phosphoserine, whichsuggests that it is phosphorylated. Analysis of MMP-2activity by zymography, gelatin dequenching assays, andmeasurement of kinetic parameters shows that thephosphorylation status of MMP-2 significantly affectsits enzymatic properties. Consistent with this, dephos-phorylation of MMP-2 immunoprecipitated fromHT1080 conditioned medium with alkaline phospha-tase significantly increases its activity. We conclude thatMMP-2 is modulated by phosphorylation on multiplesites and that protein kinase C may be a regulator ofthis protease in vivo.—Sariahmetoglu, M., Crawford,B. D., Leon, H., Sawicka, J., Li, L., Ballermann, B. J.,Holmes, C., Berthiaume, L. G., Holt, A., Sawicki, G.,Schulz, R. Regulation of matrix metalloproteinase-2activity by phosphorylation. FASEB J. 21, 2486 –2495(2007)

Key Words: dephosphorylation � gelatinase � alkaline phos-phate � protein kinase C � mass spectrometry

The matrix metalloproteinases (MMPs) are a fam-ily of zinc-containing endopeptidases best known fortheir roles in physiological and pathological remodel-ing of the extracellular matrix (ECM) during angiogen-esis, wound healing, embryogenesis, tumor metastasis,and various cardiovascular and inflammatory diseases(1, 2). The activation of MMPs requires proteolytic

removal of the N-terminal propeptide, or disruption ofthe bond between the cysteine sulfhydryl moiety in theautoinhibitory propeptide domain and the zinc ion inthe catalytic site of the enzyme (3). Proteolytic removalof the propeptide is catalyzed by proteases such asplasminogen, trypsin, or kallikrein (4, 5) as well as bymembrane-type-1 MMP activity (6). Alternatively,proMMPs can be activated post-translationally either byS-glutathiolation of the propeptide cysteine sulfhydryl,as was shown for MMP-1, -8, and -9 (7) or by S-nitrosylation as shown for MMP-9 (8). This results in anactive, full-length MMP, without proteolytic removal ofthe propeptide. Thus, MMP activity is regulated atseveral levels, including by transcriptional and post-translational mechanisms, and by the inhibition ofenzymatic activity by endogenous tissue inhibitors ofmetalloproteinases (9).

MMP-2, also known as gelatinase A or type IV colla-genase, is the most widely expressed of all the MMPsand is found in most tissues and cells. Although MMP-2is well known as an ECM degrading enzyme, it also actson several nonmatrix substrates, including interleukin1� precursor (10), big endothelin-1 (11), monocytechemoattractant protein-3 (12), and stromal cell-de-rived factor 1 (13), which have expanded the biologicalroles of this important protease. We discovered novelintracellular localization and actions for MMP-2 in theproteolytic degradation of the sarcomeric proteins tro-ponin I and myosin light chain 1 in myocardial isch-emia-reperfusion injury (14, 15). Recently we foundthat MMP-2 also associates with caveolin-1 in the mem-brane of cardiac myocytes, where caveolin-1 appears tohelp maintain MMP-2 in a membrane-bound and in-hibited state (16). Caveolin-1 also binds to other regu-latory proteins including protein kinase C (PKC) (17).MMP-2 also has a nuclear localization sequence and its

1 Current address: Department of Pharmacology, Universityof Saskatchewan, Saskatoon, SK, Canada

2 Correspondence: Departments of Pediatrics and Phar-macology, 462 Heritage Medical Research Center, Univer-sity of Alberta, Edmonton, AB T6G 2S2, Canada, E-mail:richard.schulz@ualberta.ca

doi: 10.1096/fj.06-7938com

2486 0892-6638/07/0021-2486 © FASEB

presence was verified in the nuclei of heart and livercells (18).

Because of its unexpected intracellular localizationand actions, we reasoned that there may also be novelregulatory mechanisms modulating MMP-2 activitywithin the cell. Since phosphorylation is often involvedin regulating the activities of intracellular proteins, weexamined the possibility that MMP-2 is phosphorylated.In particular, we explored the role of phosphorylationin regulating MMP-2 activity. Here we show that MMP-2is phosphorylated and that its phosphorylation statussignificantly affects its activity. Furthermore, we showthat MMP-2 is phosphorylated by PKC in vitro and thatthe PKC activator phorbol 12-myristate 13-acetate(PMA) induces the phosphorylation of MMP-2 in cellculture. Our data suggest that phosphorylation is animportant and previously unrecognized regulator ofMMP-2 activity.

MATERIALS AND METHODS

Materials

Unless otherwise specified, the reagents used were obtainedfrom Sigma-Aldrich (Oakville, ON, Canada) or Fisher Scien-tific (Ottawa, ON, Canada). Polyclonal antiphosphothreo-nine, antiphosphoserine and monoclonal MMP-2 antibodieswere obtained from Chemicon (Temecula, CA, USA). Mono-clonal antiphosphotyrosine was purchased from Cell Signal-ing Technology (Danvers, MA, USA). The polyclonal MMP-2antibody was a gift from Dr. Mieczyslaw Wozniak (Depart-ment of Clinical Chemistry, Medical University, Wroclaw,Poland). The following were purchased from the sourcesindicated: 72 kDa and 64 kDa hrMMP-2 (purified frommammalian cells) and bisindolylmaleimide (Calbiochem, SanDiego, CA, USA); the catalytic fragment of PKC and Omn-iMMP fluorogenic peptide substrate, BIOMOL, PlymouthMeeting, PA); �32P-ATP and 32P-H3PO4 (Perkin-Elmer,Wellesley, MA, USA); histone H1 protein (Stressgen, SanDiego, CA, USA); goat anti-rabbit and anti-mouse IgG conju-gated with horseradish peroxidase (Santa Cruz Biotechnol-ogy, Santa Cruz, CA, USA); DQ gelatin (Molecular Probes,Eugene, OR, USA); molecular weight markers, CoomassieBrilliant Blue R-250, Silver Stain Plus kit, Bradford ProteinAssay and PVDF membrane (Bio-Rad, Hercules, CA, USA);protein A- and G-Sepharose beads and ECL plus (Amersham,Piscataway, NJ, USA).

Prediction of phosphorylation sites within human MMP-2

The primary sequence of human MMP-2 (accession numberP08253) was obtained from the Swiss-Prot protein database(http://us.expasy.org/sprot/). NetPhos 2.0 (http://www.cbs.dtu.dk/services/NetPhos/) is based on a neural networkmethod, which predicts serine, threonine, and tyrosine phos-phorylation sites in eukaryotic proteins (19). NetPhosK 1.0(http://www.cbs.dtu.dk/services/NetPhosK/) was used topredict kinases that may act on MMP-2.

Cell culture

HT1080 cells were obtained from the American Type CultureCollection (Manassas, VA, USA). Cells were cultured inDulbecco’s modified Eagle’s medium supplemented with

penicillin (100 U/ml), streptomycin (100 �g/ml), and 10%(v/v) fetal bovine serum (Life Technologies, Inc., Gaithers-burg, MD, USA) and were passaged twice per week with a0.05% trypsin/1 mM EDTA wash (Life Technologies, Inc.).Cells were maintained at 37°C under 5% CO2 atmosphere.The medium was removed and replaced by serum-free me-dium 24 h before experimentation.

Cell homogenates

Untreated or PMA-treated (100 nM, 24 h) cells were collectedand homogenized by sonication in 50 mmol/L Tris-HCl (pH7.4) containing 3.1 mmol/L sucrose, 1 mmol/L DTT, 10�g/mL leupeptin, 10 �g/mL soybean trypsin inhibitor, 2�g/mL aprotinin, and 0.1% Triton X-100. The homogenatewas centrifuged at 10,000g (4°C for 10 min), and the super-natant was collected. Protein content in homogenates wasanalyzed using the Bradford Protein Assay, and bovine serumalbumin was used as the standard.

Two-dimensional polyacrylamide gel electrophoresis andmass spectrometry

hrMMP-2 (72 kDa, 3.2 �g) or HT1080 cell homogenates (200�g) was applied to an 11 cm immobilized pH gradient (pH5–8) strip (Bio-Rad) and equilibrated for 16–18 h at 20°C inrehydration buffer (Bio-Rad). Isoelectric focusing was per-formed using a Bio-Rad Protean isoelectric focusing cell asdescribed previously (15). Two-dimensional electrophoresiswas performed using Criterion precast gradient gels (4–20%acrylamide, Bio-Rad). After separation, proteins were de-tected by Coomassie blue or silver staining or by Western blotusing anti-MMP-2 antibody. The best resolved of the 72 kDaacidic protein spots of MMP-2 (pI 5.405 and, hence, poten-tially phosphorylated) was excised from the gel. Subse-quently, a peptide mass fingerprint was obtained by in-geldigestion with trypsin and analyzed with a Bruker DaltonicsUltraflex MALDI TOF/TOF mass spectrometer. Analysis ofexperimentally manipulated human recombinant MMP-2 wasperformed similarly but using only 400 ng of protein (neces-sitating silver staining) and mass fingerprinting of a spotwhich resolved at pI 5.64.

Calculation of the theoretical masses of MMP-2 peptidesgenerated by trypsin cleavage

Peptide-Mass (http://au.expasy.org/tools/peptide-mass.html)was used to predict the mass of trypsin digested peptides fromthe sequence of human MMP-2. For calculation of theoreticalpeptide masses we assumed: a) 1 missed cleavage level and�0.2 Da mass tolerance; b) mandatory alkylation and reduc-tion of cysteines with iodoacetamide; and c) variable oxida-tion of methionines. Only peptides bigger than 500 Da wereconsidered.

Examination of experimental peptide mass fingerprint forphosphorylation

FindMod tool was used to find potential phosphorylation(http://au.expasy.org/tools/findmod/) in tryptic peptides ofMMP-2.

MMP-2 sequence alignment

We aligned the complete amino acid sequences of MMP-2homologues from rat (NM_031054), mouse (NM_008610),human (NM_004530), and pig (NM_214192) using ClustalW.

2487PHOSPHORYLATION OF MMP-2

In vitro kinase assay

Human recombinant 64 kDa MMP-2 was diluted in reactionbuffer (50 mM Tris pH 9.5, 50 mM MgCl2, 100 mM NaCl,0.1% Tween-20 with 5 �g/ml sodium azide) to 20 ng/�l andincubated with 14 U/�l alkaline phosphatase for 1 h at 37°C.After dephosphorylation, AP activity was abolished by heatingto 80°C for 15 min. Aliquots of dephosphorylated MMP-2were then diluted to 10 ng/�l in PKC reaction buffer (100�M ATP, 40 mM MES pH 6.0, 1 mM EGTA, 10 mM MgCl2)and incubated with 6 �Ci �32P-ATP (specific activity 3000Ci/mmol) in the presence or absence of 10 ng of the catalyticfragment of PKC. Purified bovine histone H1 was used as apositive control for PKC activity. The reaction was stopped byadding an equal volume of 2� SDS-PAGE sample-loadingbuffer. Protein mixtures were then separated by SDS-PAGE(8% gel), the gel was blotted to a PVDF membrane, and themembrane was subjected to autoradiography for 30 min atroom temperature.

Metabolic labeling

On 80–90% confluence, the medium of unstimulatedHT1080 cells was replaced with serum-free, phosphate-freeDMEM. After 1 h, it was replaced with fresh phosphate-freeDMEM containing 400 �Ci/ml [32P]H3PO4 (Perkin-Elmer)and incubated for 5 h. In some experiments 100 nM bisin-dolylmaleimide was added as indicated. To prepare for im-munoprecipitation, the cells were washed once with ice-coldPBS, and each sample homogenized in the same volume oflysis buffer (50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1%Nonidet P-40, 0.5% sodium deoxycholate, 30 mM NaF, 40mM �-glycerophosphate, 20 mM sodium pyrophosphate, 1mM sodium orthovanadate, and 100 nM calyculin A). Greatcare was taken to ensure that all steps of the immunoprecipi-tation were done in an identical manner, whether usingpolyclonal MMP-2 antibody or an unrelated IgG. Equalvolumes of the final immunoprecipitate samples in 2� SDSbuffer were loaded on the gel for electrophoresis.

Immunoprecipitation

Cell homogenates or conditioned medium were incubatedeither with antiphosphoserine, antiphosphothreonine, an-tiphosphotyrosine, anti-MMP-2, or unrelated IgG overnight at4°C with mixing. Protein G-Sepharose beads (50% w/v) wereadded and incubated overnight at 4°C with mixing. Theimmunoprecipitates from cell homogenates were washed fivetimes with successive wash buffers containing 20 nM okadaicacid (two washes: 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1%Nonidet P-40, and 0.5% sodium deoxycholate; two washes: 50mM Tris–HCl, 150 mM NaCl, 0.1% Nonidet P-40, and 0.05%sodium deoxycholate; and one wash: 10 mM Tris–HCl, pH7.5, 0.1% Nonidet P-40, 0.05% sodium deoxycholate, and0.1% SDS). The immunoprecipitates from conditioned me-dia were washed 3 times at 4°C with 50 mM Tris-HCl buffercontaining 150 mM NaCl (pH 8.0). Beads were then sus-pended in 2� SDS sample buffer containing �-mercaptoetha-nol and used for autoradiography, gelatin zymography orWestern blotting. �-Mercaptoethanol was omitted, however,for gelatin zymography samples.

Alkaline phosphatase (AP) and PKC treatment

Human recombinant MMP-2 (10 ng) or immunoprecipitatedMMP-2 from conditioned medium of HT1080 cells was incu-bated in dephosphorylation buffer (0.05 M Tris-HCl and 1mM DTT, pH 8.5) with or without AP (100 U/�g target

protein) for 20 min at 37°C. For PKC treatment, hrMMP-2 wastreated with 1 ng of the catalytic fragment of PKC pernanogram of MMP-2 in PKC reaction buffer containing 200�M ATP and incubated for 1 h at 37°C. The reaction mixturewas split and stopped by adding an equal volume of 2�SDS-PAGE sample loading buffer (for AP treatment) orbisindolylmaleimide (100 nM, for PKC treatment).

Kinetic analysis of MMP-2 activity

The hydrolysis of OmniMMP fluorogenic substrate (0–66�M, prepared in 1.8% (v/v) DMSO) by either untreated,AP-treated, or PKC-treated samples of human recombinant64 kDa MMP-2 (0.2 nM in 50 mM Tris pH 7.6, 10 mM CaCl2,0.05% Brij-35, 10 �M ZnSO4) was measured at 37°C in acontinuous plate reader-based protocol. Assays were made ina total volume of 120 �l in black polystyrene half-area plates(Corning, Corning, NY, USA), and contained MMP-2 (60 �lin 2� reaction buffer) and substrate (60 �l) or DMSO (1.8%,v/v; 60 �l in blank wells). Fluorescence associated with a(7-methoxycoumarin-4-yl)acetyl-tagged cleavage product wasmeasured every 30 s for 1 h (�ex 328 nm, �em 393 nm) in aMolecular Devices SPECTRAmax Gemini XPS fluorescencemicroplate reader. Neither AP nor PKC interfered with thehydrolysis of OmniMMP or had any effect on fluorescence(results not shown). The rate of product formation in eachwell was determined through linear regression of the fluores-cence-time data by the plate reader software (SOFTmax Pro,v 4.8; Molecular Devices Inc., Sunnyvale, CA, USA). Appro-priate lag times, to preclude data obtained prior to equilibra-tion at 37°C, and end times, to preclude data obtainedfollowing a loss of linearity, were entered manually prior tolinear regression of data to obtain slopes. Rate values werethen corrected for loss of signal due to absorption by sub-strate at 393 nm, using a measured substrate extinctioncoefficient of 7627 M–1cm–1 and a measured path length of0.672 cm. Linear reaction rates (r2�0.98 for all data) werefitted to the Michaelis-Menten equation using the nonlinearcurve fitting facility of GraphPad Prism v 4.03 (GraphPadsoftware) in order to obtain KM and Vmax values.

Fluorescent gelatin dequenching assay

Untreated, AP-treated, or PKC-treated samples of 64 kDahuman recombinant MMP-2 (50 ng) were added to 190 �lreaction buffer (150 mM NaCl, 10 mM HEPES pH 7.5, 0.25mM DTT, 5 mM CaCl2, 0.1% Triton X-100 with 5 �g/mlsodium azide) with 50 ng DQ gelatin in 96-well microplates,and incubated at 37°C for 30 min. FITC fluorescence gener-ated by the cleavage of DQ gelatin was measured using aThermo Labsystems Fluoroskan Ascent microplate readerfitted with FITC excitation and emission filters. Data arereported as the percent increase above background fluores-cence observed in substrate-only blanks. Neither AP nor PKCinterfered with fluorescent dequenching (results not shown).All assays were performed in triplicate.

Gelatin zymography

Gelatin zymography for MMP activity was performed accord-ing to Sawicki et al. (20). Briefly, nonreduced immunopre-cipitated proteins using antiphospho antibodies were loadedonto an 8% polyacrylamide gel containing 2 mg/ml gelatinand electrophoresed for 60–90 min (150 V, 4°C). Afterwashing the gels with Triton X-100 (2.5% v/v, 3�20 min), thegels were incubated at 37°C for 18–24 h, stained with 0.05%Coomassie Brilliant Blue G, and destained. Gelatinolyticactivities were detected as transparent bands against the

2488 Vol. 21 August 2007 SARIAHMETOGLU ET AL.The FASEB Journal

background of Coomassie blue-stained gelatin. To quantifythe activities of the detected enzymes, zymograms were im-aged using a GS-800 Calibrated Densitometer (Bio-Rad). Theintensities of the separate bands were analyzed using Quan-tity1 measurement software (Bio-Rad) and reported as suchor expressed as a specific activity per mg protein. Condi-tioned medium from untreated HT1080 cells was used as aMMP-2 reference standard.

Western blotting

Phospho-MMP-2 levels were determined by Western blotanalysis. Briefly, immunoprecipitated proteins using anti-MMP-2 antibody were loaded and separated on 8% polyacryl-amide SDS-PAGE gels under reducing conditions (21). Afterelectrophoresis (150 V, 20°C), samples were electroblottedonto a PVDF membrane by semidry technique (25 V, 30 min,Bio-Rad). Phospho-MMP-2 content in immunoprecipitatedsamples was identified using antiphosphoserine, antiphos-phothreonine, or antiphosphotyrosine antibodies. Immuno-reactive protein bands were visualized using the ECL plusdetection system.

Computer rendering

3D computer models of MMP-2 showing putative phosphor-ylation sites were generated using PyMol (http://pymol.sourceforge.net/) and the 2.8 Å X-ray diffraction-based struc-ture of human MMP-2 (22).

Statistical analysis

Results are expressed as mean � sem or means with 95%confidence intervals. Statistical analyses were performed,where appropriate, using Student’s t test or one-way analysisof variance with Tukey’s post hoc test. P � 0.05 was consideredstatistically significant.

RESULTS

Human proMMP-2 (72 kDa) contains 29 potentialphosphorylation sites (Fig. 1). Three putative threo-

nine phosporylation sites are located within the auto-inhibitory propeptide domain. Three putative serine,two threonine, and two tyrosine phosphorylation sitesare located within collagenase-like domains-1 and -2.One threonine or tyrosine and three serine phosphor-ylation sites are located in the hemopexin-like domain,which is involved in binding to TIMPs and certainsubstrates, membrane activation, and some proteolyticactivities. The other predicted phosphorylation sites (4on serine, 4 on threonine, and 6 on tyrosine) arelocated in the collagen-binding domain (Fig. 1).

We first examined the phosphorylation status of 72kDa hrMMP-2 expressed in mammalian cells empiri-cally by 2D electrophoresis and mass-spectroscopy. Pu-rified recombinant MMP-2 (3.2 �g) was resolved by 2Delectrophoresis, resulting in 14 spots with an apparentmolecular weight consistent with MMP-2, and isoelec-tric points ranging from pH 5.29 to 6.07 (Fig. 2A). Tomaximize the probability of detecting phosphorylatedpeptides, we analyzed the best resolved of the moreacidic spots (Fig. 2A, pI5.405) by mass-spectrometry.Evidence of at least 5 phosphorylated peptides waspresent in this spectrum (data not shown). Sequenceanalysis of the detected peptides using NetPhos 2.0showed very high probability scores for phosphoryla-tion of serine 160 and 365, threonine 250, and tyrosine271. Also serine 32 was indicated as a putative phos-phorylated amino acid but with a lower probabilityscore (Fig. 2B). NetPhosK 1.0 was used to determinethe kinases most likely to phosphorylate these residues.PKC had the highest probability score (for T250),followed by protein kinase A (for S160) and glycogensynthase kinase 3� (for S32). The kinases likely to beacting on Y271 and S365 were not predicted. Theseidentified phosphorylation sites on MMP-2 and pre-dicted amino acid targets for known kinases are shownin Fig. 2B. All five of the putative phosphorylation sitesare conserved in the MMP-2 sequences of all mammals

Figure 1. Domain structure and theoretical phosphorylation sites for human MMP-2. Potential phosphorylation sites withinhuman MMP-2 were identified using the NetPhos prediction program. Vertical lines that extend above the demarcatedthreshold indicate putative amino acid phosphorylation targets, whether serine, threonine, or tyrosine. The residues discussedin the text are indicated.

2489PHOSPHORYLATION OF MMP-2

for which sequence data are available, suggesting thatthese sites may have functional significance (Fig. 3).

We tested the ability of PKC to phosphorylatehrMMP-2 in vitro in two ways. Firstly, we treated 400 ng72 kDa hrMMP-2 expressed in mammalian cells witheither PKC or AP, and resolved the MMP-2 by 2Delectrophoresis (Fig. 4A inset). Since we had used onlyone-eighth of the material in this experiment com-pared to that shown in Fig. 2A, we picked the bestresolved spot which occurred at pI 5.64. We thendetermined the phosphorylation status of the excisedspots by mass spectrometry (Fig. 4A, inset). Untreated72 kDa MMP-2 excised from gels at pI 5.64 showsevidence of phosphorylation on S160 and S365 (Fig.4A, top panel, and B). After treatment with AP, noevidence of phosphorylation was detected in the massfingerprints (Fig. 4A, middle panel, and B). Treatmentwith PKC yielded a mass fingerprint with evidence ofphosphorylation on S160, T250, S365 and T377/8 (Fig.4A, bottom panel, and B). Secondly, we assessed theability of PKC to phosphorylate MMP-2 using an in vitrokinase assay. We find that 64 kDa hrMMP-2 is effectivelyphosphorylated under these conditions (Fig. 5A).

To determine whether PKC is likely acting on MMP-2in a biological system, we used the phorbol ester PMAas a PKC activator, which activates diglyceride-depen-dent -PKC, �-PKC, and �-PKC, but not �-PKC. Weanalyzed proteins in the homogenates from untreated

or 100 nM PMA-treated HT1080 cells, by 2D electro-phoresis followed by immunoblotting with anti-MMP-2.The MMP-2 detected in 2D immunoblots of untreatedHT1080 homogenates resolves as a single broad spot,whereas two broad spots of MMP-2 immunoreactivityare detected in the PMA-treated homogenates, onewith more acidic pI (as indicated by an arrow), consis-tent with PKC-mediated phosphorylation of MMP-2(Fig. 5B). We next sought to determine the basalMMP-2 phosphorylation status in a biological system.HT1080 cells were metabolically labeled with 32P, fol-lowed by immunoprecipitation of MMP-2 and autora-diography. Figure 5C shows basal phosphorylation ofMMP-2 that is effectively inhibited by PKC blockadeusing bisindolylmaleimide. Immunoprecipitation con-trol using unrelated IgG did not show any similar band(Fig. 5C).

To investigate the effects of phosphorylation on thecatalytic behavior of MMP-2, we measured the kineticsof hrMMP-2-catalyzed hydrolysis using a synthetic flu-orogenic substrate under various conditions. Fluores-cence of this substrate, (Mca-Pro-Leu-Gly-Leu-Dpa-Ala-

Figure 3. Amino acid sequence alignments of rat, mouse,human and pig MMP-2. Alignment of the complete aminoacid sequences of MMP-2 homologues from rat, mouse,human, and pig. Boxes indicate conserved sites of putativephosphorylation discussed in the text.

Figure 2. Identification of phosphorylation sites on recombi-nant MMP-2. A) 2-D electrophoresis of 72 kDa hrMMP-2 usingnarrow range immobilized pH gradient strips (linear gradientfrom pH 5 to 8) for the first dimension (horizontal) andpolyacrylamide gradient gel 4–20% for the second dimension(vertical). Arrow indicates the protein spot used for massspectrometry analysis. B) Putative phosphorylation sites show-ing the predicted probability of phosphorylation (NetPhos),and the probability of phosphorylation by kinases (Net-PhosK).

2490 Vol. 21 August 2007 SARIAHMETOGLU ET AL.The FASEB Journal

Arg-NH2; Mca (7-methoxycoumarin-4-yl)acetyl; Dpa N-3-(2,4-dinitrophenyl)-L-,�-diaminopropionyl) isnegligible prior to cleavage, due to intramolecularquenching, but degradation by MMP-2 generates afluorescent fragment (Mca-Pro-Leu-Gly) with maximalemission at 393 nm. After correction for substrateabsorption at this wavelength (ε7627 M–1cm–1), wefound that 0.2 nM untreated 64 kDa MMP-2 had a KMof 33.3 � 1.8 �M and Vmax of 38.5 � 1.0 RFU sec–1 (Fig.6 and Table 1). Dephosphorylation of MMP-2 by APresulted in a significant increase in Vmax and a small butsignificant increase in KM, whereas phosphorylation byPKC resulted in significant decreases in both Vmax andKM (Fig. 6 and Table 1). Calculated values for Vmax/KM,based on mean values for kinetic constants, were thus1.16 (Untreated), 1.36 (AP), and 0.42 (PKC), suggest-ing an enhancement in catalytic efficiency associatedwith MMP dephosphorylation of around 320% towardthis substrate. Thus, the phosphorylation state ofMMP-2 significantly affects the enzymatic properties of

this protease. In assays of PKC-treated MMP-2, mea-sured initial velocities fell modestly as substrate concen-tration increased beyond 30 �M (see Discussion). Datapoints obtained with PKC-treated enzyme at substrateconcentrations higher than 30 �M were thus not in-cluded in nonlinear regression analyses.

Fluorescently quenched gelatin has been used as anMMP substrate both in situ and in vivo (23, 24). Due tothe complexity and variability of this substrate, it is notwell-suited to kinetic analyses. However, it does provideanother approach to quantifying MMP activity. Inend point analyses of DQ-gelatin degradation, alka-line phosphatase treated hrMMP-2 gave a small butinsignificant increase in activity (158�16% to 205�6%), whereas PKC treated hrMMP-2 was significantlyless active (118�17%, n3, P�0.05) than identicalamounts of untreated hrMMP-2.

Historically, the most widely utilized assay for MMP-2activity has been gelatin zymography. This approachhas the advantages of sensitivity and the ability todistinguish the relative electrophoretic mobility (andtherefore the identity) of the protease activities beingdetected, making it very well suited to the analysis ofcell homogenates and other complex protein mixtures.In an effort to verify the biological relevance of ourobservations, we performed gelatin zymography onmaterial immunoprecipitated from HT1080-condi-tioned media using phospho-specific antibodies. An-tiphosphothreonine, antiphosphoserine, and to a lesserextent antiphosphotyrosine antibodies are all able toimmunoprecipitate MMP-2, suggesting that MMP-2 isphosphorylated on these residues (Fig. 7A). Consistentwith this, both antiphosphoserine and antiphospho-threonine (but not antiphosphotyrosine) can detectmaterial immunoprecipitated using anti-MMP-2 anti-body on Western blots (Fig. 7B). Negative controls ofimmunoprecipitation experiments revealed no activityon gelatin zymography and no bands on Western blots(data not shown). As gelatin zymography for MMP-2activity is at least 100 times more sensitive than Westernblotting, it is not surprising that no band for MMP-2 wasobserved when using anti-phosphotyrosine antibody.Finally, we determined that dephosphorylation of im-munoprecipitated MMP-2 from HT1080 cell condi-tioned media using AP increases its gelatinolytic activityby 25-fold in zymography, consistent with our kineticanalyses of MMP-2 activity in vitro (Fig. 7C).

DISCUSSION

The regulation of MMPs has been the focus of intensiveinvestigation for decades. These efforts have yielded asophisticated understanding of the complex regulationof these proteases by several signaling pathways (9),post-translational processing by various proteases (4),and inhibition by TIMPs and RECK (REversion-induc-ing-Cysteine-rich protein with Kazal motifs) (25). De-spite all of these investigations, the role of phosphory-lation in regulating MMP activity has not been

Figure 4. Mass spectrometric analysis of the effect of eitheralkaline phosphatase (AP) or PKC treatment on the phos-phorylation status of MMP-2. A) Mass fingerprints of trypticpeptides derived from the 72 kDa hrMMP-2 spot resolving atpI 5.64 (inset). Untreated protein (top panel) shows evidenceof phosphorylated peptides (*), which are detailed in (B). APtreatment abolished phosphorylation (middle panel). Treat-ment with PKC results in increased phosphorylation (bottompanel). B) Summary of detected phosphopeptides after eachtreatment.

2491PHOSPHORYLATION OF MMP-2

considered, presumably due to the focus on the extra-cellular activities of MMPs. Given the existence ofMMP-2 splice variants lacking an N-terminal secretorysignal peptide (e.g., accession number AL832088), andempirical evidence of MMP-2 protein within cells (14,15, 18), it seems clear that the intracellular roles ofthese proteases, and their regulation, are importantresearch questions. Here we present for the first timeevidence that MMP-2 is phosphorylated, that its phos-phorylation status modulates its enzymatic activity, andthat PKC may be involved in regulating MMP-2.

Autoradiography, mass spectrometry, immunopre-

cipitation, 2D gel electrophoresis, and Western blotanalysis all confirm the bioinformatic inference thatMMP-2 may be phosphorylated on multiple residues.Several lines of evidence show that the phosphorylationstate of MMP-2 modulates its activity. We determinedthat the kinetics of MMP-2-mediated hydrolysis of asynthetic fluorogenic substrate are significantly affectedby phosphorylation. Dephosphorylation of recombi-nant MMP-2 had only modest effects on the kinetics ofsubstrate hydrolysis or the rate of fluorescent gelatindequenching. However, PKC-mediated phosphoryla-tion resulted in significant inhibitory effects in both ofthese assays, suggesting that our preparations ofhrMMP-2 were not highly phosphorylated. Also consis-tent with this is the observation that dephosphorylationof immunoprecipitated MMP-2 from HT1080 cells isprofoundly more active in zymography. One possiblereason for the greater effect of AP treatment on the 64kDa (vs. 72 kDa) secreted forms is that the propeptidedomain may interfere with access of AP to the phospho-amino acids. Phosphorylated MMP-2 activity vs. thefluorogenic substrate was lower than predicted byMichaelis-Menten analysis at substrate concentrationsabove 30 �M. This may have been due to a small degreeof residual uncorrected inner-filter effect caused by thefluorogenic substrate used, which was more evident inthese low activity samples. However, we cannot pre-clude the possibility that phosphorylation facilitateslow-affinity binding of a second molecule of this smallsubstrate to MMP-2, yielding an enzyme-disubstratecomplex of reduced catalytic competence.

Mass spectra of PKC-treated samples show at least twosites (T250 and T377/8) of phosphorylation that aredistinct from those sites detected as phosphorylated inthe untreated preparation and dephosphorylated by

Figure 6. Kinetic analysis of MMP-2-mediated hydrolysis of afluorogenic peptide substrate as a function of phosphoryla-tion status of MMP-2. Kinetic data for cleavage of OmniMMPfluorogenic peptide by 64 kDa hrMMP-2 (0.2 nM) afterphosphorylation with PKC (Œ), dephosphorylation with AP(F), or no treatment (E), fitted to the Michaelis-Mentenequation by least-squares nonlinear regression analysis. Datashown are mean � sem of 4 replicate determinations. KM andVmax values obtained are listed in Table 1. For regressionanalysis of the PKC group, only the first 6 of the 12 data pointsshown were included (see text).

Figure 5. MMP-2 phosphorylationby PKC. A) Autoradiography of 64kDa hrMMP-2 after incubation with32P-ATP with or without PKC. Ra-dioactive labeling is detected at ap-propriate molecular weights show-ing that MMP-2 is phosphorylatedby PKC. Histone H1 is used as apositive control for phosphoryla-tion. B) Immunoblots of HT1080cell homogenates separated by 2D

electrophoresis probed with MMP-2 specific antibody. Top panel: homogenates of untreated cells yield a single broad spotof MMP-2 immunoreactivity. Bottom panel: homogenates of cells treated with 100 nM PMA (a PKC activator) produced twobroad spots of MMP-2 immunoreactivity. Arrow indicates putative phosphorylated MMP-2. C) Autoradiography of MMP-2immunoprecipitated from unstimulated HT1080 cells after incubation with 32P-H3PO4 in the absence or presence of PKCinhibitor.

2492 Vol. 21 August 2007 SARIAHMETOGLU ET AL.The FASEB Journal

AP treatment. Furthermore, unlike the modest effectof dephosphorylation observed in our assays using puri-fied recombinant enzyme, dephosphorylation of MMP-2immunoprecipitated from HT1080 conditioned mediumresulted in a 20-fold increase in activity. Therefore, itseems likely that MMP-2 may be phosphorylated onmultiple sites, but either only a small fraction of these sitesare relevant to the enzymatic properties measured in ourstudies or there are multiple sites with opposing effects.Consistent with this is the observation that MMP-2 isimmunoreactive with antibodies specific for at least twodifferent phosphorylated amino acids. Elucidating whichsites are responsible for the changes we have observed inthis study, as well as the kinases and phosphatases respon-sible for their phosphorylation status, is the focus ofongoing research.

All five of the phosphorylation sites that we couldconfirm by mass spectrometry occur on residues withside chains accessible at the surface of the protein (Fig.8) and are conserved in the sequences of MMP-2 fromother mammals, which suggests that these may beevolutionarily constrained and, therefore, functionallyrelevant modifications. Parameters that phosphoryla-

tion might modulate other than substrate affinity andturnover number include substrate specificity, intracel-lular trafficking, protein stability, protein-protein inter-actions (such as with endogenous inhibitors or activat-ing proteases, phosphatases, kinases, and/or othereffectors). The majority of the predicted phosphoryla-tion sites are within the collagen binding domain,which is an essential element for substrate binding(26–28), consistent with regulation of substrate speci-ficity. Nevertheless all of these possibilities warrantfurther investigation.

Previous studies have shown a link between kinasesand the regulation of MMP-2 expression, activity andsecretion (27–32), although none of these consideredthe possibility of MMP-2 phosphorylation. We suggestthat one or more kinases and phosphatases may have arole in the direct regulation of MMP-2 activity. Asalkaline phosphatase can dephosphorylate MMP-2, it istempting to speculate that alkaline phosphatases, socharacteristically abundant in vascular tissue (29), maybe involved in activating MMP-2 during angiogenesis.

Pathological conditions affect many enzymes in-volved in either phosphorylation or dephosphorylation

TABLE 1. Kinetic constants determined by nonlinear regression analysis of data shown in Figure 6

Untreated AP PKC

Vmax (RFU/s) 38.5 (36.5 to 40.4) 65.8 (61.5 to 70.1)a 8.04 (7.24 to 8.83)a,b

KM (�m) 33.3 (29.7 to 36.9) 48.3 (42.4 to 54.2)a 19.0 (15.2 to 22.9)a,b

r2 0.986 0.988 0.978

Values are means with 95% confidence intervals obtained from a single least-squares nonlinearregression analysis of the mean � sem of 4 replicate determinations. Confidence intervals for kineticconstants were obtained from asymptotic standard errors determined by nonlinear regression. Kineticconstants for the PKC group were obtained by fitting only the first 6 points ([substrate]5–30 �M) ofthe 12 points shown in Figure 6 (see text for details). aP � 0.001 compared with untreated group; bP �0.001 compared with AP group.

Figure 7. MMP-2 phosphorylation in cell culture and effects of phosphorylation onactivity. Conditioned media of HT1080 cells were immunoprecipitated using theindicated phospho-specific antibodies and these were then subjected to gelatinzymography. The zymograms are representative of four independent experimentswith similar results. B) Conditioned media (48 h) from HT1080 cells wereimmunoprecipitated with anti-MMP-2 and subjected to Western blot analysis usingphospho-specific antibodies. The blots, representative of four independent exper-iments with similar results, show the results of triplicate samples as indicated by 1,2, 3. C) Conditioned media of HT1080 cells were immunoprecipitated usinganti-MMP-2 (or unrelated IgG, data not shown). Immunoprecipitates were treated

with or without AP (control) and subjected to gelatin zymography. The upper zymogram is a representative result from 3independent experiments which are quantified in the bar graph underneath. IP: Immunoprecipitation; A-P-Tyr:antiphosphotyrosine; A-P-Thr: antiphosphothreonine; A-P-Ser: antiphosphoserine; St: Standard (HT1080 cell conditionedmedium). *P � 0.05 vs. control (n3).

2493PHOSPHORYLATION OF MMP-2

events, which affect the regulation of cell function. Forthat reason, identification of the protein kinases andphosphatases involved in the regulation of MMP-2activity will be important to better understand thepathologies in which this ubiquitous protease is sug-gested to contribute, such as cancer (30), inflammation(31), ischemia-reperfusion (14), and cytokine-medi-ated injury to the heart (32). The finding that secretedMMP-2 is phosphorylated suggests that this may occursomewhere along its secretory pathway. As PKC local-izes to caveolin-1 of the cell membrane (17) whereMMP-2 very recently has been found to associate withcaveolin-1 (16), suggests a possible role for this kinasein the phosphorylation of secreted MMP-2. Since theeffects of phosphorylation of MMP-2 may not onlyaffect its activity but possibly also its substrate specificityand cellular localization, this may have major implica-tions in the design of rational drug therapies to lessenthe impact of MMP-2 activation in these pathologies.

In summary we have shown that MMP-2 is phos-phorylated and that its phosphorylation status mod-ulates its activity. It is hoped that further investiga-tions will reveal the identities of the kinases andphosphatases involved in regulating MMP-2 activityby this novel mechanism.

This project was funded by grants from the CanadianInstitutes of Health Research (to R. S. MOP-77524; to L. G. B.MOP-37955; to A. H. MOP-77529). M. S. was a Fellow ofNATO-Science (TUBITAK-Turkey). H. L. was a GraduateStudent Trainee of the Heart and Stroke Foundation ofCanada and the Alberta Heritage Foundation for MedicalResearch (AHFMR). L. G. B. is a Senior Scholar of theAHFMR, and R. S. is an AHFMR Scientist.

REFERENCES

1. Woessner, J. F., Jr. (1991) Matrix metalloproteinases and theirinhibitors in connective tissue remodeling. FASEB J. 5, 2145–2154

2. McCawley, L. J., and Matrisian, L. M. (2001) Matrix metallopro-teinases: they’re not just for matrix anymore! Curr. Opin. CellBiol. 13, 534–540

3. Van Wart, H. E., and Birkedal-Hansen, H. (1990) The cysteineswitch: a principle of regulation of metalloproteinase activitywith potential applicability to the entire matrix metalloprotein-ase gene family. Proc. Natl. Acad. Sci. U. S. A. 87, 5578–5582

4. Murphy, G., Willenbrock, F., Crabbe, T., O’Shea, M., Ward, R.,Atkinson, S., O’Connell, J., and Docherty, A. (1994) Regulation ofmatrix metalloproteinase activity. Ann. N. Y. Acad. Sci. 732, 31–41

5. Mignatti, P., and Rifkin, D. B. (1996) Plasminogen activatorsand matrix metalloproteinases in angiogenesis. Enzyme. Protein.49, 117–137

6. Sato, H., Takino, T., Okada, Y., Cao, J., Shinagawa, A.,Yamamoto, E., and Seiki, M. (1994) A matrix metalloproteinaseexpressed on the surface of invasive tumour cells. Nature 370,61–65

7. Okamoto, T., Akaike, T., Sawa, T., Miyamoto, Y., van der Vliet,A., and Maeda, H. (2001) Activation of matrix metalloprotein-ases by peroxynitrite-induced protein S-glutathiolation via disul-fide S-oxide formation. J. Biol. Chem. 276, 29596–29602

8. Gu, Z., Kaul, M., Yan, B., Kridel, S. J., Cui, J., Strongin, A., Smith,J. W., Liddington, R. C., and Lipton, S. A. (2002) S-nitrosylationof matrix metalloproteinases: signaling pathway to neuronal celldeath. Science 297, 1186–1190

9. Chakraborti, S., Mandal, M., Das, S., Mandal, A., andChakraborti, T. (2003) Regulation of matrix metalloprotein-ases: an overview. Mol. Cell. Biochem. 253, 269–285

10. Schonbeck, U., Mach, F., and Libby, P. (1998) Generation ofbiologically active IL-1 beta by matrix metalloproteinases: anovel caspase-1-independent pathway of IL-1 beta processing.J. Immunol. 161, 3340–3346

11. Fernandez-Patron, C., Radomski, M. W., and Davidge, S. T.(1999) Vascular matrix metalloproteinase-2 cleaves big endothe-lin-1 yielding a novel vasoconstrictor. Circ. Res. 85, 906–911

12. McQuibban, G. A., Gong, J. H., Tam, E. M., McCulloch, C. A.,Clark-Lewis, I., and Overall, C. M. (2000) Inflammation damp-ened by gelatinase A cleavage of monocyte chemoattractantprotein-3. Science 289, 1202–1206

13. Zhang, K., McQuibban, G. A., Silva, C., Butler, G. S., Johnston,J. B., Holden, J., Clark-Lewis, I., Overall, C. M., and Power, C.(2003) HIV-induced metalloproteinase processing of the che-mokine stromal cell derived factor-1 causes neurodegeneration.Nat. Neurosci. 6, 1064–1071

14. Wang, W., Schulze, C. J., Suarez-Pinzon, W. L., Dyck, J. R.,Sawicki, G., and Schulz, R. (2002) Intracellular action of matrixmetalloproteinase-2 accounts for acute myocardial ischemia andreperfusion injury. Circulation 106, 1543–1549

15. Sawicki, G., Leon, H., Sawicka, J., Sariahmetoglu, M., Schulze,C. J., Scott, P. G., Szczesna-Cordary, D., and Schulz, R. (2005)Degradation of myosin light chain in isolated rat hearts sub-jected to ischemia-reperfusion injury: a new intracellular targetfor matrix metalloproteinase-2. Circulation 112, 544–552

16. Chow, A.K., Cena, J., El-Yazbi, A. F., Crawford, B.D., Holt, A.,Cho, W. J., Daniel, E. E., Schulz, R. (2007) Caveolin-1 inhibitsmatrix metalloproteinase-2 activity in the heart. J. Mol. CellCardiol. 42, 896–901

17. Prevostel, C., Alice, V., Joubert, D., and Parker, P. J. (2000)Protein kinase C (alpha) actively downregulates through caveo-lae-dependent traffic to an endosomal compartment. J. Cell Sci.113, 2575–2584

18. Kwan, J. A., Schulze, C. J., Wang, W., Leon, H., Sariahmeto-glu, M., Sung, M., Sawicka, J., Sims, D. E., Sawicki, G., andSchulz, R. (2004) Matrix metalloproteinase-2 (MMP-2) ispresent in the nucleus of cardiac myocytes and is capable ofcleaving poly (ADP-ribose) polymerase (PARP) in vitro. FASEB J.18, 690–692

19. Blom, N., Gammeltoft, S., and Brunak, S. (1999) Sequence andstructure-based prediction of eukaryotic protein phosphoryla-tion sites. J. Mol. Biol. 294, 1351–1362

Figure 8. 3D rendering of MMP-2 showingputative phosphorylation sites. Views of the 2.8Å structure of human MMP-2 (22) show thecatalytic side and a 180° rotation rendered as aribbon diagram with the predicted electrostaticsurface shown (translucent). Putative phos-phorylation sites (S32, S160, T250, S365 andY271) are shown rendered as spheres.

2494 Vol. 21 August 2007 SARIAHMETOGLU ET AL.The FASEB Journal

20. Sawicki, G., Matsuzaki, A., and Janowska-Wieczorek, A. (1998)Expression of the active form of MMP-2 on the surface ofleukemic cells accounts for their in vitro invasion. J. Cancer Res.Clin. Oncol. 124, 245–252

21. Laemmli, U. K. (1970) Cleavage of structural proteins duringthe assembly of the head of bacteriophage T4. Nature 227,680–685

22. Morgunova, E., Tuuttila, A., Bergmann, U., Isupov, M.,Lindqvist, Y., Schneider, G., and Tryggvason, K. (1999) Struc-ture of human pro-matrix metalloproteinase-2: activation mech-anism revealed. Science 284, 1667–1670

23. Frederiks, W. M., and Mook, O. R. (2004) Metabolic mapping ofproteinase activity with emphasis on in situ zymography ofgelatinases: review and protocols. J. Histochem. Cytochem. 52,711–722

24. Wang, F. Q., So, J., Reierstad, S., and Fishman, D. A. (2006)Vascular endothelial growth factor-regulated ovarian cancerinvasion and migration involves expression and activation ofmatrix metalloproteinases. Int. J. Cancer. 118, 879–888

25. Baker, A. H., Edwards, D. R., and Murphy, G. (2002) Metallo-proteinase inhibitors: biological actions and therapeutic oppor-tunities. J. Cell Sci. 115, 3719–3727

26. Murphy, G., Nguyen, Q., Cockett, M. I., Atkinson, S. J., Allan,J. A., Knight, C. G., Willenbrock, F., and Docherty, A. J. (1994)Assessment of the role of the fibronectin-like domain of gelati-nase A by analysis of a deletion mutant. J. Biol. Chem. 269,6632–6636

27. Shipley, J. M., Doyle, G. A., Fliszar, C. J., Ye, Q. Z., Johnson, L. L.,Shapiro, S. D., Welgus, H. G., and Senior, R. M. (1996) Thestructural basis for the elastolytic activity of the 92-kDa and72-kDa gelatinases. Role of the fibronectin type II-like repeats.J. Biol. Chem. 271, 4335–4341

28. Patterson, M. L., Atkinson, S. J., Knauper, V., and Murphy, G.(2001) Specific collagenolysis by gelatinase A, MMP-2, is deter-mined by the hemopexin domain and not the fibronectin-likedomain. FEBS Lett. 503, 158–162

29. Zoellner, H. F., and Hunter, N. (1989) Histochemical identifi-cation of the vascular endothelial isoenzyme of alkaline phos-phatase. J. Histochem. Cytochem. 37, 1893–1898

30. Klein, G., Vellenga, E., Fraaije, M. W., Kamps, W. A., and deBont, E. S. (2004) The possible role of matrix metalloproteinase(MMP)-2 and MMP-9 in cancer, e.g., acute leukemia. Crit. Rev.Oncol. Hematol. 50, 87–100

31. Parks, W. C., Wilson, C. L., and Lopez-Boado, Y. S. (2004)Matrix metalloproteinases as modulators of inflammation andinnate immunity. Nat. Rev. Immunol. 4, 617–629

32. Gao, C. Q., Sawicki, G., Suarez-Pinzon, W. L., Csont, T.,Wozniak, M., Ferdinandy, P., and Schulz, R. (2003) Matrixmetalloproteinase-2 mediates cytokine-induced myocardial con-tractile dysfunction. Cardiovasc. Res. 57, 426–433

Received for publication December 11, 2006.Accepted for publication March 1, 2007.

2495PHOSPHORYLATION OF MMP-2