Inhibition of plasminogen activators or matrix metalloproteinases prevents cardiac rupture but impairs therapeutic angiogenesis and causes cardiac failure

NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999 1135

ARTICLES

More than 1.5 million people suffer acute myocardial infarction(AMI) annually in the United States. About 30 % of those pa-tients die within the first 24 hours, due to arrhythmias or pumpfailure. With improved treatments, cardiac ruture has become aserious complication, accounting for 5 to 31 % of in-hospitalmortality after AMI (refs. 1,2). Rupture mainly affects middle-aged patients with a transmural AMI but no previous history ofangina and unpredictable and is usually fatal due to absence oftreatment. Surgical repair suffers a 27% operative mortality andoffers only a 31% long-term survival3. Hypertension, unduephysical activity, diabetes, cardiac hypertrophy, fatty infiltra-tion, infarct expansion and delayed thrombolysis with streptok-inase have been related to rupture but the relevance of these riskfactors remains controversial4,5. Genetic predisposition factors orcriteria for identifying patients at risk for cardiac rupture afterAMI remain undetermined. A better understanding of the mech-anisms of cardiac rupture might lead to prevention, but this has

been precluded by lack of reproducible animal models.The plasminogen system comprises two plasminogen activa-

tors (PAs) [tissue-type PA (t-PA) and urokinase-type PA (u-PA)],their PA inhibitor-1 (PAI-1), and the matrix metalloproteinases(MMPs) [including stromelysin-1 (MMP-3), gelatinase-B (MMP-9)and tissue inhibitor-1 (TIMP-1)]; this system may be involved inmatrix remodeling after AMI (refs. 6,7). Impaired PA/MMP prote-olysis may be involved in cardiac fibrosis8 and in coronary or my-ocardial remodeling after AMI (ref. 9), whereas proteinaseinhibitors may improve cardiac function by protecting againstwall thinning or dilatation7. The role of PA/MMP proteinases inhealing and revascularization of the infarct and in cardiac rup-ture, however, remains unknown.

Here, we used a combination of gene-inactivation and gene-transfer techniques in mice to address these issues. Deficiency ofu-PA and, to a somewhat lesser extent, of MMP-9 prevented car-diac rupture after AMI. Long- term inhibition of u-PA gene func-

Inhibition of plasminogen activators or matrixmetalloproteinases prevents cardiac rupture but impairs

therapeutic angiogenesis and causes cardiac failure

S. HEYMANS1, A. LUTTUN1, D. NUYENS1, G. THEILMEIER1, E. CREEMERS2, L. MOONS1,G.D. DYSPERSIN3, J.P.M. CLEUTJENS2, M. SHIPLEY4, A. ANGELLILO1, M. LEVI1, O. NÜΒE5,

A. BAKER6, E. KESHET7, F. LUPU8, J-M HERBERT9, J.F.M. SMITS2, S.D. SHAPIRO4, M. BAES1, M. BORGERS3, D. COLLEN1,

M. J.A.P. DAEMEN2 & P. CARMELIET1

1,9Center for Transgene Technology and Gene Therapy, Flanders Interuniversity, Leuven, Belgium2Cardiovascular Research Institute, University of Maastricht, 6200 MD, The Netherlands

3Janssen Research Foundation, Tumhoutsesteenweg 30, 2340, Beerse, Belgium4Departments of Pediatrics, Cell Biology and Medicine, Washington University School of Medicine,

216 South Kinghighway Blvd., St. Louis, Missouri, USA5Division of Infectious Diseases, University Hospital Geneva, 24 Rue Mecheli-du-Crest,

CH-1211 Geneva 14, Switzerland6Departments of Medicine and Therapeutics, Western Infirmary, 44 Church Street, University of Glasgow,

Glasgow G11 6NT, United Kingdom7Department of Molecular Biology, Hebrew University-Hadassah Medical School, Jerusalem 91120, Israel

8Vascular Biology Laboratory, Weston Experimental Research Center, Thrombosis Research Institute,Emanuel Kaye Building, Canrcsa Road, Chelsea, London SW3 6LR, United Kingdom

9Sanofi Recherche, Haeobiology Research Department, 195 Route d’Espange, 31036 Toulouse Cedex, FranceA.L. and D.N. contributed equally to this study

Correspondence should be addressed to P.C.; email: [email protected]

Cardiac rupture is a fatal complication of acute myocardial infarction lacking treatment. Here,acute myocardial infarction resulted in rupture in wild-type mice and in mice lacking tissue-typeplasminogen activator, urokinase receptor, matrix metalloproteinase stromelysin-1 or metalloe-lastase. Instead, deficiency of urokinase-type plasminogen activator (u-PA–/–) completely pro-tected against rupture, whereas lack of gelatinase-B partially protected against rupture.However, u-PA–/– mice showed impaired scar formation and infarct revascularization, even aftertreatment with vascular endothelial growth factor, and died of cardiac failure due to depressedcontractility, arrhythmias and ischemia. Temporary administration of PA inhibitor-1 or the ma-trix metalloproteinase-inhibitor TIMP-1 completely protected wild-type mice against rupturebut did not abort infarct healing, thus constituting a new approach to prevent cardiac ruptureafter acute myocardial infarction.

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

1136 NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999

ARTICLES

tion, however, impaired infarct healing, predisposed to cardiacfailure under adrenergic stress and prevented therapeutic my-ocardial angiogenesis. Instead, temporary PA/MMP inhibitionby adenoviral gene transfer prevented rupture without abortinginfarct healing and therefore may constitute a new treatmentparadigm for fatal cardiac rupture.

u-PA and MMP-9 predispose to cardiac rupture after AMILigation of the left anterior descending coronary artery caused aninfarct involving 45% of the left ventricular (LV) wall (Fig. 1a). Afew subendocardial cardiomyocytes survived, probably becauseof oxygen supply from the LV cavity. Within 4 days, about 30%of infarcted male wild-type mice suffered fatal rupture of the LVwall (Fig. 1b). Only male mice died, characteristically at night, in-dicating that rupture might be induced by stress, activity or fluc-tuations in hormonal balance. Cardiac rupture also occurred inabout 30% of infarcted male t-PA–/–, uPA-receptor–/– (u-PAR–/–)MMP-3–/– or MMP-12–/– mice, but all 32 infarcted male u-PA–/–

mice and most (24 of 26) MMP-9–/– mice were protected againstventricular wall rupture (P < 0.05; Table 1). Lack of cardiac rup-ture in u-PA–/– mice was not due to reduced LV pressure (85 ± 5mm Hg, wild-type, compared with 81 ± 4 mm Hg, u-PA–/–; n = 5; P,not significant).

Infarct infiltration of leukocytes and cardiac ruptureLeukocytes infiltrated wild-type infarcts at the site and at the timeof ventricular rupture (Fig. 1c and f), first at the borders and thenin the entire infarct. Depletion of leukocytes by hydroxycar-bamide completely prevented rupture in 10 male wild-type mice(neutrophils/mm2 infarct: 520 ± 13 without and 250 ± 90 withhydroxycarbamide; P < 0.05), despite similar infarct size (45 ± 2%of LV wall). Leukocyte infiltration 4 days after AMI was similar inwild-type, t-PA–/–, u-PAR–/–, MMP-3–/– and MMP-12–/– mice, slightlyreduced in MMP-9–/– mice (by 18%) and most substantially de-creased in u-PA–/– mice (by 46%) (Table 1). Most of these cells weremyeloperoxidase-positive neutrophils (77 ± 4% in wild-type and67 ± 5% in u-PA–/– mice; P, not significant; Fig. 1c) with a smallcontribution by macrophages (19 ± 2% in wild-type and 7 ± 3% inu-PA–/– mice; P < 0.05) and lymphocytes (CD3-positive T-cells: 5 ±1% in wild-type and 3 ± 1% in u-PA–/– mice, n = 6, P < 0.05; CD19-positive B-cells: 2 ± 1% in wild-type and 0.5 ± 0.2% in u-PA–/–

mice, n = 6, P < 0.05). The impaired infarct infiltration of u-PA–/–

neutrophils did not result from intrinsic defects in chemotaxis oradhesion, or from differences in leukocytosis (not shown). Therole of u-PA in neutrophil migration in vivo did not seem to de-pend on u-PAR, as u-PAR–/– neutrophils accumulated normally ininfarcts (Table 1). This was unexpected, as chemotaxis and adhe-

a b c

e

d

f g h i

sham 4 7 14 d

MM

P ac

tivity

(A

U)

% u

PA ly

sis

area

/inf

arct

are

a

CD

45-ir

leuk

ocyt

es/m

m2

sham 2 4 7 14 d2 4 7 14 d

Fig. 1 Myocardial infarction and cardiac rupture. a, Wild-type my-ocardium 2 d after infarction, showing tropomyosin staining in viablesubendocardial layers and absent tropomyosin expression in necroticmidmyocardium (inset). *, ventricular cavity; arrows, borders of infarct. b, Macroscopic view of ruptured LV wall (arrows) in the infarct bordernear viable right ventricle (RV; dark red color) and necrotic LV (light red-dish color). c, Myeloperoxidase (MPO) staining, showing inflammatoryinfiltrate at the site of wall rupture. d, Western blot of extracts from onesham-operated heart (S) and two infarcted hearts (I), showing an in-creased ratio of active 68-kDa plasmin (Pli; arrowhead) to inactive 93-kDaplasminogen (Plg; arrow) in wild-type compared with u-PA–/– mice. Lane1, Plg standard; lane 2, Pli standard. e, Gelatin zymography of extractsfrom the infarct (I), infarct borders (BI) and non-infarcted left ventricle(LV) and right ventricle (RV), showing an increased ratio of the amounts

of the partially active 88-kDa and active 82-kDa MMP-9 (arrowhead) tothe amount of he inactive 92-kDa pro-MMP-9 (arrow) in the infarct andinfarct borders in wild-type (WT) compared with u-PA–/– mice. f, Accumulation of CD45-immunoreactive leukocytes in infarcts. Sham-operated hearts: < 10 leukocytes/mm2. *, P < 0.05, uPA–/– (�) comparedwith wild-type ( ). g and h, Activity of u-PA (g) and MMP-2 (h; �) andMMP-9 (h; �) in sham-operated and infarcted wild-type hearts. g, For u-PA activity, in situ zymographic lysis area is expressed as a percent of thetotal infarct area. �, infarct; , septum (S); �, remote non-infarcted leftventricle (LV). h, Gelatinographic MMP activity in infarct is expressed inarbitrary units (AU). *, P < 0.05 compared with S, LV or sham-operated. i, Double immunostaining for u-PA (green) and myeloperoxidase (MPO;red) in wild-type mice, 4 d after AMI. Scale bars represent 625 µm (a), 25 µm (c) and 15 µm (I).

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

sion of u-PAR–/– neutrophils to TNF-α stimulated endothelial cellsin vitro was impaired (not shown).

Mechanisms of cardiomyocyte injuryInterleukin-6 (IL-6) mediates transendothelial migration and ad-hesion of neutrophils to cardiomyocytes and subsequent myocyteinjury through the production of reactive oxygen10. The produc-tion of IL-6 was higher in wild-type than in u-PA–/– infarcts (260 ± 5 pg/mg protein in wild-type compared with 100 ± 25pg/mg protein in u-PA–/– infarcts, 4 days after AMI; P < 0.05).Oxidative stress in u-PA–/– infarcts in vivo was reduced, as demon-strated by reduced infarct levels of myeloperoxidase, involved inthe generation of hypocholorous acid (160 ± 20 units (U)/l inwild-type compared with 70 ± 5 U/l in u-PA–/– mice; n = 5; P < 0.05).The decreased IL-6 or reactive oxygen levels did not result fromabnormal IL-6 serum levels (not shown) or from an intrinsic defectin IL-6 or superoxide production by u-PA–/– inflammatory cells invitro (IL-6 in pg/ml: 13 ± 5 for wild-type and 10 ± 7 for u-PA–/– cells;superoxide production in % of wild-type: 100 ± 5% for wild-typeand 92 ± 4% for u-PA–/– cells; P, not significant).

Infarct leukocytes produce u-PA and MMP-9u-PA and MMP-9 activity were increased (by 70-fold and 17-fold,respectively) in the infarct at the site and at the time of rupture(Fig. 1g and h). t-PA activity was not affected in wild-type infarctsand did not compensatorily increase in u-PA–/– infarcts (lysis areaexpressed as % of infarct area: 63 ± 14% in wild-type and 72 ± 7%in u-PA–/–; P, not significant). Plasmin levels were higher in wild-type than in u-PA–/– infarcts (Fig. 1d). u-PA cannot directly de-grade interstitial collagens (types I and III) but can initiatecollagenolysis through activation of zymogen MMPs (ref. 11).Activation of pro-MMP-9 was reduced in u-PA–/– infarcts (Fig. 1e);the arbitrary densitometry units were 220 (inactive 92-kDa pro-MMP-9) compared with 190 (total amount of partially active 88-kDa + active 82-kDa MMP-9; ratio, 88-kDa +82-kDa/92-kDa =0.85) in wild-type infarcts, and 130 (92-kDa) compared with 60(88-kDa +82-kDa; ratio, 88-kDa +82-kDa/92-kDa = 0.45) in u-PA–/–

infarcts. Double immunostaining showed minimal expression ofu-PA and MMP-9 in sham opertated hearts but substantially up-

NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999 1137

ARTICLES

regulated expression in myeloperoxidase-positive or CD45-posi-tive leukocytes (Fig. 1i) and in Mac-3-positive macrophages in in-farcted hearts (not shown). In contrast, levels of MMP-2, afibroblast-derived MMP not activated by plasmin11, progressivelyincreased after AMI to a maximum by 14 days after AMI (Fig. 1h).

PAI-1 or TIMP-1 gene transfer prevents cardiac rupture We administered adenovirus expressing human PAI-1 (AdPAI-1;ref. 12), human TIMP-1 (AdTIMP-1; ref. 13) or a control AdRR5adenovirus12,14 to male wild-type mice 24 hours before coronaryligation. AdPAI-1 increased plasma PAI-1 levels to 54 ± 3 µg/mland AdTIMP-1 increased TIMP-1 levels to 51 ± 5 µg/ml within 5days. Human PAI-1 was deposited in infarcts, thereby inhibitingu-PA activity in the infarct by 60% (area of u-PA-mediated lysisexpressed as percent of infarct area: 90 ± 10% with AdRR5 com-pared with 30 ± 10% with AdPAI-1; n = 4; P < 0.05). All male wild-type mice treated with AdPAI-1 (n = 19) or AdTIMP-1 (n = 18) wereprotected against cardiac rupture, whereas rupture occurred in 5of 20 wild-type mice injected with control AdRR5 virus (P < 0.05).AdPAI suppressed leukocyte infiltration as efficiently as loss of u-PA gene function, whereas AdTIMP-1 was more effective thanMMP-9 gene inactivation (Table 1).

Impaired and scar formation in infarcted u-PA–/– miceRemoval of necrotic cardiomyocytes was complete by 14 daysafter AMI in all genotypes (Figs. 2a and 3a and Table 1), except inu-PA–/– mice, in which necrotic cardiomyocytes persisted for up to5 weeks as ‘mummified ghosts’ embedded in laminin-richbasement membrane (Figs. 2b and c and 3a and Table 1).Ultrastructurally, these necrotic u-PA–/– cardiomyocytes showedsigns found in cells that die of severe acute ischemia (not shown).Macrophage accumulation in infarcts was maximal by 7 days afterAMI and was significantly less in u-PA–/– than in wild-type infarcts(Mac-3 positive cells/mm2 at 7 and 14 days: 150 ± 2 and 63 ± 2 inwild-type compared with 90 ± 8 and 45 ± 2 in u-PA–/– mice; P <0.05). Ingestion of collagen fragments by macrophages, a processrequiring prior proteolysis of collagen fibers15, was found in wild-type but rarely in u-PA–/– infarcts (not shown). This did not seem toresult from an intrinsic defect in phagocytosis by u-PA–/– leuko-

Table 1 Incidence of cardiac rupture, leukocyte infiltration and collagen deposition

Cardiac rupture Leukocyte infiltration Residual necrotic area Capillary growth Coronary growth Collagen deposition(frequency; %) (cells/mm2) (%) (vessels/mm2) (vessels/mm2) (%)

Gene inactivationwild-type 17/51 (33%) 770 ± 60 1 ± 1 450 ± 40 37 ± 3 39 ± 2t-PA–/– 3/10 (30%) 710 ± 15 0.5 ± 0.3 470 ± 30 41 ± 2 41 ± 2u-PA–/– 0/32 (0%)** 420 ± 15* 60 ± 3* 340 ± 14* 18 ± 2* 23 ± 1*u-PAR–/– 4/18 (22%) 720 ± 70 0.6 ± 0.2 440 ± 24 38 ± 4 43 ± 1MMP-3–/– 3/9 (30%) 700 ± 10 0.5 ± 0.5 450 ± 20 39 ± 4 46 ± 2MMP-9–/– 2/26 (7%)** 640 ± 20* 2 ± 1 470 ± 18 41 ± 3 45 ± 3MMP-12–/– 4/18 (22%) 740 ± 60 1 ± 0.5 450 ± 8 41 ± 5 41 ± 2

Transplantationwild-type N.D. 420 ± 20 1 ± 1 450 ± 40 36 ± 5 39 ± 2wild-type in u-PA–/– N.D. 380 ± 20 1 ± 1 460 ± 30 33 ± 5 40 ± 2

Gene transferwild-type + AdRR5 5/20 (25%) 710 ± 10 1 ± 1 460 ± 12 40 ± 4 38 ± 2wild-type + AdPAI-1 0/19 (0%)** 320 ± 20* 53 ± 5* 380 ± 25# 28 ± 2* 27 ± 1*wild-type + AdTIMP-1 0/18 (0%)** 410 ± 40* 47 ± 8* 410 ± 14# 27 ± 2* 27 ± 3*

Data for cardiac rupture represent number of infarcted mice with rupture after 4 d; **, P < 0.05, compared with wild-type, by Chi-square analysis. Other data (mean ± s.e.m.) rep-resent the number of CD45+ leukocytes infiltrating in the infarct after 4 d (gene inactivation and gene transfer) or 7 d (transplantation), and the number of vessels/mm2 infarct, orthe fractional necrotic area or sirius red-stained area as percent of total infarct area after 14 d. Transplantation, wild-type bone marrow transplanted into u-PA–/– recipients. *, P <0.05, and #, P = 0.06, compared with wild-type or wild-type+AdRR5 by unpaired t-test. N.D., not done.

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

1138 NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999

ARTICLES

cytes (not shown). Depletion of leukocytes by treatment of wild-type mice with hydroxycarbamide impaired removal of necroticcells after AMI (necrotic area after 7 days: 30 ± 4% after hydroxy-carbamide compared with 15 ± 4% in control; n = 5; P < 0.05).

Smooth muscle α-actin immunoreactive myofibroblasts (notshown) and collagen (Fig. 2d and e) were minimal in sham-oper-ated hearts and progressively increased in all genotypes except inu-PA–/– infarcts (Fig. 3b and c and Table 1). In fact, the necrotic cen-ter of u-PA–/– infarcts remained devoid of fibroblasts and containedamounts of collagen similar to those in sham-operated hearts (8 ±2% compared with 7 ± 1%, respectively; P, not significant). Four-day-old u-PA–/– infarcts contained reduced levels of latent TGF-β1,a mediator of collagen deposition16 (pg latent TGF-β1/µg protein:15 ± 2 in wild-type compared with 7 ± 1 in u-PA–/– mice; n = 5; P <0.05). Furthermore, activation of latent TGF-β1 was reduced in u-PA–/– infarcts, consistent with previous suggestions16 (pg activeTGF-β1/µg protein: 9 ± 1 in wild-type compared with 3 ± 0.1 in u-PA–/– mice; n = 5; P < 0.05; ratio,f active/total TGF-β1: 0.6 ± 0.1 inwild-type compared with 0.4 ± 0.1 in u-PA–/– mice; P = 0.05).Macrophages and fibroblasts expressed u-PA during myocardialhealing (not shown).

Impaired therapeutic angiogenesis in infarcted u-PA–/– miceThe number of thrombomodulin-stained capillaries in infarcts wasonly 5% (about 250 capillaries/mm2 in all geneotypes) of the num-ber of capillaries in normal, non-infarcted myocardium. Duringsubsequent healing, fewer capillaries revascularized infarcts in u-PA–/– than in other genotypes, in particular in the subendocardium(Table 1 and Fig. 3d). VEGF-immunoreactive cells were more nu-merous in wild-type than in u-PA–/– infarcts (not shown). Smoothmuscle α-actin (SMA) staining showed that about half of the coro-

naries survived infarction in all genotypes after 2 days (Fig. 3e).During subsequent healing, the number of coronary vessels in theinfarct was restored in all genotypes except in u-PA–/– mice (Figs. 2fand g and 3e and Table 1). Both endothelial and smooth musclecells expressed u-PA during infarct revascularization (not shown).

To determine whether absent revascularization in u-PA–/– in-farcts could be rescued, we administered 1 µg recombinant VEGF,a dose inducing maximal infarct revascularization in wild-typemice (S.H. et al., unpublished data), over a period of 7 days. VEGFsubstantially improved the revascularization of wild-type infarctsbut failed to revascularize u-PA–/– infarcts (Fig. 2h and I and Table2). Increased revascularization in wild-type mice improved infarcthealing, as demonstrated by the accelerated removal of necroticcardiomyocytes and collagen deposition at 7 days (necrotic area:15 ± 4% in control compared with 6 ± 2% after VEGF; sirius red-stained area: 31 ± 2% in control compared with 39 ± 1% afterVEGF; P < 0.05), whereas healing was not improved at all in u-PA–/– infarcts (necrotic area: 61 ± 4% in control and 58 ± 3% afterVEGF; sirius red-stained area: 24 ± 2% in control and 25 ± 3% afterVEGF; P, not significant).

Bone marrow transplantation restores healing in u-PA–/– miceWhen congenic wild-type bone marrow was transplanted into ir-radiated u-PA–/– mice, leukocyte blood counts after 6 weeks intransplanted u-PA–/– mice were similar to those in wild-type mice(109 cells/l: 10 ± 2 in transplanted u-PA–/– and 8 ± 3 in wild-type; n =8; P, not significant) and u-PA activity was detectable in infarcts oftransplanted u-PA–/– mice (not shown). Transplanted u-PA–/– micehad infarct infiltration with leukocytes, removal of necrotic my-ocardium, revascularization and collagen deposition similar tothat of wild-type mice(Table 1).

PAI-1 gene transfer does not abort myocardial healing Adenoviral PAI-1 or TIMP-1 gene transfer in wild-type mice im-paired scar formation (Table 1). However, in contrast to the persis-tent lack of healing in u-PA–/– hearts, myocardial healing afterPAI-1 gene transfer resumed beyond 14 days, resulting in normalscar formation by 5 weeks after infarction (Fig. 3f). This probablyresulted from the transient expression of PAI-1 after adenoviralgene transfer (plasma PAI-1 levels: 2 ± 1 ng/ml by 14 days aftertransfer). Thus, transient PAI-1 gene overexpression only retardedbut did not abort myocardial healing, whereas it prevented my-ocardial rupture.

Impaired cardiac function in infarcted u-PA–/– miceLeft ventricular pressure (LVP) and heart rate were similar in sham-operated and infarcted wild-type and u-PA–/– mice in baseline con-ditions (not shown). Contractility (dP/dtmax) was reduced ininfarcted compared with sham-operated mice at 14 days, but to a

Table 2 Infarct revascularization and therapeutic angiogenesis

Angiogenic Genotype Capillaries Medium-sized Large vesselscompound vessels

none wild-type 410 ± 40 20 ± 4 9 ± 3u-PA–/– 270 ± 37* 28 ± 3 14 ± 5

rVEGF165 (1 µg) wild-type 590 ± 30# 53 ± 12# 28 ± 8#

u-PA–/– 260 ± 16 27 ± 5 11 ± 1

Mean ± s.e.m. of the number of endothelial-lined vessels (without smooth muscle cells)per mm2 infarct area at 7 d after infarction in five mice. *, P < 0.05, compared with wild-type; #, P < 0.05, compared with untreated, by unpaired t-test.

Fig. 2 Infarct healing. a–c, Scar formation in 14-day-old wild-type infarcts(a) in contrast to the residual presence of ‘mummified cardiomyocyteghosts’, embedded in laminin-immunoreactive basement membrane (b)and in necrotic mid-myocardium in u-PA–/– infarcts (c). d and e, Sirius redstaining of collagen, abundant in wild-type scars (d) and minimal in u-PA–/–

infarcts (e). *, ventricular cavity (a,c,d). f and g, Smooth muscle α-actin(SMA) staining, showing impaired growth of SMA-positive coronary vessels(arrowheads) in u-PA–/– infarcts (g) compared with that of wild-type infarcts(f). h and i, Thrombomodulin (TM) staining, showing many vessels (arrow-heads) in wild-type infarcts (h) but not in u-PA–/– infarcts (i) after treatmentwith rVEGF165. Scale bars represent 50 µm (a,c,d), 15 µm (f–I) and 10 µm (b).

a

b

c

gf h i

d e

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

with 0.17 ± 0.02 at 14 days in wild-type mice, P < 0.05; and 0.13 ±0.03 at 2 days compared with 0.23 ±0.03 at 14 days in u-PA–/– mice,P < 0.05; QRS-axis in degrees: 150 ± 16 at 2 days compared with130 ± 17 at 14 days in wild-type mice, P < 0.05; and 150 ± 7 at 2days compared with 170 ± 20 at 14 days in u-PA–/– mice, P, not sig-nificant). In addition, ventricular extrasystoles and atrioventricu-lar conduction disturbances occurred in two of nine infarctedu-PA–/– mice, but in none of seven infarcted wild-type mice in base-line conditions (Fig. 4a,b).

The genotypic differences became more apparent after infarctedmice were stressed with isoproterenol (1 µg/g, intraperitoneally),an adrenergic agonist known to induce arrhythmias after AMI. Allof nine infarcted wild-type mice had an accelerated heart rate, butshowed only minimal repolarization abnormalities and recoveredwithin 10 minutes after injection (Fig. 4a). In contrast, seven of tenu-PA–/– mice developed, within minutes after isoproterenol injec-tion, severe ST-segment elevation (lead II), ventricular extrasystolesand arrhythmias (Fig. 4b). u-PA–/– mice died in cardiogenic shockwithin 15 minutes of isoproterenol injection, as shown by the pro-gressive development of bradycardia, atrioventricular block andmicrovoltages, and by pulmonary congestion and gasping.

Arrhythmias in patients with ischemic heart disease have beenrelated to changes in gap junctional proteins between cardiomy-ocytes17. Levels of the gap junctional protein connexin 43 (Cx43)were reduced in the remote u-PA–/– myocardium (Cx43-im-munoreactive area as percent of total: 18 ± 1% in wild-type com-pared with 10 ± 1% in u-PA–/– mice; P < 0.05) as well as in theresidual viable subendocardial cardiomyocytes in u-PA–/– infarcts(fractional Cx43-positive area: 21 ± 1% in wild-type comparedwith 12 ± 2% in u-PA–/– mice; P < 0.05). Cx43 expression in u-PA–/–

hearts was heterogeneous and disordered, in particular at infarct

NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999 1139

ARTICLES

similar extent in both genotypes (Fig. 3g). This might explain thecomparable compensatory cardiomyocyte hypertrophy in non-in-farcted septum. The adrenergic agent dobutamine increased con-tractility (Fig. 3g) and stroke volume (Fig. 3h) in wild-type but notin u-PA–/– mice after AMI. Cardiac output was increased in wild-type compared with u-PA–/– mice (6.5 ± 0.9 ml/min in wild-typemice compared with 4.2 ± 0.8 ml/min in u-PA–/– mice at a dobuta-mine dose of 6 ng/g per minute; P < 0.05). Ventricular relaxation(dP/dtmin) and end-diastolic pressures were similar in both geno-types (not shown). Angiotensin II pretreatment increased dP/dtmax

in wild-type but not in u-PA–/– mice after AMI (LV dP/dtmax in mmHg/s: 5,000 ± 100 after saline compared with 8,000 ± 200 after an-giotensin II in wild-type mice, P < 0.05; and 5,700 ± 100 aftersaline compared with 6,000 ± 200 after angiotensin II in u-PA–/–

mice, P, not significant).

Arrhythmias and abnormal conduction in infarcted u-PA–/– miceElectrocardiograph (ECG) recordings in sham-operated miceshowed a regular sinus rhythm, a QRS axis of +60°, a positive po-larity of repolarization (lead II) and a QTc interval of about 85 mil-liseconds, but no repolarization disturbances or arrhythmias (Fig.4a and b). Within 2 days after AMI, both wild-type and u-PA–/–

mice showed Q waves and a more negative polarity of repolariza-tion in inferior leads, a rotation of the QRS axis to +150° and a pro-longation of the QTc interval to about 275 milliseconds, as seen inpatients with AMI (Fig. 4a and b). By 14 days after AMI, ECG signsnormalized in wild-type mice but deteriorated in u-PA–/– mice.Compared with the ECGs of wild-type mice, the QTc interval didnot shorten, repolarization became progressively more negativeand the QRS axis rotated further to the right in u-PA–/– mice (Fig.4a and b; T-wave amplitude in mV: 0.25 ± 0.03 at 2 days compared

WT

uPA-/-WTuPA-/-shWTMIuPA-/-MI

2 4 7 14 35 d

% n

ecro

tic a

rea/

infa

rct

area

Myo

fibro

bla

st a

rea/

infr

arct

(%

)

sham 2 4 7 14 d sham 2 4 7 14 d 2 4 7 14 d

cap

illar

ies

/mm

2

siriu

s re

d a

rea/

infa

rct

area

(%

)

2 5 weeks

siriu

s re

d a

rea/

infa

rct

area

(%

)

LV d

P/d

t m

ax (

mm

Hg

/s)

coro

narie

s/m

m2

sham 2 4 7 14 d

SV (

µl/m

in)

ng/g/min dobutamineng/g/min dobutamine

WTAdRR5WTAdPAIuPA-/-

Fig. 3 Impaired myocardial healing and cardiac function in u-PA–/– mice. a,Removal of necrotic infarct area (percent of total infarct area) is impaired inu-PA–/– mice (�). , wild-type. b, Less accumulation of smooth muscle α-actin immunoreactive myofibroblasts in u-PA–/– infarcts (�) than in wild-typeinfarcts (hashed bars). The lower number of α-actin-positive myofibroblastsin infarcts at 14 d than at 7 d probably reflects loss of smooth muscle α-actinexpression. c–e, Less deposition of sirius-red stained collagen (c), and fewer

thrombomodulin-stained capillaries (d) and SMA-stained coronaries (e) in u-PA–/– infarcts (filled bars) than in wild-type infarcts (hashed bars). f, Collagendeposition is retarded but not aborted in AdPAI-1-treated infarcted mice. g and h, dP/dtmax (g) and stroke volume (SV) (h) in sham-operated (sh) andinfarcted (MI) mice, showing that cardiac function was similar in baselineconditions, but impaired after dobutamine administration in infarcted u-PA–/–

mice. *, P < 0.05, compared with wild-type; two-way ANOVA.

a b c d

e f g h

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

1140 NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999

ARTICLES

borders, resulting in microregions with undetectable Cx43 expres-sion (Fig. 5a and b). In addition, connexin 40 immunoreactivity(present in fast-conducting Purkinje fibers) seemed reduced insubendocardial layers in u-PA–/– infarcts (not shown).

Increased ischemia in residual viable u-PA–/– myocardiumST-segment elevations in u-PA–/– mice after isoproterenol adminis-tration did not result from genotypic differences in electrolyte bal-ance, hematological parameters, blood gases, arterial pH, bodytemperature or anesthesia (not shown). Instead, they may havebeen caused by ischemia in residual viable cardiomyocytes in thesubendocardial and border regions in u-PA–/– infarcts, as shown bytheir increased expression of GRP78, a marker of cellular nutrientdeprivation18 (Fig. 5c and d) and by the swelling, vacuolizationand disruption of cristae of their mytochondriae (Fig. 5e and f).Additional signs of ischemia in remote viable u-PA–/– cardiomy-ocytes included loss of osmophilic granules (Fig. 5g and h) and in-creased ATPase activity (Fig. 5I and j) in their mitochondria19;reduced levels of ATP (µmol/g wet weight: 1.6 ± 0.2 in u-PA–/– com-pared with 2.4 ± 0.2 in wild-type mice; n = 5; P < 0.05) but in-creased levels of lactate (µmol/g wet weight: 7.9 ± 0.9 in u-PA–/–

compared with 1.4 ± 0.5 in wild-type mice; n = 5; P < 0.05); and in-creased expression of GLUT-1 and LDH-A, markers of cellular hy-poxia and indicators of anaerobic metabolism20 (number of mRNAcopies per 103 mRNA copies hypoxanthine phosphoribosyl-trans-ferase in u-PA–/– compared with wild-type mice: 190 ± 20 com-pared with 80 ± 15 for GLUT-1; 4.1 ± 0.1 compared with 2.9 ± 0.2for LDH-A; n = 5; both P < 0.05). Plasma levels of LDH, CK and tro-ponin I were not affected by treatment of infarcted wild-type or u-PA–/– mice with isoproterenol, indicating that the isoproterenoldose caused a sublethal stress for cardiomyocytes (not shown).Cellular distress in u-PA–/– mice did not result from a difference incapillary density in the remote myocardium (capillaries/mm2:4,900 ± 170 in wild-type compared with 5,100 ± 100 in u-PA–/–

mice; P, not significant).

DiscussionThis study demonstrates that u-PA and, to a lesser extent, MMP-9contribute to cardiac rupture, and shows that u-PA is essential forcardiac functional recovery after AMI. By degrading matrix macro-

molecules, PA/MMPs allow inflammatory cells (in particular neu-trophils) to infiltrate the infarct and to disrupt the collagen net-work, a prerequisite for cardiac rupture21. In a second phase,proteolysis allows infiltration of other wound cells that mediateinfarct healing22. The need for plasmin proteolysis is emphasizedby the comparably impaired infarct healing in mice lacking plas-minogen (not shown). In addition, PA/MMPs may act throughother mechanisms. PA/MMPs can produce matrix degradationfragments that affect growth and migration of inflammatory andother wound cells23, and PA/MMPs may influence the cytokine re-sponse and oxidative stress, thereby promoting cardiomyocyte in-jury24. Production of IL-6 and oxidants was normal in vitro butreduced in u-PA–/– infarcts in vivo, probably as a result of impairedneutrophil accumulation; PA/MMPs and their receptors may mod-ulate adhesion, intracellular signaling and cell differentiation24–26.However, u-PAR was not essential in vivo, indicating that the peri-cellular u-PA activity is more important27; PA/MMPs may activateand mobilize growth factors in the extracellular matrix25,28.Reduced activation of TGF-β1 probably contributed to impairedcollagen deposition in u-PA–/– infarcts; PA/MMPs may facilitatephagocytosis and lysosomal breakdown of matrix fragments bypredigesting matrix macromolecules15. Ingestion of collagen frag-ments was indeed found less frequently in u-PA–/– infarcts.

u-PA and plasmin are unable to directly degrade interstitial colla-gen. However, u-PA was coexpressed with MMP-9 in infiltratingleukocytes, and activation of zymogen MMP-9 was reduced in u-PA–/– infarcts, indicating that u-PA/plasmin activate MMP-9.Activation of MMPs may, however, also result from other mecha-nisms including reactive oxygen metabolites, elastases or cathep-sins released by neutrophils29. The finding that MMP-9–/– mice wereonly partially protected against rupture may relate to the fact thatMMP-9 only degrades basement membrane and denatured colla-gen, and that complete degradation of collagen fibers depends on aconcerted action of interstitial collagenases and gelatinases. Such acooperation between MMPs is indeed indicated by the observationthat complete prevention of rupture was only achieved after TIMP-1 gene transfer, but not after inactivation of a single MMP gene. Incontrast to its involvement in rupture, MMP-9 seemed redundantin late infarct healing, possibly because expression of gelatinase-A(MMP-2), which is able to compensate for MMP-9, was increased.

Our data do not exclude the possibilityof involvement of other proteinases incardiac destruction or healing30.

Cardiac function in infarcted u-PA–/–

mice was profoundly depressed. Thismay be related to the impaired revascu-larization of subendocardial u-PA–/– in-farct layers. In addition, the necroticmyocardial segment in infarcted u-PA–/–

mice may be dyskinetic because of itsreduced collagen content, and may im-pose a greater workload on the residualviable cardiomyocytes in the remotemyocardium. In addition, ventricularextrasystoles were often found in in-farcted u-PA–/– mice. They may resultfrom abnormal conduction throughmyocardial regions showing persistentnecrosis or heterogeneous expressionof connexin 43, known to facilitatereentry of the electric pulse17. In addi-tion, ischemia-induced after-depolar-

Fig. 4 Abnormal conduction, repolarization and arrhythmias in infarcted u-PA–/– mice. a and b, ECG record-ings of wild-type (WT; a) and u-PA–/– (b) mice after sham operation and 2 or 14 d after myocardial infarction(MI). Within 2 d, abnormal QRS complexes and negative polarity of the ST segment denivellations are present inboth genotypes. At 14 d, the negative polarity of the ST segment partially disappears in the wild-type but not inthe u-PA–/– mouse. In addition, the infarcted u-PA–/– mouse but not the wild-type mouse spontaneously developsventricular extrasystoles. After isoproterenol administration (+ ISO), the wild-type mouse shows an acceleratedheart rate but no arrhythmias or repolarization abnormalities. In contrast, the ISO-stimulated u-PA–/– mouseshows severe ST-segment elevation after 5 min and progressive atrioventricular block, bradycardia and reducedQRS voltages after 10 min. Consecutive recordings (all lead II) from the same wild-type or u-PA–/– mouse.

a b

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999 1141

ARTICLES

izations may have triggered arrhythmias. The baseline repolariza-tion disturbances in infarcted u-PA–/– mice may be due to the per-sistent infarct necrosis, whereas the acute isoproterenol-inducedrepolarization disturbances may have been caused by increasedischemia.

Our findings may indicate that increased expression of u-PA orMMP-9 could predispose to cardiac rupture, whereas increased lev-els of PAI-1 and TIMP-1 may be protective. So far, epidemiologicstudies have only related increased expression of MMP-9 and PAI-1 or reduced levels of plasminogen (due to genetic or metabolicdeterminants) to coronary atherosclerosis31 and atherothrombo-sis32–34. However, genetic linkage studies relating u-PA or MMP-9 tocardiac rupture are unavailable at present. Our findings also indi-cate that prevention of cardiac rupture by PAI-1 or TIMP-1 coulddevelop into a new non-surgical treatment for fatal rupture. Whenused timely (immediately after hospital admission), temporarily(during the first week after AMI when most ruptures occur) andcautiously (at non-toxic doses), PA/MMP-inhibitors might preventrupture, although care should be taken not to impair long-termcardiac recovery.

u-PA was essential for capillary angiogenesis and growth of col-lateral arteries during infarct revascularization. The greater re-quirement for u-PA in revascularization of the myocardium thanin other tissues may relate to its greater content of interstitial col-lagen. During revascularization, u-PA is expressed by migratingendothelial and smooth muscle cells, but inflammatory cells,which produce angiogenic factors (review, ref. 35), also express u-PA. Defective revascularization of ischemic myocardium mightprovide an additional explanation for the increased incidence ofreinfarction in patients suffering impaired fibrinolysis33. VEGFcould not improve infarct revascularization in the absence of u-PA, possibly because its action depends on endothelial u-PA ex-pression (review, ref. 35). This indicates that therapeuticangiogenesis is determined by genetic predisposition factors,such as u-PA, justifying genetic screening of patients, eligible forangiogenic treatment. It will be useful to determine whether de-creased fibrinolysis in patients also impairs therapeutic myocar-dial angiogenesis as in u-PA deficient mice. This study alsowarrants against prolonged or uncontrolled treatment of patientswith proteinase inhibitors for post-infarct remodeling7, inflam-mation or cancer25. In conclusion, transient administration ofproteinase inhibitors may provide a new approach to preventfatal cardiac rupture, but prolonged inhibition may lead to car-diac failure after AMI.

MethodsMouse model of myocardial infarction and experimental procedures.Wild-type mice and mice lacking t-PA, u-PA, u-PAR, MMP-3 (a gift from J.Mudgett), MMP-9 or MMP-12 mice, 8–12 weeks old, were used36.Myocardial infarction was done as described37: Sham-operation included allprocedures except ligation of the left anterior decending coronary artery. Fortherapeutic angiogenesis, an osmotic minipump (Alzet, type 2001, BroekmanInstitute, Someren, Netherlands), delivering 1 µg human rVEGF165

(Peprotech, Rocky Hill, New Jersey) for 7 d, was implanted subcutaneously.For leukocyte depletion, mice were pretreated with 2 mg hydroxycarbamideper g body weight intraperitoneally (Bristol-Myers Squibb, Brussels, Belgium)1 d before coronary ligation, followed by 1 mg hydroxycarbamide per g bodyweight during the subsequent 4 d.

Histology, immunostaining, morphometry and electron microscopy.Infarcted mice were anesthetized, injected with 100 µl saline containing 0.1mM cadmium chloride and perfusion-fixed with 1% paraformaldehyde atphysiological pressure. Fixed hearts were cryoembedded or embedded inparaffin. Sections 6 µm in thickness were used for histologic analysis or for im-munostaining as described27,37. Double-staining for u-PA or MMP-9 (green)and Mac3 or myeloperoxidase (red) and analysis by confocal microscopywere done as described11. Morphometric analysis and counting of immunore-active cells (in at least three optical fields of about 0.1 mm2) were done usinga Quantimet 600 image analysis system (Leica, Brussels, Belgium). Countingof capillaries and coronaries and ultrastructural analysis by transmission elec-tron microscopy were done as described27. Antibody sources were: CD3 andCD19, PharMingen (San Diego, California); myeloperoxidase, Prosan(Merelbeke, Belgium); connexin 40, Alpha Diagnostics (San Antonio, Texas).Myocardial proton-translocating ATPase activity was localized on ultrathinsections as described19.

Quantitative RT–PCR analysis. Quantitative real-time RT–PCR was done asdescribed38 using the following forward (F) and reverse (R) oligonucleotides,and probes (P) with fluorescent dye and quencher. GLUT-1: F, 5’–GGGCAT-GTGCTTCCAGTATGT–3’; R, 5’–ACGAGGAGCACCGTGAAGAT–3’; P,5’–CAACTGTGCGGCCCGTACGTCTTC–3’; LDH-A: F, 5’–CCTGGGCCATTG-GCCT–3’; R, 5’–TGCACCCGCCTAAGGTTCT–3’; P, 5’–ATGCTCTCAGC-CAAGTCCGCCACA–3’.

Biochemical and hematological analysis. In situ zymography of t-PA or u-PA activities, gelatin zymography of MMP-2 and MMP-9 activities in ex-tracts, and immunoblotting for plasminogen/plasmin were determined asdescribed11,27. ATP and lactate content (mmol/g wet weight in left and rightventricles) were determined using standard enzymatic assays. Blood levels ofCK, LDH and troponin-I, blood gases, serum electrolytes and leukocytecounts were measured using routine clinical chemistry methods. Serum lev-els of cytokines and infarct levels of latent and active TGF-β1 were measuredwith commercially available ELISAs (TNF-α: Genzyme, Cambridge,Massachusettes; IL-6: PharMingen, San Diego, California; TGF-β1 and IL-1β:

Fig. 5 Increased ischemia in residual viable u-PA–/– myocardium. a and b, Connexin 43 im-munostaining, showing reduced andheterogeneous expression in the remote my-ocardium of u-PA–/– (b) but not of wild-type (a)mice. c and d, Darkfield micrograph, showingmore in situ GRP78 mRNA expression in u-PA–/–

(d) than in wild-type (c) infarcts. e and f, Ultrastructural signs of ischemia (mitochondr-ial swelling and clarification, disruption of mito-chondrial cristae; arrowheads) in u-PA–/– (f) butnot in wild-type (e) cardiomyocytes in thesubendocardial layers within the infarct. EC, en-docardium. g–j, Increased distress of cardiomy-ocytes in the remote non-infarcted u-PA–/–

myocardium, as shown by the loss of osmophillic amorphous granules (arrowheads in g) (h) and increased staining for ATPase (arrowheads) (j) in the

a c e g i

b d jf h

mitochondria, compared with that in wild-type hearts (WT; g and i). Scalebars represent 25 µm (a and b), 1 µm (e and f) and 0.25 µm (g–j).

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

1142 NATURE MEDICINE • VOLUME 5 • NUMBER 10 • OCTOBER 1999

ARTICLES

R&D Systems, Minneapolis, Minnesota). The methods for transplantation ofbone marrow from congenic C57BL6 wild-type into u-PA–/– mice39 and mea-surement of myeloperoxidase activity30 have been described. Bone marrowneutrophils were used for chemiluminiscence measurements of superoxideproduction after stimulation with 100 nM phorbol 12-myristate 13-acetate(Sigma). For cytokine production, neutrophils were cultured in RPMI(400,000 cell per 24 wells), supplemented with 0.5% heat-inactivatedserum and cultured for 7 h before cytokine measurements.

Adenovirus mediated PAI-1 or TIMP-1 gene transfer. Adenovirus express-ing human PAI-1 (AdPAI-1; ref. 12), human TIMP-1 (AdTIMP-1; ref. 13) orcontrol AdRR5 virus (an empty adenovirus containing the CMV promoter anda polyadenylation signal that lacks a functional transgene14) have been de-scribed. One day before ligation of the left anterior descending coronaryartery, 100 µl containing 1.3 × 109 plaque-forming units of AdPAI-1, AdTIMP-1 or AdRR5 virus were injected into the tail veins of male wild-type mice. At 5d after virus injection, PAI-1 plasma levels were measured as described12.TIMP-1 plasma levels were measured using a commercially available ELISA(Amersham). Virus-injected mice were analyzed for histology or used forstaining of PAI-1, CD45, Mac3 (all at 4 or 5 d after virus injection) or collagen(sirius red; at 14 and 35 d after AMI).

Hemodynamic and electrocardiographic measurements. Pressure mea-surements and electrocardiographic recordings were made as described38.Pressure and ECG parameters were calculated according to standard criteria.The QT interval was measured according to AHA guidelines from the begin-ning of the Q wave to the intersection of the T wave with the isoelectric lineon consecutive beats, averaged during an arrhythmia-free period of 4 s. TheQTc interval (QT/(RR/100)0.5) was calculated to correct for heart rate as de-scribed. Angiotensin II (Sigma) or control vehicle (saline) were administeredusing an Alzet osmotic minipump at a dose of 0.7 mg/kg per day from day 4to day 14 after AMI. Increasing doses of dobutamine (Dobutrex; Lilly,Brussels, Belgium) were administered at 14 d by continuous infusion over 2min for each dose through a 2 French catheter (Portex Green, Baxter, UnitedKingdom) in the left external jugular vein using a Harvard microinjectionpump (Harvard Apparatus, Holliston, Massachusetts). Hemodynamic mea-surements were done after 90 s of dobutamine infusion. ECG recordings inmice 14 d after infarction were made in baseline conditions and then after in-traperitoneal injection of 1 µg isoproterenol per g body weight (Isuprel;Sanofi-Winthrop, Brussels, Belgium) over 15 min or until the death of themouse. Stroke volume and cardiac output were measured as described in ratsand were adapted for the mouse41.

AcknowledgmentsThe authors thank J.M. Herbert and F. Bono for measurements of TGFβ-1; I. VanHorebeek for help with ATP and lactate measurements; R. Gerard for the AdRR5and AdPAI-1 adenovirus, M. Van Bilzen, P. Doevedans and M. Palmen; and A.Bouché, M. De Mol, I. Cornelissen, B. Hermans, K. Deroover, A. Manderveld, S.Jansen, A. Vandenhoeck, P. Van Wesemael, S. Wyns and F. Thoné. S.H. is a Fundfor Scientific Research. research assistant; G.T. is a recipient of the GermanResearch Foundation.

RECEIVED 6 JULY; ACCEPTED 16 AUGUST 1999

1. Reddy, S.G. & Roberts, W.C. Frequency of rupture of the left ventricular free wall orventricular septum among necropsy cases of fatal acute myocardial infarction since in-troduction of coronary care units. Am. J. Cardiol. 63, 906–911 (1989).

2. Varbella, F. et al. Subacute left ventricular free-wall rupture in early course of acute my-ocardial infarction. Clinical report of two cases and review of the literature. G. Ital.Cardiol. 29, 163–170 (1999).

3. Dalrymple-Hay, M.J., Monro, J.L., Livesey, S.A. & Lamb, R.K. Postinfarction ventricularseptal rupture: the Wessex experience. Semin. Thorac. Cardiovasc. Surg. 10, 111–116(1998).

4. Zahger, D. et al. Left ventricular free wall rupture as the presenting manifestation ofacute myocardial infarction in diabetic patients. Am. J. Cardiol. 78, 681–682 (1996).

5. Peuhkurinen, K., Risteli, L., Jounela, A. & Risteli, J. Changes in interstitial collagen me-tabolism during acute myocardial infarction treated with streptokinase or tissue plas-minogen activator. Am. Heart J. 131, 7–13 (1996).

6. Knoepfler, P.S., Bloor, C.M. & Carroll, S.M. Urokinase plasminogen activator activity isincreased in the myocardium during coronary artery occlusion. J. Mol. Cell Cardiol. 27,1317–1324 (1995).

7. Rohde, L.E. et al. Matrix metalloproteinase inhibition attenuates early left ventricularenlargement after experimental myocardial infarction in mice. Circulation 99,3063–3070 (1999).

8. Robert, V. et al. Differential regulation of matrix metalloproteinases associated withaging and hypertension in the rat heart. Lab Invest. 76, 729–738 (1997).

9. Tyagi, S.C., Kumar, S., Cassatt, S. & Parker, J.L. Temporal expression of extracellularmatrix metalloproteinases and tissue plasminogen activator in the development of col-lateral vessels in the canine model of coronary occlusion. Can. J. Physiol. Pharmacol. 74,983–995 (1996).

10. Frangogiannis, N.G. et al. Cytokines and the microcirculation in ischemia and reperfu-sion. J. Mol. Cell Cardiol. 30, 2567–2576 (1998).

11. Carmeliet, P. et al. Urokinase-generated plasmin activates matrix metalloproteinasesduring aneurysm formation. Nature Genet. 17, 439–444 (1997).

12. Carmeliet, P. et al. Inhibitory role of plasminogen activator inhibitor-1 in arterial woundhealing and neointima formation: a gene targeting and gene transfer study in mice.Circulation 96, 3180–3191 (1997).

13. George, S.J., Johnson, J.L., Angelini, G.D., Newby, A.C. & Baker, A.H. Adenovirus-me-diated gene transfer of the human TIMP-1 gene inhibits smooth muscle cell migrationand neointimal formation in human saphenous vein. Hum. Gene Ther. 9, 867–877(1998).

14. Alcorn, J.L. et al. Genomic elements involved in transcriptional regulation of the rabbitsurfactant protein-A gene. Mol. Endocrinol. 7, 1072–1085 (1993).

15. Everts, V., van der Zee, E., Creemers, L. & Beertsen, W. Phagocytosis and intracellulardigestion of collagen, its role in turnover and remodelling. Histochem. J. 28, 229–245(1996).

16. Nunes, I. et al. Structure and activation of the large latent transforming growth factor-beta complex. Int. J. Obes. Relat. Metab. Disord. 20 Suppl 3, S4–8 (1996).

17. Severs, N.J. Pathophysiology of gap junctions in heart disease. J. Cardiovasc.Electrophysiol. 5, 462–475 (1994).

18. Carmeliet, P. et al. Role of HIF-1alpha in hypoxia-mediated apoptosis, cell proliferation,and tumor angiogenesis. Nature 394, 485–490 (1998).

19. Vandeplassche, G., Hermans, C., Thone, F. & Borgers, M. Stunned myocardium has in-creased mitochondrial NADH oxidase and ATPase activities. Cardioscience 2, 47–53(1991).

20. Feldhaus, L.M. & Liedtke, A.J. mRNA expression of glycolytic enzymes and glucosetransporter proteins in ischemic myocardium with and without reperfusion. J. Mol. CellCardiol. 30, 2475–2485 (1998).

21. Przyklenk, K., Connelly, C.M., McLaughlin, R.J., Kloner, R.A. & Apstein, C.S. Effect ofmyocyte necrosis on strength, strain, and stiffness of isolated myocardial strips. Am.Heart J. 114, 1349–1359 (1987).

22. Schaffer, C.J. & Nanney, L.B. Cell biology of wound healing. Int. Rev. Cytol. 169,151–181 (1996).

23. Koyama, H., Raines, E.W., Bornfeldt, K.E., Roberts, J.M. & Ross, R. Fibrillar collagen in-hibits arterial smooth muscle proliferation through regulation of Cdk2 inhibitors. Cell87, 1069–1078 (1996).

24. Plesner, T., Behrendt, N. & Ploug, M. Structure, function and expression on blood andbone marrow cells of the urokinase-type plasminogen activator receptor, uPAR. StemCells 15, 398–408 (1997).

25. Kleiner, D.E. & Stetler-Stevenson, W.G. Matrix metalloproteinases and metastasis.Cancer Chemother. Pharmacol. 43, S42–51 (1999).

26. Chapman, H.A. Plasminogen activators, integrins, and the coordinated regulation ofcell adhesion and migration. Cur. Opin. Cell Biol. 9, 714–724 (1997).

27. Carmeliet, P. et al. Receptor-independent role of urokinase-type plasminogen activatorin arterial wound healing and intima formation in mice. J. Cell Biol. 140, 233–245(1998).

28. Rifkin, D.B., Mazzieri, R., Munger, J.S., Noguera, I. & Sung, J. Proteolytic control ofgrowth factor availability. APMIS 107, 80–85 (1999).

29. Borregaard, N. & Cowland, J.B. Granules of the human neutrophilic polymorphonu-clear leukocyte. Blood 89, 3503–3521 (1997).

30. Lee, J.K. et al. A serine elastase inhibitor reduces inflammation and fibrosis and pre-serves cardiac function after experimentally-induced murine myocarditis. Nature Med.4, 1383–1391 (1998).

31. Zhang, B. et al. Functional polymorphism in the regulatory region of gelatinase B genein relation to severity of coronary atherosclerosis. Circulation 99, 1788–1794 (1999).

32. Lijnen, H.R. & Collen, D. Congenital and acquired deficiencies of components of thefibrinolytic system and their relation to bleeding and thrombosis. Fibrinolysis 3, 67–77(1989).

33. Hamsten, A. et al. Plasminogen activator inhibitor in plasma: risk factor for recurrentmyocardial infarction. Lancet 2, 3–9 (1987).

34. Henry, M. et al. Metabolic determinants are much more important than genetic poly-morphisms in determining the PAI-1 activity and antigen plasma concentrations: afamily study with part of the Stanislas cohort. Arterioscler. Thromb. Vasc. Biol. 18, 84–91(1998).

35. Carmeliet, P. & Collen, D. Vascular development and disorders: molecular analysis andpathogenetic insights. Kidney Int. 53, 1519–1549 (1998).

36. Carmeliet, P. & Collen, D. Development and disease in proteinase-deficient mice: roleof the plasminogen, matrix metalloproteinase, and coagulation system. Thromb. Res.91, 255–285 (1998).

37. Lutgens, E. et al. Chronic myocardial infarction in mice: cardiac structural and func-tional changes. Cardiovasc. Res. 41, 586–593 (1999).

38. Carmeliet, P. et al. Impaired myocardial angiogenesis and ischemic cardiomyopathy inmice lacking the vascular endothelial growth factor isoforms VEGF164 and VEGF188.Nature Med. 5, 495–502 (1999).

39. Linton, M.F., Atkinson, J.B. & Fazio, S. Prevention of atherosclerosis in apolipoprotein E-deficient mice by bone marrow transplantation. Science 267, 1034–1037 (1995).

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m