Matrix metalloproteinase-1 is associated with poor prognosis in oesophageal cancer

, . 185: 256–261 (1998)

MATRIX METALLOPROTEINASE-1 IS ASSOCIATEDWITH POOR PROGNOSIS IN OESOPHAGEAL CANCER

. 1*, . 2, ’1, . 1, . 2 . 2

1Department of Pathology, University of Aberdeen, Aberdeen, AB25 2ZD, U.K.2Department of Molecular and Cell Biology, University of Aberdeen, Aberdeen, AB25 2ZD, U.K.

SUMMARY

The matrix metalloproteinases (MMPs) are a family of closely related proteolytic enzymes which are involved in the degradation ofdifferent components of the extracellular matrix. There is increasing evidence to indicate that individual MMPs have an important rolein tumour invasion and tumour spread. Monoclonal antibodies specific for MMP-1, MMP-2, or MMP-9 have been produced, using asimmunogens peptides selected from the amino acid sequences of individual MMPs. The presence of MMP-1, MMP-2, and MMP-9 inoesophageal cancer was investigated by immunohistochemistry on formalin-fixed, wax-embedded sections of oesophageal cancers. Therelationship of individual MMPs to prognosis and survival was determined. MMP-1 was present in 24 per cent of oesophageal cancers,while MMP-2 and MMP-9 were present in 78 and 70 per cent of tumours, respectively. The presence of MMP-1 was associated witha particularly poor prognosis (log rank test 8·46, P<0·004) and was an independent prognostic factor (P=0·02). The identification ofindividual MMPs in oesophageal cancer provides a rational basis for use in the treatment of oesophageal cancer of MMP inhibitorswhich are currently undergoing clinical trial. ? 1998 John Wiley & Sons, Ltd.

J. Pathol. 185: 256–261, 1998.

KEY WORDS—collagenase; interstitial collagenase; matrix metalloproteinase; neoplasm; oesophagus

INTRODUCTION

Cancer of the oesophagus is one of the commonestmalignant tumours of the alimentary tract, with a world-wide distribution. Oesophageal cancer is characterizedby a late clinical presentation, rapid progression, andvery poor survival.1 The reason for this poor prognosisis that at the time of diagnosis, oesophageal cancerusually shows extensive local tumour invasionand frequent spread to metastatic sites, particularlyregional lymph nodes. Spread of malignant tumoursis a multi-step process and many of the stages oftumour invasion require degradation or breakdown ofthe extracellular matrix and connective tissue surround-ing tumour cells.2,3 The matrix metalloproteinases(MMPs) are a family of zinc-containing enzymes4,5

which are involved in the degradation of differentcomponents of the extracellular matrix, and there isconsiderable evidence to indicate that individual MMPshave an important role in tumour invasion and tumourspread.5–9

The MMPs have been classified into collagenases,gelatinases, and stromelysins, based on the in vitrosubstrate specificity of individual MMPs.4,10 TheseMMPs are secreted as inactive precursors which areactivated by cleavage of an N-terminal pro-peptide.With the exception of matrilysin (MMP-7), which hasonly a catalytic domain, the activated MMPs all have an

*Correspondence to: Dr Graeme I. Murray, Department of Pathol-ogy, University of Aberdeen, Foresterhill, Aberdeen, AB25 2ZD, U.K.E-mail: [email protected]

Contract grant sponsor: Scottish Hospital Endowments ResearchTrust.

CCC 0022–3417/98/070256–06 $17.50? 1998 John Wiley & Sons, Ltd.

N-terminal catalytic domain, which has structural simi-larities with the N-terminal domains of thermolysin andother zinc endopeptidases such as astacin,11,12 and aC-terminal domain which has homology with haemo-pexin.5,13 The three-dimensional structure of theC-terminal domain in porcine synovial collagenasehas been shown to be a four-bladed â-propeller, eachblade of which contains four anti-parallel â-strands.14

The catalytic and C-terminal domains of the MMPsare joined by a linker of varying length. The gelatinases(MMP-2 and MMP-9), which are also known astype IV collagenases, have an additional fibronectin-like, collagen binding domain inserted into thecatalytic domain, and in MMP-9 the linker betweenthe catalytic and C-terminal domains is especiallylong. More recently, several MMPs have been identifiedin tumour cells which are not secreted but aremembrane-bound by means of a trans-membranedomain; these have been designated membrane-typeMMPs.15,16

The gelatinases, particularly MMP-2, appear to beimportant in the initial stages of tumour invasion,17 asthey degrade components of the basement membrane,including type IV collagen. In addition, cleavage of onecomponent of the extracellular matrix, laminin-5, byMMP-2 promotes cell migration.18 Local invasion oftumours is facilitated by degradation of the extracellularmatrix and this can be performed by several MMPs,particularly MMP-1.5 Metastatic spread of tumour cellsinvolves invasion into lymphatic and blood vessels andthis process requires breakdown of the basement mem-brane surrounding these vessels. When tumour cellsreach a metastatic site the process of invasion is

Received 25 June 1997Accepted 2 March 1998

257MMP-1 AND OESOPHAGEAL CANCER

repeated, with basement membrane degradation of ves-sels followed by breakdown of the extracellular matrix.2All of these steps in the metastatic cascade involveindividual MMPs.2,5,17

In this study we raised specific monoclonal antibodieswhich allowed us to investigate the presence of MMP-1(interstitial collagenase), MMP-2 (72 kD gelatinase),and MMP-9 (92 kD gelatinase) in oesophageal can-cer. We have shown that the presence of MMP-1 isassociated with poor prognosis and is an independentprognostic factor.

MATERIALS AND METHODS

Tissue

Oesophageal tumour samples (n=46) were obtainedfrom oesophagectomy specimens from patients (agerange 34–70 years; 29 males and 17 females) who hadundergone surgery for oesophageal cancer. All speci-mens which had been submitted to the Department ofPathology, University of Aberdeen had been fixed informalin and embedded in wax. Histologically, therewere 19 squamous carcinomas and 27 adenocarcinomas.One block of tumour from each case was used forimmunohistochemical analysis of individual MMPs, andtissue sections (4 ìm in thickness) were cut and mountedon aminopropyltriethoxysilane (Sigma, Poole, Dorset,U.K.)-coated slides to ensure section adherence. Theadenocarcinomas had occurred in areas of Barrett’soesophagus. TNM staging of the tumours showed 15stage 2a tumours, four stage 2b tumours, 26 stage 3tumours, and one stage 4 tumour. All the patients had

? 1998 John Wiley & Sons, Ltd.

survived for at least 1 month following surgery and werefollowed up for 24–60 months.

Monoclonal antibodies

Monoclonal antibodies to MMP-1, MMP-2 andMMP-9 were produced using synthetic peptides corre-sponding to a region in the multiple alignment of theMMP sequences where unique sequences were present inthe MMP-1, MMP-2, and MMP-9. The amino acidsequences used were SSFGFPRTVKH (MMP-1),TSLGLPPDVQRVD (MMP-2), and KLGLGADVAQVT (MMP-9). The peptides were synthesized at theKrebs Institute, University of Sheffield, Sheffield, andwere checked by amino acid sequence analysis and massspectrometry in the Protein Facility, University ofAberdeen. The sequences of 11, 12, or 13 amino acidresidues were situated in the C-terminal domain of theMMPs, and homology modelling showed that theyformed an external loop between the second and thirdblade of the â-propeller.14 This region is distant from theactive site of the enzymes and the pro-peptide domains,and its positions on the outside of the molecule makesit likely that anti-peptide antibodies would recognizethe intact protein. The peptides were linked at theirN-termini to hen ovalbumin as carrier protein, usingglutaraldehyde for the MMP-2 and MMP-9 peptides19

and m-maleimidobenzoic acid N-hydroxysuccinimideester for the MMP-1 peptide, which had an N-terminalcysteine residue included to allow conjugation by thisreagent.20 The conjugates were used to immunizeBALB-c mice for the generation of monoclonal anti-bodies by standard methods.21,22 The peptides, linked tobovine serum albumin (BSA) by the above methods,

Fig. 1—Immunoblot showing the specificity of the anti-MMP-1, anti-MMP-2 andanti-MMP-9 antibodies. Individual MMP pro-enzymes (0·4 ìg of purified protein per lane)were subjected to SDS-PAGE and then electrophoretically transferred to PVDF mem-brane. (A) The membrane was immunostained with the MMP-1 antibody; (B) themembrane was immunostained with the MMP-2 antibody; (C) the membrane wasimmunostained with the MMP-9 antibody. Lane 1, MMP-1; lane 2, MMP-3; lane 3,MMP-2; lane 4, MMP-9

, . 185: 256–261 (1998)

258 G. I. MURRAY ET AL.

were used for initial antibody screening. The BSAconjugates were bound to an ELISA plate by incu-bation overnight at 4)C in 50 m sodium carbonate/bicarbonate buffer, pH 9·6, and the ELISA wasperformed as described previously.23

The specificity of the monoclonal antibodies wasdemonstrated by immunoblotting against purified MMPpro-enzymes (MMP-1, MMP-2, MMP-3, MMP-9).Natural human MMP-1 and recombinant humanMMP-2, MMP-3, and MMP-9 were a kind gift from Dr


Alan Galloway, British Biotech plc. Individual MMPs(0·4 ìg of purified protein per lane) were subjected toSDS-PAGE (10 per cent acrylamide) and then electro-phoretically transferred to PVDF (Problott, AppliedBiosystems, Warrington, U.K.) membrane. The mem-brane was then divided into three identical sections andeach section incubated with one of the MMP antibodies.Antibody which had reacted with antigen on themembrane was detected using alkaline phosphatase-conjugated goat anti-mouse IgG (Fc specific, Sigma)with the detection system 5-bromo-4-chloro-3-indolylphosphate/nitro blue tetrazolium (BCIP/NBT; Sigma).The antibodies were isotyped with an Isostrip kit(Boehringer Mannheim, Lewes, Sussex, U.K.) usedaccording to the manufacturer’s instructions.

Immunohistochemistry

Sections of oesophageal tumours were immunostainedwith monoclonal antibodies to MMP-1 (clone 3B6),MMP-2 (clone 4D3), and MMP-9 (clone 2C3).Immunohistochemistry for the individual MMPs wasperformed using an alkaline phosphatase anti-alkalinephosphatase technique.24 Tissue sections were subjectedto an antigen retrieval step by microwave treatment ofthe sections for 10 min (MMP-1 and MMP-9) or 5 min(MMP-2) in 0·01 sodium citrate buffer, pH 6·0, in amicrowave (Proline=) operated at full power (950 W).The sections were allowed to cool to room temperaturebefore application of the indivdual MMP antibodiesas undiluted tissue culture supernatants. Followingincubation with the monoclonal antibodies, sectionswere sequentially incubated with rabbit anti-mouseimmunoglobulin (1/100; Dako, High Wycombe, Berks,U.K.) and monoclonal alkaline phosphatase anti-alkaline phosphatase (1/100; Dako). Following eachantibody application, sections were washed in 0·05 Tris–HCl buffered saline, pH 7·6 (TBS). Alkaline phos-phatase was demonstrated using BCIP/NBT. After incu-bating the sections for 30 min at room temperature, thereaction was stopped by washing the sections for 5 minin hot tap water. The slides were then air-dried andmounted in glycerine jelly. When the immunohistochem-istry was complete, the sections were examined micro-scopically. The MMP status of the tumours was assessedas positive if any of the tumour cells showed significantimmunostaining.25 Negative controls were done byreplacing the primary antibody with TBS and by liquidphase pre-absorption of primary antibody with thecorresponding immunogen at 10 nmol/ml antibody. Thepositive control for MMP-1 was colonic adenocarci-noma which had previously been shown to be positivefor MMP-1,24 while the positive controls for bothMMP-2 and MMP-9 were lung containing intra-alveolarmacrophages.

Survival data were obtained after the assessment ofthe presence of MMPs in the oesophageal cancers werecomplete. Statistical analysis was performed using thecomputer program SPSS for Windows=. Cumulativepatient survival was assessed by the method of Kaplan–Meier, and comparison of the MMP-positive and MMP-negative survival curves was carried out using the

Fig. 2—Immunoreactivity for (a) MMP-1, (b) MMP-2, and (c)MMP-9 in oesophageal cancer. The tumour in a is a poorly differen-tiated squamous carcinoma, while the tumours in b and c aremoderately differentiated adenocarcinomas. Immunoreactivity isfound evenly distributed over all the tumour cells

, . 185: 256–261 (1998)


log-rank test. Multi-variate analysis using Cox’s methodwas performed to identify independent prognostic fac-tors. The variables which were included in the multi-variate analysis were age, sex of patients, tumour stage,histological type of tumour, and status of individualMMPs.

RESULTS

The immunoblots showed that the monoclonal anti-bodies were specific for the appropriate MMP and didnot cross-react with other MMPs (Fig. 1). The electro-phoretic mobility of the slightly lower band visible in thelane containing MMP-1 of panel A after the membranehad been incubated with the MMP-1 antibody is con-sistent with that expected for the activated form ofMMP-1. All three monoclonal antibodies had an isotypeof IgG1ê.

All the antibodies were effective on formalin-fixedsections after an antigen retrieval step. Immunoreactiv-ity for each MMP was detected in oesophageal tumourcells (Fig. 2): when staining was present, it was found inalmost all the tumour cells. Summaries of the occurrenceof MMPs in different stages and different histologicaltypes of oesophageal cancer are shown in Tables I andII. There were no significant differences in the presenceof individual MMPs between squamous carcinomaand adenocarcinoma. MMP-2 and MMP-9 were bothpresent in the majority of oesophageal tumours, withMMP-2 present in 36 (78 per cent) tumours and MMP-9in 32 (70 per cent) tumours, respectively. Only oneMMP-9-positive tumour did not also express MMP-2.MMP-1 was detected in 11 (24 per cent) oesophagealtumours, with all but two of these tumours also contain-ing MMP-2: all but one contained MMP-9. Normaloesophageal epithelium did not show immunoreactivityfor any of the MMPs, while fibroblasts consistentlyshowed MMP-1 immunoreactivity and macrophagesshowed immunostaining for MMP-2 and MMP-9.

Survival analysis showed that the presence of MMP-1was associated with more rapid progression and that all11 patients whose tumours contained MMP-1 were deadby 29 months, while nine of the patients whose tumourswere MMP-1-negative remained alive at 60 months.


The median survival in the MMP-1-positive group was7 months, while the median survival in the MMP-1-negative group was 16 months (Fig. 3). The difference insurvival was significant between the MMP-1-positiveand MMP-1-negative groups (log-rank test 8·46,P=0·004). There were no significant survival differencesbetween those patients whose tumours were MMP-2-positive or -negative: the same was true for MMP-9.Multi-variate analysis showed that MMP-1 was anindependent prognostic factor (P=0·02). The most sig-nificant prognostic factor was tumour stage (P=0·005),

Fig. 3—Cumulative survival of oesophageal cancer patients whosetumours were either positive or negative for MMP-1

Table I—The presence of MMPs in different stages of oesophageal cancer

Tumour stage

Total2a 2b 3 4

MMP-1Positive 2 (4%) 2 (4%) 6 (14%) 1 (2%) 11 (24%)Negative 13 (28%) 2 (4%) 20 (44%) 0 35 (76%)

MMP-2Positive 13 (28%) 4 (10%) 18 (39%) 1 (2%) 36 (79%)Negative 2 (4%) 0 8 (17%) 0 10 (21%)

MMP-9Positive 13 (28%) 3 (7%) 15 (33%) 1 (2%) 32 (70%)Negative 2 (4%) 1 (2%) 11 (24%) 0 14 (30%)

Table II—The presence of individual MMPs in differenthistological types of oesophageal cancer

Histological type

TotalSquamouscarcinoma Adenocarcinoma

MMP-1Positive 3 (7%) 8 (17%) 11 (24%)Negative 16 (35%) 19 (41%) 35 (76%)



, . 185: 256–261 (1998)

260 G. I. MURRAY ET AL.

while other factors (patient age, sex of patient,histological type of tumour, MMP-2 and MMP-9 status)were all non-significant.

DISCUSSION

Improved survival for patients with oesophageal can-cer needs the development of new treatment strategies,and this requires the identification of prognostic factorswhich can be targetted by therapeutic intervention. TheMMPs are one group of molecules which are potentialtargets for novel anti-cancer drugs. There is an increas-ing amount of evidence to indicate that matrix degra-dation by MMPs is an important part of the multi-stepprocess of tumour invasion and spread.3 As the MMPsare a group of closely related enzymes, it is important tohave antibodies specific for individual MMPs, and wehave demonstrated that this is the case for the antibodieswhich we have developed. Because of the location of theepitopes on an external loop of the individual MMPs ata site remote from the active site and the pro-peptidedomain, it is likely that they recognize the activatedforms as well as the pro-enzymes. These monoclonalantibodies, which recognize individual MMPs and areeffective on formalin-fixed tissue sections, will greatlyfacilitate the evaluation of MMPs in tumours.

Very few studies of MMPs in tumours have correlatedthe presence of individual MMPs to prognosis andsurvival. The first stage in the invasion of malignanttumour cells is the degradation of basement membrane.Type IV collagen is the major component of the base-ment membrane and degradation of this is mainlyachieved by MMP-2 and possibly by MMP-9.5,17 Fur-ther tumour invasion requires the action of otherMMPs, such as MMP-1, which can degrade more com-ponents of the extracellular matrix. The majority ofoesophageal cancers showed the presence of MMP-2and this is consistent with the study of Shima et al.,26

who also showed the presence of MMP-2 in oesophagealcancer cells. That study, which also localized MMP-1and MMP-3 to oesophageal tumour cells, represents theonly previous investigation of MMPs in oesophagealcancer.

In oesophageal cancer, as in colorectal cancer,25

MMP-1 appears to be important in facilitating furthertumour spread. In our study, the presence of MMP-1was associated with a particularly poor prognosis andwas an independent prognostic factor. Stage of oeso-phageal tumour was the most important prognosticfactor, while patient age, sex, MMP-2 status, MMP-9status, and histological type of oesophageal tumourwere not significant factors in influencing survival. Allthe patients who were included in the study had survivedfor at least 1 month following surgery, thereby excludingbias that post-operative mortality might introduce tosurvival analysis. Although two main types of oeso-phageal cancer are recognized histologically, bothtumour types invade and spread in a similar manner andhave an equally poor prognosis.

Low molecular weight inhibitors of MMPs have beenshown to prevent tumour spread in human tumour


xenografts27,28 and are being developed for clinical use:at least one MMP inhibitor (Marimastat) is currentlyundergoing clinical trials.29 In view of this, it is import-ant to have objective markers that could potentially beused to identify those patients who are most likely tobenefit from anti-MMP therapy. Interstitial collagenase(MMP-1)-activated 5-fluoro-uracil pro-drugs30 couldalso be appropriately used in those patients whosetumours contain MMP-1. The MMP status of oeso-phageal tumours can readily be assessed using im-munohistochemistry on formalin-fixed, wax-embeddedsections of tumours.

ACKNOWLEDGEMENTS

We thank the Scottish Hospital EndowmentsResearch Trust for financial support and Mr BryanDunbar and Mr Ian Davidson of the Universityof Aberdeen Protein Facility for amino acid sequnceanalysis and mass spectrometry.

REFERENCES

1. Morson BC, Dawson IMP, Day DW, Jass JR, Price AB, Williams GT.Morson and Dawson’s Gastrointestinal Pathology. 3rd edn. Oxford:Blackwell Scientific Publications, 1990; 53–70.

2. Hart IR, Saini A. Biology of tumour metastasis. Lancet 1992; 339: 1453–1457.

3. Kohn EC, Liotta LA. Molecular insights into cancer invasion: strategies forprevention and intervention. Cancer Res 1995; 55: 1856–1862.

4. Murphy G, Docherty AJP. The matrix metalloproteinases and theirinhibitors. Am J Respir Cell Mol Biol 1992; 7: 120–125.

5. Stetler-Stevenson WG, Liotta LA, Kliener DE. Extracellular matrix 6: roleof matrix metalloproteinases in tumor invasion and metastasis. FASEB J1993; 7: 1434–1441.

6. Davies B, Waxman J, Wasan H et al. Levels of matrix metalloproteases inbladder cancer correlate with tumor grade and invasion. Cancer Res 1993;53: 5365–5369.

7. Boag AH, Young ID. Increased expression of the 72-kd type IV collagenasein prostate adenocarcinoma. Am J Pathol 1994; 144: 585–591.

8. Muller D, Wolf C, Abecassis J et al. Increased stromelysin 3 gene expressionis associated with increased local invasiveness in head and neck squamouscell carcinoma. Cancer Res 1993; 53: 165–169.

9. Urbanski SJ, Edwards DR, Hershfield N et al. Expression pattern ofmetalloproteinases and their inhibitors changes with the progression ofhuman sporadic colorectal neoplasia. Diagn Mol Pathol 1993; 2: 81–89.

10. Woessner JF. Matrix metalloproteinases and their inhibitors in connectivetissue remodelling. FASEB J 1991; 5: 2145–2154.

11. Bode W. The X-ray crystal structure of the catalytic domain of humanneutrophil collagenase inhibited by a substrate analogue reveals theessentials for catalysis and specificity. EMBO J 1994; 13: 1263–1269.

12. Lovejoy B, Cleasby A, Hassell AM et al. Structure of the catalytic domainof fibroblast collagenase complexed with an inhibitor. Science 1994; 263:375–377.

13. Hunt LT, Barker WC, Chen HR. A domain structure common to hemo-pexin, vitronectin, interstitial collagenase and a collagenase homolog. ProtSeq Data Anal 1987; 1: 21–26.

14. Li J, Brick P, O’Hare MC et al. Structure of full-length porcine synovialcollagenase reveals a C-terminal domain containing a calcium linkedfour-bladed â-propellor. Structure 1995; 3: 541–549.

15. Sato H, Seiki M. Membrane-type matrix metalloproteinases (MT-MMPs)in tumor metastasis. J Biochem 1996; 119: 209–215.

16. Sato H, Takino T, Okada Y et al. A matrix metalloproteinase expressed onthe surface of invasive tumour cells. Nature 1994; 370: 61–65.

17. Tryggvason K, Höyhtä M, Pyke M. Type IV collagenases in invasivetumors. Breast Cancer Res Treat 1993; 24: 209–218.

18. Giannelli G, Falk-Marzillier J, Schiraldi O, Stetler-Stevenson WG,Quaranta V. Induction of cell migration by matrix metalloprotease-2cleavage of laminin-5. Science 1997; 277: 225–228.

19. Hiraga A, Kemp BE, Cohen P. Further studies on the structure of theglycogen-bound form of protein phosphatase-1 from rabbit skeletal muscle.Eur J Biochem 1987; 163: 253–258.

20. Sambrook J, Fritsch EF, Maniatis T. Molecular Cloning: A LaboratoryManual. New York: Cold Spring Harbor Laboratory Press, 1989.

, . 185: 256–261 (1998)


21. Kohler G, Milstein C. Continuous cultures of fused cells secreting antibodyof predefined specificity. Nature 1975; 256: 495–497.

22. Barnes TS, Shaw PM, Burke MD, Melvin WT. Monoclonal antibodiesagainst human cytochrome P-450 recognising different pregnenolone 16á-carbonitrile-inducible rat cytochrome P-450. Biochem J 1987; 248: 301–304.

23. Duncan ME, McAleese SM, Booth NA, Melvin WT, Fothergill JE. Asimple enzyme-linked immunosorbent assay (ELISA) for the neuron-specific ã isozyme of human enolase (NSE) using monoclonal antibodiesraised against synthetic peptides corresponding to isozyme sequence differ-ences. J Immunol Methods 1992; 151: 227–236.

24. Murray GI, Duncan ME, Melvin WT, Fothergill JE. Immunohistochemis-try of neuron specific enolase with gamma-subunit specific anti-peptidemonoclonal antibodies. J Clin Pathol 1993; 46: 993–996.

25. Murray GI, Duncan ME, O’Neil P, Melvin WT, Fothergill JE. Matrixmetalloproteinase-1 is associated with poor prognosis in colorectal cancer.Nature Med 1996; 2: 461–462.


26. Shima I, Sasaguri Y, Kusukawa J et al. Production of matrixmetalloproteinase-2 and metalloproteinase-3 related to malignant behaviourof esophageal carcinoma. Cancer 1992; 70: 2747–2753.

27. Watson SA, Morris TM, Robinson G, Crimmin MJ, Brown PD, HardcastleJD. Inhibition of organ invasion by the matrix metalloproteinase inhibitorbatimastat (BB-94) in two human colon carcinoma metastasis models.Cancer Res 1995; 55: 3629–3633.

28. Sledge GW, Qulali M, Goulet R, Bone EA, Fife R. Effects of matrixmetalloproteinase inhibitor batimastat on breast cancer regrowth andmetastasis in athymic mice. J Natl Cancer Inst 1995; 87: 1546–1550.

29. Brown PD, Giavazzi R. Matrix metalloproteinase inhibition: a review ofanti-tumour activity. Ann Oncol 1995; 6: 967–974.

30. Nichifor M, Schact EH, Seymour LW. Macromolecular prodrugs of 5fluorouracil. 2: Enzymatic degradation. J Cont Release 1996; 39: 79–92.

, . 185: 256–261 (1998)

Documents

Matrix metalloproteinase-1 is associated with poor prognosis in oesophageal cancer