J. Pathol. 188: 56–62 (1999)

MATRIX METALLOPROTEINASE-2 (MMP-2)IMMUNOREACTIVE PROTEIN—A NEW PROGNOSTIC

MARKER IN UVEAL MELANOMA?

1*, 2, 3, 4, 5 -1*

1Department of Oncology and Radiotherapy, University Hospital of Oulu, Finland2Department of Pathology, University Hospital of Oulu, Finland

3Department of Ophthalmology, University Hospital of Oulu, Finland4Department of Ophthalmology, University Hospital of Helsinki, Finland

5Diabor Ltd., Oulu, Finland

SUMMARY

Uveal melanoma is the most common primary intraocular tumour. Once haematogenous metastasis has occurred, there is no cure forthe disease and there is an obvious need for new biological prognostic markers to estimate the risk of metastasis. In this study, theexpression of matrix metalloproteinase-2 (MMP-2) was characterized immunohistochemically in 29 human uveal melanomas.Enzyme-linked immunoassays and gelatin zymographies were assessed in order to quantify the expression of gelatinases A and B, as wellas the tissue inhibitor of metalloproteinases (TIMPs), in the vitreous body. A total of 49 per cent of the uveal melanomas displayed apositive immunoreaction for MMP-2 in melanoma cells, the epithelioid cells showing the most frequent staining. There was nocorrelation between the positivity of MMP-2 staining and the size of the primary tumour, gender or age. The expression of MMP-2 wasassociated with a dismal prognosis: the 5-year overall survival rate for MMP-2-positive cases was significantly inferior to that of theMMP-2 negative cases, 49 per cent vs. 86 per cent, respectively (p=0·02). A patient group at high risk of metastatic disease wasidentified; only 38 per cent of patients with a MMP-2-positive non-spindle cell uveal melanoma survived for 5 years. The analyses ofMMPs or TIMPs in the vitreous body had no prognostic value. Positive immunostaining for MMP-2 was observed in the retinal pigmentepithelium, corneal epithelium, and fibroblasts in the ciliary body and choroid. It is concluded that immunohistochemical analysis ofMMP-2 may help to predict a risk of metastasis in uveal melanoma. Copyright ? 1999 John Wiley & Sons, Ltd.

KEY WORDS—gelatinases; uveal melanoma; metastasis; TIMP

*Correspondence to: Professor Taina Turpeenniemi-Hujanen, MD,PhD, Department of Oncology and Radiotherapy, University Hospitalof Oulu, Kajaanintie 50, 90220 Oulu, Finland.

Contract/grant sponsor: Cancer Society of Northern Finland.

INTRODUCTION

Uveal melanoma is the most common primaryintraocular neoplasm in adults, even though it is a raremalignant disease. The 5-year survival in uvealmelanoma is approximately 70 per cent, but within 15years of follow-up, 50 per cent of patients will eventuallydie of metastatic disese.1,2 Uveal and cutaneous melano-mas arise from a similar embryological origin, but theydiffer in biological behaviour. Since the eye lacks lym-phatic drainage, uveal melanomas metastasize only bythe blood stream, most commonly to the liver.

Different proteolytic enzyme systems are involved intumour development, including the matrix metallopro-teinase (MMP) system and the plasminogen activationsystem. Tumour cell adhesion, extracellular matrix pro-teolysis, and cell migration are the basic elements forinvasion and metastasis.3,4 The migration of a malignantcell through the basement membrane requires proteo-lytic activity.3 Previously, the role of MMPs has beenwell established in tumour progression and invasion inboth in vivo and in vitro studies.3,5,6 Several otherproteases, such as plasminogen activators, are also

CCC 0022–3417/99/060056–07$17.50Copyright ? 1999 John Wiley & Sons, Ltd.

involved in the degradation of basal membranes andother components of the extracellular matrix.7,8

The activity of MMPs is regulated by tissue inhibitorsof metalloproteinases (TIMPs). TIMP-2 and TIMP-1inhibit MMPs by forming a complex with both thelatent and the active forms of MMP-2 and MMP-9,respectively.9 TIMP-2 also forms a complex withmembrane-type matrix metalloproteinase (MT-MMP),which functions as a membrane receptor for latentMMP-2 on the cell surface.10 Changes in the balancebetween the free enzyme and the enzyme–inhibitor com-plex may affect the progression of a tumour to a moremalignant phenotype.11 In in vitro studies, the invasionof tumour cells was inhibited by TIMP-2;12 on the otherhand, overexpression of TIMP-2 in breast cancer corre-lated with recurrence rate,13 and with poor survival inthe case of bladder cancer.14 Thus, the role of TIMP-2 incancer metastasis is not fully understood.

Both MMPs15 and plasminogen activators16 havebeen shown to be involved in the progression ofcutaneous melanoma and there are data suggesting thatat least the plasminogen activation system is involved inuveal melanoma invasion.17 It is possible that invasionby a malignant cell would require the synergistic actionof both of these proteinase systems and possibleco-operation between MMPs. Indeed, both MMP-2and MMP-9 (72 kD and 92 kD type IV collagenases,

Received 20 April 1998Revised 15 July 1998

Accepted 8 December 1998

be o

57TYPE IV COLLAGENASE IN HUMAN UVEAL MELANOMA

Copyr

Table I—Clinico-pathological features of the 29 patients with uveal melanoma and expression of MMP-2protein

PatientNo. Gender Histology

Tumourdiameter

(mm)Scleral

infiltration

MMP-2-positivecells (%)

Site ofmetastases

Follow-up(months)

1 M Spindle 12 + 0 — 542 F Spindle ND ND 0 — 83*3 M Spindle 8 " 0 — 854 M Spindle 9 " 0 — 122*5 M Spindle 11 " 0 Lung EX/696 F Spindle 6 " 0 — 1207 F Spindle 10 " 0 — 1188 M Spindle 14 + 0 Liver EX/989 F Spindle 10 " 0 — 122

10 M Spindle 8 " 29 — 19411 M Spindle 10 " 21 Lung EX/11012 M Spindle 15 " 11 Lung EX/1013 F Spindle 7 " 20 — 15514 F Mixed 15 " 0 Liver EX/2315 F Mixed 10 + 0 Liver EX/11116 M Mixed 5 " 0 — 7217 F Mixed 12 + 0 Skin 6218 M Mixed 20 " 0 Liver EX/6019 M Mixed 13 + 0 — 5420 F Mixed 12 " 42 Liver EX/5421 F Mixed 14 + 19 Liver EX/1122 F Mixed 15 + 40 Lung EX/15223 M Mixed 15 + 47 Liver EX/3624 M Mixed 19 " 32 — 5525 M Mixed 9 " 16 — 60*26 F Epithelioid 5 " 20 Brain EX/5827 M Epithelioid 17 + 47 Liver EX/828 F Epithelioid 9 " 86 Liver EX/3129 F Epithelioid 10 " 19 Liver EX/61

*Died of a cause other than uveal melanoma.ND=not determined; EX=died of metastastic uveal melanoma.

respectively) have been detected in the culture mediumof cell lines derived from uveal melanomas.18 So far,there have been no immunohistochemical data pub-lished on MMP-2 expression in human uveal melanoma.

In uveal melanoma, there are a few well-establishedprognostic markers, such as tumour size and height, celltype, tumour location, age, and gender. Vascular net-works and cell proliferation have also been suggested.1,2

There is, however, still a need for a prognostic markerwhich could evaluate the invasive propensity ofmalignant cells and perhaps predict the occurrence ofhaematogenous metastases.

The aim of this study was to investigate the expressionof MMP-2 immunoreactive protein in primary uvealmelanoma. In addition, we evaluated the possible role ofMMP-2 as a prognostic marker in order to identifyindividual uveal melanoma patients who might benefitfrom other therapeutic modalities in conjunction withthe primary treatment. It was also of interest to find outwhether MMPs or TIMPs could be identified in thevitreous body. Since small tumours are treated withirradiation, new prognostic factors for these cases could

f clinical value.

ight ? 1999 John Wiley & Sons, Ltd.

MATERIALS AND METHODS

Materials

Altogether 29 patients were included in the finalanalysis, 15 (52 per cent) men and 14 (48 per cent)women (Table I). The mean age of diagnosis was 61·2years (range 23–83 years), (men 60·3 years, range 23–75years and women 62·2 years, range 23–75 years). Theduration of the follow-up was a minimum of 54 months.The primary therapy consisted of enucleation of theaffected eye. Fifteen patients died of metastatic diseaseduring the follow-up (Table I). One patient was excludedfrom the study due to another type of cancer 55 monthsafter enucleation of the eye.

Formalin-fixed and paraffin-embedded tissue samplesof primary uveal melanomas were obtained from thefiles of the Department of Pathology, Oulu UniversityHospital for retrospective immunohistochemical study.Consecutive uveal melanoma cases from 1978 to 1994were included in the study, according to the availabilityof adequate histological material. Primary tumourcell type, diameter, and scleral and/or optical nerveinfiltration were determined.19

J. Pathol. 188: 56–62 (1999)

58 A. VÄISÄNEN ET AL.

Vitreous samples were obtained from therapeuticvitrectomies for the enzyme-linked immunoassaysand gelatin zymography. A total of 21 samples of thevitreous body were obtained during 1990–1995 frompatients with uveal melanoma (six patients), retinaldetachment or tear (eight patients), opacities of thevitreous body (five patients) or other eye disorders (threepatients).

Immunohistochemistry

Paraffin sections (4 ìm) of uveal melanomas werestained using the avidin–biotin–immunoperoxidase tech-nique.20 The sections were incubated at 37)C for 12 h,dewaxed (Histo-Clear>, National Diagnostics, Atlanta,GA, U.S.A.), and re-hydrated. After pretreatment with0·4 per cent pepsin (Sigma, St Louis, MO, U.S.A.) at37)C for 20 min, the specimens were stained as describedearlier.21,22 A mouse monoclonal antibody for MMP-2,which recognizes the latent form of MMP-2, was appliedas a primary antibody23 (1·5 ìg/ml in 0·9 per cent NaCl,1 per cent bovine serum, and 0·01 phosphate buffer,pH 7·5). The antibody reaction was visualized by a freshsubstrate solution containing an aminoethyl-carbazolesubstrate kit (AEC kit>, Zymed, San Francisco, CA,U.S.A.). A mouse non-immuno-IgG served as a negativecontrol and MMP-2-positive cutaneous melanomasamples were used as positive controls. The sectionswere analysed microscopically by three independentobservers. If more than 5 per cent of the malignant cellsdisplayed positivity, the case was considered to bepositive for MMP-2. Semi-quantitation was performedby counting 300 consecutive malignant cells and thepercentage of MMP-2-positive melanoma cells wasdetermined.

Gelatin zymography

MMP-2 and MMP-9 enzyme activities in the vitreousbody were assessed by gelatin zymography modifiedfrom Heussen and Dowdle24,25 by using 1·0 mggelatin/ml gel as a substrate. After drying the gel, thegelatinase activity was quantified by scanning densi-tometry (MK III CS, Joyce Loebl, Gateshead on Tyne,U.K.). Culture medium samples from the humanmelanoma cell line A2058 and the human breast cancercell line MCF-7 were run in parallel as positive controlsto show the migration of MMP-2 and MMP-9 activity.

Enzyme-linked immunosorbent assays

ELISAs were performed on 96-well microtitre plates(Nunc Maxisorb, Denmark) using standard protocols. Apolyclonal antibody produced in chicken against eachof the analytes was used as a secondary antibody.A peroxidase-labelled anti-chicken IgG (Chemicon,U.S.A.) was used for detection of the bound secondaryantibody. OPD tablets (KemEnTec, Denmark) wereused to visualize the peroxidase label. The colour forma-tion was measured at 450 nm (Anhos 2000 microplatereader) and calculations were done using a Multicalcprogram (Wallac, Finland). For quantitation of

Copyright ? 1999 John Wiley & Sons, Ltd.

MMP-2, two different ELISAs were used, one tomeasure the free form, i.e. not bound to its inhibitor,and the other to measure the MMP-2/TIMP-2 complex.For measurement of the free MMP-2, the ELISA platewas coated with TIMP-2, sample and standards wereapplied, and the bound MMP-2 was detected asdescribed above. For detection of the MMP-2/TIMP-2complex, the plate was coated with a monoclonal anti-TIMP-2 antibody and the bound complex was detectedwith a polyclonal anti-MMP-2 antibody. For the detec-tion of total protein for MMP-9, TIMP-1, and TIMP-2,a protocol to quantitate the free and complex formssimultaneously was used. This was accomplished with amonoclonal antibody which recognizes both the free andthe complex forms of the analyte. The microtitre platewas coated with the monoclonal antibody, samples andstandards were added, and the bound proteins weredetected with polyclonal antibodies as described above.The monoclonal antibodies used in the ELISAs arecharacterized as follows. The monoclonal antibodyagainst MMP-9 (code GE-213) recognizes both thefree MMP-9 and that bound to its inhibitor TIMP-1.26

The monoclonal antibody against TIMP-2 (codeT2-101) recognizes both the free TIMP-2 and that in thecomplex with MMP-2.27 The monoclonal antibodyagainst TIMP-1 (code DB-102D1) recognizes boththe free TIMP-1 and that in the complex with MMP-9.It does not cross-react with TIMP-2 (data not shown).

Statistical analysis

Analysis of the clinical data was performed afterscoring of the MMP-2 immunostaining was completed.The overall and relapse-free survival rates were deter-mined by the Kaplan–Meier method and the differencesbetween the subgroups were analysed by using thelog-rank or Breslow analysis. The chi-squared test wasapplied to evaluate the association between MMP-2expression and scleral infiltration or cell type of theprimary tumour. SPSS for Windows was applied whenpossible.

RESULTS

Fourteen (49 per cent) out of 29 uveal melanomasdisplayed positive intracytoplasmic staining for MMP-2.The staining pattern was usually focal, varying fromfaintly granular to homogeneous (Figs 1A–1C). Theretinal pigment epithelium and the rods and cones werepositive (Fig. 1E). Corneal epithelial cells (Fig. 1F) andfibroblasts in the choroid and ciliary body also oftenshowed positive immunostaining for MMP-2 (Figs 1Gand 1H).

Larger size of uveal melanoma correlated with poorersurvival (p=0·0375, log-rank analysis). The 5-year sur-vival rate in small tumours (<10 mm diameter, n=9)was 78 per cent, in medium-sized tumours (10–15 mmdiameter, n=16) it was 70 per cent, and two out ofthree patients with a large tumour (>15 mm diameter)succumbed to metastatic disease. Gender, age, size of theprimary tumour or scleral infiltration did not have anyimpact on the presence of MMP-2 protein, even though

J. Pathol. 188: 56–62 (1999)

59TYPE IV COLLAGENASE IN HUMAN UVEAL MELANOMA

Copyr

Fig. 1—Immunohistochemical staining of MMP-2 protein in primary human uveal melanoma (A–D) and in normal structuresof the human eye (E–H). (A) Positive intracytoplasmic staining of MMP-2 immunoreactive protein in epithelioid uvealmelanoma. (B) The case presented in A stained with a mouse non-immuno-IgG as a negative control. (C) A spindle cell uvealmelanoma showing positive immunostaining for MMP-2. (D) A spindle cell uveal melanoma with negative immunostaining.(E) The retinal pigment epithelium, (F) the epithelium of the cornea, and (G, H) the ciliary body stained with anti-MMP-2antibody. Magnification (A–F,H) #120; (G) #30

ight ? 1999 John Wiley & Sons, Ltd. J. Pathol. 188: 56–62 (1999)

,

60 A. VÄISÄNEN ET AL.

Fig. 2—Prognostic value of MMP-2 and cell type in uveal melanoma. (A) The overall survival (p=0·02, Breslow analysis) and (B) therelapse-free survival (p=0·04, log-rank analysis) according to MMP-2 expression. (C) Effect of uveal melanoma cell type on survival (p=0·006log-rank analysis). (D) Survival of patients representing unfavourable and favourable prognosis according to MMP-2 expression and cell type(p=0·004, log-rank analysis)

two of the three large tumours (>15 mm) were positive(Table I).

A statistically significant correlation between MMP-2expression and relapse-free or overall survival wasobserved in uveal melanoma (Figs 2A and 2B). The5-year relapse-free and overall survival rates for thepatients with MMP-2-positive uveal melanoma were 38and 49 per cent, respectively (n=14), but 72 and 86 percent, respectively (n=15), for those with MMP-2-negative tumours. Positive MMP-2 immunostaining cor-related significantly with the occurrence of visceralmetastases (Table I) (p<0·05, ÷2-test).

The non-spindle cell tumours showed (63 per cent)positive staining for MMP-2 more often than spindle cell

Copyright ? 1999 John Wiley & Sons, Ltd.

tumours (31 per cent) (Table I). The intensity of MMP-2staining was stronger in epithelioid cell tumours than inspindle cell tumours (Figs 1A–1D). Interestingly, allepithelioid tumours were MMP-2-positive and the rela-tive amount of MMP-2-positive cases differed signifi-cantly from that in the spindle cell tumours (÷2=5·89,p<0·05) or spindle and mixed cell tumours (÷2=4·97,p<0·05). Patients with spindle cell tumours had a better5-year survival (92 per cent) than non-spindle cell tumourpatients (47 per cent) (Fig. 2C). The 5-year survival ratesin patients with spindle, mixed, and epithelioid tumourswere 92, 56, and 25 per cent, respectively (p=0·001).

The most distinct difference in the prognosis appearedbetween the patients with MMP-2-positive non-spindle

J. Pathol. 188: 56–62 (1999)

61TYPE IV COLLAGENASE IN HUMAN UVEAL MELANOMA

cell uveal melanoma and those with MMP-2-negativespindle cell tumours, the 5-year survival rates being 38and 100 per cent, respectively (Fig. 2D). MMP-2 posi-tivity was associated with poor prognosis in all of thedifferent cell types of primary uveal melanoma. OfMMP-2-positive patients with spindle cell tumours, 25per cent succumbed to metastatic disease during the first5 years of follow-up, whereas all MMP-2-negative casesin this histological subtype survived (Table I). Theoverall 5-year survival for the patients with MMP-2-negative mixed cell tumours was 62 per cent com-pared with 50 per cent in patients with MMP-2-positiveprimaries (not shown). All four epithelioid tumours werepositive for MMP-2 and all of these patients succumbedto metastatic uveal melanoma (Table I).

MMP-2, MMP-2/TIMP-2 complex, TIMP-1, andTIMP-2 were all found in the vitreous body in enzymeimmunoassay and gelatinase activity corresponding tothe molecular weight of MMP-2 was detectable in allcases. High levels of these proteins were observed afterdetachment or tear of the retina. In uveal melanoma, thelevels of these proteins in the vitreous body were,however, comparable to the levels of benign disorders ofthe eye other than tear or detachment of the retina (notshown).

DISCUSSION

Metalloproteinases (MMPs) have been associatedwith haematogenous metastasis in various types ofmalignancies.6 We have shown here for the first timethat MMP-2 immunoreactive protein is expressed inhuman uveal melanoma in vivo and may be associatedwith poor prognosis. Moreover, low levels ofMMP-2, TIMP-1, and TIMP-2 can be detected in thevitreous body of uveal melanoma patients with enzymeimmunoassays.

Previously, MMP-2 and MMP-9 have been found inthe medium of primary cultures of uveal melanomas.18

In a human skin melanoma cell line, MMP-2 has beenlinked to in vitro invasion.5,28 We recently showedthe association of MMP-2 protein to haematogenousmetastasis and poor prognosis in skin melanoma.15,21

Intracytoplasmic staining for MMP-2 protein wasfound here in uveal melanoma cells. The expression wasassociated with the non-spindle cell type of uvealmelanoma. All four cases representing the epithelioidcell type were positive for MMP-2. Half of the casesrepresenting the mixed cell type and a third of those withthe spindle cell type were MMP-2-positive. Similarresults have also been reported with t-PA.29

Expression of MMP-2 correlated with the overall andrelapse-free survival of uveal melanoma patients. Thedifference in the 5-year survival rates for patients withMMP-2-positive and -negative cases seemed remarkable(49 per cent compared with 86 per cent). The differencewas statistically significant, although the size of thepatient group was small. The positivity was associatedwith visceral metastases, suggesting an important rolefor MMP-2 in haematogenous metastasis in uvealmelanoma. MMP-2 expression was also associated withpoor prognosis in patient groups representing uveal

Copyright ? 1999 John Wiley & Sons, Ltd.

melanoma of different cell types. A patient group with adismal prognosis was identified in this study. The 5-yearsurvival was only 38 per cent in the patients representingMMP-2-positive non-spindle cell uveal melanomas. Incontrast, the prognosis was excellent in MMP-2-negativespindle cell tumours, which included a total of one-thirdof the patients in this study.

MMPs and their inhibitors were determined in thevitreous body using an enzyme immunoassay. VitreousMMP-2 activity could also be demonstrated by zymog-raphy. High levels of MMP-2 were observed in thevitreous body after retinal detachment or tear. In uvealmelanoma, however, the levels of specific proteins forMMPs and TIMPs were low. It is possible that MMP-2could be released from normal structures of the eye orcould be related to bleeding during detachment or tearof the retina.

MMPs most likely contribute to normal tissue turn-over. In the human eye, the turnover of the base-ment membranes in Bruch’s membrane of the retina, inBowman’s membrane of the cornea, and in the ciliarybody could involve MMP-2 activity. The retinal pigmentepithelium and corneal epithelium, as well as fibroblastsin the ciliary body and choroid, showed positive stain-ing for immunoreactive MMP-2. Retinal pigmentepithelium has been found to release t-PA in vitro.29

MMP-2 and MT1-MMP have also been identifiedrecently in post-mortem human sclera, cornea, choroid,retinal pigment epithelium, and retina by western blotanalysis;30 MMP-2 and TIMP-1 have been found in ratretinal pigment epithelium by immunohistochemistry.31

Our results are in line with these findings, supporting arole for MMP-2 in the tissue remodelling of severalstructures in the human eye.

This is the first study to demonstrate MMP-2 im-munoreactive protein expression in human uvealmelanoma. Although the number of cases was limited,the presence of MMP-2 protein was correlated withhaematogenous metastasis and poor survival. Patientswith MMP-2-positive non-spindle cell tumours formeda subgroup with an unfavourable prognosis. MMP-2might be of value as a new biochemical marker for uvealmelanoma and might help in estimating the prognosis ofindividual patients. However, since this was in thenature of a pilot study, the predictive value of MMP-2expression needs to be determined in a larger series.

ACKNOWLEDGEMENTS

We wish to express our warmest gratitude to Mrs K.Järvenpää, Mrs R. Karvonen, Mrs T. Lujala, andMs S. Koljonen for their skilful technical assistance andMr R. Bloigu for his valuable help and advice in thestatistical analysis. This study was supported partly by agrant from the Cancer Society of Northern Finland.

REFERENCES

1. Mooy CM, de Jong PTVM. Prognostic parameters in uveal melanoma: areview. Surv Ophthalmol 1996; 41: 215–228.

2. Seregård S, Kock E. Prognostic indicators following enucleation for pos-terior uveal melanoma. Acta Ophthalmol Scand 1995; 73: 340–344.

J. Pathol. 188: 56–62 (1999)

62 A. VÄISÄNEN ET AL.

3. Stetler-Stevenson WG, Liotta LA, Kleiner DE Jr. Extracellular matrix 6:role of matrix metalloproteinases in tumour invasion and metastasis.FASEB J 1993; 7: 1434–1441.

4. Mignatti P, Rifkin DB. Biology and biochemistry of proteinases in tumourinvasion. Physiol Rev 1993; 73: 161–195.

5. Höyhtyä M, Hujanen E, Turpeenniemi-Hujanen T, Thorgeirsson U, LiottaLA, Tryggvason K. Modulation of type IV collagenase activity and invasivebehaviour of metastatic human melanoma (A2058) cells in vivo by mono-clonal antibodies to type IV collagenase. Int J Cancer 1990; 46: 282–286.

6. Tryggvason K, Höyhtyä M, Pyke C. Type IV collagenase in invasivetumours. Breast Cancer Res Treat 1993; 24: 209–218.

7. Vassali J-D, Sappino A-D, Belin D. The plasminogen activator/plasminsystem. J Clin Invest 1991; 88: 1067–1072.

8. Mueller BM. Different roles for plasminogen activators and metallo-proteinases in melanoma metastasis. Curr Top Microbiol Immunol 1996;212: 65–80.

9. Birkedal-Hansen H, Moore WGI, Bodden MK, et al. Matrix metallo-proteinases. A review. Crit Rev Oral Biol Med 1993; 4: 197–250.

10. Sato K, Kida Y, Mai M, et al. A matrix metalloproteinase expressed on thesurface of invasive tumour cells. Nature 1994; 370: 61–65.

11. Cottam DW, Rees RC. Regulation of matrix metalloproteinases, their rolein tumour invasion and metastasis. Int J Oncol 1993; 2: 861.

12. Albini A, Melchioti A, Santi L, Liotta LA, Brown PD, Stetler-StevensonWG. Tumour cell invasion inhibited by TIMP-2. J Natl Cancer Inst 1991;83: 775–779.

13. Visscher DW, Höyhtyä M, Ottosen SK, et al. Enhanced expression of tissueinhibitor of MMP-2 (TIMP-2) in the stroma of breast cancer correlates withtumour recurrence. Int J Cancer 1994; 59: 339–344.

14. Grignon DJ, Sakr W, Tith M, et al. High levels of tissue inhibitor ofmetalloproteinase-2 (TIMP-2) expression are associated with poor outcomein invasive bladder cancer. Cancer Res 1996; 56: 1654–1659.

15. Väisänen A, Tuominen H, Kallioinen M, Turpeeniemi-Hujanen T. Matrixmetalloproteinase-2 (72 kD type IV collagenase) expression occurs in theearly stage of human melanocytic tumour progression and may haveprognostic value. J Pathol 1996; 180: 283–289.

16. de Vries TJ, Quax PHA, Dinjin M, et al. Plasminogen activators, theirinhibitors and urokinase receptor emerge in late stages of melanocytictumour progression. Am J Pathol 1994; 144: 70–81.

17. de Vries TJ, Mooy CM, van Balken MR, et al. Components of theplasminogen activation system in uveal melanoma—a clinico-pathologicalstudy. J Pathol 1995; 175: 59–67.

18. Cottam DW, Rennie IG, Woods K, Parsons A, Bunning RAD, Rees RC.Gelatinolytic metalloproteinase secretion patterns in ocular melanoma.Invest Ophthalmol Vis Sci 1992; 33: 1923–1927.

Copyright ? 1999 John Wiley & Sons, Ltd.

19. Yanoff M, Fine BS. Ocular melanocytic tumours. In: Yanott M, Fine BS,eds. Ocular Pathology. Barcelona: Mosby-Wolfe, 1996; 785–860.

20. Hsu SM, Raine L, Fanger H. The use of avidin–biotin–peroxidase complex(ABC) in immunoperoxidase techniques: a comparison between ABC andunlabeled antibody (PAP) procedure. J Histochem Cytochem 1981; 29:557–580.

21. Väisänen A, Kallioinen M, Taskinen PJ, Höyhtyä M, Turpeenniemi-Hujanen T. Prognostic value of MMP-2 immunoreactive protein (72 kDtype IV collagenase) in primary skin melanoma. J Pathol 1998; 186: 51–58.

22. Talvensaari-Mattila A, Pääkkö P, Höyhtyä M, Blanco-Sequieros G,Turpeenniemi-Hujanen T. MMP-2 immunoreactive protein, a marker ofaggressiveness in breast carcinoma. Cancer 1998; 83: 1153–1162.

23. Höyhtyä M, Turpeenniemi-Hujanen T, Stetler-Stevenson W, Krutzsch H,Tryggvason K, Liotta LA. Monoclonal antibodies to type IV collagenaserecognize a protein with limited sequence homology to interstitialcollagenase and stromelysin. FEBS Lett 1988; 233: 109–113.

24. Heussen C, Dowdle EB. Electrophoretic analysis of plasminogen activatorsin polyacrylamide gels containing sodium dodecylsulfate and copolymerizedsubstrates. Ann Biochem 1980; 102: 192–202.

25. Hujanen ES, Väisänen A, Zheng A, Tryggvason K, Turpeenniemi-HujanenT. Regulation of 72 kD and 92 kD type IV collagenase (gelatinase A and B)gene expression by interferons alpha and gamma in human melanoma. Int JCancer 1994; 58: 582–586.

26. Nikkari ST, Höyhtyä M, Isola J, Nikkari T. Macrophages contain 92 kDagelatinase (MMP-9) at the site of degenerated interval elastic lamina intemporal arteritis. Am J Pathol 1996; 149: 1427–1433.

27. Höyhtyä M, Fridman M, Komarek D, et al. Immunohistochemical locali-zation of matrix metalloproteinase-2 and its specific inhibitor TIMP-2 inneoplastic tissues with monoclonal antibodies. Int J Cancer 1994; 56:500–505.

28. Hujanen ES, Turpeenniemi-Hujanen T. Recombinant interferon alpha andgamma modulate the invasive potential of human melanoma in vitro. Int JCancer 1991; 47: 576–581.

29. Tripathi BJ, Park JK, Tripathi RC. Extracellular release of tissue plasmino-gen activator is increased with the phagocytic activity of the retinal pigmentepithelium. Invest Ophthalmol Vis Sci 1989; 30: 2470–2473.

30. Smine A, Plantner JJ. Membrane type-1 matrix metalloproteinase in humanocular tissues. Curr Eye Res 1997; 16: 925–929.

31. Padgett LC, Lui G-M, Werb Z, LaVail MM. Matrix metalloproteinase-2and tissue inhibitor of metalloproteinase-1 in the retinal pigment epitheliumand interphotoreceptor matrix: vectorial secretion and regulation. Exp EyeRes 1997; 64: 927–938.

J. Pathol. 188: 56–62 (1999)