Prognostic significance of stromelysin 3, gelatinase A, and urokinase expression in breast cancer

Prognostic Significance of Stromelysin 3, Gelatinase A, and Urokinase Expression in Breast Cancer BER.NARD T/=TU, MD, JACQUES BRISSON, MD, PHD, HELENE LAPOINTE, MSc, AND PASCALE BERNARD, MSc This study was aimed at testing the hypothesis that the expression of proteases essentially produced by reactive stromal cells (stromelysin-3 [ST3], gelafinase A [GELA], and urokinase [uPA]) is predictive of prognosis in patients with breast cancer. This was a study of patients with node-positive and node-negative breast cancer diagnosed from 1980 to 1986 and with an average of 10 years follow-up. ST3 (665 cases), GELA, and uPA (575 cases each) expression was obtained by in situ hybridization on formalin-fixed, paraffinembedded material using mRNA antlsense probes. ST3 was expressed by 86.6% of the cases; GELA, 77.7%; and uPA, 64.7%. A significant correlation (P < .05) was found between high (more than 10%) ST3 expression and a younger age, lymph node involvement, poor nuclear grade, ductal histology, aneuploidy, and HSP-27 expression. High GELA expression was significantly associated with c-erbB2, ductal histology, and HSP-27 expression. High uPA expression correlated with poor nuclear grade, ductal histology, lack of estrogen and

progesterone receptors, and p53 protein accumulation. High level of expression of all three proteases correlated significantly with each other and with cathepsin D expression by reactive stromal cells. By univariate analysis, both ST3 and uPA expression significantly predicted a shorter recurrence-free survival (ST3, P = .0199; uPA, P = .0269). By multivariate analyses, the prognostic significance was lost, most particularly at longer term. This study adds support to the concept that protease expression by reactive stromal cells is related to cancer cell characteristics but that their contribution to cancer progression is marginal. HuM PATnOL 29:979-985. Copyright © 1998 by W.B. Saunders Company Key words: breast neoplasms, stromelysin, gelatinase, uroklnase. Abbreviations: ECM, extracellular matrix; ST3, stromelysin-3; uPA, urokinase; GELA, gelatinase A; DMFS, distant metastasis free; OS, overall survival; HR, hazard ratio.

Invasive breast cancer is c o m p o s e d of a proliferation of cancer cells derived f r o m the m a m m a r y ductal or acinar epithelium and of a host stromal cell proliferation with benign histological features. Stromal cells are c o m p o s e d of fibroblasts and fibroblast-like cells such as myofibroblasts 1 but also of m a c r o p h a g e s 2 and b l o o d capillaries. 3 Most research on the biology and t r e a t m e n t of breast cancer focused essentially on cancer cells, and the stromal cell proliferation was often r e g a r d e d as a defense of the host against the cancer cell proliferation a n d metastasis. 4 Recent evidence, however, rather suggests that reactive stroma is essential for t u m o r growth and may actively participate in cancer invasion and metastasis through tissue remodeling. 5 Besides angiogenesis, whose role in t u m o r progression has b e e n thoroughly investigated, ~ there is little clinical evidence linking fibroblast-like cell proliferation with prognosis. Stromal tissue r e m o d e l i n g in neoplasia is t h o u g h t to be m e d i a t e d by several proteases and protease inhibitors acting on extracellular matrix (ECM) and b a s e m e n t m e m b r a n e . 6,7 At least some of t h e m were f o u n d to be expressed by reactive stromal cells in a

variety of malignancies. 79 In breast cancer, although neoplastic cells may express a n u m b e r of proteases, stromelysin-3 (ST3), urokinase (uPA) and 72 kDa type 1V collagenase, also known as gelatinase A (GELA) were m o r e specifically observed in reactive stromal fibroblastlike cells, in the close vicinity of cancer cells. 9ql ST3 has b e e n identified by subtractive hybridization of h u m a n breast cancer.la It has unusual functional properties over other matrix metalloproteinases. Thus, despite the fact that h u m a n ST3 is secreted in an active form, it has little or no proteolytic activity on major ECM components. 12 Recent studies tend to link high ST3 expression to a p o o r prognosis in breast cancer. 13,14 GELA degrades type IV collagen, gelatin, and fibronectin and is also expressed by most breast cancers.a° T h e r e is recent evidence linking a strong expression of GELA to local r e c u r r e n c e or axillary lymph nodes metastases in breast cancer, 15,16 but a relationship with survival is lacking. I6,17 uPA was observed in stromal cells a r o u n d cancer cells, and studies by immunoassays strongly relate a high uPA expression to an increased risk of relapse and death of breast cancer, a8 This study was therefore aimed at testing the hypothesis that reactive stromal cells may actively influence the prognosis of breast cancer through protease expression and that this protease p r o d u c t i o n correlates with cancer cell characteristics.

From the Department of Pathology and Groupe de Recherche en Epidtmiologie, Universit6 Laval, Qutbec, Canada. Accepted for publication February 9, 1998. Supported by the Cancer Research Society Inc. and the FRSQHydro-Qutbec joint programme of the Fonds de la Recherche en Sant~ du Qutbec (FRSQ). B.T. is recipient of a research clinical scholarship of the FRSQ and of an MRC (Canada)-INSERM (France) exchange award (1996-1997). Address correspondence and reprint requests to Bernard TEtu, MD, Centre Hospitalier Universitaire de QuEbec, Pavillon H6tel-Dieu de Qutbec, 11, C6te du Palais, QuEbec, Qutbec, Canada, G1R 2J6. Copyright © 1998 by W.B. Saunders Company 0046-8177/98/2909-001458.00/0

979

MATERIALS AND METHODS

Population The patients included in this study met the following criteria: (1) they had node-positive or node-negative disease proven by histological examination of axillary lymph nodes, but with absence of distant metastasis at diagnosis; (2) they

HUMAN PATHOLOGY Volume29, No, 9 (September 1998) had hormone receptor assays performed at the "Centre de Recherche en Cancfirologie de FUniversit6 Laval" at L'H6telDieu de Qu6bec; (3) the surgery for breast carcinoma was performed between the January 1, 1980 and the December 31, 1986. Clinical information was obtained from the patients' charts by experienced research nurses. The tumor size and n u m b e r of examined and positive lymph nodes were obtained from the pathology reports. The histological types along with the histological and nuclear grades (according to Fisher et a119) were reassessed by one of us (B.T.).

In Situ Hybridization In situ hybridization was performed on a representative section of formalin-fixed paraffin-embedded breast carcinomas. In each case, a 5-pm section stained with hematoxylin and eosin was followed by 10 unstained sections for immunohistochemistry and in situ hybridization and two to four sections of 50 pm for DNA flow cytometry. A final 5-pm fixed section was stained with hematoxylin and eosin to confirm the persistance of tumor despite the numerous sections. The method of Wolf et al1° has been used for all three probes. Five-micron-thick sections were deparaffnized, rehydrated in graded alcohols, and digested with proteinase K (4 pg/mL, 30 minutes, 37°C). After overnight hybridization (0.6 m o l / L NaC1, 50% formamide, 50°C) with 3~S labeled antisense RNA probes for ST3, GELA, and uPA, the sections have been RNAse treated (20 pg/mL, 30°rain., 37°C) and stringently washed (50% formamide, 60°C, 2 hours). Slides have been dipped in Kodak autoradiography emulsion type NBT2, exposed for 15 days, developed, and counterstained with Toluidine blue. Plasmids for ST3, GELA, and uPA were provided by Pr. Paul Basset, IGBMC, Illkirch, France. cDNA inserts were prepared from breast cancer cDNA libraries, subcloned in pBluescript II vectors, and were used as ternplates for in vitro transcription to generate sense and antisense probes. The 1,759 base pairs (bp) ST3 cDNA was subcloned in the EcoRI site and extended from nucleotides (nt) 346 to 2105; the 2124-bp GELA cDNA was subcloned in the EcoRI site and extended from nt 26 to 2150, and the 1000-bp uPA cDNA was subcloned in the PstI site and extended from nt 300 to 1300. Reduction of the probe length was achieved by partial alcaline hydrolysis. Optimal in situ hybridization was obtained with 100-bp probes for ST3 and uPA and 500 bp for GELA. The light microscopic interpretation of in situ hybridiza-

TABLE1.

Association of High Protease Expression With Other Prognostic Factors

Parameter Age (yr) <50 ->50 Tumor size (cm) <3 cm ->3 cm Involved lymph 0 nodes 1-3 >3 Peritumoral lymphovascular invasion Yes No Histologic grade I II III Nuclear grade I II III Histologic type Invasive ductal carcinoma Others S-phasefraction <8.5% ->8.5% Aneuploid Estrogen receptors Negative Positive

ST3 (No, % Pos, P)

GELA uPA (No, % Pos, P) (No, % Pos, P)

195 62.1 470 52.8

.028 171 53.2 .082 169 27.8 .516 404 45.3 403 30.5

385 53.3 246 60.2

.087 318 47.8 .615 315 29.8 .844 222 50.0 222 30.6

291 48.8 202 43.6 201 30.4 180 63.3 , 0 0 6 179 52.5 .215 179 31.3 .410 183 57.9 183 47.0 181 25.4

159 55.4 108 50.0

.390 149 52.4 .117 151 33.1 .097 84 41.7 83 22.9

80 43.8 58 44.8 56 25.0 332 57.2 .087 294 46.9 .840 294 28.2 .331 240 56.3 211 48.8 210 33.3 128 39.1 100 41.0 99 19.2 374 60.2 <.0001 327 49.9 .286 325 28.9 .003 150 56.7 136 46.3 136 39.7 574 58.9 85 30.6

.0001 501 49.5 .011 497 32.0 .003 69 33,3 70 15.7

95 47.4 68 47.1 404 60.2

.020

79 43.0 78 23.1 55 47.3 .566 54 27.8 .167 357 49.6 354 33.3

189 53.4 .678 154 52.0 .162 152 39.5 .002 449 55.2 395 45.3 394 25.6

tion has been done by one of us (B.T.) whithout knowledge of the clinical information. The scoring was assessed separately on cancer and stromal cells using a semiquantitative scale similar to that used for immunohistochemistry. For each cell compartment, the percentage of cancer or stromal cells expressing the marker (0%, <10%, 10-50%, >50%) was evaluated on the whole tumor surface of one representative section. The quality of RNA preservation has been assessed with the use ofantisense RNA probes for [3-actin, and negative controls were obtained with the use of sense probes. Cases that lacked [3-acfin expression were excluded from the analysis. In situ hybridization has been repeated in 75 randomly selected cases to insure that both the reaction and the interpretation were reproducible.

Immunohistochemistry An immunohistochemical study was performed using the avidin-biotin complex method as described by Hsu et al. 23 Primary antibodies to p53 (CM1, Signet Laboratories, Dedham, MA, and ID Labs, London, Ontario dilution of 1/50), c-erbB2 (Triton Biosciences, Alameda, CA; dilution: 1/15), HSP-27 (Hu27, Dr Jacques Landry, H6tel-Dieu de Qu6bec, dilution: 1/200), and cathepsin D (Novocastra, Newcastle, England; dilution: 1/200) were used. Positive immunostaining for p53, c-erbB2, and HSP-27 were defined by the presence of at least one cell expressing the marker. Expression for cathepsin-D

FIGURE 1. Brightfield in situ hybridization for ST3. On this figure, more than 50% of stromal cells express the protease,

980

ST3, GELA, AND uPA IN BREASTCANCER(T~tu et al) TABLE I. Parameter Progesterone receptors Negative Positive c-erbB~Expression Negative Positive p53 Negative Positive HSP-27 Negative Positive Cathepsin D (cancer cells) Negative Positive Cathepsin D (stromal cells) Negative Positive Stromelysin-3 Negative Positive Type IV collagenase Negative Positive Urokinase Negative Positive

(Continued)

ST3 (No, % Pos, P)

GELA (No, % Pos, P)

50%, >50%). We, however, selected 10% as the cutoff point between those expressing low (negative) or high (positive) protease levels by stromal cells because (1) this threshold has been commonly used in studies on other proteases such as cathepsin D2°-22; (2) Kaplan-Meier DMFS and OS curves of protease expression by stromal cells for all three proteases confirmed the reliability of using this threshold, because no difference was found between those cases with 0% and less than 10% of cells expressing the marker. DMFS and OS of cases with less than 10% of cells expressing the marker were further compared with those with 10% to 50% and more than 50% positive cells. The Cox proportional hazard model was used to evaluate the relation of protease expression with recurrence and death, adjusting for other known or suspected prognostic factors. Hazard ratios (HR) for the occurrence of distant metastases or death of patients with or without marker expression were obtained by univariate and multivariate analyses. HR compared the rate of tumor recurrence or death of patients whose tumor expressed a protease with those whose tumor did not express the marker.

uPA (No, % Pos, P)

278 52.9 .434 232 49.1 .451 231 36.8 .001 359 56.0 316 45.9 314 23.9 543 54.1 .139 465 45.4 .025 464 28.2 .112 122 61.5 110 57.3 108 36.1 464 54.5 .392 397 47.9 .984 395 26.1 .001 182 58.2 163 47.2 163 39.9 279 48.4 .001 216 45.4 .023 214 28.5 .711 198 63.1 172 57.0 172 30.2

405 53.3 .286 343 44.6 .084 342 31.0 .456 250 57.6 223 52.0 221 28.1

RESULTS

Population

364 48.6 .0001 320 41.3 .0001 319 20.7 .0001 244 56.2 242 41.7 289 63.3

We s t u d i e d 586 b r e a s t c a n c e r s f o r G E L A a n d uPA, w h e r e a s 90 m o r e cases (676) were a n a l y z e d f o r ST3. E l e v e n cases were e x c l u d e d b e c a u s e t h e y l a c k e d e x p r e s sion o f all m a r k e r s , i n c l u d i n g beta-actin. O f t h e 665 cases, t h e a g e a v e r a g e d 57 years ( r a n g e , 28 to 92). Two h u n d r e d n i n e t y - o n e w e r e n o d e negative, a n d 374 h a d n o d e invasion. A t last follow-up, 339 (50.7%) w e r e alive, a n d t h e follow-up o f t h e surviving p a t i e n t s a v e r a g e d 9.5 years a n d r a n g e d f r o m 5.2 to 14.6. Twenty-six w e r e lost to follow-up.

230 28.7 .0001 231 15.6 .0001 342 60.8 337 39.8 298 45.0 .0001 274 75.9

297 15.8 .0001 271 45.4

398 51.0 .0001 398 37.2 .0001 170 78.8 170 72.4

was defined by the presence of at least 10% cells expressing the protease. 2°

Tumor Characteristics

Flow Cytometry DNA flow cytometric analysis was performed in all cases on formalin-fixed paraffin-embedded tissues using the method described by Dressler and Bartow. 24 Ploidy and S-phase fraction (SPF) were determined on single-parameter histograms with the use of ModFit (Verity Software House inc., Topsham, ME). Debris were excluded with the use of the "Single Cut" algorithm. G0/G1 and G2/M were defined from a Gaussian curve, and SPF was evaluated with the trapezoid model. All cases with a coefficient of variation (CV) exceeding 8% were excluded, and the CV averaged 6.1% (range, 2.5% to 8.0%).

Hormone Receptor Assay Estrogen and progesterone receptors were quantitated using the hydroxylapatite method, 25 and a threshold of 10 f m o l / m g cytosol protein was used to separate positive from negative cases.

Statistical Analysis Correlation between protease expression and other recognized prognostic variables in breast cancer was determined by the chi-square test. Distant-metastasis-free (DMFS) and overall (OS) survival curves were obtained for each protease according to Kaplan and Meier. The difference between the curves was assessed using the log-rank test. DMFS and OS curves were generated using each level of expression (0%, <10%, 10%-

981

O f t h e 665 cases, 578 (86.9%) t u m o r s were invasive d u c t a l c a r c i n o m a s , a n d 0 to 29 l y m p h n o d e s w e r e involved at d i a g n o s i s ( m e a n : 3). T h e first r e c u r r e n c e was local in 12% o f cases, r e g i o n a l in 4.6%, a n d to d i s t a n t o r g a n s in 30.9%, m o s t l y to b o n e s , liver, a n d lungs.

In Situ Hybridization

Stromelysin-3. ST3 m R N A was e x p r e s s e d by 576 (86.6%) o f t h e 665 cases, a n d t h e e x p r e s s i o n was l i m i t e d to reactive s t r o m a l cells in t h e close vicinity o f c a n c e r cells (Fig 1). N o r m a l fibroblasts d i s t a n t to c a n c e r cells d i d n o t e x p r e s s t h e p r o t e a s e . T a b l e 1 shows t h e c o r r e l a t i o n o f h i g h ST3 e x p r e s s i o n with m a j o r p r o g n o s t i c i n d i c a t o r s in b r e a s t cancer. A s i g n i f i c a n t c o r r e l a t i o n ( P < .05) was f o u n d with y o u n g e r age, l y m p h n o d e i n v o l v e m e n t , p o o r n u c l e a r g r a d e , d u c t a l histology, a n e u ploidy, a n d HSP-27 e x p r e s s i o n . H i g h ST3 e x p r e s s i o n also c o r r e l a t e d with h i g h c a t h e p s i n D e x p r e s s i o n by reactive s t r o m a l cells a n d with b o t h h i g h u P A a n d G E L A e x p r e s s i o n . I n u n i v a r i a t e analysis, h i g h ST3 e x p r e s s i o n significantly c o r r e l a t e d with DMFS ( P = .0199; F i g 2) a n d with O S ( P = .0414; d a t a n o t s h o w n ) . N o d i f f e r e n c e was f o u n d b e t w e e n t h o s e with 10% to 50% a n d m o r e t h a n 50% cells e x p r e s s i n g ST3 ( d a t a n o t s h o w n ) . T a b l e 2 shows t h e H a z a r d Ratios (HR) f o r t h e o c c u r r e n c e o f d i s t a n t m e t a s t a s e s o r d e a t h o f p a t i e n t s with o r w i t h o u t

HUMAN PATHOLOGY

-. . . . . . . . . .

Volume 29, No, 9 (September 1998)

Negative ( rr=-296 ) ~o2itive ( r ~ 6 9 )

d

0

2

4

6

8

10

Years

FIGURE 2, DMFSof patients with tumors expressing ST3 in less than 10% (negative) or 10% or more (positive) cells,

marker expression by univariate and multivariate analyses. By univariate analysis, the prognosis was significantly worse for those with ST3 expression. In fact, patients with presence of ST3 had a rate of distant metastases 31% greater than that of patients whose tumor did not express the marker. However, by multivariate analysis, after adjustment for other prognostic factors (age, lymph node involvement, tumor size, nuclear grades, progesterone receptor status, DNAploidy, and c-erbB2, HSP-27 and cathepsin D expression), the HRbecame marginally significant (HR = 1.27, 95% confidence interval [CI]: 0.98 to 1.64). Similar findings were observed on rate of death (Table 2). Furthermore, when uPA was included in the model, ST3 was no longer a significant marker of recurrence or death of breast cancer (Table 2). Dichotomizing those cases with high ST3 expression into categories with 10% to 50% or greater than 50% expressing cells did not improve the prognostic significance of the marker. Gelatinase A. GELA was expressed by 447 (77.7%) cases. As for ST3, it was expressed by reactive stromal cells close to the cancer cell proliferation, but few normal stromal cells distant to the tumor site also expressed the protease. Cancer cells did not express the marker. A significant correlation (P < .05) was found between high GELA expression with c-erbB2, ductal histology, and HSP-27 expression. High GELA level also correlated with high ST3 and uPa and with cathepsin D TABLE 2.

Hazard

Stromelysin-3

Distant metastasis Death Distant metastasis Death Distant metastasis Death

Urokinase

DISCUSSION

In this study, we found a significant association of high ST3, GELA, and uPA expression by stromal cells

Hazard Ratios for the Association of Protease Expression on Occurrence of Metastases and Death

Marker

Gelatinase A

expression by reactive stromal cells. No association was found with DMFS (Fig 3) and OS (data not shown). By univariate analysis, GELA expression was associated with 15% greater risk of developing distant metastases. By multivariate analysis, the risk was not significantly different from that of tumors not expressing the marker (HR: 1.03; CI: 0.79-1.33). Similar findings were observed on OS and, therefore, GELA was not a significant marker of recurrence or death (Table 2). Urokinase. Of the 572 cases for which in situ hybridization has been performed, uPA was expressed by 370 (64.7%) cases. The pattern of expression was comparable to ST3 and was only present within stromal cells around cancer cells, but cancer cells also expressed uPA in five (1%) cases, uPA expression by cancer cells was, however, only focal. High uPA expression correlated significantly with poor nuclear grade, ductal histology, lack of estrogen and progesterone receptors, and p53 protein accumulation. It also correlated with ST3 and GELA expression and with cathepsin D expression by reactive stromal cells. High uPA expression significantly predicted DMFS (P = .0269, Fig 4A) and OS (P = .0152, not shown). By univariate analysis, the HR of developing distant metastases was 33% greater for patients whose tumor expressed uPA. By multivariate analysis (Table 2), after adjustment for other prognostic factors, the HR remained significant (HR for DMFS: 1.33; CI: 1.01-1.76) and lost significance when ST3 was included in the model (HR: 1.28; CI: 0.971.69). Similar findings were observed on OS (Table 2). By dichotomizing the percentage of cells expressing the marker, the survival was significantlyworse by univariate analysis for those patients whose tumor had more than 50% of the cells expressing the marker (P = .0165; Fig 4B). At 5 years, 64% of patients whose tumor had less than 10% of cells expressing uPA were free of disease, whereas 41% of those with more than 50% uPA expressing fibroblasts were metastasis-free. However, the difference between low and high expressors decreased over the subsequent years and, by multivariate analysis, this new subcategory did not add any significant information because of the limited number of patients.

Univariate HR* (95% CI) 1.31 1.29 1.15 1.16 1.33 1.39

(1.04-1.64) (1.01-1,65) (0.91-1,45) (0.91-1,49) (1.03-1.70) (1.06-1.82)

Multivariate * HR (95% CI) 1.27 1.20 1.03 1.07 1.33 1.33

(0.98-1.64) (0.92-1.58) (0.79-1.33) (0.81-1.41) (1.01-1.76) (0.99-1.79)

Multivariatet HR (95% CI) 1.12 (0.87-1.44) 1.05 (0.80-1.38) 1.28 (0.97-1.69) 1.31 (0.97-1.78)

Abbreviation: HR, hazard ratio. *Adjusted for age, lymph node involvement, tumor size, nuclear grade, progesterone receptor status, DNA-ploidy, and c-erbB2, HSP-27 and Cathepsin D expression. j-Adjusted for all factors and for a protease (ST3 for uPA and uPA for ST3).

982

ST3, GELA, AND uPA IN BREASTCANCER (T~tu et al) q

-Negative n=301 ) .......... Positive ( n=274 )

k ~

p = 0.2390

..... ~ N . _ ~ .

"'""\\, ~ ~ ~o "E

g~

......

2

4

6

.........,.......,.~...,_.~

...........................

8

10

Years

FIGURE 3. DMFS of patients with tumors expressing GELA in less than 10% (negative) or 10% or more (positive) cells.

with factors of p o o r prognosis by cancer ceils. We reported similar findings in our study on cathepsin D expression using immunohistochemistry, z° However, as opposed to ST3, GELA, and uPA, which were essentially expressed by reactive stromal cells, cathepsin D was o.¢ ~

~~

-Negative n~-402 ) .......... Positive ( n = 1 7 0 )

'\ ~ ~'~'--,-\ ~

p = 0.0269

~"\\~k.\\,, .

2

,.

4

. .......... ,.......,

6

8

16

Years

-< 10% (n-~102) .......... 10%-60% (n=141)

"~,...

L_~ , ~ Ii

.... •50% (n-~} ...\.

p=o.oi~

m L.................................

2

4

6

1

8

10

Years

FIGURE 4. DMFSof patients with tumors expressing uPA in (A) less than 10% (negative) or 10% or more (positive) cells and (B) less than 10%, 10% to 50% and more than 50% of cells.

983

expressed by both cancer cells and reactive stromal cells a r o u n d cancer cells. In our study on cathepsin D, protease expression by stromal cells correlated with p o o r nuclear grade, aneuploidy, absence of estrogen and progesterone receptors, and H E R - 2 / n e u oncoprotein expression, and cathepsin D expression by cancer cells did not correlate with any of these factors. 2° All of those findings strongly suggest that protease expression by reactive stromal cells responds to a stimulation from cancer cells. In vitro studies confirmed that protease production by stromal cells is at least partly u n d e r a paracrine influence of cancer cells. For example, GELA activity by fibroblasts was e n h a n c e d in cocultures with MDA-MB 231 or SKBR3 cells, two cell lines recognized for their strong invasive potential, and MCF-7, a less invasive cell line, had only a weak influence. 26 T h e exact mechanism by which cancer cells induce protease production by stromal cells is not clearly elucidated, but some information is emerging from recent literature. For instance, a 58-kDa plasma membrane glycoprotein p r o d u c e d by cancer cells has been f o u n d to induce collagenase I and GELA expression. 27 ST3 expression was f o u n d to be stimulated by various growth factors, the most potent being the basic fibroblastic growth factor, la a factor p r o d u c e d in part by cancer cells. This supports the concept that certain cancer cells may synthesize a n u m b e r of diffusable factors with a stimulatory influence on protease production by stromal cells. However, the strong correlation between each protease and cathepsin D p r o d u c e d by stromal cells suggests that these different factors are secreted simultaneously u n d e r similar conditions. All three proteases are more strongly expressed by ductal carcinomas than by any other histological subtype. This more likely reflects the more active fibroblastic activity observed in ductal carcinomas. However, the possible secretion of unknown paracrine factors specific for ductal carcinomas cannot be ruled out. In our prior study, we also showed that cathepsin D expression by stromal cells, and not cancer cells, predicted the outcome of patients with breast cancer, 2° and these data were later confirmed by three recent studies. 22,2s,29 However, in the current study, although ST3 and uPA significantly predicted a p o o r outcome by univariate analysis, this prognostic significance was lost by multivariate analyses. In fact, uPA expression only significantly predicted patients' survival in a multivariate analysis in which model ST3 was excluded. The prognostic advantage of low expressors was, however, relatively weak. When tumors with low and high expression were compared, the difference in DMFS at 5 years was only 9%. When high expressors were divided into those with 10% to 50% and more than 50% expression, the difference reached 23% between low and higher (>50%) expressors, but the long-term advantage of low expression was not maintained over the years. After inclusion of ST3 expression in the model, uPA was no more a significant prognostic marker. The proteolytic activity of uPA has been well d o c u m e n t e d in an in vitro model using the Matrigel invasion assay, modulating its

HUMAN PATHOLOGY

Volume 29, No. 9 (September 1998)

activity with antibodies to uPA and uPAR. ~° It is, however, likely that the weak prognostic influence of uPA is due to other factors such as protease inhibitors, which were f o u n d to modulate uPA activityP1 High ST3 expression predicted p o o r patients outcome in an univariate analysis, but this prognostic influence was lost in a multivariate analysis. The exact role of ST3 on breast cancer biology is unclear. ST3 is secreted in its active form, 7 and therefore ST3 mRNA levels as detected by in situ hybridization should reflect the level of active protease. It has b e e n recently observed, using transfected cells injected into nude mice, that ST3 expression does not act on tumor invasion and proliferation but rather favors cancer cell survival in a tissue environment hostile to t u m o r growthP 2 Our data and those on animal models suggest that ST3 more likely acts as a p r o m o t e r of t u m o r growth by maintaining a favorable environment at the t u m o r site or in distant organs rather than as a genuine prognostic factor. Although GELA degrades type IV collagen, gelatin, and fibronectin, our study and others 16,17failed to show any correlation of its level of expression with survival. However, GELA is secreted in an inactive pro-enzymatic form, and its mRNA or protein levels might not reflect its proteolytic activity. GELA can be activated in the extracellular milieu by cleavage of its p r o f o r m by a membrane-type MMP (MT-MMP) or can be inhibited by a tissue inhibitor of metalloproteinases. 3~ It is tempting to speculate that survival might be more d e p e n d e n t on the level of the active form of the enzyme, as obtained by zymography, or on the ratio of GELA, TIMP-2, and MT-MMP. In fact, recently, breast cancers were f o u n d to have higher levels o f activated GELA than benign lesions ~4 and breast cancers with higher GELA/ TIMP-2 ratio more likely developed lymph n o d e metastases. 35 We conclude that high ST3, uPA, and GELA expression by reactive stromal cells is strongly related to cancer cell characteristics but that, in our hands, their prognostic significance is marginal or not significant. Further studies are n e e d e d to better define their clinical usefulness.

Acknowledgment. T h e authors thank Professor Paul Basset and Professor Jean-Pierre Bellocq, IGBMC, I|lkirch, and H6pital de Hautepierre, Strasbourg, France, who provided us with plasmids to ST3, MMP2, and uPA and for their usefull comments. They also thank M i n e Blais and M i n e Pelletier, research nurses, for the data collection and Danielle M u r r y and Genevieve Beaudry for their technical assistance. T h e authors also want to acknowledge the contributions o f the following pathologists who allowed us to review their histologic material: Dr Danielle Proulx, H6tel-Dieu de L4vis, L4vis; Dr Marc Duplessis, H6pital de Chicoutirni, Chicoutimi; Dr Michel Marois, H6pital St-Frangois d'Assise, Qu6bec; Dr R6jean Canfin, H6pital de l'Enfant-J6sus, Qu6bec; Dr R o b e r t Arcand, Centre Hospitalier Sainte-Marie, Trois-Rivi6res; Dr R o b e r t P e r r o n , C e n t r e H o s p i t a l i e r Saint-Joseph, TroisRivi&es; Dr Yves Gagnon, Centre Hospitalier de Jonqui~re, Jonqui~re, Dr Frangois Milette, Dr Guy D u m o n t e t , Centre Hospitalier R4gional de Rimouski, Rimouski, Dr Marcien 984

Fournier, H6pital Laval, Qu6bec, Dr Gladys Parisien, Centre Hospitalier R6gional de la Mauricie, Shawinigan.

REFERENCES 1. Gr6goire M, Lieubeau B: The role of fibroblasts in tumor behavior. Cancer Metastasis Rev 14:339-350, 1995 2. Roger P, Montcourrier P, Maudelonde T, et al: Cathepsin D immunostaining in paraffin-embedded breast cancer cells and macrophages: Correlation with cytosolic assay. HUM PATHOL 25:863-871, 1994 3. Gasparini G, Harris L: Clinical importance of the determination of tumor angiogenesis in breast carcinoma: Much more than a new prognostic tool.J Clin Oncol 13:765-782, 1995 4. Seemayer TA, Schflrch W, Lagac~ R, et al: Myofibroblasts in the stroma of invasive and metastatic carcinoma: A host response to neoplasia. AmJ Surg Pathol 3:525-533, 1979 5. Zipori D: Stromal cells in tumor growth and regression. CancerJ 3:164-169, 1990 6. McDougall JR, Matrisian LM: Contributions of tumor and stromal matrix metalloproteinases to tumor progression, invasion and metastasis. Cancer Metastasis Rev 14:351-362, 1995 7. Basset P, Okada A, Chenard MP, et al: Matrix metalloproteinases as stromal effectors of human carcinoma progression: Therapeutic implications. Matrix Bio115:535-541, 1997 8. Poulsom R, Pignatelli M, Stetler-Stevenson WG, et al: Stromal expression of 72 kda type 1V collagenase (MMP-2) and TIMP-2 mRNAs in colorectal neoplasia. AmJ Patho1141:389-396, 1992 9. Rouyer N, Wolf C, Cbenard MP, et al: Stromelysin-3 gene expression in human cancer: An overview. Invasion Metastasis 14:269275, 1994-95 10. Wolf C, Rouyer N, Lutz Y, et al: Stromelysin 3 belongs to a subgroup of proteinases expressed in breast carcinoma fibroblastic cells and possibly implicated in tumor progression. Proc Natl Acad Sci U S A 90:1843-1847, 1993 11. Basset P, BellocqJP, Wolf C, et al: A novel metalloproteinase gene specifically expressed in stromal ceils of breast carcinomas. Nature 348:699-704, 1990 12. Pei D, Majmudar G, Weiss SJ: Hydrolytic inactivation of a breast carcinoma cell-derived serpin by human stromelysin-3. J Biol Chem 269:25849-25855, 1994 13. Engel G, Heselmeyer K, Auer G, et al: Correlation between stromelysin-3 mRNA level and outcome of human breast cancer. IntJ Cancer 58:830-835, 1994 14. Chenard MP, O'Siorain L, Shering S, et al: High levels of stromelysin-3 correlate with poor prognosis in patients with breast carcinoma. IntJ Cancer 69:448-451, 1996 15. Leone A, Correale M, Dragone CD, et al: Gelatinase A/MMMP-2 serum levels and neoplastic progression in breast cancer patients. IntJ Oncol 4:551-554, 1994 16. Daidone MG, Silvestrini R, D'Errico A, et al: LaIninin receptors, collagenase IV and prognosis in node-negative breast cancers. IntJ Cancer 48:529-532,1991 17. Visscher DW, Hoyhtya M, Ottosen SK, et al: Enhanced expression of tissue inhibitor of metalloproteinase-2 (TIMP-2) in the stroma of breast carcinomas correlates with tumor recurrence. Int J Cancer 59:339-344, 1994 18. Duffy MJ, Reilly D, McDermott E, et al: Urokinase plasminogen activator as a prognostic marker in different subgroups of patients with breast cancer. Cancer 74:2276-2280, 1994 19. Fisher ER, Sass R, Fisher B, collaborating NSABP investigators: Pathologic findings from the national surgical adjuvant project for breast cancers (Protocol No. 4). Cancer 53:712-723, 1984 20. T&u B, Brisson J, C6t4 C, et al: Prognostic signifcance of cathepsin-D expression in node-positive breast carcinoma: An immunohistochemical study. IntJ Cancer 55:429-435, 1993 21. Isola J, Weitz S, Visakorpi T, et al: Cathepsin-D expression detected by immunohistochemistry has independent prognostic value in axillary-node-negative breast cancer.J Clin Oncol 11:36-43, 1993 22. Nadji M, Fresno M, Nassiri M, et al: Cathepsin D in host stromal cells, but not in tumor cells, is associated with aggressive behavior in node-negative breast cancer. HUM PATHOL 27:890-895, 1996

ST3, GELA, AND uPA IN BREASTCANCER (T~tu et al) 23. Hsu S-M, Raine L, Fanger H: Use of avidin-biotin-peroxidase complex (ABC) in immunoperoxidase techniques: A comparison between ABC and unlabeled antibody (PAP) procedures.J Histocbem Cytochem 29:577-580, 1981 24. Dressler LG, Bartow SA: DNA flow cytometry in solid tumors: Practical aspects and clinical applications. Semin Diagn Pathol 6:5582, 1989 25. Erdos T, Best-Belpomme M, Bessada R: A rapid assay for binding estradiol to uterine receptor(s). Anal Biochem 37:244-252, 1970 26. No61AC, Polette M, LewalleJ-M, et al: Coordinate enhancement of gelatinase A mRNA and activity levels in human fibroblasts in response to breast-adenocarcinoma cells. Int J Cancer 56:331-336, 1994 27. Biswas C, Zhang Y, DeCastro R, et al: The human tumor cell-derived collagenase stimulatory factor (renamed EMMPRIN) is a member of the immunoglobulin superfamily. Cancer Res 55:434-439, 1995 28. Joensuu H, Toikkanen S, Isola J: Stromal cell cathepsin D expression and long-term survival in breast cancer. Br J Cancer 71:155-159, 1995 29. O'Donoghue AEMA, Poller DN, BellJA, et al: Cathepsin D in

primary breast carcinoma: Adverse prognosis is associated with expression of cathepsin D in stromal cells. Breast Cancer Res Treat 33:13%145, 1995 30. Hosthansen C, Johannessen B, Hoyerhansen G, et al: Urokinase-type plasminogen activation in three human breast cancer cell lines correlates with their in vitro invasiveness. Clin Exp Metastasis 14:297-307, 1996 31. Andreasen PA, Georg B, Lund LR, et ah Plasminogen activator inhibitors: Hormonally regulated serpins. Mol Cell Endocrino168:1-19, 1990 32. No61 A, Lefebvre O, Maquoi E, et al: Stromelysin-3 expression promotes tumor take in nude mice. J Clin Invest 97:1924-1930, 1996 33. Nagase H: Activation mechanisms of matrix metalloproteinases. Biol Chem 378:151-160, 1997 34. Lee KS, Rha SY, Kim SJ, et al: Sequential activation and production of matrix metalloproteinase-2 during breast cancer progression. Clin Exp Metastasis 14:512-519, 1996 35. Onisto M, Riccio MP, Scannapieco P, et al: Gelatinase A/TIMP-2 imbalance in lymph-node-positive breast carcinomas, as measured by RT-PCR. IntJ Cancer 63:621-626, 1995

985