Breast Cancer Research and Treamwnt 24: 195-208, 1993. © 1993 Kluwer Academic Publishers. Printed in the Netherlands.

Urokinase (uPA) and its inhibitor PAI-1 are strong and independent prognostic factors in node-negative breast cancer

Fritz J~inicke, MD 1, Manfred Schmitt, PhD 1 , Lothar Pache, MD 1 , Kurt Utm, PhD 2, Nadia Harbeck, MD 1, Heinz H6fler, MD 3, and Henner Graeff, MD I 1Frauenklinik and Potiklinik, 2Institut fiir Medizinische Statistik und Epidemiologie, and 31nstitut fiir Allgemeine Pathologie und Pathotogische Anatomie der Technischen Universitiit Miinchen, FRG

Key words: protease, inhibitor, urokinase, uPA, PAI-1, cathepsin D, prognosis, breast cancer, axillary node-negative patients

Summary

Evidence has accumulated that invasion and metastasis in solid tumors require the action of tumor- associated proteases, which promote the dissolution of the surrounding tumor matrix and the basement membranes. The serine protease urokinase-type plasminogen activator (uPA), which is elevated in solid tumors, appears to play a key role in these processes.

We used enzyme-linked immunoassays (ELISA) to test for uPA antigen and its inhibitor PAI-1 in tumor tissue extracts of 247 breast cancer patients who were enrolled in a prospective study. The relation of these data to known prognostic factors and to other variables such as DNA analysis and cathepsin D was studied. Disease-free and overall survival were analyzed according to Cox's proportional hazard model.

The major new finding is that breast cancer patients with either high uPA (>2.97 rig/rag protein) or high content of the uPA inhibitor PAP 1 (>2.18 ng/mg protein) in their primary tumors have an increased risk of relapse and death. Multivariate analyses revealed uPA to be an independent and strong prognostic factor. The impact of uPA is as high as that of the lymph node status. In node-negative patients the impact of uPA is closely followed by that of PAI-1. Since uPA and PAI-1 are independent prognostic factors, the node-negative patients could be subdivided further by combining these two variables. In this refined analysis, patients whose primary tumors have lower levels of both antigens evidently have a very low risk of relapse (93% disease-free survival at three years) in contrast to patients with high uPA and high PAI-1 (55% disease-free survival at three years).

The combination of uPA and PAI-1 in our group of patients with axillary node-negative breast cancer allows us to identify the 45 percent of patients having an increased risk of relapse. Consequently, more than half of the patients had less than a 10% probability of relapse and thus would possibly be candidates for being spared the necessity of adjuvant therapy.

Address for offprints: Prof. Dr. med. Fritz J~inicke, Frauenklinik der Technischen Universit~it MUnchen, Ismaningerstr. 22, D-W-8000 Manchen 80, FRG. Fax: ++49-89-41-805-146. Supported by the Deutsche Forschungsgemeinschafi (Sonderforschungsbereich 207, Projects F9 and GI0) and the Wilhelm Sander-Stiftung.

196 F Jiinicke et al

Introduction

Mortality in breast cancer is related to invasion and early hematogenic metastasis. Evidence has accumulated that invasion and metastasis in solid tumors require the action of tumor-associated proteases, which promote the dissolution of the surrounding tumor matrix and the basement mem- branes. Receptor-bound urokinase-type plasmin- ogen activator (uPA) appears to play a key role in these processes [1-5]. Tumor cells synthesize and secrete uPA as an inactive proenzyme (pro-uPA), which binds to specific receptors on the tumor cell surface [5-7]. After binding, pro-uPA is activated by plasmin or cathepsin B [8,9]. Re- ceptor-bound, active uPA converts plasminogen to plasmin. Subsequently plasmin is also bound to a different receptor on the tumor cell surface [10,11]. Plasmin then degrades components of the tumor stroma (e.g. fibrin, fibronectin, proteo- glycans, laminin) and may activate procollagenase type IV, which then degrades collagen type IV, a major part of the basement membrane [12-15]. Hence, uPA will promote the dissolution of the tumor matrix and the basement membrane, which is a prerequisite for invasion and metastasis. However, tumors also exhibit the plasminogen activator inhibitor type 1 (PAI-1), which blocks the enzymatic activity of uPA. Thus the presence of both uPA and PAI-1 will modulate the invasive and metastatic phenotype of cancer cells.

The uPA antigen content in breast cancer tissue extracts correlates with relapse-free and overall survival in the disease [16-20]. Earlier observations by Duffy et al. [21] applying activity' measurements already indicated a prognostic rel- evance of uPA. However, recent independent findings by Jfinicke et al. and Duffy et al. demonstrated that antigen measurements in breast cancer tissue extracts are far superior and deter- mined uPA antigen to be an independent prognos- tic factor [16,19]. Thus, on the basis of uPA antigen determination, high-risk patients can be

identified even within the classical risk groups defined by lymph node or hormone receptor status.

It was shown recently in an extensive meta- analysis that adjuvant therapy improves the course of the disease in breast cancer, evidently even after prolonged time of observation [22]. Never- theless, there remains considerable controversy regarding adjuvant chemotherapy in women with node-negative disease. Recommendations for treatment vary considerably. In 1990, the Nation- al Cancer Institute suggested in a clinical alert that all node-negative breast cancer patients should be considered for treatment. Alternatively, since approximately 70% of axillary node-neg- ative patients have long-term survival following primary therapy, high-risk patients should be selected for adjuvant chemotherapy by a combina- tion of different prognostic factors [23]. A majority of women with node-negative disease would thereby be spared unnecessary and poten- tially toxic adjuvant therapy. We report here on a prospective study involving 247 breast cancer patients. This study correlates different prog- nostic variables with uPA and its inhibitor PAL1, and we demonstrate that uPA and PAI-1 may serve as independent and strong prognostic factors in node-negative breast cancer patients.

Methods

Patients

In a prospective study conducted between January 1987 and July 1990, 247 patients with breast cancer and 60 patients undergoing surgery for benign breast disease were enrolled. Treatment was by modified radical mastectomy or by breast- conserving surgery including axillary lymph node dissection. The median number of lymph nodes removed was 19, and the median tumor diameter

was 2.6 cm (0.7-15) (additional clinical data are included in Table 2). Premenopausal patients with lymph node involvement received adjuvant chemotherapy, while postmenopausal patients re- ceived adjuvant hormone therapy with tamoxifen. No adjuvant therapy was given to node-negative patients. Seventeen of 247 patients (7%) had metastatic disease upon diagnosis and these patients were excluded from the prognostic evaluation. One patient was not available for follow-up. The remaining 229 patients were followed by clinical visits every 3 months for 12-54 months (median 30 months). Within this time of observation, 48 patients relapsed and 29 patients died.

Tissue extraction

Breast cancer tissue specimens were obtained at surgery, selected by the pathologist, and stored in liquid nitrogen until extraction. Specimens from patients with benign breast disease served as controls. Deep-frozen specimens of 200-500 mg wet weight were pulverized by the "Mikro-Dis- membrator" (Braun-Melsungen, Melsungen, FRG) set to 30 sec at maximal power. The resulting powder was suspended in 2 ml TBS (0.02 M Tris-HC1, 0.125 M NaC1, pH 8.5) containing 1% of the nonionic detergent Triton X-100 (Sigma, Mfinchen, FRG) under gentle shaking for 12 h at 4°C. The suspension was subjected to ultracentri- fugation (100,000 x g, 45 min, 4°C) in order to separate cell debris, nuclei, and cell membranes. The supernatants were divided into aliquots of 50 t.tl each and stored in liquid nitrogen until use. It is worth mentioning that extraction of breast tissue by a buffer containing Triton X-100 results in a higher yield of uPA antigen [19,20]. No increase of PAI-1 per mg protein was observed in the presence of the detergent, uPA and PAI-1 were determined in the Triton X-100 extract and calculated per mg of tissue protein.

uPA and PAI-1 as prognostic factors 197

Laboratory assays

The uPA-ELISA was performed as described else- where [19]. Briefly, microtiter plates (96 wells, Immulon, Dynatech, Denkendorf, FRG) were coated with monoclonal antibody to human uPA (#394; American Diagnostica Inc., Greenwich, CT) for 12 h at 4°C. After blocking with 2% BSA-TBS (30 min, 23°C) and washing with TBS- 0.05 % Tween 20, 100 gI of 1:50 diluted tissue extracts were added in 2% BSA-TBS (14 h, 4°C). The uPA was detected by biotinylated monoclonal antibody to human urokinase (#377; American Diagnostica) followed by the addition of per- oxidase-conjugated avidin (Dianova, Hamburg, FRG) and 3,3',5,5'-tetramethytbenzidine (Serva, Heidelberg, FRG) as substrate. Absorbance was measured at 450 nm by an automated microtiter plate reader (Titertek Multiskan, Flow Labora- tories, Meckenheim, FRG). Recombinant pro- uPA served as the standard in the uPA-ELISA (Grfinenthal GmbH, Stolberg, FRG), whose lower limit of detection is 10 pg uPA per ml. Different forms of uPA such as pro-uPA, HMW-uPA, and LMW-uPA were recognized. In addition, uPA complexes with plasminogen-activator inhibitor type 1 or 2 (PAl-l, PAI-2) or complexes with the uPA receptor were detected. As described else- where [24,25], exposure of uPA to various pro- teases such as plasmin, thrombin, elastase, and trypsin did not affect the determination. The interassay coefficient of variation for the uPA- ELISA is 7.5%, and intraassay variation is <2% (n=25). PAI-t was determined using a commer- cially available ELISA kit (American Diagnostica, type Imubind #822/1). All absorption values were subjected to computer-assisted analysis (EIA program, Flow Laboratories, Meckenheim, FRG). Cathepsin D was determined by applying the commercially available IRMA kit (ELSA-cath-D, CIS Bioindustries, Gif-sur-Yvette, France) to cytosols and not to Triton X-100 treated tissues, since this detergent interferes with the assay. Hormone receptor determinations were performed

198 F Jiinicke et al

using the dextran-coated charcoal technique [26]. Specimens were considered estrogen or progester- one receptor positive if they contained more than 20 fmol per mg protein. Protein content was determined by the BCA Protein Assay Reagent Kit manufactured by Pierce (Rockford, IL). The detergent Triton X-100 present in the tissue extracts does not interfere with the protein determination assay.

Cell cycle analysis

Formalin-fixed, paraffin-embedded breast cancer tissue sections (50 gm) were processed for DNA- and S-phase analysis by a modification of a method reported earlier by Hedley et al. [27]. Briefly, the sections were dewaxed in rohistol (a xylene substitute), rehydrated in ethanol at room temperature, and subsequently incubated in 2 ml 0.5% pepsin (pH 1.5, 37°C, 2 h). The resulting nuclei were washed with PBS (4°C), filtered through a 75 gm nylon mesh, and then incubated in PBS containing 5 mM EDTA and 100 U/ml RNAse (R-5000, Sigma, Munich, FRG; 15 min, room temperature). Pure, unaggregated nuclei in this preparation were verified by means of a confocal laser scanning microscope. 500 ul aliquots of suspended nuclei were stained with 20 btg propidium iodide/ml for flow cytometric analysis prior to the assay (FACScan, Becton- Dickinson, Heidelberg, FRG). Chicken erythro-

Table I. Antigen content of uPA, PAI-I, and cathepsin D

cytes and peripheral blood lymphocytes served as internal euploid standards. The "sum of broad- ened rectangles" program supplied by Becton- Dickinson was applied for DNA analysis. Only tumors with a DNA index of 1.0 + 0.1 were defined to be diploid. In this study, S-phase fractions were only calculated in those tumors in which homogeneous G0/G1 and G2/M popula- tions of nuclei could be identified by light scatter analysis.

Statistical analysis

To determine the relative prognostic impact of uPA, PAI-1, and cathepsin D in relation to the effect of known prognostic factors in a prospec- tive fashion, disease-free and overall survival were analyzed according to Cox's proportional hazard model [28] using the BMDP software package (BMDP Statistical Software, Los Angeles) and by the CART (Classification and Regression Trees) technique [30,31]. Statistical analysis included continuous as well as discrete covariates, all of which were considered as fixed (not time-dependent). Determination of the optimal cut-off for each antigen to discriminate low and high levels was performed using CART. The value with maximal log-rank test [34] was taken for the discrimination of high and low levels. The 95% confidence interval (95% CI) for this cut-off was calculated by a test-based method

in benign and malignant breast tissue extracts

Breast cancer Benign breast tissues Mean ratio Statistical Antigen n Median (range) Mean (SD) n Median (range) Mean (SD) cancer/benign significance a

uPA b 247 2.32 (0.13-15.17) 3.06 (2.52) 60 0.23 (0.02-1.22) 0.32 (0.29) 9.6 p < 0.001 (ng/mg protein)

PAI-I b 247 1.02 (0.00-27.07) 1.75 (2.75) 59 0.00 (0.00-1.20) 0.12 (0.24) 14.6 p < 0.001 (ng/mg protein)

Cathepsin D 240 41.9 (5.2-271.9) 52.9 (38.3) 26 5.0 (0.7-10.0) 5.1 (2.6) 10.4 p < 0.001 (pmol/mg protein)

a U-test Mann-Whitney b Extraction performed in the presence of the nonionic detergent 1% Triton X-100. For details see the Methods section.

uPA and PAl-1 as prognostic factors 199

[29] and boots t rap [32] techniques. Group-

oriented curves for disease-free and overall

survival were calculated according to the Kaplan-

Meier me thod [33]. The relative risks associated

with the various prognos t ic variables after dis-

cr iminat ion into high and low risk groups were

est imated by the Cox model , Correlat ions be-

tween the new factors uPA, PAI-1 , and cathepsin

D were calculated by Spea rman ' s method. The

relation o f these new factors to other prognost ic

variables, such as lymph node status, ho rmone

receptors, menopausa l status, tumor size, D N A

analysis, or vascular invasion was analyzed by the

U-test o f Mann-Whi tney or the Kruskal-Wall is

test. All tests were per formed at a s ignif icance

level o f alpha=0.05.

Table 2. Prognostic factors in 247 breast cancer patients and correlation to uPA, PAI-1 and cathepsin D

Patient subgroup n (%) uPA (pa) PAI-1 (p) Cathepsin D (p) median median median (ng/mg protein) (ng/mg protein) (pmol/mg protein)

Lymph node status negative 104 (44) 2,36 n.s, 0,93 n,s, 37,3 n.s. positive 135 (56) 2.45 1.13 45.4

Lymph nodes involved 0 104 (44) 2.36 n,s, 0.93 n.s, 37.3 n.s. 1-3 60 (25) 2.60 1.09 41,4 >4 75 (31) 2.38 1.15 49.4

Hormone receptors positive 194 (79) 2.15 n.s. 0.92 p<0.001 41.9 n.s, negative 52 (21) 2.96 1.52 41.1

ER (fmol/mg protein) _>20 171 (69) 2.13 n.s. 0.95 p<0.01 41.9 n,s. <20 76 (3I) 2.60 1,26 43,3

PR (fmol/mg protein) >-20 141 (57) 2,32 n.s, 0.93 p<0,01 41,6 n,s, <20 106 (43) 2.34 1.21 44.3

Tumor size <_2 cm 77 (31) 2,17 n.s. 0.86 n.s. 37.9 n.s. 2-5 cm 134 (54) 2.42 1.12 44,1 >5 cm 36 (14) 2.12 1.13 47.7

Menopausal status pre/peri- 88 (36) 2.03 n.s, 0.95 n.s. 39.4 n,s. post- 159 (64) 2.54 1.03 44.7

Vascular invasion absent 195 (79) 2.27 n.s. 0.97 n.s, 40.8 n,s. present 52 (21) 2.50 1,26 46.1

Ploidy b diploid 39 (25) 2,03 n.s, 1,23 n.s, 50,5 n.s. aneuploid 114 (75) 2.41 1.11 47.5

S-phase b low (_<7%) 24 (69) 1.76 n.s. 1.13 p<0.05 35.4 n.s. high (>7%) 11 (31) 2.62 1.83 65,3

n.s. = p>0.05 a p-values calculated with the U-test of Mann-Whitney or the Kruskal-Wallis test b Subgroup of 153 patients with flow cytometry data available

200 F Jiinicke et al

Results

Relation of uPA, PAl-l, and cathepsin D in primary breast cancer tissue extracts to estab- lished prognostic factors

In comparison to benign breast tissues (n = 60), extracts of primary breast cancer tissues (n = 247) were found to contain high levels of uPA, PAI-1, and cathepsin D antigen. For all three antigens, the difference between cancerous and benign

All

tissue was statistically significant at a high confidence level, although a considerable varia- tion of uPA, PAI-1, and cathepsin D content was noted (Table 1). Neither uPA nor cathepsin D content was found to be correlated with estab- lished prognostic factors such as tumor size, lymph node involvement, hormone receptors, or menopausal status (Table 2). Evidently, tumors with negative hormone receptor status, aneu- ploidy, high S-phase, or vascular invasion exhibit increased uPA content; the differences, however,

Patients

Disease-free Survival Overall Survival

uPA

1"0~ ~ l h ~ . . . . ~'~-" 1

~o ' - - t p = o.oooi

0.6

0 , 4 -

ct. 0 .2

0.0

. . . . u P A <= 2.97 ng /mg

- - u P A > 2.97 ng /mg

" 1'0 ' 2'0 " 3'0 " 4'0 ' 5'0 Months

. . . . 140 Pat ients t 7 Events - - 89 Pat ients 31 Events

PAl-1

0.8 J ~ - . . . . . ~ . . . . . . ~__

U-O 0.6- ~ p = 0.0001. >. ~ 0.4

0.2 . . . . PAl-1 <= 2.18 n g l m g

- - PAl-1 > 2.18 ng/rng

0 1 0 2 0 3 0 4 0 5 0 Months

. . . . 189 Pat ients 30 Events

1.0

0.8

O3 0 0 .6 -

0 .4

11. 0 . 2

uPA

. . . . u P A <= 2.97 ng /mg

- - u P A > 2.97 ng /mg

6'0 0.0 . , . , • , . , . , . , 0 10 2 0 30 4 0 5 0 6 0

Months

140 Patients I 0 Events - - 89 Pat ients 19 Events

PAl-1

1.0

0 0

O 0.6

0.4 ~o

0 . 2 . . . . PAl-1 < : 2.18 n g / m g

- - PAI - I > 2.18 ng /m9

6 0 0 10 2 0 3 0 4 0 5 0 6 0

Months

. . . . 189 Pat ients 19 Events - - 40 Pat ients 18 Events - - 40 Patients 10 Events

Figure 1. Disease-free and overall survival as a function of uPA or PAI-I in breast cancer patients (n = 229). Patients with high uPA or PAI- 1 content had a significantly lower rate of disease-free survival and overall survival than patients with low content of uPA or PAI-I.

uPA and PAI-1 as prognostic factors 201

1 . 0 "

0.8-

o 0 , 6 "

0 .4 -

o if- 0.2-

0.13

Cathepsin D P r e m e n o p a u s a l P a t i e n t s

" L - L I

Cathepsin D P o s t m e n o p a u s a t P a t i e n t s

0.8- ~ .................. . . . . . . . . .

p = 0,3808 ~ 1 0,6"

J 0.4"

. . . . Cathepsin D <= 41 pmol/mg 0.2" - - - Cathepsin D <= 41 prnoltmg

- - Cathepsin n > 41 pmol/mg - - Cathepsin D > 41 pmol/mg

, , • i - , , , , r , , 0 . 0 • i • , , . , , . ,

10 20 30 40 50 60 0 10 20 30 40 50 Months Months

. . . . 43 Patients 10 Events . . . . 67 Patients 7 Events - - 39 Patients 13 Events - - 72 Patients 16 Events

p = 0.0462

6 0

Figure 2. Disease-free survival as a function of cathepsin D in pre- and post-menopausal breast cancer patients (n = 221). Postmenopausal patients (n = 139) with high cathepsin D content (>41 pmol/mg protein) display a significantly higher risk of relapse than those with low cathepsin D. In premenopausal patients (n = 82), a cut-off for cathepsin D could not be calculated by CART analysis,

are o f no s ta t i s t ica l s ign i f icance . The on ly

e x c e p t i o n is PAI -1 ant igen, wh ich is i nve r se ly

re la ted to h o r m o n e r ecep to r status (p<0.001) and

d i rec t ly co r re l a t ed with h igh S -phase va lues

(p<0.05) (Table 2). W e also no ted a w e a k

cor re la t ion be tw e e n u P A and PAI-1 (r = 0.48;

p<0.001) , whereas a re la t ion be tw e e n u P A and

ca theps in D cou ld not be de mons t r a t e d (r = 0.18).

Table 3. Multivariate and univariate analysis of prognostic covariates in 229 breast cancer patients

Variable Disease-free survival Overall survival

Univariate Multivariate Relative Risk Multivariate Relative Risk p-value p-value (95% CI a) p-value (95% CI a)

uPA 0.0000 (>2.97 vs <_2.97 ng/mg protein)

Lymph node status 0.0008 (positive vs negative)

Progesterone receptor status 0.0002 (<20 vs .>20 fmol/mg protein)

PAI- 1 0.0000 (>2.18 vs <_2.18 ng/mg protein)

Menopausal status 0.0373 (peri-/pre- vs post-menopausal)

Estrogen receptor status 0.0027 (<20 vs _>20 fmol/mg protein)

Hormone receptor status 0.0004 (negative vs positive)

Cathepsin D n.s. b (>50 vs _<50 pmol/mg protein)

0.0002 3.0 (1.7-5.5) 0.0041

0.0014 2.6 (1.4-5.2) 0.0075

0.0013 2.6 (1.4-4.7) 0.0459

0.0634 1.8 (0.98-3.5) 0,4053

0.0882 - - 0.3132

0.2813 - - 0.3895

0.3655 - - 0.4024

- - - - n . s ,

3.1 (1.4-7.0)

3.2 (1.3-7.7)

2.2 (1.0-4.7)

a CI = confidence interval; b n.s. = not significant.

2 0 2 F Jiinicke et al

£1

0

1.0"

0 . 8 -

0.6"

0,4-

0.2"

0.0

Estrogen Receptors Progesterone Receptors T ~ _ 1 .0 -~- -L1._ .

-"""-~--- ] ~ ---~p o . . . . . . 0002 ;

"~ . . . . . . 1 o . s .

tp = 0.0027 - ! • 0.6 ' L 0.4 "t

. . . . pos i t i ve 0.2" - - - - posit ive - - negative - - negative

. . . . . , . i • i • 1 0 . 0 . i . , - , . i . I • •

10 20 30 40 5 0 6 0 0 10 20 30 4 0 5 0 60 Months Months

. . . . 170 P~rtlen~ 28 Events . . . . 143 Patients 20 Events - - 59 Patients 21 Events - - 86 Patients 2g Events

Figure 3. Disease-free survival as a function of estrogen or progesterone receptor status in breast cancer patients (n = 229). Patients with positive estrogen or progesterone receptor status had a significantly higher disease-free survival than those with negative receptor status. Evidently, curves initially separating do converge after a longer observation time.

Prognostic evaluation of uPA, PAl-l, and cathepsin D in all patients

In order to determine the prognostic impact of uPA, PAI-1, and cathepsin D compared to the

classical risk factors in breast cancer, univariate and multivariate analyses of disease-free and overall survival were performed using Cox's proportional hazard model. In univariate analyses the optimal cut-off of uPA for disease-free

Table 4. Multivariate and univariate analysis of prognostic covariates for disease-free survival in 101 axillary node-negative breast cancer patients

Variable Univariate Multivariate Relative risk p-value p-value (95% CI a)

uPA 0.0098 0.0136 (>2.96 vs _<2.96 ng/mg protein)

PAI- 1 0.0036 0.0189 (>2.18 vs _<2.18 ng/mg protein)

Vascular invasion 0.0267 0.2954 (present vs absent)

Cathepsin D 0.0771 0.4687 - - (>50 vs .<50 pmol/mg protein)

Estrogen receptor status 0.0503 0.6386 - - (<20 vs _>20 fmol/mg protein)

Hormone receptor status 0.0712 0.6583 - - (negative vs positive)

Tumor size 0.6218 - - - - (> 1.4 cm vs .<1.4 cm)

Progesterone receptor status 0.9336 - - - - (<20 vs >_20 fmol/mg protein)

5.5 (1.2-26.2)

4.9 (1.4-17.2)

2.5 (0.5-12.5)

a CI = confidence interval

uPA and PAl-1 as prognostic' factors 203

survival and also for overall survival was calculated to be 2.97 ng uPA/mg protein (95% confidence interval [CI]: 1.73-3.93); for PAI-1 a cut-off of 2.18 rig/rag protein (95% CI: 0.93-3.49) was obtained.

The major new finding is that breast cancer patients with either high uPA (>2.97 rig/rag pro- tein) or high content of the uPA-inhibitor PAI-1 (>2.18 ng/mg protein) in their primary tumors have an increased risk of relapse and death (Figure 1). The multivariate analyses revealed uPA to be an independent and strong prognostic

factor for both disease-free and overall survival. The impact of uPA is as high as that of the lymph node status, closely followed by the progesterone receptor status and PAI-1 (Table 3).

In contrast to uPA and PAl-l , no reliable cut-off for disease-free survival was found for cathepsin D, when analyzed in the entire group of patients. However, cathepsin D appears to be a prognostic factor in the subgroup of postmeno- pausal patients (n = 139; cut-off 41 pmol cathep- sin D/mg protein; see Figure 2). Neither in node-positive patients nor in premenopausal

1 . 0

0 . 8

0 .6

0 . 4 .

~o

0 . 2 -

uPA

p = 0.0098

1 . 0 -

0 . 8 4

0 . 6 -

0 . 4 -

0 , 2 -

PAI-I

a- 1--L ~ - - L . . . . . . . . . . .

t p = 0 .0036

. . . . u P A <= 2 .96 n g / m g . . . . PAl-1 <= 2 ,18 n g / m g

- - u P A > 2 ,96 r ig / rag - - P A I - t > 2 . 1 8 n g / m g

0 ,0 ' ' ' ' " ' ' - ~ - " , ' , 0 . 0 - . , . , . , • , •

0 1 0 2 0 3 0 4 0 5 0 6 0 1 0 2 0 3 0 4 0 M o n t h s M o n t h s

. . . . 61 Pa t ien ts 3 Even ts - - - 85 Pat ien ts 7 Even ts

- - 40 Pa t ien ts 9 Even ts - - 16 Pat ien ts 5 Even ts

5 0 6 0

C3 >,

=o #_

1 . 0 -

0 . 8 -

0 . 6 -

0 . 4 -

0 , 2 -

Cathepsin D Estrogen Receptors

p = 0 .0771 p = 0 . 0 5 0 3

1.0 ¸

0 . 8

0 . 6

0 . 4

0 . 2

O.O

. . . . C a t h e o s i n D <= 50 p m e l / m g . . . . p o s i t i v e

- - C a t h e p s i n D > 50 p m o l / m g - - n e g a t i v e

0 1 0 2 0 3 0 4 0 5 0 6 0 0 1 0 2 0 3 0 4 0 5 0 M o n t h s M o n t h s

. . . . 64 Pa t ien ts 5 Even ts . . . . 73 P a t i e n t s 7 Even ts

- - 33 Pa t i en t s 6 E v e n t s - - 28 Pa t ien ts 5 Even ts

6 0

Figure 4. Disease-free survival as a function of uPA, PAI- 1, cathepsin D, or estrogen receptor status in axillary node-negative breast cancer patients (n = 101). Patients with high uPA, high PAI-I, or negative estrogen receptor status have a significantly lower rate of disease-free survival than patients with low uPA, low PAI-I, or positive estrogen receptor status, For cathepsin D, the difference between the two curves is approaching statistical significance (p = 0.077).

204 F Jiinicke et al

1,0"

0 .8-

m, 0 . 6 -

0 .4-

n

0.2"

uPA or PAl-1 high

uPA and PAl-1 high

p = 0.0003

O.C 0 10 20 30 4 0 5 0 60

Months

... . . . . 55 Patients 2 Events

. . . . . 36 Patients 6 Events

- - 10 Patients 4 Events

Figure 5, Disease-free survival as a function of uPA plus PAI-1 in axillary node-negative breast cancer patients (n = 101). When both uPA and PAI-I are introduced into the prognostic evaluation, three groups of patients can be defined with increasing risk of relapse. A group of patients with particularly good prognosis (comprising 55 % of the node-negative patients) is thus identified by low uPA and low PAI-content.

patients could a reliable cut-off point be cal- culated for cathepsin D.

As reported earlier by others [35], we also confirm in our group of patients the transient prognostic relevance of progesterone and estrogen receptors for disease-free survival. Curves defined by estrogen or progesterone receptor status do converge after a longer time of obser- vation, whereas those of uPA and PAI-1 divide further (Figure 1 and 3).

Cell cycle analyses (ploidy and S-phase) were also performed. Tumors of 153 patients were screened; 114 patients (75%) had aneuploid tumors. Due to overlap of peaks in the DNA histograms, S-phase determination in aneuploid tumors was not attempted. In the diploid tumors (n = 39) a strong impact of S-phase (optimal cut-off = 7%) was observed in the univariate analysis. S-phase status was not subjected to multivariate analysis due to the small number of diploid tumors. In our group of patients, ploidy is not a statistically significant prognostic in- dicator.

Prognostic evaluation of uPA, PAl-l, and cathepsin D in node-negative patients

One of the concerns in using multivariate analysis to weight prognostic factors is the possible influence of adjuvant treatment. No adjuvant chemo- or hormone therapy was administered to our node-negative patients; therefore, in this group of patients, the relative impact of uPA, PAI-1, cathepsin D, and the established risk factors could be analyzed without these concerns. First, the tumor variables uPA, PAI- 1, and cathep- sin D, and the classical risk factors such as hormone receptors, tumor size, and vascular invasion, were subjected to univariate analysis. The optimal cut-off value of uPA for disease-free survival in node-negative patients was 2.96 ng uPA/mg protein (95% CI: 0.96-3.78); for PAI-1 a cut-off of 2.18 ng/mg protein (95% CI: 1.30- 3.50) was obtained. These cut-offs are identical to those obtained for the entire group of patients. In the node-negative subgroup, a cut-off value for cathepsin D was determined at 50 pmol/mg pro- tein, which was almost statistically significant (p = 0.077). In this respective analysis the prognostic relevance of vascular invasion and estrogen receptor for node-negative breast cancer patients was evident, confirming previous reports [23]. Tumor size and progesterone receptor status did not appear to be correlated to prognosis in our node-negative patients (Figure 4 and Table 4).

The multivariate analysis of disease-free survival disclosed that both uPA and PAI-1 are independent and strong prognostic factors in node-negative patients. The impact of uPA is closely followed by that of PAl-l, whereas the predictive value of cathepsin D was lower (Table 4). Since uPA and PAI-1 are independent prog- nostic factors, the node-negative patients could be grouped even further by a combination of these two variables. In this refined analysis, patients whose primary tumors have low levels of both antigens evidently have a very low risk of relapse (93% disease-free survival at three years) in

contrast to patients with high uPA and high PAI-1 (55% disease-free survival at three years) (Figure 5).

Discussion

Our investigation reveals that the uPA antigen content of tumor tissue extracts, as determined by ELISA, has a strong impact on disease-free sur- vival in axillary node-negative breast cancer patients. Since uPA is an independent prognostic factor, high-risk patients can even be identified within the classical risk groups defined by lymph node or hormone receptor status. Considering both node-negative and node-positive patients, the prognostic impact of uPA is comparable to that of the lymph node status, which up till now had been accepted as the strongest prognostic factor in breast cancer [36].

The rank of PAI-1 in multivariate analysis is close in order to that of uPA. This seems some- what contradictory, since one would expect high levels of the inhibitor PAl- 1 to act protectively by blocking the enzymatic activity of receptor-bound uPA. Nevertheless, excess release of PAL1 may be of importance for reimplantation of circulating tumor cells at distant loci, as generation and growth of metastases should be supported by formation of a new tumor stroma via preventing the uPA-mediated degradation of the extracellular matrix. This hypothesis is consistent with previous findings where a lower proteolytic uPA activity was determined in the metastases com- pared to primary tumors [37-39]. In breast cancer, this is due to the fact that higher levels of PAI-1 are expressed in the metastatic lesion than in the primary tumor, even though the uPA anti- gen content at both locations is similar [19]. The statistically significant correlation between S- phase and PAI-1 levels and also the inverse correlation of PAl-1 to hormone receptor status might even indicate a role of PAI-1 in cell prolif- eration (Table 2).

uPA and PAI-1 as prognostic factors 205

The combination of the two independent variables uPA and PAI-1 should therefore result in an even more individualized delineation of those patients having a low or high risk of relapse. Indeed, the combination of both factors considerably enhances the prognostic value for node-negative patients. In our collective of 101 node-negative patients, high levels of uPA and PAl-1 indicate a 26-fold increased risk for relapse (Table 4 and Figure 5). In addition, this type of analysis allows the delineation of patients with a good prognosis (93% disease-free survival at 36 months).

The cut-off values for uPA and PAI-1 in our prospective study were determined and confirmed by different statistical methods (CART, bootstrap) and remained stable within the 95% confidence interval, when analyzed after 12, 24, and 30 months. No statistically significant differences between the cut-off values determined in the entire group of patients and those determined separately in node-positive and node-negative patients were observed. Nonetheless, when cut- points of continuous factors are determined by finding the best value within a data set, they must be interpreted cautiously until they are indepen- dently validated. Therefore, retrospective studies on frozen tumor tissues obtained from another set of patients should be encouraged to confirm the cut-off points independently.

Another tumor-associated protease, cathepsin D, has also been shown by separate investigators in different countries to be of prognostic rel- evance in breast cancer patients, particularly in node-negative disease [40,41]. Despite approach- ing statistical significance in the univariate analysis of our node-negative patients, cathepsin D ranks below uPA, PAl-l, and vascular invasion in the multivariate analysis (Table 4). Moreover, we confirm the results of a large Danish trial in which the prognostic impact of cathepsin D is confined to postmenopausal patients [42]. In our group of node-negative breast cancer patients, not a single relapse was identified by cathepsin D

206 F Jiinicke et al

which could not have been predicted by uPA or PAI-1.

By itself, estrogen receptor status is not a sufficiently strong prognostic factor in node- negative breast cancer patients to warrant basing treatment decisions on it alone, since the 5-year disease-free survival difference (although highly significant) is only 8-9%. However, combining estrogen receptor with other variables has been recommended [23]. In our series the inclusion of estrogen receptor in risk stratification, together with uPA and PAI-1, gave no further prognostic information (Table 4).

Tumor size is an important prognostic indica- tor in axillary node-negative patients, as tumors smaller than 1 cm indicate an excellent prognosis [23]. As in other studies [41] dealing with deter- mination of prognostic factors in tumor tissues, the distribution of tumor size in our study is different from the general population. In our group of node-negative patients the median tumor diameter was 2.15 cm (0.7-7.1 cm); only one patient had a tumor smaller than 1 cm. In this confined group of node-negative patients tumor size had no significant impact on prognosis, in- dicating that size may merely reflect the age of the tumor and to a lesser degree its aggressiveness [23,41,43].

Clark et al. showed by flow cytometry that S-phase and ploidy status have a greater prog- nostic impact in axillary node-negative patients than tumor size or hormone receptor status [44]. In diploid tumors, high and low S-phase implied high and low risk of relapse, respectively (70 versus 90% probability of disease-free survival at five years). Sigurdsson et al., who were able to determine S-phase in euploid and aneuploid tumors by this approach, combined their findings in a prognostic score with progesterone receptor and tumor size [43]. They thus identified a low-risk group consisting of 63% of the axillary node-negative patients which had a survival rate comparable to the general age-related mortality in Sweden. However, flow cytometry has the dis-

advantage that within the group of aneuploid tumors, comprising approximately two-thirds of all patients, S-phase determination is difficult in those cases where overlap of histograms is evident. We were only able to assess reliably S-phase in a few cases of aneuploid tumors. Thus, our S-phase analysis is virtually restricted to diploid tumors. Similar methodological prob- lems have been encountered by other investigators [44]. Hence, a weak point in our analysis is the resulting low number of diploid patients, which was too low for multivariate regression analysis. Nevertheless, within the small subgroup of patients with diploid tumors (n = 39), a significant prognostic impact of S-phase on disease-free survival (optimal cut-off 7%, p = 0.02) was determined by univariate analysis.

Some of the recently evaluated prognostic indicators are used to analyze prognosis within defined subgroups of patients [23,45]. For example, when Her/2-neu was analyzed within a subgroup of node-negative patients with tumor size <3 cm and positive estrogen receptors, a high-risk subset could be defined (55% probabil- ity of recurrence). This small subgroup of about 10% of all axillary node-negative patients com- prises less than one-sixth of all axillary node-negative patients who will eventually have a relapse. An alternative approach to this subgroup-restricted analysis is to take advantage of factors which can be used for the entire node- negative population. In the present investigation, we have shown that the high prognostic impact of uPA and PAl- 1 distinguishes patients with high or low risk for relapse even among the entire group of node-negative patients. The combination of uPA and PAI-1 in our group of patients with axillary node-negative breast cancer allowed us to identify a group of 45% of patients having an increased risk of relapse. Consequently, more than half of the patients had a probability of relapse of less than 10 % and thus would be candidates for being spared the necessity of adjuvant therapy. The question of whether high

levels of uPA and PAI-1, which put patients into a high-risk group, may affect response to adjuvant chemotherapy can only be answered by a clinical trial.

R e f e r e n c e s

1. Dane K, Andreasen PA, Gr0ndahl-Hansen J, Kristensen P, Nielsen LS, Skriver L: Ptasminogen activators, tissue degradation and cancer. Adv Cancer Res 44: 139-266, t985 Markus G: The relevance of plasminogen activators to neoplastic growth. Enzyme 40:158-172, 1988 Blasi F: Surface receptors for urokinase plasminogen activator. Fibrinotysis 2:73-84, 1988 Mangel WF: Better reception for urokinase. Nature 344:488-489, 1990 Schlechte W, Murano G, Boyd D: Examination of the role of the urokinase receptor in human colon cancer mediated laminin degradation, Cancer Res 49:6064- 6069, 1989 Stoppetli MP, Tacchetti C, Cubellis MV, Corti A, Hearing VJ, Cassani G, Appella E Blasi F: Autocrine saturation of pro-urokinase receptors on human A431 cells. Cell 45:675-684, I986 Vassalli JD, Baccino D, Belin D: A cellular binding site for the Mr 55,000 form of the human plasminogen activator, urokinase. J Cell Biol t00:86-92, 1985 Bachmann F: Thrombosis and haemostasis. In Ver- straete M, Lijnen HR, Arnout J (eds). International Society on Thrombosis and Haemostasis, Leuven Uni- versity Press, Leuven, 1987, pp 227-265 Kobayashi H, Schmitt M, Goretzki L, Chucholowski N, Calvete J, Kramer M, Ganzler WA, J~inicke F, Graeff H: Cathepsin B efficiently activates the soluble and the tumor cell receptor-bound form of the proenzyme urokinase-type plasminogen activator (pro-uPA). J Biol Chem 266:5147-5152, t990

10. Burtin P, Chavanel G, Andr6 J: The plasmin system in human colonic tumors: an immunofluorescence study. Int J Cancer 35:307-314, 1985

11. Miles L A , Plow EF: Plasminogen receptors: Ubiqui- tous sites for cellular regulation of fibrinolysis. Fibrinolysis 2:61-71, 1988

12. Trygvasson K: Extracellular matrix and its enzymatic degradation in tumor invasion (A review). In Liotta LA (ed) Influence of Tumor Development on the Host. Kluwer Academic Publishers, Dordrecht, 1989, pp

2.

3.

4.

5.

6.

7.

8.

9.

uPA and PAI-t as prognostic factors 207

72-83 13. Dvorak HF: Tumors: Wounds that do not heal. N Engl

J Med 315:161-165, 1986 14. Wilhelm O, Hafter R, Coppenrath E, Pflanz MA,

Schmitt M, Babic R, Linke R, GOssner W, Graeff H: Fibrin-fibronectin compounds in human ovarian tumor ascites and their possible relation to the tumor stroma. Cancer Res 48:507-3514, 1988

15. Wilhelm O, Hafter R, Henschen A, Schmitt M, Graeff H: Role of plasmin in the degradation of the stroma- derived fibrin in human ovarian carcinoma. Blood 75:1673-1678, 1990

16. Duffy MJ, Reilly D, O'Sullivan C, O'Higgins N, Fennelly JJ, Andreasen P: Urokinase-plasminogen activator, a new and independent prognostic marker in breast cancer. Cancer Res 50:6827-6829, 1990

!7. Duffy MJ, Reilly D, O'Sullivan C, O'Higgins N, Fennelly JJ: Urokinase plasminogen activator and prognosis in breast cancer. Lancet 335:108, 1990

18. J~inicke F, Schmitt M, Ulm K, G{Sssner W, Graeff H: Urokinase-type plasminogen activator antigen and early relapse in breast cancer. Lancet ii:1049, 1989

19. J~inicke F, Scbmitt M, Hafter R, Hollrieder A, Babic R, Ulm K, G6ssner W, Graeff H: Urokinase-type plasmin- ogen activator (u-PA) antigen is a predictor of early relapse in breast cancer. Fibrinolysis 4:69-78, 1990

20. Jfinicke F, Schmitt M, Graeff H: Clinical relevance of the urokinase-type and the tissue-t~e plasminogen activators and of their inhibitor PAI-1 in breast cancer. Seminars in Thrombosis and Hemostasis 17:303-312, 199l

21. Duffy M, O'Grady P, Devaney D, O'Siorain L, Fennel- ly JJ, Lijnen HJ: Urokinase-plasminogen activator, a marker for aggressive breast carcinomas. Cancer 62:531-533, t988

22. Early Breast Cancer Trialists' Collaborative Group: Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune therapy. 133 randomised trials involving 31,000 recurrences and 24,000 deaths among 75,000 women. Lancet 339:1-15 + 71-85, 1992

23. McGuire WL, Tandon AK, Altred DC, Chamness GC, Clark GM: How to use prognostic factors in axillary node-negative breast cancer patients. JNCI 82:1006- 1015, 1990

24. Schmitt M, Kanayama N, Henschen A, Hollrieder A, Halter R, Gulba D, J~nicke F, Graeff H: E la s t a se released from human granulocytes stimulated with N-formyl-chemotactic peptide prevents activation of tumor cell prourokinase. FEBS Lett 255:83-88, 1989

25. Schmitt M, J~inicke F, Hailer R, Hollrieder A, Kana- yama N, Gulba D, Graeff H: Tumor-associated fibrin-

208 F Jiinicke et at

olysis in human breast cancer: Detection and quanti- ration of the urokinase-type plasminogen activator (uPA) by ELISA and immunohistochemistry. In Matsuda M, Iwanaga S, Takada A, Henschen A (eds) Fibrinogen 4. Current Basic and Clinical Aspects. Elsevier, Amsterdam, 1990, pp 213-222

26. McGuire WL, De La Garza M, Chamness GC: Evalua- tion of estrogen receptor assays in human breast cancer tissue. Cancer Res 37:637-639, 1977

27. Hedley DW: Flow cytometry using paraffin-embedded tissue: Five years on. Cytometry 10:229-241, 1989

28. Cox DR: Regression models and life-tables. J R Star Soc (B) 34:187-220, 1972

29. Hawkins DM: Testing a sequence of observations for a shift in location. J Am Star Assoc 72:180-186, 1977

30. Segal MR, Bloch DA: A comparison of estimated proportional hazards models and regression trees. Stat Med 8:539-550, 1989

31. Breiman L, Friedman JH, Olsen RA, Stone CJ: Classification and Regression Trees. Wadsworth, Belmont, 1984

32. Altman DG, Andersen PK: Bootstrap investigation of the stability of a Cox regression model. Stat Med 8:771-783, 1989

33. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958

34. Peto R, Peto J: Asymptotically efficient rank invariant procedures. J Royal Stat Soc (A) 135:185-207, 1972

35. H~ihnel R, Woodings T, Vivian AB: Prognostic value of estrogen receptors in primary breast cancer. Cancer 44:671-675, 1979

36. McGuire WL: Prognostic factors for recurrence and survival in human breast cancer. Breast Cancer Res Treat 10:5-9, 1987

37. Camiolo SM, Markus G, Piver MS: Plasminogen acti-

vator content of gynecological tumors and their meta- stases. Gynecol Oncol 26:364-373, 1987

38. Harvey SR, Lawrence DD, Madeja JM, Abbey S J, Markus G: Secretion of plasminogen activators by human colorectal and gastric tumor explants. Clin Exp Metastasis 6:431-450, 1988

39. Markus G, Camiolo SM, Kohga S, Madeja JM, Mittel- man A: Plasminogen activator secretion of human tumors in short-term organ culture, including a comparison of primary and metastatic colon tumors. Cancer Res 43:5517-5525, 1983

40. Spyratos F, Maudelonde T, Brouillet J-P, Brunet M, Defrenne A, Andrieu C, Hacene K, Desplaces A, Rouesse J, Rochefort H: Cathepsin D: An independent prognostic factor for metastasis of breast cancer. Lancet ii:l 115-1118, 1989

41. Tandon AT, Clark GM, Chamness GC, Chirgwin JM, McGuire WL: Cathepsin D and prognosis in breast cancer. N Engl J Med 322:297-302, 1990

42. Thorpe SM, Rochefort H, Garcia M, Freiss G, Christen- sen IJ, Khalaf S, Paolucci F, Pau B, Rasmussen BB, Rose C: Association between high concentrations of Mr 52,000 cathepsin D and poor prognosis in primary human breast cancer. Cancer Res 49:6008-6014, 1989

43. Sigurdsson H, Baldetorp B, Borg A, Dalberg M, Fern/3 M, Killander D, Olsson H: Indicators of prognosis in node-negative breast cancer. N Engl J Med 322:1045- 1053, 1990

44. Clark GM, Dressler LG, Owens MA, Pounds G, Oldak- er T, McGuire WL: Prediction of relapse or survival in patients with node-negative breast cancer by DNA flow cytometry. N Engl J Med 320:627-633, 1989

45. Slamon DJ, Clark GM, Wong SG, Levin WJ, Ullrich A, McGuire WL: Human breast cancer: Correlation of relapse and survival with amplification of the Her-2/neu oncogene. Science 235:177-182, 1987