Urokinase plasminogen activator - Semantic Scholar

Clinical Chemistry 44:6 1177–1183 (1998)

Enzymes and Protein Markers

Urokinase plasminogen activator: a prognostic marker in breast cancer including patients with axillary node-negative disease Michael J. Duffy,1* Catherine Duggan,1 Hugh E. Mulcahy,3 Enda W. McDermott,2 and Niall J. O’Higgins2

Urokinase plasminogen activator (uPA) is a serine protease causally involved in cancer invasion and metastasis. In this study, high concentrations of uPA in primary breast cancers were independently associated with both a shortened disease-free interval and overall survival. For the disease-free interval as endpoint, uPA was a stronger indicator of outcome than lymph node status, whereas for overall survival, nodal status was stronger than uPA. In patients without metastasis to axillary nodes, uPA was also an independent prognostic marker, using both the disease-free interval and overall survival as end points. In contrast to uPA, neither tumor size nor estrogen receptor status was prognostic in the nodenegative patients. Measurement of uPA concentrations might thus be of value in selecting the more aggressive subpopulation of node-negative breast cancer patients that could benefit from adjuvant therapy. Breast cancer is the most common fatal malignancy in women in the Western world (1). Recently, randomized clinical trials have shown that adjuvant treatment with either combination chemotherapy or hormonal therapy improved the outcome of patients with this malignancy (2). Because of these findings, it has been argued that adjuvant therapy should be given to all breast cancer patients. However, for many, this approach would be overtreatment (3). Thus, ;70% of axillary node-negative patients are cured of their disease by local surgery, radiotherapy, or a combination of these two treatments,

without any recourse to adjuvant therapy. A large number of women without disease would therefore be subjected to both the toxic side effects and costs of this therapy. In addition, because the clinical benefits of presently available therapies are relatively small, only a minority of these node-negative patients would benefit. A practical solution to this dilemma is to use prognostic markers to separate patients into low- and high-risk groups, based on metastatic potential, and focus treatment on patients with high risk of relapse (3). Urokinase plasminogen activator (uPA)4 is a serine protease with multiple functions, enabling it to play a role in cancer progression [reviewed in (4, 5)]. In particular, it can both catalyze the degradation of the extracellular matrix and stimulate cellular migration (4, 5). These two events are critical for malignant cell invasion and metastasis. Results from model systems have shown that uPA is causally involved in cancer spread (4, 5). Consistent with these findings from experimental systems, uPA has also been shown to be a marker of metastatic potential or prognosis in several human cancers, including breast, colorectal, and gastric malignancies [reviewed in (6)]. In this study, we show, using a commercially available kit, that high concentrations of uPA correlate with both shortened disease-free interval and overall survival in breast cancer patients, including women with node-negative disease.

Materials and Methods methods

Departments of 1 Nuclear Medicine and 2 Surgery, St. Vincent’s Hospital, Dublin 4, Ireland. 3 Department of Gastroenterology, St. Bartholomew’s Hospital, London EC1 M 6BQ, UK. * Address correspondence to this author at: Department of Nuclear Medicine, St. Vincent’s Hospital, Elm Park, Dublin 4, Ireland. Fax 353-1-269 6018; e-mail [email protected]. Received May 1, 1997; revision accepted February 25, 1998.

After histological examination, breast cancers were snap frozen in liquid nitrogen and then transferred to a 270 °C freezer. Tumors were homogenized in 50 mmol/L Tris-

4 Nonstandard abbreviations: uPA, urokinase plasminogen activator; ER, estrogen receptor; uPAR, uPA receptor; and PAI, plasminogen activator inhibitor.

1177

1178

Duffy et al.: Urokinase plasminogen activator

HCl (pH 7.4) containing 1 mmol/L monothioglycerol, as described previously (7). The homogenate was centrifuged at 2000g for 10 min, and the supernatant was extracted using 10 mL/L Triton X-100. Centrifugation was then carried out at 10 000g for 20 min. uPA was assayed on the supernatant using ELISA (American Diagnostica, Greenwich CT). This ELISA kit detects precursor uPA, active uPA, and uPA inhibitor complexes and has a stated detection limit of 10 ng/L. Total protein was measured by using kits obtained from Bio-Rad. uPA concentrations were expressed as ng uPA/mg protein. Estrogen receptors (ERs) were assayed as described previously (8). The cutoff point used was 200 fmol/g wet weight of tissue.

patient cohort We studied 184 consecutive patients on whom ER status, axillary node status, and tumor size were known. All of the carcinomas used were submitted for routine steroid receptor determination. Detailed characteristics of the tumors, as well as patients’ ages and adjuvant therapies administered, are listed in Table 1. Median follow-up of patients alive at the end of the study was 6.9 years (range, 0 –10.4 years). Fifty-six patients died during the follow-up period. The cutoff point used for discriminating between patients with high and low uPA concentrations was 0.81 ng/mg total protein. This was determined by using the maximal log-rank test and was identical to that found previously by us (7). Cox regression analysis was performed to determine that uPA was an independent prognostic marker. A backwards regression model was used to Table 1. Details of tumors, patients’ ages, and adjuvant therapies administered for patients with low and high levels of uPA.

Size 0–2 cm 2–5 cm .5 cm Nodal status Negative Positive ER status Negative Positive Adjuvant therapy None Tamoxifen Chemotherapy Tamoxifen and chemotherapy Age, years Mean Range

Low uPA n (%)

High uPA n (%)

49 (34.5) 68 (47.9) 25 (17.6)

15 (35.7) 23 (54.8) 4 (9.5)

72 (50.7) 70 (49.3)

15 (35.7) 27 (64.3)

62 (43.7) 80 (56.3)

17 (40.5) 25 (59.5)

36 (25.4) 85 (59.9) 16 (11.3) 5 (3.5)

8 (19.0) 28 (66.7) 5 (11.9) 1 (2.4)

51.8 30.1–86.0

50.8 34.4–82.1

achieve the most powerful set of predictor variables. Two-sided P values of ,0.05 were considered significant.

Results preliminary experiments The concentration of uPA in the breast cytosols varied widely, i.e., from 0 – 63.3 mg/L (median value, 0.66 mg/L) or from 0 –10.2 mg/g protein (median value, 0.34 mg/g protein). These values are shown in the form of a frequency distribution graph in Fig. 1. Because the concentration of calibrators in the ELISA ranged from only 0 –1 mg/L, many cytosol samples had to be diluted. It was, therefore, essential to verify that the values obtained after dilution displayed parallelism. Values corrected for dilution should give the same result, irrespective of the extent of dilution. Fig. 2 shows the linearity of values after dilution of four different breast cancer cytosols. Clearly, good parallelism was obtained over the dilutions used, i.e., 1:2, 1:4, 1:8, 1:16, and 1:32. We also checked the between-assay variation for the uPA ELISA kit. For this purpose, pooled sera from patients with metastatic breast cancers was used. This control sample was assayed on 17 different days over a period of 15 months, using kits with six different lot numbers. The mean uPA value of the 17 determinations was 0.303 mg/L, with an overall CV of 11.43%.

correlation of uPA with patient outcome All patients. As shown in Fig. 3, patients with high concentrations of uPA had both a shorter disease-free interval (P 5 0.0008) and shorter overall survival (P 5 0.0004) than patients with low concentrations of the protease. As a continuous variable, uPA concentrations were also significantly related to both the disease-free interval (P ,0.0001) and overall survival (P 5 0.0018). Table 2 compares the relative strength of uPA with established prognostic markers for breast cancer, using both univariate and multivariate analyses. With the exception of tumor size for the disease-free interval, all of the investigated parameters, (i.e., nodal status, ER status, tumor size, and uPA) correlated significantly with patient outcome in univariate analysis. By using multivariate analysis, however, only uPA and lymph node status predicted the disease-free interval, whereas uPA, lymph node status, and size correlated with overall survival. With disease-free interval as end point, uPA was a stronger prognostic marker than either nodal status, ER status, or tumor size. On the other hand, for overall survival, uPA was weaker than lymph node status but more powerful than either tumor size or ER status in predicting outcome. Axillary node-negative patients. In the axillary node-negative patients, high concentrations of uPA were also associated with both a shortened disease-free interval (P 5 0.0001) and overall survival (P 5 0.0002; Fig. 4). In this subgroup of patients, uPA was the only significant pre-

Clinical Chemistry 44, No. 6, 1998

1179

Fig 1. Frequency distribution of uPA in breast cytosols.

dictor of outcome and was significant when using both univariate and multivariate analyses (Table 3). Patients receiving different adjuvant therapies. uPA was also investigated for possible prognostic value in patients receiving different types of adjuvant treatment. For the

group of patients treated with adjuvant tamoxifen, uPA concentrations correlated significantly with both diseasefree interval (relative risk, 3.27; P 5 0.0002; x2, 13.9) and overall survival (relative risk, 4.7; P ,0.0001; x2 5 22.5). uPA was not prognostic in those groups of patients who either received no adjuvant therapy or who were given

Fig 2. Linearity of uPA values for four different breast cytosols after dilution. Dilutions were as follows: (1) 1:2; (2) 1:4; (3) 1:8; (4) 1:16; and (5) 1:32.

1180


Fig 3. Disease-free survival and overall survival of 184 patients with breast cancer, according to tumor uPA expression (u-PA low ,0.81 mg/g protein; uPA high .0.81 mg/g protein).

adjuvant chemotherapy. It should, however, be pointed out that the numbers of patients in both of these subgroups were relatively small, i.e., 44 and 21, respectively.

Discussion To our knowledge, this is the first report to show that high concentrations of uPA in node-negative breast cancers are significantly associated with both a shortened disease-free interval and overall survival. Previously, uPA was shown to be a prognostic marker for total populations of breast cancer [i.e., both node-negative and node-positive patients; reviewed in (6)]. By using a different group of

patients and a different ELISA for uPA, we demonstrated previously that high concentrations of the protease predicted a shortened disease-free interval in axillary nodenegative breast cancer patients (8). However, in that study, uPA was not shown to be an independent prognostic marker for node-negative disease. Furthermore, in our previous study (9), uPA was not related significantly to overall survival, using P ,0.05 as a criteria of significance. Others have also reported uPA to be a prognostic marker in node-negative breast cancer by using the disease-free interval as the end point (10 –12). Although multiple groups have now reported uPA to

Table 2. Univariate analysis and multivariate regression analysis of prognostic features in 184 patients with breast cancer. Disease-free survival Univariate analysis % DFa n

uPA Low (0.81 ng/mg protein) 142 High (.0.81 ng/mg 42 protein) Lymph nodes involved 0 nodes 87 1–3 nodes 49 $4 nodes 48 Tumor size 2 cm 64 2–5 cm 91 .5 cm 29 Estrogen receptor status Negative 76 Positive 108 a

DF, disease free. Log-rank test. c Log-rank test for trend. d —, excluded from regression analysis. b

Overall survival

Final regression model

Univariate analysis


at 8 y

P

Relative risk (95% CI)

P

% alive at 8 y

69.1 30.3

0.0008b

2.3 (1.3–4.0)

0.003

69.5 39.2

0.0004b

3.0 (1.7–5.3)

0.0003

0.02

81.7 55.7 37.2

,0.0001c

3.6 (2.1–6.1)

,0.0001

77.0 58.4 48.6

0.003c

2.5 (1.3–4.8)

0.006

49.2 73.7

0.02b

—d

71.6 55.6 48.1

0.003c

2.0 (1.1–3.4)

74.9 53.2 60.7

0.08c

—d

51.5 69.8

0.02b

0.6 (0.3–1.0)

0.05

P


P

1181


Fig 4. Disease-free survival and overall survival of 87 patients with node-negative breast cancer, according to tumor uPA expression (uPA low ,0.81 mg/g protein; uPA high .0.81 mg/g protein).

be prognostic in breast cancer patients, it should be pointed out that the cutoff point used by the different investigators varied widely, i.e., from 0.62 mg/g protein to 10 mg/g protein (9 –12). The reasons for these widely different cutoff points are likely to include: (a) type of tumor extract used for assaying uPA, e.g., whether whole homogenate, low speed, or high speed supernatant was used; (b) whether a detergent such as Triton X-100 was used to extract the protease; (c) different calibrators used in the various assays; and (d) specificity of the antibodies used in the different ELISAs. This last point may be particularly difficult to control because uPA can exist in multiple forms in vivo. Thus, in addition to its precursor and mature and low molecular weight forms, uPA can form complexes with its two inhibitors, PAI-1 and PAI-2,

and with its receptor uPAR. Ternary complexes such as uPA-PAI-1-uPAR and uPA-PAI-2-uPAR, as well as quaternary complexes, involving associations between these ternary complexes and the a2-macroglobulin receptor, may also exist (6). Another protease implicated in cancer progression, i.e., cathepsin D (13), has also been shown in some studies to be a prognostic marker in node-negative breast cancer patients (14, 15). However, these results with cathepsin D in node-negative patients have not been confirmed [reviewed in (6, 16)]. Similarly, other biochemical prognostic factors such as ER, progesterone receptor, c-erbB-2 and p53, although prognostic in the total population of breast cancer patients, are not consistently found to correlate with outcome in the node-negative subgroup [reviewed

Table 3. Univariate analysis and multivariate regression analysis of prognostic features in 87 patients with node-negative breast cancer. Disease-free survival Univariate analysis

uPA Low (0.81 ng/mg protein) High (.0.81 ng/mg protein) Tumor size 2 cm 2–5 cm .5 cm Estrogen receptor status Negative Positive a


Univariate analysis


n

% DFa at 8 y

P


P

% alive at 8 y

P


P

72 15

79.0 25.9

0.0001b

4.9 (2.0–12.0)

0.0004

87.9 48.8

0.0002b

62 (2.1–18.6)

0.006

43 38 6

79.0 64.6 66.7

0.16c

87.8 77.0 66.7

0.09c

—d

32 55

73.2 70.5

0.81b

75.7 86.0

0.43b

—d

DF, disease free. Log-rank test. c Log-rank test for trend. d —, excluded from regression analysis. b

Overall survival

—d

—d

1182


in (17–19)]. Interestingly, PAI-1, a naturally occurring inhibitor of uPA, has also been found to be associated with patient outcome in node-negative breast cancer patients (11, 20). According to Clark (21), a prognostic marker in breast cancer should not only predict outcome but should also be able to identify patients likely to be responsive or resistant to particular forms of treatment. uPA may also be of value in this latter area of breast cancer management. Recently, Foekens et al. (22) showed that uPA concentrations in primary breast cancers predicted response of metastatic disease to tamoxifen therapy. This effect of uPA appeared to be independent of both ER and progesterone receptor status. (22). Because uPA is an independent prognostic marker in breast cancer [reviewed in (6)], prognostic in the nodenegative subgroup of patients (6), and a possible predictor of response to hormonal therapy (22), it is a potential candidate for routine use in the management of patients with malignancy of the breast. However, before any new biological marker enters routine use, it should be evaluated properly (23). In particular, it should be evaluated with regard to both analytical performance and clinical value (24). Unlike most new biological prognostic markers described for breast cancer in recent years, these studies have either been carried out or are currently in progress for uPA. For example, in 1996, Benraad et al. (25) reported on the evaluation of six different ELISA kits. Some of the conclusions to emerge from this study were as follows: (a) Although the absolute uPA values measured with the various kits differed, good correlations were obtained between any two of the ELISAs tested. (b) In the two kits evaluated for the effects of cytosol dilution on the measured value, no systematic increase or decrease in uPA concentrations were found. With xenograft cytosols, dilutions as low as 1:400 yielded CVs ,5%. (c) Human breast cancer cytosols with added recombinant pro-uPA and lyophilized gave a mean CV of 11.5% when assayed on 26 separate days over a period of 18 months. This between-assay variation of Benraad et al. (25) is almost identical to what we found in the present study, i.e., 11.43%. In the present report, we also found an acceptable degree of parallelism for uPA after cytosol dilution. Additional work in the area of uPA methodology will require assay standardization and evaluation in External Quality Assurance trials. Regarding the latter, it should be pointed out that such studies have recently been carried out in Europe (Sweep et al., manuscript in preparation). Finally, before routine application of a new marker, randomized trials should show that determination of marker concentrations leads to either enhanced diseasefree interval, extended overall survival, or increased quality of life. To address whether assay of uPA can positively contribute to the management of breast cancer patients, a prospective randomized multicenter study was recently

initiated in Germany. In this trial, axillary node-negative breast cancer patients with high concentrations of uPA (or of its inhibitor, PAI-1) were randomized to receive either six cycles of cyclophosphamide or to be observed. Patients with low concentrations of uPA/PAI-1 were not treated. With this trial, it is hoped that the subgroup of patients that are at high risk of early disease relapse and those who will benefit from adjuvant chemotherapy can be identified. The results of this study may thus lead to a test that will prevent the administration of unnecessary and potentially toxic therapy to the majority of node-negative patients.

This work was supported by The Irish Cancer Society, The Health Research Board of Ireland, The International Association for Cancer Research, and the BIOMED 1 Program of the European Union (Project: Clinical Relevance of Proteases in Tumor Invasion and Metastasis; contract no. CT931346). Some of the uPA kits used were supplied free of charge by American Diagnostica, Inc.

References 1. Borling CC, Squires TS, Tong T. Cancer Statistics CA. Cancer J Clin 1992;42:19 –32. 2. Early Breast Cancer Trialists’ Collaborative Group. Systemic treatment of early breast cancer by hormonal, cytotoxic or immune therapy: 133 randomized trials involving 331,000 recurrences and 24,000 deaths among 75,000 women. Lancet 1992;339:1–15. 3. McGuire WL, Clark CM. Prognostic factors and treatment decisions in axillary node-negative breast cancer. N Engl J Med 1992;326:1756 – 61. 4. Danø K, Behrendt N, Bru¨nner N, Ellis V, Ploug M, Pyke C. The urokinase receptor: protein structure and role in plasminogen activation and cancer invasion. Fibrinolysis 1994;8(Suppl 1):189 – 203. 5. Duffy MJ. Urokinase plasminogen activator and malignancy. Fibrinolysis 1993;7:295–302. 6. Duffy MJ. Proteases as prognostic markers in cancer. Clin Cancer Res 1996;2:613–18. 7. Duggan C, Maguire T, McDermott E, O’Higgins N, Fennelly JJ, Duffy MJ. Urokinase plasminogen activator and urokinase plasminogen activator receptor in breast cancer. Int J Cancer 1995;61:597– 600. 8. Duffy MJ, O’Siorain L, Waldron B, Smith C. Estradiol receptors in human breast carcinomas assayed by the use of monoclonal antibodies. Clin Chen 1986;32:1972– 4. 9. Duffy MJ, Reilly D, McDermott E, O’Higgins N, Fennelly JJ, Andreasen PA. Urokinase plasminogen activator as a prognostic marker in different subgroups of patients with breast cancer. Cancer 1994;74 (8):2276 – 80. 10. Foekens JA, Schmitt M, van Putten WLJ, Peters HA, Bontenbal M, Ja¨nicke F, Klijn JGM. Prognostic value of urokinase-type plasminogen activator in 671 primary breast cancer patients. Cancer Res 1992;52:6101–5. 11. Ja¨nicke F, Schmitt M, Pache L, Ulm K, Harbeck N, Hoffer H, Graeff H. Urokinase (uPA), its inhibitor PAI-1 are strong and independent prognostic factors in node-negative breast cancer. Breast Cancer Res Treat 1993;224:195–208. 12. Ferno¨ M, Bendahl PO, Bo¨rg A, Brundell J, Hirschberg L, Olsson H, Killander D. Urokinase plasminogen activator, a strong and inde-


13. 14.

15.

16. 17. 18.

19.

pendent prognostic factor in breast cancer, analysed in steroid receptor cytosols with a luminometric immunoassay. Eur J Cancer 1996;32A:793– 801. Rochefort H. Cathepsin D in breast cancer: a tissue marker associated with metastasis. Eur J Cancer 1992;28A:1780 –3. Spyratos F, Brouillet JP, Defrenne A, Hacene K, Rousse J, Maudelonde T, et al. Cathepsin D: an independent prognostic factor for metastasis of breast cancer. Lancet 1989;2:1115– 8. Tandon AK, Clarke GM, Chamness GC, Chirgwin J, McGuire WL. Cathepsin D and prognosis in breast cancer. N Engl J Med 1990;322:297–302. Westly BR, May FEB. Cathepsin D and breast cancer. Eur J Cancer 1996;32A:15–24. Duffy MJ. Cellular oncogenes and suppressor genes as prognostic markers in cancer. Clin Biochem 1993;26:439 – 47. Gasparini G, Pozza F, Harris A. Evaluating the potential usefulness of new prognostic and predictive indicators in node negative breast cancer patients. J Natl Cancer Inst 1993;85:1206 –19. Foekens JA, Berns EMJJ, Look MP, Klijn JGM. Prognostic factors in node-negative breast cancer. In: Pasqualini JR, Katzenellenbogen

20.

21. 22.

23. 24. 25.

1183

BS, eds. Hormone dependent cancers. New York: Marcel Decker, 1996:217–53. Foekens JA, Schmitt M, van Putten WLJ, Peters HA, Kramer M, Ja¨nicke F, Klijn JGM. Plasminogen activator inhibitor-1 and prognosis in primary breast cancer. J Clin Oncol 1994;12: 1648 –58. Clarke GM. Do we really need prognostic markers in breast cancer? Breast Cancer Res Treat 1994;30:117–26. Foekens JA, Look M, Peters HA, van Putten WLJ, Portengen H, Klijn JGM. Urokinase plasminogen activator and its inhibitor PAI-1: predictors of poor response to tamoxifen therapy in recurrent breast cancer. J Natl Cancer Inst 1995;87:751– 6. McGuire WL. Breast cancer prognostic factors: evaluation guidelines. J Natl Cancer Inst 1991;83:154 –5. Blankenstein MA. Biochemical assessment of tissue prognostic factors in breast cancer. Breast 1997;6:31–7. Benraad ThJ, Guerts-Moespot J, Grøndahl-Hansen J, Schmitt M, Heuval J, de Witte JH, et al. Immunoassays (ELISA) of urokinasetype plasminogen activator (uPA): report of an EROTC/Biomed workshop. Eur J Cancer 1996;32A:1371– 81.