Assay of matrix metalloproteases types 8 and 9 by ELISA in human breast cancer.

Results from model tumour systems suggest that either increased levels of certain metalloproteases (MMPs) or decreased levels of their inhibitors correlate with metastatic potential. In this study, levels of two MMPs, i.e. MMP-8 and -9, and their inhibitor tissue inhibitor of metalloprotease type 1 (TIMP-1) were measured by enzyme-linked immunosorbent assay in human breast tumours. Levels of MMP-8 and -9 correlated significantly with each other, but neither MMP correlated with urokinase plasminogen activator. Levels of both MMP-8 and -9 were also significantly related to levels of TIMP-1. In contrast, neither MMP correlated with plasminogen activator inhibitor. No relationship was found between MMP-8, MMP-9 or TIMP-1 and either tumour size or metastasis to axillary nodes. MMP-8 and -9 levels were inversely related to levels of oestrogen receptors. MMP-8 but not MMP-9 levels were also inversely correlated with progesterone receptor levels. It is concluded that the assay for MMP-8 and -9 described here will permit the evaluation of these proteases as prognostic markers in cancer.

The matrix metalloproteases (MMPs) are a family of zinc dependent endoproteases which catalyse the degradation of several different molecules in the extracellular matrix (ECM) (for reviews, see Woesner, 1991;Aznavooran et al., 1993). The family can be divided into three main groups: interstitial collagenases, type IV collagenases or gelatinases and stromelysins. Interstitial collagenases degrade type I, II and III collagen. Neutrophil collagenase, which is also known as MMP-8, appears to have similar substrate speificity to the interstitial collagenases (Hasty et al., 1990). Type IV collagenases are so named as they cleave the helical portions of type IV collagen, a form of collagen found at high levels in basement membranes. In addition, type IV collagenase can degrade type V, VII, IX and X collagens (Aznavooran et al., 1993). Two main forms of type IV collagenase have been described with molecular weights of 72 and 92 kDa. These two forms are also known as MMP-2 and MMP-9 respectively. The stromelysins have a broad substrate specificity, degrading molecules such as fibronectin, laminin, proteoglycans and the non-helical portions of type IV collagen (Woessner, 1991;Aznavooran et al., 1993).
Considerable evidence suggests that certain members of the MMP family play a role in cancer invasion and metastasis, at least in model systems. Thus, both interstitial and type IV collagenases correlate with metastatic phenotype in various cell types (reviewed in Duffy, 1992;Aznavooran et al., 1993). In addition, inhibitory antibodies against the 72 kDa form of collagenase IV decrease the penetration of melanoma cells through reconstituted basement membranes (Hoyhtya et al., 1990). Similarly, administration of recombinant TIMP-1 to nude mice has been found to reduce the colonisation of lungs by metastatic embryo cells (Alvarez et al., 1990), while TIMP-2 has been shown to block extracellular matrix degradation by cancer cells (DeClerck et al., 1991). Furthermore, transfection of metastatic rat cells with cDNA for TIMP-2 suppresses the invasive phenotype of these cells (DeClerck et al., 1992). Finally, down-regulation of the expression of TIMP-1 by antisense mechanism confers tumorigenic and metastatic properties on Swiss T3T cells (Khokha et al., 1989). All these findings suggest that either increased levels of certain MMPs or decreased levels of their inhibitors could enhance the metastatic properties of malignant cells.
While MMPs and TIMPs have been extensively investigated in cell lines, relatively few studies have been carried out in human tumours, especially using quantitative assays. The purpose of this investigation was therefore to use an enzyme-linked immunosorbent assay (ELISA) to measure two of these MMPs, i.e. MMP-8 and -9, in a series of human breast tumours and relate these levels to established prognostic markers in breast cancer. We also assayed TIMP-1 by ELISA in the same series of tumours.

Materils and methods
Breast tumours were homogenised in 50 mM Tris-HCl buffer pH 7.4 containing 1 mM monothioglycerol. Homogenates were centrifuged at 2000 g for 10 min. ELISAs for MMP-8, MMP-9 and TIMP-1 were carried out on the supernatants, as previously described in detail (Bergmann et al., 1989;Gunther et al., 1994). The ELISAs for MMP-8 and -9 detect both precursor and active metalloprotease. Furthermore, these ELISAs detect MMP-8 and -9 complexes with TIMP-1 but not complexes of these proteases with M2-macroglobulin. The ELISA for TIMP-1 used polyclonal antibodies and detects TIMP-1 even in the presence of a 10-fold excess of either active or latent MMP-8 and -9. Assays for urokinase plasminogen activator (uPA), tissue-type plasminogen activator (tPA) and plasminogen activator inhibitor type I (PAI-1) were also carried out by ELISA using kits supplied by American Diagnostica (Greenwich, CT, USA). Oestrogen (ER) and progesterone receptors (PR) were assayed as previously described by us (Duffy et al., 1986. Pathological characteristics and steroid receptor status of the cancers used are summarised in Table I , 1992). In addition, TIMP-1 has also been detected by Northern blotting (Urbanski et al., 1992). Unlike the ELISAs described in this investigation, these techniques are difficult to quantitate. MMP-9 has however, been assayed by ELISA in plasma from patients with breast and other malignancies (Zucker et al., 1993). In this study, higher plasma levels of MMP-9 were found in patients with both breast and colorectal cancers when compared with levels in healthy subjects. In contrast, MMP-9 levels in plasma from patients with cancers of the lung, genitourinary tract and lymphomas-leukaemias did not differ significntly from those in healthy controls.
Using ELISA, we show here that levels of MMP-8 and -9 correlate significantly with one another. Since MMP-8 has only been found in leucocytes, this finding suggests that both proteases in breast tumours may be derived, at least in part, from these host cells. In squamous cell carcinomas, mRNA for MMP-9 was recently found only in eosinophils (Stahle-Backdall et al., 1993). However, MMP-9 protein was detected in neutrophils as well as in eosinophils (Stahle-Backdall et al., 1993). In another study using squamous cell carcinomas, mRNA for MMP-9 was found to be expressed by both malignant cells and macrophages (Pyke et al., 1992). In this investigation we show that levels of MMP-8 and -9 are significantly related to levels of TIMP-1. Previous studies using cell ines (Khokha et al., 1989;DeClerck et al., 1991;Aznavooran et al., 1993) have suggested that metastatic potential is related to either increased levels of MMPs or decreased levels of TIMPs. Our data using human breast tumours would not appear to support these findings from cell lines. Previously, we have shown that levels of uPA and its inhibitor PAI-i are also significantly correlated in human breast cancers (Reilly et al., 1992), while others have shown that high levels of PAI-I are significantly related to poor outcome in this disease (Janicke et al., 1993). These findings with PAM-I and TIMP-1 in human breast cancers are difficult to reconcile with the data from cell lines suggesting that certain protease inhibitors are suppressors of metastasis. Moreover, the data from human tumours would appear to suggest that some protease inhibitors (e.g. PAI-1) may be involved in or potentiate the metastatic process. In this study, neither MMP-8, MMP-9 nor TIMP-1 showed any significant relationship with pathological characteristics of the tumour, such as size or metastasis to lymph nodes. Similarly, Brown et al. (1993), using zymography and a smaller number of breast tumours, found no relationship between both MMP-9 and 72 kDA gelatinase and these histological parameters. Other proteases implicated in experimental metastasis such as uPA and cathepsin D (CD) also show no or only weak relationship with these traditional markers of breast cancer prognosis (Duffy et al., , 1991Janicke et al., 1993;Pujol et al., 1993). Despite these findings, in most reports both uPA and CD have been shown to correlate significantly with poor outcome in breast cancer (for review see Rochefort, 1992;Duffy, 1993 prognostic markers in axillary node-negative breast cancer patients (Rochefort, 1992;Duffy, 1993), the group of patients in whom new indicators of outcome are most urgently needed.
In contrast to the lack of correlation between both MMP-8 and -9 and histological features of the tumours, both proteases showed an inverse relationship with ER. Previous reports using both CD and uPA found no significant relationship between either of these proteases and ER (Dufly et al., 1991;Janicke et al., 1993). Since the presence of ER is generally associated with good prognosis in breast cancer, the inverse relationship between both MMP-8 and -9 and ER could suggest that high levels of these metalloproteases will be related to poor patient outcome. So far, relatively few studies have been carried out to evaluate MMPs as prognostic markers in cancer. In one report using node-negative breast cancer patients however, MMP-2 levels were reported to correlate with local recurrences but not with distant metastases (Daidone et al., 1991). The ELISAs described here for MMP-8 and -9 should permit an investigation on the possible prognostic value of these proteases in breast cancer. In a number of model systems, both uPA and a metalloprotease was required for metastasis (Ossowski, 1992;Montgomery et al., 1993). If the same situation applies for human breast cancer, the combined measurement of a MMP and uPA may provide more prognostic information than uPA alone.