Molecular prognostic indicators in breast cancer

EJSO 2002; 28: 467±478 doi:10.1053/ejso.2002.1258, available online at http://www.idealibrary.com on

1

REVIEW

Molecular prognostic indicators in breast cancer C. E. Rogers, R. L. Loveday, P. J. Drew and J. Greenman University of Hull Academic Surgical Unit, Castle Hill Hospital, Castle Road, Cottingham, Hull HU16 5JQ, UK

Here we review a panel of oncogene products, proteases and markers of proliferation that have shown potential as prognostic indicators in primary breast cancer. The relative merits of specific genetic mutations as well as alterations at the protein level are discussed. Finally an assessment is made of the transfer of knowledge from the laboratory to # 2002 Elsevier Science Ltd. All rights reserved. the bed-side. Key words: breast cancer; prognostic markers; oncogene; protease; proliferation.

INTRODUCTION Tumour size, histological grade and axillary nodal status are currently used to stage breast cancer. Axillary nodal status is the most important of these prognostic indicators and decisions regarding adjuvant therapy are largely based on nodal status. However, 15±20% of node-negative patients die within 5 years of diagnosis.1 It has been suggested that, in some cases, early relapse may occur in patients with micrometastatic lymph node disease that has been overlooked. However, studies investigating micrometastatic lymph node disease suggest that, even if occult disease does account for some cases of early relapse, it is not the whole story.2 The search for reliable prognostic factors that can identify high-risk patients is ongoing with the hope that targeted aggressive adjuvant therapy will improve outcome for approximately one-fifth of breast cancer patients. Growth of a tumour depends on the balance between cell proliferation and the rate of apoptosis. Pathologists currently grade tumours by assessment of the tubular arrangement of carcinoma cells, the atypical nuclear appearance and frequency of mitotic or hyperchromatic

Correspondence to: Dr J. Greenman, University of Hull Academic Surgical Unit, Castle Hill Hospital, Castle Road, Cottingham, Hull HU16 5JQ, UK. Tel.: 00 44 1482 623077; Fax: 00 44 1482 623274; E-mail: [email protected] 0748±7983/02/$35.00

nuclear figures, according to the method of Bloom and Richardson3 and modified by Elston and Ellis.4 While this grading provides important prognostic information, it is not sufficiently accurate for predicting early relapse. A number of other markers of proliferation and apoptosis have therefore been investigated for potential prognostic value. Local invasion and metastatic potential are determined, in part, by the ability to degrade the basement membrane and extracellular matrix. Factors involved in these processes have also been identified as molecules worthy of further investigation. Laboratory techniques for the study of potential prognostic markers are rapidly developing at both the gene and the protein level. Most techniques now allow the analysis of fresh or archival tissue. Genetic mutations can be studied by a variety of techniques including loss of heterozygosity, comparative genomic hybridization, fluorescence in situ hybridization (FISH), polymerase chain reaction (PCR), reverse transcriptase PCR (RT-PCR) and DNA sequencing. Protein assays include immunohistochemistry (IHC), Western blotting, flow cytometry and enzyme-linked immunosorbent assay (ELISA). There are advantages and disadvantages to each technique for studying potential prognostic markers at the DNA or protein level. For example, in some studies cDNA sequencing has been demonstrated to yield more informative prognostic information than IHC.5 However, the significance of #

2002 Elsevier Science Ltd. All rights reserved.

468

C. E. ROGERS ET AL.

Table 1 Function of potential breast cancer prognostic markers Class Oncogene products

Marker

p53

Nuclear envelope Nucleus

HER-2/neu

Cell membrane

Cyclin D1

Nucleus

Nm23

Cytoplasm

uPA

Tenascin C

Cytoplasm, stroma Cytoplasm, stroma Stroma

MMP 9

Stroma

Markers of proliferation

Ki-67

Oestrogen receptors

ERb

Proteases

Bcl-2

Location

Cathepsin D

Function

Over-expression

Induction of apoptosis

Reduced tumour proliferation

Genomic stabilizer, facilitates apoptosis Encodes growth factor receptor Regulates progression through G1 Metastasis suppressor gene

Loss of regulation of cell cycle Enhanced tumour cell growth Increased mitotic activity Decreased metastatic potential

Degradation of extracellular matrix Lysosomal enzyme

Increased metastatic potential

Augments effects of fibroblast growth factor Connective tissue remodelling

Increased metastatic potential

Nucleus

Expressed during S, G2 and M part of cell cycle

Increased proliferation

Nucleus

Activates gene transcription

Tamoxifen resistance in tumour cells

Increased metastatic potential

Increased metastatic potential

Figure 1 Cellular location of molecular markers.

a gene mutation cannot be fully addressed without looking at protein levels. IHC can be used to measure levels of protein expression and to give information on the subcellular localization. Test results must be easily reproducible for a technique to be widely applied. This review will focus on a number of the most promising oncogene products, markers of proliferation and proteases that have been investigated as potential prognostic indicators, and discuss their possible incorporation into diagnostic practice (see Table 1 and Fig. 1). We will also review the relative merits of these different

techniques and discuss how they have been combined to give additional prognostic information.

ONCOGENES AND THEIR PRODUCTS Bcl-2 Bcl-2 (B-cell leukaemia/lymphoma) is a proto-oncogene encoding a 25 kDa protein. The gene was first identified at the breakpoint of the t(14;18) translocation, an

MOLECULAR PROGNOSTIC INDICATORS IN BREAST CANCER abnormality commonly found in follicular lymphoma.6 The protein is expressed by a wide range of normal tissues and tumours, and promotes cell survival by blocking programmed cell death (PCD) or apoptosis. Bcl-2 inhibits PCD, suggesting that it may be involved in the loss of cell cycle regulation in tumour growth. A number of studies have shown that Bcl-2 expression is associated with a good prognosis. Knowlton et al. were able to show, by comparing Bcl-2 positive and negative cell lines, that Bcl-2 positive cells showed a decreased S-phase fraction and increased G1/G0 fraction. They proposed that decreased tumour proliferation might explain why Bcl-2 expression is associated with a favourable outcome in breast cancer.7 Alsabeh et al. used IHC to examine Bcl-2 expression in 208 carcinomas of the breast as well as 54 lung carcinomas and 109 gastric carcinomas. They found that 79.3% of breast carcinomas showed moderate or strong Bcl-2 staining in contrast with only 1.9% of lung and 0.9% of gastric carcinomas. In the case of breast carcinoma, Bcl-2 expression was significantly associated with oestrogen and progesterone receptor positivity, low MIB-1 expression and nuclear grades 1 or 2. Bcl-2 expression was not associated with patient age, disease stage or p53 expression and was not associated with disease-free or overall survival. It was suggested that Bcl-2 could be a useful marker in determining the origin of metastatic adenocarcinoma6 but studies on the level of Bcl-2 in a range of other tissues would be required to verify the specificity to breast. Van Slooten et al. looked at Bcl-2 expression in 441 tumours of pre-menopausal, node-negative patients, in the context of a randomized trial comparing perioperative chemotherapy with surgery alone. Patients were followed up for a median of 4 years (range 0.25±7 years). IHC analysis of paraffin embedded sections using the anti-Bcl-2 monoclonal antibody (clone 124) showed that 73.5% of tumours expressed Bcl-2 moderately or strongly, and similar staining was found in areas of ductal carcinoma in situ (DCIS) and invasive cancer within the same tumour. There was marked intratumour heterogeneity. Bcl-2 expression was again associated with low tumour grade and mitotic index, oestrogen and progesterone receptor positivity, in accordance with Alsabeh et al., but was found to be negatively correlated with large tumour size and HER-2 positivity. In univariate analysis, Bcl-2 expression was associated with diseasefree and overall survival but Bcl-2 was not an independent marker of good prognosis.8 Le et al. studied c-myc, p53 and Bcl-2 expression in the tumours of 175 patients; mean follow-up period of 9.5 years. This group also used the anti-Bcl-2 monoclonal 124 antibody and reported that Bcl-2 expression was independently related to a decreased risk of recurrence.9 Teixeira et al. investigated whether oestrogen played a role in the control of Bcl-2 expression. They demonstrated that oestrogen receptor positive MCF-7 cells only expressed Bcl-2 in the presence of oestrogen.

469

Oestrogen depletion doubled the sensitivity of the MCF-7 cells to adriamycin.10 Together these and other many other studies show that Bcl-2 is frequently over-expressed by breast carcinomas and that this expression is associated with favourable tumour characteristics such as oestrogen receptor positivity. Studies using longer follow-up periods, e.g. more than 8 years, suggest that Bcl-2 expression may predict an increased overall survival. In addition, Bcl-2 expression may have a role in determining the origin of metastatic adenocarcinoma where a primary tumour cannot be found. If Bcl-2 can be used in this way it will allow a more focused search for the primary tumour and hence more appropriate treatment.

p53 p53 is a transcriptional regulator, genomic stabilizer, inhibitor of cell cycle progression and facilitator of apoptosis.11 p53 has been christened the `guardian of the genome'.12 Abnormalities of the p53 tumour suppressor gene are the single most common molecular abnormality seen in human cancer.13 The mutant form of p53 leads to a loss of regulation of the cell cycle and has been associated with a worse prognosis in a variety of tumours including breast and bowel. The p53 gene is located on chromosome 17p and encodes for a protein consisting of 393 amino acids that can be divided into five functional domains. Both the p53 gene and protein have been extensively studied. Sjogren et al. assessed p53 activity in 316 primary breast tumours by both IHC and sequencing. The monoclonal antibody Pab 1801 was used for immunostaining; this antibody recognizes both mutant and wildtype p53. Immunostaining and cDNA sequencing revealed 22% and 20% incidence of mutations respectively. However, the groups of positive tumours were not entirely overlapping, with immunostaining showing 33% of cases as false negative and 30% of cases as false positive with respect to cDNA sequencing. When 5 year disease-free and overall survival rates were compared, both immunostaining positive and cDNA sequencing positive patients showed shorter survival, but only cDNA sequencing results were statistically significant. The authors concluded that cDNA sequencing yielded more informative prognostic information than immunohistochemistry.5 Pharoah et al. conducted a metaanalysis of 11 major studies investigating the relationship between p53 mutations and overall survival. They included all published studies looking at p53 mutations using cDNA sequencing and long-term survival and did not include studies using IHC on the basis that this method does not detect the 20% of p53 mutations resulting in protein truncation. The relative hazard (RH) ranged from 1 to 23.4. Three of the studies focused on node-negative patients and the RH ranged from 1.2 to 2.3; another three studies investigated node-positive

470 patients, where the RH ranged from 1.7 to 3.9. Pharaoh et al. were concerned about the possibility of publication bias i.e. non-publication of negative results. They calculated that a study with 1500 cases and a RH of 0.5 would be needed to change the statistically significant RH for overall survival to statistical non-significance. Pharaoh and colleagues concluded that p53 mutation screening should be included in a large trial and analysed with respect to therapy and stage to determine definitively the role of p53 as a prognostic indicator.14 Van Slooten et al. found that p53 accumulation was associated with an absence of Bcl-2 expression, an increased apoptotic and necrotic cell death, high proliferation rate and high tumour grade.15 Harbeck et al. studied a range of potential invasive and proliferative markers, including p53, in 125 node-negative breast tumours with patients followed up for a median of 76 months. Once again, p53 over-expression was associated with a higher level of Ki-67 expression but alone was not a predictor for disease-free or overall survival.16 Berns et al. looked at the relationship between p53 and tamoxifen resistance in patients who had relapsed. Cytosolic extracts from tumours were examined for p53, oestrogen and progesterone receptors and urokinase plasminogen activator (uPA). A cohort of 401 patients, none of whom had received tamoxifen prior to relapse, was followed for a median of 69 months. p53 levels were positively correlated with uPA levels and patients with high levels of p53 tended to be less responsive to tamoxifen therapy and showed reduced survival. In patients who expressed low levels of oestrogen and progesterone receptors, p53 was related to poor tamoxifen response.17 This raises the question as to whether p53 is just an indicator of poor prognosis or whether it is in some way involved with regulation of oestrogen receptors. There has been much interest in p53 as a potential target for novel therapies. The human adenovirus E1B gene encodes a protein that inactivates p53. A mutant form of the virus that does not express this protein can replicate in and lyse p53 deficient human tumour cells but not cells with functional p53. The investigators injected the mutant adenovirus into p53 deficient human cervical carcinomas in mice and caused a significant reduction in tumour size.18 This study suggests that mutant viruses could be used in the treatment of human tumours. The papers described here suggest that p53 is worthy of a large trial to assess accurately its role as a potential prognostic indicator. Authors comparing IHC with cDNA sequencing showed the latter to be more accurate and therefore the method of choice.

HER-2/neu HER-2/neu (also known as C-erbB-2) is an oncogene located on chromosome 17q. It encodes a transmembrane tyrosine kinase growth factor receptor, which is amplified in some breast tumours and has

C. E. ROGERS ET AL. been related to poor prognosis. HER-2/neu has assumed greater importance recently with FDA approval for a humanized anti-HER-2 antibody, rhuMAb HER-2 (Herceptin1), which has been shown to improve significantly the response to cytotoxic chemotherapy in 20% of HER-2/neu over-expressing tumours. The original study on HER-2 by Slamon et al. used frozen tissue and assessed gene amplification in 189 tumours by Southern blotting. Thirty per cent of tumours showed amplification of HER-2, which was correlated with overall survival and time to relapse. In node-positive patients, HER-2 amplification had greater prognostic significance than hormone receptor status.19 Paterson et al. conducted a retrospective case control study using cases identified by the Northern Alberta Breast Cancer Registry as having node-negative breast cancer. In this, 115 patients who had relapsed at between 5 and 16 years post-treatment were matched with 115 patients at an equivalent time from primary treatment with no disease. In this study, DNA was extracted from formalin-fixed tumour tissue and analysed for HER-2 copy number by slot-blot hybridization; amplification of HER-2 was a statistically significant predictor of disease-free survival.20 Charpin et al. used IHC on 148 frozen tumours with a follow-up period of 62.5 months. HER-2 over-expression was significantly related to overall survival in node-positive and nodenegative patients. In node-positive patients, HER-2 overexpression was also related to disease-free survival.21 Dittadi et al. used ELISA to measure HER-2 protein in cytosols of 115 breast cancers. They divided the data into quartiles according to HER-2 concentration. These quartiles were as follows: I (low) , 2150 U/mg protein, II and III (intermediate) 2150±5000 U/mg protein and IV (high) . 5000 U/mg protein. They showed a significantly longer disease-free survival in patients with an `intermediate' concentration of HER-2 protein. From the results of multivariate analysis, patients were grouped into `low-risk' (protein concentration 2150±30 000 U/mg) and `high-risk' (protein concentration , 2150 or >30 000 U/mg). The high-risk group correlated well with HER-2 positivity on IHC. Interestingly, there were significantly more oestrogen receptor positive cases in the intermediate group. The investigators expressed surprise at finding that very low as well as very high concentrations of the HER-2 protein were associated with poor prognosis. They suggested that progression to a more aggressive phenotype could imply selection against HER-2 over-expression or down-regulation owing to a high ligand activity.22 FISH was used by Press et al. to compare gene expression and gene amplification assays on 140 archival breast tumours. FISH compared favourably with the other methods of assay and disease-free and overall survival were significantly associated with HER-2 amplification.23 RhuMAb HER-2 (Herceptin1) is a humanized antiHER-2 monoclonal antibody expressed by a genetically

MOLECULAR PROGNOSTIC INDICATORS IN BREAST CANCER engineered Chinese hamster ovary cell line. The majority of studies so far have been on cell lines but clinical studies have been encouraging. Cobleigh et al. used Herceptin1 alone in 222 patients with HER-2 positive metastatic breast cancer, all of whom had previously received chemotherapy; the response rate was 15%.24 One multinational, randomized controlled phase III trial has been reported recently; 469 patients with HER-2 overexpressing metastatic breast cancers were randomized to receive anthracycline and cyclophosphamide alone or with Herceptin1. Patients who had previously received anthracycline-based regime were randomized to receive paclitaxel alone or with Herceptin1. At a median follow-up time of 29 months, overall survival was significantly better in both of the groups receiving Herceptin1 (25.4 months vs 20.3 months).25 HER-2 is promising as a predictor of poor prognosis. If further trials of Herceptin1 confirm its beneficial effects in treating HER-2 positive patients, HER-2 testing will become a routine part of the histological assessment.

Cyclin D1 Cyclin D1 is a gene whose product is involved in regulating progression through G1 of the cell cycle. The gene located at chromosome 11q13 is overexpressed in around 50% of breast cancers. Cyclin D1 is not over-expressed in benign breast disease but is frequently over-expressed in early forms of tumour including DCIS.26 Over-expression has been studied at both the gene and protein level. Gillet et al. found that DNA-based assays underestimate cyclin D1 over-expression as IHC identified additional tumours in a study utilizing both techniques. The IHC identified over-expression was independent of an increase in gene copy number, suggesting a number of different mechanisms for regulating expression.27 Alle et al. looked at cyclin D1 expression in normal breast, proliferative disease, DCIS and invasive carcinoma using IHC. They examined 471 samples and found that the percentage of positive cases in each subgroup increased from normal breast to invasive carcinoma. Cyclin D1 expression in DCIS was significantly higher than in proliferative disease, although expression in invasive carcinoma was not significantly higher than in DCIS.28 Van Diest et al. investigated 148 invasive breast cancers by IHC. Cyclin D1 was over-expressed in 59% of cases and showed a strong correlation with histological type. Strong staining was seen in half of ductal carcinomas and most carcinomas of special type, except medullary carcinomas in which no staining was seen. Staining was positively correlated with mitotic activity and oestrogen receptor status. In survival analyses, cyclin D1 was not a significant prognostic factor.29 Kenny et al. investigated cyclin D1 mRNA overexpression in 253 patients with a median follow-up of

471

75 months. Cyclin D1 over-expression was not associated with tumour size, grade, axillary nodal status, lymphovascular invasion, age or menopausal status. However, in the 182 oestrogen receptor positive patients, overexpression of cyclin D1 was associated with an increased risk of relapse, local recurrence, metastasis and death.30 Thus cyclin D1 may have a role in predicting patients with oestrogen receptor positive disease that have a worse prognosis. However, further work is needed to understand the regulation of the cyclin D1 gene and to establish whether it is an independent prognostic indicator.

Nm23 The nm23 `non-metastatic' gene is a family of at least two genes mapping to 17q21.3. Nm23-H1 and nm23-H2 have so far been identified. They encode 17 kDa proteins that are 90% similar and have non-specific nucleoside diphosphate kinase activity.31 Nm23 was described in 1988 by Steeg et al.32 In 1989 they described an antibody to nm23 recognizing both murine and human proteins, anti-nm23 peptide 11.33 A number of studies suggest that nm23-H1 may be a metastasis suppressor gene in human breast cancer. The most important of these studies will be considered below. In their initial study, Steeg and coworkers looked at nm23 RNA levels in the murine K-1735 melanoma cell line and also at N-nitroso-N-methylurea-induced rat mammary carcinomas. In both cases nm23 RNA levels were highest in tumours with a lower metastatic potential, suggesting that loss of activity of a gene may be associated with metastasis.33 A number of similar studies confirmed these findings. The first study looking at nm23 RNA levels in human breast cancer was published in 1989 by Bevilacqua et al. They investigated 27 ductal carcinomas by Northern blotting and in situ hybridization and reported that tumours which had metastasized to regional lymph nodes had low levels of nm23 mRNA. In the tumours that had not metastasized, 75% had high levels of nm23 mRNA. The remaining 25% had normal levels of nm23 but carried other markers of high metastatic potential such as high nuclear grade.34 Other studies have confirmed these findings including one by Royds et al. where a cohort of screen-detected lesions was examined, a population which has become more common over the past decade. Of the 128 lesions studied by IHC staining with the anti-nm23 peptide 11 antibody, 35 were benign, 26 had DCIS and 67 had invasive carcinoma. All the benign tissue showed uniform staining for nm23. The seven cases of comedo DCIS were negative for nm23 but all other DCIS cases were positive. Nm23 negativity was associated with worsening grade and lymph node stage in the invasive cancers.35 Follow-up studies of nm23 have proved less encouraging. Charpin et al. used 168 tumours from patients

472 operated on in 1986 who did not receive any adjuvant therapy. Patients were followed for 125 months or until death. Charpin and colleagues used monoclonal mouse anti-human nm23-H1, which is specific for nm23-H1, for an immunohistochemical study. Computer-assisted image analysis was used to measure the percentage of positive surface area. Nm23 positivity correlated with longer metastasis-free survival in node-negative and node-positive patients. Nm23 positivity showed borderline significance for disease-free survival only in nodepositive patients. There was no relationship between nm23 positivity, recurrence-free survival and overall survival. This group suggested that the differences in results from studies of nm23 maybe due to lack of standardization of methods.36 Nakopoulou et al. looked at nm23 expression in conjunction with steroid receptor status and HER-2. They examined 191 tumours using IHC and found that loss of nm23-H1 alone did not predict the presence or number of involved lymph nodes. Interestingly, however, nm23-H1 positive cells were likely to be progesterone receptor positive. The nm23-H1 negative/HER-2 positive phenotype was significantly associated with the number of involved lymph nodes but so were HER-2 positive phenotypes alone. In a follow-up study (range 20±36 months, median 28 months), nm23-H1 status could not be linked to outcome in any way. The authors suggested that the positive associations between nm23-H1 and clinical outcome may be due to a `bystander effect'.31 Heimann et al. used IHC to investigate nm23-H1 and E-cadherin staining with microvessel count (MVC) in a study of 168 node-negative patients treated with mastectomy. None of the patients received adjuvant chemotherapy or hormonal therapy and the median follow-up was 14 years (range 3±36 years). High levels of E-cadherin, high levels of nm23-H1 and low MVC were significantly associated with longer disease-free survival at 14 years. High MVC and low nm23-H1 did not accurately predict poor outcome at 14 years. However, when patients were grouped by E-cadherin level into high, medium or low groups, there was a significant relationship with disease-free survival. The 14 year disease-free survival was 76% for the high E-cadherin group, 61% for the intermediate group and 44% for the low group. Thus, in this study, nm23-H1 in combination with other factors helped to identify patients of good prognosis but E-cadherin alone was better at identifying patients of poor prognosis.37 In summary, although nm23 initially looked to be promising as a prognostic indicator, with high levels of expression correlating with good prognosis, the followup studies are not convincing. It may be that positive associations between nm23 and good prognosis are merely due to a `bystander effect'.31 Again, further highquality standardized studies with large numbers of patients and long-term follow-up are needed to answer these questions.

C. E. ROGERS ET AL.

PROTEASES A variety of proteases believed to be involved in metastasis have been investigated, and several have shown promise as prognostic indicators; the best studied of these are reviewed.

Urokinase plasminogen activator Plasminogen activator exists in two forms: tissue plasminogen activator and uPA. uPA catalyses the conversion of plasminogen to plasmin, and plasmin subsequently catalyses the degradation of a variety of substrates in the extracellular matrix.38 uPA has been extensively investigated by many groups and shown to be elevated in some breast tumours. In 1988, Duffy et al. reported that high levels of uPA measured by ELISA in cytosolic extracts from tumours were related to a reduced disease-free survival period. The uPA activity was calculated as the difference between total plasminogen activator (PA) activity and the activity measured in the presence of tissue PA antibodies. Patients were placed in `high' or `low' uPA groups using a cut-off point for uPA activity of 0.1 IU/mg of protein. Fifty-two patients were included with a mean follow-up of 17 months for the low uPA group and 19 months for the high uPA group. uPA activity increased with stage and increasing number of positive axillary nodes.39 In a later study by the same group, using the same methods and cut-off points, 149 patients with a median follow-up of 68 months for the low uPA group and 58 months for the high uPA group were studied. The data were analysed with respect to nodal status. In the node-positive group, high uPA levels were a significant predictor of both disease-free interval and survival whereas, in the nodenegative group, high uPA levels were only associated with disease-free interval. Also, uPA was a significant predictor of disease-free interval and survival in oestrogen receptor positive patients in both nodal groups.38 A third study by the same group published in 1998, studying 184 tumours, confirmed these results and showed that uPA was an independent prognostic indicator. In the node-negative group of patients, uPA was the only significant predictor of outcome in both univariate and multivariate analyses.40 Grùndahl-Hansen et al. measured uPA and PA inhibitor (PAI-1) levels in breast tumour cytosols of 118 premenopausal and 92 post-menopausal patients with a median follow up of 8.5 years. This group confirmed the findings of Duffy et al., with a high level of uPA being associated with shorter overall survival in pre- and postmenopausal women.41 Similar results have been found by other investigators in large, well-controlled studies.42,43 Two trials have now commenced: one in Germany and one pan-European, in which node-negative patients with a high level of uPA are randomized to receive six cycles of cyclophosphamide or to receive no chemotherapy.

MOLECULAR PROGNOSTIC INDICATORS IN BREAST CANCER Some studies have compared a number of potential prognostic indicators in the same group of patients. Harbeck et al. compared uPA, PAI-1, cathepsin D, Sphase fraction (SPF), Ki-67, p53 and HER-2/neu. They studied 125 node-negative patients with a median follow-up of 76 months. Cathepsin D was assessed using radiometric immunoassay on tumour cytosol extracts, PAI-1 and uPA levels were assessed in detergent extracts of the tumours by ELISA, SPF was assessed using flow cytometry and the other antigens were assessed using IHC on formalin-fixed tumour sections. uPA, PAI-1, cathepsin D and SPF were all identified as significant predictors of disease-free survival on univariate analysis, although only PAI-1 was significant on multivariate analysis. For overall survival, PAI-1, cathepsin D, tumour size and ploidy were significant prognostic factors but on multivariate analysis again only PAI-1 retained significance. In patients with low PAI-1, uPA was a strong prognostic indicator.16 All the studies discussed so far have assessed uPA by ELISA. However, Jahkola et al. used IHC to assess uPA, PAI-1, cathepsin D and tenascin C in 159 tumours (158 patients) with a median follow-up of 7.8 years. Positive cathepsin D staining was observed in the cytoplasm of carcinoma cells, stromal macrophages and fibroblasts. Tumours with more than 10% of either stromal or carcinoma cells showing a strong positive reaction were scored as positive. uPA and PAI-1 staining was seen in the majority of cells and was scored as negative, slight, moderate or strong. Staining was observed in the cytoplasm of carcinoma cells, the surrounding stroma and in stromal fibroblasts. Forty-seven per cent of the tumours showed cytoplasmic staining for cathepsin D and 44% of tumours showed stromal staining, with 28% of tumours showing staining of both carcinoma cell cytoplasm and stromal cells. Stromal uPA staining was seen in all tumours, being weak in 54%, moderate in 39% and strong in 7%. Cytoplasmic uPA was absent in 6%, weak in 31%, moderate in 48% and strong in 15%. High levels of stromal cell cathepsin D were associated with tumour cytoplasmic cathepsin D and fibroblastic uPA and PAI-1. Cytoplasmic cathepsin D and stromal cathepsin D were associated with proliferation, measured by Ki-67 staining, and histological grade. Fibroblastic uPA was associated with high SPF and aneuploidy. In survival analyses, cytoplasmic cathepsin D, stromal cathepsin D and fibroblastic PAI-1 were associated with metastasis in univariate analysis. Stromal uPA was associated with local recurrence. In multivariate analysis, these factors remained significant predictors for metastasis and local recurrence respectively.44 An important question that has not been fully answered yet is whether soluble uPA is equally useful as a prognostic indicator. ELISA requires fresh frozen tissue which cannot always be taken from patients with screen detected lesions. The first study examining plasma uPA in breast cancer patients was conducted

473

by Grùndahl-Hansen et al. in 1988. They demonstrated a significant elevation in plasma uPA in breast cancer patients.45 Follow-up studies are awaited eagerly. The studies described show that uPA, when assessed using ELISA, can be a predictor of overall survival in node-negative patients. This, of course, requires validation in randomized trials that are currently underway.

Cathepsin D Cathepsin D is a lysosomal enzyme synthesized by normal tissues and over-expressed by some breast tumours; it may be involved in invasion and metastasis. Some studies indicate that over-expression is associated with early relapse and reduced overall survival. Measurement of cathepsin D has been combined with assessment of other potential prognostic indicators in several studies and some of these have already been discussed.16,44 Most studies have used radioimmunoassay on tumour cytosols, though Jahkola et al. used immunohistochemistry on formalin fixed tissue44 and Tandon et al. used Western blotting.46 Spyratos et al. measured the concentration of cathepsin D in tumour cytosols by radioimmunoassay from 122 patients who were followed for a median of 4.6 years. High cathepsin D levels were associated with reduced disease-free and metastasis-free survival on univariate analysis. On multivariate analysis, high cathepsin D levels were associated with reduced metastasis-free and disease-free survival, using 45 pmol/mg protein as the cut-off. In the 68 node-negative patients, high cathepsin D levels were also associated with reduced disease-free and metastasis-free survival, using the same cut-off point.47 Duffy et al., using a similar method to assess cathepsin D levels in 169 primary breast tumours, with a cut-off point of 40 pmol/mg, identified a significant relationship between high cathepsin D levels and overall survival but the relationship with disease-free survival was not significant. They also found that high levels of cathepsin D correlated with a tumour grade of 3 and with high concentrations of uPA.48 Tandon et al. investigated 199 node-negative and 198 node-positive patients. They used Western blotting and densitometry to measure cathepsin D levels in extracts from fresh or snap-frozen tumours. Patients were followed for a median of 64 months. In univariate analysis, high cathepsin D levels (more than 75 units using a semiquantitative scale) were a significant predictor of disease-free and overall survival in nodenegative patients. Combining cathepsin D with ploidy, steroid hormone receptor status, patient age and tumour size in multivariate analysis of the node-negative group identified cathepsin D as the most powerful predictor of disease-free survival and the most important predictor of overall survival. A node-negative patient with a high level of cathepsin D had a relative risk of death of 3.9.46

474 Foekens et al. conducted a large-scale study of cathepsin D in 2810 tumour cytosols. The patients were followed for a median of 88 months. In univariate analysis, high cathepsin D levels (more than 45.2 pmol/ mg) were, once again, shown to be associated with reduced disease-free and overall survival. This association was maintained in multivariate analysis.49 Cathepsin D is another protease showing great promise as a prognostic indicator particularly as a result of the large study by Foekens et al. on a large cohort of tumours.

Tenascin C Tenascin C is an extracellular matrix glycoprotein, which is believed to promote cell growth by augmenting the mitogenic affects of fibroblast growth factor. In DCIS, it is expressed periductally, next to the basement membrane; in invasive carcinoma it is expressed in the stroma.50 In a study by Chiquet-Ehrismann et al., cells from the breast cell line MCF 7 lost their cell to cell contacts in the presence of tenascin C.51 Gould et al. investigated tenascin expression in normal, benign and malignant breast tissue by IHC. Stromal expression was reported in all three types of tissue with the last of these having the strongest staining. However, they concluded that tenascin could not be regarded as a tumour marker because of its presence in normal tissue.52 In contrast, Jahkola et al. have reported that the pattern of tenascin C staining can be used prognostically. First they investigated tenascin C expression in 137 small, node-negative tumours by IHC. Tenascin staining in the stroma did not predict metastasis but tenascin C staining of the invasion border could predict the formation of distant metastasis. Five year metastasisfree survival was 98% for tenascin-negative patients but only 85% for patients with tenascin expression in the invasion border of the tumour.53 In a follow-up study a further 238 node-negative patients were added and additional factors, HER-2 p53, Ki-67, ploidy and SPF, were also studied. In univariate analysis, tumour size greater than 13 mm, Ki-67 positivity and tenascin C expression in the invasion border were significant predictors of metastasis. In multivariate analysis, Ki-67 was the only independent predictor of metastasis.50 In summary, tenascin C is not as strong a predictor of metastasis as uPA and cathepsin D. The majority of work on this factor has been carried out by a single group and confirmation by other workers is needed.

Matrix metalloproteinase 9 The matrix metalloproteinases (MMPs) are a group of enzymes divided into three main types: the interstitial collagenases, the type IV collagenases (also known as gelatinases) and the stromolysins. In normal tissues, the metalloproteinases are responsible for connective tissue

C. E. ROGERS ET AL. remodelling. In tumours, they are thought to be responsible for connective tissue breakdown required for invasion and metastasis. MMP 9 is a 92 kDa type IV collagenase.54 Garbett et al. examined the expression of several metalloproteinases and metalloproteinase inhibitors in paired normal and tumour tissues from 43 breast and 24 colorectal patients. They found that total MMP activity was greater in tumour samples than in normal tissues, and was greater in colorectal tumours than in breast.55 Nielsen et al. studied the location of MMP 9 staining in 22 breast tumours using IHC. They found staining in neutrophils, macrophages and vascular pericytes but not in any cancer cells. The group concluded that, in breast cancer, MMP 9 is located in tumour-infiltrating stromal cells and may play a role in extracellular matrix degradation during tumour angiogenesis.56 Rha et al. investigated 28 normal, 12 benign and 126 malignant samples of breast tissue for pro-MMP 9 and MMP 9 using gelatin zymography. Pro-MMP 9 was expressed in 17.5% of malignant tumours compared with 2.5% in non-malignant breast. MMP 9 was only expressed in T2±4 tumours (90 tumours). In DCIS (14 tumours) and T1 tumours (22 tumours), only pro-MMP 9 was expressed.57 A study looking for MMP 9 in sections of primary breast tumours and brain metastases showed strong staining in both primary and secondary tumours. There was no difference in staining intensity between the tumours and therefore MMP 9 could not be used as a differentiation marker.58 While MMP 9 is expressed more strongly in larger tumours, it does not appear from published reports to be a likely candidate as a tumour marker of metastasis or early relapse; although a growing family of MMPs exist and others may prove to be of interest.

S-PHASE FRACTION AND Ki-67 DNA synthesis occurs during the S phase of the cell cycle. Therefore, the proportion of cells in S phase is an indicator of the proliferative activity of the tumour. SPF is measurable by flow cytometry or quantitative IHC. Both methods permit the analysis of formalin-fixed, paraffinembedded material.59 The Ki-67 antigen is a non-histone nuclear protein that is expressed throughout the S and G2/M phases of the cell cycle, the strongest staining occurring during the latter stages. Ki-67 is not expressed during G0, the resting phase. This molecule has become the best-studied marker of tumour cell proliferation. The anti-Ki-67 monoclonal antibody was raised by Gerdes et al. in 1983.60 They published a study in 1986 using the antibody to determine growth fractions in breast cancer.61 One limitation of this particular antibody was that it could only be used on frozen tissue. The development of a recombinant form of the Ki-67 antigen, MIB1, antibodies against which work on formalin-fixed, paraffin-embedded tissue, overcame this problem.62

MOLECULAR PROGNOSTIC INDICATORS IN BREAST CANCER Mitotic count is the easiest and simplest way to assess proliferative rate and correlates strongly with prognosis.63 In 1990, Isola et al. published results of a comparison of Ki-67 score, SPF and mitotic count, performed on frozen sections of tissue. In this study the investigators scored as follows: occasional mitotic figures (‡), two to three figures (‡‡) and four or more figures (‡‡‡). They studied 102 primary breast tumours and showed that the median Ki-67 score was significantly higher in aneuploid tumours. SPF also correlated with Ki-67 score in aneuploid tumours but not diploid tumours. The mitotic count was significantly related to Ki-67 score with the score being highest in those tumours that had the highest mitotic rate. Histological grade was significantly related to aneuploidy, SPF and Ki-67 score. This group concluded that Ki-67 analysis was relatively easy and informative and would be applicable for the routine pathology laboratory.64 The advent of MIB1 further supported these conclusions as staining could then be done on paraffin-embedded tissue rather than fresh-frozen tissue with the far superior morphology of the former fixation technique. Veilh et al. compared Ki-67 and SPF with clinical and pathological features in 148 cases. They also found a relationship between the proliferative indices, nuclear grade and mitotic index. In agreement with Isola et al.'s study, SPF only correlated with Ki-67 score in aneuploid tumours but in addition they reported a relationship between Ki-67 and axillary nodal status.65 Sahin et al. looked at Ki-67 staining in 42 node-negative tumours. Ki-67 score correlated with patient age and nuclear grade. On univariate analysis, both 5 year and overall survivals were strongly associated with Ki-67 score.66 In a much bigger study, Brown et al. correlated Ki-67 scores and S phase fraction with long term follow-up data in 674 node-negative patients. All patients were followed for 5 years. A high Ki-67 score was associated with a 1.8-fold increased risk of recurrence. In this analysis Ki-67 score was independent of tumour size and did not predict overall survival.67 Jansen et al. included node-positive patients in their study of 341 breast cancers. They looked at MIB1 score and SPF. High MIB1 was associated with aneuploidy, oestrogen receptor negativity, progesterone receptor negativity, axillary lymph node metastases and larger tumour size. In univariate analysis and multivariate analysis, MIB1 score predicted disease-free survival. MIB1 showed only borderline significance for overall survival. In univariate analysis, SPF predicted overall survival but the effect was lost on multivariate analysis.68 Keshgegian and Cnaan, acknowledging the importance of proliferative rate as a prognostic marker in breast cancer, attempted to establish the best method of measuring it. They looked at mitotic figure count (MFC), SPF, Ki-67, MIB1 and proliferating cell nuclear antigen (PCNA) by IHC in 135 breast cancers. Ki-67 only correlated with SPF whereas MIB1 correlated with MFC,

475

SPF, PCNA positivity and high nuclear grade. Only MFC and MIB1 were associated with disease-free survival.69 In most studies, SPF correlates well with Ki-67/MIB1 expression in aneuploid tumours.64,65 Likewise, MFC correlated with Ki-67/MIB1 expression64 and Ki-67/MIB1 correlated with disease-free survival in several of these studies.66±68 In addition, some studies have found that Ki67/MIB1 was associated also with overall survival.66,68 However, some inconsistencies exist. In Keshgegian and Cnaan's study, Ki-67 did not correlate with MFC, lymph node status or disease-free survival but MIB1 did correlate with disease-free survival. Keshgegian and Cnaan postulated several factors that could explain the differences; firstly, that freezing tissue may have introduced greater variation than formalin fixation, that there may be tumour heterogeneity and the level of intra- and interobserver variation.69 Dettmar et al. point out that SPF only measures cells in a small part of the cell cycle, whereas Ki-67/MIB1 staining looks at cells in all but G0.70 Measurement of cell proliferation, as determined by SPF or Ki-67 reactivity, is simple and in most studies predictive of poor prognosis. The rationale that increased proliferation is a poor prognostic factor is easily understandable but it is clear from a number of studies that cell cycle cannot be considered in isolation; an array of other tumour and host factors will be important in many instances. The questions that need to be addressed most urgently are the standardization on an optimal method of determination of one, or both, of these markers so that a number of definitive multicentre trials can be conducted. For widespread applicability the use of formalin-fixed tissue would be best. The final challenge, that is applicable to all prognostic markers, is to identify the group of tumours for which a given marker is of particular benefit.

OESTROGEN RECEPTORS  AND The oestrogen receptor was initially thought to exist in only one form, now known as oestrogen receptor a (ERa), but a second isoform (ERb) was described in 1996.71 The prognostic significance of ERa in breast cancer has stimulated similar interest in the role of ERb in breast cancer. The ERa gene is located on chromosome 6q whereas the gene for ERb is located on 14q. The oestrogen receptor binds oestrogen and then migrates to the nucleus where the complex activates gene transcription. Gene transcription is mediated via either a classical oestrogen response element or an activator protein (AP) 1 enhancer element. When signalling is mediated via AP1, oestradiol activates transcription when binding ERa but blocks transcription when binding ERb.72 Several studies have now been published looking at the role of ERb. Taylor and Al-Azzawi investigated which tissues expressed ERa and ERb using PCR and immunohistochemistry. In all cases, the oestrogen receptor

476

C. E. ROGERS ET AL.

localized to the nucleus. ERa was found in a variety of tissues including myocardium, arteries, pituitary, adrenal, ovary, bladder, uterus, cervix and breast. ERb was also found in these tissues and in addition in follicular and C-cells of the thyroid and in the prostate. Within tissues there were differences in the cell types that expressed the two receptors.73 Speirs et al. investigated the expression of both forms of the oestrogen receptor in 23 samples of normal tissue and 60 breast carcinomas using RT-PCR. ERb expression predominated in normal breast tissue whereas ERa was expressed in most tumours either alone or in combination with ERb. Nineteen out of 30 tumours expressing both receptor subtypes were node positive and there was an association with a histological grade of II. The authors suggested that the failure of some tumours to respond to anti-oestrogens may be due to agonist action when signalling via AP1 in tumours overexpressing the inhibitory signalling ERb.74 Further work by this group showed that ERb mRNA was significantly up-regulated in tamoxifen-resistant tumours when compared with tamoxifen-sensitive tumours.75 JaÈrvinen et al. investigated the association between ERa and ERb and the progesterone receptor (PR) in 92 breast cancers. They used IHC on frozen tumour sections to assess PR, ERa and ERb expression together with in situ hybridization to measure ERb mRNA. ERb expression was found in 75% of the ERa-positive, PR-positive tumours and was associated with node negativity, low grade, low SPF and pre-menopausal status.76 These results clearly are in marked contrast to those of Speirs et al. and are difficult to reconcile. The most likely explanation is the use of RT-PCR as opposed to in situ hybridization and IHC. Whether or not ERb has potential as a prognostic indicator is an area requiring more exploration. The idea of tailoring endocrine therapy more precisely to avoid the problems of resistant tumours remains an attractive option as our understanding of the basic biology increases.

a number of potential prognostic markers including uPA, cathepsin D, p53, Ki-67 and HER-2/neu have been extensively investigated (Table 2). In our opinion, a major reason for the lack of well-accepted biological markers, after almost 50 years of research, is not that such factors do not exist, but the manner in which the data are gathered. The vast majority of studies are small scale, i.e. often ,100 subjects, and are done using disparate methodologies and analyses, which instead of clarifying the situation makes it less clear. One message of this review is that standardization of techniques and grading systems is still urgently required to provide a framework for future multicentre trials. If phase III trials of Herceptin1 prove it efficacious in HER-2 positive patients, routine testing for HER-2 over-expression will need to be introduced. A trial has started in Germany to assess the impact of chemotherapy on node-negative patients with high levels of uPA.40 The other prognostic markers must also be assessed in large, randomized, multicentre trials before being included as standard prognostic indices. Some of the markers reviewed here such as MMP 9, cyclin D1, tenascin C and ERb do not look promising from the literature. McGuire and Clark,77 back in 1992, suggested some useful guidelines for evaluating new prognostic markers, which, we believe, need to be adopted as widely as possible. They recommend that each marker should Table 2 Status of prognostic markers in breast cancer Class

Marker

Oncogene products

Bcl 2 p53 HER-2/neu Cyclin D1 Nm23

No Undecided Yes Undecided No

Proteases

uPA Cathepsin D Tenascin C MMP 9

Yes Yes More studies required No

Markers of proliferation

Ki-67

Yes

Oestrogen receptors

ERb

More studies required

DISCUSSION The ideal prognostic marker is one that clearly delineates a particular prognostic group, is 100% specific, highly sensitive, inexpensive and easy to perform on a small quantity of fresh or fixed tissue. No such marker exists but

Potential as a prognostic indicator

Table 3 Clinical role of prognostic markers in breast cancer Class Oncogene products Proteases Markers of proliferation

Marker

Role

HER-2/neu

Identification of patients with poor prognosis ± likely to respond to Herceptin treatment

uPA Cathepsin D Ki-67

Identification of poor prognosis node-negative tumours ± chemotherapy can be offered Identification of poor prognosis node-negative tumours ± chemotherapy can be offered Identification of tumours with high metastatic potential ± early aggressive therapy

MOLECULAR PROGNOSTIC INDICATORS IN BREAST CANCER have a biological hypothesis, that a distinction is made between a pilot study and a definitive study and that sample size should be adequate to ensure a definitive study can be conducted. In addition, they recommend that the patient population should be appropriate and sampling bias considered, that the study should include methodological validation and optimal cut-off values in a training data set, with confirmation in a validation data set, that the relative importance of a new marker should be assessed in multivariate analyses and results externally validated by a definitive study using a different population.77 By following these guidelines, new prognostic markers can be tested in the most efficient way; avoiding, it is hoped, a number of apparently disparate claims based on a limited data set. If the study proves successful the marker can be adopted for routine use either alone or, more probably, in combination with standard clinical assessment. We believe that a number of molecular markers will make the transition from the laboratory to the clinic over the coming decades with the ultimate benefit being better prognostication and therapy of breast cancer patients (see Tables 2 and 3).

REFERENCES 1. Fisher ER, Swamidoss S, Lee CH, Rockette H, Redmond C, Fisher B. Detection and significance of occult axillary node metastases in patients with invasive breast cancer. Cancer 1978; 42: 2025±31. 2. International (Ludwig) Breast Cancer Study Group. Prognostic importance of occult axillary lymph node micrometastases from breast cancer. Lancet 1990; 335: 1565±8. 3. Bloom HJG, Richardson WW. Histological grading and prognosis in breast cancer: a study of 1409 cases of which 359 have been followed for 15 years. Br J Cancer 1957; 11: 359±77. 4. Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long term follow up. Histopathology 1991; 19: 403±10. 5. Sjogren S, Inganas M, Norberg T et al. The p53 gene in breast cancer: prognostic value of complementary DNA sequencing versus immunohistochemistry. J Natl Cancer Inst 1996; 88: 173±82. 6. Alsabeh R, Wilson CS, Ahn CW, Vasef MA, Battifora H. Expression of Bcl-2 by breast cancer: a possible diagnostic application. Mod Pathol 1996; 9(4): 439±44. 7. Knowlton K, Mancini M, Creason S, Morales C, Hockenbery D, Anderson BO. Bcl-2 slows in vitro breast cancer growth despite its antiapoptotic effect. J Surg Res 1998; 76: 22±6. 8. van Slooten HJ, Clahsen PC, van Dierendonck JH et al. Expression of Bcl-2 in node-negative breast cancer is associated with various prognostic factors, but does not predict response to one course of per-operative chemotherapy. Br J Cancer 1996; 74(1): 78±85. 9. Le MG, Mathieu M-C, Douc-Rasy S et al. C-myc, p53 and Bcl-2, apoptosis-related genes in infiltrating breast carcinomas: evidence of a link between Bcl-2 protein over-expression and a lower risk of metastasis and death in operable patients. Int J Cancer 1999; 84: 562±7. 10. Teixeira C, Reed JC, Pratt MAC. Oestrogen promotes chemotherapeutic drug resistance by a mechanism involving Bcl2 proto-oncogene expression in human breast cancer cells. Cancer Res 1995; 55(17): 3902±7. 11. Elledge RM, Allred DC. Prognostic and predictive value of p53 and p21 in breast cancer. Breast Cancer Res Treat 1998; 52: 79±98. 12. Lane DP. Cancer. p53, the guardian of the genome. Nature 1992; 358: 15±16.

477

13. Bray SE, Schorl C, Hall PA. The challenge of p53: linking biochemistry, biology and patient management. Stem Cells 1998; 16: 248±60. 14. Pharoah PDP, Day NE, Caldas C. Somatic mutations in the p53 gene and prognosis in breast cancer: a meta-analysis. Br J Cancer 1999; 80(12): 1968±73. 15. van Slooten HJ, van de Vijver MJ, van de Velde CJK, van Dierendonck JH. Loss of Bcl-2 in invasive breast cancer is associated with high rates of cell death, but also with increased proliferative activity. Br J Cancer 1998; 77(5): 789±96. 16. Harbeck N, Dettmar P, Thomssen C et al. Risk group discrimination in node-negative breast cancer using invasion and proliferation markers: 6-year median follow-up. Br J Cancer 1999; 80: 419±26. 17. Berns EMJJ, Klijn JGM, van Putten WLJ et al. P53 protein accumulation predicts poor response to tamoxifen therapy of patients with recurrent breast cancer. J Clin Oncol 1998; 16(1): 121±7. 18. Bischoff JR, Kirn DH, Williams A et al. An adenovirus mutant that replicates electively in p53-deficient human tumour cells. Science 1996; 274: 373±6. 19. 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 1987; 235: 177±82. 20. Paterson MC, Deitrich KD, Danyluk J et al. Correlation between C-erb-B2 amplification and risk of recurrent disease, in nodenegative breast cancer. Cancer Res 1991; 51(2): 556±67. 21. Charpin C, Garcia S, Bouvier C et al. C-erbB-2 oncoprotein detected by automated quantitative immunocytochemistry in breast carcinomas correlates with patients' overall and diseasefree survival. Br J Cancer 1987; 75(11): 1667±73. 22. Dittadi R, Brazzale A, Pappagallo G et al. ErbB2 assay in breast cancer: possibly improved clinical information using a quantitative method. Anticancer Res 1997; 17(2B): 1245±7. 23. Press MF, Bernstein L, Thomas PA et al. HER-2/neu gene amplification characterised by florescence in situ hybridisation: poor prognosis in node-negative breast carcinomas. J Clin Oncol 1997; 15(8): 2894±904. 24. Cobleigh MA, Vogel CL, Tripathy D et al. Efficacy and safety of Herceptin1 (humanised anti-HER-2 antibody) as a single agent in 222 women with HER2 over expression who relapsed following chemotherapy for metastatic breast cancer. J Clin Oncol 1999; 17(9): 2639±48. 25. Slamon DJ, Leyland-Jones B, Shak S. Addition of Herceptin1 (humanised anti-HER-2 antibody) to first line chemotherapy for HER2 overexpressing metastatic breast cancer markedly increases anticancer activity: a randomized multinational controlled phase III trial. Cancer Proc Am Soc Clin Oncol 1998; 17: 98a. 26. Barnes DM, Gillet CE. Cyclin D1 in breast cancer. Breast Cancer Res Treat 1998; 52: 1±15. 27. Gillet CE, Fantl V, Smith R et al. Amplification and over expression of cyclin D1 in breast cancer detected by immunohistochemical staining. Cancer Res 1994; 54(7): 1812±7. 28. Alle KM, Henshall SM, Field AS, Sutherland RI. Cyclin D1 protein is expressed in hyperplasia and intraductal carcinoma of the breast. Clin Cancer Res 1998; 4: 847±54. 29. van Diest PJ, Michalides RJAM, Jannink I et al. Cyclin D1 expression in invasive breast cancer: correlations and prognostic value. Am J Pathol 1997; 150(2): 705±11. 30. Kenny FS, Hui R, Gee JMW et al. Over expression of cyclin D1 messenger RNA predicts poor prognosis in oestrogen receptorpositive breast cancer. Clin Cancer Res 1999; 5(8): 2069±76. 31. Nakopoulou L, Tsitsimelis D, Lazaris ACh et al. Nm23, c-erbB2, and progesterone receptor expression in invasive breast cancer: correlation with clinicopathologic parameters. Cancer Detect Prev 1999; 23(4): 297±308. 32. Steeg PS, Bevilacqua G, Kopper L et al. Evidence of a novel gene associated with low tumour metastatic potential. J Natl Cancer Inst 1988; 80(3): 200±4. 33. Rosengard AM, Krutzsch HC, Shearn A et al. Reduced nm23/awd protein in tumour metastasis and aberrant Drosophila development. Nature 1989; 342: 177±80. 34. Bevilacqua G, Sobel ME, Liotta LA, Steeg PS. Association of low nm23 levels in human primary infiltrating ductal breast carcinomas

478

35. 36.

37. 38. 39. 40.

41.

42.

43. 44.

45.

46. 47. 48. 49.

50. 51.

52. 53.

54.

with lymph node involvement and other histopathological indicators of high metastatic potential. Cancer Res 1989; 49(18): 5185±90. Royds JA, Stephenson TJ, Rees RC, Shorthouse AJ, Silcocks PB. Nm23 protein expression in ductal in situ and invasive human breast carcinoma. J Natl Cancer Inst 1993; 85(9): 727±31. Charpin C, Garcia S, Bonnier P et al. Prognostic significance of Nm23/NDPK expression in breast carcinoma, assessed on 10-year follow-up by automated and quantitative immunocytochemical assays. J Pathol 1998; 184: 401±7. Heimann R, Lan F, McBride R, Hellman S. Separating favourable from unfavourable prognostic markers in breast cancer: the role of E-cadherin. Cancer Res 2000; 60: 298±304. Duffy MJ, Reilly D, McDermott E, O'Higgins N, Fennelly JJ. Urokinase plasminogen activator in different subgroups of patients with breast cancer. Cancer 1994; 74: 2276±80. Duffy MJ, O'Grady P, Devaney D, O'Siorain L, Fennelly JJ, Lijnen HJ. Urokinase plasminogen activator, a marker for aggressive breast carcinomas. Cancer 1988; 62: 531±3. Duffy MJ, Duggan C, Mulcahy HE, McDermott EW, O'Higgins N. Urokinase plasminogen activator: a prognostic marker in breast cancer including patients with axillary node-negative disease. Clin Chem 1998; 44(6): 1177±83. Grùndahl-Hansen J, Christensen JJ, Rosenquist C et al. High levels of urokinase-type plasminogen activator and its inhibitor PAI-1 in cytosolic extracts of breast carcinomas are associated with poor prognosis. Cancer Res 1993; 53(11): 2513±21. Janicke F, Schmitt M, Pache L et al. Urokinase (uPA) and its inhibitor PAI-1 are strong and independent prognostic factors in node negative breast cancer. Breast Cancer Res Treat 1993; 24: 195±208. Spyratos F, Martin PM, Hacene K et al. Multiparametric prognostic evaluation of biological factors in primary breast cancer. J Natl Cancer Inst 1992; 84: 1266±72. Jahkola T, Toivonen T, von Smitten K, Virtanen I, Wasenius V-M, Blomqvist C. Cathepsin D, urokinase plasminogen activator and type-1 plasminogen activator inhibitor in early breast cancer: an immunohistochemical study of prognostic value and relations to tenascin-C and other factors. Br J Cancer 1999; 80: 167±74. Grùndahl-Hansen J, Agerlin N, Munkholm-Larsen P, Neilsen LS, Dombernowsky P, Danù K. Sensitive and specific enzyme-linked immunosorbant assay for urokinase-type plasminogen activator and its application to plasma from patients with breast cancer. J Lab Clin Med 1988; 111: 42±51. Tandon AK, Clark GM, Chamness GC, Chirgwin JM, McGuire WL. Cathepsin D and prognosis in breast cancer. New Engl J Med 1990; 322: 297±302. Spyratos F, Maudelonde T, Brouillet J-P et al. Cathepsin D: an independent prognostic factor for metastasis of breast cancer. Lancet 1989; 2(8672): 1115±18. Duffy MJ, Brouillet J-P, Reilly D et al. Cathepsin D concentration in breast cancer cytosols: correlation with biochemical, histological, and clinical findings. Clin Chem 1991; 37(1): 101±4. Foekens JA, Look MP, Bolt-de Vries J, Meijer-van Gelder MJ, van Putten WLJ, Klijn JGM. Cathepsin-D in primary breast cancer: prognostic evaluation involving 2810 patients. Br J Cancer 1999; 79(2): 300±7. Jahkola T, Toivonen T, Virtanen I et al. Tenascin-C expression in the invasion border of early breast cancer: a predictor of local and distant recurrence. Br J Cancer 1998; 78(11): 1507±13. Chiquet-Ehrismann R, Kalla P, Pearson CA. Participation of tenascin and transforming growth factor-beta in reciprocal epithelial±mesenchymal interactions of MCF-7 cells and fibroblasts. Cancer Res 1989; 49(15): 4322±5. Gould VE, Koukoulis GK, Virtanen I. Extracellular matrix proteins and their receptors in the normal, hyperplastic and neoplastic breast. Cell Differ Dev 1990; 32(3): 409±16. Jahkola T, Toivonen T, von Smitten K, Blomqvist C, Virtanen I. Expression of tenascin in the invasion border of early breast cancer correlates with higher risk of distant metastasis. Int J Cancer 1996; 69(6): 445±7. Joachim EE, Athanassiadou SE, Kamina S, Carassavoglou K, Agnantis NJ. Matrix metalloproteinase expression in human breast cancer: an immunohistochemical study including

C. E. ROGERS ET AL.

55. 56. 57. 58.

59. 60. 61. 62.

63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76.

77.

correlation with cathepsin D, type IV collagen, laminin, fibronectin, EGFR, c-erbB-2, oncoprotein, p53, steroid receptor status and proliferative indices. Anticancer Res 1998; 18: 1665±70. Garbett EA, Reed MWR, Brown NJ. Proteolysis in human breast and colorectal cancer. Br J Cancer 1999; 81(2): 287±93. Nielsen BS, Sehested M, Kjeldsen L, Borregaard N, Rygaard J, Dano K. Expression of matrix metalloproteinase 9 in vascular pericytes in human breast cancer. Lab Invest 1997; 77(4): 345±55. Rha SY, Kim JH, Roh JK et al. Sequential production and activation of matrix-metalloproteinase-9 (MMP-9) with breast cancer progression. Breast Cancer Res Treat 1997; 43(2): 175±81. Arnold SM, Young AB, Munn RK, Patchell RA, Nanayakkara N, Markesbery WR. Expression of p53, bcl-2, E-cadherin, matrix metalloproteinase-9, and tissue inhibitor of metalloproteinases-1 in paired primary tumours and brain metastases. Clin Cancer Res 1999; 5(12): 4028±33. Hedley DW, Friedlander ML, Taylor IW. Method for analysis of cellular DNA content of paraffin-embedded pathological material using flow cytometry. J Histochem Cytochem 1983; 31: 1333±5. Gerdes J, Schwab U, Lemke H, Stein H. Production of a mouse monoclonal antibody reactive with a human nuclear antigen associated with cell proliferation. Int J Cancer 1983; 31: 13±20. Gerdes J, Lelle RJ, Pickartz HL et al. Growth fractions in breast cancers determined in situ with monoclonal antibody Ki-67. J Clin Pathol 1986; 39: 977±80. Cattoretti G, Becker MHG, Key G. Monoclonal antibodies against recombinant parts of the Ki-67 antigen (MIB 1 and MIB 3) detect proliferating cells microwave-processed formalin-fixed paraffin sections. J Pathol 1992; 168: 357±63. Baak JPA, van Dop H, Kurver PHJ, Hermans J. The value of morphometry to classic prognosticators in breast cancer. Cancer 1985; 56: 374±82. Isola JJ, Helin HJ, Helle MJ, Kallionoemi O. Evaluation of Ki-67 immunohistochemical study, DNA flow cytometric analysis, and mitotic count. Cancer 1990; 65: 1180±4. Veilh P, Chevillard S, Mosseri V, Donatini B, Magelenat H. Ki-67 index and S phase fraction in human breast carcinomas. Am J Clin Pathol 1990; 94: 681±6. Sahin AA, Ro J, Ro JY et al. Ki-67 immunostaining in node-negative stage I/II breast carcinoma. Cancer 1991; 68: 549±57. Brown RW, Allred CD, Clark GM, Osborne CK, Hilsenbeck SG. Prognostic value of Ki-67 compared to S phase fraction in axillary node-negative breast cancer. Clin Cancer Res 1996; 2(3): 585±92. Jansen RLH, Hupperets PSGJ, Arends JW et al. MIB-1 labelling index is an independent prognostic marker in primary breast cancer. Br J Cancer 1998; 78(4): 460±5. Keshgegian AA, Cnaan A. Proliferation markers in breast carcinoma. Am J Clin Pathol 1995; 104: 42±9. Dettmar P, Harbeck N, Thomssen C et al. Prognostic impact of proliferation-associated factors MIB-1 (Ki-67) and S-phase in nodenegative breast cancer. Br J Cancer 1997; 75(10): 1525±33. Kuiper GG, Enmark E, Pelto-Huikko M, Nilsson S, Gustaffson JA. Cloning of a novel receptor expressed in rat prostate and ovary. Proc Natl Acad Sci USA 1996; 93: 5925±30. Speirs V, Kerin MJ. Prognostic significance of oestrogen receptor b in breast cancer. Br J Surg 2000; 87: 405±9. Taylor AH, Al-Azzawi F. Immunolocalisation of oestrogen receptor beta in human tissues. J Mol Endocrinol 2000; 24: 145±55. Speirs V, Parkes AT, Kerin MJ et al. Coexpression of estrogen receptor a and b: poor prognostic factors in breast cancer. Cancer Res 1999; 59: 525±8. Speirs V, Malone C, Walton DS, Kerin MJ, Atkin SL. Increased expression of oestrogen receptor bmRNA in tamoxifen-resistant breast cancer patients. Cancer Res 1999; 59: 5421±4. JaÈrvinen TAH, Pelto-Huikko M, Holli K, Isola J. Estrogen receptor b is coexpressed with ERa and PR and associated with nodal status, grade, and proliferation rate in breast cancer. Am J Pathol 2000; 156(1): 29±35. McGuire WL, Clark GM. Prognostic factors and treatment decisions in axillary node-negative breast cancer. New Engl J Med 1992; 326(26): 1756±61.

Accepted for publication 11 January 2002