HISTORIAS DE USUARIO

A sensitive and reproducible method for detection and enumeration of epithelial tumour cells requires rigorous evaluation in model systems before application in the clinical setting to ensure both reliable and accurate results.

The routine ICC analysis of BM from patients with BrCa can present some problems. For example, in previous studies, ICC techniques have been used that require the examination of BM spread over several slides (Schlimok et al, 1987; Cote at al, 1988). The technique used here has some possible advantages over previous ICC techniques. The addition of predetermined numbers of cells to defined areas of coated adhesive microscope slides allowed cells to sediment and attach and may lead to better distribution of cells than previously experienced with cytocentrifuge preparations. These slides are easier to examine and as a consequence, a larger number of cells can be applied per slide. The adhesive nature of the test surfaces prevented significant cell loss during extensive liquid incubation and washing procedures. The data illustrate that in the multiple seeding experiments performed, one BrCa cell in

1,000,000 Daudi cells could reliably and reproducibly be detected in the Cell- Tak® and Cel-Line systems and one in 100,000 with the Poly-L-Lysine system. Each adhesive system offered enhanced cell attachment and spreading capabilities within the wells. Furthermore, up to 2 x 10® cells could readily be screened from clinical material obtained, without the need to examine multiple cytospin preparations, and the data further demonstrate the feasibility of ultra- high sensitivity detection of tumour cells in haematopoietic samples.

The technique also allows subsequent morphological characterisation of the malignant nature of CK"" cells, achieved by counterstaining stored frozen slides prepared at the time of sample processing, or by the application of FISH

technology (Chapter 6). Fluorescence in situ hybridization could facilitate the further biological characterisation of contaminating/disseminated tumour cells, with respect to metastatic potential, and may provide a powerful approach that can be used to detect over amplification of various genetic prognostic markers, and new insights into the course of an individual tumour. The sensitivity of the method is high, and clonal variation among the carcinoma cells can be revealed. Because FISH depends on molecular marker probes that give clear cut qualitative results, only a small number of cells are needed for analysis.

By applying ICC techniques, breast tumour cells have been detected in BM biopsies of 20-45% of patients with local disease at the time of surgery, and of up to 70% in those with métastasés (Schlimok et a/, 1987; Mansi et a/, 1991). A higher likelihood of early relapse and reduced overall survival are associated with the presence of BM micrometastasis (Cote et ai, 1991; Diel et al, 1992; Braun et al, 2000b). In multivariate analysis of prognostic factors, the results of BM biopsy behaved as an independent factor and positivity was significantly associated with early disease relapse (Harbeck et al, 1994) and reduced overall survival (Braun et al, 2000a). In other studies the best predictive value was from the combination of BM micrometastasis, nodal status and progesterone receptor status (Diel et al, 1992). Breast tumour cells in BM and peripheral blood are also capable of in-vitro clonogenic growth (Ross et al, 1995a), and in previous studies, tumour contamination of either BM grafts or PBSC collections has been associated with shorter disease-free and overall survival in stage ll-lll patients subjected to autologous transplantation (Harbeck et al, 1994; Hurd and Peters, 1995; Pedrazzoli et al, 1997; Vredenburgh et al, 1997). In follow-up studies of a subset of advanced-stage BrCa patients reported initially by Ross et al (1993), preliminary analysis at 30 months indicates that no correlation was

found with tumour contamination of the PBSC collection and time to disease progression, sites of relapse, or overall survival (Ybanez et al, 1995). Another retrospective study of a small cohort (n = 26) of advanced-stage BrCa patients with tumour-contaminated BM found a trend toward decreased overall survival (p = 0.11) at 84 months follow-up (Brockstein et ai, 1996). Insufficient power as a result of the low sample size precluded any subset analyses in this cohort. Similar trends toward decreased progression-free and overall survival in advanced-stage BrCa patients who received tumour-contaminated PBSC collections have been reported by Gluck et ai (1996).

In contrast, the retrospective study of MRD in BM by Fields et ai (1996) concluded that the probability of relapse at 36 months posttransplant was 32% for stage ll-lll and 94% for stage IV CK19 RT-PCR-positive patients. Conversely, the probability of relapse was 10% for stage ll-lll and 14% for stage IV CK19 RT-PCR-negative patients. The difference was significant (p = 0.0002) for stage IV patients. In a similar retrospective study, investigators at Duke University found that stage ll-lll BrCa patients with ICC-positive BM harvests had significantly shorter disease-free and overall survival (Vredenburgh et ai, 1997). Unfortunately, too few patients had tumour-contaminated PBSC for comparable statistical analysis. Moreover, Moss et al (1997) reported that tumour contamination of autologous BM grafts in stage IV BrCa patients was significantly associated with reduced disease-free survival (p = 0.0001). Patients who had both BM and PBSC contaminated with tumour had a worse posttransplant clinical course than did all other patients (p = 0.015). However, this study was a retrospective analysis of patients from a number of transplant centres and only included univariate statistical analysis. However, with reference to the studies cited (including this one), without multivariate analyses,

it is difficult to determine if tumour contamination of the graft is truly predictive of posttransplant outcome, or if it is a very accurate surrogate marker for treatment-resistant disease.

Some of the aforementioned studies indicate that the reinfusion of occult KR&L.<\TE:

tumour cells in autografts, as detected by ICC, do not contribute significantly to relapse following HDC in patients with BrCa. There are several possible explanations for these observations. First, the number of tumour cells may be insufficient to result in relapse and not all occult tumour cells may be clonogenic in vivo. It is possible that freezing and thawing could affect the number and clonogenicity of tumour cells. The immune system of the patient might eliminate the relatively small number of tumour cells infused and/or ultrasensitive ICC or other testing methods will need to be employed to reduce the number of possible false negative results. Any one or a combination of these mechanisms could explain the apparent lack of effect of infused tumour cells on relapse.

Recently, an RT-PCR assay for the detection of CK mRNA has been proposed as a marker of cancer cell contamination in BrCa patients. Cytokeratins are cytoskeletal intermediate filaments (Moll et ai, 1982) that are selectively expressed in cells of epithelial origin; however, CK8 and CK18 mRNA were found to be expressed also in samples of normal PB (Traweek at al, 1993), although data for K19 were more conflicting. In fact Traweek at al (1993) and Datta at al (1994) failed to detect K19 mRNA in the PB and BM samples from normal subjects or patients with haematological disease, whereas false positive results in 38-50% of normal PB samples have been reported by others (Burchill at al, 1995; Krismann atal, 1995; Hildebrandt ef a/, 1997).

Although there are several studies in which histology and/or ICC have been used for the evaluation of BM biopsies in BrCa patients, no comprehensive comparative analysis of these morphological techniques with RT-PCR for K19 has been undertaken. In accord with other studies (Fields et al, 1996), conventional histology proved inadequate to evaluate BrCa cell contamination, since >50% of BM samples with a positive result by both RT- PCR and ICC were considered to be normal by the pathologist. Conversely ICC and RT-PCR showed a high degree of concordance; additionally, it is conceivable that ICC might help in identifying cancer cells in occasional samples with negative RT-PCR results as a consequence of the uneven pattern of BM infiltration and sampling. Therefore, ICC and RT-PCR for K19 could possibly be combined. By using these sensitive techniques, the rate of detection of BM contamination may be improved in women with high-risk, localised, non-metastatic BrCa at the time of primary diagnosis.

Although PBSC collections are believed to have a lower incidence of tumour involvement than BM in BrCa patients, there are few studies to support this contention (Sharp et ai, 1992a; Ross et ai, 1991). To address this issue, a highly sensitive ICC assay was developed and validated by quantifying the number of occult tumour cells in unpaired BM and PBSC specimens collected from 19 patients with either advanced-stage metastatic or high-risk, localised,

non-metastatic BrCa.

The findings indicate that occult BrCa cells are present less frequently in PBSC than in BM (p = 0.0001, Fisher’s exact test). Furthermore, the concentration of tumour cells in immunocytochemically positive PBSC collections was significantly lower than in immunocytochemically positive BM (p = 0.0012, Mann-Whitney test). In patients with metastatic disease, the

incidence of circulating tumour cells detected in PBSC collections was independent of sites of metastatic involvement. Because patients undergoing PBSC infusion received a greater number of infused haematopoietic cells than patients receiving autologous marrow, and despite this lower contamination, this resulted in the infusion of a comparable number of tumour cells (p = 0.0503, Mann-Whitney test). Therefore, it may be postulated that not only tumour burden but phenotypic differences between BM-derived and PBSC-derived tumour cells may govern the risk of metastatic relapse following dose intensification. In addition, the data indicate that tumour cells may not be present in all PBSC pheresis collections from a single patient. Taken together, these data support the former view that PBSC collections are a less- contaminated source of haematopoietic stem cells compared to BM (Moss and Ross, 1992).

In this retrospective analysis of a small number of localised BrCa patients, a trend was found towards a statistically significant difference in DFS in patients who were inadvertently infused with contaminating tumour cells. There are at least two possible explanations. The first, and perhaps the most obvious, is that reinfused cancer cells contribute to disease relapse; the other, and perhaps the most pertinent, is that stem cell contamination reflects only a greater overall tumour burden present at diagnosis, with reinfused cancer cells having no major role on disease relapse. Additionally, with reference to the second explanation and to the patients with advanced stage IV metastatic disease, transplanted with CK-negative components, the shorter DFS observed may indicate relapse from residual endogenous tumour (greater tumour burden and/or more resistant disease). These observations suggest that when treating a subgroup of patients with a relatively high and chemoresistant residual tumour

burden, the inadequate tumour cytoreduction makes it difficult to detect the potential influence of the infusion of tumour. In contrast, when analysing patients with high risk of relapse stage II and III BrCa, potentially with a more limited and chemosensitive residual tumour burden, the presence of tumour cells in the infused haematopoietic support may more accurately predict relapse. Furthermore, with regard to relapse, preliminary evidence in patients with non-Hodgkin's lymphoma (Sharp et al, 1992b; Vose at ai, 1992) and leukaemia (Miller at a/, 1991) suggests that the removal of tumour cells either by BM purging techniques or the use of tumour-uninvolved PBSC may improve survival. The presence of contaminating tumour cells in pheresis products has been reported in recent studies to occur at a frequency of 4-20% (Brugger at a/, 1994; Fields at a!, 1996; Passos-Coelho at ai, 1996; Vogel at a!, 1996; Schulze at ai, 1997); the higher figure reported here (58% of patients) may be due to the improved sensitivity of this novel ICC assay. However, in another study (Mapara at ai, 1997) it was shown that as many as 17/21 (82%) pheresis products analysed with a pancytokeratin monoclonal antibody were found to contain tumour cells. Therefore, the use of techniques such as purging with chemotherapy (Kennedy at a/, 1991; Shpall at a/, 1991a; Passos-Coelho at a/, 1994), MoAbs (Shpall at a/, 1991b), or enriching for CD34^ cells (Berenson at a/, 1991) were explored, in the hope that this might reduce the tumour load. The immunoselection of CD34"^ cells employed by Vannucchi at a! (1998) in four cases with micrometastasis in bilateral BM biopsies at diagnosis (as demonstrated by histology, ICC and RT-PCR) and in PBSC collections (by RT- PCR) was able to yield a stem cell suspension devoid of cancer cells. Recent experience with ICC analysis of purified CD34'" cells (Hildebrandt at a/, 1997), even after ax-vivo expansion using a cocktail of haemopoietic growth factors

(Vogel et al, 1996), is in line with the observations of Vannucchi at a! (1998). However, Mapara at a! (1997) still found a significant tumour cell contamination after CD34"^ cell selection using both ICC and RT-PCR for K19 (single-round RT-PCR reaction) and the epidermal growth factor receptor (EGF-R); although the reasons for these discrepancies are unknown, the use of a different method for CD34"^ cell separation (an immunomagnetic-based procedure in that study versus an avidin device in the Vannucchi at a/, [1998] study) might have been, at least in part, responsible.

If ICC techniques are to be used for the detection of BrCa cells, methods must be developed to deal with the heterogeneity of antigen expression within and between different tumours. Substantial phenotypic heterogeneity has been observed in different breast cancers. However, the use of the broad-spectrum MoAb A45-B/B3 explored in this study supports its use for detecting cancer cells in haematopoietic samples from BrCa patients, and these assays were further utilised to study clinical material from patients with other epithelial malignancies.

In conclusion, this study shows a good correlation between the number of tumour cells added and the number of cells detected, by ICC, and this correlation has been shown to be linear, over a range of several logs in all three adhesive systems. This technique will allow quantitation of tumour cells in BM samples, and it may prove useful in monitoring the response of metastatic disease to cytotoxic or hormonal therapy and this application is explored below. The sensitivity and specificity of this ICC method supports its introduction into tumour staging classifications. Results obtained by this method may then serve as a new standard for alternative approaches to substantiate the possible

increased sensitivity over existing immunocytology methods using MoAbs to detect CK^ cells.