Review. A. C. Lambrechts, L. J. van 't Veer z & S. Rodenhuis J

Transcription

1 Annals of Oncology 9: Khmer Academic Publishers. Printed in the Netherlands. Review The detection of minimal numbers of contaminating epithelial tumor cells in blood or bone marrow: Use, limitations and future of RNA-based methods.1.2 A. C. Lambrechts, L. J. van 't Veer z & S. Rodenhuis J Division of Experimental Therapy, 2 Department of Pathology, 3 Department of Medical Oncology. The Netherlands Cancer Institute, Amsterdam. The Netherlands Summary Background: Many solid tumors commonly metastasize to the bone marrow and the presence of tumor cells in the bone marrow is associated with a poor prognosis. Detection of tumor cells in the bone marrow has been reported to be important to determine the prognosis of newly diagnosed patients and may be helpful in deciding whether or not systemic treatment is indicated. Patients and methods: The majority of the studies focus on the detection of tumor cells in non-tumor tissue using immunocytochemistry and antibodies directed against epitopes of epithelial genes. Recently, the sensitive reverse-transcriptase polymerase chain reaction (RT-PCR) has been employed for the detection of tumor cells in bone marrow, using mrna transcribed from epithelial genes as targets for RT-PCR. Results: In some studies, encouraging results were reported Introduction Many types of cancer commonly metastasize to the bone marrow. This is detectable by light microscopy and with significantly greater sensitivity by immunocytochemistry. Several reports suggest that the presence of (small numbers of) tumor cells in the bone marrow is associated with a very poor prognosis in a variety of epithelial tumors, such as breast cancer [1-5], colorectal carcinomas [6] and lung cancer [2, 7]. On the other hand, it is also clear, that a proportion of patients with minimal bone marrow involvement of breast cancer at diagnosis can achieve long-term survival [1, 8-10]. As a result, a certain degree of uncertainty exists about the significance of bone marrow involvement and about the specificity of the techniques demonstrating minimal bone marrow involvement. When a small number of cells is detected at microscopic examination that stain positively with antibodies recognizing epithelial antigens, it is assumed that these cells represent tumor cells. In some cases, this assumption can be confirmed by standard histologic or cytologic techniques, but if that is the case one might argue that the degree of bone marrow involvement is 'moderate' rather than minimal. This difference is important, when RT-PCR was used to detect expression of epithelial genes, but in many others frequent false-positive results were observed. These may result from the 'illegitimate expression' of epithelial genes in cells of non-epithelial tissues, such as bone marrow. Conclusions: Micrometastases in bone marrow can be detected with some sensitivity by antibodies directed against epithelial genes. RNA based methods, using epithelial genes as target for amplification, are less reliable. To improve these methods, a systematic approach is required to identify genes which are highly expressed in solid tumors and completely silent in blood and bone marrow of healthy individuals. Novel techniques, e.g.,'sequential analysis of gene expression' (SAGE), are now available that allow such an endeavor. Key words: epithelial tumor cells, immunocytochemistry, minimal disease, prognosis, RNA-based methods since bone marrow involvement of an epithelial tumor that can be detected without using immunocytochemistry is invariably followed by clinically disseminated disease and death. The large majority of studies focusing on minimal bone marrow involvement of epithelial tumors employ immunocytochemistry as a method to identify tumor cells among normal cells of the bone marrow. In most cases, antibodies directed against epitopes of epithelial genes are used, that are not expressed by native bone marrow cells, which are derived from mesenchymal progenitors. In patients with breast cancer [1-3], gastric cancer [11, 12], colorectal cancer [13] and lung cancer [7, 14] the presence of cells recognized by antibodies directed against epitopes of epithelial genes in the bone marrow indicates systemic disease and is reported to be a predictor for distant relapse. However, other studies have reported the absence of a correlation between the presence of cells staining positive with similar monoclonal antibodies and the probability of disease free survival and overall survival [8-10]. In those patients, later shown to have achieved long-term survival, the positively staining cells may have represented tumor cells without self-renewal capacity or normal bone marrow cells that stain false-positive [15, 16]. For example,

2 1270 monoclonal antibodies directed against cytokeratins are clearly not cancer-specific and some have been shown to cross-react with normal blood or bone marrow elements (and personal observation) [17]. Another explanation could be that cells from the macrophage/monocyte system may contain proteins derived from the primary tumor that has undergone necrosis or apoptosis and that these processed proteins are recognized by the antibodies. In fact, the capacity and the experience of the investigator to determine which of the stained cells are tumor cells has been stressed in the literature [3]. The assays are essentially subjective with considerable interobserver variation. Alternative methods to either replace or complement the existing immunocytochemical assays have been sought. The frequent occurrence of non-random chromosomal translocations has allowed the development of highly sensitive and specific DNA methods in non- Hodgkin's lymphomas [t(14;18)] [18] and leukemias [t(9;22)] [19, 20], but the absence of such chromosomal translocations in breast cancer and in other epithelial cancers has precluded the use of these molecular methods in these patients. Thus, it is not surprising that a great deal of work has been done attempting to apply the reverse-transcriptase polymerase chain reaction (RT- PCR). RT-PCR is potentially a very sensitive technique that should be able to detect minute amounts of specific RNA species in the total amount of RNA. If the expression of certain genes would indeed be absolutely confined to the cell types to be detected, RT-PCR based techniques should be able to distinguish such cells with certainty and since active rarna transcription is than documented, powerful arguments for viability and specificity could be presented. RT-PCR: Sensitivity and specificity Several groups have developed assays based on RT-PCR (rather than immunocytochemistry), using epithelial genes as targets for amplification. While some investigators have published encouraging results, many studies have encountered the problem of false-positivity and the loosely defined phenomenon of 'illegitimate expression' (see sections: target genes for epithelial tumors and the problem of illegitimate expression). In fact, however, the extreme sensitivity of the PCR and its tendency to amplify even single RNA molecules [21, 22] - including contaminating ones - may have contributed to falsepositive results. The sensitivity and specificity of RT-PCR (Figure 1) assays have frequently been determined by its application to different mixtures of cells, for instance of the breast tumor cell line MCF7 (that expresses certain target genes such as cytokeratin 19 at high level) and normal mononuclear blood cells. Small numbers of cultured breast cancer cells 'diluted' in normal blood lymphocytes are commonly used as positive controls for RNA isolation, cdna synthesis and PCR. The sensi- L-- anneal random hexamer primers I separation of strands (94 Q M 5" I annealing of sense PCR pinner (eg 60 C) I cxlcrmon (72 C) First strand cdna synthesis cdna random hexamer pnmcrs sense PCR primer antiscnse PCR primer Figure I The RT-PCR technique. cdna is synthesized using random hexamers as primers and reverse transcriptase (RT). The resulting single stranded cdna molecule is a target for amplification by PCR, using two sequence-specific primers and Taq DNA polymerase. cdna strands are separated, allowing the annealing of the primers, and these primers are extended by Taq DNA polymerase. The generated double stranded cdna is again a target for amplification. The continuous repetition of this process results in the exponential synthesis of targetspecific cdna. tivity of RT-PCR assays is than expressed as the number of tumor cells which can be detected among the mononuclear blood cells. The mononuclear blood cells alone, and the 'water' sample, are regarded as negative controls for RNA isolation, cdna synthesis and PCR, and are used to ascertain the specificity of the assay. However, when positive test results are observed in pure mononuclear blood cells, this could - in principle - be due to the presence of an extremely low expression level of the target gene in these cells. Other RNA-directed techniques A further technique to detect specific mrna expression is the Nucleic Acid Sequence Based Amplification (NASBA) assay. This technique has been described for the first time in 1991 by Compton [23]. NASBA has extensively been used in identification of mycobacteria [24] and for the qualitative and quantitative detection of HIV-1 RNA [25]. In tumor cell detection assays, NASBA has only been described for the detection of the BCR-ABL fusion mrna in Philadelphia chromosome-positive chronic myeloid leukemia [26], Studies employing NASBA in the detection of circulating epithelial tumor cells are in progress. A potential advantage of NASBA over RT-PCR is

3 1271 ss RNA 5' Table 1. Target genes used in RT PCR based assays to detect dissemination of tumor cells in bone marrow or circulating tumor cells in the peripheral blood. Gene Cytokeratin-19 Cytokeratin-20 CEA EGF-R Maspin Tumor type Lung cancer Colorectal cancer Gastric cancer Reference [36,37, ] [38] [42. 47] [35, ] [ ] [50] [48] Primer 2 5' 3' 3/VWV5 IAMV-RT 3AVW5 3/WSA/5 1 3A/VW5 3/WVV5' Cyclic Phase 3AA/VV5 3AAW5 3/WW5 RNase" H Primer 1 T7 RNA pol Sense RNA 5' 3' Sense DNA 3A/VW5' Antisense RNA 3' 5' Antisense DNA Figure 2. The NASBA amplification principle. NASBA employs AMV (Avian Myeloblastosis Virus)-reverse transcriptase (AMV-RT), RNaseH and T7-RNA polymerase in combination with two sequencespecific primers in order to produce an isothermal amplification process. Primer 1, containing a 3' target specific sequence and a 5' T7 promotor sequence, anneals to the RNA molecule and is extended by reverse transcriptase to synthesize a first strand cdna molecule. The RNA of the RNAxDNA hybrid is digested by RNaseH to allow annealing of primer 2 to the single cdna strand. Primer 2 is extended by AMV-RT to form a double stranded cdna molecule. This double stranded cdna molecule contains a T7 promotor sequence and is therefore a template for transcription. T7-RNA polymerase initiates transcription and multiple copies of antisense RNA are generated terminating at the 5' end of primer 2 and each newly synthesized antisense RNA molecule is again a target for amplification. The its unique ability to selectively amplify RNA (even in the presence of DNA) which is the result of the fact that the reaction can be performed at 41 C. At this temperature, DNA does not denature and is therefore not a target for sequence specific amplification by NASBA. Additional advantages of NASBA over RT-PCR include the sensitivity of the assay [26] and the strongly reduced risk of contamination since it requires less pipetting steps than RT-PCR. Therefore, NASBA can also be used to Table 2. Target genes used in RT-PCR based assays to detect micrometastases in regional lymph nodes. Gene Tumor type Reference Cytokeratin-19 Cytokeratin-20 Carcino embryonic antigen Mucin-1 Colorectal cancer Gastric cancer Colorectal cancer Gastric cancer [29] [28] [30-32, 34] [29] [27, 35] [27] [31,33] achieve selective amplification of tumor cell RNA (Figure 2) [23]. Target genes for epithelial tumors To be able to detect RNA species of tumor cells among the RNAs of bone marrow or blood cells, primer combinations are required that anneal to tumor RNA but not with 'bone marrow RNA'. For epithelial tumors, genes that are thought to be tissue specific for epitheliumderived cells are a logical choice. The expression of a range of epithelial and other tissue-specific genes has been studied for this purpose (Table 1). Employing one of these genes as a target, RNA expression was used to detect the presence of lymph node micrometastases in gastric cancer [27, 28], colorectal cancer [29] and breast cancer [27, 30-34]. The tumors and target genes are listed in Table 2. For one of the most frequently employed genes, the cytokeratin-19 (CK19) gene, conflicting results have been reported. False-positive results were common [29, 34] but these could sometimes be avoided by reducing the number of PCR cycles [34], suggesting that technical factors, including the sensitivity and specificity of the assay, may account for some of these. It has been reported that CK19 could be used to asses the presence of micrometastases in lymph nodes with greater sensitivity than histological examination [28, 30, 31, 34]. Cytokeratin-20 expression was evaluated in lymph nodes of patients with colorectal cancer. Expression was found in 26 of 109 lymph nodes of patients with colorectal cancer, but in none of 12 control lymph nodes [29]. The carcino-

4 1272 Table 3. Expression of 'tissue- or tumor-specific' genes in bone marrow (BM) or peripheral blood mononuclear cells (PBMC) from control individuals (false-positives). Gene Epithelial glycoprotein-40 Desmoplakin I Carcino embryonic antigen erb-b2 erb-b3 Prostate specific membrane antigen Cytokeratin-8 Cytokeratin-18 Cytokeratin-19 Episialin Percentage of false-positive samples 100% (BM) 100% (BM) 26% (BM) 42% (BM) 4% (PBMC) 71% (BM) 86% (BM) 44% (BM) 100% (BM, PBMC) 88% (PBMC) 100% (BM, PBMC) 100% (PBMC) 4% (BM) 4% (BM) 3% (PBMC) 20% (PBMC) 40% (PBMC) 60% (BM) 100% (PBMC) 100% (PBMC) Reference [43] [42] [43] [37] [40] [39] [38, 60] [42] embryonic antigen (CEA) was successfully used as a marker to evaluate the presence of micrometastases in lymph nodes, but false-positives could only be avoided if the number of PCR cycles was restricted [27, 35]. Another useful gene is reportedly the mucin-1 (MUC1) gene. Lymph nodes with no histochemically detectable micrometastases were positive by PCR evaluation [31, 33]. However, no control lymph nodes were tested in this study. In general, RT-PCR based assays may eventually prove useful to detect minimal dissemination of epithelial tumors to regional lymph nodes but standardization is lacking, the specificity of the tests is questionable and sensitivity issues have not been addressed at all. The detection of micrometastases in bone marrow or peripheral blood mononuclear cells (PBMC) using RT-PCR has attracted much interest, particularly because bone marrow and peripheral blood progenitor transplantations are increasingly employed in patients with epithelial tumors and the reinfusion of viable tumor cells in the graft is of obvious concern. A range of different epithelial and non-epithelial target genes has been tested for suitability as a tumor marker in these assays (Table 1). Again, the most extensively tested one is the CK19 gene [36-41]. Nevertheless, the evaluations of the suitability of CK19 gene expression as a tumor marker are not conclusive. CK19 transcripts were reported to be either detectable [38, 42] or absent [36, 39-41, 43] in PBMC of healthy volunteers. In patients with BM PB Figure 3. An example of CEA expression in bone marrow (upper panel) and peripheral blood (lower panel) samples of healthy individuals («= 10). RNA was amplified by NASBA, transferred to a nylon membrane and hybridized with a radioactively labeled oligonucleotide probe. The filters were exposed to a Kodak X-Omat autoradiogram usually for one day with an intensifying screen at -70 C. Strong positive signals were detected in four bone marrow samples (lanes: 1, 5, 7, 9) and faint positive signals were detected in 3 bone marrow samples (lanes: 2, 3, 8) and in one blood samples (lane 5). breast cancer, CK.19 transcripts were detectable in blood [36, 39, 41] and bone marrow [36, 37, 40, 41]. It was concluded that RT-PCR is a more sensitive method for detection of micrometastases than immunocytochemistry. The presence of CK19 transcripts in bone marrow aspirates was associated with a poor prognosis in patients with metastatic breast cancer, undergoing high-dose chemotherapy and autologous bone marrow transplantation [37]. In patients with lung cancer there was a correlation between the presence of CK19 expression in PBMC and the stage of disease, although false-positive results were reported in 20% of the control samples [38]. Less extensively tested are the CEA gene [35, 44-46] (Figure 3), the CK20 gene [42, 47], the maspin gene [48], the EGP2 gene [49] and the epidermal growth factor receptor gene [50]. For these genes, only small numbers of patient- and control samples were tested, with in some cases extremely large number of PCR cycles. Therefore, additional data are necessary to establish the suitability of these genes as marker genes for epithelial tumors. Moreover, for most target genes tested so far, positive results in control samples have been reported with worryingly high frequencies (Table 3 and Figure 3). Obviously, false-negative results represent the other side of the coin. Human tumors are known to harbor widely varying populations of cells, and this heterogeneity increases during the tumor's growth and spread. As a result, a given tumor cell in bone marrow or circulating in the blood may not express a marker gene, even if that gene has been shown to be expressed in the primary tumor. This problem could be alleviated in part by employing a small panel of target genes rather than a single one. 10

5 1273 The problem of illegitimate expression The finding of positive signals in blood or bone marrow of healthy control-individuals appears to be a universal problem of RNA-based assays designed to detect small proportions of tumor cells. At least three classes of explanations can be envisioned: 1. The amplification process replicates a sequence that is similar to but not identical with the target sequence (e.g., a transcript from a gene with a considerable homology). 2. Target genes may be expressed in a minority of non-tumor cells at significant levels. This is supported by the observation that specific staining of a small proportion of blood cells has been detected using different antibodies directed against epitopes of cytokeratin genes or the episialin gene (a gene expressed in breast cancer). 3. There may be a low degree of expression of the target gene in all or in some native blood- or bone marrow cells. The latter explanation is based on the hypothesis of 'illegitimate expression', which holds that any gene can be expressed in any cell type, albeit to a low level. The term 'illegitimate expression' has been coined by Chelly et al. [51]. These investigators had noted that transcripts of the Duchenne Muscular Dystrophy gene were not only detectable in muscle and brain cells (as expected), but also at low levels in a range of other cell types [52]. To investigate whether other 'tissue-specific' genes were also expressed in unrelated cell types, they studied the expression of anti-miillerian hormone, beta-globin, aldolase A, and factor VIIIc in fibroblasts, hepatoma cells and lymphoblasts employing the RT- PCR. Low levels of transcription could indeed be detected by RT-PCR. They also showed low-grade expression of the erythroid- and liver-type pyruvate kinase genes in several rat tissues, such as brain, lung and muscle. The expression levels were very low and were estimated to be in the order of one transcript in cells [51-53]. It has been speculated that gene expression in the absence of specific transcription factors may be very low, but not zero. Expression from methylated stretches of DNA or from DNA that is part of condensed chromatin may be impossible under most conditions, but these inactive states of DNA may change when DNA replication is taking place. At that time, any promotor could be minimally active when ubiquitous transcription factors reach their cognate DNA elements [51]. If this explanation is correct, and at least 1 transcript of any gene is present in any collection of cells, a million blood cells should contain approximately transcripts of a given gene that is 'illegitimately' expressed. In each cell type, transcripts from 14,000-20,000 different genes are present [54, 55]. The number of the individual transcripts ranges from copies per cell [54]. In colon cancer, for example, approximately 60 transcripts were present at more than 500 copies, representing 20% of the total amount of mrna, per cell [54]. Thus some tissue-specific genes may be expressed at a level of over 500 copies per cell. If one assumes that a carefully selected cancer- or tissue-specific gene leads to 500 mrna copies per epithelial cell, this would put the natural limit of detection by an RNA-based method at about four epithelial cells per million blood or marrow cells. This level of sensitivity does not exceed that of immunocytochemistry. Thus, this explanation would quite adequately explain the problem of false-positives in many reports (Table 3). At the other hand, the only systematic study of 'illegitimate expression' is that of Chelly, and even that studied only a small number of genes. If an assay system would focus on a set of abundant RNA species in epithelial cells (e.g.,fivegenes with an expression level of 5000 or more mrna copies per gene and per cell), a higher degree of sensitivity could be achievable: one tumor cell in marrow cells in this hypothetical example. This is about the maximum sensitivity that would still be practical for a clinical test: the volume of blood to be drawn in order to obtain 10 7 nucleated blood cells after a density gradient separation is about 10 ml. Very little work has been done to select appropriate genes for this purpose. The genes that have been employed to date are listed in Tables 1-3, and all have been selected on the basis of studies that have detected antigen expression in tumor cells but not in bone marrow cells at the microscopic level. The available RNA-based methods show that these are far from ideal. We suggest that systematic approaches should be taken to identify RNAs that are highly abundant in certain tumor types but not in blood, bone marrow or peripheral blood progenitor cells, in order to select more suitable genes as marker genes for epithelial tumors. The future of RNA-directed assays There are approximately 100,000 genes in the human genome which are expressed at different time points in different tissues. For the detection of minimal tumor cell contamination of blood, bone marrow or peripheral blood progenitor cells in breast cancer and other epithelial tumors it is essential to establish gene expression profiles of the tumor tissues and of blood or bone marrow of healthy volunteers. Using these expression profiles, genes can be identified which are exclusively and highly expressed in tumor tissue. These genes are potential marker genes for tumor cells provided that gene expression is absent in blood or bone marrow of healthy individuals using extremely sensitive and specific techniques such as RT-PCR or NASBA. It may well be that this approach to identify marker genes for tumor cells will be more successful than merely screening for the expression of (presumably) epithelial genes. Hopefully, this approach will allow the identification of a panel of marker genes optimally suited for both specific

6 1274 and sensitive detection of minimal tumor cell contamination. For the identification of genes, which are highly expressed in tumor tissues but are not expressed in blood or bone marrow, several molecular biological techniques have been developed recently. Differential display techniques may be used and some of these are already commercially available (e.g., Clontech). Due to the requirement that gene expression should be completely absent in blood- or bone marrow cells of healthy individuals, the sequential analysis of gene expression (SAGE) technique seems most suitable [54, 55]. Applying this technique one can generate a gene expression profile of the tissue under investigation by sequencing short sequence tags. These sequence tags can be identified using specific software in combination with the Genbank database. The expression profiles of different tissues can than be compared to each other. The specificity and sensitivity of this technique depends on the number of sequence tags that is analyzed. Also, the development of the DNA micro-array technology to generate expression profiles of different tissues is in rapid progress at this time [56, 57] and the first 'DNA chips' are becoming commercially available now. RNA-directed assays in clinical trials To establish the role of RNA-directed assays for tumor cells in the clinical setting, two requirements must be met: (1) standardization of techniques must have been achieved, together with a system for quality control, and (2) prospective clinical studies must be conducted that should optimally involve multiple centers and laboratories. Standardization should ensure that all laboratories and all workers in these laboratories employ precisely the same methods and reagents for a given test and quality control is required to continuously monitor the uniformity of methods and results over time and geographical distance. For the exquisitely sensitive techniques RT-PCR and NASBA, which are both vulnerable to the most minute of contaminations, this requirement is even more stringent than for other laboratory tests. Protocols of prospective studies should specifically state which assay(s) will be studied and which criteria will be employed to determine 'positivity'. It should also be stated what measures will be used in the quantification and what methodology will be employed to determine the predictive value of the assay(s). All these statements should be present in the protocol before the study is initiated and should be adhered to during the study and in the analysis of the results. All assays done should be included in the analysis: later exclusion of certain batches of tests is unacceptable. Employing these criteria, assays to detect tumor cells in blood, bone marrow or regional lymph nodes could be tested in large, ongoing multi-center trials not specifically designed for this purpose. For example, the significance for overall- and disease-free survival of the detection of cells expressing epithelial genes in peripheral blood progenitor cell (PCPC) collections after highdose therapy and PBPC transplantation in breast cancer could be tested in such a way. Conclusions As indicated, the detection of minimal tumor cell involvement in blood or bone marrow of patients with epithelial tumor may become critically important to determine the prognosis of newly diagnosed patients, and to predict whether systemic treatment modalities are likely to be required. Before such clinical use can be made of these new and sensitive techniques, the problem of false-positives must be addressed effectively. Doublefigure false-positive percentages are clearly unacceptable, and even occasional false-positive results would significantly impair clinical decision making. If the phenomenon of 'illegitimate transcription' is as universal as some authors believe, this problem may prove unsolvable. When technical advances, perhaps those envisaged in the previous paragraphs, have solved the false-positive problem, the assays with improved specificity must be tested in thousands of patients and in many centers. This will require rigorous standardization and quality control to avoid conflicting results due local laboratory condition and many technical factors. It is of interest, that this crucial standardization phase has not even been completed for the major competing technique, that of immunocytochemistry, despite the fact that this has been around for decades. Hence, accurate and easily replicable RNA-based tests for minimal tumor cell contamination continue to be desirable and seem to be the most appropriate choice for clinical application. Acknowledgements Supported by the Dutch Cancer Foundation (grant NKI ) and by a grant from Organon Teknika (Boxtel, The Netherlands). References 1. Mansi JL, Easton D, Berger U et al. Bone marrow micrometastases in primary breast cancer: Prognostic significance after six years" follow-up. Eur J Cancer 1991; 27: Cote RJ, Rosen PP, Lesser ML et al. Prediction of early relapse in patients with operable breast cancer by detection of occult bone marrow micrometastases. J Clin Oncol 1991; 9: Diel U, Kaufmann M. Goerner R et al. Detection of tumor cells in bone marrow of patients with primary breast cancer: A prognostic factor for distant metastasis. J Clin Oncol 1992: 10: Mansi JL, Berger U, Easton D et al. Micrometastases in bone marrow in patients with primary breast cancer: Evaluation as an early predictor of bone metastases. BMJ 1987; 295: Porro G, Menard S, Tagliabue E et al. Monoclonal antibody

7 1275 detection of carcinoma cells in bone marrow biopsy specimens from breast cancer patients. Cancer 1988: 61: Schlimok G, Funke I, Bock B et al. Epithelial tumor cells in bone marrow of patients with colorectal cancer: Immunocytochemical detection, phenotypic characterization, and prognostic significance. J Clin Oncol 1990: 8: Pantel K. Izbicki J, Passlick B et al. Frequency and prognostic significance of isolated tumour cells in bone marrow of patients with non-small-cell lung cancer without overt metastases. Lancet 1996; 347: Molino A, Pelosi G, Turazza M et al. Bone marrow micrometastases in 109 breast cancer patients: Correlations with clinical and pathological features and prognosis. Breast Cancer Res Treat 1997; 42: Salvadori B, Squicciarini P, Rovini D et al. Use of monoclonal antibody MBrl to detect micrometastases in bone marrow specimens of breast cancer patients. Eur J Cancer 1990; 26: Kirk SJ, Cooper GG, Hoper M et al. The prognostic significance of marrow micrometastases in women with early breast cancer. Eur J Surg Oncol 1990; 16: Jauch KW, Heiss MM, Gruetzner U et al. Prognostic significance of bone marrow micrometastases in patients with gastric cancer. J Chn Oncol 1996; 14: Schlimok G, Funke I, Pantel K et al. Micrometastatic tumour cells in bone marrow of patients with gastric cancer: Methodological aspects of detection and prognostic significance. Eur J Cancer 1991; 27: Lindemann F, Schlimok G, Dirschedl P et al. Prognostic significance of micrometastatic tumour cells in bone marrow of colorectal cancer patients. Lancet 1992; 340: Pantel K, Izbicki JR, Angstwurm M et al. Immunocytological detection of bone marrow micrometastasis in operable nonsmall-cell lung cancer. Cancer Res 1993; 53: Taylor-Papadimitriou J, Burchell J, Moss F et al. Monoclonal antibodies to epithelial membrane antigen and human milk fat globule mucin define epitopes expressed on other molecules. Lancet 1985; 1: Delsol G, Gatter ICC, Stein H et al. Human lymphoid cells express epithelial membrane antigen. Implications for diagnosis of human neoplasms. Lancet 1984; Pantel K, Schlimok G, Angstwurm M et al Methodological analysis of immunocytochemical screening for disseminated epithelial tumor cells in bone marrow. J Hematother 1994; 3: Rowley JD. Chromosome studies in the non-hodgkin's lymphomas: The role of the 14; 18 translocation. J Clin Oncol 1988; 6: Roth MS, Antin JH, Bingham EL et al. Detection of Philadelphia chromosome-positive cells by the polymerase chain reaction following bone marrow transplant for chronic myelogenous leukemia. Blood 1989; 74: Yokota S, Hansen-HaggeTE, Ludwig WD et al. Use of polymerase chain reactions to monitor minimal residual disease in acute lymphoblastic leukemia patients. Blood 1991; 77: Jung R, Ahmad-Nejad P, Wimmer M et al. Quality management and influential factors for the detection of single metastatic cancer cells by reverse transcriptase polymerase chain reaction. Eur J Clin Chem Clin Biochem 1997; 35: Foley KP, Leonard MW, Engel JD. Quantitation of RNA using the polymerase chain reaction. Trends Genet 1993; 9: Compton J. Nucleic acid sequence-based amplification. Nature 1991; 350: van der Vliet GM, Schukkink RA, van Gemen B et al. Nucleic acid sequence-based amplification (NASBA) for the identification of mycobacteria. J Gen Microbiol 1993; 139: van Gemen B, KievitsT, Nara Pet al. Qualitative and quantitative detection of HIV-1 RNA by nucleic acid sequence-based amplification. AIDS 1993; 7 (Suppl 2): S Sooknanan R, Malek L, Wang XH et al. Detection and direct sequence identification of BCR-ABL mrna in Ph+ chronic myeloid leukemia. Exp Hematol 1993; 21: Mori M. Mimori K. Inoue H et al. Detection of cancer micrometastases in lymph nodes by reverse transcriptase-polymerase chain reaction. Cancer Res 1995: 55: Noguchi S. Hiratsuka M. Furukawa H et al. Detection of gastric cancer micrometastases in lymph nodes by amplification of keratin 19 mrna with reverse transcriptase-polymerase chain reaction. Jpn J Cancer Res 1996: 87: Gunn J. McCall JL. Yun K et al. Detection of micrometastases in colorectal cancer patients by K19 and K20 reverse-transcription polymerase chain reaction. Lab Invest 1996; 75: Noguchi S, Aihara T, Motomura K et al. Histologic characteristics of breast cancers with occult lymph node metastases detected by keratin-19 mrna reverse transcriptase-polymerase chain reaction. Cancer 1996; 78: Noguchi S. Aihara T, Motomura K et al. Detection of breast cancer micrometastases in axillary lymph nodes by means of reverse transcriptase-polymerase chain reaction. Comparison between MUC1 mrna and keratin 19 mrna amplification. Am J Pathol 1996; 148: Schoenfeld A, Luqmani Y, Sinnett HD et al. Keratin-19 mrna measurement to detect micrometastases in lymph nodes in breast cancer patients. Br J Cancer 1996; 74: 1639^» Noguchi S, Aihara T, Nakamori S et al. The detection of breast carcinoma micrometastases in axillary lymph nodes by means of reverse transcriptase-polymerase chain reaction. Cancer 1994; 74: Schoenfeld A, Luqmani Y, Smith D et al Detection of breast cancer micrometastases in axillary lymph nodes by using polymerase chain reaction. Cancer Res 1994; 54: Mori M, Mimori K, Ueo H et al. Clinical significance of molecular detection of carcinoma cells in lymph nodes and peripheral blood by reverse transcription-polymerase chain reaction in patients with gastrointestinal or breast cancer. J Clin Oncol 1998; 16: Schoenfeld A, Kruger KH, Gomm J et al. The detection of micrometastases in the peripheral blood and bone marrow of patients with breast cancer using immunohistochemistry and reverse transcriptase polymerase chain reaction for keratin 19. Eur J Cancer 1997; 33: Fields KK. Elfenbein GJ,Trudeau WL et al. Clinical significance of bone marrow metastases as detected using the polymerase chain reaction in patients with breast cancer undergoing highdose chemotherapy and autologous bone marrow transplantation. J Clin Oncol 1996; 14: Krismann M, Todt B, Schroder J et al. Low specificity of cytokeratin 19 reverse transcriptase-polymerase chain reaction analyses for detection of hematogenous lung cancer dissemination. J Clin Oncol 1995; 13: Datta YH, Adams PT, Drobyski WR et al. Sensitive detection of occult breast cancer by the reverse-transcriptase polymerase chain reaction. J Clin Oncol 1994; 12: Moscinski LC, Trudeau WL, Fields KK et al. High-sensitivity detection of minimal residual breast carcinoma using the polymerase chain reaction and primers for cytokeratin-19. Diagn Mol Pathol 1996; 5: Kruger W, Krzizanowski C, Holweg M et al. Reverse transcriptase/polymerase chain reaction detection of cytokeratin-19 mrna in bone marrow and blood of breast cancer patients. J Cancer Res Clin Oncol 1996; 122: Burchill SA, Bradbury MF, Pittman K et al. Detection of epithelial cancer cells in peripheral blood by reverse transcriptasepolymerase chain reaction. Br J Cancer 1995; 71: Traweek ST, Liu J, Battifora H. Keratin gene expression in nonepithelial tissues. Detection with polymerase chain reaction. Am J Pathol 1993: 142: Mori M, Mimori K, Ueo H et al. Molecular detection of circulating solid carcinoma cells in the peripheral blood: The concept of early systemic disease. Int J Cancer 1996: 68: Gerhard M, Juhl H, Kalthoff H et al. Specific detection of carcinoembryonic antigen-expressing tumor cells in bone marrow

8 1276 aspirates by polymerase chain reaction. J Clin Oncol 1994, 12: Jonas S, Windeatt S, O-Boateng A et al. Identification of carcinoembryonic antigen-producing cells circulating in the blood of patients with colorectal carcinoma by reverse transcriptase polymerase chain reaction. Gut 1996; 39: Funaki NO. Tanaka J, Itami A et al. Detection of colorectal carcinoma cells in circulating peripheral blood by reverse transcription-polymerase chain reaction targeting cytokeratin-20 mrna. Life Sci 1997; 60: Luppi M, Morselli M. Bandien E et al. Sensitive detection of circulating breast cancer cells by reverse-transcriptase polymerase chain reaction of maspin gene. Ann Oncol 1996; 7: Helfrich W. ten Poele R, Meersma GJ et al. A quantitative reverse transcriptase polymerase chain reaction-based assay to detect carcinoma cells in peripheral blood. Br J Cancer 1997: 76: Mapara MY, Korner IJ, Hildebrandt M et al. Monitoring of tumor cell purging after highly efficient immunomagnetic selection of CD34 cells from leukapheresis products in breast cancer patients: Comparison of immunocytochemical tumor cell staining and reverse transcriptase-polymerase chain reaction. Blood 1997; 89: Chelly J, Concordet JP, Kaplan JC et al. Illegitimate transcription: Transcription of any gene in any cell type. Proc Natl Acad Sci USA 1989; 86: Chelly J, Kaplan JC, Maire P et al. Transcription of the dystrophin gene in human muscle and non-muscle tissue. Nature 1988; 333: Humphries S, Windass J, Williamson R. Mouse globin gene expression in erythroid and non-erythroid tissues. Cell 1976; 7: Zhang L, Zhou W, Velculescu VE et al. Gene expression profiles in normal and cancer cells. Science 1997; 276: Velculescu VE, Zhang L. Vogelstein B, Kinzler KW. Serial analysis of gene expression. Science 1995; 270: Southern EM. DNA chips: Analysing sequence by hybridization to oligonucleotides on a large scale. Trends Genet 1996; 12: Chee M, Yang R. Hubbell E et al. Accessing genetic information with high-density DNA arrays. Science 1996; 274: Neumaier M, Gerhard M, Wagener C. Diagnosis of micrometastases by the amplification of tissue-specific genes. Gene 1995; 159: Zippelius A, Kufer P, Honold G et al. Limitations of reversetranscriptase polymerase chain reaction analyses for detection of micrometastatic epithelial cancer cells in bone marrow. J Clin Oncol 1997; 15: Novaes M, Bendit I, Garicochea B et al. Reverse transcriptasepolymerase chain reaction analysis of cytokeratin-19 expression in the peripheral blood mononuclear cells of normal female donors. J Clin Pathol (Mol Pathol) 1997; 50: Received 12 May 1998; accepted 3 September Correspondence to- A. C. Lambrechts. MD Division of Experimental Therapy The Netherlands Cancer Institute Plesmanlaan CX Amsterdam The Netherlands