VI. DNA microarray technology. Principles and application to the analysis of malignant tumours (with special emphasis on lymphoma)

Abstract

The human genome sequence is now available, virtually complete. Why should oncologists bother about this data bank? Until now most research in molecular carcinogenesis has been directed at discovering and characterizing single ‘cancer’ genes. Virtually all established diagnostic techniques in molecular cancer pathology suffer from the limitation that they only tell us what we are specifically looking for. It may well be that a chosen molecular marker may be less relevant than another one which was not ordered by the clinician, or was not considered by the pathologist. Colon cancer is a nice example to illustrate this issue. For example, patients with stage III colon cancer seem to derive more benefit from adjuvant therapy if their tumours retain a wild-type K-ras sequence [1, 2]. Interestingly, colon cancer with a wild-type K-ras status will also derive particular benefit from epidermal growth factor receptor (EGFR)-targeted treatment. Colon cancers with microsatellite instability may be less aggressive than tumours with stable microsatellites. Retention of 18q alleles in microsatellite-stable tumours, and mutations of the gene for type II TGF-b1 receptor in node-positive microsatellite-instable colon cancers both point to a more favourable prognosis after adjuvant chemotherapy. A glance at this literature shows that in most papers the molecular marker of interest was carefully studied, and one or the other additional gene included in the analysis, but no such study provided an overall appraisal of all molecular markers of potential clinical value in this cancer. The established molecular diagnostic techniques are mostly too laborious to permit the comprehensive screening of a tumour biopsy sample for all possible types of genetic marker. A new approach would be to screen cancer specimens for all possible ‘gene’ problems, i.e. to obtain individual comprehensive cancer gene expression profiles, at the RNA expression level (also known as ‘signatures’). This is now possible with microarray gene profiling [3, 4]. In contrast to the study of single genes and their proteins, molecular tumour profiling is a large-scale analysis of gene expression in a tumour using DNA microarrays (Figure 1). DNA microarrays typically consist of rows and rows of oligonucleotide sequence strands, or cDNA sequences lined up in dots on a silicon chip or glass slide [3–6]. Oligonucleotide sequences or cDNAs on the chip permit specific hybridization to labelled mRNAs of interest extracted from a biopsy. Arrays can accommodate up to 30 000 specific sequences on a single chip, either chosen randomly or deliberately ‘biased’ to represent theme parks of genes typically expressed in a cell type of interest, e.g. ‘lymphoid genes’ in B cells (‘Lymphochip’) [6]. The Lymphochip is a cDNA microarray containing selected genes preferentially expressed in lymphoid cells [7, 8]. Analysis of gene expression in various lymphoid malignancies yields an orderly picture of gene expression patterns in particular types of lymphoma, reflecting lineage characteristics, stage of maturation of lymphoid cells and proliferation signatures. Diffuse large B-cell lymphoma (DLBCL, a clinically heterogeneous group of lymphomas despite their morphological similarity) can be split into subtypes with gene expression profiles typical either of germinal centre B cells, or of activated B cells. DLBCL expression signatures differ markedly between patients who were cured, and those who eventually relapsed [8]. The promise is that such expression profiles or ‘signatures’ offer more precise prognostic information than established prognostic factors, such as the International Prognostic Index in NHL (Figure 2), and eventually translate into concepts of refined differentially targeted therapy. Likewise, one invasive ductal breast cancer specimen may look deceptively similar to another one on histology, but the fate of the two women may be totally different. This is due to inherent biological differences of the two tumours hidden in their genome, which may be elusive to morphological examination. Variation in gene transcription programmes governed by specific somatic gene alterations accounts for much of the biological diversity in human tumours. The study of gene expression patterns in human breast cancer specimens displays distinct molecular portraits, or gene expression profiles providing molecular ‘fingerprints’ [9]. Tumours may be clustered by sharing gene expression patterns, and it is likely that such subgroups comprise clinically distinct subtypes or entities of breast cancer. It also turns out that the overall gene expression pattern of a breast cancer case is by and large retained in its metastases. Figure 2 shows an analysis where T1– 2 N0 tumours that had or had not relapsed within 5 years after diagnosis and primary treatment, show clearly distinct gene expression profiles, respectively. Breast cancers of the basal-like cell type, which often express neither hormone receptors nor