review nature genetics supplement • volume 21 • january 1999 25 The use of DNA microarrays for comprehensive RNA expression analysis has caused a great deal of interest recently, although the concept is not new 1,2 . Technical developments that offer increased sensitivity, the prospect that all genes for a given organ- ism could soon be scrutinized in this way and a general apprecia- tion of the need to integrate information obtained from more traditional and reductionist approaches to biology make micro- array-based expression analysis a powerful tool 3–13 . The components of a complete system can be divided into three parts (Fig. 1), involving sample preparation (which I desig- nate ‘the front end’), array generation and sample analysis (or ‘middleware’), and data handling and interpretation (the ‘back end’). Component emphasis depends heavily on the questions an investigator wishes to address. For example, there are many ‘front end’ considerations involving tissue acquisition and processing that are relevant to a researcher interested in prostate cancer but are largely irrelevant to one seeking to dissect signal transduction in established cell lines. Similarly, many ‘middleware’ issues con- cerning slide and filter generation will be irrelevant to users who opt for commercially available arrays. Despite this, many users may want to assemble all of the components involveda difficult but not impossible undertaking. The front end from sample to RNA RNA preparation from cell lines is simple and straightforward. Use of tissues and a need to microdissect adds a layer of complex- ity, and dealing with human tissues adds many more. Ethical considerations. Collection of tissue removed at surgery usually requires approval from an institutional ethics com- mittee and informed consent from individuals. In obtaining con- sent for work with tumour material, germline analysis is obviously more contentious than somatic analysis. This distinction is blurred, however, if RNA expression changes found in tumour material are characteristic of certain mutations. For example, if changes observed in breast cancer samples point to a BRCA1 muta- tion, such an analysis could be considered to be a surrogate germline test, given the low incidence of somatic mutations in this gene. Researchers performing RNA microarray analysis with human material should therefore consider duty of care, and processes for follow-up of patients, possibly including genetic counselling. Another critical ethical consideration is whether the collection of tissue impacts on diagnostic procedures. For example, determining whether clear excision margins are obtained in cancer surgery is very important but may be compromised if material is removed for research purposes. Samples obtained from primary, early stage tumours hold great value for gaining an understanding of the molecular progression of disease. Early-stage malignancies, however, may be subject to more extensive pathological scrutiny for staging than end-stage disease in which the diagnosis has been made previously. In addition, early-stage degenerative or malig- nant disease is often less apparent clinically and less accessible than late-stage or bulky disease, where surgical intervention is much more likely. Issues such as these represent logistical challenges to collecting material of high quality. Finding ways in which samples can be obtained without compromising diagnostic imperatives requires close co-operation with both surgeons and pathologists. A discussion of some of the ethical issues associated with human tis- sues is available (http://209.143.140.244/napbc/tissue.htm, http:// 209.143.140.244/napbc/irb%20review.htm, http://bioethics.gov/ cgi-bin/bioeth_counter.pl and http://www.health.gov.au/nhmrc/ ethics/consult.htm), as well as model consent forms for collecting samples from cancer patients (http://www.pmci.unimelb.edu.au/ tissforms/ and http://bioethics.gov/briefings/jan98/model.pdf). Tissue banks. Although large numbers of archival samples are available in many clinical departments, often the samples are sub- optimal with respect to RNA integrity, fixation, or critical patient information. The establishment of suitable tissue banks is a logi- cal adjunct to any in-depth RNA analysis of human tissue; reposi- tories must address issues of appropriate collection and storage and also ensure that the samples are accompanied by appropriate patient information, including treatment, outcome, epidemiolog- ical and family history data. The National Cancer Institute (NCI) coordinates a centralized tumour bank for North American researchers (http://www-chtn.ims.nci.nih.gov/). Commercial tis- sue banks, such as LifeSpan BioSciences, also provide access to a wide variety of human disease tissues (http://www.lsbio.com/). Microdissection. Diseased tissue contains a mixture of normal tissue, inflammatory cells, necrotic tissue and, in cancer samples, areas of different grade. Similarly, healthy tissue also includes a range of cell types. All of these elements can combine to produce a complex RNA expression profile. Microdissection capability is thus critical for microarray studies involving tissues (see page 38 of this issue (ref. 14)) and is also useful for associated technologies such as comparative genomic hybridization (ref. 15). Current protocols for fluorescent labelling of RNA demand large quantities of RNA, which impedes the use of microdissected RNA on GeneChip® and glass slide arrays. Laser-based microdissec- tion 16–18 offers a means of more rapidly obtaining pure material than conventional techniques. The commercially available laser capture microdissection microscope (http://www.arctur.com) is The excitement surrounding microarray technology has been tempered by the limited ability of the general biomedical research community to gain access to it. Given that the hardware required for exploitation of the technology is becoming increasingly available, it is an appropriate moment to review options, be they commercially or publically available. Here, we provide a snapshot of the rapidly changing field of microarray-based RNA expression analysis and consider the components and procedures for putting together a complete system. Options available — from start to finish — for obtaining expression data by microarray David D.L. Bowtell Peter MacCallum Cancer Institute, Locked Bag 1, A’Beckett St. Melbourne 3000, Victoria, Australia. e-mail: d.bowtell@pmci.unimelb.edu.au © 1999 Nature America Inc. • http://genetics.nature.com © 1999 Nature America Inc. • http://genetics.nature.com