Copyright © 2006, Elsevier Inc.

All rights reserved.

figure 5.1 Schematic representation of a DNA microarray hybridization comparing gene expression of a malignant epithelial cancer with its normal tissue counterpart. (Journal of Pathology, Alizadeh et al. (2001). Copyright John Wiley and Sons Ltd. Reproduced with permission.) See Plate 5.1 in Color Plate Section.

figure 5.1 Schematic representation of a DNA microarray hybridization comparing gene expression of a malignant epithelial cancer with its normal tissue counterpart. (Journal of Pathology, Alizadeh et al. (2001). Copyright John Wiley and Sons Ltd. Reproduced with permission.) See Plate 5.1 in Color Plate Section.

Most microarray systems and methods share the same general principles, but may differ in the specific details. Basically, microarray probes tethered to an immobile surface are exposed to nucleic acid targets that are generated from RNA of the sample of interest (see Fig. 5.1). The signal generated by complementary binding of target to specific microarray probes is proportional to the level of RNA expression in the sample. Microarray probes often take the form of specific cDNA sequences, which are obtained and amplified by PCR, or they may be oligonucleotides, which are smaller than cDNAs. More recently, peptide nucleic acids (PNAs) have also been used. One advantage to the smaller oligonucleotide and PNA sequences is that they can be tiled at higher density allowing a larger number of probes per array. They also do not require PCR, which can introduce errors into the probe set. There is some data to suggest that PNA chips may also have a higher target affinity [15] and, therefore, provide more specific results.

The microarrays are produced by deposition and immobilization of selected probes onto a matrix. A number of matrix substrates have been utilized, including glass, silicon, nylon, and nitrocellulose. Glass is often selected, because it is nonporous and allows covalent attachment of probes. A computer-aided robot spots a sample of each gene product, with up to 10 000 cDNA probes/spots per cm2 (Fig. 5.2).

figure 5.2 An example of an oligonucleotide-based DNA microarray is seen in this Affymetrix GeneChip array. See Plate 5.2 in Color Plate Section.

For cDNA arrays, fluorescently labeled target is generated from RNA derived from test and control samples by a single round of reverse transcription. These targets are then mixed and hybridized with the arrayed DNA spots. Image analysis of the micro-array is utilized to assess the relative amount of the targets hybridized to each probe, indicative of either increased or decreased levels of gene expression in the test relative to the control sample [16,17].

Oligonucleotide probes may be synthesized in situ on the substrate, and may consist of up to 250 000 oligonucleotide probes/spots per cm2. The arrays are hybridized with target from a single RNA sample, using differential hybridization of exact match oligo-nucleotide sequences with oligonucleotide probes that have a single base pair substitution to control for non-specific binding. The signal from exact match probes is then quantified as a measure of gene expression level.

A large number of data points are generated from each microarray experiment; all array methods require database management in order to compare the outputs of single and multiple array experiments using a number of data mining tools. These allow correlation of data into meaningful groups, which permit new hypotheses regarding possible cellular pathways and clues to disease pathophysiology. The correct interpretation of these immense datasets that result from multiple samples and from many patients in complex systems presents a major challenge. There is a clear need to prioritize the selection and investigation of candidate genes for further study.

Computational methods are used to help redefine relevant groups of tumors and genes; this hierarchical clustering can be used to identify groups of tumors with similar gene-expression profiles. These defined groups can subsequently be probed for other outcome parameters, such as survival and biologic or histologic activity [18]. Typically, interpretation of microarray data requires assistance from bioinfor-matics specialists familiar with the analysis of these very large datasets, and many resources are available on the internet to assist in this effort.

The expense of microarray experiments has decreased over time, due to the declining cost of commercial microarrays and the availability of core microarray facilities in many institutions. Moreover, studies can now be performed on a genome-wide basis for many different species as the completed genome sequences for a variety of organisms have permitted the development of DNA microarrays of model systems such as Mycobacterium tuberculosis, Escherichia coli, Drosophila melanogaster, Caenorhabditis elegans, and Plasmodium falciparum. As a result of the Human Genome Project, commercially prepared slides are now available to probe the human genome as well.

Limitations on these methods primarily rest on the availability of appropriate RNA samples that are derived from tissues of clinical interest. For brain tumors, handling of biopsy tissue currently is geared towards histopathologic examination to establish diagnosis and staging; methods of fixation and tissue embedding were not developed with micro-array-based studies in mind. RNA is relatively unstable; therefore, tumor samples used for DNA microarrays must use RNA extracted from fresh or flash-frozen tissue, which is collected in liquid nitrogen or in isopentane on dry ice. RNA which has been extracted from paraffin-embedded tissue is, in general, too fragmented for use on arrays [19].

Finally, most tumor specimens are heterogeneous, in that they contain an intermixture of normal tissue, inflammatory cells, and connective tissue, in addition to malignant cells. Analysis of homogenized tissue from this sample will generate a mix of mRNA derived from malignant tissue in addition to these normal host cell types. Validation studies are needed to separate the gene expression signals of malignant from normal cell types.

Was this article helpful?

0 0

Post a comment