The first DNA arrays used for research work were oligonucleotide-based printed microarrays. These microarrays require either PCR amplified (ds-DNA) or synthesized oligonucleotides (ss-DNA) of known gene sequences as probes. In these microarrays, glass slides are used as solid surface for spotting or printing of the probes. Cheung et al. (1999) observed that glass slides are not only observed to be stable during the varied experimental situations but also produce least background noise during hybridization, thus yielding clearer images. Probes can be printed in two ways by contact printing or by noncontact printing. Whichever the method of printing is used in both the cases an array of 100–150 μm features is created by applying few nanolitres of probe solution per spot. Contact printing means direct application of probe solution on the solid surface with the help of print pins, whereas in noncontact printing a small droplets of probe solution is placed on the solid surface by the print head. Printing is done precisely so as to avoid mixing of adjacent features. In comparison to GeneChip and high density bead arrays, oligonucleotide-based printed arrays have low density of probes ranging from 10,000 to 30,000 features per array due to the relatively large size of the features in it but has more features than suspension bead arrays and electronic microarrays.

When dsDNA is used as probes, it is required to denature them either at the time of printing or after spotting on the solid surface so that they become available for the hybridization (Tomiuk and Hofmann 2001). For getting the probes attached on the glass surface, two methods are used, first, either slides are applied with a positively charged coating that allows the electrostatic interaction with the negative charge of the phosphate backbone of the DNA (Ehrenreich 2006) and second, by UV-cross-linking of amine groups present on treated slides with the covalent bonds present between the thymidine bases in the DNA (Cheung et al. 1999). Each dsDNA probe used in these arrays represents a different gene and is usually 200–800 bp long. These probes should have high specificity and should be free from contaminants so that there is no interference while hybridization. Generally, these probes have a high sensitivity but they suffer in specificity and might be contaminated with nonspecific amplicons (Hager 2006). To select for the probe specificity, amplicons can be tested by sequencing or agarose gel electrophoresis and specificity of the hybridization data can be improved by including multiple gene segments among the probes so as to workout the redundancy. At times decreased specificity is advantageous when the genomic sequence under analysis has high natural polymorphism, although it does not help when we have to differentiate among highly similar target sequences, therefore they are not useful as far as the clinical diagnostics is concerned. At the same time, use of longer probes results in higher melting temperatures and mismatch tolerance that decreases the specificity of the probes due to random hybridization to nontarget sequences, hence for any microarray experiment probe length should be chosen carefully.

In oligonucleotide microarrays with ssDNA probes, probes are chemically synthesized short sequences usually ranging from 25–80 bp in length, in contrast to gene expression microarrays where they may be up to a length of 150 bp (Chou et al. 2004). In these microarrays due to the shorter length of probes, fewer errors are observed during probe synthesis and they help in studies focusing small genomic regions like polymorphisms. Another benefit due to the small probe length is increased specificity while targeting specific genomic regions although it reduces the sensitivity of detections. It is observed that the sensitivity and the strength of hybridization signal increases with the length of the probe used. In case of very short oligonucleotides, addition of spacers or application of a higher concentration of probe during printing improves the strength of hybridization signal (Chou et al. 2004). Ramdas et al. (2004) observed eightfold increase in sensitivity when 70 mer oligonucleotides were used as probes as compared to 30 mer probes for the genes having low levels of expression. ssDNA oligonucleotide probes are easy to manufacture but care is required that all the probes on the array should have melting temperature lying within a range of 5 °C and should not have palindromic sequences. Hence, it is required to test the probes in advance so that the hybridization data generated through any experiment using ssDNA probes remain unbiased (Chou et al. 2004). Attachment of ssDNA probes to glass surface is facilitated due to covalent linkage between modified 5′ or 3′ ends having amino group with the aldehyde or epoxy functional groups coated on the slides. Oligonucleotide microarrays manufactured by Roche NimbleGen (Madison, WI) and Agilent Technologies (Palo Alto, CA) are good examples. Each NimbleGen microarray contain >10⁶ features. The available formats are 1 × 2.1 million features, 3 × 720,000 features, 1 × 385,000 features, 4 × 72,000 features, and 12 × 135,000 features per slide. Agilent microarrays are available in the formats having 1 × 244,000 features, 2 × 105,000 features, 4 × 44,000 features, and 8 × 15,000 features per slide (Miller and Tang 2009). An oligo microarray with approximately 40,000 probes is shown in Fig. 6.3 to have an idea of how it looks after hybridization.

GeneChips are produced by in situ synthesis of oligonucleotides directly onto quartz or glass wafers having a size of 1.2 cm². Tens of thousands of oligonucleotides are placed on one wafer by synthesizing one nucleotide at a time. Each oligonucleotide probe on the array represents one gene. Length of the probes used in these arrays depends on the purpose for which these microarrays are designed. GeneChips designed by Agilent technology use longer probes having 60 bp lengths, whereas in Affymatrix GeneChips shorter probes having a length of 25 bp are used. Typically, GeneChips have >10⁶ features per microarray that actually depends on the distance between the features (Weiszhausz et al. 2006).

For binding of the in situ synthesized probes to the solid surface mainly a modification of semiconductor photolithography technology is used. In this technique, quartz surface of the GeneChip is coated with synthetic linkers modified with light-sensitive protecting groups (Fodor et al. 1991). Due to this chemical protection, addition of nucleotide to the microarray surface is restricted till its deprotection by UV light. This means nucleotides modified with a photosensitive protecting group can be added to growing oligonucleotide chains by exposing the array surface to UV light. Further, photolithographic masks are used for directing the specific nucleotides to exact probe sites. Photolithographic mask has a defined pattern of windows; those selectively transmit or block the UV light for the targeted feature on the microarray surface that is chemically protected. Due to this kind of filtration, an area of the microarray surface that does not receive UV light remains protected and does not allow addition of nucleotides, whereas areas exposed to UV light gets deprotected, and specific nucleotides can be added to these sites. Likewise, the order of nucleotide addition is directed by the pattern of windows in each photolithographic mask. Therefore, in situ probe synthesis requires cycles of masking, UV exposure, and addition of nucleotide bases (either A, C, T, or G) to the growing oligonucleotide chain on the solid surface (Fig. 6.4) (Weiszhausz et al. 2006). In addition to array synthesis using photolithography mask NimbleGen Systems has developed a technique of maskless array synthesis that provides flexibility of use with large number of probes required in in situ synthesized oligonucleotide arrays to improve sensitivity, specificity, and statistical accuracy due to shorter probe lengths.

The main advantage of GeneChips is their ability to measure the absolute expression of genes in cells or tissues. Their disadvantages are their higher costs and current inability to simultaneously compare, on the same array, the degree of expression of two related biological samples.

For expression analysis using oligonucleotide microarray hybridization (Fig. 6.5), firstly labeled RNA samples are prepared by converting extracted mRNA to double-stranded cDNA. The cDNA is then copied to antisense RNA (cRNA) by in vitro transcription reaction performed in presence of biotin-labeled ribonucleotide triphosphates (UTP or CTP). Fragmented cRNA (50–100 nucleotides) is used for hybridization. After a brief washing step to remove unhybridized cRNA, the microarrays are stained by streptavidin phycoerythrin and scanned (Aharoni and Vorst 2001).

Both types of oligonucleotide microarrays produce cleaner downstream hybridization data. Reproducibility, simpler standardization, and use of controls make these microarrays useful for clinical diagnostics. In comparison to in situ-synthesized microarrays, printed microarray seems to be relatively simple and inexpensive with flexibility to quickly adjust spotted probes on the basis of newer informations. However, in situ-synthesized microarrays are more robust due to significant control measures included in their design.

Now oligonucleotide arrays have also been developed which combine the flexibilities and qualitative advantages of synthetic probe arrays with the benefits of simultaneous analysis possible by spotted glass array. Further, custom-designed microarrays have also been made available by Nimblegen and Agilent technologies.

In high-density bead arrays silica beads of 3 μm size are used as solid support for application of probes. After hybridization, these beads can randomly self-assemble on the substrate. Two substrates available for this purpose are Sentrix Array Matrix (SAM) and the Sentrix BeadChip. These arrays facilitate high density detection of target nucleic acids. One of the examples is Illumina bead arrays, San Diego, CA.

The SAM contains a bundle of 96 fiber-optic cables of 1.4 mm diameter. Each bundle represents an array made up of 50,000 light-conducting fibers of 5 μm diameter; each of this fiber has a microwell for assembling a single bead (Fan et al. 2006). For example, in the Universal Bead array, up to 1,536 bead types can be accommodated. Here, every bead type represents a unique gene sequence and when they assemble on the fiber bundles 30 beads of each type becomes available for analysis in the array (Oliphant et al. 2002). Likewise, each SAM allows the analysis of 96 samples.

In Sentrix BeadChip up to 16 samples can be accommodated on a silicon slide that has microwells prepared for individual beads by microelectromechanical systems technology (Fan et al. 2005). Gunderson (2009) has found BeadChips more appropriate for very-high-density applications like whole-genome genotyping where 10⁵–10⁶ features are required for identifying genome-wide single nucleotide polymorphisms (SNP).

Since the beads randomly assemble on the substrate, their location is not fixed and it is required to map their position by decoding process after every single application (Gunderson et al. 2004). For the decoding purpose, each bead type is covalently attached with 700,000 copies of a unique capture sequence that acts as the identifier of the bead (Kuhn et al. 2004). For example, in Universal BeadArrays such identifier sequences are called as IllumiCodes and they are designed to specifically avoid homology with human and mouse nucleic acid sequences. The mapping of these Illumina beads requires a series of hybridization and rinse steps, so that the fluorescently labeled complementary oligonucleotides may bind to their specific bead sequence or IllumiCode and ultimately track the location of the corresponding bead type (Fan et al. 2006). This decoding process provides quality control for each feature used in the microarray (Fan et al. 2005).

It is possible to process the SAM by using a standard microtiter plate that facilitates standard automation and high-throughput processing. Similarly, individual arrays on the 16-sample BeadChip are spaced as the standard multichannel pipettor for easy handling while use. 10⁵–10⁶ features can be supported by Bead arrays. One advantage of bead arrays is their built-in redundancy that provides control for comparing data generated by more than one arrays. Further, due to the uniqueness of each microarray by altering the bead pattern spatial bias can be identified. The work on developing the analysis tools for bead arrays is in progress (Miller and Tang 2009). Bead arrays are found very useful in studies related to DNA methylation (Bibikova and Fan 2009), profiling of gene expressions (Bibikova et al. 2004; Fan et al. 2004; Kuhn et al. 2004), SNP detection, and in the International HapMap Project (Gunderson 2009).

In suspension bead arrays, microscopic polystyrene spheres, i.e., microspheres or beads are used as solid support for application of probes and flow cytometry technique is used for detection of bead and target attached to it. In these arrays, beads remain suspended in the hybridization solution and do not require fiber-optic cables or silicon slides to immobilize them before scanning. All the previously discussed microarrays were two-dimensional, or planar arrays whereas, suspension bead arrays are three-dimensional arrays.

These arrays were first described by Horan and Wheeless (1977) for detections involving antigen and antibodies. Further, it was made possible to detect multiple antibodies in one run by using different sized microsphere sets (Scillian et al. 1989; McHugh et al. 1998). In suspension bead arrays if 5.6 μm microspheres are filled with different concentrations of red (658 nm emission) and infrared (712 nm emission) fluorochromes, each bead in the set of 100 microspheres will have a distinct red to infrared ratio providing each bead a unique spectral address. This provision has strengthened multiple detections using suspension bead arrays. When, microspheres with a specific spectral address is coupled with a specific probe, it becomes equivalent to a feature of a two-dimensional microarray but when multiple microspheres with unique spectral addresses are combined with multiple specific probes it is possible to develop up to 100 combinations to study the target nucleic acids. Along with unique spectral address and specific probe suspension beads are also provided with a fluorescent reporter that helps in detecting hybridization of target with the probe. Here, when hybridization takes place reporter separates to emit fluorescence that can be detected by a flow cytometer. During flow cytometry, microsphere suspension is passed through two lasers, i.e., 635 and 532 nm. A 635 nm laser excites red and infrared fluorochromes and helps in detecting the unique spectral address of beads and hence the identity of probe and target can be worked out, whereas, 532 nm laser excites reporter fluorochromes such as R-phycoerythrin and Alexa 532 and helps in quantifying the amount of hybridization taken place on individual bead (Fig. 6.6).

Suspension bead arrays are used in following three ways for nucleic acid detections (Dunbar 2006):

Despite the low feature density due to the availability of universal capture sequences, high ability of multiplexing, relative simplicity, and affordability suspension bead arrays are highly suitable for high-throughput nucleic acid detections including clinical diagnostics if due care is taken to avoid post-amplification intra- and inter-run contaminations.

SNP’s are common DNA sequence variations having great significance for biomedical research. In human genome at least one out of every thousand nucleotide is observed to have polymorphism. Most of the SNP’s are harmless but few of them that are located in coding regions of the gene alter the composition of amino acids, thus the functions of encoded proteins. Similarly, SNP’s located in the noncoding regions are also important if they affect the splicing of the gene, or alter the promoter, enhancer, silencer or any recognition site for DNA binding proteins (Risch 2000). For example, a SNP in 5′ untranslated region of the RAD51 gene have been identified that may be associated with an increased risk of breast cancer and a lower risk of ovarian cancer among BRCA2 mutation carriers (Wang et al. 2001).

High throughput methods like microarray assays are found useful in rapid detection of mutations among large genes like BRCA1, BRCA2. For example, a high density oligonucleotide array with about 96,000 oligonucleotides of 20 bp size have been developed by Hacia et al. (1996) that could detect a wide range of heterozygous mutations in the 3.45 Kb exon 11 of the BRCA1 gene. Wen et al. (2000) found that oligonucleotide based detections are more accurate than DNA sequencing in identifying p53 mutations in ovarian tumors.

Improvements in tumor classification are central to the development of novel and individualized therapeutic approaches. Histologically indistinguishable tumors often show significant differences in clinical behavior, and sub classification of these tumors based on their molecular profiles may help explain why these tumors respond so differently to treatment (Macgregor and Squire 2002).

For example, (1) microarray technology was applied to develop innovative classifications of leukemias using microarray analysis based on neighborhood analysis and the utilization of tumor class predictors. This strategy was able to distinguish between acute myeloid leukemia and acute lymphocytic leukemia without supervisory analysis (Golub et al. 1999); (2) gene expression pattern analysis has been used to classify, at the molecular level, breast tumors (Sorlie et al. 2001), B cell lymphoma (Alizadeh et al. 2000), cutaneous melanoma (Bittner et al. 2000), and lung adenocarcinoma (Garber et al. 2001; Bhattacharjee et al. 2001); (3) by analyzing molecular profiles of 50 nonneoplastic and neoplastic prostate samples, signature expression profiles of healthy prostate, benign prostatic neoplasia, localized prostate cancer, and metastatic prostate cancer has been established (Dhanasekaran et al. 2001). These studies established the feasibility of combining large-scale molecular analysis of expression profiles with classic morphologic and clinical methods of staging and grading cancer for better diagnosis and outcome prediction.

Many tumor suppressor genes and oncogenes are regulated by the expression of other genes at transcription level. Even some of them like p53 and Myc are transcription factors. Upon activation, p53 induces growth arrest or apoptosis by transcriptionally activating its target genes. With the help of microarray-based expression profiling target genes for several gene products have been identified which directly or indirectly regulates transcription. For example, oligonucleotide array having 6,000 genes is used to identify p53 target genes (Zhao et al. 2000). Similarly, microarray-based expression profiling have been used to identify Gadd45 as one of the targets of BRCA1 tumor suppressor gene (Harkin et al. 1999).

Despite considerable advances in cancer treatment, acquired resistance to chemotherapeutic drugs continues to be a major obstacle in patient treatment and overall outcome. Anticancer drug resistance is thought to occur through numerous mechanisms, and microarrays offer a new approach to studying the cellular pathways implicated in these mechanisms and in predicting drug sensitivity and unexpected side effects. Most array studies have been carried out using cancer cell lines that are rendered resistant to commonly used anticancer drugs. Obtaining further insights into the mechanism of action of anticancer drugs and the diverse pathways involved in drug resistance may eventually be invaluable for design of more strategic treatments that are most appropriate for an individual tumor (Macgregor and Squire 2002).

For example, (1) in an attempt to obtain molecular fingerprinting of anticancer drugs in cancer cells the expression profiles of doxorubicin-induced and -resistant cancer cells were monitored (Kudoh et al. 2000); (2) similarly, in another study, a subset of 1,400 genes were analyzed to study the correlation between expression profiles and drug mechanism of action of a panel of 118 anticancer drugs (Ross et al. 2000; Scherf et al. 2000).

Conversion of noninvasive tumors into invasive tumors is known as metastasis when cancer cells start spreading from one organ to another organ or tissue with blood stream or lymphatic system. DNA microarrays have been used to identify the genes involved in metastasis.

For example, clonal relationship of 22 liver tumor foci from six patients have been investigated by Cheung et al. (2002) with the help of cDNA microarray having 23,000 genes. With an aim to identify metastasis-related genes, they could identify 63 up-regulated genes, of which 39 were known genes and 24 were expressed sequence tags and 27 down regulated genes, of which 14 were known genes and 13 were expressed sequence tags. Similar studies have been conducted with osteosarcoma, colorectal tumor and brain metastasis.

Several research groups have focused on identifying subsets of genes that show differential expression between healthy tissues or cell lines and their tumor counterparts to identify biomarkers for several solid tumors, including ovarian carcinomas, oral cancer, melanoma, colorectal cancer, and prostate cancer.

In a recent study, a subset of genes showing differential expression between healthy ovaries and ovarian tumors has been reported. Some of these genes, such as metallothionein 1G, which was found to be up-regulated in tumor samples, are implicated in resistance to the anticancer drug cisplatin and might be an indicator of pretreatment resistance of these tumors to cisplatin (Bayani et al. 2002). Other gene identified in this study is the osteopontin gene, which was strongly up-regulated in some tumor samples that have been shown to be secreted in the serum of patients with metastatic cancer, might be an excellent candidate for biomarkers of tumor progression in EOC (Singhal et al. 1997).

One of the most important challenges investigators are facing while using microarray analysis is determining which of the plethora of new differentially expressed genes is biologically relevant to the tumor system being studied. Even when rigorous efforts are made to minimize the number of variables in a microarray study, there may be an unmanageable number of differentially expressed genes that will contribute excessive background values. Therefore, combining expression microarray analysis with other approaches, particularly cytogenetics techniques, such as spectral karyotyping and comparative genomic hybridization array (CGH) (Pollack et al. 1999), offers the possibility to focus on significantly smaller subsets of genes of direct relevance to tumor biology (Bayani et al. 2002). A combination of expression arrays and CGH array techniques was used on breast cancer cell lines and has identified a limited number of genes that are both amplified and over expressed (Monni et al. 2001).

Finally, validation of the relative expression obtained from genome-wide microarray analysis is critical. Several approaches can be chosen from basic Northern analysis or semiquantitative reverse transcription-PCR to in situ hybridization (ISH) using tissue microarrays. For example, expression of several candidate genes associated with prostate cancer has been analyzed (Mousses et al. 2002) those were previously identified by cDNA microarray analysis. Tissue microarrays constructed from 544 histological biopsies were analyzed by ISH using RNA probes and/or by immunohistochemistry (IHC) using antibodies. There was excellent correlation between the cDNA microarray results and the results obtained with ISH and Northern blot analysis. In addition, protein expression assessed by IHC was also consistent with RNA expression. Similarly, comparable technologies to confirm over expression of hepsin and PIM-1 in prostate cancer have been used (Dhanasekaran et al. 2001).

DNA microarrays could potentially be an assay method to address multiple questions for species identification in both clinical and environmental settings. It identifies their phylogenetic status based on unique 16S rRNA sequences and provides information related to the presence of antibiotic markers and pathogenicity regions (Ye et al. 2001).

For example, (1) a DNA probe array have been described for species identification and detection of rifampin resistance in M. tuberculosis (Troesch et al. 1999). In this 70 mycobacterial isolates from 27 different species and 15 rifampin-resistant strains were tested. A total of 26 of the 27 species as well as all of the rifampin-resistant mutants were correctly identified. (2) For field applications, a portable system for microbial sample preparation and oligonucleotide microarray analysis has also been reported (Bavykin et al. 2001). This portable system contained three components: (a) a universal silica mini-column for successive DNA and RNA isolation, fractionation, fragmentation, fluorescent labeling, and removal of excess free label and short oligonucleotides; (b) microarrays of immobilized oligonucleotide probes for 16S rRNA identification; and (c) a portable battery-powered device for imaging the hybridization of fluorescently labeled RNA fragments on the arrays. Beginning with whole cells, it takes approximately 25 min to obtain labeled DNA and RNA samples and an additional 25 min to hybridize and acquire the microarray image using a stationary image analysis system or the portable imager. (3) Respiratory viral pathogen has been detected in connection with multiplex PCR amplification. (4) Simultaneous detection and typing of human papillomaviruses and (5) rapid detection and characterization of methicillin resistant Staphylococcus aureus, have also become possible through DNA microarrays.

Many genes associated with virulence are regulated by specific conditions. One way to determine the candidate virulence factors is to investigate the genome-wide gene expression profiles under relevant conditions, such as physiological changes during interaction with the host. A second approach would rely on comparative genomics (Ye et al. 2001).

For example, in a genome comparison study among Helicobacter pylori strains, a class of candidate virulence genes was identified by their coinheritance with a pathogenicity island using DNA microarray technique. The whole genome microarray of H. pylori was also shown to be an effective method to identify differences in gene content between two H. pylori strains that induce distinct pathological outcomes. It is demonstrated that the ability of H. pylori to regulate epithelial cell responses related to inflammation depends on the presence of an intact cag pathogenicity island (Salama et al. 2000).