Background

Top Abstract Background Materials and Methods Results Discussion References

Nearly forty years ago, a dedicated cadre of scientists observed that human response to xenobiotics is variable, and they initiated lines of research that now form the foundations of pharmacogenetics. During the last 10 years, in response to the Human Genome Project, pharmacogenetics has undergone a revolution. It has now expanded into the broader discipline of pharmacogenomics. Pharmacogenomics encompasses the study of functional variability not only in drug transport and metabolism but also in every aspect of human genetics that affects drug disposition and response. This meeting's agenda clearly reflects the impact the Human Genome Project is having on efforts to characterize variability in human response to xenobiotics.

During the past 10 years, the Human Genome Project has generated a second revolution, one of a technical nature. Its product is the wide variety of technological innovations that support pharmacogenomic research. Pharmacogenetics at its outset primarily focused on characterizing phenotypic drug responses in intact organisms using biochemical or physical endpoints (Kalow, 1997). Today pharmacogenetics has evolved into largely molecular characterization of drug responses at the cellular level. Our new capability to broadly define human variability at the genetic level and to associate genetic variability with functional diversity has largely been enabled by this recent technical revolution, including large scale functional genomic, genetic, and proteomic analyses.

One seminal invention that has had a particularly transforming effect on both genetic analysis and functional gene expression profiling is the DNA microarray (Pease et al., 1994; Southern et al., 1994; Brown and Botstein, 1999). Although there are several methods for assembling DNA arrays, they all have a common origin in the DNA blotting methods pioneered by Southern in the early 1970s (Southern, 1975). The common elements of this approach to nucleic acid analysis are an immobilized or tethered nucleic acid (DNA or RNA) species that is hybridized with a second, solution phase DNA or RNA species. The sequence of the unknown "target" nucleic acid is discerned by decoding its complementarity with the nucleic acid "probe" of known sequence. Whether the probe or target nucleic acid is immobilized varies among the different array methods, but most commonly, the probe is tethered to a surface, and the target to be analyzed is in solution. This sequence analysis method is broadly useful. It has been applied to detecting polymorphic forms of consensus sequences, detecting and quantitating RNA transcripts, and scoring genomic samples for specific, known polymorphic sequences, among other examples (Cronin et al., 1996; Hacia et al., 1996; Lockhart et al., 1996).

DNA microarrays are innovative in that they apply the principle of hybridization analysis on a physically miniaturized scale while at the same time vastly expanding parallel sampling and analysis capability. Micro chemical synthesis and robotic microfluidic delivery techniques permit manufacture of hybridization arrays capable of performing hundreds of thousands of parallel analyses on a single sample in a single assay. This makes it possible to scan an entire genome for known polymorphic variants or to query a cell type for every messenger RNA transcript expressed during the course of an experiment (DeRisi et al., 1997; Heller et al., 1997; Wodicka et al., 1997; Wang et al., 1998). Measurements of this scope are yielding vast databases of information that provide high-resolution snapshots of cellular activity or comprehensive images of genetic complexity for an entire tissue or organism. Direct comparison of parallel tests on many organisms or tissues permits population-wide genetic complexity to be measured or population-wide polymorphism distributions and allele frequencies to be assessed.

Designing hybridization arrays capable of yielding large quantities of high quality genetic information is as big a challenge as analyzing the complex data derived from these arrays. The sheer magnitude of the probe set size in array-based assays makes traditional probe design and quality control strategies ineffective. Approaches to probe and primer design developed for PCR¹ primer selection have not proven to be very effective for selecting sets of tethered probes for hybridization array-based assays. The ideal hybridization array probe set is one in which a single hybridization experiment provides maximum data quantity of the highest possible quality. Achieving this goal requires hybridization-based assays to be optimized and validated just as any other clinical laboratory test must be. To develop a functional microarray genotyping test, several components must be optimized and integrated, and then performance must be validated. The components include 1) the array probe set, which is selected based on the information needed from the hybridization test; 2) the target preparation method, which includes isolation from the sample source, amplification, and labeling; 3) a detection method that must be coordinated with the target labeling strategy chosen; and finally 4) an automated data analysis method.

Current array technologies are not generally compatible with efficient assay performance optimization and validation due to high economic penalties associated with manufacturing iterative and customized array designs. A DNA microarray format is described here that overcomes this limitation by combining in situ oligonucleotide synthesis with on-the-fly design capability enabled by surface tension chemistry localization. The standard, high-coupling yield phosphoramidite synthesis chemistry it uses supports oligonucleotide manufacturing quality comparable with traditional column-based oligonucleotide synthesis. The combined result of using traditional chemistry and surface tension localization is a unique array technology capable of supporting efficient, economic assay design optimization for any customized analysis application. Surface tension DNA microarrays are an array platform that enables research to be done on how arrays themselves are best configured; consequently, for the first time the full potential of DNA microarrays can be exploited in specific analysis applications.

Despite significant technical innovation in methods for synthesizing, immobilizing, and optimizing DNA oligonucleotide collections into array formats, equally significant hurdles remain to be overcome before this analytical method is capable of supporting routine, reliable, high-throughput assays. Most of the challenges center on developing effective methods for optimizing and standardizing array hybridization performance. Sources of variability in array hybridization assays include the purity of in situ synthesized or immobilized oligonucleotides, efficiency of in situ synthesis or DNA oligonucleotide immobilization chemistries, oligonucleotide density and availability for hybridization, the effects of the array surface chemistry on hybridization, and the relative uniformity of hybridization efficiency for all the oligonucleotides in an array. Arrays are effective research tools at this stage, but new commercial synthesis, quality assurance, and quality control strategies will have to be devised to make them useful in a clinical setting. Although arrays still hold promise as the tool of choice for high-density, high-throughput genotyping, other technologies remain as candidates to supplement or supplant this technology. Today genotyping continues to be supported by sequencing or primer extension type biochemistries coupled with gel- or capillary-based separation. Capillary electrophoresis offers speed and economy over gels but is already evolving from traditional fused silica capillary systems to microfluidic electrophoresis systems called "lab-on-a-chip" devices. The high degree of parallelism and miniaturization possible with these devices may compete quite effectively with hybridization arrays. Microfluidic chip devices have the additional attribute of being modular and adaptable to the continuously evolving landscape of genomic and genetic information that must be sampled. These devices are inherently more manufacturable and robust since they combine well developed manufacturing methods with robust, well developed biochemistries such as PCR, primer extension, and sequencing. It is not likely that any single technology will have the attributes necessary to satisfy the demands of every segment of the diverse genomics community; however, arrays, whether they are hybridization arrays or microfluidic arrays, are sure to play a significant role.

In a demonstration project, the polymorphic human NAT2 gene, a biomarker for cancer and drug metabolism, was selected as a model system to develop general strategies for designing and optimizing DNA microarrays. The goal was to develop homozygous and heterozygous DNA genotyping tests to yield a maximum of high-confidence genotype assignments. The approaches developed using this model can be generalized to design hybridization arrays of virtually any complexity for genotyping and other applications, including gene expression profiling. Furthermore, once an array is designed, it can be expanded in scope and complexity without sacrificing the development invested in the original array.

	Materials and Methods

Top Abstract Background Materials and Methods Results Discussion References

The scope of the genotyping assay design process includes the following steps: 1) selecting the polymorphisms to be genotyped; 2) collecting the sequence contexts around each polymorphism; 3) selecting the target labeling and hybridization conditions; 4) designing the hybridization probes that discriminate among the possible genotypes; 5) testing the assay performance with samples of known genotype; and 6) redesigning the probe set until the maximum genotyping performance is obtained under the chosen conditions. Once this process is complete, final test validation requires that an unknown sample set be genotyped using the microarray assay and the results to be confirmed using a reference genotyping method such as direct Sanger sequencing.

Surface tension DNA array synthesis is the combination of two processes, substrate surface preparation and in situ DNA synthesis. The substrate preparation begins with glass cleaning followed by spin coating with a layer of photoresist. Photolithographic patterning with a mask to define the desired size and distribution of the array features follows spin coating. This process is illustrated in Fig. 1.

View larger version (35K): [in this window] [in a new window]   Fig. 1.   Schematic illustrating the initial steps required for surface tension patterning the array substrates. A photoresist spin coated onto a glass surface is patterned by exposing it to light through a mask. Resist on the exposed surface is developed, leaving the intended array features protected and the rest of the surface as bare glass.

The patterned arrays are developed then immersed in a solution of fluorosilane to generate a hydrophobic silane layer surrounding the array features, which are still protected with photoresist. After the fluorosilane is cured, acetone treatment removes the remaining photoresist exposing the array feature sites. The features are coated with an aminosilane then coupled with a linker molecule that will support subsequent DNA oligonucleotide synthesis. These processed substrates display the surface tension behavior for aqueous and polar organic solvents shown in Fig. 2.

View larger version (130K): [in this window] [in a new window]   Fig. 2.   Once the glass surface is coated with fluorosilane, the remaining photoresist is removed and the array features are coated with an aminosilane. Reagents delivered to the surface are repelled by the fluorosilane and retained in the array features by surface tension.

The surface tension patterned substrates are aligned on a chuck in the robotic array synthesizer where piezoelectric nozzles are used to deliver solutions of activated standard DNA synthesis amidites as shown in Fig. 3. Washing, deblocking, capping, and oxidizing reagents are delivered by bulk flooding the reagent onto the substrate surface and spinning the chuck mount to remove excess reagents between reactions. The substrate surface is environmentally protected throughout the synthesis by a blanket of dry N₂ gas. Localizing and metering amidite delivery is mediated by a computer command file that directs delivery of the four amidites during each pass of the piezoelectric nozzle bank so the predetermined oligonucleotide is synthesized at each array coordinate. Array design iterations are accomplished by altering this synthesis command file.

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Activated phosphoramidite monomers are delivered to the array features by piezoelectric nozzles under the control of a command file that guides synthesis of the correct sequence at each array location. Washing, deblocking, capping, and oxidizing reagents are flooded onto the surface and centrifuged off between steps.

DNA array probe design was aimed at discriminating the seven polymorphic sites in the human N-acetyltransferase gene shown in Table 1 (NAT2, GenBank accession NM  000015). Specific probes were designed by selecting sequences that overlapped each polymorphic site and met additional design criteria such as length or sequence composition. In each case, probe sets were assembled that represented each polymorphism and sets of possible base pairing mismatches for both the coding and noncoding DNA strands. Additional sets of 5' and 3' probe sets were also selected for each polymorphism. An example probe set where all sequences are of a constant 17 nucleotide length is shown in Table 2.

Hybridization targets were prepared by using PCR primers (5'-GTCACACGAGGAAATCAAATGC-3' and 5'-GTTTTCTAGCATGAATCACTCTGC-3') that amplify a 1.2-kb fragment from genomic DNA that contains the entire 870 coding nucleotide, single exon of NAT2 as well as 5' and 3' noncoding sequences. The PCR product was purified, nicked with DNase to generate random fragments of about 50 to 100 nucleotides, and end-labeled with biotin-2',3'-dideoxy-ATP. This product was hybridized to the microarrays under stringent, discriminating conditions. Following washing, the biotin-labeled targets were stained with cyanine-3 fluorescent dye-streptavidin conjugates, and the array was covered with a microscope slide coverslip before fluorescence imaging.

	Results

Top Abstract Background Materials and Methods Results Discussion References

The first array design investigated was constructed with probes set at a constant 17-mer length. Twenty probes selected to identify each polymorphism on the coding strand and 20 probes on the noncoding strand were designed following the pattern shown in Table 2. Hybridization with samples homozygous for the *4 NAT2 allele (Table 1), confirmed by Sanger sequencing, showed that only a subset of these probes hybridized with enough stability to provide adequate fluorescence intensity for reliable detection. The average discrimination of each *4 polymorphism over the alternate polymorphism was about 4:1. Alternatively, when a microarray was designed using probes chosen to have a common algorithmically calculated value for the temperature at which 50% of a DNA duplex denatures of (63 ± 2°C) (Breslauer et al., 1986; Rychlik and Rhoads, 1989), all of the probes hybridized well enough to provide adequate fluorescence intensity, but the average discrimination ratio for the *4 polymorphisms decreased to approximately 3:1. The next array design incorporated probes from this set that were empirically lengthening and shortening to maximize the discrimination ratio of hybridization signals between the two polymorphisms. After several rounds of empirical optimization to improve discrimination ratios, this strategy resulted in a microarray with an average discrimination ratio for the *4 polymorphisms of greater than 6:1. A series of typical images showing this progression of array designs is shown in Fig. 4. Progressive global improvements in array discrimination ratios are summarized in Table 3.

View larger version (56K): [in this window] [in a new window]   Fig. 4.   Typical images taken from the series of array designs showing the progressive improvement of hybridization discrimination for specific positive probes relative to closely related mismatched probes. Tm, temperature at which 50% of a DNA duplex denatures.

	Discussion

Top Abstract Background Materials and Methods Results Discussion References

Oligonucleotide microarrays have most commonly been used to profile gene expression patterns; however, they are now often being used for genotyping applications. Experience with reverse dot blot membranes, which have a long history of being used for genotyping applications, shows that it is challenging to optimize oligonucleotide probes to achieve global maximum discrimination of many genetic variants simultaneously. This problem is amplified considerably in complex DNA oligonucleotide hybridization arrays. Nevertheless, it is a challenge that must be met to take full advantage of DNA array potential in genotyping applications. In situ synthesized surface tension defines DNA oligonucleotide microarrays that offer new options in approaching this problem. Because surface tension arrays do not require complex lithographic masking schemes or large libraries of presynthesized, chemically modified oligonucleotides, there is no economic penalty for testing multiple array design options to arrive at an optimal set of detection probes by an empirical process. This process has been demonstrated by developing a simple genotyping microarray assay modeled on the polymorphic human NAT2 gene. The process described here can be generalized and captured by using algorithms, then reapplied as an automated "intelligent" system to design microarray probe sets based on the polymorphisms to be detected and their local sequence compositions. Although it is unlikely that algorithms will be 100% successful in anticipating optimal probe sets, it is likely to offer a close approximation that can be completed using empirical optimization. In addition, since hybridization probe performance is very dependent on hybridization conditions and discrimination performance is dependent on labeling and detection strategies, it is also likely that these variables will introduce a need for some additional empirical array design optimization before a fully optimized assay is possible. Once a probe set has been selected to detect a given set of polymorphic variants, by using this method new polymorphisms are easily added to the set by reiterating essentially the same process. The original probe set is kept constant and a new potential set of matched probes would be selected for the new set of polymorphisms. Once these probes are fine tuned using the same empirical process that guided selection of the original probe set, they can be added, generating an expanded, optimized hybridization array. In this way, hybridization assays of great complexity can be assembled for increasingly large sets of polymorphisms by a standard, reproducible method.