Medical Device Link.

Originally Published IVD Technology January/February 2004

MOLECULAR DIAGNOSTICS

A flexible, open-platform workstation that supports custom assay development holds promise for determining multiple genotypes in a single reading.

Brian A. Dukek and Dennis J. O'Kane

The NanoChip Molecular Biology Workstation by Nanogen Inc. (San Diego) provided the team from the CLASS facility with a flexible open system for use with a moderately sized sample set in the development of molecular assays. (click to enlarge)

Medical genotyping is performed for several purposes. For example, pharmacogenetics hinges on genotyping individuals to determine drug-disposition phenotypes and to predict treatment efficacy. And epidemiological studies may involve genotyping numerous individuals in order to assess risk for or susceptibility to a specific disease. Each of these areas of application requires accurate and precise genotypic determinations.

Three of the more important factors to consider when deciding which genotyping platform to use are cost, the time required to develop assays, and the time required to perform genotyping of samples as measured against the number of polymorphisms examined and the number of samples to be processed—that is, the system throughput. The methods employed by the research team at the Core Laboratory for Assay Signal Systems (CLASS) in the Mayo Clinic Department of Laboratory Medicine and Pathology (Rochester, MN) required that genotyping assays be easy to develop. The platform had to be flexible and open to user customization, and to offer at least an intermediate level of sample throughput in order to be useful in epidemiological applications. Because it was regarded as an instrument for use in clinical research applications, it also had to be relatively inexpensive to operate.

Numerous genotyping platforms have been developed in the past decade. They range in capacity from systems that can make one or a few polymorphism determinations in a few samples, to those that perform multiplexed genotyping of up to a thousand single nucleotide polymorphisms (SNPs) in numerous samples (the Linkage III Panel from Illumina Inc. [San Diego]), to platforms capable of global genotyping of 10,000 or more SNPs on a single individual (the 10K by Affymetrix Inc. [Santa Clara, CA]). The CLASS team's genotyping needs fell between the extremes of large sample set, small genotyping panel and small sample set, large genotyping panel.

This article describes the function and capabilities of the platform that the CLASS lab decided offered the greatest potential for meeting its clinical research and epidemiological operational criteria.

The Workstation

The NanoChip Molecular Biology Workstation from Nanogen Inc. (San Diego) is an open system that facilitates the development of assays for SNPs and mutations that are not available commercially. Development time for new assays is very short; the polymerase chain reaction (PCR) is optimized, probes are designed and ordered, and, except on rare occasions, the first genotyping attempt works well enough to set the standard conditions for subsequent runs. Further, as many as 100 samples can be tested on each microarray without cross-contamination, owing to the electronic addressing system used in the application of samples.

Although other technologies examined offered advantages over Nanogen's workstation, none could provide a better combination of all of the properties the CLASS lab was looking for (see Table I). For example, the NanoChip uses restriction fragment length polymorphism (RFLP)–based SNP detection that is extremely easy to develop and is an open platform, but its throughput is low. The open-platform LightCycler from Roche Diagnostics Corp. (Indianapolis), on the other hand, potentially has higher throughput than the NanoChip, but it is more difficult to develop assays for it that involve more than just a few SNPs.

Loading Samples. The NanoChip instrument consists of two principal parts, the loader and the reader. (Computer hardware and software constitutes a third subsystem.) The loader applies biotinylated, PCR-amplified samples from 96- or 384-well plates to as many as four NanoChip microarrays, or cartridges. A cartridge consists of 100 electronic pads coated with a permeation layer of polyacrylamide gel covalently modified with streptavidin. A robotic arm aspirates a sample of PCR-amplified DNA (the amplicon) from the 96- or 384-well plate, then injects it into a port on the cartridge, dispensing the sample across the microarray. The pads to be addressed with the sample are electronically given a positive charge. The negatively charged DNA passes through the permeation layer and embeds itself. The 5' end of one strand of the amplicon is biotinylated and binds to streptavidin in the permeation layer.

Figure 1. The design of a genotyping assay. The stabilizer's 5' terminal base butts against the polymorphic site, and the Cy3-labeled wild-type and Cy5-labeled variant probe overlap the polymorphic site at the 3' terminal base. (click to enlarge)

After the PCR product of the first sample is addressed to the first pad, the array is washed, a new sample is dispensed, and the process is repeated at the next pad. When all the samples are addressed, the chip is ready to be read and the samples genotyped.

Reading the Chip. Genotyping is performed on the workstation by using two probes and a stabilizer to determine whether a specific SNP appears in an amplicon. The probes are fluorescently labeled with the cyanine dyes Cy3 and Cy5, have melting temperatures of around 37°C, and are specific on their 3' terminal base for either the wild-type or the variant allele. The stabilizer is positioned one base downstream of the polymorphic base and has a melting temperature of approximately 70°C (see Figure 1).

Figure 2. The mechanism of SNP detection. A 3' mismatch causes fewer hydrogen bonds and eliminates base-stacking interactions with the stabilizer, leading to a lower melting temperature. (click to enlarge)

During probing of the amplicon, both matched and mismatched probe-and-target annealing will occur. If the probe specific to the allele anneals, there will be a perfect match with the amplicon along the entire length of both the probe and the stabilizer. The probe has a relatively high melting temperature in this scenario because it has base pairing along its entire length and shares base-stacking energies with the stabilizer, which helps to further stabilize the probe. This is in contrast to a mismatch of probe and target, in which a pucker will occur at the 3' end of the probe. This results in one less hydrogen bond along the length of the probe, and base stacking between the probe and stabilizer does not happen in this case. These annealing events serve to depress the melting temperature of the mismatched probe relative to that of the perfect-match probe (see Figure 2).

During initial runs of an assay, the temperature inside of the reader is gradually raised, generally in 2°C increments. Each time the temperature is increased, a low-stringency buffer washes off any unattached probe. The array sites are analyzed using laser-excited two-color fluorescence detection. Then the temperature is raised another 2°C and the cycle is repeated, producing a signal/temperature trending graph from 24°C and going potentially through 50°C.

Thermal discrimination occurs at the temperature at which the mismatched probe melts while the matched probe remains. In subsequent runs of the assay, the reader temperature can be brought to the previously determined thermal discrimination temperature immediately. The temperature ramp is no longer necessary.

A 5:1 fluorescence ratio between the Cy3- and Cy5-labeled probes indicates a homozygous result (normal or variant), while a fluorescence ratio of 3:1 or less indicates a heterozygous result. Signals between 5:1 and 3:1 and results with a signal strength below 80 counts are indeterminate and considered no-calls.

Time and Cost. A major system limitation is the number of microarrays required to genotype multiple SNPs. If, for example, eight SNPs had to be genotyped for a risk assessment study, then each amplicon might be applied to a separate microarray. Loading eight microarrays with subsequent genotyping SNPs would be time consuming and costly. It takes approximately 4 minutes for the loader to transfer and electronically address each sample from a 96-well microplate to a designated microarray site; some 53 hours would be required to load 100 samples onto eight NanoChip cartridges. This loading time could be reduced by addressing amplicons from a 384-well plate onto four cartridges in the loader.

In an effort to decrease the time required for conducting genotyping studies, and the costs involved, the CLASS lab devised a method of analyzing multiple amplicons serially with the NanoChip workstation. Samples containing up to five separate amplicons, for a total of as many as eight SNPs, have been reproducibly genotyped using the method. The result has been a lowering of genotyping costs and a significant reduction in total analysis time.

System Capabilities

To illustrate the capabilities of the Nanogen platform, the CLASS lab has supported genotyping for a Parkinsonism epidemiology study. Several factors have been found to be protective against Parkinsonism, including caffeine and alcohol intake. Candidate genes involved in metabolism of these compounds are being investigated for their potential role in the development of Parkinson's disease. CYP1A2 is responsible for caffeine 3-demethylation, the initial step in caffeine metabolism.¹ CYP2E1 helps convert ethanol to acetaldehyde.² The challenge the laboratory faced was to develop genotyping assays that could be performed for all of the selected SNPs in these candidate genes, and to analyze the SNPs all on one microarray.

The objective of the experiments was to study SNPs that were either found in the promoter region or that led to an amino acid change in CYP2E1 and CYP1A2 that appeared with a frequency of 1% or greater in the general population. These criteria led to the choice of CYP2E1*2 1132 G>A; CYP2E1*4 4768 G>A; CYP2E1*5 1293 G>C and 1053 C>T; CYP2E1*7 352 A>G, 333 T>A, and 71 G>T; and CYP1A2*3 1042 G>A. Because the CYP2E1*5 and CYP2E1*7 variants were close together, it was possible to amplify all five of their SNPs with two PCRs. However, to complete all eight of the desired genotype experiments, five PCRs, and therefore five chips, would be necessary.

Prior reported NanoChip-based methods have usually involved addressing one amplicon per pad, though in one case multiple amplicons were loaded on each pad and read simultaneously.³ This simultaneous reading allowed the nonspecific detection of any mutation in the multiply addressed amplicon, but not specific genotyping. In other examples, it was possible to read serially multiple SNPs on each pad, but only when the SNPs occurred relatively close together on one amplicon.⁴ The ability to read multiple SNPs on one amplicon using the workstation under investigation would have been limited by the manufacturer's recommended maximum length of 250 base pairs for its hydrogel cartridges, but experience has shown that products with up to 1200 base pairs will work.

If one amplicon with multiple SNPs could be serially read to determine the genotype of all SNPs within that single amplicon, it seemed possible that more than one amplicon could be addressed to each pad. That would allow multiple SNPs that were either widely separated on a given gene, or situated on multiple genes, to be read on the same chip. Then, for example, a company could develop a pharmacogenomics chip that had several genes relating to drug metabolism all together in one combination assay.

Experiment and Results

The homology between genes of the cytochrome P450 superfamily necessitated designing PCRs with at least one intronic primer, or designing a nested PCR protocol using intronic primers. Primers that flanked the SNP of interest were designed using either a sense or antisense 5'-biotinylated primer. PCR cycling conditions were specific to the individual assay, and in all cases involved 10 µl total volume. After amplification, all five PCR products for a given sample were pooled and desalted using a multiscreen PCR cleanup kit from Millipore Corp. (Bedford, MA). The samples were rehydrated in a total volume of 60 µl and with a final histidine concentration of 50 mol.

The pooled amplicons were electronically addressed in the workstation loader to streptavidin-coated microarray target pads. A histidine solution containing the amplicon was dispensed over the array, covering all pads. The pads that were to be addressed were electronically biased with a positive charge. Then the negatively charged biotinylated amplicon passed through the permeation layer and embedded itself, permanently binding the 5'-biotinylated tail to the streptavidin. After the PCR product of the first sample was addressed to the first pad, the array was washed, a new sample was dispensed, and the process was repeated at the next pad. When all amplicon material was addressed, the chip was ready to be analyzed.

After loading, the samples were denatured with 0.1-mol sodium hydroxide for 10 minutes, then washed three times with water and once with high-salt buffer (50-mol sodium phosphate, pH 7.4, 500-mol sodium chloride). Five-hundred-nanomole Cy5-labeled polymorphic and Cy3-labeled wild-type specific probes were allowed to hybridize for 5 minutes.

Figure 3. Results with CYP2E1 addressed alone and addressed with four other amplicons. With data analyzed at 38°C, average signal strength and trending graphs were similar in both experiments. (click to enlarge)

The chip was read over a range of temperatures, and from the readings a signal/temperature trending graph was generated. Initially, signals were read over the range of 24°–50°C in 2°C increments. However, similar results were found to be obtainable with faster read times by reading between 32° and 44°C in 3°C increments (see Figure 3). Each time the temperature was increased, a low-stringency buffer was used to wash off any unattached probe. The array sites were analyzed using laser-excited two-color fluorescence detection. Then the temperature was raised and the cycle was repeated. Thermal discrimination occurred, and data analysis was performed, at the temperature where the mismatched probe and amplicon melted while the matched probe remained.

Nanogen recommends conducting an initial temperature-trending experiment to determine the best temperature for inducing thermal discrimination. Then, once this optimal temperature has been determined, the chip could be read at only that temperature. However, in the experience of the team at the CLASS lab, an increase in background signals caused by multiplexing can often cause call ratios above 5:1 to be missed (see Figure 4).

Figure 4. Fluorescence ratios are lowered by an increase in baseline signal, which leads to an increase in indeterminate rates. (click to enlarge)

The Cy3 and Cy5 signals are resultantly both increased, which may bring the ratio of their absolute values into the indeterminate range. Though the shape of the trending graph remains the same, the shift in absolute values changes their mathematical relationship. The CLASS lab's initial call method relies on the signal ratio, but if the ratio falls into the range between 3:1 and 5:1, an alternative call method based on the trending graph shapes comes into play. Any signal with a trending graph shape having a negative derivative—that is, the signal begins to dive more steeply as the temperature increases—indicates that its allele is present, while a signal shape with a positive derivative—the signal dives less steeply as the temperature increases—indicates that its allele is not present. Figure 4a indicates a wild-type sample by conventional call methods, while Figure 4b is indeterminate. Both samples, however, would be called wild-type samples if the derivative-based criterion is applied.

After the first probes for the first polymorphism had been used, they were stripped off with 0.1-mol sodium hydroxide for 10 minutes. The chip then was washed three times with water and once with high-salt buffer, and subsequently read to determine whether any signal remained. If either red or green signal scores over 60 were detected, the stripping and reading were repeated until signals dropped below that level. The process of hybridizing probes, reading, and stripping was repeated serially for all remaining assays.

Four known-genotype samples, confirmed by means of either bidirectional sequencing with an ABI Prism 377 automated sequencer from PerkinElmer Life and Analytical Sciences Inc. (Boston) or a combination of unidirectional sequencing and RFLP, were run 10 times for each of the eight multiplexed assays in order to check for sample carryover and accuracy. All 320 NanoChip genotyping results matched the sequencing results, suggesting achievement of the highest possible accuracy and precision.

Pad-to-pad signal carryover was checked by running four rows of a NanoChip cartridge with alternating known heterozygous and wild-type samples. No signal carryover was detected. Samples were run both separately and multiplexed, with no difference in genotyping calls, indicating that there was no intrapad crosstalk between probes for one assay and amplicon for another.

Cost Considerations

No analytical instrument is perfect, and Nanogen's is no exception. The first and foremost issue is the instrument's list price of $160,000, making it much more expensive than most other genotyping methods. The Invader from Third Wave Technologies (Madison, WI) and the AcycloPrime from PerkinElmer (Wellesley, MA) are less instrument-dependent systems, relying more on chemistry. They require a generic fluorescent and a fluorescent polarization plate reader, respectively. With these methods, the instrument cost, for an enterprise that already has the appropriate reader for other purposes, is effectively zero. If a reader must be purchased, the cost ranges from $11,000 to $50,000.

The service agreement for the molecular biology workstation costs $18,000 per instrument annually for a standard contract guaranteeing service within three days, and $24,000 per year for a priority contract providing next-day service. This service cost alone is higher than the price of some competitors' instruments.

In addition to being expensive, the NanoChip workstation is a large and extremely complex machine. It requires about seven linear feet of bench space. Within the reader and loader are many pumps and fluid lines, heating and cooling elements, red and green lasers, and photomultiplier tubes. Any of these elements can break down. Although instrument stability has improved with evolving models, it still remains a critical problem. Also, the workstation does not diagnose all problems as they occur. Only an experienced user who noticed odd data would be able to detect some of them. Once problems are detected, however, Nanogen is diligent about resolving them.

Offsetting the high cost of the instrument and the service agreement is the relatively low cost of performing assays. PCR is an expense nearly universal among genotyping methods. Additional assay-related costs involve probes, stabilizers, chips, and desalting plates. Probes and stabilizers cost approximately $200 for a set that will last hundreds to thousands of runs, with each run containing as many as 100 samples. Desalting plates cost $30 for 96 wells, and a 100-pad chip lists at $500. Not including the price of the PCR, probes, and stabilizers, a cartridge containing 100 samples and eight assays per pad would cost approximately $0.66 per result, calculated from list prices.

Conclusion

The NanoChip workstation can reproducibly use at least five amplicons per pad for a total of eight genotyping assays, reducing chip costs by a factor of five and increasing sample throughput relative to previously described methods. Other genes and their SNPs have been evaluated in a manner similar to the procedure described in this article, with similar results.

An even more efficient method of using this instrument is to replace the multiple pooled PCRs with a single multiplexed PCR. Studies involving other genes have shown that the results are the same whether multiple PCRs are pooled or a multiplexed single PCR is run. However, the latter option requires a longer development cycle owing to the greater complexity of the PCR stage. When relatively few samples are being run, pooling separate PCRs is likely to be more time-efficient. For running large numbers of samples, it is probably more efficient to spend the extra development time on getting several PCRs to work in one reaction tube.

To address the issue of possible crosstalk between multiplexed assays, the CLASS lab investigation compared individual and multiplexed signals and demonstrated that for this combination of amplicons and probes, that eventuality was not a concern. However, combinations of amplicon and probes may exist that will not work together without crosstalk. Therefore, during validation of any new NanoChip workstation procedures involving multiplexed amplicon, users should routinely examine the possibility of this complication arising.

References

1. MA Butler et al., "Human Cytochrome P-450(PA) (P-450IA2), the Phenacetin O-Deethylase, Is Primarily Responsible for the Hepatic 3-Demethylation of Caffeine and N-Oxidation of Carcinogenic Arylamines," Proceedings of the National Academy of Sciences 86 (1989): 7696–7700.

2. S Hayashi, J Watanabe, and K Kawajiri, "Genetic Polymorphisms in the 5-Prime-Flanking Region Change Transcriptional Regulation of the Human Cytochrome P450IIE1 Gene," Journal of Biochemistry 110 (1991): 559–565.

3. PS Kim, DK Tai, and KV Lu, "Combinatorial Multiplex Assay Format Using Electronic Microchip Arrays and Its Potential Application in Complex Cancer Diagnostics," Clinical Chemistry 48 (2002): 1851–1853.

4. YR Sohni et al., "Active Electronic Arrays for Genotyping of NAT2 Polymorphisms," Clinical Chemistry 47 (2001): 1922–1924.

PHOTO COURTESY NANOGEN INC.

Brian A. Dukek is development technologist and Dennis J. O'Kane, PhD, is director of the Core Laboratory for Assay Signal Systems in the Department of Laboratory Medicine and Pathology at the Mayo Clinic (Rochester, MN). They can be contacted at dukek.brian@mayo.edu and okane.dennis@mayo.edu, respectively.

Copyright ©2004 IVD Technology