Everything's Great When It Sits on a Chip

Date: May 24, 1999Microarray Products and Services GSI Lumonics Excitable Dyes I purified the DNA, ligated it, and transformed it; the gene has to be here somewhere!" In the days before positive selection vectors, a researcher might have screened thousands of clones by hand with an oligo just to find one elusive insert. Today's DNA array technology reverses that approach. Instead of screening an array of unknowns with a defined probe--a cloned gene, PCR product, or synthetic oligonucleotide--each position or "probe cell" in the array is occupied by a defined DNA fragment, and the array is probed with the unknown sample. The typical array may contain all possible combinations of all possible oligonucleotides (8-mers, for example) that occur as a "window" is tracked along a DNA sequence. It might contain longer oligonucleotides designed from all the open reading frames identified from a complete genome sequence. Or it might contain cDNAs--of known or unknown sequence--or PCR products. These DNA arrays, or biochips, represent a blossoming field, estimated to be worth $40 million a year and expected to grow to 10 times that over the next few years. Biochemical "lab on a chip" and whole-cell arrays, such as those made by Cellomics, will not be discussed in this article but are also emerging as important related technologies. We'll start with some technologies for creating DNA arrays, and then look at interrogation chemistry, scanning, analysis, and data management. Many companies produce spotters and gridders for creating your own DNA arrays. The spatial resolution of the robotics determines the density of the array. Most instruments rely on pins or needles to transfer the samples from a reservoir (usually a microtiter dish) to the substrate. The pin diameter and shape, solution viscosity, and membrane characteristics determine the volume transferred and how far the solution will spread. All this can add up to some variation in spot size, shape, and concentration. If you are considering packing as many spots onto an array as possible, you'll need to delve into the details of positional and other tolerances. Genetic MicroSystems uses a novel "pin and ring" device for fast sample spotting: the ring holds a droplet of solution picked up from the well of a microtiter dish, and the pin punches a smaller droplet from this reservoir onto the substrate. This system allows many of the smaller droplets to be spotted without the "print head" returning to the microtiter dish, and surprisingly, the company claims to have spotted a variety of solutions, ranging from volatile solvents to viscous paints, using this approach. Robotic systems for spotting arrays are offered by a number of other companies, including BioDiscovery, BioRobotics, and Cartesian Technologies. In addition to spotting you might like to try this by hand. For smaller projects, Schummer et al. describe a novel guide system for a 384-pin handheld arraying device.1 In addition to spotting DNA sequences that can be grown directly on the array membrane, Affymetrix opened the way for in situ fabrication of defined oligonucleotide arrays with the introduction of the GeneChip. One of Affymetrix's biggest strengths is its proprietary oligonucleotide synthesis technique. Affymetrix grows oligonucleotides of defined sequence on the surface of a glass wafer in a manner analogous to conventional solid-phase oligo synthesis but modified to include a light-sensitive deprotection step. By using masks--similar in concept to those used in the semiconductor industry--a series of photospecific bases can be added to selected points on the array to create a series of oligos with a variety of different sequences. Somewhere between the "gridders" and the "growers" lies Nanogen's electronic addressing technology--a "grabber" perhaps. A microchip is flooded with a solution of a DNA probe, and a row, column, or spot on the chip is electrically activated. This activation introduces a charge onto the surface; the DNA probes concentrate close to the charge and are then chemically bonded in place. The chip is then washed and another probe solution introduced. After the chip is completed, the efficiency of hybridization with a target DNA can also be improved using on-chip electronic effects. Currently, Nanogen offers 1 mm2 arrays containing 25 wire-bonded electrodes, but the company is now testing wire-bonded arrays of 100 electrodes and CMOS-prepared arrays of between 400 and 10,000 sites. Ultralow-volume (nanoliter) pipetting devices can also be used to prepare some kinds of arrays, and while pin transfer is the current state of the art, in the future, you can expect to see inkjet-like printer technology reducing the spot volume into the picoliter range. Regardless of how they are made, DNA arrays are put to a number of uses that can be loosely divided into two groups: genotyping and gene expression. Genotyping arrays are designed to examine DNA at the sequence level. Often, oligos representing all known sequence variants of a gene or collection of genes are represented in an array. A gene of unknown sequence can then be rapidly screened for a large number of deleterious changes. At its most complex, genotyping using a DNA chip involves complete determination of the nucleotide order via sequencing by hybridization (SBH). Hyseq, in collaboration with Perkin Elmer, is developing a universal chip-based SBH system that exemplifies the process. A complete set of oligonucleotides, all possible combinations of sequence in a given length, is synthesized and immobilized on a chip. The target DNA fragment to be sequenced is broken into small pieces, fluorescently labeled, and hybridized with the immobilized oligonucleotides on the chip. The sequence of the target DNA emerges from the pattern of fluorescence bound to the nested sequence. In practice, there are some pitfalls. The longer the target DNA, the longer the oligonucleotides need to be to eliminate ambiguities. As the number of probe cells needed to represent all possible combinations of nucleotides in an oligonucleotide can be calculated by four to the power of the oligo length, adding one to the oligo length quadruples the number of cells (65,536 8-mers, 262,144 9-mers, and so forth). This effectively limits the fragment size that can be analyzed. The technique is also notoriously weak for non-"random" sequences such as direct and inverted repeats. Nevertheless, the system has an enormous strength as well, in that only one chip is needed to analyze any DNA sequence. Imaging Research Color Array DNA arrays for examining gene expression can more correctly be called "gene chips" because they can involve longer fragments of synthetic or complementary DNA. While knowing what each spot in a gene expression experiment represents is beneficial--allowing, for example, correlation with proteins in a proteomics experiment--it is not a prerequisite, and probe cells often contain cDNAs that represent unknown gene sequence or function. Clontech's Atlas arrays are designed to detect expression of a specific gene. Atlas arrays are available for a range of genes of ongoing or current research interest, including cytokines, transcription factors, and cell-cycle regulators. Display Systems Biotech's DisplayARRAY membranes can be used either to study expression patterns within a variety of gene families or to search for new homologous genes. Genosys Biotechnologies' Panorama gene arrays, the first of their kind, contain DNA representative of the entire E. coli genome, represented by 4,290 PCR-amplified open reading frames. Probing these arrays is a simple way to quantify expression levels from all 4,290 E. coli genes under any imaginable growth condition.2 It seems likely that as more genomes are sequenced and analyzed, similar arrays will become available for a wide variety of organisms. NEN, in collaboration with AlphaGene, will soon be providing slides spotted with arrays of 2,400 known human genes. Other defined arrays include Research Genetics' high-density arrays of human and yeast genes and Vysis' GenoSensor arrays, which make use of comparative genomic hybridization to correlate gene expression with disease states. Once the DNA array is fabricated, some form of chemistry must occur between the sample and the array. This chemistry invariably involves hybridization of the sample to the DNA present in the array (ligation-based approaches have also been described), but the detection strategies can be quite varied. Autoradiography of radiolabeled samples is a traditional approach, but other options are available, including electronic signal transduction. Clinical Microsensors has developed "bioelectronics detection" where positive events are detected by electron transfer reactions from the DNA to the substrate, allowing fast and probe cell-specific detection events. One company (Sequenom) is even using matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry to analyze DNA fragments captured in an array format on its SpectroChip. The simplest approach, however, is direct labeling of the target DNA with a fluorescent group. Because chip-bound DNAs are often quite short and varied in sequence, a single stringent hybridization condition that is optimal for every spot on the chip is impossible. In general, however, it is possible to find temperature and salt conditions that give acceptably strong signals for the desired hybridization products and much weaker signals for mismatches. Unbound target DNA is then washed away, and the chip is ready to be scanned. A variety of conceptually simple machines is available to automate some or all of the wet work involved in probing an array. Scanning a fluorescence-labeled DNA array is conceptually quite simple: A light source excites the labeled samples and a detector system measures and records the emitted fluorescence. However, the instrumentation requirements differ based on the precise nature of the array. Most image capture instruments use a scanning detector similar to line-scanning detector systems for DNA sequencing instruments. The number of discrete points that the detector can sample across the array and the row-to-row step interval determine the size of the features that the detector can image efficiently. Currently, it is not possible to orchestrate any form of alignment between the pixels of the detector and the probe cells in the array. It is important, therefore, to be sure the detector pixels are sufficiently small to gather data from enough "significant pixels" in each probe cell to eliminate edge and other artifacts and to obtain a statistically valid estimate of the probe cell's intensity. GSI Lumonics, which manufactures confocal laser scanning systems for fluorescent microarray biochips, claims that the spatial resolution of a detector should be 1/8 to 1/10 of the diameter of the smallest element in the microarray. The company delves into this topic in greater depth on its Web site (www.genscan.com). Clearly, detector resolution is an area that must develop rapidly over the next few years. In the last year or so, probe cells have shrunk from 50 to ~25 microns, and this trend will undoubtedly continue. While the absolute number of DNA molecules that can be crammed into a small space will probably prevent the field from reaching the ~0.2 µm features found in semiconductors, 5 micron probe cells appear attainable. (This would put about four million features on a 1 cm2 chip.) Next is analysis of the scanned image. Most companies perform image analysis by placing a grid over the array to allow integration of the signal from each probe cell. If the array--and the imaging optics--are perfect, this is an easy task: overlay the grid and move it until the probe cells are centered in the grid cells. However, the grid often needs some adjustment. If the signals from the array are strong and fairly consistent, a thresholding algorithm is usually adequate, but weak signals demand a more computationally intensive approach. Thankfully, from an end-user perspective, most companies have created adequate procedures. Last but not least is the signal-to-noise ratio (S/N) of the final image. The background signal or baseline is the fluorescence from the support matrix. Signal can be weakened by insufficient concentration or labeling of sample DNA or by probe cells with too few molecules to capture adequate signal during hybridization. Assuming you can get adequate target DNA preparations, the concern should be with the probe cells. Noise arises from fluctuations in the light source, fluorescence scattered from adjacent samples, and a host of other factors. In general, the longer the detector looks at each pixel, the better the S/N, but very long signal averaging times lead to diminishing returns and excessively long overall scan times. Most companies have taken care of the major S/N issues, but if your array is likely to have a high background, or if you want to examine a large dynamic range, you should ask questions. Commercially available DNA array scanners include Hewlett-Packard's GeneArray system (reviewed by LabConsumer in August 19973), designed for Affymetrix's GeneChips. Many general purpose scanners can also be used to collect data from a wide variety of array types. These include Molecular Dynamics' Storm system, Alpha Innotech Corp.'s CCD-based FluorChem and ChemImager systems for scanning gels and blots, Nucleo Tech's 830 (fluorescent) or 920 (fluorescent and chemiluminescent) imaging workstations, and Axon Instruments' GenePix 4000 two-color array scanner for microscope slides (www.axon.com/GN_Genomics.html). Other workstations for imaging and analysis include Genomics Solutions' GeneTAC and Flexys systems and Imaging Research's M5. Finally, conventional fluorescence and storage phosphor imagers, such as Fuji's FLA-2000, can image many types of DNA arrays. Axon Software screenshot. High-density DNA microarrays often have thousands of probe cells in a variety of grid styles, but the task of managing which clone went where can become intimidating. BioDiscovery's CloneTracker software and Affymetrix's LIMS workstation are two examples of comprehensive microarray probe tracking and data analysis systems. Several other companies also have software products for DNA arrays. Silicon Genetics, for example, offers GeneSpring, a package that accepts data from a number of array formats and facilitates gene identification and expression analyses, while Scanalytics has created extensions to its comprehensive image analysis suite to accommodate information from arrayed samples. While many specialist companies provide just one part of the technology necessary to perform DNA-array experiments, few provide every necessary component. Affymetrix probably has the widest range of components, which in combination with Hewlett-Packard's array scanner add up to a complete system from array fabrication and hybridization through data analysis. Third-party vendors such as Amersham-Pharmacia Biotech resell the entire integrated system. Other companies provide a range of integrated components that cover part of the process. At the front end are several companies that make just the arrays, and a couple that make arrays and fluidics or scanners. At the back end, several make scanners and analytical software, while a few focus on just the analysis and data-mining process. Genometrix, for example, offers the Bioscanner, a sensitive, high-throughput CCD detector for fluorescence or chemiluminescence detection of hybridization results, and GeneView analysis software in addition to its array gridding systems. And besides its gridder system, Genetic MicroSystems offers an imaging system that features a confocal flying spot scanning microscope and is faster than others in this class of detector. Curagen Protein Diagram. Still other companies operate in a mode reminiscent of DNA-sequencing service laboratories. Genome Systems, for example, provides a wide range of array-related services, including clone picking, spotting, and hybridizations with your clone libraries and probes. DNA arrays appear set to become the biotechnology tool of the next decade. Resequencing, gene expression, comparative genome analysis--they seem to do it all, but the data volume they are capable of generating is overwhelming. Soon enough the expression levels of thousands of genes--as well as many mutant variants--will be documented under numerous growth and disease states. It seems inevitable that DNA arrays will drive bioinformatics software to new levels. In the meantime, here are some interesting places to delve a little deeper. The IMAGE consortium (Integrated Molecular Analysis of Genomes and their Expression, www-bio.llnl.gov/image.html) is a great place to start if you are looking for information about arrayed human and mouse libraries and genes. For skeptics (like me) who think that resequencing by hybridization is a clever idea that's too fraught with difficulties to become routinely applicable, read Gunthard et al.'s comparison of GeneChip versus ABI resequencing of HIV DNA.4 A great Web starting point can be found at www.gene-chips.com, where you'll find notes on design and applications as well as a huge list of references, meetings, and links. Nature Genetics also devoted a recent issue to the technology that can be found online at www.genetics.nature.com.5 Or you might like to build your own? You'll find instructions at cmgm.stanford.edu/pbrown/ and in an article by Cheung et al.6 Microarray Products and Services The author, Bob Sinclair, can be contacted at bobsinclair@tech-write-edit.com. M. Schummer et al., "Inexpensive handheld device for the construction of high density nucleic acid arrays," Biotechniques, 23:1087-92, 1997. C. Richmond, F.R. Blattner, "A complete set of PCR primers for the E. coli genome," Genosys Origins, 1:3-4, 1998. "Scanning the horizon: HP GeneArray scanners," The Scientist, 11[16]:17, Aug. 18, 1997. H.F. Gunthard et al., "Comparative performance of high-density oligonucleotide sequencing and dideoxynucleotide sequencing of HIV type I pol from clinical samples," AIDS Research and Human Retroviruses, 14:869-76, 1998. "The chipping forecast," Supplement, Nature Genetics, 21:supp, January 1999. V.G. Cheung et al., "Making and reading microarrays," Nature Genetics, 21:15-19, 1999.

Everything's Great When It Sits on a Chip

Interested in reading more?

Become a Member of