Microarrays are best known for measuring differential expression, that is, comparing the relative levels of gene expression between two populations of cells. DNA probes are attached to microscope slides (or other substrates) in a gridlike pattern, and when the sample is added, a complementary piece of target mRNA hybridizes to each probe. Another popular application is to use microarrays to resequence—that is, to test for the presence of a DNA sequence; they can also be used to sequence de novo, although this is not often done.

Although some integrated systems are available, most microarray instrumentation is modular. The arrayers for depositing the DNA probes in gridlike patterns on microscope slides (also called “chips”) and the readers for scanning the slides are sold separately and can be mixed and matched. A selection of arrayers is provided in Table 1,and Table 2 lists some available scanners.

Home-built versus commercial
Perhaps the most noticeable change is that building one’s own microarray system from scratch is not considered necessary anymore. “Three-and-a-half years ago, there was not much choice in terms of [commercial] instrumentation,” says Aldo Massimi, who designed and built a system for the Albert Einstein College of Medicine. “Building was the only real solution.” But recently, impressed with the improvements in scanners, Massimi purchased a commercial product as part of an upgrade.

In addition, Vivian Cheung at the University of Pennsylvania says that arrayers have improved. “Three years ago, I would not have advised people to go buy an arrayer,” she says. “Today . . . I think it would be easier to buy one than to build your own.” Building the instrument takes too much time, she explains, “[And] it’s always nice to buy something that you can use out of the box and not worry about calibrating it and servicing it yourself.”

Perhaps the middle ground between home-built and commercial systems is general-purpose instruments, such as gel documentation systems or colony-picking robots. Using these instruments for microarrays usually requires some modifications, says Cheung. For example, a colony picker would likely need to be adapted from handling square or round petri dishes to handling microwell plates, she explains. Similarly, a general imaging system would have to capture very small spots and might have to be adapted to ensure that the microscope slides are loaded properly, she says.

Because of the variability among home-built and adapted systems, the remainder of this review focuses on commercial instruments specifically targeted to microarrays.

Arrayers
The robotic instruments for making microarrays are collectively known as “arrayers”. The term “spotter” is often used interchangeably, although it is technically limited to instruments that deposit (or “spot”) cDNA probes onto a microscope slide or other substrate. These probes are synthesized off-chip and are temporarily stored in a microwell plate or other receptacle from which they are sampled. Most commercially available arrayers fall into this category.

Printing uniform spots is an arrayer’s main task, and the spots must be the same shape and size throughout a slide and from one slide to another to obtain high-quality data. In general, an array is scanned and stored as two TIFF images, Massimi explains. Each image corresponds to one of the fluorescent dyes used to indicate the presence of a probe–target match. (Many systems use two fluorescent probes, although some use more.) The analysis software superimposes the two images and allows the user to create a grid, which distinguishes the spots from the background area. The more uniform the array is, the better the chances that the software will accurately distinguish the spots and the background.

The uniformity of an array’s grid largely depends on the arrayer’s ability to precisely and accurately move to a new location. In general, spots need to be placed with 10-µm resolution in the xy plane, although there is some disagreement among vendors as to whether it is realistic to claim a resolution <10 µm. The z axis placement—that is, the distance between the print head and the slide—has some influence on the spreading of the droplet.

Because vibrations can affect the robot’s precision, Massimi recommends that a system include an optical table or some other vibration-dampening system.

Variations in spot size and shape can be caused by a variety of other factors, including the robot’s dwell time on the slide, the slide’s coating, the solution viscosity, and environmental factors, such as humidity and temperature. In addition, contaminants are a big problem because particulate matter can interfere with the deposition of spots or create false readings, says Massimi.

Printing details
The other critical component of the array system is the set of pins, pen tips, or quills that transfers the sample from a micro well plate or other receptacle to the microscope slide. Many instrument vendors obtain these printing tips from third-party manufacturers, so several instruments may use the same ones. Between samples, the arrayer must wash and dry the printing tips to avoid contamination. Massimi notes that printing slides becomes more cost-effective if many duplicates are printed at once. Because the tips pick up a fixed amount of material, making only a few slides simply throws away valuable sample in the wash.

Although some researchers prefer particular designs of printing tips, it is not clear whether their performance varies significantly, say the experts. Nevertheless, the amount of sample consumed may depend on the design. However, Massimi notes that new sensitivity-enhancing hybridization protocols may go a long way toward reducing sample consumption.

A truly critical factor is obtaining a set of printing tips that are uniform in size, shape, and height, says Massimi. Having an instrument in which the height of each pin can be adjusted individually within the print head can be an advantage, he adds. Some instruments use a relatively large number of printing tips—say, 48—to gain speed, but Cheung cautions that this gain must be balanced against the increased cost and difficulty of obtaining a matched set.

An alternative to pins, pen tips, and quills is the relatively new inkjet-style printing. Several commercial arrayers now use this approach, as do some companies that sell prefabricated microarray chips. Vendors note that inkjet print heads deposit small volumes of material, which permits high spot densities. Inkjet printing also is very flexible, allowing both on-the-fly spotting and, in some cases, in situ synthesis of probes. Some vendors say that, because inkjet printing is a noncontact method, it eliminates the problem of pins being deformed during spotting, which can affect the reproduci bility of spots. Similarly, other vendors report that inkjet arrayers are being used to make protein chips, because this approach avoids the problems of trying to spot proteins with stainless steel tips.

Nanogen

Nanogen’s microarray technology uses neither spotting nor on-chip synthesis. Instead, the new integrated arraying–scanning instrument, the NanoChip, and the chips are based on a proprietary technology called “electronic addressing”. In electronic addressing, an electrode is associated with each site on the microchip. Because DNA is a charged molecule, the prefabricated probes can be steered to specific sites by activating the electrodes. Nanogen says that this approach provides flexibility because the microarrays can be built or probed site-by-site, and many configurations are possible. Electronic addressing also permits the target molecules to be directed to specific areas of the microarray, which may reduce the hybridization time and permit the simultaneous analysis of multiple test results from a single sample, according to the company.

On-chip synthesis
The alternative to spotting cDNA arrays is Affymetrix’s GeneChip system, which is built in collaboration with Agilent Technologies. (Affymetrix also offers spotting instrumentation, which was developed separately.) These probes are not cDNAs but oligonucleotides (oligos) based on sequences from genetic databases. They are synthesized combinatorially on-chip, using a photolithography process similar to the one for patterning silicon microchips.

The GeneChip system has several advantages, says George Grills, who runs a GeneChip microarray facility at the Albert Einstein College of Medicine. It is an established system that many people are using, so it is well understood, he says. And because there is only one vendor for the instruments, the technology is more uniform than the technology for cDNA arrays.

Although the cost of the GeneChip system may be considered a disadvantage, Grills notes that comparing the cost with that of a cDNA system may be deceptive initially. He says that, in either case, users may not pay the list price for chips because they may have negotiated discounts with the chip vendors. In addition, the price tag for a cDNA chip made in a lab may not include the entire cost of the materials and labor that went into creating the microarray, he says.

The other oft-cited disadvantage of the GeneChip system is its relative inflexibility. Because the system is proprietary, users are limited to Affymetrix’s predetermined chip selection. When it is possible to obtain custom chips from the company, it is expensive and difficult. And unlike users of do-it-yourself arrayers, GeneChip users cannot evaluate chip quality for themselves during the manufacturing process, Grills adds. On the other hand, he says, Affymetrix is constantly releasing new types of chips, and the ability to obtain chips from a single source, with a uniform design and quality controls, is a major selling point of the GeneChip system.

Perhaps the most important distinguishing quality of the GeneChip system is that it has different strengths. Because of the amount of information that can be packed onto a GeneChip, the system is well suited to detecting large-scale patterns of gene expression, Grills explains. On the other hand, the lower cost of the spotted chips makes them more practical for experiments that require various conditions or multiple time points, he adds.

Scanners
“Scanners have come along nicely in the past few years,” says Massimi. “The quality has improved to the point where you can do very decent experiments and where a user’s performance—how well you learned the technique and how cleanly you did the hybridization—is probably the biggest determinant of the quality of results.”

In addition, the speed of scanners has improved, adds Cheung. A scan that used to take ~20 min on average now takes ~2 min. Nevertheless, there is a lot of variability in terms of speed, with some models showing no real improvement, she adds.

Another factor to consider is an instrument’s resolution, because that determines how many pixels make up each spot. The more pixels, the better, because the analysis software often discards pixels around the rim to clean up the data. A reasonably large depth of focus also can be important because a scanner can have difficulty detecting spots that are out of the plane of focus.

In the standard microarray system, two fluorescent dyes—usually Cy3 and Cy5—are used on the same chip. This approach provides an inherent control for variations across a chip, says Massimi. Some scanners can accommodate more colors, but using more dyes can introduce problems, he says.

Unlike the other systems, Affymetrix’s GeneChip is a one-color, one-use system, so users need two chips for a single experiment, says Grills. “That is both a disadvantage in terms of price and a disadvantage when comparing an experiment,” he says. “[However,] the answer—which is a good answer—is that [Affymetrix] includes some powerful internal controls to help users compare two chips.”

Looking to spot your own microarrays in-house?

Whether it’s to gain greater control over their experiment or to cut down on the costs of purchasing pre-existing microarrays, many researchers are now discovering the value of building their own microarrays in house. DNA that works with multiple high-throughput microarray formats is now commercially available, offering a cost-effective way to simplify microarray fabrication. If a researcher already has a microarrayer, an entire set of genes, up to 10,000 DNA elements per set, can be purchased, allowing the researcher to begin spotting the moment the DNA arrives. Incyte’s Easy-to-Spot PCR products, for example, offer preassembled DNA sets—human, mouse, and Arabidopsis— that are nonredundant and are provided in made-to-order formats that best suit a researcher’s array spotter.

Detectors
Scanners fall into two categories—confocal microscopes and charge-coupled devices (CCDs). Confocal microscopes (also called laser scanners) use lasers as excitation sources and photomultiplier tubes as detectors, whereas the CCD-based scanners use white light sources. Makers of the CCD-based instruments stress the versatility of the white light sources, which are not limited to particular laser lines, and the linearity and long integration times of the CCDs, which image the array section-by-section rather than scanning point-by-point.

However, users may find more variation in performance among scanners in one category than between categories. “When we put an array into a confocal and a CCD scanner, we have not seen a huge difference,” says Cheung.

Although some researchers have experimented with cooling the detector to reduce the noise in the electronics, most of the noise in scanners is optical, says Massimi. “The noise level of the electronics is quite a bit lower than what we see as background,” he explains. Factors such as the fluorescence of the slide’s coating or contaminants are responsible for most of the background, although the quality of the filters and lenses also may contribute, he adds.

Need for standards
As users of microarrays are already well aware, there are few standards in the field, except perhaps, the size of the microscope slides. But even that is not a universal standard. The need is most critical for compiling databases of microarray results.

The lack of standards also applies to the instrumentation. For example, there is no agreed-on way to measure a scanner’s sensitivity, and some vendors do not offer any measurements at all. Thus, the final piece of advice from the experts is to try out the instruments. When looking at arrayers, try duplicating an array that you have made before from a sample preparation that you know to be clean, and examine the array for uniformity. Similarly, when looking at scanners, take slides with comparable performance, and put them in different scanners to see how they compare.

Just don’t get too comfortable. New methods, such as covering gold nanoparticles with DNA probes, continue to be introduced, and new commercial products, including microsphere-based array instrumentation, are on the horizon. The microarray market remains a fast-paced race.

This article is adapted from Analytical Chemistry Product Review, Dec 1, 2000.

Elizabeth Zubritsky is an associate editor of Analytical Chemistry. Send your comments or questions regarding this article to mdd@acs.org or the Editorial Office by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036.