Microarrays: Mainstays of Biological Research

Within a few years of their introduction, microarrays have become a mainstay of biological research. The first commercial microarrays, produced by Affymax in 1991, featured combinatorial libraries of short peptides affixed to a glass substrate.

With the possible exception of microfluidic products, microarrays rely on the concerted input of more divergent scientific and engineering disciplines than any bioscience technology. It is not uncommon to find chemical engineers, physicists, biologists, and chemists working together on a microarray project. The breadth of scientific and engineering know-how involved in the design, fabrication, and manufacture of microarrays boggles the mind.

“We are seeing a more pronounced shift from traditional mRNA expression profiling,” comments Jennifer S. Cannon, PhD, ProtoArray Marketing Manager at Invitrogen (Carlsbad, CA). This changeover began several years back with the combination of chromosome immunoprecipitation with microarrays (i.e., ChIP on Chip), and array-based comparative genomic hybridization (aCGH). More recently, array usage is spreading into additional epigenetic research areas like methylation and miRNA profiling.

Proteomics in microarrays continues to progress, as more researchers realize the advantages, in some experiments, of studying protein expression versus genomics alone. Although protein arrays contain far fewer spots than the typical gene array, this will change as end-users demand greater density and higher throughput.

According to Microarrays: Tech nol ogy Adoption & Utilization, a recent report from BioInformatics LLC (Arlington, VA), one-third of proteomics researchers use protein microarrays, and nearly half plan to use them. Similar growth occurred earlier for gene arrays and is expected for other microarray types (tissue, chemical, oligosaccharide). The report also noted growing interest in very large arrays, particularly whole-proteome arrays.


Putting time on your side
One problem with microarray experiments is they apply time-limited analysis to time-dependent variables. In other words, genes may not be “doing their thing” when their cells were harvested.

Ziv Bar-Joseph, PhD, a professor of computer science and biology at Carnegie Mellon University (Pitts burgh, PA), describes this quandary as a sampling problem arising from the limits of taking “snapshots” of gene activity over time. Anything occurring between snapshots will often go undetected. Bar-Joseph’s technique makes it possible to detect gene activity that occurs even outside the sampling window.

Bar-Joseph has devised a computer algorithm that is similar to the checksum Internet protocol and that minimizes and overcomes the effect of time and cell asynchrony on microarray experiments. On the Internet, checksum compares the number of bits in a sent message with the bits in a received message. If the numbers do not match the message is re-sent. Bar-Joseph’s technique checks the sums of DNA microarray data points over time (a time-series experiment) against the “summary” of the temporal response.

If the two results are equal, the microarray information probably represents the biological event or system under investigation accurately. If the time series sum is lower than the microarray summary, researchers have probably missed a gene activation somewhere during the time series. If the time series exceeds the summary, an artifact has probably
occurred.

Summaries are calculated from two different types of experiment, depending on the biological process. In cases such as the cell cycle, where cells may start out in the same state but rapidly diverge over time, the summary is obtained by sampling and profiling all the cells.

Where asynchrony is not a factor, such as in response experiments, Bar-Joseph samples the cells at very high rates, mixes the samples, and analyzes them by microarray to achieve the summary profile.

To test his method, Bar-Joseph applied it to the cell cycle and discovered several new critical human genes that were not previously believed to participate.

Bar-Joseph’s algorithm allows scientists to separate genes in a time-series experiment into three groups: genes whose changes are corroborated by the summary, those that appear to be changing but are not (false positives), and active genes that were missed by looking at the time series alone (false negatives). “The second and third types cannot be identified through the time series alone,” Bar-Joseph told Bioscience Technology.


Waste not
High-throughput drug screening — even modern methods run in microtiter plates — can consume a lot of material. At this stage all critical reagents — compounds, substrates, and targets — are in very short supply. Scott Diamond, PhD, from the Dept. of Chemical and Biomolecular Engineering at the University of Pennsylvania (Philadelphia), has developed a method of printing small molecules, for example compounds used for drug screening, onto microarrays of up to 6,600 reaction sites. Since hydration of each sample spot is critical, Diamond uses simple aerosol deposition rather than more common inkjet printing.

Each spot is measured in nanoliters (vs. microliters for conventional microtiter experiments), which means researchers can squeeze 1,000 times as many experiments from the same precious batch of receptor, protein, or target. “For many screens the purified protein is rate-limiting,” says Diamond, “especially when you need to isolate it from a rare tumor or virus.” As a drug screening system, his array system shines in situations where a library of compounds needs to be screened against a family of proteins, for example kinases.



The chip can also be used in “reverse.” Diamond has printed one with fluorogenic peptides which he screened against 25 different proteases to determine the proteins’ active site preferences. One protease he identified was implicated in an early event in severe acute respiratory syndrome (SARS) infection.

Dispensing and handling sub-microliter volumes is difficult using traditional liquid handlers/dispensers, but an arrayer handles those volumes easily and accurately. Another problem, which gets worse with smaller volumes, is evaporation. In microtiter formats 1% glycerol keeps everything hydrated. Here, due to the minute quantities, 10% glycerol is required. According to Diamond he has not observed any inhibitory effect in systems he’s tested due to the higher viscosity of his solutions.

Reaction Biology (Malvern, PA) has licensed this technology through Penn, applying it to provide high-throughput screening for pharmaceutical and biotechnology companies. Diamond serves as a scientific consultant to the company.


Sugar is sweet
While most biologists concentrate on decoding genomes and proteomes, a few focus on the glycocode or “glycome” – the chemically diverse oligosaccharides that are found everywhere in biology and which have huge implications for human health. Ten Feizi, MD, at Imperial College (London, UK), arguably the top researcher in this field, studies sugars occurring in glycoproteins, glycolipids, and proteoglycans through oligosaccharide microarrays.

Feizi’s arrays provide a convenient screen for binding between sugars and receptors, microbes, soluble proteins, and other carbohydrate-binding molecules of biomedical interest such as the innate immune system, monoclonal antibodies, cytokines, and chemokines. Identifying these interactions could, for example, explain the efficacy and toxicity (or lack thereof) of glycosylated proteins in individuals or patient subgroups, and perhaps suggest ways to improve therapeutic proteins through genetic engineering, culture or process conditions.

Mass spectrometry (MS) has been a key piece of the oligosaccharide analysis side for these arrays, providing structure, molecular weight, sequence and branching information for oligosaccharides, as well as revealing the sulphate and phosphate groups that may decorate them.

Like protein microarrays, oligosaccharide arrays tend to be sparsely populated. Due to difficulties with sugar chemistry, Feizi has at present about 300 sequence-defined oligosaccharide probes with which to work. “That number may not sound impressive,” she says, “but these are difficult to make. You can’t clone them like genes. We are hoping to build up the numbers to thousands.”

Oligosaccharide synthesis is a difficult process, but advances are being made in this area. For example, new approaches are being made to synthesis through solid phase techniques similar to those for making peptides, but sugar chemistry, preparation of building blocks, and purification of the products is much more difficult. Even attaching them to array matrices is problematic. Because of their water solubility, most oligosaccharides will simply wash off most surfaces, so Feizi first attaches them to an aminolipid through the aldehyde group at the sugars’ reducing end, forming “neoglycolipids.”Then these conjugates are non-covalently linked to a nitrocellulose coated glass slide, which forms the array. Many of the oligosaccharides are obtained by isolating them from natural sources.



Feizi obtained some of her oligosaccharides through collaborations with synthetic chemists: K.C. Nicolaou at the Scripps Research Institute, La Jolla; M. Kiso at the University of Gifu, Japan; and A. Lubineau, University of South Paris, Orsay, France. Synthetic sugars will be increasingly used as they become available, but will probably not meet all of the needs of future oligosaccharide work. “Despite the advances in chemical synthesis, there will always be a place for natural oligosaccharides,” Feizi says. “It is inconceivable that we’ll be able to synthesize every sequence we need.”


Core competency
Increasingly, microarrays are being made at core facilities within research institutions to support various research projects. For example Shrikant Mane, PhD, who directs the Affymetrix GeneChip Resource at Yale Medical School (New Haven, CT), also uses arrays to identify genes implicated in epilepsy, colon cancer, and macular degeneration.

Mane’s experience is typical of best-case results. Researchers had been looking for genes implicated in macular degeneration for the last fifty years. In 2004 Affymetrix began offering a 100,000 SNP (single-nucleotide polymorphism) chip. Using that chip, Mane and coworkers identified the macular degeneration gene in less than three months. “It was like a piece of cake,” says Mane, whose April, 2005 Science paper was one of the most-cited publications of the year. He believes that before long, SNPs and other tools, when applied in microarray format, will help identify genes responsible for every significant disease.



Before the SNP chip it simply was not possible to carry out work of this type within a reasonable timeframe. “Earlier, we were hunting one gene at a time, so we could not see the big picture. Chips give you a snapshot of how genes interact with each other, and how certain events affect metabolic pathways,” says Mane, who nevertheless admits that the chasm between gene discovery and effective treatments is still significant. But even this cloud has a silver lining. “In some cases, for example cystic fibrosis, the primary gene may not be a good candidate for therapy, but gene chips will help identify genes upstream or downstream that are.”

Not all the benefits of microarrays are years away. Mane cited recent work that solved the mystery of why certain hepatitis patients develop cancer during interleukin therapy. Microarray experiments showed the effect to be dose-related, which is good news for thousands of patients.


The DiscoveryDot Platform



(A) Chemical compound or peptide libraries are microarrayed as solution phased individual reaction centers. (B) Chip is activated by a fi ne aerosol mist of analytes, such as purifi ed enzymes or cell lysates. (C) Reactions were detected and analyzed with imaging instruments.


DiscoveryDot chip



6,600 drug like small molecules are microarrayed in noliter volume on the surface of a standard 133 inch Reaction Biology microarray chip.