Why do 25 when you can go 60? Like sports car drivers hungering for speed, more microarray users are asking this question as they find out that 60- or 70-mer oligonucleotides deliver peak performance for their self-spotted arrays, outstripping the results from 25-mer oligos.
“We have done a lot of work to ask ‘what is the best probe length,?’ and we found that 60-mers were ideal,” said Dan Shoemaker of Rosetta Inpharmatics, speaking at the recent Microarray Gene Expression Database working group meeting in Boston.
Last year, Shoemaker and other scientists at Rosetta, which is now a Merck subsidiary, performed a head-to-head comparison of oligonucleotide yeast arrays made from oligos ranging in length from 20- to 60-mers, across different hybridization stringencies, and published their results in the April 2001 issue of Nature Biotechnology. Not only did the Rosetta group find that the 60-mers, hybridized at a stringency of 32 percent formamide, had the highest specificity and sensitivity of all lengths, they also found that arrays with a single oligo per gene performed as well as ones with 8 to 22 oligos. Since then, Rosetta has tested its 25,000-spot human 60-mer oligo arrays on glass slides against Affymetrix arrays, which use 25-mer oligos, and found they produce comparable results, Shoemaker said.
Rosetta’s results have resonated in both the academic and corporate worlds. Agilent, which already had licensed Rosetta’s inkjet in situ synthesis process for oligonucleotide arrays, chose to spot 60-mers on its oligo arrays, and recently released a study on its website (www.agilent.com/chem) showing its protocol for 60-mers were more reproducible, sensitive, and accurate than other protocols.
“Long oligonucleotide microarrays are becoming a widely used tool for the parallel expression analysis of thousands of genes,” Agilent’s press statement said, adding that the protocol was issued in response to this trend.
Also, scientists at institutions as diverse as GlaxoSmithKline — a key Affymetrix customer and investor — and Harvard Medical School, have gone public about their use of long oligonucleotides to spot their own arrays.
GSK Goes Long
Priti Hegde, a research investigator at GlaxoSmithKline’s R&D genetics research group, said she and other researchers used to use cDNA microarrays, but switched to oligos for performance reasons. “With cDNAs having from 500 to thousands of base pairs, cross hybridization is such a huge issue that pretty much everything hybridizes with your gene of interest and you don’t get anything significant,” said Hegde. “That is when we decided to switch to oligo-based arrays.”
Hedge and her group tested out 30-and 40-mer oligos, but found that it was difficult to design an oligo of this length that they knew would work. Oligos of 50- to 70-mers had a far lower rate of random hybridization, as a single base-pair mismatch to the probe would be less likely to occur along a 50-mer stretch than a shorter one.
Does this mean there’s something wrong with Affymetrix’s 25-mer arrays, for which GlaxoSmith- Kline is a major customer? While Hegde dismissed the company’s practice of adding single base-pair mismatches to the array, saying that it isn’t reliable given the high rate of random hybridization, she noted that Affymetrix includes a number of probes — formerly 16 to 20, now 11 probes on the new Human U133 set — for each feature on the array, making up for the lower performance per probe.
The company still uses the 25-mer Affymetrix chips for many things, Hegde added. But Affymetrix (of which GlaxoSmithKline owns a 13.4 percent, $193.4 million stake) has no probe sets or arrays for many bacterial genomes that Hegde, a former member of John Quackenbush’s lab at the Institute for Genomic Research, studies. The spotted oligo arrays also allow for more speed and versatility, Hegde said. “It’s more flexible to have your own system in-house. With Affymetrix, if you say ‘I have this one gene I am interested in,’ if it’s not in the Affymetrix probe sets you’ll never be able to find it.”
Self-spotted oligos have their own problems, Hegde acknowledged, with some translating from design to implementation better than others — a factor that can depend on the nucleotide content or some other unknown factors.
Church Lab Has Faith in 70-mer Oligos
Academic labs, such has George Church’s Harvard-Lipper Center for Computational Genomics at Harvard Medical School, have been catching on to this oligo trend recently as well. And accuracy is only part of the reason they are switching from cDNA arrays.
“When people were using PCR product clones, you had to do all these PCR reactions, and the cost of these reactions for about 6,000 cDNAs is about $12,000,” said Aimee Dudley, a postdoctoral researcher at the Church lab. “In addition to cost, there’s the labor issue. It would probably take a single person in our lab a month and a half to two months” to do the reactions.
As have others, Dudley also found that the PCR products are difficult to design so they don’t cross-hybridize with other genes.
“Oligos are cheap and easy to synthesize, you can quality-control them very well, you can control for the amount of DNA that you have in the tube, and they are easy to design so they don’t cross- hybridize,” she noted.
In choosing long oligos, the lab got encouragement from the Rosetta study as well as oligo supplier Operon, which found in its own studies that 70-mers delivered the best results and has since produced oligos of that length.
A further advantage that Dudley saw in longer oligonucleotides was their greater ability to handle the poorly understood phenomenon of RNA secondary structure than shorter oligonucleotides. “If your oligo is 20 base pairs, and the whole oligo is being taken up by secondary structure, it doesn’t have the sequence available. But if there are another 40 base pairs, those 40 base pairs would be able to hybridize to your oligo.”
Does all this long-windedness about long-mers mean that people are going to be jumping off of the Affymetrix bandwagon? Dudley doesn’t think so. “People are generally sticking with the system they feel comfortable with,” she said. “For spotted arrays or Affymetrix, you need certain equipment, and once you or your institution has invested in one system or the other, there has to be a really compelling reason to change.”