In a labyrinth of palm tree-lined lagoons and bungalow hideaways, IBC hosted its second annual International Conference on Protein Microarrays this week at San Diego’s Paradise Point Resort. This Fantasy Island setting proved oddly appropriate: While the technology offers much promise, it for the most part still hovers in the limbo between pie-in-the sky concepts and realistic commercial products.
One case in point: Cambridge Antibody Technology chief technology officer Kevin Johnson’s presentation on antibody arrays. “I see antibody arrays as eminently possible,” said Johnson. “The problem I see is how you can make any money out of them.”
At the heart of the issue lies the question of how much the idiom of DNA microarrays translates to the language of proteomics.
“Proteins are not DNA, and this opens up many new challenges technically,” Ian Humphrey-Smith, of the University of Utrecht and Glaucus Proteomics, said in an opening plenary. These challenges include the fact that proteins are less stable than nucleic acids and must be kept wet to maintain their 3D structure; their multiple binding sites; the lack of a standard signal amplification method for proteins; and the bioinformatics conundrum of protein interaction prediction.
But the payoff is worth the extra headache for many, as protein arrays can be used to study interactions between proteins, antigens and antibodies, enzymes and substrates; proteins and DNA, as well as ligands and receptors, pointed out Thomas Joos of the University of T bingen in Germany in his opening speech.
Tools companies like PerkinElmer, in fact, seem to be thriving on this promise of protein chips. The company not only snagged the successful HydroGel slides and protein chip arrayer from its Packard acquisition, at the conference it unveiled a new Protein-BiochipWorkstation and has adapted its TSA amplification reagents for proteins (see article p. 1).
Following are some of the other presentations from the first day of the conference. Additional reporting on the conference will follow in next week’s issue.
Small Molecules, Mammoth Task
Gavin MacBeath of Harvard’s Bauer Center for Genomics Research, who has served as a spokesperson for protein chips, talked about his work with chemical genomics, spotting down small molecules on arrays. Using hundreds of (virtually) identical small-molecule arrays at a time, MacBeath’s team has screened over a million possible protein-small-molecule interactions, then has verified some of the binding activity with traditional Biacore assays as well as protein arrays.
The verification protein arrays include groups of four protein targets, which are fluorescently tagged. A protein-small-molecule interaction will block the fluor. Thus, a pattern where one spot is blocked but the others remain indicates specific binding, whereas quartets of spots that disappear entirely indicate a non-specific binding event.
MacBeath also described how his group has designed microarrays that fit into the microtiter plate format, which is the standard for high-throughput screening. On top, they place a bottomless 384-well plate, then in the middle layer a silicon gasket that has a strong adhesive on the top and a weaker, reversible water-tight seal on the bottom. Under this gasket, four protein arrays on glass slides make up the lower layer. In each well, there are 100 proteins, making 25,600 assays per plate.
The advantage of this technology over just plain protein arrays, according to MacBeath, is that it can fit into existing technology for high-throughput screening assays, as well as scanning technology for protein arrays.
MacBeath wasn’t the only one who focused on adapting other technology platforms for protein arrays: This theme resonated throughout the conference, as participants said they were mindful that many genomic scientists have already spent large sums on instruments.
One original adaptation was that proposed by Lucy Holt of Diversys in the UK. Holt’s team has adapted arraying robots to deposit samples in perpendicular lines to make antibody matrices. The sample along a particular horizontal row consists of a particular heavy peptide chain, and the vertical rows deposit different light peptide chains. Together, each set of light and heavy chains forms a dimer binding site for an antibody.
The method of arraying by lines provides “efficiency savings,” for higher density arrays, as each printhead would only have to draw a sample once for a whole line, rather than having to separately go to each spot. The larger the array, the more the savings. “A thousand lines in each direction would enable a million antibody-antigen interactions to be tested,” said Holt.
Holt has validated this concept on a large-scale array, but would like to develop this approach on a smaller scale.
As a coda to this presentation, Holt threw out what she called a “wacky idea” — a proposal to take her matrix into three dimensions, wherein she could profile antibody-antigen interactions in a more detailed way. Given that proteins live in three dimensions, this idea did not seem at all wacky to some of the other participants.
Peptidomics: Cure for Protein Dyspepsia?
For Oxford GlycoSciences, the problems of protein chips were causing so much dyspepsia among researchers that they decided to digest the proteins down into their peptide components and do what they call “peptidomics” (withhold groans, please). While proteins are highly heterogenous, and thus hard to capture in a uniform manner, “when you digest peptides, they’re all in the same boat,” explained Richard Barry of the company’s department of molecular biology. Furthermore, while proteins have poor stability, peptides are more stable and can be easily bound onto a HydroGel array using peptide-specific antigens. After being immobilized, the peptides can be detected by MALDI mass spectrometry, obviating the need for a label. Signature peptides can stand as a proxy for certain proteins. “There is the potential for whole proteome screening” using this method, Barry said.
Pat Brown-Ing Proteins
Will Pat Brown’s classic cDNA microarray methods work for proteins? Brian Haab, a former student in Pat Brown’s Stanford lab and now an investigator at the Van Andel Research Institute in Grand Rapids, Mich., thinks so. Haab presented a procedure which adapted these methods nearly verbatim to antibody arrays.
In an experiment looking for prostate cancer antigen biomarkers, Haab spotted down 200 antibodies and 384 phage display clones on HydroGel slides. He then labeled the prostate cancer sera samples with Cy5 and a reference sample of healthy male serum samples with Cy3, and added them to the array. Haab then performed a replicate experiment to correct for dye bias, reversing the Cy3 and Cy5 labels on the samples, and normalized the arrays relative to a “reference” protein, IgG, that was on both of the arrays. Haab also used cluster analysis to arrive at antibodies that correlate over numerous experiments.
While borrowing these methods from DNA microarray methods seemed to work, a critical problem remained: The arrays were not very sensitive, only working at concentrations of 200 nanograms per milliliter. “We need to get down two or three orders of magnitude,” Haab said.
Whether this DNA metaphor applies, or the answer lies in peptides or matrices, conference attendees agreed that it may take years for protein array technology to find its appropriate niche. “The story of protein microarray technology is just beginning,” said Joos.