Researchers from Harvard University, the University of California, San Diego, and Agilent Technologies have developed a synthetic peptide display library covering the full human proteome and have used it to identify known and unreported autoantigens in the spinal fluid of individuals with paraneoplastic neurological disorder.
The library, which was described in a paper published this week in the online edition of Nature Biotechnology, comprises synthetic oligonucleotides expressing roughly 413,000 36-residue overlapping peptides covering all open reading frames in the human genome. These peptides are engineered for display on the surface of T7 phage and can be quantified via phage immunoprecipitation followed by next-generation sequencing.
According the Benjamin Larman, a Harvard researcher and one of the authors of the paper, the new technology offers a more sensitive, less biased, higher-throughput alternative to conventional phage display techniques.
Phage display libraries have been one of the primary technologies for investigating autoantigens tied to autoimmune diseases like multiple sclerosis or type 1 diabetes. Typically, these libraries are generated using cDNA sources from tissue tied to the disease of interest — pancreatic tissue in the case of diabetes, for instance, or brain tissue for neurological autoimmunities.
The problem with this approach is the large amount of non-coding cDNA typically incorporated into such libraries. Such regions, which, Larman said, can account for around 90 percent of the genetic material in a cDNA-based phage library, result in the expression of "basically a bunch of junk peptides that are meaningless and contribute to the background of the experiment."
Another issue is that such libraries are "extremely biased in [their] representation of different genes," he told ProteoMonitor. "Brain [tissue] expresses different genes than kidney, for example. So if you use brain as the source of your cDNA, you're going to have a selective representation of different proteins, and the level of expression can differ by many orders of magnitude. You end up with a library that's basically just the most highly expressed proteins from that tissue."
This makes it particularly difficult to detect low-abundance proteins using conventional phage display systems, Larman noted. In contrast, he said, the synthetic library provides an "unbiased and normalized representation of the proteome."
To investigate autoantigens in patients with PND, the researchers incubated the T7 library with patient cerebrospinal fluid and immunoprecipitated phage-antibody complexes. They then used next-generation sequencing to identify and quantify the oligos in the enriched phage population, giving them the levels of the specific autoantigens detected by antibodies in the CSF.
Using next-generation sequencing as a read-out platform allows the researchers to "measure the relative abundance of each clone, each peptide, in the library before and then after immunoprecipitation with patient antibodies, so we can be very quantitative," Larman said, adding that this enables detection of even low-abundance autoantigens that conventional phage display methods likely wouldn't pick up.
Protein microarrays expressing full-length proteins can also be used for research into autoimmune biomarkers, but, said Larman, these arrays typically don't feature the entire proteome. He added that they are "prohibitively expensive for most labs," offering the example of Life Technologies' 9,000-protein ProtoArray platform, which he said costs around $1,000 per patient to run. The synthetic library costs roughly $150 per patient, and could drop down to the $10-per-patient range as the researchers transfer readout of the assay from an Illumina Genome Analyzer to an Illumina HiSeq sequencer, Larman said.
"The cost [of the assay] will continue to fall along with the cost of sequencing," he added.
Using the technique, the team identified two predicted autoantigens in CFS from a PND patient as well as six additional candidate autoantigens. They also examined CSF from two patients with PND-like symptoms who had tested negative for a panel of commercially available PND autoantigens, identifying three autoantigen candidates.
Key to discovering autoantigens potentially useful as biomarkers will be screening large populations as opposed to small numbers of individuals, Larman said, noting that the new display technique, with its use of next-generation sequencing and adaptability to a 96-well format, is well-suited to such an approach.
"Any individual is going to have a personal set of autoantibodies because of the phenomenon of epitope spreading and antibodies they have against any particular infection they've had that can cross-react with [peptides] in our library," he said. "So it's really going to be important to look at populations of individuals with a particular disease and to start to look at things that are in common between them."
A potential drawback of the platform is the shortness of the peptides used. At 36 amino acids, the peptides likely lack some of the secondary structure of the full-length proteins they are meant to represent. This could cause them to bind to autoantibodies differently, leading to false positives and negatives.
[ pagebreak ]
In the Nature Biotechnology study, however, the researchers observed that the peptides retained a significant amount of secondary structure, in one case binding patient antibodies that failed to recognize the corresponding denatured full-length protein.
"Obviously the disadvantage of shorter peptides is that you have less folding units so you have less conformational information that can be presented to the antibodies," Larman said. "But we were actually surprised by the amount of conformational information that was presented on these phage particles."
Compared to typical peptide libraries, 36-residues is, in fact, fairly sizable, said Joshua LaBaer, director of Arizona State University's Biodesign Institute's Center for Personalized Diagnostics and the developer of the nucleic acid programmable protein array, or NAPPA, which is also used in autoantigen research.
"A lot of the work that you see is done with very short peptides, 20-mers and shorter than that even," he told ProteoMonitor. "And short peptides just don't sample space as well as using longer peptides. So that gives them an edge to some extent because they're working with better-folded peptides."
Still, the basic disadvantages of using peptides as opposed to full-length proteins remain, he said, noting that "a lot of the epitopes that are recognized by these autoantibodies are not linear epitopes. Many of them are discontinuous epitopes, and you're not going to see them [using] peptides."
The 36-residue length was dictated by the length of oligos that the researchers were able to generate, he said. Agilent supplied free of charge the releasable DNA microarrays used to build the library, presumably, Larman said, "to promote their oligo library synthesis technology." The company, he added, has no rights to any intellectual property stemming from the project. Agilent representatives weren't available to comment on their involvement in the study.
Without free DNA microarrays, building a similar library would cost between $10,000 and $20,000, Larman said. Much less expensive, though, is sharing the existing library, he said, noting that it is "very inexpensive to propagate because you just continue to grow the phage in bacteria and can share the library with other labs by just giving them a small aliquot" that they can grow up without ever having to print the oligos themselves.
The scientists don't have a patent on the library, Larman said, but they plan to sell licenses to companies wishing to use it. They will offer it for free to other academic researchers. They also hope to "generate intellectual property based on using the library to discover autoantigens that can be used as biomarkers" for various autoimmune diseases, he added.
Currently they are using it for discovery work exploring autoantigens tied to multiple sclerosis and type 1 diabetes, Larman said.
Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.