Protein array enthusiasts and skeptics gathered for a conference in San Diego this week to hear a variety of speakers — mostly from small start-up companies with still nascent technologies — attempt to convince them that protein arrays will soon deliver on their long-awaited promise. Several participants at this event, part of the third annual Peptalk Protein Information week sponsored by Cambridge Healthtech Institute, complained that there was too much talk of how to make an array and not enough talk of practical applications where arrays were really working. “There are all these little start-ups talking about their technologies — let’s see what they can do already!” one attendee from a large instrument company who asked not to be named told ProteoMonitor.
But kick-off speaker Steven Bodovitz, principal consultant for Select Biosciences, asserted that the obsession with initial technologies — and particularly with producing content — made sense. “It’s the protein chip content that’s holding everybody back,” Bodovitz told ProteoMonitor. “That’s the first hurdle to get over — finding applications and making it cheap and robust are factors too, but content is the big bottleneck.”
It’s Not About Antibodies
It should perhaps come as no surprise that a major focus of the array conference was on content, since the conference was held in conjunction with a simultaneous conference on protein expression. The cross-over was obvious: High-throughput production of capture agents — be they antibodies, pieces of antibodies, cell lysates, synthetic peptides, or even DNA — is an obvious starting point for any protein array platform, and one for which there is still no clear leading method or solution. Bodovitz suggested a four-part solution to the problem of inadequate capture agents: improving the speed, quality, and cost of capture agent production; making better use of content that is already available by focusing on particular targets; improving detection methods to compensate for bad affinity and selection; and concentrating on non-antibody agents such as those used to study protein interactions.
While many of the speakers addressed these points, the only observation that nearly all agreed upon was that traditional, painstaking production of antibodies on an array scale is simply not practical. Steven Pelach, president and chief scientific officer of Kinexus Bioinformatics, which is trying to synthetically produce peptide-based agents as an alternative to antibodies, acknowledged that antibodies are “the key to finding specific probes for target proteins,” but later added that “we can’t map the human proteome by this approach.”
In his talk, Pelach stressed the importance of antibodies as applied to his method of probing for kinases — which are usually not visualizable in 2D gels due to their relatively low abundance — using antibodies to probe multiplexed Western blots. Although Kinexus was finding success with this method, Pelach said, the process was prohibitively laborious, as preparation of each antibody took six months, and the total process cost $20,000 per target. As a result, he said, his company has been working to develop cheaper and faster peptide antibody mimics — 20 amino acid peptides designed in silico to bind to target sequences that he said displayed affinity comparable to antibodies. The technique, developed using software designed by Biomind, uses parameters such as hydrophobicity and bond angles to synthetically design a library of possible peptides for each given target. An initial probe is produced in a three- to four-day time frame, and then can be further refined. Petach described a one in five success rate for producing workable PAMs, and said he hopes a resultant product will launch in early 2005.
Joachim Feldwisch, project manager of arrays at Swedish start-up Affibody, also described a shortcut to the production of full-length antibodies. The company is collaborating with Mathias Uhlen, a professor at Albanova University Center at the Royal Institute of Technology in Stockholm who is working with the HUPO antibody initiative to develop a resource of antibodies or antibody alternatives to every protein in the human body. The collaborators currently boast a library of 3 billion antibody alternatives (see PM 9-19-03). The molecules, called affibodies, are synthesized on a scaffold containing Ig-binding domains of Protein A. They are 58 amino acids long (or 6 kDa, compared with 150 kDa for a typical antibody), stable to proteases and shifts in temperature, and diversified by randomly mutating 13 wildcard positions on the molecule. The surface area devoted to binding is comparable to that available on an antibody — in fact, the primary difference with natural antibodies is that there is “less junk protein,” according to Feldwisch. Feldwisch said that his company ultimately “wants to be a biotherapeutics company” that uses affibodies for “separomics” experiments, but in the meantime is collaborating with Gyros (see PM 10-21-02, 6-10-02) and Ciphergen to provide affibodies for their chips. Affibody is also in the process of producing its own affibody chips, Feldwisch said.
Michael Tainsky, director of the program in molecular biology and genetics at the Karmanos Cancer Institute at Wayne State University, introduced a method that he called “epitomics,” whereby the capture agents on a chip are cancer-specific epitopes derived from patient sera using IgG reactions, and amplified by a variation on phage display called biopanning. The idea is to tunnel in on the actual parts of an antigen that react with the patients’ immunoglobulins. “The immune system is an exquisite biosensor,” said Tainsky. “We want to capture that.” A major drawback to the method is that epitomics cannot account for post-translational modifications, he said.
Back to the Gridiron
Not everybody, however, has given up on high-throughput production of actual full-length proteins. Mingyue He, chief scientific officer of Discerna, gave a talk describing how one can use PCR to produce custom DNA constructs with attached tags, and then express these constructs in a cell-free system, with simultaneous immobiliz-ation of the resultant protein on a chip or bead surface directly off of the expression system via the tag. Since the constructs are made synthetically, there is no need for cloning, and a variety of proteins — including active antigens and enzymes — could be mass-produced in this way, He said.
Ross Chambers, assistant professor at the Center for Biomedical Inventions at the University of Texas, Southwestern Medical Center, said he employed a similar method of using PCR to make DNA constructs that encode for specific antigens, along with selected “bells and whistles” that help in expression and secretion of the gene product inside the animal. He then genetically immunized mice with the DNA construct via a gene gun, causing the mice to express the encoded antigens in vivo, and then produce antibodies against them (see PM 10-31-03). Chambers said the method had a 90 percent success rate, and, significantly, produced antibodies that recognized native proteins. “You don’t get that with commercial antibodies,” he said.
NextGen Sciences, which has been methodically producing antibodies for use in probing so-called reverse-phase arrays for breast cancer tumor markers, also announced at the conference that it is collaborating with Protagen, of Bochum, Germany, to mass produce proteins for use as content on protein arrays. Under the terms of the deal, Protagen will give NextGen access to its library of 11,000 human genes, and NextGen will use its high-throughput protein expression system, the Expression Factory, to turn the cDNA from Protagen into mass amounts of functional proteins. “We’re trying to make GenBank kind of real,” NextGen CEO Kevin Auton said at a press conference on the deal. Auton added that while the two companies will provide the expression service on a fee-for-service basis to outside customers, both NextGen and Protagen also have their own uses for the materials they produce: Protagen wants to use the proteins as content for reverse phase arrays to probe for biomarkers, while NextGen wants to use the proteins to validate the content on antibody arrays that they are already making, using commercially-available antibodies and their own antibody production methods. Auton credited the recent decision by Invitrogen to make its gene activation Gateway software an open architecture format as a major factor in the deal. “Before that it was prohibitively expensive to do this … without Invitrogen’s brave step, we couldn’t be as successful,” he said.
Auton also said that the comp-any will publish “within the next six months” results from its antibody production and validation work in breast cancer.