Skip to main content
Premium Trial:

Request an Annual Quote

Rutgers Team Devises Multi-Stage Antibody Development Workflow


NEW YORK (GenomeWeb) – A team led by researchers at the Rutgers Cancer Institute has developed a two-step antibody discovery approach that could improve development of molecules for diagnostic and therapeutic uses including imaging and drug delivery.

In a study published this month in JCI Insight, the researchers used the approach, which they termed SPARTA (Selection of Phage-Displayed Accessible Recombinant Targeted Antibodies) to generate antibodies to the known tumor cell surface protein EphA5 and GRP78 and evaluated the effectiveness of those antibodies for targeting tumors as part of antibody-drug conjugates.

The approach combines in vitro screening using phage and yeast display, followed by screening of candidate antibodies in tumor-bearing mice to assess their activity in vivo.

That final in vivo step allowed the researchers to more rigorously evaluate the actual biological activity of the candidate antibodies, said Wadih Arap, director of the Rutgers Cancer Institute and author on the JCI Insight study.

"We start with the target and there are screening steps in vivo, but the proof is in the pudding," Arap said. "We validated in vivo these pools of antibodies. And this is where it got interesting, because from these pools of antibodies we could select the best performers in vivo according to criteria we established. For instance, does it get internalized [by the cell]. Does it have a biological effect if you make it into an IDC [immunocytokine-drug conjugate]?"

The in vivo antibody display is an extension of in vivo peptide display techniques the researchers have been using for around two decades, Arap said, noting that certain technical hurdles had made the approach more challenging with antibodies.

Key to the method was moving between phage display and yeast display of the candidate antibodies, said Fernanda Staquicini, a Rutgers researcher and author on the paper. In phage display, researchers insert the gene for a candidate antibody into a bacteriophage, which then expresses it as a surface protein that can be tested for binding to the target antigen. Yeast display works similarly, fusing the gene of a candidate antibody to that of a cell surface protein, causing it to be expressed so it can likewise be tested for binding to target antigens.

The two approaches typically have different pros and cons, with phage display offering a more convenient way to generate and screen large libraries of potential antibodies and yeast display offering more quantitative assessments of antibody-target binding and more precise selection of the best antibody candidates.

Also important, Arap noted, particularly for the in vivo validation step, was using a phage display approach that didn't require use of a helper phage. Historically, helper phages have been used in phage display experiments to provide cellular machinery required for production of the displayed proteins. However, this approach has downsides including contamination of the experiment with proteins expressed by the helper phage itself, and so researchers have developed alternative methods that rely on plasmids containing the required cellular machinery in place of the helper phages.

Arap said that paper co-author Andrew Bradbury, an expert in antibody display, has been one of the leaders in developing such approaches. Formerly a researcher at Los Alamos National Lab, Bradbury is now the co-founder of Santa Fe, New Mexico-based antibody firm Specifica, which he launched in 2016.

In the JCI Insight study, the researchers started with an unbiased in vitro phage display screening against their targets, EphA5 and GRP78. After two rounds of screening, they selected a subset of candidates and cloned them into their yeast-display system and then used fluorescence-activated cell sorting to select the antibodies that best recognized the two protein targets while also not binding to similar targets.

The researchers then took these antibodies and cloned them back into a phage display system and injected these phage pools into mice with breast and lung tumors to assess their activity in vivo. Using quantitative PCR, they then measured the relative number of the different phage particles bound both to the tumor and control tissue and isolated those more highly bound to the tumor tissue for additional rounds of in vivo selection. At the end of three rounds of selection, the researchers were able to identify several candidate antibodies that targeted the EphA5- and GRP78-expressing tumors while not binding the negative control tissue. They then used recombinant protein-based ELISA and cell-based ELISA to determine that the candidate antibodies were binding to surface EphA5 or GRP78, specifically.

Having established the specific binding of the candidate antibodies, the researchers then set out to characterize their biological activity, looking specifically at whether they could effectively deliver cytotoxic drugs to EphA5- and GRP78-expressing tumors as part of antibody-drug conjugates. They identified one candidate antibody that performed particularly well as part of anti-EphA5 ADCs and another that was effective, though at higher dosage, at killing GRP78-expressing cells.

Different antibodies may exhibit different behavior in vivo, which makes it important to validate not just their binding to a given target but also their biological function, Staquicini said.

"For a target like EphA5, when you finish a screening, you may have 10 antibodies against that target, and all 10 antibodies may have completely different functions," she said. "Some of them will find the tumors in vivo. Some will not. Some will be good for drug targeting and will work as ADCs. We had examples where some of our antibodies were not effective at all as ADCs. They may be good for imaging, but not functionally in vivo."

"There are several different ways you can go about validation, depending on what application you have in mind," she said.

The researchers suggested the method laid out in the JCI Insight study could be broadly applied to the development of antibodies targeting cell surface proteins associated with particular diseases. Additionally, they are now using a monkey model of Lynch syndrome to extend the approach to the unbiased detection of antibodies to cancer-linked cell surface proteins, Arap said.

"That is the next step," he said. "Can you do this [technique] when you don't know the targets. We're in the process of trying that now."