Tackling the Affinity Gap at Amgen

August 14, 2019

August 14, 2019 | At Amgen, Agnieszka Kielczewska had a problem. The AMGN12 antibody, derived from an in vivo immunization of the XenoMouse, demonstrated single digit pM affinity to the human orthologue of the target protein, but a 200-fold weaker binding to the monkey orthologue. In order to conduct toxicology studies in non-human primate species, Amgen needed that affinity difference to be closer to 10-fold. 

Kielczewska’s team developed a novel affinity maturation approach to “close” the affinity gap without compromising binding affinity to the human target. On behalf of Bio-IT World, Kent Simmons, Senior Conference Director at Cambridge Healthtech Institute (CHI), spoke with Kielczewska about Amgen’s technique and their findings.

Editor's note: Simmons is producing a track on Antibodies Against Membrane Protein Targets as part of the upcoming Discovery on Target (DOT) conference in Boston, September 16-19. Kielczewska will be speaking on the program. Their conversation has been edited for length and clarity.

 AmgenBio-IT World: First, what makes targeting the G-protein-coupled receptors (GPCR) and ion channel target classes so difficult?

There are a number of factors that make GPCR class of membrane proteins classically challenging targets. First, GPCRs tend to be very highly conserved across species, which makes raising an immune response by utilizing an in vivo immunization very challenging due to immune tolerance. To overcome this, many groups develop custom knock out animals where the genomic locus encoding the target is disrupted. Second, most GPCRs are expressed at low levels on the surface of the cells; high, stable expression of the target is critical for both generating an immune response and for development of successful screening strategies. Third, it is exceedingly difficult to formulate GPCRs in a purified soluble protein format, though new approaches to solubilization of membrane proteins and formulation in nanoparticles are emerging (smalps, nanodics, amphipols etc), that may be utilized for phage or other display platform technologies. Fourth, there is not a lot of real estate on the surface of the cell to put an antibody on. Most published antibodies to GPCRs target the extracellular N-terminal domain. There are few published instances of antibodies interacting with loop regions, which are small, exhibit significant flexibility, and are ultimately challenging to successfully target.

What is the “affinity gap” and why is it important?

Affinity of an antibody to a target frequently (though not always) drives the potency of this antibody in subsequent bioassays for activity, and thus affects efficacy in vivo. To enable toxicology studies in non-human primate species, the affinity gap between binding to the human target orthologue and the cymonolgous monkey orthologues is desired to be small. Different industry standards exist, we generally aim for about a 10-fold difference in affinity, based on tox study dosing considerations.

In the case of antibody 731, the affinity gap was closer to 200-fold. How did you address this?

We utilized a new method for affinity maturation based on the HuTARG mammalian display platform. This technology enables targeted mutagenesis of specific regions of interest within the protein using the enzyme TdT. This allows us to focus on engineering the complementarity-determining regions (CDRs) of the antibody that is being engineered, keeping the germline and framework intact. Because generation of diversity is done by using recombination-signal sequence incorporation into targeted protein sites followed by RAG mediated recombination, the resulting mutations can comprise large deletions and insertions as well as substitutions, which differentiates this approach from others utilized in the industry. The CDR optimization process generates amino acid sequence diversity leading to a range of binding affinities to a target. Diversified variants of the antibody of interest are expressed on the surface of HuTARG library cells, enabling FACS sorting-based isolation of higher affinity variants. The power of the approach also lies in the fact that structural information of the antibody or the target is not required to be known a priori and is completely non-hypothesis driven.

We utilized the HuTARG platform for affinity maturation of antibody Ab731 towards cynomolgus IL-13 (cyIL-13). The process also resulted in dramatically increased affinity to human IL-13 and attainment of a femtomolar affinity antibody. After the discovery of this highly potent molecule we determined the crystal structure of the lead antibody complexed with the cy-IL13 and determined that the mutations introduced by the HuTARG technology were non-obvious and would be difficult to predict in silico without HuTARG platform-based engineering.

Was there an impact on potency?

There was a clear potency improvement on the cyIL-13 protein.  On the human side, the biology of the system prevented detection of the potency improvement despite the large improvement in affinity. 

What technologies has Amgen found to be most effective in working with these targets (or alternately, technologies that have generally not worked well and that are no longer in use)?  

As I mentioned before, we utilize a number of approaches to overcome difficulties associated with membrane targets. In addition to genetic immunizations, Amgen is moving to utilizing innovative protein preparations for in vivo and in vitro antibody discovery. We also utilize novel protein scaffolds and chimeric proteins for immunization.  We are also actively exploring methods to supplement this as high-quality soluble preparations of membrane proteins are frequently difficult to obtain. One such approach involves building on the mammalian cell surface display platform I described above, where the target of interest is expressed on the same cell that produces the antibody. Cells producing either high affinity variants or de novo specific binders can then be identified by FACS sorting or by phenotypic based screening. We are also utilizing approaches forgoing hybridoma completely and focusing on antibody discovery and rescue from single plasma cells. To enable this, we developed a technology based on nanofluidic optoelectronic single B-cell screening, allowing specific antibody identification in 24 hrs as compared to 3 months delivered by traditional hybridoma or phage workflows. This method was recently described in MAbs (doi.org/10.1080/19420862.2019.1624126).