Background Although cytokine therapies have demonstrated curative effects in some cancer patients, clinical use remains limited due to inflammatory toxicity profiles accompanying systemic cytokine administration. Next generation cytokine approaches include conditional signaling focused to sites of interest, such as the tumor microenvironment or specific immune cell populations. Here, we share a novel approach for generating detuned cytokine therapeutic candidates using the AlphaSeq platform, which involves the re-engineering of yeast agglutination and mating to quantitatively measure protein-protein interactions at a library-on-library scale. Using interferon alpha 2 (IFNA2) as an illustrative example, we show how AlphaSeq measures cytokine-receptor interactions and generates engineered cytokines with a broad range of affinities and signaling potencies.

Methods A saturated mutational library was created for IFNA2 and subsequently screened against a second library consisting of human IFNAR2, species orthologs, and off-target receptors, which allowed for parallel identification of hundreds of detuned variants against both human and mouse receptors in a single assay. Cytokine variants with lower affinity than parental IFNA2 were recombinantly expressed as Fc fusion proteins to orthogonally measure affinity with biolayer interferometry and characterize potency with an in vitro human PBMC phosflow assay. To localize detuned cytokine candidates to cell types of interest, we utilized our synthetic humanoid antibody discovery library to identify single chain fragment variable (scFv) binders against targets of interest and selected one candidate for AlphaBind antibody optimization. To demonstrate cell population-specific signaling, from pools of predicted affinity-improved binders we generated immunocytokine fusions with detuned variants of multiple parental cytokines and assessed specificity and signaling potency by in vitro human PBMC phosflow assay.

Results Biolayer interferometry and phosflow results of detuned cytokine-Fc fusions spanned multiple orders of magnitude of affinity and signaling potency, and demonstrated strong correlation with AlphaSeq affinity measurements. When expressed as full immunocytokines, candidate molecules showed orders of magnitude greater potency in the targeted cell population than non-targeted populations.

Conclusions Our results show the AlphaSeq platform can accurately quantitate thousands of cytokine variant affinities simultaneously against multiple relevant receptors, and when coupled with antibody discovery and AlphaBind optimization, enables the selection of candidate immunocytokine antibody fusion proteins with exquisite cell population specificity. AlphaSeq’s and AlphaBind’s rapid, comprehensive affinity determination is being used to develop a portfolio of clinically relevant therapeutic immunocytokines.