Systematic discovery of attenuated IL-2 variants optimized for bispecific antibodies enabling targeted IL-2 delivery to PD-1 + T cells

Background: Targeted delivery of an interleukin-2 variant (IL-2v) via fusion to an anti-PD-1 antibody has emerged as a promising strategy to overcome resistance to anti-PD-(L)1 therapy. However, unattenuated signaling through the IL-2 receptor beta/gamma chain (IL-2Rbeta/gamma), even in the context of cis -delivery to PD-1⁺ cells, can still lead to systemic toxicities. Fusion of IL-2v to bispecific antibodies introduces further challenges associated with dual-antigen targeting and increased risk of adverse events. To address these challenges, we screened IL-2 variants with varying degrees of IL-2Rbeta/gamma binding that either have IL-2Ralpha interaction preserved or reduced, for fusion with anti-PD-1-based bispecific antibodies to generate novel tri-specific immunocytokines. Methods: IL-2 variants were generated and screened for reduced IL-2Rbeta/gamma binding using random mutagenesis. Variants were evaluated for manufacturability after bispecific antibody fusion via mammalian cell transient expression and affinity purification for both yield and purity. Binding was determined using Octet® and ELISA, while pSTAT5 activity was evaluated on PD-1 negative and PD-1 positive cells to assess PD-1-dependent IL-2 signaling. Variants were classified into non-alpha, alpha-biased, or alpha-masked categories based on the strength of binding to IL-2Rbeta/gamma for further development. Selected tri-specific immunocytokines fused with various IL-2v configurations underwent safety and efficacy assessment in vivo using MC38 syngeneic hPD-1 knock-in mouse tumor models. Results: We generated a library of IL-2 variants with diverse sequences and, following initial screening, performed additional engineering to further optimize their compatibility with bispecific antibody formats. Selected IL-2 variants were successfully synthesized in a bispecific antibody format with an anti-PD-1 arm, generating fusion proteins that met the acceptable-yield criteria. IL-2 variants showed markedly reduced binding and pSTAT5 activity to both IL-2Rbetagamma and IL-2Ralphabetagamma compared to wildtype IL-2. The degree of binding attenuation determined whether these variants should be further developed into non-alpha, alpha-biased, or alpha-masked IL-2v. Optimized IL-2v-fused tri-specific immunocytokines showed dramatically attenuated IL-2R which was restored only upon binding to PD-1, demonstrating cis -acting PD-1-dependency. This was further verified in vivo , where selective PD-1-dependent IL-2v signaling translated into significant tumor growth inhibition with minimal toxicity. Conclusion: Our approach provides a rational framework to fine-tune IL-2v binding affinity for fusion to bispecific antibodies, which can be replicated for the discovery of other affinity-tuned cytokine variants that exert targeted immune activation and therefore, improve the therapeutic index of tri-specific immunocytokines.