Background: High-dose recombinant human interleukin-2 (aldesleukin) elicits anti-tumor immunity and is approved for the treatment of renal cell carcinoma and melanoma based on durable complete remissions. However, use of aldesleukin is limited due to treatment-related life-threatening toxicities. Recent second-generation efforts to alleviate toxicities have largely focused on eliminating binding to IL-2Rα, often with half-life extension. We have determined that mice and non-human primates (NHPs) treated with a second generation IL-2 surrogate still experience characteristic dose-limiting toxicities, including vascular leak syndrome. To overcome these toxicities and improve the therapeutic index (TI) of IL-2 as an anti-tumor immunotherapy, we employed protein engineering to generate XTX202, a highly potent third generation masked IL-2. XTX202 is unmasked in the tumor microenvironment by proteolytic activation resulting in full restoration of binding to IL-2Rβ without binding to IL-2Rα. The current study characterizes the therapeutic index of XTX202 versus aldesleukin and a second generation IL-2 surrogate. Methods: XTX202 bioactivity was measured using STAT-5 phosphorylation in human PBMCs and reporter cell lines. Anti-tumor efficacy and peripheral immune activation were evaluated in mice bearing syngeneic tumor models. Safety was evaluated in rodents and Cynomolgus monkeys. XTX200, an unmasked half-life extended IL-2 that does not bind to IL-2Rα, was used as a surrogate second generation IL-2. Results: Masked XTX202 showed limited IL-2R-dependent STAT-5 signaling in vitro. Proteolytic activation of XTX202 resulted in CD8⁺ T and NK cell activation and over 1000-fold reduction in Treg activation as compared to WT IL-2. XTX202 achieved potent tumor growth inhibition in syngeneic mouse models as a single agent with no evidence of toxicity or peripheral immune activation, thus demonstrating tumor selective activity. XTX202 efficacy in mice at 2 mg/kg dose was equivalent to that achieved with the MTD dose of 0.5 mg/kg of a second generation IL-2 surrogate. XTX202 was well tolerated in NHPs in a 4-week repeat dose study at doses up to 30 mg/kg QW whereas a second generation IL-2 surrogate was not tolerated beyond 0.7 mg/kg QW. Based on these data, XTX202 has a 10 fold improvement in TI vs second generation IL-2. Based on comparative efficacy studies with aldesleukin and literature NHP tolerability data, XTX202 is projected to have a ≥150 fold greater TI than aldesleukin. Conclusions: XTX202, a third generation, tumor-selective IL-2, inhibits tumor growth and is well tolerated in repeat dose studies in NHPs at high doses. GLP toxicity studies with XTX202 are underway and first-in-human studies are expected to initiate this year. XTX202 has the potential to be a best-in-class IL-2 immunotherapy by expanding the curative anti-tumor activity of IL-2 while minimizing dose-limiting toxicities.