IL2 Targeted to CD8+ T Cells Promotes Robust Effector T-cell Responses and Potent Antitumor Immunity

Abstract

IL2 signals pleiotropically on diverse cell types, some of which contribute to therapeutic activity against tumors, whereas others drive undesired activity, such as immunosuppression or toxicity. We explored the theory that targeting of IL2 to CD8+ T cells, which are key antitumor effectors, could enhance its therapeutic index. To this aim, we developed AB248, a CD8 cis-targeted IL2 that demonstrates over 500-fold preference for CD8+ T cells over natural killer and regulatory T cells (Tregs), which may contribute to toxicity and immunosuppression, respectively. AB248 recapitulated IL2's effects on CD8+ T cells in vitro and induced selective expansion of CD8+T cells in primates. In mice, an AB248 surrogate demonstrated superior antitumor activity and enhanced tolerability as compared with an untargeted IL2Rβγ agonist. Efficacy was associated with the expansion and phenotypic enhancement of tumor-infiltrating CD8+ T cells, including the emergence of a "better effector" population. These data support the potential utility of AB248 in clinical settings. Significance: The full potential of IL2 therapy remains to be unlocked. We demonstrate that toxicity can be decoupled from antitumor activity in preclinical models by limiting IL2 signaling to CD8+ T cells, supporting the development of CD8+ T cell-selective IL2 for the treatment of cancer. See related article by Kaptein et al. p. 1226.

Figure 1.

CD8⁺ T cells drive antitumor activity, but NK cells are responsible for toxicity with not-α-IL2 therapy. A–F, C57BL6 mice implanted with MC38 s.c. tumors were treated once with not-α IL2 (CTRL-not-α-mIL2) at the indicated doses. A, Schematic of the experimental design. Tumor size (B) and body weight (C) were assessed after treatment (n = 10). D, Peripheral blood cell expansion was assessed on day 5 after treatment with CTRL-not-α-mIL2, reported as fold change in absolute count over vehicle-treated mice (n = 5). E and F, Mice were depleted of CD4⁺ T cells, CD8⁺ T cells, or NK cells beginning 2 days prior to therapy and throughout therapy with CTRL-not-α-mIL2. Shown is tumor volume (E) and body weight (F; n = 10). Graphs in B, C, D, and F show mean ± SD. Studies are representative of 3–5 independent experiments. CR = complete response. Statistics were performed via one-way ANOVA with Dunnett multiple comparisons test versus vehicle (n.s., P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Figure 2.

Human NK cells are preferentially activated by not-α-IL2 and exhibit antigen-independent IFNγ release in vitro. A, Human PBMC subsets were assessed for IL2Rβ or IL2Rγ expression by flow cytometry. B and C, pSTAT5 was assessed by flow cytometry following a 25-minute stimulation of human blood with CTRL-not-α-hIL2. Shown is the % pSTAT5⁺ within the indicated subsets as a function of CTRL-not-α-hIL2 concentration for a representative donor (B) and EC50 values from 4 donors for the indicated cell types (C). D, PBMCs were cultured for 24 hours in the presence of rhIL2 or CTRL-not-α-hIL2, and IFNγ was quantified in the supernatant using MSD (n = 10). E, NK cells, CD4⁺ T cells, or CD8⁺ T cells were isolated from human PBMCs via flow sorting and cultured with the indicated concentrations of rhIL2 or CTRL-not-α-hIL2 for 24 hours. TransAct CD3/CD28 stimulation (1:100) was used as a positive control for T-cell populations. Plotted are geometric mean ± geometric SD (n = 10). F, PBMCs or PBMCs depleted of NK cells via CD56 were cultured for 24 hours in the presence of rhIL2 or CTRL-not-α-hIL2, and IFNγ was quantified in the supernatant using MSD (n = 3). rhIL2; recombinant human IL2. G and H, Sorted CD56^bright and CD56^dim NK subsets were cultured as in D and E. Shown is representative sort strategy (G) and IFNγ in the supernatant (H; n = 4). Data shown in B and C are representative of more than 10 independent experiments; A and D–H are representative of 2–3 independent experiments. Statistics in D were performed via paired t test versus baseline values (n.s., P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Figure 3.

Molecular design and characterization of AB248, a CD8 cis-targeted IL2. A, Overview of the objectives for a CD8⁺ T-cell cis-targeted IL2 molecule. B and C, Human PBMCs were assessed for expression of CD8α and CD8β by flow cytometry (n = 10). Shown are the constituent cell types for CD8α⁺ or CD8β⁺ lymphocytes by percentage (B), and the percentage staining positive for CD8α or CD8β within the indicated cellular subsets (C). D, Overview of the molecular characteristics of the development candidate AB248. E, pSTAT5 was assessed by flow cytometry following a 25-minute stimulation of human blood with rhIL2, CTRL-not-α-hIL2, or AB248 (n = 10). F and G, Human CD8⁺ T-cell subsets were flow sorted from peripheral blood and cultured for 24 hours with AB248, rhIL2, or media, and transcriptional changes were analyzed via RNA-seq. Shown is the sorting strategy for a representative donor (F) and gene expression (by average log₂ fold changes of cpm) over media control. APC; antigen presenting cell. (G) with each differentially expressed gene represented by a point. P value cutoffs of 0.2 are applied to fold change data. A reference dashed line with a slope of 1 is shown; R is the correlation coefficient from the linear regression model (n = 3 donors). H and I, human PBMCs were cultured with the indicated concentrations of rhIL2, CTRL-not-α-hIL2, or AB248 for 5 days and Ki-67 expression was assessed by flow cytometry. Shown is Ki-67 staining in one representative donor at 10 nmol/L (H) and Ki-67 expression (%) within the indicated cell subsets as a function of concentration (I; n = 2). J, Isolated peripheral blood CD8⁺ T cells were incubated with AB248 for 48 hours in the presence or absence of suboptimal TransAct CD3/CD28 stimulation (1:10,000) and supernatant IFNγ and TNFα were quantified using MSD (n = 5). Studies are representative of 2–3 independent experiments.

Figure 4.

Therapy with CD8 cis-targeted IL2 in mice results in potent antitumor activity. A and B, Mouse splenocytes were stimulated for 25 minutes with the indicated molecules and pSTAT5 was assessed by flow cytometry. Shown is representative pSTAT5 staining at 10 nmol/L (A) and frequency of pSTAT5⁺ for the indicated cell types as a function of concentration (B). C–H, C57BL6 mice were implanted with MC38 tumors s.c. and treated i.v. 8 days later with indicated therapy. Shown is the study schema (C), and tumor volume (D), and body weight (E) over time following dosing with CD8-mIL2. F, Peripheral blood was analyzed 5 days after CD8-mIL2 dosing by flow cytometry; shown is fold change in absolute cell count over vehicle-treated animals for the indicated cell types (n = 10). G and H, Mice were implanted with MC38 tumors and treated 8 days later once with CD8-mIL2 or CTRL-not-α-mIL2, and tumor size. (G) and body weight (H) were assessed (n = 9). I–K, 129S6 mice were implanted with T3 sarcoma cells and treated 8 days later once with the indicated treatments. Shown are the study schema (I), tumor volume (J), and body weight (K; n = 10). L and M, MC38-bearing mice were treated once with CD8-mIL2 or CTRL-not-α-mIL2 and absolute cell counts were quantified from peripheral blood and tumors 5 days after treatment. Cell counts are plotted as fold change over vehicle-treated animals (n = 6). Data represented as mean ± SD, and data shown are representative of 2–4 independent experiments. Statistics in D, G, and J were performed via one-way ANOVA with Dunnett multiple comparisons test versus control at the latest time point plotted (n.s., P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). CR = complete response.

Figure 5.

CD8 cis-targeted IL2 enhances the number and function of tumor antigen–reactive CD8⁺ T cells in mice. A and B, 129S6 mice were implanted with T3 sarcoma cells and treated 12 days later with a single dose of 1 mg/kg of CTRL-not-α-mIL2 or CD8-mIL2 as monotherapy or in combination with 5 mg/kg of anti–PD-1. Shown are the study schema (A) and tumor volume after therapy (B; n = 5). C–J, T3-bearing mice were treated with the indicated therapy and TILs were isolated and stained with mLama4 tetramer and analyzed by flow cytometry following therapy. C, Absolute counts of mLama4 tetramer⁺ CD8⁺ T cells in the tumor were assessed over time. D and E, Granzyme B expression in mLama4 tetramer⁺ TILs 2 days after therapy. Shown is granzyme B staining from representative mice (D) and across all mice (E; n = 5). F–J, Expression of exhaustion-associated markers by mLama4 tetramer⁺CD8⁺ TILs was assessed by flow cytometry 5 days following therapy. Shown are representative staining among mLama4 tetramer⁺ CD8⁺ T cells (F), the % TIM3⁺PD-1⁺ of mLama4 tetramer⁺ CD8⁺ T cells (G), TOX staining of mLama4 tetramer⁺ CD8⁺ T cells in representative mice (H) and across all mice (I), and expression of LAG3 within mLama4 tetramer⁺ CD8⁺ T cells (J: n = 5). Data, mean ± SD; data are representative of 2 independent experiments. Statistical analyses were performed via one-way ANOVA with Dunnett multiple comparisons test (n.s., P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). Statistics in C were calculated on day 4. CR = complete response.

Figure 6.

scRNA-seq after CD8-mIL2 therapy reveals dramatic rewiring of CD8⁺ T-cell immunity. A–L, 129S6 mice were implanted with d42m1-T3 (T3) sarcoma cells and treated 12 days later with a single dose of 1 mg/kg of CTRL-not-α-mIL2 or CD8-mIL2 as monotherapy or in combination with 5 mg/kg of anti–PD-1. Tumors were isolated 2 and 4 days after therapy and CD8⁺ T cells were flow sorted and scRNA-seq (10× Genomics) was performed with BEAM reagent labeling of mLama4-specific CD8⁺ T cells (n = 4 per group per day). B, UMAP visualization of CD8⁺ TILs according to the cluster. C, Relative expression (z-score of cluster average log-normalized counts) of selected genes across CD8⁺ T-cell clusters. D, Gene-set enrichment analysis of differentially expressed genes per identified CD8⁺ T-cell cluster. E, Expression profile of selected genes. F, Percentage clonally expanded, defined as detection of 3 or more cells with a given TCR sequence, by cluster. G, BEAM-labeled identification of mLama4-reactive CD8⁺ T cells. H, Proportion of mLama4-specific CD8⁺ TILs in each cluster on day 2 and day 4 by treatment condition. Colors correspond to the clusters as in B and F. I, Ki-67 expression by treatment condition in mLama4-specific CD8⁺ TILs. J, Exhaustion signature scores among mLama4-specific CD8⁺ TILs by treatment group. K and L, Volcano plots for differentially expressed genes versus vehicle control in mLama4-reactive CD8⁺ TILs for CD8-mIL2 (K) and CTRL-not-α-mIL2 (L).

Figure 7.

AB248 selectively expands CD8⁺ T cells in nonhuman primates. A–F, Cynomolgus monkeys were dosed i.v. with AB248 and peripheral blood was analyzed by flow cytometry. Shown are the study schema (A), representative Ki-67 staining in CD8⁺ T cells (B), and the percentage of peripheral blood CD8⁺ T cells staining positive for Ki-67 over time for the indicated AB248 doses (C; n = 2–4 per dose level). Absolute cell quantitation was performed using flow cytometry and hematology on the same blood draw. Shown are absolute CD8⁺ T-cell counts at the indicated dose levels following a single AB248 dose (D, n = 2–4 per dose level) and absolute counts and fold change of the indicated cell types following two AB248 doses (E, n = 6–10 per dose level). F, A cynomolgus monkey was dosed intravenously weekly with AB248 at 0.5 mg/kg for four doses and the indicated immune cell counts were assessed via flow cytometry and hematology. Doses of AB248 are indicated with arrows. Data in C, D, and E, mean ± SD.