A split, conditionally active mimetic of IL-2 reduces the toxicity of systemic cytokine therapy

Abstract

The therapeutic potential of recombinant cytokines has been limited by the severe side effects of systemic administration. We describe a strategy to reduce the dose-limiting toxicities of monomeric cytokines by designing two components that require colocalization for activity and that can be independently targeted to restrict activity to cells expressing two surface markers. We demonstrate the approach with a previously designed mimetic of cytokines interleukin-2 and interleukin-15-Neoleukin-2/15 (Neo-2/15)-both for trans-activating immune cells surrounding targeted tumor cells and for cis-activating directly targeted immune cells. In trans-activation mode, tumor antigen targeting of the two components enhanced antitumor activity and attenuated toxicity compared with systemic treatment in syngeneic mouse melanoma models. In cis-activation mode, immune cell targeting of the two components selectively expanded CD8⁺ T cells in a syngeneic mouse melanoma model and promoted chimeric antigen receptor T cell activation in a lymphoma xenograft model, enhancing antitumor efficacy in both cases.

Fig. 1. Neoleukin-2/15 can be split into two fragments that reconstitute its activity when combined.

a, Splitting strategy for Neo-2/15. Structure of Neo-2/15 (cylinder representation) and murine IL-2 receptors β and γ (beige and pink surface representations, respectively, PDB ID: 6dg5). Dashed line represents the selected split site between helix H1 (Neo2A), involved in the interface with both receptors, and fragment H32’4 (Neo2B). b, Binding of split Neo-2/15 fragments to the IL-2 receptor. Biolayer interferometry binding assay illustrating the binding kinetics of intact Neo-2/15 (black), Neo2A (blue), Neo2B (purple) and the combination of Neo2A + Neo2B (magenta) (1 μM concentration) binding to biotinylated human IL-2Rγ immobilized on an Octet streptavidin sensor in the presence of 250 nM soluble hIL-2Rβ. Full titrations are provided in Extended Data Fig. 1a. c, Signaling of split Neo-2/15 on human YT-1 cells. STAT5 phosphorylation in YT-1 cells following treatment with intact Neo-2/15 (black, EC₅₀ = 1.87 × 10⁻¹⁰M), Neo2A (blue, EC₅₀ undetermined), Neo2B (purple, EC₅₀ undetermined) or the combination of Neo2A + Neo2B (magenta, EC₅₀ = 2.83 × 10^–8M). Treatment with the split pair reconstituted Neo-2/15 activity. * indicates different EC₅₀ values with nonoverlapping 95% confidence interval ranges. Experiments were performed in triplicate three times, with similar results. Data are presented as mean ± s.d.

Fig. 2. Trans-activation of immune cells through targeted reconstitution of split Neo-2/15 on the surface of tumor cells.

a,b, In vitro assay for reconstitution of Split Neo-2/15 on the surface of engineered K562 cells. a, Neo2A and Neo2B split proteins were fused to anti-EGFR and anti-HER2 DARPin-targeting domains. K562 cells (gray) and engineered K562 cell lines transduced with HER2-eGFP (purple), EGFR-iRFP (blue) or both (pink) were mixed in an equivalent ratio then incubated with targeted split Neo-2/15 proteins at 10 nM. b, Reconstitution of Neo-2/15 binding activity on the cell surface was measured by recruitment of phycoerythrin (PE)-labeled soluble hIL-2Rβγ. *P < 0.0001. c,d, Assay performed to measure trans-presentation of split Neo-2/15 on the surface of K562 tumor cells to human YT-1 NK cells. c, Untransduced HER2^–/EGFR^– K562 cells (off-target) or HER2⁺/EGFR⁺ K562 cells (on-target) were cocultured with YT-1 cells in a 20:1 ratio (K562:YT-1) in the presence of ɑHER2-Neo2A and ɑEGFR-Neo2B fusion proteins. d, YT-1 cell activation was analyzed by measurement of STAT5 phosphorylation. Strong signaling was observed for on-target K562 cells incubated with ɑHER2-Neo2A + ɑEGFR-Neo2B (EC₅₀ = 1.57 × 10^–6). The remaining groups showed weak signaling (undetermined EC₅₀). e,f, Assay used to detect trans-activation of antigen-specific αTrp-1 CD8⁺ T cells by split Neo-2/15 on the surface of B16F10 melanoma cells overexpressing PD-L1 in coculture. e, Split Neo-2/15 proteins were fused to either a nanobody (VHH) with irrelevant specificity (Ctrl) or an αPD-L1 nanobody. f, Trans-activation of T cells was measured by expression of the activation marker CD25. All samples were incubated with an ɑCD28 antibody to provide costimulation. αCD3 was used as a positive control for T cell activation, and OVA peptide to quantify basal CD25 expression levels. g, B16 and CD8⁺ T cells were cocultured in the presence of activating proteins. Addition of excess soluble ɑPD-L1 nanobody (VHH) to competitively inhibit binding of targeted split fusion proteins to the B16 cell surface resulted in attenuated T cell activation. All data in the figure are presented as mean ± s.d. One-way ANOVA comparisons against the negative control group. NS, no statistical significance; a.u., arbitrary units.

Fig. 3. Targeting split Neo-2/15 to tumors increases safety and enhances antitumor efficacy.

a, Safety study in immunocompetent mice (n = 5 per group) treated with targeted Neo-2/15 and targeted split Neo-2/15. “Ctrl” indicates fusion to an irrelevant nanobody as untargeted control. Mice were treated daily with equivalent doses of the indicated proteins (2.6 nmol). Weight change was monitored to evaluate toxicity of the dosed molecules. One-way ANOVA was used to compare weights at day 60. b, Efficacy study in immunocompetent mice (n = 5 per group) bearing PD-L1-overexpressing B16 melanoma cells. Mice were administered daily with therapeutic doses of the test items: αPD-L1-Neo2/15 and Ctrl-Neo2/15 at 430 nmol or targeted split Neo-2/15 fragments at 8 nmol. All groups were cotreated with Ta99 biweekly. Unpaired two-tailed t-tests of tumor growth values at day 22. c, Efficacy study in mice bearing B16 PD-L1-overexpressing melanoma cells in the flank (n = 5 for PBS and Ta99 groups, n = 10 for other groups). Neo-2/15 was dosed daily at 2.6 nmol and split Neo-2/15 fusions at 8 nmol daily. d, Weight change of the mice in the study shown in c comparing the toxicity of Neo-2/15 and the αPD-L1 split Neo-2/15 fusions. c,d, Unpaired two-tailed Student’s t-tests were performed to compare tumor volume and weight change at days 19 and 29. e, Surviving αPD-L1 split Neo-2/15-treated mice were rechallenged alongside control mice receiving a primary challenge (n = 5). CD8⁺ T cells from treated mice were analyzed for recognition of tumor cells by the presence of anti-Trp1-MHCI tetramer (right). f, Survival curves for immunocompetent mice bearing B16 melanoma cells in the flank overexpressing murine PD-L1 and human HER2 and dosed daily with the following test items: PBS (n = 12), Neo-2/15 (2.6 nmol, n = 5), ɑPD-L1-Neo2A + ɑPD-L1-Neo2B (8 nmol, n = 5), ɑHER2-Neo2A + ɑPD-L1-Neo2B (8 nmol, n = 7). Statistical analysis for survival curves was performed by Mantel−Cox log rank. Data presented as mean ± s.e.m.

Fig. 4. Cis-targeting split Neo-2/15 to CD8 specifically expands CD8⁺ T cells in murine models.

a, Cis-activation mechanism of split Neo-2/15 targeted to CD8 (via fusion to an ɑCD8 VHH) on the surface of T cells. b, ɑCD8 split Neo-2/15 promotes specific CD8⁺ T cell proliferation in healthy mice. Proteins were administered daily to nontumor-bearing Foxp3-GFP mice (n = 3; n = 2 for PBS control) for 5 days at 12 µg per mouse per day (500 pmol) for intact Neo-2/15 fusions and 10 µg per mouse per day (500 pmol) for split Neo-2/15 fusions. Cell populations of spleen and both inguinal lymph nodes collected at day 6 were analyzed by flow cytometry. Unpaired two-tailed Student’s t-test was used to evaluate statistical significance between groups. c, Efficacy study in C57BL/6 J mice bearing WT B16 melanoma treated with ɑCD8 split Neo-2/15 and TA99. Test items were dosed daily at 500 pmol each, starting on day 5. Ta99 was administered biweekly starting on day 3 (150 µg per mouse). Tumor growth (right) and mouse survival (left) are shown (n = 10 mice per group for ɑCD8-targeted split Neo-2/15-treated mice, n = 5 for controls). Left: ***P = 0.006, Mantel−Cox log rank comparing Ctrl-Neo2A + Ctrl-Neo2B with ɑCD8-Neo2A + ɑCD8-Neo2B. Right: unpaired two-tailed t-test comparing Ctrl-Neo2A + Ctrl-Neo2B with ɑCD8-Neo2A + ɑCD8-Neo2B at day 19. **P = 0.049. b,c, Data presented as mean ± s.e.m.

Fig. 5. Cis-targeting split Neo-2/15 to CAR-T cells enhances CAR-T proliferation and tumor killing.

a, Cis-activation mechanism of split Neo-2/15 targeted to HER2 expressed as a transduction marker on the surface of CAR-T cells. b, STAT5 phosphorylation of untransduced HER2^– primary CD8⁺ T cells (left) or transduced HER2⁺ CD8⁺ T cells expressing the ɑCD19 CAR (right) (n = 2 cell samples). An EpCam-binding DARPin (Ec1) was fused to split Neo-2/15 fragments as an untargeted control. Data presented as mean ± s.d. * Indicates different EC₅₀ values with nonoverlapping 95% confidence interval ranges. c, In vitro coculture of CAR-T cells and NCI-H1975 (ROR1⁺) tumor cells in the presence of cis-targeted split Neo-2/15 costimulation. Anti-ROR1 CAR-T cells were engineered to express either HER2 (HER2t, on-target cells) or CD19 (CD19t, off-target control cells) as a transduction marker. CAR-Ts were cocultured with tumor cells for a total of four consecutive stimulations in the presence of the test items. End-point data showing tumor cell index (left) and CAR-T cell proliferation (right) after 35 h on the fourth stimulation of CAR-T cells with tumor cells. n = 2 independent cell cocultures. Statistical analysis performed with one-way ANOVA comparing each test group with control group (left: no CAR-T group; right: CAR-T-only groups). Data presented as mean ± s.d. d, Efficacy of ɑCD19 CAR-T cells and cotreatment with cis-targeted split Neo-2/15 fragments in a lymphoma xenograft model. Both HER2- and EpCam-targeted (negative control) split Neo-2/15 fragments were dosed at 7.5 mg/kg (n = 5 mice per group). Data presented as mean ± s.d. a–d, Statistical analysis for survival curves was performed by Mantel−Cox log rank.

Extended Data Fig. 1. IL-2 receptor binding and immune cell signaling activity of Split Neo-2/15 pairs.

Binding to immobilized IL-2Rβγ measured via biolayer interferometry (left) and STAT5 phosphorylation response (right) of YT-1 human NK cells for Neo-2/15 (black), one split fragment (blue), a second split fragment (purple) and the combination of the two split fragments (magenta). Results are shown for three split pairs: H1 + H32’4 (a); H13 + H2’4 (b); and H132’ + H4 (c). The H1 + H32’4 (Neo2A + Neo2B) pair demonstrates the optimal conditional activation profile. The calculated EC50 values for the STAT5 phosphorylation (a-c, right panels) are: Neo-2/15 0.187 nM; H1 + H32’4 14 nM; H13 + H2’4 34.6 nM; H132’ + H4 28.3 nM; H1 > 40 µM; H32’4 > 40 µM; H13 > 40 µM; H2’4 > 40 µM; H132’ 171 nM; H4 > 40 µM. * indicates different EC50 values with non-overlapping 95% confidence interval ranges. Experiments were performed in triplicate three times with similar results. Data are presented as mean values +/− s.d.

Extended Data Fig. 2. Biophysical analyses of split Neo-2/15 fragments Neo2A and Neo2B.

a, Left: Circular dichroism spectra of Neo-2/15 (black) and the Split Neo-2/15 fragments individually (blue and purple) or in combination (magenta). Neo2B shows negative ellipticity at 210 and 222 nm, indicating an alpha helical secondary structure. Right: Thermal melts of each split fragment, monitored by their signal at 222 nm during heating from 25 °C to 95 °C (heating rate ~2 °C/min). b, Biolayer interferometry binding analysis of Neo2A binding to Neo2B. Biotinylated-Neo2A was immobilized on streptavidin-coated tips and analyzed for binding in the presence of serial dilutions of Neo2B (left panel) or Neo2A (right panel). The calculated KD between Neo2A and Neo2B is 4.6 μM. Colored lines represent the concentration of analytes in nM. c, Biolayer interferometry binding analysis of Neo2A binding to Neo2B in the presence of soluble hIL-2Rβ and hIL-2Rγ. Biotin-Neo2A was bound to Streptavidin-coated tips and analyzed for binding with serially diluted Neo2B in presence of the soluble IL-2 receptor subunits hIL-2Rβ and hIL-2Rγ. The calculated KD for the complex is 50.8 nM. Colored lines represent equimolar concentrations of Neo2B, hIL-2Rβ and hIL-2Rγ in nM.

Extended Data Fig. 3. Immune cell signaling activity of targeted Split Neo-2/15 pairs.

a-c, STAT5 phosphorylation response in YT-1 human NK cells following treatment with intact Neo-2/15 or split Neo-2/15 fragments fused to HER2- or EGFR-targeted DARPins. Results are shown for three split pairs: H1 + H32’4 (a); H13 + H2’4 (b); and H132’ + H4 (c). All Split Neo-2/15 variants remain functional after fusion to DARPin targeting domains. The calculated EC50 values are: (a) Neo-2/15 0.201 nM; ɑHER2-H1 4.09 µM; ɑEGFR-H132’ 4.20 µM; ɑHER2-H1 + ɑEGFR-H32’4 43.9 nM; (b) ɑHER2-H13 5.51 µM; ɑEGFR-H2’4 6.73 µM; ɑHER2-H13 + ɑEGFR-H2’4 43.9 nM; (c) ɑHER2-H132’ 9.87 µM; ɑEGFR-H4 2.03 µM; ɑHER2-H132’ + ɑEGFR-H4 3.07 nM; d, Increasing length of the linker separating the Neo-2/15 split fragments from the DARPin targeting domain from 15 to 30 amino acids does not affect activity. The calculated EC50 values are: Neo-2/15 0.99 nM; 15 residue linker ɑHER2-H1 + ɑEGFR-H32’4 68.3 nM; 30 residue linker ɑHER2-H1 + ɑEGFR-H32’4 68.1 nM; H1 + H32’4 67.8 nM. e, Intact Neo-2/15 fused to HER2- or EGFR-targeted DARPins retains full activity, as measured by STAT5 phosphorylation response in YT-1 human NK cells. Experiments were performed in triplicate three times with similar results. Data are presented as mean values +/− s.d. * indicates different EC50 values with non-overlapping 95% confidence interval ranges.

Extended Data Fig. 4. Reconstitution of Split Neo-2/15 targeted to the surface of transduced K562 cells.

a-c, Dilution series of targeted Split Neo-2/15 variants were evaluated by flow cytometry using the experimental setup shown in Fig. 2a. Reconstitution of IL-2 receptor binding activity is measured by recruitment of fluorescently-labeled hIL-2Rβγ (resulting from a mixture of IL-2Rβ, biotinylated IL-2Rγ, and PE-conjugated streptavidin). Data are shown for HER2-targeted intact Neo-2/15 (a) and three split pairs: ɑHER2-H13 + ɑEGFR-H2’4 (b); ɑEGFR-H1 + ɑHER2-H32’4 (that is, ɑEGFR-Neo2A and ɑHER2-Neo2B (c); and ɑHER2-H132’ + ɑEGFR-H4 (d). e, Activity of the individual split Neo-2/15 fragment fusion proteins (810 nM concentration) was tested to evaluate the potential for off-target activity. The dilution series were performed in a single replicate (n = 1). All data are presented as mean values.

Extended Data Fig. 5. Trans-presentation of split-Neo-2/15 on the surface of K562 to YT-1 cells has limited potency in the absence of an immunological synapse.

Untransduced HER2-/EGFR- K562 (off-target) cells or double-positive HER2 + /EGFR + K562 (on-target) cells were cocultured with YT-1 human NK cells in varying K562:YT-1 cell ratios in the presence of ɑHER2-Neo2A and ɑEGFR-Neo2B or other split Neo-2/15 fragment pairs at varying concentrations. STAT5 phosphorylation of YT-1 cells was observed for high K562:YT-1 cell ratios, demonstrating trans-activation of immune cells from the surface of target-expressing cells. The experiments were performed in triplicate three times with similar results. Data are presented as mean values +/− s.d.

Extended Data Fig. 6. Split Neo-2/15 shows “AND” logic-gated targeted activity on B16F10 tumor cells overexpressing HER2 and PD-L1.

a, Depiction of the in vitro assay to determine reconstitution of Split Neo-2/15 binding activity on the surface of B16F10 melanoma cells overexpressing human HER2 and murine PD-L1. Reconstitution of Neo-2/15 binding activity on the cell surface is measured by recruitment of PE-labeled hIL-2Rβγ. Data are presented as mean values +/− s.d. b, Quantification of Split Neo-2/15 reconstitution on the surface of Her2+/PD-L1Hi B16F10 melanoma cells (n = 3). Combination of αHER2-Neo2A and αPD-L1-Neo2B effectively reconstituted Split Neo-2/15 activity, demonstrating AND logic-gated activity on the engineered B16F10 cells. c, Individual tumor growth data for the efficacy study in mice bearing B16 melanoma cells overexpressing murine PD-L1 and human HER2 shown in Fig. 3f. PBS (n = 12), Neo-2/15 (2.6 nmol, n = 5), ɑPD-L1-Neo2A + ɑPD-L1-Neo2B (8 nmol, n = 5), ɑHER2-Neo2A + ɑPD-L1-Neo2B (8 nmol, n = 7). Ṫ indicates surviving mice euthanized due to toxicity when the body condition score of the mice was less than or equal to two in accordance with IACUC Standard Procedure.

Extended Data Fig. 7. Activity of targeted Split Neo-2/15 is restricted to cells with high surface PD-L1 expression.

a, Quantitation of murine PD-L1 expression on B16F10 melanoma cell lines. B16F10 WT melanoma cells are low PD-L1 expressors, B16F10 WT stimulated with IFNγ are medium expressors, engineered B16 PD-L1Hi cells are high expressors. Quantitation was performed using QuantiBRITE phycoerythrin beads. n = 3 technical replicates. Data are presented as mean values +/− s.d. b, Depiction of the in vitro assay to determine reconstitution of Split Neo-2/15 binding activity on the surface of B16F10 melanoma cells with different levels of PD-L1 expression. PD-L1-targeted Neo2A and Neo2B fragments are only reconstituted on the surface of B16 cells with high PD-L1 receptor counts. Reconstitution of Neo-2/15 binding activity on the cell surface is measured by recruitment of PE-labeled hIL-2Rβγ. c, Quantification of Split Neo-2/15 binding activity on the surface of B16F10 melanoma cells with different levels of PD-L1 expression. Active Neo-2/15 on the cell surface is measured by recruitment of PE-labeled hIL-2Rβγ. Intact αPD-L1-Neo2/15 showed IL-2Rβγ binding activity on the surface of cells that were intermediate and high expressors, whereas the αPD-L1-Neo2A + αPD-L1-Neo2B were only able to reconstitute on high expressor cells. n = 3 technical replicates. Data are presented as mean values +/− s.d. d, Efficacy study in C57BL/6 J mice (n = 5/group) bearing WT B16F10 melanoma cells in the flank. Mice were dosed daily with therapeutic doses of mIL-2 (13 µg/mouse), Neo-2/15 (10 µg/mouse and 30 µg/mouse) and αPD-L1-Neo2A + αPD-L1-Neo2B at 200 µg/mouse. All groups were co-treated with TA99 bi-weekly starting on day 3. Survival analyses performed via log-rank Mantel-Cox test. Tumor growth analyses were by unpaired, two-tailed t test. 3 early ulcerated tumors (<500 mm³) removed from this analysis.* indicates P < 0.05; ** indicates P < 0.01, *** indicates P < 0.001, ns = non significant. Data are presented as mean values +/− s.e.m.

Extended Data Fig. 8. AND logic-gated targeted Split Neo-2/15 is active on human tumor cell lines with a wide range of surface receptor expression.

a, Quantitation of human HER2 and EGFR surface expression levels on multiple target cell lines (SKOV3, JIMT-1, OVCAR8, WT K562 and engineered K562 overexpressing HER2 and EGFR). The quantitation was performed using QuantiBRITE phycoerythrin beads. (*) Indicates data reported in ref. . b, Depiction of the in vitro assay performed to determine reconstitution of AND logic-gated Split Neo-2/15 binding activity on the surface of target cell lines with diverse expression of EGFR and HER2. Reconstitution of Neo-2/15 binding activity on the cell surface is measured by recruitment of PE-labeled hIL-2Rβγ. c, Quantification of AND logic-gated Split Neo-2/15 activity on the surface of multiple human tumor cell lines. Combination αEGFR-Neo2A and αHER2-Neo2B effectively reconstituted Split Neo-2/15 activity, demonstrating AND logic-gated activity on the human tumor cell lines. n = 3 technical replicates. d, Split Neo-2/15 also showed activity when both Neo2A and Neo2B fragments were targeted to the same surface tumor antigen, that is, EGFR or HER2. n = 3 technical replicates. Data are presented as mean values +/− s.d.

Extended Data Fig. 9. CD8-targeted Split Neo-2/15 mediates T cell cis-activation.

a-b, Splenocytes from WT C57BL/6 J mice (n = 3) were cultured with the indicated Neo-2/15 and Split Neo-2/15 fusion proteins at the indicated concentrations (control constructs were used at 50 nM). After four days, cells were harvested and analyzed by flow cytometry. The change in the ratio of CD8:CD4 T cells was quantified to measure selective expansion of CD8 + T cells by the targeted constructs (b). c-d, Healthy FoxP3-GFP mice (n = 3 mice, n = 2 for PBS control) were dosed daily with the indicated constructs. Intact Neo-2/15 fusions, Neo2B fusions, and PBS were administered intraperitoneally, while Neo2A constructs were given subcutaneously. After 5 days, the mice were euthanized and their spleens and lymph nodes were harvested. The CD8:CD4 ratios in the spleen cells (c) and lymph node cells (d) were quantified by flow cytometry to assess the extent of selective expansion of CD8 + T cells in vivo. Unpaired two-tailed Student’s t-test. ns = non significant. e, Anti-tumor efficacy of CD8-specific Split Neo-2/15 in a syngeneic mouse model of B16 melanoma. This efficacy experiment was carried out with the same conditions and in parallel to the study shown in Fig. 4c. Mice shown here were not co-treated with Ta99. Data in panels b-d are presented as mean values +/− s.d.; a and e (right panel) are presented as mean values +/− s.e.m.

Extended Data Fig. 10. CAR-T cells elicit antitumor activity in vitro in the presence of cis-targeted Split Neo-2/15.

a-b, Depiction of the assay performed to determine CAR-T cell proliferation and anti-tumor activity in vitro. αROR1 CAR-T cells were cocultured with NCI-H1975 tumor cells in presence or absence of cis-targeted Split Neo-2/15. CAR-T cells expressing HER2 as transfection marker (CAR HER2t) are effectively cis-activated by a combination of αHER2-Neo2A and αHER2-Neo2B, which promotes proliferation and anti-tumor activity (a). CAR-T cells expressing CD19 as transfection marker (CAR CD19t) are not cis-activated by a combination of αHER2-Neo2A and αHER2-Neo2B (b). c, End-point NCI-H1975 tumor cell counts (left) and CAR-T cell counts (right) after each round of co-culture (40 hours). Cells were co-cultured four consecutive times. n = 2 independent cell co-cultures. Data are presented as mean values +/− s.d. d, Kinetic read of NCI-H1975 tumor cell index on each co-culture tumor challenge shown in (c, left).