Exploiting a natural conformational switch to engineer an interleukin-2 'superkine'

Abstract

The immunostimulatory cytokine interleukin-2 (IL-2) is a growth factor for a wide range of leukocytes, including T cells and natural killer (NK) cells. Considerable effort has been invested in using IL-2 as a therapeutic agent for a variety of immune disorders ranging from AIDS to cancer. However, adverse effects have limited its use in the clinic. On activated T cells, IL-2 signals through a quaternary 'high affinity' receptor complex consisting of IL-2, IL-2Rα (termed CD25), IL-2Rβ and IL-2Rγ. Naive T cells express only a low density of IL-2Rβ and IL-2Rγ, and are therefore relatively insensitive to IL-2, but acquire sensitivity after CD25 expression, which captures the cytokine and presents it to IL-2Rβ and IL-2Rγ. Here, using in vitro evolution, we eliminated the functional requirement of IL-2 for CD25 expression by engineering an IL-2 'superkine' (also called super-2) with increased binding affinity for IL-2Rβ. Crystal structures of the IL-2 superkine in free and receptor-bound forms showed that the evolved mutations are principally in the core of the cytokine, and molecular dynamics simulations indicated that the evolved mutations stabilized IL-2, reducing the flexibility of a helix in the IL-2Rβ binding site, into an optimized receptor-binding conformation resembling that when bound to CD25. The evolved mutations in the IL-2 superkine recapitulated the functional role of CD25 by eliciting potent phosphorylation of STAT5 and vigorous proliferation of T cells irrespective of CD25 expression. Compared to IL-2, the IL-2 superkine induced superior expansion of cytotoxic T cells, leading to improved antitumour responses in vivo, and elicited proportionally less expansion of T regulatory cells and reduced pulmonary oedema. Collectively, we show that in vitro evolution has mimicked the functional role of CD25 in enhancing IL-2 potency and regulating target cell specificity, which has implications for immunotherapy.

Figure 1. In vitro evolution of human IL-2 variants with high affinity for IL-2Rβ

a, IL-2 displayed on yeast recapitulates cooperative receptor-binding activity. As measured by flow cytometry, IL-2 binds weakly to IL-2Rβ (left panel), undetectably to γ_c (middle panel), and cooperatively forms the IL-2Rβ/γ_c heterodimer (right panel). b, Enrichment of IL-2 variants on yeast by selection with progressively lower concentrations of IL-2Rβ. Arrows indicate an emerging population of high affinity IL-2Rβ binders (see also Supplementary Fig. 2). c, Sequences and affinities for IL-2Rβ of selected mutants from the first (mutant 6-6) and second (mutants D10 and H9) generation libraries (see Supplementary Fig. 3 for an extended table). d, On-yeast stimulation of YT-1 cells (human NK cell line) by wild-type (WT) IL-2-yeast and high affinity variants (super-2s) (see also Supplementary Fig. 4).

Figure 2. Basis of affinity enhancement for IL-2Rβ from structural and molecular dynamics characterization of D10 super-2

a, Crystal structure of the D10 super-2 at 3.1 Å with mutated residues in red (see also Supplementary Table 1 and Supplementary Fig. 7a). b, D10 in complex with human IL-2Rβ and γ_c preserves the WT receptor dimer geometry (see also Supplementary Fig. 7b). c, The unliganded D10 super-2 helix C (brown), moves towards its hydrophobic core compared to unliganded WT IL-2 (green, PDBID 3INK). This helix C position is more similar to that of helix C in IL-2 bound to IL-2Rα (purple, PDBID 1Z92) (see also Supplementary Fig. 8). d, A 40ns MD simulation shows a reduction of the average RMSD for the B and C helices, and the B-C loop in D10 versus IL-2 (see also Supplementary Fig. 8c). Error bars represent the standard error of the RMSD. e, Helix C in IL-2 (green, left panel) drifts during the MD simulation more than super-2 D10 (brown, right panel) when compared to IL-2 bound to IL-2Rα (purple).

Figure 3. Functional properties of super-2 on human NK cells in vitro

Dose-response curves of STAT5 phosphorylation on (a) CD25⁻ and (b) CD25+ YT-1 cells with WT IL-2 and three super-2s. c, Dose-response curves of STAT5 phosphorylation on CD25⁺ (solid curve) and CD25⁻(dashed curve) YT-1 cells with WT IL-2 (pink curves) and IL-2-F42A mutation (purple curves). d, Dose-response curves of STAT5 phosphorylation on CD25⁺ (solid curve) and CD25⁻(dashed curve) YT-1 cells with H9 (green curves) and H9-F42A mutation (purple curves). e, Super-2s have superior potency over IL-2 on T cells derived from CD25^−/− mice as demonstrated by dose-response curves for STAT5 phosphorylation on T cells demonstrating that potency correlates with IL-2Rβ affinity (see also Supplementary Fig. 10). f, Proliferation of human naïve CD4⁺ T cell (CD25^low) reveals similar potency profiles as seen with CD25^−/− T cells. Proliferation was measured by CFSE dilution on day 5 (see also Supplementary Fig. 10). Error bars in a–d represent SEM of mean fluorescence units for each sample at the indicated cytokine concentration.

Figure 4. Functional and anti-tumor activities of super-2 in vivo

a, Total cell counts of host CD3⁺ CD8⁺ CD44^high memory-phenotype T cells (MP CD8⁺, closed bars), and host CD3⁺ CD4⁺ CD25^high T cells (Treg, open bars) was determined in the spleens of mice receiving either PBS, 20 μg IL-2, 1.5 μg IL-2/anti-IL-2 mAb complexes (IL-2/mAb), or 20 μg H9 (see also Supplementary Fig. 14). b, Pulmonary edema (pulmonary wet weight) served to assess adverse toxic effects following IL-2 treatment, and was determined by weighing lungs before and after drying. c–f, C57BL/6 mice (n=3–4 mice/group) were injected either subcutaneously with 10⁶ B16F10 melanoma cells (B16, c), 2.5× 10⁶ murine colon carcinoma 38 (MC38, d), 10⁶ Lewis lung carcinoma (LLC, e), or mice received 3× 10⁵ B16F10 melanoma cells intravenously (B16, f), followed by daily injections of either PBS, 20 μg IL-2, 1.5 μg IL-2/mAb complexes, or 20 μg H9 for five days once subcutaneous tumor nodules became visible and palpable or from day three on for intravenously-injected tumors (see also Supplementary Fig. 15). Shown is mean tumor volume in mm³ (+/− SD) vs. time upon tumor inoculation. Error bars represent SEM. P values refer to comparisons of WT with the other treatment modalities. *, p<0.05; **, p<0.01; ***, p<0.001.