A Rational Strategy for Reducing On-Target Off-Tumor Effects of CD38-Chimeric Antigen Receptors by Affinity Optimization

Abstract

Chimeric antigen receptors (CARs) can effectively redirect cytotoxic T cells toward highly expressed surface antigens on tumor cells. The low expression of several tumor-associated antigens (TAAs) on normal tissues, however, hinders their safe targeting by CAR T cells due to on-target/off-tumor effects. Using the multiple myeloma (MM)-associated CD38 antigen as a model system, here, we present a rational approach for effective and tumor-selective targeting of such TAAs. Using "light-chain exchange" technology, we combined the heavy chains of two high-affinity CD38 antibodies with 176 germline light chains and generated ∼124 new antibodies with 10- to >1,000-fold lower affinities to CD38. After categorizing them into three distinct affinity classes, we incorporated the single-chain variable fragments of eight antibodies from each class into new CARs. T cells carrying these CD38-CARs were extensively evaluated for their on-tumor/off-tumor cytotoxicity as well as CD38-dependent proliferation and cytokine production. We identified CD38-CAR T cells of ∼1,000- fold reduced affinity, which optimally proliferated, produced Th1-like cytokines, and effectively lysed CD38²⁺ MM cells, but spared CD38⁺ healthy hematopoietic cells in vitro and in vivo. Thus, this systematic approach is highly suitable for the generation of optimal CARs for effective and selective targeting of TAAs.

Keywords: CAR design; CD38; affinity; chimeric antigen receptor; multiple myeloma; off-target effects; scFV.

Figure 1

Selection of Low-Affinity CD38 Antibodies (A) Schematic cartoon of light-chain exchange method and overview of antibody generation and selection process. (B) Representative graphs of binding assays performed with CHO cells transfected with CD38 (CHO-CD38) or WT. His-CD38 are beads coated with recombinant CD38 and detected with IgG or Fab fragments with fluorochrome beads. Rows indicate WT antibody 028 and class A–C, as defined by A, positive binding in all assays; B, positive in the cell binding and IgG; and C, only positive binding on CHO-CD38. (C) Quantified EC₅₀ values (mg/mL) of different binding assays. (D) Molecule association and dissociation curves. Interferometric profile shifts are measured and its magnitude is plotted as a function of time. (E) Interferometric profiles of the antibodies are quantified into K_D values (M) and on and off rates (k_on [1/Ms] and k_off [1/s]). n = 2 ± SD.

Figure 2

Lytic Capacity of Different Affinity CD38-CAR T Cells Lysis of cell line UM9 by different affinity CD38-CAR T cells when co-incubated with luciferase-transduced MM cell line UM9 for 16 hr; cytotoxicity was measured with BLI, n = 2. Graphs are divided into three affinity subcategories. Class A CARs are derived from class A antibodies, with the highest affinity, to class C, with the lowest affinity. CARs with the 028 VH are numbered 1–4 in each class, and CARs with 024 VH are numbered 5–8.

Figure 3

Phenotypic Profiles of Lower Affinity CD38-CAR T Cells (A) 24 hr after co-incubation with the CD38⁺ target cell line UM9 or CD38⁻ target U266, E:T ratio 1:1, cytokine secretion by mock or CD38-CAR028, A1, A4, B1, or B3 T cells was measured with a flow-cytometry-based assay in the cell-free supernatants. Graph shows the secretion of IFN-γ, TNF, and IL-2. n = 2, mean ± SEM; *p < 0.05 and **p < 0.01 using one-way ANOVA and subsequent multiple comparison. (B) CD38-CAR T cells were stimulated with MM target UM9 E:T ratio 1:3 1 week after being transduced and followed weekly. Cells were counted, and % of CAR⁺ cells was determined by flow cytometry. Figure indicated growth of CAR⁺ cells in the culture. ●, mock and open squares; □, CD38-CAR028; ▲, CARA1; ♦, CARA4; ▼, CARB1; ●, CARB3. n = 2 mean ± SEM; ns, not significant. (C) Phenotypic profile of each CD38-CAR T cell type was determined before (week 0) and after (week 1) expansion with markers CD45RA and CD62L. Percentage of total cells is depicted for naive (CD45RA⁺/CD62L⁺), central memory (CM) (CD45RA⁻/CD62L⁺), effector memory (EM) (CD45RA⁻/CD62L⁻), and effector (CD45RA⁺/CD62L⁻). n = 2 mean ± SEM. Statistical analysis was done using one-way ANOVA and subsequent multiple comparison.

Figure 4

Lytic Capacity of Lower Affinity CD38-CAR T Cells toward Multiple Myeloma Bone Marrow Bone marrow (BM-MNCs) samples of four MM patients with 20%–40% MM cells were co-incubated, mock or CD38-CAR028, A1, A4, B1 or B3 T cells for 16 hr. (A–C) The graphs depict the resulting lysis of CD138⁺/CD38⁺ cells (MM) (A), CD38⁺ MNCs (non-MM) (B), and total MNCs (C) in E:T (E:BM-MNC) ratio 1:1. Representative figure of other ratios is shown in Figure S5. ●, mock and open squares; □, CD38-CAR028; ▲, CARA1; ♦, CARA4; ▼, CARB1; , CARB3. The % lysis in these flow cytometry assays was calculated as described in the methods section. n = 4, median ± range; *p < 0.05 and **p < 0.01 using Kruskal-Wallis ANOVA and subsequent Mann-Whitney comparison. (D) Flow cytometry dot plots depicting MM-BM with CD138⁺/CD38²⁺ cells (MM) and CD138⁻/CD38⁺ healthy cells. The CD38 threshold for lysis is indicated with two horizontal bars on voltage 10³ (threshold for high-affinity CD38-CAR028) and 10⁴ for some of the lower affinity CD38-CARs.

Figure 5

CARA4 Is the Best Candidate for a Lower Affinity CAR PBMCs of a healthy donor were co-incubated with mock, high-affinity CD38-CAR028 or low-affinity CD38-CARA4 T cells for 16 hr. The graphs depict the resulting lysis of CD3⁺ (T cells), CD56⁺ (NK cells), or CD19⁺ (B cells), their total, or CD38⁺ fraction. ●, mock; □, CD38-CAR028; ▲, CARA1; ♦, CARA4. The % lysis in these flow cytometry assays was calculated as described in Materials and Methods.

Figure 6

High- and Low-Affinity CD38-CAR T Cells Are Similarly Effective In Vivo Mice were i.v. injected with 10 × 10⁶ cells of tumor cell line UM9 and treated 1 week after with i.v. injections of 5 × 10⁶ mock, high-affinity CD38-CAR028, or low-affinity CD38-CARA4 T cells. (A) BLIs are shown per group of each week. (B) Analysis of tumor load in mice by quantification of BLI measurements. Each group contained four mice, each harboring four scaffolds. ●, mock; □, CD38-CAR028; ♦, CD38-CARA4. N = 4. Results are median tumor load (cpm/cm²) of four mice ± range; *p < 0.05; **p < 0.01; and ***p < 0.001 using Kruskal-Wallis ANOVA.

Figure 7

High-Affinity CD38-CAR T Cells Affect CD34 Hematopoietic Progenitor Cells (A) Mice were intrascaffold injected with 1 × 10⁶ fluorescently (FarRed) labeled CD34⁺ and treated 1 week after with i.v. injections of 5 × 10⁶ mock, high-affinity CD38-CAR028 or low-affinity CD38-CARA4 T cells. (A) FLIs are shown per group at week 1 and 3. (B) Percentages of CD34^+/− and CD38^+/− cells in the total population of FarRed⁺ cells, which were present in the scaffold in post-mortem tissue samples analyzed by flow cytometry. Results % CD34^+/− and CD38^+/− of four mice per group, two scaffold per mouse ± range; *p < 0.05 using Mann-Whitney test on CD38⁺ percentages. (C) Total number of manually counted cells were adjusted using the percentage of FarRed⁺ cells, as measured by flow cytometry and the percentage of CD34^+/− and CD38^+/− in (B). N = 4 mice per group, median ± range.