Dual targeting of CD19 and CD22 with bicistronic CAR-T cells in patients with relapsed/refractory large B-cell lymphoma

Abstract

Relapse after CD19-directed chimeric antigen receptor T-cell (CAR-T) therapy for large B-cell lymphoma (LBCL) is commonly ascribed to antigen loss or CAR-T exhaustion. Multiantigen targeting and programmed cell death protein-1 blockade are rational approaches to prevent relapse. Here, we test CD19/22 dual-targeting CAR-T (AUTO3) plus pembrolizumab in relapsed/refractory LBCL (NCT03289455). End points include toxicity (primary) and response rates (secondary). Fifty-two patients received AUTO3 and 48/52 received pembrolizumab. Median age was 59 years (range, 27-83), 46/52 had stage III/ IV disease and median follow-up was 21.6 months. AUTO3 was safe; grade 1-2 and grade 3 cytokine release syndrome affected 18/52 (34.6%) and 1/52 (1.9%) patients, neurotoxicity arose in 4 patients (2/4, grade 3-4), and hemophagocytic lymphohistiocytosis affected 2 patients. Outpatient administration was tested in 20 patients, saving a median of 14 hospital days per patient. Overall response rates were 66% (48.9%, complete response [CR]; 17%, partial response). Median duration of remission (DOR) for CR patients was not reached and for all responding patients was 8.3 months (95% confidence interval [CI]: 3.0-not evaluable). 54.4% (CI: 32.8-71.7) of CR patients and 42.6% of all responding patients were projected to remain progression-free at ≥12 months. AUTO3 ± pembrolizumab for relapsed/refractory LBCL was safe and delivered durable remissions in 54.4% of complete responders, associated with robust CAR-T expansion. Neither dual-targeting CAR-T nor pembrolizumab prevented relapse in a significant proportion of patients, and future developments include next-generation-AUTO3, engineered for superior expansion in vivo, and selection of CAR binders active at low antigen densities.

Graphical abstract

Figure 1.

AUTO3 study design and recruitment. (A) AUTO3 CAR, CD19 CAR, and CD22 CAR are type I transmembrane proteins. Human HD37 recognizes CD19, is located at the extreme amino terminus and, in turn, connected to the CD8α stalk, transmembrane, and anchor. The intracellular endodomains are composed of costimulatory OX40 and CD3 ζ (TCRζ). Humanized LT22 recognizes CD22, is located at the extreme amino terminus and is connected to the pentameric α-helical coiled coil multimer forming domain of cartilage oligomeric matrix protein (COMP) and transmembrane anchor. The intracellular endodomains are composed from 41BB ζ and CD3 ζ (TCRζ). (B) AUTO3 trial schema. (C) AUTO3 consort diagram. Cy, cyclophosphamide; Flu, fludarabine; NT, nontransduced; scFV, single chain variable fragment; TCRz, T-cell receptor zeta.

Figure 2.

AUTO3 engraftment and persistence. (A) Peak AUTO3 using qPCR and AUTO3 dose or timing of pembrolizumab administration. (B) Peak AUTO3 using qPCR, and pre-LD disease burden using SPD <20 vs ≥20. Data missing for 3 patients. (C) Peak AUTO3 using qPCR and baseline CD19/22 status, in which CD19⁺ or CD22⁺ is reflected through H-score ≥150. CD19^lo/– and CD22^lo/– represent H-scores <150. (D) Peak AUTO3 using qPCR and disease response at month 1. (E) AUTO3 persistence using qPCR and timing of pembrolizumab. (F) AUTO3 persistence using qPCR and Tn and Tcm populations in the drug product for patients with short engraftment ≤28 days. (G) AUTO3 persistence for all patients using qPCR based on the dose. Pem, pembrolizumab. (H) AUTO3 persistence for all patients using qPCR based on the response.

Figure 3.

AUTO3 disease response. Swimmer plot, for total infused cohort. SCT, stem cell transplant; N, no; Y, yes.

Figure 4.

Duration of response. (A) Duration of response among all responders. (B) DOR based on best overall response. (C) PFS of all patients who received infusion. (D) OS of all patients who received infusion.