Development of CAR T-cell lymphoma in 2 of 10 patients effectively treated with piggyBac-modified CD19 CAR T cells

TO THE EDITOR:

CD19 -specific chimeric antigen receptor (CAR19) T cells induce remission of B cell malignancy,^1-6 but manufacture of individualized autologous products with viral vectors is expensive,⁷ delays treatment, risks contamination with malignant B cells,⁸ and is not feasible in patients with insufficient healthy T cells.⁹ Generation of CAR19 T cells using an allogeneic donor is a promising alternative strategy, but limited data exist regarding safety and efficacy.^1-3,10,11

We report first-in-human administration of CAR19 T cells manufactured using the piggyBac transposon system, an inexpensive, high-capacity, nonviral alternative for CAR T-cell generation.^12-14 Patients with relapsed or refractory CD19⁺ malignancy following HLA-matched sibling donor hematopoietic stem cell transplant (HSCT) were assessed for enrollment in a phase 1 clinical trial of escalating doses of HSCT donor-derived allogeneic piggyBac CAR19 T cells (registered at www.anzctr.org.au as ACTRN12617001579381) (Figure 1A and supplemental Methods, available on the Blood Web site). All 10 treated patients (Table 1) were included in this analysis irrespective of trial inclusion (supplemental Figure 1).

CAR19 T-cell products were generated with piggyBac in 15 days (n = 9) or 23 days (n = 1) and contained a median of 72.5% CAR19 T cells (range, 38.7% to 92.9%) (supplemental Figure 2; supplemental Methods). Most CAR19 T cells had CD62L⁻CD45RA⁻ effector memory phenotype (80.0% ± 7.1%; mean plus or minus standard deviation), and fewer had CD62L⁺CD45RA⁻ central memory phenotype (17.6% ± 6.0%). Many expressed the inhibitory marker TIM3 (55.8% ± 17.9%). Description of integration sites is detailed elsewhere (Micklethwaite et al¹⁵) but only 7% of such sites occurred within cancer-related genes and were not different to those previously described for piggyBac or viral vectors.

Nine patients survived ≥28 days and were evaluable for response (Figure 1B), with a median follow-up of 18.0 months (range, 6.0-29.2 months). The complete remission (CR) rate was 100% (example shown in supplemental Figure 3). Two patients underwent a second allogeneic HSCT while in sustained CR. At cutoff, 5 patients remained in CR of B-cell malignancy (longest, 27.4 months ongoing). Overall survival ranged from 6.0 to 29.2 months ongoing (median not reached).

Four patients relapsed, each with undetectable CAR19 T cells (median, 7.2 months). Two were eligible to receive further escalating CAR19 T-cell doses postrelapse, but did not respond. One developed CD19⁻ B cell acute lymphoblastic leukemia (B-ALL); the other lost donor marrow chimerism and may have had allorejection of donor-derived CAR19 T cells.

Two patients developed CAR19 T-cell lymphoma after achieving CR of B-cell malignancy. In each case, the diagnosis was based on identification of a new mass composed of monoclonal CAR19 T cells (according to T-cell receptor variable β chain analysis) with abnormal phenotype, which continued to increase in size and metabolic activity in the absence of residual CD19 antigen (Figure 1C; see Micklethwaite et al¹⁵ for detailed analysis). This was first detected in patient 2 on a positron emission tomography–computed tomography (PET-CT) scan at 3.2 months, but the diagnosis was only established at 15.0 months, after multiple nondiagnostic biopsies. Treatment with cyclophosphamide, vincristine, and prednisolone led to transient response, but complications of sepsis and multiorgan failure resulted in death at 18.0 months. Surviving patients subsequently underwent PET-CT screening; patient 8 was also confirmed to have CAR19 T-cell lymphoma at 11.5 months. CR was induced with 4 cycles of augmented cyclophosphamide, doxorubicin or etoposide, vincristine, and prednisolone (CHOP). He proceeded to a second allogeneic HSCT at 16.9 months and remains in remission of both B-ALL and the CAR19 T cell lymphoma at 23.2 months.

Other adverse events are summarized in Table 1. Five deaths occurred, including 2 treatment-related deaths and 2 from relapsed B-ALL. Patient 6 died on day 3 of treatment from pulmonary embolus secondary to extensive Burkitt lymphoma surrounding abdominal vessels. Four patients with high disease burden developed cytokine release syndrome (CRS) of grade ≤2,¹⁶ with elevated serum cytokines (supplemental Figure 4). Patient 2 developed exacerbation of pre-existing chronic graft-versus-host disease (GVHD) of the skin, managed with prednisone and ibrutinib. There was no immune effector cell–associated neurotoxicity syndrome (ICANS), tumor lysis syndrome, new GVHD, or early dose-limiting toxicity.

CAR19 T cells expanded in peripheral blood of all evaluable patients (Figure 1D). Following the first infusion, CAR19 T cells peaked at a median of 16.4 CAR19 T cells per μL (range, 0.9-1824) after a median of 16 days (range, 8-24 days). Greater CAR19 T-cell expansion was noted with a high disease burden, a high percentage of CAR19 T cells in the product, and the development of CRS (supplemental Figure 5). After 1 infusion, CAR19 T cells were detectable in peripheral blood for up to 4.9 months. Patient 9, who had 3 escalating doses, had continuously detectable CAR19 T cells until his death at 6.0 months, representing the longest period seen in patients not developing CAR19 T-cell lymphoma.

In this first-in-human phase 1 clinical trial, piggyBac CAR19 T cells generated from HLA-matched sibling allogeneic donors induced CR, some sustained, in all evaluable patients with relapsed and refractory CD19⁺ malignancy following HSCT. However, 2 patients developed CAR19 T-cell lymphoma, raising concern about the safety of the piggyBac CAR19 T-cell production methodology (see Micklethwaite et al¹⁵).

Like other 4-1BB–based autologous products generated with viral vectors, piggyBac CAR19 T cells expanded in proportion to disease burden, persisted for months, and effectively induced minimal residual disease–negative (MRD⁻) CR of B-ALL.^4-6 Similarly, patients with high disease burden receiving piggyBac CAR19 T cells developed CRS that was associated with elevation in cytokines but that responded to tocilizumab and dexamethasone.^16,17

We observed no ICANS or CRS grade >2. This may be a result of the intrapatient dose-escalation strategy intended to avoid large CAR T-cell doses in the presence of high disease burden, a situation known to be associated with development of severe CRS and ICANS.^17,18 All patients with B-ALL outside of the central nervous system in this study achieved MRD⁻ remission after the smallest dose of piggyBac CAR19 T cells, but 4 patients subsequently relapsed. The utility of repeat escalated CAR19 T-cell dosing remains unclear.

CAR19 T cell malignancy has not previously been documented following administration of CAR T cells produced using viral vectors, other transposon systems, or piggyBac.^{1-6,10,11,19,20} This led to the death of 1 patient in our study; however, the second patient, who remains in CR, was responsive to standard T-cell lymphoma management. Potential drivers of malignant transformation include CAR overexpression with basal activation, prolonged stimulation through the CAR or endogenous T-cell receptor leading to secondary mutational events, or transgene insertional mutagenesis. Investigations of the pathogenesis performed to date (described by Micklethwaite et al¹⁵) do not demonstrate classical insertional mutagenesis as the cause,²¹ with the pattern of integration of the CAR transgene at genomic sites by piggyBac consistent with previous reports, and not appreciably different to that seen with γ-retroviral and lentiviral vectors.^22-24 

With this first cohort of human subjects, we demonstrate that allogeneic piggyBac CAR19 T cells administered post-HSCT achieve the efficacy seen with CAR T cells generated using viral vectors. However, recruitment to trials of piggyBac CAR19 T cells at our center has been voluntarily suspended while further investigations into the pathogenesis of CAR19 T-cell lymphoma take place. Prevention of future malignant events will require changes to the production process, with animal safety testing to provide confidence before contemplating resumption of trials. The ability of piggyBac to integrate a large amount of genetic material, and the absence of a requirement for detection of competent retrovirus or lentivirus in the final product, remain attractive characteristics of this system. Determining the cause of the malignant transformation we have observed will be key to resolving whether it has a future place in human trials.