Cars in Leukemia: Relapse with Antigen-Negative Leukemia Originating from a Single B Cell Expressing the Leukemia-Targeting CAR

T cells bearing a second-generation anti-CD19 chimeric antigen receptor (CAR) induce complete remission in >90% of patients with acute lymphoblastic leukemia (ALL) at our institution. However, disease may recur and we recently identified two molecular mechanisms of relapse (PMID: 26516065). We here present a novel mechanism of antigen-negative relapse in a pediatric ALL patient. A 21 year-old male patient was in third relapse at the time of enrollment onto our CTL019 trial (ClinicalTrials.Gov #NCT01626495). The patient achieved an MRD-negative complete remission 1 month after CTL019 infusion but relapsed nine months later. Quantitative PCR analysis of the transgene and flow cytometry for CAR19 protein analysis showed the expected expansion of the CART cells followed by log-normal decay following disease eradication. At relapse, however, the transgene copy numbers had increased without a concomitant rise in CAR19 protein-expressing T cells. The CAR protein was found to be expressed by the now CD19-negative CD45^dimCD10⁺CD3^negCD22⁺ leukemia. Molecular analysis via next-generation immunoglobulin heavy chain sequencing (NGIS) of the apheresis product, used for CTL019 manufacturing, and relapse marrow at 9 months demonstrated clonal identity of the relapsed clone, which carried two rearranged IgH alleles. Sequencing of the CD19, CD21, CD81, and CD225 loci did not reveal any mutations. The analysis of lentiviral vector integration sites (LVIS) of the infusion product and post-infusion specimens showed the following: a) the infusion product carried over 15,000 unique integration sites; b) only 7 LVIS were shared between this sample and month 9 and 20 relapse specimens, none of which were near proto-oncogenes; c) the relapsed leukemia carried two LVIS, one on chromosome 10, >50 kb distal from neuropilin (NRP1) and the second in an intron of proprionyl coenzyme A carboxylase-A (PCCA). Flow cytometric and qRT-PCR analysis of leukemic cells in the apheresis and relapse showed that NRP1 levels were indistinguishable, suggesting that the lentiviral vector did not act as an enhancer for NRP1. Furthermore, qRT-PCR demonstrated that the lentiviral integration did not affect the gene expression levels of PCCA. Investigation into the origins of the leukemic CAR transduction event showed that the patient did not exhibit replication-competent lentivirus. However, NGIS analysis of infusion product revealed the leukemic clonotypes this sample, indicating that the gene transfer occurred during the manufacturing of the CTL019 cells. A retrospective analysis of 115 aphereses from ALL patients showed that the index patient had an unusually high disease burden in the apheresis product with 63% of all cells expressing CD19; at harvest, however, the CTL019 product consisted of 99.21% T cells, highlighting the purging effect of the CD19-specific T cells during manufacturing. NGIS analysis of infusion products of 17 additional ALL patients also identified the leukemic clonotype(s) in 6 more products. Only one additional patient demonstrated CAR19 protein expression on the leukemic cells, and this clone was not dominant at relapse (0.075% of all leukemic cells expressed the CAR). Our investigation into the biology of CAR19-expressing ALL cells showed the following: 1) the in vitro analysis of BBζ-signaling CAR19 showed no evidence of cytokine secretion; 2) the infusion of the baseline leukemia and CAR19-expressing leukemic cells from the same patient in mice did not demonstrate differential pharmacodynamics, even after restimulation with human CD19-expressing murine B cells in vivo; 3) the CD19 protein was detectable using flow cytometry and confocal microscopy, but only with an antibody recognizing an intracellular epitope; 4) importantly, the relapsed clone was indeed resistant to killing by CART19 cells in a xenograft model yet retained sensitivity to anti-CD22 CAR T cells. In conclusion, our data therefore show that a single leukemic cell accidentally transduced with CAR19 survived the 10-day manufacturing process and, upon reinfusion into the patient, was the sole clone at relapse 9 months later. This leukemic clone evaded CTL019 detection via downregulation of the target antigen in a cell-autonomous fashion.