Detailed Multi-Method Analysis of Bone Marrow from Pediatric Pre-B-ALL Patients Prior to CD19-CAR-T Therapy Subsequently Evidencing Overt CAR-T Resistance

Little is known about the disease-specific factors that predict responsiveness to CAR-T cell therapy, other the direct presence of the CAR-T target. Clinical outcomes at our center have demonstrated that durable responses to CD19-directed CAR-T therapy in pediatric pre-B-ALL (acute lymphoblastic leukemia) are associated with persistence of CAR-T cells in the peripheral blood, antigenic load (percent CD19-positive cells in marrow prior to CAR-T infusion), and apheresis product T-cell quality (Finney O, et al., 2019). However, in a small number of cases where both antigenic load and T-cell quality predicted a good response, the treatment failed rapidly. This led us to undertake a detailed investigation of the leukemia itself in order to discover potential disease-associated factors that correlate with resistance to CAR-T therapy.

We employed advanced exomic, and single-cell genomic and epigenomic analysis techniques to define signatures present in four CD19-CAR-T resistant bone marrow biopsy specimens, in comparison to five specimens from CD19-CAR-T responsive disease, from patients enrolled in a phase I clinical trial at Seattle Children's Hospital (PLAT-02, NCT02028455). Current cytogenic approaches to identify high risks markers (Ph+, Ph-like, MLL) were not informative as to CAR-T susceptibility, as high risk leukemias were CAR-T responsive; while a CAR-T resistant leukemia contained a marker (ETV6-RUNX1 fusion) previously associated with a good prognosis. Thus, we performed bulk whole-exome sequencing and RNAseq, single cell (sc) RNAseq, sc B-cell receptor (BCR)-seq, methylation array, H3K27ac ChIPseq, and ATACseq on these marrow samples.

Initial genomic analysis revealed a total of 5 previously described hotspot mutations in ABL1, IKZF1, EP300, and 2 in KRAS. RNAseq analyses identified actionable fusions for ABL1, ETV6, ETV5, and KMT2A. Interestingly, a therapy-sensitive leukemia harbored a KMT2A-AFF1 fusion that was shown to predispose patients to leukemic plasticity and lineage switching when treated with blinatumomab. Importantly, we identified CREBBP-fusions in leukemias that failed to achieve CD19-CAR-T cell induced B cell aplasia. CREBBP perturbations have previously been associated with relapsed and refractory ALL, but not with resistance to CAR-T therapy.

Single cell RNAseq and scBCRseq data are being analyzed for the existence of mixed lineage and gene expression-based heterogeneity that may predict clonal selection under CAR-T pressure. RNASeq analysis identified upregulation of JUN and JUND transcripts in CAR-T resistant disease, a finding which is complemented by the hypermethylation of JUND in CAR-T sensitive disease. Similarly, ATACseq and methylation data is being analyzed for lineage specification in CAR-T resistant leukemia. In comparing dysfunctional to functional CAR-T responders by ATACseq, > 10,000 unique open chromatin regions were identified in dysfunctional responders, as opposed to <500 open chromatin regions in the functional responders, indicating that CAR-T resistant disease had more open chromatin. A recent published analysis of cancer cell lines identified EP300 and CREBBP mutations that were proposed to increase substrate acetylation, and which "…may represent the first cancer-associated gain of function mutations for p300 and CBP…" (Ghandi M, et al, 2019). Our study represents one of the most comprehensive approaches to genomic profiling for B-ALL patient samples to date. The immune evasion we have described is not due to overt CD19 antigen loss, or to a long-term process of genetic alteration or drift. We propose that continued analysis of our data may reveal that epigenetic plasticity is a component of CAR-T resistance. The presence of CREBBP fusion genes or mutations, methylation array-based identification of altered JUND/JUN regulation, and ATAC Seq identification of multiple open regions of the genome in leukemias from dysfunctional responders lend support to this hypothesis. Although our analysis is preliminary and the sample number is small, we believe these in-depth analyses will highlight crucial differences in leukemia that predict responsiveness to CAR T therapy