Identifying splice variants of cell surface proteins for chimeric antigen receptor T cell therapy of Acute Myeloid Leukemia

Abstract Chimeric Antigen Receptor (CAR) T cell therapy in acute myeloid leukemia (AML) has shown modest efficacy to date. Most CAR T cell studies in AML target antigens such as CD33, CLL1 or CD123 which are also expressed on normal cells, raising concern for on-target off-tumor toxicity. As a result, there is no FDA-approved CAR therapy for AML. We develop a target discovery strategy to identify cell surface proteins bearing leukemia-specific domains derived from alternative mRNA splicing. Given the frequency of mutations in splicing factors and epigenetic modifiers in AML, we hypothesize that AML cells generate mis-splicing-derived transcripts coding for cell surface proteins with unique leukemia-specific domains that can be therapeutically targeted. Novel domains can derive from translation of non-coding regions of the genome, novel exon-exon junctions, intron retention or alternative transcription start or termination sites. From RNA-seq analysis, we build a custom transcriptome for isoform quantification with de novo assembly and transcriptome consolidation, enabling quantification of the cell transcriptome with isoform resolution and genome-wide analysis of isoform usage. We prioritize isoforms involved in isoform switches where upregulation of one isoform occurs at the expense of the most dominant form suggesting biological relevance. We estimate functional consequences computationally, run open reading frame (ORF) annotation, assess propensity for non-sense-mediated decay, annotate intron retention and predict sequence-based protein domain gain or loss, coding potential, signal peptide gain or loss, and changes in aminoacid sequence. We manually curate to exclude non-coding isoforms, single exon isoforms and lncRNAs with predicted ORFs. We further select isoforms derived from genes coding for cell surface proteins. We apply thispipeline to RNA-seq data of human primary cord blood CD34+ hematopoietic stem and progenitor cells (HSPCs)lentivirally expressing one or two recurrent driver AML mutations (i.e., SRSF2P95H and IDH2R140Q). These models represent a system to investigate mutation-dependent effects in otherwise normal human primary cells. FACS-sorted GFP+ HSPCs expressing IDH2R140Q, SRSF2P95H, and SRSF2P95H + IDH2R140Q or corresponding wild-type genes (i.e., IDH2WT, SRSF2WT, or SRSF2WT + IDH2WT) or empty vector as controls are analyzed.The group of SRSF2-mutant cells presents most mis-spliced transcripts compared to the group of IDH2-mutant cells or double mutants. We identify five isoforms that are upregulated compared to controls, involved in at least one isoform switch, derived from genes coding for cell surface proteins, have coding potential and contain unique sequences absent in the relative full-length forms. Novel splice variants of solute carrier family 12 member 7 (SLC12A7) and low-density lipoprotein receptor-related protein 5 (LRP5) are detected with high-read coverage in many AML patients from the TCGA (n=179) and Leucegene (n=263) cohorts. Three RNA-seq datasets of normal bone marrow (BM) CD34+ cells show <1% coverage with less than 3 reads, suggesting leukemia specificity. We develop monoclonal antibodies recognizing one novel sequence of the SLC12A7 splice variant (sv-SLC12A7). We analyze a cohort of 36 BM primary AML patient samples by flow-cytometry and demonstrate that this unique protein domain is detectable on blast cells (mean expression 45%) and CD34+CD38- leukemic stem cells (LSCs) (mean expression 62%). Expression in leukemic cells is significantly higher than normal CD34+ HSPCs (mean expression <5%). sv-SLC12A7 expression levels are higher in patients bearing epigenetic mutations not restricted to patients bearing SRSF2 mutation and impact clinical outcomes. Absent/minimal isoform expression in normal non-hematopoietic tissues is defined by GTEx and immunofluorescence analyses on normal tissues.Finally, we develop splice variant-directed CAR-T cells bearing a BBz architecture and demonstrate effective eradication of THP1 leukemic cells in a xenograft mouse model using NSG immunocompromised mice. This work suggests the potential of advancing current CAR therapy in AML by targeting alternatively spliced cell surface proteins.