Synthetic De Novo Modeling of Human T-Cell Acute Lymphoblastic Leukemia Reveals HOXB Genes Drive Expansion of Leukemia Cells-of-Origin and Established Tumor Clones

Mechanistic studies in human cancer have relied heavily on established cell lines and genetically engineered mouse models, but these are limited by in vitro adaptation and species context issues, respectively. More recent efforts have utilized patient-derived xenografts (PDX); however, as an experimental model these are hampered by their variable genetic background, logistic challenges in establishing and distributing diverse collections, and the fact they cannot be independently reproduced. We report here a completely synthetic, efficient, and highly reproducible means for generating T-cell acute lymphoblastic leukemia (T-ALL) de novo by lentiviral transduction of normal CD34+ human cord blood (CB) derived hematopoietic progenitors with a combination of known T-ALL oncogenes. Transduced CB cells exhibit differentiation arrest and multi-log expansion when cultured in vitro on OP9-DL1 feeders, and generate serially transplantable, aggressive leukemia when injected into immunodeficient NSG mice with latencies as short as 80 days (median 161 days, range 79-321 days). RNA-seq analysis of synthetic CB leukemias confirmed their reproducibility and similarity to PDX tumors, while whole exome sequencing revealed ongoing clonal evolution in vivo with acquisition of secondary mutations that are seen recurrently in natural human disease.

The in vitro component of this synthetic system affords direct access to "pre-leukemia" cells undergoing the very first molecular changes as they are redirected from normal to malignant developmental trajectories. Accordingly, we performed RNA-seq and modified histone ChIP-seq on nascently transduced CB cells harvested from the first 2-3 weeks in culture. We identified coordinate upregulation of multiple anterior HOXB genes (HOXB2-B5) with contiguous H3K27 demethylation/acetylation as a striking feature in these early pre-leukemia cells. Interestingly, we also found coordinate upregulation of these same HOXB genes in a cohort of 264 patient T-ALLs (COG TARGET study) and that they defined a subset of patients with significantly poorer event-free survival (Log-rank p-value = 0.0132). Patients in the "HOXB high" subgroup are distinct from those with ETP-ALL, but are enriched within TAL1, NKX2-1, and "unknown" transcription factor genetic subgroups. We further show by shRNA-mediated knockdown that HOXB gene expression confers growth advantage in nascently transduced CB cells, established synthetic CB leukemias, and a subset of established human T-ALL cell lines. Of note, while there is prior literature on the role of HOXA genes in AML and T-ALL, and of HOXB genes in normal HSC expansion, this is the first report to our knowledge of a role for HOXB genes in human T-ALL despite over 2 decades of studies relying mostly on mouse leukemia and cell line models.

The synthetic approach we have taken here allows investigation of both early and late events in human leukemogenesis and delivers an efficient and reproducible experimental platform that can support functional testing of individual genetic variants necessary for precision medicine efforts and targeted drug screening/validation. Further, since all tumors including PDXs continue to evolve during serial propagation in vivo, synthetic tumors represent perhaps the only means by which we can explore early events in cellular transformation and segregate their biology from confounding effects of multiple and varied secondary events that accumulate in highly "evolved" samples.