Identification of Novel Genes Underlying Human AML Expansion at Relapse from Chemotherapy

Introduction: Despite achievement of complete remission (CR) following chemotherapy, Acute Myelogenous Leukemia (AML) relapses in the majority of adult patients. Molecular and cellular contributors to chemotherapy resistance and to AML expansion and relapse have been identified in genomic, epigenetic and proteomic aberrations, while cellular relapse reservoirs have been identified in leukemia stem cells. Here, we set out to identify and functionally validate genes with consistent expression changes between AML at Diagnosis and Relapse aiming at a better understanding of the gene drivers and biology of AML relapse.

Methods: We employed the drop-seq 3’ single cell RNA sequencing (scRNA-seq) method (Macosko et al., Cell 2015) to analyze 20 paired AML specimens procured at diagnosis and at relapse from prior CR. We compared the gene expression in pooled (pseudo bulk RNA seq) single hematopoietic stem and progenitor cells (HSPCs) at diagnosis and relapse thus nominating a gene pool with highly concordant UP or DOWN regulation at relapse. We conducted lentiviral CRISPR-Cas9 based gene inactivation screens and cell growth assays using a custom library of ~7200 guides targeting 1550 genes and 1,000 controls. The screens were conducted in three primary human CD34+ non-malignant cell samples and in six primary human AML samples using 30-40 day cultures in stem cell media. A fraction of the genes that accelerated the growth of CD34+ and AML cells were individually re-assayed. Next, we selected 12 novel genes and 2 controls for in vivo studies in AML xenografts. Primary human AML were infected with lentiviral guide pools and following two days of culture injected into irradiated NSG-SGM3 mice. The AML engraftment was monitored serially via blood flow cytometry and bone marrows harvested in weeks 12-16. Guide frequencies in AML xenografts were determined via PCR-based library generation, followed by next generation sequencing and computationally using the MAGeCK software (Li et al., Genome Biology, 2014).

Results: The gene with the highest enrichment score (ES) following gene disruption in CD34+ cells was the known AML tumor suppressor DNMT3A (ES 1.64). The expression of DNMT3A was downregulated in AML at relapse, together supporting the conclusion that mutational as well as non-mutational DNMT3A downregulation facilitates stem and progenitor cell expansion in relapsed AML. Further, the AML tumor suppressors SETD2 (ES 1.38, rank 2), RUNX1 (ES 0.73, rank 9) and ETV6 (ES 0.79, rank 26) scored high in this assay. Remarkably, there were 150 genes with a mean enrichment score >0.5 that upon disruption resulted in cell enrichment in all three CD34+ cell samples concordantly. Serially harvested cell samples on days 27 and 40 had high concordance with a Pearson coefficient of R=0.76, p< 2 x 10^-15. We measured highly similar effects in pooled day 40 colonies harvested from lentivirally infected cells grown on large MethoCult^TM plates.

In single gene validation studies, upon genetic disruption of FOXN3, NIN1, TMX1, NCOR2, ODF3B, HMGXB4, GAPT, EHBP1L1, MGAT4B, CSNK1G2, EIF4G3, PCNT and ARHGEF2, we measured significantly increased cell numbers over time. The expression of all thirteen genes were downregulated in CMP-like leukemia cells at Re, and 10/13 of these genes were downregulated in HSC/MPP-like cells. Altogether, so far we experimentally confirmed 57% (17/30) of genes that when downregulated accelerated the growth and survival of human CD34+ stem and progenitor cells. We combined the assays results derived from primary AML day 30 ex vivo growth (N=4 AML) and in vivo AML xenograft assays (N=5 AML) and identified recurrent enrichment for guides targeting the genes: HMGXB4, GAPT, TMX1, ARHGEF2, NIN, FOXN3, and EIF4G3 and additional in vivo AML xenograft experiments are ongoing.

Conclusions: The comparative analysis of scRNA-seq data of 20 paired AML specimens procured at diagnosis and relapse, identified frequent and previously unrecognized large-scale changes in gene expression driving leukemia cell growth and relapse. Our experimental findings support a combinatorial role for reduced expression of these genes in leukemic progenitor cell fitness and expansion at relapse and revise current parsimonious models of human AML relapse.