Despite mutation acquisition in hematopoietic stem cells, JMML-propagating cells are not always restricted to this compartment

Abstract

Juvenile myelomonocytic leukemia (JMML) is a rare aggressive myelodysplastic/myeloproliferative neoplasm of early childhood, initiated by RAS-activating mutations. Genomic analyses have recently described JMML mutational landscape; however, the nature of JMML-propagating cells (JMML-PCs) and the clonal architecture of the disease remained until now elusive. Combining genomic (exome, RNA-seq), Colony forming assay and xenograft studies, we detect the presence of JMML-PCs that faithfully reproduce JMML features including the complex/nonlinear organization of dominant/minor clones, both at diagnosis and relapse. Further integrated analysis also reveals that although the mutations are acquired in hematopoietic stem cells, JMML-PCs are not always restricted to this compartment, highlighting the heterogeneity of the disease during the initiation steps. We show that the hematopoietic stem/progenitor cell phenotype is globally maintained in JMML despite overexpression of CD90/THY-1 in a subset of patients. This study shed new lights into the ontogeny of JMML, and the identity of JMML-PCs, and provides robust models to monitor the disease and test novel therapeutic approaches.

Fig. 1

Despite heterogenous distribution, phenotypically and molecularly defined stem/progenitor cell fractions are maintained in JMML. a Distribution of phenotypically defined HSC, MPP, and LMPP within the CD34⁺CD38⁻ population (left panel), and of CMP, GMP, and MEP within the CD34⁺CD38⁺ population (right panel) in BM of patients with JMML (n = 31) compared with healthy children (n = 19). b Transcriptional validation of phenotypically defined JMML stem/progenitor cell fractions (n = 14) compared with their normal counterparts sorted from healthy children BM (n = 4). RNAseq results are expressed as mean FPKM scores (±SD) for gene transcripts that are characteristic of the normal counterpart of phenotypically defined stem (HLF, MPL, ABCB1, and HOXA9) and progenitor cell (CSF1R, CSF3R, EPO, and GATA1) fractions. See Supplementary Fig. S2 for gating strategy and RNAseq results for additional genes. Anova multiple comparison, ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05; ns not significant, BM bone marrow, FPKM fragments per kilobase million, SD standard deviation

Fig. 2

CD90 is overexpressed in a subset of JMML. a Representative flow cytometry plots of a healthy child BM, and patients #92 (NRAS-JMML) and #95 (PTPN11-JMML) with respectively normal and over/ectopic expression of CD90/Thy1 across the different hematopoietic compartments. The first column shows CD90 vs CD45RA within the CD34⁺CD38⁻ fraction, the middle column shows CD135 vs CD45RA and last column shows CD90 vs CD45RA within the CD34⁺CD38⁺ fraction. Gating of cellular HSPC fractions is indicated. b Percentage of CD90/Thy1 expressing cells measured by flow cytometry in JMML compared with healthy children BM within the total CD34⁺CD38⁺ fraction, and across the different hematopoietic progenitor compartments CMP, GMP, MEP (see also Supplementary Table S2). Anova multiple comparison, ****p < 0.0001; ***p < 0.001; ns not significant

Fig. 3

PDX models accurately capture the features and clonal diversity of the disease and allow to characterize the JMML-PC. a Percentage of human CD45 out of total BMNCs present in the mouse BM at termination, in NSG (n = 12 patients, n = 35 mice), and NSG-S (n = 11 patients, n = 27 mice). In each panel, red symbols indicate JMML that are CD90^high (#53, #88, #91, and #95). b The levels of human engraftment are displayed per mutated gene between the two mice models. Matching shapes and colors represent the same patient between the two models. c The level of human CD45 engraftment out of total nucleated cells in secondary recipients (10 NSG, 8 NSG-S) for three patients (#88, #92, and #99). Samples harvested from primary NSG mice were injected into NSG secondary recipient and/or primary NSG-S into secondary NSG-S. For patient #92, cells harvested from the NSG primary mouse were injected into either NSG or NSG-S (see also Supplementary Table S3). d Heat map representation of the level of engraftment obtained after injection of the different JMML hematopoietic fractions (HSC, MPP, LMPP, CMP, GMP) of five patients with JMML (#88, #95, #154, #152, and #66). The number of cells injected per patient and per fraction is displayed for each fraction. e Correlation of variant allele frequencies (VAF) obtained from the cells post xenotransplant compared with the native JMML cells from the patient at diagnostic. Each dot representing one mutation (red dots: NSG; blue dots: NSG-S). See also Supplementary Fig. S3

Fig. 4

Clonal dynamics with time and across hematopoietic differentiation shows early clonal dominance. a Schematic of experimental procedure followed in order to delineate the origin and clonal architecture of JMML. All CFCs obtained from these experiments on naïve JMML (blue icons) or xenotransplanted JMML (brown icons) were tested for known patient mutations by targeted sequencing. See also Fig. S4. b Clonal architecture of two JMML samples (#66 and #92), as determined by combining whole exome sequencing, deep targeted sequencing, and single-cell derived colony sequencing before and after xenotransplantation. For each patient, a fish plot (left) represents clonal evolution between diagnosis and relapse. In the absence of preleukemia sample allowing to specify the kinetics of clonal emergence, subclones were represented by default as appearing simultaneously. Mutations found in each subclone are indicated (see also Supplementary Table S6). The clonal composition in total JMML mononucleated cells (MNCs) at diagnosis or relapse is also represented in circles. The larger circle represents the founding clone. Smaller circles inside represent subclones of various size and matching colors with the fish plot. Clonal composition and engraftment capacities across hematopoietic differentiation are represented on the right panels. Mutations identified in MNC were screened in sorted fractions before and after xenotransplantation using Sanger sequencing. Mouse icons tag fractions that were injected in NSG and/or NSG-S mice. Red mouse icons indicate successful engraftment whereas gray icons indicate engraftment failure. Patient (#66) KRAS-JMML showing branched evolution with independent acquisition of additional mutations targeting ASXL1. The dominant clone at diagnosis or at relapse was also dominant in corresponding xenografts. Patient (#92) NRAS-JMML showing branched evolution with independent acquisition of additional mutations targeting either NF1 or RAC2. At relapse, exome sequencing performed on the MNC evidenced the gain of an IKZF1 mutation within the clone that became dominant at relapse. Sorted relapse HSC, MPP, and LMPP engrafted in NSG mice and targeted sequencing of individual picked CFC obtained from these mice demonstrated the presence of the dominant relapse clone in HSC and MPP whereas cells retrieved from the LMPP, CMP, and GMP engrafted mice only harbored the mutations of the clone that was dominant at diagnosis but minor at relapse