JMML Fetal Identity Results Either from Retention of a Physiologic Signature or Aberrant Activation of Master Oncofetal Regulators

Objectives: Juvenile myelomonocytic leukemia (JMML) is a rare but aggressive myeloproliferative/myelodysplastic neoplasm affecting infants and young children. The narrow age-window of onset suggests that a prenatal environment is needed for JMML oncogenesis. In search of a transcriptional reminiscence of embryo-fetal characteristics that would confirm this hypothesis, we investigated how the gene expression profile of JMML hematopoietic progenitors compared to their healthy counterpart isolated at different stages of ontogeny.

Methods: Hematopoietic stem cell and progenitor cell (HSPC) fractions of JMML (n=16), bone marrow (BM) of healthy children (n=7), fetal liver (FL; n=3) and fetal BM (FBM; n=2) were phenotyped and sorted using signatures validated in the fetal and adult BM (Notta et al, Science 2011). RNAseq was performed using the TruSeq® Stranded Total RNASample preparation kit. Unsupervised hierarchical clustering analysis was done with the Bioconductor edgeR package. Differentially expressed genes were identified with the Bioconductor limma package.

Results: To eliminate the impact of variations in the HSPC distribution, the JMML transcriptome was assessed on FACS-sorted common myeloid progenitor (CMP), granulocyte-monocyte progenitor (GMP) and megakaryocyte-erythroid progenitor (MEP) cell fractions from 16 JMML and compared to healthy counterparts at different stages of ontogeny (FL, FBM, age-matched children BM). Unsupervised hierarchical clustering separated the samples into 4 groups (C1-4), primarily according to ontogeny, with 14/15 embryo-fetal fractions in C1 and all healthy post-natal progenitors in C2 (CMP, MEP) or C3 (GMP). Most JMML fractions clustered either with the prenatal fractions (C1; 17/47 fractions from 8/16 patients) or in a separate group containing no healthy sample (C4; 23/47 samples from 10/16 patients). Two groups were defined accordingly: one with JMML resembling embryo-fetal samples ('Fetal-JMML'; n=6/16), and a JMML-specific group ('Onco-JMML'; 8/16). Patients with Onco-JMML tended to be older, with a more severe presentation and elevated fetal hemoglobin levels. All PTPN11-mutated JMML were in this group whereas 5/6 Fetal-JMML had NRAS or KRAS mutations.

Analysis of differential gene expression between Fetal and Onco-JMML highlighted 344 up-regulated genes versus 19 up-regulated genes in Onco-JMML. Surprisingly, LIN28B and WT1, both known to activate fetal pathways, were the most up-regulated genes in Onco-JMML. These key transcription factors were deregulated as early as the hematopoietic stem cell (HSC) compartment. Gene Set Enrichment Analysis (GSEA) confirmed enrichment in LIN28B and WT1-related signatures and showed enrichment in an AML signature in Onco-JMML. On the other hand, Fetal-JMML showed striking overexpression of monocytic /dendritic cell markers and inflammasome and innate immunity components. GSEA confirmed the strong monocyte identity of Fetal-JMML progenitors compared to onco-JMML or healthy postnatal progenitors. Part of the monocytic markers 'aberrantly' expressed in JMML progenitors was expressed in healthy fetal progenitors. Analysis of the HSC and multipotent progenitor (MPP) fractions showed that up regulation of monocytic markers was limited to the JMML progeny compartments. As we were able to confirm the transcriptional and functional identity of the sorted progenitors, these data suggest an early monocytic priming in these JMML progenitors, reminiscent of the monocyte-biased myelopoiesis characterizing physiologic fetal hematopoiesis.

Conclusion: Our findings give a striking example of how ontogeny-related features are involved in childhood oncogenesis. They highlight a strong but complex link beween JMML and development, with a fetal identity resulting either from retention of a physiologic fetal monocytic signature or from aberrant (re)activation of master oncofetal regulators. Intriguingly, although LIN28B is thought to reprogram hematopoietic progenitors into a fetal-like state, its activation does not lead to an overall fetal profile in JMML, suggesting a regulatory mechanism distinct from that of physiological development. These two ontogeny-based signatures are likely to uncover the biology underlying previous classifiers based on AML-like profile or DNA methylation and suggest that a subset of JMML patient may benefit from immunomodulating therapies.