Spatial Transcriptomics Identifies Aberrant Communication between Dendritic Cells and Leukemia Stem Cells As a Therapeutically Targetable Axis in Juvenile Myelomonocytic Leukemia

Juvenile Myelomonocytic Leukemia (JMML) is a RAS-driven aggressive childhood leukemia characterised by aberrant proliferation of myelomonocytic cells and very poor prognosis. Unlike other leukemias, children with JMML have no disease modifying treatments and notably Bone Marrow (BM) transplantation, the only curative treatment, is associated with a 40% risk of relapse. Our previous work characterized JMML leukemia stem cells (LSC), Louka et al, J Exp Med (2021); however the mechanisms through which the JMML LSC exert clonal dominance and treatment resistance are incompletely understood. It is increasingly recognized that aberrant cell-cell communication networks play a key role in cancer development and treatment resistance; an unexplored area in JMML. We therefore set out to investigate the JMML hematopoietic compartments and malignant BM niche using a single cell(sc) multi-omic approach including scRNA seq, cell surface protein expression, CyTOF and spatial transcriptomics from primary patient BM, generating a comprehensive spatial and molecular cellular atlas from diagnosis to disease relapse.

ScRNAseq analysis (10 JMML, 4 controls, n=38506 single cells) identified expansion of HSC and progenitor cells (HSPC) and a JMML-specific differentiation trajectory leading to expansion of the myelomonocytic series; reduction of the conventional erythroid series and erythroid progenitors with aberrant expression of fetal hemoglobin, in line with the disease features. We also observed marked expansion of dendritic cells (DCs), including conventional (cDC) and plasmacytoid dendritic cells (pDC) and validated this with multiparameter flow cytometry (19 JMML cDC1 p=0.0043, cDC2 p=0.0002, pDC p= 0.0002, 6 controls) and CyTOF analysis (21 JMML, 6 controls). Differential gene expression analysis identified JMML DCs among the top 3 cellular compartments with highly diverse gene expression signatures compared to normal controls, predominantly characterised by upregulation of inflammatory pathways and impaired antigen presentation. Cell-to-cell interaction analysis demonstrated a high number of pathological interactions between JMML LSC and mature populations, many involving JMML LSC and DCs, suggesting that DCs may play a crucial role in remodelling the JMML LSC BM niche and driving disease development.

Spatial transcriptomics (ST) (n= 791564 cells) using a bespoke panel of 600 genes provided a detailed, high resolution atlas of the JMML BM architecture at diagnosis and at relapse, by capturing all hematopoietic and stroma BM elements. This confirmed a marked increase of JMML DCs compared to normal controls at diagnosis and a further expansion at disease relapse, including a marked expansion of cDCs ( cDC1 and cDC2), however at relapse numbers of pDC remained unchanged. We employed a bespoke ST analytical pipeline to accurately demonstrate increased proximity of HSPC and cDC in JMML compared to other cellular compartments, indicating a high probability of interaction between HSPC and cDC.

In order to functionally investigate these putative pathological cellular interactions, we established a JMML LSC and DC co-culture proliferation assay. This demonstrated increased proliferation of JMML LSC when cultured in the presence of disease-associated DCs, while a similar expansion was not observed when HSPC from normal controls were co-cultured with DCs from the same donor (JMML n=9 p<0.0001 , normal controls n=6). This was particularly observed when JMML LSC were co-cultured with cDC and not with pDC, in line with our ST observations at relapse. Cell-to-cell interaction network analysis identified an active AXL/MERTK (JMML-LSC) - GAS6 (DCs) axis in JMML and increased GAS6 protein was detected in supernatant of JMML LSC-DC co-cultures. We further demonstrated increased numbers of JMML LSC when cultured in the presence of exogenous GAS6, which was reversed in the presence of specific AXL/MERTK inhibitors or CRISPR-mediated MERTK knockout in primary patient CD34+cells.

In conclusion, our work represents the first report suggesting that malignant DCs can cell-extrinsically support the maintenance and proliferation of LSC. Using spatial transcriptomics we comprehensively characterized the cellular landscape in JMML BM at diagnosis and relapse and uncovered novel pathobiological cell-to-cell interactions with potential for further preclinical development.