Characterizing the Bone Marrow Stromal Niche in Pediatric Acute Myeloid Leukemia

Although current survival rates of pediatric AML (pAML) are around 80%, approximately one quarter of patients relapse. With this high relapse rate and overall survival dropping to around 50% with relapse, it is critical to understand its underlying mechanisms.

The impact of the bone marrow (BM) leukemic niche on treatment response is still incompletely understood. Not only are T cells critical for AML treatment response, but evidence points to the importance of mesenchymal stromal cells (MSCs) in creating an immune suppressive and leukemic supportive niche. In adults, the presence of AML cells results in remodeling of the BM niche, with an inflammatory stromal phenotype linked to a more favorable treatment outcome. In contrast to adult AML, studies in the stromal niche of pAML patients are scarce or based on culturally expanded stromal cells which do not faithfully resemble their in vivo native state and function.

We focus on characterizing stromal cells in pAML, investigating their role in treatment response, and studying their interactions with the leukemic niche. To achieve this, we combine single cell RNA sequencing (scRNA seq), flow cytometry and in vitro functional experiments. In order to study the dynamic changes between patients with a recurrence and those in continuous complete remission (CCR), we identified a cohort of 19 KMT2A rearranged (KMT2A::MLLT1, KMT2A::MLLT3, MLLT10, KMT2A::ENL) pediatric patients, 10 in CCR and 9 patients with a recurrence. We sequenced cells from BM aspirates at four different stages: diagnosis, before the second (TP1) and last therapy course (TP2), and at first relapse if applicable. Stromal cells are challenging to study as they are less than 0.05% of BM aspirates. To tackle this challenge, we developed a cell sorting strategy to enrich stromal cells on available material and include them in our scRNA seq analysis. To minimize blast contamination, we selected cells CD33^low and selected cells which were CD45^neg (non-hematopoietic) and CD235a^neg (non-erythroid) respectively. To estimate the stromal abundance in our sorting strategy, we included CD271 and CD90 as well-known positive MSCs markers and CD34 for endothelial cells.

Without stromal enriching, we obtained 0 – 32 stromal cells (average = 7) per patient sample with most cells from TP1 and TP2 supporting our need for enrichment. Currently, we have sorted 635 stromal cells from 8 diagnosis and 4 relapsing patients, achieving a 20-fold enrichment compared to matched samples without stromal enrichment. With the use of a publicly available dataset (Bandyopadhyay et al., 2024), our MSCs are predicted to be primed for the adipocyte lineage (LPL, APOE, PPARG) with a less abundant osteo-primed stromal cluster (IBSP, SPP1, BGLAP) and a sinusoidal-like endothelial cluster (CDH5, TEK, CD36). Surprisingly, gene set enrichment analysis revealed a negative enrichment in adipogenesis and TGFB signaling when compared to healthy MSCs which contrasts with previous tissue culture findings. pAML MSCs showed a positive enrichment in focal adhesion and inflammatory response. Preliminary analysis indicates that our sorted MSCs separate by patient into 2 main clusters, one enriched for KMT2A::MLLT10 MSCs and healthy MSCs, and the second enriched for MSCs from other KMT2A fusions. KMT2A::MLLT10 pAML MSCs show upregulation of CCL2, CXCL1, CXCL3 and CXCL8 compared to other fusions which are implicated in stromal inflammation in multiple myeloma. Cell adhesion genes LAMB1, LPP, VCAN, ACTG1 are downregulated in KMT2A::MLLT10 MSCs compared to other subtypes hinting that leukemic cells may be less dependent on direct support of MSCs by cell-cell adhesion. Furthermore, while ECM genes COL6A3, COL6A2, FN1 are upregulated in all pAML MSCs in contrast to healthy, KMT2A::MLLT10 MSCs are characterized by a downregulation of COL3A1, COL1A2, COL14A1 and MMP28 compared to other fusions implying a different matrix remodeling between fusion subtypes. Currently, we are expanding our cohort and validating our findings in co-culture models.

In conclusion, we effectively enriched primary stromal cells and our data suggests that the stromal niche is remodeled in pAML and that KMT2A::MLLT10 pAML is less dependent on MSCs by focal adhesion compared to other fusions. These data provide deeper insights into the cell-cell niche dynamics in pAML and KMT2A fusion subtypes, potentially unveiling new therapeutic targets to improve patient outcomes.