Deciphering the AML Niche: RNA Sequencing and Spatial Transcriptomics Reveal Diverse Transcriptomic Landscapes in Responders Vs. Non-Responders

Leukemic stem cells (LSC) have been shown to be responsible for relapse of acute myeloid leukemia (AML) due to their intrinsic chemoresistance. LSC exist within a highly specialized bone marrow microenvironment niche that becomes altered throughout the progression of the disease. Mechanisms of the crosstalk between LSC and the BM niche in AML development, treatment, and relapse remain obscure, and prior studies have been limited by RNA recovery from decalcified core bone marrow biopsies (cBMB). To detail the changes within the stromal and immune BM niche in AML, we utilized RNA-sequencing and spatial transcriptomics on archived cBMB and bone marrow aspirates (BMA) pre- and post-7+3 induction chemotherapy.

Through bulk RNA-seq analysis of 22 AML cBMB, we previously reported the existence of distinct subpopulations of stromal cells, correlating response to 7+3 treatment, with upregulation of osteo-mesenchymal transcripts in non-responders (NR) (no complete response after 7+3 induction, i.e. ≥5% AML blasts) (Treaba et al. 2023). Altering our RNA-seq analysis to include low count (>10 reads) transcripts allowed the detection of stromal genes at diagnosis, identifying significant upregulation of 27 mesenchymal-associated genes in responders (R) compared to NR. To further characterize transcriptomic differences between R and NR at diagnosis, we next isolated stromal cells from liquid bone marrow aspirates (BMA) of 3 R and 3 NR by fluorescence-activated cell sorting, using negative selection for hematopoietic lineage markers, and performed low-input RNA-sequencing on both stromal and hematopoietic BMA fractions. As with the cBMB, the stromal BMA exhibited upregulation of transcripts representing osteo-, mesenchymal-, and fibroblast lineages in R compared to NR (TGM2, SDC1, LRP1, PLOD1, FBLN2, VCAN, CCN2, FZD8, GPC1, MMP14, TIMP3, SDC4, COLGALT1, CALU, ITGB1, TIMP1, p value <.05), further supporting our observations of contrasting stromal composition based on response.

Next, to assess the spatial architecture of the AML stroma, we used 10x Visium CytAssist platform to perform spatial transcriptomics (ST) on cBMB. The assay was performed on decalcified FFPE cBMB from responder and non-responder samples procured at diagnosis and 14/15 days after induction chemotherapy with DV200 scores >30%. ST revealed an abundance of stromal transcripts post-treatment in both R and NR in the bone-proximal region. Intriguingly, while multi-stromal lineages were detected in R (APOE, COL1A1, COL1A2, COL3A1, CXCL12, FN1, FTL, GLUL, SLC40A1, SPARC, SPP1, TNC, VCAM1, VCAN, VIM), NR predominantly displayed upregulation of osteoblast-lineage transcripts post-treatment (COL1A1, COL1A2, SPP1, TNC). We predict that the lack of fully differentiated stroma in NR aids in maintenance of LSC.

Finally, we investigated whether a more diverse stroma in R was accompanied by alterations to hematopoietic and immune cell types. Analyses of hematopoietic transcripts in both cBMB and the hematopoietic fraction of BMA showed, in R, upregulation of an array of progenitor cell types including early erythroid and megakaryocyte progenitors at diagnosis, and HSCs and multilineage progenitors post-treatment, compared with NR. These findings suggest that the robust and diverse stroma in R supports a diversity of hematopoietic progenitor cells at diagnosis and post-treatment, with trilineage hematopoiesis, which was significantly less abundant in NR both pre- and post-treatment. Interestingly, immune profiling performed using RNA-seq deconvolution strategies showed a significant enrichment of B cells (SPIB, FCRLA, P2RX5 and GNG7) at diagnosis and myeloid dendritic cells post-treatment in the cBMB of NR, as compared to R. Additional studies into these immune cell differences may provide insight into their involvement in resistance to treatment.

As the BM niche may provide protection for AML LSC, there is an urgent need to decipher the underlying crosstalk to inform therapeutic strategies. We successful identified an array of stromal subtypes using RNA recovery from both cBMB and BMA of R AML patients undergoing 7+3 induction chemotherapy, contrasting less robust and more monoclonal stromal composition in NR. This demonstrates an opportunity to use novel RNA-seq and ST tools to characterize the molecular drivers responsible for microenvironment-driven AML chemoresistance and to inform future AML treatments.