Functional plasma cell signatures mark AML chemotherapy responders: Evidence from immune niche profiling

Abstract Chemoresistance, primarily driven by leukemic stem cells (LSC), remains a major challenge in the treatment of acute myeloid leukemia (AML). LSC exist within a highly specialized immune and non-hematopoietic bone marrow (BM) microenvironment that becomes altered throughout the progression of the disease. The mechanistic details of this crosstalk between LSC and their immune and non-immune bystander microenvironment cells remain poorly defined. To better understand the immune BM niche in AML, we utilized RNA-sequencing and spatial transcriptomics on archived core bone marrow biopsies (cBMB) and bone marrow aspirates (BMA) pre- and post-7+3 induction chemotherapy. To examine the immune microenvironment in the context of AML treatment, we performed bulk RNA sequencing on archived cBMBs and BMAs, collected at diagnosis and following 7+3 induction chemotherapy. Using CIBERSORT and a single-cell RNA-seq reference dataset (GSE253355), we deconvoluted immune populations in 22 AML cBMB, and 6 diagnostic BMA samples, the latter enriched for hematopoietic and stromal compartments. Surprisingly, there were few differences in the relative abundance of immune cell types between non-responders (NR) (no complete response after 7+3 induction, defined as ≥5% AML blasts) and responders (R) (complete response after 7+3 induction, defined as <5% AML blasts). However, there were notable alterations in B cell proportions between R and NR. NR exhibited significant enrichment of early B cell progenitors including common lymphoid progenitors (CLP) (diagnosis padj = 0.082, post-treatment padj = 0.082) and Pro B cells (diagnosis padj = 0.096, post-treatment padj = 0.027) in both diagnostic and post-treatment cBMB. However, there was a decrease in the proportion of plasma cells in NR post-treatment (padj = 0.16). The aspirate samples showed similar trends at diagnosis. These included an enrichment of CLP (padj = 0.232), Pro B cells (padj = 0.092), and Pre B cells (padj = 0.09), in NR and a decrease in Plasma cells in NR (padj. = 0.276). Notably, at both diagnosis and post-treatment, R displayed significant upregulation of IgG and IgA class-switched immunoglobulin genes, indicative of functional plasma cell activity. Together with the CIBERSORT analysis, this differential gene expression indicated an enrichment in B cell progenitors in NR and an enrichment in frequency and likely functionality of plasma cells in R. Next, to assess the spatial architecture of the AML niche, we used 10x Visium CytAssist platform to perform spatial transcriptomics (ST) on cBMB. The assay was performed on decalcified FFPE cBMB from 3 R and 3 NR samples procured at diagnosis and 14/15 days after induction chemotherapy with DV200 scores >30%. Through Seurat clustering, we identified three similar spatial domains across all post-treatment samples. We defined these as myeloid, stromal and lymphoid areas. Differential expression analysis of the lymphoid areas revealed significant differences between R and NR with plasma cell genes upregulated in R. IGKC and IGHG1 were the top upregulated genes in R compared to NR. Additionally, GSEA analysis of lymphoid areas revealed the top gene sets enriched in R to be ‘B cell mediated immunity’ and ‘Hay bone marrow plasma cells’, further supporting our observations of enriched plasma cells in R from the bulk RNA seq datasets. Together, these data suggest a divergence in B cell maturation between R and NR. NR are characterized by enrichment in immature B cell progenitors while R demonstrate an abundance of mature, class-switched plasma cells post-treatment, suggesting a more robust B-cell-driven immune response in R. These findings for the first time imply a role of B cell mediated immunity in response to AML treatment. Further work is warranted to delineate the impact of B cell immunity on resistance to chemotherapy, and to explore the potential as a therapeutic target in AML.