AML maturation stages associate with disease-permissive and immunosuppressive effects in newly diagnosed AML patients: An analysis at the single-cell level

Abstract

Abstract Introduction: The clinical outcome of acute myeloid leukemia (AML) is determined by therapy resistance and relapse. AML-induced immunosuppression has emerged as one of many cofactors and seems to be caused by distinct mechanisms, e.g. immune checkpoint interactions, altered antigen presentation and dysregulation of humoral factors. With chemotherapy, these alterations are expected to be reversible, but the exact mechanisms remain unclear. Adressing this issue, we performed single-cell RNAseq and spatial assessment to identify therapy-induced perturbations of the bone marrow (BM) microenvironment in newly diagnosed AML patients prior to and after induction therapy. Methods 43,000 single-cell RNA sequencing profiles were generated from 12 paired samples in 6 AML patients using 10x Genomics 5'-scRNA/CITE-seq (140 Total-seq C antibodies) at initial diagnosis (T1) and after induction chemotherapy with either “7+3” or CPX-351 (T2). Analyses were performed using R v4.1.1 and Seurat v4.3.0. Signaling pathway activity was calculated using PROGENy. Stemness was estimated using CytoTrace. Potential cell-cell interactions were scored using CellChat. Spatial correlation was performed by histopathology and immunohistochemistry in matched trephine biopsies using the following stains: H&E, Giemsa, PAS, Gomori silver, Prussian Blue, Glycophorin C, MPO, CD61, CD34, CD117, CD14, CD68, CD20, CD3, CD138, VISTA. All slides were microscopically assessed by a trained hematopathologist. Results Pre-treatmentBM (T1) was characterized by reduction of normal hematopoiesis (p=0.009), in particular classical dendritic cells type 1 (cDC1) (p=0.066) and monocytes (p=0.034). High levels of AML cells were associated with significantly higher levels of mature naïve B cells (p=0.002), higher levels of class-switched B cells (p=0.030) and enrichment of inflammatory monocyte-like cells with MDSC phenotype. Regarding AML cells, 8 different transcriptional subtypes were identified across patients: 4 clusters with highly immature properties (undifferentiated and LMPP-like AML cells) and 4 clusters with signs of maturation (aberrant erythroid-like, EMP-like, promyelocyte-like and monocyte-like AML cells). The immature AML clusters showed high similarities in their transcriptional and signaling patterns, high stemness, quiescence, strong expression of homing factors (CD44, ITGA4), DNA instability and signs of chemoresistance such as upregulation of beta-Catenin, Trail and TGFb pathways. Furthermore, an upregulation of NFkB, TNFa and JAK/STAT signaling was observed in these AML cell clusters, indicating the formation of a highly inflammatory and dysfunctional immune microenvironment. This environment was characterized by a decrease of cDC1, presence of IFIT-high monocyte-like MDSCs, clonal expansion of senescent T cells as well as functional disturbation of antigen-presenting cells, T and NK cells with high expression of TGFb. Interestingly, each AML cell cluster showed a distinct expression profile of immune checkpoint markers, with most prominent expression of VISTA in monocyte-like AML cells. Spatially, VISTA-positive monocyte-like AML cells surrounded other AML cell populations in the BM niche. In addition, monocyte-like AML cells showed high expression of hypoxia- and metabolism-associated signatures (e.g. OXPHOS, fatty acid/cholesterol homeostasis), which might have supported the dysfunctional microenvironment. These microenvironmental changes were largely reversible after chemotherapy (T2). Conclusion Our study reveals AML-induced dysregulation of the BM immune microenvironment, which seems reversible upon AML treatment with standard chemotherapy. Our data suggest that immature AML cells sustain the leukemic stem cell pool, whereas more mature EMP- and promyelocyte-like AML cells contribute to the bulk of AML. Furthermore, monocyte-like leukemic cells may shape the microenvironment to safeguard AML, possibly by VISTA-induced modulation of the immune system. This might explain limited therapeutic efficacy with currently available checkpoint inhibitors in AML. In the long run, identification of specific immune signatures in AML may improve our understanding of AML biology and inform therapeutic algorithms with regard to immunotherapeutic strategies in AML.