Deconvolving Immunologic Networks of the AML Marrow Microenvironment with Spatial Multi-Omic Profiling

The spatial organization of “immune hubs” in tumor microenvironments has gained recognition across histologies and tissues. Bone marrow offers a valuable platform for investigating immune hubs in the context of acute myeloid leukemia (AML) following hematopoietic stem cell transplant (HSCT). This setting may provide fresh insights into mechanisms of relapse and maintenance of remission after donor lymphocyte infusion (DLI). While it is generally understood that DLI exerts its effects through a graft-versus-leukemia (GVL) response, the specific cellular players and spatial organization driving GVL remain unidentified. Addressing this question can inform the rational design of cancer cellular therapies.

A major challenge for spatial transcriptomic assays interrogating bone marrow samples is RNA degradation due to decalcification in addition to that caused by formalin fixation and paraffin embedding (FFPE). Furthermore, existing spatial platforms lack single-cell resolution and their throughput is generally limited to a small number of samples. To address these constraints, Singular Genomics is developing the G4XTM, an in-situ next-generation sequencing (NGS) platform that enables rapid sample throughput (up to 20-fold greater than existing platforms), and enables combined readouts of transcriptomics, proteomics, and fluorescent H&E staining in the same FFPE section.

We used a pre-release version of the G4X to analyze bone marrow biopsies of patients with relapsed AML before and after treatment with DLI. We profiled 24 specimens (6 responder [R] pre-DLI, 7 R post-DLI, 5 non-responder [NR] pre-DLI, 6 NR post-DLI), aiming to elucidate the spatial relationships of immune cellular networks in the marrow microenvironment. Our targeted transcript panel included 153 immune- and marrow-related genes defining specific T cell subsets and other immune cell lineages, as well as 8 protein markers (CD3, CD4, CD8, CD45RA, HLA-DR, CD34, KI67, ATPase). We filtered out cells with transcript counts above the 95th percentile and retained cells with at least 10 transcripts and 3 unique genes. We excluded genes present in less than 5 cells and considered segmented nuclei sizes between 2.6 µm and 14.2 µm, resulting in 193,651 cells (median of 30 transcripts, 11 unique genes per cell). Further pre-processing consisted of batch correction via Harmony (Korsunsky 2019), dimensional reduction and clustering.

Cell populations across 14 distinct clusters were annotated using gene signatures derived from previously generated single-cell RNA sequencing data (Maurer et al ASH 2023) and refined by considering differentially expressed marker genes and protein expression profiles. We previously demonstrated expansion of an activated CD8 T cell population with high expression of ZNF683 and GZMB uniquely in R post-DLI (Maurer et al ASH 2023). Hence, we interrogated the dynamics and spatial relationships of this specific T cell subset, identified among other lymphoid cells in the in situ data based on gene and protein signatures.

In confirmation of our previous findings, we found that this T cell subset increased in proportion specifically in R post-DLI samples (3.6% to 6.4% in R pre to post; 2.9% to 1.9% in NR pre to post, p=ns) and exhibited decreased spatial self-association in NRs (8.79% to 2.17% in NR pre to post, 32.5 µm radius, p=0.008). In Rs, these CD8 T cells were more likely to be in the spatial vicinity of myelo-erythroid progenitor and AML cells, perhaps suggestive of active immune surveillance post-DLI (1.62% in NR post vs 6.32% in R post, 32.5 µm radius, p=0.008).

When examining spatial organization of all cell types, bone marrow of NRs pre-DLI was marked by an overabundance of myeloid cells relative to other cell types. The microenvironment of Rs was more heterogeneous before treatment and showed highest immunologic diversity post-DLI (Shannon Diversity Index: 0.744 in NR post vs. 0.943 in R post, p=0.021). These findings suggest that diversity of marrow composition pre-DLI may be predictive of therapeutic response, though this observation requires validation in larger cohorts.

We present the first analysis of post-HSCT relapsed AML marrow microenvironment with spatial multi-omic profiling using G4X and find evidence of post-DLI T cell expansion concordant with our scRNA-seq analysis. Further evaluation of spatial organization may lead to novel targets for optimizing adoptive cellular immunotherapy.

Raths:Singular Genomics: Current Employment. Gouin:Singular Genomics: Current Employment. Koh:Singular Genomics: Current Employment. Fabani:Singular Genomics: Current Employment. Livak:MBQ Pharma Inc.: Membership on an entity's Board of Directors or advisory committees. Ritz:Kite/Gilead: Research Funding; Oncternal: Research Funding; Novartis: Research Funding; Smart Immune: Membership on an entity's Board of Directors or advisory committees; LifeVault Bio: Membership on an entity's Board of Directors or advisory committees; Garuda Therapeutics: Membership on an entity's Board of Directors or advisory committees; Clade Therapeutics: Membership on an entity's Board of Directors or advisory committees; TriArm Bio: Membership on an entity's Board of Directors or advisory committees; Oncternal: Research Funding. Soiffer:Vor Biopharma: Consultancy; Neovii: Consultancy; Jasper: Consultancy; Amgen: Consultancy; Astellas: Consultancy; Smart Immune: Consultancy; Bristol Myers Squibb: Membership on an entity's Board of Directors or advisory committees. Glezer:Singular Genomics: Current Employment. Wu:Aethon Therapeutics: Membership on an entity's Board of Directors or advisory committees; Pharmacyclics: Research Funding; BioNtech, Inc: Current equity holder in publicly-traded company; Adventris: Membership on an entity's Board of Directors or advisory committees; Repertoire: Membership on an entity's Board of Directors or advisory committees.