Development of a novel multiomic sequencing approach to characterize infused donor memory-like NK cells in an immune compatible environment

Abstract

Abstract Background Cytokine-induced memory-like natural killer (mem-NK) cells have an enhanced capacity for proliferation and persistence compared to conventional NK cells and have demonstrated efficacy against myeloid leukemia. However, the absence of a robust methodology to track mem-NK cells following their infusion into a genetically identical milieu and in the absence of a donor-specific marker after hematopoietic stem cell transplant (HCT) has limited studies of their phenotype, cytotoxicity, and interaction with other cells in vivo. We sought to overcome this significant gap by deploying a novel single-cell technology to distinguish exogenously infused donor immune effector cells from their endogenous counterparts. Methods We applied a novel single-cell approach, ReDeeM (Regulatory Multiomics with Deep Mitochondrial profiling) that integrates transcriptomics and chromatin accessibility with mitochondrial DNA (mtDNA) somatic mutation-based lineage tracing. This platform was used to analyze longitudinally collected cryopreserved peripheral blood mononuclear cell (PBMC) samples from a clinical trial (NCT04024761) where donor mem-NK cells were used to treat MDS and AML relapse following HCT. Cells from 3 timepoints: recipient pre-infusion screening (SRN), donor NK-infusion product (Infusion) and recipient day 21 post infusion (day +21), in 4 donor-recipient pairs were sorted for CD3-/CD56+ and subsequently fixed and permeabilized. RNA and ATAC library preparation followed the standard Chromium Next GEM Single Cell Multiome protocol with minor modifications. Mitochondrial libraries were constructed through probe-based hybridization and subsequent library preparation based on the ReDeeM protocol. Libraries were sequenced and single-cell RNA-seq/ATAC-seq data were processed, aligned to hg38, aggregated using CellRanger (version 7.1.0), and analyzed using Seurat (version 5.0.0). Differential expression p-values were determined using the Wilcoxon rank sum test. The mtDNA sequencing data went through ReDeeM-V for preprocessing and ReDeeM-R for downstream analysis and mapped on the multiome data. A supervised machine learning algorithm was trained on specific mtDNA variants between SRN and Infusion product data. Results The mtDNA-derived clonal repertoire suggested polyclonal NK cell expansion following infusion in all recipients, originating from both the infused mem-NK cells and the reconstituting endogenous stem cell compartment after lymphodepletion. To identify transcriptomic and epigenetic pathways enriched in the context of the day +21 samples containing these polyclonally expanded NK cells, we integrated the single-cell RNA-seq and ATAC-seq data. The genes near the single-cell ATAC-seq peaks on day +21 had an enrichment of leukocyte activation as compared to the screening samples without infused mem-NK cells. The expression of genes in the Type I interferon pathway, including IFI6 and IFITM3, was associated with the day +21 vs SRN NK cells (p=1.91·10-150and p=3.69·10-111, respectively). We then overlaid the output from ReDeeM-R to characterize CD56dim NK cell clusters that likely originated from Infusion vs SRN. Infusion-origin vs SRN-derived day+21 CD56dim NK cells had a reduced expression of genes in the ubiquitin-proteasome (FBXO11, HECTD4, FBXW11, RNF125) pathways. We cross-referenced these genes with those determined using the same multiomic sequencing approach applied to healthy donor NK cells that were stimulated with IL-12, IL-15, and IL-18 to induce the memory-like phenotype in vitro. We found that downregulation of RNF-125, an E3 ubiquitin ligase that has a known role in the regulation of type I interferon associated signaling in NK cells, was associated with the memory-like phenotype both in vitro (p=7.26·10-276) and in the clinical trial samples (p=0.0027). Further analyses to characterize the epigenetic alterations associated with the memory-like NK cell phenotype are ongoing. Conclusions We successfully applied ReDeeM that integrates transcriptomics and chromatin accessibility data with mtDNA somatic mutation-based lineage tracing to distinguish exogenously infused immune effector cells within a genetically identical milieu. This has revealed insights about infused donor mem-NK cells after HCT and has the potential to better characterize the memory-like phenotype in vivo. The platform can be applied to the infusion of any cellular therapy into an immune compatible system originating from the same donor.