Bone marrow Single cell sequencing data of CAR-T treated IL-2Ra -/- mice

Despite breakthrough research and success in CAR-T therapy in heme malignancies, adverse events caused during CAR-T infusion remain a challenge. Among the various toxicities, the co-occurrence of cytokine release syndrome (CRS) along with neutropenia is the leading cause of severe life-threatening infections, increased hospitalizations, and the leading cause of non-relapse mortality. To this end, we developed an IL-2R alpha knockout model (where exons 2 and 3 that bind to IL-2 are replaced by a neomycin resistance gene) that showed the recapitulation of CRS and neutropenia seen in the clinic. This novel pre-clinical model helped us identify IFNg (one of the CRS-associated cytokines) to drive this co-occurrence. Through adoptive transfer studies, we show that excess IFNg released during CAR-T expansion triggers the inflammatory cascade that fuels the etiology of both CRS and neutropenia by triggering Th1/Th17 imbalance. IFNg is a Th1 cytokine that inhibits Th17 differentiation, which is crucial for neutrophil regulation. To corroborate the impact of IFNg on CRS-neutropenia, we processed bone marrow samples of tumor-bearing IL-2Rα -/- mice treated with CAR-T (Group 1) and IFNγ -/- CAR-T (Group 2) in the presence or absence of Th17 (Group 4) and Th1 cells (Group 5). The samples were processed for sequencing at week 4 post CAR-T infusion to study the mechanistic underpinnings of the impact of IFNg blockade on neutrophils and macrophages (M1 macrophages are known to drive CRS) that are involved in CRS and neutropenia co-occurrence during CAR-T therapy.

Methods
Single cells were isolated and then processed using the 10X Genomics Single Cell 3' v3 kit according to the manufacturer's instructions. Libraries were sequenced on the Illumina NovaSeq 6000 instrument (RRID:SCR_016387). Raw sequencing data were processed using the Cell Ranger (CR) (v6.0.0) pipeline (RRID:SCR_017344) to generate fastq files. Fastq files were aligned and quantified, generating feature-barcode count matrices. Gene-barcode matrices containing Unique Molecular Identifier (UMI) counts are filtered using CR's cell detection algorithm. Downstream analyses were performed mainly using Seurat (v5.0.0) single-cell analysis R package (RRID:SCR_016341). Eight single-cell RNA seq samples were individually read into a Seurat object (RRID:SCR_016341) to examine feature number, mitochondrial percentage, and read count distributions within each sample. Cells with fewer than 500 features or greater than 7500 features or >15% mitochondrial content were filtered out. After normalizing and finding variable features from individual samples, all samples are then integrated using FindIntegrationAnchors function. The resulting data would serve for visualization purposes. Separately, samples were merged in a SingleR (v1.6.1), a Bioconductor package (RRID:SCR_006442), which was then used to annotate individual immune cell types using the ImmGen reference database (RRID:SCR_021792) from the celldex (v1.12.0) R package. Principal component analysis was used to detect and visualize highly variable genes. Using the RNA assay, data normalization and scaling were performed using Seurat's (RRID:SCR_016341) SCTransform function regressing against mitochondrial percentage. The data were then dimension-reduced via UMAP (RRID:SCR_018217) and clustered using the Louvain algorithm for downstream visualization. Normalization, scaling, dimension-reduction, clustering, and quantification are performed on specific cell populations (i.e., Neutrophils and Macrophages) to further detect subpopulation clusters based on their expression profiles. Differential gene expression analysis was performed using Seurat's (RRID:SCR_016341) FindMarkers function on groups of interest.

Subsequent gene-set enrichment analyses were performed on comparisons of interest using the GSEApreranked procedure implemented by the clusterProfiler package (v4.6.2) (RRID:SCR_016884) using the msigDB database found in the msigdbr (v7.5.1) package (RRID:SCR_022870). Mouse versions of Hallmark, C2, and C5-GO gene sets were used to perform pathway analyses. The avg_log2FC values (obtained from the FindMarkers results) are utilized as ranked values to perform pathway analyses using prerankedGSEA. Specific cell-level signature scores are obtained using Seurat’s (RRID:SCR_016341) AddModuleScore function. The custom code generated for all single-cell sequencing analysis is available upon request.