Intratumoral Bacteroides fragilis sensitizes the combination therapy of Capecitabine and Nivolumab
Description
To elucidate the dynamic effects of Capecitabine combined with Nivolumab treatment, we applied single-cell RNA sequencing (scRNA-Seq) to longitudinally track CRC patients receiving this combination therapy throughout the treatment course. After combination therapy, the SphK2-high activity group manifested significantly elevated T cell infiltration when compared to the SphK2-low group, suggesting enhanced T cell-mediated anti-tumor immunity in the former group. Metagenome profiles of CRC patients also indicated similar α-diversity, β-diversity and dominant colonies between two groups .
Files
Steps to reproduce
The sequencing data were mapped to the human reference sequence (GRCh38). The raw gene expression matrix from each sample was aggregated and converted into a Seurat object via Seurat (v.5) package in R software. Non-immune cells with > 6000 or < 200 genes or > 40% mitochondrial genes were discarded. Immune cells with > 4000 or < 200 genes or > 25% mitochondrial genes were filtered out. To further eliminate the data of doublets, we performed the scrublet pipeline for each batch of our scRNA-seq data, which were expected to exclude doublets, and the expected_doublet_rate was set at 0.05. The gene expression matrices were normalized to the total unique molecular identifiers (UMI) counts per cell and transformed to the natural log scale. To correct the technical and biological variations and increase the accuracy of cell type designation, we applied canonical correlation analysis implemented in Seurat to all samples before cell type identification. Seurat workflow was employed to perform dimensionality reduction and unsupervised clustering. Initially, the top 2000 highly variable genes (HVGs) were selected using the FindVariableFeatures function with parameters set to method = "vst". Subsequently, the ScaleData function was applied to regress out the effects of total UMI counts and mitochondrial gene percentage from the HVG expression matrix. PCA was conducted on the scaled HVG expression matrix via the RunPCA function of Seurat, retaining the top 30 principal components for downstream analyses. A systematic two-round unsupervised clustering strategy based on the scanpy.tl.leiden function was executed to resolve cellular population architecture. During the first round, clusters were annotated through canonical lineage marker expression patterns, enabling identification of major cell types including T cells, B cells, ILCs, myeloid cells, stromal cells and epithelial cells. Clusters exhibiting high co-expression of two or more lineage markers were classified as doublets and excluded for subsequent analyses. Iterative refinement of major cell type annotation and doublet removal was performed to ensure purity across all primary cellular compartments. For each major cell type, a second round of unsupervised clustering was implemented using the aforementioned protocol to delineate fine-grained sub-types, with analyses exclusively restricted to the expression matrix of cells within the targeted cellular compartment. For each cellular sub-type, we assessed its distribution patterns across blood, adjacent normal tissue and tumor tissue by calculating the observed-to-expected (Ro/e) ratio. Pearson residuals, derived from the chi-square test, were further employed to quantify distributional differences between tumor and adjacent normal tissues. For this analysis, cells exclusively from these two tissue types were included as input for the test function.
Institutions
- Peking University First Hospital