High-throughput single-cell screening of a cross-lineage CAR library for early-stage CAR design and candidate discovery
Data files
Jun 05, 2026 version files 589.94 MB
-
README.md
4.67 KB
-
Sample1_barcodes.tsv.gz
219.01 KB
-
Sample1_features.tsv.gz
301.22 KB
-
Sample1_matrix.mtx.gz
289.67 MB
-
Sample2_barcodes.tsv.gz
186.19 KB
-
Sample2_features.tsv.gz
301.22 KB
-
Sample2_matrix.mtx.gz
299.26 MB
Abstract
Chimeric antigen receptor T cell therapy (CAR-T) has demonstrated promising efficacy in hematological malignancies, but translating that success to solid tumors remains challenging. Here, we construct a CAR library comprising approximately 1,000 variants targeting prostate-specific membrane antigen by recombining transmembrane (TM), co-stimulatory, and activation domains from Natural Killer (NK) and T cell receptors. Single-cell screening identifies ICOSTM-containing variants with improved T cell activation; NK-derived activation domains, such as DAP10ζ, DAP12ζ, and FcRγζ, further augment the effector capacity. Gene regulatory network analysis reveals that CAR variants with elevated expression of T cell-activation-related transcription factors correlates with enhanced cell function. Overall, our study advances early-stage CAR design by expanding the repertoire of structural components from diverse immune cells, providing a scalable platform for identifying candidates with functional profiles comparable to clinical benchmarks.
Dataset DOI: 10.5061/dryad.0rxwdbsdc
Description of the data and file structure
CAR+ T cells and LnCap cells were co-cultured at an effector-to-target ratio of 1:2 in X-VIVO medium supplemented with 5% fetal bovine serum. After 12h of co-culture, EGFP+ cells were isolated using flow cytometry and subsequently subjected to single-cell sequencing on the HD Chip platform (Singleron Biotechnologies), following the manufacturer’s protocol. The inserted sequence of the CAR construct was amplified from cDNA, and the resulting amplicons were sequenced immediately using a nanopore sequencing platform.
Files and variables
File: Sample1_barcodes.tsv.gz
Description: The unique cellular barcode sequences for each cell from Sample1 on the HD Chip platform (Singleron Biotechnologies). It provides a one‑to‑one mapping between each column of the expression matrix and an individual cell, as these barcodes are added to mRNAs during library preparation and later used to assign reads to specific cells.
Format(s): .tsv.gz
Variable(s):
- Barcode (character string), which uniquely identifies a cell and corresponds to the column index in the accompanying matrix file.
File: Sample1_features.tsv.gz
Description: Records the detected gene (or feature) information from Sample1 on the HD Chip platform (Singleron Biotechnologies). Each row corresponds to a gene or feature, and its position matches the row index in the accompanying expression matrix.
Format(s): .tsv.gz
Variable(s):
- Gene_id (character string): unique identifier for the gene/feature (e.g., Ensembl ID).
- Gene_name (character string): common gene symbol (e.g., TP53).
- Feature_type (character string): type of feature.
File: Sample1_matrix.mtx.gz
Description: Stores the expression count matrix in a sparse MatrixMarket format from Sample1. Each non‑zero entry represents the number of UMI (or read) counts for a specific gene (row) in a specific cell (column). This file is used together with the barcodes file (for column mapping) and the features file (for row mapping).
Format(s): .mtx.gz
Variable(s):
- Row_index: index of the gene/feature, corresponding to the row number in the features file.
- Col_index: index of the cell, corresponding to the column number in the barcodes file.
- Count: the expression count (UMI or read count) for that gene in that cell.
File: Sample2_barcodes.tsv.gz
Description: The unique cellular barcode sequences for each cell from Sample2 on the HD Chip platform (Singleron Biotechnologies). It provides a one‑to‑one mapping between each column of the expression matrix and an individual cell, as these barcodes are added to mRNAs during library preparation and later used to assign reads to specific cells.
Format(s): .tsv.gz
Variable(s):
- Barcode (character string), which uniquely identifies a cell and corresponds to the column index in the accompanying matrix file.
File: Sample2_features.tsv.gz
Description: Records the detected gene (or feature) information from Sample2 on the HD Chip platform (Singleron Biotechnologies). Each row corresponds to a gene or feature, and its position matches the row index in the accompanying expression matrix.
Format(s): .tsv.gz
Variable(s):
- Gene_id (character string): unique identifier for the gene/feature (e.g., Ensembl ID).
- Gene_name (character string): common gene symbol (e.g., TP53).
- Feature_type (character string): type of feature.
File: Sample2_matrix.mtx.gz
Description: Stores the expression count matrix in a sparse MatrixMarket format from Sample2. Each non‑zero entry represents the number of UMI (or read) counts for a specific gene (row) in a specific cell (column). This file is used together with the barcodes file (for column mapping) and the features file (for row mapping).
Format(s): .mtx.gz
Variable(s):
- Row_index: index of the gene/feature, corresponding to the row number in the features file.
- Col_index: index of the cell, corresponding to the column number in the barcodes file.
- Count: the expression count (UMI or read count) for that gene in that cell.
Code/software
All code/software used for data analysis in this study is publicly available and accessible. The specific names and versions are listed below:
minimap2 (v2.27); CeleScope (v2.0.6); R (v4.2.2) using the Seurat package (v4.3.0); UCell package (v2.2.0); pySCENIC pipeline (version 0.12.1); ClusterProfiler R package (v4.6.2).
Data acquisition
CAR+ T cells and LnCap cells were co‑cultured at an effector‑to‑target ratio of 1:2 for 12 h. EGFP+ cells were isolated by flow cytometry and then subjected to single‑cell sequencing on the Singleron HD Chip platform. Meanwhile, the inserted sequence of the CAR construct was amplified from cDNA and sequenced immediately using the Nanopore platform.
Data processing
Single‑cell RNA‑seq data were aligned to the GRCh38 reference genome using CeleScope (v2.0.6) to generate a gene expression matrix. Quality control criteria: number of expressed genes (nFeature_RNA) between 300 and 10,000, UMI count (nCount_RNA) < 15,000, and mitochondrial gene ratio < 15%.
Nanopore data were aligned to a pre‑constructed CAR reference library using minimap2 (v2.27). Cell barcodes were extracted with custom scripts, and each barcode was assigned a CAR variant. Only cells containing a single CAR variant with an average alignment score > 0.95 were retained.
A total of 34,441 single‑cell data points were obtained (34,243 cells with CAR sequences, 198 cells without CAR sequences).
Downstream analysis was performed in R using Seurat (v4.3.0): normalisation, identification of top 2,000 highly variable genes, regression of cell cycle effects, batch integration with Harmony, PCA reduction, UMAP visualisation, clustering (resolution = 0.6), and differential expression analysis (FindAllMarkers).