Single-Cell Transcriptome Profiling of Tumor Samples

Samples prepared as described above were submitted to the Cancer Genomics Shared Resource in the Wake Forest Baptist Medical Center Comprehensive Cancer Center for single-cell sequencing and analysis. Samples with ≥60% viable cells were processed for cDNA library construction using the 10× Genomics Chromium platform (10× Genomics, Pleasanton, CA, USA) with v3 chemistry. Indexed libraries were paired-end sequenced on an Illumina NovaSeq 6000 (Illumina, San Diego, CA, USA) targeting 2500 cells per sample at a median read depth of 100,000 reads per cell. Raw bcl files were converted to fastq for read demultiplexing, alignment, and counting using the CellRanger toolkit (10× Genomics, Pleasanton, CA, USA). Data QC parameters were applied as previously described [34 (link)] to select high-quality cellular transcriptomes. Data dimensionality reduction (t-SNE) and clustering (K-means) algorithms were used to assess cell-to-cell relationships. Immune cell identities were assigned as previously described [34 (link)]. Cross-sample cell populations were compared for differences in gene expression using negative binomial models with FDR correction.