Tetramer-positive progenitor exhausted or terminally exhausted cells were sorted from spleens of day 30 LCMV Cl13 infected mice or day 22 B16-OVA tumors in pools of 400 cells into a microplate containing 20 μl of RLT buffer (Qiagen) with 1% v/v β-mercaptoethanol in each well. Immediately following sorting, the plates were sealed, vortexed briefly, spun at 400g for 1 min and were flash-frozen on dry ice for storage at −80 °C until library preparation. Cellular lysates were converted to cDNA in an adapted SmartSeq2 protocol as previously described42 .
Libraries were sequenced on a NextSeq500 instrument by 37 -base-pair paired-end reads. After demultiplexing, low-quality reads were trimmed with Trimmomatic (v.0.33) using the following parameters: LEADING: 15, TRAILING: 15, SLIDINGWINDOW: 4:15 and MINLEN: 16. Trimmed reads were aligned to the mm10 mouse genome using Bowtie2 (v.2.2.4). HTSeq (v.0.6.1p1) was used to map aligned reads to genes and generate a gene count matrix. Technical replicates were averaged for each biologically independent sample. Gene counts were normalized by library size and differential expression analysis was performed using DESeq2 (v.1.18.1) (Supplementary Table 2). Following differential expression analysis between phenotypic groups, a ranking metric was calculated for each gene as R = −log10(q), where q is the FDR-adjusted P value. Preranked GSEA43 (v.3.0) was performed using the C2, C5 and C7 gene sets in the MSigDB database. To compare our data with other RNA-seq datasets from the literature, the top 150 differentially expressed genes (ranked by q value) were determined between the two indicated states for each sample generated as previously described or identified from the literature. A hypergeometric overlap test was performed to determine the overlap of each gene set with all others.