Micro-dissection and RNA-seq of epithelial cells

Tissue samples from human or mice were cut into 5–10 consecutive sections (8 μm) and the epithelial layer contained 30–50 cells on each section was micro-dissected with a Leica LMD7000 laser-capture microdissection system. RNA was isolated from the mini-bulk samples and cDNA was prepared for sequencing based on the Geo-seq protocol^{59 (link)}. Sequencing libraries were built using the TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme), and evaluated by Bioanalyzer (DNA HS kit, Agilent). RNA-seq data were mapped to GRCh38 human genome and GRCm38 murine genome by HISAT2 (version 2.1.0)^{60 (link)} with default parameter for human and murine samples, respectively. The gene expression matrix of raw reads counts after annotation by HTSeq (version 0.6.1p1) was processed using the DESeq2 (version 1.22.2)^{61 (link)} and visualized by showing the first 3 dimensions calculated by plotPCA function. We used TRIZOL to extract bulk RNA from patient biopsy (n = 45), and constructed sequencing library using NEBNext Ultra II RNA Library Prep Kit for Illumina. Sequencing data were processed by HISAT2 and HTseq as described above. The normalized expression from bulk samples of human precancerous lesions was used to estimated neutrophil fraction with CIBERSORT (version 1.06)^{62 (link)}.