Patients and tissue samples

This study was approved by the Institutional Review Board of Sun Yat-sen University Cancer Center (SYSUCC). For fibroblasts isolation, tumor tissues, nontumor tissues and metastatic lymph nodes were obtained from patients with ESCC who were not pretreated before surgery. For correlation analysis, 34 pairs of ESCC tumor tissues and metastatic lymph nodes were collected at SYSUCC. For chemoresistance analysis, 50 ESCC tumor tissues were collected from patients receiving platinum-based chemotherapy after surgery at SYSUCC. These samples are not overlapped with the above mentioned samples. None of the patients received neoadjuvant therapy. Patient characteristics are included in the supplemental data. Written informed consent was obtained from patients. The study was carried out in accordance with the Declaration of Helsinki guidelines.

Isolation of fibroblasts and primary cell culture

Fibroblasts were isolated from primary tumors, noncancerous tissues, and matched metastatic mediastinal lymph nodes as described previously [7, 8]. Tissues were prewashed with 1× penicillin-streptomycin (NCM Biotech, China), minced into small pieces of approximately 1 mm3, and digested with Collagenase IV (Sigma-Aldrich, Germany). The digested suspension was centrifuged at 180 g and washed with DMEM twice. Then, the cell pellet was resuspended and cultured in DMEM supplemented with 20% fetal bovine serum, penicillin-streptomycin, levofloxacin (Beijing Pharm. Co. Ltd., China) and fluconazole (Pfizer Inc., NY). After incubating for 30 min at 37 °C, nonadherent cells were removed, and fibroblasts were obtained because of the different adhesion times of fibroblasts and tumor cells. This step was repeated several times until the tumor cells were completely removed. The fibroblasts were subcultured for further study.

Cell lines

Esophageal squamous carcinoma cell lines, including KYSE510 (RRID: CVCL_1354), KYSE150 (RRID: CVCL_1348), KYSE410 (RRID: CVCL_1352), KYSE180 (RRID: CVCL_1349) and KYSE140 (RRID: CVCL_1347), were kindly provided by Dr. G Srivastava and Dr. GS Tsao of the University of Hong Kong [9]. 293FT cells (RRID: CVCL_6911) and NIH3T3 cells (RRID: CVCL_0594) were purchased from Invitrogen (Thermo Fisher Scientific, Waltham, MA). Mouse esophageal cancer cells (MEC2) were gift from Dr. L Fu of Senzhen University [10]. The cells were authenticated by STR profiling and confirmed to have no mycoplasma contamination.

All ESCC cells and fibroblasts were cultured in DMEM (Gibco BRL, NY) supplemented with 10% fetal bovine serum (Gibco BRL, NY) and incubated in a humidified atmosphere at 37 °C in 5% CO2.

RNA sequencing and analyses

Total RNA was extracted from cells using TRIzol (Invitrogen, Carlsbad, CA). RNA yields were quantified using a Qubit™ RNA HS Assay Kit (Invitrogen, Carlsbad, CA), and RNA quality was assessed by 4200 TapeStation (Agilent, Santa Clara, CA). RNA sequencing was conducted by Haplox (Shenzhen, China). Sequences were generated with the Illumina novaseq 6000 platform, PE150 read format. For fibroblast analysis, differentially expressed gene (DEG) analysis (fold-change>1.2 or fold-change<0.83, P < 0.05) was implemented by DESeq2. For PI16 analysis, differential analysis was performed by EdgeR (fold-change>2 or fold-change<0.5, Padj<0.05). GO enrichment analysis was performed by using DAVID (https://david-d.ncifcrf.gov/) [11, 12]. KEGG pathway analysis was performed by using the Cluster Profiler R package [13].

Conditioned medium (CM) and coculture experiments

For fibroblast conditioned medium, fibroblasts (1 × 106) were grown in 10 cm dishes and cultured in 10 ml DMEM with 10% FBS for 24 h. The medium was collected and filtered through 0.45 µm filters [14].

For PI16 enriched medium, NIH3T3 cells stably transfected with pLVX-PI16 or vector were incubated with 10 ml serum-free DMEM for 24 h. The medium was collected, filtered through 0.45 µm filters, concentrated to 200 µl using a 10 kDa Amicon® Ultra-15 Centrifugal Filter Device (Millipore, Burlington, MA), flash-frozen in liquid nitrogen, and then stored at −80 °C for further experiments [14]. When used, the concentrate was added to fresh medium to 1 ml.

KYSE150 cells (5 × 103 per well) were seeded in the bottom chamber of 24-well plates (Corning, NY), while NIH3T3 fibroblasts stably transfected with pLVX-PI16 or vector control (5 × 103 per well) were seeded into the top chamber. A coculture system was established by inserting the top chamber into a 24-well plate.

Liquid chromatography coupled with Orbitrap mass spectrometry (LC-MS/MS)

Fibroblasts at 80% confluence were pre-treated with 10 ml serum-free DMEM for 24 h. Supernatants were collected, filtered and concentrated with a 10 kDa Amicon® Ultra-15 Centrifugal Filter Device (Millipore, Burlington, MA). LC-MS/MS was conducted by Wininnovate Bio (Shenzhen, China). Protein digestion was carried out using the filter-aided sample preparation (FASP) method [15]. The peptide fractions were loaded into a nanoViper C18 (Acclaim PepMap 100, 75 μm × 2 cm) trap column. Online chromatography separation was performed on an Easy nLC 1200 system (Thermo Fisher, Waltham, MA) using a 90 min gradient on an analytical column (Acclaim PepMap RSLC, 75 μm × 25 cm C18-2 μm 100 Å). Data-dependent acquisition (DDA) mass spectrum techniques were used to acquire tandem MS data on a ThermoFisher Q Exactive mass spectrometer (Thermo Fisher, Waltham, MA) fitted with a Nano Flex ion source. The data were analyzed for protein identification and quantification using PEAKS Studio 8.5 (Bioinformatics Solutions Inc., Waterloo, Canada). The local false discovery rate at PSM was 1.0% after searching against the Homo sapiens database with a maximum of two missed cleavages. The following settings were selected: oxidation, acetylation, deamidation, pyro-glu from E, pyro-glu from Q for variable modifications and fixed carbamidomethylation of cysteine. Precursor and fragment mass tolerance were set to 10 ppm and 0.05 Da, respectively.

Clinical analysis of public datasets and online bioinformatics tools

The Cancer Genome Atlas (TCGA) Esophageal Cancer (ESCA) transcriptome data were downloaded from the UCSC genomic center. The data from a total of 68 ESCC patients (<70 years of age at time of surgery) were used to analyze overall survival, and the patients were grouped based on a cutoff (log2FPKM+1: 12.27) for PI16. The Survminer R package was used to calculate the cutoff value for PI16.

GO enrichment analysis was performed by using DAVID (https://david-d.ncifcrf.gov) software [12]. Tumor purity and the correlation between PI16 and ACTA2 in ESCA were calculated with Tumor Immune Estimation Resource (TIMER) (https://cistrome.shinyapps.io/timer/) [16]. Cell-type enrichment analysis was carried out on transcriptome data using the xCell cell type enrichment score tool (http://xcell.ucsf.edu/) [17] based on Tirosh signatures, including fibroblasts, endothelial cells, macrophages and B cells.

Single-cell RNA (scRNA) sequencing data (GSE134355 and PRJNA554845) were downloaded from GEO or Bioproject. The cell type composition of major human organs and the human cell landscape were constructed based on scRNA data. The Seurat R package was used to perform clustering analysis. 2D plots of single cells were visualized with the UMAP algorithm.

Immunohistochemistry (IHC)

Paraffin-embedded formalin-fixed sections were deparaffinized with xylene, and rehydrated with a series of ethanol concentrations, then incubated with 3% hydrogen peroxide. Antigen retrieval was performed using Sodium Citrate Antigen Retrieval Solution (Boster, China) for 8 min. After blocked with 4% BSA, the sections were incubated with primary antibodies at 4 °C overnight, then with secondary antibody for 1 h at room temperature. The staining was visualized by DAB (Dako REAL™ EnVision™ Detection System, DAKO, Glostrup, Denmark). The degree of immunostaining was evaluated by two pathologists with no prior knowledge of patient characteristics. The staining intensity was scored as four levels: negative (0), weak (1), medium (2), and strong (3). The proportion of immunopositive cells in fibroblasts was scored as: 0–25% (1), 25–50% (2), 50–75% (3), 75–100% (4). The judgment of fibroblasts excluded muscle cells, vascular smooth muscle cells and cells of muscle bundle structure and tubular structure. The IHC score was obtained by multiplying the intensity score by the proportion score. PI16 high expression was defined as IHC score>3.

Additional methodology is provided in the Supplementary information.