Targeted inhibition of RBPJ transcription complex alleviates the exhaustion of CD8+ T cells in hepatocellular carcinoma

Ethics statement

This study protocol was reviewed and approved by the Ethics Committee of Fujian Medical University Union Hospital, approval number 2021KJCX008; and the Experimental Animal Ethics Committee of Fujian Medical University, approval number IACUC FJMU 2022-0015. The liver cancer specimens used in this study were surgically resected excess tissue, and the Ethics Committee of Fujian Medical University Union Hospital has agreed to waive the patients’ written informed consent.

Animal

C57BL/6J female mice (6 weeks old, 18 ~ 20 g) were housed in a specific pathogen free environment with a 12/12 h day/night cycle. To construct a subcutaneous xenograft tumor model, 2 × 10⁶ Hepa1-6 cells were injected subcutaneously on the left side of mice, and the tumor volume was measured every 3 days using a vernier caliper. Tumor volume was calculated with the formula (length × width²/2). For induced HCC, at 8 weeks of age (six weeks after a single injection of diethylnitrosamine (DEN)), begin administration of carbon tetrachloride (CCl₄) (0.2 ml/kg i.p.) twice a week for up to 14 additional weeks (dose volume: 15 mL/kg body weight). When the tumor volume reached 250 mm³, RIN1 50 mg/kg, Bevacizumab 100 mg/kg, InVivoPlus anti-mouse PD-1 50 mg/kg, InVivoPlus anti-mouse PD-L1 50 mg/kg, L-kynurenine 30 mg/kg and clodronate liposomes 1 mL/kg were intraperitoneally injected every 3 days for a total of seven times.

Data sources and bbioinformatics analysis

We used Genomic Data Commons (GDC) download tool (https://portal.gdc.cancer.gov) from The Cancer Genome Atlas (TCGA, n = 371) to download the transcriptome data and clinical information of HCC. Fragments per Kilobase million format was used to calculate transcription spectra.

Patient specimens

From 2019 to 2022, all HCC and paired para-cancerous tissues were obtained from Fujian Medical University Union Hospital and randomly used in this experiment. All patients were diagnosed with HCC based on tissue specimens. The patients had not undergone chemotherapy, radiotherapy, or other new adjuvant therapy prior to surgery.

Cell culture

For the preparation of HCC^RIN-sup, the supernatant was collected after 48 h of RIN1-treated HCC cells and the supernatant was collected by filtration through a 0.45 μm filter to remove cells and cell debris. CD8⁺ T cells were treated with cell-free supernatant for 48 h. The mouse HCC cell line Hepa1-6, human HCC cell lines HepG2 and Huh7, and HEK-293T were purchased from American type culture collection (ATCC). HEK-293T cells were utilized for lentiviral packaging. HepG2, Huh7, and Hepa1-6 were cultured in Dulbecco’s Modified Eagle Medium (DMEM) medium containing streptomycin, penicillin, 12% (v/v) fetal bovine serum (FBS) at 37°C containing 5% carbon dioxide (CO₂). HepG2 and Huh7 were identified by STR (2020.07). HCC cells or CD8⁺ T cells were stimulated with RIN1 10 μM, Rodatristat 10 μM, IDO1-IN-18 10 μM, MHY-1485 10 μM, L-Kynurenine 200 nM and HCC^RIN1-sup for 48 h.

Weighted Gene Co-expression Network analysis (WGCNA) and hub genes identification

“WGCNA”³³ R package was employed to construct a weighted co-expression network to analyze the relationship between cancer-immune cycle and HCC gene expression data. Cancer-immune cycle score downloaded from Tracking Tumor Immunophenotype (TIP) database (http://biocc.hrbmu.edu.cn/TIP/) served as the clinical phenotype of the network. We chose fit β-value = 12 as cut off to build a scale-free network. HCC gene expression matrix was converted to a proximity matrix and then to a topology matrix. Genes were clustered using an average linkage hierarchical clustering method based on topological overlap. According to the standard of hybrid dynamic shearing tree, the minimum number of genes in each module was set to 50. The eigengenes of each module were calculated, and the cluster height was set to 0.25. Genes that could not be clustered into other modules were classified as gray module. According to the characteristic genes of each module, the correlation coefficient between each module and the cycle was calculated, then the module phenotype correlation heatmap was drawn.

Calculation of immune cell infiltration and immunotherapy response

Single sample gene set enrichment analysis (ssGSEA) algorithm in gsva R package³⁴ was performed to calculate the abundance of immune cell infiltration in HCC tissues based on the gene expression. ssGSEA was based on the specific genome of leukocytes, and HCC gene expression data was converted into the infiltrating abundance of immune cell populations, resulting in a ssGSEA score matrix, where the columns were the scores for each immune cell and the rows were patients’ IDs. “Estimate” R package³⁵ calculated “ESTIMATEScore”, “StromalScore”, and “ImmuneScore” of each sample with built-in markers. Correlation coefficients > 0.5 and P value < 0.05 were selected as criteria for statistical significance. Tumor Immune Dysfunction and Exclusion (TIDE) (http://tide.dfci.harvard.edu/)³⁶ and Submap algorithms³⁷ (http://cloud.genepattern.org/gp) were employed to explore whether RBPJ could predicted the efficiency of a patient’s response to immunotherapy. TIDE algorithm was used to measure the TIDE score of each HCC specimen in TCGA database, and then Submap algorithm was performed to compare the difference in immunotherapy efficiency between RBPJ^high and RBPJ^low expression groups.

Harvesting and processing of human liver samples

Fresh resected liver tissue samples were taken within 2 h to the laboratory to start tissue dissection and processing. First, tissue samples were thoroughly washed with phosphate buffered saline (PBS) to remove visible blood clots and to reduce blood leukocytes contamination. After tissue was minced using scalpels into ~3–5 mm diameter pieces and digested (1 mg/mL collagenase IV, 10 mg/mL Deoxyribonuclease-1 (DNase I), 10% FBS, RPMI 1640) at 37°C for 45 min using the gentle MACS Octo Dissociator with Heaters and continuous shaking. The enzymatic reaction was stopped by adding ethylenediaminetetraacetic acid (EDTA) 2 mM in PBS to a double volume of the sample. Afterward, the homogenate was filtered through a 100 mm cell strainer and centrifuged at 400 × g for 8 min at 4°C to pellet the cells and myelin. This was followed by myelin removal step by gradient centrifugation with 30% Percoll in PBS (1592 × g for 30 min at 4°C; without brakes during deceleration) using a 50 mL tube with a lid for a fixed angle rotor fitting in a centrifuge. After myelin (the top white layer) separation, the middle transparent layer without the bottom layer of red blood cells was collected and filtered once more through a 100 mm cell strainer. The single-cell suspension was washed in PBS and centrifuged at 400 × g for 8 min at 4°C to pellet the cells. Next, cells were ready for flow cytometry analysis or storing samples by freezing. For CD8⁺ T cell isolation, the human CD8 microbeads were used according to the manufacturer’s protocol, and cell purity was verified with FACSAria™ III, and sorted cells with a purity >95% were considered eligible specimens.

Harvesting and processing of mouse liver samples

Mice were euthanized with CO₂ and perfused with PBS through the left ventricle of the heart using a 25-G butterfly needle attached to a 50 mL syringe. The collected complete liver sample was dissected into ~1–3 mm diameter pieces using scissors and digested (0.4 mg/mL collagenase IV, 10 mg/mL DNase I, 10% FBS, RPMI 1640) at 37°C for 30 min applying continuous shaking. The enzymatic reaction was stopped by adding EDTA in PBS to a final concentration 5 mM. To homogenize the sample, it was repeatedly aspirated and ejected using a 5 mL syringe with a 20-G needle until a uniform homogenate was formed. Afterward, the homogenate was filtered through a 70 mm cell strainer and centrifuged at 400 × g for 8 min at 4°C to pellet the cells and myelin. This was followed by myelin removal step by gradient centrifugation with 30% Percoll in PBS (1592 × g for 30 min at 4°C; without brakes during deceleration) using a 50 mL tube with a lid for a fixed angle rotor fitting in a centrifuge. After myelin (the top white layer) separation, the middle transparent layer without the bottom layer of red blood cells was collected and filtered through a 70 mm cell strainer. The single-cell suspension was washed in PBS and centrifuged at 400 × g for 8 min at 4°C to pellet the cells. Cells were then ready for flow cytometry analysis. For CD8⁺ T cell isolation, the mouse CD8 microbeads were used according to the manufacturer’s protocol, and cell purity was verified with FACSAria™ III, and sorted cells with a purity >95% were considered eligible specimens.

Plasmid, transfection and lentivirus production and infection

Rbpj-KD (shRNA sequences were commissioned to design and synthesized by Tsingke Biotechnology Co., Ltd., shRNA primers were shown in the supplementary material) was cloned into pLKO.1-mCherry-puro vector, the scrambled sequence from our previous publication served as a non-targeting control³⁸. Open reading frames (ORF) of Cd274 and PD-L1 (ORF sequences were commissioned to design and synthesized by Tsingke Biotechnology Co., Ltd.) were cloned into pLenti-mCherry-puro vector. The stable cell lines were obtained using lentiviral infection. psPAX2-puro, pMD2.G-puro, and pLenti-puro were co-transfected into HEK-293 cells using Lipo3000. After the medium was discarded, the viral particles were collected with RPMI1640 medium, and the viral particles were collected by filtration through a 0.45 μM filter. Polybrene was added to a final concentration of 10 μg/ml when cells were infected, cells were infected for 48 h, then replaced with fresh RPMI1640 medium and subsequent experiments were performed.

Metal-isotope-tagged antibodies

Pre-conjugated antibodies to metal isotope were purchased from Fluidigm or commercial suppliers in purified form and conjugated in house using the Maxpar X8 chelating polymer kit according to the manufacturer’s instructions.

Cell surface staining for mass cytometry

To avoid nonspecific binding of antibodies, the sample was incubated at 4°C for 15 min in Human TruStain FcX (Fc Receptor Blocking Solution) or TruStain FcX™ (anti-mouse CD16/32) Antibody. Without washing, the cells were spun down, resuspended in the antibody mixture in PBS, and incubated at 4°C for 30 min. To optimize antibody staining for chemokine receptors, the sample was incubated at 37°C for 10 min, and at 4°C followed 20 min. After staining the surface antibodies, we added Cell-ID Cisplatin to the sample for 3 min to discriminate viable/dead cells. Then, the sample was washed once in PBS and centri Cisplatin fuged to pellet the cells.

Intracellular cytokine staining for mass cytometry

Cells were permeabilized using FOXP3 Fix/Perm Buffer Set according to the manufacturer’s instructions for 45 min at 4°C. Subsequently, the sample was washed once in Perm/Wash buffer and incubated in the antibody mixture in Perm/Wash buffer for 30 min at 4°C. The sample was washed once in Perm/Wash buffer and centrifuged to pellet the cells.

Cell preparation and mass cytometry acquisition

After cell surface and intracellular antibody staining, the cells were incubated in 4% paraformaldehyde aqueous solution overnight. Prior to acquisition the cells were pelleted without washing and resuspended in up to 1 mL of diluted 1:3000 Cell-ID Intercalator-Ir + Maxpar Fix and Perm Buffer for 3 h. After the sample was washed twice in PBS and twice in ddH₂O, diluted to 1.5 × 10⁶ cells/mL in ddH₂O containing 10% EQ Four Element Calibration Beads and filtered through a 40 mm filter cap FACS tube. Samples were analyzed with a Helious CyTOF2. Quality control and tuning processes on the Helios CyTOF2 were performed following the guidelines for the daily instrument operation. Data were collected as fcs files.

Removal of dead and dying cells

Leukocytes were collected and centrifuged at 300 × g for 5 min. Supernatants were removed and the cells were resuspended in 100 μL of dead cell-removal beads per 1 × 10⁷ cells as described by the manufacturer. The mixture was incubated at room temperature (RT) for 15 min and added to the MS column. The columns were then washed four times with binding buffer. Live cells were collected from the flow-through.

Preprocessing of cytometry data

Raw mass cytometry data were normalized using the MATLAB version of the Normalizer tool³⁹. Cells were assigned by manually gating on Event length and DNA (¹⁹¹Ir and ¹⁹³Ir) channels, followed by the dead cell discrimination analyzing ¹⁹⁵Pt expression using FlowJo Software. Doublets were excluded using Gaussian discrimination channels. Next, data were concatenated and de-barcoded using Boolean gating in FlowJo software. The normalized data containing living cells from every individual sample were manually exported from FlowJo Software and imported into R studio of R using the R packages “flowCore”⁴⁰ and “flowWorkspaceData”⁴¹ (R Foundation for Statistical Computing). Before automated high-dimensional data analysis, the mass cytometry data were transformed with a cofactor in the range of 5 and 60 using an inverse hyperbolic sine function⁴². For flow cytometry data, the compensation matrix was corrected using FlowJo software. After live, single, CD3 positive and compensated cells were exported and imported into R Studio. Before automated high-dimensional data analysis, flow cytometry data were transformed using an inverse hyperbolic sine function with a cofactor in the range of between 300 and 600. Additionally, all cytometry data were normalized between 0 and 1 to the 99-999^th percentile of the merged sample in each batch.

Automated population identification

To identify T cell populations accurately, we first carried out a step of FlowSOM clustering to generate a starting point of 100 nodes, on pre-processed and combined mass/or flow cytometry datasets^43,44. This was then followed by expert-guided manual meta-clustering using parameters. The respective k-value was manually chosen (in the range of between 20 and 30); identified clusters were annotated and merged based on a similarity of antigen expression in order to uphold the biological relevance of the dataset. Manually-annotated clusters were used to calculate the relative frequencies of T cell populations. Heatmaps display median expression levels of all markers per merged population and plotted using the R package “pheatmap”. From mass cytometry datasets, we pre-selected major populations and performed additional FlowSOM analysis to identify smaller cell subsets. We calculated the median antigen expression among selected cell types using the R package “dplyr”. For data visualization, we applied various dimensionality reduction techniques. For a complex overview of the compartment, we used t-Distributed Stochastic Neighbor Embedding (t-SNE)⁴⁵. To create a t-SNE of isolated T cells, we pooled equally proportioned 120,000 T cells from the datasets.

Plate clone formation assay

The HepG2 and Huh7 cells were digested and then resuspended in serum-free medium, and the cells were seeded into a 6-well culture plate at a density of 1 × 10³ cells per well. Fourteen days later, the cells were continually cultured. Every 3 days, cells and clones were observed microscopically and sub-cultured. After colony formation was completed, the colonies formed by cells were photographed under a microscope and washed three times with PBS. Then, add 1 mL of crystal violet staining solution to each well and stain for 20 min. Finally, the six-well plate that formed the clones was scanned.

Cell proliferation assay

Cell proliferation was tested using Cell Counting Kit 8. HepG2 and Huh7 cells (1000 cells per well plate) were plated into 96-well plates with RIN1 concentration of 10 nM for 48 h.

Cell cycle detection

Flow cytometry was used to detect cell cycle of HCC cells. HepG2 and Huh7 cells (1  × 10⁵ cells per well plate) were plated into 6-well plates with RIN1 concentration of 10 nM for 48 h. The cells were fixed overnight with 70% ethanol. The cells were resuspended in 250 μL of PI/RNase at a concentration of 1 × 10⁶ cells/mL, mixed, and incubated for 20 min at 4 °C. Relative light units were detected by FACSCelesta™ Flow Cytometer within 1 h.

Apoptosis detection

Flow cytometry was performed to examine apoptosis of HCC cells. HepG2 and Huh7 cells (1 × 10⁵ cells per well plate) were plated into 6-well plates with RIN1 concentration of 10 nM for 48 h. The cells were resuspended in 250 μL of binding buffer at a concentration of 1 × 10⁶ cells/mL, and then 5 μL of Annexin V-APC and PI were added, mixed, and incubated for 20 min at 4 °C. Relative light units were detected using FACSCelesta™ Flow Cytometer within 1 h.

Invasion and migration assays

We used transwell plates to assay cell invasion and migration. In the migration assay, we starved the cells in a serum-free medium for 12 h at 37 °C with 5% CO₂. We then added 700 µL of DMEM with 20% FBS to the lower well and 500 µL of serum-free medium, including 1 × 10⁶ cells to the upper transwell inserts. After culturing for 48 h at 37 °C with 5% CO₂, we counted the number of cells that adhered to the lower surface of the insert membrane. We performed the invasion assay in the same way, except that the transwell insert membrane was coated with Matrigel.

Cell scratch test assay

Once cultured, the cells reached close to 100% confluence, the cell monolayer was mechanically scratched. Scratch healing was observed at 0, 12, 24, and 48 h under a microscope at ×40 magnification (white light bright field). We used ImageJ software to calculate the migration area on the scratch, using the following formula: Scratch area rate (%) = 12, 24, and 48 h after migration scratch area/initial scratch area ×100%.

Quantitation of glucose uptake

Cells were grown to 75% confluency on 96-well plates. The glucose analog 2-NBDG (0.15 mg/mL) were added to the growth medium and incubated for 10 min at 37 °C. Excess of 2-NBDG was removed by rinsing with PBS. Plates were sealed with optical film and photographs were taken with the EVOS FLoid imaging system.

Quantification of pyruvic acid

Centrifugation at 15,000 rpm at 4 °C for 10 min, 20 μL of the supernatant was applied to a 96 well plate. Approximately 30 min after crushing, 43 μL of distilled water and 66 μL of 0.25 g/L DNPH in 1 M HCl were added to the supernatant in the 96 well plate and the plate was rolled gently for 10 min at 37 °C. After 10 min, 66 μL of 1.5 M NaOH was added. The absorbance at 515 nm was measured by the microplate reader. Standards were prepared from a 40 mM sodium pyruvate solution, ranging in concentration from 0.04 to 6 mM.

Multiplex immunohistochemistry

Tissues were fixed with 4% paraformaldehyde overnight at 4 °C, dehydrated in different concentrations of ethanol and embedded in paraffin, and cut to 5 μm with a microtome. Sections were placed in Improved Citrate Antigen Retrieval Solution and microwaved for 30 min to achieve antigen retrieval. Sections were blocked with 5% Bovine serum albumin (BSA) for 1 h and incubated with primary antibody overnight. After three washes with PBS, sections were incubated with secondary antibodies. Photographs were taken with the Vectra3 Automated Quantitative Pathology Imaging system.

Flow cytometry

To validate surface marker expression, the cells were directly stained with the indicated fluorochrome-conjugated antibodies for 30 min in Cell Staining Buffer at 4 °C and analyzed by flow cytometry. The cells were then washed in PBS two times, resuspended in Cell Staining Buffer and analyzed by flow cytometry. Samples were acquired and recorded in a FACSCelesta™ Flow Cytometer, and data were analyzed with FlowJo software. Gating strategies were provided in Supplementary Figs. 18–19.

Western blot

Cells and tissues were lysed with RIPA Lysis Buffer, centrifuged at 12,000 rpm at 4°C for 30 min. Total protein concentration in supernatant was measured by BCA Protein Quantification Kit. Equal amounts of protein were loaded on 12% polyacrylamide gel, separated, and electroporated onto polyvinylidene difuoride membranes. Proteins on the membrane were blocked with QuickBlock™ Blocking Buffer for Western Blot for 15 min. Proteins were incubated with primary antibodies at 4°C overnight followed by secondary antibodies for 1 h at RT. Visualized with SuperSignal West Dura, imaged with ChemiDoc MP Imaging System.

Real‑time quantitative PCR

Total RNA was extracted and purified with FastPure Plant Total RNA Isolation Kit and reverse transcription was employed using HiScript II Q Select RT SuperMix for qPCR. RT-qPCR was performed on Applied Biosystems 7500 and 7500 Fast Real-Time PCR Systems using ChamQ SYBR qPCR Master Mix. PCR cycling conditions were 95 °C for 30 s, followed by 40 cycles of 95 °C for 10 s, 63 °C for 10 s and 72 °C for 30 s. The cDNA melting curve was set as usual. Relative gene expression was calculated using 2^−ΔΔCT method, calibrated against GAPDH or beta Actin.

Detection of L-kynurenine content

The content of L-kynurenine secreted by Huh7 and HepG2 cells was detected using Kynurenine ELISA kit. Dilute the standard. Add 40 uL of sample or standard to the well and incubate at 37 °C for 30 min. Wash three times with PBS, add 50 uL of enzyme-labeled antibody, incubate at RT for 30 min, add 50 uL of chromogenic solution and 50 uL of stop solution. The absorbance value of each sample at a wavelength of 450 nm was measured with Multiskan™ FC System.

Detection of ATP content

ATP content was detected with ATP Assay Kit. Add 100 uL of ATP detection working solution to the well, leave it at RT for 5 min, and add 20 uL of sample or standard to the well. The relative light units of each sample were detected with Multiskan™ FC System.

Transcriptome sequencing

Total RNA was extracted and purified with FastPure Plant Total RNA Isolation Kit according to the manufacturer’s protocol. RNA-seq libraries were constructed with VAHTS® mRNA-seq V3 Library Prep Kit for Illumina. Sequencing was performed using Illumina NovaSeq 6000 platform (provided by Novogene Co., Ltd.), and the sequencing depth of each sample was 6 G bases.

Untargeted metabolomics by Liquid Chromatography (LC)-MS/MS

The metabolites in Huh7 cells were extracted. LC-MS/MS analysis was performed using Vanquish UHPLC system and Orbitrap Q Exactive HF-X mass spectrometer in both positive and negative modes (provided by Novogene Co., Ltd.).

Single-cell RNA-Seq library preparation

Single-cell suspensions were prepared from biopsies as described above. Bulk population cells were directly subjected to the 10× mass genomics chip, targeting 10,000 simultaneously captured live events. Each cell was uniquely barcoded during the cDNA library generation by Single Cell 3′ Reagent Kits v2 per the manufacturer instructions and subsequently sequenced on an Illumina HiSeq 2500 at the CCHMC DNA Sequencing and Genotyping Core.

Single-cell RNA-Seq data analysis

All single-cell RNA-Seq data were processed using Cell Ranger version 3.0.2 and the mm10 reference. The UMI count matrix was imported to Seurat⁴⁶ for postprocessing and downstream analysis. For each sample, doublets were first filtered out using Scrublet⁴⁷. Then, cells either with high mitochondrial reads, a low number of detected genes, or excess UMI counts were discarded. The UMI count matrix was normalized and scaled following the standard Seurat pipeline by adjusting for mitochondrial read counts and UMI counts. Preliminary clustering was performed using the top 20 principal components to detect major cell types first, and this information was projected onto UMAP⁴⁵. Also, marker genes were called for cell type confirmation, detecting genes exclusively expressed in distinct populations. To do so, we performed a series of pairwise differential expression analyses against each of all the other cell clusters using the “FindMarkers” function in Seurat with the parameters log₂FC > 0.25 and minimal proportion of cells expressing a gene (min.pct) >0.1, and we selected genes that were commonly highly expressed in each cluster. The proportion of each cell type was measured for each sample and was compared between the different disease statuses using a two-tailed t-test to examine the change by the groups. Pseudotime analysis was performed using Slingshot for all sample integrations to map each cell from different groups onto a common pseudotime axis, which corresponded to the mouse T cell differential trajectory⁴⁸. Cell densities along the pseudotime were visualized and compared across the different statuses using a density plot, and a two-tailed t-test was performed to test the significance.

Cleavage Under Targets and Tagmentation (CUT&Tag)

CUT&Tag was performed based on a protocol published by Kaya-Okur et.al.⁴⁹. In brief, cells were harvested, counted (50,000 cells) and centrifuged for 3 min at 600 × g at RT. Cells were washed twice in 1.5 mL Wash Buffer (20 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES) pH 7.5; 150 mM NaCl; 0.5 mM Spermidine; 1 × Protease inhibitor cocktail, EDTA free). 10 μL concanavalin A coated magnetic beads were added per sample and incubated at RT for 15 min. The supernatant was removed and bead-bound cells were resuspended in 50 μL Dig-wash Buffer (20 mM HEPES pH 7.5; 150 mM NaCl; 0.5 mM Spermidine; 1 × Protease inhibitor cocktail; 0.05% Digitonin) containing 2 mM EDTA. The primary antibody of RBPJ was diluted 1:50 in 50 μL of Dig-Wash buffer and then incubated on a rotator overnight at 4 °C. The primary antibody was removed and an appropriate secondary antibody was diluted 1:100 in 100 μL of Dig-Wash buffer and cells were incubated at RT for 30 min. Cells were washed using 1 mL Dig-Wash buffer to remove unbound antibodies. A 1:200 dilution of pAG-Tn5 was prepared in Dig-300 Buffer (0.05% Digitonin, 20 mM HEPES, pH 7.5, 300 mM NaCl, 0.5 mM Spermidine, 1 × Protease inhibitor cocktail) and incubated at RT for 1 h. Cells were washed twice with 1 mL Dig-300 Buffer to remove unbound pAG-Tn5. Then, cells were resuspended in 50 μL Tagmentation buffer (10 mM MgCl₂ in Dig-wash Buffer) and incubated at 37 °C for 1 h. Next, 1 μL of 10% sodium dodecyl sulfate (SDS) was added to 50 μL of sample and incubated at 55 °C for 10 min to stop tagmentation. To extract the DNA, 1.5 × Ampure XP beads were added to each tube. The final DNA products were eluted with 20 μL Millipore water and prepared to amplify libraries. PCR cycling conditions: 72 °C for 5 min; 98 °C for 30 s; 12 cycles of 98 °C for 10 s and 63 °C for 30 s; final extension at 72 °C for 1 min. The final libraries were purified by adding 1.1 × Ampure XP beads and eluted in 30 μL 10 mM Tris pH 8.0. The size distribution of libraries was determined by Agilent 4200 TapeStation analysis. Libraries were sequenced on Illumina Novaseq 6000 (150-bp paired ends).

Statistics and reproducibility

All data analysis was done with GraphPad Prism V.8 or SPSS V.19, and data were presented as mean ± standard deviation (χ ± s). When comparing two samples, the paired two-tailed t-test was used if the data were normally distributed and the variances were homogeneous. If the data were normally distributed with unequal variance, Wilcoxon rank sum test was used. χ²-test was used to analyze the relationship between RBPJ and clinical indicators. Pearson analysis was employed to analyze the correlation between RBPJ and immune checkpoints or abundance of leukocytes. Kaplan–Meier analysis was performed for the survival analysis. Flow cytometry data was analyzed using FlowJo software V10. P < 0.05 was set as the criterion for statistical significance.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.