2.1 Data acquisition and preparation

The clinical and transcriptomic data for PTC used in this study were sourced from The Cancer Genome Atlas (TCGA) database. All thyroid cancer patients’ transcriptomic and clinical data were downloaded via the UCSC Xena platform. Clinical pathology information filtering yielded a total of 353 PTC patients, including both clinical information and gene expression data. Additionally, PTC-related transcriptomic RNA sequencing datasets were downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) for meta-analysis, as detailed in Table S1. Since these data were obtained from public databases, obtaining ethical committee approval was not required.

2.2 GSEA analysis

In this study, Gene Set Enrichment Analysis (GSEA) was conducted using the Signal-to-Noise ratio (Signal2Noise) method for ranking metrics, with all other parameters set to their default values. GSEA was performed using the OmicShare tool. This analysis incorporated the Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG), as well as the Hallmark gene sets, to provide a comprehensive overview of the biological signals. The analyses were conducted on the Omicshare platform [20].

2.3 cBioPortal online analysis

To investigate the expression differences of CLDN1 and EGFR across various cancer samples, data analysis was carried out using the cBioPortal database (https://www.cbioportal.org/). The TCGA PTC cohort was selected to retrieve gene expression data. After data acquisition, CLDN1 and EGFR gene expression data were categorized into median groups using the built-in analysis tools of cBioPortal, and the clinical characteristics of the high and low expression groups were examined through the website [21].

2.4 HPA database analysis

To investigate the differential expression of CLDN1 and EGFR proteins across various tumor types and their corresponding normal tissues, the Human Protein Atlas (HPA) database (https://www.proteinatlas.org/) was utilized. Extensive immunohistochemical analysis data on the expression patterns of human proteins in diverse tissues and cells were provided by the HPA database. For CLDN1 and EGFR proteins, the database was searched for immunohistochemistry (IHC) results in a variety of tumor tissues and corresponding normal tissues [22].

2.5 Gene set download and correlation analysis

To explore the correlation between CLDN1 and EGFR and the genes involved in the EMT pathway, gene sets were downloaded from MsigDB with the pathway name “REACTOME_TGF_BETA_SIGNALING_IN_EMT_EPITHELIAL_TO_MESENCHYMAL_TRANSITION,” comprising 16 genes. The correlation plots for the two genes were generated using the R package ggstatsplot (v0.11.1), and the multi-gene correlation plots were visualized with the R package pheatmap (v1.0.12). Spearman’s rank correlation analysis was employed to assess the relationship between non-normally distributed quantitative variables, considering p-values less than 0.05 as statistically significant [23] (Figure S1).

2.6 Meta-analysis

Meta-analysis was conducted using the meta for the package (version 4.4.0) in R, with the effect size estimated as the mean difference (MD) alongside its 95% confidence interval (CI). The aim of heterogeneity testing (i.e., testing for consistency across studies) was to determine the feasibility of pooling the individual studies, utilizing the Q test and chi-squared test. Heterogeneity was assessed based on the I2 value and P-value. The absence of statistical heterogeneity, indicated by a P-value greater than 0.05 and an I2 value less than 50%, suggests that the variations across studies are attributable to sampling error, permitting the use of a fixed-effect model for the meta-analysis. Conversely, significant statistical heterogeneity, reflected by a P-value less than 0.05 and an I2 value exceeding 50%, necessitates the adoption of a random-effects model for pooling the results [24].

2.7 Cell culture

The normal thyroid epithelial cell line (HT-ori3)—CBP61205 and three PTC cell lines (IHH-4, SNU-790, TPC-1)—with catalog numbers CBP61201, CBP61209, and CBP60257, respectively, were purchased from Nanjing KeyGen Biotech. These cells were cultured in RPMI-1640 medium from Gibco (USA) supplemented with 10% fetal bovine serum (10099158), 100 U/mL penicillin, and 100 mg/L streptomycin (15140148). The cells were maintained at 37 °C in a 5% CO2 environment with the medium changed every 2 days. Passaging was performed when cell confluence reached 80% to 90%, and cells in logarithmic growth phase were used for experiments. Human CD8+ T cells were obtained from Huizhi Yuan Biological Technology (Suzhou) Co., Ltd. (082A04.11) and cultured in high-glucose RPMI-1640 medium (R8758, Sigma-Aldrich, USA) containing 4.5 g/L glucose for experimentation [25].

2.8 Cell processing and grouping

TPC-1 cells in logarithmic growth phase were digested with trypsin, centrifuged, and resuspended at a concentration of 5 × 104 cells/mL. The cells were then seeded into 6-well plates at a volume of 2 mL per well. Prior to constructing the in vitro cell model, lentiviruses (MOI = 10, viral titer of 1 × 108 TU/mL) were added to the cell culture medium. Post 48 h, stable cell lines were selected using 2 µg/mL puromycin (UC0E03, Sigma-Aldrich, USA) for a duration of 2 weeks. The cell transfection groups were as follows: (1) sh-NC group: transfected with sh-NC lentivirus; (2) sh-CLDN1-1 group: transfected with sh-CLDN1-1 lentivirus; (3) sh-CLDN1-2 group: transfected with sh-CLDN1-2 lentivirus; (4) sh-EGFR-1 group: transfected with sh-EGFR-1 lentivirus; (5) sh-EGFR-2 group: transfected with sh-EGFR-2 lentivirus. The lentiviral vectors used were designed and synthesized by Guangzhou Ruibo Biotechnology Co., Ltd [26].

For the co-culture and grouping: (1) sh-NC group: TPC-1 cells transfected with sh-NC lentivirus were co-cultured with CD8+ T cells; (2) sh-CLDN1 group: TPC-1 cells transfected with sh-CLDN1 lentivirus were co-cultured with CD8+ T cells; (3) sh-EGFR group: TPC-1 cells transfected with sh-EGFR lentivirus were co-cultured with CD8+ T cells. TPC-1 cells and CD8+ T cells were co-cultured in a shared system at 37 °C, 5% CO2 for 48 h. TPC-1 cells were placed in the upper chamber, while CD8+ T cells were seeded in the lower chamber of a Transwell plate (CLS3412, Corning, USA). Separated by a 0.4 µm membrane, soluble factors were allowed to pass through, while the cells were kept separate. After 3 days, the supernatant and cultured cells were collected for further experiments [27].

2.9 Flow cytometry

Cell apoptosis rate was determined using the Annexin V-FITC Cell Apoptosis Detection Kit (C1062S, Beyotime, Shanghai). Cell density was adjusted to 1 × 105 cells/mL, and 3 mL of cell suspension was centrifuged twice at 500 rpm for 5 min each time in a 10 mL centrifuge tube. After removing the culture medium, cells were resuspended in 100 μL binding buffer, and 5 μL each of Annexin V-FITC and PI were added. The mixture was gently vortexed, kept in the dark at room temperature for 15 min, and then analyzed using flow cytometry to detect FITC and PI fluorescence. Quadrants were defined as follows: Q1: upper left quadrant (UL) representing cell debris without intact membranes or cells undergoing death for other reasons; Q2: upper right quadrant (UR) indicating late apoptotic cells; Q3: lower left quadrant (LL) representing normal (viable) cells; Q4: lower right quadrant (LR) indicating early apoptotic cells. Statistical analysis focused on quadrants Q2 and Q4 to determine the apoptosis rate [28].

2.10 Enzyme-linked immunosorbent assay (ELISA)

IFN-γ (ab174443, Abcam) and Granzyme B (ab235635, Abcam) levels in the extracellular model were measured using ELISA kits according to the manufacturer’s instructions [29].

2.11 EDU experiment for assessing cell proliferation

The cells of interest were seeded in a 24-well plate and EdU (C10310-2, Guangzhou RiboBio Co., Ltd., Guangzhou) was added to the culture medium to achieve a concentration of 10 µmol/L. The cells were then incubated in a cell culture incubator for 2 h. After removing the culture medium, the cells were fixed with a PBS solution containing 4% paraformaldehyde at room temperature for 15 min. The cells were then washed twice with PBS containing 3% BSA, incubated at room temperature with PBS containing 0.5% Triton-100 for 20 min, followed by two washes with PBS containing 3% BSA. Subsequently, 100 µL of staining solution was added to each well and incubated at room temperature in the dark for 30 min. DAPI nuclear staining was performed for 5 min, and after sealing the slides, 6–10 random fields were observed under a fluorescence microscope (BX63, Olympus, Japan). The number of positive cells in each field was recorded. The EdU labeling rate (%) was calculated as the number of positive cells divided by the sum of positive and negative cells, multiplied by 100% [30].

2.12 Wound healing assay for cell migration

Using a marker pen, uniform horizontal lines were drawn on the back of a 6-well plate, spaced 1 cm apart, crossing over the wells. Approximately 5 × 105 cells were seeded in each well, cultured until confluence reached 100%, and then a straight scratch was made across the cell layer at the marked line using a pipette tip perpendicular to the cell surface. After scratching, the plate was washed three times with sterile PBS to remove non-adherent cells, ensuring the gap created by scratching was clearly visible. Fresh serum-free culture medium was added, and the cells were incubated at 37 °C in a 5% CO2 incubator. After 24 h, the width of the scratch was observed under a microscope, photographed, and the results were recorded. The migration rate analysis was performed using Image J software [31].

2.13 Transwell experiment for assessing cell invasion capability

ECM gel (E1270, Sigma-Aldrich, Germany) was added to the upper chamber of a 24-well Transwell plate (8 μm) and incubated in a 37 °C incubator for 30 min to allow the gel to solidify. Cells transfected for 48 h were collected, resuspended in serum-free culture medium at a concentration of 105 cells, and seeded in the upper chamber with 200 μL of cell suspension (2 × 104 cells/well) in the Transwell plate. The lower chamber was filled with 800 μL of culture medium containing 20% FBS. After incubating for 24 h at 37 °C, the Transwell plate was removed, washed twice with PBS, fixed with formaldehyde for 10 min, and then rinsed thrice with water. The cells were stained with 0.1% crystal violet, left at room temperature for 30 min, washed twice with PBS, and the cells on the upper surface were wiped off with a cotton ball. The invaded cells were photographed using an inverted optical microscope (CKX53, Olympus, Japan), and their invasion capability was quantified and analyzed using Image J software [32].

2.14 MTT assay for measuring cell viability

The cells of interest were seeded in a 96-well cell culture plate at a density of 3–5 × 104 cells/mL and cultured for 48 h. MTT solution (10 mg/mL, ST316, Beyotime Biotechnology Co., Ltd, Shanghai) was added to the cell suspension and incubated for 4 h, followed by the addition of DMSO with shaking for 10 min. The absorbance (OD 490 nm) was measured using a spectrophotometer (Laspec, China) to assess cell viability [33].

2.15 qRT-PCR

Total RNA was extracted from tissues or cells using Trizol reagent (15596026, Invitrogen, USA), and the concentration and purity of total RNA were assessed at 260/280 nm using NanoDrop LITE (ND-LITE-PR, Thermo Scientific™, USA). The extracted total RNA was reverse transcribed into cDNA using the PrimeScript RT reagent Kit with gDNA Eraser (RR047Q, TaKaRa, Japan). Subsequently, the expression levels of each gene were measured by RT-qPCR using SYBR Green PCR Master Mix reagents (4364344, Applied Biosystems, USA) and the ABI PRISM 7500 Sequence Detection System (Applied Biosystems).

The primers for each gene were synthesized by TaKaRa (Table S2), with GAPDH serving as the reference gene. The relative expression levels of each gene were analyzed using the 2−ΔΔCt method, where ΔΔCt = (average Ct value of the target gene in the experimental group − average Ct value of the reference gene in the experimental group) − (average Ct value of the target gene in the control group − average Ct value of the reference gene in the control group) [34].

2.16 Western blot analysis

Tissues or cells were collected and lysed with enhanced RIPA lysis buffer (containing a proteinase inhibitor) from Beyotime Biotechnology Co., Ltd, Shanghai (Product code: P0013B) to extract total proteins. The protein concentration of each sample was determined using a BCA protein quantification kit (Product code: P0012, Beyotime Biotechnology Co., Ltd, Shanghai), and 30 μg of protein was loaded per lane after adjusting with deionized water. A 12.5% SDS separating gel and a stacking gel were prepared. The samples were mixed with loading buffer, boiled at 100 °C for 5 min, cooled on ice, centrifuged, and equal amounts were loaded onto each lane for electrophoresis separation. The proteins were then transferred from the gel onto a PVDF membrane. The membrane was blocked with 5% skim milk powder at 4 °C overnight, followed by incubation with primary antibodies: rabbit anti-CLDN1 (1:800, ab307692, Abcam, Cambridge, UK), EGFR (1:200–2000, ab52894, Abcam, Cambridge, UK), TGFβ1 (1:1000, ab215715, Abcam, Cambridge, UK), TGFβR1 (1:1000, ab235578, Abcam, Cambridge, UK), TGFβR2 (1:1000, ab283230, Abcam, Cambridge, UK), and GAPDH (1:1000, ab9485, Abcam, Cambridge, UK), overnight at 4 °C. The membrane was washed thrice with PBS at room temperature for 5 min each, followed by incubation with goat anti-rabbit IgG HRP-conjugated secondary antibody (1:5000, ab6721, Abcam, Cambridge, UK) for 1 h at 37 °C. After washing the membrane thrice with PBS buffer at room temperature for 5 min each, an equal amount of Pierce™ ECL Western Blotting Substrate (32209, Thermo Scientific™, USA) components A and B were mixed in a darkroom, added to the membrane, and exposed in a gel imaging system. The bands in the Western blot images were quantified for grayscale using the Bio-Rad imaging system (BIO-RAD, USA) and ImageJ analysis software, with GAPDH as the internal control [35].

2.17 Establishment of subcutaneous tumor model in mice

Thirty-six male NOD-SCID mice aged 4–6 weeks and weighing 15–25 g were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd (Beijing, China). The NOD-SCID mice were housed in an SPF-level animal laboratory and acclimatized for 1 week to observe their status. The animal experimental processes were approved by the Ethnic Committee of Pinghu First People’s Hospital (CMUXN2023083) and conducted in strict accordance to the standard of the Guide for the Care and Use of Laboratory Animals published by the Ministry of Science and Technology of the People’s Republic of China in 2006.

Stable TPC-1 cells with knocked down CLDN1 and EGFR (sh-CLDN1, sh-EGFR) or TPC-1 cells transfected with control lentivirus (sh-NC cells) were injected subcutaneously into the flank of the mice in a volume of 100 μL PBS containing 5 × 106 cells. The mice were divided into three groups, each comprising 6 mice: sh-NC group (injected with sh-NC cells), sh-CLDN1 group (injected with sh-CLDN1 cells), and sh-EGFR group (injected with sh-EGFR cells). On the 5th day post-injection, each mouse was intravenously administered 200 µL of PBS-suspended 1 × 107 CD8+ T cells [36]. Thirty days post-injection, the mice were euthanized. Tumor diameters were measured using a caliper, and tumor volume was calculated using the formula: (longest diameter × shortest diameter2 × 0.5). Tumor tissues from the mice were collected, embedded in paraffin, and subjected to immunohistochemical analysis to evaluate the expression of CD8 and Granzyme B proteins. The animal experiments were approved by our institutional review board [37].

2.18 Immunohistochemistry

After embedding and sectioning the tumor tissues from each group of mice, the slides were baked at 60 °C for 20 min. Subsequently, the slides were sequentially immersed in xylene solution for 15 min, followed by immersion in absolute alcohol for 5 min, and then changed to fresh absolute alcohol for an additional 5 min. The slides were then hydrated by sequential immersion in 95% and 70% ethanol for 10 min each. A 3% H2O2 solution was added to each slide and allowed to react at room temperature for 10 min to block endogenous peroxidases. Citrate buffer was added, and the slides were microwaved for 3 min. Antigen retrieval solution was applied, and the slides were left at room temperature for 10 min, followed by washing with PBS three times. Normal goat serum blocking solution (E510009, Shanghai Sangon Biotech) was added and incubated at room temperature for 20 min.

Primary antibodies CD8 (MA1-7632, Thermo Fisher, 1:200) and Granzyme B (14-8822-82, Thermo Fisher, 1:200, RRID: AB_468530) were added dropwise. The slides were then incubated overnight at 4 °C, washed with PBS three times, and incubated with goat anti-rabbit IgG secondary antibody (ab6721, 1:500, Abcam, UK) for 30 min. After washing with PBS, the slides were subjected to DAB chromogenic reagent (Sigma, USA), where drops of color developers A, B, and C were added to the specimen and left for 6 min for color development. Post-coloring, the specimens were briefly stained with hematoxylin for 30 s, followed by dehydration in increasing grades of ethanol (70%, 80%, 90%, 95%), and then immersed twice in absolute ethanol for 5 min each. Finally, the specimens were sealed using neutral resin. The sections were observed under a light microscope (BX63, Olympus, Japan). The proportion of positive cells expressing the protein was measured using an image analysis system (Aperio Scanscope System, Vista, CA). For assessing the proportion of follicular tissue, the area of follicular tissue within the field of view was calculated, with the ratio determined as follicular tissue area/total tissue area. Ten random samples were taken for analysis [38, 39].

2.19 Statistical methods

In this study, the statistical analysis was conducted using GraphPad Prism software version 8.0. Descriptive statistics were reported as mean ± standard deviation for continuous variables. The comparison between the two groups was performed using an independent samples t-test. For comparisons among multiple groups, a one-way analysis of variance (ANOVA) was employed. Statistical significance was considered at a threshold of P < 0.05.