Three TC datasets (GSE29265, GSE3678, GSE33630) were collected from the Gene Expression Omnibus database (GEO; https://www.ncbi.nlm.nih.gov/gds/). The GSE3678 dataset comprises 7 TC samples and 7 paired control samples, the GSE29265 dataset contains 29 TC samples and 20 control samples, while the GSE33630 dataset contains 60 TC samples and 45 control samples. Furthermore, information collected The Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov/) database, which included 59 control samples and 512 TC samples, was also used. To identify differentially expressed genes (DEGs) in the datasets, the ASSISTANT for Clinical Bioinformatic platform (https://www.aclbi.com/static/index.html#/) was utilized. A fold-change (FC) threshold of $>$2 indicated upregulation of DEGs, whereas a FC threshold of 0.5 indicated downregulation of DEGs. The criterion for statistical significance was set at a p-value of less than 0.05.

Overlapping DEGs from the four TC datasets were identified using a bioinformatics platform (https://bioinformatics.psb.ugent.be/webtools/Venn/). Intersection analysis of both upregulated and downregulated DEGs was used to obtain DEGs that overlapped. The Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org/) was used to identify PPI networks among overlapping DEGs. Further exploration of potential functional associations and regulatory networks of candidate genes was carried out by constructing PPI networks. Analysis was performed using the Molecular Complex Detection (MCODE) algorithm. For visualization, the Cytoscape open-source software platform (version 3.7.1; Cytoscape Consortium, San Diego, CA, USA) was applied to the resulting PPI network.

Overlapping DEGs were analyzed with the “glmnet” package in R software. Tuning parameters were determined through a 10-fold cross-validation process, and the Least Absolute Shrinkage and Selection Operator (LASSO) model was applied [14, 15, 16]. To find the most predictive genes, we employed the ideal $\lambda$, which represents the minimal adjusted parameter criteria. These formed the basis of the predictive model. To find the most predictive genes, we employed the ideal $\lambda$, which represents the minimal adjusted parameter criteria. Kaplan-Meier (KM) analysis estimated overall survival (OS) probabilities, and the log-rank test evaluated differences between survival curves. To evaluate relative risk, the hazard ratio (HR) for the high-risk group was computed [17]. The “timeROC” tool was used to generate receiver operating characteristic (ROC) curves, and the area under the curve [18] was calculated to determine how well the model predicted one-, three-, and five-year survival rates. Higher area under the curve (AUC) values showed improved prediction accuracy.

Correlation analysis for significant prognostic genes was performed using R software. Single-factor and multi-factor Cox regression analyses were conducted on signature genes and clinical variables (age, gender, pathological TNM (pTNM) stage) using the “forest plot” package. For each variable, the 95% confidence intervals (CIs) [19], p-values, and HRs were computed. Statistically significant variables (p 0.05) in survival analysis were found as crucial prognostic factors. The “regression modeling strategies (RMS)” program was used to create nomograms that predicted the probability of one-, three-, and five-year survival. The concordance index was calculated to assess model accuracy. Calibration curves were generated, and the closer the alignment of the survival prediction curve to the calibration curve, the better the model performance. Gene expression levels were evaluated using R and visualized with boxplots to compare tumor and paired control groups from the TCGA-thyroid carcinoma (THCA), GSE3678, GSE29265, and GSE33630 datasets.

Otwo Biotech (Shenzhen, China) provided the normal thyroid cell line (Nthy-ori3-1), the TC cell line TPC-1, and CAL-62. The selection of TPC-1 (papillary carcinoma) and CAL-62 (mesenchymal carcinoma) was based on the unequivocal identification of their molecular features in a previous study [20] and is widely used in studies of thyroid cancer migration, DNA damage, and cell death [21]. The two cell lines can be studied to cover the heterogeneity of the main subtypes of thyroid cancer. In a humidified 37 °C atmosphere with 5% CO₂, cells were cultivated in Dulbecco’s Modified Eagle Medium (DMEM; Procell; cat. no. PM150210B; Wuhan, China) supplemented with 10% fetal bovine serum (FBS) (Gibco; Thermo Fisher Scientific, Inc.; cat. no. 16000044; Waltham, MA, USA) and 1% penicillin-streptomycin (Beyotime; cat. no. C0222; Shanghai, China). We declare that all cell lines used in this study tested for mycoplasma contamination using PCR-based detection methods, and were confirmed to be free of contamination (Supplementary Fig. 1). All cell lines were obtained from OTWO (Huatuo Biotech, China). The supplier verified the authenticity of these cell lines using short tandem repeat (STR) profiling.

To conduct transient transfection, TC cells were seeded in 24-well plates at a density of 2 $\times$ 10⁵ per well. ITGA2 was knocked down utilizing two different small interfering RNAs (siRNAs) that target it. Lipofectamine 3000 (Invitrogen; Thermo Fisher Scientific, Inc.; cat. no. L3000150; MA, USA) was used to transfect TC cells for a given period of time, as indicated by the manufacturer. si-ITGA2 siRNA sequence information: si-ITGA2-1: 5^′-UUCUCAGGAAAGCCACUCCAGGGUG-3^′ (Sense Sequence), 5^′-UUCUCAGCAGAAGGCCACUCCAG-3^′ (Antisense Sequence); si-ITGA2-2, 5^′-GAGCCUUUAGGAGCCAUCUCCAGUA-3^′, 5^′-UACUGGAGAUGGCUCCUAAAGGCUC-3^′ (Antisense Sequence); si-NC (non-targeting control) siRNA sequences: 5^′-CACAGGUGGUGUUUCGUCCACCGAA-3^′ (Sense Sequence), 5^′-UUCGGUGGACGAAACACCACCUGUG-3^′ (Antisense Sequence).

qRT-PCR was carried out as previously mentioned [22]. The relative expression of genes was determined using the 2^{-Δ⁢Δ⁢CT} technique, and the levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primer sequences utilized in the qRT-PCR tests are listed in Table 1.

To extract protein lysates from TC cells, a combination of protease inhibitor cocktail (CoWin Biosciences; cat. no. CW2200S; Nanjing, China) and phosphatase inhibitors cocktail (CoWin Biosciences; cat. no. CW2383S; Nanjing, China) was combined with RIPA lysis solution (Solarbio; cat. no. R0020; Beijing, China). Before the protein was separated using 10% SDS-PAGE and then transferred onto PVDF membranes (Beyotime; cat. no. FFP22; Beijing, China), the protein concentration was measured using a BCA Protein Assay Kit (Beyotime; cat. no. P0012; Beijing, China). Primary antibodies were then administered overnight at 4 °C after the membranes had been blocked for an hour at room temperature using 5% skim milk. These included antibodies to ITGA2 (cat. no. ab181548), BCL2-associated X protein (Bax) (cat. no. ab32503), caspase-3 (cat. no. ab32351), caspase-9 (cat. no. ab32539), BCL2 apoptosis regulator (Bcl-2) (cat. no. ab182858), poly(ADP-ribose) polymerase (PARP) (cat. no. ab191217), cleaved-PARP (cat. no. ab32561), GSDME-full length (GSDME-FL) (cat. no. ab215191), N-terminal fragment of GSDME (GSDME-N) (cat. no. ab222408), phosphorylated histone H2AX on serine 139 ($\gamma$-H2AX) (cat. no. ab81299), MYC proto-oncogene, bHLH transcription factor (C-myc) (cat. no. ab32072), CD44 molecule (CD44) (cat. no. ab51037), $\beta$-catenin (cat. no. ab32572), slug (cat. no. ab51772), snail (cat. no. ab216347), phosphorylated checkpoint kinase 2 (p-CHK2) (cat. no. ab85743), and phosphorylated ataxia telangiectasia mutated (p-ATM) (cat. no. ab36810) (Abcam; Cambridge, UK; 1:1000), as well as wingless-type MMTV integration site family, member 1 (Wnt-1) (Wuhan Sanying; cat. no. 27935-1-AP; Wuhan, China; 1:1000). Following membrane washing, the proper secondary antibodies (Kangcheng; cat. no. KC-RB-035; Shanghai, China; 1:5000) were added. The protein bands were seen using an enhanced chemiluminescence kit (Tiangen; cat. no. PA110; Beijing, China), and the data were recorded and analyzed using the ChemiDoc Imaging System (Bio-Rad, Shanghai, China). The loading control was GAPDH (Abcam; cat. no. ab8245; 1:1000).

The CCK-8 analysis method (KeyGEN; cat. no. KGA317; Nanjing, China) was used to assess cell viability. At a density of 5 $\times$ 10³ cells per well, TC cells were seeded onto 96-well plates. A microplate reader (Kehua Technologies, Shanghai, China) was used to measure the absorbance at 450 nm after the CCK-8 reagent was added to each well at 0, 1, 2, 3, 4, and 5 days.

Trypsin-EDTA (Absin; cat. no. abs42104714; Shanghai, China) was used to separate TC cells, and then washing with phosphate-buffered saline (PBS; Beyotime; cat. no. C0221B; Shanghai, China). As advised by the manufacturer, the cells were labeled with Annexin V (BD Biosciences; cat. no. 563973; San Jose, CA, USA) and propidium iodide (PI) (BD Biosciences; cat. no. 556547; San Jose, CA, USA). Viable, necrotic, and apoptotic cells were then separated using a flow cytometer (Jiyuan, Guangzhou, China). FlowJo software (version 7.6, FlowJo, Hangzhou, China) was utilized for processing the flow cytometry data.

Transwells were utilized to monitor cell invasion and migration. Transfected TC cells were placed in serum-free media in the top compartment of the Transwell instrument.

Subsequently, the medium in the lower compartment was enriched with 10% FBS. After a 24-hour incubation period, the cells were fixed with 4% paraformaldehyde (Wuhan Servicebio Biotechnology Co., Ltd.; cat. no. G1101; Wuhan, China) and stained with DAPI (Abcam; cat. no. ab104139; Cambridge, UK). The migration of cells into the bottom compartment was then assessed using inverted microscopy. The upper chamber was initially covered with matrigel for the cell invasion studies, and the same steps as for cell migration were taken.

A suitably diluted quantity of cell culture supernatant was then added to the wells of ELISA plates that had previously been coated with IL-1$\beta$ antibodies (Wuhan Sanying; cat. no. 16806-1-AP; Wuhan, China). The chromogenic substrate was added after the enzyme-linked secondary antibody had been incubated and cleaned. The reaction was subsequently terminated and absorbance at the specified wavelength was measured using a microplate reader. The absorbance results were compared to a standard curve created with known quantities of IL-1$\beta$ in order to ascertain the amount of IL-1$\beta$ present in each sample.

To create the xenograft model, TC cells transfected with si-NC and si-ITGA2-1 were suspended in serum-free media. Subcutaneous injections of a 100 µL solution containing 5 $\times$ 10⁶ cells were then administered to the flanks of ten 6-week-old female BALB/c nude mice (n = 3 per group). Subsequently, calipers were used to measure the tumor dimensions every week for 28 days. The formula used to compute the tumor volume was 0.52 $\times$ length $\times$ width². An intraperitoneal dose of sodium pentobarbital (150 mg/kg, 50 mg/mL) was administered to euthanize the animals after 4 weeks. Tumors were then surgically excised, weighed, and analyzed histologically. All animal procedures were approved by the Animal Welfare and Ethics Committee of the Laboratory Animal Science Department, Fudan University (No. 2024-MHYY-304), and were conducted following institutional ethical guidelines and complied with the 3R (Replacement, Reduction, Refinement) principles.

Lactate dehydrogenase (LDH; cat. no. P0395S) activity and reactive oxygen species (ROS; cat. no. S0035) levels were assessed following the manufacturer’s instructions. The corresponding detection kits were used, respectively (Beyotime Institute of Biotechnology, Jiangsu, China).

After embedding the subcutaneous tumor tissues of mice in sections as required, the oven temperature was set to 65 °C and baked at a constant temperature for 120 mins. This was performed according to the immunohistochemistry kit (Maxim; cat. no. Kit9710; Fuzhou, China). Sections were incubated with an anti-ITGA2 primary antibody (Abcam, cat. no. ab181548; 1:500 dilution) overnight at 4 °C, followed by detection using the HRP-conjugated secondary antibody (Kangcheng; cat. no. KC-RB-035; 1:250 dilution, Shanghai, China). The complete tissue sample was scanned at low magnification ($\times$10) to evaluate staining, and it was verified at high magnification ($\times$20).

Data was analyzed using the R program (version 4.2.2, R Foundation for Statistical Computing, Vienna, Austria). Three duplicates of each experiment were conducted, and the mean $\pm$ standard deviation (SD) was used to report the data. Tukey’s post-hoc test was used for pairwise comparisons after one-way ANOVA was used for multi-group comparisons. Statistical significance was defined as a p-value of less than 0.05.

The number of up-regulated DEGs found in the TCGA-THCA, GSE3678, GSE29265 and GSE33630 datasets was 557, 210, 247 and 430, respectively, while the number of down-regulated DEGs was 746, 209, 295 and 404, respectively (Fig. 1A–D). An integrated bioinformatics platform found that 95 up-regulated DEGs as well as 130 down-regulated DEGs overlapped across all four datasets (Fig. 1E,F). Further analysis of these overlapping DEGs was performed by constructing a PPI network. This identified 14 genes using the MCODE plugin (Fig. 1G).

Fig. 1.

Construction of the PPI network for DEGs identified in the TC datasets. (A–D) Volcano plots showing DEGs in TCGA-THCA (A), GSE3678 (B), GSE29265 (C), and GSE33630 (D) datasets. Red dots represent up-regulated DEGs, and blue dots represent down-regulated DEGs. Grey dots indicate non-significant genes. A p-value of 0.05 was the threshold for statistical significance. (E,F) Venn diagram showing the overlap of up-regulated DEGs (E) and down-regulated DEGs (F) in the four datasets. The number in each segment indicates the number of shared genes. (G) The MCODE algorithm revealed the PPI network of overlapping genes, which contained 14 nodes and 23 edges. Node size corresponds to degree of interaction, and edge thickness represents interaction confidence. PPI, protein-protein interaction; TCGA, The Cancer Genome Atlas; DEGs, differentially expressed genes; THCA, thyroid carcinoma; MCODE, Molecular Complex Detection; TC, thyroid carcinoma.

As shown in Fig. 2A,B, three critical prognostic genes were identified from the 14 genes based on the ideal lambda value ($\lambda$min = 0.014), giving the following risk score: (–0.1389) $\times$ EFEMP1 + (–0.1195) $\times$ EGR2 + (–0.2028) $\times$ ITGA2. In terms of mortality and survival, the high-risk group did better than the low-risk group (Fig. 2C). Survival analysis of these 3 genes using the KM method showed a significantly lower OS in the high-risk group, with an HR of 6.078 ($>$1) (Fig. 2D). Additionally, ROC curve analysis showed the predictive ability of this model was strongest at one year of follow-up, with an area under the curve (AUC) of 0.755 (Fig. 2E).

Fig. 2.

LASSO Cox regression and survival analysis for three key prognostic genes in TC. (A) LASSO coefficient profiles of 14 candidate genes. Each curve represents the relationship between the gene coefficient and the L1 norm. The optimal lambda value was determined by minimizing the partial likelihood deviation. (B) Plot of the correlation between log($\lambda$) and 10-fold cross-validated partial likelihood deviation for selecting the best $\lambda$ value. (C) Risk model analysis of the selected sample data. The upper panel displays the risk scores, the middle panel shows survival status (red dots indicate high-risk, blue dots indicate low-risk), and the lower panel is a heatmap of the clustering distribution of the three critical genes (ITGA2, EGR2, EFEMP1). (D) KM survival curves for the two risk-model-identified groups. The log-rank test was used to assess differences in survival probability, with a p-value of 0.000866 and a HR of 6.078 for high-risk groups compared to low-risk groups. (E) ROC curve for the patient risk model after 1, 3, and 5 years of follow-up. The real positive score is shown on the vertical axis, while the false positive score is shown on the horizontal axis. The AUC values are provided for each time point, indicating the model’s predictive accuracy. TC, thyroid carcinoma; LASSO, Least Absolute Shrinkage and Selection Operator; DEGs, differentially expressed genes; ROC, Receiver Operating Characteristic; KM, Kaplan-Meier; AUC, Area Under the Curve; HR, Hazard Ratio; CI, Confidence Interval.

Cox regression analysis and survival analysis of key prognostic factors in TC patients. The results of univariate Cox regression analysis for the 3 key candidate genes (EFEMP1, EGR2, and ITGA2) showed significant prognostic correlations, indicating that the expression levels of these genes were significantly correlated with the survival prognosis of patients (Fig. 3A). After adjusting for potential confounders, the results of multivariate Cox regression analysis for the 3 key candidate genes indicated that the ITGA2 gene still showed a significant prognostic correlation, suggesting that ITGA2 may be an independent prognostic factor (Fig. 3B). The column-line graph demonstrated a prognostic model constructed based on ITGA2 expression levels and other clinicopathologic features (Fig. 3C). The analysis showed that the predictive ability of the model was statistically significant. The calibration curves further validated the predictive accuracy of the model, showing good agreement between the predicted 1-, 3-, and 5-year survival probabilities and the actual observations (Fig. 3D). Notably, most of the actual observations in Fig. 3D are clustered near the upper-right corner, which reflects the overall high survival probability in the TC cohort. The expression level of the ITGA2 gene was significantly correlated with the prognosis of patients with thyroid cancer.

Fig. 3.

Survival analysis and prognostic assessment in TC. (A) Univariate Cox regression analysis of clinical prognostic factors and critical genes for TC patient survival. The forest plot displays HRs with 95% CI and p-values for each variable. (B) Multivariate Cox regression analysis of the same factors, adjusting for potential confounders. The forest plot shows adjusted HRs with 95% CI and p-values. (C) Nomogram for predicting 1-year, 3-year, and 5-year survival probabilities based on key prognostic factors (ITGA2, Age, Gender, pTNM-stage). The total points for each factor are summed to obtain a linear predictor, which is then translated into survival probabilities. The model’s predictive accuracy is assessed using the C-index, with a value of 0.635 (95% CI: 0.535–1) and a p-value of 0.008. (D) Calibration plot for the nomogram, showing observed versus predicted survival probabilities at 1-year, 3-year, and 5-year follow-ups. The dashed lines represent ideal prediction, while the solid lines show the actual performance of the model. The calibration plot was generated using the Clinical Bioinformatics Platform. The clustering of observed points in the upper-right region reflects the high survival probability in our TC cohort. TC, Thyroid Carcinoma; Cox, Cox proportional hazards model; HR, Hazard Ratio; CI, Confidence Interval; C-index, Concordance Index.

In this study, we analyzed the expression pattern of the ITGA2 gene in thyroid cancer and its potential as a diagnostic marker. The volcano map show that the expression level of ITGA2 was significantly up-regulated in tumor samples compared to normal samples in the TCGA database (Fig. 4A). In addition, box line plots revealed that the expression levels of ITGA2 were significantly higher in tumor samples than in normal samples in the GSE3678, GSE29265, and GSE33630 datasets, especially in the GSE29265 and GSE33630 datasets (Fig. 4B–D). Further, ROC curve analysis was performed to evaluate the potential of ITGA2 gene expression as a diagnostic marker (Fig. 4E–H). The results showed that the AUC value of ITGA2 was as high as 0.887 in the TCGA dataset, indicating a very high diagnostic accuracy. In the GSE3678 and GSE33630 datasets, the AUC values of ITGA2 were 0.898 and 0.954, respectively, which also showed good diagnostic performance. The AUC value of ITGA2 in the GSE29265 dataset was 0.907, further confirming its potential as a diagnostic marker. We analyzed the effect of ITGA2 gene expression on disease-free survival in patients using the TCGA dataset. Analysis of the Clinical BioLetter House platform (n = 512) showed that high ITGA2 expression was significantly associated with lower disease-free survival (Fig. 4I), a finding supported by results from the GEPIA database (n = 500) (Supplementary Fig. 2A). However, no significant association between ITGA2 expression and disease-free survival was found when the TCGA dataset (n = 497) was analyzed by the cBioPortal platform (Supplementary Fig. 2B). Different analytical methods or platform differences may cause this discrepancy.

Fig. 4.

Analysis of ITGA2 expression and diagnostic performance. (A–D) ITGA2 expression level in tumor and control samples from the TCGA-THCA, GSE3678, GSE29265, and GSE33630 datasets. Red dots represent tumor samples, and blue dots represent normal samples. *p 0.05, ****p 0.001. (E–H) Diagnostic performance of ITGA2 expression as assessed by ROC curves across the same datasets. (I) KM survival curves reveal a significant association between ITGA2 gene expression levels and disease-free survival in thyroid cancer patients. The red solid line indicates the high expression group and the blue dashed line indicates the low expression group. p-value was 0.0155; HR was 2.896, 95% CI 1.225 to 6.85. TCGA-THCA, The Cancer Genome Atlas Thyroid Carcinoma; GSE3678, GSE29265, GSE33630, Gene Expression Omnibus dataset identifiers; AUC, Area Under the Curve; ROC, Receiver Operating Characteristic; KM, Kaplan-Meier; HR, Hazard Ratio; CI, confidence intervals.

qRT-PCR and WB methods were used to investigate ITGA2 expression in TC cell lines TPC-1 and CAL-62, and normal thyroid cells (Nthy-ori3-1). Significant up-regulation of ITGA2 expression was observed in TC cell lines (Fig. 5A–C). Further, qRT-PCR and WB assays confirmed the efficient knockdown of ITGA2 by si-ITGA2-1 in TC cell lines (Fig. 5D–F). Cell viability declined significantly following ITGA2 knockdown in TC cell lines, as shown by the CCK-8 assay (Fig. 5G,H). Therefore, changes in ITGA2 expression were associated with altered biological behavior in TC cells, further supporting the important role of this key gene.

Fig. 5.

Validation of ITGA2 knockdown and its impact on TC cell viability. (A–C) Relative mRNA (A) and protein (B,C) expression levels of ITGA2 in Nthy-ori3-1, TPC-1, and CAL-62 cell lines. Data are normalized to GAPDH and expressed as fold change relative to control. (D–F) Confirmation of si-ITGA2-1 knockdown efficacy in TC cells via qRT-PCR and WB. Results were normalized to GAPDH and presented as fold changes relative to the si-NC (n = 3 for each group). (G,H) CCK-8 demonstrated reduced cell proliferation over a 5-day period in TC cells following ITGA2 knockdown. The absorbance at 450 nm (OD value) was measured to assess cell viability. Three independent experiments present Data as mean $\pm$ SD (n = 3). qRT-PCR, quantitative Real-Time Polymerase Chain Reaction; WB, Western Blot; si-NC, non-targeting control siRNA; CCK-8, Cell Counting Kit-8; OD, Optical Density. “*” indicates p 0.05 compared to the si-NC.

Transwell tests demonstrated that ITGA2 knockdown dramatically decreased the TC cells’ migration and invasion capacity (Fig. 6A,B). Flow cytometry analysis also demonstrated a marked increase in the apoptosis rate of TC cells with silenced ITGA2 (Fig. 6C,D). qRT-PCR analysis showed that ITGA2 knockdown in TPC-1 cells led to up-regulation of the pro-apoptotic genes Caspase-3, Bax, and Caspase-9 and down-regulation of the anti-apoptotic gene Bcl-2 (Fig. 6E). This was further confirmed at the protein level by WB analysis. Cells with ITGA2 knockdown showed significantly reduced expression of anti-apoptotic proteins relative to the control group but increased expression of a pro-apoptotic marker (Fig. 6G,I). In CAL-62 cells, ITGA2 knockdown resulted in a significant up-regulation of Caspase-3 and Bax expression, whereas the changes in Bcl-2 and Caspase-9 expression were not significant compared with the control (Fig. 6F,H,J). This suggests that in CAL-62 cells, ITGA2 knockdown may promote apoptosis mainly by affecting Caspase-3 and Bax, but not through the Bcl-2 and Caspase-9 pathways. We hypothesize that this difference may reflect the diversity of responses to ITGA2 knockdown in different thyroid cancer cell lines or suggest that CAL-62 cells may be more dependent on the Caspase-3 and Bax-mediated apoptotic pathways.

Fig. 6.

Impact of ITGA2 knockdown on cell behavior in TC cell. (A,B) Transwell test data demonstrating the effect of ITGA2 knockdown on the migratory and invasion capacities of TC cells (n = 3 per group). The bar graph represents the average number of cells that migrated or invaded through the membrane. Scale bar, 50 µm. (C,D) Flow cytometry analysis of apoptosis rates in TC cells following ITGA2 knockdown (n = 3 per group). The left panel shows flow cytometry plots, and the right panel presents a quantitative analysis of the results. (E–J) qRT-PCR and WB were used to detect changes in mRNA and protein expression of apoptosis-related genes (Bax, Bcl-2, caspase3, and caspase9) in TPC-1 cells after knockdown of ITGA2, and changes in mRNA and protein expression of Bax and caspase3 in CAL-62 cells (n = 3 per group). GAPDH was used as a loading control and was expressed as fold change relative to si-NC. si-NC, non-targeting control siRNA; qRT-PCR, quantitative Real-Time Polymerase Chain Reaction; WB, Western Blot. “*”, “**”, and “***” indicate p 0.05, p 0.01, and p 0.001, respectively, compared to the si-NC.

WB analysis showed that the expression of proteins related to apoptosis, pro-inflammatory cytokines and pyroptosis was significantly changed in TC cell lines after ITGA2 knockdown. Specifically, the full-length (FL) forms of PARP and GSDME were reduced. In contrast, the cleaved forms (cleaved-PARP and GSDME-N), as well as the pro-inflammatory cytokine IL-18, were significantly up-regulated in the si-ITGA2-1 group as compared to the control group (Fig. 7A–C). In addition, LDH activity in TC cells was notably increased after ITGA2 knockdown, indicating increased cytotoxicity (Fig. 7D). ELISA assay also showed that IL-1$\beta$ levels were significantly higher in si-ITGA2-1-treated cells compared to the control group (Fig. 7E). These results indicate that ITGA2 knockdown in TC cells induces pyroptosis and promotes inflammatory responses.

Fig. 7.

ITGA2 knockdown induces apoptosis and pyroptosis in TC cells. (A–C) qPCR and WB were performed to detect changes in mRNA and protein expression levels of apoptotic proteins (PARP and cleaved-PARP), pro-inflammatory cytokines IL-18 and pyroptotic proteins (GSDME-FL and GSDME-N) in TC cells after knockdown of ITGA2 (n = 3 per group). GAPDH was used as a loading control. (D) LDH activity assay showing changes in cell membrane integrity and cytotoxicity in TC cells after ITGA2 knockdown. The bar graph represents the mean LDH level (n = 3 per group). (E) ELISA measuring IL-1$\beta$ secretion from TC cells after ITGA2 knockdown. The bar graph shows the mean IL-1$\beta$ concentration (n = 3 per group), reflecting the inflammatory response under different treatment conditions. si-NC, non-targeting control siRNA; WB, Western Blot; LDH, Lactate dehydrogenase; ELISA, Enzyme-linked immunosorbent assay. “*”, “**”, and “***” indicate p 0.05, p 0.01, and p 0.001, respectively, compared to the si-NC.

WB analysis revealed that ITGA2 knockdown significantly upregulated the expression of key DNA damage response markers ($\gamma$-H2AX, p-CHK2, and p-ATM) compared to the si-NC control (Fig. 8A–C). Additionally, ITGA2 knockdown caused ROS to increase significantly in TC cell lines, as measured by the fluorescence intensity (Fig. 8D). These findings indicate that ITGA2 knockdown induces DNA damage and oxidative stress in TC cells.

Fig. 8.

ITGA2 knockdown induces DNA damage and ROS production in TC cells. (A–C) WB analysis of DNA damage-related protein expression (phosphorylated ataxia telangiectasia mutated (p-ATM), phosphorylated histone H2AX on serine 139 ($\gamma$-H2AX), and phosphorylated checkpoint kinase 2 (p-CHK2)) in TC cells after ITGA2 knockdown (n = 3 per group). GAPDH was used as a loading control and presented as fold-changes compared to the si-NC. (D) Measurement of ROS levels in TC cells following ITGA2 knockdown using a ROS detection kit (n = 3 per group). The bar graph represents the mean fluorescence intensity, indicating ROS production under different treatment conditions. si-NC, non-targeting control siRNA; WB, Western Blot; ROS, reactive oxygen species. “*” indicates p 0.05 compared to the si-NC.

qRT-PCR analysis revealed that ITGA2 knockdown in TC cells markedly decreased the expression of genes associated with the Wnt/$\beta$-catenin signaling pathway, such as snail, c-Myc, CD44, Wnt-1, slug, and $\beta$-catenin (Fig. 9A,B). WB assays further confirmed the down-regulation of these genes at the protein level (Fig. 9C–E), suggesting that ITGA2 knockdown inhibits stemness in TC cells by impeding the Wnt/$\beta$-catenin signaling pathway. In vivo, tumorigenesis assays were performed by injecting si-ITGA2-1-transfected TPC-1 cells into the flanks of mice. After four weeks, the ITGA2 knockdown group’s tumor weight and volume were noticeably lower than those of the control group (Fig. 10A–C). ITGA2 is localized in the cytoplasm, the, immunohistochemical analysis further confirmed that ITGA2 expression was down-regulated in tumors from the ITGA2 knockdown group (Fig. 10D).

Fig. 9.

Knockdown of ITGA2 inhibits TC stemness by suppressing the Wnt/$\beta$-catenin signaling pathway. (A–E) qRT-PCR and WB analysis of mRNA and protein expression levels of $\beta$-catenin, Wnt-1, Snail, C-myc, CD44, and Slug in TC cells after ITGA2 knockdown (n = 3 per group). GAPDH was used as a loading control, presented as fold-changes relative to the si-NC. si-NC, non-targeting control siRNA; qRT-PCR, quantitative Real-Time Polymerase Chain Reaction; WB, Western Blot. “*”, “**”, and “***” indicate p 0.05, p 0.01, and p 0.001, respectively, compared to the si-NC.

Fig. 10.

ITGA2 knockdown reduced tumor growth in vivo. (A) Representative images of tumors harvested from nude mice bearing TPC-1 cells transfected with either si-NC or si-ITGA2-1 (n = 3 per group). (B) Bar graph showing tumor weights in mice injected with TPC-1 cells transfected with si-NC or si-ITGA2-1 (n = 5 per group). (C) Bar graph displaying tumor volumes in mice injected with TPC-1 cells transfected with si-NC or si-ITGA2-1 (n = 5 per group). (D) IHC analysis of ITGA2 expression in tumor sections from the in vivo tumorigenesis assay (n = 5 per group). Scale bar, 50 µm. si-NC, non-targeting control siRNA; IHC, Immunohistochemistry; Nude mice, immunodeficient mice. “*” indicates p 0.05 compared to the si-NC.