Interplay between acute Type A aortic dissection and pan-cancer: Clinical evidence, bioinformatics, and experimental validation

Summary

Acute Type A Aortic Dissection (ATAAD) and cancer represent major global health burdens, yet shared mechanisms remain elusive. Epidemiological analyses linked ATAAD to increased risks of lung adenocarcinoma (LUAD) and head and neck squamous cell carcinoma (HNSCC). Integrated multi-omics identified mitochondrial peptidylprolyl isomerase F (PPIF) as a key ATAAD biomarker, specifically overexpressed in monocytes, with experimentally validated. Pan-cancer analyses revealed PPIF overexpression correlated with poor prognosis across malignancies, and functional assays also confirmed PPIF drives proliferation, migration, and invasion in LUAD and HNSCC. These findings nominate PPIF as a shared mechanistic node connecting ATAAD pathogenesis to cancer progression, highlighting its potential as a therapeutic target worth further investigation.

Graphical abstract

Highlights

• •Aortic dissection linked to LUAD/HNSCC risk
• •PPIF is a key mitochondrial biomarker for ATAAD, highly expressed in monocytes
• •PPIF is also overexpressed in multiple cancer types
• •PPIF promotes proliferation, migration, and invasion in LUAD and HNSCC cells

Health sciences; Medicine; Medical specialty; Cardiovascular medicine; Oncology

Introduction

Cardiovascular disease (CVD) and cancer are the two leading causes of death worldwide. Many shared factors, such as obesity and diabetes mellitus (DM), are reported. Mechanistically, CVD and cancer also share common biological mechanisms, such as inflammation, oxidative stress, hormones, cytokines et al.¹ Kathleen et al. conducted a population-based study on CVD and mortality risk in patients with cancer. Their results showed that young patients with cancer, especially in the first year after diagnosis, have a significantly elevated risk of death from aortic aneurysm and dissection. As patients age and the follow-up period extends, the risk decreases but remains higher than that of the general population.² Thus, it is very urgent to monitor and manage aortic dissection risks, particularly in younger cancer survivors and during the early post-diagnosis period.

Metabolic remodeling plays a significant role in both CVD and cancer.³ Recently, several studies have revealed the correlation between mitochondria and aortic dissection. Mie et al. found that human aortic smooth muscle cells (HAoSMCs) switch from contractile phenotype to synthetic phenotype under NOTCH1 deficiency, with the impairment of mitochondrial fusion, implying a potential target for aortic dissection.⁴ Luo et al. discovered that tumor necrosis factor (TNF) enhances AP-1 transcriptional activity via inhibiting mitochondrial oxidative phosphorylation (OXPHOS), triggering SMCs from contractile phenotype to synthetic counterparts, inducing aortic dissection finally.⁵ Furthermore, mitochondria play a pivotal role in cancer progression by supporting the metabolic reprogramming necessary for tumor growth.

Contrary to the long-held belief that mitochondrial metabolism is irrelevant in rapidly proliferating cancer cells, recent studies demonstrate its essential function in cancer. Mitochondria provide cancer cells with key metabolites for macromolecule synthesis, such as nucleotides, lipids, and amino acids, which are crucial for cell proliferation. Moreover, they generate oncometabolites that drive tumor growth. The flexibility of mitochondrial metabolism enables cancer cells to adapt throughout various stages of tumorigenesis, including metastasis. Targeting mitochondrial functions, such as the electron transport chain or the tricarboxylic acid (TCA) cycle, has emerged as a promising therapeutic strategy, with ongoing clinical trials testing inhibitors aimed at disrupting mitochondrial activity. These findings highlight mitochondria as a critical player in sustaining cancer cell survival and proliferation, making them a key target for cancer therapies.⁶ Herein, we focused on MitoDEGs that both correlated with Acute Type A Aortic Dissection (ATAAD) and cancer. To enhance the specific causality between ATAAD and cancer, we also conducted a prospective cohort study using the UK Biobank and Mendelian randomization (MR). This study may provide a therapeutic target for ATAAD and cancer, offering insights into the potential link between these diseases.

Results

Aortic aneurysms and dissections are linked with the development of certain cancers in the UK Biobank database

We conducted a prospective cohort study using the UK Biobank to examine the association between aortic aneurysms and dissections and cancers at various sites. After excluding participants with cancer at baseline and imputing missing data, we conducted a series of analyses, including Cox proportional hazards regression and Kaplan-Meier survival curve plotting. A two-sided p < 0.05 was considered statistically significant. Specific processing procedures, exposure factors, outcomes, and covariates were detailed in Figure 1A. We used Cox proportional hazards regression models to conduct multivariate analyses on the impact of aortic aneurysms and dissections on multiple cancers. The results indicated that patients with aortic aneurysms and dissections had a significantly higher overall cancer risk compared to those without these conditions (HR [95% CI] 1.14 [1.07–1.21], p < 0.001). Focusing on specific cancer types, we observed an increased risk for bladder cancer (HR [95% CI] 1.67 [1.22–2.29], p = 0.001), head and neck cancer (HR [95% CI] 1.56 [1.05–2.32], p = 0.026), kidney cancer (HR [95% CI]1.54 [1.12–2.12], p = 0.008), and lung cancer (HR[95% CI] 1.53 [1.30–1.80], p < 0.001) (Figure 1B). We selected lung cancer and head and neck cancer to plot Kaplan-Meier curves for patients with diagnoses of aortic aneurysms and dissections versus normal participants (Figures 1C and 1D). The results revealed that participants with disease developed a higher proportion of cancers and had poorer survival (log rank test p < 0.05). These data suggested an association between aortic aneurysms and dissections and the development of cancer, particularly lung and head and neck cancers.

Figure 1.

Participants with aortic aneurysms and dissections experienced a higher incidence of specific tumorigenesis

(A) Flowchart of data processing and analysis in the UK Biobank database.

(B) Multivariable-adjusted HRs (95% CIs) for cancer outcomes by aortic aneurysms and dissections status.

(C) Kaplan–Meier survival curves and log rank test of UK Biobank participants stratified by aortic aneurysms and dissections status for lung cancer endpoint.

(D) Kaplan–Meier survival curves and log rank test of UK Biobank participants stratified by aortic aneurysms and dissections status for head and neck cancer endpoint.

Mendelian randomization uncovered the causal effect of acute type A aortic dissection on lung adenocarcinoma and head and neck squamous cell carcinoma

In the MR analysis between ATAAD and LUAD, the IVW method indicated a significant causal effect between ATAAD and an increased risk of LUAD (OR [95% CI] 1.59 [1.54, 1.63], p = 0.012) (Figures 2A and 2C), with 11 SNPs adopted. The other four methods, including MR-Egger regression, Weighted mode, Simple mode, and Weighted median, showed no significant association between ATAAD and the risk of LUAD (p > 0.05). The results of Cochran’s Q-test in MR-Egger (Q = 9.310, p = 0.409) or IVW (Q = 9.354, p = 0.498) and funnel plots revealed no heterogeneity (Figure 2D). Additionally, we found no evidence of horizontal pleiotropy between instrument variables using the MR-Egger regression intercept analysis (p > 0.05) (Table S1). The leave-one-out analysis confirmed that the causal association was not influenced by any single genetic variant, supporting the robustness of the IVW results (Figure 2B). In total, all the results indicated a positive association between ATAAD and the risk of LUAD.

Figure 2.

Mendelian randomization analysis reveals the causal relationship between aortic dissection and lung adenocarcinoma risk

(A) Forest plot illustrates the combined estimates from all single-nucleotide polymorphisms (SNPs). Red dots represent the overall estimate, and horizontal lines indicate the 95% confidence intervals.

(B) Leave-one-out analysis. Black points represent the causal effect excluding a single specific variant from the analysis, with the red point showing the combined estimate using all SNPs.

(C) Scatterplot displays the estimated effects of each Mendelian randomization method, with the slope of each line representing the causal effect.

(D) Funnel plot. Vertical lines represent estimates with all SNPs. The symmetry of the funnel suggests no significant horizontal pleiotropy.

Similarly, the causality between ATAAD and an elevated risk of HNSCC was also observed by IVW method (OR [95% CI] 1.46 [1.14, 1.85], p = 0.002) and Weighted median (OR [95% CI] 1.43 [1.09, 1.88], p = 0.009) (Figures S2A and S2C), with 3 SNPs adopted. Notably, based on the calculation formula F = β²/SE,⁷ the F-statistics for rs118055578, rs2302688, and rs36029774 are 23.795, 21.742, and 47.159, respectively. All these values exceed 10, which confirms their robustness with the exposure and helps reduce the potential risk of weak instrument bias. The other three methods, including MR-Egger regression, Weighted mode and Simple mode, exhibited no obivous correlation between ATAAD and the risk of HNSCC (p > 0.05). The heterogeneity was also valued by Cochran’s Q-test in MR-Egger (Q = 0.571, p = 0.449) and IVW (Q = 0.692, p = 0.707) (Figure S2D). The p value for the pleiotropy test for MR-Egger was not statistically significant, indicating no evidence of pleiotropy (p > 0.05) (Table S2). The leave-one-out analysis demonstrated that no single genetic variant had an impact on the causal relationship, further confirming the previous findings (Figure S2B). Collectively, all the results pointed to a positive correlation between ATAAD and an increased risk of HNSCC. Herein, we can safely draw the conclusion that ATAAD could increase the risk of LUAD and HNSCC.

Identification of MitoDEGs in acute type A aortic dissection and its functional enrichment analysis

Three gene expression datasets about ATAAD, including GSE52093, GSE98770, and GSE190635 from the GEO database were used to identify DEGs via the limma package. After intersecting ATAAD DEGs with mitochondrial-related genes, MitoDEGs were shown as a volcano plot, based on the threshold of p value <0.05 and |log FC| > 0.5 (Figure 3A). There were 58 upregulated MitoDEGs and 61 downregulated MitoDEGs in total. The distribution of MitoDEGs was shown as a heatmap in the ATAAD and normal group (Figure 3B). Figure 3C clearly illustrated the expression difference of MitoDEGs between the ATAAD and the normal group. Moreover, to detect the hidden patterns among variables, a correlation heatmap illustrated the Pearson correlation between MitoDEGs (Figure 3D).

Figure 3.

The volcano plot, heatmap, boxplot, and corrplot of mitochondrial-related differentially expressed genes (DEGs) in ATAAD

(A) Volcano plot depicts mitochondrial-related DEGs in ATAAD and healthy samples.

(B) Heatmap illustrates mitochondrial-related DEGs.

(C) Boxplot shows the 119 mitochondrial-related DEGs.

(D) Pearson correlation analysis for the 119 mitochondrial-related DEGs.

In order to investigate the biological function of MitoDEGs, we conducted GO, KEGG, and GSEA enrichment analysis. According to KEGG analysis, MitoDEGs were dominantly enriched in “Carbon metabolism,” “Biosynthesis of amino acids,” and “Lysine degradation” (Figure 4A). Furthermore, based on GO analysis, MitoDEGs were enriched in “small molecule catabolic process,” “organic acid catabolic process” and “mitochondrial gene expression” (BP); “mitochondrial matrix,” “mitochondrial inner membrane” and “mitochondrial protein-containing complex” (CC); “vitamin binding,” “oxidoreductase activity, acting on the aldehyde or oxo group of donors” and “oxidoreductase activity, acting on the aldehyde or oxo group of donors, NAD or NADP as acceptor” (MF) (Figure 4B). These results clearly showed that the biological function is mainly enriched in mitochondrial-related biological processes. We also conducted GSEA analysis, the analysis reveals the significant enrichment of “catabolic processes,” “mitochondrial gene expression,” and “mitochondrial translation” in the dataset, hinting at the significance of mitochondria in ATAAD again (Figure 4C). The chordal graph further illustrated the correlation between genes and biological functions. Of note, gene PPIF is closely related to “positive regulation of mitochondrion organization” and “apoptotic mitochondrial changes,” denoting that PPIF may play an important role in regulating mitochondrial structure, function, and apoptosis. GLDC is related to “carboxylic acid catabolic process” and “small molecule catabolic process” (Figure 4D).

Figure 4.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, Gene Ontology (GO) enrichment analysis, and GSEA analysis of 119 mitochondrial-related DEGs in ATAAD

(A) KEGG pathway enrichment analysis.

(B) GO analysis focuses on Biological Process (BP), Cellular Components (CC), and Molecular Function (MF).

(C) GSEA plot for the top 8 GO terms.

(D) Chordal graph.

ScRNA-seq analysis of acute type A aortic dissection

Further, in order to elucidate the changes in ATAAD at the single-cell transcriptional level and to identify the cell types in which PPIF is highly expressed, we performed an integrated analysis based on the published scRNA-seq data GSE222318. Based on the expression of canonical marker genes, dimensionality reduction using uniform manifold approximation and projection (UMAP) identified 11 distinct cell clusters (Figures 5A and 5B). To explore differences in the gene expression of ATAAD, as shown in Figure 5C, genes including CXCL5 and CCL20 are highly expressed in the ATAAD group, which were reported as mediators in recruiting inflammatory responses in previous literature.^8,9,10,11 Moreover, upregulated genes were enriched in pathways such as “Inflammatory Response,” “TNF-α singalling Via NF-κB” and “Chemokine Signaling Pathway” (Figure 5D). Among all the upregulated genes, PPIF was also upregulated in the ATAAD group (p < 2.22e-16) (Figure 5E). To further explore the main cell types driving the progression of ATAAD, we found that the proportions of monocytes and macrophages are significantly increased in the ATAAD group, while stromal cells, T cells, and NK cells are predominantly elevated in the normal group (Figure 5F). M1 and M2 gene set scoring revealed that,¹² higher M1 scores in monocytes and cDCs suggest a pro-inflammatory role, while elevated M2 scores in macrophages indicate an anti-inflammatory and tissue repair function in ATAAD (Figure 5G).¹³ Macrophages exhibit higher M2 scores; this could be related to the origin of macrophages. For example, macrophages derived from monocytes may lean toward the M1 type, while resident macrophages in tissues may lean toward the M2 type. cDCs (conventional dendritic cells) have higher M1 scores, which may be related to their role in antigen presentation during the early stages of inflammation, in particular, the onset of ATAAD. After sensing pathogens or damage signals, cDCs can activate and migrate to lymph nodes where they present antigens to T cells, initiating an adaptive immune response.¹⁴ Monocytes play a crucial role in driving ATAAD progression. Further differential gene and enrichment analysis of monocytes revealed that monocytes in ATAAD also express the PPIF gene (Figure 5H). They are enriched in pathways such as “Cytokine-Cytokine Receptor Interaction” and “TNF-α Signaling via NF-κB” (Figure 5I), with PPIF being specifically highly expressed (p < 2.22e-16) (Figure 5J). This suggested that monocytes are a key effector cell population functioning in ATAAD, and PPIF may play a significant role in the recruitment of monocytes and the regulation of their functions. Additionally, using CellChat, we explored ligand-receptor interactions between monocytes and various cells. We found that, in ATAAD, monocytes sent a higher frequency and intensity of interactions to B/Plasma cells, cDCs, NK cells, proliferating cells, stromal cells, and T cells, while receiving more signals from proliferating cells, NK cells, stromal cells, and T cells (Figure 5K). During that process, the MIF/CD74+ axis plays an important role (Figure 5L). These results indicate that, in ATAAD, monocytes act as cellular hubs that amplify vascular inflammation through dysregulated crosstalk: (1) They may license adaptive immunity by activating B cells and cDCs, driving autoantibody production and antigen presentation; (2) Bidirectional checkpoint signaling, especially with T cells, creates an immunosuppressive storm that exacerbates tissue damage and impedes repair. This multicellular network transforms local vascular injury into systemic inflammation, revealing targetable communication axes.

Figure 5.

Single-cell RNA-seq data uncovered the PPIF-high expressed cell types and the potential mechanisms of ATAAD

(A) UMAP plot illustrates the distribution of cells from ATAAD and normal tissues.

(B) Dotplot of marker genes show high expression in each major cell type. The dot size indicates the percentage of cells expressing the markers, while the color reflects the average expression level of the markers within each cell type.

(C) Volcano plot displays the DEGs across all cell types between ATAAD and normal samples.

(D) Gene Set Enrichment Analysis (GSEA) highlighting distinct enriched pathways across all cell types in ATAAD versus normal samples. Bar chart shows normalized enrichment scores (NES) for specific pathways.

(E) Violin plot depicts PPIF expression differences between ATAAD and normal groups across all cell types.

(F) Proportion of different major cell types in ATAAD versus normal samples.

(G) Violin plot shows the M1 scores and M2 scores in major cell types.

(H) Volcano plot shows DEGs in monocytes between ATAAD and normal groups.

(I) GSEA analysis of monocyte-enriched pathways in ATAAD versus normal groups. Bar chart illustrates normalized enrichment scores (NES) for specific pathways.

(J) Violin plot compares PPIF expression in monocytes between ATAAD and normal groups.

(K) Network of cell-cell communication; Number of interactions (top); Strength of interactions (bottom).

(L) The dot plots show outgoing communication patterns and incoming communication patterns of monocytes and other cell types.

Identification and verification of hub-MitoDEG peptidylprolyl isomerase F via the intersection of five machine learning methods, weighted gene co-expression network analysis, and scRNA-seq differentially expressed genes

Based on the above 119 MitoDEGs, in order to further select key genes capable of distinguishing ATAAD and normal samples, we utilized 5 machine learning methods. The LASSO logistic regression identified 10 MitoDEGs associated with Type A aortic dissection, with non-zero coefficients indicating significant features for classification, that is, PPIF, TBL1X, SLC9A1, CASQ1, NT5DC2, HSDL2, PRELID3A, SFXN5, TRMT2B, and GLDC, respectively. The optimal regularization parameter was selected using 10-fold cross-validation, ensuring model robustness and predictive accuracy. Of note, the best λ is 0.004502466 (Figure 6A). The SVM algorithm performs gene selection using Recursive Feature Elimination (RFE) with an SVM (Radial) model. The plot shows the relationship between the number of features and the RMSE (Root Mean Squared Error) during cross-validation. The point with the smallest RMSE represents the optimal number of features, which is then used for gene selection. Therefore, the SVM algorithm filtered out four biomarkers, namely PLN, PPM1B, PPIF, and CASQ1 (Figure 6B). The Decision Tree algorithm identified the top 8 important genes for classifying Type A aortic dissection, with importance scores visualized in a bar plot. They are PLN, PPM1B, PPIF, STOML2, TBL1X, SLC9A1, CASQ1 and NT5DC2 (Figure 6C). The Random Forest algorithm identified the top 10 important genes for classifying Type A aortic dissection, namely PLN, PPM1B, STOML2, NT5DC2, PPIF, ALDH2, TBL1X, CASQ1, SLC9A1 and PRELID3A. Gene importance was determined using the MeanDecreaseAccuracy metric (Figure 6D). The Boruta algorithm identified 37 key genes for Type A aortic dissection classification through multiple iterations. The plot visualizes the importance of features, with confirmed features marked (Figure 6E).

Figure 6.

Identification of key mitochondrial-related DEGs by 5 machine learning methods

(A) Screening of key mitochondrial-related DEGs via Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression.

(B) Key mitochondrial-related DEGs identified using the Support Vector Machine (SVM) algorithm.

(C) Decision Tree algorithm-based identification of key mitochondrial-related DEGs.

(D) Key mitochondrial-related DEGs identified through the Random Forest algorithm.

(E) Screening of key mitochondrial-related DEGs using the Boruta algorithm.

To reduce the number of key genes, we also used WGCNA to identify differential genes. After filtering out the aberrant samples, the recommended soft threshold power value is 7. Figure 7A showed the construction of the weighted gene co-expression network; the co-expression modules were displayed by hierarchical clustering and dynamic branch cutting. Figure 7B displayed the correlation between these identified modules and disease status. We explored that the ME blue and ME brown modules were the most highly correlated with clinical features. In the subsequent single-cell sequencing data analysis, we also obtained differential analysis results for all cell types between ATAAD and normal tissue. After intersecting the genes from 5 machine learning methods, WGCNA ME brown module, and differentially expressed genes (DEGs) from scRNA-seq data, we eventually obtained the final hub MitoDEG, PPIF, the biomarker of ATAAD (Figure 7C; Table S3). The ROC curve validated the diagnostic efficacy of PPIF in the original ATAAD datasets, namely GSE52093, GSE98770, and GSE190635. PPIF demonstrates perfect diagnostic efficacy for ATAAD: The AUC value reaches 0.992 (95% CI: 0.972–1), while achieving nearly 100% sensitivity and 94.1% specificity. This indicates that the biomarker comprehensively captures almost all true positive cases while effectively excluding 94.1% of true negative samples (extremely low misdiagnosis rate). Its overall diagnostic performance significantly outperforms conventional clinical biomarkers, demonstrating strong potential for translation into a first-line screening tool (Figure 7D).

Figure 7.

Weighted gene co-expression network analysis (WGCNA) of ATAAD and control

(A) Cluster dendrogram shows the highly connected genes in key modules of ATAAD.

(B) Association between modules and traits in ATAAD, with correlation coefficients and p-values provided in each cell.

(C) Venn diagram illustrates the key mitochondrial-related hub gene of ATAAD, identified by the intersection of five machine learning methods, WGCNA brown modules, and DEGs from scRNA-seq data across all cell types.

(D) ROC curve for the hub gene PPIF in ATAAD, based on the above GEO datasets.

Verification of peptidylprolyl isomerase F in acute type A aortic dissection through in vitro experiments

We further examined PPIF expression in ATAAD and normal aortic tissues. WB analysis revealed that PPIF expression in ATAAD tissues is higher than normal tissues (p < 0.01, Figures 8A and 8B). Figure 8C provides representative immunofluorescence images that corroborate the WB findings. The images clearly show a higher intensity of PPIF expression in ATAAD tissue samples (as seen in the red fluorescence) compared to normal tissue samples. These results collectively validate the upregulation of PPIF in ATAAD and suggest its involvement in the pathophysiology of the disease, warranting further investigation into its functional significance and potential as a therapeutic target of ATAAD.

Figure 8.

In vitro validation of the hub gene PPIF in ATAAD samples

(A and B) Western blot (WB) analysis reveals elevated PPIF expression in ATAAD tissue samples. Data are presented as mean ± SD, analyzed by unpaired t-test (n = 3). ∗∗p < 0.01.

(C) Representative immunofluorescence images show PPIF expression in ATAAD and normal tissue samples. The nuclei are stained with DAPI (blue), PPIF is shown in the red fluorescence, and the merge of the two images allows for the visualization of PPIF expression in the context of cellular localization. Scale bar = 75 μm.

Immune infiltration analysis

As mitochondria not only integrate cellular metabolism and physiological functions but also serve as a primary source of immunity,¹⁵ moreover, concerning that PPIF is highly expressed in ATAAD monocytes, we conducted a comprehensive immune infiltration analysis to explore the immune landscape in ATAAD and control samples. Bar charts reveal significant differences in the proportions of 22 immune cell types between patients with ATAAD and controls (Figure 9A). The differential expression of immune checkpoint genes also differed significantly between groups. There were significant differences in three of the four immune checkpoint genes, including one high expression (HAVCR2 [coding T cell immunoglobulin mucin 3, TIM3]) and two low expression (ICOS [coding inducible T cell costimulator] and LAG3 [coding Lymphocyte activation gene 3 protein]) (Figure 9B). We also employed ssGSEA analysis of 28 immune cell types to show distinct immune infiltration patterns in ATAAD and control group, the results illustrated that type 2 T helper cells (p < 0.01), monocytes and effector memeory CD4 T cell (p < 0.05) are highly infiltrated in ATAAD compared with control group (Figure 9C). The correlation between PPIF expression and immune cell infiltration was also observed, indicating that PPIF is strongly correlated with monocytes, effector memory CD4 T cell, Type 2 T helper cells, type 1 T helper cells, type 17 T helper cells and T follicular helper cell in ATAAD (Figure 9D). Similar correlations were confirmed using the TIMER (Figure 9E), CIBERSORT (Figure 9F) and xCell (Figure 9G) databases in CPTAC. In common, PPIF expression in HNSCC and LUAD is negatively correlated with DC (p < 0.05) (TIMER database, Figure 9E), but positively correlated with Neutrophils (p < 0.05) (CIBERSORT database, Figure 9F). This finding suggests that high PPIF levels may contribute to an immunosuppressive niche by impairing DC-mediated antigen presentation—a critical step in initiating anti-tumor T cell responses. Conversely, the positive correlation with neutrophil infiltration (CIBERSORT) aligns with established pro-tumor roles of tumor-associated neutrophils (TANs), which often promote angiogenesis, matrix remodeling, and T cell suppression.¹⁶ This dichotomy implies that PPIF may facilitate immune evasion by simultaneously suppressing adaptive immunity (via DC inhibition) and recruiting pro-tumor innate cells (neutrophils). For most cancers, PPIF expression is positively related with common lymphoid progenitors (CLP) and pro B cells, while negatively related with myocytes, megakaryocytes, chondrocytes, endothelial cells, and hematopoietic stem cells (HSC) (xCell, Figure 9G). Beyond infiltrating immune cells, PPIF expression shows systematic associations with progenitor and differentiated cell lineages (xCell database). Its positive correlation with Common Lymphoid Progenitors (CLP) and pro-B cells suggests a potential role in promoting early lymphoid differentiation. Conversely, strong negative correlations with hematopoietic stem cells (HSC), megakaryocytes, endothelial cells, chondrocytes, and myocytes indicate the broad suppression of mesenchymal/stromal differentiation pathways. The suppression of HSCs is particularly notable, as it may reflect PPIF-driven metabolic reprogramming favoring rapidly dividing tumor cells over quiescent stem populations.

Figure 9.

Immune infiltration analysis

(A) Bar charts of 22 immune cell proportions in ATAAD and controls.

(B) Differential expression of various immune checkpoint genes between ATAAD and control groups. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(C) Immune cell infiltration analysis of 28 cell types in ATAAD and normal groups using ssGSEA. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(D) Correlation between PPIF expression and immune cell infiltration as determined by ssGSEA. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(E) The correlation between PPIF expression and immune cell infiltration in TIMER. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(F) The correlation between PPIF expression and immune cell infiltration in CIBERSORT. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(G) The correlation between PPIF expression and immune cell infiltration in xCell. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Pan-cancer analysis of peptidylprolyl isomerase F

The TCGA, GTEx, and CPTAC databases were initially used to conduct a systemic pan-cancer analysis of PPIF expression at mRNA and protein level. PPIF is upregulated in most cancers, including LUAD, HNSCC and Uterine Corpus Endometrial Carcinoma (UCEC) (Figures 10A–10C). Figure 10D showed the expression of PPIF in normal tissues based on HPA database, providing a baseline for comparison with cancer tissues. ROC curve analysis in Figures 10E and 10G demonstrated the diagnostic potential of PPIF in LUAD and HNSCC, with both cancers showing significant sensitivity and specificity for PPIF. Figures 10F and 10H depicted Kaplan-Meier survival plots for LUAD and HNSCC, indicating that higher PPIF expression is associated with poorer overall survival in both cancers, suggesting its prognostic value. Next, we analyzed the transcriptomic data of HNSCC and LUAD from the TCGA database. The cancer population was divided into high- and low-PPIF expression groups based on the median PPIF expression level (p < 2.22e-16) (Figures S3A and S3C). Differential analysis was then performed between the two groups, followed by GSEA enrichment analysis. We found that high PPIF expression in both HNSCC and LUAD was associated with mTORC1 signaling, Myc Targets V1, and Myc Targets V2 (Figures S3B and S3D). mTORC1 is a complex composed of mTOR kinase, mLST8, DEPTOR, Tti1/Tel2, Raptor, and PRAS40, it promotes tumorigenesis and development by regulating processes such as metabolic reprogramming (e.g., glycolysis, lipid synthesis, nucleotide synthesis), autophagy, and macropinocytosis. It also plays a role in tumor angiogenesis and immune regulation.¹⁷ MYC targets v1 and v2 pathways, as reported, play key roles in cancer. Both correlate with increased cell proliferation (linked to MKI67), more aggressive tumor features (higher grade, advanced stage, poorer subtypes), and worse survival.¹⁸ These pathways may clarify the mechanisms that underlie the link between PPIF and poor prognosis. Finally, Figure 10I presented survival analysis across 33 cancers, showing that PPIF expression correlates significantly with OS, DSS, DFI, and PFI, highlighting its broader role in cancer prognosis. These findings position PPIF as a potential biomarker for both cancer diagnosis and prognosis.

Figure 10.

PPIF expression and its pan-cancer analysis

(A) Pan–cancer expression of PPIF in the TCGA dataset. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(B) Pan–cancer expression of PPIF in the TCGA and GTEx datasets. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(C) Pan–cancer expression of PPIF in Clinical Proteomic Tumor Analysis Consortium (CPTAC) database (protein level). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(D) Expression levels of PPIF in normal tissues based on The Human Protein Atlas (HPA).

(E) ROC curve of PPIF in LUAD.

(F) Kaplan-Meier (KM) survival plot based on PPIF expression in LUAD.

(G) ROC curve of PPIF in HNSCC.

(H) KM survival plot based on the expression of PPIF in HNSCC.

(I) Survival analysis of PPIF in 33 cancers, including Overall survival (OS), Disease-specific survival (DSS), disease-free interval (DFI), and progression-free interval (PFI).

Verification of the expression and function of peptidylprolyl isomerase F in lung adenocarcinoma and head and neck squamous cell carcinoma through in vitro experiments

Based on the above results, we found that PPIF is highly expressed in ATAAD and has its unique pan-cancer value, while Mendelian Randomization also confirmed the causality between ATAAD and LUAD/HNSCC. Thus, we wonder whether PPIF could promote the progression of LUAD and HNSCC, making PPIF a budding star linking ATAAD and cancers.

In LUAD, to evaluate the expression of PPIF at the cellular level, we performed qRT-PCR and WB assays in BEAS-2B normal human bronchial epithelial cells, PC9 LUAD cells, and siPPIF PC9 cells. As shown in Figures 11A and 11B, compared with normal cells, PPIF is overexpressed in LUAD cells, while siPPIF effectively downregulates PPIF expression in PC9 cells (Tables 1 and 2). Next, CCK-8 and EdU assays indicated that, compared with normal cells, PPIF promoted the proliferation of LUAD cells, but the knockdown of PPIF significantly reduced the proliferation (Figures 11C and 11D). To evaluate the migrative effect of PPIF in LUAD, the Transwell migration assay and wound-healing assay were conducted. Presented in Figures 11E and 11F, PPIF stimulated the migration of LUAD, while the silencing of PPIF reversed this phenomenon. Furthermore, the Transwell invasion assay demonstrated that PPIF propelled the invasion of LUAD (p < 0.001), and the downregulated PPIF yielded similar results (p < 0.001) (Figure 11G).

Figure 11.

PPIF promoted LUAD cell proliferation, migration, and invasion

(A) QPCR assay confirmed the mRNA expression levels of PPIF in the PC9 LUAD cell line and BEAS-2B normal human bronchial epithelial cells, as well as the knockdown of PPIF expression 24 h after transfection in PC9 cells. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 6). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(B) WB confirmed the protein expression of PPIF in the PC9 LUAD cell line and BEAS-2B cells, as well as the reduced PPIF levels at 48 h after siPPIF transfection in PC9 cells. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(C) CCK-8 assay was performed to illustrate the proliferative ability of PPIF in LUAD. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(D) EdU assay demonstrated the proliferative ability of PPIF in LUAD. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(E) Transwell migration assay showed the enhanced migration potential of PPIF in LUAD cells. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(F) Wound healing assay further confirmed the migration ability of PPIF in LUAD cells. Data are presented as mean ± SD, analyzed by two-way ANOVA (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(G) Transwell invasion assay verified the invasive capability of PPIF in PC9 LUAD cell lines.

Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Table 1.

The siRNA sequences targeting PPIF

Gene			Sequences (5′-3′)
PPIF	siPPIF	Sense:	CAUCCAAGAAGAUUGUCAUTT
		Antisense:	AUGACAAUCUUCUUGGAUGTT

Table 2.

Sequences of primers used for the amplification of genes

Gene	Primer nucleotide sequence
PPIF	Forward: 5′-TGGTGACACAGGCCACAGAC-3′
GAPDH	Reverse: 5′-CCGGAGCACAGGAGCTTACA-3′
	Forward: 5′-GGAGCGAGATCCCTCCAAAAT-3′
	Reverse: 5′- GGCTGTTGTCATACTTCTCATGG-3′

As for HNSCC, qRT-PCR and WB assays were carried out in HaCaT human immortalized epidermal cells, HSC-3 HNSCC cells, and siPPIF HSC-3 HNSCC cells. Results illustrated that PPIF is upregulated in HSC-3 HNSCC cells, compared with normal HaCaT cells (p < 0.001), and siPPIF resulted in the efficient downregulation of PPIF expression (p < 0.001) (Figures 12A and 12B). Revealed by CCK-8 and EdU assays, PPIF facilitated the proliferation of HNSCC (p < 0.001) (Figures 12C and 12D). Transwell migration assay and wound-healing assay denoted that the knockdown of PPIF significantly decreased HNSCC cell migration (p < 0.001) (Figures 12E and 12F). Likewise, PPIF knockdown also depressed the invasion of HSC-3 HNSCC cells (p < 0.001) (Figure 12G).

Figure 12.

PPIF promoted HNSCC cell proliferation, migration, and invasion

(A) QPCR assay verified the mRNA expression of PPIF in HSC-3 HNSCC cell line and HaCaT human immortalized epidermal cells, as well as knocked down PPIF expression at 24h post-transfection in HSC-3 cells. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 6). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(B) Western blot (WB) assay verified the expression of PPIF protein level in HSC-3 HNSCC cell line and HaCaT human immortalized epidermal cells, as well as knockdown of PPIF expression at 48h post-transfection in HSC-3 cells. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(C) CCK-8 assay was performed to illustrate the proliferative ability of PPIF in HNSCC. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(D) EdU assay was performed to illustrate the proliferative ability of PPIF in HNSCC. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(E) Transwell migration assay verified the migration ability of the HSC-3 cell line compared with HaCaT cells, moreover, PPIF plays a vital role in HSC-3 cell migration ability. Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(F) Wound healing assay verified the migration ability of the HSC-3 cell line compared with HaCaT cells, moreover, PPIF plays a vital role in HSC-3 cell migration ability. Data are presented as mean ± SD, analyzed by two-way ANOVA (n = 3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(G) Transwell invasion assay verified the invasion ability of PPIF in HSC-3 cell lines.

Data are presented as mean ± SD, analyzed by one-way ANOVA (n = 4). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Collectively, PPIF is not only overexpressed in LUAD and HNSCC, but also plays a critical role in the proliferation, migration, and invasion of LUAD and HNSCC, suggesting that PPIF might be a promising target for these cancers.

Potential pathogenic mechanisms of peptidylprolyl isomerase F and beyond

As PPIF is dysregulated in both ATAAD and cancers, it is of significance to figure out the regulatory network of PPIF. To identify key transcriptional changes and gain deeper insights into regulatory hub proteins, we used the NetworkAnalyst platform to discover regulatory TFs and miRNAs. Figure S4A depicted the interactions between TF regulators and PPIF. The interactions of the miRNA regulators with the PPIF are shown in Figure S4B. We found that 28 TFs and 37 post-transcriptional (miRNA) regulatory signals were predicted to target PPIF, implying a significant interplay between them. We also used CPTAC to discover the correlation between PPIF expression and cell senescence markers in multiple cancers. Of note, in LUAD, there is a strong positive correlation between PPIF and MIK67 (p < 0.001), a classic proliferation marker (Figure S4C). Figure S4D displayed the association between PPIF and m6A methylation markers, indicating that PPIF may promote the progression of cancers via m6A methylation. The expression of five transcripts of PPIF, including 1 non-coding transcript and 4 protein-coding transcripts (Figure S4E). Notably, the results illustrated that the expression of protein-coding transcripts was significantly elevated om multiple cancer types, while the non-coding transcript was reduced in several cancer types (Figure S4F). Besides, except for ATAAD and cancers, with the aid of the EnrichR platform, PPIF was found to be associated with other diseases, such as Muscular Dystrophy (p < 0.001) and Myopathy (p < 0.001) (Table S4). Protein–drug interaction analyses provide insights into the structural features that influence receptor sensitivity, which can aid in drug discovery.¹⁹ Using transcriptional data from the DSigDB database, EnrichR identified ten potential therapeutic drugs, with the top 10 candidates selected based on their p-values. Table S4 lists the top 10 enriched drugs from the DSigDB database: SC-560, benfluorex, isoxsuprine, colforsin, Gly-His-Lys, IVERMECTIN, LY-294002, Diquat dication, domperidone, and H-89 (Table S5).

Discussion

To our knowledge, this study provides evidence for a causal relationship between ATAAD and cancer and reveals a potential biomarker, PPIF, that connects them. The main findings included (i) A prospective cohort study and MR analysis confirmed the association and causation between ATAAD and LUAD/HNSCC, suggesting that ATAAD may increase the risk of these two types of cancer; (ii) PPIF, highly expressed in ATAAD tissues (especially in monocytes), may be a potential diagnostic biomarker for ATAAD;(iii) PPIF is highly expressed in various cancer types and has shown significant diagnostic and prognostic potential in LUAD and HNSCC. (iv) The regulatory mechanism of PPIF may be related to the immune microenvironment, TF, miRNAs et al.

As the second leading cause of death worldwide, cancer has caused approximately 600 000 deaths in America in 2020.²⁰ Previous studies have demonstrated the association between PPIF and cancers.²¹ In our study, the pan-cancer analysis showed that PPIF was upregulated in most cancers and associated with poor cancer prognosis. In vitro experiments found that PPIF promoted LUAD and HNSCC cell proliferation, migration, and invasion. Taken together, PPIF may serve as a promising biomarker and therapeutic target for cancer. Aortic dissection, a fatal disease with high mortality worldwide, is characterized by the rupture of the aortic wall’s layers.^22,23 Data indicate that the incidence of ATAAD is 4–7.7 per 100,000 patient-years, with in-hospital mortality rates ranging from 17% to 26%.²⁴ For most patients suffer from ATAAD, immediate surgery is the gold standard.²⁵ According to the study of Liu et al., cyclophilin D (CypD), encoded by PPIF, played a role in regulating inflammation in the aorta.²⁶ Our study also found that PPIF was essential for ATAAD, highly expressed in aortic tissues in patients with ATAAD. Currently, the focus on ATAAD lies predominantly on surgical intervention, underscoring the importance of primary prevention and treatment. Thus, monitoring and targeting PPIF may achieve a viable strategy for the primary prevention and treatment of ATAAD. As surgical techniques for ATAAD advanced significantly, postoperative survival rates are expected to increase, with growing attention on postoperative outcomes.⁷ Several studies are currently investigating short-term health outcomes following ATAAD surgery, such as postoperative stroke, postoperative acute kidney injury, malperfusion, and 30-day mortality.^27,28,29 However, limited studies have addressed long-term survival conditions after surgeries. In this study, we investigated the correlation and causal relationship between the relatively rare disease, ATAAD and cancer. We also aimed to identify key molecular mechanisms or pathways linking the two diseases, which may offer diagnostic and therapeutic targets for improving postoperative survival in patients with ATAAD.

There are multiple shared risk factors between ATAAD and cancers, such as tobacco use, obesity, diabetes mellitus, alcoholism, age et al.^25,30,31 Common biological mechanisms between cardiovascular disease and cancer were also reported, for example, inflammation, clonal hematopoiesis, hypoxia, circulating microRNAs, and extracellular vesicles,³⁰ indicating the potential clinical association between them. On the one hand, the cardiovascular risk of patients with cancer has been reported widely, accounting for a large part of cardio-oncology studies. A population-based study has uncovered that patients with cancer face a significantly higher risk of aortic dissection mortality compared to the general population, especially in the 2–11 months post-diagnosis. These findings highlight the need for early and ongoing cardiovascular management in patients with cancer to prevent and monitor aortic dissection throughout survivorship.² Besides, “reverse cardio-oncology” has also drawn increasing attention, thus prompting physicians to monitor cancer status in patients with cardiovascular disease. For example, a meta-analysis showed that individuals with hypertension had a 1.6-fold increased risk of developing renal cell carcinoma compared to normotensive individuals.³² A Danish study assessed cancer risk and mortality in patients with MI. Patients of all age groups showed higher cancer incidence rates one year after MI diagnosis.³³ Compared to participants without HF, those who developed HF within one month of MI were more likely to develop cancer.³⁴ And some studies have combined patient observations with preclinical models of CVD, including MI, subsequent HF, and surgical models of aortic valve stenosis, identifying various factors driving CVD-induced tumor acceleration.³⁵ While current cardio-oncology research mainly focuses on the relationship between common CVD and cancer, our study extends this topic by exploring the link between ATAAD and cancer, particularly through identifying PPIF as a shared molecular mechanism. Given the increasing postoperative survival rates among patients with ATAAD, this conclusion is of significant importance, as long-term postoperative management has become crucial.³⁶

The mitochondrial permeability transition pore (mPTP) is a transient, non-specific channel located in the inner mitochondrial membrane, responsible for regulating molecular flux between mitochondria and the cytosol. CypD resides in the mitochondrial matrix and interacts with the β-subunit of ATP synthase (ATP5B) to modulate the opening of the mPTP.³⁷ Emerging evidence suggests that PPIF is involved in the pathogenesis of both ATAAD and cancer through multiple mechanisms. First, PPIF plays a role in immune regulation. In ATAAD, CypD expression correlates with increased monocyte infiltration and heightened inflammatory responses.³⁸ Specifically, CypD may activate the NLRP3 inflammasome, leading to the caspase-1-mediated degradation of contractile proteins in vascular smooth muscle cells (SMCs), thereby impairing aortic contractility and contributing to disease progression.³⁹ In the context of cancer, CypD modulates T cell metabolism and proliferation.⁴⁰ Th1 cell-induced immune responses is generally associated with enhanced anti-tumor activity, while an overrepresentation of Th2 cells can suppress cellular immunity.⁴¹ PPIF appears to influence the Th1/Th2 balance, thereby affecting tumor immune surveillance and potentially facilitating tumor progression.

Second, PPIF is closely associated with mitochondrial quality control and cell fate decisions. It has been shown to inhibit the FOXO3a/PINK1-Parkin pathway, thereby suppressing mitophagy and promoting the accumulation of damaged mitochondria in LUAD cells.⁴² This mitochondrial dysfunction exacerbates oxidative stress and accelerates tumor progression. Similarly, in ATAAD, impaired mitochondrial homeostasis—characterized by Ca²⁺ overload and ROS imbalance—can trigger the activation of the NLRP3 inflammasome, leading to caspase-1-dependent pyroptosis or SMC apoptosis, further contributing to aortic wall degeneration.²⁶ Studies have also found that cell-specific CypD depletion had a global antioxidant effect on the cardiovascular system. These findings suggested a previously unrecognized role for mitochondrial CypD in the regulation of superoxide and metabolism in vascular smooth muscle and endothelial cells, thereby influencing endothelial barrier function and smooth muscle vascular activity.⁴³ In cancers, CypD also contributes to the disruption of the tumor microenvironment, and excessive ROS directly promotes tumor development and metastasis.⁴⁴ Taken together, we suggested that defective mitophagy and ROS imbalance may represent a common pathological mechanism in both ATAAD and malignancies.

Finally, PPIF participates in some critical protease cascades. For example, abnormal mPTP can lead to calcium overload, activating the protease cascade and upregulating matrix metalloproteinase 9 (MMP9). MMP9 is often produced by VSMCs and macrophages and participates in extracellular matrix degradation, serving as a common mechanism in various vascular diseases.⁴⁵ In ATAAD, MMP9 activates the 8-OHdG/NLRP3/MMP9 pathway via ROS release, promoting VSMC apoptosis and elastic fiber degradation.⁴⁶ Besides, MMP9 promotes basement membrane degradation and accelerates metastasis in cancer.^47,48 Therefore, as a common mechanism between the two, PPIF may promote tumor progression after ATAAD through this pathway.

Cyclosporine A (CSA) is a widely used immunosuppressant that inhibits cyclophilin D (CypD) by binding to its active site, disrupting the interaction between CypD, the inner mitochondrial membrane, and other associated proteins. Despite its effectiveness, prolonged CSA administration is associated with toxicity. In contrast, specific CypD inhibitors exhibit lower toxicity and enhanced efficacy.⁴⁹ In this regard, mitochondrial-targeted CypD inhibitors have been developed as promising agents to mitigate mitochondrial permeability transition, myocardial reperfusion injury, and vascular oxidative stress.^50,51,52 In cancer, CypD plays a multifaceted role in tumor progression, influencing tumor cell fate and contributing to drug resistance.⁵³ Studies suggest that combining CypD inhibitors with PI3K small molecule inhibitors (PI3Ki) effectively overcomes resistance mediated by the PI3K/Akt2/CypD signaling pathway, thereby inducing significant tumor cell apoptosis.⁵⁴ However, the activation of CypD-mPTP axis has been exploited in the development of antitumor drugs as well, with the promotion of tumor cell death. Given these findings, further research into targeting PPIF for the treatment of ATAAD or cancers with high PPIF expression remains a promising avenue for therapeutic development.

Nevertheless, there are still several limitations to our study. Firstly, as a kind of non-secretory protein, blood PPIF was difficult to test in patients with ATAAD, which may limit the clinical application of PPIF as a blood biomarker for ATAAD diagnosis in emergencies. In the future, potential clinical applications may be advanced through methods such as measuring PPIF-related secreted proteins, detecting PPIF expression in monocytes using flow cytometry, or measuring extracellular vesicles of dissection blood samples. Secondly, in our study, we mainly focus on the correlation between ATAAD and LUAD/HNSCC, without exploring the relationship with other cancers, which are expected to be validated in a future study. Thirdly, further in vivo and in vitro studies are essential for confirming the molecular mechanisms by which PPIF regulates ATAAD and tumors. In addition, the hypothesis that PPIF is a true disease-hub gene also needs to be validated by in vivo functional experiments.

Limitations of the study

Several limitations need to be acknowledged in this study. First, the reliance on public datasets (e.g., GEO, TCGA) and UK Biobank data may constrain the generalizability of findings across diverse populations and clinical settings. While multi-omics integration and experimental validation strengthen the mechanistic insights, future prospective cohorts with standardized sampling could enhance clinical applicability.

Second, although in vitro assays demonstrated PPIF’s functional role in cancer proliferation and invasion, these models cannot fully recapitulate the complex in vivo microenvironment of ATAAD or metastatic tumors. Future studies using in vivo models would help validate PPIF’s pathophysiological relevance across disease contexts. Besides, clinical samples used for validation were all derived from male participants, which may cause bias.

Lastly, the clinical utility of PPIF as a non-secretory biomarker remains challenging. Blood-based detection requires specific approaches—such as PPIF expression profiling in circulating monocytes or quantification of PPIF-associated extracellular vesicles—to enable point-of-care diagnostics for ATAAD.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Weixian Yang (fwywx66@126.com).

Materials availability

This study did not generate unique reagents.

Data and code availability

• •Data: All data reported in this article can be accessed from the Gene Expression Omnibus (GEO: https://ncbi.nlm.nih.gov/geo/) and The Cancer Genome Atlas Program (TCGA, GDC: https://portal.gdc.cancer.gov/). This article analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table.
• •Code: This article does not report original code.
• •All other requests: Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments

The work was supported by the grants from National Natural Science Foundation of China (82192902), Parallel Project of Excellent Clinical Research Program in Research Wards of Beijing (BRWEP2024W012060111), and Beijing Anzhen Hospital High Level Research Funding (2024AZB1004). We also acknowledge Dr. Wen Xu from China Medical University for his advice and guidance on scRNA-seq data processing.

Author contributions

Conceptualization: F.T.; bioinformatic data processing: F.T.; methodology: F.T., Y.J., and Z.S.; cohort analysis: R.W.; formal analysis: Y.X.; resources: X.X., X.F., H.Q., and W.Y.; writing - original draft: F.T.; writing - review and editing: M.Y.; supervision: H.Q. and W.Y.; funding acquisition: W.Y.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PPIF antibody	Abcam	cat# ab110324
GAPDH antibody	Thermo Fisher Scientific	cat# 81640-5-RR
Biological samples
ATAAD ascending aorta samples	This study	NA
Normal aortic samples	This study	NA
Chemicals, peptides, and recombinant proteins
Roswell Park Memorial Institute medium 1640 (RPMI-1640) with 10% fetal bovine serum (FBS)	XYZcell	cat# SNB-TC-0753-CM
Dulbecco's modified eagle medium (DMEM) with 10% FBS	XYZcell	cat# SNB-NC-0001-CM
Minimum Essential Medium (MEM) with 10% FBS	XYZcell	cat# SNB-NC-0335-CM
DMEM/F12 with 10% FBS	XYZcell	cat# SNB-NC-0002-CM
SYBR Green Master Mix	Thermo Fisher Scientific	4309155
RIPA	Thermo Fisher Scientific	89900
PMSF	Beyotime	ST505
Phosphatase inhibitor cocktail A	Beyotime	P1081
normal blocking goat serum solution	ZSGB-BIO	ZLI-9056
HRP substrate peroxide solution and HRP substrate luminol solution	EMD Millipore Corporation	WBKLS0500
Critical commercial assays
Transcriptor First Strand cDNA Synthesis Kit	Roche	Cat. No. 04 897 030 001
BeyoClick™ EdU Cell Proliferation Kit with AF594	Beyotime	C0078S
Deposited data
UK Biobank	https://www.ukbiobank.ac.uk	NA
GWAS	https://gwas.mrcieu.ac.uk/	NA
GSE52093	GEO	https://www.ncbi.nlm.nih.gov/geo/
GSE98770	GEO	https://www.ncbi.nlm.nih.gov/geo/
GSE190635	GEO	https://www.ncbi.nlm.nih.gov/geo/
GSE222318	GEO	https://www.ncbi.nlm.nih.gov/geo/
TCGA RNA-seq data	TCGA	https://xenabrowser.net/datapages/
Experimental models: Cell lines
Human LUAD cell lines PC-9	XYZcell	cat# SNB-TC-0753
normal human lung epithelial cells BEAS-2B	XYZcell	cat# SNB-NC-0001
human HNSCC cell lines HSC-3	XYZcell	cat# SNB-TC-0335
human immortalized keratinocytes HaCaT	XYZcell	cat# SNB-NC-0002
Software and algorithms
R	https://www.r-project.org	NA
ImageJ	https://imagej.net/software/imagej/	NA
GraphPad Prism	https://www.graphpad-prism.cn/	NA

Experimental model and study participant details

Patients and tissue samples

Ascending aorta samples were obtained from patients with sporadic Stanford-A aortic dissection undergoing open surgical repair, as well as from heart transplant donors and recipients without aortic disease at Fuwai Hospital, National Center for Cardiovascular Diseases, Beijing, China. Patients with potential heritable aortic conditions or infectious aortitis were excluded from the study. Eligible patients were diagnosed with ATAAD based on repeated ultrasonography or CT angiography. Individuals with Stanford-B AAD, Loeys-Dietz Syndrome, or Marfan syndrome were also excluded. Normal aortic tissues were sourced from organ donors who had no history of aortic disease. The study was approved by the Ethics Committee of Fuwai Cardiovascular Hospital (approval number: 2021-1505). Prior to registration, all participating patients provided informed consent, in accordance with the principles outlined in the Declaration of Helsinki. All participants, containing 3 ATAAD patients and 3 controls without aortic disease were Asian males, aged 50-75 years.

Cell lines

Human LUAD cell lines PC-9 (cat# SNB-TC-0753) cultured in Roswell Park Memorial Institute medium 1640 (RPMI-1640) with 10% fetal bovine serum (FBS) (cat# SNB-TC-0753-CM), normal human lung epithelial cells BEAS-2B (cat# SNB-NC-0001) cultured in Dulbecco's modified eagle medium (DMEM) with 10% FBS (cat# SNB-NC-0001-CM), human HNSCC cell lines HSC-3 (cat# SNB-TC-0335) cultured in Minimum Essential Medium (MEM) with 10% FBS (cat# SNB-NC-0335-CM) and human immortalized keratinocytes HaCaT (cat# SNB-NC-0002) cultured in DMEM/F12 with 10% FBS (cat# SNB-NC-0002-CM) were all purchased from XYZcell (Shanghai, China). The cell lines have been authenticated with STR, Besides, the cell lines were tested for mycoplasma contamination every month. All cells were cultured in an incubator with 5% CO2 at 37°C.

Method details

UK Biobank cohort analysis

All data used in this investigation were obtained from the UK Biobank database. UK Biobank is a prospective cohort study involving over 500,000 community-dwelling adults aged 37 – 73 years old. Participants were recruited from various locations across the UK between 2006 and 2010 and completed self-administered questionnaires and relevant physical examinations. UK Biobank data has approval from the North West Multi-center Research Ethics Committee (MREC) (REC reference: 21/NW/0157). This research has been conducted with the UK Biobank Resource under project 532333.

We set the exposure factor as whether or not aortic aneurysms and dissections were presented. Cancer outcomes were defined using ICD-10 coding. Cancer data were provided by the National Cancer Registry. Outcome events were defined as the first cancer diagnosis, loss to follow-up, or end of follow-up, whichever occurred first and follow-up time was calculated as the duration from baseline enrollment to the occurrence of outcome events. First, we conducted multivariate analyses to reveal the associations between aortic aneurysms and dissections and various cancers using Cox proportional hazards regression models. The assumption of the proportional hazards model was assessed through various adjustment and validation methods, and the results suggested no significant departures from these assumptions. Some possible confounding variables listed below should be taken into account: sex (male or female), age (years), race (Non-White or White), BMI (kg/m²), Townsend deprivation index, qualification (e.g., A levels/AS levels or equivalent, College or University degree, CSEs or equivalent, NVQ or HND or HNC or equivalent, O levels/GCSEs or equivalent, Other professional qualifications, or None of the above), smoking status (Never, Previous, or Current), alcohol intake frequency (Never, Special occasions only, One to three times a month, Once or twice a week, Three to four times a week, or Daily or almost daily), physical activity (metabolic equivalent task minutes), family history (Yes or No), and diet score. Results were presented as hazard ratios (HR) and 95% confidence intervals (CI). The HRs were interpreted as an average effect during the entire follow-up period.⁵⁵ Second, we plotted Kaplan-Meier survival curves and could visually see the survival of the two groups at each time point. We also used the log-rank test to compare whether survival curves differed significantly between groups.

Mendelian Randomization

Mendelian randomization (MR) is a reliable tool to analyse the causality between exposure and clinical outcome.⁵⁶ There are three basic principles for MR. Firstly, genetic variants are strongly related with exposure. Secondly, genetic variant tools are not correlated with any confounding factors. Thirdly, outcome is influenced only through risk factors.⁵⁷ MR was implemented in R (version 4.3.2) using “TwoSampleMR” package.⁵⁸ Inverse variance weighted (IVW) method is widely considered as the primary method,⁵⁹ with the requirement of p <0.05. Data for ATAAD, LUAD and HNSCC were attained from genome-wide association studies (GWAS). For MR analysis between ATAAD and LUAD (GWAS ID: ebi-a-GCST004744), we use genome-wide significant SNPs with a p-value threshold of 5e-8 to ensure the robust result. For MR analysis between ATAAD and HNSCC (GWAS ID: ieu-b-91), due to the limited number of SNPs, a more lenient p-value threshold of 5e-6 was employed. Moreover, we harmonized the allelic directions of filtered SNPs in exposure and outcome data. IVW, MR Egger, Weighted median, Simple mode and Weighted mode were used to confirm the causality. Of note, in our study, IVW method is the primary method to determine causality. Heterogeneity analysis evaluates the consistency of causal effect estimates between exposure and outcome across various SNPs. Tests for horizontal pleiotropy assess whether a SNP influences the outcome through pathways unrelated to the exposure, potentially introducing bias. Leave-one-out analysis examines the contribution and stability of individual SNPs to the overall Mendelian randomization effect.

Data collection and processing

Three ATAAD public datasets, GSE52093,⁶⁰ GSE98770⁶¹ and GSE190635,⁶² including 17 ATAAD patients and 14 healthy donors, were downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/) database. To remove batch effect, SVA package in R software 4.3.2 was used (Figure S1). ScRNA-seq data from GSE222318 were also downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/) database.⁵ The Cancer Genome Atlas (TCGA) was used to attain RNA-seq data and the clinical information of 33 different cancer types. Genotype-Tissue Expression (GTEx) expression was also utilized to gain gene expression data of normal tissue.⁶³ On protein level, Clinical Proteomic Tumor Analysis Consortium (CPTAC) database was also employed to gain gene expression data in both cancer and normal tissue.⁶⁴ The Human Protein Atlas (HPA) offered expression data in normal tissue.⁶⁵ Mitochondrial-related genes was obtained from MitoCarta 3.0 database (https://www.broadinstitute.org/mitocarta/mitocarta30-inventory-mammalian-mitochondrial-proteins-and-pathways) and Gene Set Enrichment Analysis (GSEA, http://www.gsea-msigdb.org/gsea/index.jsp).^66,67,68

Identification of differentially expressed genes

Data were downloaded from GSE52093, GSE98770 and GSE190635, and box plot was used to present the batch-corrected matrix. Using “Limma” package of R software, genes with p <0.05 and |log fold change (FC)| >0.5 are regarded as DEGs.⁶⁹ The illustrations of DEGs, such as volcano, heatmap plot, box plot and pearson correlation plot, were generated by “ggplot2”, “pheatmap”, “ggcorrplot” and “ggthemes” packages of R software (version 4.3.2).^70,71,72,73

Functional enrichment analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted using “ClusterProfiler” R package.⁷⁴ GO includes cell composition (CC), biological process (BP) and molecular functions (MF) modules. Furthermore, GSEA (http://software.broadinstitute.org/gsea/index.jsp) was also used for exploring related pathways and potential mechanisms.

Single-cell transcriptome data preprocessing

Standard procedures were employed for single-cell RNA sequencing data analysis. Stringent quality control was first performed, retaining cells expressing >200 genes, with UMI counts below the 99th percentile, and mitochondrial gene proportion <25%. Batch effects were corrected using Harmony, followed by dimensionality reduction and visualization via UMAP. Cell clustering was performed using the FindClusters function in Seurat (v4.3.0).⁷⁵ Differential expression analysis was based on the Wilcoxon rank-sum test. Genes expressed in >25% of cells in either comparison group were considered, with adjusted P-value < 0.05 and |log₂FC| > 0.25 set as significance thresholds. Functional enrichment analysis was conducted using clusterProfiler (v4.6.2),⁷⁶ with significance defined as Benjamini-Hochberg (BH) adjusted P-value < 0.05. Analysis of cell-cell interactions was performed using CellChat (v1.5.0).⁷⁷ Ligand-receptor pairs with a P-value < 0.05 were selected for downstream analysis and visualization to systematically decipher signaling mechanisms between cell subpopulations.

Machine learning

By intersecting DEGs and mitochondrial-related genes, genes that are differentially expressed and associated with mitochondria were referred to as MitoDEGs. Five machine learning methods, least absolute shrinkage and selection operator (LASSO), support vector machine (SVM) algorithm, decision tree algorithm, random forest algorithm and Boruta algorithm were used. In LASSO regression (“glmnet” R package), we fit a LASSO model using glmnet with a binomial family (binary logistic regression), performed 10-fold cross-validation.⁷⁸ The optimal lambda value that minimizes the cross-validated error is 0.004502466. In SVM algorithm, rfe function from “caret” R package executed recursive feature elimination (RFE), in combination with a Support Vector Machine (SVM) model that uses a radial basis function kernel (svmRadial) and cross-validation for performance evaluation.⁷⁹ In random forest algorithm, we fit a random forest model to the training data using the randomForest function (“randomForest” R package), the number of trees were set to 500.⁸⁰ For decision tree algorithm, rpart function (“rpart” R package) was used to fit a decision tree model to the training data, varImp function computes the importance of each variable in the model based on how much each variable contributes to decreasing the node impurity in the tree.⁸¹ The Boruta algorithm (“Boruta” R package), a feature selection method using Random Forest, enhances the original dataset with shuffled “shadow features” to identify significant variables. It builds a forest with these features, evaluates their importance, and iteratively excludes the less crucial ones, ensuring only impactful features are retained.⁸²

Construction of WGCNA

Using WGCNA R package, samples are filtered through hclust function. The soft-thresholding power was 7, recommanded by pickSoft Threshold function, in order to make the established network a power-law distribution in the scale-free topology network. The WGCNA analysis utilizes Pearson correlation to quantify associations between continuous module eigengenes and binarized disease status, with results visualized in a heatmap displaying Pearson's r-values and statistical significance. In labeled heatmap plot, correlation coefficients > 0.7 are regarded as strong correlations. Finally, we took the intersection of genes identified by the five machine learning methods and WGCNA, obtaining the hub genes.

Diagnostic and prognostic analysis

The R package pROC was used to generate receiver operating characteristic (ROC) curves and verify the diagnostic performance of the hub genes in ATAAD.⁸³ Area Under Curve (AUC) is defined as the area under the ROC curve. We often use the AUC value as the evaluation standard for the models, with classifiers with larger AUC values performing better. Generally speaking, an AUC of at least 0.9 is considered Excellent; an AUC between 0.8 and 0.9 is Good; an AUC between 0.7 and 0.8 is Fair; an AUC between 0.6 and 0.7 is Poor; and an AUC between 0.5 and 0.6 is deemed a Fail.⁸⁴ To uncover PPIF’s diagnostic value (ROC curves and Kaplan-Meier curves) and prognostic significance, including Overall survival (OS), Disease-specific survival (DSS), Disease Free Interval (DFI), Progression Free Interval (PFI) in cancers, TCGAplot R package was employed.⁸⁵

Immune infiltration analysis

CIBERSORT was used for predicting immune cell proportion in ATAAD patients and healthy control, based on gene expression profile.⁸⁶ The ssGSEA algorithm in “GSVA” R package, was employed to evaluate the abundance of 28 immune cells in ATAAD patients and healthy controls. Meanwhile, ssGSEA algorithm also evaluated the correlation between hub-genes and 28 immune cells.⁸⁷ The immune infiltration analysis employs Pearson correlation to quantify hub gene-immune checkpoint relationships and hub gene-immune cell score associations, while implicitly leveraging point-biserial correlation for groupwise comparisons of immune cell proportions through boxplots and split-violin plots.

To discover the internal relationship among hub-gene PPIF, immune infiltration and cancers, we used protemic data from CPTAC database, based on ProteoCancer Analysis Suite (PCAS) platform,⁸⁸ then performed TIMER, CIBERSORT and xCell algorithms on pan-cancer level. Besides, the correlations between PPIF, cell senescence and m6A methylation on pan-cancer level were also performed on PCAS.

Gene-regulatory network analysis

Hub-genes-microRNA (miRNA)-interaction networks and Hub-genes-TF-interaction networks were discovered using NetworkAnalyst platform.⁸⁹

Quantitative real-time PCR (qRT-PCR)

Cells were seeded into 6-well plates. After siPPIF (GenePharma, Suzhou, China) trancfection, RNA was extracted with Trizol reagent and was reverse transcribed into cDNA (Roche, Shanghai, China). Then Real time PCR was performed using SYBR Green Master Mix (Thermo Fisher Scientific, Massachusetts, USA). The siRNA targeting PPIF were designed by GenePharma (Suzhou, China) and the sequences were listed in Table 1. Primers were listed in Table 2.

Western Blot

Protein was extracted using RIPA (Thermo Fisher Scientific, Massachusetts, USA), PMSF (Beyotime, Shanghai, China) and Phosphatase inhibitor cocktail A (Beyotime, Shanghai, China). Equal amount of protein was subjected to SDS-PAGE (4-12% gel) and transferred using iBlot™ 2 Transfer Stacks, PVDF membrane (Thermo Fisher Scientific, Massachusetts, USA). After blocking 30 min using QuickBlock™ Western Blocking buffer at room temperature, membranes were incubated with primary antibodies against PPIF (1:1000, Abcam, UK) and GAPDH (1:3000, Thermo Fisher Scientific, Massachusetts, USA) at 4 °C overnight. Subsequently, membranes were washed by TBST three times and incubated with their corresponding secondary antibodies for 1 h at room temperature. Then, membranes were washed by TBST three times, 5 min each time. The bands were visualized by HRP substrate peroxide solution and HRP substrate luminol solution (EMD Millipore Corporation, Burlington, USA). The immunoblot bands intensity were quantified with ImageJ software (NIH, MD, USA).

Immunofluorescence

To perform aortic tissue immunofluorescence, tissue sections were subjected to dewaxing and rehydration procedures, followed by antigen retrieval through boiling with sodium citrate (pH = 6.0). Then washed with ice-cold PBS and fixed in 4% paraformaldehyde solution for 20 min. Subsequently, the cells were permeabilized with 0.3% Triton X-100 and blocked with normal blocking goat serum solution (ZLI-9056, ZSGB-BIO, Beijing, China). After blocking, tissue sections were incubated with primary antibody against PPIF (1:1000, Abcam, UK) overnight at 4 °C, followed by PBS washing and incubation with secondary antibodies at room temperature for 1 h. The cell nuclei were counterstained with DAPI and examined using a fluorescence microscope. The mean fluorescence intensity was quantified using Image J software (NIH, MD, USA).

Cell Counting Kit-8 (CCK-8) assay

Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8) assay, cells were seeded in 96-well plates at a density of 5 × 10³ cells per well and incubated at 37°C in a humidified 5% CO2 incubator for 24 hours. After incubation, 10 μL of CCK-8 reagent was added to each well, and the plate was further incubated for 1–2 hours at 37°C. The absorbance was measured at 450 nm using a microplate reader. The cell viability was calculated as a percentage of the control group.

EdU proliferation assay

The EdU (5-ethynyl-2′-deoxyuridine) incorporation assay was used to assess cell proliferation. Briefly, EdU was added to the culture medium at a final concentration of 10 μM, and cells were incubated with the reagent for another 2 hours to allow incorporation into newly synthesized DNA. Following incubation, cells were fixed with 4% paraformaldehyde for 15 minutes at room temperature. To detect EdU incorporation, the Click reaction was performed according to the manufacturer’s instructions (Beyotime, Shanghai, China). Briefly, the cells were permeabilized with 0.5% Triton X-100, followed by a Click reaction cocktail containing fluorescent azide and copper sulfate. The reaction was allowed to proceed for 30 minutes at room temperature in the dark. After washing with PBS, the cells were stained with DAPI for nuclear visualization. EdU-positive cells were analyzed using a fluorescence microscope, and the percentage of proliferating cells was quantified by counting the EdU-positive cells relative to the total cell number.

Wound healing assay

At 90% confluence of cells in 6-well plates, a wound was scratched by 1000-μL tips. After washing with PBS, the cells were then incubated at 37°C in a humidified 5% CO2 incubator. The images were photographed under the inverted microscope.

Transwell migration and invasion assay

Transwell migration and invasion assays were performed to assess the migratory and invasive abilities of the cells. For migration, cells were seeded in the upper chamber of a Transwell insert (8-μm pore size) in serum-free medium, while the lower chamber was filled with medium containing 10% FBS as a chemoattractant. After 24 hours of incubation, non-migrated cells on the upper surface of the membrane were removed with a cotton swab, and migrated cells on the lower surface were fixed with 4% paraformaldehyde and stained with crystal violet. For invasion assays, the Transwell inserts were pre-coated with Matrigel to mimic the extracellular matrix. The procedure was otherwise identical to the migration assay. The number of cells that migrated or invaded through the membrane was counted in five randomly selected fields under a microscope.

Quantification and statistical analysis

All bioinformatic analyses were performed using R software (version 4.3.2) and online databases. All experiments were performed at least three times and the data were analyzed using GraphPad Prism software and shown as mean ± SD. ∗ p < 0.05, ∗ ∗ p < 0.01, ∗ ∗ ∗p < 0.001 mean that the difference is statistically significant.

Published: September 30, 2025