Programmed cell death-related transcriptomic features are associated with aortic valve calcification progression: an integrated analysis of public datasets and clinical samples

Introduction

Calcific aortic valve disease (CAVD) is the most prevalent valvular disorder in older adults, with its prevalence rising to 12–26% among individuals over the age of 75 years (1,2). As the disease transitions from early sclerosis to advanced stenosis, patients may develop angina, heart failure, with an annual mortality risk of as high as 25–50% (1,3). Despite the substantial clinical burden, no pharmacological therapy has been proven to halt or reverse valve calcification. Although surgical or transcatheter valve replacement can relieve outflow obstruction, these interventions neither target early-stage disease nor modify the underlying pathobiology (4,5). Understanding the molecular determinants of CAVD progression is therefore essential for improving risk stratification and developing new therapeutic strategies.

Current evidence indicates that aortic valve calcification is an active, cell-mediated pathological process rather than a passive degenerative phenomenon (6). In response to diverse pathological cues, valvular interstitial cells (VICs) can undergo osteogenic reprogramming and express bone-related markers, including Runx2 and BMP2, promoting mineral deposition within the leaflet (7,8). At the same time, infiltration of macrophages, T cells, and dendritic cells amplifies inflammatory signaling and accelerates extracellular matrix remodeling (9,10). Although these pathophysiological events have been partially characterized, the molecular triggers that drive the transition from early sclerosis to advanced stenosis remain poorly defined.

Within this complex cellular and inflammatory milieu, programmed cell death (PCD) has emerged as a crucial regulatory process governing cell fate and tissue remodeling (11). Multiple PCD modalities—such as apoptosis, ferroptosis, autophagy, and pyroptosis—have been implicated in regulating oxidative stress, inflammatory cytokine release, metabolic homeostasis, and matrix turnover, all of which may contribute to calcific pathology. For example, ferroptosis-associated lipid peroxidation can intensify oxidative damage, while dysregulated autophagy may enhance osteogenic differentiation (12,13). Pyroptosis-driven cytokine release further promotes inflammation-induced tissue injury (14). However, most existing studies examine individual PCD pathways in isolation, and few have systematically characterized PCD-related transcriptional features across different stages of CAVD. Moreover, the key molecular drivers linking PCD to calcification progression remain unclear.

From a systems biology perspective, crosstalk among PCD pathways, immune microenvironment remodeling, and osteogenic signaling may collectively influence the progression of valve calcification (15). PCD may modulate immune cell recruitment and polarization by altering chemokine production, regulating extracellular vesicle cargo, or reshaping local metabolic substrates. These effects may, in turn, promote activation of VICs and drive calcific matrix deposition (16-18). Profiling PCD-associated transcriptomic alterations may therefore reveal pivotal molecular nodes that connect intracellular death programs with tissue-level pathological remodeling.

In light of these knowledge gaps, the present study aims to comprehensively characterize PCD-related transcriptomic features across normal, sclerotic, and stenotic aortic valve tissues and to identify candidate genes associated with disease progression. By integrating differential expression analysis, gene set variation analysis, weighted gene co-expression network analysis (WGCNA), dynamic clustering, and machine learning-based feature selection, we established an exploratory framework for identifying and ranking key PCD-associated genes. Furthermore, peripheral blood leukocyte samples from clinical cohorts were used to preliminarily evaluate whether the identified tissue-derived candidate genes showed consistent expression trends in an accessible clinical sample type. This integrative strategy seeks to provide association-based evidence regarding PCD-related molecular alterations during aortic valve calcification and to generate hypotheses for future mechanistic studies. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0273/rc).

Methods

Datasets

Two publicly available transcriptomic datasets related to aortic valve calcification were obtained from the Gene Expression Omnibus (GEO) database. The training dataset GSE83453 contains 27 aortic valve tissue samples, including normal valves (n=8), mildly calcified bicuspid aortic valves (BAVs; n=10), and severely stenotic aortic valves (SAVs; n=9). The validation dataset GSE51472 includes five normal, five sclerotic, and five calcified aortic valve samples.

Raw expression matrices and platform annotation files were downloaded from the GEO repository. Data preprocessing consisted of background adjustment, log2 transformation, and quantile normalization. When multiple probes corresponded to a single gene, the probe with the highest average expression was selected. Sample grouping based on disease severity was retained as originally annotated and used consistently in all subsequent analyses.

Differential expression analysis

Differential expression analysis was performed to identify transcriptional alterations associated with aortic valve calcification. After normalization, gene expression matrices from GSE83453 were analyzed using the limma package. Linear models were fitted for each gene, and empirical Bayes moderation was applied to improve variance estimation in groups with limited sample sizes. Contrast matrices were constructed to compare normal valves with mildly and severely calcified valves, respectively. Genes with P<0.05 were considered differentially expressed and were carried forward to downstream analyses.

Gene set variation analysis (GSVA)

GSVA was conducted to evaluate the enrichment of PCD-related gene sets across samples. Four PCD-associated gene sets were retrieved from the Molecular Signatures Database (MSigDB), including GOBP_FERROPTOSIS.v2025.1.Hs, HALLMARK_APOPTOSIS.v2025.1.Hs, KEGG_REGULATION_OF_AUTOPHAGY.v2025.1.Hs, and REACTOME_PYROPTOSIS.v2025.1.Hs. The normalized expression matrix of the GSE83453 dataset was used as input, and enrichment scores were computed for each gene set in each sample. These PCD-related enrichment scores were then incorporated into subsequent analyses to characterize pathway-level transcriptomic differences across disease stages.

WGCNA

WGCNA was performed to identify co-expression modules associated with aortic valve calcification. Genes were first filtered based on expression variability by calculating the median absolute deviation (MAD) across samples, and the top 10,000 most variable genes were retained for network construction. Using the WGCNA R package, a scale-free network was established by selecting an appropriate soft-thresholding power through the pickSoftThreshold function, which evaluates the fit to scale-free topology and average connectivity. An adjacency matrix was then transformed into a topological overlap matrix (TOM), followed by hierarchical clustering to detect gene modules using the dynamic tree-cutting algorithm. Module eigengenes were subsequently calculated, and modules with highly correlated eigengenes were merged to obtain the final set of biologically meaningful modules.

Mfuzz clustering

Time-series-based soft clustering was performed using the Mfuzz package to characterize dynamic expression patterns across disease stages. All genes were first evaluated for expression variability, and those exceeding the predefined filter.std threshold were retained to enhance the biological relevance of clustering. The normalized expression matrix was then subjected to fuzzy c-means clustering. Membership values were calculated for each gene to define the degree of assignment to individual clusters, and genes were ultimately grouped into six clusters, each representing a distinct expression trajectory across the three sample groups.

Construction of the risk scoring model

A risk scoring model was constructed using the candidate genes obtained from the integrated analyses. The least absolute shrinkage and selection operator (LASSO) regression model was implemented to select key features and build the classifier. Sample groups were used as the categorical outcome variable, and 10-fold cross-validation was applied to determine the optimal penalty parameter λ. Genes with non-zero regression coefficients in the final model were retained, and their coefficients were combined to generate the risk score formula. Risk scores were subsequently calculated for all samples in the training dataset and used for downstream stratification and model evaluation.

Immune cell infiltration analysis

Immune cell infiltration was evaluated using the Cell-type Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORT) algorithm. The normalized gene expression matrix was used as input, and immune cell fractions were estimated based on the LM22 signature matrix, which defines 22 human immune cell subtypes. CIBERSORT was run with default parameters to perform the deconvolution analysis, and samples with valid output were included for subsequent comparisons. Differences in immune cell composition among groups were examined through statistical testing, and correlations between key genes and immune cell subsets were further assessed.

Clinical samples

Peripheral blood leukocyte samples were obtained from archived residual blood specimens collected during routine clinical evaluation at The Third Affiliated Hospital of Sun Yat-sen University (Guangzhou, China). According to echocardiographic findings, subjects were categorized into three groups: healthy controls, patients with mild aortic valve sclerosis, and patients with severe aortic valve stenosis. Leukocyte fractions were obtained after erythrocyte lysis and used for subsequent RNA extraction. All specimens were anonymized and handled under standard laboratory conditions to ensure RNA integrity and consistency across groups. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The Third Affiliated Hospital of Sun Yat-sen University (No. A2025-525-01). Informed consent was waived in this retrospective study.

Quantitative real-time polymerase chain reaction (qRT-PCR)

Total RNA was isolated from peripheral blood using a TRIzol-based protocol. Whole blood was first diluted 1:3 with phosphate-buffered saline (PBS) and incubated with red blood cell (RBC) Lysis Buffer at room temperature for 5–10 min, followed by centrifugation at 1,200–1,500 rpm for 5 min. The lysis step was repeated to ensure complete removal of erythrocytes. The resulting leukocyte pellet was washed once with PBS and lysed in TRIzol reagent at three times the sample volume, with incubation at room temperature for 5 min.

Chloroform was added at a ratio of 200 µL per 1 mL TRIzol, mixed vigorously for 15 s, incubated for 2–3 min, and centrifuged at 12,000 ×g, 4 ℃ for 15 min to separate the aqueous phase. The RNA-containing upper phase was transferred carefully to a fresh tube, avoiding disturbance of the interphase. RNA was precipitated by adding 500 µL pre-chilled isopropanol per 1 mL TRIzol, mixing by inversion, incubating for 10 min, and centrifuging at 12,000 ×g, 4 ℃ for 10 min. The RNA pellet was washed with 1 mL of 75% ethanol, centrifuged at 7,500 ×g, 4 ℃ for 5 min, air-dried for 10 min, and dissolved in 50 µL RNase-free water. RNA concentration and purity were measured using a microvolume ultra violet (UV) spectrophotometer.

qRT-PCR was performed using the HiScript II One Step RT-PCR Kit (Vazyme, Nanjing, China). Each 50 µL reaction contained 16.5 µL 2× One Step Mix, 0.5 µL One Step Enzyme Mix, 2.5 µL forward primer (10 µM), 2 µL reverse primer (10 µM), 2 µL RNA template, and RNase-free water to volume. PCR was performed using a standard two-step protocol (50 ℃ for 30 min; 94 ℃ for 30 s; 60 ℃ for 30 s; 72 ℃ for 2 min) for 35 cycles, followed by a final extension at 72 ℃ for 5 min. Primers for CAV1, IGF1, IL7R, and JAK1 were synthesized according to National Center for Biotechnology Information (NCBI) reference sequences (listed in Table 1). Relative mRNA expression levels were calculated using the 2^−ΔΔCt method with GAPDH as the internal control.

Statistical analysis

Statistical analyses for clinical and qRT-PCR data were performed using SPSS version 23.0 (IBM Corp., Armonk, NY, USA). All data were expressed as mean ± standard deviation (SD) from three independent experiments (n=3). Continuous variables were analyzed using one-way analysis of variance (ANOVA) or the Kruskal-Wallis test, depending on data distribution. Post hoc pairwise comparisons were applied when appropriate. Correlations between gene expression and immune cell fractions were assessed using Spearman’s rank correlation analysis. A two-sided P<0.05 was considered statistically significant.

Results

Differentially expressed genes and progressive activation of PCD pathways in aortic valve calcification

To characterize transcriptional alterations associated with calcification severity, we first compared gene expression profiles among control, mildly calcified (BAV), and severely calcified (SAV) valve tissues. Differential expression analysis revealed a graded pattern of transcriptomic disturbance across disease stages. In the BAV group, 524 genes were upregulated and 615 were downregulated relative to controls; in the SAV group, the number of dysregulated genes increased markedly to 2,531 upregulated and 2,606 downregulated transcripts (Figure 1A,1B). This expanding set of perturbed genes reflected the escalating molecular disruption accompanying progressive calcification.

Figure 1 Differentially expressed genes and progressive activation of PCD pathways in aortic valve calcification. (A,B) Volcano plot of differentially expressed genes between control and BAV (A) or SAV (B) samples. Red indicates up-regulated genes, while blue indicates down-regulated genes. (C) Violin plot showing increasing PCD GSVA scores among the three groups. (D) Line plot illustrating the upward trend of PCD pathway scores across disease stages. BAV, bicuspid aortic valve; FC, fold change; GSVA, gene set variation analysis; PCD, programmed cell death; SAV, stenotic aortic valve.

Given the reported relevance of PCD-related genes in tissue degeneration, we next evaluated pathway-level gene set enrichment using GSVA. PCD scores differed significantly among the three groups (Kruskal-Wallis, P=0.01), with both BAV and SAV samples showing higher enrichment relative to controls (Figure 1C). The line-plot trajectory further illustrated a clear upward shift in PCD pathway activity from control to BAV and from BAV to SAV, mirroring the transcriptomic gradients observed in the differential expression analysis (Figure 1D).

Taken together, these findings indicate that as calcification advances, the valve tissue undergoes escalating transcriptional disturbances accompanied by a coordinated elevation in PCD-related pathway activity.

Identification of key co-expression modules associated with calcification severity and PCD activity

Building on the transcriptional gradients observed in “Differentially expressed genes and progressive activation of PCD pathways in aortic valve calcification” section, we next applied WGCNA to uncover coordinated gene expression patterns underlying aortic valve calcification. Using the top 10,000 most variable genes, network topology analysis indicated that a soft-thresholding power of 13 yielded a scale-free structure with a fitting index R²>0.80, along with stable mean connectivity (Figure 2A). This ensured that the resulting network accurately captured biologically meaningful co-expression relationships.

Figure 2 Co-expression network analysis identifies key modules linked to calcification severity and PCD activity. (A) Determination of the optimal soft-thresholding power based on scale-free topology and mean connectivity. (B) Gene dendrogram and module assignment obtained through dynamic tree cutting. (C) Correlations between module eigengenes and clinical traits. PCD, programmed cell death.

Hierarchical clustering and dynamic tree cutting grouped genes into multiple initial modules, which were subsequently merged into six major co-expression modules based on eigengene similarity (Figure 2B). These modules displayed distinct expression signatures, suggesting the activation of different transcriptional programs across disease stages.

Correlation analysis revealed that the brown, pink, and black modules were most strongly associated with both PCD activity and calcification severity (|Cor| >0.4, P<0.05) (Figure 2C). Notably, eigengene values of these modules increased progressively from control to BAV and peaked in SAV tissues, mirroring the stepwise elevation in PCD pathway activity shown in Figure 1C,1D. The consistent directionality across datasets indicates that these modules capture the dominant transcriptional response accompanying disease progression.

In summary, WGCNA identified the brown, pink, and black modules as the co-expression units most sensitive to calcification severity and PCD activity. Their progressively elevated eigengene patterns suggest that these modules may represent coordinated transcriptional programs associated with degenerative changes in aortic valve tissues.

Dynamic clustering identifies six representative expression patterns and highlights 56 convergent candidate genes

To further characterize coordinated transcriptional responses across disease stages, dynamic clustering was performed using Mfuzz on the genes retained from prior analyses. This approach resolved the expression landscape into six distinct temporal clusters, each representing a characteristic trajectory across control, BAV, and SAV samples (Figure 3A). Several clusters—such as clusters 1, 3, and 6—exhibited progressively increasing patterns, whereas others (clusters 4 and 5) displayed stepwise attenuation, indicating that valve calcification involves multiple divergent regulatory programs rather than a single linear response. These temporal signatures underscore the presence of parallel biological processes that are differentially engaged as the disease advances.

Figure 3 Dynamic clustering reveals six distinct temporal expression patterns and identifies 56 core candidate genes. (A) Six dynamic expression clusters identified by Mfuzz. Each panel represents one Mfuzz cluster (Clusters 1–6). Different colored lines indicate individual genes, and the color intensity reflects the membership degree of each gene within the corresponding cluster (higher membership is shown in warmer colors). (B) Venn diagram integrating DEG, WGCNA modules, PCD genes, and Mfuzz clusters. BAV, bicuspid aortic valve; DEG, differentially expressed gene; PCD, programmed cell death; SAV, stenotic aortic valve; WGCNA, weighted gene co-expression network analysis.

To refine the gene set with the greatest biological relevance, we integrated four independent sources of evidence: differentially expressed genes, WGCNA module genes, PCD-associated genes, and Mfuzz cluster members. Venn analysis revealed that the intersection of these datasets yielded 56 genes consistently supported across analytical layers (Figure 3B). These genes are positioned at the convergence of stage-dependent expression changes, co-expression network relevance, pathway linkage, and temporal stability.

Overall, dynamic clustering revealed multiple disease-related transcriptional trajectories, and the 56 genes identified through multi-source intersection represent a stable, biologically convergent candidate set with high relevance to calcification progression.

LASSO-based feature selection identifies four key genes and establishes a robust risk scoring model

Following the convergence of 56 candidate genes from multi-dimensional analyses, we applied LASSO regression to further refine the most informative predictors of calcification severity. The coefficient profiles showed that only a subset of genes maintained non-zero weights as the penalty increased (Figure 4A). Ten-fold cross-validation identified an optimal λ value that minimized binomial deviance (Figure 4B), yielding four key genes—CAV1, IGF1, IL7R, and JAK1—as the final features incorporated into the risk model.

Figure 4 LASSO-based feature selection identifies four key genes and establishes a robust risk scoring model. (A) LASSO coefficient profiles of 56 candidate genes across increasing penalty parameters. (B) Cross-validation plot identifying the optimal λ for LASSO feature selection. (C) Distribution of risk scores across disease stages. (D) ROC curve showing strong discriminatory ability of the risk model. (E) Calibration curve demonstrating good agreement between predicted and observed probabilities. (F) Nomogram integrating gene-based risk score for predicting calcification probability. (G) Expression levels of four LASSO-selected genes across control, BAV, and SAV groups. **, P<0.01; ***, P<0.001 vs. control group. AUC, area under the curve; AV, aortic valve; BAV, bicuspid aortic valve; LASSO, least absolute shrinkage and selection operator; ROC, receiver operating characteristic; SAV, stenotic aortic valve.

Based on these four genes, a composite risk score was calculated for each sample. The distribution of risk scores demonstrated a clear stepwise elevation from controls to BAV and further to SAV (Kruskal-Wallis, P=8.7×10⁻⁵), indicating strong stage discrimination (Figure 4C). Receiver operating characteristic analysis showed excellent classification performance, with an area under the curve (AUC) of 0.944 for distinguishing calcified from non-calcified valves (Figure 4D). Calibration analysis confirmed good agreement between predicted and observed probabilities in the internal dataset (Figure 4E), providing preliminary support for the model performance; however, given the limited sample size of the training cohort, the possibility of overfitting cannot be fully excluded.

To enhance clinical interpretability, the risk score was incorporated into a nomogram that integrates gene expression features with predicted disease probability (Figure 4F). This tool provides a visual framework for estimating individual calcification risk based on molecular signatures.

Expression profiling of the four genes further validated their biological relevance. CAV1 and JAK1 displayed progressive upregulation across disease stages, whereas IGF1 and IL7R showed stage-dependent downregulation, with all comparisons reaching statistical significance (P<0.01) (Figure 4G). These opposing patterns reflect distinct regulatory roles of these genes in calcification progression and are consistent with their selection by the LASSO algorithm.

Overall, the integration of LASSO feature selection, risk scoring, performance validation, and expression profiling supports CAV1, IGF1, IL7R, and JAK1 as candidate molecular indicators associated with aortic valve calcification and provides an exploratory framework for disease stratification that requires further validation in larger cohorts.

External validation confirms the discriminative power and calibration performance of the risk model

To assess the generalizability of the four-gene risk model, we applied it to an independent validation dataset GSE51472. Consistent with the findings from the training cohort, the calculated risk scores demonstrated a clear gradient across control, sclerotic, and calcified valve tissues, with scores increasing progressively along the disease continuum. Group comparisons showed significant differences in risk score distributions (Kruskal-Wallis, P=0.004), and pairwise analyses confirmed higher scores in sclerotic and calcified valves relative to controls (Figure 5A).

Figure 5 External validation confirms the discriminative power and calibration performance of the risk model. (A) Risk score distribution across control, sclerotic, and calcified valve samples in the external validation dataset. (B) ROC curve showing the discriminative performance of the model in the validation cohort. (C) Calibration curve evaluating agreement between predicted and observed probabilities. *, P<0.05; **, P<0.01. AUC, area under the curve; AV, aortic valve; ROC, receiver operating characteristic.

The model retained strong discriminatory ability in the validation cohort, achieving an AUC of 0.875 for distinguishing calcified from non-calcified valves (Figure 5B). Although slightly lower than the performance in the training set, the AUC remained relatively high, providing preliminary evidence of model reproducibility in an independent dataset. Calibration analysis further showed that the predicted probabilities closely matched the observed proportions, with the bias-corrected curve aligning well with the ideal diagonal line (Figure 5C). This result suggests that the model provides not only good classification accuracy but also reliable probability estimation in external samples.

Collectively, these validation results provide preliminary support that the four-gene risk model maintains discriminatory performance and acceptable calibration in an independent dataset, while larger cohorts are still required to confirm its generalizability and reduce the potential influence of overfitting.

Immune infiltration analysis reveals microenvironmental remodeling and its association with the four-gene signature

To characterize immune microenvironment remodeling during aortic valve calcification, we estimated the proportions of 22 immune cell types across control, BAV, and SAV samples using CIBERSORT. The overall distribution of immune infiltration exhibited marked shifts along disease progression (Figure 6A). Calcified valves contained higher proportions of innate immune cells—including macrophages and dendritic cells—whereas several lymphocyte-related populations showed relative reductions. This pattern indicates a gradual transition from immune homeostasis to an inflammation-dominant state as calcification advances.

Figure 6 Immune infiltration analysis reveals microenvironmental remodeling and its association with the four-gene signature. (A) Immune cell composition across control, BAV, and SAV samples. (B) Correlation heatmap showing coordinated relationships among immune cell subsets. (C) Differential abundance of immune cells across groups. (D) Positive correlations between hub genes and immune cell subsets. *, P<0.05; **, P<0.01 vs. control and SAV groups. BAV, bicuspid aortic valve; SAV, stenotic aortic valve.

Correlation analysis among immune cell subsets revealed a structured, modular organization within the immune landscape (Figure 6B). Multiple T-cell subsets formed a tightly correlated cluster, representing coordinated adaptive immune activity. Similarly, macrophage M0, M1, and M2 subsets showed strong positive intercorrelations, consistent with a continuum of macrophage polarization in diseased tissue. Activated and resting dendritic cells, together with neutrophils, formed another positively correlated module. These coordinated clusters suggest that immune activation in calcified valves is not random, but reflects systematic reorganization centered on inflammation amplification and antigen presentation.

Group comparisons of immune cell abundance further identified subsets most strongly associated with disease severity (Figure 6C). Macrophages M0 were markedly increased in the BAV group compared with controls and SAV (P<0.01), indicating early recruitment and accumulation of undifferentiated macrophages. Resting dendritic cells showed slightly higher proportions in diseased valves than in controls (P<0.05), consistent with enhanced antigen-presentation capacity. In contrast, macrophages M2 were significantly reduced in BAV/SAV compared with controls (P<0.05), suggesting that the reparative macrophage compartment is relatively depleted as calcification progresses. Monocytes were significantly elevated in SAV valves compared to controls (P<0.05), supporting their role as a circulating source for tissue macrophages in the diseased microenvironment.

We next evaluated the relationship between the four hub genes (CAV1, IGF1, IL7R, JAK1) and immune infiltration patterns (Figure 6D). All significant associations were positive correlations, indicating synchronized changes between gene expression and immune activation. Specifically: JAK1 showed a positive correlation with monocytes (P<0.05), suggesting JAK-STAT involvement in monocyte recruitment or activation; IL7R correlated positively with resting CD4 memory T cells (P<0.05), consistent with its role in T-cell homeostasis; IGF1 correlated positively with macrophages M2 (P<0.05), implying a link with reparative or remodeling-related macrophage functions; CAV1 showed positive correlations with resting mast cells and plasma cells (both P<0.05), suggesting potential roles in mast cell stabilization or humoral immune modulation.

Collectively, these findings demonstrate that aortic valve calcification is accompanied by pronounced immune-cell compositional shifts and network reorganization. The coordinated positive associations between the four-gene signature and key immune subsets further support their biological relevance, linking PCD-related genes to the immune microenvironment and suggesting a potential association between PCD-related gene expression and immune microenvironment remodeling during valve calcification.

qRT-PCR validation evaluates the expression patterns of the four-gene signature in peripheral blood leukocyte samples

To preliminarily evaluate the translational relevance of the transcriptional patterns observed in valve tissue datasets, we quantified the expression of CAV1, IGF1, IL7R, and JAK1 in peripheral blood leukocyte samples from healthy controls, patients with mild aortic valve sclerosis, and patients with severe aortic valve stenosis (Figure 7A-7D). All four genes demonstrated disease-stage-related mRNA expression trends consistent with the bioinformatic findings, supporting their potential relevance as accessible peripheral blood molecular indicators.

Figure 7 qRT-PCR validation evaluates the expression patterns of the four-gene signature in peripheral blood leukocyte samples. qRT-PCR validation of CAV1 (A), IGF1 (B), IL7R (C), and JAK1 (D) expression in peripheral blood leukocytes from healthy controls, patients with mild aortic valve sclerosis, and patients with severe aortic valve stenosis. **, P<0.01; ***, P<0.001 vs. control group. qRT-PCR, quantitative real-time polymerase chain reaction.

Among the upregulated genes, CAV1 expression increased progressively across disease stages (P<0.001), with the highest levels observed in peripheral blood leukocytes from patients with severe aortic valve stenosis (Figure 7A). This stepwise elevation aligns with its role as a risk-enhancing gene in the model and may reflect systemic molecular alterations associated with disease progression. Similarly, JAK1 expression was significantly elevated in peripheral blood leukocytes from patients with mild aortic valve sclerosis (P<0.01) and further increased in patients with severe aortic valve stenosis (P<0.001) (Figure 7D), suggesting its potential association with systemic inflammatory changes during disease progression.

Conversely, both downregulated genes showed a marked decline with increasing disease severity. IGF1 expression decreased substantially in peripheral blood leukocytes from patients with mild aortic valve sclerosis (P<0.001) and was further reduced in patients with severe aortic valve stenosis (P<0.001) (Figure 7B), suggesting a peripheral blood expression trend consistent with the tissue transcriptomic analysis. IL7R also displayed a disease-stage-related decrease in peripheral blood leukocytes, with significant reductions in patients with mild aortic valve sclerosis (P<0.01) and severe aortic valve stenosis (P<0.001) (Figure 7C), suggesting altered immune-related transcriptional patterns in peripheral blood.

Collectively, these qRT-PCR results provide preliminary mRNA-level peripheral blood evidence supporting the translational relevance of the four-gene signature. The consistent pattern—CAV1 and JAK1 upregulated, IGF1 and IL7R downregulated with increasing disease severity—supports the potential utility of this signature as an accessible peripheral blood molecular indicator of disease progression, while further tissue-level and protein-level validation remains necessary.

Discussion

This study systematically investigated the role of PCD in the progression of CAVD by integrating transcriptomic analysis with clinical validation. Through a multi-level analytical framework involving differential expression analysis, PCD pathway scoring, WGCNA, and time-series clustering, we identified four key genes—CAV1, IGF1, IL7R, and JAK1—whose expression levels were significantly associated with disease severity. The four-gene risk score model constructed based on these features demonstrated promising but exploratory discriminatory ability in both the training and validation datasets. Moreover, qRT-PCR results from peripheral blood leukocyte samples showed stepwise mRNA expression patterns of these genes that were broadly consistent with the tissue transcriptomic findings. These findings suggest that PCD-related transcriptomic features are associated with the progression from early sclerosis to advanced valve calcification and may serve as candidate molecular indicators for disease stratification.

At the transcriptomic level, we observed increased enrichment of four PCD-related gene sets—including apoptosis, pyroptosis, autophagy, and ferroptosis—along the disease continuum, suggesting that PCD-associated transcriptional patterns are altered during CAVD progression (13). However, because GSVA-based pathway scores may also capture overlapping inflammatory, oxidative stress, and cellular stress responses, these results should be interpreted as association-based evidence rather than direct proof of increased PCD activity. For instance, CAV1 was significantly upregulated in severely calcified valves. Prior studies have shown that CAV1 is associated with osteogenic differentiation of VICs through lipid raft-mediated modulation of TGF-β/BMP signaling pathways (19,20). JAK1 also exhibited progressive elevation, which may reflect transcriptional changes related to inflammatory cascades via the JAK-STAT signaling axis, consistent with its known role in monocyte activation and immune regulation (21). In contrast, the gradual decline of IGF1 expression may represent an association with reduced protective signaling, given its known anti-apoptotic, antioxidative, and metabolic homeostatic functions in cardiovascular tissue (22,23). Similarly, the reduced expression of IL7R, a key receptor for T cell homeostasis, may suggest a possible association with altered adaptive immune regulation, a phenomenon reported in other chronic inflammatory conditions (24). These findings support a hypothesis-generating model in which PCD-related transcriptional alterations are associated with degenerative and inflammatory remodeling in the valvular microenvironment.

In addition, we observed significant remodeling of the local immune microenvironment in calcified aortic valves. Specifically, there was a progressive increase in undifferentiated M0 macrophages and monocytes, alongside a marked decrease in reparative M2 macrophages, indicating a potential shift toward a pro-inflammatory state. Previous studies have emphasized the importance of macrophage polarization in valve calcification, especially the imbalance between M1 and M2 phenotypes (25,26). However, few have addressed the accumulation of M0 macrophages during early disease stages. Our findings may complement this body of work by highlighting the potential role of precursor immune cell recruitment in the initiation of valvular inflammation. This suggests that monocyte recruitment and precursor macrophage loading may precede and condition subsequent immune polarization.

Importantly, all four signature genes demonstrated positive correlations with specific immune cell subtypes, such as JAK1 with monocytes, CAV1 with mast cells and plasma cells, and IGF1 with M2 macrophages. Some of these associations are consistent with existing literature—for example, JAK1 has been implicated in leukocyte activation (21), and IGF1 has been linked to tissue repair via M2 macrophage regulation (27). However, the correlation between CAV1 and mast cells/plasma cells has not been widely reported in CAVD.

This novel association may suggest a previously unrecognized link between caveolar signaling and humoral or mast cell-mediated immune processes. Given CAV1’s role in membrane trafficking and signal transduction, it may influence vesicle-mediated mediator release in mast cells or antibody responses via plasma cells. These hypotheses merit experimental validation.

Based on these observations, we cautiously propose a putative association between PCD-related transcriptomic alterations and immune microenvironment remodeling in CAVD. The correlations between the four-gene signature and immune cell subsets may reflect coordinated inflammatory and stress-related transcriptional programs rather than direct regulatory interactions. Although such interactions have been suggested in cancer and autoimmune diseases, direct evidence in valvular pathology remains limited. The present study provides preliminary support for this concept, but further functional experiments are required to elucidate its biological relevance and causal significance in CAVD progression.

Unlike previous studies that have focused on isolated PCD pathways (e.g., pyroptosis or ferroptosis), our study is, to our knowledge, the first to integrate multiple PCD modalities and link them to both gene expression and immune infiltration patterns in CAVD. This multi-dimensional framework enabled the identification of a four-gene transcriptomic signature associated with disease stage and immune microenvironmental remodeling. The model showed preliminary calibration and discrimination in both internal and external datasets, and its mRNA expression patterns were observed in peripheral blood leukocytes, suggesting potential translational relevance that requires further validation.

Nevertheless, this study has several limitations. First, both the transcriptomic training cohort and the qRT-PCR validation cohort were relatively small. which may increase the risk of model overfitting even with cross-validation. Although external validation was performed, larger multi-center cohorts are needed to assess the model’s generalizability, stability, and clinical utility. Second, qRT-PCR validation was performed in peripheral blood leukocytes rather than valve tissues. Therefore, these results provide only preliminary peripheral evidence for the translational relevance of the four-gene signature and cannot directly confirm its tissue-level expression. Future studies using matched valve tissue and blood samples are needed. Third, our analyses and qRT-PCR validation were performed at the transcriptomic and mRNA levels; therefore, it remains unclear whether changes in gene expression translate into protein abundance or functional activity. Future studies should include protein-level assays, such as Western blotting, immunohistochemistry, or enzyme-linked immunosorbent assay (ELISA), to further validate the four-gene signature. Fourth, as an observational study, the associations identified do not imply causation. Confounding factors such as age, comorbidities, and medication use may influence gene expression. Finally, heterogeneity in the GEO datasets—arising from differences in sample origin, technical platforms, and geographic background—may introduce bias and affect reproducibility.

In summary, this study highlights the potential contribution of PCD activation and immune remodeling to the pathogenesis of aortic valve calcification. The four-gene signature—CAV1, IGF1, IL7R, and JAK1—provides a promising molecular tool for disease stratification and mechanistic exploration. If validated in larger cohorts and experimental models, these findings may inform the development of early diagnostic strategies or therapeutic interventions targeting PCD-immune interactions in CAVD, thereby advancing the prospects of precision medicine in valvular heart disease.

Conclusions

This study identified PCD-related transcriptomic features associated with the progression of aortic valve calcification and proposed an exploratory four-gene signature—CAV1, IGF1, IL7R, and JAK1—that reflected disease severity in the analyzed datasets and clinical samples. By integrating molecular profiling with immune infiltration analysis, we found that increased enrichment of PCD-related gene sets was accompanied by immune microenvironment remodeling in calcified valves. These findings provide association-based evidence for a potential link between PCD-related transcriptional alterations and CAVD progression.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0273/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0273/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0273/prf

Funding: This study was supported by National Health Commission Capacity Building and Continuing Education Center (No. GWJJMB202510021075).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0273/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The Third Affiliated Hospital of Sun Yat-sen University (No. A2025-525-01). Informed consent was waived in this retrospective study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kraler S, Blaser MC, Aikawa E, et al. Calcific aortic valve disease: from molecular and cellular mechanisms to medical therapy. Eur Heart J 2022;43:683-97. [Crossref] [PubMed]
2. Sengeløv M, Cheng S, Biering-Sørensen T, et al. Ideal Cardiovascular Health and the Prevalence and Severity of Aortic Stenosis in Elderly Patients. J Am Heart Assoc 2018;7:e007234. [Crossref] [PubMed]
3. Otto CM, Prendergast B. Aortic-valve stenosis--from patients at risk to severe valve obstruction. N Engl J Med 2014;371:744-56. [Crossref] [PubMed]
4. He Y, Yang YJ, Wang ZJ, et al. Bibliometric analysis of treatment modalities in calcific aortic valve stenosis. Front Pharmacol 2025;16:1431311. [Crossref] [PubMed]
5. Chong T, Lan NSR, Courtney W, et al. Medical Therapy to Prevent or Slow Progression of Aortic Stenosis: Current Evidence and Future Directions. Cardiol Rev 2024;32:473-82. [Crossref] [PubMed]
6. Moncla LM, Briend M, Bossé Y, et al. Calcific aortic valve disease: mechanisms, prevention and treatment. Nat Rev Cardiol 2023;20:546-59. [Crossref] [PubMed]
7. Blaser MC, Kraler S, Lüscher TF, et al. Multi-Omics Approaches to Define Calcific Aortic Valve Disease Pathogenesis. Circ Res 2021;128:1371-97. [Crossref] [PubMed]
8. Qin Z, Bäck M. Advances in single-cell transcriptomics: unraveling the pathogenesis of calcific aortic valve disease. Cardiovasc Res 2026;121:2838-59. [Crossref] [PubMed]
9. Issa N, Blot G, Candellier A, et al. Macrophages as key modulators of calcific aortic valve disease. Front Cardiovasc Med 2025;12:1664067. [Crossref] [PubMed]
10. Broeders W, Bekkering S, El Messaoudi S, et al. Innate immune cells in the pathophysiology of calcific aortic valve disease: lessons to be learned from atherosclerotic cardiovascular disease? Basic Res Cardiol 2022;117:28. [Crossref] [PubMed]
11. Li M, Wang ZW, Fang LJ, et al. Programmed cell death in atherosclerosis and vascular calcification. Cell Death & Disease 2022;13:467.
12. Li XZ, Xiong ZC, Zhang SL, et al. Potential ferroptosis key genes in calcific aortic valve disease. Front Cardiovasc Med 2022;9:916841. [Crossref] [PubMed]
13. Zhao Y, Linkermann A, Takahashi M, et al. Ferroptosis in cardiovascular disease: regulatory mechanisms and therapeutic implications. Eur Heart J 2025;46:3247-60. [Crossref] [PubMed]
14. Yu C, Zhang Y, Yang L, et al. Identification of pyroptosis-associated genes with diagnostic value in calcific aortic valve disease. Front Cardiovasc Med 2024;11:1340199. [Crossref] [PubMed]
15. Bian W, Wang Z, Sun C, et al. Pathogenesis and Molecular Immune Mechanism of Calcified Aortic Valve Disease. Front Cardiovasc Med 2021;8:765419. [Crossref] [PubMed]
16. Liu J, Hong M, Li Y, et al. Programmed Cell Death Tunes Tumor Immunity. Front Immunol 2022;13:847345. [Crossref] [PubMed]
17. Xiong M, Chen Z, Tian J, et al. Exosomes derived from programmed cell death: mechanism and biological significance. Cell Commun Signal 2024;22:156. [Crossref] [PubMed]
18. Shi B, Phan TK, Poon IKH. Extracellular vesicles from the dead: the final message. Trends Cell Biol 2025;35:439-52. [Crossref] [PubMed]
19. Hiepen C, Mendez PL, Knaus P. It Takes Two to Tango: Endothelial TGFβ/BMP Signaling Crosstalk with Mechanobiology. Cells 2020;9:1965. [Crossref] [PubMed]
20. Halloran D, Durbano HW, Nohe A. Bone Morphogenetic Protein-2 in Development and Bone Homeostasis. J Dev Biol 2020;8:19. [Crossref] [PubMed]
21. Giryes S, Wong C, Bridgewood C, et al. Paradoxical Activation of Entheseal Myeloid Cells by JAK1 and Tyk2 Inhibitors via Interleukin-10 Antagonism. Arthritis Rheumatol 2025;77:1373-86. [Crossref] [PubMed]
22. Obradovic M, Zafirovic S, Soskic S, et al. Effects of IGF-1 on the Cardiovascular System. Curr Pharm Des 2019;25:3715-25. [Crossref] [PubMed]
23. Kasprzak A. Insulin-Like Growth Factor 1 (IGF-1) Signaling in Glucose Metabolism in Colorectal Cancer. Int J Mol Sci 2021;22:6434. [Crossref] [PubMed]
24. Chen D, Tang TX, Deng H, et al. Interleukin-7 Biology and Its Effects on Immune Cells: Mediator of Generation, Differentiation, Survival, and Homeostasis. Front Immunol 2021;12:747324. [Crossref] [PubMed]
25. Li G, Qiao W, Zhang W, et al. The shift of macrophages toward M1 phenotype promotes aortic valvular calcification. J Thorac Cardiovasc Surg 2017;153:1318-1327.e1. [Crossref] [PubMed]
26. Dabravolski SA, Popov MA, Maltseva ON, et al. Exploring Macrophage Contribution to Vascular Calcification in Atherosclerosis. IUBMB Life 2025;77:e70064. [Crossref] [PubMed]
27. Chi Y, Jiang H, Yin Y, et al. Macrophage Signaling Pathways in Health and Disease: From Bench to Bedside Applications. MedComm (2020) 2025;6:e70256.