The transcriptomic dataset GSE60438 (platform GPL6884) was retrieved from the Gene Expression Omnibus (GEO) database (Yong et al., 2015). This dataset was selected for its inclusion of clinically defined severe preeclampsia/eclampsia samples. For specific details, please refer to Supplementary Table S1. Probe IDs were mapped to official gene symbols using the ID Conversion tool available in SangerBox (http://sangerbox.com/). Differential expression analysis between the eclampsia patient group (N = 25) and the normotensive control group (N = 23) was performed using the limma R package. To control for false discoveries in this high-dimensional analysis, the resulting p-values were adjusted using the Benjamini-Hochberg false discovery rate (FDR) method. Genes meeting the thresholds of FDR-adjusted p-value <0.05 and an absolute log2 fold change (|log2FC|) > 1 were defined as significantly differentially expressed genes (DEGs). All data processing and statistical analyses were implemented in R, with data visualization conducted using the ggplot2 and pheatmap packages.

To identify hub genes associated with eclampsia pathogenesis, we performed WGCNA on the BioWinford platform (http://biowinford.site/). The gene expression data were normalized and filtered, focusing on the top 5000 most variable genes. A weighted neighbor-joining matrix was constructed based on gene correlations, and appropriate soft-thresholding was applied to ensure a scale-free network. Hierarchical clustering was used to divide the network into modules, which were then correlated with clinical features using Pearson’s coefficients. Hub genes were identified by high connectivity within modules, as determined by module membership (MM) and gene significance (GS). Co-expression networks were visualized using heatmaps and network diagrams to highlight key hub genes. Statistical analyses were performed in R, with P values <0.05 considered significant.

To explore the biological functions and pathways associated with eclampsia, we conducted functional annotation and visualization using the BioWinford Platform. DEGs were subjected to Gene Ontology (GO) enrichment analysis, with a minimum of 3 overlapping genes and a p-value ≤0.01 considered statistically significant. Additionally, Gene Set Enrichment Analysis (GSEA) was performed to assess the differential expression of genes in eclampsia, focusing on GO terms. Ranked gene lists, based on logFC or t-values, were input into GSEA, with gene set sizes ranging from 15 to 500 and 1000 permutations. A significance threshold of p < 0.05 was applied.

To assess immune infiltration in the control and eclamptic patient groups, we used the xCell algorithm on the BioWinford Platform, which provides detailed enrichment analysis of 64 immune and stromal cell types. The analysis was based on normalized gene expression data, and the results were visualized using box plots to compare immune infiltration levels. Statistical significance was assessed using the Wilcoxon rank-sum test, with p < 0.05 considered significant.

LASSO (Least Absolute Shrinkage and Selection Operator) regression was employed to refine the feature set derived from WGCNA, selecting the most parsimonious and discriminative genes associated with eclampsia. This analysis was conducted using the BioWinfordMR online platform (Wang et al., 2025). The optimal penalty parameter (λ) was determined through 10-fold cross-validation, which minimized the mean cross-validated error and thereby prevented model overfitting. This process resulted in a simplified model that retained only the most critical predictive features.

Eclampsia patients and healthy controls were recruited from the Department of Obstetrics. Eclampsia was diagnosed according to international guidelines (ISSHP/ACOG): new-onset tonic-clonic seizures in a woman with preeclampsia (hypertension ± proteinuria/organ dysfunction after 20 weeks of gestation), after excluding other causes. Exclusion of alternative seizure etiologies (e.g., epilepsy, stroke, metabolic disorders) was performed through detailed history, neurological examination, cranial MRI, and consultation with a neurologist. For specific details, please refer to Supplementary Table S2.

Clinical samples, including serum and placental tissues, were collected from patients diagnosed with eclampsia and healthy controls after obtaining informed consent. The collection of these samples was approved by the institutional review board (IRB) under ethical approval number 20230006. For serum collection, blood samples were drawn from participants and centrifuged at 3,000 rpm for 10 min to separate the serum, which was stored at −80 °C until further analysis. Placental tissues were obtained during delivery, immediately placed in liquid nitrogen, and stored at −80 °C.

Total RNA was extracted from placental tissues and serum samples using TRIzol reagent (Thermo Fisher Scientific, Catalog No. 15596026, United States) according to the manufacturer’s instructions. cDNA synthesis was performed with 1 µg of RNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Catalog No. K1622, United States). Quantitative PCR was carried out using SYBR Green Master Mix (Thermo Fisher Scientific, Catalog No. 4367659, United States) on an Applied Biosystems 7500 Real-Time PCR System (Thermo Fisher Scientific, United States). Primers for DPP4, IL6, TNF-α, and GAPDH were designed as follows: DPP4 (Forward: 5′-CTC​CTG​CGG​AGG​AGA​GTC​A-3′, Reverse: 5′-GAG​CCA​TGG​AAG​GAG​GAG​A-3′); IL6 (Forward: 5′-TGA​GGA​GAC​TTG​CCT​GGT​G-3′, Reverse: 5′-AAG​GAG​GAG​ACT​TGA​GGA​AGG-3′); TNF-α (Forward: 5′-CCC​TGA​AAG​GAG​ACG​AGA​C-3′, Reverse: 5′-GAG​GAG​GAA​GGA​GGA​GAG-3′); and GAPDH (Forward: 5′-GGA​GCC​AAA​CGG​GTC​ATC​T-3′, Reverse: 5′-GAC​AGT​CAG​CCG​CAT​CTT​CT-3′).

Proteins were extracted from placental tissues and serum samples using RIPA buffer (Thermo Fisher Scientific, Catalog No. 89901, United States), supplemented with protease and phosphatase inhibitors (Thermo Fisher Scientific, Catalog No. A32959, USA). Protein concentrations were determined using the BCA Protein Assay Kit (Thermo Fisher Scientific, Catalog No. 23225, United States). Equal amounts of protein (30 µg) were separated by 10% SDS-PAGE (Bio-Rad Laboratories, Catalog No. 4561033, United States) and transferred to PVDF membranes (Millipore, Catalog No. IPVH00010, United States) using a Trans-Blot Turbo Transfer System (Bio-Rad Laboratories, United States). Membranes were blocked with 5% non-fat milk (Thermo Fisher Scientific, United States) in TBST for 1 h at room temperature, followed by overnight incubation at 4 °C with primary antibodies: DPP4 (Abcam, Catalog No. ab133605, UK), IL6 (Cell Signaling Technology, Catalog No. 12153, United States), TNF-α (Cell Signaling Technology, Catalog No. 3707, USA), GAPDH (Cell Signaling Technology, Catalog No. 2118, United States), Caspase-1 (Cell Signaling Technology, Catalog No. 2267, United States), p-Caspase-1 (p17) (Cell Signaling Technology, Catalog No. 3866, United States), ASC (Santa Cruz Biotechnology, Catalog No. sc-514414, United States), p-ASC (PhosphoSolutions, Catalog No. 352-P-ASC, United States), NLRP3 (Abcam, Catalog No. ab214185, UK), p-NLRP3 (Cell Signaling Technology, Catalog No. 12291, United States), p65 (Cell Signaling Technology, Catalog No. 8242, United States), and p-p65 (Cell Signaling Technology, Catalog No. 3033, United States). After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies (Cell Signaling Technology, Catalog No. 7076, United States) for 1 h at room temperature. Protein bands were visualized using the Chemiluminescent HRP Substrate Kit (Bio-Rad Laboratories, Catalog No. 1705061, United States) and detected using a ChemiDoc Imaging System (Bio-Rad Laboratories, United States). Band intensities were quantified using ImageJ software (NIH, USA), and protein levels were normalized to GAPDH. Relative protein expression was calculated by comparing the target protein intensities between experimental and control groups.

Human Umbilical Vein Endothelial Cells (HUVECs) and the Human Trophoblast Cell Line (HTR-8) were obtained from ATCC (USA). HUVECs were cultured in Endothelial Cell Growth Medium-2 (EGM-2) (Lonza, Catalog No. CC-3162, United States), while HTR-8/SVneo cells were cultured in RPMI-1640 medium (Thermo Fisher Scientific, Catalog No. 11875093, USA), both supplemented with 10% fetal bovine serum (FBS) (Gibco, Catalog No. 16140-071, United States) and 1% penicillin-streptomycin (Thermo Fisher Scientific, Catalog No. 15140122, United States) under standard culture conditions (37 °C, 5% CO₂). For DPP4 gene knockdown, cells were transfected with siRNA targeting DPP4 (si-DPP4#1 and si-DPP4#2), using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific, Catalog No. 13778075, United States). The siRNA sequences were designed as follows: si-DPP4#1 (5′-GGA​GGA​GAG​GAA​GAG​AAU-3′) and si-DPP4#2 (5′-GGA​AGA​GGA​GGA​AGA​GAA​U-3′), and were transfected at a final concentration of 50 nM for 48 h.

To induce an inflammatory response, cells were treated with lipopolysaccharide (LPS) (Sigma-Aldrich, Catalog No. L2880, United States) at 1 μg/mL or TNF-α (Sigma-Aldrich, Catalog No. T0157, United States) at 20 ng/mL for 24 h. For the Transwell migration assay, Transwell inserts (Corning, Catalog No. 3422, United States) with 8 µm pores were used. Cells (1 × 10⁴) were seeded in the upper chamber in serum-free medium, and the lower chamber was filled with complete medium containing 10% FBS as a chemoattractant. After 24 h of incubation, the cells on the lower surface of the membrane were fixed with 4% paraformaldehyde (Sigma-Aldrich, Catalog No. 158127, USA) and stained with Crystal Violet (Sigma-Aldrich, Catalog No. C3886, United States). For the Transwell invasion assay, the procedure was similar, but the upper chamber of the Transwell inserts was pre-coated with Matrigel (Corning, Catalog No. 354234, United States) at 50 μg/mL for 2 h at 37 °C. After 48 h of incubation, invaded cells were fixed with 4% paraformaldehyde and stained with Crystal Violet. The number of migrated and invaded cells was quantified by counting the cells in five random fields under a light microscope (Olympus, Japan). All assays were performed in triplicate, and the data were expressed as the average number of cells per field.

Cells were plated in 6-well plates at a density of 5 × 10⁵ cells per well. Once cells reached confluence, a sterile 1000 µL pipette tip was used to create a uniform scratch (wound) across the center of the well. After scratching, cells were washed twice with phosphate-buffered saline (PBS) (Thermo Fisher Scientific, Catalog No. 10010023, United States) to remove any detached cells and then incubated in serum-free medium for 24 h to allow wound closure. The migration of cells into the wound area was observed and photographed at different time points (0, 12, and 24 h) under a light microscope (Olympus, Japan).

Growth Factor Reduced Matrigel (Corning, Catalog No. 354230, United States) was added to a 96-well plate and incubated at 37 °C for 30 min to allow gel solidification. After pre-treatment with lipopolysaccharide (LPS) (Sigma-Aldrich, Catalog No. L2880, United States) at 1 μg/mL or TNF-α (Sigma-Aldrich, Catalog No. T0157, USA) at 20 ng/mL for 24 h, HUVECs (5 × 10⁴ cells per well) were seeded onto the solidified Matrigel and incubated for 6–8 h at 37 °C in a humidified incubator with 5% CO₂. Tube formation was observed and photographed under a light microscope (Olympus, Japan).

Statistical analyses were performed using R (version 4.3.1). For experimental validation, data are presented as the mean ± SEM of at least three independent biological replicates unless otherwise specified. The Student’s t-test was applied for comparisons between two groups with normally distributed data, whereas the Wilcoxon rank-sum test was used for non-normally distributed data, which are presented as medians with interquartile ranges. The specific sample size (n) for each experiment, representing the number of biologically independent samples or replicates, is stated in the corresponding figure legends. The results were considered statistically significant with the following p-value thresholds: p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***).

To identify hub genes associated with the progression of eclampsia, we performed WGCNA. First, to explore the differential gene expression in eclampsia and its correlation with clinical traits, we generated a clustering plot comparing eclampsia samples with normal controls. The results showed clear clustering with no significant outliers (Figure 1A). A scale-free co-expression network was then constructed using a soft threshold of 4, selected based on scale independence and average connectivity (Figures 1B,C). Additionally, we created a hierarchical clustering dendrogram based on gene correlations, which revealed a total of nine gene modules (Figure 1D). These modules were color-coded for identification (Figure 1E). Finally, we assessed the correlation between the gene modules and clinical traits. The red module exhibited the highest positive correlation with eclampsia (r = 0.32, p = 0.02), while the magenta module showed the strongest negative correlation (r = −0.43, p = 0.002) (Figure 1F).

To identify DEGs between the eclamptic and control groups, we analyzed the transcriptome data using R and Perl with the limma package. The thresholds for selecting DEGs were set at P < 0.05 and log |FC| > 1. For visualization, we used the ggplot2 and pheatmap packages. The heatmap analysis revealed a total of DEGs, including 4,642 upregulated genes and 2,193 downregulated genes in the eclamptic group compared to the control group (Figure 2).

To investigate the biological functions and pathways associated with DEGs in eclampsia, we performed GO and GSEA. GO enrichment analysis revealed that eclampsia DEGs were significantly associated with chromosomal regions, mitochondrial protein-containing complexes, chromosomes, telomeric regions, ribosomal large subunits, catalytic step 2 spliceosomes, organelle ribosomes, mitochondrial ribosomes, and DNA repair complexes, among other relevant processes (Figure 3A).

Further GSEA-based GO analysis identified key enrichments in processes such as cell activation, cell killing, cytokine production, and the regulation of cytokine production (Figure 3B). KEGG pathway analysis using GSEA highlighted enrichments in chemokine signaling, hematopoietic cell line development, natural killer (NK) cell-mediated cytotoxicity, osteoblast differentiation, and rheumatoid arthritis (Figure 3C). Additionally, Reactome pathway analysis identified enrichment in antimicrobial peptides, immune system cytokine signaling, hemostasis, and neutrophil degranulation (Figure 3D).

To assess the impact of DEGs on immune cell populations in eclampsia, we used the xCELL algorithm to explore immune cell involvement in disease progression. The results revealed significant differences in the correlation of endothelial cells, erythrocytes, keratinocytes, macrophages, megakaryocytes, monocytes, neurons, and preadipocytes between the eclampsia and normal control groups (Figure 4).

To identify key genes associated with the progression of eclampsia, we performed a λ-based LASSO regression analysis (Figure 5A), which identified DPP4 as a critical gene involved in disease progression (Figure 5B). Subsequently, we collected serum and placental samples from both eclamptic patients and normal controls to assess DPP4 expression through qRT-PCR and WB assays. We found that DPP4 was significantly overexpressed in the placental tissues of eclamptic patients (Figure 5C), and a similar upregulation was observed in their serum samples (Figure 5D). Additionally, WB analysis revealed a significant increase in DPP4 protein levels in placental tissues from eclamptic patients compared to controls (Figure 5E).

Building on our previous bioinformatics analysis, which highlighted the activation of inflammatory responses in eclampsia, we further examined the expression of pro-inflammatory cytokines IL-6 and TNF-α. Both mRNA and protein levels of IL-6 and TNF-α were found to be significantly elevated in the placental tissues of eclamptic patients (Figures 5F,G). Similarly, Western blot analysis confirmed the significant upregulation of IL-6 and TNF-α proteins in the placental tissues of these patients (Figures 5H,I).

To investigate the potential biological functions and mechanisms of DPP4 in eclampsia progression, we developed a cellular model of eclampsia-induced inflammation. Cells were stimulated with lipopolysaccharide (LPS) to induce an inflammatory response. Concurrently, we used siRNA technology to knock down DPP4 and assessed the expression levels of inflammatory cytokines IL-6 and TNF-α.

Our results showed that the mRNA expression of DPP4 was significantly elevated following the induction of the inflammatory response, and this increase was effectively suppressed upon DPP4 knockdown using siRNA (Figure 6A). Similarly, mRNA levels of the pro-inflammatory cytokines IL-6 and TNF-α were markedly elevated after the inflammatory response was induced, and these levels were significantly reduced upon DPP4 inhibition (Figures 6B,C).

At the protein level, DPP4 expression was also significantly upregulated following the induction of the inflammatory response, and this upregulation was reversed following DPP4 knockdown (Figure 6D). In parallel, the protein levels of IL-6 and TNF-α were significantly increased after the induction of inflammation but were substantially reduced upon DPP4 inhibition (Figure 6E). These changes were statistically significant, reinforcing the role of DPP4 in modulating the inflammatory response.

Following successful inhibition of DPP4 in HTR-8 cells, we evaluated the impact on cell biological functions using Transwell migration, invasion, scratch, and lumen formation assays.

In the Transwell migration assay, we observed a significant reduction in the migratory capacity of HTR-8 cells upon DPP4 inhibition, compared to the control group (Figures 7A,B). Similarly, in the Transwell invasion assay, DPP4 inhibition led to a marked decrease in the invasion ability of HTR-8 cells (Figures 7C,D). In addition, a scratch assay further confirmed that DPP4 inhibition significantly reduced the migratory potential of HTR-8 cells at both 0 and 24 h, relative to the control group (Figures 7E,F). Finally, we assessed the lumen-forming ability of HTR-8 cells through a lumen formation assay. DPP4 inhibition resulted in a significant restoration of the cells’ lumen-forming capacity compared to the control group (Figure 7G).

To investigate the mechanisms underlying the effects of DPP4 on biological functions in an inflammatory response model of HTR-8 cells, we examined the protein expression and phosphorylation levels of key components in the p65/NLRP3/ASC/Caspase-1 signaling pathway. Our results demonstrated that phosphorylation levels of p65, NLRP3, ASC, and Caspase-1 were significantly reduced following DPP4 depletion, compared to the model group (Figure 8A). Statistical analysis of grey-scale data further confirmed that these changes were significant (Figure 8B).