Abstract

   Preeclampsia (PE) and endometrial cancer (EC) are two distinct
   conditions that share common genetic and molecular mechanisms involving
   immune dysregulation, endothelial dysfunction, and angiogenesis. This
   study aimed to investigate the potential genetic links between PE and
   EC, identify key prognostic genes, and develop a risk model to predict
   overall survival in EC patients. We conducted comprehensive genetic and
   molecular analyses, revealing significant overlaps in immune and
   angiogenic pathways between PE and EC. Through LASSO regression and
   multivariate Cox analysis, we identified five core prognostic
   genes—FSTL3, PRSS23, IGFBP4, MYDGF, and TSC22D3—that were used to
   construct a risk model. This model effectively stratified EC patients
   into high- and low-risk groups, with significant differences in overall
   survival. Patients in the low-risk group exhibited better 1-, 3-, and
   5-year survival outcomes and had higher immune cell infiltration and
   expression of immune checkpoint-related genes, indicating a more
   favorable tumor microenvironment. Additionally, the analysis showed
   that these genes are also implicated in the pathogenesis of PE,
   highlighting potential shared molecular mechanisms. Our findings
   suggest that these PE-related genes may serve as valuable prognostic
   biomarkers for EC and could lead to improved prognostic tools and
   personalized treatment strategies for EC patients.

   Keywords: Preeclampsia, Endometrial cancer, Immune dysregulation,
   Angiogenesis, Prognostic biomarkers

Introduction

   Preeclampsia (PE) is a pregnancy-specific hypertensive disorder that
   affects 5–8% of pregnancies and is a leading cause of maternal and
   fetal morbidity and mortality worldwide [[30]1, [31]2]. PE is
   characterized by new-onset hypertension and proteinuria after 20 weeks
   of gestation, and it can lead to severe complications, including
   eclampsia, placental abruption, organ failure, and even maternal and
   fetal death [[32]3, [33]4]. The exact cause of PE is still not fully
   understood, but it is believed to be a multifactorial condition
   involving abnormal placentation, endothelial dysfunction, and immune
   dysregulation [[34]5]. There is growing evidence that genetic factors
   play a significant role in the development of PE, with numerous studies
   suggesting a genetic predisposition to the disorder.

   On the other hand, endometrial cancer (EC) is the most common
   gynecological malignancy in developed countries and the second most
   common worldwide [[35]6, [36]7]. It primarily affects postmenopausal
   women, but a growing number of cases are being diagnosed in younger
   women, especially those with a history of obesity, hypertension, and
   diabetes [[37]8]. Like PE, EC has complex etiological factors that
   include hormonal imbalances, metabolic disturbances, and genetic
   predispositions [[38]9]. The relationship between PE and EC has been a
   topic of increasing interest, as both conditions share several common
   risk factors, including obesity and metabolic syndrome, and appear to
   involve similar pathways of endothelial and immune dysregulation
   [[39]10, [40]11].

   PE and EC, though occurring at different life stages—PE during
   pregnancy and EC typically postmenopause—may share pathophysiological
   mechanisms involving vascular dysfunction, immune regulation, and
   angiogenesis [[41]12–[42]14]. In PE, abnormal placentation and poor
   spiral artery remodeling lead to hypoxia and release of antiangiogenic
   factors such as sFlt-1 and sEng, contributing to endothelial
   dysfunction and clinical manifestations like hypertension and
   proteinuria [[43]15–[44]17]. Similarly, in EC, dysregulated
   angiogenesis, marked by overexpression of VEGF, promotes tumor growth
   and metastasis [[45]18–[46]20]. Immune dysregulation is critical in
   both conditions; PE involves an exaggerated inflammatory response with
   increased IL-6 and TNF-α, leading to placental injury [[47]21, [48]22],
   while EC features immune cell infiltration, including TAMs and Tregs,
   facilitating immune evasion [[49]23]. These parallels in immune
   dysregulation suggest common genetic pathways that may influence
   inflammation and immune responses in both conditions.

   The genetic basis of PE has been the subject of extensive research,
   with studies identifying several candidate genes and genetic variants
   associated with the disorder. These include genes involved in immune
   regulation, endothelial function, and angiogenesis. For example,
   polymorphisms in genes such as FLT1, which encodes for VEGF receptor 1,
   have been implicated in the development of PE [[50]24]. Additionally,
   maternal and fetal genetic interactions are thought to influence the
   risk of developing PE, with fetal genetic factors contributing to
   abnormal placental development and maternal factors modulating immune
   responses to the fetus [[51]25]. Similarly, genetic predisposition
   plays a crucial role in the development of EC. Studies have identified
   several genetic mutations and variants associated with increased risk
   of EC, including mutations in DNA mismatch repair genes (MLH1, MSH2,
   MSH6) and oncogenes such as PTEN, PIK3CA, and KRAS [[52]26, [53]27].
   These genetic mutations contribute to the development of endometrial
   hyperplasia, a precursor to EC, and are involved in key signaling
   pathways that regulate cell growth, apoptosis, and angiogenesis.

   Emerging evidence suggests that the genetic mechanisms underlying PE
   and EC may overlap, particularly in the regulation of angiogenesis and
   immune responses. Genes such as VEGF and its receptors, which are
   critical for both placental development and tumor growth, may
   contribute to both conditions. Additionally, alterations in
   immune-regulatory genes that influence inflammation may predispose
   individuals to both PE and EC [[54]28]. Women with a history of PE
   often have underlying metabolic conditions like obesity, insulin
   resistance, and hypertension, all of which are risk factors for EC. The
   endothelial and immune dysfunction observed in PE may have long-term
   effects, potentially promoting endometrial tumorigenesis. Furthermore,
   shared genetic predispositions may link the two conditions, increasing
   the likelihood of developing EC later in life. Studies also show that
   women with a history of PE have a higher prevalence of metabolic
   syndrome and cardiovascular disease, which are associated with chronic
   inflammation and dysregulated metabolism, further contributing to the
   risk of EC [[55]29].

   Despite the growing recognition of the potential link between PE and
   EC, there is still much to be learned about the genetic and molecular
   mechanisms underlying this association. Identifying shared genetic
   pathways between PE and EC could provide new insights into the etiology
   of both conditions and lead to the development of novel biomarkers for
   early detection and risk stratification. Moreover, understanding how
   the immune system is dysregulated in both conditions could open up new
   avenues for therapeutic interventions aimed at modulating immune
   responses and reducing the risk of both PE and EC. In this context, our
   study seeks to explore the genetic mechanisms that may link PE and EC,
   focusing on immune-related genes and pathways involved in inflammation,
   endothelial function, and angiogenesis. By integrating genetic data
   from patients with a history of PE and EC, we aim to identify common
   genetic variants and molecular signatures that contribute to the
   development of both conditions. This research could have important
   implications for the prevention and management of PE and EC,
   particularly in identifying women at higher risk and developing
   targeted therapies that address the underlying genetic and immune
   dysregulation shared by both conditions.

Methods

Data collection

   For this study, we gathered multiple datasets to investigate the
   genetic links between PE and EC. Single-cell RNA sequencing (scRNA-seq)
   data related to EC were obtained from the GEO database ([56]GSE173682),
   which includes five EC samples sequenced using the 10X Genomics
   platform. For bulk RNA-seq data, we selected samples from the TCGA-UCEC
   cohort, which included 23 adjacent normal samples and 554 EC samples
   used for differential expression analysis. Additionally, 522 patients
   with complete clinical and survival information were used for subgroup
   classification and the construction of the risk model. In relation to
   PE, we sourced data from the [57]GSE114691 dataset, which consists of
   18 control and 18 PE samples. To validate the findings externally, we
   utilized the IMvigor210, [58]GSE78220, [59]GSE135222, and [60]GSE91061
   datasets, particularly in the context of immune therapy responses and
   survival analyses.

Data processing

   For the PE data, we performed differential expression analysis using
   the “limma” package, with the selection criteria set as log fold change
   (logFC) greater than 2 or less than -2, and an adjusted p-value less
   than 0.05. For scRNA-seq data, we utilized Seurat version 4.2.2 for
   analysis. After reading the dataset, we performed quality control by
   excluding cells with nFeature_RNA counts less than 500 or greater than
   5000, cells with a mitochondrial gene content exceeding 20%, and those
   with hemoglobin gene (HB) expression greater than 3. The NormalizeData
   function was used for data normalization. Dimensionality reduction and
   clustering were conducted using the top 20 principal components (PCs)
   and a resolution of 0.02, resulting in seven clusters. The RunUMAP
   function was applied to compute and visualize the clustering results.
   Cell types were annotated using SingleR and validated with previous
   literature. To identify marker genes for each cell type, we employed
   the FindAllMarkers function. Lastly, the Harmony package was used to
   correct for batch effects across samples, ensuring uniformity for
   downstream analyses.

PE scoring

   The PE-related differentially expressed genes were converted into a GMT
   file for further analysis. ScRNA-seq data were used to calculate PE
   scores using five different methods: AUCell, UCell, singscore,
   AddModuleScore, and average expression values. The PE scores were
   compared across different cellular subpopulations to identify
   differences in PE scoring levels. Based on the median PE score, cells
   were divided into high-PE and low-PE scoring groups. Differentially
   expressed genes between the high- and low-PE groups were identified
   using the FindMarkers function.

Prognostic network of PE-related genes

   To analyze PE-related genes associated with prognosis, we used the
   “limma” package to perform differential expression analysis of PE
   score-related genes in EC samples from the TCGA database. Univariate
   Cox regression analysis was performed to assess the prognostic
   significance of these differentially expressed genes in EC patients
   (p < 0.05). The maftools package was employed to analyze the mutation
   frequency of these prognostic genes in EC, along with their chromosomal
   locations and copy number variations (CNVs).

Subtype analysis based on PE-related genes

   Using the expression of prognostic genes identified from PE scoring,
   the ConsensusClusterPlus package was applied to classify EC patients
   into two subtypes. The expression levels of prognostic genes in the two
   subtypes were compared, and boxplots were generated using ggplot2 to
   visualize the differences. The Survival package was used to analyze the
   survival outcomes of patients in the different subtypes. Additionally,
   the single-sample gene set enrichment analysis (ssGSEA), a method used
   to estimate the enrichment of gene sets in individual samples, was used
   to calculate immune cell infiltration scores. The estimate package,
   which assesses the proportion of immune and stromal components in tumor
   samples, was employed to compute immune scores. These scores were then
   compared between the two subtypes to assess differences. Finally, GSVA
   was used to perform enrichment analysis between the subtypes.

Prognostic model construction

   Based on the differentially expressed genes identified, we used LASSO
   (Least Absolute Shrinkage and Selection Operator) regression, a method
   for variable selection and regularization that helps identify the most
   significant core genes by minimizing overfitting, for model
   development. EC patients were randomly divided into training and
   validation sets in a 6:4 ratio. A multivariate Cox regression analysis
   was performed to construct a prognostic model, where the risk score was
   calculated as the sum of each prognostic gene's expression multiplied
   by its risk coefficient. Based on the median risk score, samples were
   classified into high- and low-risk groups. The Survival and survminer
   packages were used for survival analysis and to generate survival
   curves, while the timeROC package was used to plot ROC curves to assess
   the predictive accuracy of the model.

Immune-related analysis

   ssGSEA was employed to calculate immune cell infiltration scores and
   immune function scores for EC samples, allowing for the assessment of
   immune activity at the single-sample level. Boxplots and correlation
   scatterplots were generated using the ggpubr package to visualize the
   data. The TIDE website ([61]http://tide.dfci.harvard.edu/) was used to
   evaluate immune exclusion and dysfunction. Additionally, the IPS
   website was utilized to estimate the potential efficacy of immune
   checkpoint inhibitor treatments. The TCGAbiolinks package was used to
   download SNP data for EC patients, and the maftools package was applied
   to calculate tumor mutational burden (TMB) scores for each patient.
   Boxplots were generated using the ggpubr package to illustrate the
   expression levels of immune checkpoint molecules.

Drug sensitivity analysis

   We preliminarily compared the survival differences and immunotherapy
   responses in different risk groups within the IMvigor210 cohort.
   Survival differences between risk groups in the [62]GSE78220 and
   [63]GSE135222 datasets were also analyzed, as well as the treatment
   sensitivity in the [64]GSE91061 dataset. To predict drug sensitivity,
   the pRRophetic algorithm was used to calculate IC50 values, which were
   compared between risk groups. The IC50 values represent the sensitivity
   of patients to chemotherapy and targeted therapies, with higher IC50
   values indicating lower sensitivity to these treatments. The
   differences in IC50 values between the high- and low-risk groups were
   analyzed to evaluate potential differences in drug sensitivity.

Results

Differential gene analysis in PE reveals 52 genes

   In the differential analysis of PE, we identified 52 differentially
   expressed genes, among which 5 were downregulated and 47 were
   upregulated (Fig. [65]1A). Before analyzing the PE score in endometrial
   carcinoma, we integrated and quality-controlled the single-cell data,
   retaining 31,442 cells for downstream analysis following stringent
   methodological quality control standards (Fig. [66]1B). Based on the
   ElbowPlot results, we selected the top 20 principal components for
   dimensionality reduction and clustering. With a resolution set to 0.02,
   we identified seven clusters (0–6) (Fig. [67]1C). After batch effect
   removal, the cells from different samples were evenly distributed and
   ready for further analysis (Fig. [68]1D). According to singleR and
   previous literature, the seven clusters were annotated into six cell
   types, including endothelial cells (PLVAP, PECAM1, SELE, CLDN5),
   epithelial cells (WFDC2, KRT18, KRT8, EPCAM), fibroblasts (PCOLCE, LUM,
   DCN, COL1A1), T cells (PTPRC, CCL5, CD7, CD2), macrophages (TYROBP,
   APOC1, C1QB, C1QA), and mast cells (TPSAB1, CPA3, TPSB2, GATA2)
   (Fig. [69]1E, F). To assess the differences in PE scores among cell
   types, we calculated scores using multiple algorithms and defined the
   median of four scoring methods as the PE score. The results showed that
   macrophages had significantly higher PE scores than other cell types
   (Fig. [70]1G).

Fig. 1.

   [71]Fig. 1
   [72]Open in a new tab

   Differential gene expression and clustering analysis of PE in
   endometrial cancer (EC). A Volcano plot showing the differentially
   expressed genes between PE and control samples. Genes with logFC > 2 or
   logFC < −2 and an adjusted p-value < 0.05 were considered significant.
   Highlighted are the 5 downregulated and 47 upregulated genes. B Uniform
   Manifold Approximation and Projection (UMAP) visualization of
   single-cell RNA sequencing data after integration and quality control,
   showing the distribution of cells from different EC samples. C UMAP
   plot showing clustering of cells into seven distinct clusters (0–6)
   based on dimensionality reduction and resolution 0.02. D UMAP plots
   displaying the distribution of cells before and after batch effect
   correction using Harmony, showing that cells from different samples are
   evenly distributed. E UMAP plot showing the annotation of the seven
   clusters into six cell types based on known markers: endothelial cells,
   epithelial cells, fibroblasts, T cells, macrophages, and mast cells. F
   Dot plot representing the expression levels of marker genes across the
   different cell types. G Dot plot showing the PE scores across different
   cell types calculated using four scoring methods, with macrophages
   displaying significantly higher scores compared to other cell types

PE score-related genes in uterine corpus endometrial carcinoma (UCEC)

   Among the 302 genes associated with PE scores, 106 genes exhibited
   significant differential expression in endometrial carcinoma.
   Univariate Cox analysis of these 106 genes identified 14 genes
   associated with endometrial carcinoma prognosis (Fig. [73]2A). To
   illustrate the complex relationship between PE score-related genes and
   endometrial carcinoma prognosis, we constructed a network diagram
   (Fig. [74]2B). The genomic locations of these 30 genes are displayed in
   Fig. [75]2C. Further analysis revealed widespread CNVs among these
   genes, with FSTL3 and LIMD2 showing deletions and TIMP2 showing
   amplification (Fig. [76]2D). Investigation of somatic mutation
   frequencies in these 14 genes in cervical cancer revealed that COL6A3
   had the highest mutation rate (up to 72%), while the other genes had
   lower mutation frequencies (Fig. [77]2E). Comparative analysis showed
   that most PE score-related genes were significantly upregulated in
   endometrial carcinoma tissue compared to normal tissue (Fig. [78]2F).

Fig. 2.

   [79]Fig. 2
   [80]Open in a new tab

   PE score-related genes in uterine corpus endometrial carcinoma (UCEC).
   A Forest plot of univariate Cox regression analysis showing hazard
   ratios and p-values for the 14 genes associated with EC prognosis. B
   Network diagram illustrating the relationship between PE-related genes
   and their association with risk factors (red) and favorable factors
   (green) in endometrial carcinoma prognosis. C Circos plot displaying
   the chromosomal locations of the 30 PE score-related genes. D Copy
   number variation (CNV) frequency analysis of the 14 genes, showing
   gains (red) and losses (green) across endometrial carcinoma samples. E
   Waterfall plot illustrating the somatic mutation frequencies of the 14
   PE-related genes in cervical cancer patients, with COL6A3 having the
   highest mutation rate (72%). F Boxplot comparing the expression levels
   of the 14 PE score-related genes between normal and tumor tissues,
   indicating significantly higher expression in endometrial carcinoma

Clustering of UCEC patients into C1 and C2 groups

   Using consensus clustering analysis, we divided 522 endometrial
   carcinoma patients into two groups, C1 and C2 (Fig. [81]3A–C).
   Figure [82]3D illustrates the differential expression of the 14 genes
   between the C1 and C2 groups. Kaplan–Meier analysis revealed a
   significant difference in survival rates between the C1 and C2 groups
   (P < 0.001, Fig. [83]3E), with C2 patients having a better prognosis.
   Immune cell infiltration analysis showed that C2 had higher
   infiltration of various immune cells, including regulatory T cells, CD8
   T cells, activated NK cells, and dendritic cells (Fig. [84]3F).

Fig. 3.

   [85]Fig. 3
   [86]Open in a new tab

   Consensus clustering analysis and characterization of endometrial
   carcinoma subtypes. A Consensus clustering matrix of 522 endometrial
   carcinoma patients, divided into two clusters (C1 and C2) based on gene
   expression profiles, with k = 2 as the optimal number of clusters. B
   Delta area plot showing the relative change in the area under the
   cumulative distribution function (CDF) curve, supporting k = 2 as the
   optimal cluster number. C CDF plot illustrating consensus distribution
   for different values of k, with k = 2 providing a clear separation. D
   Differential expression analysis of 14 genes between C1 and C2 groups.
   E Kaplan–Meier survival analysis showing that patients in the C2 group
   have a significantly better prognosis than those in the C1 group
   (p < 0.001). F Immune cell infiltration analysis demonstrating that the
   C2 cluster has higher infiltration of immune cells. G Tumor
   microenvironment (TME) analysis revealing that C2 has higher stromal,
   immune, and ESTIMATE scores, indicating a more favorable immune
   microenvironment. H Pathway enrichment analysis showing that C2 subtype
   is negatively associated with several signaling pathways

   Regarding the differences in the tumor microenvironment (TME) between
   the two groups, C2 exhibited higher immune and mechanism scores
   (Fig. [87]3G). Finally, enrichment analysis revealed that the C2
   subtype was negatively correlated with signaling pathways such as the
   WNT and Notch pathways (Fig. [88]3H).

Construction of a risk model using 14 prognostic genes

   Given the significant impact of the 14 PE-related prognostic genes on
   endometrial carcinoma prognosis, we performed LASSO regression analysis
   and subsequently utilized multivariate Cox regression to construct a
   risk model based on the remaining core risk genes (Fig. [89]4A, B).
   Ultimately, five genes—FSTL3, PRSS23, IGFBP4, MYDGF, and TSC22D3—were
   selected as core risk genes for the final risk model (Fig. [90]4C). In
   both the TCGA cohort and the training and validation sets, we observed
   significant differences in overall survival between the high- and
   low-risk groups stratified by risk score. Patients in the low-risk
   group consistently showed better overall survival rates. Additionally,
   the risk model effectively predicted 1-, 3-, and 5-year overall
   survival and represented characteristic information for certain
   patients (F[91]ig. [92]4D-I).

Fig. 4.

   [93]Fig. 4
   [94]Open in a new tab

   Construction of a risk model using 14 prognostic genes. A LASSO
   regression plot showing the coefficient paths for each gene in relation
   to log lambda. B Partial likelihood deviance of the LASSO regression
   model with tenfold cross-validation, identifying the optimal lambda
   value. C Coefficients of the five core prognostic genes selected for
   the final risk model (FSTL3, PRSS23, IGFBP4, MYDGF, TSC22D3). D–F
   Kaplan–Meier survival curves comparing overall survival between
   high-risk and low-risk groups in the training, validation, and TCGA
   cohorts, showing significantly better survival in the low-risk group
   (p < 0.001). G–I ROC curves evaluating the predictive accuracy of the
   risk model for 1-, 3-, and 5-year overall survival in the training,
   validation, and TCGA cohorts, demonstrating robust performance with AUC
   values above 0.7

Independent prognostic value of risk score in endometrial carcinoma

   To investigate the independent prognostic role of the risk score
   alongside other clinical features such as age, grade, and stage, we
   performed univariate and multivariate Cox regression analyses. These
   analyses demonstrated that the risk score independently predicted
   overall survival in endometrial carcinoma patients (Fig. [95]5A, B).
   Using TCGA data, we constructed a nomogram to predict 1-, 3-, and
   5-year OS, incorporating the risk score, age, grade, and stage as
   predictive parameters (Fig. [96]5C). The calibration curve, concordance
   index (C-index), and decision curve analysis (ECA) indicated good
   consistency between the predicted and actual outcomes (Fig. [97]5D–F).
   The heatmap revealed clear clustering differences between the risk
   groups in terms of stage, immune score, and stromal score
   (Fig. [98]5G). When analyzing the differences in risk scores between
   PE-related clusters, we found that the C2 subgroup had lower risk
   scores (Fig. [99]5H). Furthermore, a Sankey diagram showed that the C2
   group, characterized by lower risk scores, correlated with more
   favorable survival outcomes (Fig. [100]5I).

Fig. 5.

   [101]Fig. 5
   [102]Open in a new tab

   Independent prognostic value of risk score in endometrial carcinoma. A
   Univariate Cox regression analysis showing hazard ratios and p-values
   for clinical features (age, grade, stage, and risk score) in predicting
   overall survival. B Multivariate Cox regression analysis showing that
   the risk score independently predicts overall survival, alongside
   clinical features. C Nomogram predicting 1-, 3-, and 5-year overall
   survival based on risk score, age, grade, and stage in EC patients. D
   Calibration curves comparing observed overall survival with
   nomogram-predicted survival at 1, 3, and 5 years, indicating good
   predictive accuracy. E Concordance index (C-index) comparing risk
   score, age, grade, and stage over time, showing the highest concordance
   for the risk score. F Decision curve analysis (ECA) for the risk score
   and clinical features, showing the net benefit of the risk score in
   predicting overall survival. G Heatmap displaying clustering
   differences in risk score, age, grade, stage, immune score, and stromal
   score between high- and low-risk groups. H Violin plot showing
   significantly lower risk scores in the C2 cluster compared to the C1
   cluster. I Sankey diagram illustrating the association between
   clusters, risk groups, and survival outcomes, showing that the C2
   cluster is associated with lower risk scores and better survival

Immune characteristics of low-risk group

   In the low-risk group, most genes from the HIL family exhibited higher
   expression levels (Fig. [103]6A). Using multiple methods, including
   ssGSEA, we performed enrichment and functional analysis of immune cells
   and their functions. The results indicated that nearly all immune
   functions, except for the type II IFN response, were significantly
   enriched in the low-risk group (Fig. [104]6B). Figure [105]6C, D show
   that the infiltration of almost all immune cells was more pronounced in
   the low-risk group compared to the high-risk group. Additionally,
   immune scores indicated that the low-risk group had higher immune and
   stromal component scores (Fig. [106]6E). Finally, we compared the
   differences in immune types between risk groups and found a significant
   variation in immune types between the two groups (Fig. [107]6F).

Fig. 6.

   [108]Fig. 6
   [109]Open in a new tab

   Immune characteristics of the low-risk group. A Boxplot comparing the
   expression levels of HLA genes between high-risk and low-risk groups,
   showing higher expression in the low-risk group. B Boxplot comparing
   immune function scores between high-risk and low-risk groups, showing
   significant enrichment of immune functions (except for type II IFN
   response) in the low-risk group. C Boxplot comparing immune cell
   infiltration scores between high-risk and low-risk groups, showing
   higher infiltration of almost all immune cells in the low-risk group. D
   Violin plot comparing immune component scores (StromalScore,
   ImmuneScore, and ESTIMATEScore) between high-risk and low-risk groups,
   indicating higher scores in the low-risk group. E Heatmap displaying
   immune-related pathway enrichment analysis, comparing differences
   between high- and low-risk groups in the TCGA cohort. F Heatmap
   displaying the distribution of immune subtypes between the high-risk
   and low-risk groups, showing a significant difference in immune subtype
   composition between the two groups within the TCGA cohort

Immune therapy as a treatment option for advanced malignancies

   Immunotherapy is becoming a treatment option for patients with advanced
   malignancies. Our analysis of the TIDE results revealed that TIDE and
   Exclusion scores were lower in low-risk patients. However, we also
   found that the immune therapy response rates between high- and low-risk
   patients were similar (Fig. [110]7A, B). Previous studies suggest that
   TIDE scores are inversely correlated with the efficacy of immunotherapy
   in cancer patients. Given the potential of immune checkpoint inhibitors
   (ICIs) to block CTLA4/PD-1 interactions and treat certain tumors, we
   used IPS scores based on IFNG expression levels to evaluate the
   potential for treatment. Notably, low-risk patients exhibited higher
   IPS scores across multiple assessments (Fig. [111]7C). Microsatellite
   instability (MSI) is widely used to predict patient prognosis and
   immunotherapy response. In our study, we compared the MSI status of
   patients with different risk types and found that MSI-H patients were
   more common in the low-risk group, which further explained their better
   prognosis (Fig. [112]7D). Figure [113]7E highlights significant
   differences in the expression of checkpoint genes between the two
   groups, with higher expression observed in the low-risk group. This
   includes several well-known immunotherapy targets, such as CD274
   (programmed death-ligand 1, PD-L1) and CTLA4.

Fig. 7.

   [114]Fig. 7
   [115]Open in a new tab

   Immune therapy as a treatment option for advanced malignancies. A
   Violin plots comparing TIDE, IFNG, dysfunction, and exclusion scores
   between high- and low-risk groups, showing lower scores in the low-risk
   group. B Bar plot showing the proportion of responders and
   non-responders in the high- and low-risk groups. Boxplot comparing risk
   scores between immune therapy responders and non-responders. C Violin
   plots comparing IPS scores (based on CTLA4 and PD-1 status) between
   high- and low-risk groups, showing significantly higher scores in the
   low-risk group, indicating a greater potential for immune checkpoint
   inhibitor therapy response. D Bar plot showing the proportion of MSI-H,
   MSS, and MSI-L patients in the high- and low-risk groups. Boxplot
   comparing risk scores between patients with different MSI statuses,
   showing that MSI-H patients are more common in the low-risk group. E
   Boxplot comparing the expression levels of immune checkpoint-related
   genes between high- and low-risk groups, with higher expression
   observed in the low-risk group, including key immunotherapy targets
   such as PD-L1 and CTLA4

Tumor mutation burden and prognosis

   Tumor mutations play a critical role in influencing patient prognosis.
   As shown in Fig. [116]8A and B, the mutation frequency between high-
   and low-risk groups was similar. Additionally, Fig. [117]8C
   demonstrates that patients with high TMB had better survival rates
   compared to those with low TMB. Figure [118]8D indicates that patients
   in the high-risk group with low TMB had the worst prognosis.

Fig. 8.

   [119]Fig. 8
   [120]Open in a new tab

   Tumor mutation burden (TMB) and prognosis. A Waterfall plot showing the
   mutation frequency of common cancer-related genes in high-risk
   patients, with PTEN, PIK3CA, and ARID1A among the most frequently
   mutated genes. B Waterfall plot showing mutation frequency in low-risk
   patients, with a similar mutation profile to high-risk patients. C
   Kaplan–Meier survival curve comparing overall survival between patients
   with high and low TMB, showing significantly better survival in the
   high-TMB group (p < 0.001). D Kaplan–Meier survival curve comparing
   overall survival between high-risk and low-risk groups, stratified by
   TMB levels, showing that high-risk patients with low TMB have the worst
   prognosis (p < 0.001)

Comparative analysis of treatment outcomes between risk groups

   We compared the treatment outcomes between patients in different risk
   groups. In the IMvigor-210 cohort, we evaluated the restricted mean
   survival (RMS) at 6 and 12 months to account for the delayed clinical
   effects of immunotherapy. We also assessed differences in long-term
   survival after 3 months of treatment (P < 0.05; Fig. [121]9A, B). The
   results indicated that low-risk patients had better prognoses,
   suggesting a greater benefit from immunotherapy. Additionally, the
   distribution of risk scores across patients with different treatment
   responses showed that the responders (complete response [CR]/partial
   response [PR]) had significantly lower risk scores compared to
   non-responders (progressive disease [PD]/stable disease [SD])
   (P = 0.021, Fig. [122]9C). Consistently, in the [123]GSE78220
   (p = 0.027, Fig. [124]9D) and [125]GSE135222 (p = 0.0011, Fig. [126]9E)
   cohorts, low-risk patients demonstrated better prognoses after
   immunotherapy, and in [127]GSE91061 (p = 0.012, Fig. [128]9F), it was
   further suggested that the low-risk group tends to have better
   immunotherapy outcomes.

Fig. 9.

   [129]Fig. 9
   [130]Open in a new tab

   Comparative analysis of treatment outcomes and drug sensitivity between
   risk groups. A, B Kaplan–Meier survival curves comparing restricted
   mean survival (RMS) at 6 and 12 months between high-risk and low-risk
   groups in the IMvigor210 cohort, showing better outcomes in the
   low-risk group. C Boxplot comparing risk scores across different
   treatment response groups (PR, PD, CR, SD), with significantly lower
   risk scores in responders (CR/PR). D Kaplan–Meier survival curve from
   the [131]GSE78220 cohort, comparing overall survival between high-risk
   and low-risk groups, showing significantly better survival in the
   low-risk group (p = 0.027). E Kaplan–Meier survival curve from the
   [132]GSE135222 cohort, comparing overall survival between high-risk and
   low-risk groups, showing better survival in the low-risk group
   (p = 0.0011). F Boxplot from the [133]GSE91061 cohort comparing risk
   scores between responders (R) and non-responders (NR) to treatment,
   with significantly lower scores in the responder group (p = 0.012). G–J
   Violin plots comparing IC50 values for cisplatin, docetaxel,
   paclitaxel, and tamoxifen between high-risk and low-risk groups,
   showing significantly lower IC50 values in the low-risk group,
   indicating higher sensitivity to these drugs

Drug sensitivity analysis

   We conducted drug sensitivity analyses and compared the IC50 values of
   cisplatin, docetaxel, paclitaxel, and tamoxifen. Consistently, we found
   that the low-risk group had lower IC50 values, indicating higher
   sensitivity to these drugs (Fig. [134]9G-J).

Discussion

   In this study, we explored the potential genetic and molecular links
   between PE and EC, two conditions that share significant commonalities
   in terms of immune dysfunction, endothelial disruption, and angiogenic
   imbalance. Our findings suggest that the genetic mechanisms underlying
   PE and EC may be interconnected through pathways involving immune
   regulation, inflammation, and vascular development, offering novel
   insights into how these seemingly unrelated conditions might be related
   at a molecular level. Additionally, we constructed and validated a risk
   model based on core prognostic genes, which could predict overall
   survival in EC patients and possibly provide insights into the role of
   PE-related genes in tumorigenesis.

   The connection between PE and EC is likely underpinned by shared
   genetic factors and pathways. Both conditions involve aberrant
   angiogenesis and immune dysregulation, which are crucial for placental
   development and tumor growth. Genes regulating VEGF, immune checkpoint
   molecules, and other factors related to immune tolerance and
   inflammation are significant in the pathogenesis of both PE and EC. The
   involvement of VEGF is particularly noteworthy because it highlights
   how disruptions in angiogenic signaling can have broad implications. In
   PE, the overexpression of sFlt-1, a soluble form of the VEGF receptor,
   disrupts angiogenic balance, leading to impaired placentation and
   systemic endothelial dysfunction. Similarly, VEGF overexpression in
   endometrial tumors supports tumor angiogenesis, which is essential for
   tumor growth and metastasis. This shared role emphasizes that
   dysregulated vascular development and remodeling may predispose
   individuals to both conditions, linking pregnancy complications to
   future cancer risk.

   Immune dysfunction is another pivotal factor in both PE and EC. In PE,
   an exaggerated inflammatory response marked by increased
   pro-inflammatory cytokines such as TNF-α and IL-6 contributes to
   endothelial damage and placental injury. In EC, the TME is influenced
   by immune cells, including TAMs, Tregs, and MDSCs, which facilitate
   immune evasion and tumor progression. The observed overlap in immune
   dysregulation suggests that the immune system's inability to maintain
   balance during pregnancy and tumorigenesis may be driven by shared
   genetic predispositions or molecular triggers. These findings
   underscore the potential long-term implications of immune and
   angiogenic dysregulation, connecting the pathophysiology of PE with
   that of endometrial cancer.

   The epidemiological evidence indicating that women with a history of PE
   are at higher risk for developing EC later in life aligns with our
   genetic findings [[135]30]. PE is often associated with underlying
   metabolic disorders, such as obesity, insulin resistance, and chronic
   hypertension, all of which are established risk factors for EC. The
   shared risk profile between these two conditions suggests that
   metabolic and hormonal imbalances may contribute to a predisposition
   for both PE and EC.

   One explanation for this increased risk is the long-term vascular and
   immune effects of PE. Women who experience PE may have persistent
   endothelial dysfunction, chronic inflammation, and metabolic
   disturbances, all of which could contribute to the development of
   endometrial hyperplasia and, eventually, EC. Additionally, genetic
   factors that predispose women to PE, particularly those involving
   angiogenesis and immune regulation, may also predispose them to EC
   through similar mechanisms of immune escape, chronic inflammation, and
   dysregulated cell growth.

   Our study identified several PE-related genes that also play roles in
   EC, including FSTL3, PRSS23, and IGFBP4. These genes were found to be
   involved in both endothelial and immune processes, highlighting their
   potential as biomarkers for predicting not only the risk of developing
   PE but also the subsequent risk of EC. Understanding the dual roles
   these genes play in both conditions could help develop early detection
   strategies for EC in women with a history of PE, allowing for earlier
   intervention and potentially improved outcomes.

   One of the key contributions of this study is the construction of a
   robust risk model based on five core prognostic genes (FSTL3, PRSS23,
   IGFBP4, MYDGF, and TSC22D3), which successfully stratified EC patients
   into high- and low-risk groups. These genes were identified through
   LASSO regression and multivariate Cox regression analyses and were
   selected for their strong association with overall survival. Notably,
   these genes are also involved in processes such as immune modulation,
   angiogenesis, and cell proliferation, all of which are relevant to both
   PE and EC [[136]31–[137]33].

   The risk model demonstrated excellent predictive power for 1-, 3-, and
   5-year overall survival in EC patients across multiple cohorts,
   including both the training and validation sets. Patients classified as
   low-risk based on this model consistently showed better overall
   survival, while high-risk patients had significantly worse outcomes.
   This stratification highlights the potential for these PE-related genes
   to serve as prognostic biomarkers, providing valuable information for
   personalized treatment strategies.

   Moreover, the model's ability to predict immune-related outcomes could
   have clinical implications for immunotherapy in EC. Our analysis showed
   that low-risk patients, who had higher expression levels of immune
   checkpoint-related genes (such as CD274 and CTLA4), were more likely to
   benefit from ICIs. Given that PE involves immune dysregulation, it is
   possible that targeting immune pathways with ICIs could be an effective
   treatment strategy for a subset of EC patients, particularly those with
   genetic profiles linked to PE.

   This study also sheds light on the immune landscape of EC in the
   context of PE-related gene expression. Low-risk patients exhibited
   higher levels of immune cell infiltration, including CD8 + T cells,
   regulatory T cells, and NK cells, which are critical for anti-tumor
   immunity. In contrast, high-risk patients had lower immune infiltration
   and higher expression of immunosuppressive markers, indicating a more
   immune-evasive TME.

   These findings suggest that the immune dysregulation seen in PE might
   persist in some women, predisposing them to a more immunosuppressive
   TME in EC. The presence of immunosuppressive cells, such as Tregs and
   MDSCs, in both PE and high-risk EC patients further supports the idea
   that immune evasion mechanisms are shared between these two conditions.
   This connection underscores the importance of targeting the immune
   system in therapeutic strategies for EC, particularly in patients with
   a history of PE.

Limitations and future directions

   While this study provides valuable insights into the genetic and
   molecular connections between PE and EC, several limitations must be
   acknowledged. First, although we identified several key genes involved
   in both conditions, the precise mechanisms by which these genes
   contribute to the progression from PE to EC remain unclear. Further
   functional studies are needed to elucidate the specific roles of these
   genes in immune regulation, angiogenesis, and tumor development.
   Second, the majority of data used in this study were derived from
   retrospective cohorts, which may introduce biases related to patient
   selection and clinical outcomes. Prospective studies are needed to
   validate our findings in larger, more diverse populations and to
   explore whether these genetic markers can be used in clinical practice
   for early detection or risk stratification. Finally, the role of
   environmental and lifestyle factors, such as diet, physical activity,
   and exposure to toxins, in modulating the risk of both PE and EC was
   not addressed in this study. Future research should consider these
   factors to gain a more comprehensive understanding of the interplay
   between genetics and the environment in these conditions.

Conclusion

   In summary, this study provides compelling evidence that PE and EC
   share genetic and molecular mechanisms related to angiogenesis, immune
   regulation, and endothelial dysfunction. The identification of
   overlapping pathways between these two conditions highlights the
   potential for shared biomarkers and therapeutic targets. Our risk
   model, based on core PE-related genes, demonstrated strong prognostic
   value in EC and could guide personalized treatment approaches,
   particularly in the context of immunotherapy. Further research is
   needed to validate these findings and to explore the clinical
   applications of these genetic insights, with the ultimate goal of
   improving outcomes for women affected by both PE and EC.

Acknowledgements