Genetic variation perspective reveals potential drug targets for subtypes of endometrial cancer.

The primary proteomic data analyzed in this study were derived from the deCODE Genetics study dataset, which identified 4,907 plasma proteins from GWASs involving 35,559 Icelandic participants 21 . To enhance the robustness of our findings, we utilized two independent protein GWAS datasets for replication: the UKBPPP and the Gudjonsson A study. From the UKBPPP study, we selected 2923 plasma protein information from 54,219 European participants 22 . Instrumental variables (IVs) were extracted from the Gudjonsson A study, which included 5361 European participants and data on 2091 plasma proteins 23 (Sup. Table 1). The EC data were obtained from O’Mara TA’s GWAS meta-analysis, comprising 12,906 EC cases and 108,979 controls of European ancestry, sourced from 13 studies 24 . For histological subtype analysis, the data were stratified into 8758 endometrioid carcinoma cases with 46,126 controls, and 1230 non-endometrioid carcinoma cases with 35,447 controls (Sup. Table 1). Colocalization analysis was conducted to identify shared causal variants between protein expression and EC within specific genomic regions. We performed Bayesian colocalization analysis using the COLOC package (V.5.2.3) to ascertain the probability that both traits share a causal variant. The posterior probabilities were categorized as follows: PPH0 indicates no association with either protein trait; PPH1 is associated with protein expression only; PPH2 is linked to EC risk but not with protein expression; PPH3 suggests both traits are associated, but not at the same locus; and PPH4 indicates both traits share a common causal variant at the same locus 25 . The sum of these posterior probabilities is equals one. Plasma proteins with PPH3 + PPH4 > 0.8 or PPH4 > 0.5 were considered strong candidates for colocalization. MR is a natural experiment that leverages genetic variation to investigate the causal relationships between exposures and outcomes. MR addresses the limitations of traditional observational studies, being less prone to confounding and reverse causality bias, offering reliability between observational studies and intervention trials 26 . Summary-data-based MR (SMR), a method that focuses on the causal relationship between gene or protein expression and disease, was employed in this study. The analysis was conducted using the SMR software with default parameters, and IVs with minor allele frequencies (MAFs) below 0.01 were excluded to minimize bias from rare mutations. The SMR analysis focused on identifying drug targets for EC and its two subtypes (endometrioid carcinoma and non-endometrioid carcinoma) using the deCODE genetics database. Heterogeneity in dependent instruments (HEIDI) test was applied to SMR results. A p-value greater than 0.05 indicated that the association of proteins with EC was not due to linkage disequilibrium, suggesting no significant association. Multiple comparisons were adjusted for using the false discovery rate (FDR), and the results were validated through the UKBPPP and Gudjonsson A database. Protein interactions between therapeutic targets for EC were further explored using the GeneMANIA tool. Drug-gene interactions were then examined using the DGIdb database. PheWAS was performed to assess potential side effects and pleiotropy not captured in the SMR analyses. Gene-phenotype correlations with a p-value below 1e − 8 were considered significant. To identify translational-level drug targets for EC, differential expression analyses (mRNA and protein) were conducted using the UALCAN database for the key targets. Additionally, we obtained RNA-seq data and corresponding clinical information of EC patients from the TCGA database. Patients were categorized into high and low-expression groups based on the median expression level of IGF2R and CST3. Kaplan-Meier survival curves were employed to evaluate the predictive prognostic value of these targets with hazard ratios (HRs) calculated through the Cox proportional hazard model. A total of 51 paraffin-embedded endometrioid carcinoma tissue samples, 25 non-endometrioid carcinoma samples (including 18 endometrial serous carcinoma, 5 clear-cell carcinoma, and 2 carcinosarcomas), 20 normal endometrial tissue samples were obtained from patients who underwent surgery between October 2022 and May 2024 at the Department of Obstetrics and Gynaecology of Jingjiang People’s Hospital. The clinical data were sourced from medical records, and none of the patients had received radiotherapy, chemotherapy, or biological treatments before surgery. The study was approved by the Ethics Committee of Jingjiang People’s Hospital based on the declaration of Helsinki (2023-KY-019-01). Informed consent was obtained from all participants or their legal guardians. RNA was extracted from 24 endometrioid carcinomas and adjacent normal tissues, as well as from 13 non-endometrioid carcinomas and adjacent normal tissues. Total RNA was isolated using a TIANGEN (Beijing, China) extraction kit. Reverse transcription was conducted with a TaKaRa (Japan) kit, and PCR amplification was performed using the PrimeScript RT reagent Kit with gDNA Eraser (TaKaRa, Japan). Primers were synthesized by Sangon Biotech (Shanghai, China), with sequences provided in Sup. Table 2. Relative RNA expression was quantified with β-actin as an internal reference. Each sample was analyzed in triplicate, with three replicate wells per analysis, to ensure accuracy and minimize variance. For the IHC, tissue sections from endometrioid carcinomas and normal controls were incubated overnight at 4 °C with an anti-IGF2R antibody (1:100 dilution, Cat No. 20253-1-AP, Proteintech, China). Similarly, sections from non-endometrioid carcinomas and normal controls were incubated with an anti-CST3 antibody (1:100 dilution, Cat No. 12245-1-AP, Proteintech, China) under the same conditions. Following this, the samples were treated with a secondary antibody derived from the primary antibody (1:200 dilution, G1213, Servicebio, China), stained using DAB, and counterstained with hematoxylin. Two senior pathologists independently evaluated the samples in question using a double-blind manner. For each case, at least five high-power fields of view containing a minimum of 200 cells per field were examined. The assessment followed a two-tier scoring system: (1) Based on the percentage of positive cells, scores were assigned as follows: 0 points are assigned for no positive cells, 1 point for less than 10% positivity, 2 points for 10–50% positivity, 3 points for 50–75% positivity, and 4 points for more than 75% positivity. (2) Based on staining intensity, scores were assigned as 0 points for no color, 1 point for yellow, 2 points for brown-yellow, and 3 points for dark brown. The total score is obtained by multiplying the two scores. A score of greater than five was classified as high expression, while a score of five or less indicated low expression. Additionally, the study analyzed the correlation between IGF2R and CST3 expression and clinical-pathological characteristics of different EC subtypes. Statistical analyses were conducted using R software (version 4.2.1) and GraphPad Prism 9 (GraphPad, Dotmatics, MA). Wilcoxon matched pairs analyzed the RT-qPCR data signed rank test, and IHC scores were analyzed using the Mann-Whitney test. The clinical data were evaluated using Pearson’s Chi-square or Fisher’s Exact test. A p-value of less than 0.05 was considered the threshold for statistical significance.

The overall study design is illustrated in Fig.  1 , with detailed data sources and methodologies outlined below. Briefly, we utilized pQTL data from three extensive proteomics studies and employed Bayesian co-localization analysis, SMR, and HEDI test to elucidate their causal relationship with EC through both discovery and replication analyses. The identified potential targets were further examined for protein interaction analysis, druggability assessment, and drug toxicity evaluation. Ultimately, the principal drug targets were validated for their expression levels in EC. Fig. 1 Flowchart of the study design in our study. TSMR two-sample Mendelian randomization, SMR summary data-based Mendelian randomization analysis, HEIDI heterogeneity in the dependent instrument, pQTL protein quantitative trait loci. Flowchart of the study design in our study. TSMR two-sample Mendelian randomization, SMR summary data-based Mendelian randomization analysis, HEIDI heterogeneity in the dependent instrument, pQTL protein quantitative trait loci. Bayesian colocalization analysis identified 82 plasma proteins associated with EC, 75 with endometrioid carcinoma, and 13 with non-endometrioid carcinoma all potentially sharing a causal variant (Sup. Tables 3–5). To strengthen the colocalization findings, five candidate drug targets for EC, seven for endometrioid carcinomas, and seven for non-endometrioid carcinomas (Sup. Tables 6–8) passed SMR analysis ( p   0.05). The results revealed that the high expression of CBR3 (OR = 1.150; 95%CI 1.058–1.250; p  = 2.097 × 10 − 2 ), GSTO1 (OR = 1.073; 95%CI 1.032–1.117; p  = 1.096 × 10 − 2 ), and IGF2R (OR = 1.128; 95%CI 1.047–1.216; p  = 2.481 × 10 − 2 ), as well as the low expression of HHIP (OR = 0.689; 95%CI 0.544–0.872; p  = 2.511 × 10 − 2 ) and MMP10 (OR = 0.831; 95%CI 0.730–0.905; p  = 6.410 × 10 − 3 ) were positively correlated with an increased risk of EC (Fig.  2 ). For endometrioid carcinoma, ACAP2 (OR = 3.997; 95%CI 1.681–9.506; p  = 1.599 × 10 − 2 ), CBR3(OR = 1.175; 95% CI 1.065–1.297; p  = 1.446 × 10 − 2 ), GSTO1 (OR = 1.100; 95% CI 1.050–1.152; p  = 2.044 × 10 − 3 ), and IGF2R (OR = 1.165; 95% CI 1.067–1.272; p  = 1.046 × 10 − 2 ) were identified as risk proteins contributing to disease progression. Conversely, HHIP (OR = 0.625; 95%CI 0.469–0.831; p  = 1.446 × 10 − 2 ), MMP10 (OR = 0.762; 95%CI 0.669–0.868; p  = 2.044 × 10 − 3 ), and TLR2 (OR = 0.211; 95%CI 0.073–0.609; p  = 2.589 × 10 − 2 ) were found to reduce the risk of endometrioid carcinomas (Fig.  2 ). In non-endometrioid carcinoma, high expression of FSTL5 (OR = 4.142; 95%CI 1.375–12.474; p  = 1.811 × 10 − 2 ) and MAPK9 (OR = 3.070; 95%CI 1.695–5.561; p  = 2.352 × 10 − 3 ) was significantly associated with increased risk, while CST3 (OR = 0.523; 95%CI 0.339–0.804; p  = 7.010 × 10 − 3 ), DNAJB14 (OR = 0.069; 95%CI 0.014–0.346; p  = 6.371 × 10 − 3 ), GMPR2 (OR = 0.547; 95%CI 0.371–0.808; p  = 1.010 × 10 − 3 ), IFI16 (OR = 0.062; 95%CI 0.010–0.378; p  = 7.010 × 10 − 3 ), and NEO1 (OR = 0.464; 95%CI 0.274–0.786; p  = 7.873 × 10 − 3 ) appeared to serve as protective factors (Fig.  2 ). Fig. 2 The discovery phase uses SMR to search for therapeutic targets for EC (endometrioid and non-endometrioid carcinoma). Discovery dataset deCODE genetics database. OR odds ratio, CI confidence interval, HEIDI heterogeneity in the dependent instrument. The discovery phase uses SMR to search for therapeutic targets for EC (endometrioid and non-endometrioid carcinoma). Discovery dataset deCODE genetics database. OR odds ratio, CI confidence interval, HEIDI heterogeneity in the dependent instrument. In replication analysis of the UK Biobank Pharma Proteomics Project (UKBPPP) and Gudjonsson A database, nine proteins (GSTO1, CBR3, IGF2R, MMP10, MAPK9, GMPR2, CST3, DNAJB14, and NEO1) were successfully validated with consistent directionality, in line with the primary analysis (Fig.  3 ). Notably, the results from all three databases provided evidence supporting IGF2R as a risk factor for endometrioid carcinoma and CST3 as a protective factor for non-endometrioid carcinoma. Fig. 3 Robustness evaluation of multi-data source replication analysis and SMR replication analysis. Replication datasets: UKBPPP and Gudjonsson A datasets. The highlighted background indicates therapeutic targets supported by the primary and replication analysis. Robustness evaluation of multi-data source replication analysis and SMR replication analysis. Replication datasets: UKBPPP and Gudjonsson A datasets. The highlighted background indicates therapeutic targets supported by the primary and replication analysis. We used GeneMANIA to construct protein-protein interactions (PPI) networks for the potential therapeutic targets, identifying only two interactions (co-expression and genetic interactions) (Fig.  4 A). Our study identified the potential for co-expression among IGF2R, GSTO1, and CST3. Furthermore, we observed that IGF2R may engage in genetic interactions with MAPK9, while CST3 may interact genetically with NEO1 (Sup. Table 9). Fig. 4 PPI network, druggability evaluation, and PheWas results of drug targets. ( A ) The PPI network of potential therapeutic targets. ( B ) Search for drug candidates related to potential drug targets. ( C ) Manhattan plot for IGF2R. ( D ) Manhattan plot for CST3. (Different colors represent different phenotypes, with each point or triangle representing a specific phenotype. Upward-facing triangles indicate that an increase in gene expression level is associated with an increased risk of the corresponding phenotype. When a point lies above the dashed line corresponding to the suggested threshold, the PheWAS result exceeds the recommended significance level). PheWAS phenome-wide association study. PPI network, druggability evaluation, and PheWas results of drug targets. ( A ) The PPI network of potential therapeutic targets. ( B ) Search for drug candidates related to potential drug targets. ( C ) Manhattan plot for IGF2R. ( D ) Manhattan plot for CST3. (Different colors represent different phenotypes, with each point or triangle representing a specific phenotype. Upward-facing triangles indicate that an increase in gene expression level is associated with an increased risk of the corresponding phenotype. When a point lies above the dashed line corresponding to the suggested threshold, the PheWAS result exceeds the recommended significance level). PheWAS phenome-wide association study. In the druggability evaluation, 12 proteins (CBR3, GSTO1, HHIP, IGF2R, MMP10, TLR2, CST3, FSTL5, GMPR2, IFI16, MAPK9, and NEO1) were predicted as druggable targets (Fig.  4 B). To explore drug interactions, the DGIdb tool identified 43 drugs targeting five of these proteins (CBR3, IGF2R, TLR2, CST3, MAPK9). Two proteins, TLR2 and MAPK9, have been identified as promising drug development targets. Tomaralimab, which targets TLR2, is currently in clinical phase 1 trials. It plays a crucial role in the immune system and shows potential for treating central nervous system (CNS) disorders, including Parkinson’s disease. Meanwhile, several MAPK9 inhibitors have been developed for the treatment of inflammatory diseases and tumors. For example, Bentamapimod has completed a phase 2 trial of endometriosis, and Tanzisertib has been investigated in multiple clinical trials, including for idiopathic pulmonary fibrosis. In contrast, clinical trials of CC-401 on myeloid leukemia have stalled in phase 1 (Sup. Table 10). Since most drugs are administered via the bloodstream, we applied PheWas to IGF2R and CST3 which had the strongest supporting evidence, to assess potential pleiotropic and toxic side effects. Among the 17 phenotypic associations linking IGF2R and CST3 to protein expression modulation (Fig.  4 ), IGF2R appears to influence the expression of proteins such as CSTO1, LGMN, CPVL, TPP1, and NPC2. In contrast, CST3 predominantly affects cystatin C expression, potentially leading to abnormal laboratory test outcomes. A preliminary safety assessment of these drug targets was conducted using PheWas analysis, with significant findings summarized in Sup. Table 11. However, further validation through animal models and clinical trials is required. Further analysis demonstrated that mRNA and protein expression levels of IGF2R were higher in endometrioid carcinoma tissues compared to normal tissues (Fig.  5 A and C), and patients with higher IGF2R expression had a poorer progression-free survival (PFS) (Fig.  5 E). In contrast, there was no significant difference in the mRNA and protein levels of CST3 in non-endometrioid histology (Fig.  5 B and D). However, CST3 appeared to have a protective effect on overall survival (OS) in EC (Fig.  5 H). Nevertheless, a trend was observed in which high IGF2R expression correlated with worse OS in EC patients (Fig.  5 G), while high CST3 expression was associated with better PFS, though this was not statistically significant (Fig.  5 F). Fig. 5 Differential and prognosis analysis of drug targets IGF2R and CST3 in endometrial carcinoma. ( A , B ) IGF2R and CST3 mRNA expression levels in EC patients matched adjacent normal samples. ( C , D ) IGF2R and CST3 protein expression levels in EC patients matched adjacent normal samples. ( E – H ) KM plots showing IGF2R and CST3’s (mRNA) influence on PFS and OS. PFS progress-free survival, OS overall survival. Differential and prognosis analysis of drug targets IGF2R and CST3 in endometrial carcinoma. ( A , B ) IGF2R and CST3 mRNA expression levels in EC patients matched adjacent normal samples. ( C , D ) IGF2R and CST3 protein expression levels in EC patients matched adjacent normal samples. ( E – H ) KM plots showing IGF2R and CST3’s (mRNA) influence on PFS and OS. PFS progress-free survival, OS overall survival. We validated the differential expression of IGF2R and CST3 in EC samples through RT-qPCR and IHC. RT-qPCR results showed that IGF2R was up-regulated in endometrioid carcinomas ( p  = 0.0001) (Fig.  6 A), while CST3 was down-regulated in non-endometrioid carcinomas ( p  = 0.0302) (Fig.  6 B). Similarly, the IHC results showed that IGF2R was significantly overexpressed in both early and advanced-stage endometrioid carcinoma compared to normal samples ( p  = 0.0026, p  < 0.0001) (Fig.  6 E-G), while CST3 was down-expressed in advanced-stage non-endometrioid carcinoma than in normal samples ( p  = 0.0081) (Fig.  6 H-J). IHC scores were detailed in Sup. Tables 12–14. Additionally, we observed that IGF2R expression was higher in advanced-stage endometrioid carcinoma than in early-stage ( p  = 0.0092). When examining the correlation between IGF2R expression and CST3 expression with clinicopathological features, including age, diabetes mellitus type 2, hypertension, FIGO stage, myometrial invasion, lymph node metastasis, and p53 mutation status, we found that IGF2R expression was significantly associated with the FIGO stage ( p  = 0.018) and myometrial invasion ( p  = 0.002) (Table  1 ). We observed that among the 51 endometrioid carcinomas, those with advanced-stage disease and lymph node metastasis all exhibited high IGF2R expression. Similarly, CST3 expression was significantly related to the FIGO stage ( p  = 0.043) (Table  2 ). Among the 25 non-endometrioid carcinomas, 8 cases were advanced-stage, of which 7 cases exhibited low CST3 expression and 1 case exhibited high expression. Fig. 6 Expression of IGF2R and CST3 in subtypes of EC. ( A ) Differences in IGF2R mRNA between endometrioid carcinoma and adjacent normal tissues. ( B ) Differences in CST3 mRNA between non-endometrioid carcinoma and adjacent normal tissues. ( C ) Differences in IGF2R IHC scores between endometrioid carcinoma and normal endometrial samples. ( D ) Differences in CST3 IHC scores between non-endometrioid carcinoma and normal endometrial samples. ( E – G ) The immunohistochemical staining of IGF2R protein in different stages of endometrioid carcinomas and normal endometrial samples (100x and 400x magnification, respectively). ( H – J ) The immunohistochemical staining of CST3 protein in different stages of non-endometrioid carcinomas and normal endometrial samples (100x and 400x magnification, respectively). Wilcoxon matched pairs analyzed the RT-qPCR data signed rank test, and IHC scores were analyzed using the Mann-Whitney test. Data are expressed as means with SEM. Expression of IGF2R and CST3 in subtypes of EC. ( A ) Differences in IGF2R mRNA between endometrioid carcinoma and adjacent normal tissues. ( B ) Differences in CST3 mRNA between non-endometrioid carcinoma and adjacent normal tissues. ( C ) Differences in IGF2R IHC scores between endometrioid carcinoma and normal endometrial samples. ( D ) Differences in CST3 IHC scores between non-endometrioid carcinoma and normal endometrial samples. ( E – G ) The immunohistochemical staining of IGF2R protein in different stages of endometrioid carcinomas and normal endometrial samples (100x and 400x magnification, respectively). ( H – J ) The immunohistochemical staining of CST3 protein in different stages of non-endometrioid carcinomas and normal endometrial samples (100x and 400x magnification, respectively). Wilcoxon matched pairs analyzed the RT-qPCR data signed rank test, and IHC scores were analyzed using the Mann-Whitney test. Data are expressed as means with SEM. Table 1 Clinicopathological characteristics of endometrioid histological patients with IGF2R expression. Characteristics IGF2R p -value High expression Low expression n 31 20 Age, mean ± sd 60.323 ± 11.957 55.35 ± 8.0412 0.108 Diabetes mellitus type 2, n (%) 0.838 Yes 5 (71.4%) 2 (28.6%) No 26 (59%) 18 (41%) Arterial hypertension, n (%) 0.469 Yes 9 (69.2%) 4 (30.8%) No 22 (57. 9%) 16 (42.1%) Tumor stage (FIGO), n (%) 0.018 I 21 (51.2%) 20 (48.8%) II 3 (100%) 0 (0%) III 7 (100%) 0 (0%) Histological grading, n (%) 0.145 G1 17 (53.1%) 15 (46.9%) G2-G3 14 (73.7%) 5 (26.3%) Myometrial invasion, n (%) 0.002 No 2 (33.3%) 4 (66.7%) Shallow 16 (50%) 16 (50%) Deep 13 (100%) 0 (0%) Lymph node metastasis, n (%) 0.254 Yes 4 (100%) 0 (0%) No 27 (57.4%) 20 (42.6%) p53 1 Wild 30 (60%) 20 (40%) Mutation 1 (100%) 0 (0%) Clinicopathological characteristics of endometrioid histological patients with IGF2R expression. Table 2 Clinicopathological characteristics of non-endometrioid histological patients with CST3 expression. Characteristics CST3 p -value High expression Low expression n 7 18 Age, mean ± sd 65.714 ± 5.4685 67.611 ± 6.6078 0.507 Diabetes mellitus type 2, n (%) 0.274 Yes 0 (0%) 5 (100%) No 7 (35%) 13 (65%) Arterial hypertension, n (%) 0.672 Yes 4 (33.3%) 8 (66.7%) No 3 (23.1%) 10 (76.9%) Tumor stage (FIGO), n (%) 0.043 I 6 (54.5%) 5 (45.4%) II 0 (0%) 6 (100%) III-IV 1 (12.5%) 7 (87.5%) Myometrial invasion, n (%) 0.231 No 1 (50%) 1 (50%) Shallow 5 (38.5%) 8 (61.5%) Deep 1 (10%) 9 (90%) Lymph node metastasis, n (%) 0.354 Yes 1 (11.1%) 8 (88.9%) No 6 (37.5%) 10 (62.5%) p53, n (%) 0.179 Mutation 1 (10%) 9 (90%) Wild 6 (40%) 9 (60%) Clinicopathological characteristics of non-endometrioid histological patients with CST3 expression.

This study has identified several plasma proteins associated with EC. These proteins have the potential to facilitate the early diagnosis of EC, recognition of EC recurrence, assessment of risk and prognosis, and development of related drugs. However, further experimentation and clinical studies are needed to evaluate the efficacy of these druggable targets in the future.

As EC enters an era characterized by advanced molecular profiling, the challenge of identifying novel therapeutic targets and developing new drugs remains significant. In this study, we identified a total of 14 potential drug targets by co-localization analysis and SMR: 7 genes for endometrioid carcinoma and 7 genes for non-endometrioid carcinoma. Additionally, our findings elucidated that elevated expression of IGF2R represents a significant clinical risk factor for endometrioid cancers, while CST3 appears to have a protective effect in non-endometrioid cancers. These discoveries contribute to the theoretical foundation for developing clinically targeted therapeutic strategies. Numerous studies have established a strong link between IGF2R and tumorigenesis, with the protein being detected in both tumor tissues 27 – 31 and the urine of cancer patients 32 . Our research revealed that IGF2R is significantly overexpressed in EC and is associated with a poorer prognosis. This aligns with previous studies indicating that elevated IGF2R expression correlates with unfavorable outcomes in laryngeal squamous cell carcinoma and cervical cancer 27 , 31 . Interestingly, in certain cancers including bladder cancer, IGF2R functions as a tumor suppressor 30 . As a key receptor in the IGF signaling pathway, IGF2R regulates energy metabolism. Moreover, prior studies have also shown that genetic variants of IGF2R, such as IGF2R167, may facilitate the development of high-mutation-rate cancers like colorectal, endometrial, and breast cancers 33 . Our study further confirms that IGF2R upregulation is linked to an increased genetic susceptibility to EC. The overexpression of IGF2R can disrupt lysosomal homeostasis and cellular autophagy 31 , accelerating EC progression. Therefore, IGF2R could serve as a critical therapeutic target in EC treatment, and inhibiting its activity may help mitigate disease progression in EC patients. CST3, a cysteine protease inhibitor found in various body fluids, plays a pivotal role in regulating cell proliferation, migration, and apoptosis, which significantly impacts tumor development 34 . Previous researches have linked CST3 to the prognosis of multiple cancers, including multiple myeloma 35 and breast cancer 36 . In our study, reduced CST3 levels were observed in non-endometrioid carcinomas, consistent with existing literature. Specifically, single-cell analyses by Qianhua Wu et al. also showed downregulation of CST3 protein expression in non-endometrioid carcinoma 37 . CST3 expression may be influenced by genetic factors, cytokines, inflammatory responses, and disease states. Previous research has identified TGF-β as a critical player in tumorigenesis, potentially affecting CST3 expression 38 . Moreover, CST3 may be involved in epigenetic regulatory mechanisms, such as DNA methylation, which could inhibit its transcriptional activity 34 . The methylation status of CST3 may thus be relevant to tumor-suppressive function in EC, but further studies are required to elucidate the specific relationship. Additionally, we have identified proteins such as GSTO1, MMP10, HHIP, CBR3, and ACAP2 that are associated with the prognosis of endometrioid carcinomas. GSTO1 is highly expressed in various cancers, including colon cancer 39 , bladder cancer 40 , melanoma 41 , hepatocellular carcinoma 42 , and non-small cell lung cancer 43 – 45 , and is implicated in drug resistance in colon and breast cancer cells 46 . MMP10, a matrix metalloproteinase, plays a complex role in cancer, contributing to cancer stem cell maintenance and tumor metastasis 47 . However, our study found that high MMP10 expression is associated with a better prognosis in endometrioid carcinoma, contrary to previous studies linking MMP10 to worse outcomes in other cancers, possibly due to differences in tumor microenvironments and cancer stages 48 – 50 . In addition, HHIP, a known negative regulator of the hedgehog signaling pathway, may inhibit tumorigenesis 51 , aligning with our findings. While research on CBR3 and ACAP2 in cancer is limited, related molecules such as LncRNA CBR3-AS1 and CircACAP2 have been studied. Our research has identified CBR3 and ACAP2 as potential risk factors for endometrioid carcinoma, consistent with previous research showing high expression of CBR3-AS1 52 – 54 and CircACAP2 55, 56 in various cancers, where they promote cell proliferation and metastasis. However, further investigation is needed to fully understand the roles of these proteins in endometrioid carcinoma. In addition, we have newly discovered that proteins such as MAPK9, FSTL5, and DNAJB14 are associated with the prognosis of non-endometrioid carcinomas. MAPK9 (JNK2) has been associated with a poor prognosis in non-endometrioid cancers, potentially due to its role in promoting tumorigenesis via the JNK signaling pathway 57 , 58 . Our findings suggest that FSTL5 and DNAJB14 may serve as favorable prognostic markers in non-endometrioid cancers. FSTL5 has been shown to inhibit tumor progression by suppressing epithelial-to-mesenchymal transition 59 . While DNAJB14, a multifunctional molecular chaperone, is essential for maintaining intracellular protein homeostasis, participating in stress responses, and playing roles in various disease-associated processes 60 . The functions of these drug targets in tumors are complex and may vary across cancer types, highlighting the need for further research to elucidate their roles in tumorigenesis and their potential as therapeutic targets for cancer treatment. The strengths of this study include its contribution to identifying new therapeutic targets for EC by evaluating the druggability of targets and their interactions. However, some limitations must be acknowledged. First, the study population was limited to Europeans, and further validation in diverse racial groups is necessary. Second, the sample size for this study was relatively limited. Future studies should incorporate larger sample sizes and refine statistical models to achieve greater precision in elucidating variable relationships and conducting comprehensive analyses. Third, the EC samples were derived from the past two years, and the follow-up period was insufficient for a comprehensive multifactorial analysis. The follow-up period is planned to be extended to five years, and future research will perform multivariate analyses of IGF2R and CST3 to evaluate their independent effects on EC prognosis. Additionally, validation of these proteins’ levels in plasma is crucial, given the potential differences between tissue and plasma proteins. Future studies should also design functional experiments to explore the mechanisms by which targets like CBR3 and ACAP2 may contribute to cancer development.

Endometrial cancer (EC) is one of the most prevalent gynecological malignancies globally. Its rising morbidity and mortality rates pose a significant threat to women’s health. Key risk factors for EC include both lifestyles and genetic influences, such as obesity, metabolic syndrome, aging, prolonged exposure to unopposed estrogen, and inherited genetic mutation 1 . Traditionally, EC is categorized into two primary histological subtypes: endometrioid carcinoma (estrogen-dependent, accounting for approximately 80% of cases) and non-endometrioid carcinoma (estrogen-independent, making up about 20% of cases) 2 . Non-endometrioid carcinomas are generally more aggressive variants like endometrial serous carcinoma, clear-cell carcinoma, and carcinosarcoma 3 . With the advent of molecular heterogeneity of EC, the cancer genome atlas (TCGA) has further categorized EC into four distinct molecular subtypes: POLE-mutated EC (5–15%), microsatellite instability-high (MSI-H)/deficient mismatch repair (dMMR) EC (25–30%), copy-number high EC (serous-like, 5–15%), and copy-number low EC (endometrioid-like, 30–40%) 1 . These subtypes differ not only in their genetic profiles but also in their clinical outcomes, with copy-number high tumors often harboring P53 mutations or the presence of frequent FBXW7 and PPP2R1A mutations and increased levels of cell cycle dysregulation (due to alterations in CCNE 1, PIK3CA, MYC, and CDKN2A) showing the worst prognosis 4 . Despite these advancements in molecular classification, the identification of effective drug targets for EC remains a challenge, highlighting the need for further research into novel treatment strategies. Recent progress in proteomics and genomics has revealed the critical role of human plasma proteins in various biological processes, especially their potential as therapeutic targets 6 . Genetic studies, such as genome-wide association studies (GWAS) 7 – 11 have uncovered numerous protein quantitative trait loci (pQTL) that influence protein expression levels and disease susceptibility, exploring the genetic relationships of complex traits and diseases by identifying genome-wide loci, analyzing genetic associations, and crossing trait loci 12 . As the cost of genotyping and sequencing has decreased, the sample pool for GWAS has grown to more than 1 million participants, including more than 5,700 GWAS performed on more than 3,300 traits 13 . GWAS investigates potential therapeutic targets and predicts disease risk by examining the correlation of genes with genetic phenotypes 14 . Mendelian randomization (MR), a powerful epidemiological tool employing genetic variants as instrumental variables, is increasingly used to infer a causal relationship between risk factors and disease outcomes. When combined with proteomic data, this approach has proven valuable in identifying novel therapeutic targets for several diseases, including cancer 15 , 16 . Epidemiological studies have demonstrated that EC exhibits a notable hereditary influence, with heritability within families estimated to range from approximately 27–52% 17 – 20 . The study of genetic factors in EC is still in its early stages, and expanding GWAS data may reveal new biomarkers and therapeutic targets. In this study, we aim to identify potential therapeutic targets for EC using MR and proteomic approaches. By integrating data from pQTLs, GWAS meta-analyses, and functional studies, we will explore novel therapeutic targets for both endometrioid and non-endometrioid subtypes of EC. Furthermore, key targets will be validated through Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) and immunohistochemistry (IHC) of clinical samples. Our research will provide solid data to support the development of new treatment strategies and drug targets for EC.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4 Supplementary Material 5 Supplementary Material 5 Supplementary Material 6 Supplementary Material 6 Supplementary Material 7 Supplementary Material 7 Supplementary Material 8 Supplementary Material 8 Supplementary Material 9 Supplementary Material 9 Supplementary Material 10 Supplementary Material 10 Supplementary Material 11 Supplementary Material 11 Supplementary Material 12 Supplementary Material 12 Supplementary Material 13 Supplementary Material 13 Supplementary Material 14 Supplementary Material 14