Therapeutic Targets for Gastric Cancer: Mendelian Randomization and Colocalization Analysis

Abstract

Background

Gastric cancer (GC) is one of the most prevalent malignancies in the world. Most patients are diagnosed at advanced stages of the disease, primarily attributable to the insidious nature of early symptoms and the infrequent occurrence of routine screening. Further biomarkers are still needed for more comprehensive analysis, targeted prognostication, and effective treatment strategies. Plasma proteins are promising biomarkers and potential drug targets in GC. This study aims to identify potential therapeutic targets for GC by conducting a comprehensive proteome-wide Mendelian randomization (MR) and colocalization analyses.

Methods

Plasma proteins were obtained from the UK Biobank Pharma Proteomics Project (UKB-PPP), including Genome-Wide Association Study(GWAS)data of 1463 plasma proteins. Genetic associations with cancer were derived from the European Bioinformatics Institute (EBI) database, including 1029 patients and 475,087 controls (dataset: ebi-a-gcst90018849). MR analysis was conducted to assess the association between plasma proteins and the risk of developing cancer. Additionally, colocalization analysis was employed to investigate whether the identified proteins and gastric cancer exhibited shared incidental variants. Finally, using the extensive Finnish database in the R9 version, the potential harmful effects of target proteins on the treatment of gastric cancer were explored through the whole phenomenon association study (PheWAS).

Result

The results showed that 15 proteins may be associated with the risk of gastric cancer, and one protein is expected to become a therapeutic target for gastric cancer. There was a positive genetic association between plasma levels of 11 proteins and increased GC risk, while 4 proteins exhibited an inverse association with GC risk (P < 0.05). Colocalization analysis revealed that PPCDC and GC exhibited shared genetic loci among the 15 proteins examined, indicating that PPCDC may serve as potential direct target for intervention in GC. Further phenotype wide association studies showed that PPCDC (P < 0.05) could be associated with certain potential side effects.

Conclusion

Our research examined the causal relationship between plasma proteins and gastric cancer, shedding light on potential therapeutic targets. These findings have significant implications for the development of early diagnostic markers and targeted therapies for GC, potentially improving patient outcomes and survival rates. Future studies should validate these findings in diverse populations and explore the clinical applications of these targets.

Introduction

Gastric cancer (GC) ranks as the fifth most prevalent malignancy in the world and stands as the third foremost cause of cancer-related mortality [1], thereby representing a significant risk to human health and well-being.

Based on the data released by GLOBOCA, the global annual incidence rate of GC in 2020 reached 1,089,000 (corresponding to an age-standardized incidence rate of 11.1 per 100,000 individuals). ranking fifth among all malignant neoplasms. The incidence rate of GC varies among regions, with the highest in East Asia, Central Europe, Eastern Europe and South America, and the lowest in North America and Africa. In the same year, GC caused 769,000 deaths (equivalent to an age standard mortality rate of 7.7 per 100,000 individuals), ranking fourth among all types of cancer, second only to lung cancer, colorectal cancer, and liver cancer [2, 3]. A study conducted in 2022 estimates that the global number of new cases will rise by 62%, reaching approximately 1.77 million by 2040 [4].

The early detection and diagnosis of GC are pivotal in the endeavor to diminish its incidence and mortality rates. Systemic chemotherapy serves as the primary treatment modality for metastatic GC, yielding a median overall survival (OS) of approximately 12 months for patients undergoing conventional chemotherapy [5]. Consequently, enhancing the diagnostic accuracy of early-stage GC and its precursor lesions has emerged as a central focus within the contemporary research landscape. This pursuit is driven by the imperative to identify and characterize the subtle indicators of GC at its most treatable stages, thereby optimizing patient outcomes and informing therapeutic strategies.

Carcinoembryonic antigen, carbohydrate antigen 19 − 9 and carbohydrate antigen 72 − 4 are commonly used biomarkers for clinical diagnosis of GC, as well as novel biomarkers such as some peptides, DNA, non-coding RNA, and circulating tumor cells. Despite the non-invasive, simple, and affordable advantages of biomarker testing, the diagnostic capabilities of established biomarkers remain limitations, there is still a scarcity of biomarkers demonstrating high sensitivity and specificity. Various potential new biomarkers for GC remain in the early stages of fundamental research [6]. Currently, the primary focus of targeted therapy for advanced gastric cancer is on the inhibition of HER2 and VEGFR2 pathways. The scope of its applicable population is restricted, and drug resistance issues are gradually emerging. Consequently, there is a pressing necessity to establish novel and more dependable therapeutic targets [1].

Plasma proteins are essential components in numerous biological processes, including signaling, transportation, growth, tissue repair, and immune defense. Dysregulation of these proteins is frequently observed in a variety of diseases, thereby rendering them significant targets for pharmacological development. Zhou et al. [7] used liquid chromatography coupled with tandem mass spectrometry labeling method and found that there were 11 differentially expressed plasma proteins in gastric cancer patients, among which, p08493, q9h939, a0a087wty6, a0a0g2jmc9, p14207, q86ud1, q8nbp7 were up-regulated, while p00441, p16157, p62979, a0a2r8y7 × 9 were down regulated, which confirmed that there was differential expression of plasma proteins in gastric cancer patients. The above illustrates the potential value of plasma proteins in the diagnosis and treatment of cancer. With the continuous updating of proteomic research methods and technologies, more plasma proteins will become new targets and new choices for clinical treatment in the future.

Previous epidemiological studies have found a certain correlation between the levels of some plasma proteins and the risk of GC [8]. However, these conclusions were obtained from observational studies that may be prone to confusion and reverse causality. Moreover, in the absence of conclusive evidence, randomized controlled trials cannot explore the causal relationship between thousands of proteins and GC.

Recently, Mendelian randomization (MR) analysis is a reliable analytical approach to evaluate the causal association between exposures and outcomes, has become a a novel epidemiological method for elucidating causal associations [9,10,11], as well as serving as a widely utilized instrument for the repurposing of approved pharmaceuticals and the identification of new therapeutic targets [10]. MR study design follows the Mendelian inheritance rule of “random assignment of parental alleles to offspring”. If genotype determines phenotype, then genotype is associated with disease through phenotype. Therefore, genotype can be used as an instrumental variable (IV) to infer the association between phenotype and disease. The MR method uses genetic variants as IV to build models and derive causal effects. The MR analysis is predicated on three fundamental assumptions: (1) Correlation assumption: The genetic IV are strongly correlated to the exposure. (2) Independence assumption: The IV is not affected by confounding factors related to both the exposure and outcome. (3) Exclusivity assumption: The IV is only allowed to exert effect on the outcome via the exposure [12]. In traditional observational clinical studies, the presence of confounding factors greatly interferes with the causal inference of exposures and outcomes. Theoretically, MR can effectively avoid the influence of eliminating confounding factors, eliminate the interference of reverse causality, and provide definitive causal relationships.

In this study, we performed whole proteome-wide MR analysis and colocalization analysis of GC to explore potential biomarkers and therapeutic targets of the disease. Furthermore, the implementation of phenome-wide association studies (PheWAS) enabled the prediction of adverse reactions linked to these targets.

Methods

Study Design

The study design is shown in Fig. 1. pQTLs (Protein Quantitative Trait Loci) of plasma protein were extracted, pQTLs were selected as instrumental variable, MR analysis was performed to evaluate the relationship between plasma protein and GC risk, and multi SNP based SMR test was used to study the association between relevant exposure and outcome traits.

Fig. 1

Study design

Colocalization analysis was used to examine whether the identified proteins and GC shared incidental variants. Finally, the potential harmful effects of target proteins on the treatment of GC were explored by whole phenomenon association study (PheWAS).

Related databases: UK Biobank Pharma Proteomics Project (UKB-PPP) database: (https://www.synapse.org/#!Synapse:syn51364943/files/). European Bioinformatics Institute (EBL):https://www.ebi.ac.uk/, including 1029 patients and 475,087 controls (dataset: ebi-a-gcst90018849). eQTLs( Expression Quantitative Trait Loci) data: https://cnsgenomics.com/data/SMR/#eQTLsummarydata.

Plasma Protein Quantitative Trait Loci

The research utilized plasma proteome data sourced from the UKB-PPP database. These datasets contain a large number of plasma proteins and their related genetic information. Proteome analysis was performed on plasma samples from 54,306 participants, and data on 1463 proteins were collected.

We selected known protein quantitative trait loci (pQTLs) of plasma proteins as instrumental variables to screen genome-wide pQTLs of 1463 plasma proteins in the ukb-ppp database. These pQTLs represent genetic variations associated with plasma proteins. When extracting pQTLs, specific screening criteria are applied, and these criteria are summarized as follows:

• 1. The pQTLs exhibited a genome-wide significant association(p < 5e-8).

• 2. The assumption of independence was satisfied. (Perform linkage disequilibrium (LD) clumping procedure: r2 < 0.001.)

• 3. The pQTLs are analyzed within a 10,000 kb range surrounding the relevant protein-coding sequences.

• 4. Harmonize data to align the allele directions of exposed data and outcome data; remove palindromic SNPs, which are single-nucleotide polymorphisms that appear the same in both strands of the DNA sequence, in order to mitigate potential bias in the analysis.

• 5. Filter rare variations to ensure the validity of instrumental variables. (Minor allele frequency > 0.01.)

• 6. Excluded: SNP < 3.

MR Analysis

Our MR analysis was performed according to the guidelines outlined in the 《Strengthening the reporting of observational studies in epidemiology using mendelian randomisation》, which used STROBE-MR checklist [13–14]. We employed a two-sample MR approach to evaluate the causal association between levels of specific plasma proteins and the risk of GC. In this context, pQTLs were utilized as instrumental variables. By examining the genetic effects associated with these pQTLs, we derived inferences regarding alterations in protein levels, which were subsequently utilized to draw causal conclusions about the risk of GC.

In this study, ukb-ppp protein was used as the exposure and GC as the outcome. The R software package “TwoSampleMR”V.0.5.7, MendelianRandomization V.0.9.0” were used for two sample MR analysis. We used seven different methods for MR analysis: inverse variance-weighted (IVW), weighted median (WM), MR-Egger regression, weighted mode, containment mixture method, robust adjusted profile score (RAPS), and constrained maximum likelihood. IVW method was selected as the main method for MR analysis because of its high statistical efficiency, while the other six methods were used as complementary methods. Five MR methods (one of which was IVW) were all < 0.05 and no level pleiotropy and heterogeneity were taken as positive proteins.

The Cochran’s Q statistic (MR-IVW) and Rucker’s Q statistic (MR-Egger) were employed to detect the heterogeneity of MR analysis, and P > 0.05 indicated that there was no heterogeneity [15]. The intercept test of MR-Egger to detect horizontal pleiotropy, P > 0.05 indicates that there is no horizontal pleiotropy [16]. The “leave one out” analysis was used to investigate whether the causal relationship between protein and GC was affected by a single SNP [17].

SMR Analysis

In order to explore the relationship between protein levels and GC, we performed a multi SNP based SMR test to investigate the association between relevant exposure and outcome, and whether the effect of SNPs on phenotype is mediated by gene expression.

In SMR analysis, eQTLs were used as instrumental variables of gene expression, and GC was used as the outcome for SMR analysis.

SMR software tools (https://cnsgenomics.com/software/smr/).

Colocalization Analysis

To examine whether plasma proteins share the same genetic variation sites with GC, we performed colocalization analysis. Colocalization analysis was used to determine whether there were genetic variants associated with protein levels and gastric cancer risk at the same gene locus. We screened out those proteins that shared genetic variation by colocalization test.

For proteins with positive MR results, we conducted colocalization analysis to test whether the identified protein association with GC was driven by linkage disequilibrium.

The colocalization analysis involves five hypotheses [18]: H0 indicates that the selected SNP within the locus is unrelated to both protein A and disease B; H1 suggests that the SNP within the selected locus is associated with protein A but not with disease B; H2 implies that the SNP within the selected locus is related to disease B but not to protein A; H3 states that the SNP within the selected locus is associated with either protein A or disease B, but the two are independent SNPs; H4 signifies that the SNP within the selected locus is concurrently associated with both protein A and disease B, and is a shared SNP. Due to limited power in the co-localization analysis, we focused our examination on genes with the posterior probability of hypothesis 4 (PPH4) equal to or greater than 0.5. Proteins that are positive in colocalization analysis (PPH4 ≥ 0.5) are considered potential drug targets, and a PheWAS analysis is performed on the target proteins.

Phenome-wide Association Study

PheWAS, often referred to as reverse GWAS, represent a methodological approach employed to investigate the relationships between SNPs or specific phenotypes and a diverse range of phenotypes encompassing the entire phenome.

We conducted a comprehensive PheWAS analysis to explore the potential impact of target proteins on other traits and diseases. This analysis encompasses multiple phenotypes to assess whether the target proteins associated with gastric cancer have detrimental effects on other diseases or biological traits.

Results

Plasma pQTLs

After stringent application of the IV screening criteria, the mendelian randomization (MR) analysis incorporated 6202 pQTLs. Relevant SNP data can be found in Supplementary 1.

MR Analysis

MR analysis utilizing UK Biobank Pharma Proteomics Project (UKB-PPP) data identified a significant positive correlation between 14 proteins and GC risk, including Angiopoietin-2 (ANGPT2), All-trans retinoic acid-induced differentiation factor (ATRAID), BPI fold-containing family A member 2 (BPIFA2), T-cell differentiation antigen CD6 (CD6), Alpha-(1,6)-fucosyltransferase(FUT8), Protein PBMUCL2 (HCG22), Immunoglobulin lambda constant 2 (IGLC2), Leukocyte immunoglobulin-like receptor subfamily B member 1 (LILRB1), Matrilin-3 (MATN3), Protein kinase C-binding protein NELL1 (NELL1), Phosphopantothenoylcysteine decarboxylase (PPCDC), Repulsive guidance molecule A (RGMA), Toll-like receptor 3 (TLR3) and Tumor necrosis factor receptor superfamily member 10 C (TNFRSF10C), with elevated plasma levels correlating to a higher likelihood of developing GC.

Conversely, the analysis also indicated a negative correlation between certain proteins and GC risk, including Complement factor H-related protein 2 (CFHR2), Carboxypeptidase Q (CPQ), Lymphotoxin-beta (LTB), Nicotinamide/nicotinic acid mononucleotide adenylyltransferase 1 (NMNAT1), Phenazine biosynthesis-like domain-containing protein (PBLD), Tumor necrosis factor receptor superfamily member 11 A (TNFRSF11A) and Tumor necrosis factor ligand superfamily member 12 (TNFSF12), suggesting a potential protective role against GC. Comprehensive data, including odds ratios and 95% confidence intervals for these associations, were presented in Table 1; Fig. 2.

Table 1 MR analysis between the serum proteins and gastric cancer

Fig. 2

Forest map of MR analysis

CI: confidence interval; OR: odds ratio

Sensitivity Analysis

To validate the robustness of the identified associations, numerous sensitivity analyses were conducted, indicating that the roles of these proteins in GC risk remained unaffected by pleiotropy, with no significant heterogeneity detected. Consequently, the credibility of these results was reinforced. Table 2 presented the 21 proteins, all exhibiting no signs of pleiotropy or heterogeneity.

Table 2 Sensitivity analysis of the MR analysis results of exposures and outcomes

SMR Analysis

Further SMR analysis validated the proteins identified initially, confirming a significant causal association between 15 proteins and GC. These proteins remain statistically significant in relation to GC, indicating their potential as therapeutic targets.

The 15 identified proteins included ANGPT2, ATRAID, CD6, CPQ, FUT8, HCG22, IGLC2, LILRB1, LTB, PPCDC, RGMA, TLR3, TNFRSF10C, TNFRSF11A, TNFSF12. Table 3 showed the detailed data.

Table 3 The SMR analysis results of exposures and outcomes

Colocalization Analysis

To explore the mechanistic relationship between these proteins and GC, gene colocalization analysis was performed within a 1 MB interval surrounding the 15 validated protein loci. This analysis assessed whether genetic variants associated with GC coincided with alterations in protein levels. Multiple genetic variation sites were shared with GC, with a PPH4 value exceeding 50%, suggesting that both may be influenced by the same genetic variation. Results indicated that the PPCDC protein exhibited significant colocalization signals at loci linked to GC risk, with a PPH4 value of 51.8%.

These findings support a potential causal relationship between PPCDC and GC, offering new evidence for considering PPCDC protein as a therapeutic target in GC treatment.

For further details, please refer to Fig. 3 and Table 4.

Table 4 Results of colocalization analysis

Fig. 3

Colocalization analysis results of PPCDC and Gastric Cancer

The x-axis usually represents the position on the chromosome, the y-axis represents statistical significance (-log10(p-value))

Phenome-wide Association Study

To evaluate the effects of GC-related PPCDC on various phenotypes, the Finnish database (version R9) phenotype library facilitated a phenome-wide association study (PheWAS). P < 0.05 signified a potential causal relationship with PPCDC. Results indicated that PPCDC protein correlated not only with GC but also influenced several other diseases and biological traits, such as ganglion, asthma-related acute respiratory infections, gingivitia andperodontal diseases, aspergillosis, biologicalmedicines for asthma, vestibularneuronitis, pruritus, vascular dementia, benign neoplasmof meninges, nonhereditary hypogammaglobulinemia, lipomatosis, rheumatic heart disease, henoch-Schonlein purpura nephritis, congenital malformations of pulmonary and tricuspid valves and wide developmentaldisorders.

While PPCDC may aid in GC treatment, its negative implications for other traits warrant careful consideration during therapeutic applications.

The Manhattan plot of Fig. 4 identified 16 diseases across each system potentially impacted by PPCDC. The threshold line denoted P < 0.05, while the red markers highlighted the diseases with the lowest P values within each system.

Fig. 4

Manhattan plot of result of PheWAS analysis of associations between PPCDC and other disease outcomes

The Manhattan plot shows the diseases in various systems that may be affected by PCDCC. The x-axis represents diseases from each system, and the y-axis shows the -log10 of the p-value for each genetic variant tested. The blue line represents p < 0.05, and the red dots represent the diseases with the smallest p-values in each system.

Discussion

GC ranks among the five most common malignancies globally, and is a leading cause of cancer-related deaths [19]. In 2020, 27,600 new GC cases were reported, with a five-year survival rate of merely 32% (2010–2016) [20]. Presently, patients undergoing early surgical resection for stage I GC experience a five-year survival rate of 85–95%, establishing surgery as the preferred treatment modality [21–22]. Nonetheless, challenges persist in the screening and management of early GC due to limitations in preventive measures, including gastroscopy and serological tests.

This study utilized the UKB-PPP data to investigate the correlation between plasma protein levels and GC. Utilizing the Olink Explore platform, the UKB-PPP characterized the plasma proteomic profiles of 54,219 UK Biobank participants, conducted extensive pQTLs analysis for 2,923 proteins, and identified 14,287 significant genetic associations, 81% of which were novel. Additionally, ancestry-specific pQTLs analysis was performed for non-European participants. This initiative offers the scientific community a comprehensive open-access proteomics resource, significantly advancing the understanding of the biological mechanisms underlying protein genomic discoveries and supporting the development of biomarkers, predictive models, and therapeutic strategies [23].

This study identified 15 plasma proteins potentially causally linked to GC through MR and colocalization analyses, including ANGPT2, ATRAID, CD6, CPQ, FUT8, HCG22, IGLC2, LILRB1, LTB, PPCDC, RGMA, TLR3, TNFRSF10C, TNFRSF11A and TNFSF12.

Among them, ANGPT2, ATRAID, CD6, FUT8, HCG22, IGLC2, LILRB1, PPCDC, RGMA, TLR3, and TNFRSF10C are likely correlated positively with GC risk, while CPQ, LTB, TNFRSF11A, and TNFSF12 appeared to correlate negatively with this risk.

Proteins Positively Correlated With the GC Risk

In this study, 11 plasma proteins—ANGPT2, ATRAID, CD6, FUT8, HCG22, IGLC2, LILRB1, PPCDC, RGMA, TLR3, and TNFRSF10C—correlated positively with GC risk and may influence its development through various pathways. For instance, ANGPT2 is a known angiogenic factor that plays a pivotal role in tumor growth and metastasis, with its high expression widely reported in various cancers. This indicates that ANGPT2 may elevate the risk of gastric cancer by enhancing tumor angiogenesis. Notably, the colocalization analysis of PPCDC protein revealed significant overlap with multiple genetic variation sites associated with GC, suggesting a vital role in its onset and progression.

ANGPT2, a member of the angiopoietin (Ang) family, plays a significant role in vascular remodeling and inflammation. Its mechanisms may involve destabilizing quiescent blood vessels, allowing VEGF to stimulate proliferation and chemotactic migration of vascular buds, and activating TIE2-expressing monocytes/macrophages (TEMs), thereby promoting angiogenesis, tumorigenesis, metastasis, and immunosuppression [24,25,26]. ANGPT2 functions as both a therapeutic indicator for immune checkpoint inhibitors and a target for the precise selection of chemotherapeutic agents. ANGPT2 overexpression in GC is associated with a poor prognosis, significantly contributing to tumor initiation and promoting angiogenesis [27]. Notably, ANGPT2 levels in GC tissues surpass those in adjacent paracancerous and normal tissues [28].

ATRAID, also known as APR3, exhibits substantially increased mRNA expression in HL-60 cells following all-trans-retinoic acid (ATRA) treatment.

Located on human chromosome 2p23.3, ATRAID is involved in various cellular processes. Overexpression of ATRAID has been shown to inhibit cyclin D1 expression during the G1/S phase [29], although the precise mechanisms by which ATRAID influences GC remain unclear.

CD6, a transmembrane receptor integral to signal transduction, is predominantly expressed in T cells and specific subsets of B and NK cells. Endogenous ligands such as CD166/ALCAM, CD318/CDCP-1, and Galectins 1 and 3 are often overexpressed in various malignant cell types. Immunomodulatory receptors on lymphocyte surfaces have demonstrated potential in novel cancer immunotherapies, including CTLA-4 and PD-1; thus, CD6 emerges as a promising target for innovative treatments of malignant tumors [30].

Glycosylation alterations constitute critical molecular events in tumorigenesis and progression, with glycosyltransferases playing a vital role in this process. FUT8, a member of the fucosyltransferase family, functions as a key enzyme in N-glycan core fucosylation. Notably, FUT8 and related core fucosylation proteins are frequently upregulated in liver, lung, colorectal, pancreatic, and thyroid tumors, while exhibiting downregulation in GC. This differential expression positions them as potential biomarkers for cancer diagnosis, progression, and prognosis. Additionally, core fucosylation EGFR, TGFBR, E-cadherin, PD1/PD-L1, and α3β1 integrin present promising targets for tumor therapy [31].

HCG22 significantly influences the progression of multiple human diseases, including steroid-induced abnormal intraocular pressure and late-onset asthma. Bioinformatics analysis indicates that lncRNA HCG22 contributes to the development of esophageal squamous cell carcinoma, head and neck squamous cell tumors, as well as thyroid and cervical cancer [32].

IGLC2 is involved in the variable region structural domains of the immunoglobulin light chain, which is critical for antigen recognition [33]. It presents a notable mutation risk in the pan-cancer context [34] and is expressed in tumor cells, playing a key role in the prognosis of triple-negative breast cancer [35].

Immune checkpoint blockade represents a novel approach in tumor immunotherapy. Blocking of T cell inhibitory pathways exhibits clinical efficacy across diverse cancer types and may enhance innate immune responses. LILRB1 has emerged as a promising target in immune checkpoint therapy, belonging to the leukocyte immunoglobulin-like receptor superfamily. It is expressed by various immune cells, including macrophages and specific cytotoxic lymphocytes, and plays a role in regulating immune responses through interactions with both classical and nonclassical human leukocyte antigen (HLA) class I molecules. Currently, LILRB1-specific antibodies are undergoing various phases of preclinical and clinical development [36].

Coenzyme A (CoA) functions as a vital cofactor in numerous metabolic pathways. PPCDC, a key enzyme in CoA synthesis, belongs to the homooligomeric flavin-containing Cys decarboxylase (HFCD) family. Limited research exists on the relationship between PPCDC and GC, though case reports indicate that pathogenic variants can lead to autosomal recessive dilated cardiomyopathy [37].

RGMs play a significant role in various essential biological processes, such as cell migration, differentiation, iron homeostasis, and apoptosis. Three RGM subtypes (RGMa, RGMb/DRAGON, and RGMc/hemojuvelin) are implicated in the pathogenesis of several diseases, including multiple sclerosis (MS), cancer, and juvenile hemochromatosis (JHH), yet the underlying molecular mechanisms remain largely unknown [38].

RGMA expression is markedly downregulated in breast cancer tissues, especially within metastatic samples. Low RGMA expression correlates with poor patient prognosis and may contribute to breast cancer progression through activation of the FAK/Src/PI3K/AKT signaling pathway. However, its relationship with GC remains uninvestigated, leaving the mechanisms underlying its role in GC progression unclear [39].

TLR3 exhibits anti-tumor, pro-tumor, and dual effects in cancer, positioning it as a potential prognostic biomarker and immunotherapy target. A retrospective cohort study involving 564 gastric adenocarcinoma patients indicated that elevated TLR3 expression may influence GC prognosis [40].

Located at 8p21.3 (23.01–23.03 Mb), tumor necrosis factor receptor superfamily member 10c (TNFRSF10C) is among the most frequently deleted loci in colon cancer. Variations in TNFRSF10C copy number correlate with metastatic colorectal cancer [41], while its hypermethylation significantly increases colon cancer risk [42]. Nevertheless, the mechanism underlying its role in GC remains unclear.

Proteins Negatively Correlated With GC Risk

Four plasma proteins—CPQ, LTB, TNFRSF11A, and TNFSF12—exhibited a negative correlation with GC risk, suggesting a potential role in inhibiting or defending against the disease.

CPQ, also referred to as plasma glutamate carboxypeptidase (PGCP), belongs to the M28 family of metallocarboxypeptidases. Increased mRNA expression of CPQ in glioblastoma tissue compared to normal tissue positions it as a significant prognostic biomarker for glioblastoma patients [43]. However, its mechanism in GC remains unclear.

LTβR signaling correlates with inflammation and tumor development. LTβR presents a novel target for mitigating gastric inflammation and pathology associated with Helicobacter pylori, potentially elucidating its negative correlation with GC [44].

TNFRSF11a plays a crucial role in regulating cell differentiation, proliferation, and survival. It functions as an inducer of activated dendritic cells and is essential for maintaining immune tolerance. Research indicates that TNFRSF11a inhibits the motility and migration of breast cancer cells while promoting the proliferation of cervical cancer cells and suppressing apoptosis [45]. Furthermore, TNFRSF11A expression correlates positively with the survival of GC patients, leading to its incorporation into prognostic models for predicting high-risk GC cases [46].

Single-cell and tissue transcriptome studies reveal that TNFSF12 stimulation in A549 human lung cancer cells induces significant migration and aggressive phenotypes, along with upregulation of genes associated with glycolysis and epithelial-mesenchymal transition. However, the specific mechanisms by which TNFSF12 exerts effects in GC remain unclear [47].

Identification of Novel Drug Targets

Given the strong colocalization evidence, PPCDC emerges as a promising therapeutic target for GC. Its involvement in metabolic pathways influencing T cell immunity suggests that modulating PPCDC activity could enhance anti-tumor immune responses, potentially improving immunotherapy outcomes. Targeting PPCDC might complement existing treatments, such as immune checkpoint inhibitors, by optimizing metabolic conditions that support T cell function.

However, potential risks must be carefully evaluated. Insights from PheWAS studies indicate that PPCDC modulation could elevate susceptibility to asthma-related respiratory infections, implying possible immune system dysregulation [48]. Therefore, precision medicine strategies—such as patient stratification based on genetic and immunological profiles—should be considered to maximize therapeutic efficacy while minimizing adverse effects.

Future studies should focus on mechanistic investigations of PPCDC in GC and preclinical validation of PPCDC-targeted drugs. These efforts will be crucial for assessing its feasibility as a drug target and ensuring its safety in clinical applications.

Study Significance

This study demonstrates several strengths: the application of MR and SMR to estimate the causal effect of circulating proteins on GC through genetic variations enhances the robustness of the findings, while subsequent colocalization analysis further validates the results. Additionally, PheWAS analysis allows for an in-depth exploration of potential drug targets’ side effects, offering new insights into GC treatment. The use of a recent database (UKB-PPP) enriches the sample diversity, contributing to the overall rigor of the research.

These findings not only offer new insights into the biological mechanisms of gastric cancer but may also lay the groundwork for developing personalized treatment strategies. Given the complex roles of multiple proteins, future research should continue to explore their functions in various cancer subtypes, as well as their interactions with other known cancer-related molecules. Additionally, proteins identified as negatively correlated in the study should be further validated as potential targets for prevention, to assess their application in gastric cancer prophylaxis.

Study Limitations

This study yields significant preliminary results, though several limitations are evident. Although the UKP-PPP dataset includes 2,923 plasma proteins, only 1,463 proteins are integrated into the MR analysis due to restrictions associated with instrumental variables. Given the limited power of colocalization analysis, the research concentrates on cases where PPH4 is 0.5 or higher.

Another limitation concerns the lack of cell and animal model validation. While the study’s colocalization and Mendelian randomization analyses suggest plausible causal relationships between genetic variants and GC, the absence of experimental validation in cell-based or animal models significantly limits the ability to directly confirm these findings in a biological context. The findings, therefore, remain hypothetical and may not fully reflect the biological reality, as gene-environment interactions and tissue-specific effects are difficult to capture purely through bioinformatic analyses. Cell culture models would be particularly valuable for investigating the functional role of specific proteins, such as PPCDC, in gastric cancer progression. Additionally, animal models could enable more in vivo verification of the metabolic and immune-related pathways implicated in the study, including the investigation of the immune modulation by PPCDC. The absence of these experimental steps means that further mechanistic insights into these proteins’ roles in GC are still pending.

Future research should address these gaps by integrating experimental models to test the hypotheses generated through computational analysis. These experiments would also help validate the instrumental variables used in MR analysis, providing more robust causal evidence. Collaborations with research groups specializing in animal models and cellular biology will be key to advancing these aspects. Specifically, in vivo experiments using genetically modified mice or patient-derived xenografts could be employed to assess the physiological relevance of these proteins in GC development.

Moreover, while the colocalization analysis has identified potential shared genetic loci between PPCDC and GC, further functional studies are needed to determine the exact molecular mechanisms through which these loci influence cancer progression. Understanding whether these loci act through T cell immunity or other metabolic pathways will be critical in determining the therapeutic potential of targeting PPCDC.

Conclusion

In conclusion, while this study offers a promising foundation for exploring PPCDC as a potential therapeutic target in gastric cancer, it is clear that further experimental validation is essential to confirm the causal relationships and functional relevance of the identified proteins. Future research efforts, particularly through collaborations involving experimental biology, will be crucial to overcome the current study’s limitations and move towards clinical applications.