Exploring the causal relationship between serum uric acid and gastrointestinal cancer based on Mendelian randomization studies and bioinformatics approaches.

1. Introduction
In the world today, gastrointestinal (GI) tract malignant cancers are the most deadly and sick types of tumors. The incidence of colorectal cancer (CRC) and stomach cancer (GC) has increased dramatically, particularly in the last few decades. CRC ranks second in terms of death rate and third in terms of new incidence, while GC ranks fifth in terms of both.[1] Complex genetic and environmental variables, including lifestyle choices, genetic diversity, and environmental exposure, are involved in the pathophysiology of GC and CRC.[2] Furthermore, some new research indicates that the incidence of GC and CRC may be directly associated with metabolic illnesses.[3]
In recent years, increasing attention has been paid to the tumor microenvironment (TME) in gastrointestinal malignancies.[4] The TME consists of various non-cancerous components, such as immune cells, stromal fibroblasts, blood vessels, and the extracellular matrix, which closely interact with tumor cells.[5] These interactions, often mediated by cytokines, metabolites, and signaling pathways, play a crucial role in cancer initiation, progression, and immune evasion. Among various metabolic factors involved, uric acid (UA) – a byproduct of purine metabolism – has been suggested to influence the local tumor environment.[6,7] Depending on the surrounding biochemical context, UA may either act as an antioxidant or contribute to oxidative stress, potentially modulating inflammatory responses and cellular behavior within the TME. These dual properties make it a molecule of interest when exploring the metabolic dimensions of cancer development.[8] The final byproduct of the body’s cells’ metabolism of purines and nucleic acids is UA. The body’s serum uric acid (SUA) levels often stay within a fairly steady range. However, SUA levels may be raised, resulting in hyperuricemia, when purine metabolism is aberrant or UA excretion is compromised.[9] Recent in vitro cytological investigations have demonstrated that SUA can display both oxidative and antioxidant properties contingent upon its chemical milieu. Different theories on SUA’s involvement in disease have been developed due to its dual role.[10] SUA has been demonstrated to have some protective properties in addition to having potent antioxidant actions that scavenge free radicals in the body.[11] Nonetheless, SUA might be intimately linked to oxidative stress and chronic inflammation, 2 pathological processes that might encourage the growth and spread of tumors.[12,13]
While the exact mechanism underlying the correlation between SUA and GI cancer is still unknown, earlier observational studies have shown that SUA levels are linked to GC and CRC and have an impact on the disease’s prognosis and course.[14–16] It is unknown, though, what causes the other 2. It is notable that in recent years, MR approaches have been applied extensively in epidemiological investigations. For example, recent studies have applied MR to investigate causal links in various diseases, including the relationship between lipid metabolism and cardiovascular disease,[17] the effect of body mass index on cancer risk,[18] and the impact of inflammatory markers on neurodegenerative disorders.[19] MR uses single nucleotide polymorphisms (SNP) as instrumental variables (IVs) to evaluate the causal link between exposure factors and illness outcomes.[20] Reverse causation and confounding variables can be successfully avoided since genes are randomly assigned at conception, much as randomized grouping in randomized controlled trials. This ensures the validity of the research.[21,22]
Therefore, our goal was to first discuss with SUA the possible causal relationship between GC and CRC. We then conducted enrichment analysis and protein–protein interaction (PPI) network analysis to investigate the possible biological pathways connecting them, providing additional proof to support GI cancer clinical care.

2. Methods

2.1. Two-sample Mendelian randomization study

2.1.1. Study design
Using pooled data from the publicly available genome-wide association study (GWAS), we investigated the possible causal link between SUA and both GC and CRC in this investigation. We have to adhere to the 3 main MR research presumptions, as with the majority of MR studies: The chosen IVs must, according to the following 3 hypotheses: they must exhibit a strong correlation with the exposure factor; they cannot exhibit a correlation with other confounders; and they must only pass through the exposure factor and thereby impact the outcome, i.e., the assumptions of exclusivity, independence, and correlation. Since the study’s results were derived from GWAS summary data that was made accessible to the public, no further authorization or informed permission was needed (Fig. 1).

2.2. GWAS datasets for SUA and GC/CRC
Using genomic data from the IEU OpenGWAS project (https://gwas.mrcieu.ac.uk/), the MR study was conducted. In the IEU database, we looked for the most recent GWAS data with a sizable sample size and the same ethnicity. We chose GWAS summary data from SUA and GC/CRC (Table 1).

2.3. Selection of IVs
To meet the main assumptions of the MR study, we selected SNPs based on prior studies. We identified SNPs that were strongly associated with the exposure and had a significance criterion threshold of P < 5 × 10−8. To reduce the influence of linkage disequilibrium between SNPs, we set r2 = 0.001 and kb = 10,000. We next removed the intermediate allele frequency of the back-text SNPs to ensure allelic consistency across IVs and outcome variables. Palindromic SNPs with intermediate allele frequencies are excluded to guarantee that IVs and outcome variables have the same allelic distribution. We estimated the F-value of SNPs, set a threshold of F < 10, and removed SNPs with F < 10 to prevent bias induced by weak IVs. Finally, we employed the MR-PRESSO test to select SNPs with high dispersion values to eliminate results bias due to individual SNP heterogeneity.

2.4. Statistical analysis
In this study, we used the 2-sample MR approach to investigate the causal association between SUA levels and GC and CRC. The causal link between SUA and GC/CRC was investigated using the Wald ratio, inverse-variance weighted (IVW), MR-Egger, weighted median, and weighted mode methods. Among other reasons, the IVW technique implies that all SNPs are legitimate IVs and that they can be used to obtain the most accurate causal impact estimations. As a result, we adopted the IVW approach as our major analytical tool to evaluate the overall effect of exposure factors on the outcome variables. Following that, we used the MR-PRESSO approach to determine the presence of extensive horizontal pleiotropy, as well as to discover and delete any abnormal outliers. To ensure the reliability of causation, MR analyses were repeated after outliers were removed. In addition, we used the MR-Egger intercept test to determine the horizontal multiplicity of results, with P > .05 indicating no horizontal multiplicity. To analyze the heterogeneity of individual IVs, we used Cochran Q statistic, which shows no heterogeneity if P > .05, showing strong agreement between IVs. The heterogeneity was also shown and analyzed using a funnel plot. The leave-one-out strategy was used to delete specific SNPs one at a time to see if the effect values changed considerably, and the results were shown. To ensure that the results were robust, we conducted additional sensitivity analyses to rule out any bias. All analyses were performed using R version 4.2.2, with data processing and result visualization provided by the “TwoSampleMR” package.

2.5. Bioinformatics analyses
First, we obtained information about the matching IV genes for MR analysis from the National Library of Medicine (https://www.ncbi.nlm.nih.gov/). Genes relevant to SUA metabolism and GC/CRC were collected from the GeneCards database (https://www.genecards.org/). Overlapping SNP genes, SUA metabolism-related genes, and GI cancer-related genes were identified using Venn plots after eliminating non-protein-coding genes. The STRING database was used to do PPI network research. The top 10 hub genes were then found by calculating each gene’s degree, or the number of other genes related to it, using Cytoscape software. Finally, we conducted gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of hub genes using the online bioinformatics analysis website (https://www.bioinformatics.com.cn/).

2.6. Institutional review board statement
This study did not require ethical approval. This is because all the data involved in this study were obtained in public databases. All original studies from which the data were obtained had received approval from the respective local ethics committees.

3. Results

3.1. Selection of IVs
In the MR studies of SUA and GC, we identified a total of 259 SNPs as IVs based on stringent quality control. In the same way, 249 SNPs were found to be IVs for use in the SUA and CRC MR analyses. The F-value range of 30 to 18,757 demonstrated that the IVs were statistically robust and free from the weak IV bias, guaranteeing the reliability of the study’s findings. (Table S1, Supplemental Digital Content, https://links.lww.com/MD/P495).

3.2. Two-sample Mendelian randomization analysis
We discovered a causal connection between SUA level and GC using the IVW technique (OR = 0.804, 95% Cl: 0.700–0.922, P = .002). It was not possible to determine a clear causal link between SUA level and CRC at this time (OR = 0.999, 95% Cl: 0.997–1.000, P = .080) (Table 2). When the MR-PRESSO test was run, 2 outliers (rs2581824, rs7773175) were found in SUA with CRC, and 3 outlier SNPs (rs11264341, rs28755272, rs7773175) were found in SUA with GC. The MR analyses were run again independently after the outliers were eliminated, and the outcomes remained mostly unchanged (Table 2). The results of the other 4 approaches (weighted mode, simple mode, MR-Egger, and weighted median) were broadly in line with the IVW direction, despite having larger 95% Cl ranges (Fig. 2) (Table S2, Supplemental Digital Content, https://links.lww.com/MD/P495).

3.3. Sensitivity analysis
Cochran Q test was used to investigate heterogeneity in the MR analysis; the results indicated that there was no heterogeneity between the SUA and GC groups (P = .149). The horizontal pleiotropy in the MR analyses was next tested using MR-Egger analysis, and it was shown to be absent in the SUA and GC groups (Egger intercept = −0.002, SE = 0.0023, P = .414) (Table 3). We also conducted Cochran Q test and MR-Egger analysis to test the results of the SUA and CRC groups. The results suggested that there was heterogeneity in the MR results of the SUA and CRC groups (P = .034). However, there was no obvious horizontal pleiotropy (Egger intercept = <0.001, SE < 0.001, P = .927). But in this study, we used the IVW random effects model, which could reduce the influence of the heterogeneity of the data itself and make the results more robust. Furthermore, leave-one-out sensitivity analysis was used to investigate the impact of particular SNPs on the IV results. The findings showed that, even after deleting specific SNPs, the results remained substantially consistent and that individual SNPs had minimal bearing on the outcomes (Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/P496). Better consistency and less IV heterogeneity were shown in the forest plot (Fig. S2, Supplemental Digital Content, https://links.lww.com/MD/P496). Funnel plot shows no pleiotropy in SNPs (Fig. 3).

3.4. Bioinformatics analysis of SUA and GC
178 Mapped genes were collected from the National Library of Medicine based on the SNPs chosen by MR analysis (Table S3, Supplemental Digital Content, https://links.lww.com/MD/P495). The GeneCards database yielded 3684 genes associated with SUA metabolism and 13,898 genes related to GC (Table S4, Supplemental Digital Content, https://links.lww.com/MD/P495). 99 of the 106 genes that overlapped between the 3 were protein-coding genes, according to the Venn diagram (Fig. 4A; Table S5, Supplemental Digital Content, https://links.lww.com/MD/P495). Subsequently, the PPI network was utilized to exclude each of the 60 interacting genes (Fig. 4B). Ten hub genes were found to be peroxisome proliferator-activated receptor gamma (PPARγ), IGF1R, LRP2, ABCG2, APOE, HNF1A, ABCA1, ARID1A, HNF4A, and GNAS (Fig. 4C). GO enrichment analysis showed that hub gene bioprocess (BP) was mainly enriched in response to lipoprotein particle, cellular response to lipoprotein particle stimulus, amyloid-beta clearance, lipid homeostasis, anion homeostasis; cellular component is mainly enriched in brush border membrane, endocytic vesicle, apical plasma membrane, external side of plasma membrane, apical part of cell; molecular function is mainly enriched in insulin-like growth factor I binding, cholesterol transfer activity, sterol transfer activity, hormone receptor binding, insulin-like growth factor binding (Fig. 4D). KEGG enrichment analysis suggested that hub genes were mainly enriched in Cholesterol metabolism, AMPK signaling pathway, Maturity onset diabetes of the young, ABC transporters, Ovarian steroidogenesis, and other pathways (Fig. 4E).

4. Discussion
This work used 2-sample MR, mostly based on a large-scale GWAS, to analyze the causal connection between SUA levels and GC/CRC in a European population. Higher SUA levels may be linked to a lower risk of GC, according to the results, which indicated a strong inverse causal relationship between SUA levels and GC. The MR revealed no discernible causal relationship between SUA levels and CRC. Several sensitivity analyses verified the accuracy of the findings. These investigations shed new light on the connection between blood UA levels and various gastrointestinal cancers.
Prior research has disputed the function of UA in the human body; at first, it was believed to be an “inert waste product” of purine metabolism.[23] Subsequent research revealed that UA exhibited potent antioxidant properties, scavenging singlet oxygen and free radicals while safeguarding mitochondrial function.[24,25] Nevertheless, UA has also been shown in multiple studies to be pro-oxidant in the intracellular environment, where it increases the expression of angiotensin and activates the renin–angiotensin system locally.[26,27] It also induces a variety of oxidative stress responses by producing reactive oxygen species.[28] Epidemiological research indicates a close relationship between hyperuricemia and several ailments, including insulin resistance, metabolic syndrome, hypertension, and problems with the metabolism of fats and carbohydrates.[15,16] Nonetheless, it has been discovered in certain research that elevated SUA levels may encourage brain growth and lower the prevalence of mental illness.[29]
Epidemiological research on SUA and cancer has also revealed significant confusion and debate. A prospective study found that cancer patients had significantly reduced SUA levels compared to the normal group (P < .0001).[30] In another meta-analysis, high SUA levels were found to increase total cancer incidence (RR = 1.03, 95% CI: 1.01–1.05, P = .007) and mortality (RR = 1.17, 95% CI: 1.04–1.32, P = .010).[31] More observational studies on SUA, GC, and CRC indicate that greater SUA levels increase the incidence of GC and CRC and have a worse prognosis.[16,32,33]
In the present study, we noticed by MR analysis that SUA reduced the risk of GC (OR = 0.804, 95% Cl: 0.700–0.922, P = .002), however, there was no significant causal link with CRC (OR = 0.999, 95% Cl: 0.997–1.000, P = .080). This outcome was consistent with past observational research.[14,30] The findings of sensitivity analysis did not reveal any heterogeneity or pleiotropy (P > .05).
Previous research on the relationship between SUA and disease has focused primarily on the antioxidant and pro-oxidant properties of UA, but the direct biological mechanisms are unknown. As a result, we used a bioinformatics approach to identify potential key genes and biological processes between SUA and GC. We discovered the first 10 hub genes in the PPI network: PPARγ, IGF1R, LRP2, ABCG2, APOE, HNF1A, ABCA1, ARID1A, HNF4A, and GNAS. Among the hub genes identified, PPARγ functions as a tumor suppressor in gastric cancer. It encodes the PPARγ, a member of the nuclear hormone receptor superfamily, which primarily regulates cell differentiation, proliferation, and metabolism.[34] Several studies have shown that upregulation of PPARγ expression in gastric cancer cells significantly reduces their proliferative and migratory capacities, highlighting its tumor-suppressive role.[35] As such, therapeutic strategies aimed at activating PPARγ may offer a promising avenue for gastric cancer treatment.[36] IGF1R, by binding to its ligands IGF-1 and IGF-2, activates downstream signaling pathways such as PI3K/Akt and MAPK, thereby promoting cell growth and survival. In gastric cancer, excessive activation of IGF1R has been shown to enhance cancer cell proliferation and resistance to apoptosis.[37,38] LRP2 has been reported to undergo epigenetic silencing, which has been associated with poor outcomes in various malignancies, including clear cell renal cell carcinoma, papillary renal cell carcinoma, mesothelioma, papillary thyroid carcinoma, and invasive breast cancer.[39] HNF4A and APOE are markers of poor prognosis in GC.[40,41] The results of GO and KEGG enrichment analyses indicated that these genes play important roles in multiple metabolic and signaling pathways, which may influence SUA and GC progression by regulating key biological processes and functions such as lipid metabolism, cell membrane dynamics, and signaling pathways. The AMPK signaling pathway and the cholesterol metabolism route have been found to be significantly enriched in pathway analysis. Of these, it has been demonstrated that the AMPK signaling pathway plays a role in the emergence of multiple malignancies, including gastric cancer.[42,43] The development of cancer is also closely linked to cholesterol metabolism.[44] Moreover, complex interactions between genes and pathways are revealed by network analysis, highlighting the necessity of cooperative multigene investigations to comprehend the mechanisms involved in cancer.
Currently, available clinical data does not support the use of SUA level testing to predict GC risk. Nonetheless, in patients with lower SUA levels, doctors should be on the lookout and do routine GC screenings, particularly in those with other GC risk factors such as a family history of GC, an HP infection, or a history of chronic long-term gastritis.[45]
Using genetic variations as IVs, we conducted 2-sample MR analyses in this study to eliminate the reverse causality and confounding bias associated with typical observational studies. Additionally, our sensitivity analysis yielded consistent results. To guarantee accurate results, we also employed large-scale GWAS data for genetic prediction. Lastly, we looked at the possible biological causes of the 2 using bioinformatics and offered some recommendations for further research.
The current study still has certain limitations because our findings are based on data from a European population, and because gene expression varies throughout populations, it is necessary to further verify whether the conclusions drawn can be applied to other groups. Second, more research using cell or animal models is required to confirm any potential biological pathway linking SUA and GC.

5. Conclusion
In this work, we comprehensively examined the causative association between SUA levels and the risk of GC and CRC using a 2-sample MR analysis based on large-scale GWAS summary data. According to our findings, there was a substantial reduction in the risk of GC with greater SUA levels, indicating that SUA might be protective against the onset and progression of GC. There was no discernible link discovered between SUA and CRC risk. Subsequent research endeavors have to persist in investigating the correlation between SUA and additional forms of cancer and verify the possibility of utilizing these discoveries in therapeutic settings.

Author contributions
Conceptualization: Kaichen Han, Yongwen Li, Guangzhi Wu.
Data curation: Kaichen Han, Guangzhi Wu.
Formal analysis: Huaizhu Chu, Guangzhi Wu.
Investigation: Huaizhu Chu, Yongwen Li, Guangzhi Wu.
Methodology: Kaichen Han, Hengheng Zhang, Haigang Liu.
Project administration: Kaichen Han, Haigang Liu.
Resources: Kaichen Han, Yongwen Li, Hengheng Zhang, Haigang Liu, Jianguo Xu.
Software: Kaichen Han, Huaizhu Chu, Yongwen Li, Hengheng Zhang, Haigang Liu, Jianguo Xu.
Supervision: Yongwen Li, Jianguo Xu.
Validation: Kaichen Han, Yongwen Li, Hengheng Zhang, Haigang Liu, Jianguo Xu.
Visualization: Huaizhu Chu, Yongwen Li, Hengheng Zhang, Jianguo Xu.
Writing – original draft: Kaichen Han, Huaizhu Chu, Yongwen Li, Jianguo Xu.
Writing – review & editing: Kaichen Han, Huaizhu Chu, Yongwen Li, Hengheng Zhang, Jianguo Xu.

Supplementary Material