Causal Relationship Between Butyrate and Dyspepsia: Evidence From Two‐Sample Mendelian Randomization Analysis of CSF Metabolites

ABSTRACT

Background

Dyspepsia is a prevalent gastrointestinal disorder with complex pathogenesis involving the gut‐brain axis. While alterations in gut microbiota have been linked to dyspepsia, the role of central nervous system metabolites, particularly those in cerebrospinal fluid (CSF), remains unexplored.

Objective

To investigate the potential causal relationship between CSF metabolites and dyspepsia using a two‐sample Mendelian randomization (MR) approach.

Methods

We conducted a two‐sample MR analysis using genome‐wide association study (GWAS) summary statistics. CSF metabolite data were derived from 532 individuals across two cohorts, and dyspepsia outcome data were obtained from the UK Biobank (7586 cases and 353 608 controls). Instrumental variables (SNPs) were selected based on genome‐wide significance (p < 5 × 10⁻⁸), with clumping to eliminate linkage disequilibrium. The inverse‐variance weighted (IVW) method was the primary analytical approach, supplemented by MR‐Egger, weighted median, and weighted mode methods. Sensitivity analyses, including heterogeneity tests and MR‐PRESSO, were used to assess the robustness of the findings.

Results

Among 71 CSF metabolites tested, only butyrate (4:0) showed a significant inverse causal association with dyspepsia (IVW OR = 0.997, 95% CI: 0.996–0.998, p < 0.001; P _FDR = 0.007). Sensitivity analyses indicated no evidence of heterogeneity or pleiotropy. Additional enrichment analysis revealed involvement of genes associated with serine‐type peptidase and protein catabolic processes.

Conclusions

Our study provides the first genetic evidence linking elevated CSF butyrate levels to a reduced risk of dyspepsia, highlighting a potential neuroprotective role within the gut‐brain axis.

1. Introduction

Dyspepsia, or called functional dyspepsia, is a prevalent chronic gastrointestinal disorder. According to a large cross‐sectional population‐based study, approximately 10% of the adult population meets the symptom‐based criteria for Rome IV dyspepsia [1]. Various factors are associated with the onset of dyspepsia, including Helicobacter pylori infection and dysbiosis of the gut microbiome [2, 3]. However, as dyspepsia is recognized as a gut‐brain axis disorder, it is critical to understand the neurological mechanisms underlying this condition [4]. The spinal cord plays a pivotal role in the nervous system, acting as a central conduit for transmitting signals from the brain to target organs and providing feedback to the brain regarding organ status [5]. Duan et al. explored the therapeutic potential of Yokukansan (YKS) in treating dyspepsia using an animal model. Their immunohistochemical analysis revealed that YKS treatment reduced stress‐induced gastric hypersensitivity and lowered phosphorylated extracellular signal‐regulated kinase (p‐ERK) 1/2 levels in the spinal cord, suggesting a strong link between spinal cord factors and the pathophysiology of dyspepsia [6]. Cerebrospinal fluid (CSF), the fluid surrounding the brain and spinal cord, contains metabolites closely tied to central nervous system (CNS) activity [7]. In studies by Ballester et al. and Ratuszny et al., CSF metabolites have shown potential as diagnostic tools for CNS‐related conditions, including primary or metastatic CNS tumors and enteroviral meningitis [8, 9]. However, the relationships between CSF metabolites and dyspepsia remain unknown.

Clinically, conducting large cohort studies to investigate the link between CSF metabolites and diseases is challenging, and establishing causal associations is even more difficult. Mendelian randomization (MR) provides a robust framework for investigating such associations and has been widely applied across a variety of diseases [10]. By using single‐nucleotide polymorphisms (SNPs) as instrumental variables (IVs), MR allows for the assessment of causal relationships between exposures and outcomes, with data derived from publicly available genome‐wide association studies (GWAS), eliminating the need for additional clinical samples [11]. MR has been applied to explore causal associations between CSF metabolites and neurological conditions such as Alzheimer's disease and Glioblastoma multiforme [12, 13]. Building on this approach, our study aims to investigate the causal relationship between CSF metabolites and dyspepsia using Mendelian randomization. The goal is to identify specific metabolites that may contribute to the pathogenesis of dyspepsia, potentially offering new insights into the causal associations between them.

2. Methods and Martials

2.1. Study Design

This study aimed to investigate the causal associations between CSF metabolites and dyspepsia using a MR approach. The study adhered to the three fundamental assumptions of MR: (1) the IVs for the CSF metabolites are not related to confounders, (2) the IVs for CSF metabolites influence dyspepsia solely through the exposure‐outcome pathway, and (3) the IVs for CSF metabolites are not directly associated with dyspepsia. Based on these assumptions, the MR results suggest that the selected CSF metabolites have a causal relationship with dyspepsia. The overall study design is illustrated in Figure 1.

FIGURE 1.

Mendelian randomization study design.

2.2. Data Sources

The CSF metabolite GWAS data were originally sourced from the study conducted by Panyared et al. In their research, 532 participants were drawn from two cohorts and a total of 213 718 SNPs are presented in supplementary materials of their study [14].

The GWAS data for dyspepsia were extracted from the UK biobank database. This cohort includes 7586 dyspepsia and 353 608 controls. Further details can be found at https://nealelab.github.io/UKBB_ldsc/h2_summary_K30.html.

2.3. IVs Selection

The IVs for the exposure were selected based on a pvalue threshold of < 5 × 10⁻⁸. To avoid linkage disequilibrium, clumping parameters were set at r ² = 0.001 and a distance of 10 000 kb. The Steiger test was performed to ensure the direction of causality, and any SNPs associated with dyspepsia (Steiger test p value < 0.05) were excluded. Additionally, weak IVs could potentially affect the final outcome, so SNPs with an F statistic < 10 were excluded. The F value was calculated using the following equation: F = (beta/SE) [2], where beta represents the effect estimate and SE is the standard error [15].

2.4. Enrich Analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed for genes mapped from the selected significant SNPs. Enrichment analysis was conducted using the clusterProfiler package [16], with p values < 0.05 considered statistically significant and adjusted p values < 0.05 indicating robust significance.

2.5. Statistics Analysis

The inverse‐variance weighted (IVW) method was employed as the primary analysis approach, and the results obtained through IVW were considered the main findings [17]. In addition, three other analytical approaches—MR‐Egger, weighted median, and weighted mode—were applied [18, 19, 20]. The results were evaluated from two key perspectives: if the findings from the additional three methods were statistically significant and if the odds ratios (ORs) were consistent in direction with the IVW results, this would further strengthen the conclusions. Analyses using all methods were only conducted when the number of SNPs exceeded two; if there were exactly two SNPs, only the IVW method was used.

The robustness of the MR results was validated through sensitivity analysis. A heterogeneity test was performed, with a Q‐test p‐value greater than 0.05 indicating no heterogeneity [21]. A pleiotropy test was also conducted, where a p value greater than 0.05 suggested the absence of pleiotropy. Additionally, the MR‐PRESSO Global Test was conducted in conjunction with the pleiotropy test, with a p value greater than 0.05 further confirming the absence of pleiotropy. However, if the number of SNPs was fewer than four, the Global Test could not be performed due to limitations [22].

All analyses, including IVW, MR‐Egger, weighted median, and weighted mode, considered pvalues less than 0.05 as suggested statistically significant and FDR p values less than 0.05 as statistically significant. The statistical analysis was performed using R software version 4.4.1, specifically with the TwoSampleMR package (version 0.6.0), the MendelianRandomization package (version 0.8.0), and the MR‐PRESSO package (version 1.0).

3. Results

3.1. Butyrate Levels in CSF Was Negative Causal Associated With Dyspepsia

In total, 71 CSF metabolites had a sufficient number of IVs to conduct the MR analysis with dyspepsia. After screening, a total of 2401 SNPs were available for these 71 CSF metabolites, and the mean F‐statistic for these SNPs was 37.90. Detailed information on these IVs is provided in Table S1.

When CSF metabolites were used as exposures and dyspepsia as the outcome, the results indicated that only the level of butyrate (4:0) in CSF was negatively and causally associated with dyspepsia. As shown in Figure 2A, the MR analysis using the IVW method yielded an OR of 0.997 (95% CI: 0.996–0.998, p < 0.001, P _FDR = 0.007). The MR‐Egger method produced an OR of 0.998 (95% CI: 0.993–1.003, p = 0.452). Figure 2B presents the forest plots of the MR analyses between butyrate and dyspepsia, where the weighted median method showed an OR of 0.998 (95% CI: 0.996–1.000, p = 0.086), and the weighted mode method yielded an OR of 0.999 (95% CI: 0.995–1.003, p = 0.603). Figure 3 illustrates the directionality of OR values across the four MR methods: light blue represents IVW, dark blue denotes MR‐Egger, light green indicates the weighted median, and dark green refers to the weighted mode. The consistent trend of the lines across all methods suggests robust and reliable results. Additionally, the funnel plot in Figure 4 shows that the effect sizes of individual SNPs are symmetrically distributed around the IVW and MR‐Egger estimates, with a Q‐test p value of 0.496, indicating no significant heterogeneity. Figure 5 visualizes the leave‐one‐out analysis, demonstrating that the overall effect size remained on the same side of the null line even when individual SNPs were sequentially excluded, further supporting the absence of heterogeneity. Moreover, tests for horizontal pleiotropy showed no significant evidence of bias, with a pleiotropy test P value of 0.690 and MR‐PRESSO global test p = 0.690.

FIGURE 2.

Forest plots: (A) Forest plot of SNP effects of butyrate (4:0) levels on dyspepsia; (B) Forest plot showing Mendelian randomization estimates.

FIGURE 3.

Scatter plots of causal effect estimates for Butyrate (4:0) levels on dyspepsia.

FIGURE 4.

Funnel plots of causal effect estimates for Butyrate (4:0) levels on dyspepsia.

FIGURE 5.

Leave‐one‐out sensitivity analysis of causal effect estimates for Butyrate (4:0) levels on dyspepsia.

The remaining forward MR analysis results are presented in Table S2, where none of the associations remain statistically significant after FDR correction. Tables S3 and S4 display the sensitivity analyses for the forward MR analyses. Table 1 shows the SNPs related to butyrate used in the MR analysis, together with their mapped genes.

TABLE 1.

SNPs information of butyrate used in MR analysis with mapping gene.

SNP	Mapping gene	Effect allele	Other allele	eaf	beta	se	p	F power
rs16922855	ARMC3	A	G	0.0212	−0.2655	0.0405	5.30E‐11	42.975309
rs79417281	CMIP	G	T	0.0119	−0.335	0.0504	2.99E‐11	44.180288
rs12500575	a	T	C	0.0214	−0.2895	0.0496	5.23E‐09	34.066992
rs116952419	DPP6	A	G	0.0272	−0.2237	0.0391	1.08E‐08	32.732445
rs188830987	LDLRAD3	A	G	0.0058	−0.4548	0.0821	3.06E‐08	30.687012
rs12628385	LOC107985590	C	T	0.0069	−0.3924	0.0707	2.90E‐08	30.804855
rs73694251	MMP16	A	G	0.0156	−0.2692	0.0477	1.70E‐08	31.850287
rs140958854	ENTPD5	C	G	0.0081	−0.4152	0.0707	4.24E‐09	34.488624
rs71315482	a	G	A	0.0196	−0.2632	0.0471	2.34E‐08	31.226978
rs73238622	CORIN	C	T	0.0151	−0.317	0.057	2.69E‐08	30.929209
rs147867908	a	A	G	0.009	−0.429	0.0585	2.22E‐13	53.777778
rs141142047	TMPRSS7	C	T	0.0055	−0.4492	0.0813	3.35E‐08	30.528004
rs140849076	FBXW7	T	C	0.0068	−0.5743	0.0646	6.08E‐19	79.033751
rs78516608	GPC5	T	G	0.009	−0.4031	0.0584	4.91E‐12	47.643088
rs74024498	RASGRF1	T	C	0.0248	−0.2552	0.0404	2.82E‐10	39.902363
rs139117943	LOC105372121	G	A	0.0121	−0.3076	0.0558	3.56E‐08	30.38815
rs117350993	a	G	A	0.0092	−0.3172	0.0511	5.53E‐10	38.532267
rs979582	a	C	A	0.9932	0.4673	0.072	8.62E‐11	42.123706
rs114242793	AGBL4	T	C	0.0123	−0.3049	0.0551	3.17E‐08	30.620456
rs189212987	a	T	C	0.0249	−0.2427	0.0394	7.34E‐10	37.944349
rs56401047	a	C	T	0.0212	−0.2916	0.0479	1.13E‐09	37.05988
rs11905541	a	T	C	0.0108	−0.409	0.0549	9.41E‐14	55.501143
rs143579526	a	G	A	0.0085	−0.4629	0.0673	6.03E‐12	47.309051
rs72682115	a	A	G	0.0168	−0.3005	0.0468	1.36E‐10	41.228473
rs12145681	a	T	C	0.0349	−0.1847	0.0331	2.42E‐08	31.137074
rs72914649	HECW2	A	G	0.0109	−0.3515	0.0594	3.19E‐09	35.016906
rs10997796	PRKG1	T	C	0.018	−0.2797	0.0475	3.84E‐09	34.673502
rs78165339	LOC105372666	T	C	0.0114	−0.4058	0.0601	1.48E‐11	45.590583
rs115889085	a	T	G	0.0138	−0.3388	0.0559	1.38E‐09	36.733574
rs76523291	a	A	G	0.0136	−0.293	0.0535	4.43E‐08	29.993537
rs181959785	a	T	C	0.0281	−0.2473	0.0408	1.30E‐09	36.739049
rs11236439	GDPD5	T	C	0.0373	−0.167	0.0299	2.39E‐08	31.1954
rs112536160	a	C	A	0.0115	−0.3568	0.0602	3.11E‐09	35.128266
rs62172304	THSD7B	C	G	0.0258	−0.2402	0.0412	5.39E‐09	33.990032
rs141262213	a	G	T	0.0284	−0.2696	0.0409	4.17E‐11	43.450338
rs114467160	a	A	T	0.0192	−0.2523	0.0463	4.99E‐08	29.694261

^a Means no mapping gene available.

3.2. KEGG and GO Analysis

To investigate the potential biological functions and signaling pathways associated with the differentially expressed genes (DEGs), we performed GO and KEGG enrichment analyses (Figure 6). In the Gene Ontology (GO) analysis, under the Molecular Function (MF) category, DEGs were significantly enriched in terms such as serine‐type peptidase activity (P _adj < 0.001) and serine hydrolase activity (P _adj < 0.001), involving the genes DPP6, CORIN, and TMPRSS7, suggesting a possible role in serine protease‐mediated protein degradation. In addition, metalloexopeptidase activity was also significantly enriched (P _adj = 0.001), with contributions from MMP16 and AGBL4. Within the Biological Process (BP) category, DEGs were significantly enriched in positive regulation of protein catabolic process (P _adj < 0.001) and protein processing (P _adj = 0.001), involving genes such as FBXW7, AGBL4, HECW2, LDLRAD3, MMP16, and CORIN, indicating their potential involvement in regulating protein degradation and post‐translational modifications. The KEGG pathway analysis further revealed significant enrichment in pathways such as Regulation of lipolysis in adipocytes (P _adj = 0.034), mediated by PRKG1, as well as Pyrimidine metabolism (P _adj = 0.035) and Long‐term depression (P _adj = 0.036), associated with ENTPD5 and PRKG1, respectively. Other pathways, including Nucleotide metabolism (P _adj = 0.051) and Gap junction (P _adj = 0.053), showed borderline significance. Collectively, these findings suggest that the DEGs are primarily enriched in functions and pathways related to protease activity, protein catabolism, lipid metabolism, and nucleotide metabolism, highlighting their potential roles in cellular metabolic regulation and function, and all enrichment analysis were displayed in Table S9.

FIGURE 6.

Bubble plots of GO and KEGG enrichment analysis for genes mapped from the selected significant SNPs.

3.3. Reverse Mendelian Randomization Analysis

When dyspepsia was treated as the exposure and CSF metabolites as the outcomes, no significant associations were observed, either before or after adjustment. The instrumental variables used in these analyses are listed in Table S5. All results from the reverse MR analysis are presented in Table S6, and the corresponding sensitivity analyses are shown in Tables S7 and S8.

4. Discussion

Our study employed Mendelian randomization to identify causal relationships between 86 CSF metabolites and dyspepsia. The findings offer a novel perspective on dyspepsia by linking central metabolites, particularly CSF butyrate, to its pathogenesis. The results offer a new perspective on dyspepsia by connecting central metabolites, especially CSF butyrate, to its development. Unlike most prior dyspepsia studies that mainly focused on peripheral factors like Helicobacter pylori infection and changes in gut microbiota, our approach takes a different perspective. Recent research indicates that dyspepsia patients often have imbalances in duodenal microbiota, such as increased Streptococcus and decreased butyrate‐producing Butyricicoccus that are associated with symptom severity [23]. A reduction in Butyricicoccus, a key producer of the short‐chain fatty acid butyrate, appears characteristic of dyspepsia [24]. These findings align with our observation that higher CSF butyrate levels are linked to a lower risk of dyspepsia, implying a protective effect. Both microbiota studies and our CSF‐focused analysis support the idea that butyrate might be beneficial in dyspepsia, highlighting a deficiency of butyrate‐producing microbes in the gut and providing genetic evidence that butyrate within the CNS can causally decrease dyspepsia risk. This complementary perspective emphasizes the microbiota‐gut‐brain axis as a unified framework: disturbances in the gut, such as dysbiosis and lower butyrate production, along with alterations in CNS butyrate levels, may reflect different aspects of the same underlying mechanism in dyspepsia development.

Our research is the first to link cerebrospinal fluid metabolites with dyspepsia, unlike earlier metabolomic studies which primarily examined neurological disorders. CSF metabolite profiles have shown diagnostic or disease relevance in various CNS conditions; for instance, certain CSF metabolites serve as biomarkers for brain tumors and enteroviral meningitis. Recent genetic studies utilizing MR have connected CSF metabolite levels to neurodegenerative diseases like Alzheimer's and brain cancers. However, such studies had not been applied to functional gastrointestinal disorders prior to our work. The lack of comparable research underscores the uniqueness of our findings. We bridge neurology and gastroenterology by proposing that CNS metabolites, reflected in CSF, can influence a functional gut disorder. This novel insight supports earlier evidence that dyspepsia involves central dysregulation, such as stress‐induced changes in the spinal cord, like altered ERK signaling which can impact gastric sensitivity in animal models. Overall, our study adds a biochemical perspective to the brain‐gut axis in dyspepsia, consistent with existing gut microbiome findings, such as the beneficial role of butyrate in dyspepsia.

These findings have important implications for understanding and treating dyspepsia. Recognizing butyrate as a protective factor supports the gut‐brain axis model of dyspepsia, where metabolic signaling is involved. Traditionally, dyspepsia was linked to gastric motility issues, visceral hypersensitivity, low‐grade inflammation, and psychosocial factors. Our study suggests that central metabolite levels can directly influence symptom development, implying that biochemical mediators play a role alongside neural and immune interactions within the gut‐brain axis. The inverse relationship between butyrate and dyspepsia risk indicates that it may exert neuroprotective or modulatory effects within this axis. Mechanistically, short‐chain fatty acids like butyrate can cross the blood–brain barrier and exert anti‐inflammatory and neurotrophic effects in the CNS. For example, butyrate can decrease neuroinflammatory cytokine production and promote brain‐derived neurotrophic factor release, potentially reducing central sensitization or neuroinflammation associated with dyspeptic symptoms [25]. There is growing support from mechanistic studies, animal experiments, and clinical metabolomic data for the biological plausibility of how gut‐derived butyrate can influence the central nervous system and ultimately affect dyspepsia. First, butyrate produced by the gut microbiota can be efficiently absorbed through colonic epithelial cells, enter the portal vein and systemic circulation, and be stably detected in human plasma, indicating that butyrate can reach the central nervous system [26]. Second, studies have shown that butyrate injected via the carotid artery can rapidly accumulate in brain tissue within seconds, a process that depends on the expression of monocarboxylate transporter MCT1 and fatty acid transporter FAT/CD36 on brain microvascular endothelial cells [27]. Third, clinical metabolomic studies have confirmed that butyrate can be detected in human cerebrospinal fluid, where it coexists with other SCFAs such as acetate and propionate [28]. In addition, studies of paired serum and cerebrospinal fluid samples have shown that multiple gut‐derived metabolites can be detected in both body fluids, and that their cerebrospinal fluid concentrations correlate with indices of blood–brain barrier integrity, suggesting dynamic exchange between gut microbiota–derived metabolites, especially SCFAs and the CNS [29]. In summary, our findings expand the dyspepsia model by integrating CNS metabolite homeostasis and encourage future research on the gut‐brain axis to include metabolomic profiling.

However, the limitations of this study include. First, focusing on a European population introduces some geographic and genetic constraints. Second, compared to other GWAS studies on metabolites, the sample size of the original CSF metabolite GWAS used in this study is relatively small, which might lead to less precise estimates of metabolite‐SNP associations, potentially biasing the MR analysis or reducing its ability to detect causal effects. Third, there's a risk of instrument selection bias and pleiotropy in our MR approach. We selected instrumental SNPs and performed clumping, but if a SNP affects dyspepsia via another pathway or is linked to a related metabolite, the MR assumptions could be violated. To avoid this, we conducted pleiotropy tests and found no widespread pleiotropy for butyrate, but subtle effects cannot be completely ruled out. Because genetic variants often impact multiple metabolites, caution must be exercised when attributing causality to a single metabolite like butyrate. Future research may benefit from multivariable MR or network analyses. Fourth, given multiple hypothesis tests, 71 metabolites examined, only the butyrate association remained significant after false discovery rate correction. While this emphasizes the robustness of the butyrate finding, it also raises the chance of missing other true associations. Some metabolites showed nominal significance, but no further claims were made. Larger samples or more sensitive data could reveal additional causal factors.

5. Conclusion

This study is the first to explore the relationship between CSF metabolites and dyspepsia, and we have identified four CSF metabolites with statistically significant causal associations with the condition. We encourage future research to investigate the effects of metabolite interventions in the CSF of animal models to further elucidate the mechanisms of the gut‐brain axis in dyspepsia.

Funding

The authors have nothing to report.

Ethics Statement

Our study used a publicly available GWAS database that had been approved by the Ethics Committee and no new data were collected, so no additional ethical approval was required.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: SNP information in the forward MR study.

Table S2: Full outcomes of SNP effects of CSF metabolites on dyspepsia.

Table S3: Heterogeneity test results from the forward MR analysis.

Table S4: Horizontal pleiotropy assessment in the forward MR analysis.

Table S5: SNP information in the reverse MR study.

Table S6: Full outcomes of SNP effects of dyspepsia on CSF metabolites.

Table S7: Heterogeneity test results from the reverse MR analysis.

Table S8: Horizontal pleiotropy assessment in the reverse MR analysis.

Table S9: Enrichment analysis results.

Cai Y., Jiang Y., Zhou H., et al., “Causal Relationship Between Butyrate and Dyspepsia: Evidence From Two‐Sample Mendelian Randomization Analysis of CSF Metabolites,” JGH Open 9, no. 12 (2025): e70333, 10.1002/jgh3.70333.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.