Serotonin, Adenosine Metabolism and Multiple Sclerosis: A Multi-Omics Perspective
The study by Huang and colleagues investigates a provocative hypothesis: serotonin may contribute to multiple sclerosis (MS) pathogenesis by suppressing the activity of adenosine deaminase (ADA), an enzyme that regulates extracellular adenosine metabolism. Rather than treating serotonin solely as a neurotransmitter involved in mood, sleep and behaviour, the authors examine it as an immunologically active metabolite capable of influencing inflammatory signalling. Their central model proposes that genetically elevated serotonin reduces ADA activity, while lower ADA activity is associated with greater susceptibility to MS. To explore this model, the investigators combine Mendelian randomisation, single-cell RNA sequencing, functional enrichment analysis and machine-learning-based biomarker selection. This integrative design is the defining feature of the article, allowing evidence from population genetics to be connected with cell-type-specific transcriptional patterns and diagnostic modelling. The study consequently positions the serotonin–ADA axis at the intersection of neurotransmission, purine metabolism, immune regulation and neurodegeneration.
The Biological Context of Multiple Sclerosis
MS is a heterogeneous autoimmune disease characterised by inflammatory demyelination, axonal injury and progressive neurodegeneration within the central nervous system. Autoreactive T cells, particularly Th1 and Th17 populations, contribute to inflammatory injury, whereas regulatory T cells normally restrain excessive immune activation. B cells, antibodies, microglia and other innate immune populations also participate in lesion formation and sustained tissue damage. Within this complex network, serotonin can signal through several 5-hydroxytryptamine receptors expressed by immune cells, producing context-dependent pro-inflammatory or anti-inflammatory effects. ADA is equally relevant because it converts adenosine and deoxyadenosine into inosine-related metabolites, thereby helping determine the concentration and duration of adenosinergic signals. Changes in ADA activity can therefore affect cytokine secretion, lymphocyte activation, blood–brain barrier function and glial responses. The article argues that an interaction between these two systems could provide a mechanistic bridge between systemic metabolism and immune-mediated neurological injury. Nevertheless, serotonin signalling is biologically pleiotropic: its effects differ by receptor subtype, cell population, tissue compartment and disease stage, meaning that a simple classification of serotonin as universally harmful or protective would be inappropriate.
An Integrative Multi-Omics Study Design
The investigators first conducted two-sample Mendelian randomisation to estimate causal relationships among genetically predicted serotonin levels, ADA activity and MS. Genetic instruments were obtained from genome-wide association and protein quantitative trait locus resources, while the MS outcome dataset included 47,429 cases and 68,374 controls of European ancestry. Instrumental variants were required to be strongly associated with the exposure, independent of relevant confounders and free from pathways affecting MS outside the exposure of interest. The primary estimator was inverse-variance weighting, supplemented by weighted-median and MR–Egger approaches, heterogeneity testing and assessments of horizontal pleiotropy. The workflow diagram on page 3 illustrates how this genetic analysis was integrated with single-cell clustering, differential-expression analysis, Gene Ontology and KEGG enrichment, three feature-selection algorithms and nomogram construction. For the cellular component, the authors analysed the GSE194078 single-cell dataset after excluding cells with very low or very high gene counts or excessive mitochondrial expression. Principal-component analysis, UMAP and t-SNE were used to resolve cellular heterogeneity, after which immune populations were annotated and stratified according to receptor-signalling activity.
Genetic Evidence for the Serotonin–ADA Relationship
The Mendelian-randomisation results supported three connected associations. Genetically predicted serotonin was positively associated with MS, with an inverse-variance-weighted coefficient of β=0.350 and a nominal p value of 3.63×10⁻⁵. Genetically predicted ADA activity was inversely associated with MS, producing β=−0.395 and p=2.73×10⁻⁴. Serotonin was also negatively associated with ADA activity, with β=−0.089 and p=8.70×10⁻³, whereas the reverse analysis did not support a causal effect of ADA on serotonin. Sensitivity analyses reportedly found no substantial horizontal pleiotropy or heterogeneity, and all retained instruments had F-statistics above 10, reducing concern about weak-instrument bias. The scatter plots and forest plot on page 4 visually reinforce the direction of these estimates. Taken together, the results are compatible with a model in which higher serotonin suppresses ADA and thereby increases MS susceptibility. A critical terminological qualification is necessary, however: the outcome dataset was based on case–control MS status rather than longitudinal measurements of disability accumulation, relapse frequency, lesion expansion or neurodegeneration. The analysis therefore provides more direct evidence regarding MS risk or susceptibility than clinical disease progression. The authors acknowledge this distinction, noting that progression was inferred indirectly because suitable longitudinal GWAS data were unavailable.
Cellular Resolution of ADA Expression
Single-cell transcriptomic analysis identified 18 clusters representing six major immune-cell categories: T cells, monocytes, natural killer cells, B cells, common myeloid progenitors and platelets. ADA expression was not uniformly distributed across these populations. It was highest in the combined T–NK compartment and lowest in platelets, directing subsequent analyses toward lymphocyte-associated receptor signalling. When immune cells were divided into high- and low-receptor-signalling groups, ADA expression was higher in the low-signalling group. This observation suggests a relationship between ADA abundance and the activation state of immunometabolic receptor networks, although the cross-sectional transcriptomic data cannot establish that serotonin directly causes the observed expression pattern. Differentially expressed genes were enriched in cytoplasmic translation, RNA splicing, ATP synthesis coupled to electron transport and protein–RNA complex organisation. KEGG analysis highlighted ribosomal pathways, reactive-oxygen-species-associated processes and pathways shared across several neurodegenerative diseases. These findings broaden the proposed mechanism beyond cytokine signalling alone: serotonin-mediated disruption of ADA and adenosine homeostasis could theoretically influence cellular energetics, translational machinery and stress responses in addition to classical immune activation. The UMAP plots and cell-proportion diagrams on page 5 are especially informative because they show both the diversity of the sampled immune compartment and the localisation of ADA expression within specific cell types.
Biomarker Discovery and Diagnostic Modelling
The study also sought to convert its transcriptomic observations into a diagnostic signature. Support Vector Machine–Recursive Feature Elimination identified 16 candidate variables, regression-based selection identified 15 genes, and random forest analysis prioritised 30 genes. Intersecting these outputs yielded three shared features: IK, UBA52 and CCDC25. A nomogram constructed from their expression achieved an area under the receiver operating characteristic curve of 1.000 in the GSE13551 training dataset and 0.976 in the independent GSE21942 validation dataset. Individually, the genes also showed substantial discriminatory performance, although their validation AUCs were lower than that of the combined model. Functionally, the article associates IK with immune and NF-κB-related regulation, UBA52 with ubiquitination and protein homeostasis, and CCDC25 with cytoskeletal organisation and immune-cell migration. The GeneMANIA network displayed on page 9 connects the three markers to genes involved in ribosomal function, protein targeting and localisation to the endoplasmic reticulum. These results are promising but should be interpreted as exploratory rather than clinically definitive. Near-perfect classification in a relatively limited transcriptomic dataset can arise from small sample sizes, dataset-specific batch structure, preprocessing choices or subtle information leakage. Validation in large, prospectively recruited cohorts—including clinically relevant neurological controls—will be required before the nomogram can be regarded as a robust diagnostic instrument.
Translational Significance, Limitations and Future Directions
The study provides a coherent hypothesis-generating framework in which serotonin, ADA-dependent adenosine metabolism and immune-cell signalling converge in MS. Its principal strength is methodological triangulation: genetic instruments reduce some forms of observational confounding, single-cell analysis identifies plausible cellular contexts, and machine learning proposes measurable biomarker candidates. Nonetheless, several limitations restrict immediate clinical interpretation. The genomic datasets were predominantly derived from individuals of European ancestry, potentially limiting transferability to other populations. Mendelian randomisation remains dependent on assumptions concerning instrument validity and horizontal pleiotropy, while serum serotonin-related genetic proxies may not fully represent serotonin concentrations in the central nervous system or within individual immune-cell compartments. Single-cell datasets are vulnerable to technical, batch and annotation effects, and gene expression does not necessarily correspond to enzymatic activity. Most importantly, the proposed pathway has not yet been demonstrated experimentally through controlled manipulation of serotonin, ADA and adenosine signalling in MS-relevant cells or animal models. Future research should therefore quantify serotonin and ADA longitudinally in blood and cerebrospinal fluid, resolve receptor-specific effects, test the pathway in human immune-cell systems and validate IK, UBA52 and CCDC25 in multicentre cohorts. The serotonin–ADA axis is thus best regarded as a biologically plausible therapeutic hypothesis—not yet a validated treatment target—and the reported diagnostic signature as a promising candidate requiring rigorous prospective evaluation.
Disclaimer: This blog post is based on the provided research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.
References:
Huang, L., Shi, J., Li, H., & Lin, Q. (2025). Mendelian randomisation and single-cell transcriptomic analyses reveal serotonin promotes multiple sclerosis progression by suppressing adenosine deaminase activity. BMJ open, 15(9), e102876.
