Introduction

Bipolar disorder (BD) is a chronic and disabling psychiatric illness marked by recurrent depressive and (hypo) manic episodes, affecting about 1.1% of individuals over their lifetime [1]. Research indicates that around 70% of individuals with BD exhibit clinical manifestations by the age of 25, with the highest risk of onset occurring between 15 and 19 years [2]. Beyond its profound impact on individual well-being, BD also imposes significant social and economic burdens, contributing to reduced quality of life, disability, and even premature mortality [3]. As a result, the early identification of reliable biomarkers for the diagnosis of BD is of paramount importance.

Psychiatric disorders such as BD are closely linked to gut-brain axis dysfunction, with the gut contributing to digestion, nutrient absorption, and maintenance of internal homeostasis [4]. BD significantly impacts both neural and gastrointestinal systems, with inflammation potentially serving as both a cause and consequence of alterations in the gut microbiota [5]. Notably, a causal relationship has been suggested between altered mucosal permeability and intestinal inflammation in BD patients, with marked changes in β-Proteobacteria [6]. These alterations involve higher levels of E. coli and Bifidobacterium adolescentum with reduced Faecalibacterium. Moreover, BD is associated with disturbances in gut metabolism [7], particularly affecting amino acid, carbohydrate, and lipid metabolism [8].

Several studies have investigated gut microbiota changes in BD patients. Aizawa et al. reported that although Bifidobacterium and Lactobacillus are not the major pathological factors of BD, these bacteria may be associated with patients’ sleep and stress [9]. Chen et al. examined gut microbiota profiles among young adults with schizophrenia, bipolar disorder, and healthy controls, and identified potential diagnostic biomarkers for these disorders [10]. Evans et al. found a significant reduction of Faecalibacterium in BD patients versus healthy controls., and further observed an association between its levels and clinical manifestations [11]. Taken together, these studies underscore microbial alterations in BD, although most investigations have relied on 16S rRNA sequencing.

Proteomics based on mass spectrometry provide a highly effective and reliable technique for accurately quantifying numerous proteins in biological samples [12]. This method enables the identification of differentially expressed proteins (DEPs) between pathological and normal conditions, making it a valuable tool for discovering potential biomarkers of psychiatric disorders [13]. Altered protein profiles across the gut–brain axis impact various biological functions, such as metabolic processes and inflammatory responses [14]. We propose that BD significantly impacts the protein profile of the gut, influencing both metabolic and inflammatory pathways.

Most current studies on the proteome of BD have primarily focused on cerebrospinal fluid, blood, and urine. Previous studies have identified four cerebrospinal fluid proteins—testican-1, CLEC1B, draxin, and TNFSF21—as associated with BD [15], which also revealed notable differences in the serum metabolome of BD patients compared with healthy controls (73.54% variation). These differences include altered levels of microbial-derived neuroactive metabolites, B vitamins, stilbenes, γ-aminobutyric acid, and short-chain fatty acids [16].

Building on prior research, using metaproteomics, we examined gut microbiota, microbial proteins, and host proteins in adolescents with BD (N = 15) versus healthy controls (N = 58). We further explored alterations in gut proteins and their associations with microbial composition, and proposed potential biomarkers that may aid in the diagnosis of BD.

Method

Participants

This study was approved by the Ethics Committee of the Fifth Affiliated Hospital of Sun Yat-sen University (Approval No. 2020 K32-1). Fifteen adolescents with BD and 58 healthy controls (HCs) were recruited between May 2020 and October 2024 for this study. HCs were slightly older (p.adj ≤ 0.05) whereas sex and BMI showed no significant differences between groups. Participant characteristics are listed in Table 1 and Table S1. Two senior psychiatrists independently confirmed BD diagnoses using the Brief International Neuropsychiatric Disorder Interview. Severity of depression and mania in participants was measured using the Hamilton Depression Scale (HAMD) and Young’s Mania Scale (YMRS), respectively.

Table 1 Demographic characteristics.
Full size table

The inclusion criteria included: (i) Aged 12–35 years old, any sex; (ii) Adolescent patients with BD diagnosed with depressive episodes according to The International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) criteria; (iii) Current 17-item Hamilton Depression Scale (HAMD-17 item) score ≥8; (iv) Current Young Mania Rating Scale (YMRS) score ≤20 and (v) Voluntary informed consent signed by the person and/or his/her guardian.

The exclusion criteria included: (i) Combination of any mental disorders, such as organic brain disorders, mental retardation, substance dependence or abuse, personality disorders; (ii) Serious physical diseases, such as cerebrovascular, respiratory, hematologic, hepatic, renal, endocrine, neurological, and other systemic disorders; and (iii) Pregnancy.

Healthy subjects, matched with age and sex, were required to have no DSM-IV Axis I or II disorders, neurodegenerative, severe endocrine, or metabolic diseases (e.g., diabetes, gout), no family history of psychiatric disorders, no hormone therapy, and no pregnancy. Moreover, no antibiotics, probiotics, or prebiotics were consumed during the preceding month.

Sample collection and protein extraction

Samples were collected, processed, and stored according to standardized protocols. Briefly, the fresh fecal sample was collected and stored at −80 °C within 30 min. Approximately 50 mg of feces was weighed into a 2 mL tube and dissolved in lysis buffer containing 10 mM tris(2-carboxyethyl) phosphine (TCEP), 40 mM chloroacetamide (CAA) and 6 M Guanidinium hydrochloride (GdmCI) in 100 mM Tris buffer (pH = 8.5) for 15 min at 95°C for one-step reduction and alkylation. The protein lysis solution was sonicated for 15 min, centrifuged at 16,000 × g for 15 min at 4 °C, and the resulting supernatant was concentrated to approximately 200 µl using a 30 kDa cutoff filter. From this concentrated supernatant, 50 µl was taken and dissolved in digestion solution containing 4 µg/ml trypsin, achieving a final enzyme-to-protein ratio of 1:100. After mixing the digestion solution in a thermomixer at 37 °C and 500 rpm for 10–18 h, the obtained peptide mixture sample was purified using a C18 extraction column, subsequently subjected to drying treatment, and stored at −80 °C. Peptide mixtures were resuspended in a loading buffer containing 2% (v/v) acetonitrile (ACN) and 0.01% (v/v) trifluoroacetic acid (TFA) prior to liquid chromatograph mass spectrometer (LC-MS) acquisition.

Proteomic data acquisition

Analyses were performed using an M-class liquid chromatography system (Waters) coupled to a ZenoTOF 7600 mass spectrometer (SCIEX) with an OptiFlow ion source. The Luna 5 μm C18 micro trap column (300 μm × 20 mm, Phenomenex) ran at 10 μL/min, while the Kinetex XB-C18 analytical column (300 μm × 150 mm, 2.6 μm, 100 Å, Phenomenex) was used at 5 μL/min with a 55-min gradient. The ion source was set with the following parameters: curtain gas 50, CAD gas 7, Gas 1 (nebulizer) 30, Gas 2 (heater) 60, temperature 350 °C, positive polarity, ion spray voltage 4500 V, declustering potential 80 V, and collision energy (CE) 10 V. Data were collected in data-dependent acquisition (DDA) mode, with a full MS scan from m/z 350–1500 (100 ms accumulation), followed by MS/MS scans of the top 40 precursor ions (charge 2–6, intensity >100 cps) over m/z 100–2000 (20 ms accumulation).Precursor ions were excluded for 6 s after one MS/MS acquisition with a mass tolerance of 50 mDa. Collision energy for MS/MS was dynamically determined using the equation CE = (slope) × (m/z) + (intercept), with slope and intercept set to 0.049 and 5 for charge 2⁺, 0.048 and –2 for charge 3⁺, and 0.05 and –2 for charges 4–6⁺, respectively. A collision energy spread of 5 V was applied for MS/MS, and Zeno pulsing was enabled to enhance sensitivity (Zeno threshold: 100,000 cps).

Taxonomic and functional analysis of metaproteomics data

Using the Lowest Common Ancestor (LCA) algorithm, microbial taxonomic and functional analysis of peptides identified in the metaproteomics workflow was further performed using UniPept (https://unipept.ugent.be/). The search settings were as follows: equate I and L; filter duplicate peptides; advanced missing cleavage handling. Peptides were annotated with Gene Ontology (GO) terms and EC numbers, and the normalized intensities of peptides were used to quantify the relative abundances of microbial taxa and functional groups.

Metaproteomics data processing

Proteins were identified using the PEAKS search engine. Proteins were identified by matching tandem mass spectrometry (MS/MS) spectra against a comprehensive database comprising host sequences (UniProt human reference proteome), microbial sequences, and common dietary organism sequences, with duplicate peptide removal performed using USEARCH v11.0.667. To enhance taxonomic coverage, microbial protein sequences were compiled from multiple public repositories of culture-dependent and culture-independent microbial sequences, including all archaeal, bacterial, and fungal sequences from NCBI RefSeq and UniProtKB, 1520 reference genomes derived from over 6,000 human fecal strains, 215 bacterial genomes isolated from human fecal samples [17], and an integrated gene catalog from human fecal metagenomes [18].

Bioinformatic analysis

Bioinformatic analyses were performed in R (v4.3.2). Raw LC–MS data were denoised and normalized, and α-diversity was assessed using the diversity and specnumber functions from vegan. β-diversity was evaluated with Bray–Curtis distances (vegan::vegdist) followed by principal coordinates analysis (ape::pcoa). Between-group differences were tested using two approaches: linear modeling with empirical Bayes correction (limma::lmFit and eBayes) and Wilcoxon rank-sum tests (stats::wilcox.test). To evaluate the diagnostic potential of differentially expressed host proteins, receiver operating characteristic (ROC) curves and precision–recall (PR) curves were generated using the ROCR::roc and PRROC::pr.curve functions, respectively, with diagnostic performance quantified by the area under the curves (AUC). Spearman rank correlations were calculated using psych::corr.test to construct core molecular interaction networks. Multiple testing correction was applied using fdrtool, with adjusted p values (P.adj) ≤ 0.05 (two-sided) considered significant. Data visualization was conducted with ggplot2, pheatmap, and Cytoscape. Regarding code availability, analyses were performed using standard workflows of publicly available R packages, as detailed above. The scripts are available from the corresponding author upon reasonable request.

Result

Mass spectrometry-based proteomics analysis of gut microbiota reveals significant differences between BD patients and HCs

A total of 73 participants were enrolled, comprising 15 in the BD group and 58 in the HC group. Fecal samples were analyzed using mass spectrometry-based proteomics to determine the intestinal microbiota composition. Statistical analyses revealed that Shannon and Simpson indices did not differ significantly between the groups (Fig. 1A, B). In contrast, the Chao1 index indicated that, compared with HCs, the BD group showed reduced species richness (p.adj ≤ 0.05, Fig. 1C). This indicates that BD patients have lower gut microbiota species richness than healthy controls.Furthermore, no significant differences in overall gut microbiota composition between BD and HC groups were observed via Principal Coordinates Analysis (PCoA) analysis(Fig. 1D), indicating that the general community structure was largely comparable.

Fig. 1: Changes in the general composition of the gut flora in teenagers suffering from BD.
Fig. 1: Changes in the general composition of the gut flora in teenagers suffering from BD.
Full size image

Shannon (A), Simpson (B), and Chao 1 (p.adj < 0.05) (C) indices of intestinal flora species richness in BD and HC groups. (D) Bray-curtis distance-based PCoA visualisation analysis of gut microbiota species richness between BD and HC groups.Composition of intestinal microorganisms at phylum (E), order (F), and class (G) taxonomic hierarchies between BD and HC groups.

The phylum-level composition was dominated by Bacteroidota and Bacillota, which constituted 69.84% of the BD group and 83.6% of the HCs (Fig. 1E). At the class level, the gut microbiota was predominantly composed of Clostridia, Bacteroidia, Erysipelotrichia, and Bacilli, accounting for 48.46% in BD group and 54.96% in HCs (Fig. 1F). At the order level, the gut microbiota was mainly composed of Eubacteriales, Bacteroidales, and Erysipelotrichales, accounting for 48.29% in BD groud and 52.73% in HCs (Fig. 1G).

Classification and functional alterations of differential gut microbiota in BD

Wayne’s intestinal microbiota profile (Fig. 2A) demonstrates that the intestinal microbiota of BD group and HC group share overlapping components. Analysis of variance using limma and FDR correction identified 23 differentiating gut microbiota between the two groups (p.adj ≤ 0.05, |log2FC| ≥ 1) (Fig. 2B). Specifically, the microbes reduced in the intestines of adolescent BD patients include: two classes (Erysipelotrichia, Bacilli), two orders (Erysipelotrichales, Lactobacillales), one family (Erysipelotrichaceae), one genus (Faecalitalea), and three species (Parabacteroides distasonis, Bacteroides uniformis, Faecalitalea cylindroides). Interestingly, the abundance of Lactobacillales in the HC group accounted for 2.25%, whereas in the BD group, the abundance of this bacterium decreased sharply to nearly zero. In contrast, the microbes enriched in the intestines of adolescent BD patients include: one phylum (Actinomycetota), one class (Alphaproteobacteria), four orders (Actinomycetales, Bifidobacteriales, Hyphomicrobiales, Veillonellales), four families (Bifidobacteriaceae, Rikenellaceae, Nitrobacteraceae, Veillonellaceae), three genera (Alistipes, Bradyrhizobium, Megasphaera), and one species (Alistipes putredinis). Overall, the gut microbiota of adolescents with BD exhibits significant compositional changes.

Fig. 2: Analysis of differences in gut flora between adolescent BD patients and healthy controls.
Fig. 2: Analysis of differences in gut flora between adolescent BD patients and healthy controls.
Full size image

(A) Wayne’s chart of gut microbes counts. (B) Limma difference analysis and fdr correction of gut microbes in adolescent BD patients and healthy controls. (C)The Wilcoxon rank-sum test revealed a significant difference (p.adj < 0.05, |log2FC| > 1) in the bacterial function GO between the group of adolescent BD patients and the group of healthy controls.

Furthermore, the Wilcoxon rank-sum test revealed significant differences in bacterial functional GO terms between adolescent BD patients and healthy controls (p.adj ≤ 0.05, |log2FC| ≥ 1). Specifically, ten biological processes were up-regulated, while nine biological processes were down-regulated (Fig. S2A). In terms of cellular localization, the motile cilium, beta-galactosidase complex, and pore complex were down-regulated, whereas the cell outer membrane and small ribosomal subunit were up-regulated (Fig. S2B). Additionally, sixteen molecular functions were found to be up-regulated, and nineteen were found to be down-regulated (Fig. 2C).

Proteomic profiling reveals altered bacterial protein expression in BD patients

By applying limma analysis of variance with FDR correction, we were able to further investigate the differences in intestinal bacterial proteins between adolescents with BD and HCs. The comparison between the two groups revealed statistically significant changes in bacterial protein expression (p.adj ≤ 0.05, |log2FC | ≥ 1). As shown in Fig. 3, BD samples exhibited significantly higher levels of several proteins, including 60 kDa chaperonin/Chaperonin GroEL (GroEL), DNA-directed RNA polymerase subunit beta (RpoB), TonB-dependent receptor (TBDR), and bifunctional acetaldehyde-CoA/alcohol dehydrogenase (AdhE). In contrast, BD samples also demonstrated significant down-regulation of proteins such as glutamate dehydrogenase (GDH), D-proline reductase proprotein (PrdA), malate dehydrogenase (MDH), bifunctional 4-hydroxy-2-oxoglutarate aldolase/2-dehydro-3-deoxy-phosphogluconate aldolase (HOGA / EDA), succinate-CoA transferase (ScoAT), glycine cleavage system protein H (GcvH), OmpA family protein/porin (OmpA), and OmpA family protein (OmpA).

Fig. 3: Proteomic profiling reveals altered bacterial protein expression in BD patients.
Fig. 3: Proteomic profiling reveals altered bacterial protein expression in BD patients.
Full size image

Analysis of variance by limma and correction by fdr showed significant differences in bacterial proteins between adolescent BD patients and healthy controls (p.adj < 0.05, |log2FC| > 1). (A) Heat map showing differential gut bacterial proteins in adolescent BD patients versus healthy controls. (B) Volcano plot showing specific changes in gut bacterial proteins in adolescent BD patients compared to healthy controls.

Proteomic profiling reveals altered human protein expression in BD patients

A limma analysis of variance with FDR correction showed significant differences in intestinal host proteins between adolescents with BD and HCs (p.adj ≤ 0.05, |log2FC| ≥ 1). As shown in Fig. 4A, B, several proteins were significantly upregulated in BD samples, including Ig mu chain C region (IGHM), annexin A2 (ANXA2), immunoglobulin heavy constant alpha 1 (IGHA1), isoform Somatic-2 of angiotensin-converting enzyme (ACE), Ig kappa chain V-III region NG9 (Igκ V-III NG9), and tropomyosin alpha-1 chain (TPM1). In contrast, BD samples showed significant downregulation of proteins such as kallikrein-1 (KLK1), alpha-amylase 2B (AMY2B), lithostathine-1-beta (REG1B), chymotrypsin-like elastase family member 2 A (CELA2A), elastase 3 A, pancreatic (ELA3A), DEFA3, IgGFc-binding protein (FCGBP), and chymotrypsin-C isoform X1 (CTRC-X1). These results indicate significant alterations in both microbial and host proteins in the gut of BD patients. To assess the diagnostic potential of host proteins while accounting for potential bias from imbalanced sample sizes, ROC and PR curves were applied (Fig. 4C, D). ROC-AUC values of CELA2A, DEFA3, and KLK1 were 0.905, 0.897, and 0.897, respectively, with corresponding PR-AUC values of 0.978, 0.975, and 0.975.These findings suggest that CELA2A, DEFA3, and KLK1 may serve as promising biomarkers for BD.

Fig. 4: Proteomic profiling reveals altered human protein expression in BD patients.
Fig. 4: Proteomic profiling reveals altered human protein expression in BD patients.
Full size image

(A) Heat map showing differential gut-host proteins in adolescent BD patients versus healthy controls. (B) Volcano plot showing specific alterations in gut human proteins in adolescent BD patients compared to healthy controls. (C) ROC curve and AUC values showing the diagnostic capability of the human intestinal proteins. (D) PR curve and AUC values showing the diagnostic performance of human intestinal proteins.

Correlation analysis of differential gut microbiota, microbial GO functions, microbial proteins, and host proteins

Spearman correlation analyses were performed on gut microbiota, microbial GO function data, and proteins (including microbial and host proteins) (Fig. 5). Results indicated a strong positive correlation between CELA2A and both Lactobacillales and Bacteroides uniformis. CELA2A, KLK1, MDH, LacI, OmpA/Porin, GcVH, REG1R, and AMY2B were positively correlated with Faecalitalea cylindroides as well as its corresponding family (Erysipelotrichaceae) and class (Erysipelotrichia). DEFA3 and KLK1 were strongly positively correlated with Parabacteroides distasonis, whereas IGHM exhibited a strong negative correlation with Parabacteroides distasonis.

Fig. 5: Correlation Analysis of Differential Gut Microbiota, Microbial GO Functions, Microbial Proteins, and Host Proteins.
Fig. 5: Correlation Analysis of Differential Gut Microbiota, Microbial GO Functions, Microbial Proteins, and Host Proteins.
Full size image

Spearman correlations were calculated for microbiota, microbial GO, and protein (microbial and host) data, and interaction networks were visualized in Cytoscape.

Disscussion

We used metaproteomics to investigate alterations in intestinal proteins and gut microbiota in the feces of adolescents with BD. Our analysis revealed significant changes in both intestinal flora and human protein levels. The altered intestinal flora were primarily involved in lactic acid production, short-chain fatty acid metabolism, and microbial regulation. Furthermore, the altered human proteins were associated with anti-inflammatory processes and gut barrier maintenance [19,20,21]. Notably, the host proteins CELA2A, DEFA3, and KLK1 achieved high ROC-AUC values (0.905, 0.897, and 0.897) and PR-AUC values (0.978, 0.975, and 0.975). These findings suggest that they may serve as early and reliable biomarkers for adolescents with BD.

Significant differences in gut microbial species richness were detected in BD patients relative to healthy controls. A decline in the abundance of Firmicutes, including Bacilli and Erysipelotrichi, was observed in BD patients, consistent with previous findings [22]. Notably, in our study, in the gut of BD patients, lactate-producing bacteria, including Bifidobacteriaceae [23], and lactate-metabolizing bacteria, such as Megasphaera [24], were found to be elevated. In certain cases, lactate accumulation in the gut can lead to symptoms such as acidosis, arrhythmias, and neurotoxicity [25]. However, lactate, as a substrate, can participate in “cross-feeding” interactions with bacteria that produce short-chain fatty acids [26], such as Megasphaera and Faecalibacterium. Bacteria belonging to the Megasphaera genus convert lactate into short-chain fatty acids, including propionate and acetate [27]. Experimental evidence suggests that propionic acid may produce antidepressant effects [28], whereas acetic acid enhances colonic tryptophan hydroxylase 1 expression, the enzyme that limits serotonin (5-HT) synthesis [29]. These findings suggest that Megasphaera plays an active role in host 5-HT biosynthesis. Megasphaera may be associated with mood improvement. However, its excessive proliferation may result from unfavorable conditions, such as poor dietary adaptation or metabolic stress [30].

In addition, Studies on adult BD patients have found that Faecalibacterium abundance is frequently reduced [10], and its relative levels are positively correlated with clinical symptom severity [11]. However, in our results, Faecalibacterium did not show differential changes, which contrasts with previous studies. One possible explanation for this discrepancy lies in the methodological approach. 16S rRNA sequencing profiles microbial taxonomy, whereas proteomics reflects functional protein activity. As genomic abundance does not necessarily align with protein expression or function [31], the two approaches may yield different results. In addition, the younger age and shorter disease duration of the adolescent BD patients in our study may also have contributed to the observed differences. A study on patients with depression demonstrated that the abundance of Faecalibacterium decreased in those with active depressive episodes but not in treatment-responsive patients [7]. All BD patients in our study were enrolled during hospitalization, and medication use was not controlled. Therefore, we cannot exclude the possibility that pharmacological treatment contributed to the lack of changes in Faecalibacteriuma bundance. This represents a limitation of our study, and future investigations should focus on the gut microbiota of drug-naïve adolescents with BD.

Adolescents with BD are constantly under the stress [32] and fatigue [33] caused by their condition. Research in Norway has observed that individuals with chronic fatigue syndrome show significantly higher levels of Alistipes [34]. Similarly, in a mouse model subjected to long-term stress, Alistipes was also notably elevated [35]. Our results similarly show that Alistipes, a gut microbiota associated with stress, is found at higher levels. Alistipes is commonly isolated from the gut and is linked to various gut-related and extra-intestinal diseases, such as anxiety, depression, cirrhosis, and chronic fatigue syndrome [36]. Stress-induced increases in Alistipes may be linked to alterations in the gut-brain axis.

Symptoms of the gastrointestinal tract, including bloating, constipation, or diarrhea, are commonly observed in patients with BD, suggesting potential impairment of intestinal function [37]. In the present study, we found a decrease in KLK1, a serine protease, in the intestines of BD patients. Zhou et al. reported that intestinal KLK1 progressively decreases with disease progression in inflammatory bowel disease (IBD), exacerbating gut barrier dysfunction [19]. These findings suggest that changes in KLK1 in BD may reflect impaired gut function or a low-grade inflammatory state [38]. Such results are consistent with prior research demonstrating a positive link between IBD and BD [39].

Interestingly, Motta et al. reported that CELA2A is expressed and secreted by colonic epithelial cells, and its activity is upregulated in IBD, contributing to mucosal inflammation [20]. In contrast, our study found a marked decrease in CELA2A expression in adolescents with BD. Considering the influence of gut microbiota on CELA2A, germ-free mice show a substantial reduction in CELA2A levels [40]. Consistent with this, we observed a decreased abundance of Bacteroides uniformis, which was positively correlated with CELA2A. Higher levels of this bacterium have been associated with reduced intestinal inflammation [41] and improved mood [42]. These findings suggest that, in the context of BD characterized by low-grade intestinal inflammation [38], alterations in CELA2A are more likely driven by changes in microbial composition rather than by inflammatory signaling.

DEFA3, an α-defensin, is part of the antimicrobial and cytotoxic peptide family involved in host defense [43]. DEFA3 likely contributes to gut microbiota homeostasis by adjusting to changes in the intestinal ecosystem, thereby regulating microbial balance and promoting intestinal health [21]. In a mouse model of depression, a reduction in intestinal DEFA3 led to microbial dysbiosis and translocation [44]. In our results, DEFA3 was strongly positively correlated with Parabacteroides distasonis. Polysaccharides are converted into succinate and acetate by P. distasonis, which additionally plays a role in bile acid regulation and metabolic processes [45]. Since bile acid imbalance can impair intestinal barrier function and integrity [46], it remains to be clarified whether the effect of DEFA3 on gut homeostasis is mediated through P. distasonis. Further studies are warranted to elucidate the underlying mechanisms.

Limitations

First, gut microbiota are influenced by multiple factors. In this study, we controlled for antibiotic, probiotic, and prebiotic use in healthy controls, but no standardized dietary control was applied. Implementing short-term standardized diets in future studies may help reduce inter-individual variability. Second, the BD and HC groups included relatively small and unequal numbers of participants. Increasing the sample size and ensuring balanced groups in future studies would improve the reliability of the results.Finally, future work will focus on validating host proteins with potential diagnostic value, such as CELA2A, DEFA3, and KLK1. Their potential roles in the gut of patients with BD will also be further explored

Conclusions

Adolescents with BD exhibited pronounced changes in gut microbiota, microbial proteins, and host proteins. Notably, Bifidobacteriaceae, Megasphaera, and Alistipes were enriched in BD patients. CELA2A, DEFA3, and KLK1 demonstrated high diagnostic potential, as reflected by their ROC-AUC and PR-AUC values. In addition, correlation analyses between differential microbial taxa and host proteins suggested potential interactions, highlighting the possible roles of CELA2A, DEFA3, and KLK1 in the pathophysiology of BD.