Multi–omic Longitudinal Profiling Reveals Coordinated Gut Microbiota, Metabolomics, and Host Proteomics Signatures Predictive of Antidepressant Response in Drug–Naïve Depression

Multi–omic Longitudinal Profiling Reveals Coordinated Gut Microbiota, Metabolomics, and Host Proteomics Signatures Predictive of Antidepressant Response in Drug–Naïve Depression | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Multi–omic Longitudinal Profiling Reveals Coordinated Gut Microbiota, Metabolomics, and Host Proteomics Signatures Predictive of Antidepressant Response in Drug–Naïve Depression Po-Hsiu kuo, Shih-Kai Lin, Shiau-Shian Huang, Mong-Liang Lu, Chun-Hsin Chen This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8836038/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Thirty to fifty percent of patients with major depressive disorder (MDD) show inadequate response to antidepressants, yet biological predictors remain elusive. We conducted a longitudinal multi–omic study of 28 antidepressant–naïve patients with MDD sampled at baseline, one month, and two months, integrating shotgun metagenomics, microbial functional pathways, targeted fecal metabolomics, and plasma proteomics to identify response–predictive signatures. While global microbial diversity remained stable, treatment responders exhibited distinct microbial trajectories characterized by early reductions in inflammation–associated taxa and restoration of butyrate–producing bacteria, and baseline microbial composition predicted two–month response with 96.5% cross–validated accuracy. Cross–lagged panel analyses revealed bidirectional coupling between symptom improvement and microbial dynamics. Functionally, responders maintained pathway–level redundancy with distributed species contributions to core biosynthetic pathways, whereas non–responders showed progressive functional narrowing and taxonomic dominance. Metabolomically, responders exhibited restoration of secondary bile acids and butyrate, while non–responders showed depletion patterns and divergent hexanoic acid trajectories. Plasma proteomics identified treatment–responsive proteins with divergent longitudinal patterns, enriched in immune defense, epithelial remodeling, and intracellular signaling pathways. Serial mediation analyses demonstrated that microbial changes influenced treatment response through bile acid and short–chain fatty acid intermediates that modulated host proteomic networks. These findings indicate that antidepressant efficacy is associated with coordinated functional resilience across the gut–microbiome–host axis rather than compositional restructuring, suggesting targets for microbiome–informed precision psychiatry. Health sciences/Biomarkers/Predictive markers Biological sciences/Physiology Biological sciences/Biochemistry Major depressive disorder antidepressant treatment response shotgun metagenome gut microbiome metabolimics plasma proteomics longitudinal multiomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1 Introduction Major depressive disorder (MDD) affects approximately 5.7% of adults globally and is consistently ranked among the leading causes of disability worldwide 1 . Although antidepressants remain the primary treatment modality, clinical responses are highly heterogeneous, with 30–50% of patients exhibiting inadequate improvement despite appropriate therapy 2 . Elucidating the biological underpinnings of this variability remains a critical unmet need in psychiatry. The gut microbiota influences brain function through metabolic, immune, and neuroendocrine signaling pathways 3 , 4 . Microbial metabolites, particularly short–chain fatty acids (SCFAs), bile acids, and tryptophan derivatives, serve as key neuromodulatory and immunoregulatory intermediaries 3 , 5 and have been increasingly implicated in the response to antidepressant treatment 6 . Dysbiosis observed in depression 7 , 8 may shape pharmacological responsiveness, while antidepressants can reciprocally alter microbial composition and function 9 – 12 . At the host level, plasma proteomic studies have identified alterations in inflammatory, metabolic, and neuroendocrine markers in first–episode MDD patients 13 , many of which normalize—or fail to normalize—in parallel with antidepressant treatment response 14 , 15 . Together, these findings suggest that microbial metabolites and host systemic responses constitute interconnected determinants of antidepressant efficacy 16 – 18 . Building on this framework, recent studies have begun to examine how the gut microbiota contribute to heterogeneity in antidepressant treatment outcomes 19 – 21 . Despite accumulating evidence linking the gut microbiota to depression and treatment response, critical gaps remain. Most studies employ cross–sectional designs or assess single time points, limiting the identification of temporal dynamics that distinguish responders from non–responders 12 , 22 . Existing work has predominantly relied on taxonomic profiling, providing limited mechanistic insight into microbial functional pathways, metabolic outputs, or host molecular responses 23 . Multi–omic integration spanning microbiome composition, functional capacity, metabolite production, and host proteomic changes is increasingly recognized as critical for understanding microbiota–host interactions 24 , 25 . However, such integrative analyses remain scarce, and few studies have systematically examined pathway–level organization, functional redundancy, or bidirectional coupling between microbiota dynamics and clinical symptom trajectories. Although multi–omic approaches have been applied in controlled animal models to dissect mechanistic host–microbiota interactions, comparable integrative frameworks in human studies remain limited 26 – 28 . To address these gaps, we conducted a longitudinal study of antidepressant–naïve MDD patients sampled at three time points during treatment, integrating shotgun metagenomics, microbial functional pathway analysis, targeted fecal bile acid and SCFA quantification, and high–dimensional plasma proteomics. We applied complementary statistical frameworks—including mixed–effects models, cross–lagged panel analyses, machine learning prediction, and serial mediation—to (1) identify temporal microbial, metabolic, and proteomic signatures distinguishing responders from non–responders; (2) assess bidirectional relationships between symptom change and microbial dynamics; (3) characterize functional pathway organization and redundancy; and (4) map microbiota–metabolite–host protein networks underlying treatment response. This integrated approach aims to delineate the biological mechanisms underlying heterogeneity in antidepressant response and inform microbiome–based precision psychiatry strategies. 2 Materials and Methods This longitudinal discovery study was designed to characterize multi–omic dynamics across antidepressant treatment with sufficient temporal resolution (three time points) and biological depth (four omics layers) to identify mechanistic signatures of treatment response. While the sample size (n = 28 participants, 81 complete observations across time points) is modest, it provides adequate statistical power for our primary aims. For longitudinal mixed–effects models, we achieve >80% power to detect medium–to–large effect sizes (|Cohen’s d| ≥ 0.7) for within–subject temporal changes at α = 0.05 with 81 observations 29 . For cross–sectional between–group comparisons (responders vs. non–responders), we have 80% power to detect large effect sizes (d ≥ 0.9) for 17 responders and 11 non–responders 30 . Finally, the depth of multi–omic profiling (288 species, 90 pathways, 31 metabolites, 9,458 proteins) enables systems–level pattern detection across correlated biological layers 31 . We emphasize that this study represents a discovery cohort designed to generate hypotheses and identify candidate biomarkers requiring validation in larger, independent samples. 2.1 Participants and Clinical Assessment Patients aged 20–65 years with antidepressant–naïve MDD (DSM–5) 32 were recruited from psychiatric clinics in Taipei. The cohort included both first–episode (n = 21) and recurrent depression (n = 7). Exclusions followed established criteria (schizophrenia, schizoaffective disorder, intellectual disability, drug–induced mood disorders, recent gastrointestinal surgery (within 2 months), current infections, or recent use of probiotics, prebiotics, symbiotics, or antibiotics). All procedures were IRB–approved with written informed consent. Participants were assessed at BL (baseline), TP1 (one–month), and TP2 (two–month) 33 . Demographics were collected at BL, and clinical measures and stool/blood samples were obtained at each visit (Fig. S1). Stool was transported at 4 °C with a 200–mg pellet reserved for DNA extraction, and plasma was isolated from EDTA blood (centrifuged at 3000 g, 4 °C). After exclusions, 25 participants contributed complete longitudinal data (75 samples across BL/TP1/TP2), and three provided partial follow–up (BL and TP1 only; six samples). Among all participants, antidepressant selection followed clinical judgment by treating psychiatrists and included SSRIs (n = 10, 35.71%), SNRIs (n = 2, 7.14%), and other antidepressants (n = 16, 57.14%). Medication type remained unchanged throughout the study period, and antidepressant dose remained stable in 21 participants. Concomitant psychotropic medications were documented but not considered exclusionary if stable for ≥2 weeks before baseline. Baseline diet was evaluated with a validated short–form food–frequency questionnaire 34 , and participants received guidance to maintain stable dietary patterns during the study; interim dietary changes were not formally assessed. Depressive symptoms were assessed at each visit with the 17–item Hamilton Depression Rating Scale (HAMD)35. Treatment response was defined by percent reduction in HAMD and by categorical response (≥50% reduction), evaluated at TP1 and TP2 36,37 . 2.2 Shotgun metagenomic sequencing, quality control, and functional profiling Stool DNA was extracted using the QIAamp® PowerFecal® Pro Kit following the manufacturer’s protocol. Metagenomic libraries were prepared using the Illumina Nextera XT kit and sequenced on an Illumina NovaSeq 6000 platform to generate paired–end reads. Raw reads underwent standard quality control and trimming, followed by human–read removal (Figs. S2–3; Table S1). Taxonomic profiling was performed with MetaPhlAn 4 using the CHOCOPhlAn SGB database to obtain species–level relative abundances 38 . Functional profiling was conducted with HUMAnN 3.2 39 , mapping reads to ChocoPhlAn and UniRef90 to derive MetaCyc pathway abundances 40 . Species contribution profiles were generated to quantify taxon–function coupling. All microbial features were pseudocount–adjusted and log–transformed before analysis 41 . Laboratory personnel conducting multi–omic profiling were blinded to response status and time–point labels during data generation. 2.3 Stool metabolite profiling Bile acids (BAs) were quantified from ~20 mg stool using a methanol/acetonitrile/water/formic acid extraction containing isotope–labeled internal standards. Extracts were analyzed by LC–MS/MS (positive–ion ESI, MRM) and quantification was performed using TargetLynx. SCFAs were profiled from ~50 mg stool using derivatization and analyzed by GC–MS (EI mode). Metabolite levels were log–transformed for downstream analysis. 2.4 Plasma proteomics profiling Targeted plasma proteomics was performed using the Illumina Protein Prep (IPP) SOMAmer platform (~9,500 proteins; CV ~5.5%) 42 . Approximately 130 µL plasma per sample was processed following the standard IPP workflow and sequenced on an Illumina NovaSeq system. Protein values were log–transformed before analysis. 2.5 Bioinformatic and data processing Shotgun metagenomics and targeted metabolomics generated three data layers: taxonomic profiles (MetaPhlAn 4), functional pathways (HUMAnN 3.2), and targeted BA/SCFA quantifications. Taxonomic features were aggregated to phylum, genus, and species levels, and low–prevalence features (<20%) were removed, yielding 149 genera and 288 species (Fig. S4) 43 . HUMAnN pathway tables were curated by excluding low–coverage (<0.3) and noninformative pathways, resulting in 90 high–confidence MetaCyc pathways 40 (Figs. S5–6). Targeted metabolomics initially quantified 46 BAs and 11 SCFAs; after LOQ–based filtering and preprocessing, 21 BAs and 10 SCFAs were retained (Tables S2–3; Fig. S7) 44 . Plasma proteomics identified 10,272 proteins; after QC and excluding unmapped targets 45 , 9,458 proteins were retained and annotated against Reactome curated pathway knowledgebase 46 . 2.6 Statistical analysis All statistical analyses were performed in R and RStudio. Demographic and clinical characteristics were compared between responders and non–responders using non–parametric tests (continuous variables) and χ² or Fisher’s exact tests (categorical variables); α– and β–diversity indices were evaluated but were not a primary focus. To characterize microbial shifts, we used: time–as–category mixed models (Model I); time–as–continuous models incorporating ΔHAMD (TP1–BL; TP2–BL) with Mundlak adjustment (Model II); and responder vs non–responder fold–change analyses (TP1–BL, TP2–BL) (Model III). Baseline predictors of TP2 response were evaluated using Wilcoxon screening and a Random Forest classifier with repeated 5–fold cross–validation and permutation testing (Model IV). All models were adjusted for baseline dietary covariates; nominally significant species in Models II–III were selected as targeted taxa. Species–pathway coupling models predicted pathway fold–changes from targeted species fold–changes in mixed–effects models; HUMAnN stratified outputs quantified species contributions 39 . Functional redundancy was characterized using Shannon entropy and Gini coefficients 47 . Δ–based cross–lagged panel models assessed bidirectional temporal associations between symptom changes and targeted species. Serial mediation tested sequential mediation via metabolites and proteins (1,000 bootstraps). Limma–trend tests was applied for cross–sectional metabolite/protein comparisons 48 . Multi–omics integration used DIABLO (mixOmics) 49 , which identifies correlated features across species, metabolites, and proteins while optimizing discrimination between responders and non–responders. We set α = 0.05 and controlled FDR using Benjamini–Hochberg 50 . 3 Results 3.1 Clinical characteristics and treatment response patterns Of the 28 participants (mean age 36.11 ± 13.68 years; 64.29% female), 17 (60.71%) achieved a ≥50% reduction in HAMD scores by their latest follow–up visit and were classified as responders. Treatment response increased from 50% at TP1 to 64% at TP2. Baseline demographic, dietary, and clinical characteristics did not differ significantly between eventual responders and non–responders, including age, sex, BMI, or baseline HAMD severity. Antidepressant class distribution and concomitant medication use showed no association with response status. Only responders demonstrated significant HAMD reductions across time points (BL: 16.73 ± 4.85 → TP1: 9.53 ± 5.37 → TP2: 4.14 ± 4.35; p < 0.001), whereas non–responders showed minimal change (BL: 17.45 ± 5.41 → TP2: 13.22 ± 4.68; p = 0.16) (Tables S4–5; Figs. S8–9). 3.2 Taxonomic dynamics: temporal changes and response–specific trajectories Global community structure remained stable during treatment. The α– and β–diversity metrics showed no significant temporal changes or differences between responders and non–responders (Figs. S10–11), indicating that antidepressant treatment did not induce community–wide restructuring. Despite stable diversity metrics, Model I exhibited treatment–associated species temporal dynamics. Early reductions (BL→TP1) occurred in facultative anaerobes and inflammation–associated taxa: Weissella confusa/cibaria , Lactococcus lactis , Klebsiella pneumoniae , Intestinibacter bartlettii, Turicibacter bilis , and Actinomyces naeslundii (Table S6; Fig. S12). Model II displayed symptom–linked microbial changes, and Mundlak–adjusted models disentangled pure temporal effects from symptom–related changes while revealing substantial overlap between the two (Tables S7–8; Fig. S13). Time–as–continuous models incorporating ΔHAMD revealed that 27 species changed in association with either treatment duration or symptom improvement. Notably, several taxa—including Dialister hominis , Anaerostipes hadrus , Phascolarctobacterium faecium , Ruminococcus lactaris , Lachnospira spp., and Clostridiaceae bacterium Marseille–Q3526 —were associated with both time progression and changes in depressive symptoms (ΔHAMD), indicating that these taxa exhibit dynamic trajectories jointly shaped by pharmacological exposure and clinical improvement. However, none of these associations remained statistically significant after multiple testing correction. Model III revealed response–group divergent trajectories (Table 1; Fig. S14). At TP1 (early divergence) responders showed stronger decreases in Hungatella hathewayi, Faecalicatena contorta, Wansuia hejianensis, and Blautia producta , along with increases in Streptococcus sanguinis and Megamonas funiformis . At TP2 (late divergence) responders exhibited increases in Clostridium leptum, Clostridium saudiense, and Oscillospiraceae _unclassified, and decreases in Eubacterium ventriosum, Blautia glucerasea, and Catenibacterium tridentinum . Model IV identified microbial signatures at baseline predictive of TP2 treatment response. A Random Forest classifier achieved high discriminative accuracy (cross–validated AUC = 0.965, 95% CI: 0.84–1.00) confirmed by permutation testing (p_perm < 0.001). Top predictive features included Raoultibacter timonensis, Blautia faecis, and Escherichia coli (Fig. 1; Tables S9–10; Figs. S15–16). No results remained significant after correction for multiple testing. 3.3 Bidirectional microbiota–symptom relationships Cross–lagged panel models revealed temporally ordered bidirectional associations between ΔHAMD and changes in 17 targeted species (Fig. 2; Table S11). When modeling symptom change predicting subsequent microbial abundance (ΔHAMD → future microbial abundance) at the nominal level (P < 0.05), greater symptom improvement predicted higher abundance in Anaerostipes hadrus , Megasphaera sp. NM10 , Dialister hominis , and Phascolarctobacterium faecium at TP1. During TP1–TP2, less symptom improvement predicted higher abundance in Actinomyces graevenitzii and Ruminococcus lactaris at TP2, while the effect in Blautia hansenii was reversed. Across BL–TP2, all significant species—including Alistipes indistinctus , Alistipes shahii , Clostridiaceae bacterium Marseille–Q3526 , Lachnospira sp. NSJ_43 , and Ruminococcus lactaris —displayed positive coefficients. A subset of these relationships also survived FDR correction, including Anaerostipes hadrus, Actinomyces graevenitzii, Clostridiaceae bacterium Marseille–Q3526, and Ruminococcus lactaris . In the reverse models predicting HAMD scores from Δabundance at the nominal level, most BL–TP1 associations were negative (e.g., Alistipes communis , Anaerostipes hadrus , Dialister hominis ), whereas TP1–TP2 showed mixed effects, with Actinomyces graevenitzii positive and Blautia hansenii and Dialister hominis negative. During BL–TP2, most species demonstrated positive coefficients, except for Dialister hominis and Mogibacterium diversum , which were negative. Together, these results reveal temporally structured and species–specific bidirectional relationships, with taxa such as Dialister hominis , Ruminococcus lactaris , and Clostridiaceae bacterium Marseille–Q3526 showing the most consistent cross–lagged patterns, indicating dynamic feedback between these taxa and depressive symptoms. Among these relationships, those involving Anaerostipes hadrus, Dialister hominis, Ruminococcus lactaris, and Clostridiaceae bacterium Marseille–Q3526 remained robust after FDR correction. 3.4 Functional pathway dynamics and redundancy Functional profiling identified 90 high–confidence MetaCyc pathways dominated by core biosynthetic functions: amino acid biosynthesis (29%), nucleotide biosynthesis (17%), carbohydrate metabolism (11%), and cell structure biosynthesis (10%) (Fig. S17). Species–pathway modeling identified functional links associated with the 38 targeted species. Among 3,420 species–pathway pairs tested, 397 were nominally significant and 12 remained significant after FDR correction (Fig. S18; Table S12). Alistipes shahii showed the broadest associations, negatively linking to multiple core biosynthetic pathways (L–lysine biosynthesis VI, peptidoglycan biosynthesis I–II, UMP biosynthesis I–II, S–adenosyl–L–methionine salvage I). Additional associations were observed for Wansuia hejianensis (inosine–5’–phosphate degradation), Mogibacterium diversum (L–threonine biosynthesis), and Weissella confusa (UDP–glucose–derived O–antigen biosynthesis). Species–level contribution analysis identified Prevotella copri, Megamonas funiformis, Ruminococcus bromii, Roseburia faecis, and Klebsiella pneumoniae as dominant contributors (>1%) to targeted pathways. Seven targeted species ( Clostridium scindens , Megamonas funiformis , Phascolarctobacterium faecium , Clostridium leptum , Ruminococcus lactaris , Streptococcus sanguinis , and Weissella confusa ) were listed as contributors to our targeted pathways (Tables S13–14; Fig. S19). Sankey diagrams revealed striking divergence in pathway contribution structure between response groups. Responders maintained broader, multispecies contributions to core biosynthetic pathways, with parallel functional inputs from multiple taxa—especially M. funiformis (1.8–2.1% contribution) , P. faecium (1.1–1.5%) , and R. lactaris (~0.1%). Non–responders exhibited progressive functional narrowing, with TP2 characterized by increased dominance of R. lactaris (0.2–0.4% contribution), reduced contributions from P. faecium (0.3–0.4%), and depleted contributions of C. scindens and S. sanguinis (Fig. 3, Table S15). For functional redundancy (Shannon H) and dominance (Gini G), stratified analyses showed opposing patterns: in pathways such as inosine–5’–phosphate degradation and L–lysine/L–threonine biosynthesis, responders exhibited a transient TP1 reduction followed by recovery at TP2 (BL = 0.59 → TP1 = 0.39 → TP2 = 0.47), whereas non–responders showed a continuous decline (BL = 0.60 → TP1 = 0.39 → TP2 = 0.35) (Fig. S20). Gini coefficients demonstrated reciprocal patterns. Polynomial trend analysis revealed a significant quadratic trajectory of functional redundancy in responders (p = 0.003), as measured by both Shannon entropy (H) and the Gini index, whereas no significant temporal trend was observed in non–responders. 3.5 Metabolite signatures: bile acids and short–chain fatty acids Spearman correlations between Δabundance of 38 targeted species and 31 metabolites revealed two major co–association modules (Figs. 4A–B; Table S16). Module 1 (positive associations) included bile–acid–producing and butyrate–associated taxa — B. producta, A. shahii, C. leptum, and C. scindens positively correlated with secondary/deconjugated bile acids and branched–chain fatty acids. Module 2 (negative associations) included inflammation–associated taxa — W. confusa, P. faecium, H. hathewayi, W. hejianensis, M. funiformis, and F. contorta showed negative correlations with conjugated bile acids and several SCFAs. Representative associations that remained significant after false discovery rate (FDR) correction included W. confusa with 3α–hydroxy–7–ketolithocholic acid (ρ = −0.546), P. faecium with glycochenodeoxycholic acid (ρ = −0.501), and B. producta with 2– and 3–Methylbutanoic acid (ρ = 0.581, 0.575). Two bile acids—deoxycholic acid and 5β–cholanic acid–3β,12α–diol—inversely correlated with ΔHAMD, indicating higher levels when symptoms improved (Fig. 4C). Limma–trend differential testing identified response–dependent metabolic remodeling (Fig. 4D; Fig. S21). Responders showed BL → TP2 increases in multiple secondary bile acids (e.g., deoxycholic acid, glycolithocholic acid, isolithocholic acid, lithocholic acid), whereas non–responders exhibited parallel declines (e.g., deoxycholic acid and glycolithocholic acid). For SCFAs, responders displayed a TP1 → TP2 rebound of butyric acid (TP2 p = 0.041), while non–responders remained flat. Conversely, hexanoic acid rose transiently in non–responders but declined in responders (TP2 p = 0.027). Although these effects were detected at the nominal significance level, none of the metabolite changes remained significant after FDR correction. 3.6 Treatment–responsive plasma proteomic signatures and multi–omic integration Among 9,458 quantified plasma proteins, 173 proteins showed differential abundance at TP2 between responders and non–responders, including proteins significant at BL and TP2 (p<0.05, Fig. 5A). Eighteen proteins exhibited consistent group differences at both BL and TP2, whereas the majority displayed heterogeneous longitudinal trajectories. Notably, longitudinal trajectory analysis revealed heterogeneous patterns: progressive divergence, in which responder–non–responder differences were increasingly amplified (e.g., BCHE, CD2, MICA, and NOTCH2), and cross–over dynamics, characterized by attenuation and reversal of baseline differences during treatment (e.g., PAXIP1 and TFPI2). Importantly, such trajectory patterns were also evident among proteins without significant baseline differences but showing progressive separation, including ARFRP1, BCHE, CAMP, GRN, IGF1, PCYOX1, PRRT4, RNF7, SERGEF. (all p < 0.05 at TP2), suggesting that longitudinal divergence and cross–over dynamics can emerge during treatment (Fig. S22). Although these trajectory patterns were identified at the nominal significance level, none of the protein–level differences remained significant after false discovery rate (FDR) correction. Serial mediation analyses (1,000 bootstrap iterations) were used to prioritize treatment–responsive species–metabolite–protein pathways that potentially link microbial changes to host proteomic responses via bile acid and SCFA intermediates (Table 2; Table S17). Among the top 10 ranked pathways (based on total effect size and nominal bootstrap significance), Cloacibacillus porcorum , Megasphaera sp. NM10, and Dialister hominis were repeatedly linked to host protein changes through specific metabolic intermediates. Bile acid–mediated pathways (e.g., isolithocholic acid and glycolithocholic acid) connected these species to epithelial– and immune–related proteins, whereas hexanoic acid–mediated pathways linked microbial shifts to proteins involved in immune and stress–related responses, including HOXB8, CDH17, KLK3, HSPA1L, and CYP2S1. Collectively, these pathways nominate bile acids and SCFAs as candidate metabolic intermediates linking microbial shifts to immune–, epithelial–, and stress–response protein networks. Multi–omics integration using DIABLO revealed coordinated signatures across treatment–responsive microbial species, metabolites, and host proteins that partially discriminated responders from non–responders (species: R² = 0.53; metabolites: R² = 0.41; proteins: R² = 0.76), with the strongest discrimination observed in the protein block (Fig. 5B; Fig. S23). Cross–omic loadings revealed co-associated features across microbial taxa, metabolites, and proteins. Focusing on Reactome pathways most relevant to antidepressant treatment response and host–microbiota interactions, treatment–responsive proteins were primarily enriched in immune defense–related and host structural remodeling processes (Fig. 5C). Notably, pathways related to antimicrobial peptides and alpha–defensins, as well as extracellular matrix degradation and matrix metalloproteinase activation, were consistently highlighted. Metabolic and nutrient–handling pathways, including cobalamin (vitamin B12) uptake and transport and regulation of IGF transport by IGF–binding proteins, were also enriched. In addition to these core pathways, enrichment was observed for cytokine– and intracellular signaling–related processes, including cytokine signaling and MAPK–related cascades, as well as pathways involved in protein turnover and homeostasis, such as proteasome assembly and neddylation (Table S18). These enrichments suggest that antidepressant response involves coordinated modulation of immune defense, structural plasticity, and stress–response systems influenced by microbiota–derived metabolites. 4 Discussion This study presents a longitudinal multi–omic analysis of antidepressant treatment, integrating gut microbiota, microbial functions, fecal metabolites, and host plasma proteomics. Despite stable global microbial diversity, antidepressant treatment response in drug–naïve MDD patients was characterized by coordinated reorganization across the gut–microbiome–host axis and time–dependent shifts in specific bacterial taxa that tracked depressive symptom trajectories and temporally predicted clinical improvement. While overall functional profiles remained stable, responders exhibited greater functional redundancy and resilience, accompanied by selective modulation of biosynthetic pathways. Metabolomic analyses further identified bile acids and short–chain fatty acids as key intermediates linking microbial dynamics to treatment response. At the host level, plasma proteomics revealed heterogeneous protein trajectories and pathway–level alterations related to immune regulation, epithelial defense, and intracellular signaling, which distinguished responders from non–responders and integrated with microbial and metabolic changes. Together, these findings indicate that effective antidepressant treatment depends on functional resilience of the microbiome–host axis, the capacity to maintain metabolic flexibility and coordinated host adaptation, rather than compositional change per se. Consistent with previous reports 51 , gut microbial diversity remained stable throughout antidepressant treatment, with no significant temporal changes or differences by treatment response, supporting selective rather than community–wide effects of antidepressants 52 . However, beneath this stable surface, longitudinal modeling revealed a structured microbial reorganization driven jointly by time and changes in depressive severity. Early treatment phases featured reductions in facultative anaerobes enriched in inflammatory conditions (e.g., Weissella, Lactococcus, Klebsiella ), followed by a gradual enrichment of obligate anaerobes associated with fiber fermentation and bile–acid metabolism (e.g., Clostridium leptum, C. scindens, Eubacterium ), consistent with progressive restoration of microbial homeostasis 53 – 56 . Notably, microbial changes associated with depressive symptom improvement largely followed the same temporal directions as treatment–associated changes, indicating that symptom improvement modulated the magnitude, rather than the direction, of longitudinal microbial trajectories. Taxa that increased or decreased over time showed more pronounced changes in individuals with greater symptom improvement, suggesting that clinical response amplifies the underlying microbial shifts associated with treatment, rather than altering their overall direction. Importantly, response-stratified analysis revealed longitudinal microbial divergence that was not captured by time– or symptom–linked models alone. Rather than reflecting broad restoration across functional groups, these response–specific patterns suggest that effective treatment is characterized by the capacity to engage late–phase ecological reorganization. In this context, selective enrichment of specific obligate anaerobes, including the butyrate–producing taxon Clostridium leptum 57 , may represent a hallmark of successful ecological adaptation during treatment, whereas failure to exhibit such late–phase changes in non–responders points to incomplete transition following initial treatment–associated perturbation. When considered alongside the temporal and symptom–related trends observed in Models I and II, these response–specific patterns are consistent with a two–phase framework, in which early treatment–associated perturbation is followed in responders by selective late–phase ecological reorganization that may confer metabolically beneficial effects 58 . Additionally, baseline microbiome profiles predicted treatment response with high accuracy (cross–validated AUC = 0.965). Beyond longitudinal and response–associated microbial dynamics, our findings suggest that pretreatment microbiome states may influence sensitivity to antidepressant therapy, consistent with prior attempts to leverage baseline microbiome features for response prediction 11 , 21 . Baseline predictive signatures did not overlap with taxa exhibiting treatment– or response–linked trajectories, indicating that pretreatment microbial configurations likely modulate host pharmacometabolic responsiveness rather than directly mediating downstream microbial restoration. Our cross–lagged panel analyses advance beyond associational findings by imposing temporal order, revealing that microbiota changes and symptom improvement engage in dynamic reciprocal feedback. Early symptom improvement predicted subsequent microbial shifts and increases in SCFA–producers (e.g. Anaerostipes hadrus , Phascolarctobacterium faecium , Dialister hominis , and Megasphaera sp. NM10 ), while prior enrichment of these taxa forecasted later clinical improvement, suggesting a rapid restoration of fermentative and short–chain–fatty–acid–producing taxa in response to early therapeutic effects 59 , 60 . This bidirectionality aligns with mechanistic models wherein antidepressant–induced neurobiological changes (e.g., reduced inflammation, restored HPA–axis function) create a more hospitable gut environment for beneficial microbes, which in turn produce metabolites that reinforce therapeutic effects. Importantly, clinical improvement was associated not only with the expansion of SCFA–producing taxa but also with the contraction of taxa previously associated with dysbiotic or alternative metabolic states (e.g., Alistipes indistinctus, Clostridiaceae bacterium Marseille–Q3526, Lachnospira sp. NSJ–43, and Ruminococcus lactaris ), underscoring that treatment response reflects a reconfiguration of microbial modules rather than a unidirectional change. These findings support a dynamic model in which gut microbial changes are not merely passive correlates of symptom improvement but participate in a dynamic, time–structured host–microbe equilibrium in which microbial reorganization both responds to, and actively shapes, clinical recovery. Species-guided pathway analyses moved beyond taxonomic shifts and highlighted functional reorganization underlying treatment response. Core biosynthetic pathways, including L–lysine, peptidoglycan, and UMP biosynthesis, were repeatedly linked to treatment–responsive taxa identified in longitudinal and causal models, suggesting coordinated shifts in microbial energy and nucleotide metabolism during antidepressant exposure. Notably, these pathways were predominantly anchored to A. shahii , whose negative associations across multiple biosynthetic routes are consistent with downregulation of energetically costly metabolic programs and potential attenuation of kynurenine–related inflammatory signaling 61 , 62 . Within this same functional context, species–level contribution analyses further indicated complementary involvement of SCFA–associated taxa, including P. faecium and R. lactaris , supporting links between microbial fermentation, inflammation, and neurotransmission relevant to depressive symptoms 63 . Responders maintained distributed, multi–taxon contributions to core biosynthetic functions, whereas non–responders exhibited progressive taxonomic dominance and reduced functional evenness, indicating reduced ecological resilience under pharmacological perturbation. This pattern parallels ecological principles linking biodiversity and functional redundancy to ecosystem resilience 64 , 65 and indicates that a system with multiple taxa capable of performing essential functions better withstand perturbations. In the gut microbiome context, functional redundancy may enable metabolic flexibility under pharmacological stress, allowing continued production of neuroactive metabolites despite compositional fluctuations. The observation that non–responders lose redundancy suggests that antidepressant efficacy depends partly on the microbiome's capacity to reorganize functionally while maintaining metabolic output that may be compromised in treatment–resistant individuals. Shifts in bile acids and SCFAs support a microbiota–metabolite–host signaling axis underlying treatment response. Responders showed longitudinal increases in secondary bile acids (particularly deoxycholic acid, lithocholic acid, and related intermediates) and butyrate, whereas non–responders showed progressive depletion. Secondary bile acids are known to signal via FXR, TGR5, and VDR, thereby suppressing inflammatory tone, enhancing epithelial barrier function, and regulating HPA–axis activity 66 , 67 , processes central to proposed antidepressant mechanisms. Butyrate similarly exerts anti–inflammatory, neuroprotective, and epigenetic effects relevant to mood regulation 68 . The divergent hexanoic acid trajectories between groups suggest differential fermentation pathway utilization, potentially reflecting distinct microbial community states. Serial mediation analyses linked these metabolic changes to host proteomic responses, supporting a model in which microbiota–associated alterations in bile acid and SCFA availability may contribute to treatment efficacy through modulation of systemic immune and stress–response systems. Proteomic analyses suggest that antidepressant treatment response is characterized by progressive, time–dependent host molecular adaptation rather than static baseline differences. Heterogeneous protein trajectories—including divergent and cross–over patterns that separate responders from non–responders during treatment—are consistent with antidepressant efficacy emerging through gradual normalization of multiple physiological systems 69 , 70 . Enrichment of immune defense (antimicrobial peptides, defensins), epithelial remodeling (ECM degradation, MMPs), intracellular signaling (MAPK cascades), and proteostasis pathways aligns with established antidepressant mechanisms involving inflammation resolution, structural plasticity, and cellular stress management 69 , 71 , 72 . Importantly, serial mediation analyses situate these proteomic changes within a microbiota–metabolite–host framework, identifying bile acids and SCFAs as key intermediates through which microbial alterations may modulate host protein networks relevant to treatment response. For instance, SCFAs modulate immune activation, stress signaling, and neuroendocrine pathways through GPCR signaling and epigenetic regulation, providing a plausible route by which gut microbial shifts translate into systemic protein–level adaptations during treatment 73 . These results indicate that shifts in gut microbial activity shape systemic host responses through metabolite–mediated signaling. They support an integrative framework in which effective treatment requires coordinated adaptation across microbial, metabolic, and host molecular layers. Overall, the data suggest that antidepressant efficacy is associated with the gut microbiota’s capacity to engage bile acid– and SCFA–related host signaling pathways that recalibrate immune and stress–response systems. Several limitations should be noted. The modest sample size, while adequate for discovery–phase multi–omic research with rigorous cross–validation, limits generalizability, as multi–omics analyses often face challenges in reproducibility and external validation with small cohorts 74 . Although longitudinal cross–lagged and mediation analyses strengthen inference, causality remains observational and precludes definitive causal inference. The heterogeneous antidepressant treatments, though reflecting real–world practice, preclude drug–specific conclusions. Independent validation in larger cohorts with standardized treatment regimens is essential. Diet was assessed only at baseline and not monitored during treatment, potentially confounding microbial changes. The observational design lacks placebo control, limiting separation of specific antidepressant effects from natural symptom trajectories or non–specific factors. Fecal metabolites reflect luminal exposure, metagenomic functions are relative estimates, and plasma proteomics captures peripheral rather than brain–specific or central host responses, though systemic inflammation and metabolic changes likely influence central processes 75 . Within these constraints, this study shows that antidepressant response reflects coordinated reorganization across the gut–microbiome–host axis. Rather than global microbial change, effective treatment involves targeted taxonomic shifts linked to symptom improvement, preservation of functional redundancy, restoration of secondary bile acids and butyrate, and adaptation of host immune and signaling pathways. These findings highlight functional resilience, the ability to maintain metabolic flexibility and coordinated adaptation under treatment, as a key determinant of antidepressant efficacy. Although validation is needed, the results point toward microbiome–informed precision approaches to improve depression treatment. Declarations Acknowledgement We thank National Science and Technology Council for supporting our research. We appreciate Taipei Municipal Wanfang Hospital, Taipei City Psychiatric Center, and Taipei Veterans General Hospital for assistance in sample collection. We gratefully acknowledge BIOTOOLS Co., Ltd. for their technical support in shotgun metagenomic sequencing and fecal metabolomics. We also truly appreciate the support from the Center of Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Novascope Diagnostics Inc., and Genomics BioSci & Tech. Co., Ltd. for their assistance with plasma proteomics profiling on Illumina platforms. We sincerely thank Chou Yuan–Tung for the valuable assistance in subject recruitment for this study. Funding This study was funded mainly by National Science and Technology Council (NSTC 108–2314–B–002–136–MY3; 110–2314–B–002–067–MY3, MOST 111–2314–B–038–062–MY2, NSTC 113–2314–B–038–091–; 114–2314–B–038–105–) and the National Taiwan University Career Development Project (109L7860). This study was also supported by Population Health Research Center from Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education in Taiwan (grant number NTU–112L9004, NTU–113L9004, NTU–114L9004, NTU–115L9004). The finding agents are not involved in the study design, sample collecting, experimental protocol, data analysis, article drafting, or choice to submit to a journal. Conflicts of interest The authors declare that there is no conflict of interest regarding the publication of this paper. Data and analysis codes availability The data cannot be provided due to the data usage rights mentioned in informed consent. The authors will supply codes in response to reasonable requests Author contributions SKK Lin designed the study, managed follow up workflow, biosample extraction to final strategies of analyses, and wrote the first draft of the manuscript. PH Kuo and CH Chen contributed to study design, administrative support, funding acquisition, and patient recruitment, they also revised the final manuscript critically. SS Huang, ML Lu, and CH Chen helped with the subject recruitment. All authors contributed to and approved the final manuscript. Ethics approval and consent to participate All methods were performed in accordance with the relevant guidelines and regulations of Taipei Medical University Wan Fang Hospital, Taipei City Hospital Heping Songde Branch, and Taipei Veterans General Hospital. The study protocol was approved by the Institutional Review Board of Taipei Medical University Wan Fang Hospital (Approval No. N202207051), Taipei City Hospital Heping Songde Branch (Approval No. TCHIRB–11105013), and Taipei Veterans General Hospital (Approval No. 2023–05–005A). Informed consent was obtained from all participants included in the study. Written informed consent for publication of identifiable images was not applicable, as no identifiable images were included in this study. References GBD 2019 Mental Disorders Collaborators. Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Psychiatry 9 , 137–150 (2022). Thomas, L. et al. Prevalence of treatment–resistant depression in primary care: cross–sectional data. Br. J. Gen. Pract. J. R. Coll. Gen. Pract. 63 , e852–858 (2013). Cryan, J. F. et al. The Microbiota–Gut–Brain Axis. Physiol. Rev. 99 , 1877–2013 (2019). Li, J. et al. An integrated catalog of reference genes in the human gut microbiome. Nat. Biotechnol. 32 , 834–841 (2014). Clarke, G. et al. The microbiome–gut–brain axis during early life regulates the hippocampal serotonergic system in a sex–dependent manner. Mol. Psychiatry 18 , 666–673 (2013). Chen, C., Wang, G.–Q., Li, D. & Zhang, F. Microbiota–gut–brain axis in neurodegenerative diseases: molecular mechanisms and therapeutic targets. Mol. Biomed. 6 , 64 (2025). Lin, S.–K. K. et al. Dysbiosis and depression: A study of gut microbiota alterations and functional pathways in antidepressant–naïve mood disorder patients. Transl. Psychiatry 15 , 290 (2025). Liu, L. et al. Gut microbiota and its metabolites in depression: from pathogenesis to treatment. eBioMedicine 90 , (2023). Macedo, D. et al. Antidepressants, antimicrobials or both? Gut microbiota dysbiosis in depression and possible implications of the antimicrobial effects of antidepressant drugs for antidepressant effectiveness. J. Affect. Disord. 208 , 22–32 (2017). Gamboa, J. et al. Impact of antidepressants on the composition of the gut microbiome: A systematic review and meta–analysis of in vivo studies. J. Affect. Disord. 369 , 819–833 (2025). Lee, S. M. et al. The intestinal microbiota as a predictor for antidepressant treatment outcome in geriatric depression: a prospective pilot study. Int. Psychogeriatr. 34 , 33–45 (2022). Borgiani, G. et al. The bidirectional interaction between antidepressants and the gut microbiota: are there implications for treatment response? Int. Clin. Psychopharmacol. 40 , 3–26 (2025). Gędek, A. et al. Altered Cytokine Levels in the First Episode of Major Depression and in Antidepressant–Naïve Patients: A Systematic Review and Meta–Analysis. Int. J. Mol. Sci. 26 , 10362 (2025). Kouba, B. R. et al. Role of Inflammatory Mechanisms in Major Depressive Disorder: From Etiology to Potential Pharmacological Targets. Cells 13 , (2024). Nava, R. G. et al. Modulation of neuroimmune cytokine networks by antidepressants: implications in mood regulation. Transl. Psychiatry 15 , 314 (2025). Huang, C.–T. et al. Impact of Gut Microbiota Alterations on Mitochondrial Bioenergetics in Cortical Astrocytes and Sensorimotor Impairment in a Rat Model of Lipopolysaccharide–Associated Encephalopathy. Shock Augusta Ga https://doi.org/10.1097/SHK.0000000000002637 (2025) doi:10.1097/SHK.0000000000002637. O’Riordan, K. J. et al. Short chain fatty acids: Microbial metabolites for gut–brain axis signalling. Mol. Cell. Endocrinol. 546 , 111572 (2022). Kim, C. H. Complex regulatory effects of gut microbial short–chain fatty acids on immune tolerance and autoimmunity. Cell. Mol. Immunol. 20 , 341–350 (2023). Jiang, Y. et al. The role of gut microbiota and metabolomic pathways in modulating the efficacy of SSRIs for major depressive disorder. Transl. Psychiatry 14 , 493 (2024). Ruohan, Z. et al. Gut microbiota as a novel target for treating anxiety and depression: from mechanisms to multimodal interventions. Front. Microbiol. 16 , (2025). Dong, Z. et al. Gut microbiome: A potential indicator for predicting treatment outcomes in major depressive disorder. Front. Neurosci. 16 , 813075 (2022). Gamboa, J. et al. Impact of antidepressants on the composition of the gut microbiome: A systematic review and meta–analysis of in vivo studies. J. Affect. Disord. 369 , 819–833 (2025). Knight, R. et al. Best practices for analysing microbiomes. Nat. Rev. Microbiol. 16 , 410–422 (2018). Lloyd–Price, J. et al. Multi–omics of the gut microbial ecosystem in inflammatory bowel diseases. Nature 569 , 655–662 (2019). Muller, E., Shiryan, I. & Borenstein, E. Multi–omic integration of microbiome data for identifying disease–associated modules. Nat. Commun. 15 , 2621 (2024). Chetty, A. & Blekhman, R. Multi–omic approaches for host–microbiome data integration. Gut Microbes 16 , 2297860. Sabih Ur Rehman, S. et al. Integrative systems biology approaches for analyzing microbiome dysbiosis and species interactions. Brief. Bioinform. 26 , bbaf323 (2025). Jiang, Y. et al. Multi–omics analysis reveals mechanisms of FMT in Enhancing antidepressant effects of SSRIs. Brain. Behav. Immun. 126 , 176–188 (2025). Ferdous, T. et al. The rise to power of the microbiome: power and sample size calculation for microbiome studies. Mucosal Immunol. 15 , 1060–1070 (2022). Kelly, B. J. et al. Power and sample–size estimation for microbiome studies using pairwise distances and PERMANOVA. Bioinforma. Oxf. Engl. 31 , 2461–2468 (2015). Hasin, Y., Seldin, M. & Lusis, A. Multi–omics approaches to disease. Genome Biol. 18 , 83 (2017). American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders . (American Psychiatric Association, 2013). doi:10.1176/appi.books.9780890425596. Palleja, A. et al. Recovery of gut microbiota of healthy adults following antibiotic exposure. Nat. Microbiol. 3 , 1255–1265 (2018). Chou, Y.–C. et al. Dietary intake of vitamin B(6) and risk of breast cancer in Taiwanese women. J. Epidemiol. 21 , 329–336 (2011). Hamilton, M. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23 , 56–61 (1960). Lin, C.–H., Park, C. & McIntyre, R. S. Early improvement in HAMD–17 and HAMD–7 scores predict response and remission in depressed patients treated with fluoxetine or electroconvulsive therapy. J. Affect. Disord. 253 , 154–161 (2019). McIntyre, R. S. et al. Measuring the severity of depression and remission in primary care: validation of the HAMD–7 scale. CMAJ Can. Med. Assoc. J. 173 , 1327–1334 (2005). Blanco–Míguez, A. et al. Extending and improving metagenomic taxonomic profiling with uncharacterized species using MetaPhlAn 4. Nat. Biotechnol. 41 , 1633–1644 (2023). Beghini, F. et al. Integrating taxonomic, functional, and strain–level profiling of diverse microbial communities with bioBakery 3. eLife 10 , e65088. Caspi, R. et al. The MetaCyc database of metabolic pathways and enzymes – a 2019 update. Nucleic Acids Res. 48 , D445–D453 (2020). Kaul, A., Mandal, S., Davidov, O. & Peddada, S. D. Analysis of Microbiome Data in the Presence of Excess Zeros. Front. Microbiol. 8 , 2114 (2017). Illumina Protein Prep | Detect and quantify proteins with NGS. https://sapac.illumina.com/products/by–type/sequencing–kits/library–prep–kits/protein–prep.html. Nearing, J. T. et al. Microbiome differential abundance methods produce different results across 38 datasets. Nat. Commun. 13 , 342 (2022). Sun, J. & Xia, Y. Pretreating and normalizing metabolomics data for statistical analysis. Genes Dis. 11 , 100979 (2024). Kraemer, S. et al. Crossing the Halfway Point: Aptamer–Based, Highly Multiplexed Assay for the Assessment of the Proteome. J. Proteome Res. 23 , 4771–4788 (2024). Croft, D. et al. Reactome: a database of reactions, pathways and biological processes. Nucleic Acids Res. 39 , D691–D697 (2011). Eliazar, I. & Sokolov, I. M. Maximization of statistical heterogeneity: From Shannon’s entropy to Gini’s index. Phys. Stat. Mech. Its Appl. 389 , 3023–3038 (2010). Ritchie, M. E. et al. limma powers differential expression analyses for RNA–sequencing and microarray studies. Nucleic Acids Res. 43 , e47 (2015). Singh, A. et al. DIABLO: an integrative approach for identifying key molecular drivers from multi–omics assays. Bioinforma. Oxf. Engl. 35 , 3055–3062 (2019). Benjamini, Y. & Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J. R. Stat. Soc. Ser. B Methodol. 57 , 289–300 (1995). Brown, L. C., Bobo, W. V., Gall, C. A., Müller, D. J. & Bousman, C. A. Pharmacomicrobiomics of Antidepressants in Depression: A Systematic Review. J. Pers. Med. 13 , 1086 (2023). Gamboa, J. et al. Impact of antidepressants on the composition of the gut microbiome: A systematic review and meta–analysis of in vivo studies. J. Affect. Disord. 369 , 819–833 (2025). Liu, D. et al. Anaerostipes hadrus, a butyrate–producing bacterium capable of metabolizing 5–fluorouracil. mSphere 9 , e0081623 (2024). Wu, F. et al. Phascolarctobacterium faecium abundant colonization in human gastrointestinal tract. Exp. Ther. Med. 14 , 3122–3126 (2017). Kang, J. D. et al. Bile Acid 7α–Dehydroxylating Gut Bacteria Secrete Antibiotics that Inhibit Clostridium difficile: Role of Secondary Bile Acids. Cell Chem. Biol. 26 , 27–34.e4 (2019). Daniel, S. L. & Ridlon, J. M. Clostridium scindens: history and current outlook for a keystone species in the mammalian gut involved in bile acid and steroid metabolism. FEMS Microbiol. Rev. 49 , fuaf016 (2025). Kabeerdoss, J., Sankaran, V., Pugazhendhi, S. & Ramakrishna, B. S. Clostridium leptum group bacteria abundance and diversity in the fecal microbiota of patients with inflammatory bowel disease: a case–control study in India. BMC Gastroenterol. 13 , 20 (2013). Singh, V. et al. Butyrate producers, ‘The Sentinel of Gut’: Their intestinal significance with and beyond butyrate, and prospective use as microbial therapeutics. Front. Microbiol. 13 , 1103836 (2022). Facchin, S. et al. Short–Chain Fatty Acids and Human Health: From Metabolic Pathways to Current Therapeutic Implications. Life 14 , 559 (2024). Markowiak–Kopeć, P. & Śliżewska, K. The Effect of Probiotics on the Production of Short–Chain Fatty Acids by Human Intestinal Microbiome. Nutrients 12 , 1107 (2020). Gao, K., Mu, C., Farzi, A. & Zhu, W. Tryptophan Metabolism: A Link Between the Gut Microbiota and Brain. Adv. Nutr. 11 , 709–723 (2020). Parker, B. J., Wearsch, P. A., Veloo, A. C. M. & Rodriguez–Palacios, A. The Genus Alistipes: Gut Bacteria With Emerging Implications to Inflammation, Cancer, and Mental Health. Front. Immunol. 11 , (2020). Louis, P. & Flint, H. J. Formation of propionate and butyrate by the human colonic microbiota. Environ. Microbiol. 19 , 29–41 (2017). Galland, T., Pérez Carmona, C., Götzenberger, L., Valencia, E. & de Bello, F. Are redundancy indices redundant? An evaluation based on parameterized simulations. Ecol. Indic. 116 , 106488 (2020). Louca, S. et al. Function and functional redundancy in microbial systems. Nat. Ecol. Evol. 2 , 936–943 (2018). Li, T. & Chiang, J. Y. L. Bile acid signaling in metabolic disease and drug therapy. Pharmacol. Rev. 66 , 948–983 (2014). Wahlström, A., Sayin, S. I., Marschall, H.–U. & Bäckhed, F. Intestinal Crosstalk between Bile Acids and Microbiota and Its Impact on Host Metabolism. Cell Metab. 24 , 41–50 (2016). Dalile, B., Van Oudenhove, L., Vervliet, B. & Verbeke, K. The role of short–chain fatty acids in microbiota–gut–brain communication. Nat. Rev. Gastroenterol. Hepatol. 16 , 461–478 (2019). Duman, R. S., Aghajanian, G. K., Sanacora, G. & Krystal, J. H. Synaptic plasticity and depression: new insights from stress and rapid–acting antidepressants. Nat. Med. 22 , 238–249 (2016). Miller, A. H. & Raison, C. L. The role of inflammation in depression: from evolutionary imperative to modern treatment target. Nat. Rev. Immunol. 16 , 22–34 (2016). Hodes, G. E., Kana, V., Menard, C., Merad, M. & Russo, S. J. Neuroimmune mechanisms of depression. Nat. Neurosci. 18 , 1386–1393 (2015). Autry, A. E. et al. NMDA receptor blockade at rest triggers rapid behavioural antidepressant responses. Nature 475 , 91–95 (2011). Parada Venegas, D. et al. Short Chain Fatty Acids (SCFAs)–Mediated Gut Epithelial and Immune Regulation and Its Relevance for Inflammatory Bowel Diseases. Front. Immunol. 10 , 277 (2019). Sherwani, M. K. et al. Multi–omics time–series analysis in microbiome research: a systematic review. Brief. Bioinform. 26 , bbaf502 (2025). Loh, J. S. et al. Microbiota–gut–brain axis and its therapeutic applications in neurodegenerative diseases. Signal Transduct. Target. Ther. 9 , 37 (2024). Tables Table 1 and 2 are available in the Supplementary Files section. Additional Declarations The authors have declared there is NO conflict of interest to disclose Supplementary Files Supplementaryinformationver120250119.docx Supplementary Information MainResultTable1ver120250119.docx MainResultTable2ver120250119.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewer # 2 agreed at journal 10 Apr, 2026 Review # 1 received at journal 13 Mar, 2026 Reviewer # 1 agreed at journal 24 Feb, 2026 Reviewers invited by journal 19 Feb, 2026 Editor assigned by journal 16 Feb, 2026 Submission checks completed at journal 16 Feb, 2026 First submitted to journal 14 Feb, 2026 Unknown event 12 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8836038","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":593974635,"identity":"a7e72022-220b-4090-8db4-57853d859564","order_by":0,"name":"Po-Hsiu kuo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsElEQVRIiWNgGAWjYDACCTYgUQFlk6DlDEQ1kDAgUgtjGylazKXb0iQ+zrtTZ3CA+eBtHoY/iQ2EtFjOOXZMcua2ZxIGB9iSrXkYDAhrMbiR3ibNu+0wUAuPmTRQSy6RWuaAtPB/I1ZL2jFp3gawLWzEabGckZZsOePYYcmZh9mMLecYGNcT1GIukWZ440PNYX6+480Pb7ypkDMmpAMpHphRucRoGQWjYBSMglGACwAAkQ43SWd+F3wAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0003-0365-3587","institution":"College of Public Health, National Taiwan University, Taipei, Taiwan","correspondingAuthor":true,"prefix":"","firstName":"Po-Hsiu","middleName":"","lastName":"kuo","suffix":""},{"id":593974636,"identity":"56104971-7769-4907-95ef-8e3048edb244","order_by":1,"name":"Shih-Kai Lin","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Shih-Kai","middleName":"","lastName":"Lin","suffix":""},{"id":593974637,"identity":"90825bd4-a210-4b06-92b1-41b22a09b1c1","order_by":2,"name":"Shiau-Shian Huang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Shiau-Shian","middleName":"","lastName":"Huang","suffix":""},{"id":593974638,"identity":"3410067b-4f0c-4f24-b306-b7c71f51bd24","order_by":3,"name":"Mong-Liang Lu","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Mong-Liang","middleName":"","lastName":"Lu","suffix":""},{"id":593974639,"identity":"de09d36a-7181-404f-b3b5-8499e966147e","order_by":4,"name":"Chun-Hsin Chen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Chun-Hsin","middleName":"","lastName":"Chen","suffix":""}],"badges":[],"createdAt":"2026-02-10 03:55:37","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8836038/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8836038/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103398108,"identity":"6bfee47c-42da-4d7b-9ec5-a6c12acd8e62","added_by":"auto","created_at":"2026-02-25 08:58:48","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1675886,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eBaseline gut microbial species predictive of antidepressant treatment response identified by random forest modeling. \u003c/strong\u003eLog-transformed relative abundances of 14 baseline taxa (5 genera and 9 species) identified in Model IV were used as predictors to classify treatment response at follow-up. \u003cstrong\u003e(A) \u003c/strong\u003eReceiver operating characteristic (ROC) curve showing the classification performance of the random forest model; \u003cstrong\u003e(B)\u003c/strong\u003e Variable importance scores of baseline microbial taxa contributing to discrimination between responders and non-responders.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/093cc63722bd4f928776a9d6.png"},{"id":103397986,"identity":"a2b8ed37-7b87-4f62-81e8-59f90ffaba61","added_by":"auto","created_at":"2026-02-25 08:58:36","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":186041,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eBidirectional Δ-based cross-lagged panel models linking targeted microbial species and depressive symptom changes. \u003c/strong\u003eEach panel displays regression coefficients from cross-lagged models assessing temporal associations between changes in species abundance and changes in HAMD scores across consecutive time intervals, with adjustment for prior values of both variables. Species abundances were log-transformed, and pseudocounts (1e-20) were added before modeling.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/07cbf073990b41b05068a59d.png"},{"id":103398128,"identity":"4bc8e889-8a1a-4f0e-b767-cf330f2b15fe","added_by":"auto","created_at":"2026-02-25 08:58:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":3929275,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpecies-level functional contributions to targeted microbial pathways stratified by treatment response.\u003c/strong\u003e Sankey diagrams depict the relative contributions of targeted microbial species (left) to targeted MetaCyc pathways (right), stratified by treatment response at TP2. Node colors indicate individual species or pathways, and link widths represent the magnitude of species-level contribution to pathway activity derived from HUMAnN stratified outputs. Upper: Non-responders; Lower: Responders.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/6f22d28f2ed59e0ddc615159.png"},{"id":103397951,"identity":"55c79206-ee54-4e8c-86fb-9b9bfc28f692","added_by":"auto","created_at":"2026-02-25 08:58:31","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3141185,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCorrelation structure and longitudinal dynamics of fecal metabolites and targeted microbial species. (A, B)\u003c/strong\u003eHierarchically clustered Spearman correlations (ρ) between targeted species and bile acids (A) or short-chain fatty acids (B). \u003cstrong\u003e(C)\u003c/strong\u003e Associations between changes in bile acid levels (Δmetabolite) and depressive symptom changes (ΔHAMD) from baseline to TP2; shaded areas indicate 95% confidence intervals; \u003cstrong\u003e(D)\u003c/strong\u003eLongitudinal trajectories of representative bile acid and SCFA metabolites stratified by treatment response across BL, TP1, and TP2; points denote group means ±95% confidence intervals.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/775cf2f687b04214d3ecf7aa.png"},{"id":103397979,"identity":"ebb16cde-f569-4709-9329-9120e4ce93f7","added_by":"auto","created_at":"2026-02-25 08:58:33","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3875639,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eProteome-wide screening, multi-omics integration, and pathway interpretation of treatment-responsive plasma proteins.\u003c/strong\u003e \u003cstrong\u003e(A) \u003c/strong\u003eProteome-wide volcano plot comparing responders and non-responders at TP2. The x-axis shows log₂ fold change (Responder/Non-responder) and the y-axis shows −log₁₀(p-value). Proteins significant at both baseline and TP2 are highlighted in orange; TP2-specific significant proteins (p \u0026lt; 0.05) in blue; extended candidates (0.05 ≤ p \u0026lt; 0.1 with |log₂FC| ≥ 1.2) in green; and non-significant proteins in grey. Triangles indicate proteins retained in serial mediation analysis (TP2 p \u0026lt; 0.05 and |log₂FC| ≥ 1.3). Horizontal and vertical reference lines denote p-value and fold-change thresholds, respectively. The analysis includes 9,458 plasma proteins; \u003cstrong\u003e(B)\u003c/strong\u003e DIABLO circos plot of multi-omics associations related to treatment response. Significant correlations (|r| \u0026gt; 0.5) among microbial species (blue), bile acid–related metabolites (green), and treatment-responsive proteins (red) are shown, with red and blue links indicating positive and negative correlations. Outer tracks display scaled abundance trajectories in responders and non-responders; \u003cstrong\u003e(C)\u003c/strong\u003e Top reactome pathway enrichment of treatment-responsive proteins. Pathways were selected by the significance p \u0026lt; 0.01 (n = 11). The x-axis represents the combined enrichment score, and the y-axis lists enriched Reactome pathways. Dot size reflects the number of overlapping proteins, and dot color indicates statistical significance (−log₁₀ P), with darker shading indicating stronger enrichment.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/c40dd205ef920c469f0f376e.png"},{"id":103398257,"identity":"460b672b-13c1-4315-8d88-bf1d558e6ed1","added_by":"auto","created_at":"2026-02-25 08:59:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":13568653,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/755886f5-08a1-4737-a78e-2f3391e8cc24.pdf"},{"id":103398120,"identity":"6b0aa7b3-31f8-498a-9837-3d6e3139e0ff","added_by":"auto","created_at":"2026-02-25 08:58:49","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5781848,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Information\u003c/p\u003e","description":"","filename":"Supplementaryinformationver120250119.docx","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/cac3dea7d77cc7f56f676ddf.docx"},{"id":103397949,"identity":"535a67cf-fadb-417d-ba98-4c1a0ff2f307","added_by":"auto","created_at":"2026-02-25 08:58:31","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":28208,"visible":true,"origin":"","legend":"","description":"","filename":"MainResultTable1ver120250119.docx","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/dcb380133e4e78ec82e755c3.docx"},{"id":103398142,"identity":"76e2cd8c-b0b2-466a-9919-2e9f9a85c6da","added_by":"auto","created_at":"2026-02-25 08:58:56","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":26208,"visible":true,"origin":"","legend":"","description":"","filename":"MainResultTable2ver120250119.docx","url":"https://assets-eu.researchsquare.com/files/rs-8836038/v1/ad4b799b1a94458bbb868c0c.docx"}],"financialInterests":"The authors have declared there is \u003cb\u003eNO\u003c/b\u003e conflict of interest to disclose","formattedTitle":"Multi–omic Longitudinal Profiling Reveals Coordinated Gut Microbiota, Metabolomics, and Host Proteomics Signatures Predictive of Antidepressant Response in Drug–Naïve Depression","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eMajor depressive disorder (MDD) affects approximately 5.7% of adults globally and is consistently ranked among the leading causes of disability worldwide\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Although antidepressants remain the primary treatment modality, clinical responses are highly heterogeneous, with 30\u0026ndash;50% of patients exhibiting inadequate improvement despite appropriate therapy\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Elucidating the biological underpinnings of this variability remains a critical unmet need in psychiatry.\u003c/p\u003e \u003cp\u003eThe gut microbiota influences brain function through metabolic, immune, and neuroendocrine signaling pathways\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Microbial metabolites, particularly short\u0026ndash;chain fatty acids (SCFAs), bile acids, and tryptophan derivatives, serve as key neuromodulatory and immunoregulatory intermediaries\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e and have been increasingly implicated in the response to antidepressant treatment\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Dysbiosis observed in depression\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e may shape pharmacological responsiveness, while antidepressants can reciprocally alter microbial composition and function\u003csup\u003e\u003cspan additionalcitationids=\"CR10 CR11\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. At the host level, plasma proteomic studies have identified alterations in inflammatory, metabolic, and neuroendocrine markers in first\u0026ndash;episode MDD patients\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, many of which normalize\u0026mdash;or fail to normalize\u0026mdash;in parallel with antidepressant treatment response\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Together, these findings suggest that microbial metabolites and host systemic responses constitute interconnected determinants of antidepressant efficacy\u003csup\u003e\u003cspan additionalcitationids=\"CR17\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBuilding on this framework, recent studies have begun to examine how the gut microbiota contribute to heterogeneity in antidepressant treatment outcomes\u003csup\u003e\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Despite accumulating evidence linking the gut microbiota to depression and treatment response, critical gaps remain. Most studies employ cross\u0026ndash;sectional designs or assess single time points, limiting the identification of temporal dynamics that distinguish responders from non\u0026ndash;responders\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Existing work has predominantly relied on taxonomic profiling, providing limited mechanistic insight into microbial functional pathways, metabolic outputs, or host molecular responses\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Multi\u0026ndash;omic integration spanning microbiome composition, functional capacity, metabolite production, and host proteomic changes is increasingly recognized as critical for understanding microbiota\u0026ndash;host interactions\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. However, such integrative analyses remain scarce, and few studies have systematically examined pathway\u0026ndash;level organization, functional redundancy, or bidirectional coupling between microbiota dynamics and clinical symptom trajectories. Although multi\u0026ndash;omic approaches have been applied in controlled animal models to dissect mechanistic host\u0026ndash;microbiota interactions, comparable integrative frameworks in human studies remain limited\u003csup\u003e\u003cspan additionalcitationids=\"CR27\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo address these gaps, we conducted a longitudinal study of antidepressant\u0026ndash;na\u0026iuml;ve MDD patients sampled at three time points during treatment, integrating shotgun metagenomics, microbial functional pathway analysis, targeted fecal bile acid and SCFA quantification, and high\u0026ndash;dimensional plasma proteomics. We applied complementary statistical frameworks\u0026mdash;including mixed\u0026ndash;effects models, cross\u0026ndash;lagged panel analyses, machine learning prediction, and serial mediation\u0026mdash;to (1) identify temporal microbial, metabolic, and proteomic signatures distinguishing responders from non\u0026ndash;responders; (2) assess bidirectional relationships between symptom change and microbial dynamics; (3) characterize functional pathway organization and redundancy; and (4) map microbiota\u0026ndash;metabolite\u0026ndash;host protein networks underlying treatment response. This integrated approach aims to delineate the biological mechanisms underlying heterogeneity in antidepressant response and inform microbiome\u0026ndash;based precision psychiatry strategies.\u003c/p\u003e"},{"header":"2 Materials and Methods","content":"\u003cp\u003eThis longitudinal discovery study was designed to characterize multi\u0026ndash;omic dynamics across antidepressant treatment with sufficient temporal resolution (three time points) and biological depth (four omics layers) to identify mechanistic signatures of treatment response. While the sample size (n = 28 participants, 81 complete observations across time points) is modest, it provides adequate statistical power for our primary aims. For longitudinal mixed\u0026ndash;effects models, we achieve \u0026gt;80% power to detect medium\u0026ndash;to\u0026ndash;large effect sizes (|Cohen\u0026rsquo;s d| \u0026ge; 0.7) for within\u0026ndash;subject temporal changes at \u0026alpha; = 0.05 with 81 observations\u003csup\u003e29\u003c/sup\u003e. For cross\u0026ndash;sectional between\u0026ndash;group comparisons (responders vs. non\u0026ndash;responders), we have 80% power to detect large effect sizes (d \u0026ge; 0.9) for 17 responders and 11 non\u0026ndash;responders\u003csup\u003e30\u003c/sup\u003e. Finally, the depth of multi\u0026ndash;omic profiling (288 species, 90 pathways, 31 metabolites, 9,458 proteins) enables systems\u0026ndash;level pattern detection across correlated biological layers\u003csup\u003e31\u003c/sup\u003e. We emphasize that this study represents a discovery cohort designed to generate hypotheses and identify candidate biomarkers requiring validation in larger, independent samples.\u003c/p\u003e\n\u003ch3\u003e2.1 Participants and Clinical Assessment\u003c/h3\u003e\n\u003cp\u003ePatients aged 20\u0026ndash;65 years with antidepressant\u0026ndash;na\u0026iuml;ve MDD (DSM\u0026ndash;5)\u003csup\u003e32\u003c/sup\u003e were recruited from psychiatric clinics in Taipei. The cohort included both first\u0026ndash;episode (n = 21) and recurrent depression (n = 7). Exclusions followed established criteria (schizophrenia, schizoaffective disorder, intellectual disability, drug\u0026ndash;induced mood disorders, recent gastrointestinal surgery (within 2 months), current infections, or recent use of probiotics, prebiotics, symbiotics, or antibiotics). All procedures were IRB\u0026ndash;approved with written informed consent. Participants were assessed at BL (baseline), TP1 (one\u0026ndash;month), and TP2 (two\u0026ndash;month)\u003csup\u003e33\u003c/sup\u003e. Demographics were collected at BL, and clinical measures and stool/blood samples were obtained at each visit (Fig. S1). Stool was transported at 4 \u0026deg;C with a 200\u0026ndash;mg pellet reserved for DNA extraction, and plasma was isolated from EDTA blood (centrifuged at 3000 g, 4 \u0026deg;C). After exclusions, 25 participants contributed complete longitudinal data (75 samples across BL/TP1/TP2), and three provided partial follow\u0026ndash;up (BL and TP1 only; six samples).\u003c/p\u003e\n\u003cp\u003eAmong all participants, antidepressant selection followed clinical judgment by treating psychiatrists and included SSRIs (n = 10, 35.71%), SNRIs (n = 2, 7.14%), and other antidepressants (n = 16, 57.14%). Medication type remained unchanged throughout the study period, and antidepressant dose remained stable in 21 participants. Concomitant psychotropic medications were documented but not considered exclusionary if stable for \u0026ge;2 weeks before baseline. Baseline diet was evaluated with a validated short\u0026ndash;form food\u0026ndash;frequency questionnaire\u003csup\u003e34\u003c/sup\u003e, and participants received guidance to maintain stable dietary patterns during the study; interim dietary changes were not formally assessed.\u003c/p\u003e\n\u003cp\u003eDepressive symptoms were assessed at each visit with the 17\u0026ndash;item Hamilton Depression Rating Scale (HAMD)35. Treatment response was defined by percent reduction in HAMD and by categorical response (\u0026ge;50% reduction), evaluated at TP1 and TP2\u003csup\u003e36,37\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3 id=\"_Toc213360226\"\u003e2.2 Shotgun metagenomic sequencing, quality control, and functional profiling\u003c/h3\u003e\n\u003cp\u003eStool DNA was extracted using the QIAamp\u0026reg; PowerFecal\u0026reg; Pro Kit following the manufacturer\u0026rsquo;s protocol. Metagenomic libraries were prepared using the Illumina Nextera XT kit and sequenced on an Illumina NovaSeq 6000 platform to generate paired\u0026ndash;end reads. Raw reads underwent standard quality control and trimming, followed by human\u0026ndash;read removal (Figs. S2\u0026ndash;3; Table S1). Taxonomic profiling was performed with MetaPhlAn 4 using the CHOCOPhlAn SGB database to obtain species\u0026ndash;level relative abundances\u003csup\u003e38\u003c/sup\u003e. Functional profiling was conducted with HUMAnN 3.2\u003csup\u003e39\u003c/sup\u003e, mapping reads to ChocoPhlAn and UniRef90 to derive MetaCyc pathway abundances\u003csup\u003e40\u003c/sup\u003e. Species contribution profiles were generated to quantify taxon\u0026ndash;function coupling. All microbial features were pseudocount\u0026ndash;adjusted and log\u0026ndash;transformed before analysis\u003csup\u003e41\u003c/sup\u003e. Laboratory personnel conducting multi\u0026ndash;omic profiling were blinded to response status and time\u0026ndash;point labels during data generation.\u003c/p\u003e\n\u003ch3 id=\"_Toc213360228\"\u003e2.3 Stool metabolite profiling\u003c/h3\u003e\n\u003cp\u003eBile acids (BAs) were quantified from ~20 mg stool using a methanol/acetonitrile/water/formic acid extraction containing isotope\u0026ndash;labeled internal standards. Extracts were analyzed by LC\u0026ndash;MS/MS (positive\u0026ndash;ion ESI, MRM) and quantification was performed using TargetLynx. SCFAs were profiled from ~50 mg stool using derivatization and analyzed by GC\u0026ndash;MS (EI mode). Metabolite levels were log\u0026ndash;transformed for downstream analysis.\u003c/p\u003e\n\u003ch3\u003e2.4 Plasma proteomics profiling\u003c/h3\u003e\n\u003cp\u003eTargeted plasma proteomics was performed using the Illumina Protein Prep (IPP) SOMAmer platform (~9,500 proteins; CV ~5.5%)\u003csup\u003e42\u003c/sup\u003e. Approximately 130 \u0026micro;L plasma per sample was processed following the standard IPP workflow and sequenced on an Illumina NovaSeq system. Protein values were log\u0026ndash;transformed before analysis.\u003c/p\u003e\n\u003ch3\u003e2.5 Bioinformatic and data processing\u003c/h3\u003e\n\u003cp\u003eShotgun metagenomics and targeted metabolomics generated three data layers: taxonomic profiles (MetaPhlAn 4), functional pathways (HUMAnN 3.2), and targeted BA/SCFA quantifications. Taxonomic features were aggregated to phylum, genus, and species levels, and low\u0026ndash;prevalence features (\u0026lt;20%) were removed, yielding 149 genera and 288 species (Fig. S4)\u003csup\u003e43\u003c/sup\u003e. HUMAnN pathway tables were curated by excluding low\u0026ndash;coverage (\u0026lt;0.3) and noninformative pathways, resulting in 90 high\u0026ndash;confidence MetaCyc pathways\u003csup\u003e40\u003c/sup\u003e (Figs. S5\u0026ndash;6). Targeted metabolomics initially quantified 46 BAs and 11 SCFAs; after LOQ\u0026ndash;based filtering and preprocessing, 21 BAs and 10 SCFAs were retained (Tables S2\u0026ndash;3; Fig. S7)\u003csup\u003e44\u003c/sup\u003e. Plasma proteomics identified 10,272 proteins; after QC and excluding unmapped targets\u003csup\u003e45\u003c/sup\u003e, 9,458 proteins were retained and annotated against Reactome curated pathway knowledgebase\u003csup\u003e46\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3 id=\"_Toc213360230\"\u003e2.6 Statistical analysis\u003c/h3\u003e\n\u003cp\u003eAll statistical analyses were performed in R and RStudio. Demographic and clinical characteristics were compared between responders and non\u0026ndash;responders using non\u0026ndash;parametric tests (continuous variables) and \u0026chi;\u0026sup2; or Fisher\u0026rsquo;s exact tests (categorical variables); \u0026alpha;\u0026ndash; and \u0026beta;\u0026ndash;diversity indices were evaluated but were not a primary focus.\u003c/p\u003e\n\u003cp\u003eTo characterize microbial shifts, we used: time\u0026ndash;as\u0026ndash;category mixed models (Model I); time\u0026ndash;as\u0026ndash;continuous models incorporating \u0026Delta;HAMD (TP1\u0026ndash;BL; TP2\u0026ndash;BL) with Mundlak adjustment (Model II); and responder vs non\u0026ndash;responder fold\u0026ndash;change analyses (TP1\u0026ndash;BL, TP2\u0026ndash;BL) (Model III). Baseline predictors of TP2 response were evaluated using Wilcoxon screening and a Random Forest classifier with repeated 5\u0026ndash;fold cross\u0026ndash;validation and permutation testing (Model IV). All models were adjusted for baseline dietary covariates; nominally significant species in Models II\u0026ndash;III were selected as targeted taxa.\u003c/p\u003e\n\u003cp\u003eSpecies\u0026ndash;pathway coupling models predicted pathway fold\u0026ndash;changes from targeted species fold\u0026ndash;changes in mixed\u0026ndash;effects models; HUMAnN stratified outputs quantified species contributions\u003csup\u003e39\u003c/sup\u003e. Functional redundancy was characterized using Shannon entropy and Gini coefficients\u003csup\u003e47\u003c/sup\u003e. \u0026Delta;\u0026ndash;based cross\u0026ndash;lagged panel models assessed bidirectional temporal associations between symptom changes and targeted species. Serial mediation tested sequential mediation via metabolites and proteins (1,000 bootstraps). Limma\u0026ndash;trend tests was applied for cross\u0026ndash;sectional metabolite/protein comparisons\u003csup\u003e48\u003c/sup\u003e. Multi\u0026ndash;omics integration used DIABLO (mixOmics)\u003csup\u003e49\u003c/sup\u003e, which identifies correlated features across species, metabolites, and proteins while optimizing discrimination between responders and non\u0026ndash;responders. We set \u0026alpha; = 0.05 and controlled FDR using Benjamini\u0026ndash;Hochberg\u003csup\u003e50\u003c/sup\u003e.\u003c/p\u003e"},{"header":"3 Results","content":"\u003ch3 id=\"_Toc213360232\"\u003e3.1 Clinical characteristics and treatment response patterns\u003c/h3\u003e\n\u003cp\u003eOf the 28 participants (mean age 36.11 \u0026plusmn; 13.68 years; 64.29% female), 17 (60.71%) achieved a \u0026ge;50% reduction in HAMD scores by their latest follow\u0026ndash;up visit and were classified as responders. Treatment response increased from 50% at TP1 to 64% at TP2. Baseline demographic, dietary, and clinical characteristics did not differ significantly between eventual responders and non\u0026ndash;responders, including age, sex, BMI, or baseline HAMD severity.\u0026nbsp;Antidepressant class distribution and concomitant medication use showed no association with response status. Only responders demonstrated significant HAMD reductions across time points (BL: 16.73 \u0026plusmn; 4.85 \u0026rarr; TP1: 9.53 \u0026plusmn; 5.37 \u0026rarr; TP2: 4.14 \u0026plusmn; 4.35; p \u0026lt; 0.001), whereas non\u0026ndash;responders showed minimal change (BL: 17.45 \u0026plusmn; 5.41 \u0026rarr; TP2: 13.22 \u0026plusmn; 4.68; p = 0.16)\u0026nbsp;(Tables S4\u0026ndash;5; Figs. S8\u0026ndash;9).\u003c/p\u003e\n\u003ch3 id=\"_Toc213360233\"\u003e3.2 Taxonomic dynamics: temporal changes and response\u0026ndash;specific trajectories\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eGlobal community structure remained stable during treatment. The \u0026alpha;\u0026ndash; and \u0026beta;\u0026ndash;diversity metrics showed no significant temporal changes or differences between responders and non\u0026ndash;responders (Figs. S10\u0026ndash;11), indicating that antidepressant treatment did not induce community\u0026ndash;wide restructuring. Despite stable diversity metrics, Model I exhibited treatment\u0026ndash;associated species temporal dynamics. Early reductions (BL\u0026rarr;TP1) occurred in facultative anaerobes and inflammation\u0026ndash;associated taxa: \u003cem\u003eWeissella confusa/cibaria\u003c/em\u003e, \u003cem\u003eLactococcus lactis\u003c/em\u003e, \u003cem\u003eKlebsiella pneumoniae\u003c/em\u003e, \u003cem\u003eIntestinibacter bartlettii, Turicibacter bilis\u003c/em\u003e, and \u003cem\u003eActinomyces naeslundii\u003c/em\u003e (Table S6; Fig. S12). Model II displayed symptom\u0026ndash;linked microbial changes, and Mundlak\u0026ndash;adjusted models disentangled pure temporal effects from symptom\u0026ndash;related changes while revealing substantial overlap between the two (Tables S7\u0026ndash;8; Fig. S13). Time\u0026ndash;as\u0026ndash;continuous models incorporating \u0026Delta;HAMD revealed that 27 species changed in association with either treatment duration or symptom improvement. Notably, several taxa\u0026mdash;including \u003cem\u003eDialister hominis\u003c/em\u003e, \u003cem\u003eAnaerostipes hadrus\u003c/em\u003e, \u003cem\u003ePhascolarctobacterium faecium\u003c/em\u003e, \u003cem\u003eRuminococcus lactaris\u003c/em\u003e, \u003cem\u003eLachnospira\u003c/em\u003e spp., and \u003cem\u003eClostridiaceae\u003c/em\u003e \u003cem\u003ebacterium Marseille\u0026ndash;Q3526\u003c/em\u003e\u0026mdash;were associated with both time progression and changes in depressive symptoms (\u0026Delta;HAMD), indicating that these taxa exhibit dynamic trajectories jointly shaped by pharmacological exposure and clinical improvement. However, none of these associations remained statistically significant after multiple testing correction.\u003c/p\u003e\n\u003cp\u003eModel III revealed response\u0026ndash;group divergent trajectories (Table 1; Fig. S14). At TP1 (early divergence) responders showed stronger decreases in \u003cem\u003eHungatella hathewayi, Faecalicatena contorta, Wansuia hejianensis,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Blautia producta\u003c/em\u003e, along with increases in \u003cem\u003eStreptococcus sanguinis\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Megamonas funiformis\u003c/em\u003e. At TP2 (late divergence) responders exhibited increases in\u003cem\u003e\u0026nbsp;Clostridium leptum, Clostridium saudiense,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Oscillospiraceae\u003c/em\u003e_unclassified, and decreases\u003cem\u003e\u0026nbsp;\u003c/em\u003ein\u003cem\u003e\u0026nbsp;Eubacterium ventriosum, Blautia glucerasea,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Catenibacterium tridentinum\u003c/em\u003e. Model IV identified microbial signatures at baseline predictive of TP2 treatment response. A Random Forest classifier achieved high discriminative accuracy (cross\u0026ndash;validated AUC = 0.965, 95% CI: 0.84\u0026ndash;1.00) confirmed by permutation testing (p_perm \u0026lt; 0.001). Top predictive features included\u003cem\u003e\u0026nbsp;Raoultibacter timonensis, Blautia faecis,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Escherichia coli\u003c/em\u003e (Fig. 1; Tables S9\u0026ndash;10; Figs. S15\u0026ndash;16). No results remained significant after correction for multiple testing.\u003c/p\u003e\n\u003ch3 id=\"_Toc213360236\"\u003e3.3 Bidirectional microbiota\u0026ndash;symptom relationships\u003c/h3\u003e\n\u003cp\u003eCross\u0026ndash;lagged panel models revealed temporally ordered bidirectional associations between \u0026Delta;HAMD and changes in 17 targeted species (Fig. 2; Table S11). When modeling symptom change predicting subsequent microbial abundance (\u0026Delta;HAMD \u0026rarr; future microbial abundance) at the nominal level (P \u0026lt; 0.05), greater symptom improvement predicted higher abundance in \u003cem\u003eAnaerostipes hadrus\u003c/em\u003e, \u003cem\u003eMegasphaera sp. NM10\u003c/em\u003e, \u003cem\u003eDialister hominis\u003c/em\u003e, and \u003cem\u003ePhascolarctobacterium faecium\u003c/em\u003e at TP1. During TP1\u0026ndash;TP2, less symptom improvement predicted higher abundance in \u003cem\u003eActinomyces graevenitzii\u003c/em\u003e and \u003cem\u003eRuminococcus lactaris\u003c/em\u003e at TP2, while the effect in \u003cem\u003eBlautia hansenii\u003c/em\u003e was reversed. Across BL\u0026ndash;TP2, all significant species\u0026mdash;including \u003cem\u003eAlistipes indistinctus\u003c/em\u003e, \u003cem\u003eAlistipes shahii\u003c/em\u003e, \u003cem\u003eClostridiaceae bacterium Marseille\u0026ndash;Q3526\u003c/em\u003e, \u003cem\u003eLachnospira sp. NSJ_43\u003c/em\u003e, and \u003cem\u003eRuminococcus lactaris\u003c/em\u003e\u0026mdash;displayed positive coefficients. A subset of these relationships also survived FDR correction, including \u003cem\u003eAnaerostipes hadrus, Actinomyces graevenitzii, Clostridiaceae bacterium Marseille\u0026ndash;Q3526,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Ruminococcus lactaris\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eIn the reverse models predicting HAMD scores from \u0026Delta;abundance at the nominal level, most BL\u0026ndash;TP1 associations were negative (e.g., \u003cem\u003eAlistipes communis\u003c/em\u003e, \u003cem\u003eAnaerostipes hadrus\u003c/em\u003e, \u003cem\u003eDialister hominis\u003c/em\u003e), whereas TP1\u0026ndash;TP2 showed mixed effects, with \u003cem\u003eActinomyces graevenitzii\u003c/em\u003e positive and \u003cem\u003eBlautia hansenii\u003c/em\u003e and \u003cem\u003eDialister hominis\u003c/em\u003e negative. During BL\u0026ndash;TP2, most species demonstrated positive coefficients, except for \u003cem\u003eDialister hominis\u003c/em\u003e and \u003cem\u003eMogibacterium diversum\u003c/em\u003e, which were negative. Together, these results reveal temporally structured and species\u0026ndash;specific bidirectional relationships, with taxa such as \u003cem\u003eDialister hominis\u003c/em\u003e, \u003cem\u003eRuminococcus lactaris\u003c/em\u003e, and \u003cem\u003eClostridiaceae bacterium Marseille\u0026ndash;Q3526\u003c/em\u003e showing the most consistent cross\u0026ndash;lagged patterns, indicating dynamic feedback between these taxa and depressive symptoms. Among these relationships, those involving \u003cem\u003eAnaerostipes hadrus, Dialister hominis, Ruminococcus lactaris,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Clostridiaceae bacterium Marseille\u0026ndash;Q3526\u003c/em\u003e remained robust after FDR correction.\u003c/p\u003e\n\u003ch3 id=\"_Toc213360239\"\u003e3.4 Functional pathway dynamics and redundancy\u003c/h3\u003e\n\u003cp\u003eFunctional profiling identified 90 high\u0026ndash;confidence MetaCyc pathways dominated by core biosynthetic functions: amino acid biosynthesis (29%), nucleotide biosynthesis (17%), carbohydrate metabolism (11%), and cell structure biosynthesis (10%) (Fig. S17). Species\u0026ndash;pathway modeling identified functional links associated with the 38 targeted species. Among 3,420 species\u0026ndash;pathway pairs tested, 397 were nominally significant and 12 remained significant after FDR correction (Fig. S18; Table S12). \u003cem\u003eAlistipes shahii\u003c/em\u003e showed the broadest associations, negatively linking to multiple core biosynthetic pathways (L\u0026ndash;lysine biosynthesis VI, peptidoglycan biosynthesis I\u0026ndash;II, UMP biosynthesis I\u0026ndash;II, S\u0026ndash;adenosyl\u0026ndash;L\u0026ndash;methionine salvage I). Additional associations were observed for \u003cem\u003eWansuia hejianensis\u003c/em\u003e (inosine\u0026ndash;5\u0026rsquo;\u0026ndash;phosphate degradation), \u003cem\u003eMogibacterium diversum\u003c/em\u003e (L\u0026ndash;threonine biosynthesis), and \u003cem\u003eWeissella confusa\u003c/em\u003e (UDP\u0026ndash;glucose\u0026ndash;derived O\u0026ndash;antigen biosynthesis).\u003c/p\u003e\n\u003cp\u003eSpecies\u0026ndash;level contribution analysis identified \u003cem\u003ePrevotella copri, Megamonas funiformis, Ruminococcus bromii, Roseburia faecis,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Klebsiella pneumoniae\u003c/em\u003e as dominant contributors (\u0026gt;1%) to targeted pathways. Seven targeted species (\u003cem\u003eClostridium scindens\u003c/em\u003e, \u003cem\u003eMegamonas funiformis\u003c/em\u003e, \u003cem\u003ePhascolarctobacterium faecium\u003c/em\u003e, \u003cem\u003eClostridium leptum\u003c/em\u003e, \u003cem\u003eRuminococcus lactaris\u003c/em\u003e, \u003cem\u003eStreptococcus sanguinis\u003c/em\u003e, and \u003cem\u003eWeissella confusa\u003c/em\u003e) were listed as contributors to our targeted pathways (Tables S13\u0026ndash;14; Fig. S19). Sankey diagrams revealed striking divergence in pathway contribution structure between response groups. Responders maintained broader, multispecies contributions to core biosynthetic pathways, with parallel functional inputs from multiple taxa\u0026mdash;especially \u003cem\u003eM. funiformis\u0026nbsp;\u003c/em\u003e(1.8\u0026ndash;2.1% contribution)\u003cem\u003e, P. faecium\u0026nbsp;\u003c/em\u003e(1.1\u0026ndash;1.5%)\u003cem\u003e,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;R. lactaris\u0026nbsp;\u003c/em\u003e(~0.1%). Non\u0026ndash;responders exhibited progressive functional narrowing, with TP2 characterized by increased dominance of \u003cem\u003eR. lactaris\u0026nbsp;\u003c/em\u003e(0.2\u0026ndash;0.4% contribution), reduced contributions from \u003cem\u003eP. faecium\u0026nbsp;\u003c/em\u003e(0.3\u0026ndash;0.4%), and depleted contributions of \u003cem\u003eC. scindens\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;S. sanguinis\u003c/em\u003e (Fig. 3, Table S15). For functional redundancy (Shannon H) and dominance (Gini G), stratified analyses showed opposing patterns: in pathways such as inosine\u0026ndash;5\u0026rsquo;\u0026ndash;phosphate degradation and L\u0026ndash;lysine/L\u0026ndash;threonine biosynthesis, responders exhibited a transient TP1 reduction followed by recovery at TP2 (BL = 0.59 \u0026rarr; TP1 = 0.39 \u0026rarr; TP2 = 0.47), whereas non\u0026ndash;responders showed a continuous decline (BL = 0.60 \u0026rarr; TP1 = 0.39 \u0026rarr; TP2 = 0.35) (Fig. S20). Gini coefficients demonstrated reciprocal patterns. Polynomial trend analysis revealed a significant quadratic trajectory of functional redundancy in responders (p = 0.003), as measured by both Shannon entropy (H) and the Gini index, whereas no significant temporal trend was observed in non\u0026ndash;responders.\u003c/p\u003e\n\u003ch3 id=\"_Toc213360241\"\u003e3.5 Metabolite signatures: bile acids and short\u0026ndash;chain fatty acids\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eSpearman correlations between \u0026Delta;abundance of 38 targeted species and 31 metabolites revealed two major co\u0026ndash;association modules (Figs. 4A\u0026ndash;B; Table S16). Module 1 (positive associations) included bile\u0026ndash;acid\u0026ndash;producing and butyrate\u0026ndash;associated taxa \u0026mdash; \u003cem\u003eB. producta, A. shahii, C. leptum,\u003c/em\u003e and \u003cem\u003eC. scindens\u003c/em\u003e positively correlated with secondary/deconjugated bile acids and branched\u0026ndash;chain fatty acids. Module 2 (negative associations) included inflammation\u0026ndash;associated taxa \u0026mdash; \u003cem\u003eW. confusa, P. faecium, H. hathewayi, W. hejianensis, M. funiformis,\u003c/em\u003e and \u003cem\u003eF. contorta\u003c/em\u003e showed negative correlations with conjugated bile acids and several SCFAs. Representative associations that remained significant after false discovery rate (FDR) correction included \u003cem\u003eW. confusa\u003c/em\u003e with 3\u0026alpha;\u0026ndash;hydroxy\u0026ndash;7\u0026ndash;ketolithocholic acid (\u0026rho; = \u0026minus;0.546), \u003cem\u003eP. faecium\u003c/em\u003e with glycochenodeoxycholic acid (\u0026rho; = \u0026minus;0.501), and \u003cem\u003eB. producta\u003c/em\u003e with 2\u0026ndash; and 3\u0026ndash;Methylbutanoic acid (\u0026rho; = 0.581, 0.575).\u003c/p\u003e\n\u003cp\u003eTwo bile acids\u0026mdash;deoxycholic acid and 5\u0026beta;\u0026ndash;cholanic acid\u0026ndash;3\u0026beta;,12\u0026alpha;\u0026ndash;diol\u0026mdash;inversely correlated\u0026nbsp;with \u0026Delta;HAMD, indicating higher levels when symptoms improved (Fig. 4C).\u0026nbsp;Limma\u0026ndash;trend differential testing identified response\u0026ndash;dependent metabolic remodeling\u0026nbsp;(Fig. 4D; Fig. S21).\u0026nbsp;Responders showed BL \u0026rarr; TP2 increases in multiple secondary bile acids (e.g., deoxycholic acid, glycolithocholic acid, isolithocholic acid, lithocholic acid), whereas non\u0026ndash;responders exhibited parallel declines (e.g.,\u0026nbsp;deoxycholic acid and glycolithocholic acid). For SCFAs, responders displayed a TP1 \u0026rarr; TP2 rebound of butyric acid (TP2 p = 0.041), while non\u0026ndash;responders remained flat. Conversely, hexanoic acid rose transiently in non\u0026ndash;responders but declined in responders (TP2 p = 0.027). Although these effects were detected at the nominal significance level, none of the metabolite changes remained significant after FDR correction.\u003c/p\u003e\n\u003ch3\u003e3.6 Treatment\u0026ndash;responsive plasma proteomic signatures and multi\u0026ndash;omic integration\u003c/h3\u003e\n\u003cp\u003eAmong 9,458 quantified plasma proteins, 173 proteins showed differential abundance at TP2\u0026nbsp;between responders and non\u0026ndash;responders, including proteins significant at BL and TP2 (p\u0026lt;0.05,\u0026nbsp;Fig. 5A). Eighteen proteins exhibited consistent group differences at both BL and TP2, whereas the majority displayed heterogeneous longitudinal trajectories. Notably, longitudinal trajectory analysis revealed heterogeneous patterns: progressive divergence, in which responder\u0026ndash;non\u0026ndash;responder differences were increasingly amplified (e.g., BCHE, CD2, MICA, and NOTCH2), and cross\u0026ndash;over dynamics, characterized by attenuation and reversal of baseline differences during treatment (e.g., PAXIP1 and TFPI2). Importantly, such trajectory patterns were also evident among proteins without significant baseline differences\u0026nbsp;but showing progressive separation, including\u0026nbsp;ARFRP1,\u0026nbsp;BCHE,\u0026nbsp;CAMP, GRN, IGF1, PCYOX1, PRRT4, RNF7, SERGEF. (all p \u0026lt; 0.05 at TP2), suggesting that longitudinal divergence and cross\u0026ndash;over dynamics can emerge during treatment (Fig. S22). Although these trajectory patterns were identified at the nominal significance level, none of the protein\u0026ndash;level differences remained significant after false discovery rate (FDR) correction.\u003c/p\u003e\n\u003cp\u003eSerial mediation analyses (1,000 bootstrap iterations) were used to prioritize treatment\u0026ndash;responsive species\u0026ndash;metabolite\u0026ndash;protein pathways that potentially link microbial changes to host proteomic responses via bile acid and SCFA intermediates (Table 2; Table S17). Among the top 10 ranked pathways (based on total effect size and nominal bootstrap significance), \u003cem\u003eCloacibacillus porcorum\u003c/em\u003e, \u003cem\u003eMegasphaera\u003c/em\u003e sp. NM10, and \u003cem\u003eDialister hominis\u003c/em\u003e were repeatedly linked to host protein changes through specific metabolic intermediates. Bile acid\u0026ndash;mediated pathways (e.g., isolithocholic acid and glycolithocholic acid) connected these species to epithelial\u0026ndash; and immune\u0026ndash;related proteins, whereas hexanoic acid\u0026ndash;mediated pathways linked microbial shifts to proteins involved in immune and stress\u0026ndash;related responses, including HOXB8, CDH17, KLK3, HSPA1L, and CYP2S1.\u0026nbsp;Collectively, these pathways nominate bile acids and SCFAs as candidate metabolic intermediates linking microbial shifts to immune\u0026ndash;, epithelial\u0026ndash;, and stress\u0026ndash;response protein networks.\u003c/p\u003e\n\u003cp\u003eMulti\u0026ndash;omics integration using DIABLO revealed coordinated signatures across treatment\u0026ndash;responsive microbial species, metabolites, and host proteins that partially discriminated responders from non\u0026ndash;responders (species: R\u0026sup2; = 0.53; metabolites: R\u0026sup2; = 0.41; proteins: R\u0026sup2; = 0.76), with the strongest discrimination observed in the protein block (Fig. 5B; Fig. S23). Cross\u0026ndash;omic loadings revealed co-associated features across microbial taxa, metabolites, and proteins.\u003c/p\u003e\n\u003cp\u003eFocusing on Reactome pathways most relevant to antidepressant treatment response and host\u0026ndash;microbiota interactions, treatment\u0026ndash;responsive proteins were primarily enriched in immune defense\u0026ndash;related and host structural remodeling processes (Fig. 5C). Notably, pathways related to antimicrobial peptides and alpha\u0026ndash;defensins, as well as extracellular matrix degradation and matrix metalloproteinase activation, were consistently highlighted. Metabolic and nutrient\u0026ndash;handling pathways, including cobalamin (vitamin B12) uptake and transport and regulation of IGF transport by IGF\u0026ndash;binding proteins, were also enriched. In addition to these core pathways, enrichment was observed for cytokine\u0026ndash; and intracellular signaling\u0026ndash;related processes, including cytokine signaling and MAPK\u0026ndash;related cascades, as well as pathways involved in protein turnover and homeostasis, such as proteasome assembly and neddylation (Table S18). These enrichments suggest that antidepressant response involves coordinated modulation of immune defense, structural plasticity, and stress\u0026ndash;response systems influenced by microbiota\u0026ndash;derived metabolites.\u003c/p\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eThis study presents a longitudinal multi\u0026ndash;omic analysis of antidepressant treatment, integrating gut microbiota, microbial functions, fecal metabolites, and host plasma proteomics. Despite stable global microbial diversity, antidepressant treatment response in drug\u0026ndash;na\u0026iuml;ve MDD patients was characterized by coordinated reorganization across the gut\u0026ndash;microbiome\u0026ndash;host axis and time\u0026ndash;dependent shifts in specific bacterial taxa that tracked depressive symptom trajectories and temporally predicted clinical improvement. While overall functional profiles remained stable, responders exhibited greater functional redundancy and resilience, accompanied by selective modulation of biosynthetic pathways. Metabolomic analyses further identified bile acids and short\u0026ndash;chain fatty acids as key intermediates linking microbial dynamics to treatment response. At the host level, plasma proteomics revealed heterogeneous protein trajectories and pathway\u0026ndash;level alterations related to immune regulation, epithelial defense, and intracellular signaling, which distinguished responders from non\u0026ndash;responders and integrated with microbial and metabolic changes. Together, these findings indicate that effective antidepressant treatment depends on functional resilience of the microbiome\u0026ndash;host axis, the capacity to maintain metabolic flexibility and coordinated host adaptation, rather than compositional change per se.\u003c/p\u003e \u003cp\u003eConsistent with previous reports\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e, gut microbial diversity remained stable throughout antidepressant treatment, with no significant temporal changes or differences by treatment response, supporting selective rather than community\u0026ndash;wide effects of antidepressants\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e. However, beneath this stable surface, longitudinal modeling revealed a structured microbial reorganization driven jointly by time and changes in depressive severity. Early treatment phases featured reductions in facultative anaerobes enriched in inflammatory conditions (e.g., \u003cem\u003eWeissella, Lactococcus, Klebsiella\u003c/em\u003e), followed by a gradual enrichment of obligate anaerobes associated with fiber fermentation and bile\u0026ndash;acid metabolism (e.g., \u003cem\u003eClostridium leptum, C. scindens, Eubacterium\u003c/em\u003e), consistent with progressive restoration of microbial homeostasis\u003csup\u003e\u003cspan additionalcitationids=\"CR54 CR55\" citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. Notably, microbial changes associated with depressive symptom improvement largely followed the same temporal directions as treatment\u0026ndash;associated changes, indicating that symptom improvement modulated the magnitude, rather than the direction, of longitudinal microbial trajectories. Taxa that increased or decreased over time showed more pronounced changes in individuals with greater symptom improvement, suggesting that clinical response amplifies the underlying microbial shifts associated with treatment, rather than altering their overall direction. Importantly, response-stratified analysis revealed longitudinal microbial divergence that was not captured by time\u0026ndash; or symptom\u0026ndash;linked models alone. Rather than reflecting broad restoration across functional groups, these response\u0026ndash;specific patterns suggest that effective treatment is characterized by the capacity to engage late\u0026ndash;phase ecological reorganization. In this context, selective enrichment of specific obligate anaerobes, including the butyrate\u0026ndash;producing taxon \u003cem\u003eClostridium leptum\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e, may represent a hallmark of successful ecological adaptation during treatment, whereas failure to exhibit such late\u0026ndash;phase changes in non\u0026ndash;responders points to incomplete transition following initial treatment\u0026ndash;associated perturbation. When considered alongside the temporal and symptom\u0026ndash;related trends observed in Models I and II, these response\u0026ndash;specific patterns are consistent with a two\u0026ndash;phase framework, in which early treatment\u0026ndash;associated perturbation is followed in responders by selective late\u0026ndash;phase ecological reorganization that may confer metabolically beneficial effects\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. Additionally, baseline microbiome profiles predicted treatment response with high accuracy (cross\u0026ndash;validated AUC\u0026thinsp;=\u0026thinsp;0.965). Beyond longitudinal and response\u0026ndash;associated microbial dynamics, our findings suggest that pretreatment microbiome states may influence sensitivity to antidepressant therapy, consistent with prior attempts to leverage baseline microbiome features for response prediction\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Baseline predictive signatures did not overlap with taxa exhibiting treatment\u0026ndash; or response\u0026ndash;linked trajectories, indicating that pretreatment microbial configurations likely modulate host pharmacometabolic responsiveness rather than directly mediating downstream microbial restoration.\u003c/p\u003e \u003cp\u003eOur cross\u0026ndash;lagged panel analyses advance beyond associational findings by imposing temporal order, revealing that microbiota changes and symptom improvement engage in dynamic reciprocal feedback. Early symptom improvement predicted subsequent microbial shifts and increases in SCFA\u0026ndash;producers (e.g. \u003cem\u003eAnaerostipes hadrus\u003c/em\u003e, \u003cem\u003ePhascolarctobacterium faecium\u003c/em\u003e, \u003cem\u003eDialister hominis\u003c/em\u003e, and \u003cem\u003eMegasphaera sp. NM10\u003c/em\u003e), while prior enrichment of these taxa forecasted later clinical improvement, suggesting a rapid restoration of fermentative and short\u0026ndash;chain\u0026ndash;fatty\u0026ndash;acid\u0026ndash;producing taxa in response to early therapeutic effects\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e,\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e. This bidirectionality aligns with mechanistic models wherein antidepressant\u0026ndash;induced neurobiological changes (e.g., reduced inflammation, restored HPA\u0026ndash;axis function) create a more hospitable gut environment for beneficial microbes, which in turn produce metabolites that reinforce therapeutic effects. Importantly, clinical improvement was associated not only with the expansion of SCFA\u0026ndash;producing taxa but also with the contraction of taxa previously associated with dysbiotic or alternative metabolic states (e.g., \u003cem\u003eAlistipes indistinctus, Clostridiaceae bacterium Marseille\u0026ndash;Q3526, Lachnospira sp. NSJ\u0026ndash;43, and Ruminococcus lactaris\u003c/em\u003e), underscoring that treatment response reflects a reconfiguration of microbial modules rather than a unidirectional change. These findings support a dynamic model in which gut microbial changes are not merely passive correlates of symptom improvement but participate in a dynamic, time\u0026ndash;structured host\u0026ndash;microbe equilibrium in which microbial reorganization both responds to, and actively shapes, clinical recovery.\u003c/p\u003e \u003cp\u003eSpecies-guided pathway analyses moved beyond taxonomic shifts and highlighted functional reorganization underlying treatment response. Core biosynthetic pathways, including L\u0026ndash;lysine, peptidoglycan, and UMP biosynthesis, were repeatedly linked to treatment\u0026ndash;responsive taxa identified in longitudinal and causal models, suggesting coordinated shifts in microbial energy and nucleotide metabolism during antidepressant exposure. Notably, these pathways were predominantly anchored to \u003cem\u003eA. shahii\u003c/em\u003e, whose negative associations across multiple biosynthetic routes are consistent with downregulation of energetically costly metabolic programs and potential attenuation of kynurenine\u0026ndash;related inflammatory signaling\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e,\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e. Within this same functional context, species\u0026ndash;level contribution analyses further indicated complementary involvement of SCFA\u0026ndash;associated taxa, including \u003cem\u003eP. faecium and R. lactaris\u003c/em\u003e, supporting links between microbial fermentation, inflammation, and neurotransmission relevant to depressive symptoms\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e. Responders maintained distributed, multi\u0026ndash;taxon contributions to core biosynthetic functions, whereas non\u0026ndash;responders exhibited progressive taxonomic dominance and reduced functional evenness, indicating reduced ecological resilience under pharmacological perturbation. This pattern parallels ecological principles linking biodiversity and functional redundancy to ecosystem resilience\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e,\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e and indicates that a system with multiple taxa capable of performing essential functions better withstand perturbations. In the gut microbiome context, functional redundancy may enable metabolic flexibility under pharmacological stress, allowing continued production of neuroactive metabolites despite compositional fluctuations. The observation that non\u0026ndash;responders lose redundancy suggests that antidepressant efficacy depends partly on the microbiome's capacity to reorganize functionally while maintaining metabolic output that may be compromised in treatment\u0026ndash;resistant individuals.\u003c/p\u003e \u003cp\u003eShifts in bile acids and SCFAs support a microbiota\u0026ndash;metabolite\u0026ndash;host signaling axis underlying treatment response. Responders showed longitudinal increases in secondary bile acids (particularly deoxycholic acid, lithocholic acid, and related intermediates) and butyrate, whereas non\u0026ndash;responders showed progressive depletion. Secondary bile acids are known to signal via FXR, TGR5, and VDR, thereby suppressing inflammatory tone, enhancing epithelial barrier function, and regulating HPA\u0026ndash;axis activity\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e,\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e, processes central to proposed antidepressant mechanisms. Butyrate similarly exerts anti\u0026ndash;inflammatory, neuroprotective, and epigenetic effects relevant to mood regulation\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e. The divergent hexanoic acid trajectories between groups suggest differential fermentation pathway utilization, potentially reflecting distinct microbial community states. Serial mediation analyses linked these metabolic changes to host proteomic responses, supporting a model in which microbiota\u0026ndash;associated alterations in bile acid and SCFA availability may contribute to treatment efficacy through modulation of systemic immune and stress\u0026ndash;response systems.\u003c/p\u003e \u003cp\u003eProteomic analyses suggest that antidepressant treatment response is characterized by progressive, time\u0026ndash;dependent host molecular adaptation rather than static baseline differences. Heterogeneous protein trajectories\u0026mdash;including divergent and cross\u0026ndash;over patterns that separate responders from non\u0026ndash;responders during treatment\u0026mdash;are consistent with antidepressant efficacy emerging through gradual normalization of multiple physiological systems\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e,\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e. Enrichment of immune defense (antimicrobial peptides, defensins), epithelial remodeling (ECM degradation, MMPs), intracellular signaling (MAPK cascades), and proteostasis pathways aligns with established antidepressant mechanisms involving inflammation resolution, structural plasticity, and cellular stress management\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e,\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e,\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e. Importantly, serial mediation analyses situate these proteomic changes within a microbiota\u0026ndash;metabolite\u0026ndash;host framework, identifying bile acids and SCFAs as key intermediates through which microbial alterations may modulate host protein networks relevant to treatment response. For instance, SCFAs modulate immune activation, stress signaling, and neuroendocrine pathways through GPCR signaling and epigenetic regulation, providing a plausible route by which gut microbial shifts translate into systemic protein\u0026ndash;level adaptations during treatment\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e. These results indicate that shifts in gut microbial activity shape systemic host responses through metabolite\u0026ndash;mediated signaling. They support an integrative framework in which effective treatment requires coordinated adaptation across microbial, metabolic, and host molecular layers. Overall, the data suggest that antidepressant efficacy is associated with the gut microbiota\u0026rsquo;s capacity to engage bile acid\u0026ndash; and SCFA\u0026ndash;related host signaling pathways that recalibrate immune and stress\u0026ndash;response systems.\u003c/p\u003e \u003cp\u003eSeveral limitations should be noted. The modest sample size, while adequate for discovery\u0026ndash;phase multi\u0026ndash;omic research with rigorous cross\u0026ndash;validation, limits generalizability, as multi\u0026ndash;omics analyses often face challenges in reproducibility and external validation with small cohorts\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e. Although longitudinal cross\u0026ndash;lagged and mediation analyses strengthen inference, causality remains observational and precludes definitive causal inference. The heterogeneous antidepressant treatments, though reflecting real\u0026ndash;world practice, preclude drug\u0026ndash;specific conclusions. Independent validation in larger cohorts with standardized treatment regimens is essential. Diet was assessed only at baseline and not monitored during treatment, potentially confounding microbial changes. The observational design lacks placebo control, limiting separation of specific antidepressant effects from natural symptom trajectories or non\u0026ndash;specific factors. Fecal metabolites reflect luminal exposure, metagenomic functions are relative estimates, and plasma proteomics captures peripheral rather than brain\u0026ndash;specific or central host responses, though systemic inflammation and metabolic changes likely influence central processes\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWithin these constraints, this study shows that antidepressant response reflects coordinated reorganization across the gut\u0026ndash;microbiome\u0026ndash;host axis. Rather than global microbial change, effective treatment involves targeted taxonomic shifts linked to symptom improvement, preservation of functional redundancy, restoration of secondary bile acids and butyrate, and adaptation of host immune and signaling pathways. These findings highlight functional resilience, the ability to maintain metabolic flexibility and coordinated adaptation under treatment, as a key determinant of antidepressant efficacy. Although validation is needed, the results point toward microbiome\u0026ndash;informed precision approaches to improve depression treatment.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAcknowledgement\u003c/h2\u003e\n\u003cp\u003eWe thank National Science and Technology Council for supporting our research. We appreciate Taipei Municipal Wanfang Hospital, Taipei City Psychiatric Center, and Taipei Veterans General Hospital for assistance in sample collection. We gratefully acknowledge BIOTOOLS Co., Ltd. for their technical support in shotgun metagenomic sequencing and fecal metabolomics. We also truly appreciate the support from the Center of Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Novascope Diagnostics Inc., and Genomics BioSci \u0026amp; Tech. Co., Ltd. for their assistance with plasma proteomics profiling on Illumina platforms. We sincerely thank Chou Yuan\u0026ndash;Tung for the valuable assistance in subject recruitment for this study.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis study was funded mainly by National Science and Technology Council (NSTC 108\u0026ndash;2314\u0026ndash;B\u0026ndash;002\u0026ndash;136\u0026ndash;MY3; 110\u0026ndash;2314\u0026ndash;B\u0026ndash;002\u0026ndash;067\u0026ndash;MY3, MOST 111\u0026ndash;2314\u0026ndash;B\u0026ndash;038\u0026ndash;062\u0026ndash;MY2, NSTC 113\u0026ndash;2314\u0026ndash;B\u0026ndash;038\u0026ndash;091\u0026ndash;; 114\u0026ndash;2314\u0026ndash;B\u0026ndash;038\u0026ndash;105\u0026ndash;) and the National Taiwan University Career Development Project (109L7860). This study was also supported by Population Health Research Center from Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education in Taiwan (grant number NTU\u0026ndash;112L9004, NTU\u0026ndash;113L9004, NTU\u0026ndash;114L9004, NTU\u0026ndash;115L9004). The finding agents are not involved in the study design, sample collecting, experimental protocol, data analysis, article drafting, or choice to submit to a journal.\u003c/p\u003e\n\u003ch2\u003eConflicts of interest\u003c/h2\u003e\n\u003cp\u003eThe authors declare that there is no conflict of interest regarding the publication of this paper.\u003c/p\u003e\n\u003ch2\u003eData and analysis codes availability\u003c/h2\u003e\n\u003cp\u003eThe data cannot be provided due to the data usage rights mentioned in informed consent. The authors will supply codes in response to reasonable requests\u003c/p\u003e\n\u003ch2\u003eAuthor contributions\u003c/h2\u003e\n\u003cp\u003eSKK Lin designed the study, managed follow up workflow, biosample extraction to final strategies of analyses, and wrote the first draft of the manuscript. PH Kuo and CH Chen contributed\u0026nbsp;to study design, administrative support, funding acquisition, and patient recruitment, they also revised the final manuscript critically. SS Huang, ML Lu, and CH Chen helped with the subject recruitment. All authors contributed to and approved the final manuscript.\u003c/p\u003e\n\u003ch2\u003eEthics approval and consent to participate\u003c/h2\u003e\n\u003cp\u003eAll methods were performed in accordance with the relevant guidelines and regulations of Taipei Medical University Wan Fang Hospital, Taipei City Hospital Heping Songde Branch, and Taipei Veterans General Hospital. The study protocol was approved by the Institutional Review Board of Taipei Medical University Wan Fang Hospital (Approval No. N202207051), Taipei City Hospital Heping Songde Branch (Approval No. TCHIRB\u0026ndash;11105013), and Taipei Veterans General Hospital (Approval No. 2023\u0026ndash;05\u0026ndash;005A). Informed consent was obtained from all participants included in the study. Written informed consent for publication of identifiable images was not applicable, as no identifiable images were included in this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGBD 2019 Mental Disorders Collaborators. Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990\u0026ndash;2019: a systematic analysis for the Global Burden of Disease Study 2019. \u003cem\u003eLancet Psychiatry\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 137\u0026ndash;150 (2022).\u003c/li\u003e\n\u003cli\u003eThomas, L. \u003cem\u003eet al.\u003c/em\u003e Prevalence of treatment\u0026ndash;resistant depression in primary care: cross\u0026ndash;sectional data. \u003cem\u003eBr. J. Gen. Pract. J. R. Coll. Gen. Pract.\u003c/em\u003e \u003cstrong\u003e63\u003c/strong\u003e, e852\u0026ndash;858 (2013).\u003c/li\u003e\n\u003cli\u003eCryan, J. F. \u003cem\u003eet al.\u003c/em\u003e The Microbiota\u0026ndash;Gut\u0026ndash;Brain Axis. \u003cem\u003ePhysiol. Rev.\u003c/em\u003e \u003cstrong\u003e99\u003c/strong\u003e, 1877\u0026ndash;2013 (2019).\u003c/li\u003e\n\u003cli\u003eLi, J. \u003cem\u003eet al.\u003c/em\u003e An integrated catalog of reference genes in the human gut microbiome. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 834\u0026ndash;841 (2014).\u003c/li\u003e\n\u003cli\u003eClarke, G. \u003cem\u003eet al.\u003c/em\u003e The microbiome\u0026ndash;gut\u0026ndash;brain axis during early life regulates the hippocampal serotonergic system in a sex\u0026ndash;dependent manner. \u003cem\u003eMol. Psychiatry\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 666\u0026ndash;673 (2013).\u003c/li\u003e\n\u003cli\u003eChen, C., Wang, G.\u0026ndash;Q., Li, D. \u0026amp; Zhang, F. Microbiota\u0026ndash;gut\u0026ndash;brain axis in neurodegenerative diseases: molecular mechanisms and therapeutic targets. \u003cem\u003eMol. Biomed.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 64 (2025).\u003c/li\u003e\n\u003cli\u003eLin, S.\u0026ndash;K. K. \u003cem\u003eet al.\u003c/em\u003e Dysbiosis and depression: A study of gut microbiota alterations and functional pathways in antidepressant\u0026ndash;na\u0026iuml;ve mood disorder patients. \u003cem\u003eTransl. Psychiatry\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 290 (2025).\u003c/li\u003e\n\u003cli\u003eLiu, L. \u003cem\u003eet al.\u003c/em\u003e Gut microbiota and its metabolites in depression: from pathogenesis to treatment. \u003cem\u003eeBioMedicine\u003c/em\u003e \u003cstrong\u003e90\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eMacedo, D. \u003cem\u003eet al.\u003c/em\u003e Antidepressants, antimicrobials or both? Gut microbiota dysbiosis in depression and possible implications of the antimicrobial effects of antidepressant drugs for antidepressant effectiveness. \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e \u003cstrong\u003e208\u003c/strong\u003e, 22\u0026ndash;32 (2017).\u003c/li\u003e\n\u003cli\u003eGamboa, J. \u003cem\u003eet al.\u003c/em\u003e Impact of antidepressants on the composition of the gut microbiome: A systematic review and meta\u0026ndash;analysis of in vivo studies. \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e \u003cstrong\u003e369\u003c/strong\u003e, 819\u0026ndash;833 (2025).\u003c/li\u003e\n\u003cli\u003eLee, S. M. \u003cem\u003eet al.\u003c/em\u003e The intestinal microbiota as a predictor for antidepressant treatment outcome in geriatric depression: a prospective pilot study. \u003cem\u003eInt. Psychogeriatr.\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 33\u0026ndash;45 (2022).\u003c/li\u003e\n\u003cli\u003eBorgiani, G. \u003cem\u003eet al.\u003c/em\u003e The bidirectional interaction between antidepressants and the gut microbiota: are there implications for treatment response? \u003cem\u003eInt. Clin. Psychopharmacol.\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 3\u0026ndash;26 (2025).\u003c/li\u003e\n\u003cli\u003eGędek, A. \u003cem\u003eet al.\u003c/em\u003e Altered Cytokine Levels in the First Episode of Major Depression and in Antidepressant\u0026ndash;Na\u0026iuml;ve Patients: A Systematic Review and Meta\u0026ndash;Analysis. \u003cem\u003eInt. J. Mol. Sci.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 10362 (2025).\u003c/li\u003e\n\u003cli\u003eKouba, B. R. \u003cem\u003eet al.\u003c/em\u003e Role of Inflammatory Mechanisms in Major Depressive Disorder: From Etiology to Potential Pharmacological Targets. \u003cem\u003eCells\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eNava, R. G. \u003cem\u003eet al.\u003c/em\u003e Modulation of neuroimmune cytokine networks by antidepressants: implications in mood regulation. \u003cem\u003eTransl. Psychiatry\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 314 (2025).\u003c/li\u003e\n\u003cli\u003eHuang, C.\u0026ndash;T. \u003cem\u003eet al.\u003c/em\u003e Impact of Gut Microbiota Alterations on Mitochondrial Bioenergetics in Cortical Astrocytes and Sensorimotor Impairment in a Rat Model of Lipopolysaccharide\u0026ndash;Associated Encephalopathy. \u003cem\u003eShock Augusta Ga\u003c/em\u003e https://doi.org/10.1097/SHK.0000000000002637 (2025) doi:10.1097/SHK.0000000000002637.\u003c/li\u003e\n\u003cli\u003eO\u0026rsquo;Riordan, K. J. \u003cem\u003eet al.\u003c/em\u003e Short chain fatty acids: Microbial metabolites for gut\u0026ndash;brain axis signalling. \u003cem\u003eMol. Cell. Endocrinol.\u003c/em\u003e \u003cstrong\u003e546\u003c/strong\u003e, 111572 (2022).\u003c/li\u003e\n\u003cli\u003eKim, C. H. Complex regulatory effects of gut microbial short\u0026ndash;chain fatty acids on immune tolerance and autoimmunity. \u003cem\u003eCell. Mol. Immunol.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 341\u0026ndash;350 (2023).\u003c/li\u003e\n\u003cli\u003eJiang, Y. \u003cem\u003eet al.\u003c/em\u003e The role of gut microbiota and metabolomic pathways in modulating the efficacy of SSRIs for major depressive disorder. \u003cem\u003eTransl. Psychiatry\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 493 (2024).\u003c/li\u003e\n\u003cli\u003eRuohan, Z. \u003cem\u003eet al.\u003c/em\u003e Gut microbiota as a novel target for treating anxiety and depression: from mechanisms to multimodal interventions. \u003cem\u003eFront. Microbiol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, (2025).\u003c/li\u003e\n\u003cli\u003eDong, Z. \u003cem\u003eet al.\u003c/em\u003e Gut microbiome: A potential indicator for predicting treatment outcomes in major depressive disorder. \u003cem\u003eFront. Neurosci.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 813075 (2022).\u003c/li\u003e\n\u003cli\u003eGamboa, J. \u003cem\u003eet al.\u003c/em\u003e Impact of antidepressants on the composition of the gut microbiome: A systematic review and meta\u0026ndash;analysis of in vivo studies. \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e \u003cstrong\u003e369\u003c/strong\u003e, 819\u0026ndash;833 (2025).\u003c/li\u003e\n\u003cli\u003eKnight, R. \u003cem\u003eet al.\u003c/em\u003e Best practices for analysing microbiomes. \u003cem\u003eNat. Rev. Microbiol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 410\u0026ndash;422 (2018).\u003c/li\u003e\n\u003cli\u003eLloyd\u0026ndash;Price, J. \u003cem\u003eet al.\u003c/em\u003e Multi\u0026ndash;omics of the gut microbial ecosystem in inflammatory bowel diseases. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e569\u003c/strong\u003e, 655\u0026ndash;662 (2019).\u003c/li\u003e\n\u003cli\u003eMuller, E., Shiryan, I. \u0026amp; Borenstein, E. Multi\u0026ndash;omic integration of microbiome data for identifying disease\u0026ndash;associated modules. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 2621 (2024).\u003c/li\u003e\n\u003cli\u003eChetty, A. \u0026amp; Blekhman, R. Multi\u0026ndash;omic approaches for host\u0026ndash;microbiome data integration. \u003cem\u003eGut Microbes\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 2297860.\u003c/li\u003e\n\u003cli\u003eSabih Ur Rehman, S. \u003cem\u003eet al.\u003c/em\u003e Integrative systems biology approaches for analyzing microbiome dysbiosis and species interactions. \u003cem\u003eBrief. Bioinform.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, bbaf323 (2025).\u003c/li\u003e\n\u003cli\u003eJiang, Y. \u003cem\u003eet al.\u003c/em\u003e Multi\u0026ndash;omics analysis reveals mechanisms of FMT in Enhancing antidepressant effects of SSRIs. \u003cem\u003eBrain. Behav. Immun.\u003c/em\u003e \u003cstrong\u003e126\u003c/strong\u003e, 176\u0026ndash;188 (2025).\u003c/li\u003e\n\u003cli\u003eFerdous, T. \u003cem\u003eet al.\u003c/em\u003e The rise to power of the microbiome: power and sample size calculation for microbiome studies. \u003cem\u003eMucosal Immunol.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1060\u0026ndash;1070 (2022).\u003c/li\u003e\n\u003cli\u003eKelly, B. J. \u003cem\u003eet al.\u003c/em\u003e Power and sample\u0026ndash;size estimation for microbiome studies using pairwise distances and PERMANOVA. \u003cem\u003eBioinforma. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 2461\u0026ndash;2468 (2015).\u003c/li\u003e\n\u003cli\u003eHasin, Y., Seldin, M. \u0026amp; Lusis, A. Multi\u0026ndash;omics approaches to disease. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 83 (2017).\u003c/li\u003e\n\u003cli\u003eAmerican Psychiatric Association. \u003cem\u003eDiagnostic and Statistical Manual of Mental Disorders\u003c/em\u003e. (American Psychiatric Association, 2013). doi:10.1176/appi.books.9780890425596.\u003c/li\u003e\n\u003cli\u003ePalleja, A. \u003cem\u003eet al.\u003c/em\u003e Recovery of gut microbiota of healthy adults following antibiotic exposure. \u003cem\u003eNat. Microbiol.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 1255\u0026ndash;1265 (2018).\u003c/li\u003e\n\u003cli\u003eChou, Y.\u0026ndash;C. \u003cem\u003eet al.\u003c/em\u003e Dietary intake of vitamin B(6) and risk of breast cancer in Taiwanese women. \u003cem\u003eJ. Epidemiol.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 329\u0026ndash;336 (2011).\u003c/li\u003e\n\u003cli\u003eHamilton, M. A rating scale for depression. \u003cem\u003eJ. Neurol. Neurosurg. Psychiatry\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 56\u0026ndash;61 (1960).\u003c/li\u003e\n\u003cli\u003eLin, C.\u0026ndash;H., Park, C. \u0026amp; McIntyre, R. S. Early improvement in HAMD\u0026ndash;17 and HAMD\u0026ndash;7 scores predict response and remission in depressed patients treated with fluoxetine or electroconvulsive therapy. \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e \u003cstrong\u003e253\u003c/strong\u003e, 154\u0026ndash;161 (2019).\u003c/li\u003e\n\u003cli\u003eMcIntyre, R. S. \u003cem\u003eet al.\u003c/em\u003e Measuring the severity of depression and remission in primary care: validation of the HAMD\u0026ndash;7 scale. \u003cem\u003eCMAJ Can. Med. Assoc. J.\u003c/em\u003e \u003cstrong\u003e173\u003c/strong\u003e, 1327\u0026ndash;1334 (2005).\u003c/li\u003e\n\u003cli\u003eBlanco\u0026ndash;M\u0026iacute;guez, A. \u003cem\u003eet al.\u003c/em\u003e Extending and improving metagenomic taxonomic profiling with uncharacterized species using MetaPhlAn 4. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 1633\u0026ndash;1644 (2023).\u003c/li\u003e\n\u003cli\u003eBeghini, F. \u003cem\u003eet al.\u003c/em\u003e Integrating taxonomic, functional, and strain\u0026ndash;level profiling of diverse microbial communities with bioBakery 3. \u003cem\u003eeLife\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e65088.\u003c/li\u003e\n\u003cli\u003eCaspi, R. \u003cem\u003eet al.\u003c/em\u003e The MetaCyc database of metabolic pathways and enzymes \u0026ndash; a 2019 update. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, D445\u0026ndash;D453 (2020).\u003c/li\u003e\n\u003cli\u003eKaul, A., Mandal, S., Davidov, O. \u0026amp; Peddada, S. D. Analysis of Microbiome Data in the Presence of Excess Zeros. \u003cem\u003eFront. Microbiol.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 2114 (2017).\u003c/li\u003e\n\u003cli\u003eIllumina Protein Prep | Detect and quantify proteins with NGS. https://sapac.illumina.com/products/by\u0026ndash;type/sequencing\u0026ndash;kits/library\u0026ndash;prep\u0026ndash;kits/protein\u0026ndash;prep.html.\u003c/li\u003e\n\u003cli\u003eNearing, J. T. \u003cem\u003eet al.\u003c/em\u003e Microbiome differential abundance methods produce different results across 38 datasets. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 342 (2022).\u003c/li\u003e\n\u003cli\u003eSun, J. \u0026amp; Xia, Y. Pretreating and normalizing metabolomics data for statistical analysis. \u003cem\u003eGenes Dis.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 100979 (2024).\u003c/li\u003e\n\u003cli\u003eKraemer, S. \u003cem\u003eet al.\u003c/em\u003e Crossing the Halfway Point: Aptamer\u0026ndash;Based, Highly Multiplexed Assay for the Assessment of the Proteome. \u003cem\u003eJ. Proteome Res.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 4771\u0026ndash;4788 (2024).\u003c/li\u003e\n\u003cli\u003eCroft, D. \u003cem\u003eet al.\u003c/em\u003e Reactome: a database of reactions, pathways and biological processes. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, D691\u0026ndash;D697 (2011).\u003c/li\u003e\n\u003cli\u003eEliazar, I. \u0026amp; Sokolov, I. M. Maximization of statistical heterogeneity: From Shannon\u0026rsquo;s entropy to Gini\u0026rsquo;s index. \u003cem\u003ePhys. Stat. Mech. Its Appl.\u003c/em\u003e \u003cstrong\u003e389\u003c/strong\u003e, 3023\u0026ndash;3038 (2010).\u003c/li\u003e\n\u003cli\u003eRitchie, M. E. \u003cem\u003eet al.\u003c/em\u003e limma powers differential expression analyses for RNA\u0026ndash;sequencing and microarray studies. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, e47 (2015).\u003c/li\u003e\n\u003cli\u003eSingh, A. \u003cem\u003eet al.\u003c/em\u003e DIABLO: an integrative approach for identifying key molecular drivers from multi\u0026ndash;omics assays. \u003cem\u003eBioinforma. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 3055\u0026ndash;3062 (2019).\u003c/li\u003e\n\u003cli\u003eBenjamini, Y. \u0026amp; Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. \u003cem\u003eJ. R. Stat. Soc. Ser. B Methodol.\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 289\u0026ndash;300 (1995).\u003c/li\u003e\n\u003cli\u003eBrown, L. C., Bobo, W. V., Gall, C. A., M\u0026uuml;ller, D. J. \u0026amp; Bousman, C. A. Pharmacomicrobiomics of Antidepressants in Depression: A Systematic Review. \u003cem\u003eJ. Pers. Med.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1086 (2023).\u003c/li\u003e\n\u003cli\u003eGamboa, J. \u003cem\u003eet al.\u003c/em\u003e Impact of antidepressants on the composition of the gut microbiome: A systematic review and meta\u0026ndash;analysis of in vivo studies. \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e \u003cstrong\u003e369\u003c/strong\u003e, 819\u0026ndash;833 (2025).\u003c/li\u003e\n\u003cli\u003eLiu, D. \u003cem\u003eet al.\u003c/em\u003e Anaerostipes hadrus, a butyrate\u0026ndash;producing bacterium capable of metabolizing 5\u0026ndash;fluorouracil. \u003cem\u003emSphere\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e0081623 (2024).\u003c/li\u003e\n\u003cli\u003eWu, F. \u003cem\u003eet al.\u003c/em\u003e Phascolarctobacterium faecium abundant colonization in human gastrointestinal tract. \u003cem\u003eExp. Ther. Med.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 3122\u0026ndash;3126 (2017).\u003c/li\u003e\n\u003cli\u003eKang, J. D. \u003cem\u003eet al.\u003c/em\u003e Bile Acid 7\u0026alpha;\u0026ndash;Dehydroxylating Gut Bacteria Secrete Antibiotics that Inhibit Clostridium difficile: Role of Secondary Bile Acids. \u003cem\u003eCell Chem. Biol.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 27\u0026ndash;34.e4 (2019).\u003c/li\u003e\n\u003cli\u003eDaniel, S. L. \u0026amp; Ridlon, J. M. Clostridium scindens: history and current outlook for a keystone species in the mammalian gut involved in bile acid and steroid metabolism. \u003cem\u003eFEMS Microbiol. Rev.\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, fuaf016 (2025).\u003c/li\u003e\n\u003cli\u003eKabeerdoss, J., Sankaran, V., Pugazhendhi, S. \u0026amp; Ramakrishna, B. S. Clostridium leptum group bacteria abundance and diversity in the fecal microbiota of patients with inflammatory bowel disease: a case\u0026ndash;control study in India. \u003cem\u003eBMC Gastroenterol.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 20 (2013).\u003c/li\u003e\n\u003cli\u003eSingh, V. \u003cem\u003eet al.\u003c/em\u003e Butyrate producers, \u0026lsquo;The Sentinel of Gut\u0026rsquo;: Their intestinal significance with and beyond butyrate, and prospective use as microbial therapeutics. \u003cem\u003eFront. Microbiol.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1103836 (2022).\u003c/li\u003e\n\u003cli\u003eFacchin, S. \u003cem\u003eet al.\u003c/em\u003e Short\u0026ndash;Chain Fatty Acids and Human Health: From Metabolic Pathways to Current Therapeutic Implications. \u003cem\u003eLife\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 559 (2024).\u003c/li\u003e\n\u003cli\u003eMarkowiak\u0026ndash;Kopeć, P. \u0026amp; Śliżewska, K. The Effect of Probiotics on the Production of Short\u0026ndash;Chain Fatty Acids by Human Intestinal Microbiome. \u003cem\u003eNutrients\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 1107 (2020).\u003c/li\u003e\n\u003cli\u003eGao, K., Mu, C., Farzi, A. \u0026amp; Zhu, W. Tryptophan Metabolism: A Link Between the Gut Microbiota and Brain. \u003cem\u003eAdv. Nutr.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 709\u0026ndash;723 (2020).\u003c/li\u003e\n\u003cli\u003eParker, B. J., Wearsch, P. A., Veloo, A. C. M. \u0026amp; Rodriguez\u0026ndash;Palacios, A. The Genus Alistipes: Gut Bacteria With Emerging Implications to Inflammation, Cancer, and Mental Health. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eLouis, P. \u0026amp; Flint, H. J. Formation of propionate and butyrate by the human colonic microbiota. \u003cem\u003eEnviron. Microbiol.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 29\u0026ndash;41 (2017).\u003c/li\u003e\n\u003cli\u003eGalland, T., P\u0026eacute;rez Carmona, C., G\u0026ouml;tzenberger, L., Valencia, E. \u0026amp; de Bello, F. Are redundancy indices redundant? An evaluation based on parameterized simulations. \u003cem\u003eEcol. Indic.\u003c/em\u003e \u003cstrong\u003e116\u003c/strong\u003e, 106488 (2020).\u003c/li\u003e\n\u003cli\u003eLouca, S. \u003cem\u003eet al.\u003c/em\u003e Function and functional redundancy in microbial systems. \u003cem\u003eNat. Ecol. Evol.\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 936\u0026ndash;943 (2018).\u003c/li\u003e\n\u003cli\u003eLi, T. \u0026amp; Chiang, J. Y. L. Bile acid signaling in metabolic disease and drug therapy. \u003cem\u003ePharmacol. Rev.\u003c/em\u003e \u003cstrong\u003e66\u003c/strong\u003e, 948\u0026ndash;983 (2014).\u003c/li\u003e\n\u003cli\u003eWahlstr\u0026ouml;m, A., Sayin, S. I., Marschall, H.\u0026ndash;U. \u0026amp; B\u0026auml;ckhed, F. Intestinal Crosstalk between Bile Acids and Microbiota and Its Impact on Host Metabolism. \u003cem\u003eCell Metab.\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 41\u0026ndash;50 (2016).\u003c/li\u003e\n\u003cli\u003eDalile, B., Van Oudenhove, L., Vervliet, B. \u0026amp; Verbeke, K. The role of short\u0026ndash;chain fatty acids in microbiota\u0026ndash;gut\u0026ndash;brain communication. \u003cem\u003eNat. Rev. Gastroenterol. Hepatol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 461\u0026ndash;478 (2019).\u003c/li\u003e\n\u003cli\u003eDuman, R. S., Aghajanian, G. K., Sanacora, G. \u0026amp; Krystal, J. H. Synaptic plasticity and depression: new insights from stress and rapid\u0026ndash;acting antidepressants. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 238\u0026ndash;249 (2016).\u003c/li\u003e\n\u003cli\u003eMiller, A. H. \u0026amp; Raison, C. L. The role of inflammation in depression: from evolutionary imperative to modern treatment target. \u003cem\u003eNat. Rev. Immunol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 22\u0026ndash;34 (2016).\u003c/li\u003e\n\u003cli\u003eHodes, G. E., Kana, V., Menard, C., Merad, M. \u0026amp; Russo, S. J. Neuroimmune mechanisms of depression. \u003cem\u003eNat. Neurosci.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1386\u0026ndash;1393 (2015).\u003c/li\u003e\n\u003cli\u003eAutry, A. E. \u003cem\u003eet al.\u003c/em\u003e NMDA receptor blockade at rest triggers rapid behavioural antidepressant responses. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e475\u003c/strong\u003e, 91\u0026ndash;95 (2011).\u003c/li\u003e\n\u003cli\u003eParada Venegas, D. \u003cem\u003eet al.\u003c/em\u003e Short Chain Fatty Acids (SCFAs)\u0026ndash;Mediated Gut Epithelial and Immune Regulation and Its Relevance for Inflammatory Bowel Diseases. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 277 (2019).\u003c/li\u003e\n\u003cli\u003eSherwani, M. K. \u003cem\u003eet al.\u003c/em\u003e Multi\u0026ndash;omics time\u0026ndash;series analysis in microbiome research: a systematic review. \u003cem\u003eBrief. Bioinform.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, bbaf502 (2025).\u003c/li\u003e\n\u003cli\u003eLoh, J. S. \u003cem\u003eet al.\u003c/em\u003e Microbiota\u0026ndash;gut\u0026ndash;brain axis and its therapeutic applications in neurodegenerative diseases. \u003cem\u003eSignal Transduct. Target. Ther.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 37 (2024).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1 and 2 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"molecular-psychiatry","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"mp","sideBox":"Learn more about [Molecular Psychiatry](http://www.nature.com/mp/)","snPcode":"41380","submissionUrl":"https://mts-mp.nature.com/cgi-bin/main.plex","title":"Molecular Psychiatry","twitterHandle":"@molpsychiatry","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Major depressive disorder, antidepressant treatment response, shotgun metagenome, gut microbiome, metabolimics, plasma proteomics, longitudinal multiomics","lastPublishedDoi":"10.21203/rs.3.rs-8836038/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8836038/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThirty to fifty percent of patients with major depressive disorder (MDD) show inadequate response to antidepressants, yet biological predictors remain elusive. We conducted a longitudinal multi\u0026ndash;omic study of 28 antidepressant\u0026ndash;na\u0026iuml;ve patients with MDD sampled at baseline, one month, and two months, integrating shotgun metagenomics, microbial functional pathways, targeted fecal metabolomics, and plasma proteomics to identify response\u0026ndash;predictive signatures. While global microbial diversity remained stable, treatment responders exhibited distinct microbial trajectories characterized by early reductions in inflammation\u0026ndash;associated taxa and restoration of butyrate\u0026ndash;producing bacteria, and baseline microbial composition predicted two\u0026ndash;month response with 96.5% cross\u0026ndash;validated accuracy. Cross\u0026ndash;lagged panel analyses revealed bidirectional coupling between symptom improvement and microbial dynamics. Functionally, responders maintained pathway\u0026ndash;level redundancy with distributed species contributions to core biosynthetic pathways, whereas non\u0026ndash;responders showed progressive functional narrowing and taxonomic dominance. Metabolomically, responders exhibited restoration of secondary bile acids and butyrate, while non\u0026ndash;responders showed depletion patterns and divergent hexanoic acid trajectories. Plasma proteomics identified treatment\u0026ndash;responsive proteins with divergent longitudinal patterns, enriched in immune defense, epithelial remodeling, and intracellular signaling pathways. Serial mediation analyses demonstrated that microbial changes influenced treatment response through bile acid and short\u0026ndash;chain fatty acid intermediates that modulated host proteomic networks. These findings indicate that antidepressant efficacy is associated with coordinated functional resilience across the gut\u0026ndash;microbiome\u0026ndash;host axis rather than compositional restructuring, suggesting targets for microbiome\u0026ndash;informed precision psychiatry.\u003c/p\u003e","manuscriptTitle":"Multi–omic Longitudinal Profiling Reveals Coordinated Gut Microbiota, Metabolomics, and Host Proteomics Signatures Predictive of Antidepressant Response in Drug–Naïve Depression","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-25 08:56:23","doi":"10.21203/rs.3.rs-8836038/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"This content is not available.","date":"2026-04-10T20:06:54+00:00","index":2,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2026-03-13T04:44:11+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2026-02-24T09:23:56+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewersInvited","content":"","date":"2026-02-19T14:58:36+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-16T18:06:13+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-16T17:52:51+00:00","index":"","fulltext":""},{"type":"submitted","content":"Molecular Psychiatry","date":"2026-02-14T05:28:59+00:00","index":"","fulltext":""},{"type":"checksFailed","content":"","date":"2026-02-12T16:17:40+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"molecular-psychiatry","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"mp","sideBox":"Learn more about [Molecular Psychiatry](http://www.nature.com/mp/)","snPcode":"41380","submissionUrl":"https://mts-mp.nature.com/cgi-bin/main.plex","title":"Molecular Psychiatry","twitterHandle":"@molpsychiatry","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"40a5b61f-ce84-4162-bb7f-eccc713ea9a1","owner":[],"postedDate":"February 25th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":63208233,"name":"Health sciences/Biomarkers/Predictive markers"},{"id":63208234,"name":"Biological sciences/Physiology"},{"id":63208235,"name":"Biological sciences/Biochemistry"}],"tags":[],"updatedAt":"2026-02-25T08:56:23+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-25 08:56:23","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8836038","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8836038","identity":"rs-8836038","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}