Abstract

Objective

   This study aims to explore the relationship between gut microbiota and
   fecal metabolomic profiles in patients with systemic lupus
   erythematosus (SLE), with and without lupus nephritis (LN), in order to
   identify potentially relevant biomarkers and better understand their
   association with disease progression.

Methods

   Fecal samples from 15 healthy controls (HC) and 36 SLE patients (18
   SLE-nonLN and 18 SLE-LN) were analyzed using 16S rRNA gene sequencing
   and untargeted metabolomics. Differential microbial taxa and
   metabolites were identified using Linear Discriminant Analysis Effect
   Size (LEfSe) and Orthogonal Partial Least Squares Discriminant Analysis
   (OPLS-DA). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and
   Receiver Operating Characteristic (ROC) curve analyses were used to
   assess the potential clinical relevance of selected metabolites.

Results

   Beta diversity analysis demonstrated distinct microbial clustering
   between groups (p < 0.05). SLE-LN samples showed an increased relative
   abundance of Proteobacteria and decreased Firmicutes compared to
   SLE-nonLN. Metabolomic profiling identified multiple differentially
   abundant metabolites, with notable enrichment in primary bile acid
   biosynthesis pathways (e.g., Glycocholic acid, AUC = 0.951). In the
   SLE-nonLN group, increased Glycoursodeoxycholic acid levels
   (AUC = 0.922) were observed in pathways related to taurine and
   hypotaurine metabolism. Correlation analysis indicated a negative
   association between Escherichia-Shigella and bile acid levels
   (p < 0.01).

Conclusion

   This integrative analysis suggests that patients with SLE and LN harbor
   distinct gut microbiota and metabolomic profiles. The identified
   microbial taxa and metabolites may have potential as non-invasive
   biomarkers and could contribute to a better understanding of SLE
   pathogenesis and progression.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12866-025-03995-5.

   Keywords: SLE, LN, Gut microbiota, Metabolomics, Biomarkers, Fecal
   samples, 16S rRNA sequencing, Metabolic pathways

Introduction

   Systemic lupus erythematosus (SLE) is a chronic autoimmune disorder
   characterized by the immune system’s aberrant attack on healthy
   tissues. Such dysregulation leads to systemic inflammation and
   multiorgan damage [[36]1]. SLE manifests with heterogeneous clinical
   presentations and variable disease severity, affecting multiple organs.
   The global prevalence of SLE ranges from 30 to 50 cases per 100,000
   individuals [[37]2]. In China, the reported prevalence is 6.17 per
   100,000 males and 67.78 per 100,000 females [[38]3]. Lupus nephritis
   (LN), a severe renal complication of SLE, is characterized by
   immune-mediated kidney damage with diverse pathological types and
   significant clinical manifestations, making it one of the most serious
   forms of SLE [[39]4, [40]5]. Despite extensive research, the underlying
   pathogenesis of LN remains incompletely understood. A primary goal in
   SLE management is preventing irreversible organ damage, which requires
   identifying key molecular contributors to disease progression [[41]6].
   Accurate diagnosis, timely intervention, and early relapse management
   are crucial for effective LN treatment. Although renal biopsy is the
   gold standard for LN diagnosis, its invasiveness limits its utility for
   continuous disease monitoring, underscoring the need for reliable
   non-invasive biomarkers [[42]7, [43]8]. Current routine biomarkers,
   such as serum creatinine and complement component C3b, have limited
   utility in assessing LN disease activity or facilitating real-time
   diagnosis [[44]9].

   Emerging evidence implicates the gut microbiota as a critical modulator
   of autoimmune processes, including SLE [[45]10, [46]11]. Gut microbiota
   metabolize dietary components into bioactive metabolites that modulate
   systemic immune responses [[47]12]. Microbial metabolites, such as
   short-chain fatty acids and bile acids, are key mediators of
   host-microbiota interactions [[48]13, [49]14]. For example, Zhang et
   al. reported that fecal samples from SLE patients exhibited
   significantly elevated metabolic activities, including enhanced amino
   acid biosynthesis, vitamin B1 metabolism, nitrogen cycling, tryptophan
   degradation, and cyanoamino acid metabolism, compared to healthy
   controls [[50]15]. However, single-omics approaches (e.g., metagenomics
   or metabolomics alone) fail to capture the complexity of disease
   mechanisms. Integrative analysis of gut microbiota and host metabolome
   dynamics provides a holistic understanding of microbial and metabolic
   interactions in pathogenesis [[51]16]. Despite progress in SLE
   metabolomic profiling, studies exploring microbiota-metabolome
   interplay, particularly in differentiating SLE with nephritis (SLE-LN)
   from SLE without nephritis (SLE-nonLN), remain limited [[52]17].

   In this study, we applied 16S rRNA gene sequencing and liquid
   chromatography-tandem mass spectrometry (LC-MS/MS)-based untargeted
   metabolomics to analyze fecal samples from SLE patients (SLE-LN and
   SLE-nonLN) and healthy controls (HC). Our objective was to identify
   potential biomarkers and explore gut microbiota-metabolome interactions
   to provide novel insights into the mechanisms underlying SLE
   pathogenesis and progression.

Materials and methods

Sample collection

   We recruited 36 patients diagnosed with SLE from Nanjing Drum Tower
   Hospital, affiliated with Nanjing University Medical School. All
   patients were newly admitted and were not receiving immunosuppressive
   therapy at the time of sample collection. Diagnosis was based on the
   1997 revised classification criteria of the American College of
   Rheumatology (ACR) and further confirmed through clinical evaluation,
   including serological markers and renal biopsy where applicable. The
   cohort consisted of 18 SLE-LN and 18 SLE-nonLN. Clinical
   characteristics, including laboratory test results and medical history,
   were obtained from the hospital’s electronic medical records system.
   Additionally, 15 age-, gender-, and BMI-matched HC were recruited from
   the hospital’s health examination center. These individuals had no
   prior history of autoimmune diseases, infections, metabolic disorders,
   or malignancies. All participants provided written informed consent,
   and the study was approved by the Ethics Committee of Nanjing Drum
   Tower Hospital. Fecal samples were collected using sterile containers
   upon admission and immediately processed to maintain sample integrity.
   Each sample was divided into two aliquots and immediately stored at
   -80 °C for subsequent 16S rRNA sequencing and untargeted metabolomics
   analysis. The overall study design is illustrated in Fig. [53]1.

Fig. 1.

   [54]Fig. 1
   [55]Open in a new tab

   Grouping Design and Analysis Flowchart

DNA extraction and 16S rRNA sequencing

   Total genomic DNA was extracted using the QIAamp DNA Stool Mini Kit
   (QIAGEN) with modifications, including an extended 10-minute
   bead-beating step to enhance the lysis of Gram-positive bacteria. DNA
   purity (A260/A280 ratio 1.8-2.0) and integrity were verified via
   NanoDrop 2000 spectrophotometry and 1% agarose gel electrophoresis. The
   V3-V4 region of the 16 S rRNA gene was amplified using primers
   341 F/806R under the following conditions: 98 °C for 1 min; 30 cycles
   of 98 °C for 10 s, 55 °C for 30 s, 72 °C for 30 s; final extension at
   72 °C for 5 min. PCR products were purified (AxyPrep DNA Gel Extraction
   Kit) and quantified (Qubit dsDNA Assay Kit). Equimolar pooled libraries
   were sequenced on an Illumina MiSeq PE300 platform (2 × 300 bp) in a
   single run to minimize batch effects.

Sequencing data processing and microbial diversity analysis

   Raw sequencing reads were preprocessed using Cutadapt (v4.0) to trim
   adapters and remove low-quality bases. Quality filtering was performed
   with FASTP (v0.23.4), retaining reads with Phred scores ≥ 20 and
   lengths ≥ 200 bp. Paired-end reads were merged using FLASH (v1.2.11;
   min overlap = 20 bp, max mismatch = 0.1). Further filtering steps were
   conducted to remove ambiguous bases (N), sequences with homopolymer
   runs (> 8 bp), and chimeric reads using USEARCH (v11.0). Operational
   taxonomic units (OTUs) were de novo clustered at 97% similarity using
   VSEARCH (v2.21.1) and taxonomically assigned via the QIIME2 Naïve
   Bayesian Classifier (v2023.2) against the SILVA 142 database
   (confidence threshold = 80%). Alpha diversity (Chao, Shannon, Simpson)
   was calculated after rarefaction to 10,000 reads/sample. Beta diversity
   was assessed using weighted/unweighted UniFrac distances and visualized
   via principal coordinate analysis (PCoA) and non-metric
   multidimensional scaling (NMDS). Linear discriminant analysis Effect
   Size (LEfSe) analysis was performed using Python (v3.9.7) to identify
   differentially abundant taxa, applying significance thresholds of 0.05
   for both the Kruskal-Wallis and Wilcoxon tests, and an LDA score
   threshold of 4.

Metabolite extraction and LC-MS/MS analysis

   Fecal metabolites were extracted by homogenizing 20 mg of sample with
   400 µL methanol: water (7:3, v/v) on ice. After sonication (10 min),
   vortexing (1 min), and centrifugation (12,000 × g, 10 min, 4 °C), 200
   µL of supernatant was analyzed via LC-MS/MS (Waters ACQUITY UPLC HSS T3
   C18 column, 1.8 μm, 2.1 × 100 mm). The mobile phase consisted of 0.1%
   formic acid in water (solvent A) and acetonitrile (solvent B), with a
   gradient elution (0.4 mL/min): 0–11 min, 5–90% B; 12–14 min, 5% B. Raw
   data were converted to mzML format using ProteoWizard version
   3.0.23136. Peak detection and retention time alignment were performed
   using XCMS (v4.7). Metabolites were identified through our in-house
   database and the Kyoto Encyclopedia of Genes and Genomes (KEGG) online
   database.

Statistical analysis

   Constrained Principal Coordinate Analysis (CPCoA) was performed using
   the prcomp function in R (v4.2.3). Hierarchical clustering analysis was
   conducted with the ComplexHeatmap package, and results were visualized
   as heatmaps with dendrograms. Normalized signal intensities of
   metabolites were displayed as color spectra following unit variance
   scaling. Differential metabolites were defined by VIP ≥ 1, |Log2FC|≥1,
   and p < 0.05. Data were log2-transformed and mean-centered prior to
   orthogonal partial least squares discriminant analysis (OPLS-DA).
   Functional and pathway analysis was conducted using the KEGG database,
   with pathways considered significantly enriched at p < 0.05.

Results

Altered microbiota composition among groups

   The DNA from fecal samples of 15 HC, 18 SLE-nonLN, and 18 SLE-LN
   patients was examined using 16 S rRNA gene sequencing. Species
   accumulation curves (Fig. [56]S1) and rarefaction curves (Fig. [57]S2)
   indicated sufficient sequencing depth and coverage. To assess bacterial
   diversity differences across the three groups, both within-sample
   (alpha) and between-sample (beta) diversity metrics were calculated. No
   significant differences in ACE, Chao, Shannon, and Simpson indices were
   observed among the HC, SLE-nonLN, and SLE-LN groups (Fig. [58]2A-D),
   suggesting a lack of significant variations in microbial richness or
   diversity. In contrast, beta diversity analysis using CPCoA and NMDS
   demonstrated distinct clustering patterns between SLE groups (nonLN,
   LN) and HC (Fig. [59]S3A-B). These findings suggest that although alpha
   diversity differences were not significant, SLE influenced gut
   microbiota composition.

Fig. 2.

   [60]Fig. 2
   [61]Open in a new tab

   Alpha diversity analysis across HC, SLE-nonLN, and SLE-LN groups. The
   ACE (A), Chao (B), Shannon (C), and Simpson (D) indices represent
   different measures of bacterial diversity among the three groups

Altered microbial composition associated with SLE

   At the phylum level, Firmicutes was the predominant phylum in all
   groups, with a significantly higher relative abundance in SLE-nonLN
   (59.08%) compared to HC (49.35%) and SLE-LN (39.50%, Fig. [62]3A).
   Conversely, Proteobacteria abundance was markedly elevated in SLE-LN
   (28.02%) versus SLE-nonLN (12.93%) and HC (23.18%). At the genus level,
   the relative abundance of Faecalibacterium was highest in the HC group
   (10.31%) and progressively declined in the SLE-nonLN and SLE-LN groups
   (4.06% and 1.90%, respectively). Conversely, Bacteroides abundance was
   lower in the HC group (9.57%) and increased in the SLE-nonLN and SLE-LN
   groups (11.26% and 14.73%, respectively). Compared to the HC group, the
   genera Blautia decreased, whereas Streptococcus, Enterococcus,
   Akkermansia, and Lactobacillus were enriched in both the SLE-nonLN and
   SLE-LN groups (Fig. [63]3B).

Fig. 3.

   [64]Fig. 3
   [65]Open in a new tab

   Altered gut microbiota composition among HC, SLE-nonLN, and SLE-LN
   groups. (A) Relative abundances at the phylum level. (B) Relative
   abundances at the genus level

LEfSe analysis among groups

   LEfSe identified 27 differentially abundant taxa across groups
   (LDA > 4, p < 0.05), with 6 taxa enriched in the SLE-nonLN group, 11 in
   the SLE-LN group, and 10 in the HC group. SLE-nonLN was characterized
   by Bacilli and Lactobacillales, while SLE-LN showed enrichment of
   Enterobacteriaceae, Enterobacterales, Proteobacteria,
   Gammaproteobacteria, Escherichia_Shigella, and Escherichia_coli (LDA
   score > 4.7). HC exhibited higher Clostridia abundance (Fig. [66]4A-B).
   Pairwise comparisons confirmed Enterobacter as the most distinct genus
   among groups (Kruskal-Wallis p < 0.001; Fig. [67]S4), with
   Escherichia-Shigella significantly differentiating SLE-LN from
   SLE-nonLN (p = 0.0208, Fig. [68]S5).

Fig. 4.

   [69]Fig. 4
   [70]Open in a new tab

   Microbial community analysis using LEfSe. (A) Histogram of LDA scores
   showing differentially abundant microbial taxa (LDA score > 4). (B)
   Cladogram depicting specific differential taxa

Changes in metabolome and key metabolites

   Gut microbial metabolites play a crucial role in modulating host
   physiological functions. To characterize metabolic alterations among
   HC, SLE-nonLN, and SLE-LN groups, metabolite abundance in fecal samples
   was analyzed using LC-MS. Untargeted metabolomics revealed distinct
   clustering among HC, SLE-nonLN, and SLE-LN groups via OPLS-DA, with
   clear separation between groups (Fig. [71]5A). Permutation tests
   validated the model’s robustness, with the Y-intercept of the
   regression line for Q² values below zero (R^2Y = 0.995, Q^2 = 0.788;
   Fig. [72]5B). A total of 177 metabolites were differentially abundant
   in HC vs. SLE-LN, 159 in HC vs. SLE-nonLN, and 94 in SLE-nonLN vs.
   SLE-LN (VIP ≥ 1, p < 0.05, |Log2FC| ≥1). Differential metabolites were
   identified based on the OPLS-DA model and visualized using volcano
   plots (Fig. [73]5C).

Fig. 5.

   [74]Fig. 5
   [75]Open in a new tab

   Metabolic profile analysis. (A) OPLS-DA score plot demonstrating
   significant differences in metabolic profiles among groups. (B)
   Permutation test for model validation. The x-axis represents the
   correlation coefficient, and the y-axis represents the predictive
   performance of the model. (C) Volcano plot of differential metabolites.
   The x-axis denotes the groups, while the y-axis represents the log₂
   fold change of metabolites. Metabolites upregulated in each group are
   shown in red

Metabolic pathways analysis

   Pearson correlation analysis of the top 50 differential metabolites
   (ranked by VIP scores) revealed distinct co-regulation patterns among
   HC, SLE-nonLN, and SLE-LN groups (Fig. [76]S6). KEGG pathway enrichment
   analysis identified taurine and hypotaurine metabolism as the most
   significantly altered pathway in SLE-nonLN vs. HC, followed by primary
   bile acid biosynthesis and histidine metabolism (Fig. [77]6A). In
   SLE-nonLN vs. SLE-LN comparisons, key pathways included primary bile
   acid biosynthesis, thiamine metabolism, and sulfur metabolism
   (Fig. [78]6B). After excluding metabolites of dietary or pharmaceutical
   origin (e.g., Hydroxychloroquine, Myristoleic acid, Polygodial,
   Furosemide), seven metabolites were consistently altered across all
   groups (Table [79]1). Heatmap analysis highlighted three
   metabolites—Glycocholic acid, Glycochenodeoxycholic acid, and
   5,8,11-Eicosatrienoic acid—that were dysregulated in both HC vs.
   SLE-nonLN and SLE-nonLN vs. SLE-LN comparisons (Fig. [80]S7).

Fig. 6.

   [81]Fig. 6
   [82]Open in a new tab

   KEGG pathway enrichment analysis. KEGG pathway enrichment analysis
   highlighting major metabolic pathways affected in (A) HC vs. SLE-nonLN
   and (B) SLE-nonLN vs. SLE-LN comparisons. The y-axis lists the enriched
   pathways, while the x-axis represents the enrichment factor (ratio of
   significantly altered metabolites to total metabolites in a given
   pathway). The bubble size corresponds to the number of enriched
   metabolites, whereas the color gradient denotes statistical
   significance, with darker shades indicating higher significance

Table 1.

   Metabolites with intergroup differences in fecal samples
   Compounds Class HC vs. SLE-nonLN SLE-nonLN vs. SLE-LN
   VIP FC p Trend VIP FC p Trend
   Glycocholic acid Bile acids 2.99 5.06 0.00 ↑ 3.10 0.37 0.00 ↓
   5,8,11-Eicosatrienoic acid Fatty Acyls 1.66 0.31 0.01 ↓ 1.80 3.11 0.00
   ↑
   Stachydrine Amino acid and Its metabolites 1.23 4.31 0.03 ↑ — — — —
   Glycine deoxycholic acid Organic acid And Its derivatives 1.83 11.02
   0.04 ↑ — — — —
   Glycochenodeoxycholic acid Bile acids 2.52 8.88 0.00 ↑ 2.56 0.28 0.01 ↓
   Methylsuccinic acid Organic acid And Its derivatives — — — — 1.06 0.35
   0.02 ↓
   Palmitoylcarnitine Alkaloids — — — — 1.28 2.78 0.03 ↑
   Ursocholic acid Bile acids — — — — 1.41 6.95 0.02 ↑
   Oxypurinol Heterocyclic compounds — — — — 1.54 0.15 0.04 ↓
   [83]Open in a new tab

   Notes: ↑ Represents upregulation of metabolite; ↓ represents
   downregulation of metabolite

   Abbreviations: VIP, variable importance in the projection; FC, fold
   change

Identification of metabolite biomarkers to distinguish SLE-LN from SLE-nonLN

   Receiver operating characteristic (ROC) was conducted to evaluate the
   discriminatory power of key metabolites in distinguishing SLE-LN from
   SLE-nonLN. Seven differential metabolites involved in major metabolic
   pathways were assessed, revealing that Glycocholic acid (AUC = 0.951),
   Glycochenodeoxycholic acid (AUC = 0.827), Oxypurinol (AUC = 0.769),
   Methylsuccinic acid (AUC = 0.685) and 5,8,11-Eicosatrienoic acid
   (AUC = 0.716) exhibited varying degrees of classification accuracy
   between SLE-nonLN and SLE-LN (Fig. [84]7A-E). Among them, Glycocholic
   acid and Glycochenodeoxycholic acid exhibited the strongest predictive
   capability. For SLE-nonLN vs. HC discrimination, Glycine deoxycholic
   acid (AUC = 0.844) and Glycochenodeoxycholic acid (AUC = 0.922)
   exhibited high accuracy, whereas Soyasaponin I (AUC = 0.689) showed
   limited utility (Fig. [85]S8).

Fig. 7.

   [86]Fig. 7
   [87]Open in a new tab

   ROC analysis of potential biomarkers distinguishing SLE-LN and
   SLE-nonLN. (A) Glycocholic acid, (B) Methylsuccinic acid, (C)
   Oxypurinol, (D) 5,8,11-Eicosatrienoic acid, (E) Glycochenodeoxycholic
   acid

Correlation between microbiota and metabolites

   Spearman’s correlation analysis between the top 16 microbial genera and
   key metabolites revealed significant associations. The correlation
   heatmap (Fig. [88]8A) highlighted key relationships:
   Escherichia-Shigella and Enterobacter abundances were negatively
   correlated with bile acids (Glycochenodeoxycholic acid and Glycocholic
   acid), but positively correlated with 5,8,11-eicosatrienoic acid.
   Additionally, Bacteroides, Faecalibacterium, and Parabacteroides were
   positively correlated with Methylsuccinic acid, indicating potential
   metabolic interactions. Notably, Escherichia-Shigella was inversely
   associated with Glycochenodeoxycholic acid, Glycocholic acid and
   Stachydrine in HC vs. SLE-nonLN comparisons, while Streptococcus
   correlated positively with Glycocholic acid (Fig. [89]8B). Furthermore,
   Soyasaponin I demonstrated a negative correlation with Subdoligranulum,
   Agathobacter, and Faecalibacterium, while showing a significant
   positive correlation with Erysipelatoclostridium (p < 0.01). These
   findings suggest that alterations in gut microbial composition
   significantly impact host metabolic profiles, particularly in the
   regulation of bile acid metabolism. Notably, key genera such as
   Escherichia-Shigella and Enterobacter appear to contribute
   substantially to bile acid synthesis, further underscoring their
   potential role in SLE pathogenesis.

Fig. 8.

   [90]Fig. 8
   [91]Open in a new tab

   Spearman correlation analysis between gut microbiota and differential
   metabolites. (A) Correlation analysis in SLE-nonLN and SLE-LN groups.
   (B) Correlation analysis in HC and SLE-nonLN groups. The x-axis
   represents differential metabolites, while the y-axis denotes bacterial
   genera identified at the 16 S rRNA gene level. Red indicates a positive
   correlation, whereas blue represents a negative correlation. Asterisks
   indicate statistical significance: *p < 0.05, **p < 0.01

Discussion

   SLE is a chronic autoimmune disease that primarily affects young women
   and leads to systemic inflammation. LN, a severe renal complication of
   SLE, significantly worsens patient prognosis [[92]18]. While our
   previous work identified tRNA-derived small noncoding RNAs (tsRNAs) as
   potential biomarkers for LN [[93]19, [94]20], the gut
   microbiota-metabolome axis remains underexplored in SLE progression. In
   the present study, we integrated 16S rRNA sequencing with untargeted
   metabolomics to explore gut microbial and metabolic alterations across
   different stages of SLE. Our findings reveal stage-specific microbiota
   and metabolite patterns, which may provide new insights into SLE
   pathophysiology and support the identification of non-invasive
   biomarkers.

   The gut microbiota, functioning as a dynamic “virtual organ,” plays
   pivotal roles in immune regulation, metabolic homeostasis, and barrier
   integrity [[95]21]. In our study, alpha-diversity metrics remained
   stable in SLE-nonLN compared to HC, whereas beta-diversity analysis
   revealed significant compositional shifts between SLE groups and HC.
   These results suggest that alterations in microbial composition, rather
   than overall richness, may be associated with disease status. Further
   subgroup comparisons revealed differing microbial patterns between
   SLE-nonLN and SLE-LN, with SLE-LN characterized by an increased
   abundance of taxa previously linked to inflammatory phenotypes.

   At the phylum level, we observed a shift from Firmicutes to
   Proteobacteria during disease progression. Firmicutes were predominant
   in SLE-nonLN (59.08%), while Proteobacteria were significantly enriched
   in SLE-LN (28.02%). This contrasts with a Spanish cohort reporting
   Bacteroidetes enrichment in SLE [[96]22], likely reflecting geographic
   and dietary influences on microbial ecology. The depletion of
   Firmicutes (particularly butyrate-producing Faecalibacterium) and
   enrichment of Proteobacteria (e.g., Escherichia-Shigella) in SLE-LN may
   contribute to gut barrier dysfunction, permitting translocation of
   pro-inflammatory metabolites (e.g., lipopolysaccharides) into systemic
   circulation—a mechanism implicated in LN-related renal inflammation
   [[97]23]. Similar phylum-level dysbiosis patterns observed in systemic
   sclerosis and rheumatoid arthritis further support the potential
   involvement of these taxa in autoimmune dysregulation [[98]24–[99]26].

   LEfSe analysis further delineated key taxonomic differences. Class
   Bacilli and order Lactobacillales were enriched in SLE-nonLN, whereas
   SLE-LN samples showed a predominance of Enterobacteriaceae and
   Gammaproteobacteria. These findings are consistent with previous
   reports indicating increased fecal Lactobacillus abundance in SLE
   patients [[100]27, [101]28]. Notably, preclinical models have
   demonstrated strain-specific effects of Lactobacillus spp., with L.
   reuteri exhibiting protective immunomodulatory effects [[102]29], while
   other strains may exacerbate autoimmunity [[103]30]. In our cohort, the
   consistent enrichment of Enterobacteriaceae, Enterobacterales, and
   Escherichia-Shigella in SLE-LN (LDA > 4.7, p < 0.05) underscores the
   potential contribution of these taxa to disease severity. While the
   exact mechanisms remain to be elucidated, their expansion in LN
   patients suggests they may serve as markers of microbial dysbiosis
   associated with disease progression [[104]31, [105]32].

   Metabolomics, with its high resolution and sensitivity, has emerged as
   a powerful tool for unraveling disease-specific metabolic perturbations
   [[106]33, [107]34]. Our untargeted fecal metabolomics analysis
   identified profound alterations in lipid and amino acid metabolism
   across SLE stages, underscoring their pivotal roles in disease
   progression. Notably, Mead acid (5,8,11-Eicosatrienoic acid), a
   biomarker of essential fatty acid deficiency [[108]35, [109]36], was
   significantly elevated in SLE-LN compared to SLE-nonLN (FC = 3.11,
   p = 0.004). Elevated Mead acid levels, often indicative of a deficiency
   in dietary essential fatty acids, particularly arachidonic acid, may
   contribute to autoimmune disease progression [[110]37]. These findings
   align with Zhang et al.’s report of Mead acid as a prognostic marker in
   LN [[111]38], suggesting its potential utility for identifying SLE
   patients at higher risk of renal involvement.

   Changes in lipid metabolism are closely associated with lipid-induced
   nephrotoxicity and play a significant role in SLE-LN pathophysiology.
   Bile acids, essential regulators of lipid metabolism, mediate this
   process through TGR5 modulation [[112]39]. Previous studies have
   reported correlations between specific bile acids—including Glycocholic
   acid and Glycochenodeoxycholic acid—and SLE disease activity [[113]40].
   In our cohort, both metabolites were significantly elevated in
   SLE-nonLN compared to HC (FC = 5.06 and 8.88, respectively), suggesting
   potential involvement in early-stage disease. These elevations may be
   linked to dysregulation of primary bile acid biosynthesis and lipid
   handling in SLE [[114]41]. Interestingly, both bile acids showed
   markedly reduced levels in SLE-LN (FC = 0.37 and 0.28, respectively),
   implying a shift in bile acid metabolism during disease progression.
   While ROC analysis indicated high classification performance for these
   metabolites in distinguishing SLE subgroups (AUC = 0.951 and 0.827),
   their diagnostic utility warrants further validation in larger,
   independent cohorts. These findings support the relevance of bile acid
   metabolism in SLE and highlight Glycocholic acid and
   Glycochenodeoxycholic acid as promising candidates for non-invasive
   monitoring of disease progression [[115]42].

   Beyond lipid dysregulation, we observed stage-specific shifts in amino
   acid metabolism. SLE-nonLN patients exhibited elevated fecal levels of
   glucogenic amino acids (e.g., glycine, proline), indicative of a
   metabolic shift toward gluconeogenesis—a potential adaptation to
   chronic inflammation-induced energy demands [[116]43]. In contrast,
   glycine and proline levels were reduced in SLE-LN patients compared to
   those with SLE-nonLN. This reduction was accompanied by distinct shifts
   in gut microbial composition, suggesting a potential association
   between amino acid metabolism and the progression of disease. Among the
   taxa showing the most notable differences across the groups were
   Escherichia-Shigella and Enterobacter, both of which were significantly
   more abundant in the SLE-LN group (p < 0.001). Escherichia-Shigella
   exhibited a negative correlation with bile acid metabolites, including
   Glycochenodeoxycholic acid and Glycocholic acid, across all three
   groups. Similarly, Enterobacter was negatively correlated with nearly
   all differential metabolites, such as Pterine, L-cysteine, and
   Glycoursodeoxycholic acid, which are strongly associated with bile acid
   biosynthesis and thiamine metabolism pathways. These findings suggest
   that the expansion of these microbial taxa may contribute to
   alterations in host metabolic pathways linked to disease severity.
   Nevertheless, further experimental studies are needed to confirm these
   associations and clarify the underlying mechanisms.

   However, this study has several limitations. The relatively small
   sample size reduces the statistical power of subgroup analyses and may
   limit the generalizability of our findings. Additionally, although
   participants were newly admitted and not receiving immunosuppressive
   therapy at the time of sampling, it was not feasible to completely rule
   out prior medication effects, which may have impacted the gut
   microbiota and metabolome. Future studies should incorporate larger and
   more diverse patient cohorts, including those with other autoimmune or
   inflammatory conditions, to validate and expand upon these findings.
   Longitudinal analyses are also warranted to explore dynamic changes in
   microbiome–metabolome interactions throughout the course of SLE and LN
   progression.

Conclusion

   This study identified distinct alterations in gut microbiota
   composition and fecal metabolite profiles in patients with SLE,
   particularly those with LN. SLE was associated with increased
   Firmicutes and decreased Proteobacteria, while SLE-LN showed enrichment
   of potentially pro-inflammatory taxa such as Enterobacteriaceae (e.g.,
   Escherichia-Shigella), which may be linked to renal inflammation.
   Metabolomic profiling revealed stage-specific metabolic shifts,
   including elevated levels of Mead acid and reduced concentrations of
   key bile acids such as Glycocholic acid and Glycochenodeoxycholic acid
   in SLE-LN patients. Notably, the abundance of Escherichia-Shigella was
   inversely correlated with key bile acid levels, suggesting possible
   microbiota–metabolite interactions relevant to LN progression. While
   causal relationships cannot be inferred, these findings may enhance our
   understanding of the microbiome–metabolome axis in SLE and support the
   potential use of non-invasive biomarkers for monitoring and stratifying
   disease severity.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [117]Supplementary Material 1^ (1.7MB, docx)

Acknowledgements