Metallomics Reviews
Integrated multi-omics and MRE model for Crohn’s bowel damage
Clinical Overview
Prospective two-centre study of 309 adults with Crohn’s disease and 30 healthy controls integrates magnetic resonance enterography, fecal 16S rRNA profiling, and targeted faecal/serum metabolomics to predict cumulative bowel damage defined by Lémann Index >4.8. Key signatures include reduced microbial diversity, enrichment of Erysipelatoclostridium and Ruminococcus_gnavus_group, elevated fecal aromatic amino acids, and depleted serum glycerophospholipids/sphingolipids. The optimized XGBoost classifier combining three genera, nine metabolites, and three MRE features achieved AUC 0.857 internally and 0.829 in an external cohort, with perianal disease, Erysipelatoclostridium, and faecal alanine the dominant contributors guiding accurate prediction of bowel damage risk and severity in routine practice.
What was reviewed and who was studied
This original prospective two-centre study enrolled 230 adults with Crohn’s disease and 30 healthy controls in a derivation cohort, plus 79 additional Crohn’s patients in an external test cohort at a tertiary hospital in China. Bowel damage was defined by Lémann Index >4.8. Patients underwent standardised 3.0 T MRE, faecal 16S rRNA sequencing, and targeted UPLC–MS/MS metabolomics on faecal (233 metabolites) and serum (528 metabolites) samples, with multi-omics integration and machine learning used to predict bowel damage.
Major findings
| Finding | Details |
|---|---|
| Bowel-damage phenotype on MRE | Seven MRE features differed between BD and NBD: greater bowel wall thickness (median 7.0 vs 6.0 mm, P=0.010), longer involved segments (>15 cm more frequent, P=0.042), higher mural T2 signal, more perienteric effusion, more comb sign, more penetrating lesions and markedly more perianal disease (perianal fistula/abscess 84.5% vs 45.7%, P<0.001). |
| Microbial diversity and composition | CD patients had significantly reduced α-diversity vs controls (all P<0.001), with further reductions in Simpson and Pielou’s evenness in BD vs NBD (P<0.05). Community structure differed across HC, NBD, BD by Bray–Curtis PCoA (P≤0.038). |
| Key bacterial genera | Ruminococcus gnavus group and Erysipelatoclostridium emerged as BD-associated genera, increasing from controls to NBD, then decreasing in BD but remaining above control levels. Beneficial genera (Romboutsia, Prevotella_9, Barnesiella, Fusicatenibacter, Butyricicoccus) declined progressively HC→NBD→BD (all P<0.05). |
| Faecal metabolomic shifts | Of 233 faecal metabolites, BD showed enrichment of amino acids, organic acids and peptides (FDR<0.001). Eighteen shared discriminatory metabolites were mainly amino acids and fatty/carboxylic acids; 16 were increased in BD, whereas pimelic and suberic acid were decreased. Pathway analysis highlighted phenylalanine/tyrosine/tryptophan biosynthesis (Impact 0.500, P=0.035). |
| Serum lipidomic disruption | Among 528 serum metabolites, 67 overlapping differentials (mainly SM, PC, PE, CE, Cer, LPC) showed a stepwise decrease from HC to NBD to BD. Pathways for glycerophospholipid, ether lipid and sphingolipid metabolism were significantly perturbed (P≤0.016). |
| Microbe–metabolite correlations | Erysipelatoclostridium, R. gnavus and Clostridium_innocuum_group correlated positively with multiple PCs, PEs, SMs and PSs (e.g. ePE(38:1) r>0.34, P<0.01), while Fusicatenibacter was negatively correlated with several faecal amino acids (glycine r=−0.456, P<0.001). |
| Linking micro-omics to MRE | Twenty-nine biomarkers (6 genera, 5 faecal, 18 serum metabolites) correlated with 7 MRE features. Erysipelatoclostridium correlated with bowel wall thickness (r=0.380, P=0.002), T2 signal (r=0.299) and perienteric effusion (r=0.331). Serum PCs and ePEs were negatively correlated with penetration and high T2 signal. |
| Mediation and causal pathways | Mediation analyses suggested Prevotella_9 reduced penetration and perianal disease via serum Cer(d18:2/23:1) and PE(38:1), with 22.8–26.6% mediation; Fusicatenibacter influenced penetration via faecal 2-hydroxybutyric acid (mediation 52.7%). Reduced R. gnavus linked to lower PC(O-38:2) and SM(d17:1/24:1) (PACME 0.007–0.010). |
| Multi-omics prediction model | An XGBoost model combining 3 genera, 6 faecal metabolites, 3 serum metabolites, and 3 MRE features plus clinical covariates (Model 7) achieved AUC 0.857 (derivation) and 0.829 (external test), with PPV 81.8% and NPV 73.7%. SHAP analysis prioritised perianal disease, Erysipelatoclostridium, faecal alanine, serum ePS(38:2) and faecal arachidonic acid. |
Implications for Microbial Metallomics
The study links specific gut community structures and organic metabolite signatures to structural bowel damage, using a multi-omics plus imaging framework that is directly adaptable to future work incorporating microbial and host metallomic layers in Crohn’s disease.
| Concept | Implication |
|---|---|
| Defined BD-associated dysbiosis (e.g. Erysipelatoclostridium, R. gnavus, Prevotella_9) | These genera can anchor future metallomic profiling to test whether metal-handling pathways co-vary with BD-linked taxa, enabling organism-specific hypotheses about metal-dependent inflammatory or fibrotic processes. |
| Aromatic amino acid–enriched faecal metabolome in BD | The demonstrated pipeline could be extended to quantify metal-binding or metal-modified metabolites alongside aromatic amino acids, permitting joint modelling of redox-active small molecules and bowel structural injury. |
| Systemic depletion of glycerophospholipids and sphingolipids in BD | Lipid-centric serum signatures show that systemic biochemistry can be linked to MRE features; analogous serum metallomic panels (e.g. total vs protein-bound species) could be integrated into similar machine-learning frameworks. |
| Strong microbe–metabolite–MRE correlations and mediations | The mediation strategy provides a template for testing whether microbial metal-utilisation pathways, measured by targeted metalloproteomic or speciation assays, mediate imaging-defined bowel damage. |
| Multi-omics+MRE XGBoost classifier with external validation | The modelling architecture is directly re-usable with additional metallomic inputs, allowing evaluation of whether adding metal species improves discrimination of BD over purely organic metabolite signatures. |
| Use of LI-defined BD as outcome | A validated structural damage index provides a robust endpoint for future metallomic studies to relate microbial metal-related functions to long-term bowel injury rather than transient activity alone. |
Limitations
The authors note that the modest, two-centre sample may limit generalizability and that broader, multi-centre cohorts are needed. Mechanistic in vitro and in vivo studies were not performed, so causal pathways linking specific taxa, metabolites and MRE features to bowel damage remain inferential. Inflammatory markers such as faecal calprotectin and detailed clinical activity indices were not incorporated, as the focus was cumulative structural damage, potentially under-representing concurrent inflammatory burden.
Future perspectives
Logically extending this work, larger multicentre cohorts with greater geographic and treatment diversity should validate the XGBoost multi-omics model and recalibrate it for different clinical environments. Serial sampling could test whether microbial and metabolic signatures track evolving Lémann Index and MRE features or predict response to biologics and surgery. Mechanistic experiments exposing key taxa (Erysipelatoclostridium, R. gnavus, Prevotella_9, Fusicatenibacter) to host-derived substrates could dissect how they modulate aromatic amino acid and lipid pathways linked to MRE phenotypes. Methodologically, the same pipeline could accommodate additional omics layers, including targeted metallomic or proteomic measurements, to explore whether expanding beyond organic metabolites further sharpens bowel damage stratification.
Key takeaways for Researchers and Clinicians
In adult Crohn’s disease across two hospital campuses, bowel damage defined by Lémann Index >4.8 was associated with a distinct pattern of MRE changes, gut dysbiosis and coordinated fecal and serum metabolic disturbances. Amino acid–rich faecal profiles and systemic depletion of glycerophospholipids and sphingolipids emerged as key biochemical correlates of damage, tightly linked to shifts in genera such as Erysipelatoclostridium, RuminococcusRuminococcusRuminococcus_gnavus_group, Prevotella_9 and Fusicatenibacter, and to MRE features including wall thickening, comb sign, perianal disease and penetrating lesions.
The most clinically actionable insight is methodological: integrating fecal microbiome, fecal and serum metabolomics, and carefully selected MRE features into an externally validated XGBoost model provided robust discrimination of bowel-damaged versus non-damaged patients (AUC 0.829 in the external cohort), with perianal disease and a small set of microbial and metabolic markers driving predictions. This suggests a plausible future pathway toward non-invasive, mechanism-informed BD stratification that could inform escalation decisions, imaging intervals and trial enrichment. In translational terms, the study demonstrates how a multi-omics–imaging framework can serve as a scaffold onto which future metallomic measurements might be layered to link microbial community function, systemic chemistry and structural gut injury.
Citation
Huang L, Meng J, Lin S, et al. Integrating gut microbiome and metabolomics with magnetic resonance enterography to advance bowel damage prediction in Crohn’s disease. Journal of Inflammation Research. 2025;18:7631–7649. doi:10.2147/JIR.S524671