Metabolic network restoration, not donor microbial engraftment, associates with resolution of recurrent Clostridioides difficile infection

Recurrent Clostridioides difficile infection (rCDI) remains one of the most intractable consequences of antibiotic-associated dysbiosis. Standard therapies such as vancomycin suppress the pathogen but simultaneously erode the commensal communities that normally provide colonization resistance, trapping patients in a cycle of relapse. Fecal microbiota transplantation (FMT) has transformed rCDI care with durable clinical response, yet the mechanisms by which it restores intestinal health have remained unclear. A new multi-omics study published in iScience now suggests that resolution of rCDI is associated with the restoration of gut metabolic networks rather than with simple donor microbial replacement, offering a fresh conceptual framework for next-generation microbiome-based therapeutics.

The investigation builds on a large randomized trial of 138 participants that compared two fundamentally different interventions: lyophilized fecal microbiota transplantation (LFMT), which contains live donor bacteria, and lyophilized sterile fecal filtrate (LSFF), manufactured by removing live bacteria from fecal suspension through a series of filters down to a pore size of 0.2 micrometers before lyophilization. While LSFF was less effective at preventing rCDI (65 percent) than full-spectrum LFMT (88 percent), the possibility remained that other components such as metabolites, bacteriophages, and extracellular proteins could contribute to clinical benefit without donor microbial engraftment. To address this question, the researchers analyzed paired stool samples collected from the first seven participants in the trial, four LFMT recipients and three LSFF recipients, at week 0 before treatment and at week 24 after treatment. All seven participants received investigational products manufactured from a single healthy donor.

The team combined proximity-ligation (Hi-C) metagenomics, metaproteomics, and targeted metabolomics in a systems-level framework rarely applied in this context. Hi-C sequencing crosslinks DNA molecules that are physically close in intact cells, allowing researchers to link viral contigs to their bacterial hosts and to reconstruct metagenome-assembled genomes with strain-level resolution. The LSFF products were first verified to contain minimal non-viable bacterial DNA: two complementary metagenomic pipelines, Kraken2 and MetaPhlAn4, showed 74.71 percent fewer bacterial reads in LSFF relative to LFMT, with bacterial richness decreased by 54 percent with Kraken2 and 90.1 percent with MetaPhlAn4. Gram staining and culturomics detected no organisms or growth, indicating that the trace bacterial signals likely represented extracellular DNA fragments or non-viable cell debris.

Surprisingly, the bacteria-free filtrate was enriched for bacteriophages, exhibiting a six-fold increase in viral sequence richness compared with LFMT. The viral community was dominated by members of the class Caudoviricetes, the most prevalent double-stranded DNA bacteriophages in the human gut virome, with Klebsiella phage VLCpiS11a the most abundant species at 23.6 percent. Bacterial community assembly followed distinct trajectories after the two treatments. By week 24, LFMT recipients had recovered bacterial richness and Shannon diversity approaching donor levels, and principal coordinates analysis showed a clear shift toward the donor cluster. Fourteen bacterial lineages present in donors and recipients at week 24 but absent at week 0 showed average nucleotide identity (ANI) values above 0.995, supporting strain-level donor engraftment for taxa including Alistipes putredinis, Agathobacter rectalis, and Blautia stercoris. In contrast, LSFF recipients displayed only partial recovery, and only four bacterial taxa were shared between donors and recipients at week 24 with limited genomic similarity, suggesting that the modest bacterial recovery following LSFF was largely driven by endogenous or environmentally acquired taxa.

Phage-host associations also diverged between the groups. Across LFMT recipients, the majority of phage linkages retained the same predicted bacterial host species as their donor counterparts, with host concordance above 60 percent in three of four recipients, indicating that whole-stool transplantation preserves native phage-host structures. Host-change events in LFMT recipients were concentrated almost exclusively within the Firmicutes phylum, particularly among Lachnospiraceae, Oscillospiraceae, and Ruminococcaceae, and involved phylogenetically closely related genera such as Faecalibacterium and Eubacterium. LSFF recipients showed more heterogeneous patterns, with species-concordant linkages ranging from 0 percent to 61.8 percent, and host transitions in one participant involving Bacteroidota including Bacteroides uniformis and Bacteroides stercoris. This suggests that phages transferred without their original bacterial hosts may interact with a broader range of potential hosts, although the small participant number warrants caution.

Despite these marked compositional differences, functional metagenomic profiling identified a striking convergence toward donor-like functional capacity. At week 0, 112 functions overlapped with donor profiles in LFMT recipients and 151 in LSFF recipients; by week 24, 27 additional donor-like functions emerged in LFMT and 32 in LSFF, with 22 functions shared between the two treatment groups. Heatmap analysis integrating functional annotations with bacterial strain-level identity revealed that the newly acquired donor-like functions were not predominantly carried by high-ANI donor-derived strains. Instead, most convergent functions were linked to low-ANI bacterial lineages, emphasizing the role of non-donor bacteria and the distributed nature of functional redundancy within the recovering microbiome. In LSFF recipients, donor-like functions including N-acetylglucosamine utilization and multiple sugar transport systems were consistently associated with low-ANI taxa, indicating that functional recovery followed different ecological routes in the two groups.

Metaproteomic analysis corroborated these findings at the level of realized protein expression. Deep profiling identified 18,690 protein groups, of which 17,814 were prokaryotic, 368 human, 157 viral, and 351 plasmid. Metaproteome richness was primarily influenced by time rather than treatment: both groups showed markedly lower protein identification at week 0 than at week 24, recovering to donor-comparable levels by week 24, driven predominantly by increased microbial protein expression. Comparative KEGG pathway analysis demonstrated a time-dependent metabolic transition. At week 0, enriched pathways reflected stress and survival functions, including two-component regulatory systems, RNA degradation, biofilm formation, beta-lactam resistance, oxidative phosphorylation, and the glyoxylate/dicarboxylate cycle. By week 24, these stress-associated pathways were supplanted by nutrient and energy flux modules such as carbon fixation, ABC transporters, coenzyme A biosynthesis, and 2-oxocarboxylic acid metabolism, with particularly notable changes in amino acid biosynthetic and degradation pathways, including branched-chain amino acids, cysteine, and arginine metabolism, pathways linked to C. difficile growth and FMT outcome. Taxon-function analysis identified 915 genus-level taxa-function pairs differentially regulated between week 0 and week 24, with 797 upregulated at week 24 and enriched for functions derived from health-associated genera including Phocaeicola, Anaerobutyricum, Blautia, and Eubacterium.

Metabolomic profiling using chemical isotope labeling liquid chromatography-mass spectrometry (CIL LC-MS) connected these genomic and proteomic alterations with metabolic output. Across all samples, 7,237 unique peak pairs were detected, with 1,504 metabolites increased and 826 decreased after intervention. The most prominent alterations occurred in dipeptides and tripeptides, amino acids and their derivatives, bile acids, short-chain fatty acids, and vitamin B-related metabolites. Small peptides such as leucyl-lysine and tyrosyl-glycine were markedly elevated at week 24, suggesting enhanced peptide turnover or microbial proteolytic activity, while polyamines such as cadaverine and putrescine were elevated at week 0, reflecting dysregulated nitrogen metabolism. Origin analysis using MetOrigin classified 8 metabolites as host-derived, 83 as microbiota-derived, and 70 as co-metabolized, and pathway enrichment revealed that only 4 host-specific pathways were significantly altered, whereas 59 co-metabolic and 38 microbiota-associated pathways were enriched. Cysteine and methionine metabolism emerged as a central node, representing the sole pathway common to all three origins. Heatmap analysis of 19 proteinogenic amino acids showed that L-arginine, threonine, serine, and glutamic acid were significantly elevated at week 24, while cysteine, alanine, and glycine were highest at week 0, mirroring the metaproteomic evidence for coordinated host-microbiota amino acid exchange.

The authors conclude that durable rCDI resolution was accompanied by restoration of gut microbiome metabolic function despite distinct compositional trajectories, with both interventions converging toward recovery of amino acid metabolism, redox balance, and host-microbiota co-metabolism. These observations support a model in which rCDI resolution is associated with metabolic network reprogramming that may arise through multiple ecological processes, including donor microbial engraftment, extracellular microbial products and metabolites, and activity of endogenous or environmentally acquired bacteria. The coordinated shifts away from oxidative stress and survival pathways toward anaerobic, nutrient-processing functions suggest re-establishment of an energetically favorable metabolic niche that supports colonization resistance and suppresses C. difficile outgrowth. Amino acid circuits, particularly branched-chain amino acids, cysteine, and arginine pathways, are known to modulate C. difficile toxin production, spore germination, and immune signaling, and the coordinated rise in amino acid derivatives, dipeptides, and short-chain fatty acids is consistent with re-establishment of nitrogen and carbon exchange between host and microbiota.

The study has important limitations that the authors emphasize. The small number of participants limits statistical power and increases susceptibility to inter-individual variability, the study did not include a placebo-treated comparator group, and only participants who failed neither treatment were profiled, making it impossible to determine whether the identified signatures are specific to successful treatment or simply associated with recovery over time. Because samples were profiled at week 0 and week 24 only, the ecological dynamics between these time points cannot be resolved, and only a single donor was included, limiting generalizability. The authors stress that the results should be viewed as hypothesis-generating and require validation in larger cohorts including both successful and unsuccessful treatment outcomes, together with complementary in vitro and in vivo experimental studies. Nonetheless, the metabolic signatures identified here, particularly those involving amino acid and redox pathways, may warrant evaluation as potential biomarkers for treatment monitoring or as targets for future development of synthetic consortia and defined postbiotic formulations, positioning functional restitution as a key endpoint of microbiota-based therapy.

Subject of Research: Microbiome

Article Title: Metabolic network restoration, not donor microbial engraftment, associates with resolution of recurrent Clostridioides difficile infection

Article References: Whitman, B. T., Karimianghadim, R., Auch, B., Liachko, I., Monaghan, T. M., Pakpour, S., & Kao, D. (2026). Metabolic network restoration, not donor microbial engraftment, associates with resolution of recurrent Clostridioides difficile infection. iScience, 29(11), Article 117907. https://doi.org/10.1016/j.isci.2026.117907

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117907

Keywords: Clostridioides difficile, fecal microbiota transplantation, metagenomics, metaproteomics, metabolomics, bacteriophages, dysbiosis, microbiome therapeutics, Metabolic, network, restoration, donor