Phage-microbe dynamics after sterile faecal filtrate transplantation in individuals with metabolic syndrome: a double-blind, randomised, placebo-controlled clinical trial assessing efficacy and safety

Metabolic syndrome (MetSyn), affecting about a quarter of the global population, is characterized by insulin resistance and elevates risks for type 2 diabetes and cardiovascular disease. The gut microbiota contributes to these conditions through metabolites and inflammation. Beyond bacteria, the gut ecosystem includes viruses, predominantly bacteriophages (phages), which modulate microbial community structure via lytic and lysogenic cycles and have been implicated in gastrointestinal diseases and diabetes. Prior work described reduced richness and diversity of the gut phageome in MetSyn with altered composition. Given phages’ capacity to modulate bacteria, interest has grown in phage-based interventions, including phage therapy and faecal virome transplantation (FVT). In mice, FVT can mirror effects of faecal microbiota transplantation (FMT), and sterile faecal filtrate transplantation (FFT) has cured recurrent Clostridioides difficile infection in a small human study, potentially offering safety advantages over FMT by removing live bacteria. As lean-donor FMT improved insulin sensitivity in MetSyn and murine FVT improved weight and glucose tolerance, the study aimed to test whether transferring faecal phages via FFT alters glucose metabolism and gut phage-bacteria dynamics in humans with MetSyn.

Key prior findings include: (1) phages significantly shape microbial communities and the human gut virome is abundant and diverse; (2) alterations in the gut virome associate with inflammatory bowel disease, colorectal cancer, and type 2 diabetes; (3) in MetSyn, phageome richness and diversity are reduced with greater inter-individual variation; (4) phage therapy can target specific pathogens but may not broadly remodel microbiomes; (5) FVT in mice reduced weight gain and improved glucose tolerance, with effects mitigated by antibiotics (suggesting phage-bacteria dependence); (6) sterile faecal filtrate transfer cured recurrent C. difficile in humans; and (7) lean-donor FMT improved insulin sensitivity in MetSyn. Collectively, these studies support exploring bacteriophage-containing, bacteria-depleted transplants to beneficially modulate gut microbiota and metabolic outcomes, while highlighting safety and standardization advantages over traditional FMT.

Study design: Prospective, double-blind, randomised, placebo-controlled pilot clinical trial at Amsterdam UMC (AMC), Netherlands. Participants: 24 European Dutch adults (18–65 years) with MetSyn (NCEP criteria), BMI ≥25 kg/m², not on medications, and without significant comorbidities or substance use; randomized to FFT (n=12) or placebo (n=12). Donors: Lean healthy European Dutch donors screened per European FMT Working Group guidelines, including questionnaires, stool screening for parasites/pathogens/MDROs, calprotectin, blood tests, and (from May 2020) SARS-CoV-2 testing; 5 donors used. Interventions: FFT produced from 50 g donor stool homogenized in 500 ml sterile saline, filtered by gauze, centrifuged twice (1 h, 10,000×g), and passed through a sterile 0.2 µm tangential flow filtration membrane (Vivaflow 50); stored overnight refrigerated. Administration via nasoduodenal tube; ~300 ml infused over 15–20 min. Placebo: sterile saline with brown color, identical appearance. Pre-treatment: bowel cleansing with Klean-Prep the day before. Blinding: opaque tubing and blinded syringes; randomisation by block (sizes 4/6/8) with stratification for age/sex via Castor EDC. Outcomes: Primary—change in glucose metabolism assessed by OGTT (75 g) total glucose AUC from day 0 to day 28. Secondary—fasting glucose, insulin, HOMA-IR, HbA1c; continuous glucose monitoring (CGM; Freestyle Libre) one week before to one week after intervention; safety (AEs, clinical labs for renal/liver function, hematology, inflammation); microbiome/virome dynamics. Sampling: Blood for OGTT glucose and C-peptide at 0, 15, 30, 45, 60, 90, 120 min; fasting labs at baseline and day 28. Stool collected at baseline and days 2, 4, 7, 14, 28. CGM data analyzed with CGDA. Sequencing and bioinformatics: Whole-genome shotgun (WGS) sequencing on fecal DNA (NEBNext Ultra II; Illumina HiSeq 2×150 bp; ~6 Gb/sample). Viral-like particles (VLP) isolated (0.45 µm filtration, free-DNA digestion, dsDNA phage only), libraries (NEBNext Ultra II FS) sequenced on NovaSeq 6000 S4 2×150 bp. Read QC (fastp), error correction and deduplication (bbmap tadpole/clumpify); WGS reads cross-assembled per participant with metaSPAdes; VLP assemblies with MEGAHIT. Viral contigs (>5 kb) identified using VirSorter2 and CheckV with stringent criteria; deduplicated and clustered into viral populations (VPs; 90% ANI) and viral clusters (vContact2); abundance by mapping (bowtie2) with coverage thresholds; RPKM computed. Bacterial profiling with mOTUs; binning to MAGs (MetaBAT2), quality checked (CheckM), taxonomy (GTDB-Tk). Phage–host links via prophage identification (viral contigs within MAGs) and CRISPR spacer matches (≤2 mismatches). Statistics: Richness and alpha diversity (Shannon); composition via clr-transformed PCA and principal response curves (vegan); PERMANOVA with age/sex covariates; multiple testing via Benjamini–Hochberg. Differential abundance of VLP VPs on day 2 by ANCOM-BC (age/sex adjusted). Phage–host interaction dynamics assessed by Spearman correlations of species-level mean changes in VP and MAG abundances (days 0–2, 2–28, 0–28). Safety/quality controls: Donor FFT VLP counts ~1.25×10^8 VLPs/ml (SD 0.45×10^8); qPCR confirmed ~10^5-fold reduction of bacterial DNA; culture showed no colony-forming units in 100 ml filtrate. Eukaryotic/human viral reads were low (mean 0.044%±0.3%; median 0%).

- Participants: 24 MetSyn subjects randomized (FFT n=12; placebo n=12); baseline characteristics balanced except screening systolic BP higher in placebo but not at baseline/follow-up. - Safety: No serious adverse events; more intervention-related AEs in FFT than placebo (subjects: 6 vs 2, p=0.19; total AEs: 8 vs 2, p=0.11); AEs mild GI (diarrhoea, constipation, bloating, nausea) resolving in median 3 days. Clinical safety labs showed no between-group differences; urea increased in both groups, likely due to laxative. - Primary outcome: No significant difference between groups in OGTT glucose AUC change from day 0 to day 28. Within-group small increases in glucose AUC were observed and were nominally significant in FFT before multiple-testing correction; C-peptide AUC changes not different between groups. - Secondary metabolic outcomes: Fasting insulin increased within both groups from day 0 to day 28 (placebo p=0.036; FFT p=0.007), with corresponding HOMA-IR increases (placebo p=0.047; FFT p=0.008). Fasting glucose and HbA1c similar between groups. BMI, BP, and lipid parameters largely unchanged between groups (some within-group changes noted in Table 3). - CGM: Overall similar glucose levels and variability between groups; within FFT group, time in range (3.9–10 mmol/L) increased from 95.5% to 97.5% (p=0.02, nominal significance; lost after multiple-testing correction), suggesting transient improvement in glucose variability post-intervention. - Microbiome/virome diversity: Bacterial and WGS phageome richness and alpha diversity showed slight, non-significant decreases immediately post-intervention in both groups, returning toward baseline by days 14–28. VLP phageome richness increased slightly by day 2 in both groups; VLP alpha diversity decreased slightly only in placebo (non-significant). - Donor-shared and new phages: WGS phageome showed a non-significant increase in donor-shared VPs up to day 14 after FFT (p=0.2). VLP phageome donor-shared VPs decreased non-significantly after FFT (p=0.3). The relative abundance of new VPs increased over time in both groups; increases were slightly higher in FFT, particularly VLPs on day 2 (p=0.2). - Composition changes: Significant difference in VLP phageome composition at day 2 between FFT and placebo (PRC and PCA; PERMANOVA p=0.02 and p=0.028, respectively). ANCOM-BC identified 216 differentially abundant VPs on day 2. Bacterial hosts enriched among these VPs included six species and five genera; species with significant enrichment included Sutterella wadsworthensis (more prevalent in placebo), Scatocola faecigallinarum, Roseburia intestinalis, Faecalibacterium sp000434635, CAG-882 sp003486385, and CAG-267 sp001917135 (several more prevalent in FFT). - Phage–host dynamics: Between days 0–2, changes in VP vs host MAG abundances were negatively correlated in FFT (Spearman R = -0.13, p=0.0048) and positively correlated in placebo (R = 0.17, p=3.3×10^-6), indicating more antagonistic (lytic) dynamics in FFT and more protagonistic/lysogenic dynamics in placebo. Effects diminished by days 2–28 (FFT R = -0.043, p=0.24; placebo R = 0.12, p=0.0041) and absent over days 0–28. - Overall: FFT safely and transiently perturbed the VLP phageome and phage–bacterium interaction dynamics without producing a significant improvement in OGTT-based glucose metabolism at 28 days.

The trial addressed whether transferring a bacteria-depleted faecal filtrate containing phages from lean donors could improve glucose metabolism and modulate gut phage–bacterium dynamics in MetSyn. FFT was safe and well-tolerated, supporting feasibility and potential safety advantages over FMT. Although the primary metabolic outcome (OGTT glucose AUC) did not differ between groups at 28 days, FFT recipients exhibited a nominal short-term improvement in CGM time-in-range and, critically, showed significant, short-lived alterations in the VLP phageome composition at day 2 alongside antagonistic phage–host dynamics. These findings suggest that gut phages can transiently reshape the virome and induce lytic interactions with bacterial hosts in humans. The lack of significant improvement in glucose metabolism may reflect several factors: small sample size and heterogeneity limiting statistical power; the use of FFT rather than a more purified/concentrated FVT; the bowel laxative pre-treatment potentially altering host-bacterial availability and baseline insulin/HOMA-IR; and donor–recipient mismatches reducing effective phage–host encounters in a highly individual-specific human virome. In contrast to murine studies, human microbiomes are more heterogeneous, and the inclusion of bacterial communities in FMT may be necessary to achieve metabolic benefits in MetSyn. Nonetheless, the observed day-2 VLP compositional shift and antagonistic phage–host correlations indicate that phage-based interventions can modulate gut microbial ecology, albeit transiently under current protocols.

This double-blind, randomised, placebo-controlled pilot trial demonstrates that sterile faecal filtrate transplantation from lean donors is safe and can transiently alter the gut phageome and induce antagonistic phage–bacterium interactions in individuals with metabolic syndrome. However, FFT did not significantly improve glucose metabolism as measured by OGTT at 28 days. The study provides a basis for future trials to optimize efficacy, including omitting laxative pre-treatment, refining filtrate preparation toward cleaner and more concentrated viromes, pooling donors to enhance phage diversity, and matching donors to recipients based on bacteriome/phageome profiles. Future work should broaden virome profiling to include ssDNA and RNA viruses and explore targeted phage strategies against specific bacterial taxa, potentially in combination with approaches that restore broader microbial dysbiosis.

- Small sample size and participant heterogeneity likely limited power to detect metabolic effects; original power assumed FFT efficacy comparable to FMT, which may not hold. - Laxative pre-treatment likely affected baseline fasting insulin/HOMA-IR and may have reduced available bacterial hosts, potentially diminishing FFT efficacy and contributing to metabolic parameter variability. - FFT (0.2 µm filtration) not equivalent to a purified/concentrated FVT; filtrates may contain non-phage components (bacterial debris, metabolites, peptides) influencing outcomes; 0.2 µm pore may exclude some larger phages; eukaryotic viruses were present at very low levels but effects cannot be fully excluded. - Sequencing focused on dsDNA phages, potentially missing ssDNA and RNA viruses; some VLP assemblies were challenging; no mock positive controls were included. - Donor–recipient matching was limited; high inter-individual virome specificity may hinder phage engraftment; donor pooling not implemented. - Short follow-up (28 days) may miss longer-term effects; observed virome and interaction changes were transient. - Population limited to Dutch European subjects; generalizability may be limited; small sample precluded sex-based analyses.