Abstract

Background

   Fecal microbiota transplantation (FMT) is widely used to treat severe
   infections and investigated for the treatment of complex diseases. The
   therapeutic efficacy of FMT is related to the successful engraftment of
   bacteriophages from healthy donors to recipients. However, gut
   bacteriophage contributions to FMT engraftment and treatment outcomes
   remain unclear.

Methods

   The gut phageome from previously published metagenomes of donors and
   recipients across 23 FMT studies was assembled and functionally
   annotated for a meta-analysis.

Results

   Gut phageome profiles of FMT recipients, especially those with
   recurrent Clostridioides difficile infection (rCDI), shifted toward
   donor phageomes, accompanied by increased phageome alpha diversity.
   Engraftment of donor phages varied between recipient conditions with
   the highest engraftment rate, overrepresented by putative temperate
   phage, in patients with rCDI. Consistently, a higher proportion of
   auxiliary metabolic genes (AMGs), with the potential to support and
   modulate bacterial metabolism, were annotated on putative temperate
   phages.

Conclusions

   FMT leads to significant taxonomic, functional, and lifestyle shifts in
   recipient phageome composition. Future FMT studies should include gut
   phageome characterization and consider it as a potential factor in
   microbial community shifts and treatment outcomes.
   Your browser is not supporting the HTML5 <video> element.
   You may download that video as a file and play it with a player of you
   choice.

   [36]Download video stream.
   [37]Download video file^ (141.8MB, mp4)
   [38]Open in a new tab

   Video Abstract

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s40168-025-02046-5.

Introduction

   Fecal microbiota transplantation (FMT) is investigated for the
   treatment of inflammatory conditions of the gastrointestinal tract, but
   the exact way it works and the changes it causes in the microbiome are
   not well understood [[39]1]. In clinical practice, FMT has a nearly 90%
   success rate for treating recurrent Clostridioides difficile infection
   (rCDI) [[40]2]. Additionally, FMT has been investigated for treating
   symptoms or modulating the disease processes in other conditions
   associated with gut microbial imbalance [[41]3, [42]4], including
   multidrug-resistant bacteria (MDRB) infections [[43]5–[44]7], Crohn’s
   disease (CD) [[45]8, [46]9], ulcerative colitis (UC) [[47]5,
   [48]10–[49]12], anti-PD-1 therapy-resistant melanoma [[50]13, [51]14],
   metabolic syndrome [[52]15], obesity [[53]3], tyrosine kinase
   inhibitors (TKI)-related diarrhea [[54]16], and Tourette syndrome
   [[55]17]. Common findings from these studies include that FMT
   facilitates microbiome recovery [[56]18] and enhances microbial ability
   to respond to other treatments [[57]13]. While species-specific
   engraftment patterns in bacteria have been widely characterized
   [[58]16], only a few studies, to date, have investigated viral
   engraftment [[59]19–[60]21].

   Viruses, mostly bacteriophages (phages) that are viruses specific to
   bacteria, constitute a major component of human gut microbiomes. In
   addition to killing their host bacteria, phages mutually coexist with
   their bacterial host while providing them with fitness-promoting
   auxiliary genes. Intriguingly, an FMT study for rCDI using sterile
   fecal filtrate with no bacteria showed equal treatment efficacy when
   the FMT included or lacked bacteria [[61]22], proposing that phages may
   play important roles in FMT treatment efficacy. Supporting this idea, a
   fecal virome transplantation (FVT) from lean donors decreased weight
   gain and normalized blood glucose tolerance in mice with obesity
   [[62]23]. In humans, FVT for patients with metabolic syndrome prompted
   significant changes of phage composition with downstream effects in
   virulent phage-bacteria interactions [[63]24]. While metagenomic data
   contains information about phages, most FMT studies lack evaluation of
   the impact of confounders, especially the sequencing depth, on phage
   profiling and engraftment outcomes. Post-FMT phageome shifts and
   potential effects to metabolic regulation of bacteria remain unclear.

   Here, we metagenomically investigated commonalities and differences of
   gut phageome profiles, phage engraftment, and its implications in gut
   microbial metabolism in a meta-analysis of 23 FMT studies. We found
   that FMT was followed by prominent phageome shifts. Phage engraftment
   outcomes varied according to different patient conditions while
   putative temperate phages tended to be involved in different
   engraftment outcomes after FMT. We further found that putative
   temperate phages might contribute more to regulate bacterial
   metabolism. Functional commonalities in metabolic regulation assisted
   by phage AMGs across studies might implicate in phageome shifts and
   phage engraftment after FMT. This study provides a new perspective to
   explore mechanisms of gut phageome shift and phage engraftment in
   future FMTs and phage-based therapies.

Methods

Datasets

   As of March 8, 2024, a total of 39 clinical FMT studies with
   metagenomic profiling were retrieved using previously established
   search terms: ((fecal microbiota transplant*) OR (fecal microbial
   transplant*) OR (faecal microbiota transplant*) OR (faecal microbial
   transplant*) OR (fecal transplant*) OR (faecal transplant*) OR (stool
   transplant*) OR (fecal microbiota transfer) OR (fecal microbial
   transfer) OR (faecal microbiota transfer) OR (faecal microbial
   transfer) OR (fecal transfer) OR (faecal transfer) OR (stool transfer)
   OR (fecal microbiota donation) OR (fecal microbial donation) OR (faecal
   microbiota donation) OR (faecal microbial donation) OR (fecal donation)
   OR (faecal donation) OR (stool donation) OR (fecal microbiota
   suspension) OR (fecal microbial suspension) OR (faecal microbiota
   suspension) OR (faecal microbial suspension) OR (fecal suspension) OR
   (faecal suspension) OR (fecal microbiota infusion) OR (fecal microbial
   infusion) OR (faecal microbiota infusion) OR (faecal microbial
   infusion) OR (fecal infusion) OR (faecal infusion) OR (stool infusion)
   OR (FMT) OR (bacteriotherapy)) AND ((shotgun) OR (Metagenom*)) [[64]5].
   Seventeen studies were excluded for the following reasons: (1)
   sequencing data and metadata were not publicly available, (2)
   poor-quality sequencing data were deposited without recorded reason, or
   (3) donor samples were not directly from healthy individuals within
   independent cohorts but rather standardized human gut microbiota or
   biobank. The included cohorts were recruited in the USA (n = 671),
   Canada (n = 187), Netherlands (n = 454), Italy (n = 45), France
   (n = 115), Russia (n = 17), Israel (n = 75), China (n = 25) and New
   Zealand (n = 381), and in total included 1970 fecal metagenomic samples
   from healthy donors (n = 342), healthy recipients (n = 14, [[65]25])
   and patients with C. difficile infection (n = 444, [[66]5,
   [67]26–[68]32]), antibiotic resistance or MDRB infections (n = 157,
   [[69]5–[70]7]), CD (n = 156, [[71]8, [72]9]), UC (n = 43, [[73]5,
   [74]10–[75]12]), anti-PD-1 therapy-resistant melanoma (n = 223,
   [[76]13, [77]14]), metabolic syndrome (n = 210, [[78]15]), obesity
   (n = 323, [[79]3]), TKI-related diarrhea (n = 43, [[80]16]), and
   Tourette syndrome (n = 15, [[81]17]). After quality control (trimming
   and decontamination), the average sequencing depth per sample was 5.54
   Gb. The cohorts are described in Supplementary Table 1 and samples in
   Supplementary Table 2.

Quality control

   Trim Galore [[82]33] was used to trim off low-quality reads and
   sequencing adapters. Subsequently, clean reads were filtered to remove
   human genome contamination using KneadData v0.12.0 (–bypass-trf,
   –reference-db hg37dec_v0.1, –remove-intermediate-output). For each
   metagenome, high-quality reads (the number of clean reads is not less
   than 50% of the average of total reads in each dataset/cohort, the
   proportion of Q30 bases is not less than 85%, the proportion of
   non-host reads is not less than 20%, and the proportion of paired reads
   is not less than 70%) were sorted and split into forward, reverse and
   unpaired output files.

Viral identification and abundance

   Both paired and unpaired reads were assembled using MEGAHIT v1.2.8
   [[83]34] (–min-contig-len 500). An earlier Standard Operating Procedure
   (SOP) [[84]35] was modified to identify putative viral contigs.
   Briefly, the VirSorter2 v2.2.4 [[85]36] was used to detect dsDNA and
   ssDNA phages from assembled contigs (–min-length 5000, –min-score 0.5).
   The contamination and completeness were assessed by the CheckV v1.0.1
   using the default parameters. As putative viral contigs, contigs were
   required to contain at least one viral gene identified by CheckV or
   meet the following criteria: (1) host gene count = 0, or (2) VirSorter2
   score ≥ 0.95, or (3) viral hallmark gene count > 2. The viral
   operational taxonomic units (vOTUs) were clustered using MMseqs2
   v15.6f452 [[86]37] (–min-seq-id 0.95, -c 0.85, –cov-mode 1) based on
   Minimum Information about an Uncultivated Virus Genome (MIUViG)
   standards (average nucleotide identity (ANI) ≥ 95% over alignment
   fraction (AF) ≥ 85% of the smaller contig) [[87]38]. Only vOTUs defined
   as complete genomes or those with completeness greater than 50% were
   kept for building an index and mapping sequencing reads. The
   quantification of trimmed vOTUs was performed using coverM v0.7.0.

vOTUs characterization

   Prodigal v2.6.3 [[88]39] (-p meta) was used to predict the open reading
   frames (ORFs) of vOTUs. The trimmed vOTUs were classified into viral
   clusters (VCs) and taxonomic annotated on of family and genus levels
   using vConTACT2 [[89]40]. The lifestyle (putative temperate or
   virulent, labeled as undefined if lifestyle scores were equal) was
   predicted using Bacphlip v1.0 [[90]41] using the default parameters.
   Phage host prediction was performed using iPHoP v1.3.3 [[91]42] with
   the iPHoP_db_Aug23_rw database. Confidence scores of predicted hosts
   were calculated according to five host-based tools (Blast, CRISPR,
   VirHostMatcher, WIsH, and PHP) and a phage-based tool (RaFAH).

AMG annotation and abundance

   The trimmed vOTUs were processed through VirSorter2
   (–seqname-suffix-off, –viral-gene-enrich-off, –provirus-off,
   –prep-for-dramv). Processed sequences were used to annotate AMGs using
   DRAM v1.4.6 [[92]43] (DRAM-v annotate) based on PFAM, dbCAN,
   RefSeq_viral, VOGDB, and MEROPS peptidase databases. A summary of
   potential AMGs was created from the DRAM-v annotations (DRAM-v
   distill). All annotated AMGs were further classified into KEGG modules
   and pathways using the KEGGREST package v1.44.0. To avoid the
   possibility of host-derived reads affecting the abundances of
   viral-encoded AMGs, the abundances of vOTUs were used as
   representations for the abundances of AMGs [[93]44]. The abundances of
   KEGG orthologs (KO) were determined by adding the abundances of the
   AMGs assigned to each KO as previously reported [[94]45].

Classification and quantification of phage engraftment outcomes after FMT

   Phage engraftment outcomes at each time point were determined using an
   allogenic donor sample, a donor-matched recipient pre-FMT baseline
   sample, and the same recipient post-FMT sample. Four studies with
   multiple donors [[95]3, [96]8, [97]32] were excluded from engraftment
   analyses. In total, 551 recipient-donor pairs (n = 1595 metagenomes)
   were included in engraftment analyses. The classification and
   quantification of engraftment outcomes were carried out according to
   [[98]46] with minor modifications. Proportions of engraftment outcomes
   after FMT were quantified based on the prevalence and abundance of each
   vOTU. Detailed information is given in Supplementary Table 3.

Statistical analyses

   Intergroup comparisons of the Shannon index and viral abundance between
   donors and recipients pre-FMT baselines were carried out using unpaired
   Student’s t-test with Welch’s correction or Mann–Whitney test according
   to data distribution. Paired t-test or Wilcoxon matched-pairs signed
   rank test was applied on comparisons between lifestyles, or when
   samples were from the same recipient at different time points.
   Chi-squared test was used for rate comparison. Graphpad Prism v9.0.0
   was used to calculate the statistics above. Other statistical analysis
   and visualization were performed in R v4.1.3. The Bray–Curtis distance
   with normalization was used to generate PCoA ordination plots, which
   were visualized by ggplot2 package v3.3.6 and ggthemes package v4.2.4.
   The marginal histograms on both sides were added using ggExtra package
   v0.10.0. Adonis2 from the vegan package v2.6–4 was used to perform
   permutational multivariate analysis of variance (PERMANOVA) analysis.
   An evaluation of confounders was conducted through SIAMCAT package
   v2.8.0 [[99]47]. Differential analysis with correction of random
   effects were performed by MaAsLin2 package v1.6.0 [[100]48] with a
   threshold of FDR corrected p-value (q-value) < 0.05 (transform = “LOG,”
   normalization = “NONE,” standardize = TRUE, analysis_method = “LM,”
   correction = “BH”). The confounders including sequencing depth
   (measured by the number of post-QC sequencing reads), country and
   cohort (dataset) were included as fixed effects in MaAsLin2 (linear
   mixed models). Any vOTUs that were significantly correlated with these
   fixed effects were considered as false positives and excluded. Linear
   associations between donors’ Shannon index and viral outcomes in
   recipients after FMT were calculated based on the Pearson correlation
   coefficient. An adjustment for multiple comparisons was made using
   Benjamini-Hochberg (BH) correction. The KEGG pathway enrichment
   analysis of bacterial metabolism was performed by MicrobiomeAnalyst
   v2.0 [[101]49] and the significance of the pathway was expressed as
   p-value < 0.05.

Results

Gut phageome meta-analysis of 23 FMT studies

   We analyzed 23 FMT studies, spanning North America (11 from the USA and
   one from Canada), Europe (3 from Netherlands, 2 from Italy, one from
   France and one from the European part of Russia), Asia (2 from Israel
   and one from mainland China) and Oceania (one from New Zealand)
   including 1970 fecal metagenomes from healthy donors and FMT recipients
   across 10 conditions (patients with 9 diseases and healthy volunteers)
   (Fig. [102]1a). Metagenomes were assessed for quality and assembled to
   identify viral contigs (Fig. [103]1b). We detected a total of 103,402
   viral operational taxonomic units (vOTUs), which were further filtered
   according to the Uncultivated Virus Genome (MIUViG) standards to obtain
   29,208 vOTUs with medium-to-high quality and completeness
   (Fig. [104]1c). Only 1690 of filtered vOTUs were classified into at
   least one viral cluster (VC) with known reference genomes of uncultured
   viruses (Fig. [105]1d). Among these classified vOTUs, > 95% of vOTUs
   belonged to families Kyanoviridae (former Myoviridae, 51.95%),
   Drexlerviridae (former Siphoviridae, 31.54%) or Autographiviridae
   (former Podoviridae, 11.72%). (Fig. [106]1e, Supplementary Table 4).
   13,905 (47.61%) of vOTUs we predicted to have a putative temperate
   lifestyle and 15,273 (52.29%) vOTUs were predicted to have a putative
   virulent lifestyle (Fig. [107]1f, Supplementary Table 5). Most of the
   classified vOTUs were predicted to be bacterial hosts
   Gammaproteobacteria, Clostridia, Bacilli, and Bacterodia at the class
   level (Fig. [108]1g, Supplementary Table 6).

Fig. 1.

   [109]Fig. 1
   [110]Open in a new tab

   A meta-analysis of 1970 metagenomes from 23 FMT studies. a Distribution
   of metagenomic samples from clinical FMT studies by region worldwide
   and the number of samples in each patient condition after quality
   control. b Density plots of post-QC sequencing reads, assembled
   contigs, reads mapping to viral contigs, and the number of vOTUs per
   sample. c Proportion of vOTU quality categories according to CheckV. d
   Proportion of vOTUs that were classified into at least one viral
   cluster (VC) or none according to vConTACT2. e) Proportion of vOTUs by
   family level taxonomic assignments. f Density plot of predicted vOTU
   lifestyle scores according to Bacphlip (0: virulent, 1: temperate). g
   Number of classified vOTUs with different predicted bacterial hosts at
   the class level according to iPHoP

FMT lead to increased phageome diversity and shift towards donors

   We first compared phageome diversity, measured by the Shannon index,
   between different recipient conditions and donor groups. FMT recipients
   with rCDI consistently showed lower phageome diversity at baseline
   compared to donors and shifted to higher phageome diversity after FMT
   (Fig. [111]2a). Although we observed similar trends in FMT recipients
   with MDRB infections and IBD, we did not observe statistically
   significant differences between recipients and donors, or shifts
   between pre- and post-FMT diversities in other cohorts (Supplementary
   Table 7). Gut phageome of pre-FMT recipients under different conditions
   was significantly differentiated from donors and recipients after FMT
   (Fig. [112]2b, PERMANOVA based on Bray–Curtis dissimilarities,
   P < 0.001, R^2 = 3%). Especially, samples from rCDI patients clustered
   separately from other FMT recipients (Fig. [113]2c). We observed
   significant differences in beta-diversities within donors (PERMANOVA,
   P < 0.001, R^2 = 36%) and pre-FMT recipients (PERMANOVA, P < 0.001,
   R^2 = 28%) between different cohorts (Supplementary Fig. 1a, b),
   highlighting potential geographical, technical and other confounding
   effects present in these data. Phage profiles significantly shifted
   towards those of their donors in approximately 75.81% and 69.09% of
   recipients with rCDI and MDRB infections, respectively (Fig. [114]2d).
   After correcting for differences in sequencing depth (Supplementary
   Fig. 2) and excluding those vOTUs that were significantly correlated
   with the sequencing depth, country or study cohort, approximately
   23.41% (6837 of 29,208) of vOTUs showed a statistically significant
   difference between donors and recipients (pre- or post-FMT). Among
   these, 93.92% (6,421 of 6,837) of differentially abundant vOTUs
   displayed greater relative abundance in both donors and post-FMT
   recipients compared to recipients at baseline (Fig. [115]2e,
   Supplementary Table 8). For those with taxonomic classification, 24.55%
   (205 of 835) of putative temperate phages and 24.24% (207 of 854) of
   putative virulent ones showed significant differences between donors
   and recipients (pre- or post-FMT) (Fig. [116]2e). This indicates major
   phageome shifts post-FMT and a wide scale donor phage engraftment in
   FMT recipients.

Fig. 2.

   [117]Fig. 2
   [118]Open in a new tab

   Changes in phage diversity and gut phageome profiles after FMT. a
   Comparisons of the Shannon index between FMT donors and recipients
   (pre- and post-FMT) per FMT study. Data are represented as
   median ± interquartile range (IQR). Outliers are marked as black
   points. For statistical comparisons, see Supplementary Table 7. b–c
   PCoA ordination based on Bray–Curtis dissimilarities between phageome
   profiles of samples from all FMT studies. Differences between donor and
   recipient groups (PERMANOVA, P < 0.001, R^2 = 3%) and FMT studies
   (PERMANOVA, P < 0.001, R^2 = 12%) were statistically significant. d
   Phageome shift of FMT recipients per FMT studies. The phageome shift
   was defined as difference between the Bray–Curtis dissimilarity between
   donor and matching pre-FMT recipient and the similarity between donor
   and matching post-FMT recipient (positive values indicate shifts
   towards donors). FMT studies with multiple donors per recipient were
   excluded from this analysis. Data are represented as median ± IQR.
   Outliers are marked as black points. *p < 0.05, **p < 0.01, ***
   p < 0.001, ****p < 0.0001. e Heatmap of relative abundances of phages
   with statistically significant differences between donors and
   recipients (either pre- or post-FMT). The row and column were split
   according to phage lifestyle and sample type, respectively

Phageome engraftment outcomes varied across patient conditions

   We next compared post-FMT samples to samples with their matched donors
   to determine different engraftment and persistence outcomes of phages
   in recipients (Fig. [119]3a). We identified six outcomes for each phage
   in post-FMT recipient metagenomes: colonization (donor phages observed
   in the recipients post-FMT); novel phages (either phages not observed
   in donors and pre-FMT recipients were introduced, or the low-abundance
   phages that had colonized were expanded); coexistence of phages shared
   by donor and recipient; persistence (recipient phages that were not
   detectable in donors and persisted through FMT); rejection (absence of
   detectable engraftment of donor phages); and loss of recipient phages
   after FMT (Fig. [120]3b, Supplementary Table 3). The resilience of
   recipient phages accounted for only 3.51 ± 2.19% to 10.46 ± 2.85%,
   while 37.32 ± 21.04% to 75.69 ± 6.48% of phage outcomes could not be
   determined (Fig. [121]3c). More phages from donors or those from either
   environmental or previously undetected took over phage communities of
   recipients with rCDI after FMT, while novel donor and environmental
   phages accounted for, on average, 16.11 ± 8.2% to 39.37 ± 18.91% of the
   total post-FMT phageome in other recipient groups (Fig. [122]3c). Phage
   engraftment outcomes varied considerably in different conditions
   (Fig. [123]3d). There was a higher rate of engraftment in FMT cases of
   rCDI (colonization and novel phages, accounting for 25.79 ± 12.26% and
   33.37 ± 13.59% post-FMT phages, respectively). In contrast, coexistence
   (24.57 ± 14.15%) was the most common outcome case where phages were
   present in both donors and pre-FMT recipients, mainly in FMT cases of
   healthy volunteers and noninfectious diseases (except rCDI and MDRB
   infections, 31.89 ± 10.08%) (Fig. [124]3d). We observed phage from all
   six engraftment outcome groups in each donor-recipient pair and none of
   the recipients showed complete phage turnover or complete rejection
   after FMT, despite persistent recipient phages or engrafted phages
   being rare (Fig. [125]3d). These results show that phage engraftment
   varies in different diseases. Recipients with infectious diseases, such
   as rCDI or MDRB infections, may experience more phage engraftment after
   FMT.

Fig. 3.

   [126]Fig. 3
   [127]Open in a new tab

   Classification and comparisons of phage engraftment outcomes in FMT.
   a–b Six phage engraftment outcomes in FMT based on the presence or
   absence of each vOTU in donor and recipient samples. c Proportion of
   different summarized scenarios from panel b in each recipient
   condition. **** p < 0.0001. d Proportion of different phage engraftment
   outcomes across FMT studies. FMT studies with multiple donors per
   recipient were excluded from this analysis

More abundant putative temperate phages were associated with engraftment
outcomes

   To determine whether phage lifestyles were associated with engraftment
   outcomes, we compared the phage lifestyles before and after FMT across
   diseases. Overall, putative temperate phages were more abundant
   compared to putative virulent phages in both donors and recipients
   (Fig. [128]4a). Phages belonging to different engraftment outcome
   groups showed relatively even proportions of putative temperate and
   virulent phage with few exceptions (Fig. [129]4b–e, Supplementary
   Fig. 3a). Post-FMT healthy volunteers exhibited a higher proportion of
   putative virulent phages within novel phages with an unknown source
   (Fig. [130]4b) and post-FMT samples from rCDI patients were colonized
   with slightly more putative temperate than virulent donor phages
   (Fig. [131]4c, Supplementary Fig. 3a). Notably, putative temperate
   phages (59.63 ± 5.99%) showed a predominance in phages where the
   colonization outcome was undetermined (i.e., the vOTU was detected in
   the donor and pre- and post-FMT recipient samples) (Fig. [132]4d,
   Supplementary Fig. 3a). Among vOTUs persistent in FMT recipients
   (detected both pre- and post-FMT), putative temperate phages were more
   common among patients with rCDI (Fig. [133]4e, Supplementary Fig. 3a).
   In contrast, we observed more persistent putative virulent phages in
   healthy volunteers and recipients with melanoma or metabolic syndrome
   (Fig. [134]4d). No statistically significant pairwise differences
   between donor and recipient samples emerged within either lifestyle
   before or after FMT when all FMT studies were combined (Fig. [135]4a).
   However, relative abundances of putative temperate phages in
   approximately 85.38% of donors (292 from 14 datasets) were
   significantly higher than putative virulent phages (Supplementary
   Fig. 3b, Supplementary Table 9). Particularly, 80.2% of recipients with
   rCDI (243 from 8 datasets) showed higher relative abundances of
   putative temperate phages post-FMT compared to pre-FMT (51.77%, 141
   from 3 datasets) (Supplementary Fig. 3b, Supplementary Table 9). As an
   exception, healthy volunteers with equal relative abundances of
   putative temperate and virulent phages pre-FMT presented a trend
   towards donors with a relatively higher abundance of putative virulent
   phages post-FMT, albeit the difference was not no statistically
   significant (Supplementary Fig. 3b, Supplementary Table 9). The
   phageome shift within phage lifestyles varied between patient
   conditions and FMT studies (Fig. [136]4f, Supplementary Fig. 3c). Both
   putative temperate and virulent recipient phage profiles shifted
   towards those of their donors in approximately 56% of recipients
   (Supplementary Fig. 3c). On average, both putative temperate and
   virulent phage profiles of recipients with infectious diseases (rCDI
   and MDRB infections) significantly shifted towards those of their
   donors after FMT (95% CI of putative temperate phages: 0.029–0.048 and
   0.005–0.03, respectively; putative virulent phages: 0.034–0.059 and
   0.009–0.04, respectively) (Fig. [137]4f). Taken together, FMT might
   favor colonization of putative temperate phages, especially in patients
   with rCDI, potentially through stable engraftment of their bacterial
   hosts.

Fig. 4.

   [138]Fig. 4
   [139]Open in a new tab

   Post-FMT shifts in temperate and virulent phages vary between patient
   conditions. a Relative abundances of phages per lifestyle between
   donors and recipients (pre- and post-FMT) combining all FMT studies.
   b–e Average proportion of virulent and temperate phages according to
   phage engraftment outcomes, as in Fig. [140]3b: b novel phages, c phage
   colonization, d coexistence, and e persistence. * p < 0.05, **
   p < 0.01, *** p < 0.001, **** p < 0.0001. f Post-FMT phageome shift
   towards donor phageome profiles per patient condition. Data are
   represented as mean with 95% confidence interval (CI)

Phage AMG carriage and diversity differed by lifestyle

   The auxiliary metabolic genes (AMGs) have been found in many phages,
   which can be traced back to bacterial hosts and assist in regulating
   host metabolism [[141]45]. We annotated a total of 11,505 AMGs across
   26.53% (7750 of 29,208) of the vOTUs (Fig. [142]5a). Approximately
   22.82% (2625 of 11,505) and 8.59% (988 of 11,505) of AMGs were assigned
   to the DNA (cytosine-5)-methyltransferase (dcm) and
   S-adenosylmethionine synthetase (metK), respectively (Fig. [143]5b).
   Both of them are involved in the cysteine and methionine metabolism. At
   the KO level, AMGs were enriched in 22 pathways (Fig. [144]5c,
   P < 0.05). Among these, the glycolysis/gluconeogenesis pathway was
   statistically most significant (Fig. [145]5c), implying its potential
   role in providing upstream compounds of amino acid biosynthesis. We
   further asked whether putative temperate and virulent phages differed
   in terms of AMG carriage and whether this had potential implications in
   the metabolism of post-FMT microbiomes. AMG compositions differed
   significantly between putative temperate and virulent phages
   (Fig. [146]5d, PERMANOVA, P < 0.001, R^2 = 3%). In addition, both
   Shannon and richness indexes of AMG profiles were significantly higher
   in putative temperate than virulent phages (Fig. [147]5e, f), implying
   that putative temperate phages harbor more AMGs compared to putative
   virulent phages. Most pathways, especially the cysteine and methionine
   metabolism, harbored a higher proportion of AMGs from putative
   temperate phages compared to putative virulent phages (Fig. [148]5g).
   More abundant and highly diverse AMGs in putative temperate phages
   might contribute more to regulate bacterial metabolism, especially the
   glycolysis/gluconeogenesis pathway and the metabolism of cysteine and
   methionine.

Fig. 5.

   [149]Fig. 5
   [150]Open in a new tab

   Diversity and functional distribution of AMGs in phageome differ
   between phage lifestyles. a Proportion of high-level AMG classes
   according to DRAM. b Number of AMGs per DRAM functional categories. c
   Enrichment analysis of bacterial metabolic pathways based on observed
   AMGs. Pathways with p < 0.05 are shown. d PCoA ordination of
   Bray–Curtis dissimilarities between AMG profiles stratified by phage
   lifestyles (PERMANOVA, P < 0.001, R.^2 = 3%). e–f Comparisons of
   Shannon index (e) and richness (f) of AMGs between phage lifestyles.
   Data are represented as mean ± standard error of the mean (SEM), ****
   p < 0.0001. g Distribution and proportion of phages with different
   lifestyles in differential metabolic pathways of bacterial hosts. *
   p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns p ≥ 0.05

AMGs in phageome may support and regulate gut microbial metabolism after FMT

   We found that cysteine and methionine metabolism genes were enriched
   among AMGs with potential implications throughout the human lifespan.
   For example, early-life fecal cysteine levels are positively associated
   with neurological and physical development [[151]50]. On the other
   hand, AMGs that assist bacterial methionine metabolism in centenarians
   may be related to their longevity [[152]51]. In order to investigate
   these pathways in detail, we combed 33 AMGs related to cysteine and
   methionine metabolism with potential to regulate microbial metabolism
   and their abundance in donors and changes in recipients after FMT
   (Fig. [153]6 and Supplementary Fig. 4). In glycolysis, AMGs encoding
   glucose-6-phosphate isomerase (pgi), fructose-bisphosphate aldolase
   (class II, fbaA), glyceraldehyde 3-phosphate dehydrogenase (gapA),
   phosphoglycerate kinase (pgk), 2,3-bisphosphoglycerate-dependent
   phosphoglycerate mutase (gpmA), pyruvate kinase (pyk) and lactate
   dehydrogenase (ldh) were significantly increased after FMT potentially
   promoting the metabolism of α-D-Glucose 6-phosphate to lactate
   (Supplementary Fig. 5a). Pyruvate produced from glycolysis is
   oxidatively dehydrogenated to form acetyl-CoA, which is further
   metabolized to acetate through phosphate acetyltransferase (pta) and
   acetate kinase (ackA). AMGs encoding these enzymes were all elevated in
   post-FMT samples (Supplementary Fig. 5b). 3-Phospho-D-glycerate derived
   from glycolysis can be interconverted with 3-phosphonooxypyruvate,
   which is an important intermediate in the synthesis of 3-phosphoserine.
   The significantly increased AMG encoding cysteine synthase (cysK) in
   post-FMT samples may promote the conversion of 3-phosphoserine to
   cysteine (Supplementary Fig. 6a). As a precursor of cysteine,
   cystathionine can be metabolized from aspartate and the process may be
   promoted by elevated levels of AMGs encoding homoserine dehydrogenase
   (hom) and homoserine O-acetyltransferase (metA) observed in post-FMT
   samples (Supplementary Fig. 6b). Moreover, the cystathionine can be
   mutually transformed with homocysteine, which is methylated to form
   methionine by the 5-methyltetrahydrofolate-homocysteine
   methyltransferase (metH) and
   5-methyltetrahydropteroyltriglutamate-homocysteine methyltransferase
   (metE). In contrast, the methionine can be converted to
   S-adenosylmethionine (SAMe) through S-adenosylmethionine synthetase
   (metK). After that, the methyl group of SAMe is transferred to a range
   of receptor molecules by DNA (cytosine-5)-methyltransferase 1 (dcm,
   encoded AMG was decreased post-FMT) to generate S-adenosylhomocysteine
   (SAH), which is then hydrolyzed into homocysteine and adenosine by
   adenosylhomocysteinase (ahcY, encoded AMG was decreased post-FMT)
   (Supplementary Fig. 6c). As one of the starting compounds of the
   shikimate pathway, phosphoenolpyruvate from glycolysis might be
   accelerated to shikimate by dehydration, cyclization, and
   dehydrogenation through elevated AMGs encoding 3-dehydroquinate
   dehydratase II (aroQ) and shikimate dehydrogenase (aroE) after FMT
   (Supplementary Fig. 7a). Chorismate, which is metabolized from
   shikimate by the shikimate kinase (aroK/aroL), is a key intermediate in
   the biosynthesis of aromatic amino acids. Among these, tryptophan
   synthesis might be enhanced by the tryptophan synthase (alpha chain,
   trpA; beta chain, trpB; encoded AMGs were increased post-FMT)
   (Supplementary Fig. 7b). These results suggest commonalities in
   metabolic regulation shared among gut phageomes across FMT studies,
   potentially implicated in FMT induced post-FMT phageome shifts.

Fig. 6.

   [154]Fig. 6
   [155]Open in a new tab

   Phage AMG-encoded genetic potential to assist bacterial metabolic
   regulation post-FMT. Pairwise comparisons of AMG abundances between
   donors and recipients are shown in Supplementary Fig. 4–7. Significance
   of AMGs was defined with 95% CI and FDR corrected p-value < 0.05

Discussion

   We investigated gut phageome profiles, phage engraftment, and its
   implications in gut microbial metabolism in a meta-analysis of 23 FMT
   studies with metagenomic data. We observed major phageome shifts and a
   wide-scale donor phage engraftment in FMT recipients. Recipients with
   infectious diseases had a higher rate of donor phage engraftment after
   FMT. Especially in patients with rCDI, FMT led to the colonization of
   putative temperate phages, potentially through stable engraftment of
   their bacterial hosts. We further compared phage AMGs between
   lifestyles and found that AMGs in putative temperate contribute
   functions with the potential to regulate bacterial metabolism. These
   commonalities in metabolic regulation across studies may further
   support persistent gut microbiome shifts potentially implicated in
   clinical outcomes in FMT.

   We observed decreased phage diversity in most FMT recipients at
   baseline, across diseases, compared to healthy donors. More diverse and
   rich gut phageomes from FMT donors promoted the recovery of phage
   diversity in post-FMT phageomes. Previous studies suggested that the
   viral richness and diversity were significantly increased after FMT,
   especially in recipients with rCDI [[156]52, [157]53]. In patients with
   obesity, especially those with type 2 diabetes, the gut phage diversity
   and richness were also significantly reduced [[158]54]. The resolution
   of diarrhea in rCDI recipients after FMT has been associated with
   greater phage diversity [[159]55]. Consistent with our results,
   post-FMT recipients with rCDI tended to have gut phage communities more
   similar to those of donors than recipients at baselines [[160]56,
   [161]57].

   We observed differences in post-FMT engraftment of donor phages across
   different patient conditions. Specifically, we found the highest
   turnover of pre-FMT recipient phages in patients with rCDI accompanied
   by large scale engraftment of recipient phages, consistent with
   previous studies [[162]56, [163]58]. The phages observed in pre-FMT
   samples accounted for the majority of post-FMT gut phageomes in patient
   conditions other than rCDI. Data from fecal viral transplantation and
   phage-bacteria correlations suggest that phages
   play a significant role in clinical outcomes following FMT [[164]59].
   In addition, an open-label clinical trial showed that the effectiveness
   of FMT in autism spectrum disorders was accompanied by engraftment of
   phage communities from the donor [[165]60]. Another recent randomized
   clinical trial using FMT to treat metabolic syndrome found a
   significantly positive correlation between phage diversity and donor
   phage engraftment [[166]47]. Clearly, detailed information on phage
   engraftment, the long-term persistence of engrafted phages, and the
   correlation between engraftment and therapeutic effect need to be
   investigated in future FMT trials [[167]61]. In the longer term, a
   phage-specific modification of FMT may have the potential to address
   donor variability, unstable efficacy, and infection risks.

   We further compared commonalities and differences in phage lifestyles
   across diseases. Uniformly, the relative abundance of putative virulent
   phages was significantly reduced in recipients after FMT, tending to be
   more similar to the donor. An increase in extracellular virulent phages
   was reported as one of the signs of gut dysbiosis [[168]62]. It has
   been demonstrated that virulent phages can weaken the intestinal
   colonization of bacteriome and can coexist for a long time [[169]63].
   In contrast, temperate phages are more likely to introduce new
   functional genes to host genomes naturally via generalized or
   specialized transduction, which favors bacterial evolution [[170]26].
   An earlier study suggested that a significant proportion of FMT
   transfers were from temperate phages, which formed stable associations
   with their hosts [[171]64]. Due to the fact that temperate phages are
   capable of exchanging their genetic material, FMT may transfer
   high-risk genetic information from donor phages to recipients that will
   maintain their host health. This also explains that the symbiosis
   between Bacteroidetes and Firmicutes with temperate and virulent phages
   is an important step toward dysbiosis alleviation [[172]65].
   Expectedly, our meta-analysis showed that more engraftment of putative
   temperate phages was present in most of recipients with rCDI after FMT.
   Nevertheless, about 90% of healthy individuals did not have a dominant
   phage sequence harboring a temperate phage genome marker, suggesting
   that human gut phages might persist for long periods without being
   integrated into their host genomes [[173]66]. Of note, the coexisting
   phages in post-FMT recipients were mainly putative temperate ones,
   which were consistent across conditions. By introducing temperate
   phages into the community, related bacterial hosts may be killed, but
   not the original hosts [[174]67].

   To explore these post-FMT commonalities associated with putative
   temperate phages, we annotated and compared AMGs in phages with
   different lifestyles. AMGs are essential for a wide range of microbial
   functions, such as nutrient metabolism, biofilm formation, bacterial
   motility, and transportation [[175]68]. As expected, higher richness
   and diversity of AMGs were observed in temperate phages. In phage
   evolution, horizontal gene transfer (HGT) is a major driving force, and
   the frequency of HGT varies with different lifestyles [[176]69].
   Temperate phages often experience more HGT than virulent ones, as more
   frequent HGT enables phage genomes to become more functionally diverse
   [[177]70]. Additionally, we found that a higher proportion of AMGs was
   identified from putative temperate rather than putative virulent
   phages. Under gut dysbiosis, a low bacterial diversity could lead to
   low phage diversity, indicating phages adapt strategies that are not
   dependent on their hosts for abundance, possibly favoring the temperate
   lifestyle [[178]61]. Meanwhile, the presence of AMGs in temperate
   phages might enhance the survivability of their bacterial hosts
   [[179]46].

   This meta-analysis outlined some commonalities in phage-assisted
   metabolic regulation of bacterial hosts after FMT. Our results
   supported a hypothesis that FMT might enhance phage AMGs assisted in
   microbial glycolysis and biosynthesis of lactate and acetate across
   diseases. It has been reported that genes involved in the glycolysis
   were associated with more physical activity, which were transmissible
   from exercised donors to recipients after FMT [[180]71]. Lactate, a
   major product derived from carbohydrates, inhibited the reproduction of
   pathogenic bacteria [[181]72]. AMGs involved in microbial short-chain
   fatty acid biosynthesis have been found in low-risk patients undergoing
   allogeneic stem cell transplants, which may assist in providing
   protective immunomodulatory metabolites during FMT [[182]73]. Acetate,
   another potential beneficial factor with immunomodulatory effects, was
   reduced in chronic disease states and was significantly elevated by
   FMTs [[183]74, [184]75]. Moreover, we speculate that phage AMGs might
   assist in regulating bacterial DNA methyltransferase. SAMe might
   transfer less methyl group under the action of DNA
   (cytosine-5)-methyltransferase that was significantly reduced after
   FMT. Previous studies have shown that the bacterial DNA
   methyltransferase is a transcriptional regulator that enables bacteria
   to adapt more quickly to environmental changes, for example, to resist
   antibiotics [[185]76]. In addition, FMT promoted AMGs-assisted
   microbial biosynthesis of shikimate and tryptophan. Numerous studies
   have demonstrated that tryptophan and its derivatives play important
   roles in maintaining gut health and alleviating inflammation [[186]77].
   The shikimate pathway that produces aromatic amino acids (mainly
   including tyrosine, phenylalanine, and tryptophan) in microbes has been
   identified as a highly conserved metabolic module in the human gut
   [[187]78]. Interestingly, gut phages from donors have the potential to
   regulate microbial shikimate pathways through FMT, although the
   detailed mechanism is currently unexplained. Overall, we hypothesize
   that engrafting gut phages after FMT may assist in the recovery of
   recipient microbial communities perturbed by antibiotics and disease
   factors through promoting microbial glycolysis and the biosynthesis of
   lactate, acetate, and aromatic amino acids, and regulating microbial
   environmental adaptability associated with DNA methylation. These
   results highlight the importance of phages in the gut ecosystem and
   provide a novel perspective for future FMT/FVT research to investigate
   underlying mechanisms.

Limitations

   This study merits further consideration in the following aspects.
   Despite our efforts to minimize the effects of confounders and batch
   effects, there were differences in sample handling, DNA extraction and
   other laboratory procedures between the FMT studies that were difficult
   to account for due to other systematic biases between studies. For
   example, some FMT studies derive bacteria supplemented in FMT through
   centrifugation which could deplete bacteriophage quantities [[188]79,
   [189]80]. Differences in bioinformatic algorithms, databases, and
   standard operating procedures (SOPs) for phage genome assembly,
   annotation, and lifestyle prediction may introduce biases and
   variability between studies, potentially affecting the comparability of
   results. Phage identification and lifestyle prediction based on
   incomplete or fragmented phage genomes derived from bulk metagenomes
   may lead to an underestimation of temperate phages, as hallmark
   lysogenic genes required for accurate lifestyle predictions may be
   missing. Many patient conditions had a limited sample size and lack of
   metadata leading to limited statistical power and lack of findings.
   Longitudinal statistical comparisons in each dataset, condition, or
   sample type were not performed due to the lack of a uniform follow-up
   scheme and protocol in different diseases.

Supplementary Information

   [190]40168_2025_2046_MOESM1_ESM.pdf^ (566.1KB, pdf)

   Supplementary Material 1: Supplementary Fig. 1. PCoA plots of
   Bray–Curtis dissimilarity between phageome profiles of donors (a,
   PERMANOVA, P < 0.001, R^2 = 36%) and recipients (b, PERMANOVA,
   P < 0.001, R^2 = 28%) in each dataset.
   [191]40168_2025_2046_MOESM2_ESM.pdf^ (129.8KB, pdf)

   Supplementary Material 2: Supplementary Fig. 2. Evaluation of
   confounding covariates (post-QC sequencing read count, country and FMT
   study ID) on an inter-group comparison between donors and recipients at
   baseline according to SIAMCAT.
   [192]40168_2025_2046_MOESM3_ESM.pdf^ (539.2KB, pdf)

   Supplementary Material 3: Supplementary Fig. 3. a) Proportions of
   temperate and virulent phages in different phage engraftment outcomes
   in each recipient. b) Comparisons of phage relative abundance between
   lifestyles. Data are represented as median ± IQR. Outliers are marked
   as black points. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
   c) Phageome shift towards donor phageome profiles in different patient
   conditions and FMT studies. Data from studies with multiple donors per
   recipient [[193]3, [194]8, [195]29] were excluded from this graph. Data
   are represented as median ± IQR. Outliers are marked as black points.
   [196]40168_2025_2046_MOESM4_ESM.pdf^ (497.6KB, pdf)

   Supplementary Material 4: Supplementary Fig. 4. Phage AMG-encoded
   genetic potential to assist bacterial metabolic regulation post-FMT.
   [197]40168_2025_2046_MOESM5_ESM.pdf^ (1,013.2KB, pdf)

   Supplementary Material 5: Supplementary Fig. 5. Pairwise comparisons of
   AMG relative abundances involved in glycolysis (a) and pyruvate
   metabolism (b) of bacterial hosts between donors and recipients (pre-
   and post-FMT).
   [198]40168_2025_2046_MOESM6_ESM.pdf^ (1.1MB, pdf)

   Supplementary Material 6: Supplementary Fig. 6. Pairwise comparisons of
   AMG relative abundances involved in cysteine biosynthesis (a),
   methionine biosynthesis (b) and methionine degradation (c) of bacterial
   hosts between donors and recipients (pre- and post-FMT).
   [199]40168_2025_2046_MOESM7_ESM.pdf^ (877.6KB, pdf)

   Supplementary Material 7: Supplementary Fig. 7. Pairwise comparisons of
   AMG relative abundances involved in shikimate pathway (a) and
   tryptophan biosynthesis (b) of bacterial hosts between donors and
   recipients (pre- and post-FMT).
   [200]Supplementary Material 8.^ (2MB, xlsx)

Acknowledgements