Introduction

Anorexia nervosa (AN) is a complex disorder characterized bydisturbances in the way one’s body weight or shape is experienced, difficulty in gaining weight, restrictive food intake, and depression and anxiety. The evidence base for treatment is still very limited and AN carries one of the highest fatality rates amongst psychiatric disorders1. Fewer than half of treated adults with AN achieve recovery2. In fact, there is little evidence that the prognosis for AN in adults has improved over the last decades3. So far, attempts at pharmacologic treatment have focused on a narrow range of compounds and only a few controlled studies have been completed4. Furthermore, large placebo-controlled trials have proven to be very difficult to conduct in individuals with AN because of its egosyntonic nature, low treatment acceptability, and low participant retention5.

In some cases AN is preceded by depression, trauma, gastrointestinal symptoms or infection. However, its etiology is multifactorial with both genetic and environmental factors playing a role6. The role of genetic factors is well-established7,8. Further, AN is associated with multiple endocrine alterations9, and it has been proposed that AN should be reframed as a metabo-psychiatric disorder10.

The human digestive tract harbors complex microbial communities, with the gut microbiota (GM) playing key roles in host metabolism. The GM aids nutrient and energy acquisition, synthesizes amino acids, vitamins, and metabolites, and drives enteroendocrine hormone signaling. These processes and GM-host interactions support homeostasis, affecting immunity, pathogen defense, gastrointestinal function, metabolism, appetite, emotions via the gut-brain axis, and functions beyond the gastrointestinal tract11,12.

Growing evidence shows that in the active phase of illness, individuals with AN have a dysbiotic GM, with lower microbial diversity and reduced stool levels of serotonin, GABA, dopamine, butyrate, and acetate compared to healthy controls (HCs)13,14,15. A multi-omics study uncovered profound and complex disruptions of the GM and altered serum metabolites in individuals with active AN15. While constipation and dyspepsia stem from neurologic and hormonal adaptations to starvation16,17, returning to a normal diet with more complex foods, like fiber, does not fully reverse dysbiosis in AN18.

Two recent reviews on GM in individuals with AN highlight overlapping findings. A 2021 review (n = 9) identified reduced butyrate-producing taxa, particularly Roseburia, and increased levels of Parabacteroides, Alistipes, and mucin-degrading bacteria like Akkermansia in AN patients19. A 2024 review (n = 14) similarly found lower abundances of Roseburia inulinivorans and Faecalibacterium prausnitzii, key butyrate producers linked to healthy GM, and higher levels of Methanobrevibacter smithii in AN20. F. prausnitzii, an anti-inflammatory bacterium, and a potent butyrate producer associated with mental health, is notably reduced in AN21. Increased Methanobrevibacter smithii may enhance energy extraction and contribute to symptoms like constipation and gas22.

Fecal microbiota transplantation (FMT) involves the transfer of fecal matter (the GM including all metabolites, hormones and more) from chosen donors to recipients with GM dysbiosis. Two single human case studies have been published where FMT was used as experimental treatment of AN23,24. Both studies suggest that FMT can overcome persistent dysbiosis, metabolic, and clinical symptoms if implemented as an adjunctive to treatment as usual. FMT may “reset” gut-brain signaling, boost appetite, and relieve gastrointestinal symptoms like constipation caused by the lack of essential GM-derived molecules and enzymes needed for complex food digestion.

During the last decade FMT has become well established as a highly effective routine treatment for recurrent Clostridioides difficile gut infections (rCDI) in many countries25. FMT trials for recurrent C. difficile infection (rCDI) have demonstrated over 90% cure rate, regardless of the preparation method or administration route26. However, our recent findings indicate that capsules were more effective than both FMT enemas and rectal bacteriotherapy27. Despite the non-specific nature of an FMT, it has shown remarkably few mild side effects and a very low rate of serious adverse events28.

This study aimed to assess the feasibility of using FMT to restore the GM in adult females with AN, exploring a single FMT administered orally or rectally based on participant preference, without pretreatment GM clearing. We used metagenomic sequencing to analyze GM changes before and one week after FMT, assessing shifts toward the donor’s GM, including diversity, species transfer, and functional gene pathways. We examined whether a single FMT could influence serum levels of sex-, appetite-, and metabolism-related hormones. We examined gut microbial β-glucuronidase (gmGUS), which can reactivate conjugated hormones excreted in the gut like estrogen, influencing metabolism. Its reduction in AN may impair estrogen reabsorption and contribute to symptoms10,29.

We also hypothesized that individuals with AN have low fecal gmGUS levels, which might be restored by FMT. Furthermore, we assessed participants self-reported stool consistency and transit time using the Bristol Stool Form Scale (BSFS)30 and eating disorder symptomatology. Participant evaluations and observations were collected to provide qualitative insights and guide future study designs.

Here, we show that a single oral or rectal FMT in adult women with AN is feasible, well-tolerated, and leads to shifts in GM composition toward donor profiles, along with improved stool consistency, while producing no significant changes in appetite- sex-related hormones or gmGUS.

Results

Feasibility

Over a nine-month recruitment period in 2023, 69 individuals expressed interest in participating. Nineteen were excluded during pre-interview screening due to age (n = 4), sex (n = 3), or BMI outside our criteria (n = 12). Of the 50 who completed screening, 28 were further excluded based on our inclusion criteria, most commonly due to BMI (n = 20), laxative abuse (n = 7), or recent probiotic use (n = 2); several individuals met more than one exclusion criterion. Ultimately, 22 females with AN were enrolled and provided informed consent, with 18 (81%) completing all study procedures. These numbers suggest overall feasibility, though the strict inclusion criteria limited final enrollment. The dropouts either failed to deliver the second fecal sample or complete the 1 month follow-up questionnaire (Supplementary Fig. 1).

No serious FMT-related adverse effects were reported. However, one participant with a history of depression was hospitalized for severe depression post-FMT, underscoring the need for careful monitoring in vulnerable populations. For the administration route 19 chose oral FMT with capsules and only 3 chose the rectal route. None had problems swallowing the frozen capsules within the required time. Overall, most participants were young students recruited from Odense or Ballerup hospitals or through social media with a mean age of 24.5 ( ± 5.7 years) and mean BMI of 16.1 (±1.9).

Psychopathology measures of quality of life (EDQLS), depression (MDI) eating disorder symptoms (EDE-Q) and participation data

There were no significant differences in psychopathology measures of eating disorder symptoms (EDE-Q), depression (MDI), or disease-specific quality of life (EDQLS), as tested by paired t-tests (Supplementary data 1). The small sample size limits detection of associations with donor, closer-to-donor-status, or clinical variables. All 19 participants completed the FMT feasibility questionnaire one week post-FMT. Among capsule recipients (16/19), none reported disgust, found swallowing acceptable, and preferred capsules for future use. They considered a 6-week regimen reasonable and 30–50 capsules per FMT acceptable. Most would adhere to treatment if no alternatives existed. However, the three enema recipients (3/19) remained uncomfortable with swallowing capsules.

Gut Transit Time

Based on self-reported stool type using the BSFS at the time of fecal sampling—one day prior to FMT and one week following it—seven participants reported no change in stool type, while 11 participants reported looser stools. (Wilcoxon test P < 0.01; Fig. 1).

Fig. 1: Bristol Stool Form Scale scores.
Fig. 1: Bristol Stool Form Scale scores.The alternative text for this image may have been generated using AI.
Full size image

Change in Bristol Stool Form Scale scores from pre-FMT (w0) to 1 week post-FMT (w1), colored by donor (Donor 1 = blue, Donor 2 = green). Each line represents an individual participant (n = 18), showing the direction and magnitude of changes. A significant decrease in stool consistency is observed (Wilcoxon test **p = 0.006). Scores range from 1 (hard lumps) to 7 (Entirely liquid), with lower scores indicating firmer stools and longer transit times.

Fecal moisture content

To validate the BSFS scores and analyse differences stratified on BSFS response, we measured the fecal moisture content in the pre-post-FMT samples. A Spearman’s rank correlation test showed a strong positive association between fecal moisture content and self-reported BSFS scores both before (rho = 0.74, p = 0.0017) and after FMT (rho = 0.88, p < 0.0001). Overall, no significant difference in fecal moisture content was seen comparing pre-FMT (68.45 ± 12.02%) and post-FMT (69.32 ± 12.29%) (paired t-test: p = 0.55) regardless of BSFS scores. Pre-FMT, the fecal moisture content ranged from 53.67% to 96.22% and post-FMT from 47.03% to 89.64%. Fecal moisture content did not change significantly after FMT in participants, regardless of stool consistency (when based on BSFS response) or microbiota closer-to-donor status (when based on Bray-Curti’s dissimilarity metrics) (See Supplementary Figs. and Methods).

Microbiome analysis

The GM diversity of the FMT donations (n = 13) from two distinct donors (Don1: n = 3; Don2: n = 10) and of participants (n = 18) pre- and post-FMT is illustrated in Fig. 2A, using both the Chao1 and Fig. 2B Shannon indices. Individual participants’ pre- (w0) and post- (w1) FMT are shown connected with lines. The Chao1 (total richness) index of both donors was significantly different from that of their eventual recipients. In particular Don2 recipients also showed a significant difference in diversity between pre-FMT (w0) and post-FMT (w1), while the Shannon diversity (evenness) index no statistically significant changes were observed between pre- and post-FMT. Using a Distance-based Redundancy Analysis (dbRDA) model FMT-recipients experience a clear shift toward donor GM composition post-FMT (Fig. 2C, p = 0.001).

Fig. 2: Overview of microbial diversity and changes associated with FMT.
Fig. 2: Overview of microbial diversity and changes associated with FMT.The alternative text for this image may have been generated using AI.
Full size image

A–F Overview of microbial diversity and compositional changes associated with FMT. A, B Alpha diversity of donor samples and participant microbiota pre- (w0) and post (w1)-FMT measured using Chao1 richness (A) and Shannon evenness (B) indices, stratified by donor source (Don1 = blue (n = 3), Don2 = green (n = 10)) using 1 biological sample for each. Donor versus w0 is tested by one-sample t-test towards the mean of the donor, while change from w0 to w1 is by mixed linear model with patient and time as factors. Both tests are two-sided. Significant changes are indicated (*p < 0.05, ***p < 0.001). Actual p values are: 2a: p(Donor1 vs w0) = 0.020, p(Donor1, w0 vs w1) = 0.153, p(Donor2 vs w0) = 0.020, p(Donor2, w0 vs w1) = 0.0317. 2b: p(Donor1 vs w0) = 0.058, p(Donor1, w0 vs w1) = 0.602, p(Donor2 vs w0) = 0.010, p(Donor2, w0 vs w1) = 0.0500, 2d: p = 4.1e-04. C Beta diversity visualized by dbRDA plot, showing Bray-Curtis dissimilarities of microbial communities. Points are colored by donor and timepoint (Purple=w0 and dark blue (w1_don1) or dark green (w1_don2)=w1), illustrating shifts (n = 18) toward donor profiles after FMT. D Distance to donor microbiota pre- (w0) and post (w1)-FMT, stratified by donor (Don1 = blue, Don2 = green). We used a linear mixed model with patient as random effect, and time as systematic factor. A significant reduction in distance post-FMT is observed (***p < 0.001). E Volcano plots of differentially abundant taxa between donor and participant samples, highlighting key families associated with each donor (e.g., Lachnospiraceae and Bacteroidaceae for Don1, Christensenellaceae and Eggerthellaceae for Don2). Vulcano-plot based on DEseq2 parametric differential abundance testing with wald-test for specific contrasts. F Differentially abundant microbial pathways post-FMT, showing changes in pathway categories across donor groups based on log2 fold change. Boxplots of log-fold change as function of inference (p value categories). Box from 25–75 percentile (IQR), line in box is median, whiskers are from min to max, and points reflect outliers defined as 1.5*IQR above (below) 75 (25) percentiles.

To assess whether the FMT treatment effectively shifted the recipients’ GM composition toward that of the donor, we calculated the distance to the donor both pre- and post-FMT. Prior to FMT, the average distance between recipients and donors was 0.65 (interquartile range [0.65; 0.69]), indicating a notable distinction between their microbiota compositions (Fig. 2D). Following FMT, this distance significantly decreased for all but two individuals, reaching an average of 0.59 [0.53; 0.64] (Wilcoxon test, P < 0.0001), demonstrating a substantial shift in recipients’ microbiota composition toward that of the donor. To further assess donor specificity, we also calculated distances to the ‘other’ donor and found no significant change from baseline to w1 (p = 0.46). Additionally, recipients’ w1 profiles were significantly closer to their assigned donor than to the ‘other’ donor (p < 0.001), supporting a donor-specific shift (see Supplementary Figs. 9 and 10).

To assess GM compositional changes from pre- to post-FMT differential abundance analysis was conducted separately for each donor due to their distinct microbial profiles. Figure 2E shows a volcano plot, where positive effect sizes indicate taxa (family level) enriched by FMT, and negative values indicate suppression. Figure 2F recasts these results, displaying effect sizes relative to significance. Together, these plots suggest FMT generally enriches a range of taxa without depleting or suppressing specific ones. Most enriched taxa are abundant in donor samples, except Ruminococcaceae at low abundance and Christensenellaceae, that were significantly increased in Don2 recipients, despite being absent in Don2 samples.

FMT-closer-to-donor status was defined by evaluating the changes in IQR (Interquartile Range) Bray-Curtis dissimilarity between donor and recipient pre- and post-FMT. Closer-to-donors were participants with a reduction of at least 1.5 standard deviations in Bray-Curtis dissimilarity metrics, classifying recipients as closer-to-donor (n = 9) and not-closer-to-donor (n = 9). All three participants receiving FMT via enema were not-closer-to-donor. Among Don1 recipients (n = 4), 2 were closer-to-donor and 2 were not-closer-to-donor.

We also explored potential microbial predictors of not-closer-to-donor status before FMT. To achieve this, we applied a sparse partial least squares discriminant analysis (sPLS-DA) model to baseline stool samples to identify bacterial families associated with FMT response. The regression coefficients from this model (Fig. 3) highlight bacterial families associated with either an increased or decreased likelihood of FMT success. Positive coefficients identify families promoting closer-to-donor status, including some Lachnospiraceae, Bacteroidaceae, and UBA1381. Conversely, negative coefficients highlight families such as Coriobacteriaceae, Oscillospiraceae other Lachnospiraceae and Rikenellaceae, which are associated with reduced response likelihood.

Fig. 3: Predicting closer-to-donor status: Figure shows the regression coefficients from a sparse partial least squares discriminant analysis model predicting closer-to-donor status (yes/no) based on baseline stool composition prior to FMT.
Fig. 3: Predicting closer-to-donor status: Figure shows the regression coefficients from a sparse partial least squares discriminant analysis model predicting closer-to-donor status (yes/no) based on baseline stool composition prior to FMT.The alternative text for this image may have been generated using AI.
Full size image

Positive coefficients indicate bacterial families associated with an increased probability of being a closer-to-donor, while negative coefficients represent families linked to reduced likelihood of response. Colors denote family-level taxonomy, with key taxa highlighted.

Furthermore, we performed a species-level analysis to pinpoint key differences between the two FMT donors and their respective recipients, both pre- and post-FMT, offering exploratory insights into the microbial factors involved in the FMT treatment. A qualitative comparison between healthy donors (Don1 (n = 3) and Don2 (n = 10)) revealed notable differences in 34 taxa, including significant higher abundances of Catenibacterium misuokai and Methanobrevibacter smithii in Don2, and Ruminococcus bicirculans and Bifidobacterium longum in Don1 (Supplementary Fig. 2).

We further analyzed the specific species level differences between donors and eventual recipients before FMT. Notably, Don1 had a higher abundance of Prevotella copri, Ruminococcus callidus, and Coprococcus eutactus, while the Don1 recipients showed higher abundances of Fecalibacillus intestinalis and a Ruminococcus species sp003526955 (Supplementary Fig. 3) pre-FMT. For Don2, Catenibacterium misuokai was markedly more abundant compared to eventual recipients. The recipients had higher abundances of the same Ruminococcus sp003526955 and Bifidobacterium longum compared to Don2 (Supplementary Fig. 4).

For Don1-recipients (n = 4), a comparison between baseline stool samples (w0) and those collected one week after FMT (w1) revealed a significant increase in the abundance of three taxa: Ruminococcus callidus, Prevotella copri, and Coprococcus eutactus following FMT. (Supplementary Fig. 5). For Don2 recipients (n = 14) the most prominent significant change from pre- to post-FMT, was a higher abundance in Catenibacterium mitsuokai (out of 10 significantly increased taxa) post-FMT. Two taxa CAG-41 sp900066215 (Firmicutes) and Phocaeicola dorei (Bacteroides) were at a lower abundance (Supplementary Fig. 6).

The final analysis of changes in individual taxa from pre-FMT (w0) to post-FMT (w1), we focused on identifying bacteria that had 1: higher abundance post-FMT, 2: were absent in the recipient before FMT, and 3: were present in the corresponding donor. This was to investigate possible direct transfer of species from donor to recipient. Following FMT, we observed a moderate transfer of 11 species with significant changes in abundance that were originally present in the donors but absent in the recipients pre-FMT. These species include Catenibacterium misuokai, Alloprevotella sp900541575, Anaerotardibacter sp900553605, CAG-177 sp003538135 (Ruminococcus), Cryptobacteroides sp000434935, Duodenibacillus massiliensis, Faecousia sp900546415, Phocaeicola sp000434735, UBA1829 sp900548385, UMGS1502 sp900552365 (Eggerthellaceae), and [UMGS1590]_unknown (Peptococcaceae). See (Supplementary Fig. 7) for further details.

Blood samples and gut microbial β-glucuronidase

We assessed the effects of FMT on sex hormones (estradiol, FSH, and LH), appetite-regulating hormones (GLP-1, insulin, glucagon, C-peptide, PYY, GiP, and leptin) using blood samples, and gmGUS. Figure 4 depicts the pre- (w0) and post-FMT (w1) changes for each biomarker, stratified by microbiota-derived closer-to-donor status. While biomarker responses were highly individual, no clear patterns linked to closer-to-donor status emerged. Notable changes included a two-fold reduction in estradiol levels across participants, though this did not reach statistical significance (OR = 0.51, CI: 0.26–1.01, p = 0.054). Insulin levels showed a modest 10% increase overall (OR = 1.1, CI: 1.00–1.21, p = 0.047), while PYY levels in closer-to-donors exhibited a small but significant increase (OR = 1.22, CI: 1.02–1.46, p = 0.037).Exploratory analyses showed that most metabolites and hormones were unrelated to self-reported BSFS scores, except blood estradiol, which modestly correlated with BSFS at week 1 (Supplementary Table 1). No significant changes pre- to post-FMT in GmGUS were observed (Fig. 4).

Fig. 4: Explored biomarkers.
Fig. 4: Explored biomarkers.The alternative text for this image may have been generated using AI.
Full size image

Scatter plots showing correlations between measured variables at week 0 (w0) and week 1 (w1) for closer-to-donor (blue circles) and not-closer-to-donor (green triangles). A Correlations for metabolic markers, including C-peptide, GIP, GLP-1, Glucagon, Insulin, Leptin, and PYY, with corresponding Pearson correlation coefficients (R). B Correlations for hormonal markers, including Estradiol (Estra), FSH, and LH. C Correlation for Glucuronidase activity. The pink line represents the linear regression fit for the data points.

Metagenome gene pathway KEGG modules and single gene analysis

The KEGG modules analysis was conducted to investigate the changes to metabolic pathways and functional gene pathways present within the GM pre- to post-FMT. Overall, we see no changes to module relative abundances pre- to post-FMT. For more details see Supplementary Fig. 3. We further performed a single-gene analysis to assess functional gene changes in the GM pre- and post-FMT. While the total annotated gene count remained stable, constrained correspondence analysis revealed a significant reduction in distance to Don2 post-FMT for Don2 recipients (p = 0.01), indicating greater similarity to the donor’s gene composition (Fig. 5). Don1 post-FMT distance to Don1 recipients was not significantly different from pre-FMT (P > 0.05).

Fig. 5: Annotated gene shifts related to FMT.
Fig. 5: Annotated gene shifts related to FMT.The alternative text for this image may have been generated using AI.
Full size image

A PCoA plot with Bray-Curtis dissimilarity of annotated genes across sample groups related to Don2, analyzed using constrained correspondence analysis (CCA). Sample points are color-coded: donor samples (orange), pre-FMT (green), and post-FMT (purple), with dashed lines enclosing each group. Axes represent principal coordinates CCA1 and CCA2, explaining 79.5% and 20.5% of fitted variance (8% and 2.1% total variance). Omnibus CCA statistics indicate a significant model fit (p = 0.01, Pseudo-F = 1.97, R² = 10.1).

DiscussionOral FMT, delivered as a single and low-cost intervention, demonstrated feasibility and safety in adults with AN, showing significant shifts in GM composition toward that of the donor one week after treatment with no reported adverse effects. Most participants preferred oral FMT, completed all study requirements, and were satisfied with the administration method; those using capsules found them easy to swallow and were open to future doses. Recruiting 22 subjects in 9 months suggests that any larger RCT will likely need to include multiple centers to achieve adequate recruitment. As this was an open-label study, we cannot rule out the potential influence of a placebo effect, which may confound interpretation of subjective outcomes. Production, freezing, and administration of FMT capsules at the hospital, as well as home stool collection, were feasible and well-tolerated, with high participant compliance and no major logistical issues. If future trials involve home administration, considerations should include tamper-resistant placebo capsule design, secure cold storage, and measures to reduce capsule comparison, such as staggered distribution. Placebo content also needs careful planning, as individuals with AN may be sensitive to perceived caloric content; glycerol could be used as a minimal filler in both FMT and placebo capsules. Analysis of GM shifts revealed that the GM of recipients became more closer to GM of donor within one week, demonstrating the feasibility and impact of FMT in individuals with AN without prior GM manipulation. We observed a significant change in the self-reported BSFS scores a proxy of gut transit time (TT), based on self-reported stool consistency, from pre- to post-FMT. Although fecal moisture content (MC) significantly correlated with the BSFS scores, there were no significant changes in MC overall or based on BSFS response. It is a limitation that BSFS is on self reported data and MC was done on once frozen samples. This is noteworthy, as GI issues like constipation, gas, and bloating are likely to contribute to the negative feedback loop in AN31. If a single FMT can improve TT, this warrants further investigation. Future studies should measure stool energy content, such as through bomb calorimetry along with TT, to determine if decreased TT leads to energy loss, especially in these sensitive patients. Early findings without TT measurements do show that energy content in stool increases after inpatient refeeding32. We did observe a modest correlation between estradiol levels and BSFS scores pre-post FMT, suggesting that TT may possibly be influenced by menstrual cycle–related or other fluctuations in estradiol levels. In healthy individuals, stool energy density is associated with transit time and microbial enterotypes33, and microbial richness shows a strong correlation with colonic TT34.

As a feasibility study, this trial was not designed to assess significant changes in psychopathology measures such as eating disorder symptoms, depression, or quality of life. The lack of observed changes does not necessarily indicate that a single FMT is ineffective but rather reflects the study’s limited follow-up period, sample size, and single-treatment design. Given the complexity of AN, future research should explore longer follow-up periods and multiple FMT sessions to better evaluate potential psychological effects. We chose a one-month follow-up for the psychometric measures to reduce practice effects, as shorter intervals could bias responses and may not capture clinically relevant changes. The limited diversity in our samples restricts the generalizability of our findings. Future studies should include participants of different ages, ethnicities, socioeconomic backgrounds, and sexes to better reflect the full spectrum of AN. Excluding postmenopausal individuals, those with somatic comorbidities, and younger or older age groups leaves gaps in understanding the diverse presentations of AN. Future recruitment should broaden inclusion criteria, reconsider age restrictions, and allow select comorbidities with proper supervision and power calculations, ensuring more representative sampling and improving the external validity of FMT trials for AN. In this study, we did not pre-treat the FMT recipients with antibiotics or colonic irrigation, which is a common procedure for FMT against rCDI. Considering that AN patients already suffer from a disorder and have low body weight, pre-treatment might be unwise. Consequently, in our study, we are transplanting a presumed fully functional donor GM into a pre-existing, presumed dysbiotic recipient GM without any functional prescreening of donor or recipient. Since pre-treating is believed to enhance the engraftment of the donor species in both mice and humans35,36,37, this is still worth considering for future studies. In addition to assessing FMT feasibility, we analyzed GM changes one week post-FMT and evaluated blood and stool biomarkers. Sampling fecal matter and blood was done with a one-day gap due to the expected participant’s inability to provide a fecal sample at the hospital. As a result, fecal and blood measurements are not directly comparable. This should be considered in future metabolomics studies of FMT interventions. Despite the small sample size, this study aimed to explore any biological effects as proof of principle for future placebo-controlled trials.

Donors and recipients did show differences in GM composition prior to FMT, and one week after the procedure, recipients’ GM shifted significantly closer to that of the donors. This provides proof-of-principle for using FMT without prior GM manipulation in individuals with AN. However, without further follow-up, it remains unclear how long this shift is sustained and whether it has lasting clinical relevance. Additional time points at later stages are needed to evaluate the sustainability of this shift, account for variations in GM samples, and consider potential background drift over the week. Due to limitations of MAG-based profiling, we avoid conclusions about engraftment or definitive presence/absence of specific taxa. According to a previously reported single-case study23 GM shift did not persist long-term after transplantation in a single AN patient. In our study, we utilized two donors donating 13 different samples in total, used for our 22 FMTs, which is not ideal and presents a limitation for the statistical analysis and interpretation of the results. Future FMT trials sould balancedonor selection and preparation by prioritizing diversity over cost, using multiple donors with limited donations per donor, assigning each donation to one patient, or prioritizing streamlining inocula for purposeful heterogeneity. There are both advantages and disadvantages to using single-donor-donation per FMT38. An alternative approach to ours, as suggested by a study protocol39, is to have multiple donors contribute equally to each FMT bolus. Beyond ethical and safety concerns, we chose single-donor FMTs to preserve the transfer of a complete, functional microbial ecosystem. However, donor selection—including age, sex, and overall compatibility—likely plays a crucial role. Future research should compare different donor selection strategies and recipient matching, particularly for conditions beyond rCDI, to optimize FMT effectiveness. The safety and tolerability of repeated FMT capsule treatments in AN remain unknown and should be evaluated in future studies involving repeated dosing.

In reviewing the literature on GM in AN in context to our results several considerations emerge. Limitations include the substantial impact of both long-term and short-term dietary intake on GM composition40,41. Previous studies have largely relied on qPCR and 16S rRNA gene amplicon sequencing techniques, while our study utilizes full-genome shotgun metagenomics for a more detailed analysis. However, our study is not intended to compare AN patients with HCs; rather, it focuses on individuals with AN following a single FMT using a lower inoculum than typical protocols, observed over one week, with minimal background GM drift. In our study, we analyzed and compared our two FMT donors and identified 34 significantly different taxa between them. Of presumed AN relevance, Don2 showed notably higher abundances of Methanobrevibacter smithii, as well as species of Prevotella and Blautia (Lachnospiraceae). Additionally, Catenibacterium mitsuokai (not associated with AN) was highly abundant in Don2 and appeared at high levels in Don2 recipients following FMT, despite being initially absent pre-FMT. In contrast, Don1 exhibited higher levels of Alistipes putredinis and Gemmiger. A. putredinis has been associated with AN and linked through bacterial genetics to metabolism-related bioclinical variables, while Gemmiger was observed at higher abundances in HCs compared to AN patients42.

We analyzed GM differences between donors and their respective recipients, identifying differences relevant to AN. In our study Don1 had lower abundances of Anaerostipes (Lachnospiraceae), a butyrate producer43 compared to eventual recipients. In other studies, Anaerostipes was found to be associate with changes in IL-15 levels in adolescents with AN44 and were significantly increased at admission compared to HC and was further increased at discharge by 50%45. This aligns with findings in adolescents with AN, showing a significant increase in Anaerostipes abundance during weight rehabilitation18. Anaerostipes was also found to be increased in relative abundance after a single case of AN-FMT along with genera from order Clostridiales: Roseburia spp., Ruminococcus spp., Blautia spp. Faecalibacterium prausnitzii, Clostridium spp.23. In contrast, other findings have previously shown a reduced abundance of Anaerostipes in adults with AN46,47. For Don1, we also found higher Coprococcus eutactus (butyrate producer) compared to Don1-recipients, which aligns with the higher abundances in HC vs AN, found previously42. Our Don2 exhibited a higher abundance of Gemmiger species compared to the Don2-recipients, consistent with findings showing lower levels of Gemmiger in AN versus HCs42. Conversely, Don2-recipients had increased abundance of several Alistipes species compared to Don2, aligning with prior studies reporting higher Alistipes levels in AN relative to HCs14. However, unlike Fan et al.‘s (2023) findings, our results reveal a lower abundance of Roseburia intestinalis in Don2 than in Don2-recipients before FMT42. Notably, there is a substantial overlap between species previously associated with gut transit time in functional constipation such as lower Faecalibacterium and Roseburia and those found correlated to AN-related dysbiosis. This suggests that the overall significant species identified in AN might represent a signal linked to constipation48.

As a way of analysis, we defined an FMT-closer-to-donor status by evaluating the change in beta diversity distance between the recipient and their respective donor pre-post-FMT. This method allowed us to explore whether certain bacterial species could predict a positive shift toward the donor’s microbiome composition. We found that within the Lachnospiraceae family, the genus Blautia has emerged as a potential predictor for positive closer-to-donor status. Blautia species have been associated with elevated C-reactive protein (CRP) and are more abundant in patients with major depressive disorder and AN49. Their presence was positively correlated with CRP levels, depression scores, and AN symptoms. On the other hand, predictors for not-closer-to-donor status can be found within the Oscillospiraceae family, particularly in the genera Butyricicoccus (a butyrate producer), Faecalibacterium (including Faecalibacterium prausnitzii), and Ruminococcus. Additionally, the Rikenellaceae family, associated with leptin resistance in mice (which may influence appetite), contains Alistipes species50. The presence of different species of Coriobacteriaceae is also associated with not-closer-to-donor status. A study has found Coriobacteriaceae as the only phylotype significantly enriched in individuals with AN compared to athlete HCs, overweight, or obese body types51. Additionally, FMT from heavier donors promoted increased growth in pigs, likely linked to Faecalibacterium and Coriobacteriaceae52, underscoring the possible “readiness” of recipient GM towards FMT outcomes. The focus here is on donor-recipient compatibility, emphasizing baseline GM composition and function in FMT success, perhaps especially in non-pretreated individuals. Without antibiotic pretreatment, responses likely depend on the existing microbiome and identifying microbial predictors could help pre-stratify patients as closer-to-donor or not-closer-to-donor, guiding microbiota optimization through diet, prebiotics, or antibiotics before FMT.

We also explored the impact of FMT on sex hormones (Estradiol, FSH, and LH) in blood samples, pre and post-FMT. Measuring sex hormones in relation to FMT treatments in AN is intriguing but requires careful consideration. Firstly, LH secretion is pulsatile, limiting the accuracy of single blood samples. Secondly, in childbearing-age females, AN often causes hypothalamic amenorrhea, with low gonadotropin and estrogen levels. Lastly, hypothalamic amenorrhea is an adaptive response to extreme malnutrition, enhancing survival9. Therefore, even if altering the microbiome stimulates the hypothalamus-ovarian axis, it may not be beneficial without continuous increased energy availability from diet. Three (17%) of the participants reported use of hormonal contraceptives; however, this was not taken into account. We also analyzed appetite-regulating hormones (GLP-1, insulin, glucagon, C-peptide, PYY, GiP, and leptin) to compare with findings from our previous, mouse study53 In that study, mice receiving multiple FMTs from individuals with AN showed lower food intake and increased PYY and leptin levels, suggesting AN-associated microbiota influence appetite regulation. However, in this human study, no significant changes in these hormones were observed one week post a single FMT. The reported trends in Estradiol, Insulin and PYY are likely of limited clinical relevance in AN and should be interpreted cautiously, especially given the small sample size and lack of dietary or weight data.

We improved a protocol54 to measure gmGUS activity in fecal samples from individuals with AN. We hypothesized that reduced microbiota diversity in AN might compromise gmGUS, potentially restorable by FMT10,29. While gmGUS levels only showed increasing trends in closer-to-donor post-FMT, along with elevated estrogen levels, further investigation into gmGUS levels in a larger population comparing AN and HCs, or following FMT procedures in general, would still be interesting.

We find testing the feasibility of FMT as a treatment for AN present a complex challenge. AN is characterized by ambivalence and often lack of insight, where patients may desire to feel better, but are ambivalent towards interventions that may lead to weight gain. This inherent contradiction complicates the implementation of randomized trials, despite the critical need for such studies to establish evidence-based treatments5. The ambivalence observed in patients with AN underscore the difficulty in achieving informed consent and adherence to treatment protocols, especially when the anticipated outcomes conflict with fears related to weight gain. This is a barrier that needs to be addressed in the design and completing of clinical trials in this population. The lack of clinical data impedes the ability to draw definitive conclusions about the safety, efficacy, and long-term outcomes of FMT in treating AN.

Limitations

This feasibility study included a small, highly selected sample of adult biological females with AN, excluding adolescents, men, postmenopausal individuals, and those with medical comorbidities, which limits generalizability to the broader AN population. The open-label, non-randomized design without a control group, together with the short one-week follow-up, restricts interpretation of both microbiome and clinical outcomes. The use of two donors across multiple FMT batches complicates donor-specific analyses. Finally, as only a single FMT was administered, the findings may not reflect effects achievable with repeated dosing or by longer-term monitoring.

This feasibility study provides a foundational step in exploring FMT as a potential adjunct treatment avenue for AN. A single, low-cost oral FMT intervention demonstrated feasibility, safety, and significant shifts in GM composition toward that of the healthy donor, with no adverse effects. Improved stool consistency among participants suggests possible gastrointestinal benefits. These findings support the feasibility and safety of FMT as a potential tool to address GM dysbiosis in AN. Our study was not powered to test the effect of FMT on AN symptoms and psychopathology, a larger, more comprehensive, well-powered study will need to determine the therapeutic potential of FMT in AN treatment. Future research should focus on a placebo-controlled, double-period crossover design with repeated weekly FMT treatments alongside treatment as usual and extended follow-up to assess clinical impact and sustainability. Broader participant recruitment and optimized protocols could further elucidate the long-term therapeutic role of FMT in AN.

Methods

This study was conducted in accordance with all relevant ethical regulations and was approved by the Danish National Committee on Health Research Ethics (VEK protocol H-22030034), registered at Clinicaltrials.gov (NCT05834010) and written informed consent was obtained from all participants. Further ethical oversight provided through Copenhagen University Hospital Hvidovre (CUHH). This study was an open-label, non-randomized feasibility study performed in 2023 at CUHH, Denmark. For the clinical trial protocol see supplementary materials”Protocol”. Individuals with AN were recruited55,56. Participants completed validated questionnaires before and four weeks after FMT, as specified below. After inclusion, participants were given the choice of receiving a single FMT either via enema or via oral capsules. Blood and stool samples were collected before and again 7 days after the single FMT dose. Likewise, a custom participation questionnaire (Supplementary File) about the FMT procedure and stool type according to the Bristol Stool chart was collected30. Blood samples were analyzed for appetite- and sex-related hormones, while stool samples were examined for GM composition, β-glucuronidase, bacterial species and genes transfer from donor to recipient.

Statistics & reproducibility

As a feasibility study, this work lays the foundation for future power calculations. We predetermined the sample size based on the power of planned GM analysis, no data were excluded, and the study was neither randomized nor blinded.

Recruitment

Participants were recruited actively from Odense and Ballerup hospitals, Denmark and passively through public web postings and trial promotion. Interested individuals received recruitment documents detailing inclusion/exclusion criteria and instructions to confirm their participation. After a prescreening of age, sex, and body mass index (BMI) they were instructed to fill out the initial Eating Disorder Examination Questionnaire (EDE-Q)57. Based on these scores suitable candidates were invited to a meeting for the informed consent process. During the meeting, participants received informed consent documents, Danish pamphlets on rights and ethics, and details about study procedures, time commitments, potential risks, and data collection. All participants were in psychotherapy, many with prior hospitalizations, and their AN diagnoses were confirmed by their treatment centers. Travel expenses were reimbursed, but no monetary compensation was provided.

Inclusion criteria

Adult (age >18 years) biological females referred to the specialized centers at Ballerup or Odense, Denmark, or by advertising who met Diagnostic and Statistical Manual of Mental Disorders version 5 diagnostic criteria for AN58.

Exclusion criteria

Age <18 and >45 years, known ongoing use of laxatives, antibiotic treatment within the last three months, probiotics within the last month, inflammatory bowel disease, colorectal cancer, other chronic bowel disorders, or any other somatic chronic diseases requiring medical treatment.

FMT production and administration

Fecal donors were screened and tested according to the international guidelines for organization of FMT and donor recruitment including questions on normal BMI (18.5–25.0 kg/m2), normal bowel habits (1–2 daily of normal consistency) and with no use of antimicrobial therapy in the past 6 months59,60. We did not specifically ask questions about histories of EDs or other psychopathology. Donors were further selected to be female and in the reproductive age. Manufacturing of capsules was done according to patent61. One portion of FMT capsules consist of 50 grams of un-lyophilized single donor material stored at −80 °C freezer and transferred to −18 °C freezer on the morning of treatment. Frozen FMT capsules (20-25), from a single donor portion, were swallowed under supervision using either fruit juice or flat, diet soda within 10-15 min after 6 h of fasting. Participants opting for rectal FMT received approximately 150 ml (50 g) of donor FMT, instilled 30–40 cm rectally following standard procedures62. Both FMTs as enema or capsules were administered once as a single treatment.

Sample collection

Fecal samples were collected at home by the participants. Participants were instructed to collect about 4 × 1-5 g in 4 specialized collection tubes with an in-tube-spoon attached and immediately store their samples in their own freezer at −18 °C. Samples were transported in a cold, insulated box with freezing elements, delivered within 24 hours, and stored at −60 °C until DNA extraction. Standard blood samples were taken twice, one week apart at the hospital using 10 mL EDTA, 10 mL, Heparin-tubes, 10 mL PAX.

Hormone and appetite biomarker measurements

From the blood plasma, we analysed the following biomarkers: active GLP-1, insulin, Glucagon, C-peptide, PYY, GiP and leptin levels using the U-PLEX® Multiplex sandwich immunoassay (MSD®, Rockville, MD, USA), following the manufacturer’s protocol. We measured Estradiol, P-Follitropin (FSH) and Lutropin (LH) pre- and post-FTM treatment using the CUHH standard pipeline electrochemilumininscence (ECL) immunoassay (Cobas 8000, Roche Denmark).

Gut microbial β-glucuronidase

To assess gmGUS levels, we modified a previously published method54. For full methodology, see Supplementary Methods.

Participation questionnaire and fecal sample consistency according to Bristol Stool Form Scale (BSFS)

Participants completed a questionnaire (Supplementary File) on “gut feeling”, FMT evaluation, future participation limits, and stool consistency using the BSFS30,a validated but indirect proxy for gut transit time (TT)30.

Fecal moisture content measurements

To further validate the BSFS we measured fecal moisture content. Samples (pre-post-FMT) from 15 of 18 participants were thawed (3 had insufficient material), homogenized, and briefly vortexed. Moisture content was determined by drying samples at 50 °C for 48 hours and reweighing using a Mettler Toledo AT200 scale (±0.1 g). Data normality was assessed with the Shapiro-Wilk test.

Psychopathology

We utilized several validated psychometric instruments to assess various aspects of psychopathology: (A) Eating Disorder psychopathology: The assessment of eating disorder symptoms was conducted using the Eating Disorder Inventory (EDI-3), which has been validated for Danish patients63. (B) Eating Disorder symptoms: We administered the 22-item Eating Disorder Examination Questionnaire (EDE-Q), a self-report version of the semi-structured Eating Disorder Examination (EDE) interview, to assess eating disorder symptoms. (C) Quality of Life: The Eating Disorder Quality of Life Scale (EDQLS), validated for Danish anorexia nervosa populations, was used to measure the impact of eating disorders on patients’ quality of life64,65. (D) Depressive Symptoms: To assess the severity of depressive symptomatology, we employed the Major Depression Inventory (MDI), a widely used self-report instrument66,67.

Microbiome analysis

DNA extraction and library preparation: Fecal genomic DNA extraction was conducted using (250 mg of stool) QIAamp PowerFecal Pro DNA kit according to manufacturer. Purity was tested on Nanodrop, ≥10 ng/μL, OD260/280 = 1.8-2.0. DNA sequencing was done by Novogene: The DNA libraries were prepared for the Whole Genome Shotgun Sequencing (WGS) NovaSeq X Plus Series (PE150) (15 GB reading depth per sample). The metagenomics DNA was randomly sheared into short fragments and end repaired, A-tailed and further ligated with Illumina adapter. This was then PCR amplified, size selected, and purified. For further info on metagenomics assembly, binning, taxonomy functional annotations and bioinformatics see Supplementary methods 2.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.