There has been much excitement about the potential role of the gut microbiome in influencing systemic metabolism and the development of diabetes and other cardiometabolic disorders [1,2]. The relationship between obesity, diet, metabolic diseases, and the microbiome is complex, and despite intense interest in this topic, there are few clinical studies to establish causality. The most intriguing data are derived from preclinical mouse models, which have demonstrated that genetic-, diet-, and medication-induced obesity result in microbiome shifts that confer susceptibility to obesity and negative metabolic outcomes when transferred to a new host [3–11]. Remarkably, weight gain has been shown to occur without mice switching to an obesity-inducing diet, suggesting an outsized role of the gut microbiome in dictating body weight. However, the germ-free and conventional animal models used in these studies do not directly replicate human diet, microbiome, or gastrointestinal physiology [12], and therefore these provocative results require clinical validation in human studies.

In humans, the potential relationships between obesity, metabolic disease, and the microbiome are less clear. Antibiotic exposure early in life increases later risk for obesity, presumably via undesirable alterations in the gut microbiome during childhood development [13,14]. Initial small cohort studies suggested that adults with obesity [15–17] and those with type 2 diabetes [18,19] have a different gut microbiome signature than lean controls, with decreased bacterial and/or genetic diversity. Larger cross-sectional cohorts of >1,000 patients showed mixed results, with some finding no consistent diversity or compositional differences between lean and obese adults [20] and others noting small but significant associations between microbiome and body mass index, metabolism, and body composition [21,22].

The most provocative clinical evidence supporting the metabolic potential of microbiota alterations is derived from 2 randomized clinical trials from the Netherlands. In a pilot study with 18 participants (n = 9 FMT recipients, n = 9 controls), and a subsequent larger follow-up study with 38 participants (n = 26 FMT recipients, n = 12 controls), one research group has observed that fecal microbiota transplantation (FMT) from lean donors can transiently improve peripheral insulin sensitivity in men with obesity and metabolic syndrome [23,24]. Nevertheless, these studies found reversion of the gut microbiome and insulin resistance in the FMT recipients back to baseline within 12–18 weeks after FMT administration, indicating a short-lived effect.

Prior clinical studies of FMT have typically relied on direct administration of fresh stool suspensions via upper or lower endoscopic procedures (e.g., esophagogastroduodenoscopy, colonoscopy), often with preceding gastrointestinal lavage and in the setting of pretreatment antibiotics [25]. While these procedures are effective delivery systems, their moderate invasiveness limits considerations of repeated FMT administrations. We have pioneered a novel encapsulation technique to safely deliver oral encapsulated frozen FMT inocula in a clinical setting [26]. FMT capsules have proven to be as efficacious as endoscopically delivered FMT for recurrent Clostridium difficile colitis, and with better patient acceptability ratings [27]. Encapsulated FMT is now offered as standard care for recurrent C. difficile at our institution, and over 400 patients (including 202 formally reported [28]) have been treated without serious related adverse events (AEs).

Given the excellent safety and tolerability profile of FMT capsules, we sought to investigate whether repeated FMT administration could be a viable treatment strategy for durably modifying the gut microbiome and improving human metabolism. We therefore conducted a pilot double-blind randomized placebo-controlled trial that involved weekly administration of oral FMT capsules derived from healthy lean donors delivered to adults with obesity and mild–moderate insulin resistance. We hypothesized that weekly oral FMT would (1) safely and sustainably alter the microbiome among recipients with obesity and (2) improve metabolic endpoints, including insulin sensitivity as assessed by hyperinsulinemic euglycemic clamps.

Between August 2016 and April 2018, we screened 145 individuals to recruit 24 adults with obesity and mild–moderate insulin resistance to participate in this randomized controlled trial (Fig 1). At baseline, the FMT and placebo groups were well balanced in terms of age, sex, weight, and bionutritional measures (Table 1). Our study population was predominantly female, with an average BMI of 38.8 ± 6.7 kg/m² and 41.3 ± 5.1 kg/m² in the FMT and placebo groups, respectively. Of the 24 randomized participants, 23 (96%) completed the 12-week study, including all weekly supervised capsule administrations. Within the FMT group, 1 participant dropped out after week 2 due to gastrointestinal symptoms and did not attend subsequent study visits, and 1 participant missed the 6-week visit due to a family emergency but attended the 12-week visit. Four metabolically healthy lean donors (3 women, 1 man; BMI range 19.5–21.8 kg/m²) provided material for the FMT capsules that were delivered to 12 FMT recipients, with a range of 1 to 5 recipients per donor. The baseline characteristics of the donors are shown in Table A in S1 Data.

After 6 weeks of treatment with study capsules, there were nonsignificant improvements in insulin sensitivity in the FMT group as compared to the placebo group (percentage change in insulin-stimulated glucose uptake: +5% ± 12% FMT versus −3% ± 32% placebo; mean difference 9%, 95% CI −5% to 28%; p = 0.16; Fig 2). Results were similar when insulin-stimulated glucose uptake was corrected for steady-state insulin level (Fig A in S1 Data, p = 0.14). Furthermore, sensitivity analysis using a linear mixed model with adjustment for baseline insulin sensitivity yielded a similarly nonsignificant difference between groups (p = 0.46). Exclusion of the outlier in the placebo group led to a suggestion of improvement in insulin sensitivity in the FMT group as compared to the placebo group that nevertheless was not statistically significant (mean difference 14%, 95% CI −1% to 30%, p = 0.06).

There were no differences between the FMT and placebo groups in change in HOMA-IR, fat mass, fasting lipids, or resting energy expenditure over the 12-week study (Table 2). Body weight was similarly unchanged throughout the study. There was a statistically significant but clinically minor greater reduction in HbA1c at 12 weeks (mean difference −0.1%, 95% CI −0.3 to −0.01%, p = 0.04) and increase in C-reactive protein at 6 weeks (mean difference 1.8 mg/l, 95% CI 0.3 to 3.3, p = 0.02) in the FMT group as compared to the placebo group (Table 2). There were subtle changes in caloric intake throughout the study, but these were not statistically different between groups (Table 2). In exploratory post hoc analyses, change in insulin-stimulated glucose uptake did not appear to differ by donor (p = 0.88) or by baseline demographics (sex, p = 0.24; ethnicity, p = 0.70) or BMI (p = 1.00) of the recipients.

Baseline microbiome analysis revealed that, with the exception of donor 1, preps from the same donor tended to cluster separately from baseline samples of participants with obesity (Fig B in S1 Data). Preps from donor 1 demonstrated distinctively high diversity, with approximately 50% more unique 16S V4 DNA sequences (ASVs) than preps from the other 3 donors (Fig 3), whereas the median diversity of the preps from the other 3 donors fell within the interquartile range of the baseline participant samples.

As expected, we observed temporal variability in placebo participant microbiomes, as well as some background similarity between placebo samples and donor material (Fig 4). We thus used the variability in placebo participant microbiome data to define the background level of endogenous microbiome variation in our analyses. The microbiomes of FMT recipients following dosing were more similar in composition to their paired donor material and less similar in composition to their own baseline sample, compared to placebo participants (Fig 4). These data suggest that FMT shifted recipients’ microbiomes away from their respective baseline compositions and towards donor compositions. Subdividing FMT recipients by their donor material revealed that this shift was observed among recipients of FMT from donors 1, 3, and 4 (Fig C in S1 Data). Although the microbiome of the single donor 2 FMT recipient did not exhibit a large shift towards the microbiome composition of donor 2 (Fig C in S1 Data, panel A), this recipient’s microbiome did demonstrate a potential shift away from baseline following FMT (Fig C in S1 Data, panel B).

We further defined specific 16S V4 DNA sequences as donor-specific (engrafting) ASVs if they were present in both donor material and post-dosing recipient samples but not in baseline recipient samples for donor–recipient pairs. ASVs observed in post-baseline recipient samples but not observed in paired sequenced donor material were categorized as newly detected, not engrafting. All FMT recipients exhibited engraftment of donor-specific ASVs (Fig 5), but the relative abundances of the engrafting ASVs were highly variable. For example, the percentage of total reads in post-dosing recipient microbiome samples that mapped to engrafting ASVs was notably high among donor 1 FMT recipients, with a sample median of 47%, whereas median abundances for the other donors were around 9.5%. In the majority of FMT recipients, total abundance of engrafting ASVs exceeded the background level of newly detected ASVs (Fig 5). The exceptions were participants 13 and 18, for whom the majority of post-dosing samples suggested that the rate of newly detected and engrafting ASVs did not exceed the expected background variation [33].

Overall, our data suggest that bacterial strains from donor FMT capsules successfully engrafted in the majority of our participants, although engraftment was markedly strongest among donor 1 FMT recipients and may not have occurred at all in participants 13 and 18. When occurring, engraftment persisted throughout the 12-week study, including 6 weeks after the cessation of dosing. Participant microbiome similarity to donor material plateaued after week 3, suggesting that dosing for more than 3 weeks does not result in additional engraftment. However, the additional weeks of dosing could have contributed to maintenance of engraftment.

The engrafting ASVs from all donors represented a diverse set of enteric genera (Fig D–G in S1 Data). Strikingly, we observed that 2 Prevotella ASVs strongly engrafted in donor 1 FMT recipients, increasing the relative abundance of Prevotellaceae in donor 1 FMT recipient microbiomes by at least 8-fold post-dosing and thus shifting the enterotype signature of these participants towards a higher Prevotella abundance relative to Bacteroides (P/B ratio) (Fig H in S1 Data). No enterotype shifts were consistently observed among other recipients. Despite the evidence for engraftment, compared to placebo participants, FMT-treated participants did not display a notable increase in microbiome diversity from baseline to week 6, suggesting that the FMT treatment did not impact participant microbiome diversity (Fig I in S1 Data).

A large percentage of ASVs were identified in both donor and baseline recipient material. It is unclear whether those ASVs represent the same bacterial strain present in both samples or whether the donor and participant microbiomes each contained different bacterial strains that could not be differentiated due to the limited taxonomic resolution provided by 16S V4 sequencing. To address this, we sequenced a subset of donor 1 FMT recipient and placebo samples using shotgun metagenomics, which can provide finer taxonomic resolution. These data recapitulated ASV engraftment results, showing that the 3 donor 1 FMT recipients had more bacterial species in common with donor material, and fewer bacterial species in common with their own baseline sample relative to the 3 placebo participants (Fig J in S1 Data). Additionally, among the bacterial species found in both baseline and post-FMT shotgun metagenomic samples, approximately 50% the bacterial strains of those species identified in weeks 6 and 12 were more phylogenetically related to strains found in the donor material than the baseline recipient sample (Fig K in S1 Data), suggesting those organisms originated from the FMT dose.

We did not observe any statistically significant correlations between changes in participant microbiome composition and changes in metabolic outcomes. Given prior studies that have shown low bacterial diversity as predictive of greater metabolic response to FMT [24], we performed exploratory subset analyses. Among participants with low baseline microbiome diversity (Table B in S1 Data), analyses suggested greater improvements in several metabolic outcomes at 12 weeks for those who received FMT (n = 4) versus placebo (n = 7), including total cholesterol (mean difference −0.6 mmol/l, 95% CI −1.0 to −0.1 mmol/l), HbA1c (mean difference −0.2%, 95% CI −0.4 to −0.01%), and fasting glucose (mean difference −0.6 mmol/l, 95% CI −1.1 to −0.1 mmol/l).

There were no serious AEs reported in either group throughout the 12-week study (Table C in S1 Data). More study participants reported at least 1 episode of diarrhea in the FMT group than in the placebo group, although this difference was not statistically significant (10 versus 5, p = 0.09). The majority of diarrheal symptoms were rated as mild, and there were no CTCAE grade 3+ AEs. Intriguingly, the only 4 diarrheal events rated as moderate occurred in 2 participants who received FMT from donor 4, one of whom dropped out of the study following their second event. Upon inspection of engrafting ASVs, no organisms stood out as the potential cause of the moderate AEs (Fig F in S1 Data) albeit several days separated the timing of events and stool collection. There were no imbalances between FMT and placebo groups in other symptoms.

This double-blind randomized placebo-controlled trial was designed to test the safety of FMT capsules and their ability to alter the gut microbiome in adults with obesity, and to probe for a causal link between microbial changes and metabolism. We found that 6 weeks of FMT capsule administrations sustainably altered gut microbiome composition for the majority of participants without serious adverse effects, and absent any antibiotic pretreatment or lavage. Despite the encouraging engraftment signal, we did not find statistically significant differences between the FMT and placebo groups in insulin resistance, body weight, or most other metabolic markers in these adults with obesity and without diabetes. HbA1c modestly decreased at 12 weeks in the FMT group as compared to the placebo group, although the magnitude of improvement was small. Both metabolic and microbiome responses to FMT were highly variable, suggesting a complex host–recipient dynamic.

To date, only Nieuwdorp and colleagues have published interventional FMT trials in patients with obesity. In a small pilot trial, conducted in the Netherlands, 9 men with obesity who received endoscopically delivered FMT infusions from lean donors had significantly improved peripheral insulin sensitivity over 6 weeks as assessed by hyperinsulinemic euglycemic clamp [23]. The median level of insulin-mediated glucose uptake was approximately 73% higher at 6 weeks than at baseline, and was accompanied by increased gut microbial diversity. A larger follow-up study by the same group once again demonstrated a statistically significant increase in peripheral insulin sensitivity among 26 adults with obesity receiving endoscopic FMT infusion from lean donors, although the magnitude of improvement was more modest (approximately 12%) [24]. This follow-up study also documented a small decline in HbA1c at 6 weeks after FMT, which is similar to the minor improvement in HbA1c at 12 weeks in our trial. However, engrafting clades were different in the 2 prior studies, and the studies also documented different patterns of change in fecal short chain fatty acids and bile acids. In both prior studies, metabolic and microbiome changes were short-lived, having disappeared by 12–18 weeks after FMT infusion [23,24].

The aforementioned small studies as well as our current pilot trial are not definitive and should be considered hypothesis-generating. Indeed, one possible interpretation of our study is that the gut microbiome does not regulate human metabolism in the same dramatic manner as has been shown in preclinical studies. Alternatively, it is possible that a greater magnitude of FMT engraftment is required to effect systemic changes. Mouse models have shown that obesity phenotypes can be transferred to germ-free mice through colonization with obese mouse microbiota [4,5]. However, there is little evidence to date informing whether obese or lean phenotypes can be transferred to already colonized models or the extent to which the native microbiome needs to be replaced to induce a phenotypic change.

The closest human equivalent to colonizing a germ-free mouse would be the total replacement of an individual’s native microbiome, which to our knowledge has never been demonstrated. Strategies have been proposed to further improve FMT engraftment in clinical studies. For example, treating our study participants with broad-spectrum antibiotics prior to FMT dosing would likely have increased the ratio of donor to baseline microbes after FMT [34]. However, this approach could be associated with side effects and raise ethical and antibiotic stewardship concerns. It has been suggested that bowel cleansing might enhance FMT engraftment, but this strategy has not been rigorously studied, and bowel preps only minimally impact microbiome composition [35]. For these reasons, combined with the absence of a clear experimentally supported engraftment level target, we elected to begin testing the impact of FMTs on metabolic outcomes in human participants with the most minimally invasive approach. Indeed, we found evidence of engraftment in the majority of the FMT recipients without any gut preparation or pretreatment antibiotics. It nevertheless remains unclear if achieving a greater number or relative abundance of engrafting strains in our clinical study would have yielded positive metabolic outcomes more akin to results from germ-free mouse models.

It is helpful to contrast our pilot trial with the prior published clinical trials of FMT and metabolism. Baseline age, BMI, and HbA1c in our study were roughly comparable to those of participants in the Netherlands projects. However, microbiomes of donors and recipients can strongly vary in diversity, composition, and engraftment strength, as is evident in our study data, and not all donor material may be equally efficacious, nor all recipient microbiomes equally responsive to a microbiome therapy. Notably, our trial was conducted in the US, and we expect the recipients and donors from the US and Netherlands studies to have different microbiome compositions due to regional, race/ethnicity, and dietary differences in the study populations [36–38]. Furthermore, Nieuwdorp and colleagues enrolled men exclusively, whereas our recipients receiving active FMT treatment were 67% women, and 3 out of our 4 donors were women. There are some data to suggest differences in the gut microbiome between men and women [39], and thus it is possible that there are sex-specific differences in donor FMT material, or altered metabolic responses to FMT among female recipients. The routes of FMT delivery (capsule versus endoscopy), type of FMT (frozen versus fresh), choice of controls (non-microbiome placebo versus autologous FMT), and antecedent gastrointestinal preparation (none versus bowel prep) differed from prior studies. These have not been important factors in studies of FMT for recurrent C. difficile colitis [40], although appropriate caution should be taken when extrapolating from FMT outcomes in other disease conditions.

Underlying causes of obesity and insulin resistance are multifactorial and likely vary among individuals. It is possible that only a subset of individuals may respond to alterations of the microbiome. Of note, the second Netherlands study reported that lower gut microbiota diversity among obese recipients at baseline predicted metabolic response to FMT [24]. In our exploratory analyses, we observed that FMT capsules led to possible improvements in total cholesterol, fasting glucose, and HbA1c among those with low microbiome diversity at baseline, although these results should be interpreted with caution given the small numbers of participants studied. It is also important to note that the magnitude of clinical improvement after FMT in these exploratory studies is modest. We found that donor microbial diversity was positively associated with better engraftment, although changes in microbial composition were not specifically correlated with metabolic outcomes. We used a different method for characterizing microbial communities than that used in the Netherlands studies; thus, direct comparisons between cohorts are difficult to make. Variable engraftment by donor and variable metabolic changes among FMT recipients raise the question of whether selecting donors with specific microbiome signatures or designing a targeted doseable microbial consortium could yield metabolic improvements.

In a noteworthy study, gut microbiome composition helped to predict glycemic responses to various diets [41]. A specific enterotype defined by a high abundance of Prevotella or a high ratio of Prevotella relative to Bacteroides (P/B ratio) [42] has also shown promise in predicting response to dietary interventions. For example, Kovatcheva-Datchary et al. observed that participants with a high abundance of Prevotella experienced greater improvements in glucose metabolism following fiber intervention than participants with low Prevotella abundance [43]. In addition, a study by Hjorth et al. found that participants with P/B ratios above 0.01 lost more weight after switching to a high-fiber healthy diet than participants with a P/B ratio below that threshold [44]. Thus, it is possible that additional selection criteria may be required to produce a microbiome therapeutic to optimize engraftment and metabolic responses, perhaps in the setting of a concurrent dietary intervention.