Fecal microbiota transplantation from psychiatric patients to mice - systematic review of methodologies and a call for standardization

Abstract

Background

Fecal microbiota transplantation (FMT) has emerged as a key tool to explore the role of the microbiome-gut-brain axis in psychiatric disorders. However, the field is hindered by significant methodological inconsistencies.

Methods

A comprehensive literature search identified 31 studies performing FMT from human patients with psychiatric conditions into rodent models.

Results

None of the 31 studies followed an identical FMT protocol. Significant heterogeneity was observed across studies in rodent model selection, including germ-free, antibiotic-pretreated, or specific pathogen-free approaches, in antibiotic regimens, timing and microbiota depletion verification, as well as in FMT donor strategy, dosage, frequency, engraftment assessment, and behavioral testing schedules.

Conclusions

This review highlights the necessity for standardized methodologies in microbiome research. Evidence-based recommendations are provided to promote reproducibility in future work. Investigators are encouraged to publish transparent and rigorous protocols, to enhance the translational potential of microbiome-gut-brain axis research.

Introduction

The microbiota-gut-brain axis has been a source of intense investigation for several years. A growing body of evidence has revealed specific microbial signatures associated with various psychiatric conditions [1–3]. The gut-brain axis has been widely studied in rodents, mainly mice and rats, to uncover its mechanisms, often through fecal microbiota transplantation (FMT), which transfers gut microbes from donors to recipients via oral gavage [4, 5]. Human to mouse FMT is highly complex and numerous steps have the potential to introduce variability [4, 6–8]. Microbiome research is still relatively new and suffers from widespread methodologic inconsistencies. These methodological limitations not only hinder cross-study comparisons but also restrict the translational potential of findings, emphasizing the urgent need for standardized approaches in microbiota-gut-brain axis research.

The aim of this systematic review is to identify common themes in research methodologies and make evidence-based recommendations. Our goal is to provide a foundation for standardizing experimental approaches, ultimately enhancing the reproducibility and reliability of studies exploring psychiatric disease through FMT.

Materials and methods

Experimental design

The present systematic review followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (10). A comprehensive literature search was conducted to identify relevant studies. Databases including PubMed, Scopus, Cochrane, and Web of Science were searched from 01/01/2004 to 07/16/2024. The literature searches were conducted by adapting the search method according to the specific database used. The following search string was applied in PubMed: “(psychiatry OR psychology OR learned helplessness OR affect OR anxiety OR depression OR emotional regulation OR psychological distress OR euphoria OR fear OR pleasure OR sadness OR motivation OR anhedonia OR catatonia OR mania OR psychomotor disorders OR delirium OR hallucinations OR psychomotor agitation OR temperament OR resilience OR anxiety disorders OR separation anxiety OR neurotic disorders OR obsessive-compulsive disorder OR panic disorder OR phobic disorders OR agoraphobia OR social phobia OR feeding and eating disorders OR anorexia nervosa OR binge-eating disorder OR bulimia nervosa OR mood disorders OR bipolar disorder OR depressive disorder OR major depressive disorder OR treatment-resistant major depressive disorder OR dysthymic disorder OR seasonal affective disorder OR premenstrual dysphoric disorder OR postpartum depression OR cyclothymic disorder OR neurodevelopmental disorders OR autism spectrum disorder OR autism OR autistic disorder OR attention deficit disorder OR hyperactivity OR ADHD OR schizophrenia OR catatonic schizophrenia OR disorganized schizophrenia OR paranoid schizophrenia OR treatment-resistant schizophrenia OR psychotic disorders OR substance-related disorders OR alcohol OR amphetamine OR cocaine OR inhalant OR marijuana OR narcotic OR phencyclidine OR substance abuse OR tobacco OR traumatic stress disorders OR post-traumatic stress disorders OR psychological trauma) AND (fecal microbiota transplantation OR fecal microbiome transplantation OR FMT OR human microbiota associated mouse) AND (mice OR rat OR murinae) AND (human)”. This strategy was tailored to ensure the comprehensive identification of relevant studies across different databases. The search terms were modified as necessary to fit the syntax and structure of each database, maximizing the retrieval of pertinent literature. In addition to database searching, we also screened the reference lists of relevant articles and reviews. One additional study, known to the authors and meeting all inclusion criteria, was identified through manual search and included.

PICO strategy

The PICO (Population, Intervention, Comparison, and Outcome) framework was employed to structure and guide the systematic review process. In this systematic review, the Population refers to patients with psychiatric diagnoses and mouse models, while the Intervention encompasses FMT using samples from psychiatric patients into a mouse/rat model. The Comparison component does not reflect an a priori hypothesis testing of treatment effects between experimental and control groups; rather, it was applied as a methodological filter to ensure the inclusion of studies adopting a rigorous experimental design with an appropriate control group (i.e., FMT from healthy donors). Finally, the Outcomes focused on identifying methodological similarities and differences across studies. This structured approach ensured the identification of relevant studies and the systematic assessment of their outcomes in relation to the research question.

Eligibility criteria

Studies were included if they: (1) involved FMT from human donors with a psychiatric disorder (primary or comorbid) to mice or rats; (2) used fecal matter from psychiatric patients, regardless of diagnosis; (3) compared FMT from psychiatric patients to healthy or other relevant controls; and (4) were peer-reviewed translational research articles. Excluded were studies that: (1) used only non-psychiatric donors; (2) used animals other than mice or rats; (3) did not perform FMT; (4) used bacterial cultures instead of FMT; (5) used only rodent-to-rodent FMT; or (6) were reviews, protocols, case reports, or conference abstracts.

Study selection

A.M.D. and A.G.N. screened titles and abstracts. Full texts were assessed by G.C. and F.S. Discrepancies were resolved through discussion or by consulting F.C. and A.B.

Data extraction

Data were extracted from the included studies by two independent reviewers using a standardized form. The extracted data included: 1. authors, year of publication, 2. patient’s diagnosis, 3. mouse/rat model, 4. germ-free state/induction, 5. pooled/single donor (patient: mouse), 6. hours from last antibiotic treatment, 7. FMT protocol, 8. behavioral tests performed, 9. time between last gavage and first behavioral test. Any discrepancies in data extraction were resolved by consensus or consultation with a third reviewer.

Data synthesis

A qualitative synthesis was conducted to integrate and interpret the findings from the included studies. The synthesis aimed to provide a comprehensive understanding of the effects of fecal microbiota transplantation from psychiatric patients to mice. Due to the heterogeneity in study designs, outcomes, and methods, no meta-analysis was performed.

Results

Study selection

The study selection process is summarized in a PRISMA flow diagram (Fig. 1) [9]. A detailed overview of all information collected from the 31 studies is provided in supplementary table 1.

Fig. 1. Study design.

PRISMA flow diagram illustrating the study selection process.

Psychiatric diagnosis

All selected studies evaluated the effects of FMT on mouse or rat models from psychiatric patients with various diagnoses: depression (n = 13) [5, 10–20], schizophrenia (SCZ) (n = 2) [21, 22], anxiety (n = 3) (further sub-categorized as Generalized Anxiety Disorder (GAD) [21], Social Anxiety Disorder (SAD) [22], and anxiety in comorbidity with a medical condition [23]), anorexia nervosa (AN) (n = 3) [24–26], alcohol use disorder (AUD) (n = 4) [27–30], autism spectrum disorder (ASD) (n = 3) [31–33], Attention-deficit/hyperactivity disorder (ADHD) (n = 1) [34], and bipolar disorder (BD) (n = 1) [35].

Depression

The studies involving patients with depression were the most numerous and highly heterogeneous (Supplementary Figure 1). Some focused solely on female patients [16, 36]. In some, depression was diagnosed based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders [5, 10, 14, 15, 17, 19, 20, 36], while in others it was assessed as a symptom using specific psychometric tools [11–13, 18]. Some studies involved patients who were not receiving psychopharmacological treatment [13, 16, 19], whereas one study had two groups of patients differing in whether they were receiving psychopharmacological treatment, but the selection of samples for FMT was random [17]. In others depression, as either a symptom or diagnosis, was comorbid with chronic inflammatory bowel diseases [12, 18].

Schizophrenia

The three studies on schizophrenia differ markedly: one used samples from patients on antipsychotics [37], the other from drug-free patients [38], and a third used randomly selected samples from a clinical cohort in which most patients were medicated, although a minority were drug-free at the time of collection [39].

Anxiety

As previously mentioned, studies on anxiety were also heterogeneous: one included patients with GAD diagnosed using the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria [21], another with SAD also used the DSM-5 [22], and a third assessed anxiety as a symptom in IBS-D patients using the Hospital Anxiety and Depression Scale (HADS) [23].

Anorexia nervosa

All studies used fecal samples from female patients, likely due to epidemiological reasons. The only difference between these studies was the diagnostic criteria: two studies used DSM-5 for diagnosis [24, 25], while one study used DSM-IV-TR [26]. Information regarding purging behaviors or binge eating were not explicitly stated.

Alcohol use disorder

In studies involving patients with AUD, all patients were male. The diagnostic methods varied significantly between studies. In two studies, the diagnosis was made using the Alcohol Use Disorder Identification Test (AUDIT) [27, 28]. One study used DSM-IV criteria [29], and another used the International Classification of Diseases (ICD-10) criteria [30].

Autism spectrum disorder

Studies involving patients with ASD likely included only male patients, although gender is specified in only one of these studies [31]. Given that ASD is a neurodevelopmental disorder, it is crucial to note the differences in patient recruitment age. In one study, patients were aged 5–10 years [31], while in another, they were aged 2–6 years [32]. Only one study made the diagnosis through structured interviews according to DSM-5 criteria [32]. In other studies, the method of diagnosing ASD is not specified [31, 33], or the demographic characteristics of the sample are not reported [33].

ADHD and bipolar disorder

ADHD and bipolar disorder (BD) were each investigated in a single study. For ADHD, the diagnosis was made following DSM-IV criteria using the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS) [34]. The BD study involved patients with BD who were experiencing a depressive episode, diagnosed according to DSM-IV-TR criteria [35].

Rodent models

The selected studies utilized a variety of mouse models to investigate the effects of FMT from psychiatric patients (Supplementary Table 2). The C57BL/6 mouse, used in 17 studies, was the most common model [10–13, 18, 21, 22, 25, 28–31, 33, 34, 36–38]. The other mouse models, listed in descending order of frequency, were: Kunming (6 studies) [14, 15, 17, 19, 35, 39], BALB/C (2 studies) [26, 32], Swiss Webster (1 study) [24], and NIH Swiss (1 study) [23]. As for the rats, 1 study used the Wistar rat [27], 1 study used FSL and FRL rats [16], 1 study used the Sprague-Dawley rat [5], and 1 study used an unspecified rat model [20].

Gut microbiota depletion

Prior to FMT, some studies administered a broad-spectrum antibiotic treatment (AbT) to induce a GF state, while others used an established GF colony (Supplementary Table 3). 14 studies used a GF model [14, 15, 17, 19, 20, 23–26, 28, 32–34, 39], 12 studies used AbT [5, 10, 11, 13, 21, 22, 27, 29, 30, 36–38], and 5 studies employed a specific pathogen free (SPF) model [12, 16, 18, 31, 35].

Antibiotic regimen

A large degree of heterogeneity was identified in studies using antibiotics prior to FMT (Fig. 2). Only 3 out of 12 studies employed the same antibiotic regimen. The majority of studies (7/12) utilized a cocktail containing metronidazole, vancomycin, ampicillin, and neomycin (MANV). Amphotericin B was added in 2/7 studies using the MANV cocktail. Despite using a similar antibiotic cocktail, the route of administration and duration of treatment differed significantly amongst the studies. Metronidazole was used in 4/5 studies not using the MANV cocktail.

Fig. 2. Antibiotic regimens used.

Overview of the antibiotic regimens used to induce a germ-free state in the analyzed studies. The figure highlights the variability in treatment protocols, with only 3 out of 12 studies using the same antibiotic combination. For the study by Zhao et al. [30], no reference to the timing of the therapy was provided. Credit: biorender.com.

Differences were also noted in the route of administration and duration of therapy. Antibiotics were administered through the drinking water (5/12), gavage (2/12), or a combination of the two (5/12). The treatment duration varied between 7 to 28 days, averaging 14 days.

Of the 12 studies that employed antibiotic treatment, only 5 explicitly reported verifying microbiota depletion through 16S rRNA sequencing. The remaining studies relied on treatment protocols that have been generally shown to be effective in depleting the gut microbiota.

In studies using mouse models pretreated with antibiotics, we investigated the time interval between the last day of antibiotic treatment and the first gavage with fecal samples from psychiatric patients. One study performed the first gavage after 12 hours, 3 studies after 24 hours, 5 studies after 48 hours, 2 studies after 72 hours, and in 1 study, the timing was not specified (Supplementary Figure 2).

FMT Protocol

As with antibiotic treatment, the FMT protocol also showed significant variability across studies. Differences included the use of pooled feces or feces from a single patient, the dosage of gavage administered to each mouse, the time interval between the last day of antibiotic treatment and the start of colonization, and finally, perhaps most importantly, the number of gavages performed and the timeframe in which they were carried out. Additionally, we occasionally observed that information regarding the FMT protocol was not clearly and transparently reported. As a result, in some instances, we had to infer that, for example, pooled feces were used and that only a single gavage was performed. Specifically, eight studies were inferred to be pooled (Medina 2023, Yoo 2022, Jang 2021, Glenny 2021, Xiao 2021, Zhao 2020, Tengler 2020, Hata 2019) and 5 studies were assumed to use a single gavage (Lai 2021, Liu Y 2021, Glenny 2021, Xiao 2021, Hata 2019).

In 18 studies, FMT was performed using pooled samples and in 13 studies animals were transplanted with feces from a single donor sample. In studies utilizing a single donor or multiple donors (not pooled samples) to transplant gut microbiota from psychiatric patients into mice, we analyzed the number of mice per donor ratio (Supplementary Figure 3). Between one and fourteen animals were used for each human donor, with an average of six mice per donor. In Fig. 3, we summarize the FMT protocols, highlighting the type of samples used and whether the animal model was GF or pretreated with antibiotics before the FMT. As can be seen from the figure, there is significant variability in protocols used for FMT.

Fig. 3. Summary of FMT protocols used.

The frequency of gavage during the colonization period varies from one to four weeks. Each circled/squared dot represents a gavage. *1 gavage/day for 3 days and after 2 weeks 1 gavage/week for 6 weeks. Credit: biorender.com.

Engraftment efficiency

We additionally assessed whether engraftment was evaluated in the included experiments. Among the 31 studies, 13 directly verified engraftment, employing approaches such as shotgun metagenomics (Fan et al. [24], Medina-Rodriguez et al. [11], Zhu et al. [38]), 16S rRNA qPCR (alone or in combination with shotgun metagenomics) (Fan et al. [24], Medina-Rodriguez et al. [11], Wang et al. [27], Wolstenholme et al. [28], Knudsen et al. [16], Glenny et al. [25], Zhu et al. [38], Leclercq et al. [29], Zhao et al., 2019), short-chain fatty acid (SCFA) quantification (Wolstenholme et al. [28]) together with 16S rRNA qPCR, or comparative analyses of relative OTU abundances. Some studies also employed advanced analytical approaches, applying Principal Coordinates Analysis (PCoA) based on various beta-diversity distance matrices (Leclercq et al. [29]), including weighted and unweighted UniFrac (Sharon et al. [33], Hata et al. [26], Zheng et al. [17]) and Bray–Curtis dissimilarity (Fan et al. [24], De Palma et al. [23]), to compare the microbiota profiles of human donors and transplanted mice. Three studies evaluated engraftment indirectly, for example through partial genus-level overlap and enrichment of specific genera in FMT recipient mice (Wei et al. [37], Zheng et al. [39]) or microbial beta-diversity analyses (Tengler et al., 2020). In contrast, 15 of the 31 studies did not assess engraftment.

Behavioral tests

We considered it crucial to analyze the time between the last gavage and the start of the behavioral tests (Supplementary Figure 4). Again, we observed considerable variability among the studies, with some conducting behavioral tests during the gavage period. 3/30 studies did not perform BTs. On average, behavioral tests were performed 10 days after the final gavage. Latency period ranged from the same day as the final gavage to 21 days after final gavage. Several studies performed behavioral tests during the colonization period.

For readers interested in a detailed overview of the behavioral tests used in relation to the psychiatric condition of the human FMT donors, we provide a comprehensive summary in (Supplementary Table 4). This specific aspect was not further discussed in the main Discussion section, as the topic is too broad and would compromise the clarity and focus of this review; in our view, it would warrant a dedicated meta-study on its own.

Discussion

The importance of diagnosis: from symptoms to syndrome

In psychiatry, diagnosis is particularly challenging due to the complexity of mental health disorders. Unlike other fields of medicine, where diagnostic tools such as imaging or laboratory tests can offer concrete data, psychiatric diagnoses often rely on the observation of symptoms and patient-reported experiences. The boundaries between different psychiatric conditions can be blurred [40, 41]. This complexity emphasizes the importance of moving from a mere symptom-based approach to a more comprehensive understanding of syndromes, which considers not only the symptoms but also their duration, severity, and impact on daily functioning.

The studies reviewed showed a marked lack of diagnostic standardization (Supplementary Figure 1), with some using DSM criteria and others relying solely on symptom presence. This inconsistency limits comparability and raises concerns about diagnostic validity and reliability.

Certainly, the categorical system of mental disorders presents imperfections. The two most used systems, the DSM and ICD, show significant differences and favor a diagnosis based on diagnostic criteria that sometimes are too vague and at other times too strict. Moreover, diagnosis alone is not sufficient, as psychiatric disorders are uniquely influenced by various social and psychological factors [42]. In contrast, though not entirely opposed to the categorical approach, we have the dimensional approach which allows practitioners to identify different subtypes within a psychiatric syndrome [42]. This perspective highlights the heterogeneity within mental health conditions, suggesting that while certain underlying mechanisms may be shared, the expression of symptoms can vary significantly, requiring a deeper understanding of diagnosis and treatment. The dimensional approach also appears to be the most suitable for studies that aim to highlight how biological variables (e.g., neurotransmitter levels, neuroimaging findings, or microbiological factors) are correlated with specific psychopathological manifestations, rather than solely based on diagnosis. Indeed, studies exploring the influence of the microbiota on mental disorders have highlighted how various dimensions of psychopathology are closely linked to alterations in gut microbiota [43–45].

It is important to avoid the mistake of interpreting a dimensional diagnosis as something solely based on symptoms. Instead, it should be seen as a more comprehensive characterization of a disorder. Studies that recruit patients based solely on the presence of a single symptom are not examples of best practice, as this approach neglects the broader picture of the psychiatric syndrome.

We would also like to highlight that studying the microbiota in conditions like anorexia nervosa and alcohol use disorder is, in our opinion, particularly challenging due to the profound metabolic disturbances intrinsic to these disorders. In anorexia, psychological factors such as body image distortion and fear of weight gain (48, 54–56) lead to malnutrition and metabolic alterations (57). In alcohol use disorder, it becomes difficult to separate the psychological components, such as cravings and distress, from the biological impact of chronic alcohol consumption. This raises the question of whether microbiota alterations stem from psychological aspects or are primarily driven by metabolic imbalances (58). Although evidence suggests a role of the microbiota in influencing liver dysfunction (59) and eating behavior (60), the strong overlap between mental and physical symptoms in these disorders complicates both interpretation and translational modeling, particularly when comparing human and animal data. By contrast, psychiatric conditions such as depression, schizophrenia, bipolar disorder, autism spectrum disorder, anxiety, and ADHD do not inherently include metabolic dysfunction as a diagnostic feature. Immunometabolic disturbances can occur in some subgroups—either in drug-naïve patients [46] or as a consequence of psychopharmacological treatments [47], particularly antipsychotics [48]—but they are not universal. This difference implies that, while confounding from metabolic disturbances can often be minimized in these populations by excluding major medical comorbidities, in anorexia nervosa and alcohol use disorder such confounding is intrinsic and therefore largely unavoidable.

Having made these considerations, it is crucial to exclude patients with comorbid medical conditions, such as metabolic or autoimmune diseases, as it is known that these conditions can lead to alterations in the microbiota [2, 24] and could therefore represent potential confounding factors in the interpretation of study results. In this regard, most studies applied similar exclusion criteria to minimize potential confounders, including severe physical comorbidities (except when psychiatric conditions were comorbidities), uncontrolled metabolic diseases, immunosuppression, active infections, active gastrointestinal diseases, and current or recent antibiotic use. This reflects a higher level of standardization in the definition of inclusion and exclusion criteria across studies.

As with any study, the choice of controls is crucial to the experimental design. While the specifics of which control to pick will depend on the study, using both dimensional and criteria-based characterizations will assist in appropriate selection of controls. For example, a study may compare patients with differing diagnoses (e.g. depression and schizophrenia) that share dimensional characteristics or vice versa. As with any diagnosis, the presence or absence of treatment must be included in demographic data. As the microbiome is known to metabolize some psychotropic drugs [49], incorporating pharmacologic therapies into analysis may provide useful insight.

We recommend enrolling patients with diagnoses based on the diagnostic criteria of the latest version of the DSM and characterizing them from a psychopathological perspective using validated clinician-administered scales. Doing so will allow investigators to more precisely determine whether a specific psychiatric condition (diagnosed using criteria) with particular psychopathological characteristics (dimensional characterization) exhibits features that are reflected in the gut microbiota and whether the microbiome contributes to the pathophysiology of disease.

How crucial is the choice of mouse or rat model?

There are two points we would like to address regarding choice of animal model: (1) the known behavioral differences between wild-type models, and (2) the utility of mouse models of behavioral disease in human-to-mouse FMT.

While WT models are the default controls in experimental design there are well known differences. For example, T cells from C57BL/6 mice preferentially produce a Th1 response whereas BALB/c is pushed towards a Th2 response [50, 51]. WT mice also display inherent differences in their behavior [52, 53]. Behavioral assays, in particular, are known for high degrees of variability even within the same inbred strain [54]. Direct comparison of the C57BL/6 and BALB/c has shown that C57BL/6 mice display less anxiety-like behavior and are less vulnerable to social defeat stress [55–57].

WT mice were used for all studies analyzed in this review. Our search did not yield any studies that utilized transgenic mouse models or other models to predispose animals to a behavioral phenotype. While the microbiome clearly has an important role in the gut-brain-axis, it is important to acknowledge the contribution of the host. The native mucosal immune system and host genetics have an important role in shaping and responding to the microbiome [58]. The importance of the host has been recognized in other disease paradigms. In inflammatory bowel disease (IBD), for example, studies have demonstrated that healthy controls can harbor a microbiome that is pro-inflammatory to mouse models of IBD [59]. While animal models of psychiatric conditions are not meant to be a perfect mirror of human disease, there is still utility in models that exhibit features of the disease [60, 61].

While mouse models represent the vast majority of rodent models in human-to-rodent FMT experiments, some studies have also compared the engraftment efficiency of rat models. Despite lacking a gall bladder [62], the engraftment efficiency of rats has been shown by some to be superior to murine models [63, 64]. However, the repertoire of murine models of neurologic and behavioral conditions still vastly outnumbers rat models [65, 66]. Furthermore, rat colonies require significantly more space and resources making it logistically more difficult to maintain compared to murine colonies [67].

The vast differences in behavioral phenotypes between WT mouse strains makes proposing a definitive recommendation difficult. Each strain presents its own advantages and disadvantages. While we do not recommend a specific WT strain it is important that investigators are consistent and recognize the inherent differences between the strains. Likewise, readers are advised to be mindful of these differences when interpreting results from behavioral assays. During experimental design, we also recommend considering the use of models of behavioral disease. Doing so in conjunction with a dimensional approach for human subject selection, as previously discussed, may provide a more accurate representation of human disease.

Germ-free, antibiotic-treated, or specific pathogen-free mice

Though the ultimate goal of germ-free reared and antibiotic-treated mice is similar – to enable durable engraftment of human derived microbes in the host – there are important advantages and disadvantages to both methods.

Germ-free colonies are generated by surgically removing animals from the womb. The animals are then raised and bred in sterile isolators with frequent monitoring to ensure no microbial colonization has occurred [68]. Germ-free animals are a powerful tool in the study of the microbiome. It is akin to generating knock-out animals to study the function of a specific protein.

The use of germ-free animals presents notable disadvantages. Gnotobiotic facilities are resource-intensive and require specialized personnel to ensure sterility and prevent contamination. Germ-free animals also have significant alterations in their development. The mucosal immune system and peripheral tolerance is particularly affected by the lack of exposure to microbes [69]. Various studies have shown that germ-free mice have impaired formation of gut-associated lymphoid tissue (GALT), impaired Treg function, increased colonic Th17 cells, decreased secretory IgA, and other deficits [70]. As a consequence, germ-free mice may have an unexpected response to colonization.

The microbiome has also been implicated in the development of the central nervous system (CNS) and behavior. To further complicate the issue, the effect is not consistent amongst strains. Germ-free naval medical research institute (NMRI) mice displayed decreased anxiety-like behavior compared to their SPF counterparts [71]. Conversely, germ-free C57BL/6 mice have increased anxiety on the step-down test [72].

Compared to germ-free models, antibiotic pre-treatment before FMT is less costly and logistically simpler, requiring only standard training in oral gavage. It also preserves normal immune and CNS development. However, antibiotics can cause significant side effects and are not without risks. Antibiotics alter host metabolism, reduce tissue weight [73], and chronic exposure to broad-spectrum antibiotics can cause liver toxicity. The antifungal Amphotericin B is known for its severe nephrotoxic side effects [74]. As discussed in the subsequent section, antibiotic cocktails generally do not include anti-viral agents. While antibiotic pre-treatment does significantly reduce the microbial burden of the intestinal tract, it is not a true “germ-free” state [75].

Some studies use SPF mice without pre-conditioning, minimizing logistical and financial constraints. While this may better mimic natural disease, it creates competition between resident and transplanted microbiota and ultimately limits accuracy. Unconditioned mice show poor engraftment of human microbes during FMT [76] likely due to colonization resistance that acts to oppose engraftment of potentially harmful invading pathogens [77]. Additionally, the microbiome of SPF mice frequently exhibits significant cage-to-cage variability both before and after FMT [78]. We direct the reader to other informative reviews that highlight complexities of host-microbe physiology that modify the stability and efficacy of FMT in humans [79, 80].

Given the available literature describing poor engraftment efficiency and colonization resistance, we do not recommend the use of unconditioned mice for human-to-mouse FMT studies. The decision to use germ-free animals or antibiotic pre-treatment cannot be prescribed. Germ-free models may not be accessible, and both methods can affect immune and CNS development. We urge investigators to carefully weigh the benefits and limitations of germ-free models or antibiotic treatment for their specific experimental paradigm. Regardless of method, we recommend collecting DNA from fecal samples before and after FMT to understand microbial community structure across cages and between experimental treatment groups. Moreover, microbial profiling of the inoculum and the recipients would help assess differences in engraftment efficiency between the recipients.

Optimal antibiotic regimen for microbiota depletion

Across studies, different antibiotic regimens were used for microbiota depletion, varying in duration and administration method (drinking water or gavage). Commonly used antibiotics included ampicillin, ciprofloxacin, imipenem, kanamycin, metronidazole, neomycin, and vancomycin, sometimes combined with the antifungal amphotericin B. Notably, metronidazole was used in all studies except one (Ritz et al. [22]).

The most common regimen combined metronidazole, ampicillin, neomycin, and vancomycin to target gram-positive, gram-negative, anaerobic bacteria, and methicillin-resistant Staphylococcus aureus (MRSA). However, the rationale for combining antibiotics with overlapping spectra remains unclear. For example, the simultaneous use of neomycin and kanamycin, both primarily active against gram-negative bacteria, or ampicillin and imipenem, both β-lactam antibiotics, appears redundant and may unnecessarily increase the risk of off-target effects and complexity.

Some studies added amphotericin B to the antibiotic cocktail, as fungi are part of the gut flora and antibiotics alone may lead to fungal overgrowth [81]. Though less studied than the microbiome, the mycobiome may influence behavior [82], and fungal-bacterial interactions are well documented [83]. In humans, amphotericin B can cause nephrotoxicity within days [84]; similar effects occur in mice but may be reduced at lower doses [85]. Compared to intravenous administration, the oral bioavailability of Amphotericin is very low (0.2–0.9%) due to its eponymous amphoteric nature [86]. Oral use has effectively reduced fungal load with minimal side effects, even at high doses [87–90].

It is also worth noting that the gut contains a virome and phageome [91] which are not targeted by antibiotic cocktails. Due to their nature, effective antiviral regimens are challenging to implement and are not widely used in FMT studies.

Studies comparing drinking water and oral gavage in mice show differences in efficiency and consistency. Two studies highlight gavage as more effective for microbiota depletion. Ochoa-Reparaz et al. found that gavage improved antibiotic tolerance compared to ad libitum delivery [92], and Hill et al. showed that daily gavage more effectively reduced bacterial 16S DNA than ad libitum protocols [93]. In contrast, a study by Tirelle et al. demonstrated that administering antibiotics either by twice-daily oral gavage or through drinking water results in a more robust and consistent depletion of fecal bacteria in mice compared to once-daily oral gavage [90]. Two other studies support twice-daily gavage, though they did not compare with once-daily gavage: Reikvam et al. and Zarrinpar et al. showed twice-daily gavage effectively depleted cultivable fecal microbiota and reduced bacterial DNA 400-fold [89, 94]. Additionally, the study from Reikvam et al demonstrated that ad libitum administration in the drinking water increases mortality and decreases water consumption.

Though no method is conclusively superior for human-derived microbial engraftment, the most effective appear to be antibiotics in drinking water or twice-daily gavage. However, oral gavage is not without risks: stress, injury, aspiration pneumonia from misplacement into the trachea [95], and esophageal damage [96, 97]. Antibiotics administered in drinking water also show high mortality, possibly due to bitter taste (e.g., metronidazole) and the resulting reduced intake [89].

Regarding the treatment duration, in the studies analyzed, it ranged from a minimum of 7 days up to a maximum of 28 days. Tirelle et al. found that extending antibiotic treatment beyond 4 days did not further reduce fecal bacterial loads. No significant differences were observed on Days 4, 7, and 12 in either gavage or drinking water groups. Additionally, it was noted that fungal overgrowth occurs between Days 7 and 12, despite the use of amphotericin-B in the antibiotic cocktail [90]. Another study by Amorim et al. found that 7-day antibiotic treatment was most effective for gut decontamination, with no added benefit from continued treatment up to day 21 [98].

Based on current evidence, we recommend a 7-day antibiotic regimen with twice-daily gavage (10 ml/kg), performed by a trained investigator. The cocktail should include metronidazole (10 mg/mL), ampicillin (10 mg/mL), neomycin (10 mg/mL), vancomycin (5 mg/mL), and amphotericin B (0.1 mg/mL). FMT should follow 12–72 hours after the last antibiotic dose to avoid antibiotic-inoculum contact and minimize the risk of recolonization [90].

Pooled, or not pooled, that is the question

The use of pooled or individual samples for performing FMT varied significantly between studies. 17 studies used pooled feces, while 13 studies used feces from a randomly selected patient. As one might expect, there are significant differences in the microbiome between patients. To further complicate the issue, an individual’s microbiome and microbial metabolites can change from day to day [7]. Some studies have investigated the utility of pooling samples in an effort to decrease variability and perform FMT with a more consistent donor sample [99–101]. Indeed, these studies show that pooling donor fecal samples reduces the variability seen from individual donors.

It is important that investigators recognize what is being modeled with FMT and what the implications of pooling are. The goal of FMT is to determine if there is a correlation between the microbiome and a specific phenotype. Previous perspectives have framed FMT using the terms biologic unit (BU) and observational unit (OU). In this context, donor samples are BUs and the recipient mice are OUs [102]. Combining donor samples creates an artificial BU; while this may eliminate individual variability, it is also unrepresentative of the microbiome of the patient population it is attempting to replicate [6]. This experimental design philosophy emphasizes analysis of the OU at the expense of the BU. It reduces the sample size (N) of the BU to 1. Although there is substantive immunologic and behavioral variability that manifests between co-housed mice with identical genetic backgrounds, increasing the number of mice that receive a given microbial inoculum only increases OUs; this is analogous to increasing technical replicates only and has been termed “pseudoreplication” [103]. Therefore, pooling donor samples does not provide a rigorous experimental paradigm to answer how the microbiome correlates to disease.

We recommend performing FMT from individual donors over pooling samples. While using samples from individual donors is preferred over pooling samples, there are major logistical hurdles to overcome. As described previously, investigators should consider mice to be technical replicates; the sample size is equivalent to the number of patients, not the number of mice. Carrying this logic forward, power analysis and sample size determination should be performed using the number of donors. The number of technical replicates used is generally determined by the variability of the assay. In this paradigm, the assay is engraftment of the donor microbiome. There is an added layer of complexity given the natural variability found among rodent models of the same strain. A separate power analysis must be performed for the specific assay of interest (eg behavioral tests). Performing a priori power analysis of both the assay of interest and engraftment will ensure experiments are sufficiently powered.

How many gavages are necessary for optimal gut colonization?

FMT protocol had the greatest variability in all studies analyzed. Amongst the 30 studies, there were 30 different protocols. Calls for standardizing FMT methodology have become increasingly more prevalent [4, 98, 104]. Studies have repeatedly shown that difference in FMT protocols have a material effect on the outcome.

The main goal of FMT is stable microbiome engraftment. While it seems intuitive that repeated FMTs ensure stability, studies show that a single FMT in germ-free mice can maintain engraftment for 28–60 days [59, 104]. However, stability may vary by housing or mouse model, and single inoculations may be less effective in mice with an existing microbiome. On the other hand, increasing FMT frequency offers no added benefit [105] and may even disrupt graft stability [104].

In both germ-free and antibiotic-treated mice, engraftment takes time and varies across the gut. Several studies agree that approximately 28 days are needed for stability [4, 6, 59, 104–106].

It is crucial that investigators report the engraftment efficiency of their FMT and the method of validating engraftment efficiency. The equivalent analogy in other fields of research would be omitting a western blot showing the protein of interest has indeed been knocked out. Bray-Curtis dissimilarity and weighted/unweighted UniFrac distances are frequently used to measure engraftment efficiency. A detailed discussion of the benefits and shortcomings of these methods is outside the scope of this review. We point the readers to recently published reviews and studies that comprehensively address the subject [79, 107–109].

Our recommendations differ slightly between antibiotic-treated and germ-free mice. In studies not using germ-free animals (ie antibiotic or laxative pre-treatment), performing FMT once per week for 3–4 weeks seems to provide the best balance of engraftment stability without overly perturbing colonization [6, 98, 104, 110]. In studies using germ-free animals [59, 105], performing a single gavage appears to be sufficient to achieve stable engraftment. Increasing FMTs to three times in the first week (day 0, 4, and 7) may increase similarity to the donor and does not perturb stability. It is also crucial to report engraftment efficiency to confirm the stable transfer of the donor microbiome and to facilitate power analysis.

How long should we wait after colonization before performing behavioral tests?

While behavioral test selection is beyond this review’s scope, the timing of tests post-FMT is relevant. Studies varied widely, with tests starting immediately after the last gavage to three weeks later, and some continuing gavage during testing. Though no standard waiting time exists, gavage is stressful, and time is needed for engraftment and stabilization. Additionally, behavioral testing in non-sterile settings may introduce external microbes, increasing variability and reducing interpretability.

Gavage is a known stressor for lab animals. Brown et al. showed that gavage with corn oil in rats raises corticosterone in a dose-dependent manner [111]. Balcombe et al. found gavage, handling, and blood collection trigger stress responses like elevated blood pressure and altered behavior, with no habituation [112]. Additionally, Bonnichsen et al. reported gavage increases heart rate, blood pressure, and body temperature in rats, with greater distress at higher doses [113].

Given the stress caused by gavage, choosing an appropriate delay before behavioral testing is essential. Gavage can elevate corticosterone, mean arterial pressure (MAP), heart rate, and alter behavior for hours. Walker et al. found MAP stayed high for up to 5 hours post-gavage, and fecal corticosterone rose only in the gavage group, indicating hypothalamic-pituitary-adrenal (HPA) axis activation. However, MAP and heart rate normalized within 24 hours [114]. Similarly, Jones et al. found that stress-related markers like corticosterone and immune cell ratios were unaffected 24 hours post-gavage, especially when animals were anesthetized, which could reduce acute stress responses [115].

A minimum waiting period of 24 hours following the final gavage is recommended prior to initiating behavioral assessments. This interval allows for the normalization of stress-related physiological parameters, thereby reducing the potential for confounding effects. Omitting this delay may compromise the validity of behavioral outcomes due to the influence of residual stress responses.

Conclusion

Microbiome research is a relatively recent field compared to other areas of biomedical science. Consequently, research methods exhibit significant heterogeneity. This systematic review highlights the need for standardized methodologies and offers evidence-based recommendations to achieve this goal. We encourage future microbiome-gut-brain axis researchers to adopt these recommendations to prioritize methodological transparency and we encourage journal editors to require the consistent reporting of these experimental variables in published articles. Enhancing research rigor and reproducibility is integral to our shared goal of advancing translational research. Our full recommendations are provided in Table 1.

Table 1.

Key recommendations.

Topic	Recommendations
Psychiatric diagnosis	• Enroll patients with diagnoses based on the diagnostic criteria of the latest version of the DSM. • Characterize them from a psychopathological perspective using validated clinician-administered scales
Mouse/rat model	• Consider differences in behaviour of WT mice • Consider the use of non-WT mice to represent psychiatric condition being studies
Germ-free state	• Investigators should carefully consider the advantages and disadvantages of each model system. • We do not recommend using mice without pretreatment for human-to-mouse FMT studies.
Antibiotic regimen	• For the drug cocktail, we recommend metronidazole (10 mg/mL), ampicillin (10 mg/mL), neomycin (10 mg/mL), vancomycin (5 mg/mL), and amphotericin B (0.1 mg/mL). • It is essential to verify microbiota depletion following treatment, for example by performing 16S rRNA qPCR or sequencing on fecal samples, to confirm the efficacy of the decontamination protocol before proceeding with FMT or other experimental interventions.
Method of antibiotic administration	• 7-day regimen using twice-daily gavage with a gavage volume of 10 ml/kg. This recommendation is contingent on the presence of a well-trained investigator that can safely and consistently perform oral gavage.
Pooled vs single donor	• We recommend performing FMT from individual donors over pooling samples.
Donor/mouse ratio	• Researchers should treat mice as technical replicates, with the sample size reflecting the number of donors (patients/microbiomes) rather than the number of mice. • Power analysis and sample size calculations should focus on the number of donors. • The number of mice per donor depends on the variability of the assay, which in this case is the engraftment efficiency of the donor microbiome. • Consideration should also be afforded to intrinsic variability within a mouse strain.
Time after antibiotic treatment before the first gavage	• FMT can be performed between 12–72 hours following final administration of antibiotics. This provides sufficient time to avoid direct contact between the antibiotics and the inoculum without risking recolonization with environmental microbes.
Number of gavages for colonization	• In studies not using germ-free animals (i.e. antibiotic or laxative pre-treatment), performing FMT once per week for 3–4 weeks seems to provide the best balance of engraftment stability without overly perturbing colonization. Verification of engraftment is strongly recommended, ideally by profiling the murine microbiota (e.g., via 16S rRNA sequencing or qPCR) and comparing it to the human donor microbiota to confirm colonization success and fidelity. • In studies using germ-free animals, performing a single gavage appears to be sufficient to achieve stable engraftment; however, engraftment should still be verified by assessing the similarity between the recipient murine microbiota and the original human donor microbiota before initiating behavioral or biological assessments. • Bray-Curtis dissimilarity index and UniFrac distances are amongst the most common methods for assessing efficiency.
Time after colonization before performing behavioral tests.	• We recommend a waiting period of at least 24 hours after the last gavage administration may be ideal before starting behavioral testing. This timeframe allows physiological stress markers to normalize, thereby minimizing the influence of gavage-related stress on the outcomes of interest.

This table summarizes key suggestions aimed at improving research methodology and clinical practice considering the evidence discussed in this review.

Author contributions

A.M.D. and A.G.N. conceptualized the study, performed the literature search, wrote the main manuscript text, and prepared figures. A.B., G.C., F.S., and F.C. provided expertise contributing to the final recommendations and edited the manuscript. All authors read and reviewed the final manuscript.

Funding

A.G.N. and F.C. were supported by the National Institute of Diabetes and Digestive and Kidney Diseases, NIDDK097948. A.B. was supported by R01AG085316, 5R03AG080175-02, and Target ALS New Academic Investigators Award.

Data availability

All data are available in the main text or the supplementary materials.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-026-03847-4.