Fecal Microbiota Transplantation (FMT) From a Human at Low Risk for Alzheimer's Disease Improves Short‐Term Recognition Memory and Increases Neuroinflammation in a 3xTg AD Mouse Model

ABSTRACT

Human microbiota‐associated murine models, using fecal microbiota transplantation (FMT) from human donors, help explore the microbiome's role in diseases like Alzheimer's disease (AD). This study examines how gut bacteria from donors with protective factors against AD influence behavior and brain pathology in an AD mouse model. Female 3xTgAD mice received weekly FMT for 2 months from (i) an 80‐year‐old AD patient (AD‐FMT), (ii) a cognitively healthy 73‐year‐old with the protective APOEe2 allele (APOEe2‐FMT), (iii) a 22‐year‐old healthy donor (Young‐FMT), and (iv) untreated mice (Mice‐FMT). Behavioral assessments included novel object recognition (NOR), Y‐maze, open‐field, and elevated plus maze tests; brain pathology (amyloid and tau), neuroinflammation (in situ autoradiography of the 18 kDa translocator protein in the hippocampus); and gut microbiota were analyzed. APOEe2‐FMT improved short‐term memory in the NOR test compared to AD‐FMT, without significant changes in other behavioral tests. This was associated with increased neuroinflammation in the hippocampus, but no effect was detected on brain amyloidosis and tauopathy. Specific genera, such as Parabacteroides and Prevotellaceae_UGC001, were enriched in the APOEe2‐FMT group and associated with neuroinflammation, while genera like Desulfovibrio were reduced and linked to decreased neuroinflammation. Gut microbiota from a donor with a protective factor against AD improved short‐term memory and induced neuroinflammation in regions strategic to AD. The association of several genera with neuroinflammation in the APOEe2‐FMT group suggests a collegial effect of the transplanted microbiome rather than a single‐microbe driver effect. These data support an association between gut bacteria, glial cell activation, and cognitive function in AD.

Transplanting gut microbiota from a human donor with the APOEe2 allele, protective against Alzheimer's disease (AD), improved visual memory in a mouse model of AD and increased neuroinflammation in the hippocampus. Parabacteroides and Prevotellaceae_UGC001 genera were more prevalent in mice receiving APOEe2 microbiota and linked to the observed neuroinflammation.

1. Introduction

Alzheimer's Disease (AD) is characterized by extracellular accumulation of beta‐amyloid (Aβ), intracellular deposition of hyperphosphorylated tau, and neuroinflammation [1]. Like other common chronic conditions, AD likely develops due to complex interactions among multiple factors, including age, genetics, environment, lifestyle, and coexisting medical conditions. On one hand, risk factors such as old age, female biological sex, genetic variants (e.g., APOEe4 allele), cardiometabolic diseases (e.g., hypertension, obesity, diabetes), traumatic brain injury, disrupted sleep, social isolation, hearing loss, excessive alcohol consumption, smoking, and air pollution have been consistently associated with an increased risk of developing Alzheimer's pathology. On the other hand, genetic factors (such as the APOEe2 allele), a high level of education, a healthy diet (e.g., Mediterranean diet), and engaging in diverse physical and social activities have been shown to reduce the incidence of AD [2, 3, 4, 5]. Despite these known factors, the mechanisms underlying susceptibility or protection remain elusive, and the combined effects of all the above‐mentioned require further investigation.

The term gut microbiota (GM) refers to the symbiotic microorganisms that colonize the human gut. GM's involvement in promoting AD onset and progression is supported by a large and growing body of both preclinical and clinical studies [6, 7, 8, 9]. Interestingly, many AD risk factors are also linked to alterations in gut microbiota. For instance, studies analyzing GM compositions in older individuals have found a general decrease in bacterial alpha diversity, together with an increase in opportunistic agents that could be linked to age‐related chronic diseases [10, 11]. Similarly, cardiovascular disease, obesity, and alcohol consumption have been associated with GM dysbiosis [12, 13, 14, 15].

Additionally, the main genetic risk factor for AD, the APOEe4 genotype, has been associated with differentially abundant genera depending on the carried allele in several human and rodent studies, although no consensus was observed [ ¹⁶ , ¹⁷ , ¹⁸ ].

FMT is now accepted as treatment for certain pathological conditions (e.g., Clostridium difficile infection) and has recently shown promising results in treating other neurodegenerative diseases such as Parkinson's disease [19].

Our hypothesis is that GM communities isolated from donors with different factors of protection against AD pathology can directly affect behavior and brain pathophysiology in a transgenic (Tg) mouse model of AD. We generate HMA 3xTgAD mice with bacteria isolated from stools of (i) an amyloid‐negative cognitively healthy APOE2 subject (genetic‐related protection), (ii) a young healthy donor (age‐related protection), and (iii) an amyloid‐positive AD patient (no protection).

The overall aim of this study is to identify if FMT from individuals with protective factors against AD can cause cognitive and pathophysiological improvement and what would be the associated genera.

2. Materials and Methods

2.1. Human Donor Selection

FMT donors were selected from a cohort at the Geneva Memory Center, ranging from cognitively unimpaired individuals to those with dementia. The cohort is part of the gMAD study (Geneva Ethics Committee Ref: CCER_2016–01346), which collects variables like brain amyloidosis (via amyloid‐PET), APOE genotyping, and neuropsychological assessments. Three participants with varying protective factors against AD were chosen as described in Figure 1A.

FIGURE 1.

Experimental settings and behavioral tests. (A) Demographic table of the FMT donors. (B) Experimental setting: Animals were administered acid suppressants and FMT treatments for 2 months after 2 weeks of antibiotic treatment and one administration of laxative. Behaviors were measured at the end of the treatment period. (C, D) Discrimination index (C) and novel object recognition index (time spent with novel object—known object in sec). (D) from the novel object recognition test. (E)% of good alternance in the Y‐maze test. (F) Open field test, distance done in 60 min (cm), and (G) time spent in the center (min). (H) Elevated plus maze test, arm preference ratio (open/close). Statistics: T‐test with *p < 0.05. Bar plots represent mean and SD. Abbreviations: ABT, antibiotics; AD, Alzheimer's disease; COPD, chronic obstructive pulmonary disease; FMT, Fecal microbiota transplantation; HCTZ, candesartan cilexetil + hydrochlorothiazide; HTN, hypertension; LAX, laxative; na, not available; PET, positron emission tomography; PMR, polymyalgia rheumatica.

2.2. Study Design and Experimental Timeline

Twelve‐months‐old female 3xTgAD (APP^SWE, PS1^M146V and Tau^P301L) mice (B6;129‐Tg(APPSwe, tauP301L)1LfaPsen1tm1Mpm/Mmjax) were kept in a conventional facility under a 12 h light/dark cycle with food and water access ad libitum and housed by groups of 2–3 mice per cage. They were bred with C57BL/6J, and the genotype was validated to select the animals. Animals were randomly assigned and tested blind to experimental conditions. Experimental approval was obtained from the ethics committee for animal experimentation of the canton of Geneva, Switzerland.

In order to maximize the colonization [20], fecal microbiota transplantation (FMT) protocol was performed as follow, (1) Antibiotic treatment for 14 days in the drinking water (100 μg/mL ampicillin, 100 μg/mL cefoperazone sodium salt, 100 μg/mL clindamycin hydrochloride), renewed 3 times per week. (2) Laxative administration (Moviprep 30 mg/kg), 24 h before a high dose of FMT (10⁹cells in 100 ul). (3) Then, a stomach acid suppressant (Omeprazole 30 mg/100 mg Madidrop Sucralose) was administered once every working day and a low dose of FMT (10⁶cells/100 ul) once per week for 2 months (Figure 1B).

Behavioral tests were performed during the light phase 1 week before the sacrifice. After 2 months of treatment, mice were sacrificed by intracardiac saline perfusion. The brain was collected, and one hemisphere was used for TSPO autoradiography; the other was fixed with 4% paraformaldehyde for immunohistochemistry analysis. Cecum was collected and stored at −80°C until further use for microbiome analysis.

2.3. FMT Preparation

The selected donors provided fresh stools in a Feconcontener that were processed within 2 h. For the mice‐FMT, stools were collected from the same mice before the initiation of the experiment. All mice were placed in clean cages for 1 h, and fecal pellets were all collected and pooled for FMT preparation, once just before the start of the experiment (10 mL falcon tube filled up). The donated material was entirely processed under anaerobic conditions, following standardized methods used for C. difficile infection treatment. 50 g of stool sample was homogenized in 250 mL of sterile anaerobic PBS and filtered through 2, 1, 0.5, and 0.25 mm sieves. The resulting material was centrifuged at 6000 × g for 15 min and resuspended in 20 mL of sterile anaerobic PBS. Living bacteria were counted with the Live/Dead Backlight Bacterial Viability and Counting Kit (ref. L34856, invitrogen), and diluted to 10⁹cells/100 ul and 10⁶cells/100 ul in sterile anaerobic PBS 10% glycerol before aliquoting and freezing at −80°C until further use. 100ul of FMT preparation, thawed for 1 h before use, was administered by gavage with a sterile plastic cannula.

2.4. Behavior Tests

The first day, the mice were place in an open field box (45 × 45 × 45 cm) for 60 min in order to measure general locomotor activity (measured by the distance covered in 60 min), and estimate the mice anxiety level by the time spent in the center [21]. This test was used as habituation for the novel object recognition (NOR) test performed the day after [22]. Mice were placed 3 times for 5 min separated by 10 min in an open field with 2 objects distinct in shape and color. After 10 min, they were placed a fourth time in the open field with one of the two objects modified. Zones 2 cm larger than the objects were drawn on the analysis software to define the measurement areas. The analysis detects the head, body and tail of the animals. The time taken for the head to be detected in the measurement zones was used as an index of object exploration. The time spent with the new object minus the time spent with the object used during habituation represents the object recognition index and provides a readout of short‐term recognition memory. Discrimination index to evaluate no bias was present in the habituation period measured as the average delta of the time spent with each object during the 3 sessions. On the next day, they were placed in a Y‐maze (20 × 5 × 30 cm) for 5 min. The animal was positioned in one of the arms and remained free of movement for 5 min. The percentage of spontaneous alternations in the Y‐maze is used to assess spatial working memory [23]. The last day, the anxiety index was evaluated using the elevated plus maze, a plus‐shaped apparatus elevated 50 cm above the ground. It consisted of two open arms (20 cm × 5 cm long and width, respectively), two enclosed arms (20 cm × 5 cm × 30 cm, long, width, and high, respectively), and a middle compartment (5 cm × 5 cm) [24, 25]. The animal was positioned in the central area and remained free of any movement for 5 min. The number of entries in the open versus closed arm was used to measure anxiety level. A camera placed above the mazes and open field box recorded the movements of the animals, and the Ethovision (Noldus) software was used for automatic analysis of behaviors. All behavioral tests were performed in dim light settings.

2.5. In Situ TSPO Autoradiography

Brain TSPO autoradiography, including [¹²⁵I]CLINDE synthesis, was performed as previously described [26]. Briefly, brain hemispheres sections (20 μm) were immersed in a Tris‐MgCl2 buffer alone (20 min), then in the same buffer containing [¹²⁵I]CLINDE (0.11 MBq/mL, 90 min). Non‐specific binding was estimated by the presence of 10 μM of unlabeled CLINDE on adjacent sections. Slides were then exposed to gamma‐sensitive phosphor‐imaging plates (Fuji BAS‐IP MS2325). Autoradiograms were analyzed with Aida Software V4.06 (Raytest Isotopenmessgerate GmbH) together with home‐made calibration curves.

2.6. Immunohistochemistry and Immunofluorescence

Paraffin‐embedded brain blocks were cut on a microtome (5 μm). After an antigen retrieval procedure (95°C, 20 min, 0.1 M Tris‐Citrate buffer), sections were incubated with primary antibody, 4G8 (Biolegend 800,710, 1/500), and AT8 (Thermofisher MN1020, 1/500) in PBS 0.2% Triton X‐100 1% BSA for 48 h incubation at 4°C. The secondary antibody was added for 90 min at RT, and slides were mounted with Fluorsafe. Images were acquired with the Zeiss Axio Scan.Z1 slide scanner and analyzed with QuPath 0.5.0 software.

2.7. Microbiota Analysis

DNA was extracted with the MagPurix Bacterial DNA Extraction kit using the automated Zinexts system from the donor FMT preparation (1 mL of 10⁹cell/100 ul preparation was spined and resuspended in RLT buffer) and 50 mg of frozen cecum content from the mice, according to the manufacturer's instructions. Prior to DNA extraction, stool samples were mechanically disrupted with bead‐beating using the Precellys lysing kit soil grinding SK38, 3 × 40 sec at 6000 rpm, followed by 5 min at 95°C shaking at 600 rpm. The regions V3 and V4 [27] of the bacterial 16SrRNA gene were amplified and purified according to the 16S Metagenomic Sequencing Library Preparation protocol by Illumina. The sequencing was performed on the Illumina Miseq Sequencer with a 300 bp paired‐end run using the MiSeq V3 reagent with 20% PhiX and a loading concentration of 5,25 pM.

The raw demultiplexed sequences were processed using QIIME2 [28] (version 2021.8) run on a high‐performance computing (HPC) platform. The reads were quality filtered and merged using the QIIME 2 Plugin dada2 [29] (version 2021.8.0). Silva138.1 SSU Ref NR99 database was formatted with the QIIME2 rescript plugin and adapted to the 16S V3‐V4 region and thereafter used to classify the reads with the feature‐classifier plugin from QIIME2 (classifysklearn method) [30]. Sequences classified as mitochondria, chloroplasts, or eukaryotes and amplicon sequence variants (ASVs) without phylum‐level assignment were filtered from the data sets. Samples with less than 100,000 reads were removed from further analyses.

2.8. Statistical Analysis

Alpha diversities, indicating the richness and abundance of ASVs within each individual, were calculated on a rarified data set (rarified to 102,300 reads) using the QIIME2 pipeline. Beta diversity distance metrics showing the similarity or difference in microbiota composition between individuals (Bray Curtis distance) were calculated on relative abundance data sets and visualized in R using the phyloseq package (version 1.44.0) [31]. Statistics were performed with Permanova using the adonis function in the vegan package (version 2.6–8) and pairwise comparison using the pairwiseAdonis package (version 0.4). Maaslin2 (version 1.14.1) [32] was used for differential abundance analysis on a rarefied dataset with TSS normalization, prevalence set to 15%, and cage household set as a random effect. Statistical analyses were performed using R (v.4.4.0 and v.4.3.0) and GraphPad Prism. The comparison among groups was performed using standard t‐tests or ANOVA with Tukey post hoc testing when multiple groups were compared. Associations of genera, behavioral and biological variables were assessed with Spearman correlation. Weighted engraftment efficiency was measured as the percentage of the sum of donor ASV reads present in recipient mice/ total rarefied number of reads (102,300).

3. Results

3.1. APOEe2‐FMT Improves Short‐Term Recognition Memory in the Novel Object Recognition Test

Donors were carefully selected to highlight characteristics that reflect different levels of protection to AD (they are referred to throughout the manuscript as the main characteristic trait of this protection): (i) AD: no protection to AD, with clinical diagnosis of dementia, brain amyloidosis, APOEε4 carrier genotype, female gender, and advance age (80 yo) (ii) APOEe2: genetic‐related protection to AD (APOEε2 carrier), associated with no cognitive decline diagnosed, no brain amyloidosis, a high level of education, male gender, and (iii) Young: age‐related protection to AD (22 yo), with no APOE gene‐related risk (ε3/ε3), no cognitive complaints, male gender. Other lifestyle and dietary habits, medication, and medical history that could influence the GM composition were also collected and described in Figure 1A. Stool samples isolated from the mice before treatment were used as a control for FMT (mice). We used a published antibiotic‐based model that has been proven to result in stable engraftment of human microbiota in mice [33] to generate a human microbiota‐associated 3xTgAD mouse model (Figure 1B). The overall engraftment efficiency was moderate for the human donors, with a mean of 12.1% (SD: 2.1); 18.5% (SD: 11.8) and 19.6% (SD: 7.7) for the APOE‐e2‐FMT, AD‐FMT, and Young‐FMT groups, respectively. This moderate efficiency is expected in a human microbiota‐associated mouse model due to interspecies differences in microbiome composition [34]; the age of the mice can also reduce engraftment efficiency [35], and the use of cecal readouts following fecal transplantation [36]. As anticipated, this percentage was higher in mice—92.6% (SD: 2.4)—that received their own microbiota transplantation.

While there is no object preference during habituation sessions (assessed with ANOVA and t‐tests comparison), APOEe2‐FMT was significantly associated with improved short‐term recognition memory compared to the AD‐FMT (p = 0.033 at the NOR test) (Figure 1C,D). In contrast, the spatial working memory index (% of alternation behavior in the Y‐maze test) did not differ among groups (Figure 1E). The open field test indicated that the AD‐FMT group showed a tendency for reduced locomotor activity (as shown by the distance in 60 min, Figure 1F), although not reaching significance, and was associated with a reduced time spent in the center (p = 0.035 compared to APOEe2‐FMT, Figure 1G). This could indicate a slight increase in anxiety levels that was, however, not confirmed in the elevated plus maze test (Figure 1H).

3.2. APOEe2‐FMT Increases Neuroinflammation

To understand the mechanisms behind the improved short‐term recognition memory in the APOEe2‐FMT, TSPO in situ autoradiography was performed to assess the hippocampus neuroinflammation. APOEe2‐FMT was significantly associated with increased TSPO density compared to the other groups. Young‐FMT's TSPO density was intermediate and was observed as significantly increased compared to AD (p = 0.038) (Figure 2A). This was the case in all subregions of the hippocampus (2‐way ANOVA: FMT effect: p < 0.0001; subregions effect: p = 0.7; Figure 2B). Brain amyloidosis and tauopathy were measured by immunohistochemistry, and no statistical significance was determined among the FMT groups (Figure 2D,E). However, a slight correlation between TSPO and 4G8 signal in the subiculum was detected (Pearson correlation r = −0.34 and p = 0.05; Figure 2F).

FIGURE 2.

Neuroinflammation and AD pathophysiology. (A–C) In situ autoradiographies of TSPO with the specific binding ratio (density) in the hippocampus (A) and its subregions (subiculum, dorsal and ventral hippocampus) (B) and its representative image (C). (D) Amyloid density in the subiculum measured by % 4G8 labelled area. (E) pTau density in the hippocampus measured by % AT8 labelled area. (F) Scatter plot of the Pearson correlation between amyloid and TSPO density in the subiculum. Statistics: (A) ANOVA and (B) 2‐way ANOVA with Tukey's multiple comparison; *p < 0.05; **p < 0.01; ***p < 0.001. Barplots represents mean and SD.

3.3. Bacterial Genera Associated With APOEe2‐FMT Are Also Associated With Neuroinflammation

Cecal contents were collected after 2 months of recurrent FMT of the different donors, and microbiota characterization was performed using 16S rRNA library preparation and sequencing. Beta diversity was measured with Bray Curtis distances, and as expected, the human donors cluster differently from the mouse recipients due to the differences in species and the different sections of the gastrointestinal tract (cecal for the recipient's vs. colonic for the donors) (Adonis test for FMT group: Pr(>F) = 0.001 and pairwise comparison of Mouse versus Young, AD, and APOEe2 with adjusted p value of 0.006, 0.006, and 0.042, respectively). However, no clear segregation was observed among the human FMT recipients (Figure 3A). Alpha diversity was measured by the Shannon entropy, and a significant reduction in diversity was observed in the APOEe2‐FMT group compared to the Mice‐FMT and AD‐FMT groups (Figure 3B). This was in line with the APOEe2 donor preparation that indicated a reduced Shannon entropy compared to the other donors. The mice donor preparation had an increased diversity eventually due to the pooling of different donors from the same cage. The Maaslin2 package in R was used to measure the differential abundance of the genera between the FMT recipient mice. Two approaches were used, one comparing the groups with the mice‐FMT group as a neutral control (Figure 3C) and one comparing with the AD‐FMT as a pathological control (Figure 3D). Using a detection threshold (qval < 0.02), 5 genera were associated with human‐FMT as decreased in all human‐FMT groups compared to the mice‐FMT one. Rikenella, Eggerthellaceae_uncultured, and Desulfovibrio genera were reduced only in the “protected” FMT (APOEe2 and young); three genera were increased only in the young‐FMT, Parabacteroides was increased in the APOEe2‐FMT group and both Ligilactobacilus and Coprostanoligenes groups were specifically reduced in the AD‐FMT group when compared to the mice‐FMT group (Figure 3C).

FIGURE 3.

Gut microbiota profile and association matrix. (A) Principal coordinate analysis (PCoA) indicating the beta diversity measured by Bray‐Curtis distances of cecal microbiomes from mice receiving FMTs and the fecal microbiomes of each donor. Ellipses indicate the normal distribution per group. (B) Alpha diversity of cecal microbiome from mice receiving FMTs or the fecal microbiome of each donor measured by Shannon entropy. Statistical analysis with ANOVA and Tukey post hoc testing *p < 0.05. (C, D) Heatmap depicting the coefficients of associations between the FMT groups and the significantly associated genera compared to the Mice‐FMT group (q < 0.02) (C) and the AD‐FMT group (p < 0.05) (D). Coefficients were calculated as (−log(qval)*sign(coeff)) based on the Maaslin2 R package and show enrichment in red or depletion in blue. (E) correlation matrix between the normalized abundance of the genera (log(TSS)) and the measured variables in all groups of mice receiving FMT. The association is indicated in red when positive and blue when negative. Only the statistically significant associations (Sperman test) are indicated in the corresponding cells (*p < 0.05; **p < 0.01; ***p < 0.001).

Some of these associations were also identified when the comparison was made against the AD‐FMT group, using a less stringent detection threshold (p < 0.05) (Figure 3D). Indeed, the Ligilactobacilus and Coprostanoligenes genera were reduced in AD‐FMT when compared to mice‐FMT and increased in both APOEe2 and Young‐FMT when compared to AD‐FMT. Similarly, Phascolactobacterium was significantly associated with Young‐FMT when compared to both mice and AD‐FMT, and Parabacteroides seems to be specifically associated with the APOEe2‐FMT.

Figure 3E shows a correlation matrix between the selected genera associated with both Young and APOEe2‐FMT compared to AD‐FMT (see Figure 3D) and the behavioral test and brain pathophysiology measured in the same mice (the 4 groups are included). It appears that the neuroinflammation measured in the brain with TSPO autoradiography is highly associated with specific genera found associated with the APOEe2‐FMT group. Indeed, Biophila, Muribaculum, Parabacteroides, and Prevotellaceae_UGC_001 are positively associated, and Desulfovibrio, f_Ruminococcaceae g_uncultured, and g_Anaerotruncus are negatively associated with both neuroinflammation and APOEe2‐FMT. Interestingly, the genera specifically reduced in APOEe2‐FMT also tend to be associated with increased neurofibrillary tangles and amyloid plaques and lower short‐term recognition memory (NOR test), despite no significance.

4. Discussion

In this study, we explored the potential protective properties of the GM from human donors against AD pathophysiology. We transplanted fecal microbiota from three distinct donors with different AD‐risks/protection‐associated patterns into a 3xTgAD mouse model. The primary protective feature of each group was emphasized to enhance readability throughout the manuscript: (1) a donor with a genetic protective factor (APOEε2‐FMT) (2) a donor protected by his age (young‐FMT) and an AD patient, a non‐protected donor (AD‐FMT). However, this study cannot establish a direct causal link between specific donor characteristics and the effects of FMT, as the donor should be considered in their entirety rather than through isolated attributes. The AD donor, for instance, has cumulative risk factors for Alzheimer's disease, including advanced age, female gender, brain amyloidosis, and the APOE ε3/ε4 genotype [2, 3, 4, 5]. Similarly, the APOE ε2 donor has additional factors associated with protection against Alzheimer's disease, such as a higher level of education. Other factors, including dietary and lifestyle habits, medications, and comorbidities, are also known to influence GM composition and should be considered [37]. In particular, physical activity levels differed between the donors but were also in line with their respective age categories. Although no differences in dietary habits or lifestyle were identified, the potential influence of medications and comorbidities on microbiota composition cannot be ruled out at this stage.

Nevertheless, our study reveals that FMT from the APOEe2‐FMT donor enhances short‐term recognition memory, as assessed by the NOR test, when compared to AD‐FMT. While we observed potential effects on locomotion and anxiety, these did not reach statistical significance. Previous research had already demonstrated the protective effects of FMT in APP/PS1 and 5xFAD models, where wild‐type (WT) donors' FMT led to significant improvements in cognitive function and reduction of Aβ plaques [36, ³⁸ , ³⁹ ]. Elangovan et al. also investigated FMT from young (8–10 weeks old) and old (30–32 weeks old) WT mice, finding that FMT from young donors yielded stronger improvements [ ³⁸ ]. These findings suggest that identified protective factors, such as age, also known to be associated with change in microbiota community, could induce risk mitigation to AD through the gut microbiota. In our study, FMT from the young donor did not show any discernible effect on cognition; however, we are the first to show risk mitigation to AD can happen from a donor with other protective factors (such as APOEe2).

It is interesting to note that when Shen et al. [40] transplanted microbiota from AD patients into APP/PS1 mice, they found that recipient mice performed significantly worse in cognitive measures compared to transgenic littermates who did not receive FMT. However, when they treated them a second time with healthy human donor, they observed improvements in memory, indicating that overall, the modulation of the microbiota towards a healthier or “protected” donor can improve the cognitive outcome in mice. In human, a few case studies reported significant improvement in cognitive outcome in AD patient following FMT treatment to clostridium difficile infection [ ⁴¹ , ⁴² ] (e.g., An 82 years old male scored from 20 to 29 at Mini mental states evaluation (MMSE) after the FMT from his healthy wife [41]). However, the mechanistic link between this FMT and the cognitive improvement remain elusive.

Here we measured neuroinflammation with in situ autoradiography of the TSPO protein, a readout of glial cell activation [43]. We observed that both APOEe2 and, to a lesser extent, Young‐FMTs were associated with increased neuroinflammation in the hippocampus when compared with AD or mice‐FMTs. Increasing evidence suggests a dual role of neuroinflammation in AD pathophysiology, more specifically for microglial activation that can either assist the clearance of amyloid accumulation or can promote extensive inflammation in reaction to amyloid or tau, eventually causing widespread neurodegeneration [44, 45]. Importantly, this difference might differ depending on the stage of the disease. Indeed, microglial activation seems to protect patients with MCI from deterioration but has an opposite role at the AD stage when tau pathology predominates [46]. Here we show that repeated FMT from protected donors increases neuroinflammation and tends to correlate with reduction of amyloid plaques in the subiculum. At 12 months of age, the 3xTgAD mice model has been described to be at the early stage of the pathophysiology with an increasing number of amyloid plaques appearing from 6 to 12 months of age and neurofibrillary tangles accumulation starting around 12 months [47]. This model develops cognitive impairment before the onset of brain pathophysiological features of AD [ ⁴⁷ , ⁴⁸ ]. This might explain why no differences could be detected in amyloid or tau accumulation in the brain. More analyses are needed to better characterize the nature of the neuroinflammation and the specific role of the microglial cells in this context.

When we characterized the GM of the recipient mice as well as the one of the donors. We observed that specific genera were reduced after a human FMT compared to a mouse one. This is in line with previously published literature demonstrating the species‐specific microbiomes [49]. We also noticed a reduction in alpha diversity from the APOEe2‐donor that was reflected in the recipient mice. This is counterintuitive to the generally shown healthy association of a microbiome with a high level of alpha diversity [50]. We could identify specific genera associated with either AD‐FMT, Young‐FMT, or APOEe2‐FMT. We confirmed the association of some genera already associated with AD. Desulfovibrio [7, 51] was previously shown to be more abundant in the AD population when Parabacteroides [52] was reduced in MCI and AD. We selected the differentially abundant genera compared to the AD‐FMT and proceeded with correlation matrix analysis. We observed that irrespective of the FMT administration (all groups were included in the correlations), we could identify a pattern where genera associated with the APOEe2‐FMT group were also associated with neuroinflammation and overall tended to correlate with protective features such as reduced brain tauopathy and improved object recognition index. However, the precise mechanisms through which these microbes influence pathophysiological effects remain to be elucidated.

A key limitation of this study is that the results are based on a single donor, reflecting current clinical practices in validated FMT models, such as those used to treat Clostridium difficile infections. However, although the use of a single donor limits our ability to draw conclusions about specific protective factors that may influence FMT effects, our study highlights an area for therapeutic investigations.

More studies will be required to determine if the presence of the APOEe2 allele can be a causal mediator of the effect we observed. Future analyses will focus on investigating the functional activity of the transplanted microbiome and the specific genera associated with APOEe2‐FMT to better understand these connections. Seo et al. recently identified that glial cell activation or homeostatic states can be modulated by short‐chain fatty acids (metabolites produced by the GM) in a microbiota‐ and APOE‐dependent manner and influence tau pathology in an AD mice model [16]. This study confirmed the intricate role of the gut‐brain axis, with an important mediator being the immune system. Further studies are needed to better characterize the microbiota‐associated neuroinflammation and its dual role in the pathophysiology of AD.

5. Conclusion

FMT from a donor with a genetic protective factor (APOEe2 carrier) against AD into 3xTg‐AD mice could improve short‐term recognition memory and increase neuroinflammation. This was associated with specific bacteria that showed a consistent correlation with neuroinflammation and tended to be associated with brain pathophysiology and short‐term recognition memory, without reaching significance. Together, these results confirm the dual role of neuroinflammation in the pathophysiology of AD and its intricate relation with the gut microbiome. They also support FMT as a possible therapeutic avenue for AD.

Consent

All human subjects provided informed consent.

Conflicts of Interest

The authors declare no conflicts interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.