Dietary Fiber and Microbiota Metabolite Receptors Enhance Cognition and Alleviate Disease in the 5xFAD Mouse Model of Alzheimer’s Disease

Animal model and treatment

Heterozygous 5xFAD mice (B6SJL background) carry mutant human transgenic amyloid precursor protein (Swedish mutation K670/671NL, London mutation V717I, and Florida mutation I716V) and presenilin 1 (PSEN1, M146L and L286V) under the control of the neuron-specific mouse Thy1 promoter. Gpr41/43/109a^−/−, Gpr41/43^−/−, Gpr109a^−/− WT/5xFAD mice were generated by CRISPR at Monash Genome Modification Platform as described in Gpr41/43/109a CRISPR-Cas9 KO. Mouse genotypes from tail biopsies were determined using real-time PCR with specific probes designed for each gene (Transnetyx). Both male and female mice were studied. Mice used in this study are available on request. All mice were housed in an environmentally controlled, specific pathogen-free facility (Monash Animal Research Platform) on a 12 h light-dark cycle, provided with food and water ad libitum. All experimental protocols were approved by Monash Animal Ethics Committee (Ethics #17271) and conformed to national and institutional guidelines.

5xFAD mice and WT mice were randomly divided and were fed with different diets: zero fiber diet (Specialty Feeds, SF09-028), control diet (Specialty Feeds, AIN93G), and HAMSAB diet (Specialty Feeds, SF18-086), High-fiber diet (Specialty Feeds, SF11-025), HAMSA (Specialty Feeds, SF18-084), HAMSB (Specialty Feeds, SF17-171). Zero fiber diet does not contain dietary fiber. AIN93G is the control diet with 5% cellulose. HAMSAB diet contains 15% acetylated high-amylose resistant starch, 15% butyrylated high-amylose resistant starch, and 5% cellulose. High-fiber diet contains 9.7% crude fiber and AD fiber, and all CHO as gel crisp starch. HAMSA contains 15% acetylated high-amylose resistant starch and 5% cellulose. HAMSB contains 15% butyrylated high-amylose resistant starch and 5% cellulose.

Gpr41/43/109a CRISPR-Cas9 KO

Gpr41 and Gpr43, Gpr109a, or all three GPCRs were globally knocked out in mice on WT and 5xFAD backgrounds using CRISPR-Cas9 (Monash Genome Modification Platform). The UCSC Genome Browser (https://genome.ucsc.edu/) was used to identify guide RNA target sites flanking the Gpr109a gene and Gpr41/Gpr43 gene. The following guide RNA were used: 95 bp upstream of the ATG of Gpr109a (5′-CTGACATGACATAAAGCCGA-3′) and 1282 bp downstream of the STOP codon of Gpr109a (5′-TAATCTGTATCCAGCTCGGG-3′) to knock out GPR109a gene; 868 bp upstream of the ATG of Gpr41 (5′-TACCTGTAACCCAGACGTTA-3′) and 1145 bp downstream of the STOP codon of Gpr43 (5′-AAGCCAGCTACGGGCTACAC-3′) to knock out both Gpr41 and Gpr43 genes. CRISPR RNAs (crRNA, IDT) were annealed with trans-activating crRNA (tracrRNA) to form a functional crRNA:tracrRNA guide RNA duplex. Cas9 nuclease (IDT, #1081058) was incubated with the guide RNAs to form a ribonucleoprotein complex. Cas9 nuclease (30 ng/μl) and the crRNA:tracrRNA guide RNA duplexes (30 ng/μl) were microinjected into the pronucleus/cytoplasm of the zygotes at the pronuclei stage. Injected zygotes were transferred into the uterus of pseudo pregnant F1 females. Both male and female mice were studied.

Behavioral tests

• Novel object recognition

The protocol for novel object recognition was adapted from a previous study (Leger et al., 2013). One week before testing, mice were handled 3 times/day at 5 min intervals by the experimenter. Three to four days of handling was performed until mice were well habituated and relaxed. This reduces stress and anxiety from handling, which may interfere with object exploration and recognition (Deacon, 2006). Novel object recognition consists of habituation, a familiarization session, and a test session. Mice were first placed in the empty open field (40 cm L × 40 cm W × 35 cm H, black plastic box without lid) and allowed to explore the open field for 5 min. At 24 h later, two identical objects were placed in the open field, and mice were allowed to explore the two objects freely until 20 s of total exploration time was reached. Exploration was defined when the mouse sniffed the object or touched the objects while looking at it. Sitting and climbing onto the objects or chewing the objects was not considered as exploration. After familiarization, mice were immediately removed and returned to their home cage. One hour later, one of the objects was replaced by a novel object and the test session was performed with the same procedure as the familiarization session. Mice were excluded if they could not reach 20 s exploration time within 10 min in either familiarization session or test session. The open field and objects were cleaned with 70% (v/v) ethanol between animals and between different sessions. Recognition index was calculated for each mouse, which is a reflection for recognition memory. Recognition index = Time (novel)/Time (novel) + Time (familiar). For mice performing multiple novel object recognition tests at different ages, different sets of objects were used. Objects were chosen based on the recommendations of Heyser and Chemero (2012). The two objects have similar size (both can be climbed by the mouse) and odor (odor was minimized by cleaning with 70% (v/v) ethanol), but different shapes, brightness, and texture. For the 5 and 3.5 month novel object recognition test (NOR), a symmetric tower of Lego bricks (6.3 cm H × 3.3 cm L × 3.3 cm W, built in yellow, red, green, blue, black, green brick), and a white bottle with blue bottom (7 cm H × 3 cm L × 3 cm W) were used. For the 6 month NOR, a pink toy bear (5 cm H × 5 cm L × 4 cm W) and an asymmetric Lego tower (5.7 cm H × 5.1 cm L × 5.1 cm W, built in blue, green, yellow, red Lego bricks) were used. For the 8 month NOR, a transparent ball with pink bottom (6 cm H × 4 cm L × 4 cm W) and a pink toy pig (5.3 cm H × 4 cm L × 4.5 cm W) were used. All behavioral tests were performed in the dark phase (active phase between 19:00 P.M. and 23:00 P.M.). To minimize the stress caused by bright lighting, dim lighting illuminated at ∼15 lux at the center of the NOR box was used. Mouse behavior was video-recorded using Mi Home Security Camera 360° 1080P (MJSXJ02CM). Data were analyzed manually by an experimenter who was blinded to object novelty and genotype.

• Spontaneous T maze alternation

The protocol for spontaneous T alternation was adapted from previous reports (Deacon et al., 2002; Deacon and Rawlins, 2006). Spontaneous T maze alternation was used to assess the short-term memory of mice. The T maze consisted of one start arm and two goal arms (each arm is 30 cm L × 10 cm W × 20 cm H). A removable central partition (7 cm L × 20 cm H) extended into the start arm. Guillotine doors could close the goal arm entrances from the choice area. The floor was covered with a thin layer of wooden bedding to facilitate running (Deacon et al., 2002). First, the central partition was placed in the T maze, and all guillotine doors were raised. A mouse was placed at the end of the start arm facing away from the goal arms. The mouse was allowed to choose freely between two goal arms. Once the mouse entered a goal arm, the mouse would be confined in the chosen arm for 30 s by closing the guillotine door. After 30 s, the central partition was removed and guillotine doors were all raised, the mouse was returned to the end of the start arm and allowed to choose freely again. The alternation was recorded. Successful alternation was defined as the mouse entering the goal arm that was not visited before. Criteria for entering the goal arm were met when the whole animal, including the tail, entered the goal arm and stayed in the arm for at least 2 s. When an animal immediately left the goal arm after entering, this was not considered as entering. After each trial, any urine or feces were removed before starting the next trial. Mice from the same cage used the same wooden bedding, and the bedding was changed when starting a new cage of mice. Each mouse performed 10 trials over 3 d (four trials on the first day, three trials on the second and third day) with a minimum 30 min interval and the alternation rate was calculated. Alternation rate = number of successful alternations/number of total trials. Any trial that could not be completed within 3 min was aborted. Alternation rate provided an index for short-term memory. All behavioral tests were performed in the dark phase (active phase between 19:00 P.M. and 23:00 P.M.). To minimize the stress caused by bright lighting, dim lighting illuminated at ∼15 lux at the center of the T maze was used. The experimenter was blinded to genotype and diet treatment.

Brain immunohistology

Mice were humanely killed by CO₂ asphyxiation, then transcardially perfused with ice-cold PBS. Brains were harvested; left brain hemispheres were dissected and fixed in 4% PFA in PBS overnight at 4°C. Right brain hemispheres were dissected into hippocampus and cortex, followed by immediate freezing in liquid nitrogen and stored at −80°C. Fixed brains were cryoprotected by incubating in 30% (w/v) sucrose solution for 24-48 h at 4°C and then frozen in liquid nitrogen. Frozen brains were stored at −80°C until use. Frozen brains were sectioned into 40- or 20-μm-thick sagittal sections, which were stored in anti-freeze solution (16.7% (w/v) sucrose, 33% (v/v) ethylene glycol, Sigma, #293237) in 0.05 m PB, pH 7.6) at −20°C until use.

Before performing immunofluorescent staining, brain sections were washed 4 times with PBS for at least 15 min at room temperature. Sections were then permeabilized and blocked for 1.5 h in PBST (1 × PBS containing 0.05% (v/v) Tween-20, Biochemicals, #BIO0777) with 0.5% (v/v) Triton X (Sigma, #T9284) and 2% (w/v) BSA (Sigma, #A1470) followed by 1 h incubation with Mouse on Mouse Blocking reagent (Vector Laboratories, MKB-2213, 1:200). Sections were stained with primary antibodies overnight at room temperature. Primary antibodies (diluted in 1% (w/v) BSA and 0.25% (v/v) Triton X) include the following: rabbit anti-Iba-1 (Wako, #019-19741, 1:500), mouse anti-6E10 (Biolegend, #803001, 1:500), rabbit anti-doublecortin (DCX, Cell Signaling, #4604, 1:200). The following day, brain sections were washed 3 times with PBST (10 min each). Then the sections were incubated with secondary antibodies diluted in 1% (w/v) BSA and 0.25% (v/v) Triton X for 2 h at room temperature, including AlexaFluor-488 goat anti-rabbit IgG (H + L) (Invitrogen, #A11008, 1:800), AlexaFluor-568 donkey anti-mouse IgG (H + L) (Invitrogen, #A11037, 1:800). Brain sections were washed once with PBS followed by nuclei staining using DAPI (Invitrogen, D1306; 1:5000 in PBST) for 20 min. After washing, sections were mounted with Mowiol 4-88 mounting medium (Merck Millipore, 475904) and coverslipped. For 6E10 staining, 30 min citrate antigen retrieval was performed before permeabilization and blocking. In brief, the slides are immersed in Target Retrieval Solution (DAKO, #S1699, diluted 1:10 with distilled water) and heated in DAKO PT Link (98°C, #PT20040). The temperature is maintained for 30 min before cooling down. Slides were washed 3 times in PBST before permeabilization and blocking.

Confocal image analysis

• Microglia morphology, microglia-Aβ clustering analysis

The 25-30 μm z stack images were taken at dentate gyrus (DG) with the 40× oil 1.25 NA objective on Leica SP8 confocal microscope (Leica Application Suite X 3.5.7.23 225) at 2× zoom with 0.37 μm z step size. All images were taken at 1024 × 1024 resolution, resulting in a pixel size of 142 nm. To quantify microglia morphology, the filament Tracer function from Imaris was used. Microglia were masked with surfaces based on the Iba-1 channel. A starting point was placed at the microglial cell body, and the seed point threshold was adjusted to fill the processes of microglia. The branching structure of microglia was generated by Filament Tracer. The average process length, number of branch points, and number of terminal points for each animal were calculated. All images were quantified with the same parameters. Cell body diameter was measured manually using maximum projection images created in FIJI. Two sagittal brain sections per animal were quantified. The experimenter was blind to diet treatment. All images were quantified with the same parameters.

To quantify the number of microglia clustering around plaques, the spot module from Imaris was used to detect microglia based on the Iba-1 signal, and the surface module was used to construct 3D surfaces of Aβ plaques based on the 6E10 signal. Microglia located within the 20 μm threshold from the plaque surface were counted using spots close to surface Xtension. The algorithm measures the distance from the center of the microglia spot to the nearest point of the plaque surface, allowing for the quantification of microglial clustering around plaques. Plaque volume was also measured using surface function. Finally, the number of microglia around the plaque was normalized to plaque volume. To analyze microglial engulfment of Aβ, microglia were given a surface based on the Iba-1 channel. 6E10 were masked based on the selected microglia surface. The volume of masked 6E10 was quantified (microglia volume is also recorded) using surface function in Imaris. The volume of engulfed plaque was normalized to microglia volume. Two sagittal brain sections per animal were quantified. All images were quantified with the same parameters. The experimenter was blind to diet treatment.

• Aβ plaque and microglia density

To quantify Aβ plaques and microglia density, 40 μm brain sections were stained with 6E10 and Iba-1, images were taken with the 20× oil 0.75 NA objective on Leica SP8 confocal microscope. All images were at 1024 × 1024 resolution, resulting in a pixel size of 568 nm. Images were taken at the DG and cortex. Plaque % area, plaque size, plaque number, and Iba % area were quantified with FIJI. Two sagittal brain sections per animal were analyzed. All images were quantified with the same parameters. The experimenter was blind to diet treatment.

• DCX⁺ neuron quantification

To quantify neurogenesis, 40 μm brain sections were stained with DCX. The 25-30 μm z stack images were taken from DG with the 20× oil 0.75 NA objective on Leica SP8 confocal microscope with 2 μm z step size. All images were at 1024 × 1024 resolution, resulting in a pixel size of 568 nm. DCX⁺ neurons were detected using spots function in Imaris. The number of DCX⁺ neurons were normalized to image z stack steps. Two images were taken from each brain section. Two sagittal brain sections per animal were analyzed. Average number was calculated for each animal. All images were quantified using the same parameters. The experimenter was blind to diet treatment.

Food intake measurement

The food intake was measured for WT and 5xFAD mice at 4 months. Female mice fed zero fiber diet, control diet, and HAMSAB diet were single housed. After 2 d of habituation, daily food intake was measured. In short, food was weighed before giving to the mouse and food was weighed again after 24 h. Food intake was measured for a consecutive 4 d, and average daily food intake was calculated.

Gut histologic analysis

Gut sections were fixed in neutral buffer formalin for 24 h before processing. Fixed gut tissue was processed in Leica PELORIS following a 6 h cycle: formalin 15 min; 80% (v/v) ethanol 15 min; ethanol 15, 15, 15, 30, 45 min; xylene 20, 20, 45 min; paraffin 30, 30, 45 min. The paraffin-embedded gut tissues were sectioned into 4-μm-thick sections and stained with Masson’s trichrome and Alcian blue/PAS at Monash Histology Platform.

For Masson’s trichrome staining, sections were first placed in Bouin’s fixative for 1 h at 60°C. After that, sections were stained in Celestin Blue for 5 min and then in Weigert’s hematoxylin for 30 min. Sections were differentiated briefly in acid alcohol before washing in tap water for 10 min. After washing, sections were stained with Biebrich scarlet acid fuchsin for 5 min and applied 5% tungstophosphoric acid for 5 min. Finally, sections were stained with Aniline blue staining solution for 5 min and washed in 1% aqueous acetic acid for 3 min before dehydrating and mounting. Fibrotic areas were in blue, cytoplasm and muscle were in red, while nuclei were in dark brown to black.

For Alcian blue/PAS staining, sections were first incubated in Alcian blue (1% Alcian blue in 3% acetic acid) for 5 min and then oxidized with 1% periodic acid for 10 min. After that, sections were immersed in Schiff’s reagent for 15 min and washed thoroughly in water for 10 min. Finally, sections were counterstained in hematoxylin for 30 s and blued in Scott’s tap water for 30 s before dehydrating and mounting. Acid mucins are stained in blue and neutral mucins are magenta. Mucin combinations are purple or blue purple depending on the dominant mucin type. Nuclei are dark blue.

Stained sections were imaged with Olympus DotSlide microscope at 20× magnification; all images were taken at 1024 × 1024 resolution, resulting in a pixel size of 322 nm. ImageJ was used to analyze the images. For gut fibrosis, three 322 μm × 322 μm fields were analyzed for each animal, and the color deconvolution plugin was used to extract fibrotic area. The percent fibrotic area was quantified using the same threshold, and the average fibrotic area was calculated for each animal. To measure villi length in the small intestine, 8 villi from each animal were measured. The average villi length was calculated for each animal. Goblet cells were counted manually by an experimenter blinded to treatment group and genotype. In brief, 8 villi were selected randomly for each animal, and the total number of goblet cells in the villi was counted manually. The number of goblet cells was normalized to the width of the villi.

Cecal DNA extraction, library preparation, and sequencing

Cecal content was collected from 6 month WT mice and 5xFAD mice fed with zero fiber, control, and HAMSAB diet. Cecal contents were frozen immediately in liquid nitrogen after collection. The protocol for library preparation and sequencing was previously described (Jama et al., 2020). In brief, cecal DNA was extracted using QIAamp PowerFecal Pro DNA kit (QIAGEN, #51804) following the manufacturer’s instructions. The V4 region of the bacterial 16S rRNA was amplified using previously described PCR methods (Thompson et al., 2017). Briefly, 20 ng of sample DNA was mixed with 515F and 806R primers (Bioneer, 515F forward primer: GTGYCAGCMGCCGCGGTAA, 806R reverse primer: GGACTACNVGGGTWTCTAAT), Platinum Hot Start PCR Master Mix (Fisher Scientific, #13000014) and amplified in a Veriti Thermal Cycler (Fisher Scientific, #4375786) in triplicates. The resulting PCR product was assessed for quality and concentration obtained using the MultiNa MCE-202 Microchip Electrophoresis system (Shimadzu); 240 ng of PCR product per sample was pooled and purified using the PureLink PCR Purification Kit (Fisher Scientific, #K310002). Samples were sequenced in Illumina MiSeq Sequencer (300 bp paired end reads) at the Australian Genome Research Facility.

Microbiome data analysis

Sequenced data in the form of FASTQ files were obtained from the sequencing facility and passed initial quality control using FASTQC version 3 (available from: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). QIIME2 (version 2019-10) workflow was used for downstream analysis (Bolyen et al., 2019). Raw reads were first trimmed (to ensure the quality of at least 50% of reads were >20 Phred score). Reads were then merged, dereplicated, and chimeric reads were removed. DADA2 QIIME2 plugin was used to denoise the reads (Callahan et al., 2016). The Greengenes database (gg-13-8-99-515-806-nb-classifier.qza) was used as a classifier for all taxonomy-related plots and tables (DeSantis et al., 2006). The α-diversity rarefaction curves using the Shannon Index were generated to determine the ideal sequencing depth for further analyses. A minimum sampling depth of 9980 annotated sequence variants was used.

Both qualitative (Jaccard and unweighted UniFrac distance) and quantitative (Bray-Curtis dissimilarity and weighted UniFrac distance) were used to determine β-diversity. The permutational ANOVA (PERMANOVA, 999 permutations for pseudo-F distribution) was performed to test the separation between groups. p < 0.05 was considered significant. p values were Bonferroni-corrected for multiple testing, where false discovery rate q < 0.05 was considered significant.

SCFA analysis

The protocol for SCFA analysis was previously described (Yap et al., 2021). Perfused brain samples were stored at −80 until processing. Acetate and butyrate were measured by gas chromatography after liquid-liquid extraction. SCFAs were directly measured after solvent extraction using specialized polar phase gas chromatography-mass spectrometry Phenomenex Zebron ZBFFAP column (Phenomenex); 100 µg of brain samples was mixed with internal standards (50 µl of 200 μm heptanoic acid), and then transferred into glass test tubes containing 30 µl of 0.2 m NaOH. Samples were mixed vigorously with 3 ml of diethyl ether, and then centrifuged for 5 min, and left at room temperature until the upper ether layer was removed. The residue was redissolved in cold 30 μl of 1 m phosphoric acid in a glass insert GC vial. Samples were analyzed on an Agilent 7890A gas chromatograph (Agilent Technologies). Peaks were detected with a flame ionization detector at 210°C and identified and quantitated against calibration standards. Intra-assay CVs were 14.2% and 10.3% for sodium acetate and sodium butyrate, respectively.

Ex vivo microglia isolation

The protocol for ex vivo microglia isolation was previously described (Grubman et al., 2021). Mice were transcardially perfused with ice-cold PBS, and brains were harvested (cerebellum and olfactory bulb were excluded). Single-cell suspensions were prepared by mechanical dissociation in Dounce buffer (15 mm HEPES, 0.5% (w/v) glucose in HBSS, Invitrogen, 14175095) followed by filtering through 70 μm mesh. Single-cell suspensions were washed with Dounce buffer and centrifuged at 850 × g for 5 min at 4°C. Cell pellets were resuspended in 6 ml 70% (v/v) isotonic Percoll in PBS (isotonic Percoll: 1:9 ratio of 10× PBS to Percoll, Sigma-Aldrich, P4937), overlaid with 6 ml 37% (v/v) isotonic Percoll in PBS. Percoll gradient was centrifuged at 2000 × g for 25 min at 4°C with no brake. After spinning, cells were harvested from 37% 70% Percoll interphase and washed in PBS. The cells were incubated with anti-mouse Fc block (BD Biosciences, 553141; 1:200) for 10 min and stained with CD11b-PECy7 (Biolegend, 101216; 1:200), CD45-BV786 (BD Biosciences, 564225; 1:100), and CX3CR1-FITC (Biolegend, 149020; 1:100) for 30 min on ice. Propidium iodide (PI, Sigma-Aldrich, 1:500) was added to the cells before sorting to distinguish live and dead cells. Fc block, antibodies, and PI used in this protocol were diluted in FACS buffer (0.5% (w/v) BSA, 2.5 mm EDTA in PBS). Microglia were sorted using FACSAria III cell sorter (BD FACSDiva Software version 8.0.1). Microglia were gated as singlets → viable cells (PI^–)→ CD11b⁺, CD45^low→CX3CR1⁺.

Microglial RNA-seq

RNA extraction, library preparation, and RNA-seq were previously described (Grubman et al., 2021). RNA extraction from 1 × 10⁵ to 4 × 10⁵ microglia was performed on QIAcube (QIAGEN) using the RNeasy Micro Kit (QIAGEN, #74004). The quality of RNA was determined using Bioanalyser (Agilent RNA 6000 Pico kit, #5067-1513); 0.5-2 ng of RNA samples with RIN value > 8 was selected for library preparation. An 8 bp sample index and a 10 bp unique molecular identifier were added during initial poly (A) priming. Twenty-four RNA samples were multiplexed (23 microglia samples and 1 iPSC control for normalization for one run, totally two runs with 46 microglia samples and 2 iPCS control). Pooled samples were amplified using a template switching oligonucleotide. The Illumina P5 (5′ AAT GAT ACG GCG ACC ACC GA 3′) and P7 (5′ CAA GCA GAA GAC GGC ATA CGA GAT 3′) sequences were added by PCR and Nextera transposase, respectively. The library was designed so that the forward read (R1) utilizes a custom primer (5′ GCC TGT CCG CGG AAG CAG TGG TAT CAA CGC AGA GTA C 3′) to sequence directly into the index and then the 10 bp unique molecular identifier. The reverse read (R2) uses the standard R2 primer to sequence the cDNA in the sense direction for transcript identification. The library pools were with a mean size of ∼330 bp at a concentration of ∼3 nm. Following denaturation, 1.6 pm of library pools was sequenced on the NextSeq550 (Illumina), using the NSQ V2.5 High output kit (Illumina # 20024906) in accordance with the Illumina Protocol 15046563 version 02, generating 2 reads per cluster composed of a 19 bp forward read and 72 bp reverse read. Quality control, library preparation, and sequencing were performed at the MHTP Medical Genomics Facility by Trevor Wilson.

RNA-seq analysis

Sequencing reads were mapped to the mouse transcriptome reference genome (GRCm38) using STAR (version 020201). We then established read counts for each gene using featureCounts (version 1.5.2). We analyzed the bulk RNA-seq read counts using R (version 3.6.3) and limma (version 3.42.2), and edgeR (version 3.28.1). Data handling and plotting were performed using tidyverse (version 1.3.0). We first removed lowly or nonexpressed genes by requiring a log2 CPM count of >1 in at least 3 samples. Next, we calculated TMM normalization factors to remove composition bias using calcNormFactors. We established the sample S2830 as an outlier and removed it from further analysis. To visualize dimensionality reduction of the sequencing data, we first removed unwanted variation in the data with removeBatchEffect and selected sample sex and age to be regressed out from the CPM values of the read counts. The multidimensional scaling plot used plotMDS and ggplot2 to visualize the remaining variance in the data. To determine differentially expressed genes (DEGs), we first determined the gene-wise variance trends using voom. Then, we built a linear model with the lmFit function and genotype, diet, age, and sex as covariates. We also analyze the data with the diet modeled as a continuous variable (1, zero fiber; 2, control; 3, HAMSAB) alongside sample age and sex. The diet-specific DEGs are visualized in a heatmap (using the pheatmap function, version 1.0.12), after creating normalized read counts with above described modeling and the removeBatchEffect function. We further tested the diet contrast for enrichment of disease-associated microglia (DAM) signature genes using the roast function and the same design matrix as for the continuous diet DEG analysis. Additionally, we visualized those gene sets as volcano plots by using the diet contrast’s log fold change and adjusted p value results from the topTable function and the limma results from above.

Retro-orbital intravenous CD45 labeling

Mice were anesthetized with isoflurane by inhalation. Under deep anesthesia, retro-orbital intravenous injection of 2.5 μg PECy7-labeled anti-CD45 antibody (BD Biosciences, #552848, diluted in 200 μl DPBS) was performed using a 27G insulin syringe (Terumo, #SS05M2713KA). Mice were allowed to regain consciousness before being humanely killed by CO₂ asphyxiation 3 min after injection.

Flow cytometry for brain

Mice were humanely killed by CO₂ asphyxiation. Brains were harvested and collected into cold PBS (cerebellum and olfactory bulb were excluded). Brain tissue was finely minced in 500 μl digestion buffer (1% (v/v) DNase I (Merck, D4527-40KU) and 1% (w/v) collagenase IV (Invitrogen, 17104019) in 1× PBS and digested for 30 min at 37°C. Digested tissue was mechanically disrupted, filtered through 70 μm mesh, and enriched for leukocytes by Percoll (Sigma, GE17-0891-01). In short, the brain pellet was resuspended in 4 ml of 30% (v/v) Percoll (diluted in PBS) and was layered onto 4 ml of 70% (v/v) Percoll (diluted in PBS). Percoll gradient was centrifuged at 900 × g for 20 min at 20°C with no acceleration and brake. After spinning, cells were harvested from the interphase and washed in the FACS buffer. Single-cell suspensions were counted before labeling dead cells with fixable viability dye FVS620 (BD Biosciences, #564996, 1:1000) for 15 min on ice. Nonspecific binding was blocked using mouse FcR blocking reagent (Miltenyi Biotec, #130-092-575, 1:50) for 10 min on ice before proceeding to surface marker staining. Cells were stained with surface marker-specific antibodies for 30 min on ice and fixed and permeabilized with Foxp3/Transcription factor Staining Buffer Set (eBioscience, #00-5523-00) followed by FoxP3 staining. Data were acquired on a BD FACSymphony A3. The flow cytometry staining panel included KLRG1- BUV395 (BD Biosciences, #740279,1:100), CD11c-BUV496 (BD Biosciences, #750483, 1:200), NK1.1-BV421 (BD Biosciences, #562921 1:100), CD69-BUV563 (BD Biosciences, #741234, 1:50), CD4-BUV661 (BD Biosciences, #612974, 1:200), CD8-BUV737 (BD Biosciences, #612759, 1:200), CD11b-BUV805 (BD Biosciences, #741934, 1:400), MHC II-BV510 (BD Biosciences, #742893, 1:100), Ly6C-BV570 (Biolegend, #128030, 1:200), Ly6G-BV605 (BD Biosciences, 563005, 1:100), F4/80-BV711 (BD Biosciences, 565612, 1:100), CD19-BV786 (BD Biosciences, 563333,1:200), FoxP3-AF488 (Biolegend, #320012, 1:100), CD3-PerCPCy5.5 (BD Biosciences, #551163, 1:100), TCRβ-PerCPCy5.5 (BD Biosciences, #560657, 1:100), CD115-APC (Biolegend, #135510, 1:200), MHC II-PE (Biolegend, #107608, 1:1000), and CD45.2-APCCy7 (BD Biosciences, #560694, 1:100).

Statistical analysis

No statistical methods were used to predetermine sample sizes, but the sample sizes were determined based on similar, previously conducted studies (Choi et al., 2018; Lourenco et al., 2019; Martorell et al., 2019). Animals were randomized into different groups with comparable numbers of animals in each group whenever possible. Datasets were assessed for normality and group variance using the Shapiro–Wilk normality test before statistical testing. All analyses were performed with Prism 9 (GraphPad). Values are expressed as mean ± SEM unless otherwise stated. Specific statistical parameters are detailed in the figure legends.

Data availability statement

16S rRNA-seq: https://dataview.ncbi.nlm.nih.gov/object/PRJNA770817?reviewer=m7gtk382nagnes3hcnsnar70fr. Microglial RNA-seq: To review GEO accession GSE186210: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE186210.

Code availability statement

Code is available from the authors by reasonable request.