Young versus aged microbiota transplants to germ-free mice: increased short-chain fatty acids and improved cognitive performance

Figure 2 shows principal component analyses (PCAs) of the Aitchison distance, a measure of beta diversity, in mice with young or aged microbiomes at 30, 60, and 90 d post FTG. Note the clear separation between the young and aged microbiomes at each of the three time points (PERMANOVA, P = .001 for each time). Also, note the separation in the points on the plot for the aged microbiome at 30 d; 6 samples cluster to the left of the centroid and 5 samples cluster to the right of the centroid (Figure 2, arrow in top panel). The separation in the aged group at 30 d was traced back to a single cage (points to the right of the centroid). In the group housed in a single cage, there was an expansion of Parabacteroides and possibly Akkermansia (see bracket in Supplemental Figure II).

Alpha diversities for fecal samples were calculated using Shannon, Simpson, and Chao1 indices in mice with aged and young microbiomes (Supplemental Figure I). All three indices were significantly decreased in mice with an aged microbiome at 30, 60, and 90 d after the initial FTG. These indices suggest that there is both a decrease in diversity and possibly a decrease in evenness with an aged microbiome.

Figure 3 shows differences in the abundance of bacterial taxa in feces resulting from a young FTG at 30, 60, and 90 d compared to that in mice with an aged FTG. An unclassified genus or genera in the Muribaculaceae family significantly decrease at all three time points and an unclassified genus or genera in the Peptococcaceae family significantly increased at all three time points in mice with a young microbiome compared to that with an aged microbiome. In some instances, the young microbiome was either consistently increased or consistently decreased across the time points, although statistical significance was not achieved for all of the times. Lachnospiraceae_UCG-006 significantly decreased at 60 and 90 d. Mice with young microbiomes showed taxa increases in SCFA or putative SCFA producers including genera in the family, Lachnospiraceae, an unclassified genus or genera in the Peptococcaceae family, Ruminococcaceae_UBA1819, and Oscillospiraceae_Colidextribacter.^22–24 Although some bacteria capable of producing SCFAs increased in mice with an aged microbiome, the overall effect suggests more abundant SCFA-producing taxa in mice with a young microbiome. A stacked bar chart of taxa abundance for individual mice is shown in Supplemental Figure II.

Functional abundance of important bacterial enzymes used in the metabolism of acetate, propionate, and butyrate was determined from 16S rRNA gene analysis of feces harvested at 30, 60, and 90 d using the software program, PICRUSt2 Figure 4. This heat map does not necessarily predict if SCFAs should be decreased in mice with an aged microbiome since many of these enzymatic reactions can operate in either the forward or reverse direction. However, the heat map does indicate that enzymes involved in SCFAs metabolism in gut are different in the 2 groups of mice and predicts potential differences of the SCFA production with young and aged microbiomes. Using PICRUSt2 we were able to determine which of the bacteria taxa that we found significantly different in Figure 3 contributed to the significant changes in SCFA-related enzymes shown in Figure 4.²⁴Table 1 shows the enzymes that significantly changed (column 1), the sampling day when a difference in the enzyme expression occurred (columns 2–4), the taxa potentially responsible for the change (column 5), and the effect of the taxa on increasing or decreasing the enzyme in mice with a young microbiome (column 5). Note that Ruminococcaceae_UBA1819, genera of the Lachnospiraceae family, an unclassified genus or genera of the Peptococcacaeae family, the genus, Colidextribacter, and an unidentified members of the class, Bacilli, increased at relevant times and could be responsible for the changes in enzymes involved with SCFA metabolism. The analysis of bacterial taxa in young and aged mice would predict that the microbiota in young mice secrete more SCFAs than that of mice with an aged microbiome.

Figure 5 shows relative SCFA concentration in feces at 0, 30 and 60 d after the initial gavage. Note that concentrations of SCFAs were substantially lower in the GF mice at baseline “0 days” as would be expected in animals lacking gut bacteria. Compared to mice with a young microbiome, mice with an aged microbiome showed significant decreases in fecal acetate (P = .005), propionate (P < .001), and butyrate concentrations (P = .009; N = 11 per group; two-way repeated measures ANOVA). [Note: Supplemental Table I provides a more complete analysis for all two-way ANOVA in this study]. Holm-Sidak post hoc analysis was used to compare individual groups where “*” represents a significant difference in mice with an aged microbiome compared to a young microbiome at the same time point. With the exception of acetate and possibly butyrate, cecal concentrations of SCFAs at 90 d were similarly decreased by ~50% (Supplemental Figure III). Note that the SCFAs in the 30 d aged group showed a dichotomy as described under beta diversity above. The five samples that are to the right of the centroid in Figure 2 show significant decreases for all SCFAs compared to the other six samples, which are left of the centroid. Dichotomy in the SCFAs did not occur in any other group.

Figure 6 summarizes the results of the novel object recognition test which is a measure of short-term memory. With short-term recall mice will spend more time at a novel object than a familiar object. Results of the novel object recognition test demonstrate that mice with a young microbiome spend more time on the novel object than mice with aged microbiomes (P < .001; N = 10–11 per group; two-way repeated measures ANOVA, see Supplemental Table I).

Figure 7a shows the results of the tail suspension test, a measure of depressive-like behavior. Mice that spend less time attempting to escape (more immobility time) are considered to have a more depressive-like phenotype. Mice with an aged microbiome had more immobility time (P = .008; N = 9–11; two-way repeated measures ANOVA, see Supplemental Table I) compared to those with a young microbiome. Post hoc Holm-Sidak analysis also demonstrated significant effects of the aged microbiome (“*” and “**” represent P < .05 and 0.001, respectively) compared to the group with the young microbiome for the same time period. Note that the time of immobility in the 30 d aged group showed a dichotomy as described under beta diversity above. The five samples to the right of the centroid in Figure 2have statistically significant increases in time of immobility compared to the other six samples, which are left of the centroid (149 ± 10 versus 83 ± 21, P = .027). None of the other behavioral tests regardless of time after the initial gavage showed a dichotomy based on mice housing. Figure 7b shows the results of the Barnes Maze test, a test for spatial memory. Overall mice with an aged microbiome showed a longer latency to reach the entry zone for an escape hole (P = .030; N = 11; two-way repeated measures ANOVA).

Finally, there were no differences in distance traveled in an open field test or the time mice could hang on a wire without falling (Supplemental Figure IV, A and B), indicating that neither activity nor strength should be factors affecting performance in other tests in the aged mice. In the open field test, there were no differences in either the total moving time or the velocity of movement between mice with young and aged microbiomes at any of times following the days after the initial FTG.

Flow cytometry was conducted using brains of mice with a young or aged microbiome. Supplemental figure V shows the gating strategy and corresponding results for flow cytometry studies from brains 90 d after the initial FTG. There were no statistical differences in the number of CD45⁺ cells (leukocyte), CD4⁺ helper T cells, CD8⁺ cytotoxic T cells, and γδ T cells (CD45⁺CD4⁻CD8⁻TCR γδ⁺) in brains from mice with a young and aged microbiome (N = 5 per group). Supplemental Figure VI shows plasma inflammatory cytokines in the plasma of mice 90 d after the initial FTG (N = 6–10 for each group). There were no significant differences in IL-5, IL-6, IL-9, IL-10, IL-12 (p40), eotaxin, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), IFN-γ, keratinocyte chemoattractant (KC, also known as CXCL1) or, monocyte chemoattractant protein-1 (MCP-1), macrophage inflammatory protein-1α (MIP-1α), RANTES, or TNF-α. We note that inflammatory markers in brain and plasma (Supplemental Figures V and VI) were only measured at the termination of the study, 90 d after the initial FTG. Thus, we do not know if the inflammatory status was different between groups at earlier time points.

This study was performed under the guidelines of the National Institute of Health and all experiments were approved by the Institutional Animal Care and Use Committees at the University of Texas Health Science Center at Houston (UTHSCH) and the Baylor College of Medicine. Given the differences in the microbiome and inflammatory responses between male and females, sex must be treated in separate groups. We included only males in the present study since inclusion of both sexes would be prohibitive for a single study. C57BL/6-GF male mice (12 weeks) were obtained from the Baylor College of Medicine Gnotobiotic Rodent Facility and shipped in autoclaved shipping crates to UTHSCH. The crates were assembled, bedded with Alpha-Dri bedding and autoclaved with the lid closed. The mice were removed from the isolator in the gnotobiotic facility using sterile transfer bags, housed in the shipping crates under a biosafety cabinet, and immediately shipped. The two medical schools are only a few hundred yard apart precluding lengthy and stressful travel. Upon arrival at UTHSCH, fecal samples were collected to confirm the absence of bacteria in the gut. To assess the “germ-free” status of the mice upon arrival, an internal standard was serially diluted and the copy number of the 16S rRNA gene in feces from each transfer crate was analyzed using qPCR. 16S rRNA gene was not detected in feces from any of the transfer crates confirming that the mice were germ-free. Once the aseptic crates were opened at the UTHSCH, the mice were transferred to autoclaved cages, provided with autoclave chow and water, and maintained in a conventional animal holding room.

Fresh fecal samples were collected at 9–10 am from wild-type young (2–3 months) or aged (18–20 months) donor male mice (N = 5–10 per group). The samples were pooled and immediately homogenized using ice-cold PBS (120 mg feces/1 ml) followed by centrifugation at 800 g for 3 min at 4°C. The supernatant was used for transplantation. One hundred microliters of the fecal supernatant was gavaged on arrival and at day 7, 14, 30, and 60 Figure 1. Mice were housed with other mice receiving the same FTG and placed in cages with filtered tops. Cages with aged FTG mice were separated from the young FTG mice but were housed in the same conventional animal holding room at the UTHSCH. Precautions were taken to avoid cross contamination between groups during behavioral testing and when changing cages, food, water, and bedding.

The time line for the study is shown in Figure 1 with each collection time noted. Bacterial 16S rRNA gene and SCFAs were analyzed from feces at 30, 60, and 90 d and from cecal content at 90 d after the initial FTG. Tissues were harvested at 90 d after the initial gavage.

Fecal samples were collected from the recipient mice and immediately frozen at −80°C. 16S rRNA sequencing was performed on collected fecal samples. DNA was extracted using MO BIO PowerMag Soil DNA Isolation Kit (MO BIO Laboratories), according to the manufacturer’s protocol. 16S rRNA gene sequence libraries were generated using the V4 region primers 515 F and 806 R on the Illumina MiSeq platform.

Taxonomic analysis: The raw data files in binary base call (BCL) format created by the MiSeq run were converted into FASTQ format using the Illumina ‘bcl2fastq’ software. Raw pair-ended 16s rRNA sequences (FASTQ) were demultiplexed by QIIME2 framework core function.³⁵ Reads were de-noised and merged into amplicon sequence variants (ASVs) by DADA2 pipeline in R.^36,37 Taxonomic annotations were also generated against DADA2-formatted training FASTA files derived from SILVA138 Database.³⁸ ASVs with identical taxonomic assignment were grouped into taxonomic bins. Alpha diversity (Shannon Index, Simpson Index and Chao1 Index) was calculated using R package Vegan.³⁹ Student’s t-test was used to compare alpha diversity at each time point. For principal component analysis (PCA), permutational multivariate analysis of variance (PERMANOVA) with the adonis function in Vegan package was used on Aitchison distance matrices generated by ALDEx2 package.^40–42 ALDEx2 package was also used to calculate differential abundance of centered log-ratio transformed taxa counts, represented as effect size. Kruskal–Wallis test with post hoc Benjamini-Hochberg correction was used for statistical analysis.

Functional abundance of SCFA-related enzymes: The table of ASV counts and sequences of ASVs were input into PICRUSt2 for metagenomic prediction.^24,43-46 Relative abundance of predicted sample gene family profiles was calculated using humann2_renorm_table function of HUMAnN2 pipeline.^47–49 Statistical analysis was performed using Python package scipy and statsmodels. P values were adjusted using the Benjamini-Hochberg correction. A false discovery rate (FDR) of <0.05 was considered statistically significant.

Mice were acclimated in a conventional behavioral testing room for a minimum of 1 hour prior to initiation of any tests. Investigators were blinded to the treatment (aged or young microbiome) of each mouse and tests were conducted at the same time each day. Mice were allowed to recover 30 minutes between tests. The order of the tests for a given day was dependent on the perceived level of stress ranging from low to high stress. For instance, the tail suspension test was conducted last. To protect against cross contamination of bacteria between mice and to remove olfactory cues, all equipment was thoroughly cleaned with 70% ethanol between animals.

Fecal and cecal samples were submitted to the Metabolomics Core at Baylor College of Medicine for processing and analysis. SCFAs were measured by LC-MS (Agilent LC-QQQ-MS system) and compared against a known reference. SCFAs were normalized to the measure obtained in the group with the young microbiome. When an SCFA was measured over several time points, the values were normalized to that obtained for the “30 day” time point for young microbiome.

Ninety days after the initial FTG, mice were anesthetized with Avertin (250 mg/kg, i.p.) and blood was collected by cardiac puncture using a heparinized syringe. Each mouse was perfused transcardially using ice cold PBS solution and the brain was removed. The right hemisphere was homogenized and incubated in RPMI-1640 (Corning) containing 1 mg/ml of collagenase/dispase (Roche) and 10 mg/ml of DNase I (Roche) for 45 min at 37°C. The cell suspension was filtrated through a 70-μm cell strainer, washed and applied to a Percoll (GE Healthcare) gradients of 30% and 70% using the centrifugation at 500 g for 20 min. The interphase between the gradients was collected. Cells were stained with anti-CD45 (clone: 30-F11, BioLegend 103139), anti-CD4 (clone: RM4-5, BioLegend 100516), anti-CD8α (clone: 53–6.7, BioLegend 100737) and anti-TCRγδ (clone: GL3, BioLegend 118124). An amine reactive Live/Dead Aqua viability stain (Invitrogen L34966↗) was used to identify only live cells. Fluorescence-minus-one controls were used to distinguish positively stained cells for each antibody. The cells were analyzed using CytoFLEX (BECKMAN COULTER). The data were analyzed using FlowJo software (Tree Star, Inc.).

Cytokines were measured from plasma obtained 90 d after the initial FTG. Plasma samples were collected from both young and aged FTG groups at post-FTG day 90 (N = 10 for each group) using Bio-Plex Pro Mouse Cytokine Standard 23-Plex (Bio-Rad) according to the manufacturers protocol.

Two-way repeated measures ANOVA with post hoc Holm-Sidak test where appropriate (Figure 5, 6, 7 and Supplemental Figure IV) and the t-test (Supplemental Figure III, V, and VI) were used. P < .05 was considered statistically significant.Funding This work was supported by grants AG058436 from the National Institute of Aging (Bryan, McCullough, Co-PIs), NS103592 from the National Institute of Neurological Diseases and Stroke (McCullough, Bryan, Co-PIs), and grant HL134838 from the National Institute of Heart, Lung, and Blood (DJD), DK56338 from National Institute of Diabetes and Digestive and Kidney Diseases which supports the Texas Medical Center Digestive Diseases Center and its support to VRV and DJD; 15SDG23250025 from the American Heart Association (VRV); 19POST34410076 from the American Heart Association and the American Brain Foundation (JL). The metabolomics core was supported by CPRIT Proteomics and Metabolomics Core Facility (NP) (RP170005), NIH (P30-CA125123), and Dan L. Duncan Cancer Center.