Despite numerous studies investigating how the gut microbiota is altered in AD, both in human and murine models, few studies have extensively sampled longitudinally to identify the dynamic gut microbiota signatures in 3xTg-AD mice. Previous studies have shown that 3xTg-AD mice have a distinct bacterial signature compared to age-matched controls.41–43 However, there are only two studies to our knowledge that investigate gut microbiota in 3xTg-AD mice at more than one time point; the authors evaluated the gut microbiome at two42,44 and four42,44 time points. In one study, the gut microbiota of 3xTg-AD and WT mice were assessed at 8, 12, 18, and 24 weeks.44 They too demonstrate compositional differences that were highlighted at the 8 week timepoint, but the specific taxa that were depleted in the 3xTg-AD mice differed from our study. In the second study, the gut microbiota of 3xTg-AD and WT mice were assessed at 16 weeks and 24 weeks. They similarly demonstrate alterations in the gut microbiome prior to development of pathologies, but did not report taxonomic changes to the genus level.42 Here, we assessed the temporal dynamics by dense longitudinal sampling of microbial communities in the gut of 3xTg-AD mice over the course of a year to better understand compositional changes that correlate with disease pathologies. Our study characterizes gut microbiota composition at 25 time points (n = 1,717 total samples), with multiple time points corresponding to pre-pathology development, plaque deposition, and one time point corresponding to plaque deposition and hyperphosphorylated tau. Several bacteria, including Bacteroides acidifaciens, Prevotella sp., Akkermansia muciniphila, Turicibacter sp., and Lactobacillus salivarius, differed in relative abundance between 3xTg-AD and WT mice temporally. Turicibacter sp. and Akkermansia muciniphila were enriched in the gut microbiota in 3xTg-AD mice at early time points, preceding pathology development, while Bacteroides acidifaciens and Prevotella sp. were enriched in the gut microbiota of 3xTg-AD mice at later time points. Critically, these features in the gut microbiota were used to successfully predict strain of mice early on in life, showing potential for unique signatures in the gut microbiota composition to be used as a predictor of AD prior to pathology development.
Previous studies support that perturbations to the gut microbiota composition alter host immune responses, thereby shifting towards a proinflammatory environment in the colon and hippocampus.45 To quantify changes in the inflammatory profile of 3xTg-AD mice, we assessed the expression of relevant neuroinflammatory and inflammatory genes at each body site. Significant increases in TNF-α, IL-6, IL-1β, IFN-γ gene expression via RT-qPCR on brain tissue have been observed in 3xTg-AD mice at 16 months of age.46 In our study, we found significant upregulation of IL-6 gene expression in the colon of 52 week 3xTg-AD mice when compared to 52 week WT mice, but no changes in TNF-α, IL-1β, IFN-γ were observed at 52 weeks of age. We also observed significant upregulation of glial fibrillary acidic protein (GFAP), a marker of astrogliosis, in the hippocampus and colon of 3xTg-AD mice at 52 weeks when compared to 52 week WT mice. Enteric glial cells (EGCs) are resident to the enteric nervous system, which aids in regulation of the gastrointestinal tract via modulation of immune and endocrine function.47 EGCs resemble astrocytes in the brain in their morphology, ability to secrete cytokines, and their expression of glial fibrillary acidic protein. Increased gene expression of GFAP in the colon of rats 4 hours after intravenous LPS injection suggests that GFAP upregulation is a result of acute exposure to a systemic inflammatory environment.47 Interestingly, GFAP has also been identified as a blood biomarker in AD patients and correlates with cognitive impairment.48 These findings support our hypothesis that the GI tract and the brain are communicating via transport of chemical mediators in the bloodstream. Finally, MRC1 (also known as CD206) was elevated in the hippocampus at 24 weeks of age in 3xTg-AD mice, indicating microgliosis. We hypothesize that the upregulation of MRC1 at 24 weeks of age is associated with increased phagocytosis in response to the deposition of amyloid-β, which is documented at 6 months of age.49
In this study, we demonstrate distinct microbial compositions in 3xTg-AD mice prior to the development of AD pathologies. As mice aged, the gut microbiota of 3xTg-AD and WT mice became more similar. Unweighted metrics (Jaccard and Unweighted Unifrac) demonstrated significant differences at 8 and 24 weeks, but not at 52 weeks of age. We did observe significant differences using weighted beta diversity metrics (Bray Curtis and Weighted UniFrac) which account for abundance of observed features at 8 weeks, but not at 24 and 52 weeks. This indicates that lower abundance bacterial microbiota features are strong drivers of changes in gut microbiota composition. Similar findings of compositional differences early in life were reported in female 3xTg-AD mice at 3 and 5 months of age when compared to B6129SF1/J mice.42 Early-life gut microbiota composition perturbations in mice have been associated with aging-associated health and disease, including neurodegenerative diseases like AD.50 Our findings indicate compositional differences in microbial communities, driven by rare taxa early in life, are present prior to amyloidosis and tauopathy development.
Alpha diversity is frequently used as a marker of disease status, and is decreased in several diseases associated with the gut-microbiota brain axis, including depression,51 Autism Spectrum Disorder,52 Parkinson’s disease,53 and in some studies, AD.16,54 In humans, alpha diversity was reported to be decreased in elders with AD compared to age-matched healthy participants.16 When we analyzed alpha diversity metrics by subsampling our data to include one mouse at each timepoint, we did not find significant differences. These findings align well with other studies that have been performed in mice. In one, no differences in alpha diversity were reported when comparing 3 and 5 month 3xTg-AD female mice42 and in another, no differences in alpha diversity were reported in 8, 12, 18, and 24 week old 3xTg-AD male mice compared to age-matched WT mice.44 However, when we leveraged dense longitudinal sampling using LME, we demonstrate that genotype has an effect on Faith’s PD, where WT mice have a higher alpha diversity than 3xTg-AD mice. These findings suggest lower alpha diversity in 3xTg-AD mice may be a predictor of disease status when assessed during onset and progression of AD pathologies.
To identify key features of the gut microbiota composition that differentiate 3xTg-AD mice from WT mice, we used a Random Forest machine learning classifier on a feature table of the fecal microbiota. Our analysis demonstrated successful discrimination between 3xTg-AD and WT mice using gut microbiota compositions from 4 to 52 weeks of age, but prediction accuracy was improved when we included only samples from pre-pathology timepoints. We selected samples at 2 months of age (6 and 8 weeks) and 6 months of age (22 and 24 weeks) to increase sample size due to loss of samples during the sample classifier training. Several of the features that were most important for predicting strain were also significant in our other analyses, including Lactobacillus sp., Lactobacillus salivarius, and Bacteroides sp. The predictive power of these models indicates unique bacterial communities early in life and throughout life in 3xTg-AD mice modeling AD disease pathologies. Interestingly, Haran and colleagues were able to discriminate between elders with AD and elders with different types of dementia using a random forest model using strain-level features of the gut microbiome generated using shallow shotgun metagenomic sequencing.55 Both B. fragilis and B. vulgatus were important features in classifying participants in their study. Bacteroides were also enriched in 3xTg-AD mice in our study. These findings suggest certain microbes identified in the cohort with AD in this study, including Bacteroides sp., may play a mechanistic role in the key pathologies of AD. We are performing additional studies to evaluate the role of Bacteroides in AD progression.
We observed concordance in feature importance across our Random Forest classifier, longitudinal volatility analysis, and differential abundance testing (ANCOM). Analysis of feature volatility revealed taxa at the bacterial genus- and species- level resolution that are predictive of age within each strain. Akkermansia muciniphila, a mucin-degrading bacteria associated with intestinal inflammation in mice, is present in 3xTg-AD early in life, but not in WT mice.56 Turicibacter sp., shown to be involved in intestinal serotonin production, is increased in 3xTg-AD mice early in life compared to WT mice.57 Prevotella sp., associated with reduction in short chain fatty acid production and intestinal inflammation in mice,58 was increased later in life of 3xTg-AD mice. Lactobacillus salivarius, a bacteria shown to positively influence immune cell development, was present in greater relative abundance in WT mice for the first 32 weeks of life.59 Taken together, these results indicate potential contributions from 3xTg-AD mice gut microbial communities in inflammatory processes and neurological health.
All three statistical approaches used in our study (ANCOM, Random Forest machine learning, and volatility analysis) demonstrated increased relative abundance in Bacteroides in 3xTg-AD mice. Notably, Random Forest identified the low abundance taxon, Bacteroides acidifaciens, as highly important in predicting mouse strain. Other species of Bacteroides have been implicated in health status and are likely key contributors to host-microbial interactions via the gut microbiome-brain axis. Bacteroides fragilis and Bacteroides stercosis function ecologically as keystone species, indicated by low relative abundance and disproportionately numerous interactions on microbial community dynamics.60 B. fragilis can influence the gut microbiome-brain axis and reduce autism-like behaviors by modulating serum metabolites and GI inflammation.61 Bacteroides were also increased in abundance in mice expressing a variant of human APP (APPswe [Tg2576]) compared to control mice, and administration of B. fragilis promoted amyloid deposition in the APP/PS1 mice.62 Another study, this time using 5xFAD mice, which model amyloidosis at an earlier time point than 3xTg-AD mice, demonstrated increased relative abundance of Prevotella sp., Bacteroides acidifaciens, and Turicibacter sp. in 5xFAD mice at 10 weeks of age. The 10 week time point in 5xFAD mice and the 24 week time point in 3xTg-AD mice (where we observed the first increase in B. acidifaciens) both represent development of amyloidosis in the respective models. This may indicate that changes in relative abundance of certain microbes is critical during the onset of amyloid-β exposure. These findings suggest the potential for amyloidosis to alter microbial communities in the gut of mice modeling AD amyloid-β plaques.
Bacteroides have also been observed as differentially abundant in human studies of AD, though the association with health or disease are conflicting. In one study of participants with AD and age-matched human controls, Vogt et al. demonstrated increased relative abundance of Bacteroides in patients with AD. Interestingly, this increase positively correlated with greater amyloid burden in the brain and CSF phospho-tau, indicating a greater disease burden.16 In another study, Haran and colleagues also observed increased Bacteroides in patients with dementia compared to age-matched controls.55 However, Zhuang et al. found that Bacteroides relative abundance decreased in patients with AD.63 Taken together, these findings in humans support our findings in a mouse model, and suggests a role for gut-associated Bacteroides in progression of AD pathologies.
Mechanistically, species in the genus Bacteroides might influence neuroinflammatory processes in the brain. Bacteroides fragilis produces an endotoxin, lipopolysaccharide, that is unique to the species of Gram-negative bacteria (BF-LPS). BF-LPS may cross the gut epithelium and enter the bloodstream, inducing systemic inflammation and upregulation of pro-inflammatory cytokines via the NF-κB pathway.64 BF-LPS is recognized by TLR-2, TLR-4, and CD41 microglial cells, potentially inciting microgliosis in the brain. We are currently investigating the role of B. acidifaciens in the ecology of the gut microbiota and hypothesize that it may also function as a keystone species and influence neurological health status through the gut microbiome-brain axis.
The complexity of the host-microbe interactions in 3xTg-AD mice was demonstrated in this study by the dynamic microbial communities and immune profiles. Our study characterized the gut microbiota temporally in 3xTg-AD mice modeling amyloid-β plaques and hyperphosphorylated tau to identify key changes in composition correlated with disease pathogenesis. The present study shows upregulation of biomarkers for microgliosis, astrogliosis, and intestinal inflammation. Analysis of the gut microbiome demonstrated an altered gut microbiota composition associated with 3xTg-AD early in life, including prior to pathology development, that is predictive of disease state. This is the first study of its kind to characterize the gut microbiota at 25 time points, ranging from pre-pathology to modeling of both amyloidosis and tauopathy. Additionally, it will provide a reference for future studies to determine frequency of fecal sampling in longitudinal gut microbiota analysis based on the well characterized evolving gut microbiota composition in the present study. It is critical for future studies on the role of the gut microbiota-brain axis and AD to investigate multiple time points throughout disease progression due to changes in the gut microbiome and inflammatory profile as exemplified in the current study. Furthermore, focus on the functional microbiome through a multi-omics approach is essential in better understanding host - microbe interactions via the gut microbiota-brain axis in AD.