The role of the gut microbiome in the regulation of astrocytes in Alzheimer's disease

Abstract

Alzheimer's disease (AD) is the most common neurodegenerative disorder and is the most common cause of dementia. AD is characterized pathologically by proteinaceous aggregates composed of amyloid beta (Aβ) and tau as well as progressive neurodegeneration. Concurrently with the buildup of protein aggregates, a strong neuroinflammatory response, in the form of reactive astrocytosis and microgliosis, occurs in the AD brain. It has recently been shown that the gut microbiome (GMB), composed of trillions of bacteria in the human intestine, can regulate both reactive astrocytosis and microgliosis in the context of both amyloidosis and tauopathy. Many studies have implicated microglia in these processes. However, growing evidence suggests that interactions between the GMB and astrocytes have a much larger role than previously thought. In this review, we summarize evidence regarding the gut microbiome in the control of reactive astrocytosis in AD.

Alzheimer's Disease

AD is the most common cause of dementia, which is the loss of cognition sufficient to impair social and occupational functioning [1]. Cognitive impairment caused by AD occurs on a clinical continuum which includes no cognitive impairment, mild cognitive impairment (MCI), and dementia. The typical cognitive domains affected by AD are memory, language, visuospatial function, and executive function. The typical clinical course of AD is usually one that starts with amnestic MCI and eventually progresses to dementia. AD patients also often experience neuropsychiatric symptoms, including depression, anxiety, social withdrawal, delusions, hallucinations, and aggression [1].

The most common form of AD is the late onset form, which occurs in patients over the age of 65. Early onset AD (EOAD), which occurs before the age of 65, can be caused by autosomal dominantly inherited mutations in the APP (encoding amyloid precursor protein), PSEN1 (encoding presenilin 1), and PSEN2 (encoding presenilin 2) genes [2]. These genes encode proteins that are essential for the formation of amyloid beta (Aβ) plaques, a hallmark pathological lesion in the AD brain. Aβ is formed by the cleavage of amyloid precursor protein (APP) by beta secretase and gamma secretase, while cleavage by alpha secretase and gamma secretase precludes Aβ formation [2]. Aβ is a normal physiologic product of the brain, but for unknown reasons accumulates with age to form Aβ plaques in some individuals. While EOAD can be caused by autosomal dominantly inherited mutations, late onset AD (LOAD) is not associated with causal mutations. However, there are several genetic risk factors that have been detected by genome wide association studies (GWAS) that are associated with LOAD [3]. The E4 allele of apolipoprotein E (APOE4) is the most significant LOAD genetic risk factor, while APOE E2 allele is protective against AD. APOE4 is thought to potentially increase seeding of Aβ plaques and also may play a role in astrocyte and microglial activation [4]. Several other genes are genetic risk factors for AD, such as CLU, TREM2, and CD33. Many of AD genetic risk factor genes are expressed in astrocytes and microglia and appear to play a role in the neuroinflammation response to amyloid and tau pathology [3].

Neuroinflammation has emerged as an important aspect in the pathophysiologic progression of AD. As Aβ plaques begin to build up in the brain, microglia and astrocytes initiate and sustain an innate immune response in which they detect, engulf, and wall off Aβ plaques. Furthermore, these cells secrete cytokines and chemokines to recruit additional CNS cells with inflammatory function and peripheral immune cells to Aβ plaques. While this innate immune response is likely protective in the short term, a sustained chronic immune response from these cells likely results in hyperactivated states which lead to inefficient Aβ clearance as well as synaptic, neuronal, and circuit level damage. Additionally, when microglia and astrocytes take on inflammatory roles, they often lose their homeostatic functions, which negatively impacts brain homeostasis. These mechanisms of neuroinflammatory damage are likely essential for cognitive decline observed in AD, prompting vigorous research efforts to more fully understand the mechanisms regulating neuroinflammation and to therapeutically target these mechanisms [5,6].

In addition to Aβ plaque accumulation, neurofibrillary tangles composed primarily of abnormally phosphorylated and post-translationally modified tau protein are observed in the AD brain [7]. Tau tangles are thought to play a significant role in neurodegeneration and correlate better with neurodegeneration and cognitive impairment in AD than Aβ plaques [8,9]. Tau pathology may play a role in neurodegeneration through its activation of microglia, astrocytes and T-cells [10]. Increasing evidence indicates that pathologic tau spreads through the brain via synaptically interconnected regions [11,12]. Aβ may facilitate tau seeding and spreading in the brain through several mechanisms [13], such as Aβ-mediated activation of CDK-5 and GSK3-β kinases that phosphorylate tau [14,15].

Astrocytes in Homeostasis

Astrocytes are the most numerous glial cells in the central nervous system (CNS). These cells were first described by Rudolf Virchow in 1856 and originally called “neuroglia”. Camillo Golgi later described astrocytic endfeet contact with brain vasculature and postulated that “neuroglia” had a primary role of metabolic support and exchange with the vasculature in the CNS. The term “astrocyte” was coined by Michael von Lenhossék in 1891 [16].

Astrocytes are derived from a type of neural progenitor cell (NPC) called radial glial cells (RGCs). Once RGCs differentiate into mature astrocytes, they begin to express particular genes like glial fibrillary acidic protein (GFAP), extend their processes, and establish functional connections with neurons and vasculature. Mature astrocytes are important in blood brain barrier (BBB) regulation and maintenance, ionic buffering, neurotransmitter recycling, glycogen storage, and regulation of synapses [17].

Over 99% of gray matter astrocytes contact blood vessels and are an important component of the neurovascular unit (NVU), which also consists of endothelial cells, smooth muscle cells, pericytes, and neurons [18,19]. The NVU maintains the integrity of the BBB and regulates cerebral blood flow [19]. Perivascular astrocytic endfeet contact capillary endothelial cells and release factors, such as Sonic hedgehog (SHH), fibroblast growth factor (FGF), and glial derived neurotrophic factor (GDNF), to modulate tight junction production in endothelial cells [[20], [21], [22], [23], [24]]. Perivascular astrocytes also express high levels of gap junction proteins called connexins, such as connexin 43 (Cx43) and 30 (Cx30), which allow them to cross-communicate and modulate endothelial cells [25]. In addition to modulating levels of endothelial tight junction protein synthesis, astrocytes also release factors that can increase brain vascular angiogenesis [26,27].

Ions in the CNS are important for the regulation of neuronal action potentials and subsequent neurotransmitter release. Astrocytes play a major role in buffering levels of extracellular ions to maintain brain homeostasis and physiologic neuronal firing. Potassium and sodium ions are important for action potential generation as they increase membrane potential. Excess extracellular concentrations of these ions lead to neuronal hyperexcitability. However, astrocytes prevent excess buildup of these ions through expression of inward rectifying ion channels, such as Kir4.1 or P2X channels [28]. Astrocytes are also important for regulation of chloride ions in the extracellular space [29].

Astrocytes express high levels of the glutamate transporters excitatory amino acid transporter 1 (EAAT1) and excitatory amino acid transporter 2 (EAAT2) and thus play a major role in neurotransmitter regulation and recycling [30]. These cells take up excess synaptic glutamate and convert it to glutamine via glutamine synthetase. Astrocytes can then release glutamine to neurons where it can be converted back to glutamate. The uptake of excess glutamate in the synaptic cleft prevents excitotoxicity. Astrocytes also express GAT transporters for uptake of the inhibitory neurotransmitter gamma-aminobutyric acid (GABA). Astrocytes can also convert GABA to glutamine and release it to neurons to be converted into GABA or glutamate. Interestingly, although astrocytes primarily release glutamine instead of glutamate or GABA, a subset of astrocytes can actively participate in neural circuit function by directly releasing glutamate or GABA through calcium dependent exocytosis in a process known as gliotransmission [31,32].

Glycogen is a glucose polymer that is used for energy in animals. In the CNS, astrocytes represent the major source of glycogen storage [33]. When CNS energy demand exceeds glucose availability, astrocytes can convert glycogen to lactate which can then be shuttled to neurons to use for energy expenditure. Lactate is primarily released by astrocytes via the MCT1 transporter and is captured by neurons also primarily through the MCT transporter. Lactate can be converted to pyruvate where it can then be converted to adenosine triphosphate (ATP) through oxidative phosphorylation.

Astrocyte processes directly contact synapses, and consequently these cells play an important role in the formation, function, and pruning of synapses [34]. Astrocytes secrete substances that promote synaptogenesis, such as hevin, brain-derived neurotrophic factor (BDNF), and TGF-β. Conversely, they can also release proteins that negatively regulate synaptogenesis, such as secreted protein acidic rich in cysteine (SPARC). Several proteins released by astrocytes can influence presynaptic and postsynaptic function. Astrocytic release of cholesterol complexed with apolipoprotein E can increase presynaptic strength [35,36], but astrocyte-derived thrombospondin and SPARC have a negative effect on presynaptic strength [37,38]. Astrocyte derived proteins can also affect α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and n-methyl-d-aspartate (NMDA) receptor production on the postsynaptic density. Glypican 4 and 6, extracellular matrix proteins, activity dependent neurotrophic factor (ADNF), and TNF-α increase synaptic AMPA or NMDA receptors [[39], [40], [41], [42], [43], [44]]. However, thrombospondin and SPARC reduce AMPA receptors on the postsynaptic compartment [45]. In addition to modulating synaptic formation and function, astrocytes play an active role in synapse elimination through the phagocytic receptors MEGF10 and MERTK [46].

Reactive Astrocytosis

Astrocytes serve a supportive function and are important for homeostasis in the CNS. They also serve a function in responding to pathologic insults in the context of disease or injury [47]. In these contexts, astrocytes alter their morphological, transcriptional, and functional phenotypes in a process known as reactive astrocytosis or reactive astrogliosis (Fig. 1). Significant heterogeneity exists in reactive astrocytosis depending on the context and can manifest as astrocytes that take on a pro-inflammatory or anti-inflammatory phenotype that promote or limit CNS inflammation, respectively [47]. The early rise of transcriptomic technology initially led to a discovery of distinct responses of reactive astrocytes, termed A1 (neurotoxic, defined in the context of lipopolysaccharide-mediated neuroinflammation) and A2 (neurotrophic, defined in the context of ischemic stroke models) after previously defined macrophage terminology (M1/M2) [48,49]. However, due to the development of more sophisticated transcriptomic technologies, such as single-cell RNA sequencing, it is now clear that reactive astrocytes possess a significant heterogeneity of response rather than responses that are binary [47,[50], [51], [52]]. Some of the primary signaling pathways initiating and sustaining reactive astrocytosis are Janus kinase/signal transducer and activator of transcription (JAK/STAT) [[53], [54], [55]], Nuclear factor-κB (NF-κB) [56], and calcineurin/nuclear factor of activated T cells (NFAT) [57]. However, there are many proposed transcription factors mediating reactive astrocyte signaling [58].

Fig. 1.

Continuum of astrocyte reactivity with respect to time. Astrocytes are supportive cells in the brain that function in blood-brain barrier maintenance, ionic buffering, neurotransmitter recycling, glycogen storage, and synaptic regulation during homeostasis. In the context of CNS insults or disease, they change their functional state to respond to pathology and become “reactive astrocytes”. Reactive astrocytes lose many of their homeostatic functions and gain pro-inflammatory functions. While reactive astrocytes may be protective in acute contexts, such as in stroke, they are generally detrimental in chronic contexts, such as in neurodegenerative disease, where they may cause neurotoxicity, increase protein aggregation, or recruit pro-inflammatory peripheral immune cells.

Astrocytes in AD

Astrocytes play a major role in the neuroinflammatory response in AD and surround Aβ plaques, as observed in postmortem AD brains and in mouse models of amyloidosis [59]. Astrocytes contribute to amyloidosis through participation in Aβ clearance, production, and through their interaction with other CNS cell types [60] (Fig. 2). Astrocytes can clear Aβ through the release of proteolytic enzymes, such as insulin degrading enzyme (IDE), neprilysin (NEP), matrix metallopeptidase 2 (MMP2), and matrix metallopeptidase 9 (MMP9) that can cleave Aβ [[61], [62], [63], [64]]. Astrocytes are phagocytic and can directly degrade Aβ through several mechanisms [65]. Aβ can also be produced by astrocytes as they express APP, β-site amyloid precursor protein cleaving enzyme-1 (BACE1), and the components of γ-secretase. The expression of these genes has been shown to be increased in reactive astrocytes in AD [[66], [67], [68], [69]].

Fig. 2.

Possible interactions between the gut microbiome and astrocytes in Alzheimer's disease. Genetic and environmental factors, such as diet, sleep, and exercise, can modulate GMB composition over an individual's lifetime. GMB composition likely tunes levels of GMB-derived byproducts and metabolites, such as short chain fatty acids (SCFAs), amino acids, and hormones. The balance of these GMB-derived molecules influences peripheral and central nervous system immunity. Astrocytes are well positioned to uptake GMB-derived molecules or cytokines from the periphery because of their endfeet contact with the brain vasculature. Astrocytes can be directed by GMB-derived molecules or peripheral inflammatory stimuli to modulate their functional state. Astrocytes can contribute to Aβ pathology through their expression of Aβ-producing machinery, their ability to directly phagocytose Aβ plaques, and their ability to regulate microglial phagocytosis of Aβ plaques. Astrocytes may also influence tau seeding and spreading as well as contribute to neurodegeneration through release of neurotoxic substances. Figure adapted from Chandra et al., Molecular Neurodegeneration, 2023.

In addition to astrocytes, microglia are also phagocytic and are involved in limiting Aβ plaque accumulation [70]. Astrocytes can modulate microglial phagocytosis by the release of inflammatory mediators. For example, astrocytes release complement C3 that binds to C3aR on microglia, which compromises Aβ plaque phagocytosis [71]. Conversely, astrocytes release IL-3 and IL-33 that bind to their respective receptors on microglia and increases Aβ plaque phagocytosis [72,73]. Similarly, astrocytic release of programmed death-ligand 1 (PD-L1) binds to the programmed cell death protein 1 (PD-1) receptor on microglia to increase Aβ plaque clearance [74]. Astrocytes also release substances, such as interferon-induced transmembrane protein 3 (IFITM3) and cholesterol, which can increase neuronal production of Aβ plaques [75,76]. In the context of AD, reactive astrocytes take on a pro-inflammatory transcriptional program and lose many of their homeostatic functions [77]. The loss of physiologic astrocytic functions is detrimental in the AD brain as it may lead to dysregulation of the BBB, ion buffering, neurotransmitter buffering, and energy regulation [47,49,77]. Reactive astrocytes can also release substances that can kill neurons or compromise neuronal function. Reactive astrocytes release saturated lipids via APOJ that are neurotoxic [78]. Additionally, reactive astrocytes release C3 that can bind to C3aR on neurons, which leads to aberrant neuronal network function [56]. Together with microglia, astrocytes can prune synapses via the MEGF10 and MERTK receptors and have been found to be important for compensation whenever microglial phagocytosis of synapses is compromised [46,79]. However, excessive synaptic pruning by microglia and astrocytes during pathological states is likely detrimental [80].

Astrocytes are the primary producers of APOE in the brain and the APOE4 allele is the greatest risk factor for AD. APOE4 produced by astrocytes has been shown to potentiate Aβ pathology and neurodegeneration [81]. APOE4 derived from astrocytes has also been shown to promote BBB impairment as well as induce cerebrovascular dysfunction [[82], [83], [84]]. Astrocyte-derived APOE4 is also implicated in tau-mediated neurodegeneration [85].

Tau pathology drives a strong immune response in astrocytes as evidenced by transcriptomic studies in tau mouse models and tauopathy patient brains [86,87]. Astrocytes may play a role in the propagation of tau via the phagocytosis of dying neurons with tau inclusions and subsequent release of pathogenic tau seeds [88]. Aβ is thought to be important for downstream seeding and spreading of tau pathology and astrocytes potentially participate in this process [89]. Astrocytes also express Mapt, the gene encoding tau, and the production of astrocytic tau may play a role in AD [90].

GMB and AD

Human evidence of GMB alterations in AD

Gut microbial changes in human and animal models are extensively reviewed in Chandra et al. [91], but are briefly reviewed below. Two studies in 2017 showed that individuals positive for Aβ/AD patients have an altered GMB composition compared to individuals without Aβ/AD. Cattaneo and colleagues measured plasma levels of cytokines and stool abundance of particular GMB taxa (Escherichia/Shigella, Pseudomonas aeruginosa, Eubacterium rectale, Eubacterium hallii, Faecalcterium prausnitzii, Bacteroides fragilis) using quantitative real time PCR [92]. They found an increase in mRNA expression of the pro-inflammatory cytokines IL-6, CXCL2, NLRP3, and IL-1β and a decrease in the expression of anti-inflammatory cytokine IL-10 in Aβ-positive cognitively impaired patients compared to Aβ-negative cognitively normal individuals. A positive correlation was observed between the pro-inflammatory cytokines tested and Escherichia/Shigella and a negative correlation with Eubacterium rectale. Furthermore, Vogt and colleagues, took an unbiased approach and performed 16s ribosomal RNA amplicon sequencing on DNA isolated from fecal matter of AD patients and healthy control subjects [93] (HC). This group found a decrease in overall GMB microbial diversity in AD patients, as well as a decrease in Firmicutes and Bifidobacterium and increased levels of Bacteroidetes compared with HC. Since these initial studies, several other groups have found alterations in GMB composition between AD patients and HC [94,95]. Zhuang et al. found a decrease in Bacteroidetes and an increase in Actinobacteria in AD patients compared with HCs [94]. Liu et al. found a reduction in Firmicutes and an increase in Proteobacteria in AD compared to control [95]. Recently, Ferreiro et al., recruited 115 healthy and 49 preclinical AD patients (Aβ+ but clinical dementia rating of 0) and performed metagenomic sequencing on DNA extracted from fecal matter. The authors observed that preclinical AD patients have a distinct GMB profile compared to healthy controls. They also found that GMB composition correlated with Aβ and tau PET but not with markers of neurodegeneration, such as hippocampal volume or total tau in CSF. This result suggests the GMB is associated more with earlier AD pathologic events, such as Aβ and tau pathology, rather than neurodegeneration which occurs later in the AD disease course. Lastly, they found that GMB composition information improved machine learning prediction of preclinical AD status [96].

The results from these collective studies suggest that GMB composition is altered in AD patients, which may play a role in the pathogenesis and progression of AD. However, these studies are correlative and results from AD microbiome-targeted therapeutic clinical trials will be needed to assess whether the alterations in the GMB directly influence human AD pathogenesis. Additionally, there seems to be little consensus between the particular bacterial changes in AD patients in these studies. Moreover, the sample sizes in the published studies are quite small (generally less than 50–100 participants per group) and there are several confounding factors that might influence patient GMB composition including geographical location, diet, and environmental exposures. To resolve these confounding issues, it will be critical to carry out large, international studies that assesses GMB composition between age and sex-matched AD patients and HCs. In order to resolve which bacteria may actually modify AD-relevant pathologies, bacteria found to correlate with pathologies in human studies will need to be tested in animal models. Furthermore, to determine whether probiotic strains of bacteria could be useful as therapeutics for AD, they will need to be tested in large clinical trials.

Mouse studies implicating GMB involvement in AD models

Consistent with alterations of GMB composition observed in AD patients, differences in the GMB have also been observed in several Aβ amyloid mouse models, including 5XFAD [97,98], APPSwe/PSEN1dE9 [[99], [100], [101], [102]] (APP/PS1), and APPSwe/PSEN1L166P [103] (APPPS1-21) models compared with nontransgenic mice. Brandscheid and colleagues reported an increase in Firmicutes and a decrease in Bacteroidetes phyla at 9 weeks of age in 5XFAD mice compared to nontransgenic mice [97]. Conversely, Chen and colleagues reported a decrease in Firmicutes and an increase in Bacteroidetes phyla at 3 months of age in 5XFAD mice compared with nontransgenic mice. At 6 months of age, there was a substantial increase in Bacteroidetes, Proteobacteria, and Deferribacteres and a decrease in alpha diversity in 5XFAD mice [98]. Shen and colleagues profiled the GMB composition in 3, 6, and 8-month-old APP/PS1 mice. Changes over the course of aging in the GMB roughly correlated with Aβ pathology. Particularly, an increase in Odoribacter and Helicobacter genera and decreases in Prevotella species correlated positively and negatively with Aβ, respectively. Furthermore, a decrease in GMB diversity was also observed [99]. Chen and colleagues assessed GMB composition in 1, 2, 3, 6, and 9-month-old APP/PS1 mice and observed GMB changes as early as 1 month that increased over time, including Escherichia-Shigella, Desulfovibrio, Akkermansia, and Blautia [100]. Zhang and colleagues found Verrucomicrobia and Proteobacteria increased in 8–12-month-old APP/PS1 mice. Conversely, Ruminococcus and Butyricicoccus were significantly decreased in 8–12-month-old APP/PS1 mice compared with nontransgenic mice. Interestingly, several short chain fatty acids (SCFAs) were decreased in the feces and brain of APP/PS1 mice compared with nontrangenic mice [101]. Cuervo-Zanatta and colleagues reported sex-specific GMB changes in APP/PS1 mice compared with nontransgenic mice. Male APP/PS1 mice had a larger alteration of their GMB profiles compared with females [102]. Harach and colleagues reported significant increases in Bacteroidetes and Tenericutes and a decrease in Firmicutes, Verrucomicrobia, and Proteobacteria at 8 months in APPPS1-21 mice compared with nontrangenic mice [103]. Overall, these studies show that there are likely age and sex-dependent changes in GMB composition and diversity between mouse models of Aβ amyloidosis and their nontransgenic counterparts. While of interest to study how bacterial composition changes overtime in AD mouse models, it should be noted that the mice in these studies are housed in different facilities, given different diets, and possess different genetic backgrounds and will most certainly have differing GBMs in respective mouse colonies. The changes in bacteria also are correlative with the progression of pathology and do not imply causation.

In addition to determining how GMB composition changes in AD mouse models over time, several studies have tried to determine the role of the GMB in AD through antibiotic (abx) and germ-free (GF) based manipulations. Abx has been shown in several AD mouse models to alter GMB composition, including the APP/PS1 [104,105], APPPS1-21 [[106], [107], [108], [109]], 5XFAD [[110], [111], [112]], and APP NL−G−F [113] models. For example, Minter and colleagues [104] treated mice with a high dose of broad spectrum abx cocktail (kanamycin, gentamicin, colistin, metronidazole, vancomycin) between postnatal day (PND) 14-PND 21 and found a change in GMB composition with a large increase in Akkermansia and Lachnospiracea at 6 months compared with vehicle-treated controls. Abx-mediated alteration in GMB composition requires an abx cocktail, as individual abx are not effective at altering either GMB composition or amyloidosis [107]. Moreover, the abx used in the cocktail do not cross the blood brain barrier at appreciable levels, thus implying that the effects on Aβ amyloidosis are likely mediated by the GMB rather than by direct effects in the brain [107]. Additionally, three studies have raised APPPS1-21 [103,114] and 5XFAD mice [110] in GF animal facilities, where the mice are devoid of GMB, and compared them to conventionally raised animals that have an unperturbed GMB. Manipulation of the GMB with abx consistently results in a reduction of Aβ plaque deposition [[104], [105], [106], [107], [108], [109], [110]]. Notably, this reduction in Aβ deposition is sex-specific and only occurs in males in the APPPS1-21 and APP/PS1 mouse models [105,106]. There are several potential reasons for these sex-specific outcomes that need further investigation. For example, differences in GMB-hormone interactions [115,116] as well as differences in immune response between sexes [117] might be responsible. A recent study by Saha and colleagues suggests that estrogen may mediate these sex differences. Abx causes rises in circulating estrogen levels in female APPPS1-21 mice. When these mice are ovariectomized (OVX), reducing their estrogen levels, amyloid is decreased. When OVX female mice are supplemented with estrogen, amyloid levels are increased [118]. Abx may have effects that are GMB independent. It has been reported that fecal matter transplants (FMT) from untreated amyloid model mice into abx-treated amyloid model mice restores their GMB and Aβ pathology, indicating that abx-mediated reduction in amyloid pathology likely occurs through a GMB-dependent mechanism [106,108,109]. The abx-mediated reduction in amyloid is consistent with the phenotype observed in GF mice [103,110,114]. GF mice also have a strong reduction in amyloid beta pathology, but surprisingly, this effect is not sex-specific [103,110,114]. The mechanism by which abx treatment decreases Aβ pathology in males but not females as opposed to equal reduction of amyloid in GF mice is not known. One important aspect of abx-treated animals versus GF animals is that in contrast to GF animals that lack GBM, abx does not necessarily lead to GBM depletion but rather to a change in bacterial composition and diversity. In addition to abx and GF models, another method to perturb the GMB is to colonize GF rodents with human fecal matter through FMT. A recent study found FMT of human AD fecal matter into GF rats induced cognitive deficits and decreased hippocampal neurogenesis, suggesting there may be a causal role of the GMB in inducing AD symptomology [119].

In addition to changes in amyloidosis, abx or GF-mediated GMB depletion or absence, respectively, results in changes in microglial inflammatory state in amyloidosis models [[103], [104], [105], [106], [107],110,114]. Several studies have shown that at the time of sacrifice, microglia appear to become less pro-inflammatory and more phagocytic as determined by RNAseq and microglial morphological analysis in the context of abx-mediated GMB alteration [[105], [106], [107]]. For example, APP/PS1 and APPPS1-21 mice treated with abx have altered microglial morphology in which microglia have increased process length and number, consistent with a more homeostatic state [[105], [106], [107]]. APPPS1-21 mice also have a reduction in expression of genes associated with microglial cell activation as determined by bulk RNAseq [106,108]. Microglial changes are restored when abx-treated mice are given FMT from non-treated mice [106,108], suggesting that the GMB mediates abx-induced changes in microglia. Taken together, these results suggest that microglia in the context of abx may lose deleterious pro-inflammatory function, become more efficient at phagocytosis, and may be part of the mechanism whereby abx leads to a reduction in amyloid. To this point, colony stimulating factor 1 receptor (CSF1R) inhibitor-mediated microglial depletion prevents abx-mediated reduction in amyloid pathology [108]. This result implies that microglia must be required for the abx-mediated decrease in amyloid pathology. Microglia are also strongly affected in GF AD mice compared to AD mice that are housed in conventional environments. Harach and colleagues [103] showed that GF APPPS1-21 mice had reduced ionized calcium binding adapter 1 (IBA1) ​+ ​microglia in the brain at 3.5 and 8 months compared to conventionally housed mice [103]. In contrast, Mezo and colleagues [110] observed an increase in IBA1+ microglia in GF 5XFAD mice in the hippocampus at 4 months of age compared to conventionally housed controls. Microglial bulk RNAseq revealed an activated microglial transcriptomic signature in GF 5XFAD mice characterized by upregulation of Apoe, Trem2, Axl, Cst7, Cd9, Itgax, and Clec7a and downregulation of P2ry12. Lastly, GF 5XFAD mice increased microglial phagocytosis of Aβ compared to controls [110]. The cause of the differing results between Harach and Mezo requires further investigation. Colombo and colleagues used GF APPPS1-21 mice to study the role of GMB-derived SCFAs in the regulation of microglial transcriptomic state and their response to Aβ pathology [114]. GF APPPS1-21 have a decrease in SCFAs, plaque associated microglia, and Aβ plaques. Interestingly, administration of SCFAs via drinking water to GF APPPS1-21 mice resulted in an increase in plaque associated microglia and Aβ plaques. Nanostring transcriptomic analysis of SCFA treated APPPS1-21 mice revealed an activated microglial state, characterized by increased expression of several genes in the APOE-TREM2 pathway. These results suggest that gut-derived SCFAs can mediate microglial transcriptomic state and potentially modulate levels of Aβ plaques in the brain [114]. While GF mice are useful models for studying GMB contribution to disease, they possess several developmental defects that may affect disease phenotype in ways that may not be translationally relevant [[120], [121], [122]]. Peripheral [123] and central immune development [121], neurotransmission [124], and neurogenesis [125] may be altered by GF conditions and could confound studies using these models. It is useful to combine GF-based GMB manipulation with other less severe GMB manipulations to validate the importance of findings of disease models.

In the context of tauopathy, Seo and colleagues found several changes in microgliosis upon GMB perturbations. The authors found a reduction in IBA1+ and CD68⁺ microglia in APOE4 PS19 mice housed in a GF environment compared those raised in a conventional environment at 40 weeks of age. Additionally, APOE3 PS19 mice treated with abx from P16-22 and aged to 40 weeks exhibited microglia with longer and more abundant processes, reflecting a more homeostatic phenotype. However, there was no change in microglial morphology in APOE4, APOE KO, or Tau KO groups. Interestingly, microglial process length positively correlated with hippocampal size, suggesting that homeostatic microglial phenotype is associated with hippocampal preservation. Additionally, the authors performed single nucleus RNA sequencing (snRNAseq) and found 3 subclusters of microglia in their dataset. They found a reduction in microglial cluster 1 and an increase in cluster 0 in APOE3 PS19 mice treated with abx compared to controls. Cluster 1 was associated with microglial cell activation, while cluster 0 was associated with homeostatic functions, such as lipid metabolism, chemotaxis, and development. APOE4 PS19 animals had an increase in Cluster 1 upon abx treatment compared to controls and Tau KO animals had a similar phenotype to the APOE3 mice.

GMB Regulation of Astrocytes

Because astrocytes contact almost all blood vessels in the brain, it is logical that they are well positioned to uptake GMB-derived byproducts from the circulation (Fig. 2). The primary evidence for a link between gut microbiome and astrocytes comes from mouse models of multiple sclerosis (MS). It has been shown in MS mouse models that gut-derived metabolites, such as derivatives of tryptophan can directly influence astrocyte activation. Rothhammer and colleagues sorted astrocytes from mice subjected to experimental autoimmune encephalitis (EAE) and performed RNA sequencing and found type I interferon (IFN-I) signaling was significantly increased. Deletion of Ifnar1, the receptor for IFN-ɑ, in astrocytes was associated with decreased CNS inflammation and protection from EAE. IFN-I signaling drove increased aryl hydrocarbon receptor (Ahr) expression, which is a receptor for gut-derived tryptophan metabolites. The authors then found that Ahr activation reduced NFκB signaling through activation of suppressor of cytokine signaling 2 (Socs2). Interestingly, administration of a tryptophan depleted diet worsened EAE clinical scores which could be rescued by tryptophan supplementation in wildtype but not astrocytic Ahr knockout mice. Administration of gut-derived tryptophan derivatives, such as indoxyl-3-sulfate, indole-3-propionic-acid, and indole-3-aldehyde, phenocopied astrocytic Ahr-dependent protection from EAE. This study suggests that gut-derived metabolites can directly regulate astrocyte responses during CNS inflammation [126]. Additionally in MS, gut-derived metabolites can modulate transcriptional activity in other CNS cells, such as microglia [127] or natural killer (NK) cells [128], which can also regulate astrocyte reactivity. Another study by Rothhammer et al. (2018) showed that gut-derived tryptophan metabolites can modulate TGFɑ and VEGF-B signaling in microglia via Ahr [127]. These signaling pathways then regulate astrocyte inflammatory gene expression. Sanmarco et al. (2021) found that the GMB regulates the expression of IFN-γ in meningeal NK cells. NK-derived IFN-γ regulates the induction of LAMP1 ​+ ​TRAIL ​+ ​astrocytes, which induce pro-inflammatory T-cell apoptosis [128]. Furthermore, SCFAs can modulate astrocytic gene expression in a sex-dependent manner in vitro [129]. This collection of studies indicates that the GMB is capable of influencing astrocyte phenotype.

GMB and Astrocytes in AD

The GMB has been shown to regulate astrocyte phenotypes in the context of both amyloidosis and tauopathy [91,109,130]. Our group recently studied the role of the GMB on reactive astrocytosis in the APPPS1-21 mouse model of amyloidosis [109]. We found postnatal treatment of APPPS1-21 mice with a broad-spectrum antibiotic cocktail from P14-21 led to a sex-specific reduction in GFAP ​+ ​reactive astrocytes and GFAP protein levels in the male brain cortex. Additionally, we also observed an abx-mediated reduction in plaque-associated GFAP ​+ ​reactive astrocytes. Furthermore, we found that abx-induced alterations in astrocyte morphology characterized by an increase in branch number and length corresponded with ezrin protein levels. Ezrin was recently shown to be a regulator of astrocytic morphology [52]. In addition to the morphologic changes, we found abx treatment reduced astrocytic C3 expression, which is a marker of neurotoxic reactive astrocytes and has been shown to impair microglial and neuronal function [56,71]. The abx-mediated changes in astrocyte phenotypes were associated with abx-induced GMB bacterial compositional changes. Notably, several gram-negative bacteria were decreased, such as Odoribacter, Anaeroplasma, and Paraprevotella, which have pro-inflammatory lipopolysaccharide (LPS) in their outer cell walls. There was also a striking increase in Akkermansia, which has previously been reported after abx treatment [105]. Akkermansia has been shown to lower Aβ42 levels, Aβ plaques, and ameliorate cognitive dysfunction in APP/PS1 mice [131]. In order to determine whether the abx treatment effects on astrocyte phenotypes were mediated through GMB changes, we gave male abx-treated APPPS1-21 water vehicle (VHL) or fecal matter transplants (FMT) from untreated age-matched APPPS1-21 donor mice to restore the GMB. Those mice that received abx ​+ ​FMT had increased GFAP ​+ ​reactive astrocytes, GFAP ​+ ​plaque-associated astrocytes (PAAs), less astrocytic branching, and increased astrocytic C3 levels compared to abx ​+ ​VHL treated mice. These results suggest that abx-mediated effects on astrocyte phenotypes are dependent on the GMB. We found similar astrocyte phenotypes in APPPS1-21 male mice raised in a GF environment compared to those raised in a specific pathogen free environment (SPF) as we did in abx-treated mice compared to controls. Lastly, we depleted microglia from APPPS1-21 mice using CSF1R inhibition and treated mice with abx to determine the role of microglia in abx-induced astrocyte changes. We found that there were no astrocyte morphologic changes after abx treatment when microglia were depleted, implying that morphologic changes were microglial dependent. However, abx-mediated reductions in GFAP ​+ ​astrocytes, PAAs, and C3 expression in astrocytes were still decreased by abx treatment when microglia were depleted. This suggests abx-mediated reductions in GFAP ​+ ​astrocytes, PAAs, and C3 levels are independent of microglia. While microglia are not required for these abx-induced effects on astrocytes, depletion of microglia prevents abx-mediated reduction in amyloid pathology at certain timepoints. We speculate that GMB control of astrocytosis may regulate microglial phagocytosis of amyloid pathology (Fig. 2). We found that abx decreases astrocyte C3 expression and increases IL-33 production. C3 has been shown to impair microglial phagocytosis, while IL-33 increases it, suggesting these are possible mechanisms whereby astrocytes may increase microglial phagocytosis of amyloid [71,73]. The identity of possible gut-derived intermediates that regulate astrocytosis in amyloidosis are currently under investigation.

In the context of tauopathy, Seo and colleagues, found that GMB perturbations alter astrocyte morphological and transcriptional phenotypes coincident with neurodegeneration and phosphorylated tau in the PS19 model in an APOE isoform and sex-specific manner [130]. They found a reduction in GFAP ​+ ​astrocytes in APOE4 PS19 mice housed in GF conditions compared to those raised in a conventional environment at 40 weeks of age. Additionally, mice that were initially raised in a GF environment and then moved to conventional housing at 12 weeks had increased GFAP ​+ ​astrocytes at 40 weeks of age compared to those fully raised under GF conditions. Furthermore, PS19 mice treated with abx from P16-22 had an increase in astrocytic process length and number of processes in APOE3 mice but not in APOE4 or APOE KO mice at 40 weeks of age. There was no change in male mouse astrocyte morphology at 12 weeks of age following early life abx. Interestingly, female PS19 mice treated with abx had a decrease in astrocytic process number and length at 40 weeks of age. Seo and colleagues also performed single nucleus RNA sequencing (snRNAseq) to determine the transcriptional profile of brain nuclei in PS19 mice after P16-22 abx treatment at 40 weeks of age. Four distinct transcriptional clusters of astrocytes were found. Notably, Cluster 1 was drastically reduced in APOE3 mice that were treated with abx, while Cluster 0 was increased. Cluster 1 was enriched in reactive astrocyte pathways, such as gliogenesis and regulation of neuron death, while Cluster 0 was enriched in neurotrophic homeostatic pathways such as glutamate receptor signaling and regulation of synapse assembly. A similar response was observed in tau APOE KO mice, but there was little change in APOE4 tau mice that were treated with abx compared to controls. Together, these results suggest that perturbation of the GMB in the context of tauopathy has an influence on astrocyte phenotype. Because astrocytes are associated with neurodegeneration, GMB control of astrocytosis may regulate neurodegeneration associated with AD. However, further investigation using models to modulate reactive astrocytosis following GMB perturbations will have to be performed to fully elucidate the relationship between the GMB and astrocytes in the context of tauopathy.

Astrocytes are important cells that function primarily as a source of support for the brain during homeostasis [16]. These cells function in blood brain barrier (BBB) regulation and maintenance, ionic buffering, neurotransmitter recycling, glycogen storage, and regulation of synapses [16]. However, during disease they shift their functional state to respond to pathology and become pro-inflammatory and in a chronic context become deleterious [47]. The GMB has emerged as master regulator of immunity [132]. It has become increasingly clear that the GMB regulates brain innate immunity in addition to peripheral immunity through direct and indirect mechanisms [133]. Most studies investigating the role of the GMB in controlling brain innate immunity have focused on microglia, but astrocytes also have neuroinflammatory functions and are uniquely positioned to take up GMB-derived byproducts, metabolites, and cytokines from the periphery because of their endfeet that contact brain vasculature [18,121]. Studies from the Multiple sclerosis (MS) field have illuminated the influence of the GMB on astrocytes in physiology and in the context of CNS disease [[126], [127], [128]]. In addition to MS, GMB-mediated control of astrocyte function likely plays an important role in Alzheimer's disease. While GMB control of astrocyte phenotypes has been understudied, especially in AD, recent studies clearly show that abx and GF housing-based GMB modifications have large impacts on astrocytes in both amyloid and tau mouse models of AD [109,130]. There are several open questions which need to be answered to better understand the role of the gut microbiome in regulating astrocytes in AD. These questions include (1) which specific bacteria and bacteria derived metabolites may play a role in GMB-astrocyte communication? (2) Is GMB-astrocyte communication through a direct or indirect mechanism (is the peripheral immune system potentially involved in mediating this connection)? (3) What are the specific molecular pathways in astrocytes which are tuned by the GMB? (4) Does the GMB play a role in regulating astrocyte communication with other CNS cells? (5) How do molecular pathways in astrocytes which are regulated by the GMB relate to AD pathology? (6) How can the GMB-astrocyte axis be therapeutically targeted to benefit AD patients? Future studies will need to elucidate the exact mechanisms responsible for GMB-astrocyte communication and how these mechanisms may be perturbed in AD.

Author Contributions

S Chandra and R Vassar wrote and edited the manuscript.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Robert Vassar reports financial support was provided by Northwestern University Feinberg School of Medicine. Robert Vassar reports a relationship with Cure Alzheimer's Fund that includes:. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.