Alzheimer’s disease (AD) is the most common neurodegenerative disease worldwide. Amyloid-β (Aβ) accumulation and neurofibrillary tangles are two key histological features resulting in progressive and irreversible neuronal loss and cognitive decline. The macrophages of the central nervous system (CNS) belong to the innate immune system and comprise parenchymal microglia and CNS-associated macrophages (CAMs) at the CNS interfaces (leptomeninges, perivascular space and choroid plexus). Microglia and CAMs have received attention as they may play a key role in disease onset and progression e. g., by clearing amyloid beta (Aβ) through phagocytosis. Genome-wide association studies (GWAS) have revealed that human microglia and CAMs express numerous risk genes for AD, further highlighting their potentially critical role in AD pathogenesis. Microglia and CAMs are tightly controlled by environmental factors, such as the host microbiota. Notably, it was further reported that the composition of the gut microbiota differed between AD patients and healthy individuals. Hence, emerging studies have analyzed the impact of gut bacteria in different preclinical mouse models for AD as well as in clinical studies, potentially enabling promising new therapeutic options.

On November 3, 1906, Alois Alzheimer presented a clinical course of a disease, later named after him. More than 100 years later, Alzheimer’s disease (AD) is one of the greatest health care challenges of the twenty-first century worldwide, with more than 55 million patients to date and with forecasts of 150 million patients by 2050 [115]. The clinical symptoms of AD patients include irreversible progressive loss of cognitive function, including memory, spatial orientation, vision and language, ultimately leading to impaired ability to perform activities of daily life [106]. Macroscopically, AD brains exhibit signs of atrophy and enlarged ventricles due to neuronal loss. Histologically, AD is characterized by extracellular amyloid-beta (Aβ) deposition, abnormally phosphorylated tau protein aggregation, synaptic degeneration, and the activation of microglia and astrocytes [102]. Although the pathogenesis is well described, the potential cause(s) of AD remain elusive, hence, preventative or disease-modifying treatments are still limited.

For the past decade, it has been proposed that microorganisms residing in the gut (collectively called microbiota) may be associated with AD. Notably, it has been reported that patients with AD exhibit changes in their microbiota composition and diversity. There is increasing evidence that crosstalk between the gut microbiota and the brain, especially through bacteria-derived mediators such as short-chain fatty acids (SCFAs), plays a critical role in host health, homeostasis and disease [30]. Given the importance of this topic, we review recent findings on how the gut microbiota shapes neurodegenerative diseases by modulating CNS macrophages in mouse models of AD and the potential translation of these findings to therapeutic options for humans.

Hallmarks of AD: Amyloid-β and phosphorylated tau pathology

The detection of Aβ pathology is a hallmark for AD diagnosis, whereby this 40–42 amino acid peptide is derived from proteolytic cleavage of the amyloid precursor protein (APP). APP is a family of conserved type I membrane proteins that are highly expressed in neuronal and glial cells, with orthologs identified across different species [118]. APP can be processed via α- (ADAM10), β- (BACE1) and γ- (PSEN1/2) secretases into its cleavage products [192]. During pathological amyloidogenic processing by β- and γ-secretases, APP is sequentially cleaved, resulting in C-terminal fragments (CTFs), which are further processed into Aβ40 and Aβ42 fragments [177]. CTF accumulation during APP processing is neurotoxic and can be detected in human cerebrospinal fluid (CSF) [92]. A critical role for Aβ is also evident by the fact that Aβ deposits are now targeted by monoclonal antibodies [81]. Monoclonal antibodies such as aducanumab, lecanemab, and donanemab were generated to target Aβ. Lecanemab, also a humanized IgG1 monoclonal antibody, can bind with high affinity to small soluble Aβ protofibrils, therefore delaying cognitive impairment in the early stage of the disease. This antibody has passed a phase III trial in the USA, although it is still in the approval process in Europe. Currently, one limitation is that the treatment starts at a relatively late stage of the disease and these new therapeutic options are often given after a substantial number of neurons are already lost [172]. Therefore, molecular biomarkers are critical for an early and reliable diagnosis of AD. CSF and blood biomarkers show high potential to optimize diagnostic strategies, with several candidates currently being investigated such as Aβ42/40 as well as p-tau 231 and p-tau217 [6].

In addition to Aβ plaques, tau-mediated neurofibrillary tangles (NFTs) constitute the second main pathological hallmark of AD. The occurrence of NFTs progressively during the course of AD [100], whereby the preclinical phase is characterized predominantly by early Aβ deposition with an onset of clinical symptoms at least 10–20 years later [99]. Compared with Aβ, NFTs are more closely associated with synaptic loss, neurodegeneration, and cognitive decline [121]. Tau is expressed predominantly in neurons and to a lower degree in oligodendrocytes and astrocytes [76]. The main function of tau is to bind to microtubules to stabilize and support their assembly. Furthermore, tau participates in axonal transport [157], axonal elongation, neurogenesis [61], and synaptic plasticity. In AD, tau loses its physiological function with its binding equilibrium for microtubules [7], resulting in an increased cytosolic concentration of unbound tau, causing misfolding, aggregation and the formation of NFTs. NFTs impair regular axonal transport and lead to synaptic dysfunction and neurodegeneration [135].

One percent of all AD cases are familial autosomal dominant, with onset as early as 30 years of age. This familial AD is caused by mutations in APP genes or those associated with APP processing, such as PSEN1 and PSEN2 [95]. However, the etiopathology of the much more frequent sporadic AD remains unclear (reviewed in [139]), probably because the disease pathogenesis is heterogeneous and caused by a complex interaction of genetic and environmental risk factors. Genome-wide association studies (GWAS), revealed specific mutations within risk genes for AD, including the apolipoprotein E (APOE4) gene on chromosome 19 [88,151,184].APOE has critical functions in homeostasis, with APOE4 carriers showing accelerated breakdown of the blood–brain barrier (BBB) [9,113]. Additionally, several risk genes, such asTREM2, ABCA7,CD33 andMS4A6A, are expressed by CNS macrophages [88,151,183]. In the human brain,TREM2, which encodes an immunoreceptor tyrosine-based activation motif-containing cell surface receptor, is among the most highly expressed receptors on microglia [28]. Heterogeneous mutations withinTREM2 increase the risk of late-onset AD by 2–fourfold [52,154], suggesting a potentially critical role for microglia during AD already from disease onset on [102].

The role of microglia in AD

Microglia are the resident macrophages of the CNS with a plethora of immune and homeostatic functions, including innate immune response, neuroprotection, synaptic pruning and phagocytosis of cellular debris [124]. In 1919, Río-Hortega published a series of papers, where he first described `microglia` as an independent cell type and defined their distribution and morphological phenotype. Interestingly, he also noted the putative mesodermal origin of these cells and recognized their phagocytic capacity [33,152]. In 1999, it was proposed that microglia originate from the yolk sac [122], which was later confirmed by two fate-mapping studies [48,144]. Moreover, microglia are derived from CD45⁻ c-kit⁺ erythromyeloid progenitors in the yolk sac, and their developmental process is characterized by the maturation and differentiation of microglia progenitors via CX3CR1⁻ and CX3CR1⁺ stages (Kierdorf et al. 2013). Microglia are highly plastic and regulate the stability of their microenvironment in the healthy CNS independently of hematopoietic stem cell (HSC)-derived cells [57]. They are long-living cells with low turnover rates across different brain regions, however they can locally proliferate during perturbation [164].

With the help ofCx3cr1^GFP/WT reporter mice [72], it was demonstrated that microglia are constantly active, probing an area tenfold larger than their cell body [32,116]. To recognize microbes, metabolites, chemokines and cytokines, microglia express a variety of pattern recognition receptors (PRRs), such as toll-like receptors (TLRs), NLRP3 and scavenger receptors, including the Aβ-sensing CD36 [102]. Upon Aβ recognition, microglia accumulate around Aβ plaques and clear Aβ deposits, e.g., through phagocytosis. Concomitantly, these plaque-associated microglia are in an activated state and initiate proinflammatory signaling cascades as well as inflammasome formation [179]. During disease progression, microglial morphology changes as they become less ramified, with larger cell bodies and shorter processes and a decrease in their process dynamics [85]. During disease progression, microglia continuously produce neurotoxic cytokines such as interleukin (IL)-1β, IL-8, tumor necrosis factor alpha (TNF-α) and reactive oxygen species (ROS) [54]. However, they lose their ability to efficiently clear Aβ over time [84]. This chronic immune activation can lead to an exhausted microglial state and downregulation of Aβ-degrading enzymes such as insulin-degrading enzyme (IDE) and neprilysin (NEP) (Wyss-Coray et al. 2002). Furthermore, homeostatic genes, such asP2ry12 orTmem119, are downregulated upon microglial activation, whereasApoe, Clec7a andTrem2 are upregulated in plaque-associated microglia [78]. Specifically,TREM2 upregulation has been found both in patients with AD and additionally in transgenic mouse models [71] [78] (Box 1). A potentially TREM2-dependent neuroprotective role of plaque-associated microglia in the early stages of plaque formation was reported in mouse models as well as human AD patients and may attenuate tau seeding in neuritic plaques [75,96,189].

In addition to beneficial roles of Aβ recognition and clearing, microglia may also enable the spread of Aβ [34], and in addition the tau protein between neurons via phagocytosis and exocytosis [5,163], although the exact mechanisms are not yet fully understood. A study with human and murine samples suggested that microglia can take up tau but do not completely degrade it, consequently triggering tau aggregation in recipient cells [62]. Furthermore, a study using a mouse model for tauopathy described microglial NF-κB activation and the resulting release of cytokines, such as IL-1, IL-6 and TNF-α, as a possible cause for microglial-mediated tau spreading [175].

CNS-associated macrophages

CNS-associated macrophages (CAMs), also known as border-associated macrophages (BAMs), have become increasingly important in recent years. CAMs are present at the interfaces between the periphery and brain parenchyma, including the leptomeninges (harboring leptomeningeal macrophages, mΜΦ), perivascular space (containing perivascular MΦ, pvΜΦ) and choroid plexus (comprising stromal choroid plexus MΦ, cpΜΦ and Kolmer epiplexus cells), controlling these potential gateways into the CNS [129]. CAMs and microglia share a common progenitor during embryonic development, and the final cell fate occurs subsequently and locally in the developing CNS [49]. Interestingly, leptomeninges function as an intermediate environmental niche during early postnatal development, from where mΜΦ migrate into the perivascular space to give rise to pvMΦs [104]. In contrast, cpΜΦs are an ontogenically and transcriptionally mixed population partially replenished with HSC-derived cells in adulthood [104]. In contrast to microglia, the development of CAMs has recently been shown to be independent of transforming growth factor β (TGF-β) [19,171]. In the context of AD, pvΜΦs seem to play a role in clearing Aβ across the BBB [58,161]. Notably, several risk genes for AD are also strongly expressed in human pvΜΦ [183]. In a slow-progressingApp^NL−F AD mouse model, secreted phosphoprotein 1 (SPP1/osteopontin) was upregulated predominantly in pvΜΦs and proposed to be required for microglia to engulf synapses and enhance phagocytosis in the presence of Aβ oligomers [143].

Astrocytes and AD

In addition to microglia and CAMs, astrocytes feature certain immune regulatory functions. These distinct glial cells with a neuroectodermal origin, firstly termed astrocytes 1895 by Michael von Lenhossék, are the most abundant cells within the CNS [46]. Astrocytes maintain plenty of essential homeostatic functions during development and adulthood, especially the control of the BBB (reviewed in great detail in [91]). Aβ and tau accumulation lead to astrocytic activation [147] and the upregulation of glial fibrillary acidic protein (GFAP) [40]. Elevated GFAP levels can be measured in the blood and CSF of AD patients, whereby plasma GFAP is suggested to increase early in AD progression [10]. Activated astrocytes undergo functional and molecular changes [43] and contribute to neurodegeneration via the release of neurotoxic cytokines, nitric oxide (NO) and ROS [15]. Interestingly, astrocytes also express major AD risk genes, includingAPOE,CLU andFERMT2 [4], further supporting a pivotal role for glial cells in AD.

The gut microbiota steers CNS immunity

Increasing evidence suggests that environmental factors, including the intestinal microbiota, critically shape the host’s physiology and are essential for immune function, including the CNS [42]. The word microbiota is composed of ancient greek μικρός(mikrós) 'small' and βίος(bíos) 'life' and includes bacteria, archaea, viruses, fungi and protozoa that reside in and on humans, animals and plants. Approximately 2500 years ago, the ancient Greek physician Hippocrates proclaimed `all disease begins in the gut`. Two millennia later, the development of new techniques boosted microbiological research, underscoring the importance of the gut microbiota in health and disease. Today it is known, that the gut-brain axis enables bidirectional communication between the CNS and the gut. This communication occurs through direct and indirect neuronal connections and endocrine, metabolic and immune pathways via vitamins, neurotransmitters, and microbial metabolites such as SCFAs [21]. Increasing evidence suggests that SCFAs, including acetate, propionate and butyrate, produced predominantly by gut bacteria through the fermentation of otherwise indigestible dietary fibers are important gut-derived mediators able to substantially shape immune function in general [162]. Accordingly, SCFAs have also been associated with a variety of (neurodegenerative) diseases [30]. Specifically, acetate was identified as the essential SCFA driving microglia maturation and regulating their homeostatic metabolic state. Acetate normalized the immature phenotype of microglia that was present in germ free (GF) mice as well as an impairment of complex II in the mitochondrial respiratory chain [41].

Already during embryonic and early postnatal development, the maternal microbiota significantly influences microglial properties in a time- and sex-dependent manner (Fig. 1) [168]. Later in life, the human and mouse gut microbiota display an age-specific composition [132]. Furthermore, intestinal permeability is partially increased, resulting in higher blood concentrations of gut-derived metabolites. It was shown that gut-derivedN⁶-carboxymethyllysine (CML) augmented oxidative stress and mitochondrial dysfunction in microglia during aging [114]. Moreover, gut-derived tryptophan metabolites can directly act on microglia via aryl hydrocarbon receptors (AhR) [133]. Microglial AHR activation further regulates astrocytes via transforming growth factor α (TGF-α) and vascular endothelial growth factor B (VEGF-B). In summary, gut-derived molecules, including tryptophan, CML, and SCFAs, can affect CNS immunity during steady state and disease.

Gut-derived molecules affect the CNS immune system during development and aging. Before birth the microbiota of the mother affects the microglia gene expression of the embryo, e.g. by increasing genes associated with LPS stimulation. After birth the pub is colonized and bacterial amount and diversity increases. During adulthood a stabile microbiota composition is reached, which modulates microglia and astrocytes via microbial metabolites, including the SCFA acetate and tryptophan-derived metabolites. In aged miceN⁶-carboxymethyllysine (CML) accumulates in the brain, which results in oxidative stress and mitochondrial dysfunction in microglia

Full size image

Induced microbiota modulation by antibiotics results in diminished Aβ accumulation

Initially, Minter and colleagues uncovered that alterations in the gut microbiota affect Aβ pathology in the CNS [111]. They used the APP/PS1 mouse model (Box 1) and treated them with a cocktail of several antibiotics (ABX) by oral gavage, starting at postnatal day 14 (P14) until P21, followed by ABX supplementation in the drinking water until the age of 6 months (Table 1) [111]. Compared with vehicle-treated APP/PS1-21 mice, long-term ABX-treated male APP/PS1 mice showed a significantly decreased Aβ burden with smaller plaques in the hippocampus and cortex. In follow-up studies, they confirmed the influence of oral ABX treatment on Aβ deposits in the same mouse model [36,37]. Interestingly, they reported that only a broad cocktail of ABX (composed of kanamycin, gentamicin, colistin, metronidazole and vancomycin) ameliorated the Aβ burden, whereas the application of the individual substances alone had no effect [36]. These results may indicate that a rather broad reduction of several bacterial families is necessary to affect Aβ accumulation.

Notably, ABX treatment exclusively during P14 and P21 (without subsequent long-term ABX-treatment) was sufficient to diminish the Aβ plaque load at 6.5 months of age in the hippocampus and cortex of male APP/PS1 mice [111]. Surprisingly, these findings were not evident in 3- and 5-month-old female APP/PS1 and APP/PS1-21 mice, indicating sex-dependent effects [37,111].

Constant absence of the microbiota reduces Aβ depositions at the early and late stages of AD in germ-free housed mice

As an alternative to ABX-induced depletion of intestinal bacteria, GF mice can be used to analyze the effects of the absence of microbiota on AD-related pathology. In contrast to ABX-treated mice, GF mice lack any colonization with microorganisms, which allows researchers to investigate the effects of a life-long absence of microorganisms or to recolonize mice with a defined set of microorganisms at defined time points. By using GF mice, the potential side effects of ABX on innate immune cells can be avoided. However, GF-housed mice are more artificial and do not reflect a scenario that may occur outside the laboratory. However, the combination of several microbiota manipulation strategies helped to identify new mechanistic insights and improved the understanding of microbiota-dependent effects on AD.

The first publication in the context of AD research that took advantage of GF housing was published in 2017 by Harach and colleagues [55]. They reported that, compared with SPF-housed mice, GF APP/PS1 mice accumulated less Aβ in males and females [27,55]. These results were later confirmed in 5xFAD mice housed under GF conditions (Box 1) [109]. Additionally, treatment of SPF-housed 5xFAD mice with an ABX mixture supplemented in the drinking water 2 months before the terminal endpoint of the experiment caused a reduced Aβ plaque burden compared to SPF mice at 4 months of age. However, GF 5xFAD mice displayed less Aβ pathology, in comparison to ABX-treated 5xFAD mice, indicating that the lifelong absence of microbes might have a more pronounced effect on the Aβ pathology. Additionally, GF housing and ABX treatment prevented neuronal cell loss and cognitive dysfunction during chronical disease stage (10 months of age) [109]. These findings indicate that AD can be modulated both at early stages and in a more chronic phase of the disease in a microbiota-dependent manner. Complementary, colonization of 4-month-old GF-housed APP/PS1 mice with the gut microbiota of aged WT or APP/PS1 mice increased Aβ pathology, with a more pronounced effect when the mice were colonized with microbiota derived from APP/PS1 mice [55].

In, these findings indicate that at early stages of the disease, the microbiota is involved in modulating the disease onset of AD in different mouse models via complementary microbiota manipulation strategies [55,109,110,111]. Notably, this effect is dynamic and can be induced with the help of ABX. Additionally, it was revealed that GF housing and the depletion of the microbiota at late stages of the disease have beneficial effects on further disease progression in 5xFAD mice, preventing neuronal loss and memory function [27,109].

APP processing is not affected by the gut microbiota

One possible mechanism by which the gut microbiota may affect Aβ deposits in transgenic mouse models is by altering APP processing and subsequent Aβ production. However, no differences in APP or BACE1 expression were detected between ABX-treated and control APP/PS1 mice [110]. Further analysis revealed no changes in APP, CTF-α, CTF-β, BACE1, ADAM10 or components of γ-secretase in the hippocampal tissue extracts of 4- and 10-month-old GF, ABX-treated and SPF control 5xFAD mice [109]. These observations indicate that changes in the microbiota do not affect Aβ production in these transgenic mouse models of AD.

However, Harach and colleagues reported increased levels of the Aβ-degrading enzymes NPE and IDE in APP/PS1 GF mice compared to APP/PS1 SPF housed mice [55], which might partially explain the decreased Aβ pathology.

Microbiota-dependent functions of microglia in mouse models of AD

After excluding APP processing as a primary mode of action, the interplay of the microbiota and CNS glial cells might impact AD pathology. Minter et al. reported that long- and short-term ABX treatment of male APP/PS1 mice reduced the number of plaque-associated microglia and astrocytes and altered microglial morphology (Table 1) [110,111]. Additionally, modulating the interaction between the gut and the brain by housing female and male APP/PS1 mice under GF conditions reduced microglial density in the neocortex compared to colonized controls [55].

In contrast, ABX treatment of APP/PS1-21 mice did not significantly alter the number of plaque-associated microglia in 7-month-old, 3-month-old females or males [37], whereas GF 5xFAD mice showed an increase in hippocampal microglia density and plaque-associated microglia [109], indicating overall partially divergent microglial phenotypes across different AD mouse models. Furthermore, a more ramified microglial morphology was found in GF 5xFAD males than in SPF controls, indicating a rather reduced reactive state [109]. Interestingly, transcriptome analysis of cortex tissue homogenates from APP/PS1-21 mice revealed altered expression of genes associated with either homeostatic (e.g.,Mef2a,Junb,Bhlhe41,Fos, andTnfrsf11a) or neurodegenerative microglial phenotypes (includingLgals3,C1qa,C1qb,Cd63, andLag3) in male ABX-treated APP/PS1-21 [37]. Genome-wide bulk RNA sequencing of FACS-isolated microglia from 4-month-old GF 5xFAD males revealed increased expression of genes associated with AD-related activation (Axl,Cst7,Itgax,Cd9 orClec7a), Aβ detection and clearance (Apoe andTrem2), and reduced expression of the homeostatic markerP2ry12 compared with those in SPF controls [109]. Similarly, Aβ uptake was elevated in microglia from GF 5xFAD mice [41,109], which may explain the attenuated Aβ pathology in GF-housed AD model mice.

However, the effects of the microbiota on microglia seem to be age dependent, as GF housing did not alter the (plaque-associated) microglial density or Aβ uptake in aged 5xFAD mice (10 months) compared with SPF controls [109]. The fact that ABX treatment did not affect the number of plaque-associated microglia or Aβ uptake in 5xFAD mice indicated potentially microglia-independent clearing mechanisms [109]. To expand the knowledge of these potential mechanisms, CAMs and their potential contributions to the Aβ burden were examined [141]. Although the density of these CAM populations was lower in GF and ABX-treated 5xFAD mice than in SPF 5xFAD mice, the Aβ uptake of pvMΦ was increased in GF 5xFAD mice as well as upon ABX treatment. Therefore, pvMΦ might contribute to microbiota-dependent Aβ clearance, potentially explaining the reduced Aβ pathology at disease onset as well as at a more progressed stage of the disease [109,141]). Furthermore, pvMΦs participate in vascular Aβ uptake in a PDAPP mouse model during anti-Aβ (ED6) immunotherapy. Compared with IgG-treated controls, anti-Aβ immunotherapy promoted the recruitment of Mac387⁺ and Siglec1⁺ pvMΦs to vascular deposits and was thereby suggested to increase hemorrhagic events [165].

In addition to the Aβ clearance function of microglia, which is boosted under GF housing conditions, some studies highlight a rather harmful role of microglia in AD progression. Microglia depletion via the CSF1R inhibitor PLX5622 from 1.5 months of age reduced Aβ pathology in the cortex and thalamus of 4- and 7-monthold SPF 5xFAD mice when compared to non-treated controls, suggesting a detrimental role of microglia in controlling the Aβ burden [155]. Later, when 10-month-old 5xFAD mice were treated for 1 month with the less effective CSF1R inhibitor PLX3397, no effect on Aβ pathology was reported. However, a depletion of 80% of the microglia ameliorated dendritic spine and neuronal cell loss [156]. Additionally, in 9-week-old APP/PS1-21 males, cortical Aβ plaque load was also reduced after microglia depletion (using PLX5622) when compared to non-depleted controls [38]. In this study, the plaque-modulating effect of microglia depletion was less pronounced in 3-month-old mice, potentially indicating time-dependent effects and a detrimental role during initial plaque formation [38].

In summary, these data support, on the one hand beneficial microglial and pvMΦs functions to eliminate Aβ that are tightly controlled by gut bacteria. However, microglia and pvMΦs may detrimental upon chronic activation, whereby microglia may even contribute to Aβ propagation and spread (d´Errico et al. 2021).

Gut-derived SCFAs modulate microglial function

Since SCFAs are known to constantly modulate microglia biology under physiological conditions [42], several studies have analyzed their impact on microglia in the context of AD. The first indication that SCFA supplementation might impact the pathogenesis of AD was published by Govindarajan et al. [51]. They injected 14-month-old APP/PS1-21 mice for 6 weeks with butyrate (1.2 g/kg body weight) and described an improved associative memory function, compared to vehicle-treated mice. Butyrate is known to inhibit histone deacetylases (HDAC) and accordingly several elevated histone acetylation (H3K14, H4K5, and H4K12) in the hippocampus and cortex of APP/PS1-21 mice compared with WT controls were detected. However, the injection of butyrate did not alter the hippocampal or cortical Aβ plaque load [51]. A more recent study showed that short-term supplementation with a mixture of acetate, propionate and butyrate enhances the Aβ plaque load in GF APP/PS1, without affecting APP processing when compared to controls [27]. This finding was confirmed in GF 5xFAD mice by supplementing acetate alone [41]. Furthermore, in both mouse models, microglia from GFhoused mice that were supplemented with SCFAs presented a more activated amoeboid morphology compared to non-supplemented GF mice [27,41]. Importantly, in SPF housed mice, SCFA supplementation further reduced microglial Aβ uptake compared to non-supplemented SPF mice [27]. Concordantly, GF 5xFAD mice supplemented with solely acetate showed a diminished Aβ uptake comparable to the levels found in SPF mice [41]. Thereby, acetate was identified as the microbiota-derived SCFA modulating microglia Aβ uptake [41].

Additionally, in APP/PS1, an altered expression of genes associated with pathogen recognition, such asTlr7,Tyrobp andCd14, and downregulation ofMyd88, was observed. Furthermore, they showed an upregulation of activation (CD86) and phagocytosis markers (CD68), as well as an activation of the AD relevant APOE-TREM2 pathway [27].

However, Colombo et al. and Erny et al. reported different effects on microglial density in different mouse models. In GF APP/PS1 mice, short-term SCFA supplementation resulted in an increase of plaque-associated microglia [27], whereas acetate supplementation decreased the number of microglia in a GF 5xFAD mouse model [41]. Interestingly, Erny and colleagues reported equal (plaque-associated) microglia density and Aβ pathology in SPF 5xFAD mice and GF 5xFAD mice supplemented with acetate, indicating a regulatory effect of acetate [41]. Furthermore, in acetatetreated GF 5xFAD mice microglia exhibited an increased cytokine expression and reduced phagocytic capacity when compared to non-treated GF housed mice.

Under physiological conditions, the absence of the microbiota with subsequently reduced acetate levels in the brain alters the metabolic state of microglia toward an abnormal amino acid metabolism with impaired arginine, proline and purine metabolism. Notably, H3K4me3 and, more prominently, H3K9ac in microglia were affected by the gut microbiota, including metabolic genes, indicating that the metabolic fitness of microglia is at least partially epigenetically controlled (Fig. 1). Furthermore, metabolite profiling of FACS-isolated microglia revealed a decrease in various fatty acids and lipids in GF mice compared with SPF mice. Moreover, microglia from GF mice have an increased number of mitochondria with a reduced mitochondrial membrane potential and an increased production of mitochondrial ROS compared with SPF microglia [41]. These mitochondrial defects can be ameliorated by acetate supplementation, which fuels the mitochondrial TCA cycle and reduces GF-associated impairments in complex II of the respiratory chain. Interestingly, complex II defects could not be detected under pathological conditions in the 5xFAD mouse model, suggesting an altered metabolic state during disease conditions.

In addition to studies performed in mice, SCFAs have various crucial functions across species, as highlighted in a study that housed pigs under GF conditions. The absence of a microbiota and the subsequent lack of SCFAs impaired physiological functions in the gut, including colonic mobility, blood flow and gastrointestinal pH, which affected the absorption of electrolytes and nutrients. The transplantation of a healthy microbiota rescued the nutrient digestibility and improved the intestinal development and barrier function [193].

Microbiota-derived tryptophan metabolites ameliorate AD pathology

It has been suggested that tryptophan-derived derivatives such as indole-3-lactic acid, indole-3-acetic acid and indole-3-carbinol can affect AD pathology [125]. Recently, tryptophan and indole-3-lactic acid (ILA) were identified as critical gut-derived molecules that probably reduce Aβ accumulation and cognitive impairment via AhR signaling [82]. Additionally, the oral administration ofStreptococcus thermophilus, Lactobacillus reuteri, andLactobacillus delbrueckii to 5xFAD mice could reduce soluble Aβ42 levels in aged animals. This administration correlated with increased levels of tryptophan and ILA in the plasma. Furthermore, a clinical data set analyzing feces, showed thatLactobacillus reuteri is more abundant in healthy individuals than in patients with an early stage of Aβ accumulation. Moreover, postmortem brain analyses revealed increased expression of genes associated with the AhR signaling pathway, suggesting a potential protective effect against AD progression, recapitulating the murine findings [82].

Microbiota-dependent functions of astrocytes in AD mouse models

In addition to microglia, astrocytes are modulated by the gut microbiota in the APP/PS1 mouse model. Short-term ABX treatment altered the number of plaque-associated astrocytes and their morphology(Fig. 2) [110]. Recently, Chandra et al. performed a more in-depth analysis of the effects of short-term ABX treatment and GF housing on astrocytes in the APP/PS1-21 mouse model [24]. Compared with vehicle control-treated male APP/PS1-21 mice, 9-week- and 3-month-old short-term ABX-treated male APP/PS1-21 mice presented a reduced number of plaque-associated astrocytes in the cortex. Fecal microbiota transplantation (FMT) from non-ABX-treated APP/PS1-21 mice into short-term ABX-treated APP/PS1-21 mice restored the ABX-induced changes in astrocyte cell numbers. No differences were observed in females, indicating sex-specific effects. Compared with those in mice raised under SPF conditions, plaque-associated astrocytes were decreased in 9-week-old APP/PS1-21 males housed with GF, whereas no differences were detected in the number of nonplaque-associated astrocytes [24]. Additionally, ABX treatment and GF housing similarly increased the morphology of astrocytes compared with vehicle-treated and SPF-housed APP/PS1-21 mice. Interestingly, in short-term ABX-treated microglia-depleted (PLX5622) 3-month-old APP/PS1-21 mice, no astrocytic morphological alterations were observed compared with those in vehicle-treated mice. However, the previously described reduction in plaque-associated astrocytes was maintained, indicating the existence of microglia-independent and microglia-dependent mechanisms [24]. In general, the gut microbiota-dependent interactions of astrocytes and microglia are not yet fully understood in the context of AD.

Gut-microbiota driven alteration in the development of AD. The absence of gut microbiota via germ-free (GF) housing or its reduction via antibiotic-treatment (ABX), results in a reduction of gut microbiota produced short-chain fatty acids (SCFAs) and tryptophan-derived metabolites. Microglial Aβ clearance is controlled by gut-derived acetate. Additionally, the cell density of perivascular macrophages (pvMΦ) is reduced, accompanied with an increased Aβ uptake upon GF-housing and ABX-treatment, resulting in ameliorated AD pathology

Full size image

Several studies have focused on the impact of SCFAs on microglia; however, astrocytes can also be modulated by SCFAs. Sun and colleagues first demonstrated this effect by performing long-term dietary SCFA supplementation in APP/PS1 mice [159]. They reported a significant improvement in learning and spatial memory abilities, as well as a reduction in Aβ pathology. Additionally, the SCFA diet upregulated astrocytic glutamine synthetase, indicating increased metabolic astrocyte–neuron coupling [159].

Microbiota-derived effects on tau pathology

Previous studies focused mainly on Aβ pathology, whereas a recent study determined the effect of the gut microbiota on tau pathology in a P301S tau transgenic mouse model with humanized ApoE (Box 1) [146]. HumanAPOE is expressed in three major genetic isoforms, wherebyAPOE4 is considered to be the most important genetic risk factor for AD [137]. Short-term ABX treatment and GF housing ameliorated tau pathology in 40-week-old P301S tau mice, resulting in improved behavior and inhibited neuronal cell loss. ABX treatment showed sex differences, with reduced levels of phosphorylated hippocampal tau found only in males. Furthermore, and in accordance with previous studies, SCFA supplementation in GF P301 tau mice with theAPOE4 isoform exacerbated AD pathology [146]. Overall, this study hypothesized that the gut microbiota can regulate the immune response to tau pathology potentially via SCFAs, but the underlying mechanisms remain to be determined.

Microbiota dysbiosis and its impact on cognitive function

Several studies have shown that the microbiota composition of AD mice is altered compared to WT mice. To address the question if the transmission of the AD-associated microbiota is sufficient to impact cognitive function in WT animals, Zhang and colleagues performed cohousing studies [191]. Specifically, they cohoused two-month-old WT female mice with sex- and age-matched 5xFAD mice for a period of 3 months, caused a cognitive impairment of the WT animals. Furthermore, an increased abundance ofDubosiella andPeptococcaceae was reported. These bacteria are known to be associated with inflammation, decreased amounts of butyrate, increased levels of tau phosphorylation and increased accumulation of Aβ42. Additionally,Bacteroides fragilis was reported to trigger an AD-like pathology with cognitive dysfunction in mice [181]. However, there are also contradictory studies suggesting a decrease inBacteroides species [97,194]. Additionally, new studies have shown that Agathobacter, a butyric acid-producing bacteria belonging to the Lachnospiraceae family, might have a neuroprotective effect. In APP/PS1 mice, the administration ofA. rectalis reduced Aβ accumulation by inhibiting microglial activation [101]. The authors suggested that the underlying mechanism could be the butyrate-dependent regulation of the Akt/NF-κB pathway.

A critical role for the microbiota was further determined in zebrafish, whereby zebrafish housed under GF conditions failed to develop physiological social behavior due to impaired microglial remodeling [18]. Interestingly, microglial density was lower in GF zebrafish larvae compared to colonized controls, which contrasts with findings in mice [18,42]. Furthermore, monocolonization with distinct bacterial strains (e.g.,Aeromonas veronii,Enterobacter cloacae, andStaphylococcus sp.) could restore neurodevelopmental features, indicating a dynamic bacterial effect on microglia and CNS development [18].

Microbiota-dependent alteration of the peripheral immune system

Data from clinical and preclinical studies indicate that the peripheral immune system might be involved in AD. Epidemiological studies revealed a dysregulated peripheral immune system in AD patients [86], with various inflammatory markers, peaking during the initial symptomatic phase [149]. Additionally, the risk to develop AD is increased by the dysregulation of peripheral immune markers, including C-reactive protein, IL-6 and IL-1β [31,149]. Furthermore, patients with a history of infectious disease necessitating hospitalization have an increased AD risk [153]. In addition to epidemiological studies, animal studies underline the interplay of the peripheral immune system and AD. Generally, it is known that peripheral bacterial and viral infections can induce inflammatory pathways in the brain [8,74]. In APP/PS1 mice, a systemic inflammatory stimulus via LPS activated microglia and impaired their Aβ clearance capacity [166]. In contrast, sequential low-dose LPS stimulation can induce immune memory in microglia cells, causing a tolerance effect [178]. Furthermore, repetitive low-dose LPS stimulation reduced Aβ pathology and neuronal cell death in the APP23 AD mouse model [178]. These effects are reportedly mediated via epigenetic alterations of microglia. Increased acetylation of H3K27 and methylation of H3K4, indicating active enhancers, were altered and associated with endocytosis and phagocytosis. In contrast, a single LPS stimulation worsens Aβ pathology and alters the epigenetic landscape of microglia toward the upregulation of pathways associated with TNFα and HIF-1 signaling.

To investigate the impact of the microbiota on the peripheral immune system and subsequently on the pathogenesis of AD, further studies were performed in APP/PS1 and APP/PS1-21 mice. In both mouse models, short-term and long-term ABX-treatment altered the serum concentrations of various cytokines and chemokines in a sex-specific manner [37,110,111]. However, these studies reported partially contrary results, and these findings need to be clarified in future studies.

Alterations of the human gut microbiota in AD patients

In addition to data obtained from rodents, studies with human AD patients have shown a potential interplay between the gut microbiota and AD (Table 2). Initially, two studies published in 2017 revealed alterations in the gut microbiota composition in AD patients. First, Cattaneo et al. reported an increased abundance ofEscherichia andShigella species and decreased levels ofEubacterium rectale compared with non-AD controls [22]. Second, Vogt et al. performed 16S rRNA sequencing of fecal samples and reported decreased microbiota diversity in AD patients. Furthermore, they described an increase inBacteroidetes and a decrease inFirmicutes. [173]. Reduced abundance ofFirmicutes was also found in another Chinese study including patients with mild cognitive impairment (MCI) and AD [98]. However, other studies reported a decrease inBacteroidetes [97,194]. It was further suggested that the gut microbiota composition might change already early in the disease process and correlated rather with Aβ and tau levels, and to a lesser extent with the degree of neurodegeneration [44].

In summary, these correlative studies describe alterations of the microbiota composition of AD patients. However, the alterations seem to be partially inconsistent across studies, and it remains unclear if and to what extent the reported dysbiosis affects AD progression including bacteria-derived molecules that may affect the AD pathology. In addition to AD, several other CNS diseases have been shown to be affected by the gut microbiota.

Parkinson’s disease and the role of the gut microbiota

In addition to AD, other neurodegenerative diseases, including Parkinson’s disease (PD), are influenced by the microbiota. Among other symptoms, PD patients develop bradykinesia, rigidity, postural instability, and tremors in the hands, arms and legs. Pathologically, PD is characterized by a loss of dopaminergic neurons in the substantia nigra and an accumulation of intracellular protein deposits (Lewy bodies). Interestingly, in PD patients, constipation is among the first symptoms, partially preceding neurological symptoms by decades [2]. Furthermore, the composition of the gut microbiota is altered in PD patients, with an increase inEnterobacteriaceae,Lactobacillus,Bifdobacterium, andAkkermansia [117,148,169]. Additionally, opportunistic pathogens, includingCorynebacterium,Porphyromonas,Alistipes,Bacteroides,Escherichia,Megasphaera andDesulfovibrio, are elevated in PD patients. Moreover, various bacteria, e.g.,Blautia,Coprococcus,Roseburia,Lachnospira, and the SCFA-producingFusicatenibacter andFaecalibacterium, are decreased [117,148,169]. Consistent with this reduction, decreased fecal SCFA levels can be found in PD patients. Interestingly, the reduction of SCFA-producing bacteria in PD patients is correlated with the severity of cognitive and motor symptoms [26,170]. However, SCFA levels in the plasma of PD patients are increased, potentially because of elevated intestinal permeability. In addition, PD patients presented increased expression of several inflammatory markers in the colon and feces, including CCL2, CCL5, CCR5, IL-1β, IL-6, IL-8, IL-17A, IFN-β, IFN-γ, TNFα, TLR2, and TLR4, as well as increased numbers of CD3⁺ T cells in the colon [65,126,145]. Furthermore, in rodent and nonhuman primate studies, the intraintestinal injection of various forms of misfolded alpha-synuclein can reach the locus coeruleus and substantia nigra via the vagus nerve [60]. This transmission can be mitigated by surgical vagotomy [83]. Moreover, a cohort of patients in Denmark who underwent vagotomy from 1977–1995 presented a decreased risk for PD [160].

In a mouse model overexpressing α-syn (Thy1-aSyn), ABX treatment (from 5–6 weeks of age until 12–13 weeks) ameliorated α-syn-dependent motor deficits and reduced microglia activation compared to non-treated mice [140]. In the same study it was shown that GF housing reduced motor deficits, α-syn accumulation and microglia activation in comparison to SPF mice, effects which could be reversed via SCFA supplementation. Additionally, transplantation of faces obtained from PD donor patients worsened motor deficits more prominently in Thy1-aSyn mice than in mice transplanted with faces from matched healthy donors [140]. FMT from toxin-induced PD mice (MPTP) into WT mice can also impair their motor function [158].

Microbiota-dependent effects on multiple sclerosis

Multiple sclerosis (MS) is considered an autoimmune disease of the CNS whose pathogenesis is also influenced by the microbiota. MS is characterized by demyelination and inflammation of the CNS and affects approximately 2.8 million people worldwide [174]. Various studies have reported alterations in the microbiota of MS patients, with increases inAkkermansia,Methanobrevibacter,Ruthenibacterium lactatiformans,Hungatella hathewayi, andEisenbergiella tayi [11,68] and decreases in SCFA-producingButyricimonas,Faecalibacterium, andClostridium cluster IV and XIVa [68,112]. Similarly, fecal SCFA levels are slightly lower in MS patients than in controls [68]. In the serum of MS patients, butyrate levels are dampened [142], however, acetate levels are increased [127]. These results indicate that alterations in the gut microbiota are associated with MS risk and disease progression. However, a robust conserved pattern of microbiota alterations remains to be defined.

Additionally, in a monophasic experimental autoimmune encephalomyelitis (EAE) mouse model, the gut microbiota affects MS pathogenesis. ABX treatment [185] and the absence of complete microbiota caused by GF- housing attenuated EAE and spontaneous MS pathogenesis while reducing the expression of proinflammatory cytokines [12,94]. Mechanistically, these beneficial effects are caused by a reduced T and B-cell response as well as a more pronounced T reg response [12,94]. Interestingly, SCFA supplementation (especially propionate) increased the Treg population, decreased Th1 and Th17 cells, and ameliorated the annual relapse rate and brain atrophy in human MS patients [39].

In a cuprizone toxin-induced demyelination model, GF housing reduced microglial cell numbers and increased oligodendrocyte density, an effect that was reversible upon colonization [107]. Additionally, the colonization of GF-housed EAE mice with MS patient faces worsened symptoms more intensely than FMT from healthy individuals did [23].

In addition to SCFAs, AhR ligands were identified as further microbiota-derived metabolites (produced by, e.g.,Peptostreptococcus russellii andLactobacillus spp. [3]) that modulate the pathogenesis of EAE [133]. AhR signaling increases TGF-α and VEGF-B production in microglia, eliciting proinflammatory neurotoxic astrocyte activities and thereby subsequently worsening EAE [133,134]. In addition to the impact of the gut microbiota on EAE, the microbiota residing in the lung influences the outcome in a rat model of EAE. Lung dysbiosis was induced by intratracheal neomycin treatment and subsequently caused decreased infiltration of B cells and CD4⁺ and CD8⁺ T cells into the CNS. Furthermore, microglial type I IFN activation is dampened, hence ameliorating the severity of EAE [64].

Outlook: from mouse models to potential AD treatment strategies

Modulating microglia and CAMs directly or indirectly might be a promising strategy to treat AD and other CNS diseases. The gut microbiota is able to impact microglia; therefore, altering the gut microbiota composition or providing directly beneficial bacteria-derived molecules might be a promising strategy to adjust microglia for the treatment of CNS diseases. The strategies used to alter the gut microbiota might include probiotics/prebiotics, FMT, diet or the direct application of beneficial bacterial molecules (Duscha et al. 2020).

In APP^NL−G−F mice, oral supplementation withBifidobacterium breve (for 4 months beginning at 3 months of age) reduces microglial cell density in the hippocampus and ameliorates the hippocampal Aβ plaque load and memory impairment [1]. The beneficial effects of probiotics were further shown in (i) APP/PS1 mice (supplementation withBifidobacterium Lactis Probio-M8 for 45 days starting at 4 months of age decreased the cortical Aβ plaque load and improved cognitive performance, as determined via the Y-maze test) [20] and (ii) 3xTG AD mice (supplementation with SLAB51, a mixture of 9 bacterial strains, for 4 months starting at 8 weeks of age decreased the Aβ plaque load and improved cognitive performance, as determined via novel object recognition) [13].

Furthermore, modulating the gut microbiota composition via the drug sodium oligomannate (GV-971) was reported to ameliorate AD pathogenesis in 5xFAD mice. Wang et al. gavaged 6-month-old 5xFAD mice for one month with GV-971 (Table 1) and reported decreased hippocampal Aβ and tau accumulation, improved cognitive performance, and decreased microglial density in the hippocampus when compared to untreated 5xFAD mice [176]. A recent publication confirmed the beneficial effects of sodium oligomannate treatment on AD pathogenesis in APP/PS1 mice. Daily treatment with sodium oligomannate starting at the age of 8 weeks reduced cortical Aβ pathology in 12-week-old males and females. Although the beneficial effect on Aβ accumulation was less pronounced in females [14].

In addition to probiotics and prebiotics, diet is an important influencing factor on the microbiota composition and has been shown to influence the pathogenesis of AD in various mouse models. A high-fat diet was reported to exacerbate the Aβ plaque load in APP/PS1 [17], 5xFAD [108] and APP^NL−F mice [105]. In contrast, a ketogenic diet can decrease the Aβ plaque load and ameliorate cognitive performance and neuronal cell loss in 5xFAD mice [182]. A ketogenic diet is characterized by nearly complete elimination of carbohydrates, which results in ketone body production and their consumption as main energy source. Among various ketone body species, β-hydroxybutyrate is the most abundant. β-Hydroxybutyrate is known to inhibit the NLRP3 inflammasome [188], which is chronically activated in various AD mouse models [59]. Compared with vehicle treatment, β-hydroxybutyrate supplementation via drinking water (8 weeks, 0.01875 g/ml) not only reduced inflammasome activation in 5xFAD mice but also decreased the Aβ plaque load. [150]. Furthermore, β-hydroxybutyrate supplementation decreases the density of (plaque-associated) microglia in 5xFAD mice. Interestingly, β-hydroxybutyrate supplementation had no effect on the microglia of WT mice [150].

In addition to dietary interventions, intermittent fasting can modulate the gut microbiota and, e.g., increase the amount ofFirmicutes while decreasing Bacteroidetes when compared to mice fed ad libidum. Furthermore, intermittent fasting reduces microglial density and subsequently ameliorates Aβ pathology and cognitive dysfunction in 5xFAD mice. Furthermore, supplementation with amino acids found to be increased during intermittent fasting (sarcosine and dimethylglycine) can mimic the beneficial effects of intermittent fasting [123].

Treatment strategies might include the modulation of microglia, potentially by altering the microbiota via prebiotics, probiotics or diet intervention. Furthermore, the phagocytic activity of microglia can be increased in aging mice via CD22-blocking antibodies. Thereby, the uptake of myelin debris, amyloid-β oligomers and α-synuclein fibrils could be increased, subsequently improving cognitive function [128]. A further strategy is based on the depletion of dysfunctional microglia, followed by the engraftment of hematopoietic cells. A recent study reported exacerbated Aβ pathology and reduced microglia clustering around plaques in TREM2-deficient 5xFAD mice. They then depleted the microglia in TREM2-deficient 5xFAD mice via CSF1R inhibition and replaced them by infusing TREM2^WT hematopoietic cells. After engraftment, these circulation-derived myeloid cells clustered around plaques and reduced the Aβ pathology when compared to mice transfused with TREM2^KO hematopoietic cells [186].

In summary, the abovementioned studies pave the way for an improved understanding of how microglia and their interactions with the microbiota can support the development of future AD, PD and MS treatments.

No datasets were generated or analysed during the current study.

We thank J. Porter-Maguire for comments and proofreading. Figures were created with BioRender.com. We thank all members of the group for critical and constructive data discussions.

Open Access funding enabled and organized by Projekt DEAL. DE is supported by the Deutsche Forschungsgemeinschaft, Germany, Project-ID 259373024 – TRR 167 and Project number 491676693 – TRR 359. DE was further supported by the Berta-Ottenstein-Programme for advanced Clinician Scientists, the Else Kröner-Fresenius-Stiftung and Ministry of Science, Research and Arts, Baden-Wuerttemberg under the aegis of JPND.

1. Philipp Schaible and Julia Henschel have contributed equally.

Authors and Affiliations

1. Institute of Neuropathology, Medical Faculty, University of Freiburg, Breisacher Str. 64, 79106, Freiburg, Germany

   Philipp Schaible, Julia Henschel & Daniel Erny

2. Faculty of Biology, University of Freiburg, Freiburg im Breisgau, Germany

   Philipp Schaible & Julia Henschel

Contributions

PS, JH and DE wrote the manuscript. PS and JH created the figures.

Corresponding author

Correspondence toDaniel Erny.

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visithttp://creativecommons.org/licenses/by/4.0/.

Reprints and permissions