Gram-negative bacteria and their lipopolysaccharides in Alzheimer’s disease: pathologic roles and therapeutic implications

Under healthy conditions, most intestinal microbiomes interact with the brain through several mechanisms, including neurotransmitter generation, and contribute to the maintenance of brain homeostasis [62]. Increasing evidence suggests that the gastrointestinal tract is the bridge between the microbiota and the CNS [63]. The link between the microbiome and brain disorders emerged from the impact of gastrointestinal microbes on the development of microbial byproducts in the brain [64]. The microbial-derived byproducts are active mediators of gut-brain communication and may be potential therapeutic targets for neurodevelopmental and neurodegenerative disorders. Particularly, bacterial byproducts and exotoxin molecules, such as capsular proteins, flagellin, short-chain fatty acids (SCFA), fimbrillin, peptidoglycan, proteases, gingipain, vacuolating cytotoxin A (VacA), and methylglyoxal (MG), may be considered pathogen-associated molecular patterns. Moreover, exotoxins interact with pattern recognition receptors, such as toll-like receptors (TLR) 2 and 4. Thus, bacteria imbalance and gut exotoxins induce neuroinflammatory reactions, such as microglial cell activation, affecting the function of neuronal cells [65]. The dysregulated microbiota-induced inflammation may also lead to the invasion of microbes or microbial byproducts into the brain, neuroinflammation, and production of Aβ and phosphorylated tau [66]. One study reported that alterations of the composition of gut microbiota in APP/PS1 mice are related to the increased Aβ levels in the brain and impairment of cognitive function. Moreover, another study found that the dysregulation of microbiota, intestinal epithelial barrier dysfunction, and vascular Aβ deposition occur in the intestine before the onset of cerebral Aβ deposition in Tg2576 mice [67]. These reports suggest that microbiota dysregulation is related to the development and progression of AD.

The precipitate, which shows a β-folded sheet structure located vertical to the fibrous axis and is rich in aggregated insoluble lipoproteins, was designated as amyloid. Because of the hydrophobic nature of the aromatic amino acid peptides that compose the primary sequence of APP-derived amyloid, self-aggregation of amyloid monomers compiles over time into dimers, oligomers, and fibrils. Significant inflammatory reactions and neurodegeneration from amyloid accumulation appear in the brain, which affects neurodegenerative diseases, such as AD, Parkinson's disease, and prion disease [68]. Surprisingly, many studies have revealed the presence of bacteria-produced amyloids [68‐70] (Fig. 2). The bacterial amyloids have a biophysical nature that is highly similar to human-derived amyloids, including the aggregate-forming ability [71, 72]. Although the amyloid produced by bacteria differs in its primary structure from the amyloid produced in the CNS, they have similarities in their tertiary structure [73]. Exposure to bacterial amyloid proteins in the gut could provoke systemic inflammation [74]. Moreover, bacterial amyloid affects disease progression by interacting with Aβ present in the AD brain (Table 3). The interaction of bacterial amyloid with neuronal amyloid, which is endogenously produced in the brain, promotes either their aggregation or cross-seeding in the AD brain [75, 76]. Furthermore, bacterial amyloid significantly promotes Aβ pathology in AD [69]. Curli is a well-known gram-negative bacterial amyloid. It is an integral part of the biofilm extracellular matrix produced by certain strains of enterobacteria, such as E. coli [77]. CsgA and CsgB are not only two major structural components of curli fibers, but are also essential components of biofilms [78]. In particular, the structure of the fibril produced by CsgA and CsgB of curli proteins is very similar to that of amyloid [79]. Interestingly, the fibril includes a β-sheet structure similar to Aβ and has been reported to form fibrillar aggregates [80]. The bacterial amyloid from curli may be a potential contributor to Aβ pathology in AD since several amyloid proteins interact with Aβ to induce co-aggregation or cross-seeding [37, 75]. In addition, FapC, a bacterial amyloid from Pseudomonas, is another strong contributor to Aβ pathology [81]. The FapC fibril is a powerful accelerator of Aβ fibrillization in AD [82]. In the co-culture of Aβ and FapC seeds, the rate of increase of the β-sheet ratio was more than three times above the Aβ-only culture [82]. Furthermore, FapC has been reported to promote Aβ-associated pathology several times faster in the AD zebrafish model [82]. Surprisingly, the bacterial amyloid and Aβ can bind to the same receptor due to their structural similarity. For example, similar to Aβ, the bacterial amyloid from curli can bind to the TLR2-TLR1-CD14 (cluster of differentiation 14) complex, which facilitates the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) pathway [83]. Moreover, MG, an intermediate metabolite of glucose metabolism in cells, is widely secreted by gram-negative bacteria such as E. coli, and distributed at a high rate in the body [84]. The MG acts as a ligand for the receptor for advanced glycation end products (RAGE), resulting in increased beta-site amyloid precursor protein cleaving enzyme 1 (BACE-1) expression and Aβ levels [85, 86]. In addition, it has been shown that P. gingivalis increases Aβ production by increasing the gene expression of APP and BACE1 and decreasing the gene expression of disintegrin and metalloproteinase domain-containing protein 10 (ADAM10) [59, 87]. Moreover, gingipain, a family of proteases secreted by P. gingivalis, can affect to activate γ-secretase through cleavage of caspase-3 [87]. In addition, three proteinase genes that contribute to the virulence of P. gingivalis, RgpA, RgpB, and Kgp, have been associated with Aβ production. In particular, RgpB has been reported to induce massive generation of Aβ by activating the metalloproteinase meprin β, which is an alternative BACE1 cleavage of APP [88]. Furthermore, the load of KgpB is particularly increased in the AD brain compared to the healthy brain, and RgpB has been reported to co-localize with Aβ [59]. Collectively, these findings suggest that bacterial molecules from gram-negative bacteria could induce the production and aggregation of Aβ, affecting the onset and progression of AD.

Hyperphosphorylated tau aggregation is the main pathological hallmark of AD [98]. Neurofibrillary tangles (NFTs) consist of hyperphosphorylated and aggregated microtubule-associated protein tau [99], while the intermediate form of tau causes cytotoxicity and cognitive impairment [100]. Tau hyperphosphorylation is regulated by various kinases that are affected by numerous factors, including gram-negative bacteria [101, 102] (Fig. 2). Helicobacter pylori induces tau hyperphosphorylation by activating the glycogen synthase kinase-3β (GSK-3β) [102]. Moreover, it has been revealed that P. gingivalis can increase tau hyperphosphorylation at, e.g., Thr231 and Ser396 residues, in human iPSC-differentiated neuronal cells and C57BL/6 mice [87, 103]. Particularly, P. gingivalis infection activates GSK-3β [104]. Furthermore, gingipain can contribute to tau hyperphosphorylation by regulating the protein kinase B (Akt)/GSK-3β activity by cleaving procaspase-3 to caspase-3 [59, 105]. Gingipain is involved in tau fragmentation and generation of paired helical filament through tau proteolysis; the tau fragments can induce tau aggregation and phosphorylation [106‐108] (Table 3). Interestingly, Kgp has been reported as a trigger or accelerator of tau pathology [59]. The hexapeptide motif-containing tau peptide generated by Kgp can be easily hyperphosphorylated and contribute to the formation of paired helical filaments and NFT. Similarly, MG has been reported to induce tau hyperphosphorylation through the GSK-3β activity [89]. Furthermore, DNA derived from several species of gram-negative bacteria, such as E. coli and P. gingivalis, has been reported to promote tau pathology [109]. In particular, the gram-negative bacterial DNA—frequently reported in patients with AD—strongly induces tau misfolding and aggregation [109]. Therefore, gram-negative bacteria could initiate or exacerbate tau pathology by inducing tau hyperphosphorylation and aggregation in AD.

Neuroinflammation is a pathological hallmark induced by abnormally aggregated Aβ peptides in AD [110, 111]. Moreover, microglia activated by Aβ can accelerate neurodegeneration in the brain during AD [111]. Gram-negative bacteria are known triggering factors for inflammatory responses [112]. A study confirmed that the effect of microbiota on microglial maturation in germ-free mice could be regulated by SCFA, a byproduct of bacterial metabolism [94]. Furthermore, the microbial metabolites of tryptophan can modulate astrocyte activity [97]. These results show that the bacterial-derived byproducts, such as VacA, SCFA, phosphorylcholine, and tryptophan [90, 94‐97], are involved in neuroinflammation by modulating microglia and astrocyte activity. Although this evidence demonstrates the pivotal role of byproducts from gram-negative bacteria in neuroinflammation, only a few studies have investigated their effects in AD. A recent study showed that the P. gingivalis oral infection causes strong microglial activation in the brains of apolipotein E (ApoE)^−/− mice [60]. In particular, gingipain from P. gingivalis can lead to the release of neuroinflammatory cytokines in an AD brain [59] (Fig. 2). In addition, respiratory infection of Bordetella pertussis in APP/PS1 mice increases brain infiltration of T cells and activation of microglia and macrophages [113]. Helicobacter pylori infection stimulates the secretion of pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α) and IL-6, causing inflammation-related neurodegeneration [61]. Moreover, it has been reported that monocytes infected with Chlamydia pneumoniae (C. pneumoniae) may contribute to late-onset AD by inducing secretion of pro-inflammatory cytokines and chemokines from microglia and astrocytes [114]. Furthermore, the C. pneumoniae-infected microglia show increased levels of TNF-α [115], a critical neuroinflammatory factor in AD. These findings imply that the gram-negative bacteria introduced into the CNS can aggravate AD pathology through strong neuroinflammation.

Neuronal loss is a prominent pathological feature of AD [3]. Surprisingly, gram-negative bacteria, H. pylori, can induce neuronal cell death by secreting VacA [91]. Moreover, MG from gram-negative bacteria—a cell death-related toxin—has been reported to induce oxidative stress and apoptosis through activation of nicotinamide adenine dinucleotide phosphate oxidase and production of reactive oxygen species [64]. In AD, MG is a potential key factor that triggers neuronal death through the decrease of mitochondrial membrane potential, down-regulation of Bcl-2, and up-regulation of pro-apoptotic proteins, such as caspase-3 and Bax [116, 117]. These mitochondrial dysfunctions caused by MG can accelerate AD neurodegeneration. TLR4 activates the nod-like receptor protein 3 by recognizing gram-negative bacteria; it is a key receptor for the onset of neurodegenerative diseases, including AD [118]. Therefore, activation of TLR4 signaling by gram-negative bacteria can lead to neuronal cell death [119]. Importantly, gram-negative bacteria can trigger neuronal cell death through not only their byproducts but also the bacteria themselves (Fig. 2). Taken together, both gram-negative bacteria and their byproducts could play a key role in neuronal cell death and neurodegeneration in AD.

Aβ accumulation is significantly reduced in sterile APP mice, but consistently increased in LPS-treated APP mice [32, 137], highlighting the role of LPS, apart from established AD pathogenic factors, and presenting LPS as a potential risk factor, equally strong as AD’s genetic components. LPS modulates Aβ production by significantly increasing the activity of APP-cleaving enzymes, such as BACE-1 and γ-secretase, while decreasing α-secretase activity [138] (Fig. 4). Moreover, LPS increases mRNA expression of APP and contributes to the production of Aβ in the hippocampus through the cathepsin B-related mechanism [139]. In LPS-injected rodents, BACE1 immunoreactivity and Aβ accumulation were found in the ipsilateral cerebral cortex and hippocampal formations [140]. Moreover, LPS increases Aβ oligomers by promoting Aβ aggregation [18, 141]. Furthermore, the LPS-induced systemic inflammation could provoke Aβ clearance impairment via (1) down-regulated expression of low-density lipoprotein receptor-related protein 1 (LRP-1); (2) inhibition of Aβ entry into the blood vessels in the brain; and (3) dysfunction of p-glycoprotein [142]. These studies demonstrate that the LPS-induced Aβ burden and Aβ plaques could play key roles in Aβ-related AD pathology.

LPS is potentially instrumental in exacerbation of tau pathology (Fig. 4). First, LPS induces tau phosphorylation through not only hypoactivation of tau phosphatase but also activation of cyclin-dependent kinase 5 (CDK-5) and GSK-3β [143, 144]. LPS may stimulate the activation of GSK-3β through regulating the phosphoinositide-3-kinase (PI3K)/Akt pathway, increasing the production of phosphorylated tau [104, 145]. In particular, LPS can increase phosphorylation of tau in the hippocampus [144, 146, 147]. Second, LPS induces tau aggregation in the AD brain [143]. Several studies have reported that LPS contributes to the tau aggregation through activation of CDK-5 and GSK-3β in the 3xTg model [143, 144]. Accumulating evidence suggests that LPS accelerates tau pathology by acting as an exogenous regulator of kinases such as mitogen-activated protein kinase (MAPK), GSK-3β, c-Jun N-terminal kinases, and p38 [143, 148]. These results suggest that LPS may affect tau pathology by inducing hyperphosphorylation and aggregation of tau. Moreover, LPS could play a key role in tau hyperphosphorylation and aggregation, which is considered a major factor and therapeutic target for AD pathology.

Accumulating evidence suggests that LPS contributes to AD pathology through glial activation modulation (Fig. 4). It has been reported that LPS administration increases microglial density in the brain [141]. In particular, LPS is an agonist of glial TLR4, activating the myeloid differentiation primary response 88/toll/interleukin-1 receptor-domain-containing adapter-inducing interferon-β (MyD88/TRIF) pathway and promoting pro-inflammatory responses by activating NF-κB [149‐151]. In contrast, some studies have suggested that LPS promotes the anti-inflammatory response [152, 153]. It has been proposed that continuous mild LPS exposure reduces inflammatory responses in the brain by suppressing pro-inflammatory mediators and boosting anti-inflammatory mediators [152]. In addition, several studies have revealed that low-dose LPS can induce anti-inflammatory responses in AD animal models [154‐156]. The possible underlying mechanism is that the low-dose LPS exposure reduces inflammation via upregulating the expression of MyD88-dependent signaling pathway inhibitors like IRAK-M, Ship, and Tollip [157]. Unfortunately, the precise mechanism by which mild LPS exposure provokes the anti-inflammatory response is not known. Although the role of LPS-exposed microglia in the AD brain is not clear, accumulating evidence suggests that LPS may exacerbate AD pathology through interactions with several receptors related to neuroinflammation [158, 159]. Notably, LPS significantly increases the expression level of RAGE [160], a receptor critically involved in AD pathology such as Aβ production and clearance, tau pathology, and synaptic degeneration [161]. In particular, the microglial RAGE-dependent signaling pathway plays a causative role in neuroinflammation, Aβ deposition, and cognitive impairment in AD [162]. Interestingly, it has been reported that stimulation of RAGE by LPS increases endothelial permeability and activates NF-κB [163]. The glial NF-κB activation by LPS leads to secretion of pro-inflammatory cytokines. In addition, recent studies have suggested that LPS can induce leukocyte infiltration into the brain and microglial activation by increasing the generation of reactive oxygen species through NADPH oxidase 2 (NOX2) activation [164]. The widespread high-level LPS in the AD brain alone can cause pathological and excessive neuroinflammatory reactions. Interestingly, LPS is also a ligand for triggering receptors expressed on myeloid cells 2 (TREM2), a receptor that regulates microglial phenotype switching [165]. Stimulation of TREM2 by LPS can convert the microglial phenotype from an anti-inflammatory phenotype to a pro-inflammatory phenotype [166]. Particularly, not only Aβ-induced neuroinflammation but also LPS-induced neuroinflammation could facilitate neurodegeneration and cognitive impairment [167]. These results suggest that the LPS-induced neuroinflammation could act as a direct and fatal factor in AD pathology and cognitive dysfunction.

LPS induces synaptic loss within the CNS [168] and reduces synaptic plasticity in the brain [169]. Moreover, LPS affects the inhibitory and excitatory synapses of adult-born hippocampal neurons, induces neuronal and synaptic loss, and reduces cognitive function [170, 171], supporting the hypothesis that LPS in AD is involved in neurodegeneration (Fig. 4). In addition, LPS can inhibit neuronal function via damage to myelin in AD [172]. It has been reported that LPS causes damage to oligodendrocytes and increases myelin basic protein degradation in the AD brain. The first mechanism for LPS-induced neurodegeneration is that LPS induces the activated p38α MAPK signaling pathway in microglia and increases TNF-α secretion [173, 174]. In addition, LPS is one of the potent factors capable of activating NOX2 in the CNS [164, 175]. The NOX2 activation in glia and neurons can induce neuronal cell death through massive oxidative stress, which has been suggested as a contributor to several neurodegenerative diseases, including AD [176, 177]. The LPS-induced neuroinflammation such as NOX2 activation can be a possible additional contributor to neurodegeneration in AD pathology. Second, LPS facilitates neurodegeneration through mitochondrial dysfunction [178]; namely, LPS may affect mitochondrial fusion genes, such as mitofusin (Mfn)1, Mfn2, and OPA1, which are important in neurodegenerative diseases, including AD. In addition, LPS leads to neurodegeneration by inducing oxidative stress and triggering the mitochondrial apoptotic pathway [179]. Third, LPS can directly induce neuronal cell death through neuronal TLR4, which is a major receptor that plays a key role in the activation of the inflammatory response on AD. LPS not only increases TLR4 expression but also acts as a ligand for neuronal TLR4, inducing the transcription of caspase-11 and promoting the activation of the inflammasome [180, 181]. Since TLR4 expression is also increased by aging and Aβ, the interaction between LPS and TLR4 may be more fatal to AD [182]. NF-κB, a well-known downstream molecule of the TLR4/Myd88/TRIF signaling pathway, is also known to be important for neuronal survival and acts either as a pro-apoptotic or anti-apoptotic factor [183]. Consequently, LPS can cause neuronal death by directly acting on neurons, such as acting on neuronal receptors followed by influx into neuronal cells through OMV [181]. Taken together, these studies demonstrate that LPS can induce neurodegeneration and affect the initiation and progression of AD.