Potent of strategic approaches for tauopathies ranging from single-cell transcriptome to microbiome

ABSTRACT

Tauopathies as one of the typical neurodegenerative diseases are characterized by pathologically aggregated tau in neuron and glia, which causes dysfunction, synapse loss, and cognitive impairments. There are more than 26 cases that are classified according to major tau isoform aggregates production or abnormal tau generation processes. Although the various concepts for targeting tau mechanisms have been investigated, the obvious limitations still remain and consequently, hindering their applicability. Therefore, additional strategies are required in the therapeutics and diagnosis. Here, we briefly summarized the etiological tau in the hyperphosphorylation, and the present concepts of tau-targeted disease therapies and diagnosis. We suggest the potential of approaches from deep learning-based single-cell transcriptome to microbiome and discuss the encountered challenges for the applicability and practicality.

KEYWORDS:

• Tauopathies
• deep learning
• single-cell transcriptome
• microbiome

Introduction

In recent years, neurodegenerative diseases influence more than 50 millions of people worldwide (Ehrenberg et al. 2020). The number of patients is increasing as the aging population is growing (Kim et al. 2023; Van Schependom and D’Haeseleer 2023; Wilson et al. 2023). In the effort to develop treatment and preventive methods, numerous studies have attempted to uncover underlying mechanisms of neurodegenerative diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD) and Huntington’s disease (HD), revealing the common risk factors such as tau, α-synuclein, and amyloid β (Aβ) (Hansson 2021; Khan et al. 2023).

The microtubule-associated protein tau (MAPT) gene, which consists of 16 exons on human chromosome 17q21, encodes tau protein organized with four main domains: N-terminal domain, proline-rich region, microtubule-binding repeat region (MTBR) and C-terminal domain (Garcia-Escudero et al. 2017; Barbier et al. 2019; Sallaberry et al. 2021). Alternative splicing-derived six isoforms of tau are determined by exon 2 and 3 insertion in the N-terminal (0N, 1N or 2N) and exon 10 insertion as the second repeat region that leads to three or four repeat regions (3R or 4R) in the MTBR (Lester et al. 2021; Sexton et al. 2022) (Figure 1). Tau 3R and 4R isoforms are normally detected at a 1:1 ratio in the brain. However, the imbalanced ratio may cause tauopathies with cognitive impairment (Espindola et al. 2018). Furthermore, when dysregulated, common post-translational modifications such as phosphorylation, acetylation, glycosylation and methylation can play a role in the acceleration of tau aggregation. Aggregated tau reduces binding affinity to microtubules and leads to neurofibrillary tangles (NFTs) in the neuron and glia, causing the synaptic dysfunction, cognitive impairment, and neuronal loss (Wang and Mandelkow 2016).

Figure 1. Tau aggregation-associated with diverse mechanisms. There are six tau isoforms, and unregulated 3R or 4R isoform promote the tau accumulation, which is grouped into the 3R or 4R tauopathies. Dysregulated post-translational modification of tau involving the phosphorylation, acetylation, glycosylation, and methylation, also caused the aggregation, especially hyperphosphorylation is known as typical tau pathology.

Throughout the last decade, multifaceted methods involving imaging or fluid-based diagnosis, inhibitor drugs, immunization, antisense oligonucleotides (ASOs), and CRISPR systems have been reported for tau-directed diagnostics or therapeutics (Jadhav et al. 2019; Sexton et al. 2022). Several tau-targeting therapies have entered clinical trials phase II. Unfortunately, limitations such as the toxicity, off-target effect, and complications associated with the delivery through blood–brain barrier (BBB) make tauopathies incurable, thus additional therapeutic approaches are required to advance (Zhang et al. 2022).

Conspicuously progressed sequencing technology and genome-wide association studies (GWAS) have guided the comprehensive elucidation of genetic interactions between tau and several disease-specific genes that can be potential candidates in neurodegenerative diseases involving tauopathies (Kim et al. 2021; Lim et al. 2021a, 2021b; Kim et al. 2023). To date, many studies have focused on the relevance between neurodegenerative diseases and individual cells to explain the brain region-specific pathological features through the sequencing analysis, including single-cell transcriptome analysis (Luo et al. 2020; Olah et al. 2020; Park and Jung 2022).

On the other hand, the recent studies are reported the microbiota-gut-brain axis, which is the complex communication of direct or indirect signaling network through the neuronal metabolic, endocrine, and immunity in the brain development and pathology of neurodegenerative diseases (Dash et al. 2022; Damiani et al. 2023). The microbiome has been highlighted to be diagnostic biomarkers and therapeutic objectives in neurodegenerative diseases (Zhu et al. 2021; Chandra et al. 2023). These efforts have revealed the branches of broader mechanism in tauopathies, which may provide promising directions in the practical applications.

In this review, we summarize the interaction between hyperphosphorylation of tau proteins and tau-associated factors in tauopathies and overview the current therapeutic and diagnostic strategies to discuss their advantages. Finally, we discuss the potential of meaningful approaches using the deep learning-based single-cell transcriptome and microbiome.

Tau pathology

Tauopathies are categorized into a broad range of neurodegenerative diseases, possessing a hallmark of insoluble tau fibrils and diverse spatial patterns of deposits. Because native tau itself rarely folds or promotes aggregate formation, it can be detected when the native tau structure is compromised due to multiple causes, including dysregulated phosphorylation (Wang and Mandelkow 2016). Subsequent to these tau abnormalities, tau proteins are accumulated and knotted inside neurons, called NFTs. Tau NFTs are strongly associated with local atrophy and the exacerbation of neurodegenerative manifestations, including neuronal loss, synaptic weakening, cognitive impairment, and eventual clinical decline (Bejanin et al. 2017; Xie et al. 2018; Berron et al. 2021). The spreading and propagation of tau tangles, which characterize tauopathies, have been observed in both human patients and tau-mice through brain imaging, and in the overall regions of the human brain using a predictive model (Fung et al. 2020; Vogel et al. 2020; Buciuc et al. 2023). Tangles have long been considered a proximate cause of pathological functioning in neurodegenerative diseases. However, the exact mechanisms and further relations remain elusive and also debated as insufficient to induce neurodegeneration.

The precise etiologies remain the subject of controversial, however, numerous studies have described tau hyperphosphorylation as an essential clue. Microtubule structural integrity enables neurons to maintain proper morphological dynamics throughout their axons and dendrites (Iwanski and Kapitein 2023). Tau phosphorylation occurs in the context of representative amino acid residues, such as serine (S), tyrosine (Y), or threonine (T), and counts for over 85 potential sites (Xia et al. 2021). The primary phospho-epitope on tau has been identified as 17 distinct sites consisting of serine-proline or threonine-proline, which are suspected to regulate tau and microtubule disassembly due to their interdependency. The initial phosphorylated sites on tau (e.g. tau-181) named ‘master sites’ were demonstrated to enhance distal phosphorylation at phospho-available sites that note a hierarchy among phospho-sites (Stefanoska et al. 2022). When tau is hyperphosphorylated at specific sites, its affinity for microtubules decreases, with the risk of microtubule destabilization consequently increasing. Tau homeostasis is ascribed to the balance of tau-associated kinases and phosphatases, whereby overactivated kinases empower phosphorylation rather than dephosphorylation (Wainaina et al. 2014).

Several classes of kinases and phosphatases can add or remove phosphate groups from tau. Phosphate is also added by tyrosine kinases (e.g. Fyn, LCK, SYK, and ABL1), proline-directed kinases (e.g. GSK3, CDK5, MAPK, and JNK), and non-proline-directed kinases (e.g. PKA, TTBK and MARK) (Hanger et al. 2009). Glycogen synthase kinase 3 (GSK3), which mostly attaches a phosphate group to S/T residues, is a ubiquitously expressed kinase that is particularly enriched in the brain. The regulatory cascade of both isoform GSK3α and β to tau has been examined regarding tau phosphorylation at specific residues and its subsequent physiological effects, such as promotion of neuronal impairment and cognitive dysfunction (Fan et al. 2022; Zhou et al. 2022). Fyn is another tau regulatory-associated kinase that implicates a phosphorylated tyrosine site known as Tyr-18, which has been shown to decrease calcium influx and tau accumulation in synapse (Miyamoto et al. 2017; Briner et al. 2020). The cAMP-dependent protein kinase (PKA) also targets tau to phosphorylate, which results in NFTs formation and calcium dysregulation (Carlyle et al. 2014). The intriguing aspect is that PKA primes tau for subsequent phosphorylation by GSK3 and CDK5 catalytic activity (Liu et al. 2006). Otherwise, tau dephosphorylation is performed by protein phosphatase 2A (PP2A; dominant activity of nearly 71%), PP1, PP5, and PP2B (Noble et al. 2013; Taleski and Sontag 2018). PP2A is a response to diminished tau phosphorylate activity and is considered to be linked to Alzheimer’s disease (Rudrabhatla and Pant 2011). PP2A imposes dephosphorylation at its specific S/T sites on tau (e.g. S214, S262, T205, and T212), which shows different catalytic efficiencies and indirect correlations with other kinases (Kruse et al. 2020).

In addition to hyperphosphorylation, multiple classes of RNAs and RNA-binding proteins (RBPs) can also contribute to tau aggregates formation, accumulation, and mislocalization, and comprise an emerging field of study (Patrick et al. 2017; Kavanagh et al. 2022). Nuclear speckle-related RNA, snRNAs, and snoRNAs are revealed to be enriched in tau aggregates and a colocalization-mediated alteration in soma has been identified (Lester et al. 2021). The long noncoding RNA, NEAT1, has also been investigated to participate in the tau pathway by regulating microtubule stabilization (Zhao et al. 2020). The GTPase-activating protein-binding protein 2 (G3BP2) has recently been demonstrated to inhibit tau inclusion following its direct binding to the MTBR (Wang et al. 2023). With regard to the tau regulatory cascade, multiple factors, such as the residue of phosphorylation, the participating kinases/phosphatases, and other molecules orchestrate tau pathophysiology. However, the mechanisms of pathogenesis and further progress of tau still remain unclear, despite the intensive research, which has been figured out dynamic correlations in tau cell biology.

Current approaches to tauopathies

In recent years, many therapeutic and diagnostic avenues have been suggested to target the tau to regulate the abnormal isoform and expression, aggregation, and several kinases through the four therapeutic concepts; tau immunotherapy, inhibitor drugs, ASOs, CRISPR/Cas9 and tau clearance (Table 1), as well as one diagnosis concept; deep learning-based images (Vega et al. 2021; Sexton et al. 2022) (Figure 2).

Figure 2. Summary of current tau therapies and diagnosis. The graphical diagram summarizes the current tau-targeting therapies and diagnosis methods. Therapeutic concepts are composed the ASOs, CRISR/Cas9, inhibitor drugs, tau immunotherapy, and clearance. Diagnosis concepts are applied to various tau images, including the hyperspectral images, PET, tau-immunostaining, and fluorescence resonance energy transfer (FRET) (Kim et al. 2023) that can be used to train the deep learning for the prediction of tau pathology.

Table 1. Current mechanisms and clinical phases of tau-targeting therapeutic strategies.

Therapeutic strategy	Drug	Mechanism		Clinical phase	Clinical trial identifier	Indication
			Binding site
Tau vaccine	AADvac1	Decrease the total tau or p-tau via their specific binding motif (Plotkin and Cashman 2020; Song et al. 2022)	Tau 294–305	Phase II	NCT02579252	Mild AD
	ACI-35		p-tau 393–408	Phase I/II	NCT04445831	Early AD
Anti-tau antibody	Semorinemab		Tau 2–24	Phase II	NCT03289143 NCT03828747	Mild AD and moderate AD
	Gosuranemab		Tau 15–24	Phase I	NCT02460094	PSP and AD
	Tilavonemab		Tau 25–3	Phase II	NCT02880956 NCT03712787	Early AD
	Bepranemab		p-tau 235–250	Phase II	NCT04867616	MCI or mild AD
	Zagotenemab		Tau 7–9/312–322	Phase II	NCT03518073	Early AD
	JNJ-63733657		pThr217	Phase II	NCT04619420	Early AD
	PNT001		pThr231	Phase I	NCT04096287	Healthy
	LuAF87908		p-tau 386–408	Phase I	NCT04149860	Healthy and AD
	RG7345		pSer422	Phase I	NCT02281786	Healthy
Kinase inhibitor	Saracatinb	Reduce the tau phosphorylation through the Fyn inhibitor		Phase II	NCT02167256	AD
	Tideglusib	Decrease the tau phosphorylation using the GSK3β inhibitor (Islam et al. 2022)		Phase II	NCT01049399 NCT00948259 NCT01350362	PSP and mild to moderate AD
	LM11A	Diminish the tau phosphorylation through the p75^NTR inhibitor (Islam et al. 2022)		Phase II	NCT03069014	Mild to moderate AD
Glycosylation inhibitor	MK-8719	Decrease the amount of tau aggregation using the O-GlcNAcase (OGA) inhibitor (Islam et al. 2022; Zhang et al. 2022)		Phase I	–	PSP
	ASN120290			Phase II	–	PSP
Acetylation inhibitor	Salsalate	Suppress the acetyltransferase p300-induced tau acetylation (Islam et al. 2022; Zhang et al. 2022)		Phase I	NCT02422485 NCT03277573	Mild to moderate AD
Aggregation inhibitor	LMTM	Inhibit the tau aggregation via blocking the interaction of tau-tau repeat domains (Islam et al. 2022)		Phase III	NCT01689246 NCT01626378 NCT03446001	AD, bvFTD, and MCI-AD
	GPSE	Interrupt the tau oligomerization via inhibiting tau fragment peptide (Islam et al. 2022)		Phase II	NCT02033941	AD
Microtubule stabilizer	Epothilone D	Maintain the stability of microtubule with tau protein (Islam et al. 2022; Zhang et al. 2022)		Phase I/II	NCT01492374	Mild AD
	Davunetide			Phase II	NCT01056965	PSP, FTLD, and CBS
Gene therapy	BIIB080	Regulate the MAPT gene through the antisense oligonucleotides (Zhang et al. 2022)		Phase II	NCT03186989	Mild AD
	–	Modulate the MAPT gene through the CRISPR/Cas9 (Zhang et al. 2022)		–	–	–
Tau clearance	USP13	Activate the ubiquitin-proteasome system through the ubiquitin-associated components (Tang et al. 2019)		–	–	–
	AZD2014	Activate the autophagy-lysosome system using the mTOR kinase inhibitors (Tang et al. 2019)		–	–	–
	AZD8055			–	–	–
	OSI-027			–	–	–
	rapamycin			–	–	–

Note: bvFTD: behavioral variant of frontotemporal dementia; CBS: corticobasal syndrome; FTLD: frontotemporal lobar degeneration; GPSE: grape seed polyphenolic extract; LMTM: leucomethylthioninum; MCI-AD: mild cognitive impairment due to AD; PPA: primary progressive aphasia; PSP: progressive supranuclear palsy.

Tau immunotherapy and inhibitor drugs

Tau immunotherapy utilizes the passive or active immunization strategies to decrease the hyperphosphorylation and accumulation of tau (Vaz and Silvestre 2020). The tau vaccine is the active immunotherapy that facilitates to produce endogenous antibody to response the pathogenic tau proteins through several tau-specific epitopes carrying antigens. The AADvac1 and ACI-35 are representative tau vaccine, which only remained clinical research in a decade, and have been entered the phase II trial (Novak et al. 2021; Song et al. 2022). Anti-tau antibodies are the passive immunotherapy that designed to bind the specific tau epitopes and can interact the intracellular tau after endocytosis. Semorinemab, Gosuranemab, Tilavonemab, Bepranemab, Zagotenemab, JNJ-63733657, PNT001, LuAF87908, and RG7345 has been investigated in phase I or II trial up to date. Consequently, both tau vaccine and anti-tau antibodies reduced the total tau or diverse p-tau through their specific binding motif to tau protein that also inhibited the diffusion of extracellular tau. However, they need to understand the exact timing for treatment and therapeutic efficiency (Plotkin and Cashman 2020; Panza and Lozupone 2022).

Inhibitor drugs are designed to target the post-translational modifications of tau to disturb the progression of tauopathies. They have been investigated to confirm the efficiency in clinical trials; The kinase inhibitors are the Fyn inhibitor (Saracatinb), GSK3β inhibitor (Tideglusib), and p75^NTR inhibitor (LM11A) in trial phase II; The glycosylation inhibitors are MK-8719, and ASN120290 in trial phase I or II; The acetylation inhibitor is Salsalate in trial phase I; The aggregation inhibitors are leuco-methylthioninium-bis (LMTM), and grape seed polyphenolic extract (GPSE) in trial phase II or III; The microtubule stabilizers are Epothilone D, and Davunetide in trial phase I or II (Islam et al. 2022).

ASOs and CRISPR/Cas9

The ASOs are the synthesized short single-stranded oligodeoxynucleotides, which are designed to pair complementary to target mRNA, and this complex allows to degrade by the endonuclease RNase H1, or interrupt the interaction with RBPs involving the ribosomal subunits via blocking the AUG start site, thereby inhibit translation to subsequent target protein (Rinaldi and Wood 2018). Additionally, ASOs can alter splicing during RNA processing to bind the splice site or factors, thus it manipulates the target exon to be skipped or retention. Therefore, it has been developed to target MAPT gene to regulate the tau expression or isoform. However, ASOs have a common considerable impediment for the clinical application in brain disorders, which are their poor permeability across the BBB because of negative charge and size. Thereby, they are used to deliver via intrathecal and intraventricular administration (Chakravarthy et al. 2020; Damianich et al. 2021; Easton et al. 2022).

CRISPR/Cas9 is a remarkable gene-editing technology from the prokaryotic adaptive immune system. It is composed of single guide RNA (sgRNA) and Cas9 protein; the sgRNA guides Cas9 to bind target sequences, and the Cas9 works as the CRISPR-associated endonuclease to cleave the DNA to edit or repair the sequence (Konermann et al. 2018; Song and Koo 2021). It has been investigated to apply the AD therapeutics and developed to modify the construct to cross the BBB, such as adeno-associated virus and nanocomplex (Park et al. 2019; Bhardwaj et al. 2022).

Tau clearance

Eliminating aggregated tau is an expeditious method to suppress the progression in tauopathies. The various mechanisms of tau clearance have been reported in the degradation of protein pathway; ubiquitin-proteasome system and autophagy-lysosome system in intracellular tau, as well as microglia and astrocytes uptakes in extracellular tau (Tang et al. 2019). The ubiquitin-proteasome system components have been investigated for therapies, including the ubiquitin-specific peptidase 12 (USP12), USP13, USP14, USP30, E3 enzymes, and ubiquitin C-terminal hydrolase L1 (UCHL1). USP13 knockdown reduces the p-tau level via activating tau clearance in AD (Liu et al. 2019; Schmidt et al. 2021; Kim et al. 2023). The autophagy-lysosome system targeting the mammalian targets of rapamycin (mTOR) kinase shows reduced the autophagy signal, in turn, the mTOR kinase inhibitors, such as AZD2014, AZD8055, OSI-027 and rapamycin, alleviated aggregation of tau (Salama et al. 2019; Silva et al. 2020). The endocytosis has been reported to activate glial cells cascades through tau tangles attachment to microglia surface factors, the CX3C motif chemokine receptor 1 (CX3CR1) in microglia, and integrins on astrocytes (Perea et al. 2020; Seitkazina et al. 2022).

Diagnosis with tau images

Diagnostic methods for tau using the brain tissue have high accuracy, but collecting tissue from living organisms is not suitable for routine or early diagnosis. Imaging and fluid-based tau biomarkers have been used to diagnostic strategies in patients with tauopathies (Lee et al. 2019; Bergauer et al. 2022). Deep learning-based imaging analysis has been raised a remarkable approach that trained the deep learning using the diverse images of tau to understand better the tau aggregation and localization across tauopathies in diagnosis (Ushizima et al. 2022). Current studies were exhibited the applicability the diagnosis in tauopathies using the deep learning-based positron emission tomography (PET) hyperspectral imaging, and tau-immunostaining (Jo et al. 2020; Du et al. 2022; Koga et al. 2022). In addition, recent studies have been developed the fluorescence resonance energy transfer (FRET)-based biosensor cell lines, which are stably expressed the tau protein with cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP). The aggregated tau has been excited by the YFP signal by the emission of CFP as the FRET signal. Therefore, the analysis of FRET-based tau aggregation may be a diagnostic avenue (Kim et al. 2023; Maina et al. 2023).

Novel applicability to tau therapy

Deep learning-based single-cell transcriptome

Intriguingly, several studies were demonstrated the distinct tau lesions in the brain and different cell types of accumulated tau within AD, corticobasal degeneration (CBD), and progressive supranuclear palsy (PSP). These showed the cell-to-cell interaction and spread over several brain compartment of tau (Narasimhan et al. 2020; Mate de Gerando et al. 2021). Several genetic signatures have revealed that susceptibility against abnormal tau depends on regional and cellular vulnerability. For example, Bcl2-associated athanogene 3 (BAG3) gene known to be involved in tau homeostasis was a preferential expression in excitatory neuron compare to inhibitory neuron (Fu et al. 2019). These evidences are suggested the individualized properties of deposited tau that may cause distinct effect on cell types or brain regions (Figure 3).

Figure 3. Potency of deep learning-based single-cell transcriptome and microbiome in tauopathies. Schematic diagram showed the procedures of deep learning-based single-cell transcriptome and microbiota-gut-brain axis interaction in brain to enhance the comprehensive understand of tau pathology. These approaches suggested the applicability of tauopathies therapy that the cell type-specific properties and genetic interaction of tau, as well as crosstalk between brain and metabolites of gut microbiota, such as SCFAs, AAM, and TMAO.

Single-cell transcriptome analysis is a more specific technology than the traditional whole genome sequencing from bulk sampling, especially in the gene expression, thus it provides a crucial understanding of the cell type, state or unique distribution. Currently, various evidences have been reported that the cell type-specific differences of pathological gene expression may be associated with clinical severity in neurodegenerative diseases (Han and Luo 2023). Using the single-cell transcriptome analysis from the prefrontal cortex of 48 individuals with diverse conditioned AD, 1031 differentially expressed genes (DEGs) were separated from the excitatory, inhibitory neurons, astrocytes, oligodendrocytes, oligodendrocyte precursor cells, and microglia; high percentages of DEGs are 75% downregulated in excitatory neurons and 95% in inhibitory neurons, while 53–63% of DEGs are upregulated in astrocytes, oligodendrocytes, and microglia. Moreover, the myelination-related genes expression was showed mainly the cell type-specific DEGs. For example, the LINGO1 was commonly altered the expression in almost all cell types, while BEX1, CNTNAP2, ERBIN, NEGR1, and PRNP were only changed in neurons, as well as CRTAB in only glia. Similarly, APOE as the AD risk marker, was significantly increased in microglia, while decreased in astrocytes. Besides, the cell subpopulation of specific markers expression also changed in AD (Mathys et al. 2019; Luo et al. 2020). In 2022, novel subtypes of early stage AD-associated microglia (EADAM), and late stage AD-associated microglia (LADAM) were identified from the mice with Aβ and tau, as well as several siglec genes were selectively regulated in the EADAM or LADMA that related to functions of immune systems (Kim et al. 2022). These cell type-specific DEGs or disease-specific cell population was identified not only in the neuronal cells but also in the immune cells in AD. Using the 36,849 peripheral blood mononuclear cells (PBMCs) from AD patients, different subpopulation of five immune cells was reported that CD4+ T cells were upregulated, while CD8+ T cells, natural killer cells, and monocytes-macrophages cells were downregulated. In the B cells, there were not specific changes. There were also showed several key of DEGs, but need to further studies to explain the exact mechanisms (Xu and Jia 2021). While other study was showed the decrease of B cell in the PBMCs of both early and late stage AD patients. In the B cells of early and late stages AD, KIR3DL2, QPCT, PPP2R2B genes were increased and FRAT2, WWC3, SPG20 were decreased commonly. These DEGs in the late stage of AD were more significantly changed the expression than the early stage of AD in the B cells (Xiong et al. 2021).

Although they found the specific DEGs, there were limitations to define their functions because of the lack of cerebrospinal fluid (CSF) or brain samples from diverse stages, inaccuracy due to background noise during the single-cell isolation, and unintegrated massive sequencing data. To improve the problems, the deep learning-based approach has been suggested for the single-cell transcriptome analysis (Yan et al. 2021). Deep learning belonging the artificial intelligence (AI) has deep neural networks composed of multiple layers that lead to learn from large datasets as the human brain, thereby it can perform more complicated functions than machine learning (Yang et al. 2023). Application of deep learning in single-cell transcriptome analysis has operated to obtain more accurate and enhanced information from the massive complex, noisy, and heterogeneous single-cell transcriptome data. Then, it allows to overcome the crucial challenges of single-cell transcriptome analysis, including the upstream process as the production of quality samples, standardized diverse noisy and low quality cells, as well as downstream process as the analysis of biological features with cell, gene, and pathway from a large amount of data. However, since its relatively lack of datasets, it is insufficient to demonstrate the correlation within the cell or disease-specific properties of tau and biological mechanisms as described as cell-tocell connection and DEGs, thus further studies are still required (Brendel et al. 2022; Erfanian et al. 2023).

Microbiome and tau pathology

Examining the microbiota-gut-brain axis has been suggested as a valuable insight regarding the role of microbiota in numerous physiological regulations in the brain, including neurodevelopment, microglial immune function, myelination, and other fundamental activities (Cryan et al. 2019) (Figure 3). Commensal gut microbes are known to perform several physiological functions in the host and comprise a highly diverse and complex community (Thursby and Juge 2017). The uniqueness of gut microbiomes due to their compositional pattern has been demonstrated to reflect a condition of human. In this context, comparisons among healthy, aging, pathogenic and diseased people throughout the entire body can reveal distinct in regulatory effects on microbiota and microbiota-derivatives dependent manner (Wilmanski et al. 2021; Hou et al. 2022).

Several distinctive microbial taxa have been known to exist comparatively more abundant in AD patients’ stool than in normal individuals that mostly contain thousands of species. Such taxa include Escherichia/Shigella and Eubacterium rectale, which potentially lead to progressively more detrimental AD phenomena (Cattaneo et al. 2017). A recent study has shown the relation between the AD preclinical signature tau by PET and the gut microbiome, by examining the altered phosphorylation of tau-181 in CSF (Ferreiro et al. 2023). Distortions in the controlling system of microbiota throughout the gastrointestinal tract and brain axis result from the production of unbeneficial metabolites for the host. Certain bacterial metabolites can permeate a loosened BBB, while other bacterial signals can be conveyed through the vagus nerve via the enteric nervous system (Braniste et al. 2014; Bonaz et al. 2018). The vagus nerve provides a direct communication route with microbiota to the brain facilitated by neuroactive molecule-mediated regulation, which may stimulate cellular signaling in the central nervous system. The vagus nerve recognizes microbiota-derivatives through its afferent fibers, and the delivered information can perpetuate dysfunctional conditions, such as the production and secretion of microglial cytokine as a consequence (Bonaz et al. 2018; Han et al. 2022).

Metabolites of gut microbiota have been exerted to be involved in the brain via their absorbance through the gut-brain axis cognate receptors, by which short-chain fatty acids (SCFAs), amino acid metabolites (AAM), trimethylamine N-Oxide (TMAO), and other metabolites can manipulate the crosstalk with the brain system (Swer et al. 2023) (Table 2). SCFAs representing butyric acid, acetic acid, and propionic acid, are small organic compounds produced by microbiota in the gastrointestinal environment, which confer local protection. In addition, SCFAs can modulate enterochromaffin cells to increase 5-HT synthesis cascades and may cross the BBB by circulation in the bloodstream with the support of monocarboxylate transporters (MCTs). Following the SCFAs infiltration into the brain, G protein-coupled receptors (FFAR2 and FFAR3) activation and histone deacetylase (HDAC) inhibition have been reported as their major mechanisms (Guo et al. 2022). SCFAs have been associated with the underlying mechanisms in several brain and neuronal disorders, and exacerbating amyloid beta deposition (Colombo et al. 2021). More recent research has highlighted an interdependency between SCFAs and tau-mediated disorder through ApoE isoform carrying tau transgenic mice model (Seo et al. 2023). AAM is amino acids from gut bacterial fermentation, are converted by metabolizing phenylalanine and tryptophan, which transduce signals related to host physiological regulation. These two essential amino acids are processed by the host-microbial metabolic pathway; for example, phenylalanine is a precursor of several neurotransmitters, such as dopamine, whereas tryptophan is a precursor of serotonin and indole (Dodd et al. 2017; Liu et al. 2020). Tryptamine from tryptophan can easily cross BBB and has been detected to induce NFTs in human cells and mice (Paley et al. 2007). The TMAO has also been reported to have positively correlated with tau phosphorylation in the CSF, but not in NFTs (Vogt et al. 2018).

Table 2. Microbiota-derived metabolites and relations to tauopathies and brain.

	Substrate (Neurotransmitter)	Effects on tau regulation	Effect on brain mechanisms
			BBB cross	Physiological function
SCFA	Butyric acid	Ameliorate tau hyperphosphorylation in AD-mimic mouse model (Cao et al. 2018) Reduce tau phosphorylation in enteric neuron (Chapelet et al. 2023)	Positive	Inhibit histone deacetylase (HDAC) and activate GPCR (Bourassa et al. 2016) Restore synaptic plasticity (Jiang et al. 2021) Modulate microglia activity and cytokine secretion (Matt et al. 2018)
	Acetic acid		Positive	Regulate hippocampal synaptic plasticity and cognitive function (Zheng et al. 2021) Modulate microglia phagocytosis via metabolic sufficiency (Erny et al. 2021)
	Propionic acid	Potential to manipulate tau phosphorylation in ApoE isoform-associated tauopathy (Seo et al. 2023)	Positive	Protect BBB integrity against LPS-mediated perturbation and oxidative stress (Hoyles et al. 2018) Increase risk of AD through its excessive concentration (Killingsworth et al. 2020)
	Phenylalanine		Positive
	Dopamine	Relation to tau phosphorylation that deposited in the intestinal mucosa (Tian et al. 2023)	Negative	Regulate dopamine transporter and receptor-mediated pathway on synaptic cleft (Hamamah et al. 2022)
	Tryptophan		Positive
	Serotonin		Negative	Relation to children with autism spectrum disorder (Xiao et al. 2021) Weaken chronic unforeseeable mild stress stimulation-mediated depress and strengthen neurogenesis (Ma et al. 2023)
	Indole	Inhibit tau phosphorylation via AhR pathway to form NLRP3 inflammasome in APP/PS1 mouse model (Sun et al. 2022)	Positive	Promote neurogenesis via AhR signaling in adult mouse hippocampus (Wei et al. 2021) Affect depressive emotion and locomotion in rat model (Jaglin et al. 2018)
	Tryptamine	Induce formation of PHFs and NFTs (Paley et al. 2007)	Positive	Activate Aryl hydrocarbon receptor (AhR) signaling (Salminen 2023) Inhibit tryptophanyl-tRNA synthetase (TrpRS) that implicates neuronal dysfuction (Paley 2011)
TMAO		Positive correlation with tau phosphorylation in cerebrospinal fluid (Vogt et al. 2018) Induce tau phosphorylation at S202 and T205 of midbrain organoid model (Lee et al. 2022)	Positive	Influence ageing-related neurodegenerative phenotype of midbrain organoid model (Lee et al. 2022) Disturb synaptic plasticity via PERK cascade (Govindarajulu et al. 2020)

Note: GPCR: G protein-coupled receptors; LPS: lipopolysaccharide; PERK: protein kinase R-like endoplasmic reticulum kinase; PHFs: paired helical filaments.

Regarding the microbiota-gut-brain dependency in disease research, methods for visceral bacterial composition changes through fecal microbiota transplantation (FMT) have been used (Matheson and Holsinger 2023). This method has mostly been employed to transit the healthy microbiota into a pathological environment for non-healthy recipients. Multiple studies have supported the positive results deriving from this method, for example, mouse models presenting with amyloid and NFTs display a distinct attenuation of NFTs formation following FMT, together with improvement in cognitive function (Kim et al. 2020). Similarly, APPswe/PSEN1dE9 mice exhibit alleviated tau phosphorylation after FMT (Sun et al. 2019). Moreover, FMT from young donor mice was found to reverse age-associated exceedingly harmful effects in the nervous system of recipient mice by modulating microglial activity, inflammatory cytokines, and metabolic pathways (Jing et al. 2021; Parker et al. 2022). Meanwhile, systemic dysbiosis may occur via an indeterminate mechanism-mediated cascade, resulting in adverse outcomes when a detailed understanding of the microbiota community composition in different contexts is lacking with substantial strategies (Marrs and Walter 2021).

Although the results of the microbiota transition are controversial, the unique composition of the microbiome depending on host physiology has been verified by metagenomic sequencing and profiling (Zhu et al. 2021). Since the axis focuses on bidirectional communication-derived influence, coupling longitudinal genetic analysis to other molecular disciplinary omics is expected to further delineate the interaction machinery of the microbiota-gut-brain axis (Cryan and Mazmanian 2022; Diener et al. 2022). Relativeness with several neurological disorders including psychiatric, developmental, and degenerative has also been accumulating through emerging evidence (Warner 2019; Socala et al. 2021; Cutuli et al. 2022). Here, functional characterization using next-generation sequencing (NGS) was utilized in order to obtain high-throughput data for extensive analysis (Wensel et al. 2022). Although certain limitations, such as bias and missing information of the unique taxon reference sequence database remain. Advanced sequencing technologies are continuously improving the access to contextual information, by which spatial resolution of the microbiome, metagenomics at the single-cell scale, and gut mucosal-specific IgA antibody-based shotgun sequencing (Yen and Johnson 2021).

One of the ultimate goals of interconnectivity studies is to reveal underlying causality and address the clinical treatment of brain disorders. Across the highly complex microbiota ecosystem, blood-secreted metabolites, their particular compositions, and molecular intervention could contribute as signatures for diagnosis and further therapies (Zhan et al. 2023), and phenylacetylglutamine (PAGln) has been found as a cardiovascular disease marker (Huynh 2020; Nemet et al. 2020). Numerous questions remain unanswered, especially in terms of the microbiota-dependent signal transmission for tau phosphorylation and NFTs manifestation. A hitherto research on the largely unknown interplay between tauopathies and the microbiome is still in its infancy and requires further insights.

Conclusion

Through the continuous research regarding on tauopathies, numerous methods have been made and attempted to evolve the tau-targeted disease-modifying therapy as a promising therapeutics. The deep learning-based single-cell transcriptome and microbiome have been suggested the novel strategies that showed the potent cell type-specific properties and genetic cooperation with tau in tauopathies, as well as the interplay between the microbiota and aberrant tau. Nevertheless, these approaches are still in the stage of infancy and are not enough to explain the circumstantial connection with tau pathology. In addition, abnormal tau-induced DEGs, unique cell populations or microbiota in tauopathies need to demonstrate the exact pathway what the main functions are among the protective response or pathogenic response. Moreover, longitudinal research are also required to discriminate the properties of pathological tau that depend on the stages of progression in diseases. Taken together, while these approaches need further comprehensive understanding, it is carefully suggested the strategic avenues that may provide meaningful criteria for tauopathies.

Declaration of interest

The authors report no declarations of interest.

Acknowledgements

Conceptualization, J. Y. Joo; data curation, M. Kang; writing – original draft preparation, S. H. Kim and S. Yang; writing – review and editing, S. Yang and E. Yang; visualization, S. H. Kim and E. Yang; supervision, project administration and funding acquisition, J. Y. Joo. All authors have read and agreed to the published version of the manuscript. We thank Jeehye Jung for assistance of graphical diagrams. The graphical figures were made with biorender.com.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This works was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) [grant number RS-2023-00217123] and the Ministry of Science and ICT [grant number 2022R1A2C1002984].