Microglia: Friend and foe in tauopathy

Abstract

Aggregation of misfolded microtubule associated protein tau into abnormal intracellular inclusions defines a class of neurodegenerative diseases known as tauopathies. The consistent spatiotemporal progression of tau pathology in Alzheimer’s disease (AD) led to the hypothesis that tau aggregates spread in the brain via bioactive tau “seeds” underlying advancing disease course. Recent studies implicate microglia, the resident immune cells of the central nervous system, in both negative and positive regulation of tau pathology. Polymorphisms in genes that alter microglial function are associated with the development of AD and other tauopathies. Experimental manipulation of microglia function can alter tau pathology and microglia-mediated neuroinflammatory cascades can exacerbate tau pathology. Microglia also exert protective functions by mitigating tau spread: microglia internalize tau seeds and have the capacity to degrade them. However, when microglia fail to degrade these tau seeds there are deleterious consequences, including secretion of exosomes containing tau that can spread to neurons. This review explores the intersection of microglia and tau from the perspective of neuropathology, neuroimaging, genetics, transcriptomics, and molecular biology. As tau-targeted therapies such as anti-tau antibodies advance through clinical trials, it is critical to understand the interaction between tau and microglia.

1. Introduction

Aggregation of misfolded microtubule associated protein tau (MAPT, hereafter “tau”) into abnormal cellular inclusions, both neuronal and glial, defines a molecular class of neurodegenerative disease known as tauopathies, including Alzheimer’s disease (AD), frontotemporal lobar degeneration tau (FTLD-tau), corticobasal degeneration (CBD), and progressive supranuclear palsy (PSP), among others. Tau aggregates are characterized by stable morphologies (Holmes et al., 2014), differential cell tropism (Guo et al., 2016), and distinct tau proteoforms (Dujardin et al., 2020) which vary across different tauopathies. For example, the neurofibrillary tangle (NFT), an intraneuronal aggregate of misfolded tau with a flame or globose shape, is a prominent hallmark of AD and PSP (Kovacs, 2015). It is well established that aberrant post-translational modifications (predominantly hyperphosphorylation, but also acetylation, methylation, and other modifications) and/or mutations (e.g., P301S) cause tau to misfold and form aggregates. Misfolded monomeric tau aggregates to form dimers, trimers, multimers, and eventually large filamentous or fibrillar aggregates (e. g. NFTs) which accumulate within the cell. The mechanistic relationship between tau inclusions and the clinical presentation in tauopathies is still unclear. Nevertheless, there is a strong correlative relationship between the development of tau pathology in specific neuroanatomical regions and the functional deficits observed in tauopathy patients.

The phenomenon that tau lesions affect brain regions sequentially in a stereotypic progression suggests that affected brain areas are selectively vulnerable to tau pathology, or that tau pathology spreads from one brain regions to next (Mudher et al., 2017). Numerous studies support the hypothesis that tau spreads in a prion-like manner, often referred to as “tau seeding and spreading” (Clavaguera et al., 2013, 2009; Falcon et al., 2015; Frost et al., 2009; Iba et al., 2015; Kaufman et al., 2016; Wu et al., 2016). “Seeding” refers to the ability of certain conformers or strains of misfolded tau to enter recipient cells and trigger multimerization and conformational conversion of native intracellular tau, and “spreading” intercellular propagation of these tau seeds. In vitro studies have confirmed that tau can travel transynaptically from neuron to neuron (Takeda et al., 2015). In vivo mouse models also show that tau aggregates injected into the mouse brain can seed tau aggregation in connected regions (Iba et al., 2013); viral gene transfer of tau in mice also leads to rapid spreading of tau to connected neurons(Asai et al., 2015; Wegmann et al., 2015). Furthermore, expression of a human tau transgene in entorhinal cortex neurons of the mouse leads to propagation of tau aggregates into the synaptically connected hippocampus (de Calignon et al., 2012). Studies with human postmortem brain tissue show that tau seeds precede formation of NFTs along the Braak tau pathway, also suggesting that tau propagates in humans (DeVos et al., 2018a). Neuronal tau release and uptake is considered the putative mechanism of tau propagation (de Calignon et al., 2012; Liu et al., 2012; Takeda et al., 2015; Wegmann et al., 2015) and released tau in the extracellular space can be found as both a free protein and in extracellular vesicles (EVs) (Brunello et al., 2020). Notably, glia are also involved in tau seeding and spreading (Asai et al., 2015; Ferrer et al., 2019; Hopp et al., 2018; Wang and Ye, 2020; Zareba-Paslawska et al., 2021).

Microglia are the resident immune cells of the central nervous system and are derived from early myeloid progenitors that migrate to the brain during development (Ginhoux et al., 2010). Microglia constantly survey the microenvironment, and rapidly respond to injury (Nimmerjahn et al., 2005) as well as changes in neuronal activity (Liu et al., 2019; Umpierre et al., 2020). Microglia express numerous transmembrane receptors and ion channels to sense and respond to microenvironment alterations (Kettenmann et al., 2011). Microglia display a range of phenotypes or functional states which are defined by their transcriptomic signature, via expression of specific proteins, or sometimes a distinct cellular morphology. The microglial homeostatic phenotype is characterized by immune-surveillance and tightly regulated inflammatory functions which allow microglia to maintain tissue homeostasis without exacerbating the damage themselves. A progressive transition away from a microglia homeostatic phenotype and toward a dysregulated reactive phenotype correlates with the progression of cognitive deficits during tauopathies (Arriagada et al., 1992; Gerhard et al., 2006, 2004; Pascoal et al., 2021). Genetic polymorphisms that modify microglia phenotype and function are common in AD (Karch and Goate, 2015) and may represent a shared etiology amongst tauopathies. Microglia may contribute to disease progression by secretion of inflammatory factors that directly and indirectly promote neurodegeneration and protein aggregation (Cameron and Landreth, 2010). However, emerging literature suggests that microglia senescence or dampened responsiveness rather than an over-reactive response could be a contributing factor to neurodegeneration during tauopathy (Angelova and Brown, 2019). Surprisingly, microglia can also directly contribute to the progression of tau pathology by transmitting aggregate-forming tau seeds intercellularly (Asai et al., 2015; Clayton et al., 2021; Hopp et al., 2018). Conversely, microglia can also reduce tau seeding activity, which would suggest that microglia may potentially be harnessed therapeutically to restrict tau spread (Andersson et al., 2019; Hopp et al., 2018; Luo et al., 2015; Majerova et al., 2014). Cumulatively, these data establish microglia as both friend and foe during tauopathy, capable of restricting or promoting tau spread. With this complex duality in mind, the overall goal of this review is to describe how interactions between microglia and tau aggregates in AD and other tauopathies contribute to disease progression, and to demonstrate how the pivotal bidirectional interaction between microglia and tau facilitates the pathogenesis, progression, and possibly even treatment of tauopathy.

2. Progression of tau pathology is related to microglia phenotype

Tauopathies are broadly divided as either “primary”, where tau aggregates are the predominant neuropathological lesion (example: PSP), or “secondary”, where tau aggregates are observed as a co-pathology alongside another protein aggregate (example: AD) (Irwin, 2016). The former is often grouped together using the umbrella term FTLD-tau. There are many other tauopathies that are discussed in greater detail elsewhere (Kovacs, 2015; Orr et al., 2017). Typically, primary tauopathies result in behavioral and personality disturbances, cognitive impairment and/or extrapyramidal motor dysfunction. PSP and CBD often manifest as atypical Parkinsonian disorders. The progression of symptomatology correlates with the neuroanatomical distribution and burden of tau pathology which occurs in stereotypical spreading patterns, operationally defined through postmortem histopathology staging systems. For AD, Braak NFT staging classifies the progression of tau deposits from the entorhinal cortex (Braak stage I/II) to limbic regions (Braak stage III/IV) to neocortical areas (Braak stage V/VI) (Braak and Braak, 1991). For PSP, a proposed staging schema suggests six stages in which tau pathology progresses from the globus pallidus, subthalamic nuclei, and/or striatum to the frontal cortex, cerebellum, and occipital lobe while also considering whether the tau inclusions are localized to neurons, oligodendrocytes, or astrocytes (Kovacs et al., 2020). For CBD, a staging system does not exist, although generally pathology initially affects the dorsolateral prefrontal cortex and basal ganglia circuits, progressing to posterior regions later in the disease (Constantinides et al., 2019). In AD, tau-immunoreactive NFTs but not Aβ-positive plaques, positively correlate with dementia severity (Arriagada et al., 1992; Bejanin et al., 2017; Schöll et al., 2016). Dementia severity and cognitive domain specific impairment also associate with Braak NFT stage, especially in middle-to-late stage transition from medial temporal and limbic brain regions to diffuse neocortical distribution (Nelson et al., 2012). Other tauopathies, such as PSP, similarly show negative correlations between global cognitive and executive functioning with total tau burden (Koga et al., 2017). Alternatively, the NFTs and other tau aggregates themselves may not be the direct cause of neurodegeneration and cognitive decline, and some have hypothesized that NFTs could instead be a sign of protective sequestering of toxic tau (Bretteville and Planel, 2008). The species of tau responsible for neurodegeneration is still largely debated, but there is growing consensus that soluble, intermediate tau multimers (often termed “oligomers”) drive neurotoxicity (Flach et al., 2012; Fox et al., 2011; Frost et al., 2014; Gerson et al., 2014; Lasagna-Reeves et al., 2011; Niewiadomska et al., 2021; Sun et al., 2018). Overall, these data demonstrate that tau burden is associated with functional decline and generally proceeds through brain regions sequentially.

The stereotypical anatomical progression of different tauopathies lead to two competing hypotheses: (1) affected brain regions and cell subpopulations are selectively vulnerable to tau pathology and (2) that tau pathology spreads from one interconnected brain regions to another (Mudher et al., 2017). It is likely that the progression of tau pathology results from a combination of both selective vulnerability and tau seeding and spreading at different disease stages. A recent study showed that tau spreading drives early tau propagation in early-stage AD but that local replication drives late tau progression (Meisl et al., 2021). The specific factors involved in either selective regional vulnerability or tau spreading are not fully characterized, although several lines of data point to microglia as potential cellular mediators of vulnerability to and spread of tau pathology. Histopathological and positron emission tomography (PET) neuroimaging studies demonstrate that regional changes in microglia phenotype correlate with cognitive decline and/or tau pathology in AD and other tauopathies (Cherry et al., 2016; Gerhard et al., 2006, 2004; Minett et al., 2016; Parachikova et al., 2007; Pascoal et al., 2021; Perez-Nievas et al., 2013; Sanchez-Mejias et al., 2016), while RNA sequencing studies define specific molecular changes in microglia phenotype in humans as well as mouse models of tauopathy (Rexach et al., 2020; H. Wang et al., 2018).

2.1. Histopathological characterization of microglia phenotype in tauopathy

Several histopathological microglia markers are associated with the progression of tau pathology in human patients. Minett and colleagues found that higher levels of CD68 (a marker of phagocytic microglia), macrophage scavenger receptor, and Fcγ receptor I (FcγR, also known as CD64) and lower levels of ionized calcium binding adaptor molecule 1 (Iba1) were associated with worsened cognition as measured by the mini mental state exam (MMSE) (Minett et al., 2016), while Parachikova and colleagues found that higher levels of major histocompatibility complex class II (MHC-II) was associated with worsened cognition by the MMSE (Parachikova et al., 2007). Similarly, Cherry and colleagues found that CD68 + cell density could predict diagnosis of dementia in cases with repetitive head trauma (Cherry et al., 2016). Perez-Nievas and colleagues examined non-demented individuals without probable AD pathology, non-demented individuals with intermediate and high probably AD pathology (“mismatch” cases), and demented individuals with probable AD pathology to identify factors that could provide resilience to cognitive deficits caused by accumulation of amyloid β (Aβ) plaques and tau tangles. They found that mismatch cases had significantly fewer CD68-positive microglia in addition to lower levels of hyperphosphorylated tau in synapses (Perez-Nievas et al., 2013). These data are in line with data from Parachikova and colleagues showing that mismatch cases in their dataset also have lower levels of MHC-II-positive microglia compared to AD cases with cognitive deficits (Parachikova et al., 2007), since both CD68 and MHC-II are traditionally considered markers of pro-inflammatory microglia activation. Histopathological studies that focused on microglial morphological characterization rather than expression of inflammatory markers yielded very different results (Sanchez-Mejias et al., 2016; Streit et al., 2009, 2014). Indeed, these authors found very few microglia exhibiting the hypertrophic morphology characteristic of activated microglia, and instead an abundance of dystrophic microglia with fragmented cytoplasm (cytorrhexis), spheroids, and a reduced area coverage (Sanchez-Mejias et al., 2016; Streit et al., 2009, 2014). These dystrophic microglia are thought to be hypofunctional and possibly senescent rather than activated (Sanchez-Mejias et al., 2016; Streit et al., 2009, 2014). Interestingly, one of these groups observed that dystrophic microglia appear in brain regions prior to the spread of NFTs based on the stereotypical anatomical progression of AD (Streit et al., 2009). As the authors point out, this would suggest that loss of microglial protective function in a brain region renders it more susceptible to tau pathology (Streit et al., 2009). Though these data seem to present a paradoxical role of microglia in neurodegenerative disease, it is likely that each story presents a different facet of a multidimensional cell. Indeed, there is a growing consensus that both pro-inflammatory and hypofunctional microglia contribute to tau pathology and neurodegeneration in different ways (Angelova and Brown, 2019; Streit et al., 2014).

2.2. Transcriptomic characterization of microglia phenotype in tauopathy

The transcriptomic signature of microglia during neurodegenerative diseases is reviewed elsewhere (Deczkowska et al., 2018). Briefly, microglia display a unique homeostatic transcriptomic signature (Butovsky et al., 2014) that transitions to a “primed” or “neurodegenerative” signature during aging, AD, tauopathy, and other neurodegenerative processes such as amyotrophic lateral sclerosis referred to as “disease associated microglia” (DAM) or “neurodegenerative microglia” (MGnD) (Butovsky et al., 2015; Friedman et al., 2018; Holtman et al., 2015; Keren-Shaul et al., 2017; Krasemann et al., 2017); notably, this microglia phenotype is unique from that which is induced by an acute inflammatory insult such as lipopolysaccharide (LPS) (Holtman et al., 2015). The homeostatic microglial transcriptomic signature is characterized by minimal expression of pro-inflammatory cytokines and high expression of immune surveillance and clearance pathways (e. g. P2RY12 and growth-arrest specific gene 6 (Gas6), respectively) (Butovsky et al., 2014; Schuh et al., 2014). The low expression of pro-inflammatory cytokines is a sign that immune activation is resolved following clearance of the immunological insult (e. g. cell debris) (Hanisch and Kettenmann, 2007). Contrastingly, MGnDs/DAMs down-regulate immune checkpoint and “homeostatic” genes (e. g. CX3CR1 and P2RY12) and upregulate stress-related, inflammatory, phagocytic, and lipid homeostasis genes (Keren-Shaul et al., 2017; Krasemann et al., 2017)including genes that are known AD risk factors (Efthymiou and Goate, 2017; Pimenova et al., 2018). Most notably, the TREM2-APOE pathway serves as a key regulator in DAM phenotype conversion (Krasemann et al., 2017). Transcriptomic studies like those by Keren-Shaul et. al. and Krasemann et. al., pave the way for research aimed at understanding how diverse microglial subtypes interact with tau pathology.

RNA sequencing reveals a complex evolution of changing microglia phenotypes throughout tauopathy pathological stages that is reflective of changes observed in histopathological studies. Wang and colleagues recently found that rTg4510 P301L mice display little phosphorylated tau at 2 months of age and had the same number of Iba1 + microglia as WT animals at 2 months of age (H. Wang et al., 2018). In a principle component analysis of the transcriptomic data, microglia from the 2-month old rTg4510s clustered closely with WT microglia indicating that their phenotypes are undiscernible at this stage (H. Wang et al., 2018) during which these models exhibit hyperphosphorylated tau but not NFTs (Santacruz et al., 2005). A clear microglial phenotypic difference emerged at 4 months of age (H. Wang et al., 2018) coinciding with the onset of NFT pathology (Santacruz et al., 2005). They observed early changes in genes related to tumor necrosis factor (TNF) signaling, interleukin (IL)-1α and IL-1β, and colony stimulating factor 1 (CSF1) that remained elevated at all timepoints, as well as changes in AD risk genes, phagocytosis genes, complement components, scavenger receptors, and microglia polarization genes at different timepoints (H. Wang et al., 2018). Rexach and colleagues used this signature to bioinformatically identify two competing tauopathy microglia expression modules: immune activation and immune suppression (Rexach et al., 2020); these modules are also present in human AD, FTD, and PSP transcriptomic datasets. Rexach and colleagues posit that immune activation begins early during tauopathy in microglia via extracellular detection of damage signals via toll-like receptor (TLR) and NLRP3 inflammasome signaling, while immune suppression is activated later in microglia via type 1 interferon (IFN) signaling (e.g., IFN-β) with subsequent type 2 IFN response (e.g., IFN-γ) (Rexach et al., 2020). Thus, the two counteracting phenotypes cause a state of immune dysregulation where the normal homeostatic clearance functions of microglia are increasingly suppressed. Overall, these findings support observations from histopathological microglia phenotyping by confirming that there are disease-stage specific changes in microglia, progressing from pro-inflammatory activation of microglia to suppression of microglia function.

Curiously, the immune suppression microglia phenotype is characterized by upregulation of viral response genes (Rexach et al., 2020). As noted by Rexach and colleagues, studies show that tau induces transposable element expression through chromatin relaxation and PIWI-interacting RNA depletion (Guo et al., 2018; Sun et al., 2018). Endogenous retroviral double stranded RNA (dsRNA), a product of transposable element transcription, is known to drive a type of IFN response (Chuong et al., 2016). So, Rexach and colleagues hypothesize that dsRNA serves as the upstream signal for the immune suppression phenotype (Rexach et al., 2020). Considering extracellular tau itself is thought to activate microglia (Pampuscenko et al., 2020; Stancu et al., 2019), this paradigm provides a rationale for seemingly paradoxical effect of tau on microglial phenotype. Thus, perhaps blockade of type 1 IFN signaling in microglia would remove the breaks on microglial activation and promote immune clearance of pathological substrates. However, inhibition of immune suppression could prove dangerous, as it is well established that chronically activated microglia are neurotoxic (Banati et al., 1993). Though intriguing, the hypothesis presented by Rexach and colleagues requires more experimental rather than correlative evidence. Future studies may aim to determine whether microglial phenotype dysregulation in rTg4510 mice could be rescued by inhibition of the IFN response. These data illustrate the concept that microglia are both “friend and foe” in tauopathies, as both too little and too much action by microglia can be damaging to neurons.

2.3. Genetic risk variants for tauopathy regulate microglia function

In addition to transcriptomic changes in microglia phenotype associated with tauopathy, there are several variants in genes associated with microglia that modulate risk for tauopathies (Pimenova et al., 2017). The well-known AD risk factor, apolipoprotein E (ApoE), regulates microglia phenotype during AD (Krasemann et al., 2017) and the ApoE4 risk genotype is correlated with high tau seeding activity (Dujardin et al., 2020). Moreover, both microglia and microglial ApoE are necessary for tau-mediated neurodegeneration in mice (Shi et al., 2019). ApoE’s immunological function in microglia is also linked to the function of the gene encoding triggering receptor expressed on myeloid cells 2 (Trem2) (Shi and Holtzman, 2018), another microglia-specific gene harboring AD risk variants (Guerreiro et al., 2013) that influences tau burden in tauopathy mice (Gratuze et al., 2020). Together, microglial ApoE and Trem2 are thought to interact to transition microglia from a homeostatic state to a neurodegenerative disease phenotype (Krasemann et al., 2017). The gene encoding phospholipase C gamma 2 (PLCγ2) is also linked to ApoE/Trem2 signaling and a P522R mutation in this gene has been associated with reduced late onset AD (LOAD) risk (Andreone et al., 2020; Conway et al., 2018; Sims et al., 2017). Curiously, one study reported that the same PLCG2 variant was positively associated with PSP though these data lacked statistical power (Conway et al., 2018). Larger validation studies would be highly valuable given recent experimental evidence that shows Aβ pathology-dependent roles of Trem2 signaling on tau pathology (Gratuze et al., 2021). Bridging integrator 1 (BIN1), the second most prevalent LOAD risk locus behind APOE (Seshadri et al., 2010), has been shown to promote EV release by microglia (Crotti et al., 2019). Microglia-derived EVs have been implicated in the spread of tau pathology (Asai et al., 2015; Clayton et al., 2021). These are just a few examples of tauopathy risk genes linked to microglial function and phenotype. Overall, while we do not comprehensively understand how these genetic risk factors lead to aggregation and spread of misfolded tau or how this leads to neurodegeneration, microglia clearly play an important role.

2.4. Microglia phenotype in tau PET imaging

PET neuroimaging of neuroinflammation, tau, and Aβ permits assessment of pathologic processes during life. PET-based metrics of neuroinflammation are chiefly derived from microglial marker translocator protein 18 kDa (TPSO otherwise known as peripheral benzodiazepine receptor). Microglia activation, measured via [¹¹C]-PBR28, correlates with cerebrospinal fluid (CSF) TREM2 and other inflammatory biomarkers (Pascoal et al., 2021) that capture the neuroinflammatory phenotype. PET studies on PSP, CBD, and AD cases demonstrate microglia activation generally parallels the distribution of tau deposition (Cagnin et al., 2006, 2001; Gerhard et al., 2006, 2004). Similar to histopathological studies, microglia activation correlates with declining MMSE scores, while amyloid burden does not (Edison et al., 2008). PET studies combining amyloid, tau, and microglia neuroimaging also reflect histopathological studies: increased brain binding of [¹¹C]-PBR28 for activated microglia spatially correlates with [¹⁸F]-AV1451 NFT binding more so than [¹⁸F]-flutemetamol Aβ plaque binding (Dani et al., 2018). Regression analysis in a recent PET study using [¹⁸F]-AZD4694 (Aβ), [¹⁸F]-MK-6240 (tau), and [¹¹C]-PBR28 revealed progression of tau pathology depended on the microglia activation network, mirrored Braak NFT staging, and that this relationship is potentiated by Aβ-mediated microglia activation (Pascoal et al., 2021). Importantly, this study demonstrates that microglia may drive tau pathology in human patients. This is supported by earlier studies suggesting that changes in microglia phenotype precede tau pathology in a similar somatotopic pattern in human (López-González et al., 2015) and mouse studies (Jaworski et al., 2011; Yoshiyama et al., 2007). Altogether, these data provide critical support for the hypothesis that microglia influence progression of tau pathology, these descriptive studies cannot fully uncover whether microglia influence the progression of tau pathology via modulation of selective vulnerability or directly via influencing tau pathology.

2.5. Manipulation of microglia phenotype can alter vulnerability to tau pathology

Several experimental studies have demonstrated that proinflammatory microglia can initiate tau pathology (Bhaskar et al., 2010; Ghosh et al., 2013; Jiang et al., 2015; Lee et al., 2015; Li et al., 2003; Maphis et al., 2016, 2015a, 2015b; Quintanilla et al., 2004). Pro-inflammatory activation of microglia with LPS induces tau hyperphosphorylation via IL-1 dependent mechanisms in WT rodent neurons in vivo and in vitro (Bhaskar et al., 2010; Lee et al., 2015; Li et al., 2003). Interestingly, IL-1β overexpression significantly increased tau hyperphosphorylation in 3xTgAD mice with P301L tau, even though Aβ plaque pathology was suppressed by 70–80 % and microglia plaque coverage was increased (Ghosh et al., 2013). Additionally, adoptive transfer of microglia from human tau transgenic mice lacking the anti-inflammatory fractalkine receptor (CX3CR1) into wild-type (WT) mice induces tau pathology in recipient mice, which is reduced with anti-IL1β treatment (Maphis et al., 2015a). Overall, these data demonstrate an important role for activated microglia in the initiation of tau pathology, particularly via pro-inflammatory IL-1β cytokine signaling. Conversely, tau pathology is ameliorated when neuroinflammation is suppressed or reduced by manipulating microglia phenotype via suppression of TNFα in 3xTgAD mice (Gabbita et al., 2015), or blockade of IL-1 signaling in 3xTgAD mice (Kitazawa et al., 2011). Additionally, Mapt knockout neurons and mice are protected from LPS induced neuroinflammatory toxicity in Cx3cr1 knockout mice (Maphis et al., 2015a) suggesting that tau mediates the neuronal toxicity of activated microglia. Overall, these data demonstrate that there is a toxic reciprocal cycle between vulnerability to tau pathology and microglia phenotype.

Over the years, studies examining TREM2 have made it increasingly clearer that TREM2 and the microglial MGnD/DAM phenotype it regulates play context specific roles in AD. In mouse models of Aβ plaque pathology, TREM2 deficiency exacerbates plaque-associated toxicity most likely due to the loss of the protective plaque compaction performed by MGnD/DAM microglia (Jay et al., 2017; Ulrich et al., 2014; Wang et al., 2016; Yuan et al., 2016). But studies using double transgenic tauopathy-TREM2 deficient models suggest a very different and more nuanced role of TREM2 in the context of tau pathology (Bemiller et al., 2017; Gratuze et al., 2020; Jiang et al., 2015; Leyns et al., 2017, 2019; Sayed et al., 2018). Deletion of murine TREM2 (mTREM2) reduces the brain atrophy and microgliosis in P301S mice without affecting tau phosphorylation (Leyns et al., 2017; Sayed et al., 2018). However, 5-month-old P301S mice intracerebrally injected with lentivirus expressing a TREM2 silencing cassette displayed heightened tau phosphorylation and microgliosis at 2 months post-injection (Jiang et al., 2015). This is concordant with what was observed in hemizygous mTREM2 knockout P301S mice (Sayed et al., 2018). Strikingly, TREM2 overexpression in P301S mice, attenuated microglial expression of pro-inflammatory cytokines, neuronal loss, tau phosphorylation, and decline in spatial memory (Jiang et al., 2016). More recently, studies have utilized human TREM2 knock-in mice with either R47H mutation (LOAD risk variant) or common variant crossed with P301S mice (Gratuze et al., 2020; Sayed et al., 2021). Gratuze and colleagues observed that the homozygous R47H-hTREM2;P301S genotype attenuated microglial reactivity, brain atrophy, and expression of the MGnD/DAM signature (Gratuze et al., 2020). Sayed and colleagues reported that the heterozygous R47H-hTREM2;P301S genotype had the opposite effect (Sayed et al., 2021), thus recapitulating what was observed in mTREM2 suppressed (Jiang et al., 2015) and hemizygous knockout mice (Sayed et al., 2018). Together these studies suggest that dampened or haploinsufficient expression of TREM2 exacerbates tauopathies. It is still unclear which downstream signaling mechanisms underpin this phenomenon. Both pro- and anti-inflammatory signaling pathways have been linked to TREM2 activation (Forabosco et al., 2013). In accordance with these observations, transcriptomic studies have teased apart “pro-inflammatory” and “anti-inflammatory” subsets of MGnD/DAM phenotype (Rangaraju et al., 2018). This could partially explain the observations that deletion and overexpression have the same outcome with respect to pro-inflammatory cytokine expression. Curiously, TREM2 hemizygosity has no impact on pro-inflammatory cytokine expression in Aβ plaque models (Ulrich et al., 2014). These context-dependent differences might be explained by the differences in extracellular concentration of putative ligands of TREM2 ligands or other parallel immune receptors that influence microglial phenotype. Studies which focused on tau spread in the context of Aβ pathology reported loss of TREM2-dependent MGnD/DAM signaling exacerbated tau-spread to peri-plaque dystrophic neurites (Gratuze et al., 2021; Leyns et al., 2019). Based on these data one may speculate that the MGnD/DAM phenotype is protective during the early stages of AD by mitigating plaque-associated toxicity and tau spread, but concurrent chronic inflammation has deleterious repercussions on the progression of tau pathology. These studies underscore the importance of further investigating microglia-tau interaction as improper timing of TREM2-targeted treatments could have serious consequences in a patient population.

Despite the experimental findings that show suppressing certain pro-inflammatory signaling pathways ameliorates tau pathology, one should avoid making the generalization that inflammation is intrinsically harmful in tauopathies. Indeed, microglial surveillance and clearance functions are essential in neurodegenerative disease where extracellular protein aggregates and cell debris are rampant. For example, microglial activation and clearance of neuronal TDP-43 protein aggregates was shown to be necessary for recovery of motor neuron function in a reversibly inducible TDP-43 expression model of amyotrophic lateral sclerosis (Spiller et al., 2018). Important to the topic of this review, TDP-43 aggregates can be a feature of secondary tauopathies, including AD (Huang et al., 2020). Further, Rexach and colleagues show that an immune suppression rather than activation phenotypic module is correlated with tau pathological progression (Rexach et al., 2020). Therefore, studies aiming to generate immunomodulatory treatments for tauopathies should consider the importance of preserving the homeostatic immune function of microglia.

Another important microglial phenotype with emerging mechanistic relevance to tauopathies is senescence. Cellular senescence is characterized by cell cycle arrest (Hayflick and Moorhead, 1961) and the senescence-associated secretory phenotype (SASP) (Coppé et al., 2008). As reviewed elsewhere, the number of senescent and/or dystrophic microglia in the brain increases with advanced age and neurodegenerative disease (Angelova and Brown, 2019; Streit et al., 2014). Microglia senescence is characterized by cytorrhexis, shortened processes, reduced motility, reduced phagocytic capacity, and increases in pro-inflammatory cytokines and reactive oxygen species (Angelova and Brown, 2019). Microglia with dystrophic morphology, which may represent senescence, accumulate in brain regions with high tau burden in human AD (Sanchez-Mejias et al., 2016; Streit et al., 2009, 2014). Notably, senescent microglia and astrocytes, characterized by p16^INK4A expression, accumulate in P301S tauopathy mice; genetic or pharmacological elimination of senescent microglia (and astrocytes) from these mice reduces tau phosphorylation, neurodegeneration, and cognitive decline (Bussian et al., 2018), although a conflicting study showed that pharmacological elimination of senescent cells in rTg4510 tauopathy mice was protective by eliminating senescent NFT-containing neurons rather than senescent glia (Musi et al., 2018). More recently, it was shown that after ingesting tau aggregate-bearing neurons, SASP microglia secrete tau extracellularly and exhibit reduced phagocytic capacity (Brelstaff et al., 2021). Taken together, these data suggest that tau pathology contributes to microglial senescence, but that microglia senescence also contributes to tau pathology in a feedforward cycle. It is currently unclear what, if any, overlapping characteristics are shared between senescent microglia and MGnD/DAM microglia. A recent transcriptomic study of nuclei from human AD brain identified subpopulations of senescent cells that corresponded with tau pathology, although excitatory neurons but not microglia or other glia expressed markers of senescence (Dehkordi et al., 2021). However, single nuclei transcriptomic studies show poor sensitivity for detection of microglia phenotype in human brain (Thrupp et al., 2020), warranting further investigation into the intersection of the MGnD/DAM phenotype and senescence to determine whether they are distinct or overlapping microglia phenotypes.

3. Microglia internalize tau

Tau is released into the extracellular space of the brain as a mixture of free tau and vesicular tau (Brunello et al., 2020). Several studies have observed microglia take up free tau in vitro (Andersson et al., 2019; Asai et al., 2015; Bolós et al., 2015; Funk et al., 2015; Luo et al., 2015; Majerova et al., 2014; Sanchez-Mejias et al., 2016; Zilkova et al., 2020) and when injected directly into the mouse brain (Bolós et al., 2015). Additionally, microglia can internalize neurons that contain tau in vitro (Sanchez-Mejias et al., 2016). There is also evidence that this uptake occurs with endogenously expressed tau in vivo: tau is observed within microglia in human cases of AD(Bolós et al., 2015; Ghoshal et al., 2001; Odawara et al., 1995), mouse models of tauopathy (Asai et al., 2015; van Olst et al., 2020), and aged Tupaia belangeri (tree shrews) which develop tau pathology naturally with age (Rodriguez-Callejas et al., 2020). Microglia do not express the tau gene Mapt (Hopp et al., 2018; Y. Zhang et al., 2014); however, microglia isolated from the rTg4510 mouse model of tauopathy do contain tau protein(Hopp et al., 2018). Several receptors facilitating tau uptake in neurons and astrocytes have been identified (De La-Rocque et al., 2020; Martini-Stoica et al., 2018; Wang and Ye, 2020), and there may be therapeutic potential in targeting them to halt the spread of tau. There is a relative dearth of information on the mechanisms by which microglia internalize tau. Microglial heparan sulphate proteoglycan (HSPG) (Funk et al., 2015), CX3CR1 (Bolós et al., 2017), and P2RY12 (Das and Chinnathambi, 2021) have all been shown to bind tau, albeit only blockade of CX3CR1 and HSPGs was shown to inhibit uptake of tau. HSPGs have also been shown to mediate tau uptake in neurons (Holmes et al., 2013) and astrocytes (Martini-Stoica et al., 2018). Additionally, low density lipoprotein receptor related protein 1 (LRP1), an important immune regulator in microglia (Chuang et al., 2016), has been shown to mediate tau uptake in neurons (Rauch et al., 2020). Therefore, it will be important to verify experimentally whether microglial LRP1 also mediates tau uptake, as future therapies that target these receptors to block neuronal tau propagation may also affect microglial uptake of tau. Given microglial uptake of tau represents a beneficial clearance mechanism, consideration of microglia-specific tau internalization mechanisms is warranted.

3.1. Phagocytosis

Classical work, such as that of Gey et al. in 1954 and W.H. Lewis in 1931, distinguished phagocytosis (cell “eating”) from pinocytosis (cell “drinking”) (Gey et al., 1954; Lewis, 1931). Yet, in the fields of tauopathy and microglia, the term “phagocytosis” is frequently and inaccurately used to describe microglial uptake of extracellular, soluble tau multimers. The biomechanics of phagocytosis dictate why this process is better suited for uptake of large particles (≥0.5 μm in diameter) such as foreign bodies (e. g. bacteria) or cell debris (e. g. apoptotic bodies) over solutes, including tau multimers. Indeed, phagocytosis initiates upon the cooperative activation of multiple phagocytic receptors which bind to different sites on the surface of the particle directly or indirectly (Rosales and Uribe-Querol, 2017). More receptors are recruited to the phagocytic cup to facilitate the anchoring of the membrane protrusion as the membrane progressively covers the surface of the particle (Jaumouillé and Waterman, 2020). Mathematical modeling complemented by experimental observations indicate that particles with a diameter of 2–3 μm are most efficiently internalized via phagocytosis (Champion et al., 2008). Therefore, microglia utilize phagocytosis to clear apoptotic neurons (Grommes et al., 2007) and prune synapses (Paolicelli et al., 2011) while employing liquid-phase pinocytic mechanisms for uptake of small protein aggregates (Mandrekar et al., 2009). We speculate microglial phagocytosis of tau seeds is restricted to engulfment of tau seed containing synapses and whole neurons, rather than small soluble extracellular aggregates.

Opsonic phagocytosis involves binding of opsins (e. g. complement and immunoglobulin G (IgG)) to a particle before it can be internalized (Flannagan et al., 2012). Immunoreactivity of complement has been observed near Aβ-plaques, dystrophic neurites, neuropil threads, and NFTs in histological sections of AD brains (Loeffler et al., 2008). Synapse pruning by microglia is carried out using the complement system (Stephan et al., 2012). Some have postulated that phagocytosis of tau-containing synapses tagged with complement component 1q (C1q) is a means by which microglia internalize tau (Clayton et al., 2021). Interestingly, C1q is known to interact with LRP1 (Duus et al., 2010), a receptor that has been implicated in the uptake of tau by neurons (De La-Rocque et al., 2020). Another major opsonic receptor is FcγR which mediates phagocytosis of IgG coated “meals” (Flannagan et al., 2012). Both Aβ- and tau-targeting IgG antibodies have been in development as potential immunotherapies to treat AD for over a decade (Cummings et al., 2014). However, it is unlikely that anti-tau monoclonal antibodies (mAbs) targeting a single specific region of tau, such as Semorinemab (which binds the N-terminus of tau), can coat the surface of a tau multimer. Since only large, opsonized particles are internalized via FcγR-mediated phagocytosis, these smaller immune complexes are more likely to be internalized by a different form of FcγR-mediated endocytosis (Huang et al., 2006; Tse et al., 2003).

Apoptotic cell receptors allow phagocytes to clear apoptotic cells while suppressing the immune response (Szondy et al., 2017). Examples include (α_vβ_3/5 integrin and Axl/Mer which bind to milk-fat globule epidermal growth factor 8 (MFG-E8) and Gas6, respectively (Grommes et al., 2007; Neniskyte and Brown, 2013). Bridging proteins MFG-E8 and Gas6 bind to phosphatidylserines (PS), which are normally found on the inner leaflet of the membrane but are flipped to the outer leaflet in apoptotic cells (Bratton et al., 1997). Thus, PS provide an “eat me” signal which helps microglia distinguish living from dying neurons. Experimental studies suggest that tau-filled living neurons erroneously express these “eat me” signal, and this results in their aberrant phagocytosis by microglia via MFG-E8 bridging protein (Brelstaff et al., 2018, 2021). Intriguingly, microglia displaying a senescence-like phenotype, become hypophagocytic, and secrete tau after ingesting tau filled neurons in vitro (Brelstaff et al., 2021).

3.2. Macropinocytosis

Macropinocytosis is a form of pinocytosis (cellular “drinking”) in which membrane ruffles protrude from the surface and then join to form a nascent macropinosome (diameter ≈0.25–5 μm) (Canton, 2018). Like phagocytosis, macropinocytosis is actin dependent (Canton, 2018). However, unlike phagocytosis, macropinocytic membrane ruffles (e. g. lamellipodia, circular dorsal ruffles) do not adhere to the surface of the engulfed particle(s), and thus the size of the macropinosome is not dependent on the size of the particle (Jaumouillé and Waterman, 2020). In most cell types, macropinocytosis is triggered by some intra- or extracellular stimulus, but it can also occur constitutively in certain myeloid lineage cells (Canton, 2018; Lim and Gleeson, 2011). For example, macrophages (Bohdanowicz et al., 2013; Steinman et al., 1976; Yao et al., 2009), and immature dendritic cells (Garrett et al., 2000; Norbury et al., 1997) employ constitutive macropinocytosis to survey the extracellular milieu (Doodnauth et al., 2019). Homeostatic microglia, like macrophages, serve as sentinels monitoring their local environment through continuous sampling via pinocytosis (Booth and Thomas, 1991; Canton, 2018; Canton et al., 2016; Fitzner et al., 2011; Z. Liu et al., 2020; Mandrekar et al., 2009; Ranson and Thomas, 1991; Redka et al., 2018). In fact, these surveying, ramified microglia are the most highly pinocytic cells in the brain (Booth and Thomas, 1991; Ranson and Thomas, 1991). Such propensity for liquid phase endocytosis, suggests that microglia activate macropinocytosis constitutively like their myeloid kin (macrophages). Further, a recent publication showed that microglial macropinocytosis involves the same signaling pathways that mediate macrophage and dendritic cell macropinocytosis (Z. Liu et al., 2020). Microglial macropinocytosis has been shown to mediate the uptake of soluble Aβ (Mandrekar et al., 2009), and most likely contributes to uptake of soluble tau (Funk et al., 2015). Microglial macropinocytosis has been shown to mediate uptake of oligodendrocyte derived exosomes (Fitzner et al., 2011). Tau containing exosomes have been shown to contribute to tau spread (Asai et al., 2015; Clayton et al., 2021). Further, it has been shown that exosomes permeabilize the endolysosomal compartment, providing tau a means to escape degradation (Polanco et al., 2021). Therefore, it would be valuable to investigate whether neuron-derived exosomes containing tau are also taken up via macropinocytosis. Moreover, considering the fundamental importance of macropinocytosis to myeloid cells, more studies characterizing the role of macropinocytosis in microglia are warranted.

Tau aggregates have been shown to bind to HSPGs on the surface of neurons (Holmes et al., 2013), astrocytes (Martini-Stoica et al., 2018) and microglia (Funk et al., 2015). HSPGs are ubiquitous glycoproteins present on the surface of cells and the constitute an important component of the extracellular matrix and are internalized via macropinocytosis (Sarrazin et al., 2011). Internalization of HSPGs occurs constitutively, so HSPG-mediated tau internalization may be a passive process of tau simply binding to HSPGs and “hitching a ride” to the endosomal compartment (Sarrazin et al., 2011); however, there is evidence that HSPGs can act as a receptor to induce endocytosis of specific ligands (Wittrup et al., 2009). Ultrastructural analysis via electron microscopy images revealed that tau bound to the neuronal surface HSPGs enters the cell via lamellipodia-like projections that form large non-specific vesicles resembling macropinosomes (Holmes et al., 2013). In microglia, one group showed similar lamellipodia-like structures form in response to extracellular tau in vitro (Das et al., 2020; Desale and Chinnathambi, 2021; Uhlemann et al., 2016). The same group also demonstrated that treatment with dietary supplement α-linolenic acid (ALA) mimicked these effects and promoted tau endocytosis by microglia in vitro (Desale and Chinnathambi, 2021). The authors posit that ALA could be used to slow tauopathy progression by promoting microglial tau clearance. But, if ALA indeed induces tau endocytosis via lamellipodia projection and formation of macropinosomes, then it would likely also promote neuronal (Holmes et al., 2013) and astroglial (Martini-Stoica et al., 2018) tau uptake. This underscores the importance of recognizing that microglia, despite being the “phagocytes of the brain,” utilize non-phagocytic forms of endocytosis that overlap with those used by other cells.

3.3. Micropinocytosis

Micropinocytosis, which includes clathrin-mediated, caveolae-dependent, and clathrin/caveolae-independent endocytosis, is a form of endocytosis in which smaller vesicles take up liquid-phase particles through either receptor-mediated or receptor-independent processes (Iversen et al., 2011; Saha et al., 2013; Solé-Domènech et al., 2016). Neuronal micropinocytic mechanisms of tau internalization include clathrin-mediated endocytosis (CME) via muscarinic acetylcholine receptors and LRP1, and clathrin-independent endocytosis via flotillin-dependent and flotillin-independent lipid microdomains (De La-Rocque et al., 2020). To the best of our knowledge, no studies have specifically examined the role of any of these receptors or endocytic mechanisms in microglial uptake of tau. Like neurons, microglia undergo various forms of micropinocytosis (Minami et al., 2012; Solé-Domènech et al., 2016) and express some of the same cell surface receptors (e. g. LRP1) as neurons (Rauch et al., 2020).

Chemokine receptors are internalized either constitutively or upon ligand binding via CME or caveolae-dependent endocytosis (Bonecchi et al., 2010). Bolos and colleagues showed that knockout of microglial fractalkine receptor CX3CR1 reduced microglial tau uptake in vitro and in vivo, and that monomeric tau has high affinity for CX3CR1, albeit reduced with tau phosphorylation by the kinase glycogen synthase 3 β (GSK3β) (Bolós et al., 2017). Further, they showed that endogenous ligand CX3CL1 competitively inhibits binding of monomeric tau to CX3CR1 (Bolós et al., 2017). The biological relevance of CX3CR1 as a microglia receptor of tau is unclear, since neurons, the main producer of tau protein in the brain, more efficiently release tau dimers and multimers compared to tau monomers (Wegmann et al., 2016) and high-molecular weight phosphorylated tau strains are thought to be the main tau strains involved in propagation of tau seeds (Takeda et al., 2015). Nevertheless, the findings of Bolós and colleagues are intriguing considering CX3CR1 activation by endogenous ligand CX3CL1 results in suppression of the immune activation (Biber et al., 2007). Bolós and colleagues also found that coincubation of CX3CL1 and tau monomers reduces binding affinity of tau monomers to CX3CR1 (Bolós et al., 2017) which might suggest that tau monomers bind to the orthosteric site of CX3CR1. Other studies have shown that aggregated tau induces microglial pro-inflammatory activation (Das and Chinnathambi, 2021; Morales et al., 2013). Thus, it would be informative to investigate how tau monomers and aggregates modulate dose-dependent responding of CX3CR1 to CX3CL1. Tau acting as either an antagonist or inverse agonist at this receptor would provide a mechanistic explanation for induction of microglial activation by tau. Future work is needed to determine which receptors bind and promote the internalization of tau and the other downstream pathways are affected, particularly in the context of microglial activation state.

FcγR-mediated phagocytosis is restricted to the engulfment of large, IgG coated particles (Huang et al., 2006; Tse et al., 2003). Smaller antigen-antibody complexes and aggregates of IgG are internalized via FcγR-mediated actin-independent micropinocytic mechanisms, such as CME (Huang et al., 2006; Tse et al., 2003). In mouse brain slice cultures expressing human tau treated with fluorescently labeled 4E6G7 anti-tau mAb, pharmacological inhibition of clathrin but not actin blocked antibody uptake (Congdon et al., 2013). Similar results were observed in primary culture neurons treated with the same antibody (Congdon et al., 2013). Others have found that anti-tau antibodies robustly enhance uptake of small, soluble tau aggregates by microglia (Andersson et al., 2019; Funk et al., 2015; Zilkova et al., 2020) most likely via actin-independent micropinocytosis rather than phagocytosis (Huang et al., 2006; Tse et al., 2003). The same pharmacological inhibitors used in Congdon et. al. could be applied to cultured microglia in the presence of tau antibodies to implicate or disqualify actin-independent micropinocytosis involvement.

The protein product of LOAD risk gene BIN1 has been shown to be a negative regulator of CME in neurons (Calafate et al., 2016). The neuronal isoform of the BIN1 protein (Bin1) is unique in that it contains a clathrin and AP-2-binding (CLAP) domain which has been shown to mediate the interaction between clathrin and its adaptor protein, AP2 (Micheva et al., n.d.; Ramjaun and McPherson, 1998; Slepnev et al., n.d.; Wigge et al., 1997). In an in vitro model of neuronal tau spread, overexpression and knockdown of BIN1 decreased and increased tau spread, respectively (Calafate et al., 2016). Interestingly, the microglial isoform of BIN1 has been shown to promote the release of tau in EVs (Crotti et al., 2019) which may protect it from lysosomal degradation (Polanco et al., 2021). The role of CME and the various clathrin-independent micropinocytic mechanisms have not been explored for microglial uptake of tau. Indeed, CME is implicated in both microglia and tau pathology during AD: PICALM, a clathrin adaptor protein, is localized to both microglia and tau during AD, and polymorphisms in PICALM are risk factors for AD (Ando et al., 2013). Since micropinocytic mechanisms have been identified to mediate neuronal uptake of tau, it will be important to determine whether the same mechanisms in microglia facilitate tau uptake and whether this leads to tau clearance or spread.

4. Microglia release tau seeds

Both mouse and human microglia release seed-competent tau into their conditioned media ex vivo (Hopp et al., 2018), suggesting that microglia release tau that can cause tau aggregation in recipient cells. Tau protein and bioactive tau seeds are still present in conditioned media from microglia isolated from tauopathy mice up to at least 6 days in vitro. This is longer than the previous reported half-life of tau in vitro which ranges from 3- to 20-hours (David et al., 2002; Shimura et al., 2004), indicating (1) that seed competent tau strains are surprisingly long lived within microglia compared to previous in vitro half-life data; (2) that microglia are not fully processing tau into non-toxic forms; and (3) that seed competent tau strains are released by microglia. Microglia may release tau seeds via intracellular tau mediated microglia toxicity (Sanchez-Mejias et al., 2016) leading to release of not-yet-degraded tau seeds. There may be common mechanisms mediating endosomal-lysosomal dysfunction across various amyloid aggregates (Aβ, NFTs, α-synuclein), including but not limited to intracellular vesicle rupture (Flavin et al., 2017; Polanco et al., 2021). Several studies suggest that microglia secrete tau-containing exosomes in vitro (Crotti et al., 2019) and in vivo (Clayton et al., 2021) which may be how microglia propagate spread of tau directly. However, it still unknown whether the microglia-derived exosomes contain seed-competent aggregates, and it is unclear how much tau-containing exosomes contribute to tau spread in vivo. Additionally, microglia secrete factors in addition to tau or tau-containing exosomes that may exacerbate tau propagation in neurons. For example, neurons express a variety of cytokine receptors that trigger specific intracellular signaling cascades that may confer specific vulnerability to the combination of tau and cytokines release by microglia, since cytokines themselves cause tau seeding (Gorlovoy et al., 2009) and phosphorylation (Maphis et al., 2015b).

Exosomes are EVs derived from the endosomal comparment which are used to dispose of waste or send long-distance signals (Colombo et al., 2014). Internalized exosomes are trafficked to the late endosome where they may either by transported to the cell membrane for exocytosis or fuse with the lysosome for degradation (Colombo et al., 2014). However, the endolysosome does not efficiently degrade the exosomes and can rupture following exosome uptake (Polanco et al., 2021). This suggest that exosomal-tau has higher potential for spread than exosome-free tau. Blockade of exosome release and depletion of microglia can mitigate tau spread in mouse models of tau seeding and coadministration of LPS and ATP can stimulate exosome release by microglia (Asai et al., 2015). The authors of this study postulate that LPS stimulation drives ubiquitination via TLR4 activation, ATP activates P2X7 receptor (P2X7R) to promote release of exosomes, and ubiquitinated proteins are incorporated in multivesicular bodies which are later released as exosomes (Asai et al., 2015). Indeed, in the same study they were able to show that LPS priming of microglia led to ubiquitination of tau (Asai et al., 2015), and P2X7R activation has been shown elsewhere to induce exosome release (Dubyak, 2012). More recent work from the same groups has shown that Aβ, a putative TLR4 ligand (Y. Liu et al., 2020), is associated with microglial exosome release in a mouse model of Aβ-plaque pathology and tau spread (Clayton et al., 2021). Thus, the coincidence of extracellular ATP released by stressed neurons and Aβ-plaques in the microenvironment may serve as a trigger for microglial exosome release. Given the recent results from PET imaging of human AD patients demonstrating that microglia drive the progression of tau pathology via potentiation by Aβ (Pascoal et al., 2021), these animal studies suggest a fascinating mechanism by which Aβ and tau pathology synergize via microglia dysfunction.

BIN1, the second most prevalent LOAD risk locus (Seshadri et al., 2010) plays a key role in microglia exosome release (Crotti et al., 2019) in addition to its aforementioned role in CME. BIN1 human isoform 6 is predominantly expressed in microglia and shares 95 % sequence similarity with the murine microglia-predominant variant, murine isoform 2 (Crotti et al., 2019). Human embryonic kidney 293 T (HEK293T) cells transfected with plasmid encoding the sequences of tau and murine isoform 2 resulted in greater release of tau-containing exosomes than HEK293T transfected with tau alone (Crotti et al., 2019). Further, microglia-specific knockout of BIN1 in PS19 mice reduces tau spread in male but not female mice (Crotti et al., 2019). Further studies are warranted to determine the cause of these sex differences. Nevertheless, these findings taken together with the in vitro findings that neuronal BIN1 loss potentiates tau spread (Calafate et al., 2016) presents a divergent role of microglial and neuronal BIN1. Although only the neuronal isoforms of Bin1 expresses a CLAP domain which interacts with the CME machinery (Micheva et al., n.d.; Ramjaun and McPherson, 1998; Slepnev et al., n.d.; Wigge et al., 1997), all isoforms of Bin1 contain a BAR domain which senses membrane curvature and an SH3 domain which interacts with dynamin. This presents evidence of crosstalk between the endocytic pathways of tau internalization and tau-containing exosome release in microglia.

It is important to point out that two recent studies modeling tau propagation in the context of Aβ pathology have reported seemingly contrasting effects of ablating microglia on tau spread (Clayton et al., 2021; Gratuze et al., 2021). Gratuze et. al. reported that ablation of microglia increased tau spread, and Clayton et. al. reported that ablation of microglia reduced tau spread. In both studies, tau pathology was seeded via stereotaxic injection in a discrete brain region and a CSF1 receptor (CSF1R) inhibitor was used to ablate microglia (Clayton et al., 2021; Gratuze et al., 2021). However, there are several differences in their experimental design that could account for these contradictory observations. For instance, Gratuze et. al. injected human AD derived tau aggregates into the hippocampus and cortex of a 5XFAD mouse, whereas Clayton et. al. injected a viral vector driving the expression of P301L tau into the entorhinal cortex of a humanized amyloid precursor protein (APP) knock-in APP^NL-G-F mouse. Further, Clayton et. al. used a more CSF1R-selective microglia depleting drug (PLX5562) than the one used by Gratuze et. al. (PLX3397), which also inhibits c-Kit (DeNardo et al., 2011). Notably, the two Aβ pathology mouse models differ in that transgenic 5XFAD mice overexpress human APP and presenilin 1 (PS1) with familial AD mutations under the Thy1 promoter, while the knock-in model, APP^NL-G-F, expresses human familial AD mutant APP at lower levels and does not express human PS1. Thus, it is possible that regional, mouse, or pharmacological differences may affect the role that microglia play in tau spread. Also, it should be noted that not all microglia are ablated by CSF1R inhibitor treatment. Indeed, Gratuze et. al. reported that approximately 20 % of microglia survived ablation and 80 % of these treatment-resistant microglia were associated with Aβ-plaques. Ablation-resistant microglia have been characterized via single-cell RNA sequencing and were shown to have a highly similar transcriptomic profile to microglial progenitor cells (Zhan et al., 2020). This explains why microglia rapidly repopulate after withdrawal of CSF1R inhibition (Elmore et al., 2014; Gratuze et al., 2021; Zhan et al., 2020). Curiously, ablation-resistant progenitor-like microglia highly express Mac2, also known as Galectin-3 (Zhan et al., 2020) a purported marker of the plaque-associated microglia implicated in the spread of tau pathology (Clayton et al., 2021). Given this information, one might postulate that these plaque-associated, Mac2-positive microglia implicated in tau spread constitute a population of microglial progenitor-like cells. Future studies are warranted to characterize the transcriptome of the microglial population implicated tau spread to determine if distinct subpopulations of microglia respectively promote and mitigate tau spread. Nevertheless, data from Gratuze et al. highlight an important protective role for microglia and microglial TREM2 in suppression of tau spread. Microglial degradation of tau is likely an important mechanism by which microglia exert protective functions during tauopathy.

5. Microglia degrade tau

Early studies in human tissues suggest that microglia may process tau into specific forms based on immunohistochemical labeling with conformation-specific anti-tau antibodies. For example, the anti-tau antibody Tau-2 labels a subset of microglia-like cells in AD and dementia with Lewy bodies (Odawara et al., 1995) as well as microglia-like cells surrounding ischemic foci (Uchihara et al., 2005, 2004, 2000). The Tau-2 antibody also labels NFTs and neuropil threads in AD tissue (Odawara et al., 1995; Uchihara et al., 2005, 2004, 2000). However, unlike NFTs, microglia-like Tau-2 positive cells did not show labeling with other anti-tau antibodies, were devoid of argyrophilic fibrillar characteristics, and underwent a reversible conformation change upon exposure to detergent (Uchihara et al., 2005, 2004, 2000). Similarly, the anti-tau antibody Tau-66 labels a subset of microglia in AD but not control brains which were not labeled by other anti-tau antibodies such as Tau5 (total tau, aa210–241) (Ghoshal et al., 2001). The Tau-66 antibody also labels diffuse reticulated plaques as well as NFTs and neuronal cell bodies in both a punctate and ribosomal-like pattern. Notably, the ribosomal-like staining is like that which Tau-2 stains in neurons. The Tau-66 antibody recognizes a discontinuous epitope of tau containing residues aa155–244 and aa305–314 (Ghoshal et al., 2001) and is associated with loss of the N- and C- termini in late-stage conformational changes in AD tau pathology (Ibarra-Bracamontes et al., 2020) but does not label PSP or CBD tau pathology (Berry et al., 2004). More recently, Iba1-labeled microglia were found to colocalize with tau-containing structures labeled with phospho-specific (PHF-1 (pS396, pS404), AT100 (pT212, pS214_, Tau396 (pS396), AT180 (pT231), Tau404 (pS404), and AT8 (pS202, pT205)) and total tau antibodies (Tau-1 (aa162–210) and HT7 (aa159–163)) (Bolós et al., 2015). However, colocalization or apposition between microglia and these phospho- and total-tau antibodies was partial (suggestive of active engulfment), whereas the Tau-66 and Tau-2 labeling showed microglia that were filled with cytoplasmic tau (perhaps suggestive of failed degradation). Overall, these observations confirm that microglia internalize tau in vivo, and suggest that microglia may process tau by cleaving at specific epitopes, leading to specific conformational changes recognized by Tau-2 and Tau-66. Currently, little is known about the fragmentation, conformation, phosphorylation, or other post-translational modifications that microglia perform on tau during degradative processing or preparation for release in exosomes and whether these modifications represent an increase or decrease in tau toxicity or propensity for spread.

Several studies have shown that microglia are capable of degrading tau under different conditions in vitro (Andersson et al., 2019; Luo et al., 2015; Majerova et al., 2014). Majerova and colleagues found that heparin-induced recombinant tau oligomers (tau151–391) were internalized and degraded by primary rat microglia, rat primary blood monocyte derived macrophages, immortalized microglia-like mouse BV2 cells, and immortalized macrophage-like mouse TIB67 cells, but macrophages were more efficient than microglia and microglia degradation of tau was only efficient after LPS exposure (Majerova et al., 2014). Luo and colleagues found that human sarkosyl-insoluble tau fibrils from human AD cases were internalized and degraded by primary mouse microglia and that this was enhanced by the MC1 anti-tau antibody in an Fc-dependent manner (Luo et al., 2015). Andersson and colleagues found that rTg4510-derived sarkosyl-insoluble tau was internalized by primary mouse microglia in an Fcγ-dependent fashion. They also found that inhibition of lysosomal acidification in microglia with chloroquine or bafilomycin A1 reduced sarkosyl-insoluble hyperphosphorylated tau degradation, while inhibition of the proteosome did not (Andersson et al., 2019). Lysosomal acidification is important for pH-sensitive proteolytic cleavage. Notably, microglia express a specific cysteine protease, asparaginyl endopeptidase (AEP, otherwise known as legumain) that has been implicated in the formation of aggregate-prone tau fragments as well as physiological tau degradation, pointing to a dual role of microglia-mediated tau proteolysis (Behrendt et al., 2019). Legumain cleaves tau at four different sites (N167, N255, N296 and N368)(Schlegel et al., 2019). Previous studies indicated that legumain knockout mice show increased AT8 tau inclusions and that legumain-resistant tau prevents synaptic dysfunction (Z. Zhang et al., 2014). Studies point to both increased (Behrendt et al., 2019; Z. Zhang et al., 2014) or unchanged (Schlegel et al., 2019) legumain activity in human AD brains; the study that found unchanged levels controlled better for vascular issues which can drive legumain activity. Further, follow-up studies have not found accumulation of N368 tau fragments in tau aggregates in AD brains (Schlegel et al., 2019). Notably, microglia are enriched for legumain in the brain compared to other cell types and microglia are able to degrade tau without releasing cleaved fragments (Behrendt et al., 2019). When microglia are depleted from the brain via CSF1R inhibition, there is a significant reduction in legumain expression (Spangenberg et al., 2019). Cathepsin (in Greek: “to digest”) is a family of widely expressed proteases which are found within the endosomal/lysosomal compartments during homeostatic conditions and may also be secreted upon induction (Cavallo-Medved et al., 2011), such as through pro-inflammatory activation in the case of microglia (Lowry and Klegeris, 2018). Cathepsins are differentially classified based on the amino acid residue within the proteolytic site which confers their function as serine, cysteine, or aspartic proteases (Siklos et al., 2015; Turk et al., 2012). The role of microglial cathepsins in neurodegenerative disease is dependent upon context (Lowry and Klegeris, 2018). Microglial cathepsins serve not only immune clearance functions, but are also immune regulatory roles (Lowry and Klegeris, 2018). For example, intracellular cysteine protease cathepsin B has been implicated in tau processing (Bendiske and Bahr, 2003; Farizatto et al., 2017), but has also been shown to enhance the NF-κB and caspase-1 activation (Halle et al., 2008; Ni et al., 2015; Zhou et al., 2016). Pharmacological enhancement of cathepsin B has been shown to reduce hyperphosphorylated tau accumulation in cultured hippocampal slices (Bendiske and Bahr, 2003; Farizatto et al., 2017). However, interpretation of this data is difficult as cathepsin B is also highly expressed in neurons (Bernstein et al., 1990) and astrocytes (Oberstein et al., 2021). Like legumain, cathepsin B is pH sensitive, but it differs in that lower pH favors its exopeptidase activity while higher (neutral/alkaline) pH favors its endopeptidase activity (Cavallo-Medved et al., 2011). Thus, cathepsin B processing of tau may differ depending upon its location intra- or extracellularly and the acidification state of the lysosomal compartment. However, enhancement of mature cathepsin B expression via treatment with Z-Phe-Ala-diazomethylketone (PADK) was capable of reversing hyperphosphorylated tau accumulation induced by chloroquine treatment, despite the lysosome deacidifying property of chloroquine (Bendiske and Bahr, 2003). Further, PADK treatment has been shown to prevent hyperphosphorylated tau accumulation in the context of proteosome inhibition (Farizatto et al., 2017). Thus, it is possible that cathepsin B can degrade certain pathological forms of tau in both acidic and non-acidic contexts. However, due to the differing function of exo- or endopeptidases, it is likely that its cleavage products are dependent upon these two contexts. Importantly, extracellular cathepsin B has been shown to be neurotoxic independent of tau pathology (Kingham and Pocock, 2001). In addition to cathepsin B, cathepsin D (Bednarski and Lynch, 1996; Johnson et al., 1991) and cathepsin S (Nübling et al., 2017) have been shown to cleave tau. Notably, incubation of tau with cathepsin S alone results in tau oligomer formation (Nübling et al., 2017), suggesting that partial degradation of tau can be deleterious. Within the lysosome, cathepsins and other lysosomal proteases most likely work in concert to degrade tau into its constituent amino acids. Overall, the role of microglial cathepsins in the context of tauopathy remains poorly understood. Thus, further work is warranted to determine how modulation of cathepsin activity may affect microglial tau processing and tauopathy neurodegeneration.

6. Role of microglia phenotype in tau processing

Microglia phenotype changes throughout the course of tauopathy (Rexach et al., 2020; H. Wang et al., 2018), which may affect microglial processing of tau. When assessing the role of the MGnD/DAM and inflammatory phenotypes in tauopathy, it is important to remember that neither phenotype represents a monolith. Indeed, their constituent subpopulations may affect tau pathology in different ways. This may explain why experimental models of these two phenotypes have been shown to both restrict and propagate tau pathology under different contexts. For example, ApoE, an AD risk factor that is crucial for triggering the MGnD/DAM phenotype during AD (Krasemann et al., 2017), is correlated with high tau seeding activity in human cases of tauopathy (Dujardin et al., 2020). Ablation of microglia or microglia-specific knockout of ApoE reduces NFT-mediated neurodegeneration in tauopathy model mice suggesting that MGnD/DAM contribute to tau-mediated neurodegeneration (Shi et al., 2019). Additionally, microglia activation triggers tau release by microglia (Asai et al., 2015; Clayton et al., 2021). Similarly, activated microglia may contribute to tau pathology by exacerbating tau pathological phosphorylation, aggregation, and expression in mouse and cellular models (Bhaskar et al., 2010; Ghosh et al., 2013; Kitazawa et al., 2005; Lee et al., 2015; Li et al., 2003; Quintanilla et al., 2004). However, activated microglia can also restrict tau seeding and spreading: ablation of microglia or suppression of microglia activation by knockout of TREM2 in the context of Aβ pathology enhances tau spreading (Gratuze et al., 2021). This is also consistent with data showing that microglia need to be activated to internalize and degrade tau (Andersson et al., 2019; Luo et al., 2015; Majerova et al., 2014), but these studies themselves are not entirely consistent with each other. For example, Majerova and colleagues observed that microglia activation with LPS modestly enhances the proportion of microglia that internalize tau pre-formed fibrils from 26.4 % to 42.6 % with 24 h of LPS pretreatment with no effect with a 2–6 h LPS pretreatment (Majerova et al., 2014), while Andersson and colleagues did not note any significant change in intracellular tau levels with LPS treatment at an undescribed timepoint (Andersson et al., 2019). Notably, peripheral macrophages and monocytes are more effective than microglia at internalization and degradation of tau, suggesting there are specific properties of different myeloid cells that make them more efficient at tau degradation (Majerova et al., 2014). This could be due to reduced lysosomal acidity in microglia compared to other macrophages (Majumdar et al., 2007). Moreover, in the chronic inflammatory environment of AD, microglia lose key degradative enzymes (Hickman et al., 2008). Additionally, inflammatory activation is known to alter endocytosis in myeloid-lineage cells (Redka et al., 2018). These observations warrant a deeper examination into which cellular processes drive tau degradation without also causing improper processing and packaging and release of tau into extracellular vesicles. Such findings would not only help to reconcile contradictions on the influence of phenotype on tau processing, but also provide therapeutic targets for promoting tau clearance.

7. Strategies to promote tau clearance

The overall mechanism of tau clearance from the brain is unknown, although clearance into the glymphatic system (Patel et al., 2019), and vasculature (J. Wang et al., 2018) are involved. Notably, tau pathology induces vascular dysfunction that may affect clearance (Bennett et al., 2018). It is not known whether tau pathology leads to glymphatic dysfunction (Silva et al., 2021), but in mouse models of aging (Kress et al., 2014) and Aβ pathology (Peng et al., 2016) the glymphatic system is impaired as a result of astrogliosis and mislocalization of astrocytic aquaporin 4 (AQP4). Loss of AQP4 results in increased tau aggregation after traumatic brain injury (Iliff et al., 2014). Importantly, microglia compensate clearance of extracellular aggregates when the glymphatic system is impaired (Feng et al., 2020). Pre-plaque AQP4-deficient APP/PS1 mice exhibited more intraneuronal accumulation of Aβ and enhanced microglia activation compared to AQP4-intact APP/PS1 mice, but only loss of AqP4 combined with local depletion of microglia via clodronate liposomes resulted in plaque deposition (Feng et al., 2020). Whether microglia would compensate similarly with extracellular tau is unknown. However, it was shown that intrahippocampally injected monomeric tau is cleared from the brain parenchyma and deposited into the systemic circulation much more efficiently in mice with an intact glymphatic system than in mice in which the glymphatic system was ablated genetically (Patel et al., 2019). Future studies will need to investigate the role of microglia in extracellular tau clearance in the context of glymphatic and vascular dysfunction during tauopathy.

Therapeutics that target tau protein, such as anti-tau mAbs, are also in clinical trials (e. g. NCT03828747). Tau aggregates are reduced in tauopathy models after treatment with anti-tau mAbs (Ayalon et al., 2021; Castillo-Carranza et al., 2014; Yanamandra et al., 2013; Zilkova et al., 2020). Studies have shown that microglia are important for antibody-mediated tau clearance putatively via FcγR effector function (Andersson et al., 2019; Funk et al., 2015; Zilkova et al., 2020). However, a group at Genentech reported that microglial activation via effector function of anti-pS409-tau immunotherapy negates its therapeutic efficacy (Lee et al., 2016). The authors noted that only effectorless mutant anti-pS409-tau antibodies were able to block intraneuronal tau fragmentation induced by exogenously applied tau multimers when neurons were cocultured with microglia (Lee et al., 2016). They also observed that effectorless mutant anti-pS409-tau antibodies were modestly more effective in blocking tau spread in a mouse model of tauopathy than WT anti-pS409-tau antibodies, albeit both were significantly more effective than IgG control (Lee et al., 2016). Based on these findings, they conclude that FcγR effector function attenuates the therapeutic efficacy of tau immunotherapies by causing neurotoxicity via pro-inflammatory activation of microglia (Lee et al., 2016). Thus, passive immunotherapies with an IgG4 backbone which confers reduced effector function relative to other IgG isotypes (e. g. IgG1, IgG3) (Vidarsson et al., 2014) have dominated clinical trials (NCT03289143, NCT03828747, NCT03068468, NCT02880956). Thus far, most of these therapies have since been scrapped due to a failure to demonstrate disease modifying benefits except for one (Semorinemab) for which clinical trials are ongoing (NCT03828747). Lee and colleagues’ rationale for choosing the IgG4 backbone is contested by a more recent study reporting anti-tau antibody IgG isotype had no impact on pro-inflammatory cytokine secretion by primary human microglia (Zilkova et al., 2020). The same study also found that IgG1 isotype more effectively promoted tau uptake by human microglia than the IgG4 isotype (Zilkova et al., 2020). Furthermore, several experimental studies have shown that antibodies promote microglial degradative capacity for tau (Andersson et al., 2019; Funk et al., 2015; Luo et al., 2015). Although antibodies might also facilitate tau degradation by neuronal lysosomes (Congdon et al., 2013), it may be more favorable to design therapies that direct tau aggregates away from neurons and toward microglia. More work is required to elucidate the overall therapeutic mechanism of anti-tau mAbs, so that more effective treatments may be developed.

Therapeutics that reduce Mapt gene expression, such as antisense oligonucleotides (ASOs), are currently in clinical trials (e.g., NCT03186989). Mapt-targeting ASO treatment in tauopathy mice reduces existing tau aggregate pathology (DeVos et al., 2017). Theoretically these therapies may act primarily intracellularly to restrict replication of tau multimers which one study suggests may play a larger role than spread in late-stage tau pathological progression (Meisl et al., 2021). However, microglial clearance of extracellular tau may also contribute to its therapeutic efficacy. Notably, when the Mapt gene is silenced in the rTg4510 mouse model of tauopathy, the presence of Aβ pathology attenuates tau clearance after tau gene silencing compared to mice that only exhibit tau pathology (DeVos et al., 2018b) which could be due to disrupted microglia endocytosis due to Aβ pathology (Hickman et al., 2008). These data suggest that microglia may play a role in therapeutic tau clearance after Mapt gene silencing. Future studies will need to investigate the role of microglial endocytosis in tau clearance after Mapt gene silencing.

8. Concluding remarks

Herein we have discussed the importance of microglia in tauopathies and ways in which these cells may help or harm in the context of future tau-targeted therapies. There are currently no disease-modifying treatments for tauopathies, and halting tau-spread may be one of the most effective strategies for preventing widespread neurodegeneration. Ultimately, tau-based therapeutic strategies aimed at eliminating or mitigating cognitive impairment will be most efficacious in prodromal, asymptomatic, or early disease phases, before the spread and degree of tau pathology leads to irreparable cell loss and atrophy. Therefore, understanding factors that influence the propagation or restriction of tau pathology, such as microglia, is vital for developing effective treatments for tauopathy. Theoretically, promoting tau clearance by inducing microglial endocytosis and degradation of tau could be an effective therapeutic strategy. However, microglial endocytosis of tau may not always be protective, as improper processing of tau could enhance its propensity for spread (Asai et al., 2015; Clayton et al., 2021; Hopp et al., 2018). Future therapies will want to avoid triggering microglial pro-inflammatory cytokine cascades, improper proteolysis of tau, and exosome release. Anti-tau mAbs afford a means to enhance microglial degradation of extracellular tau (Andersson et al., 2019; Funk et al., 2015; Luo et al., 2015), but the microglial contribution to tau clearance therein is still unclear. Determining outcomes that potentiate anti-tau therapeutic efficacy, via microglia or other mechanisms, is paramount to successful design of future therapies. We present a model for microglia-tau interactions in Fig. 1, acknowledging unanswered questions remain. Our current understanding suggests that microglia act as both friend and foe within the context of tauopathy. Overall, promoting microglia beneficial functions will be key to harnessing their therapeutic potential in tauopathy. In adapting microglia’s therapeutic potential, the approach of handling these cells may be akin to a common approach in life: ‘to keep your friends close, but your enemies closer’.

Fig. 1.

Possible pathways by which tau seeds are internalized and processed by microglia. Black arrows: Tau seeding and spread pathway – Tau seeds are released from the axon terminal as either free or EV tau. Released tau is taken up by the postsynaptic neuron and internalized seeds serve as a structural template for native tau to misfold and aggregate. Blue arrows: Tau endocytic pathways – Free tau, EV tau, mAb-bound tau, or tau within dendritic spines can be taken up by various endocytic mechanisms. Single immune complexes of tau and anti-tau mAb are endocytosed via FcγRs and most likely involve CME. C1q, Axl/Mer and Gas6 have been implicated in the phagocytosis of dendritic spines. Tau binds microglial HSPGs which are internalized via macropinocytosis. Microglia internalize EVs through macropinocytosis. The endocytosed tau seeds are transported via the nascent endocytic vesicle to the EE. The EE membrane invaginates to form a MVB and matures to form an LE. Green arrows: Tau degradative pathways – Tau may be degraded extracellularly by CatB released by a secretory lysosome or intracellularly upon fusion of the MVB/LE with a LY containing legumain. Red arrows: Tau endosomal escape/exosome release – Tau has been shown to escape ELs. Co-activation of TLR4 and P2X7R, and the downstream inflammasome pathway induces the release of exosomes from the MVB/LE. This causes the release of tau-containing exosomes in the context of tau pathology. Abbreviations: Bin1 = bridging integrator 1; CatB = cathepsin B; C1q = complement component 1 q; CME = clathrin-mediated endocytosis; EE = early endosome; EL = endolysosome; EV = extracellular vesicle; Exo = exosome, FcγR = Fc gamma receptor; HSPG = heparan sulfate proteoglycan; LE = late endosome; LY = lysosome; Macro = macropinocytosis; MVB = multivesieular body; NLRP3 = NOD-, LRR-, and pyrin domain-containing protein 3; P2X7R = purinergic P2X7 ionotropic receptor; Phago = phagocytosis; TLR4 = toll-like receptor 4.

Abbreviations:

MAPT

Microtubule associated protein tau

AD

Alzheimer’s disease

FTLD-tau

frontotemporal lobar degeneration tau

CBD

corticobasal degeneration

PSP

progressive supranuclear palsy

NFT

neurofibrillary tangle

EVs

extracellular vesicles

Aβ

amyloid β

PET

positron emission tomography

Iba1

ionized calcium binding adaptor molecule 1

MMSE

mini mental state exam

DAM

disease associated microglia

MGnD

neurodegenerative microglia

Gas6

growth-arrest specific gene 6

TNF

tumor necrosis factor

IL

interleukin

CSF1

colony stimulating factor 1

TLR

toll-like receptor

IFN

interferon

dsRNA

double stranded RNA

ApoE

apolipoprotein E

Trem2

triggering receptor expressed on myeloid cells 2

PLCγ2

phospholipase C gamma 2

LOAD

late onset AD

BIN1

Bridging integrator 1

TPSO

translocator protein 18 kDa

CSF

cerebrospinal fluid

LPS

lipopolysaccharide

WT

wild-type

mTREM2

murine TREM2

SASP

senescence-associated secretory phenotype

HSPG

heparan sulphate proteoglycan

LRP1

low density lipoprotein receptor related protein 1

IgG

immunoglobulin G

C1q

complement component 1q

mAb

monoclonal antibody

MFG-E8

milk-fat globule epidermal growth factor 8

PS

phosphatidylserines

ALA

α-linolenic acid

CME

clathrin-mediated endocytosis

GSK3β

glycogen synthase 3 β

CLAP

clathrin and AP-2-binding

APP

amyloid precursor protein

PS1

presenilin 1

HEK293T

Human embryonic kidney 293T

CSF1R

CSF1 receptor

AEP

asparaginyl endopeptidase

AQP4

astrocytic aquaporin 4

ASOs

antisense oligonucleotides