Microglia and amyloid plaque formation in Alzheimer ’ s disease – Evidence, possible mechanisms, and future challenges

Alzheimer ’ s disease (AD) is a neurodegenerative disease characterized by cognitive decline that severely affects patients and their families. Genetic and environmental risk factors, such as viral infections, synergize to accelerate the aging-associated neurodegeneration. Genetic risk factors for late-onset AD (LOAD), which accounts for most AD cases, are predominantly implicated in microglial and immune cell functions. As such, microglia play a major role in formation of amyloid beta (A β ) plaques, the major pathological hallmark of AD. This review aims to provide an overview of the current knowledge regarding the role of microglia in A β plaque formation, as well as their impact on morphological and functional diversity of A β plaques. Based on this discussion, we seek to identify challenges and opportunities in this field with potential therapeutic implications.


Introduction
Alzheimer's disease (AD) is a neurodegenerative disease that accounts for most dementia cases [1].The underlying pathology is characterized by abnormal amyloid beta (Aβ) accumulation and deposition in the form of Aβ plaques, as well as hyperphosphorylated tau accumulation and formation of neurofibrillary tangles (NFT) [2].These changes eventually lead to neuronal cell death and synapse loss, paralleling the onset of cognitive symptoms.In recent years, neuroinflammation has emerged as another core pathological feature of AD involving both Aβ-dependent and -independent mechanisms, and with microglia, the brain's resident immune cells, being a major player [3].
Aβ peptides are generated through sequential proteolytic cleavage of the membrane protein amyloid precursor protein (APP).First, β-secretase cleaves off the extracellular part of APP.The remaining transmembrane part is further cleaved by the γ-secretase complex, which contains presenilin 1 and 2 (PSEN1 and PSEN2), to release Aβ peptides of varying length.Aβ1-40, Aβ1-42, and Aβ4-42 are the most abundant Aβ isoforms in AD brain [4] and the most hydrophobic of these Aβ peptides, Aβ1-42, is especially prone to aggregate into βsheet conformations to form oligomers, protofibrils, and fibrils which are found in AD brain.APP as well as the processing machinery are expressed by a wide variety of cell types in the CNS.Of these, Aβ is produced in large amounts by excitatory neurons [5,6] but also by oligodendrocytes [5,7] and probably to a lesser extent by glial cells.Aβ accumulation elicits a profound response by microglia which is paralleled by Aβ clearance (uptake and degradation of Aβ), as well as plaque formation and interaction (Fig. 1).The initial microglial response may be beneficial, and contribute to Aβ clearance, but excessive microglia activation including inflammatory responses may cause neuronal damage and disease progression at later stages.
Genetic variants with varying penetrance and population minor allele frequency play a substantial role in the development of AD.Common genetic variants, which individually have low penetrance, are associated with late-onset AD (LOAD) which accounts for most AD cases where symptoms appear after the age of 65.Recent genome-wide association studies (GWAS) have identified more than 70 risk loci for LOAD [8][9][10][11] including for example TREM2, CD33, MS4A6A, ABCA7, and SORL1.Interestingly, most of these are genes predominantly expressed by microglia [12,13].This highlights the central importance of microglia, and emphasizes the need to understand their role in AD.
Here, we will review the role of microglia in Aβ plaque formation as well as potential mechanisms which may influence plaque seeding, plaque compaction, and plaque isolation and highlight remaining challenges.

Different types and stages of Aβ plaques and the current knowledge on their "activities"
The accumulation of Aβ in the brain follows a well-described spatial-temporal pattern from neocortex to subcortical regions, including the hippocampus [14].The γ-secretase-cleaved peptide is monomeric but can adopt aggregated conformations to form short fibrillar oligomers, globular nonfibrillar oligomers, and amyloid fibrils.Aβ self-aggregation is required for toxicity [15].Especially soluble nonfibrillar oligomers have been suggested to possess high neurotoxic activity [16][17][18][19][20], which correlates with smaller size and high surface hydrophobicity.Various Aβ plaque morphologies have been identified in the AD brain and can be crudely categorized into diffuse and fibrillar plaques [21].Fibrillar plaques are tightly associated with AD and comprise the typical dense-core plaques.An important neuropathological feature of Aβ plaques is the presence of neurodegenerative activity.These neuritic plaques( [22], reviewed in [23]) which occur at later stages of AD contain degenerating axons and dendrites, referred to as dystrophic neurites, and correlate with disease severity.Additional plaque types include dense-core coarse-grained plaques [24] and diffuse cotton-wool plaques which are abundant in patients with certain PSEN1 mutations [25,26].Collectively, amyloid pathology is well-described along the AD continuum but the functional consequences and neurotoxic properties of the diverse Aβ aggregates and deposits remain less clear.

Microglia and their roles in Aβ plaque formation
Due to their close association with Aβ plaques, microglia have been suggested to be causally involved in various steps of Aβ plaque formation, including plaque seeding, plaque compaction, and plaque isolation (Fig. 1).The role of microglia in Aβ plaque formation has been studied in various mouse models of AD which exhibit differential patterns of pathological progression [27].Initial studies in APP/PS1 mice using pharmacogenetic microglia ablation by introducing a suicide gene under control of a microglial promoter (CD11b), which induced astrocyte activation, before or upon Aβ plaque development did not find differences in plaque load or neuritic dystrophy [28].Later studies have primarily used colony stimulating factor 1 receptor (CSF1R) inhibitors to efficiently deplete microglia from mouse brain.In 5xFAD mice, depletion of microglia before AD onset reduced Aβ plaque seeding [29][30][31], in particular the formation of dense-core plaques within cortical regions [29] and neuritic plaques [30], and improved cognition [30].Another study has found that microglia depletion before plaque deposition increased neuritic dystrophy and induced a shift from compact to more diffuse plaques in 5xFAD mice [31].Interestingly, microglial repopulation led to Aβ plaque remodeling, resulting in more compact plaques in microglia-repopulated regions [31].Microglia depletion in 5xFAD and APP/PS1 mice upon amyloid plaque establishment on the other hand did not alter Aβ plaque load [32,33].Similarly, microglia depletion in 3xTg-AD mice upon Aβ plaque establishment did not alter plaque load but improved cognition [34].Inhibition of microglial proliferation in APP/PS1 mice resulted in improved memory performance and reduced synaptic degeneration despite no difference in the number of Aβ plaques [35].Microglia-deficient mice show a profound shift from parenchymal Aβ plaques to cerebral amyloid angiopathy (CAA).CAA involves the accumulation of Aβ within blood vessels in the brain, which is accompanied by hemorrhages and premature death [36].CAA appearance has also been described in mouse models of microglia depletion [29].This suggests that microglia play a role in regulating vascular health and CAA, a prominent AD copathology [37].Collectively, these studies further support the idea that microglia might be involved in the different steps of Aβ accumulation in mice but also suggest that there might be parallel mechanisms influencing the disease course such as synaptic removal by microglia, and that the exact role of microglia might depend on the disease stage.
Interestingly, plaque-associated microglia remain dynamic and internalize injected amyloidbinding dye (methoxy-X04) at high rates [38] suggesting a role for microglia in plaque maintenance.In addition, the clustering of microglia around plaques has been suggested to exert a barrier function that increases plaque compaction while decreasing neuritic dystrophy [39][40][41].
Microglia require functional triggering receptor expressed on myeloid cells 2 (TREM2) to acquire the typical disease-associated microglia (DAM) signature which has been observed in mouse models of AD [42,43] and to a lesser extent in human AD brain [44,45].TREM2 controls key cellular functions in microglia such as phagocytosis [46], migration [47], and modulates inflammatory signaling [48].TREM2 binds various ligands, including Aβ, and

J o u r n a l P r e -p r o o f
Journal Pre-proof exerts its functions through downstream signaling cascades (reviewed in [49]).Hypofunctional TREM2 variants are associated with increased risk of LOAD [50,51].Deletion of Trem2 in mouse models of AD reduced initial Aβ plaque accumulation [52,53], similar to microglia depletion, but promoted plaque formation at later stages [52].Interestingly, NLF mice expressing the R47H TREM2 risk variant showed a selective accumulation of very small plaques compared with NLF mice [54].Microglial processes, which highly express TREM2, surround early Aβ plaques promoting their insulation and compaction [41].Interestingly, this has been shown to be markedly decreased in Trem2deficient mice [41,55], in mice expressing the R47H Trem2 variant [56], and in humans with R47H TREM2 mutations [41], resulting in less compact plaques and more severe neuritic tau hyperphosphorylation and axonal dystrophy.In later stages of pathology, however, TREM2 signaling has been shown to promote Aβ deposition into established plaque structures [40,57].Another study used overexpression of non-cleavable TREM2 in APP23/PS45 mice which increased numbers of small plaques, but not large plaques, and aggravated neuroinflammation compared to control mice at 3-months of age [58].Further supporting these findings, overexpression of human TREM2 in 5xFAD mice increased compact inert Aβ plaques while decreasing filamentous Aβ plaques and ameliorated behavioral deficits [59].These results suggest a dual role of TREM2.In earlier stages, TREM2 prevents plaque formation and seeding, likely through the ingestion and degradation of soluble Aβ.In later stages of the disease, TREM2 promotes the deposition of Aβ into compacted, potentially less harmful dense core plaques [40] which limits exposure of neurons to toxic Aβ [57].
In contrast to the studies discussed above, Aβ plaque seeding was found to be increased in the absence of functional TREM2.This was accompanied by decreased microglial clustering around newly formed plaques as well as reduced plaque-associated apolipoprotein E (ApoE) [60].Interestingly, injection of Trem2 knockdown antisense oligonucleotides (ASOs) reduced Aβ plaque load by half in 10-month-old APP/PS1 mice [61].However, ASO injection had no effect on Aβ plaque load at pre-plaque (4-month-old) and early-plaque stages (7month-old) [61].
In addition, agonistic anti-TREM2 antibodies, which have been developed to therapeutically stimulate protective TREM2 signaling (reviewed in [62]), have been shown to ameliorate Aβ plaque deposition in mouse models of amyloidosis [43,63,64].Finally, sTREM2 injection into brains of 5xFAD mice has been shown to reduce plaque load [65] and higher sTREM2 CSF levels seem to exert a protective effect in humans (reviewed in [66]).The spleen tyrosine kinase (SYK) coordinates downstream functions and signaling of several receptors including TREM2, CLEC7A, and CD33.As such it is not surprising that activation of Syk in microglia has been reported in the presence of AD pathology [67].Microglia-specific Syk deletion in 5xFAD mice increased Aβ plaque load [68,69].These plaques were found to be less compact and instead more filamentous [68] or dynamic [69] in nature compared to 5xFAD control mice.This was accompanied by a decrease in plaque-associated microglia, less Aβ phagocytosis by microglia, and impaired DAM transition [68,69].Conversely, injection of an antibody against CLEC7A, which directly activates Syk, increased number of

J o u r n a l P r e -p r o o f
Journal Pre-proof dynamic and inert Aβ plaques while it decreased number of filamentous plaques without affecting overall plaque load in TREM2 R47H 5xFAD mice [69].
A recent study investigated the effects of genetic ablation of the receptors Axl and Mer, which are highly expressed by microglia, on amyloid pathology in APP/PS1 mice [70].Double knockout mice developed fewer dense-core plaques and in turn had about 50 percent more diffuse cotton-wool plaques and increased cerebral amyloid angiopathy [70].Collectively, these studies suggest that TREM2 as well as other receptors and the associated downstream pathways regulate microglial responses to Aβ in AD including the association of microglia with plaques, the exact consequences of these alterations however remain less clear.
ApoE is a lipid-binding protein in lipoprotein particles, which participates in lipid transport.The APOE4 allele remains the strongest genetic risk factor for LOAD in humans and ApoE isoforms have been shown to exert differential effects on Aβ metabolism and microglial responses to Aβ plaques [71][72][73].For example, ApoE4 has been shown to accelerate Aβ plaque seeding in mice [74].Furthermore, ApoE is strongly upregulated by microglia in the presence of amyloid pathology [42,71].ApoE deficiency, independent of isoform, in APPPS1ΔE9 and APPPS1-21 mice reduced fibrillar plaque deposition, reduced plaque compaction, reduced clustering of microglia around plaques as well as increased neuritic dystrophy [75,76].In contrast, expression of ApoE3 in microglia in ApoE knockout mice increased clustering of microglia around plaques and reduced amyloid pathology and neuritic dystrophy whereas expression of ApoE4 led to impaired microglial responses by impairing lipid metabolism [72].
Collectively, the current literature suggests a complex role for microglia in Aβ plaque formation with an early function in elimination of Aβ and a late function in compacting and containment of the plaques (Fig. 1).Although there still is a significant volume of conflicting data, possibly due to differences in model systems as well as differences in the stages of amyloid pathology, reduced ability of microglia to envelop amyloid deposits is most often found to be accompanied by an increase in less compact plaques.Local responses around these deposits are associated with more severe neuritic tau hyperphosphorylation and axonal dystrophy.

Mechanisms for microglia-mediated modulation of Aβ plaque formation
Microglia are professional phagocytes and possess numerous receptors and mechanisms to internalize Aβ.The interaction with Aβ can activate microglia through for example CD36 and toll-like receptors (reviewed in [77]) and drive a strong neuroinflammatory response.Importantly, microglial activation has been found to correlate with amyloid pathology in the living human brain [78].The overabundance of Aβ in AD likely triggers cellular pathways which might contribute to formation of Aβ plaques (Fig. 2).
Plaque formation has been shown to depend on the presence of endocytosis-or phagocytosis-competent cells [79].Internalized Aβ can be routed to multivesicular bodies where further Aβ fibrilization can lead to penetration of the vesicular membrane and cell

J o u r n a l P r e -p r o o f
Journal Pre-proof death [79].The acidic pH in the endocytic and lysosomal pathway is known to promote fibril formation [80].It has further been shown in 5xFAD mice that Aβ accumulation within lysosomes can induce microglial cell death which in turn contributed to plaque formation and growth [81].Moreover, accumulation of intraneuronal Aβ and consequent neuronal cell death have been suggested to be the origin of neuritic plaques [30,82].Remaining cell debris and dying cells that expose phosphatidylserine (PS) can also be taken up by microglia via receptors such as TREM2 and TAMs.In addition, microglia might contribute to seeding of new Aβ plaques by transporting phagocytosed Aβ elsewhere, for example to sites of injury [83].
The NLRP3 inflammasome is an important innate immunological sensor in microglia [84] and has been shown to be activated in AD and by Aβ [84,85].Microglial Aβ phagocytosis can lead to lysosomal damage which in turn activates NLRP3 [86].Nlrp3 deletion in APP/PS1 mice reduced interleukin-1β activation, decreased Aβ plaque load, and improved cognitive performance [85].These data suggest that Aβ not only induces inflammation, but the inflammatory response in turn amplifies Aβ plaque accumulation.Inflammasome-activated cells release ASC-specks upon pyroptosis, which in turn can activate the inflammasome in neighboring microglia.It has further been shown that ASC-specks can act as nucleation sites for Aβ to further aggregate and fibrilize [87].Finally, additional mechanisms may be involved in microglia-mediated promotion of Aβ plaque seeding and formation, including apoE aggregation as a primer of Aβ aggregation [88].Ingested Aβ enters the endo-lysosomal pathway where it can be degraded but also further aggregate due to a pH decrease.This can lead to lysosomal damage which in turn activates the NLRP3 inflammasome and ASC speck formation and release through pyroptosis.ASC specks, exocytosis of aggregated Aβ material, and deposition through cell death might contribute to plaque seeding.

Role for infections and inflammation in Aβ plaque formation
There is substantial evidence for a role of inflammation and infection in AD, but less is known about how the immune response influences Aβ plaque formation specifically.While the early innate immune response by microglia aims to protect the brain, an excessive chronic neuroinflammation, as seen in neurodegenerative diseases, can be detrimental and further drive disease progression [89].
Repeated LPS injections before Aβ plaque deposition have been shown to decrease plaque load in APP23 mice and has been suggested to be due to epigenomic and transcriptomic alterations in microglia as part of innate immune memory [90].LPS priming before plaque deposition decreased microglial activation in response to Aβ, increased Aβ internalization, and decreased Aβ plaque size in prefrontal cortex in the 5xFAD AD mouse model [91].Similarly, intrahippocampal LPS injections reduced Aβ load in APP/PS1 mice [92].Mechanistically, this is consistent with the idea that primed microglia have enhanced phagocytosis and degradation activity, hence eliminating Aβ and leaving limited peptide material for cerebral accumulation.Consistent with this, other studies have shown that overexpression of tumor necrosis factor α (TNFα) [93] or interferon gamma (IFNγ) [94] increased microglial phagocytotic activity and reduced plaque deposition.On the other hand, inflammation and inflammatory microglial cell death can also promote Aβ plaque formation, as already indicated in the previous section.
The idea that virus infections can contribute to AD development was proposed over 40 years ago [95].A lot of the work has focused on herpes simplex virus type 1 (HSV1) which is an endemic human neurotropic herpesvirus that infects over 79% of adults worldwide [96].HSV1 infections lead to the establishment of lifelong latent infection of sensory neurons of the trigeminal ganglion (TG).Periodic HSV1 reactivation results in virus spread to the oral mucosa and the central nervous system (CNS) [97,98].Viral activity in the aging brain has been linked to genetic (e.g., APP processing genes), clinical (e.g.clinical dementia rating), and neuropathological (e.g., neuritic plaque density) aspects of AD [99].Our group has recently discovered that HSV1 infection suppresses TREM2 expression in iPSC-derived cell models as well as in the mouse brain [100].Furthermore, additional AD risk genes have been linked to HSV1 infection (reviewed in [101]).Some evidence indicates that CNS HSV1 infection promotes seeding of Aβ deposits in neural 3D cultures and in young 5xFAD mice [102,103] whereas other studies did not find a link between HSV1 and Aβ deposition [104,105].An additional study of mild recurrent HSV1 infection reported an HSV1-induced AD-like phenotype with intraneuronal Aβ accumulation and Aβ plaque formation [106].The innate antiviral immune response which highly depends on type I interferon (IFN-I) production by microglia is critical to restrict HSV1 infection in the CNS [107,108].Recent single cell sequencing analyses provided insight into the diverse microglial activation states and subpopulations in human and mouse AD brain one of which is characterized by high expression of IFN-stimulated genes (ISGs) [109].Interestingly, microglia expressing ISGs were found to surround nucleic acid-containing neuritic plaques in postmortem brains of patients with AD [110].Short-term IFN-I receptor blockade in 5xFAD mice reduced microgliosis and synapse loss but did not affect plaque load [110].
Altogether, further mechanistic studies, specifically evaluating the role of microglia, are needed to pinpoint if there is a causal relationship between immune activation and the different stages of amyloid pathology.Here, system inflammation triggered by for example viral infection or LPS must be distinguished from Aβ-induced inflammatory processes.Indepth characterization of the microglial cytopathic response to repeated recurrent HSV1 infections may reveal a microglia-driven modulation of Aβ plaque deposition.

Outlook
Numerous genetic and experimental studies have highlighted the central importance of microglia in AD.However, potential conflicting functions of microglia with regards to Aβ plaque pathology remain.As discussed above, there is evidence to suggest that microglia may actively be involved in various steps of Aβ plaque formation depending on several factors, including the stage of AD progression.
Although microglial activation is associated with AD pathology, it has been difficult to dissect whether and when this process is damaging or protective.It is likely that common genetic risk variants [11] functionally translate into differences in how microglia handle and react to Aβ pathology, but we are just beginning to understand the biological functions of all these variants.The results from numerous studies discussed in this review suggest that dysfunctional microglia might be less able to contain Aβ within compact plaques with less neurotoxic activity.
Anti-Aβ immunotherapies constitute another facet of microglia-Aβ plaque interaction which was not covered by this review but which we would like to mention.During anti-Aβ immunotherapies, microglia are believed to take part in Aβ clearance via FcγR-mediated phagocytosis of antibody-Aβ complexes [111][112][113][114]. Furthermore, evaluation of long-term effects of anti-Aβ immunotherapy identified that initially activated microglia convert into a deactivated state which might impair their ability to reactivate later [115].To be able to assess these dynamics of microglial activation in patients, microglia-specific biomarkers [116] are needed.
Most of the work on this topic has been done using transgenic mouse models of AD and given the obvious differences between human and mouse microglia [117][118][119], the question remains how translatable these results are.Novel model systems to study Aβ plaque formation such as human induced pluripotent stem cell-derived models [120,121] and J o u r n a l P r e -p r o o f Journal Pre-proof chimeric mouse models [122] might overcome this challenge.Furthermore, a thorough investigation of plaque quality in addition to plaque quantity including the interrogation of functional differences and toxicity between plaque types and the different stages of plaque formation is needed.Such work may reveal that microglia play multiple roles in the development of AD, including both beneficial and deleterious.To get to the bottom of this question, there is a need for indepth understanding of both the structure-functional relation of different types and stages of Aβ plaques as well as the dynamic impact on microglia activity during AD development.

Figure 1 .
Figure 1.Contribution of microglia to different stages of plaque formation.Microglia can take up and degrade Aβ to some extent which involves phagocytic receptors such as TREM2.Through several mechanisms (described below) microglia might be involved in plaque

Figure 2 .
Figure 2. Mechanisms for microglia-mediated modulation of Aβ plaque formation.Microglia ingest Aβ primarily via receptors which directly bind Aβ and through uptake of Aβ-