Microglia and its genetics in Alzheimer’s Disease

Abstract

Alzheimer’s Disease (AD) is the most prevalent form of dementia across the world. While its discovery and pathological manifestations are centered on protein aggregations of amyloid-beta (Aβ) and hyperphosphorylated tau protein, neuroinflammation has emerged in the last decade as a main component of the disease in both pathogenesis and progression. As the main innate immune cell type in central nervous system (CNS), microglia play a very important role in regulating neuroinflammation, which occurs commonly in neurodegenerative conditions including AD. Under inflammatory response, microglia undergo morphological changes and status transition from homeostatic to activated forms. Different microglia subtypes displaying distinct genetic profiles have been identified in AD, and these signatures often link to AD risk genes identified from the genome-wide association studies (GWAS), such as APOE and TREM2. Furthermore, many of AD risk genes are highly enriched in microglia and specifically influence the functions of microglia in pathogenesis, e.g. releasing inflammatory cytokines and clearing Aβ. Therefore, building up a landscape of these risk genes in microglia, based on current preclinical studies and in the context of their pathogenic or protective effects, would largely help us to understand the complexed etiology of AD and provide new insight for the unmet need of effective treatment.

Introduction

Alzheimer’s Disease (AD) is a chronic, devastating neurodegenerative disorder characterized by progressive impairment of behavioral and cognitive functions. It is the most common subtype of age-related dementia, affecting ~50 million individuals worldwide [1, 2]. Of note, the global prevalence of AD is still increasing and predicted to double every 20 years [2, 3]. As a major contributor to death and disability among the aging population, it places a formidable burden on the health-care systems [1, 4]. Despite enormous efforts in drug development, no effective cure currently exists [5]. Thus, there is an urgent need for an in-depth understanding of the etiology and pathogenesis of AD, which will facilitate the development of novel disease-modifying therapeutic strategies to prevent or slow down the progression of AD.

AD has cardinal pathological features including extracellular Aβ depositions often in the form of plaques, and formation of intracellular neurofibrillary tangles (NFTs) from hyperphosphorylated tau [5]. However, it is a heterogeneous disorder with complex etiology, particularly in the vast majority of sporadic cases. Accumulating evidence demonstrated that gliosis and neuroinflammation not only strongly correlate with AD pathological hallmarks, but also exert a causal role in disease pathogenesis [6, 7]. Instead of being a mere bystander activated by emerging senile plaques and NFTs, these processes may contribute as much to neuronal dysfunction and AD pathogenesis [6, 7]. Therefore, a comprehensive characterization of gliosis and neuroinflammation will provide critical insights into the underlying molecular mechanisms and identify potential effective targets for therapeutic intervention in AD.

Microglia, the brain-resident myeloid cells, play pleiotropic physiological roles in both developing and adult central nervous system (CNS) [8], by promoting phagocytic clearance and providing trophic support for brain homeostasis and tissue repair [9]. Aberrations in the cellular properties and functions of microglia contribute to a list of neurological conditions, ranging from neurodevelopmental disorders to neurodegenerative diseases, such as autism and AD, respectively [8, 10]. In particular, evidence from clinical and pre-clinical studies has clearly demonstrated the critical roles of microglia in the pathogenesis of AD [11]. Not only these data indicate a microglial diversity in AD, such as disease-associated microglia (DAM), plaque-associated microglia (PAM), and dark microglia [12–14], but also a variety of AD risk genes, such as TREM2, CD33, APOE4, MS4A, and PLCG2, which are preferentially or selectively expressed in microglia [11, 15]. Moreover, these risk genes are correlated with amyloid plaque burden and disease severity [16]. For example, mutations or genetic variants of specific genes encoding important immune receptors such as TREM2 (triggering receptor expressed on myeloid cells 2) and CD33 are differentially associated with risks for developing AD [17, 18]. These findings reveal the complexed interactions between risk genes and other signaling pathways in microglia, which might be a key to understand the neuroinflammation in AD.

In this review, we explored recent advances regarding microglial genetics and how phenotypes influence the pathogenesis of AD. Specifically, the risk genes highly expressed in microglia and their pathophysiological implications in AD pathology were emphasized. Further elucidation of the genetic mechanisms that regulate microglial activation and function is particularly important, which will lead to uncovering potential new targets for future therapeutic intervention.

Historical Perspectives

The name of microglia came from more than a century of research [19]. In 1856, “neuroglia” was defined by pathologist Rudolf Virchow for the first time based on their spacing around neurons and the ability to hold together like glue [20], while in fact, “neuroglia” refers more as a macrophage population comprising of astrocytes and oligodendrocytes. In the end of 1800’s, W. Ford Robertson introduced the term of “mesoglia” to describe mesoderm-derived phagocytic elements in the nervous system that has a distinct origin from “neuroglia” [21] and claimed that “mesoglia” appear differently from “neuroglia” because they seemed to be phagocytic based on platinum substitution modification of Golgi technique [22]. Later, Santiago Ramon y Cajal successfully used the gold chloride sublimate method to separate astrocytes from “apolar cells” and referred to “mesoglia” as “the third element of nervous system” [23]. In 1919, Pio del Rio-Hortega, a student of Cajal, found that the apolar cells consisted of two different cell types during histological examinations using the ammoniacal silver carbonate method [24], for which he named “microglia” and “interfascicular” oligodendrocytes to describe the non-neural, non-astrocytic “third element” [24].

Although the microglia have been studied for decades, their developmental origin is still debated. Evidence from initial studies, including those performed from Rio-Hortega, indicated that microglia arose from embryonic progenitors [24]. Based on similarities in morphology and phagocytic features, microglia were believed to be of myeloid origin like other macrophage populations, and originate from blood-circulating monocytes [19]. In the last 50 years, the Yolk Sac (YS) origin of microglia has been extensively discussed. By applying markers Mac-1 and F4/80 to label the cells dissociated from the forebrain, Alliot et al. found that embryonic brain parenchyma contained potential progenitors for microglia in C57BL/J6 mice [25]. Further investigations from the same group suggested that YS-derived microglia progenitors migrate into head mesenchyme and brain rudiment at embryonic day (ED) 8, where they further proliferate [26]. This theory of YS origin has been confirmed with a wide range of tracing techniques, including fate-mapping. For example, Ginhoux et al. used in vivo lineage tracing to demonstrate that Runx1-positive mouse myeloid progenitors in the YS seed the brain between ED8.5 and ED9.5 [27]. As Ncx^−/− embryos, which have normal YS hematopoiesis but lack heartbeat and normally functional blood circulation [28], failed to develop any microglia at ED9.5 [27], which suggested that Runx1⁺ myeloid progenitors migrate to brain from YS through blood circulation before the establishment of the blood-brain barrier. Taken together, the endogenous microglial population is derived predominantly from primitive myeloid progenitors in the YS during early development, and maintained by local self-renewal, which is independent of hematopoiesis under pathophysiological conditions [29]. The potential contribution of peripherally derived microglial/macrophage-like cells in neurological diseases is more complicated and remains to be explored [30, 31].

From Homeostasis to activation

Microglia are the sentinels of the brain’s innate immune system, continuously monitoring the surrounding microenvironment [32]. Under normal physiological conditions, they exhibit a highly “ramified” morphology with a small soma; while in response to brain injury or in disease conditions, they can be activated rapidly and swiftly change to an “amoeboid” phenotype, which is characterized by shortened and extensively branched processes, as well as hypertrophy of the cell body [33]. There is a well-established classification system for various phenotypes microglia (Figure 1). M0 microglia, also known as the “resting” microglia, maintain the homeostasis of the CNS [33], which can further acquire either M1 or M2 phenotypes upon activation. The classical M1-like microglia exhibit cytotoxic and pro-inflammatory activities [6], and carry signature markers such as CD16, CD32, iNOS, IL-1β, IL-6 and TNF-α in mouse microglia and CD40, CD86 in human microglia [34–37]; in contrast, the alternatively activated M2-like microglia express signature markers such as CD206 (also known as macrophage mannose receptor 1), YM-1, FIZZ1, IL-10 as mouse marker and Arginase 1 (Arg-1), TREM2, and TGF-β for both human and murine. In addition, M2-like microglia often promote inflammation resolution and tissue regeneration [38–40]. According to distinct activation mechanisms and functions, M2-like microglia can be further divided in three subtypes, M2a, M2b, and M2c [41]. The M2a phenotype, typically stimulated by exposure to IL-4 and IL-13, is closely associated with anti-inflammatory and restorative processes [42]. By contrast, the M2b phenotype, often induced by the fusion of Fc gamma receptors (FcγRII) and immunoglobulin G (IgG) complexes, regulates selective phagocytosis and inflammatory responses [43]. Alternatively, the M2c phenotype, which occurs in response to IL-10 and TGF-β, plays crucial roles in immunosuppression, phagocytosis and tissue remodeling [44]. It is noteworthy that microglial phenotypes may be intermittent and dynamic in different pathophysiological conditions, and therefore further elucidation of these cellular and molecular changes will provide insights into the fundamental functions of microglia in CNS injuries and diseases.

Figure 1. Microglia subtypes and activation process.

Microglia can be activated from homeostatic status (M0) to activate forms (M1 or M2), which can be further categorized into different subtypes. After activation, PAM expresses abundant of IL-1β and shows characteristic gene profile changes. Increased intracellular Ca²⁺ contributes to its hyperreactive immune response. DAM can be activated by various damaged-associated molecular patterns (DAMPs), which impair the phagocytosis and proteostasis functions. Dark Microglia can be activated by aging, chronic stress, AD pathologies or CX₃CR1 deficiency. TREM2 is increasingly expressed in dark microglia near Aβ plaques. While the M2 subtypes, including M2a, M2b and M2c, in general are anti-inflammatory and promote proteostasis and repair. The inflammatory cytokines and neurotoxic factors produced by activated M1 microglia lead to neuro damage, which can stimulate more Aβ plaques made by neuron and go back to activate microglia further. Abbreviations: IL-1β: interleukin-1β; Aβ: β-amyloid; sAβ: soluble β-amyloid; PAM: Plaque-associated microglia; DAM: Disease-associated microglia.

Microglial transcriptomics in AD

Accumulating evidence has revealed microglia as an indispensable contributor to neuroinflammation and neurodegeneration [45], therefore understanding the diversity and subtype categorization of microglia is a key towards deciphering their fundamental roles in AD. Notably, microglial transcriptomes by bulk population sequencing have characterized the microglial diversity in age, neurodegenerative diseases and psychiatric disorders [46, 47]. However, the specific contributions from different microglia subtypes to brain pathologies remain unclear from bulk sequencing results, sometimes even controversial [33]. The advances in single-cell techniques, such as cytometry by time-of-flight mass spectrometry (CyTOF) as well as single-cell RNA sequencing, offer a good solution, as they detect cell types and intermediate cell states in an unbiased fashion. Such single-cell RNA technologies have revealed the spatio-temporal diversity of microglia under both homeostatic and pathological conditions in human and animal models [48]. For example, Hammond et al. discovered that there are at least 8 subtypes of microglia in the naïve mouse brain, and their transcriptomic changes are associated with age and gender [49], regional heterogenicity [50], and AD pathogenesis [51–53]. Therefore, in the following section, we will survey important microglial subtypes that are closely associated with AD at both pathological and transcriptomic levels.

Disease-associated microglia (DAM) is the most recently identified subtype of microglia [54], which has been detected in the 5xFAD transgenic mouse model by single-cell and single-nucleus RNA-seq analysis [52]. As reported, Keren-Shaul et al. sorted CD45⁺ immune cells from 5xFAD mouse brains and compared the gene profiling differences with control groups via massively parallel single-cell RNA-seq (MARS-Seq). In this study, five groups of immune cells including DAM were identified, which exhibited distinct cellular states in AD brains compared to that of wild-type controls. In addition, histological analysis confirmed that DAM is uniquely localized around intercellular Aβ plaque particles in both mouse and human AD samples, but are not present in healthy brain tissue [52]. More interestingly, the gene expression of TREM2, CTSD, TYROBP and APOE is significantly upregulated while the transcripts for P2RY12, P2RY13, CX3CR1, CD33 and TMEM119 are heavily downregulated in DAM. These findings indicate a potential crucial role of DAM in microglial phagocytic and/or endocytic functions, as well as inflammatory cascades [52].

Plaque-associated microglia (PAM) is known as a cell population of microglia which are stimulated by insoluble Aβ [54]. They can be detected in human AD brains by using immunohistochemical techniques [55]. Also, the transformed morphological features of PAM and their spatial distance to Aβ plaques was further revealed by 3D cell reconstruction in TgCRND8 mouse [56, 57]. Furthermore, Brawek et al. revealed substantial dysfunction of PAM in both APP23/PS45 [58] and APP/PS1 [59] mouse models, and demonstrated that aberrant intracellular calcium signaling may trigger release of toxic species from microglia in the plaque vicinity [60].

Dark microglia were discovered as a new subtype of microglia in APP/PS1 mice with electron microscopy (EM) by Bisht et al. [12] Ultrastructural analysis demonstrated that they are strikingly different from the other phenotypes of microglia. Dark microglia show signs of oxidative stress including condensed cytoplasm and nucleoplasm with “dark” appearance under electron microscopy [12]. In addition, they are rarely observed in healthy brains, but often detected in individuals with chronic stress, aging and AD. Besides morphological changes, dark microglia have a specific gene profile with downregulated of homeostatic markers such as Iba1, P2RY12 and CX3XR1 [54], and upregulation in TREM2 expression, as well as increased immunoreactivity to the microglial-specific antibody 4D4 near amyloid plaques [12, 61]. Currently, the pathophysiological contribution of dark microglia to AD, however, remains unclear and warrants further investigations [54].

Human Alzheimer Microglia (HAM) refers to a special subtype of microglia found in AD patients which exhibits a unique gene expression patterns when compared to normal human microglia [54] or DAM from mouse AD models [62]. Srinivasan et al. applied a novel method for sorting myeloid cells from frozen post-mortem specimens of superior frontal gyrus (SFG) and fusiform gyrus in human AD tissue, the gene expression changes observed by RNA-Seq in these myeloid cells were not similar with DAM profile in mouse AD models. They named this new profile in human Alzheimer’s microglia/myeloid cells as HAM signature, in which half of these changes are consistent with healthy human aging profile while the other half only altered in AD-related genes like APOE, highlighting the significant differences of myeloid cells between human diseases and mouse models [62]. The potential reasons of such distinct differences between HAM and DAM may be driven by different immune system responses, timeline of AD progression or specific effects of microglia activation in human and mice [54].

AD-Associated Risk Genes in Microglia

While most AD cases are late-onset and sporadic, genetic inheritance is still a major determinant, and over 40 genes associated with the disease [63, 64] or loci have been identified over the last decade [15, 65]. The increased AD GWAS projects and samples size lead to a larger scale of AD discovery as well as improving statistics power of AD relative variants. In this part, we will focus on a subset of AD risk genes that are enriched in microglia and discuss their potential molecular mechanisms in AD pathology in the context of microglia phenotypes. We are fully aware that most of AD associated SNPs are unique to the human genome, and cannot be simply modeled in mice, and the biological difference between human and mice is substantial. They differ not only at the nucleotide sequences level, but also on epigenetic regulations. Although most of these are human SNPs, their relevant murine models have offered substantial insights to their biological functions in microglia and relationship to AD.

APOE

Apolipoprotein E (APOE) is a lipid-binding protein that transports cholesterol and other lipids between cells and organs. The human APOE gene has three major alleles, ε2, ε3 and ε4, which confer amino acid variations at position 112 and 158. APOE3 (Cys-112 and Arg-158) is the most common isoform in population and considered as the “neutral” isoform, while APOE4 (Arg-112 and Arg-158) is the most significant genetic risk factor for late onset of AD (LOAD). Two copies of ε4 increase the likelihood of developing AD by >12 folds, and lowers the age of onset by nearly 17 years [66]. On the other hand, APOE2 (Cys-112 and Cys-158) is considered to be neuroprotective and associated with a nearly 50% of reduction in AD risk. Clinical and preclinical evidence has demonstrated an isoform-dependent (APOE4 > APOE3 > APOE2) and dose-dependent association with AD risks and pathophysiology [67], however, both APOE4 and APOE2 are linked to cerebral amyloid angiopathy (CAA)-related hemorrhage [68]. The role of APOE4 in AD can be attributed to both loss of protective function and gain of toxic function [66, 69]. In contrast to APOE3, APOE4 harms multiple cellular functions in increasing APP expression and Aβ production from neuronal cells [70, 71], decreasing Aβ clearance [72, 73], promoting Aβ aggregation [74, 75], reducing cerebrovascular functions and BBB integrity [76, 77], as well as impairing synaptic function [78, 79] and neuronal activities [71, 80], and causing brain atrophy [81, 82]. However, the exact mechanisms of APOE isoform-dependent actions on different brain cell types are not fully understood yet.

APOE is mainly produced by astrocytes in normal brain [66]. However, it is highly upregulated in DAM, as revealed by recent single cell RNA sequencing studies in both humans and mice [51–53]. In addition, using single-molecule fluorescence in situ hybridization (smFISH) in APP knock-in mouse model (App^NL-G-F) [83], Frigerio et al. found the microglia were enriched in Apoe transcripts, and the intensity of signals increased in cells located closer to Aβ plaques [84], suggesting that DAM are potentially the major local contributor of APOE around the plaques. Besides amyloid pathology, APOE also influences tauopathy [66, 67]. By comparing P301S tau transgenic mouse with human APOE knock-in (ε2, ε3 or ε4) mouse model, or with murine Apoe knock-out (EKO) mice, Holtzman’s lab found P301S/ε4 mice had significantly higher level of tau pathology and neuroinflammation, when compared with P301S/ε2, P301S/ε3 or P301S/EKO [85], demonstrating a direct link between APOE4 and tauopathy in AD. This phenotype attributes to a pro-inflammatory effect which promotes activation of both microglia and astrocytes [85], as well as the upregulation of serpin superfamily of protease inhibitors, Serpina3 in particular [86].

TREM2

As a member of the triggering receptors expressed on myeloid cells (TREM) family, TREM2 is mainly expressed in microglia and infiltrating macrophages in brain. TREM2 protein belongs to the immunoglobin superfamily and plays a critical role in the innate immune system. The extracellular ligands for TREM2 include various lipoproteins like low density lipoprotein (LDL), malondialdehyde-modified LDL (MDA-LDL), high density lipoprotein (HDL), apolipoprotein E and clusterin (CLU, also known as apoJ), as well as phospholipids, glycolipids, apoptotic neurons and even bacteria [87] (Figure 2). The transmembrane domain of TREM2 interacts with the DNAX-activating protein 10 (DAP10) and 12 (DAP12), which are adaptor proteins that mediate intercellular signaling through Phosphoinositide 3-kinase (PI3K) and Spleen tyrosine kinase (SYK), respectively [87, 88]. TREM2 can be cleaved by disintegrin and metalloproteinase domain containing proteins ADAM10 and ADAM17 and shred its ectodomain sTREM2, which will deactivate the TREM2/DAP12 signaling [87]. sTREM2 levels in the cerebrospinal fluid (CSF) are correlated with tau and microglia changes in early stage of AD, and can be potentially used as a fluid biomarker [89, 90].

Figure 2. AD risk genes and signaling pathways in microglia.

As a triggering receptor expressed on myeloid cells, TREM2 have various ligands from extracellular, including Aβ, LDL, HDL, apoptotic cells and ApoE. The combination of TREM2 and its ligand can phosphorylate ITAM of DAP12, then triggering SYK, PI3K/Akt and GSK3β downstream signaling. The interaction between APOE and TREM2 triggers the transcription regulation and microglia dysfunction. Microglia speck-like protein containing a CARD (ASC) and Capase-1 upregulate proinflammation cytokines IL-1β and IL-18. In the presence of sialic-acid, CD33 inhibits the TREM-dependent uptake of Aβ. LY294002, an inhibitor of PI3K/Akt pathway; Abi3 variants is associated with Aβ or tau aggregation in microglia and ABI3 may combine with CYFIP1 and WAVE2 to form wave regulator complex (WRC) and regulate microglia activation in AD. PLCγ2 can be phosphorylated by SYK and stimulate IP3, which then triggering Ca²⁺ signaling, modulating phagocytosis of microglia or activating NF-κB pathway. Abbreviations: Aβ: β-amyloid; LDL: low density lipoprotein; HDL: high density lipoprotein; MDA-LDL: malondialdehyde-modified LDL; DAP12: DNAX-activating protein; IL-1β: interleukin-1β; IL-18: interleukin-18.

Nonsense mutations in TREM2 and DAP12 are linked to Nasu-Hakola disease, a rare autosomal recessive condition of early onset frontotemporal dementia with leukoencephalopathy and basal ganglia calcification [91]. In 2013, a rare R47H variant of TREM2 (rs75932628) was identified as a risk allele for AD [18, 92], and the R47H mutation was predicted to be partial loss-of-function [18]. Data from NIMH AD Genetics Initiative Study and AD Sequencing Project indicated additional risk rare coding variants including R62H, D87N and H157Y, and a protective mutation R62C [93]. Yeh et al. used bio-layer interferometry (BLI) to test the effects of missense mutations on apolipoprotein and lipoprotein binding, and found R47H, D87N and R62H on hTREM2 all exhibited significant loss of binding abilities with LDL, CLU and APOE [94]. In addition, Atagi et al. confirmed R47H mutation significantly reduced the binding ability of TREM2 on all three isoforms of APOE [95].

During AD pathogenesis, microglia cluster around plaques forms a physical barrier limiting amyloid pathology and neurotoxicity [96, 97]. Microglia can efficiently uptake Aβ-LDL complex; however, this ability is dramatically inhibited in Trem2-KO microglia in vitro [94]. Although the amyloid phenotypes in Trem2-KO mice when crossed with different AD models were reported to be model-specific, age- and region-dependent [98], impaired microglia responses to amyloid are commonly observed regardless of model and disease stage [96–100]. The TREM2-dependent microglial endophenotype in vivo is consistent with transcriptomic changes at the single-cell level [51–53]. For example, Keren-Shaul et al. reported that the activation of DAM in AD requires a TREM2-dependent pathway in Trem2^−/−; 5xFAD mice [52], for example, the regulation of APOE to DAM requires TREM2-dependent pathway [101]. In addition, the TREM2-APOE signaling associated with neurodegeneration can be characterized by inhibition of homeostatic transcription factors such as PU.1, MEF2A, MAFB, SALL1 and SMAD3, and activation of pro-inflammatory transcription factors such as BHLHE40, TFEC and ATF3, and MIR-155 [101]. Therefore, targeting TREM-APOE axis may represents a novel approach in restoring microglial homeostasis in AD.

CD33

CD33, a member of the sialic-acid binding immunoglobulin-like lectins (SIGLETs), anchors on the cytomembrane of microglia, macrophages, myeloid progenitor cells and monocytes. The extracellular part of CD33 consists of a N-terminal immunoglobulin domain is responsible for sialic acid recognition, as well as a C2-type immunoglobulin repeat. Intracellularly, human CD33 has two conserved cytoplasmic tyrosine-based motifs: a membrane-proximal ITIM and a membrane-distal ITIM-like motif; while the murine form lacks the ITIM motif [102]. CD33 plays important roles in immune cells, including cell adhesion, endocytosis, and inhibition of cytokines release [103].

CD33 was found modulating microglia activation and inhibiting Aβ by regulating its expression and alternatively splicing [104]. GWAS has identified that rs3865444 and rs12459419 are the two main CD33 variants that confer the risk for AD [105–107]. The major allele rs3865444^C is linked to an increased risk of AD in Chinese, European, and North American populations [108] and is associated with cognitive decline in AD [109]. In contrast, the minor allele rs3865444^A exerts a protective effect against AD which is also associated with increased CD33 expression level [110, 111].

CD33 is enriched on the surface of microglia, and inhibits microglia uptake of Aβ in the presence of sialic-acid [112, 113]. The amount of CD33-immunoreactive microglia positively correlates with Aβ burden in the same brain region [113]. Interestingly, TREM2 and CD33 exert opposite effects on Aβ pathology and microglia activation in 5xFAD mice. Compared to the exacerbated pathologies in Trem2^−/−;5xFAD mice, CD33 deletion in Cd33^−/−;5xFAD mice promoted microglia phagocytosis and neuroprotection [17]. This protection is abolished in the double knockout Cd33^−/−;Trem2^−/−;5xFAD mice, suggesting that TREM2 acts downstream of CD33 in modulating microglial pathology in AD [17].

PICALM

Phosphatidylinositol binding clathrin assembly protein (PICALM) regulates the formation of the clathrin lattice during endocytosis. Multiple single nucleotide polymorphisms (SNPs) within or close to the PICALM gene, such as rs3851179, have been associated with LOAD [114, 115]. It was first functionally linked to AD for its key function in mediating Aβ clearance across the blood-brain barrier [116]. Specifically, PICALM regulates endothelial internalization of Aβ that is bound to the low-density lipoprotein receptor-related protein-1 (LRP1), guides Aβ trafficking to Rab5-positive early endosomes and Rab11-positive sorting endosomes, and eventually leads to Aβ transcytosis from brain to circulation. Therefore, animal model with Picalm deficiency develops accelerated AD pathology. PICALM is also known to be highly expressed in microglia, and further upregulated in LOAD patients [117]. Although its exact function in AD-related microglia phenotypes is yet to be investigated, single cell sequencing has identified two gene-trait correlation modules in AD microglia, carrying a common molecular signature specific to AD risk genes including APOE, TREM2 and PICALM [53], suggesting a potential interaction between these genes in microglia.

CR1

Complement receptor 1, also known as c3b/c4b receptor or CD35, is a member of receptors of complement activation (RCA) family. The human CR1 gene is located in the complement related genes cluster on chromosome 1. It encodes type 1 membrane glycoprotein enriched in innate immune cells including microglia. CR1 modulates complement activation in multiple pathways contributing to inflammation and AD pathogenesis, for example, CR1 expressed in microglia is participated in immune clearance of Aβ in AD brain, it is upregulated when exposed to Aβ, and its blockage with neutralizing antibody attenuated microglia phagocytosis of Aβ [118]. In 2009, studies in nearly 4,000 AD cases from Belgium, Finland, Italy and Spain found CR1 rs6656401 polymorphism is associated with AD risk, with odds ratio (OR) of 1.21 [119]. Similar association between multiple CR1 polymorphisms and AD risks were later confirmed in different populations [120, 121], and remain highly significant in GWAS meta-analysis [114, 115]. Yet, the exact mechanism how CR1 variants are linked to AD pathogenesis remains unknown.

MS4A

Membrane-spanning 4-domains subfamily A (MS4A) is a cluster of genes that encode CD-20 like proteins with four transmembrane-spanning regions. Members of MS4A such as MS4A1, MA4A2, and MS4A4B participate in regulation of intracellular Ca²⁺ signaling. Two SNPs of the MS4A gene cluster, rs610932 and rs670139, were first reported as susceptibility loci associating with AD in 2011 with OR 0.91 and 1.08, respectively [106]. One of them, rs610932, is significantly associated with AD in Chinese Han population [122]. In addition, rs1562990 was found in Spain general population with an OR of 0.88[123]; and rs1582763 was the top SNP reached by GWAS using sTREM2 concentration in patient CSF samples (p=1.15×10⁻¹⁵) [124]. Furthermore, human macrophages treated with MS4A4A antibody showed decreased sTREM2 production, suggesting a functional interaction between MS4A and TREM2 in microglia.

ABCA7

ATP-binding cassette subfamily A member 7 (ABCA7) is a multi-pass transmembrane protein that harnesses energy from ATP hydrolysis and actively transfers molecules across cell membranes. It is expressed in microglia, macrophages, and subtypes of neurons in both humans and mice. ABCA7 variants with potential loss-of-functions are associated with AD risk, which nearly doubled in African-Americans [125]. Abca7 haplodeficiency in microglia impairs the immune response by disrupting the expression of CD14 after acute lipopolysaccharide stimulation in mouse models, as well as leading excessive microglial Aβ accumulation with increased endosomal compartments [126], suggesting Abca7 loss-of-functions in microglia may contribute to AD pathogenesis. ABCA7 also directly regulates phagocytic pathways in mouse microglia [126], and Abca7 deficiency in Abca7^−/−;APP/PS1 mice accelerated amyloid pathology and neurodegeneration [127, 128].

ABI3

ABI3 is a member of Abelson (Abl) interactor (Abi) family of adaptor proteins. The ABI family regulates the actin cytoskeleton mammalian cells and commonly has three functional domains: a homeobox homology domain, a proline rich region and a Src-homology 3 (SH3) domain, besides that, ABI3 also contains a coiled-coil domain for dimerization. In invasive cancer cell lines, ABI3’s SH3 domain interacts with p21-activated kinase (PAK) to modulate ABI3 expression and cell motility [129]. ABI3 is also a key component of the ABI3/WAVE2 complex (AWC), which regulates actin polymerization during T cell activation [130]. The ABI3 common variant rs55978930 is association with decreased t-tau level in CSF in progressive mild cognitive impairment (pMCI) patients, and rs16947151 is associated with cognition decline in MCI patients [131]. More importantly, a missense change p.S209F in ABI3 at rs616338 was found by Sims et al. as a rare-coding risk variant using GWS association signals with LOAD. p.S209F is consistently associated with increased LOAD risk cross all the stages of AD [132]. ABI3 is enriched in microglia, however its role in microglia-related AD pathology remained underexplored.

PLCG2

Phospholipase C gamma 2 (PLCG2) encodes the 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase gamma-2 (PLCγ2), which is a transmembrane signaling enzyme producing inositol 1,4,5-trisphosphare (IP3) and diacylglycerol (DAG) as second messengers in immune cells (Figure 2). PLCγ2 has multiple conserved domains including PH, PU-PLC X-box, SH2 1, SH2 2, SH3, PI-PLC Y-box and C2. In microglia, PLCγ2 can promote many cellular processes like survival, proliferation and phagocytosis, as well as production of cytokines and chemokines[133]. A missense variant of PLCG2 p.P552R at rs72824905 was identified as a protective variant in AD by Sims et al [132]. Rare coding variants of ABI3 and PLCG2 are differentially association with AD risks in different cohorts across the world [134, 135], which indicates that ABI3 p.S209F functions as pathogenic variant while PLCG2 p.P522R is protective. This is confirmed in a recent report [136] demonstrating that PLCG2 p.P522R carriers had a much slower progression in cognition decline and lower pTau₁₈₁ levels in CSF compared to non-carriers. In addition, there is a functional protein interplaying network among PLCG2, APOE and TREM2 in microglia, as the gene co-expressed with PLCG2 significantly overlap with the ones that co-expressed with TREM2 (P=1.37X10⁻³³) and APOE (p=7.49×10⁻³⁴) [136]. However, how these variants influence the protein normal functions in microglia and contribute to AD pathogenesis needs further understanding.

NLRP3 inflammasome

Increasing evidence suggested that innate immunity and inflammasome response strongly contribute to AD pathogenesis in brain. NLRP3 inflammasome is highly important for microglia activation, and its indispensable roles in Aβ plaque formation and tauopathy have been demonstrated [137]. The nucleotide-binding oligomerization domain-, leucine-rich repeat- and pyrin domain-containing 3 (NLRP3) inflammasome is a subcellular multiprotein complex, which can sense a wide range of exogenous and endogenous stimuli such microbes, aggregated and misfolded proteins, and adenosine triphosphate, which result in activation of caspase-1 [138]. NLRP3 inflammasome is comprised of a sensor NLRP3, an adaptor protein apoptosis associated speck-like protein containing a CARD (ASC) and an effector caspase-1[139]. Activated caspase-1 subsequently leads to the processing of pro-inflammatory cytokines interleukin-1β (IL-1β) and interleukin-18 (IL-18) and mediates rapid cell death [140]. Both IL-1β and IL-18 drive inflammatory responses through diverse downstream signaling pathways leading to neuronal damage and other neurological pathogenesis [139, 141–143].

Based on single cell RNA sequencing in humans, NLRP3 is mainly expressed in microglia [144]. It plays a critical role for initiating and maintaining neuroinflammation, as well as in the development of neurological disorders including AD [143]. Specifically, Nlrp3 is activated in AD and contributes to pathology in APP/PS1 mice because NLRP3 or caspase-1 deficiency leads to the alteration of microglia phenotypes and the decreased Aβ deposition [145]. Recent evidence showed that loss of NLRP3 inflammasome function in microglia reduced tau hyperphosphorylation and aggregation in Tau/Nlrp3^−/− mouse model [146], identifying an important role of microglia and NLRP3 inflammasome activation in the pathogenesis of tauopathies. These studies clearly demonstrate that the NLRP3 inflammasome activation increases the risk for driving AD pathology and suggest that the manipulation of NLRP3 inflammasome activity is potentially promising way to treat AD. Indeed, targeting NLRP3 inflammasome signaling with small molecular inhibitors has been proven to ameliorate proinflammatory response in microglia and alleviate AD pathogenesis in preclinical studies [142, 147].

Developing AD therapeutic strategies on AD risk genes

There is currently no effective treatment for AD. Most AD therapeutic strategies mainly focus on preventing Aβ aggregation by modulating its enzymatic processing or clearance [148]. The genetic studies on TREM2 and its variants in microglia have enabled new advances in AD therapeutics. Monoclonal antibody AL002, which activates the TREM2 receptor and microglial immune response, has shown promising efficacy in preclinical studies for Aβ clearance and cognitive improvement [149]. Its human variant is now in phase 2 clinical trials. On the other hand, anti-APOE antibodies, such as HJ6.3 that binds to APOE associated with Aβ plaques [150] and HAE-4 that binds to non-lipidated APOE [151], have shown success in preclinical animal models by recruiting microglia to plaques and significantly reducing Aβ deposition. In addition, increasing APOE’s lipidation, designing APOE mimetics or interfering APOE and Aβ interaction all performed effective in mouse models [152]. As NLRP3 activation in microglia has been proved as a contributor for neuroinflammation and pathogenesis in AD [145, 146], therapeutics antagonizing the NLRP3 inflammasome are also under development right now.

Conclusion and Future Directions

In this review, we briefly introduced historical perspectives on microglia including the initial discovery and definition, demonstrated the classic activation of microglia from homeostatic (M0) to activated (M1/M2) status and summarized the diverse subtypes of microglia after being activated. Most importantly, we discussed the heterogeneity of the AD risk gene profile in microglia, like TREM2 and APOE which are well-established genetic risk factors, or ABI3 and PLCG2 which are novel contributors for AD progression. Among these risk genes closely associated with AD or other neurodegenerative disorders, some own their specific pathway to trigger pathologies of AD; some directly interacte with Aβ and modulate the phagocytosis of microglia; in addition, some risk genes closely collaborate within complex network stimulating microglia changing from homeostatic status to DAM. However, there are still many of them remaining unknown about their mechanisms for being an AD risk gene. The post-GWAS analysis is important for assessing the potential role of AD risk genes, the various hypotheses will lead more comprehensive biological mechanism uncovered and better characterize functional variants within AD [63]. Understanding microglia’s dynamics on morphology, activation, spatial distribution and gene expression profile enable the discovery of neurodegenerative diseases much easier and widely. With the advanced and continually improved technologies and analysis based on large clinical data from different populations of world-wide range, countless risk genes will show up for people to study, even though nowadays many of them keep remain unclear and unknown for us, identification and characterization of them can provide promising targets and benefits for developing treatments in clinical therapy for AD and other neurodegenerative diseases.

Table 1.

Alzheimer’s Disease Risk genes

Gene	Activities in AD	Cellular Functions
		Microglia	Other cell types
TREM2	TREM 2 variants R47H, D87N, R62C and H157Y cause partial loss of functions on microglia’s phagocytosis and clearance of Aβ, and increase AD risks.	Interacting with and phosphorylating DAP12 adaptor, mediating SYK, PI3K/Akt, GSK3β pathways and rising inflammasome NLRP3.	Cleaved by ADAM10 in dendritic cells; a negative regulator for TLR-4 expression in astrocytes; attenuate macrophage activation.
CD33	Increased CD33 expression on microglia reduces the uptake of Aβ. rs3885444A allele is protective for AD.	A receptor on microglia activated by sialic acid-containing glycoproteins and glycolipids, recruiting proteins to inhibit phagocytosis.	Phosphorylates ITIM, docks SH2 domain-containing protein and inhibits phagocytosis in myeloid cells.
APOE	APOE4 is the strongest genetic risk for LOAD, it initiates and accelerates Aβ accumulation, aggregation and deposition in brain	ApoE is a main lipid ligand for TREM2 on microglia.	ApoE4 decreases uptake of palmitate and increases fatty acid oxidation and accumulates in astrocytes; stimulates expression of ADP-ARF6 and traps ABCA1 in endosome.
PICALM	Clearance of Aβ, playing a protective role in recognizing and shipping APP-CTF in autophagic degradation process.	Mediating trafficking endocytic protein during endocytosis.	PLCALM regulates internalization of Aβ, and guides Aβ trafficking and clearance in brain endothelial cells.
ABI3	ABI3 variants may cause neurodegeneration through adjusting total level of tau protein in CSF or modulating cognitive function.	ABI3 participates in regulation of actin cytoskeleton and inhibits ectopic metastasis of tumor cells, as well as cell migration.	ABI3, CYFIP1 and WAVE constitute WRC, and regulates cytoskeleton dynamics in follicular thyroid cells, which is related to cancer development and metastasis. ABI3 also regulate cell migration in endothelial cells.
PlCG2	Variant rs72824905G is a protective variant in AD and other neurodegenerative diseases.	PLCγ2 catalyzes the hydrolysis of phospholipids and produces diacylglycerols and water-soluble phosphorylated derivatives of the lipid head groups.	PLCγ2 involved in cell proliferation, transformation and tumor growth.
CR1	The activity of microglial phagocytosis on Aβ is positively related with activation of CR1.	CR1 is a glycoprotein expressed in microglia and a main receptor for C3b which participates in immune activation.	CR1 regulates the binding between immune complex in erythrocytes. In B cells, CR1 forms a complex with CR2 to facilitate C3b or iC3b coated protein antigen uptake.
ABCA7	The premature termination codon mutation in ABCA7 can cause increased risk on AD. ABCA7 haplodeficiency leads to excessive microglial Aβ accumulation.	ABCA7 has an important function to regulate homeostasis of phospholipids and cholesterol in CNS and peripheral tissues and involves in phagocytosis process.	N/A
MS4A	MS4A involves in AD pathology through controlling intercellular Ca2+, MS4A family overexpression activates T cell, in which Th1 and Th17 can increase microglial production of inflammatory cytokines.	MS4A family is adaptor protein or ion transport modulator in regulating cell activation, growth and development.	MS4A4 express at the plasma membrane in monocytes, dendritic cell and macrophages. MS4A4 is important for macrophage differentiation and polarization, and involves in dectin-1-dependent activation of NK cell-mediated resistance to metastasis.

Abbreviations: ADP-ARF6, ribosylation factor 6; ABCA1, ATP binding cassette A1; ALS: Amyotrophic lateral sclerosis; APP-CTF, APP C-terminal fragment; CYFIP1, Cytoplasmic FMR1 Interacting Protein 1; WAVE, WASP-family verprolin-homologous protein.