Microglial activity, INPP5D/SHIP1 signaling, and NLRP3 inflammasome activation in Alzheimer's disease

Alzheimer’s disease (AD) is a neurodegenerative disease that accounts for an estimated 60–80% of dementia cases in the United States. Alzheimer’s dementia presents with memory loss, apathy, confusion, impaired communication and behavioral changes. Pathologically, AD is characterized by the presence of extracellular amyloid β (Aβ) plaques, intracellular neurofibrillary tau tangles, neuroinflammation, synapse loss and cell death. Most AD cases arise due to an interplay between complex polygenic mechanisms and non-genetic factors [1, 2]. In recent years, genome wide association studies (GWASs) and analyses of rare-coding variants have identified over 70 loci associated with conferring risk for late-onset AD (LOAD). Many of these loci are in the vicinity of genes that are strongly expressed in microglia and implicated in microglial functions, including ABCA7, ABI3, APOE, CD2AP, CD33, CLU, CR1, EPHA1, HLA-DRB5/DBR1, MS4A6A/MS4A6E, PLCG2, SORL1, TREM2 and INPP5D [3–11]. An active area of investigation is geared towards unraveling the precise mechanisms by which these variants increase risk for AD. Elucidating the normal function of each of the genes associated with these loci will enlighten our understanding of microglia’s role in the adult brain. Here we focus primarily on what is known about one of these genes, INPP5D, in microglial function and dysfunction.

Microglia are the resident myeloid cells of the brain parenchyma. They survey and sense their local environment using their dynamic, fine processes and respond to stimuli including neuronal and other cellular activity, cell death, injury, infection and other inflammatory insults [12, 13]. During development, microglia are involved in shaping neuronal circuits by pruning and remodeling synapses, clearing dead or dying cells, and regulating myelination and vasculature development by interacting with other glial and endothelial cells of the central nervous system (CNS) [12, 14, 15]. In adulthood, microglia exist as a heterogenous population of cells that can display distinct phenotypes and functions. While some microglia continue to exhibit a homeostatic phenotype to maintain brain homeostasis, some may change dramatically as they respond to injury or insults by releasing pro-inflammatory cytokines, causing neurotoxicity directly or inducing neurotoxic phenotypes in astrocytes [16–18]. While “healthy” microglia play important roles in protecting the brain, in neurodegenerative disease, “dysfunctional” microglia can contribute to further pathology. An array of cell surface receptors and signaling pathways in microglia have been implicated in mechanisms underlying AD. While the literature can be confusing at times with seemingly conflicting data regarding the precise mechanisms at play in microglia, examination of the data leads to two high level conclusions: (1) there is complex regulation of microglial activation that involves both negative and positive feedback and (2) cellular and molecular context plays a critical role in determining the precise pathway(s) affected in microglia in AD. In this review, we attempt to integrate findings regarding INPP5D with the growing literature regarding microglial pathways that impinge upon risk for AD.

One mechanism by which multiple cell types contribute to inflammatory processes that is increasingly gaining attention is the activation of the nucleotide-binding domain (NOD)-like receptor protein 3 (NLRP3) inflammasome. Inflammasomes are intracellular molecular platforms that induce the secretion of pro-inflammatory cytokines to engage the innate immune system against infection or cellular stress [19]. In this review, we describe the formation and functions of the microglial NLRP3 inflammasome, particularly in the context of neuroinflammatory processes in AD. We then summarize the known functions of a gene expressed exclusively in myeloid cells that is associated with LOAD through GWAS, INPP5D. We discuss the role of SHIP1, the protein encoded by INPP5D, in phosphoinositide signaling and its involvement in microglial functions, and recent evidence suggesting that reduction of SHIP1 activity results in activation of the NLRP3 inflammasome. Finally, we hypothesize a mechanistic link between SHIP1’s regulation of the NLRP3 inflammasome and microglial function, and the implications of this regulation on different cell types in the brain.

Aging, which remains the largest risk factor for LOAD, is associated with low-grade systemic inflammation (termed “inflammaging”) [20, 21]. Senescent cells acquire the senescence-associated secretory phenotype (SASP), characterized by increased secretion of pro-inflammatory cytokines, including IL-1β, IL-6, IL-8, chemokines, growth factors and several other freely diffusible molecules [22]. Levels of these pro-inflammatory cytokines (including IL-1β, IL-6, IL-8, IL-18 and TNF) are upregulated in serum, cerebrospinal fluid (CSF) and post-mortem brain tissue of AD patients [23–25]. Thus, aging is associated with a basal inflammatory state, which is exacerbated in AD.

AD is characterized by a state of sustained glial activation and neuroinflammation. Prolonged immune activation results in part from ineffective clearing of pathological aggregates such as Aβ. This immune response in turn activates glia and induces the secretion of pro-inflammatory cytokines, which can drive further aggregation of Aβ by affecting the capacity of microglia and astrocytes for phagocytosis and degradation [26]. Microglia migrate to sites of Aβ-rich plaques, surrounding them and phagocytosing Aβ that is dynamically released from the plaques. In line with this, increased numbers of microglia can be detected near Aβ plaques in AD. These disease-associated microglia (DAMs), which express plaque-induced genes (PIGs) in the context of AD, are activated, inflammatory and display increased phagocytic activity to clear Aβ and limit plaque-associated neurotoxicity [27]. Transcriptionally, the DAM signature includes downregulation of genes expressed in homeostatic microglia such as CX3CR1 and P2RY12 and upregulation of genes including APOE, AXL, SPP1 and TREM2, as well as major histocompatibility complex class II (MHC-II), a gene essential for the initiation of antigen-specific immune response [27–29].

Several AD risk genes expressed primarily in microglia and other myeloid cells have important roles in immune function and inflammatory response both inside and outside of the brain. One such gene is triggering receptor expressed on myeloid cells 2 (TREM2). TREM2 is a transmembrane receptor with an immunoglobin-like ectodomain that associates with the adaptor protein DNAX-activating protein of 12 kDa (DAP12, also known as TYRO protein kinase-binding protein or TYROBP) to transduce downstream signaling. TREM2 ligands include zwitterionic and anionic phospholipids, glycolipids, apoptotic cells, lipoproteins such as HDL, LDL, APOE and APOJ/CLU, as well as Aβ [30, 31]. Autosomal recessive loss-of-function mutations in TREM2 cause Nasu-Hakola disease, which is a rare genetic disorder characterized by bone cysts and progressive presenile dementia [32]. Coding variants in TREM2 confer a high risk for developing AD. The most common coding variant, R47H, results in a reduction of TREM2 function and was found to increase risk for developing AD three- to five-fold, suggesting that TREM2 is protective in AD. Several other TREM2 variants associated with AD and frontotemporal dementia (FTD) have also been found to confer partial loss-of-function, as they impair TREM2 binding to ligands, maturation, ectodomain shedding and microglial phagocytic activity [33]. Microglia with TREM2 loss-of-function cannot transition to the DAM state and are locked in a homeostatic state, in which they no longer respond normally to pathological challenges [34]. Microglia with TREM2 reduction do not cluster and become activated around plaques, which was shown to result in elevated plaque burden [35]. TREM2 binding to Aβ promotes its phagocytosis by microglia, which in turn induces secretion of pro-inflammatory cytokines IL-6 and CCL3, increases microglial proliferation, migration and apoptosis and reduces intracellular K⁺ concentrations [31, 36]. While the role of TREM2 activity in AD pathogenesis appears complex, some studies have suggested that TREM2’s exact role may depend on the context and stage of the disease. Loss of Trem2 in an APP/PS1 AD mouse model accelerated amyloidogenesis in early disease pathogenesis, whereas Trem2 deficiency did not affect amyloid load at a later disease stage [37]. On the other hand, another study reported reduced plaque numbers early, but increased plaque area later in disease progression in APP/PS1/Trem2^-/- mice [38]. In line with this, prolonged TREM2 activation by inhibiting its proteolytic cleavage in another AD mouse model increased plaque deposition and promoted neuroinflammation at an early stage of amyloidogenesis [39]. Moreover, loss of TREM2 function has been found to increase tau pathology, with elevation in tau seeding around Aβ plaques in APP/PS1 mice injected with tau isolated from human AD brain [40]. On the other hand, in PS19 tau transgenic mice that express human tau containing the P301S mutation (in the absence of plaque pathology), loss of TREM2 was protective against brain atrophy and synapse loss [41, 42], and this protective effect was abolished in the presence of ApoE4 [43]. Together, these findings point to a dynamic and complex role for TREM2 during different stages and contexts of AD pathogenesis, possibly mediated by distinct signaling pathways.

Phospholipase-Cγ2 (PLCG2) is another gene associated with AD risk which has a role in microglial inflammation. Mutations in the gene cause the immune disorder PLCG2-associated antibody deficiency and immune dysregulation (PLAID) and its autoimmune version APLAID, which result in immunodeficiency [44]. PLCγ2 is a membrane-associated enzyme that cleaves phosphatidylinositol-4,5-bisphosphate (PIP₂) to generate diacyl-glycerol (DAG) and inositol-1,4,5-trisphosphate (IP₃), which play central roles in intracellular signal transduction. While missense hypermorphic variants of PLCG2 confer a reduced risk for AD, the mechanistic role that PLCγ2 plays in AD pathogenesis is unclear [44–48]. In human iPSC-derived microglia, PLCγ2 was shown to act downstream of TREM2 to modulate cell survival, phagocytosis, lipid metabolism and clearance of neuronal debris, while also acting downstream of toll-like receptors (TLRs) to mediate pro-inflammatory response upon microglia stimulation with lipopolysaccharide (LPS) [48]. PLCG2 RNA expression was found to be increased in post-mortem brain tissue of LOAD patients and was positively correlated with increased amyloid burden [47]. These results suggest that like TREM2, PLCγ2 activity could be beneficial or harmful depending on the disease stage.

INPP5D is a gene specifically expressed in myeloid cells that encodes the protein src homology 2 (SH2) domain containing inositol polyphosphate 5-phosphatase 1 (SHIP1). While INPP5D has been repeatedly implicated in AD risk, it had first been recognized for being essential for proper hematopoiesis [49]. Loss-of-function mutations in INPP5D have been linked to myeloproliferative cancers such as acute myeloid leukemia [50–52]. Germline ubiquitous knockout of this gene in mice leads to the overproliferation of peripheral myeloid cells that results in early postnatal lethality [53]. Together, these studies point to the importance of SHIP1 function in myeloid cells. Evidence of potential dysfunction of INPP5D in AD will be discussed in later sections.

Together, the roles of several AD risk genes in peripheral immune disorders and inflammatory microglia activation validate the importance of inflammation and the innate immune system in AD pathogenesis. Recently, a role for inflammasomes has been recognized for mediating neuroinflammation in neurodegenerative diseases including AD. Dysregulated inflammasome activation had been previously linked to autoimmune diseases and cancer. Inflammasomes are a part of the innate immune system that coordinate host defense response against pathogens and host-derived “danger” signals that communicate cellular stress or toxicity [19, 54, 55]. The next section provides an overview of the inflammasome and its involvement in inflammation.

The canonical inflammasome complex is composed of a sensor, an adaptor, and a downstream effector. The cytosolic sensor can be either of the nucleotide-binding domain and leucine-rich-repeat-containing (NLR), absent in melanoma 2 (AIM2) or pyrin proteins. These proteins assemble the formation of a multiprotein complex, called the inflammasome, upon sensing molecular signals associated with pathogens, also called pathogen-associated molecular patterns (PAMPs), or endogenous damage/danger-associated molecular patterns (DAMPs). NLRP3 is a member of the NLR protein family that can recognize a variety of DAMPs, PAMPs, as well as environmental substances. NLRP3 inflammasomes are currently the best characterized inflammasomes and are predominantly expressed by microglia in the CNS [55]. NLRP3 protein consists of a C-terminal leucine-rich repeat (LRR), followed by an NLP family apoptosis inhibitor protein (NAIP) nucleotide-binding domain (NACHT) and an N-terminal pyrin domain (PYD) (Fig. 1). The LRR domain recognizes activating signals, the NACHT domain mediates the self-oligomerization and activation of NLRP3, while PYD mediates the downstream protein-protein interactions. The adaptor protein, apoptosis-associated speck-like protein containing a CARD (ASC, also known as PYCARD) consists of PYD and a caspase recruitment domain (CARD). Finally, the main effector caspase, caspase-1, initiates the secretion of the pro-inflammatory cytokines IL-1β and IL-18 [19, 54, 56].

NLRP3 inflammasome activation is a two-step process. The first step, often referred to as the priming step, prepares the inflammasome for activation through a series of transcriptional and post-translational mechanisms that allow the generation of components of the inflammasome. This priming step can be induced in a variety of ways including activation of TLRs upon ligand binding, which includes bacterial LPS and other PAMPs, as well as the activation of tumor necrosis factor receptor (TNFR) and IL-1 receptor 1 (IL-1R1) by exogenous TNF and IL-1β, respectively. The priming step involves upregulation of expression of inflammasome components, including NLRP3, ASC, as well as pro-IL-1β, all of which are upregulated through NF-kB signaling to levels necessary for inflammasome activation. In the second, activation step, the now primed NLRP3 inflammasome can be directly activated by any of the various signals including ATP, reactive oxygen species (ROS) and DNA released from damaged mitochondria, protein aggregates, uric acid crystals and ceramide, bacterial or viral RNA, airborne pollutants such as asbestos and silica, decreased intracellular concentrations of K⁺ and Cl⁻ as well as Ca²⁺ flux [57–60] (Fig. 1).

The inflammasome is tightly regulated and autoinhibited at the resting state. In the priming step, the inflammasome remains in the autoinhibited state while becoming sensitized to activating signals. Once the inflammasome components are activated, NLRP3 undergoes oligomerization through homotypic interactions between NACHT domains, which leads to the loss of the autoinhibited state. The serine-threonine kinase NIMA-related kinase 7 (NEK7) interacts and forms a complex with NLRP3, which is essential for NLRP3 activation. The oligomerized NLRP3 recruits and induces the filament assembly of ASC through homotypic PYD interactions. ASC acts as a molecular platform to recruit pro-caspase-1 through homotypic CARD interactions. Pro-caspase-1 then undergoes proximity-induced autoproteolysis, which leads to active caspase-1 processing pro-IL-1β and pro-IL-18 into their mature forms IL-1β and IL-18 and the subsequent secretion of these cytokines [55, 58, 61–63] (Fig. 1).

Inflammasome activation can lead to a form of cell death called pyroptosis, although sublethal inflammasome activation can also occur [64, 65]. During sublethal inflammasome activation, activated caspase-1 cleaves gasdermin D (GSDMD) to generate N-terminal GSDMD fragment (GSDMD-N), which inserts into and forms pores in the plasma membrane, resulting in the release of inflammatory cytokines into the extracellular environment [64, 66] (Fig. 1). While the N-terminal domains of other members of the gasdermin family (GSDMA, GSDMB, GSDMC and GSDME) can also form pores and induce necrotic cell death, GSDMD is the only gasdermin protein with a cleavage site for caspase-1 [67]. Pyroptosis may occur downstream of GSDMD pore formation and is mediated by Ninjurin1 (NINJ1), a cell surface protein that induces plasma membrane rupture upon its oligomerization [68, 69]. Cell membrane rupture upon pyroptosis leads to the release of DAMPs such as HMGB1 as well as ASC specks into the extracellular space [69, 70]. These ASC specks released from the pyroptotic cell accumulate in the extracellular space, where they can potentially continue to process and activate pro-caspase-1 and pro-IL-1β [70]. They can also be recognized by neighboring microglia as a danger signal and induce inflammatory responses including activation of the NLRP3 inflammasome, thereby propagating inflammation [70–72].

NLRP3 inflammasome activation is implicated in the systemic low-grade inflammation in the brain and periphery during normal aging. Various DAMPs that accumulate throughout an organism’s lifespan activate the NLRP3 inflammasome, which leads to elevated IL-1β levels associated with aging [73]. Interestingly, ablation of Nlrp3 in mice reduced circulating levels of IL-18 and IL-1β and reduced microglial activation and astrogliosis in the hippocampus. Nlrp3 ablation also increased expression of genes related to learning and memory and reduced age-related cognitive decline in “wild type” mice [73]. Elevated levels of IL-1β in the brain can itself induce neuroinflammation, neurotoxicity and learning and memory impairment [55, 74]. Therefore, increased baseline activation of the NLRP3 inflammasome in aging may predispose the brain to neurodegenerative diseases, in which neuroinflammation is a common feature [55].

Concordantly, aberrant NLRP3 inflammasome activation has been found to mediate AD-related neuroinflammation and exacerbate pathogenesis. Evidence for activation of functional NLRP3 inflammasomes has been detected in early and intermediate stages of AD from studies of post-mortem brains. On the other hand, only an increase in NLRP3 levels, but not in ASC or caspase-1, was observed in monocytes isolated from mildly cognitively impaired (MCI) individuals. In these monocytes, NLRP3 and caspase-1 were also not co-localized and IL-18/IL-1β secretion was not detected, suggesting that the NLRP3 inflammasome might not be active at this stage of disease [75, 76]. However, another study reported strongly increased levels of active caspase-1 in early-onset AD and LOAD patients as well as in MCI individuals [77]. These findings suggest that elevated NLRP3 inflammasome activation is associated with AD neuropathology. Indeed, distinct forms of Aβ act as DAMPs that can activate the inflammasome. Fibrillar Aβ (fAβ), following phagocytosis by microglia, activates the NLRP3 inflammasome and leads to increases in secretion of IL-1β and IL-18. fAβ phagocytosis results in lysosomal dysfunction and release of lysosomal contents into the cytoplasm, including the lysosomal protease cathepsin B, which is a mediator of inflammasome activation induced by Aβ [78]. Oligomeric Aβ (oAβ) also can activate the NLRP3 inflammasome and induce IL-1β secretion in LPS-primed microglia, but this effect was not dependent on oAβ phagocytosis and release of cathepsin B from damaged lysosomes. Instead, it was suggested that oAβ might activate the inflammasome by inducing ROS generation by damaged mitochondria and NADPH oxidase 2 (Nox2), formation of pores in the cell membrane, disruption of K⁺ channels and subsequent K⁺ efflux [79].

While Aβ can activate the NLRP3 inflammasome, it has been suggested that NLRP3 inflammasome activation also can promote Aβ aggregation. As previously mentioned, ASC specks are released into the extracellular milieu after pyroptosis, where they may propagate inflammation. In one study, ASC specks in the extracellular space could bind Aβ which then led to further Aβ aggregation both in vivo and in vitro. In that study, ASC-bound Aβ was detected in brain samples from AD patients, but not in samples from age-matched controls without dementia. Co-localization of ASC and Aβ also was found in brain tissue from patients with MCI that preceded AD, indicating that ASC speck-induced Aβ aggregation may occur in the brain [80]. ASC-binding to Aβ induced pyroptosis and impaired microglial uptake and degradation of Aβ in primary mouse microglia, contributing to further pathology driven by these aggregates [72].

The NLRP3 inflammasome has a bidirectional relationship with AD-related tau pathology as well. Evidence suggests that while some forms of tau can activate the NLRP3 inflammasome, inflammasome activation can exacerbate tau pathology by promoting elevated phosphorylation of tau [81, 82]. Treatment of wild-type primary microglia with recombinant wild-type tau and recombinant tau with a P301S mutation increased IL-1β secretion, an effect that was attenuated in Nlrp3^−/− and Asc^−/− microglia, suggesting that tau induces IL-1β secretion in a NLRP3-dependent manner [81]. On the other hand, treatment of mouse primary hippocampal neurons with conditioned media from wild-type, but not Nlrp3^−/− and Asc^−/−, microglia primed with LPS resulted in elevated total and phosphorylated tau levels, suggesting that NLRP3 inflammasome activation in microglia can induce tau pathology in neurons. Moreover, treatment with brain homogenates from APP/PS1 mice induced tau hyperphosphorylation in a mouse model of tau aggregation pathology (Tau22) but not in Tau22/Nlrp3^−/− and Tau22/Asc^−/− mice, indicating the NLRP3 inflammasome as a mediator of Aβ-induced tau pathology [81]. In line with these findings, another study showed that aggregated tau seeds activated the NLRP3 inflammasome in primary mouse microglia by inducing lysosomal destabilization and release of the lysosomal protease cathepsin B, similar to fAβ [78, 82]. In the same study, exogenously seeded tau induced tau pathology in Tau P301S transgenic mice, but not in Asc-deficient or NLRP3 inhibitor (MCC950)-treated Tau transgenic mice, demonstrating the ability of the NLRP3 inflammasome to aggravate tau pathology [82].

Inhibiting or downregulating the NLRP3 inflammasome ameliorates AD neuropathology in animal models, which opens up a therapeutic avenue for modulators of NLRP3 activation in AD. Ablation of Nlrp3 or Casp1 in APP/PS1 mice rescued spatial memory and behavioral deficits, LTP suppression and spine loss observed in these mice [77]. APP/PS1/Nlrp3^−/− and APP/PS1/Casp1^−/− mice also displayed decreased Aβ plaque deposition, suggesting that NLRP3 inflammasome activation impairs microglial capacity to clear Aβ [77]. The decreased plaque deposition was also related to increased levels of the proteolytic enzyme insulin-degrading enzyme (IDE) [77], previously reported to degrade Aβ [83]. In Tau22 mice, ablation of Nlrp3 or Asc reduced levels of hyperphosphorylated tau and tau aggregates and rescued tau-induced spatial memory deficits [81]. In light of the potential of NLRP3 ablation in alleviating AD pathology and related phenotypes, small molecule inhibitors of NLRP3 are currently being investigated as a therapeutic strategy [84]. For example, the NLRP3-specific small molecule inhibitor MCC950 was found to reduce microglial activation, IL-1β secretion and Aβ plaque numbers, increase microglial phagocytosis of Aβ and improve cognitive function in APP/PS1 mice [85, 86]. Whether these inhibitors will prove clinically useful remains to be investigated.

Given SHIP1’s potential role in regulating the signaling mechanisms discussed above, a reduction in SHIP1 protein levels is poised to promote microglial activation and inflammation related to AD pathogenesis. GWASs have identified three single nucleotide polymorphisms (SNPs), rs35349669, rs10933431, and rs7597763, at the INPP5D locus that are associated risk for AD [8–10]. While variant rs35349669 is proposed to confer increased risk for AD, rs61068452 is proposed to confer protection [166, 167]. These SNPs are located in introns of INPP5D, and it is unclear whether and how they affect INPP5D mRNA levels or splicing.

One study found elevated INPP5D mRNA expression in human AD brains that positively correlated with amyloid plaque density, and elevated Inpp5d mRNA and protein expression in 5xFAD mice throughout disease progression [168]. SHIP1 protein expression was identified in plaque-associated microglia of 5xFAD mice, with phagocytic microglia showing a reduction of Inpp5d expression compared to non-phagocytic microglia [168, 169]. Another study similarly found an upregulation of INPP5D RNA expression in both human AD patients and an App knock-in mouse model [170]. A recent study investigated RNA expression of the different INPP5D isoforms in the post-mortem brains of AD and non-AD individuals. While they also found an increase in the expression of INPP5D isoforms in the AD brain, the expression of a truncated isoform that lacks the phosphatase domain was increased, suggesting that the increase in INPP5D expression may not necessarily lead to increased SHIP1 activity [171]. Of note, the investigation of AD-associated SNPs in INPP5D in this study did not reveal a clear SNP effect on overall INPP5D expression [171].

While data from our group agrees with the elevation of INPP5D mRNA levels in human AD post-mortem brains, we also found that aqueous-soluble SHIP1 protein levels are reduced in brain tissue from individuals diagnosed with AD compared with non-AD individuals [161]. However, we found that AD brains displayed increased immunostaining for SHIP1 compared to age and sex-matched brain tissue from non-cognitively impaired individuals when a C-terminally directed antibody to SHIP1 was used. Plaque-associated microglia clearly displayed higher SHIP1 immunostaining than microglia distal to plaques, SHIP1 levels significantly correlated with microglial ramification, where less ramified and putatively more activated plaque-associated microglia had lower levels of SHIP1 immunostaining [161]. Moreover, the staining pattern for SHIP1 was diffuse in some microglia, while a more punctate pattern was observed in other microglia. In AD brain tissue, SHIP1 localization was shifted from diffuse to punctate patterns in microglia, suggesting that SHIP1 may be aggregated (and perhaps aqueous-insoluble) in the AD brain [161]. Peptide-level mass spectrometry analyses of TBS- and urea-solubilized postmortem brain lysates revealed a reduction of peptides encoding the phosphatase domain of SHIP1 in the AD brain [161]. Further, in agreement with a reduction in function of INPP5D in AD microglia we observe some concordance between expression profiles of iMGs with biallelic loss-of-function alleles in INPP5D and in microglia from AD brain tissue. For example, iMGs with biallelic INPP5D LOF alleles and microglia from AD brain tissue both show a reduction in RNA levels of NF-kB pathway genes. In accord, overexpression of INPP5D in iMGs showed an elevation in RNA levels of NF-kB pathway genes, suggesting that microglia from AD brains may display changes relevant to reduced SHIP1 activity [161]. Together, these findings support the hypothesis that there are multiple pools of SHIP1 in the brain, and that in the AD brain, enzymatically active SHIP1 is reduced. Additional analyses are warranted to develop sensitive methods to measure SHIP1 activity in microglia present in postmortem brain samples to determine if activity is altered in AD brain tissue.

In contrast to the notion that SHIP1 activity is reduced in the AD brain, recent studies in mice point to a largely beneficial role for reduction of Inpp5d expression. However, data regarding the impact of reduced Inpp5d expression on Aβ plaque load are complicated. Conditional knock out of Inpp5d in myeloid cells in 3-month-old APP/PS1 mice resulted in an increase in plaque burden and recruitment of microglia to plaques. In this same study, significant congruence was observed between plaque-associated gene expression changes induced by Inpp5d reduction and previously identified human AD gene networks [136]. However, another recent study that knocked out Inpp5d in myeloid cells at 6 weeks of age found reduced plaque load, accompanied by a reduction in dystrophic neurites and an increase in GFAP⁺ immunoreactivity [172]. Investigation of 5xFAD mice with germline deletion of one copy of Inpp5d similarly showed increased microglia engagement with plaques but decreased plaque deposition. Further, Inpp5d reduction rescued a behavioral phenotype that is observed in 5xFAD mice WT for Inpp5d at 6 months of age, reduced spontaneous alternation [164]. In another recent study, conditional knock out of Inpp5d in 3-week-old 5xFAD mice similarly enhanced microglia recruitment to plaques, decreased microglial ramification, and ameliorated neuritic dystrophy but had no impact on plaque burden [173]. The contradictory impacts of Inpp5d loss on plaque load in these studies may be attributed to the time point in disease progression at which Inpp5d expression is reduced as well as the differences in the AD mouse models, as also discussed recently in Gandy and Ehrlich (2023) [174].

Taken together, while these studies in humans and mice all support a role for SHIP1 in AD, there is seemingly contradictory evidence regarding whether therapeutic strategies for AD should aim to increase or decrease SHIP1 activity. While SHIP1 inhibition is currently being investigated by the field in light of the largely beneficial effects of Inpp5d knockdown in AD mouse models, SHIP1 agonism also has been proposed as a therapeutic strategy for AD. The differences in human and mouse studies may reflect genuine biological differences between species and/or may be due instead to differences in the disease stage interrogated and context of other, currently unknown extenuating factors. Deep analyses to disentangle these possibilities will clarify the optimal strategy for SHIP1 targeting while at the same time illuminating a better understanding regarding the fundamental biological processes at play within microglia in the healthy and diseased brain.

Corroborating SHIP1’s role in modulating microglial activation and inflammation, we recently discovered a direct connection between SHIP1 and the NLRP3 inflammasome in iMGs. Acute pharmacological inhibition of SHIP1’s phosphatase activity as well as genetic ablation of a single copy of INPP5D in iMGs induced the secretion of IL-18 and IL-1β that was blocked by NLRP3 inhibition with MCC950 and also with caspase-1 inhibitors [161]. Though reduction of INPP5D increased the secretion of IL-18 and IL-1β, it lowered their mRNA expression, suggesting that reduction of SHIP1 activity positively regulates the activation step but may act through a feedback mechanism to inhibit the priming step for the NLRP3 inflammasome [161] (Fig. 3D).

Evidence from post-mortem analyses of human AD brains support the inverse relationship between SHIP1 levels and the activation of the inflammasome. When levels of secreted IL-1β and IL-18 were compared to aqueous-soluble SHIP1 levels in brain lysates, lower SHIP1 levels were signficantly associated with elevated IL-18 protein levels, with the correlation driven by data from AD brain tissue. Further, comparison of immunoreactivity for SHIP1 and ASC specks in brain tissue from aged individuals revealed that lower SHIP1 levels were significantly correlated with a higher percentage of microglia with ASC specks in the AD brain [161]. Furthermore, a recent single-nucleus RNA sequencing profiling study of microglia from AD and non-AD brains identified an “inflammatory” cluster that was over-represented in the AD brains. While this cluster showed significant downregulation of INPP5D, IL-1β expression was significantly elevated, and this cluster was enriched for NLR signaling [175]. Similarly, the gene expression profiles of microglia from an App knock-in mouse model identified an inflammatory reactive microglia cluster that was enriched with AD risk genes, including Apoe, but this cluster displayed decreased expression of Inpp5d, suggesting Inpp5d reduction as a potential mediator of inflammatory response in this microglia subtype [176]. Therefore, we posit that SHIP1 is a negative regulator of NLRP3 inflammasome activation and that reduction of SHIP1 activity contributes to the activation of the NLRP3 inflammasome in AD. In parallel, reduction of SHIP1 activity is poised to disinhibit signaling pathways that are involved in the phagocytosis of Aβ, which is a required step for Aβ-induced activation of the NLRP3 inflammasome [78].

Collectively, these results suggest an important role for SHIP1 in microglial biology at the non-disease state and in AD pathogenesis. SHIP1 levels might be differentially altered based on disease stage, microglia state or levels of pathological burden, and studying SHIP1 biology with these caveats in mind might reveal therapeutic avenues that can be tailored to these different contexts.

Although the inflammasome can found in many cell types in the brain, the inflammasome, and in particular the NLRP3 inflammasome, has been most commonly studied in microglia [177, 178]. While details of the inflammasome pathway have been deeply studied, much less is known about the functional consequences of inflammasome activation in immune cells on other cell types in the brain and the signaling pathways altered in these other cell types. In this section, we consider the consequences of microglial NLRP3 inflammasome activation on other brain cell types.

Microglia actively regulate neuronal synapses in a healthy physiological setting, and processes mediating this function can become aberrant in a neurodegenerative disease context, as in early stages of AD [15, 143, 179]. Microglial NLRP3 inflammasome activation can perturb neuronal morphology and function during the chronic inflammation that is observed in AD. Multiple groups have shown that inhibition of the NLRP3 inflammasome can rescue deficits in neuronal synaptic transmission and plasticity as well as reduction of neurite number and integrity that results from LPS administration in mice [180–182]. These results provide evidence that the NLRP3 inflammasome mediates microglial contributions to neurodegeneration, as inflammasome activation has deleterious effects on neuron physiology and health.

Microglial activation or inflammatory insults induce astrocytes into a reactive state, which can have both neuroprotective and neurotoxic effects [17, 183–185]. It is thought that the “neurotoxic” reactive astrocytes associated with AD release pro-inflammatory cytokines and can no longer support neuronal synapse formation, facilitate synaptic pruning and clearance of myelin debris and may even induce cell death in neighboring neurons, thereby promoting further neurodegeneration and inflammation [17, 186]. Conversely, the “neuroprotective” phenotype of astrocytes promotes neuronal survival and homeostasis, as these astrocytes secrete neurotrophic factors and anti-inflammatory cytokines [184, 187, 188]. Currently, published studies suggest that NLRP3 inflammasome activation and subsequent secretion of IL-18 and IL-1β from microglia drives astrocytes into the neurotoxic reactive state, which induces upregulation of the secretion of the complement protein C3 and results in lower synapse numbers [180, 189–191]. However, these studies evaluated astrocyte state upon NLRP3 inflammasome activation based on the binary “A1/A2” astrocyte description, which does not fully capture the diversity of astrocyte states and phenotypes in different contexts [183]. Therefore, further studies are warranted to evaluate changes in astrocyte state and function upon microglial NLRP3 activation in an unbiased manner, without, for example, testing the expression of a selected “cocktail” of marker genes.

Microglial activation of the NLRP3 inflammasome can also have deleterious effects on oligodendrocyte biology. Aberrant activation of the microglial NLRP3 inflammasome has been implicated in the onset and progression of multiple sclerosis (MS), an autoimmune demyelinating disease of the CNS [192, 193]. In a commonly employed mouse model for MS, experimental autoimmune encephalomyelitis, microglial NLRP3 inflammasome activation resulted in pyroptosis of oligodendrocytes, promoted inflammation and induced demyelination [194, 195]. These results invite speculations for a potential role of the microglial NLRP3 inflammasome activation in loss of oligodendrocytes and demyelination also seen in AD [196].

Genetic studies have revealed new insights into the importance of microglia in pathways leading to AD. These studies have inspired the development of a multitude of new rodent and human stem cell experimental systems for studying the function of these genes. The use of these systems, coupled to studies of brain tissue from AD patients, have yielded insights into aspects of the mechanisms that link dysfunction of LOAD genes to the accumulation of plaques and tangles, loss of synapse and dementia. Additionally, mechanistic studies of diseases of peripheral myeloid cells have provided clues into the signaling pathways involved in inflammasome activation and the molecular consequence of dysfunction of genes such as INPP5D. While these studies have yielded new insights regarding AD pathogenesis, there remain many unanswered questions that will need to be answered to translate these findings into new therapeutic strategies for AD.

Studies of the function of myeloid-expressed LOAD candidate genes have revealed that the molecular pathways involved are complex and tightly regulated. A recurring theme is that alterations of these genes have context-dependent effects that can be differential with disease stage and exposure to different environmental cues. This appears to be true for SHIP1 function and the role of phosphoinositide signaling broadly in brain health, as well as for the signaling pathways involved in priming and activation of the inflammasome. Understanding the intricate interplay between Aβ and tau clearance, phosphoinositide signaling and inflammasome activation will be important for identifying appropriate therapeutic strategies for AD. In addition, key next steps will involve better understanding the mechanisms linking these processes in microglia and other immune cells to consequences on other cell types in the brain that lead to dementia.

The authors would like to thank Richard Pearse for critical reading of the review, and all Young-Pearse lab members for their input on the interpretations of the literature.

GT and TLY-P wrote the manuscript.

This work was supported by NIH grants P01AG015379,RF1NS117446,R01AG055909, U01AG072572 and U01AG061356.

Not applicable.