CX3CR1 deficiency aggravates amyloid driven neuronal pathology and cognitive decline in Alzheimer’s disease

Background Despite its identification as a key checkpoint regulator of microglial activation in Alzheimer’s disease, the overarching role of CX3CR1 signaling in modulating mechanisms of Aβ driven neurodegeneration, including accumulation of hyperphosphorylated tau is not well understood. Methodology Accumulation of soluble and insoluble Aβ species, microglial activation, synaptic dysregulation, and neurodegeneration is investigated in 4- and 6-month old 5xFAD;Cx3cr1+/+ and 5xFAD;Cx3cr1−/− mice using immunohistochemistry, western blotting, transcriptomic and quantitative real time PCR analyses of purified microglia. Flow cytometry based, in-vivo Aβ uptake assays are used for characterization of the effects of CX3CR1-signaling on microglial phagocytosis and lysosomal acidification as indicators of clearance of methoxy-X-04+ fibrillar Aβ. Lastly, we use Y-maze testing to analyze the effects of Cx3cr1 deficiency on working memory. Results Disease progression in 5xFAD;Cx3cr1−/− mice is characterized by increased deposition of filamentous plaques that display defective microglial plaque engagement. Microglial Aβ phagocytosis and lysosomal acidification in 5xFAD;Cx3cr1−/− mice is impaired in-vivo. Interestingly, Cx3cr1 deficiency results in heighted accumulation of neurotoxic, oligomeric Aβ, along with severe neuritic dystrophy, preferential loss of post-synaptic densities, exacerbated tau pathology, neuronal loss and cognitive impairment. Transcriptomic analyses using cortical RNA, coupled with qRT-PCR using purified microglia from 6 month-old mice indicate dysregulated TGFβ-signaling and heightened ROS metabolism in 5xFAD;Cx3cr1−/− mice. Lastly, microglia in 6 month-old 5xFAD;Cx3cr1−/− mice express a ‘degenerative’ phenotype characterized by increased levels of Ccl2, Ccl5, Il-1β, Pten and Cybb along with reduced Tnf, Il-6 and Tgfβ1 mRNA. Conclusions Cx3cr1 deficiency impairs microglial uptake and degradation of fibrillar Aβ, thereby triggering increased accumulation of neurotoxic Aβ species. Furthermore, loss of Cx3cr1 results in microglial dysfunction typified by dampened TGFβ-signaling, increased oxidative stress responses and dysregulated pro-inflammatory activation. Our results indicate that Aβ-driven microglial dysfunction in Cx3cr1−/− mice aggravates tau hyperphosphorylation, neurodegeneration, synaptic dysregulation and impairs working memory. Supplementary information The online version contains supplementary material available at 10.1186/s13024-022-00545-9.


Background
Neuroinflammation and glial activation in Alzheimer's disease (AD) precede the onset of extracellular β-amyloid (Aβ) plaque deposition and evolve during the development of neurofibrillary tangles (NFTs) and neurodegeneration. These underlying signaling pathways are postulated to lay the framework for cognitive impairment seen in AD [1]. Single-nuclei RNA sequencing (snRNA-seq) data have indicated that distinct clusters of activated microglia are associated with Aβ vs. NFTs vs. inflammatory signaling in AD [2,3], and molecular networks between these unique diseaseassociated glial clusters can reciprocally shape their activation profiles [4]. Lastly, large scale GWAS studies identifying single-nucleotide polymorphisms (SNPs) proximal to microglial-enriched, innate immune genes like CR1, CD33, MEF2C, HLA-DRB5-DRB1, PTK2b and TREM2 that significantly increase the risk of AD [5] have further underscored the critical role of microglia in neurodegeneration and disease progression.
Recently, loss-of-function variants in the microglial fractalkine receptor (CX3CR1) have been associated with worsened Braak staging in AD and neurodegeneration along with reduced survival in amyotrophic lateral sclerosis (ALS), implicating these SNPs as disease-modifying variants in neurodegenerative diseases [6][7][8]. Signaling via CX3CR1 and its neuronal ligand CX3CL1 represents one of multiple important neuro-glial communication axes that maintain microglial homeostasis [9,10]. Cx3cr1 deficiency is associated with a transient deficit in microglial abundance during early post-natal development [11,12]. This correlates with reduced synaptic engulfment and deficits in synapse maturation and elimination during early post-natal development [11,12], resulting in decreased functional brain connectivity, impaired social-interaction and increased autism-like, repetitive behaviors in adult mice [11]. CX3CR1-signaling not only actively dampens microglial phagocytosis and neurotoxic/pro-inflammatory activation during healthy aging, but also inhibits NMDA and glutamate dependent Ca 2+ influx into neurons, thereby protecting against neuronal excitotoxicity [13][14][15]. Indeed, disruption of Cx3cr1 increases pro-inflammatory microglial signaling that correlates with heightened loss of dopaminergic neurons in the substantia nigra in murine models of Parkinson's disease [16,17]. Similar results have been reported in the SOD1-G93A model of ALS, where Cx3cr1 deficiency is associated with increased SOD1 aggregation, neuronal loss and inflammatory microglial activation with impaired autophagy-lysosomal degradation pathways and autophagosome maturation [18,19]. These studies highlight the context-dependent nature of neuroprotective, microglial CX3CR1 signaling.
Single-cell RNA sequencing (scRNA-seq) studies have suggested that downregulation of Cx3cr1 in plaqueassociated microglia is a primary event in the neuroinflammatory cascade in neurodegenerative diseases [20][21][22]. Attenuated Cx3cr1 expression represents a shift in microglia towards a protective phenotype associated with increased expression of Trem2, Apoe, CD68, Axl, MerTK and Lpl, indicative of enhanced capacity for TREM2-dependent plaque compaction, Aβ phagocytosis and lipid metabolism [20]. In contrast, activation of TREM2-APOE signaling in AD results in a subset of microglia that display a 'neurodegenerative phenotype' , characterized by suppression of homeostatic genes like Cx3cr1 and P2ry12 and upregulation of pro-inflammatory markers like Clec7a, Ccl5, Ccl2, IL1b, and Nos2. These neurodegenerative microglia are associated with neuritic plaques, degenerating neurons and trigger a loss of tolerogenic responses in EAE and ALS [21]. Thus, downregulation of Cx3cr1 has distinct effects on plaque clearance and subsequent neurotoxicity in AD.
Our previous studies using murine models of AD deficient in Cx3cr1 have provided insights into the divergent role of fractalkine signaling in amyloidosis and tauopathy [23][24][25]. In the APPPS1 and R1.40 mouse models of amyloidosis, Cx3cr1 deficiency results in a gene-dose dependent reduction in fibrillar Aβ (fAβ) burdens in early disease, suggesting that a loss of CX3CR1 is beneficial in Aβ pathogenesis. Interestingly, 4 month-old APPPS1;Cx3cr1 −/− mice display impaired microglial plaque engulfment with altered activation of plaqueassociated microglia as evidenced by reduced CD68 and CD45 immunoreactivity [25]. By contrast, in the hTau model of tauopathy, loss of Cx3cr1 enhances microglial activation, increases accumulation of Gallays + neuronal NFTs and exacerbates cognitive dysfunction [23]. Furthermore, reactive microglia in hTau;Cx3cr1 −/− mice drive the spread of pathological pTau, possibly via the IL-1β signaling pathway [24]. Overall, our data suggests that while the loss of microglial CX3CR1 may enhance plaque clearance in early stages of AD, it may aggravate long term neurodegeneration. However, given that pathological Aβ is the primary trigger of the neuropathological cascade in AD, insights on how Aβ-triggered microglial activation shapes subsequent Keywords: Amyloid, CX3CR1, Microglia, Neurodegeneration, Tau Page 3 of 21 Puntambekar et al. Molecular Neurodegeneration (2022) 17:47 plaque associated neurotoxicity and how CX3CR1 signaling affects these pathways are still lacking.
In this report, we investigate how microglial responses to Aβ shaped by CX3CR1 drive long term neurotoxicity in the 5xFAD model of amyloid disease. Our results indicate that Cx3cr1 deficiency exacerbates neurodegeneration and cognitive impairment with disease progression in 5xFAD animals by driving increased accumulation of neurotoxic oligomeric Aβ (oAβ), fibrillar Aβ (fAβ) plaques, intraneuronal inclusions of hyperphosphorylated tau (pTau) and enlarged foci of neuritic dystrophy. These effects are potentiated, in part, due to microglial dysfunction in 5xFAD;Cx3cr1 −/− mice as evidenced by impaired microglial Aβ-phagocytosis and clearance as well as aberrant TGFβ-signaling, inflammatory activation and reactive oxygen species (ROS) metabolism. Lastly, in contrast to 5xFAD;Cx3cr1 +/+ mice in which pTau pathology strongly correlates with accumulation of compact plaques and small foci of neuritic dystrophy, aggravated deposition of pathological tau in the absence of Cx3cr1 is correlated with, oligomeric Aβ, filamentous Aβ and large dystrophic neurites.

Materials and methods
Animals 5xFAD mice on the C57BL/6 J background were obtained through Jackson Laboratories in collaboration with the Model-AD Center at the IU School of Medicine (Stock # 34,848-JAX). 5xFAD mice were maintained as hemizygotes for the APP and PSEN1 transgenes (5xFAD ± ). B6.129P2(Cg)-Cx3cr1tm1Litt/J mice on the congenic background were purchased through Jackson Laboratories (Cx3cr1 −/− : Stock # 005,582). In these mice, the first 390 base-pairs of the coding exon 2 of the Cx3cr1 gene are replaced by the enhanced green fluorescent protein (EGFP) sequence, thereby resulting in a loss of Cx3cr1 gene expression. Trem2 −/− mice on the C57BL/6 J background were purchased through Jackson Laboratories (Stock # 027,197) and mated with 5xFAD ± mice to generate 5xFAD;TREM2 −/− mice. 5xFAD ± mice were mated with Cx3cr1 −/− animals, and progeny were subsequently intercrossed to generate 5xFAD ± ;Cx3cr1 −/− cohorts (henceforth referred to as 5xFAD;Cx3cr1 −/− ). Pathology in the absence of Cx3cr1 was compared to age-matched 5xFAD ± ;Cx3cr1 +/+ (henceforth referred to as 5xFAD;Cx3cr1 +/+ ) mice throughout the study. Agematched littermates that do not express 5xFAD mutations (C57BL/6 J;5xFAD −/− , henceforth referred to as B6) were used as controls throughout. All animals were housed in animal facilities within Stark Neurosciences Research Institute (SNRI) at The Indiana University School of Medicine (IUSM), accredited by the Association and Accreditation of Laboratory Animal Care. Animals were maintained according to USDA standards and the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All experiments were approved by the IUSM Institutional Animal Care and Use Committee.

Immunohistochemistry
Three Male and 3 female mice of each genotype were used for all histochemical analyses, based on power analyses for an 80% probability of detecting a 25% change in AD pathological outcomes, based on previous publications [25][26][27]. Brain tissues fixed in 4% PFA were cryoprotected in 30% sucrose and embedded in OCT. Embedded brains were processed into free-floating, 30µms thick, sagittal sections using a Leica Cryostat and stored at -20ºC in cryostorage solution. Immunofluorescence staining was done as previously described [25]. Briefly, sections were washed with 1XPBS and incubated with a 10 mM sodium citrate solution containing tween-20 (pH = 6.00) for 15 min, at 90ºC to quench microglial EGFP. Sections were cooled for 30 min at room temperature (RT) and incubated in a blocking solution containing 5% normal donkey serum (Sigma Aldrich; Cat # D9663-10ML) and 0.5% Triton-X100 (Sigma Aldrich; Cat # X100-100ML) for 1 h at RT. Mouse on Mouse (MOM) Blocking Reagent (1:1000, Vector Laboratories) was added to the blocking solution if primary antibodies used were generated in mouse or rat. Tissues were incubated overnight at 4ºC in blocking solution with primary antibodies, followed by incubation species-specific Alexa-fluor conjugated secondary antibodies (1:1000, Life Technologies). Sections were mounted on Superfrost Plus glass slides and air-dried for 30-45 min. For visualization of fibrillar Aβ 42 plaques, slides were dipped in a 1% Thioflavin-S (ThioS) solution, de-stained with 70% ethanol and washed with 1XPBS prior to being coverslipped with ProlongGold.
For 3,3′-diaminobenzidine (DAB) staining, sections were incubated in a solution of 1% H 2 O 2 in 1XPBS for 30 min at RT to quench endogenous peroxidases. Antigen retrieval was performed by incubating sections in 10 mM sodium citrate (pH = 6.00) at 90ºC for 10 min. Sections were incubated in blocking solution containing 5% normal goat serum, 0.5% Triton-X100 and Mouse-On-Mouse (M.O.M) blocking reagent (1:1000) for 1 h at RT. Tissues were incubated overnight at 4ºC in blocking buffer with Mouse α-AT8. Following incubation with biotinylated α-mouse IgG (1:200, Invitrogen, Cat # B-2763), slices were developed using the VECTASTAIN Elite ABC kit (Vector Laboratories) and DAB. Sections were mounted and coverslipped using Permount (FisherScientific, Cat # SP15-100). Details for all primary antibodies used for immune-fluorescent and DAB staining are listed in Supplementary Table 1.

Imaging and analysis
High resolution fluorescence imaging was done using the Nikon AR1 confocal microscope. High-resolution, brightfield images were acquired using the CTR5000 upright Leica microscope. Post processing and analysis was done using ImageJ (National Institutes of Health).
Branching and junction analysis for Iba1 + microglia in B6;Cx3cr1 +/+ and B6;Cx3cr1 −/− mice was done using the 'Skeletonize' and ' Analyze Skeleton' plugins in ImageJ. For analysis of plaque circularity and Iba1 occupancy, 15-20 μm Z-stacks were imaged at a 60X magnification. Circularity analysis on ThioS + plaques was done using the 'Shape Descriptors' plugin in ImageJ. Cut offs for plaque circularity were defined as previously published by Yuan P. et al. [28], where filamentous plaques had a circularity score of 0.00-0.14 and compact plaques had circularities greater than 0.30. Plaques with circularity scores between 0.15-0.28 were classified as having 'intermediate' phenotypes. To quantify microglial process-engagement with ThioS + plaques, regions-of-interest (ROIs) were traced along ThioS + plaque borders in serial sections co-stained with ThioS and Iba1. Defined ROIs were applied to the Iba1 layer, and the percentage of area within the ROI positive for Iba1 immunoreactivity was quantified. Neuronal numbers within the subiculum were counted automatically using the particle analysis feature. The watershed plugin was used for segmentation of tightly packed NeuN + cell bodies. Dystrophic neurites were counted manually. For size classification, the line selection tool was used to manually threshold out particles less than 500 µm (50-500 µm), followed by subtraction of the number of dystrophic neurites > 500 μm (500-1000 µm) from the total number of dystrophic neurites.

Cx3cr1 deficiency does not alter the homoeostatic microglial phenotype in adult mice
Studies have shown a transient decrease in microglial densities in brains of B6;Cx3cr1 −/− mice between postnatal days 8 through 28 [12]. Similarly, sc-RNA seq studies that have demonstrated that transcriptomic differences observed in FACS purified, CD11b + CD45 + microglia from 2 month-old B6;Cx3cr1 −/− mice are not evident in aged microglia from 12 and 24 month-old Cx3cr1 deficient mice as compared to cells isolated from agematched B6;Cx3cr1 +/+ animals [31]. These studies imply that Cx3cr1 deficiency has transient and subtle effects on the transcriptional landscape of homeostatic microglia. To ascertain that no overt microglial defects persist into adulthood, we examined 6 month-old B6;Cx3cr1 +/+ and B6;Cx3cr1 −/− mice for microglial abundance and homeostatic activation. qRT-PCR analyses using cortical mRNA revealed similar expression of canonical microglial genes namely Pu.1, Iba1, P2ry12 and Cd11b (Supplementary Fig. 1Ai-Aiv). Furthermore, flow-cytometry analyses revealed similar numbers of CD11b + cells in the brains of B6 mice with and without Cx3cr1 (Supplementary Fig. 1B). Interestingly, while no differences were observed in the number of cellular processes between the two genotypes, Cx3cr1 −/− microglia displayed a modest increase in the number of process junctions as compared to microglia from B6;Cx3cr1 +/+ mice ( Supplementary  Fig. 1C). Lastly, with the exception significantly reduced levels of Cst7, we observed no significant differences in the expression of genes associated with homeostatic or inflammatory microglia in 6 month-old B6;Cx3cr1 +/+ and B6;Cx3cr1 −/− mice ( Supplementary Fig. 4A, B). Taken together with these published studies, our data show similar homeostatic microglial signatures in 6 month-old B6 mice with and without Cx3cr1.

Cx3cr1 deficiency leads to accelerated plaque deposition
To investigate the kinetics of Aβ accumulation in the absence of Cx3cr1, brain sections from 4 and 6 month-old 5xFAD;Cx3cr1 +/+ and 5xFAD;Cx3cr1 −/− mice were immunolabeled with anti-Aβ 1-42 antibodies. As seen in our previous studies using APPPS1 mice [25,26], loss of Cx3cr1 resulted in significantly reduced MOAB2 + plaque load in the cortex and hippocampus in 4-month-old 5xFAD mice ( Fig. 1A-C). By contrast, at 6 months of age, the number of MOAB2 + plaques were significantly increased in the cortex and hippocampus of 5xFAD;Cx3cr1 −/− mice as compared to age-matched 5xFAD;Cx3cr1 +/+ animals ( Fig. 1A-C). 6-month-old 5xFAD;Cx3cr1 +/+ mice showed an ~ 1.5 fold and ~ twofold increase in Aβ 42 plaque loads in the cortex and hippocampus respectively as compared to their 4 month-old counterparts. By contrast, an ~ 2.8 fold and ~ sixfold higher plaque burden in the cortex and hippocampus of 6 month of 5xFAD;Cx3cr1 −/− mice over 4 monthold 5xFAD;Cx3cr1 −/− animals indicated that plaque deposition is accelerated with disease progression in the absence of Cx3cr1 ( Fig. 1A-C).

Cx3cr1 deficiency exacerbates the accumulation of neurotoxic species of Aβ
Insoluble aggregates of fibrillar Aβ (fAβ) along with soluble, oligomeric Aβ (oAβ) are associated with neurotoxicity in AD [32][33][34]. To assess whether the loss of Cx3cr1 alters the accumulation these neurotoxic Aβ species, we first stained serial brain sections with Thioflavin S (ThioS) to visualize fAβ plaques. ThioS + plaques deposited in 5xFAD;Cx3cr1 −/− mice appeared significantly more diffuse when compared to those in 5xFAD;Cx3cr1 +/+ mice (Fig. 1D). Using circularity analysis, to distinguish compact plaques from plaques with a filamentous/diffuse or an intermediate phenotype (Fig. 1E), we observed that Cx3cr1 deficiency resulted in a significant reduction in the proportion of compact plaques in the cortex (Fig. 1F) and hippocampus ( Fig. 1G) of 6-month-old 5xFAD mice, with a concomitant increase in accumulation of plaques with intermediate and diffuse morphologies. To investigate the accumulation of soluble oAβ species, we immune-stained for OC + oAβ at 4-and 6-months of age.
In-situ quantification of OC + Aβ accumulation revealed that increased proportion of cortical areas were positive of OC immunoreactivity in 5xFAD;Cx3cr1 −/− mice throughout the course of the disease (Fig. 1H). Consistent with these results, high-resolution confocal microscopy revealed larger deposits of OC + oAβ surrounding compact and filamentous ThioS + plaques in the cortices of 6-month-old 5xFAD;Cx3cr1 −/− mice compared to similar plaques in 5xFAD;Cx3cr1 +/+ mice (Fig. 1I). While female 5xFAD mice displayed significantly increased plaque burdens compared to males (Fig. 1B, C), no significant differences in OC + oAβ loads were observed in female and male 5xFAD cohorts. Taken together, our results indicate that Cx3cr1 deficiency shifts Aβ dynamics towards increased accumulation/generation of toxic species of soluble oAβ associated with highly filamentous fAβ plaques.

ROS metabolism and oxidative stress responses
Given the aberrant accumulation of toxic Aβ and impaired microglial plaque engagement in 5xFAD;Cx3cr1 −/− mice, we hypothesized that the Aβ driven neuropathological milieu is altered in the absence of Cx3cr1. To investigate how CX3CR1 shapes glial activation and the neuroinflammatory microenvironment, we ran the nCounter ® Neuroinflammation Panel which queries 770 genes involved in neuron-glia interactions, inflammation, and neuroplasticity (Supplemental Files 8,9). Transcriptional analyses using cortical RNA from 6 month-old 5xFAD animals revealed upregulation of genes associated with cellular apoptosis (Casp9, Casp3, Casp8) and pro-survival signaling (Bcl2l1) in 5xFAD;Cx3cr1 −/− mice when compared to Cx3cr1 +/+ counterparts (Fig. 3Ai-Aii). Interestingly, we observed increased expression of pro-inflammatory genes (Ccl2, Ccl5), along with increased Cst7, P2ry17 and Tgfbr1 levels in 5xFAD;Cx3cr1 −/− mice ( Fig. 3Ai-Aii). Lastly, genes associated with nitric-oxide signaling, ROS production and oxidative stress responses were differentially altered in 5xFAD;Cx3cr1 −/− animals (Pten, Pink1, Nostrin, Sod2, Anxa1, Lcn2) ( Fig. 3Ai-Aii). Gene-ontology analysis of differentially expressed genes (DEGs) revealed that in comparison with their female counterparts that displayed regulation of programmed cell death/apoptosis, regulation of ROS metabolism and regulation of TGFβ3 signaling as the key biological pathways affected by the loss of Cx3cr1, top processes affected in male 5xFAD;Cx3cr1 −/− mice were associated with signaling related to cell cycle arrest in response to DNA damage and ER stress (Supplemental Fig. 3A). Despite these differences, common pathways altered in female and male 5xFAD;Cx3cr1 −/− mice indicated increased cellular apoptosis/necroptosis, altered oxidative/ER stress, increased DNA damage and cell cycle arrest (Supplemental Fig. 3B). Additionally, KEGG enrichment analysis revealed that loss of CX3CR1 signaling may result in altered phagocytosis along with alterations in key signaling pathways with known involvement in intracellular protein shuttling, protein phosphorylation, cellular senescence, synaptic plasticity/ transmission, and cognition (Supplemental Fig. 3C).

Discussion
Distinct microglial clusters associated with Aβ plaques, NFTs and dystrophic/degenerating neurons, which are further critically impacted by signaling pathways such as CX3CR1, TREM2, PLCγ2 etc. have been identified in the AD brain. These data indicate that the dynamic neurodegenerative micro-environment is defined not only by microglial interaction with specific pathological AD features, but also by interactions between multiple, microglia-specific signaling pathways [2][3][4]. While current literature in the field is largely focused on investigating microglial pathways that confer increased risk for developing AD (e.g. TREM2, APOE, PLCγ2), diseasemodifying SNPs around microglial genes, particularly the loss-of-function CX3CR1-V249I variant, have been associated with increased neuronal loss and disease severity in macular degeneration, ALS and AD [7,8,38]. Taken together with sc-RNA seq studies identifying downregulation Cx3cr1 in plaque associated microglia as a molecular event that shapes microglial activation in AD [20][21][22], CX3CR1-signaling has emerged as a critical determinant of long-term pathological outcomes in AD. However, the downstream effects of attenuated CX3CR1 on the microglial phenotype and subsequent mechanisms that impact neurodegenerative pathology in AD remain elusive, in part due to the lack of comprehensive, transgenic animal models [14,39]. Using 5xFAD animals, we demonstrate that in addition to skewing the microenvironment towards the accumulation of increasingly neurotoxic Aβ species, the loss of CX3CR1 signaling in AD results in impaired microglial plaque engagement, dampened microglial Aβ phagocytosis along with impaired lysosomal activation and skewing of microglia towards a 'degenerative' phenotype. Accumulation of dysfunctional microglia is associated with aggravated neuritic dystrophy, synaptic dysfunction, tau hyperphosphorylation, increased neurodegeneration and cognitive impairment. Disease phenotypes observed in the absence of Cx3cr1 are reminiscent of pathology seen in mouse models with AD associated SNPs in key microglial genes such as TREM2 and APOE [2,21,22,28]. These data underscore that highly dynamic microglial responses in AD are simultaneously shaped by interactions between multiple signaling pathways. For instance, our data suggests elevations in microglial Apoe and Il-1β in 5xFAD;Cx3cr1 −/− mice (Fig. 3), which may drive neurotoxic microglial responses via TREM2-APOE signaling axis [21]. Studies using the PS19 mouse model of tauopathy have demonstrated that TREM2 signaling aggravates NFT pathology, associated with pro-inflammatory, Il-1β + microglia [40]. Taken together with published studies showing that increased pTau pathology in hTau;Cx3cr1 −/− mice is driven by microglial IL-1β [23,24], our data suggests that Cx3cr1 deficiency may drive pTau accumulation in 5xFAD mice via microglial TREM2-IL1β signaling. Neuronal CX3CL1, the ligand for microglial CX3CR1, is cleaved by ADAM10, ADAM17 and BACE1, resulting in a membrane-bound C-terminal fragment (-ct). CX3CL1-ct is in-turn cleaved by γ-secretase to release the intracellular domain of CX3CL1 (-ICD). Using transgenic 5xFAD mice that overexpress CX3CL1ct (5xFAD;CX3CL1-ct), Fan et. al. have recently demonstrated that CX3CL1-ICD promotes neurogenesis, potentially via the TGFβ3-Smad2 pathway. 5xFAD;CX3CL1-ct mice show decreased plaque loads and reduced neuronal loss [27,41]. Lastly, overexpression of CX3CL1 in the PS19 model of tauopathy, reduces synaptic and neuronal loss and improves memory and cognition [41]. Interestingly, TGFβ3-signaling, regulation of synaptic plasticity and learning/memory/cognition are among the top biological/cellular processes altered in 6-month-old 5xFAD;Cx3cr1 −/− mice (Supplemental Fig. 3). However, in contrast to results described above, we see a significant increase in neurotoxicity in the subiculum (Fig. 7), and increased accumulation of pTau + neurons and dystrophic neurites in the subiculum, cortex and hippocampus (Fig. 6) in the absence of Cx3cr1. Thus, while CX3CL1 signaling to neurons and TGFβ3mediated neurogenesis may potentially be upregulated in 5xFAD;Cx3cr1 −/− mice, our data suggest that this is not sufficient to reverse/reduce neuronal loss and ameliorate long-term cognitive decline. We hypothesize that synergistic effects of CX3CL1 signaling via microglial-CX3CR1 and CX3CL1 back-signaling to neurons are required for amelioration of pathology and increased neuronal preservation in AD.
Ultrastructural analyses of microglia in mice exposed to chronic social stress have revealed that Cx3cr1 deficiency aggravates the accumulation of 'dark microglia' . These cells display signs of oxidative stress, including an electron-dense cytoplasm and nucleoplasm, cytoplasmic fragmentation, dilation of Golgi-and ER membranes and mitochondrial alterations [42]. Pathological, Cx3cr1 −/− microglia show downregulation of homeostatic markers like P2RY12 and display extensive engulfment of presynaptic axon terminals and postsynaptic dendritic branches. Similar 'dark' microglia associated with Aβ plaques in APPPS1 mice [42][43][44] are postulated to be responsible for aberrant synaptic stripping and pathological remodeling of synaptic circuits. Interestingly, our transcriptomic analysis has identified dysregulation of NO/ROS metabolism, responses to oxidative and ER stress, synaptic plasticity and synaptic vesicle cycling in 6 month old 5xFAD;Cx3cr1 −/− mice (Supplemental Fig. 3B-C), along with increased mRNA expression for Pten and Cybb/Nos2 with a decrease in Pink1 levels in microglia purified from 5xFAD;Cx3cr1 −/− mice (Fig. 3). Furthermore, while we demonstrate a preferential loss of post-synaptic densities in 5xFAD;Cx3cr1 −/− mice (Fig. 7A, D), 5xFAD;Cx3cr1 +/+ and 5xFAD;Cx3cr1 −/− animals display modest but significant increases in pre-synaptic Sv2a and Synaptophysin levels as compared to their genotype matched B6 controls (Fig. 7A, C). We hypothesize that a) the loss of Cx3cr1 may result in a significant reduction in pre-and post-synaptic coupling into active synapses and b) an increased recruitment of signaling proteins to pre-synaptic terminals may be reflective of compensation for the lack of synaptic coupling/plasticity observed in our mice (Supplemental Fig. 3). Why post-synaptic terminals are particularly vulnerable in AD, and what drives aggravated post-synaptic loss in the absence of Cx3cr1 is not completely understood and remains an important question for targeted neuroprotective therapies. We show that the pathological microenvironment in 5xFAD;Cx3cr1 −/− mice comprises of increased levels of highly diffuse, fAβ plaques as well as toxic, soluble (OC + ) oAβ (Fig. 1), suggesting an impairment in amyloid clearance mechanisms. Studies have indicated that prolonged exposure to toxic Aβ peptides downregulates expression of phagocytic receptors such as CD47, CD36 and RAGE [45], impairs microglial autophagy and induces lysosomal dysfunction in AD [46], an observation corroborated invivo by significant alteration of endocytosis/phagocytosis and AGE-RAGE signaling pathways (Supplemental Fig. 3) coupled with reduced fAβ uptake and lysosomal acidification in CD11b + microglia from 5xFAD;Cx3cr1 −/− mice in this study (Supplemental Fig. 6, Fig. 4).
Studies using stereotaxic injections of pathological human-AD tau into 5xFAD and APP-KI mice have shown that seeding and spread of pTau is affected by plaque burdens, where increased Aβ loads exacerbate MAPT pathology and lead to worsening of memory deficits [47]. Interestingly, studies in the ArcTau, Tg2576 and hAPP-J20 mouse models of AD have shown that soluble, OC + oAβ, rather than insoluble Aβ plaques facilitate pTau pathology and cognitive deficits [37,48]. Recent GWAS studies that have implicated CX3CR1-V249I (rs3732349), a putative loss-of-function variant previously associated with age-related macular degeneration, in worsened NFT pathology and Braak staging in AD, and neurodegeneration in ALS [6-8, 38, 49]. In line with these data, we observe an overall increase in tau hyperphosphorylation and accumulation of AT8 + pTau in neuronal soma along with mislocalization of pTau into axons, that significantly correlates with OC + oAβ accumulation in the absence of Cx3cr1. Furthermore, our previous work in hTau;Cx3cr1 −/− mice has implicated reactive, inflammatory, CD68 + , IL1β + microglia in active spread of neurotoxic tau [23,24]. Our studies also show an increase in pro-inflammatory microglia in 5xFAD;Cx3cr1 −/− mice with increased transcript levels of Ccl2, Ccl5 and Il1β along with a downregulation of Tgfβ1 mRNA expression. Coupled with increased Clec7a, Apoe and Cybb/ Nos2 levels, microglia in 5xFAD;Cx3cr1 −/− mice show a phenotype reminiscent of 'degenerative' microglia, associated with apoptotic neurons in AD and EAE. This degenerative microglial phenotype correlates with an overall increase in neurodegenerative pathology including severe neuritic dystrophy, aggravated neuronal loss in the subiculum, and worsened memory deficits in 6 month-old 5xFAD;Cx3cr1 −/− mice. Interestingly, we observed increased levels of total tau in soluble fractions of 5xFAD brain lysates, as compared to their respective B6 controls. We hypothesize that this increase is reflective of the release of tau from degenerating / dying neurons into the interstitial space, which in-turn potentiates the seeding and spreading of neurotoxic tau [50]. We postulate that the increased availability of these soluble tau seeds, in conjunction with elevated oAβ and filamentous fAβ levels (Fig. 1, Fig. 6D) drive exacerbated pTau pathology observed in 5xFAD;Cx3cr1 −/− mice. Interestingly, Fuhrmann et. al. have demonstrated that the loss of Cx3cr1 reduces neuronal loss in cortical layer III in the 5xTg mouse model of AD [51]. This study showed that active polarization of microglial processes towards layer III neurons is required for neuronal elimination in 5xTg mice. Unlike our results which show heightened neuronal pTau aggregation in layer III (Fig. 6), eliminated neurons in 5xTg mice display increased aggerates of Aβ 42 but not pTau. While we have not pursued an in-depth stereological quantification of cortical neurons, our preliminary observations indicate heightened neurodegeneration in cortices of 5xFAD;Cx3cr1 −/− mice.
Taken as a whole, these data indicate that a) active elimination of neurons by neurodegenerative microglia may be dictated by the quality of intraneuronal pathological protein aggregates and b) microglial neurodegenerative responses modulated by CX3CR1 may be subject to region-dependent microglial heterogeneity along with regional differences in neuronal susceptibility [44,52,53]. We hypothesize that aggravated pTau pathology and subsequent neuronal loss in 5xFAD;Cx3cr1 −/− mice may be potentiated, in part, by elevated levels of neurotoxic oAβ along increasingly neurotoxic, filamentous plaques, resulting in aggravated spread of neurotoxic tau by pathological, inflammatory microglia.

Conclusions
Clinical outcomes in AD are a net sum of a range of pathological microglial activation states that are dynamically controlled by CX3CR1-signaling in response to the diverse pathological features of AD. Thus, CX3CR1 may simultaneously mediate distinct microglial phenotypes based on contextual cues received via their interaction with Aβ, dysfunctional synapses, dystrophic neurites or distressed neurons within their microenvironment. While our understanding of the overarching role of microglial CX3CR1 on disease mechanisms that affect neuronal and synaptic health is still primitive, this study is the first to provide an in-vivo link between CX3CR1-dependent microglial activation, aberrant homeostasis of soluble and insoluble Aβ and subsequent effects on chronic neurodegeneration in AD. This work suggests that the long term neurodegenerative changes including accumulation of pathological tau, synaptic dysregulation and altered neuronal homeostasis in the absence of Cx3cr1 correlate with the quality along with the abundance of extracellular Aβ. We hypothesize that Cx3cr1 deficiency in AD impairs microglial endolytic activation, thereby affecting uptake and degradation of fibrillar Aβ, and triggering an accumulation of neurotoxic Aβ species. Lastly, the neurotoxic Aβ microenvironment in the absence of Cx3cr1 further drives increased microglial dysfunction typified by aberrant inflammatory activation, ROS metabolism and a skewing of the microglial response to a neurodegenerative phenotype.