Csaba Fekete1,2, Csaba Vastagh1, Ádám Dénes3, Erik Hrabovszky1, Gábor Nyiri4,Imre Kalló 1,5, Zsolt Liposits1,5 and Miklós Sárvári1*

Abstract
Microglia are instrumental for recognition and elimination of amyloid β1-42 oligomers (AβO), but the long-term consequences of AβO-induced inflammatory changes in the brain are unclear. Here, we explored microglial responses and transciptome-level inflammatory signatures in the rat hippocampus after chronic AβO challenge. Middle-aged Long Evans rats received intracerebroventricular infusion of AβO or vehicle for 4 weeks, followed by treatment with artificial CSF or MCC950 for the subsequent 4 weeks. AβO infusion evoked a sustained inflammatory response including activation of NF-κB, triggered microglia activation and increased the expression of pattern recognition and phagocytic receptors. Aβ1-42 plaques were not detectable likely due to microglial elimination of infused oligomers. In addition, we found upregulation of neuronal inhibitory ligands and their cognate microglial receptors, whilst downregulation of Esr1 and Scn1a, encoding estrogen receptor alpha and voltage-gated sodium-channel Na(v)1.1, respectively, was observed. These changes were associated with impaired hippocampus-dependent spatial memory and resembled early neurological changes seen in Alzheimer’s disease. To investigate the role of inflammatory actions in memory deterioration, we performed MCC950 infusion, which specifically blocks the NLRP3 inflammasome. MCC950 attenuated AβO-evoked microglia reactivity, restored expression of neuronal inhibitory ligands, reversed downregulation of ERα, and abolished memory impairments. Furthermore, MCC950 abrogated AβO-invoked reduction of serum IL- 10. These findings provide evidence that in response to AβO infusion microglia change their phenotype, but the resulting inflammatory changes are sustained for at least one month after the end of AβO challenge. Lasting NLRP3-driven inflammatory alterations and altered hippocampal gene expression contribute to spatial memory decline.

Keywords:amyloid β1-42 oligomer, hippocampus, microglia, NLRP3 inflammasome, estrogen receptors, spatial memory

Introduction
Microglia constantly survey the brain parenchyma (Nimmerjahn, Kirchhoff, & Helmchen, 2005) and react promptly to disturbances
to maintain homeostasis (Hanisch & Kettenmann,2007). Microglia are equipped with receptors to monitor neurotransmission, neuronal damage, and to recognize any foreign or harmful substances (Lucin & Wyss-Coray, 2009). Microglial cells are instrumental in the recognition and elimination of amyloid beta (Aβ). Cell surface receptors including scavenger receptors (El Khoury et al., 1996), CD14, TLR2 and TLR4 (Reed-Geaghan, Savage, Hise, & Landreth, 2009) bind Aβ, resulting in activation of NF-κB (Li, Long, He, Belshaw, & Scott, 2015). Activation of NF-κB triggers transcription of a set of proinflammatory genes (Muruve et al., 2008) including IL1B and NLRP3 . Thereafter, encounter of Aβ lysosomal fragments and NLRP3 facilitates formation and activation of the NLRP3 inflammasome complex which promotes the release of IL-1β and IL-18, modulates eicosanoid synthesis, autophagy, metabolism and gene transcription (Rathinam & Fitzgerald, 2016). Activation Immunochemicals of the NLRP3 inflammasome has been demonstrated in Alzheimer’s disease (AD), and suggested that it plays a causative role in disease pathogenesis (Halle et al., 2008; Heneka et al., 2015). Recent preclinical studies have supported this notion. Genetic ablation of Nlrp3 improves memory in APP/PS1 mice (Heneka et al., 2013), whilst pharmacological inhibition of the NLRP3 inflammasome reduces genotype-related microglia activation and enhances memory in transgenic mouse models (Daniels et al., 2016; Dempsey et al., 2016).Aβ peptides form soluble oligomers of various sizes, and aggregate into amyloid fibrils that gradually accumulate as insoluble senile plaques in the AD brain (Ahmed et al., 2010; Burdick et al., 1992; Halverson, Fraser, Kirschner, & Lansbury, 1990; Kirschner et al., 1987; Kuo et al., 1996). The presence of activated microglia, many of which express IL-1 (Griffin et al., 1989), has been demonstrated at the plaques throughout the cerebral cortex (Itagaki, McGeer, Akiyama, Zhu, & Selkoe, 1989; McGeer, Itagaki, Tago, & McGeer, 1987). While Aβ1-42 fibrils are unable to evoke a strong microglial response, soluble Aβ1-42 oligomers (AβO) activate microglia at the low nanomolar concentrations (Maezawa, Zimin, Wulff, & Jin, 2011). Soluble AβO has been implicated in the pathogenesis of AD (Bilousova et al.,2016; Lesne et al., 2006; Muller-Schiffmann et al., 2016; Shankar et al., 2008; Wang et al., 2017). The correlation between the degree of dementia in AD and the level of soluble AβO (Lue et al., 1999) supports this notion.

Infusion of Aβ1-42 into the cerebral ventricles of young adult rats is an established approach to mimic the neurological phenotype of late-onset AD (Forny-Germano et al., 2014; Garcia- Matas et al., 2010; He et al., 2012; Lecanu, Greeson, & Papadopoulos, 2006). Early studies suggest that Aβ fibrils alone are not effective to induce memory loss, but pro-oxidant agents potentiate the effect of Aβ1-42, and the mixture of Aβ1-42 and pro-oxidants causes microglia activation and memory disturbances (Lecanu et al., 2006). Recent studies have demonstrated that AβO diffuse into the brain parenchyma (Forny-Germano et al., 2014) and evoke a more robust microglial response and memory loss than do Aβ1-42 fibrils (He et al., 2012). Aging remains the major risk for the development of dementia. As age-related changes of the rat hippocampal transcriptome already occur at 6 month of age (Kadish et al., 2009), middle-aged rats have also been used in Aβ infusion experiments (Frautschy et al., 200 1; Frautschy, Yang,Calderon, & Cole, 1996). However, it is currently unclear whether AβO itself induces long- lasting inflammatory changes in the brain and memory deficits in an NLRP3-dependent manner.

In this study, we examined the impact of a 4-week AβO infusion into the lateral cerebral ventricle of middle-aged Long Evans rats on microglia phenotype, inflammatory changes and transcriptomic fingerprints involved in neuron-microglia communication, estrogen signaling, inhibitory interneuron function and spatial memory. We also explored putative mechanisms that may link the microglial and inflammatory responses to AβO infusion and spatial memory impairments sulfoximine and iron sulfate heptahydrate were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and Sigma (St. Louis, MO, USA), respectively. Rabbit polyclonal antiserum to Iba1 was ordered from Wako Chemicals GmbH, (Neuss, Germany). Monoclonal anti-Aβ antibody was obtained from Abcam (Cambridge, UK). Artificial CSF and MCC950 (Coll et al., 2015) were purchased from Tocris (Bristol, UK) and Sigma, respectively.

Aβ1-42 peptide (Bachem, H-1368) was dissolved in 10% (w/v) NH4OH (Ryan et al., 2013). Aliquots were prepared, freeze dried, and stored at -20oC. Prior to use, 1 mg of Aβ1-42 was dissolved by adding 20 µ l of 60 mM NaOH and immediately diluted with 80 µl water. Then, the alkaline solution was neutralized with 1.011 ml of low salt buffer (10 mM HEPES pH7, 10 mM NaCl) to make a 0.2 mM Aβ1-42 solution. The solution was incubated for 4 hours on ice to promote oligomer formation (Ahmed et al., 2010). As pro-oxidant agents potentiate the effect of Aβ (Garcia-Matas et al., 2010; Lecanu et al., 2006), we added 2.5 ml low salt buffer containing 10 mg FeSO4·7H2O and 96 mg buthionine sulfoximine to the AβO solution. The resulting solution consisted of 61.5 µM AβO, 10 mM FeSO4 and 0.12 M buthionine sulfoximine. In parallel, the same solution without Aβ1-42 was prepared and used for the treatment of control animals. Alzet 2004 minipumps (Durect, Cupertino, CA, USA), filled with these solutions, were incubated at room temperature for 16 hours in saline before use.Young male Long Evans rats were obtained from Charles River Laboratories (Calco, Italy). Animals were housed in the animal care facility of Institute of Experimental Medicine for the subsequent months. Eight month old rats were used in the study. APPNL-F/NL-F mice were generated by Takaomi Saido’s laboratory (Nilsson, Saito, & Saido, 2014; Saito et al., 2014). For generation of the transgenic animal, the targeting vector was the mouse APP gene which included the Swedish and the Iberian human APP mutations (Mullan et al., 1992). The Aβ sequence within the mouse APP was humanized. These mice have normal levels of full length APP, but show increased Aβ1-42 :Aβ1-40 ratio characteristic for AD pathology (Hardy, 1997). Animals were kept under normal conditions, with 12/12 hour light cycles, food and water was available ad libitum. Unlike in any other existing AD mouse models, the APPNL-F/NL-F mice reproduce human AD pathology with the highest fidelity. All studies were carried out with permission from the Animal Welfare Committee of Institute of Experimental Medicine (Permission Number: A5769-01), and in accordance with the legal requirements of the European Community (Directive 2010/63/EU). Animal experimentation described was
conducted in accord with accepted standards of animal care.

Surgery was carried out under isoflurane anesthesia, and with buprenorphine pain relief. Stainless steel infusion guide cannula (22G, Plastics One, Roanoke, VA, USA) was implanted into the right lateral cerebral ventricle (D=5.3 mm, L=1.4 mm, AP=0.8 mm). Alzet 2004 minipumps filled with either 0.2 ml of pro-oxidant-containing 61.5 µM AβO solution or vehicle, were connected to the cannula via PE50 tubing (Plastics One) and placed under the skin in the interscapular region. Pumps released 1.67 µg AβO daily for 4 weeks. Dosage Childhood infections was selected based on the work of Frautschy et al. (Frautschy et al., 2001). After AβO infusion, minipumps were replaced with new ones filled with either 0.2 ml of 0.167 µ g/ml MCC950 in artificial CSF (aCSF), or 0.2 ml of aCSF. Dose of MCC950 was calculated based on published data on IC50 for NLRP3 (Coll et al., 2015). Pumps released 1.0 µg MCC950 daily for 4 weeks. Consecutive treatments generated four experimental groups including control group (vehicle + aCSF, n=6), drug control group (vehicle + DRUG, n=3), AβO challenged group (AβO + aCSF, n=6) and MCC950 treatment group (AβO + DRUG, n=8) were examined.

On the fourth week of the second treatment, we started to perform Morris water maze (MWM). We tested spatial acquisition for 4 days, reversal on the next day and spatial learning for 2 days. The maze used in this study was a black circular pool (180 cm in diameter) filled to a 40 cm-depth with 24oC water. The black platform (10 cm in diameter) was hidden 2 cm below the water surface in the SW quadrant. On each day, rats were given four successive trials with a maximum duration of 90 sec, the inter-trial interval was 1 minute. Starting points (N, E, NW, SE) were randomized (Vorhees & Williams, 2006). Multiple color distal cues were placed on the wall and remained there throughout the experiment. Performance was recorded and analyzed by the EthoVision software (Noldus, Wageningen, Netherlands).At the completion of the experiment, animals were deeply anaesthetized with a mixture of ketamine and xylazine (50 mg/kg ketamine, 10 mg/kg xylazine). After collection of CSF from the cisterna magna, animals for gene expression analysis were transcardially perfused with ice-cold PBS containing 10% RNAlater (Invitrogen, Carlsbad, CA, USA). Brains were quickly removed from the skull and placed into anice-cold RBM-4000C rat brain matrix (ASI Instruments, Warren, MI, USA). A 4 mm thick coronal slice was dissected with two blades positioned at bregma -2.2 and -6.2. Hippocampal formations were dissected from this slice.Samples were collected in Eppendorf tubes containing 1 ml of ice-cold RNAlater, incubated on ice for 1 hour, frozen on dry ice and stored at -80oC. Total RNA was isolated using the RNeasy Lipid Tissue Mini Kit (Qiagen, Hilden, Germany). RNA analytics included capillary electrophoresis using Nano RNA Chips on 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). cDNA synthesis was performed using the High Capacity cDNA Reverse Transcription Kit according to manufacturer’s instructions (Applied Biosystems, Foster City, CA, USA).

TaqMan inventoried gene expression assays (Applied Biosystems) were used to study transcriptional regulation of 75 selected genes by real-time PCR. Glyceraldehyde-3-phosphate dehydrogenase (Gapdh), hypoxanthine phosphoribosyl-transferase (Hprt1) and peptidyl- prolyl isomerase A (Ppia) were used as house-keeping genes. Each assay consisted of a FAM dye-labeled TaqMan MGB probe and two PCR primers. Real-time PCR was performed on a ViiA7 Real-Time PCR System (Applied Biosystems) as described earlier (Sarvari, Kallo, Hrabovszky, Solymosi, & Liposits, 2014). The ViiA7 RUO software and relative quantification against calibrator samples (ΔΔCt) were used for data analysis. A computed internal control corresponding to the geometric mean of Ct values of selected house-keeping genes was used for ΔCt calculation. Relative quantity (RQ = 2-ΔΔCt) was used to characterize gene expression in the various experimental groups. PCR experiments conformed to minimum information for publication of quantitative real-time PCR experiments (MIQE) guidelines (Bustin et al., 2009).

After behavioural tests, animals were deeply anaesthetized with a mixture of ketamine and xylazine, and CSF was collected from the cisterna magna. Three rats per group were perfused transcardially with PBS followed by 120 ml of 4% paraformaldehyde (PFA) in 0.1 M PBS. Brains were quickly removed from the skull and post-fixed overnight in 4% PFA. Brains were cryoprotected overnight in 30% sucrose and frozen on powdered dry ice. Coronal, 30 µm thick sections were prepared using a freezing microtome (SMR2000R, Leica Biosystems, Nussloch, Germany). Sections were collected into a series of six wells, thus each well contained consecutive sections that were 180 μm apart rostro-caudally from one another. For Iba1 immunohistochemistry (IHC), after pretreatment with a solution containing 0.5% Triton X-100 and 0.5% H2O2 in PBS for 20 minutes, sections were immersed in a rabbit polyclonal antiserum against Iba1 (1:2,000 dilution, #019-19741, Wako Chemicals) for 72 hours at 4ºC, followed by 1-hour incubations in biotin conjugated anti-rabbit IgG (1:500 dilution, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and ABC elite reagent (1:1,000 dilution, Vector Laboratories, Burlingame, CA, USA). The immunoreaction product was visualized with nickel-enhanced diaminobenzidine (0.15% NiNH4SO4, 0.005% H2O2, 0.05% diaminobenzidine in Tris pH7.6). Sections were mounted on glass slides and coverslipped with DPX mounting medium (VWR, Radnor, PA, USA). Slides were analyzed using a Zeiss Axioskop-2 (Carl-Zeiss Microscopy, Jena, Germany) microscope. Sections were photographed with an AxioCam HRc digital camera controlled by AxioVision 4.6 software (Carl-Zeiss). Digital pictures were taken using 20x and 40x objectives.

For immunofluorescent detection of Aβ1-42,sections were pretreated with a solution containing 1% human serum albumin, 0.1% Triton X-100 and 1mg/ml digitonin for 1 hour and incubated with mouse monoclonal antibody against Aβ1-42 (#126649, Abcam) at the dilution of 1:200 for 72 hours at 4oC. Cy3-conjugated anti-mouse IgG (1:500 dilution, Jackson ImmunoResearch Laboratories) was diluted in PBS containing 2% normal horse serum and applied on the sections for 2 hours. Similarly prepared sections from the cerebral cortex of APPNL-F/NL-F mice (Nilsson et al., 2014) were used as positive controls. Sections were mounted on glass slides and coverslipped with VectaShield mounting medium (Vector Laboratories). Slides were investigated using a Zeiss Axioskop-2 microscope under epifluorescent illumination using a filter set for Cy3 dye. Specificity of the Aβ1-42 antibody was confirmed by the lack of staining on wild-type mouse tissue, and the characteristic staining in APPNL-F/NL-F mice with Aβ plaques.The measurement of rat IFNγ, IL-1α, IL-10 and TNFα was performed by cytometric bead array (CBA) as earlier (Denes et al., 2015), using BDTM CBA Flex Sets (BD Biosciences, Franklin Lakes, NJ, USA) according to manufacturer’s protocol. Samples were acquired on a BD FACSVerseTM flow cytometer and data analyzed using FCAP Array v3 software (BDBioscience).In PCR data evaluation, group data were expressed as relative quantity: RQ(mean)±standard deviation. Statistical significance of the changes in gene expression was analyzed using one- way ANOVA followed by Newman-Keuls post-hoc test (Statistica 13.1, Dell Software, Round Rock, TX, USA). In the assessment of MWM data, we used one-way ANOVA with Newman-Keuls post-hoc test to determine probability of the changes. Differences were considered significant at p<0.05.

AβO infusion evoked lasting microglial activation and inflammation in the hippocampus We characterized changes of the microglia phenotype in the hippocampus after intracerebro- ventricular (icv.) infusion AβO and artificial CSF by measuring the expression of Iba1, Cd11b, Cd68, Cd80, Cd86 and RT1-EC2 (Fig. 1A-F). All selected microglia reactivity markers were upregulated in the AβO challenged group compared to controls. IHC and quantitative evaluation of Iba1 immunoreactivity were performed to verify the presence of activated microglia (Fig. 1G). In accordance with elevated expression of reactivity marker genes, Iba1 integrated density was significantly higher in the CA1 region of animals challenged with AβO compared to controls. To characterize further the microglia phenotype, we investigated mRNA expression of a set of microglia related genes encoding receptors for recognition, recruitment and phagocytosis (Table 1). Expression of all selected recognition (Tlr2, Tlr3, Tlr4, Tlr9, co-receptor Cd14), recruitment (C3ar1, C5ar1, Ccr2, Cxcr4), phagocytic and scavenger receptor genes (Cd11b, Cd11c, Cd163, Cd18, Cd68, Crry, Fcgr1a, Fcgr2a, Fcgr2b, Fcgr3a, Mfge8, Mrc1) was enhanced in the AβO challenged group compared to controls.MCC950 treatment after AβO infusion reversed mRNA expression of all selected reactivity marker genes (Fig. 1A-F). MCC950 also decreased Iba1 integrated density (Fig. 1G) supporting the results of gene expression analysis described above. MCC950 fully reversed upregulation of recognition, recruitment and phagocytosis receptor genes (Table 1).

We studied expression of selected chemokine, complement and cytokine genes, some of them related to the status of the transcription factor NF-κB (Table 2). Infusion of AβO and CSF robustly enhanced chemokine (Ccl2, Cxcl10) and complement (C3, Cfb) expression in the hippocampus. AβO also increased Nlrp3 and Il1b expression indicating activation of NF-κB. Upregulation of additional proinflammatory cytokines (Tnf, Il12b) and Nos2 supported the notion that microglia remained activated after icv. infusion of AβO and CSF. On the other hand, icv. infusion of MCC950 immediately after AβO infusion reduced expression of all pro- inflammatory genes to control levels (Table 2) indicating that the NLRP3 inhibitor abolished NF-κB-dependent transcriptional activation.We couldn’t detect extracellular Aβ1-42 plaques in the hippocampus after AβO and CSF infusion (Fig. 2), although we detected Aβ1-42 immunoreactivity in the cerebral cortex of APPNL-F/NL-F mice (Nilsson et al., 2014). This finding indicated that microglia may have contributed to the elimination of AβO or prevented its accumulation in the brain, but remained activated after AβO challenge.

Neurons control microglia reactivity by neuronal inhibitory ligands via their cognate microglial receptors (Hanisch & Kettenmann, 2007; Tian, Rauvala, & Gahmberg, 2009). After AβO and CSF infusion, expression of Cx3cl1 (Fig. 3A), Cd200 (Fig. 3B), and Cd22 (Fig. 3C) increased moderately, while their cognate microglial receptors, Cx3cr1 (Fig. 3D), Cd200r1 (Fig. 3E), and Cd45 (Fig. 3E) enhanced robustly compared to controls. Expression of other neuronal inhibitory ligands such as Cd47 (RQ=1.67±0.33, p=0.01) and Sema3a (RQ=1.63±0.15, p=0.001) elevated moderately. Transcription of Cd86, which is a key player in the modulation of the microglia phenotype (Louveau et al., 2015), also enhanced significantly (Fig. 1E). MCC950 treatment restored expression of both neuronal ligands and their microglial receptors (Fig. 3).Estradiol is synthesized from testosterone by aromatase in the hippocampus. Its effects are mediated by estrogen receptors, mainly via estrogen receptor alpha (ERα) and beta (ERβ), which are encoded by Esr1 and Esr2, respectively. AβO challenge inversely regulated the expression of estrogen receptors, decreased Esr1 (Fig. 4A) and increased Esr2 (Fig. 4B) expression. Inhibition of NLRP3 inflammasome activation restored Esr1 mRNA expression and attenuated upregulation of Esr2.Interneuron-specific and parvalbumin-positive basket cell predominant voltage-gated sodium channel subunit Nav1.1 shows decreased expression in AD (Verret et al., 2012). As the process takes place at early phase of the disease, we followed mRNA expression Scn1a encoding the Na(v)1.1 subunit in the hippocampus after icv. infusion AβO and CSF. Scn1a expression was decreased (RQ=0.79±0.02, p=0.032),but MCC950 restored Scn1a (RQ=0.95±0.06) expression, although the change did not reach statistical significance
(p=0.09).Hippocampus-dependent spatial memory was tested in MWM tasks. During spatial acquisition, decreasing escape latency times indicated fast spatial learning in the case of the parameters, therefore only latency times of the vehicle control group were shown. The AβO challenged group showed higher latency times than controls from day 2, and the difference became statistically significant on day 4 (Fig. 5A).

On day 5, the hidden platform was moved to the opposite quadrant (reversal) and spatial acquisition was followed for three days (Fig. 5B). Control animals learned the new position of the platform fast, while the AβO challenged group showed impaired learning. The difference between latency times reached statistical significance on day 3 (Fig. 5B).In the MCC950 treatment group, we measured decreased latency times compared to the AβO challenged group, similar to control groups (Fig. 5A). After reversal, the MCC950 treatment group learned the new position of the platform fast, while the AβO challenged group showed
impaired learning (Fig. 5B). The performance of control groups and the MCC950 treatment group was similar (and was not different statistically) indicating that MCC950 completely reversed AβO-evoked memory impairments.Blockade of the NLRP3 inflammasome reversed AβO-induced decreases in serum IL-10 Activation of microglia and the NLRP3 inflammasome may influence immune-related cells in the circulation.Therefore,we investigated serum cytokine profiles by measuring concentrations of IL-1α, TNFα, IL-10 and IFNγ. TNFα and IFNγ levels were below the detection limit of the CBA assay. In contrast to the low serum levels of proinflammatory cytokines, concentrations of IL-10 (Fig. 6A) were 105±1.7pg/ml in control animals, which decreased to 83±14.0pg/ml in AβO challenged animals, but the change did not reach statistical significance (p=0.064). MCC950 treatment normalized serum IL-10. Serum levels of IL-1α (Fig. 6B) showed no difference among treatment groups.

The microglia driven innate immune response against AβO in the hippocampus is implicated in the pathogenesis of late-onset AD (Akiyama et al., 2000; Heneka et al., 2015; Lucin & Wyss-Coray, 2009).To explore the AβO-induced microglia response and putative downstream mechanisms acting in the aging hippocampus, we induced an early AD-like neurological phenotype in middle-aged male rats by AβO (and CSF) infusion into the lateral cerebral ventricle. Then, we studied spatial memory, as well as selected molecular elements of the innate immune response and interneuron-related processes. To clarify the impact of the microglia driven innate immune response to AβO on spatial memory impairment in our middle-aged infusion model, we blocked activation of the NLRP3 inflammasome by icv. infusion of MCC950.Infusion of Aβ1-40 and Aβ1-42 into the cerebral ventricles of middle-aged rats is an established approach to mimic neurological phenotype of late-onset AD (Frautschy et al., 2001; Frautschy et al., 1996). As familial AD mutations lead to increased Aβ1-42 : Aβ1-40 ratio (Hardy, 1997), and Aβ1-40 and Aβ1-42 are the predominant species in neuritic plaques of AD patients (Masters et al., 1985; Mori, Takio, Ogawara, & Selkoe, 1992), we chose Aβ1-42 to form oligomers. Of note, structure and composition of AβO are markedly different from the neurotoxic oligomers of Aβ1-40 (Ahmed et al., 2010).

This structural difference goes with distinct prefibrillar to fibrillar oligomer transitions, and while AβO are stable at low temperature and low ionic strength, Aβ1-40 oligomers slowly form protofibrils and fibrils under the same condition (Ahmed et al., 2010; Glabe, 2008). Our oligomer preparation under low temperature-low salt condition results in mainly pentamers (and decamers), but we cannot exclude the possibility that the structure and composition of AβO may change during infusion, although we found no sign of fibril formation. Using expression analysis and evaluation of Iba1 immunoreactivity, we provided evidence that 4 weeks after AβO infusion still there is microglia activation in the hippocampus (Fig. 1). Although extracellular Aβ1-42 plaques were not present (Fig. 2), we found evidence for upregulation of chemokines, complement and proinflammatory cytokines indicating the activation of NF-κB (Table 2). Robust transcriptional activation of genes encoding C3-Cd11b/Cd18 ligand-receptor pairs indicated increased microglial phagocytosis.In accordance, upregulated production and activation of complement have been demonstrated in the AD brain (Eikelenboom & Stam, 1982; Yasojima, Schwab, McGeer, & McGeer, 1999). Early activation of complement in AD has been proved recently (Hong et al., 2016; Zanjani et al., 2005). Upregulation of complement genes in the hippocampus after icv. infusion AβO(Table 2) lends further support for the role of complement in early AD. Although we observed enhanced expression of some microglia receptors and their neuronal inhibitory ligands (Fig. 3), regulatory mechanisms were unable to prevent microglia activation after the
encounter with AβO. Uncontrolled microglia activation coincided with downregulation of Esr1 (Fig. 4) and Scn1a expression, and impairment of spatial memory (Fig. 5).

Estradiol is synthesized from testosterone by aromatase in the hippocampus (Hojo et al., 2004). Its effects are mediated by estrogen receptors, mainly via ERα and ERβ, both are expressed in the rat hippocampus (Shughrue & Merchenthaler, 2000). In microglia, the expression of ERα and ERβ is still controversial. While both have been observed in vitro, only ERα has been observed in vivo (Sierra, Gottfried-Blackmore, Milner, McEwen, & Bulloch, 2008). We found downregulation of Esr1 indicating possible impairment of powerful microglia-dependent anti-inflammatory mechanisms (Vegeto et al., 2003). The regulatory role of ERα-mediated E2 action in neuroinflammation and NLRP3 inflammasome activation remains largely unexplored in AD. On the other hand, reduced ERα expression may also have an impact on neuromodulation (B. McEwen et al., 2001), as ERα is localized in dendrites, dendritic spines, and presynaptic terminals (B. S. McEwen, Akama, Spencer-Segal, Milner, & Waters, 2012). ERα mediates translational regulation of synaptogenesis via activating protein synthesis for dendritic function (Akama & McEwen, 2003). In addition, ERα, associated with clusters of vesicles in perisomatic inhibitory boutons, regulates vesicle trafficking at inhibitory synapses (Hart, Snyder, Smejkalova, & Woolley, 2007). These effects make E2 a powerful regulator of the hippocampus. Cessation of ovarian E2 synthesis renders the hippocampus vulnerable to AβO and may contribute to the increased risk of AD in postmenopausal women (Henderson & Buckwalter, 1994). Decreased mRNA expression of ERα isoforms (Ishunina, Fischer, & Swaab, 2007) and reduced number of both cytoplasmic and nuclear ERα positive neurons (Hu et al., 2003) have been reported in the AD hippocampus.Scn1a, which encodes the voltage-dependent sodium channel alpha subunit Na(v)1.1 expressed predominantly in parvalbumin-positive basket cells, is downregulated in AD brains at both mRNA and protein levels (Verret et al., 2012). The

Na(v)1.1 expression, hypersynchrony and memory impairment has been established in AD (Verret et al., 2012). AβO-induced downregulation of Esr1 and Scn1a may interfere with the function of inhibitory interneurons, which represent a link between NLRP3 inflammasome activation and memory decline. AβO-mediated memory decline became significant on the 4th day of spatial acquisition in MWM (Fig. 5), in accord with other studies using the middle- aged rat infusion model (Frautschy et al., 2001). We found that inhibition of the NLRP3 inflammasome by icv. infusion of MCC950 after AβO challenge attenuated AβO-evoked microglia reactivity (Fig. 1) and NF-κB activation, and abolished spatial memory impairment (Fig. 5). These observations agree with the results of recent pharmacological studies in APP transgenic mouse models (Daniels et al., 2016; Dempsey et al., 2016). An important novel observation that adds to these earlier studies is that AβO is sufficient to induce long-lasting inflammation and memory deficits in an NLRP3- dependent manner, even in the absence of major amyloid deposits in the brain parenchyma. In addition, we found that MCC950 restored transcriptomic changes of genes involved in neuron-microglia communication, and reversed downregulation of Esr1 and Scn1a, although in the latter case the effect did not reach statistical significance. As MCC950 infusion leads to the resolution of inflammation and restoration of spatial memory, it is likely that AβO-evoked inflammatory responses play causative role in the suppression of estrogen signaling, alteration of basket cell specific sodium transport and impairment of spatial memory. Based on these findings, we propose that AβO activates microglia, which persists even after the elimination of the pathogenic substance. Enduring activation of microglia interferes with ERα signaling and subunit composition of voltage-gated sodium channels in basket cells among others, resulting in reversible impairment of spatial learning and memory.

Central inflammatory reactions may have an impact Asciminib in vitro on the peripheral immune system, and viceversa. Cognitively impaired patients with amyloidosis have shown higher serum levels of proinflammatory cytokines and lower levels of IL-10 compared to patients without brain amyloidosis (Cattaneo et al., 2017). Therefore, activation of microglia in the hippocampus may influence immune-related cells in the circulation and alter the serum cytokine profiles. Although statistical significance was not achieved in our study, serum levels of IL-10 tended to decrease in middle-aged rats following AβO challenge (Fig. 6). Importantly, serum IL-10 levels are affected in individuals with AD (Asselineau et al., 2015; Remarque et al., 2001; Swardfager et al., 2010). In peripheral blood mononuclear cells, IL-10 synthesis is almost ceased in fast cognitive decliners, whereas its level is slightly increased in slow cognitive decliners with AD (Asselineau et al., 2015). In our middle-aged rat infusion model, MCC950 restored normal serum IL-10 levels (Fig. 6).This finding gives support to the idea that AβO- induced microglia activation results in a decrease in serum IL-10 levels. Emerging evidence indicates that IL-10 acts in a negative feedback loop to regulate the NLRP3 inflammasome during chronic stimulations (Gurung et al., 2015; Ip, Hoshi, Shouval, Snapper, & Medzhitov,
2017; Yao et al., 2015). A very recent paper has demonstrated that IL-10 promotes mitophagy that eliminates dysfunctional mitochondria in macrophages (Ip et al., 2017). However, the functional contribution of IL-10 to the transcriptomic, inflammatory and behavioral changes observed has not been tested in this study.

In conclusion, these results demonstrate that icv. infusion of AβO evokes lasting microglial and inflammatory responses in the hippocampus of middle-aged rats leading to reversible impairment of spatial memory. The underlying mechanisms might include interference with Esr1 and Scn1a expression in the hippocampus, and peripheral IL-10-mediated action could potentially contribute to the changes observed, which will need to be tested in further studies.