CRISPR/Cas9-mediated CysLT1R deletion reverses synaptic failure, amyloidosis and cognitive impairment in APP/PS1 mice

As a major pathological hallmark of Alzheimer’s disease (AD), amyloid-β (Aβ) is regarded as a causative factor for cognitive impairment. Extensive studies have found Aβ induces a series of pathophysiological responses, finally leading to memory loss in AD. Our previous results demonstrated that cysteinyl leukotrienes receptor 1 (CysLT1R) antagonists improved exogenous Aβ-induced memory impairment. But the role of CysLT1R in AD and its underlying mechanisms still remain elusive. In this study, we investigated CysLT1R levels in AD patients and APP/PS1 mice. We also generated APP/PS1-CysLT1R-/- mice by clustered regulatory interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9)-mediated CysLT1R deletion in APP/PS1 mice and studied the effect of CysLT1R knockout on amyloidogenesis, synapse structure and plasticity, cognition, neuroinflammation, and kynurenine pathway. These attributes were also studied after lentivirus-mediated knockdown of CysLT1R gene in APP/PS1 mice. We found that CysLT1R knockout or knockdown could conserve synaptic structure and plasticity, and improve cognition in APP/PS1 mice. These effects were associated with concurrent decreases in amyloid processing, reduced neuroinflammation and suppression of the kynurenine pathway. Our study demonstrates that CysLT1R deficiency can mediate several beneficial effects against AD pathogenesis, and genetic/pharmacological ablation of this protein could be a potential therapeutic option for AD.


INTRODUCTION
Amyloid-β (Aβ) is the main pathological hallmark of Alzheimer's disease (AD), and neurotoxic effects induced by Aβ have been well documented to be a causative factor of the disease. AD is also characterized by neurofibrillary tangles made up of hyperphosphorylated tau, and elevated neuroinflammation in the brain, which induce impairment of hippocampal synaptic plasticity and may lead to neuronal apoptosis, finally leading to cognitive decline in the affected persons [1]. Neuroinflammation has become an increasingly stressed topic in AD pathology which is characterized by the activation of astrocytes and microglia that trigger a vicious cycle of inflammatory responses through the increased release of pro-inflammatory cytokines [2]. These inflammatory cytokines have been reported to augmenting Aβ formation and resulting in neuronal damage [3]. Importantly, treatments blocking the inflammatory responses can reverse Aβ-induced neuroinflammation and memory impairment [4]. Therefore, therapeutic targeting of neuroinflammation in AD is becoming the focus for future studies.
Cysteinyl leukotrienes (Cys-LTs) are a group of the inflammatory lipid molecules that include LTC4, LTD4, and LTE4, and are generated by the activation of 5lipoxygenase (5-LOX) pathway of the arachidonic acid metabolism [5]. Cys-LTs initiate inflammatory signaling AGING cascades by two major G-protein coupled receptors: CysLT1R and CysLT2R, the roles of which have traditionally been examined in relation to asthma and other allergic diseases [6]. However, in the recent years, increasing evidence has backed their possible involvement in the pathophysiology of AD, Parkinson's disease (PD), and other neurological disorders [7,8]. We previously have found the involvement of CysLT1R in Aβ-induced neurotoxicity, as the neuroinflammation and memory impairment induced by Aβ1-42 are ameliorated through administration of selective CysLT1R antagonists such as montelukast and pranlukast [9]. Moreover, we have suggested that CysLT1R is strongly related with LPS-induced neuroinflammation, apoptosis and memory impairment, and through pretreating mice with pranlukast [10]. However, the precise mechanisms by which CysLT1R mediates AD-related learning and memory impairment were unknown.
Many studies have indicated that elevated Aβ1-42 is correlated with proinflammatory cytokines inducedkynurenine pathway (KP) activation [11]. KP is the main metabolic pathway of tryptophan degradation producing a series of metabolites [12]. Quinolinic acid (QUIN), one metabolite from KP, is considered to be an endogenous N-methyl-D-aspartate receptor (NMDAR) agonist with strong excitotoxic properties, which leads to glutamatergic excitotoxicity through overactivation of NMDARs in AD [13]. In additions, QUIN is present in neurons, microglia and astrocytes around senile plaques and neurofibrillary tangles in the hippocampus of AD patients [14]. As a rate-limit enzyme involved in the main pathway towards NAD + synthesis, kynureninase (KYNU) promotes the production of QUIN in KP and is also involved in AD pathogenesis. Since CysLT1R mediates Aβ-induced neurotoxicity following by elevated proinflammatory cytokines, which in turn induces kynurenine pathway dysregulation, we hypothesized that CysLT1R could be important for KP regulation, leading to neuroinflammation and related cognitive impairment. To test this hypothesis and specifically elucidate the role of CysLT1R during AD pathogenesis, we downregulated the expression levels of CysLT1R by knockout or knockdown of the CysLT1R gene in APP/PS1 mice for the first time, and examined the effects of CysLT1R manipulation on amyloidosis, synaptic plasticity, cognition, neuroinflammation, and KP regulation.

CysLT1R expression is upregulated in APP/PS1 mice and AD patients
We first measured CysLT1R levels in APP/PS1 mice hippocampus. Both protein and mRNA levels of CysLT1R were elevated to about 2-fold and 3-fold in the hippocampus of 6-month-old (for protein, F [1,6] Figure 1A-1D).
We next confirmed the elevated CysLT1R levels in brain sections from 10-month old APP/PS1 mice by immunofluorescence staining. APP/PS1 mice exhibited significant upregulation of CysLT1R in the dentate gyrus (DG; F [1,10]  We further confirmed elevated protein levels of CysLT1R in AD patients by immunohistochemical analyses. A 2-fold elevation in CysLT1R levels (F [1,14] = 4.491, P<0.01, Figure 1J, 1K) in the prefrontal cortex of AD patients was observed compared to agematched controls. The detail information of AD patients and controls has been shown in the Supplementary Table 1. Altogether, CysLT1R levels are upregulated in both APP/PS1 mice and AD patients compared to controls. Moreover, CysLT1R protein and mRNA levels in APP/PS1 mice are increased with age.

CysLT1R knockout ameliorates cognitive decline in APP/PS1 mice
To evaluate the effects of CysLT1R deficiency on spatial learning and memory, mouse behavior was tested by the MWM task. During day 1-2 (visible platform training), all groups exhibited the similar latencies, suggesting similar visual and motor functions among all the groups (effect of day, F [3,284] Figure 4B). In the probe trial, CysLT1R deletion significantly increases the percentage of time stayed in the target quadrant (F [3, 28] = 4.405, P < 0.05, Figure 4C-4E) and the numbers of platform crossings in APP/PS1 mice (F [3, 28] = 6.55, P < 0.05, Figure 4D, 4E). Y-maze task evaluation also showed that impairment of working memory was alleviated (for the number of correct choices, F [3, 28] = 5.751, P < 0.05, Figure 4F; and for the escape latency, F [3, 28] = 5.053, P < 0.05, Figure 4G) by CysLT1R deficiency. Moreover, this was further confirmed by novel object recognition test (F [3, 28] = 6.045, P < 0.05, Figure 4H). Importantly, all groups traveled similar total distance in the open field test (F [3, 28] = 0.119, P > 0.05, Figure 4I), suggesting similar exploration capability and motor functioning.

CysLT1R knockdown mediates similar effects on AD pathologies
To know whether CysLT1R knockdown could also yield similar effects seen in the knockout studies, we reduced CysLT1R levels in the DG of APP/PS1 mice hippocampus by bilateral injection of LV-CysLT1R shRNA-EGFP (Supplementary Figure 2A Figure 5E-5H) were decreased in the hippocampus of CysLT1R-shRNAtreated APP/PS1 mice. Moreover, hippocampal knockdown of CysLT1R had no effect on behavioral tests, hippocampal synaptic plasticity, and amyloidogenesis in WT mice.

DISCUSSION
In our current study, elevated levels CysLT1R are firstly confirmed in the brains of AD patients and APP/PS1 mice. Increased levels of CysLT1R protein and mRNA were found in both hippocampus and prefrontal cortices, which were found to be age-dependent and region-specific with disease progression in APP/PS1 mice. Consistent with previous studies demonstrating age-dependent increases in glial and neuronal expressions of CysLT1R [17], our data also revealed APP/PS1 mice exhibited higher CysLT1R localization in microglia, astrocytes, and neurons compared with WT mice. How CysLT1R mediates AD-related deficits in APP/PS1 mice still remains elusive, though it is possible that accumulating Aβ induces the upregulation of 5-LOX [18], which in turn produces Cys-LTs, inducing CysLT1R expression.
Functional synapse formation is critical for the synaptic transmission in the brain, and synaptic plasticity is the important neurobiological foundation for learning and memory. Marked synapse loss was found in the brain, especially in hippocampus and cerebral cortex, of AD patients in previous findings. APP/PS1 mice at 3 months of age also showed deficits in synaptic plasticity [19]. In our study, we found impaired LTP, declined dendritic spine density, synapse number, and pre-and postsynaptic marker proteins in APP/PS1 mice. NMDAR, an ionotropic glutamate receptor in the hippocampus, is essential for cognition and dysfunction of NMDAR is closely correlated with excitotoxicity [15,20]. In additions, NMDARs functions are influenced by interaction with PSD-95, an important postsynaptic scaffolding protein that is involved in protein assembly, synaptic development and neural plasticity [21]. The functions of NMDARs and PSD-95 play the critical role in LTP induction at CA3-CA1 synapses [22]. As such, NMDARs dysfunction and PSD-95 reduction in APP/PS1 mice may explain the corresponding LTP reduction we observed in the current study. We speculate that increased NR2B occurs in extrasynaptic locations in APP/PS1 mice, and overactivation of extrasynaptic NMDARs lead to a variety of aberrant transcriptional cascades and posttranslational modifications associated with neurotoxicity by stimulating cell death pathways [23]. Notably, we demonstrated that CysLT1R gene knockout or knockdown effectively restored LTP and reversed elevated NR2B levels as well as reduced PSD-95 expression in APP/PS1 mice hippocampus. Interestingly, despite the improvement in spine density, synapse number, and the levels of NMDARs due to CysLT1R deletion in APP/PS1 mice, CysLT1R -/mice did not exhibit any alterations compared to WT mice ( Figure  2C-2L), suggesting CysLT1R may mediate neurotoxicity only under pathological conditions. AGING Aβ deposits develop in APP/PS1 mice from 6 months and it progressively increase with age. Abundant plaques are observed in both hippocampus and cortex of 9-month-old APP/PS1 mice and continue to increase to around 12 months of age [24]. In the present study, Aβ pathology was ameliorated through CysLT1R deficiency without altering total APP levels in APP/PS1 mice. Thus, we investigated whether APP processing and Aβ degradation were changed. Our data indicated that CysLT1R gene knockout or knockdown didn't affect the level of CTFα produced by α-secretase cleavage in APP/PS1 mice, whereas CTFβ produced by BACE1 cleavage was decreased, which could be explained by a reduction in BACE1 levels. Additionally, PS1, the essential part of the γ-secretase complex, was reduced after CysLT1R gene knockout or knockdown, which might also explain the decrease in Aβ levels. Furthermore, the proteolytic enzymes, IDE and NEP, that degrade Aβ were not altered significantly by CysLT1R gene knockout or knockdown [25]. Moreover, we observed that APP/PS1-CysLT1R -/mice showed inhibited activation of astrocytes, microglia and decreased release of IL-6 and TNF-α. It is possible that the levels of BACE1 and PS1 are increased in reactive astrocytes of AD brain. Our previous studies have suggested that CysLT1R promotes the release of proinflammatory cytokines though NF-κB signaling pathway, and proinflammatory cytokines upregulate BACE1 activity and Aβ production [26]. Therefore, CysLT1R deletion ameliorates amyloidogenesis in APP/PS1 mice most likely by decreasing the proinflammatory cytokines expression and inhibiting astrocytes activation via NF-κB signaling. Moreover, recent study has indicated that Aβ clearance is promoted by converting microglia from the M1 state to the M2 state [27]. The activation states of microglia could be M1 phenotype representing pro-inflammatory activity or M2 microglia exhibiting an anti-inflammatory phenotype. Therefore, the markers for M1 phenotype could be pro-inflammatory cytokines, such as TNFα and IL-1β, and the markers for M2 phenotype could be transforming growth factor-β (TGF-β) against inflammation [28]. Importantly, the marker of M1 microglia phenotype, TNFα, was elevated in APP/PS1 mice but decreased by CysLT1R deletion, implicating a possible shift of microglia phenotype from M1 to M2. Consequently, CysLT1R deficiency could promote Aβ clearance by microglial phagocytosis.
Besides NF-κB signaling pathway and following neuroinflammation and apoptosis mediated by CysLT1R, our data revealed that KP induced by proinflammatory cytokines is also involved in the CysLT1R-mediated AD pathogenesis. Substantial evidence has accumulated in the past decades depicting that dysregulation of KP and the production of neurotoxic metabolites are major contributors to the pathogenesis of several neurodegenerative diseases such as AD [29]. The initial rate-limiting step of the KP from tryptophan is determined by Indoleamine 2,3-dioxygenase (IDO) activity. The proinflammatory cytokines, such as IFNγ, TNF-α and IL-1β, synergistically stimulate IDO expression and activity [30][31][32]. Beyond this, RelB-p52 could directly bind to the IDO promoter region and subsequently induce IDO transcription in noncanonical NF-kB pathway [33]. Aβ induces IDO expression and increased levels of IDO have been observed in hippocampal tissue from AD patients. Consistent with these studies, our data showed IDO level was elevated in APP/PS1 mice at 10 months age. As a key enzyme lying on the main pathway towards NAD + synthesis, KYNU oxidizes both L-KYN and 3-OH-L-KYN to anthranilic acid and 3-OH-ANA, respectively [34], followed by the synthesis of QUIN. QUIN, a specific competitive agonist for NMDAR [35], is identified as a potential neurotoxin causing glutamatergic excitotoxicity [36]. Thus, upregulation of KYUN would favor the production of increased amount of toxic QUIN. As expected, we found elevated KYNU levels and activities in APP/PS1 mice, and such increases were reversed through CysLT1R knockout. Importantly, QUIN impairs postsynaptic elements, promotes neurodegenerative lesions, as well as induces neuroinflammation and apoptosis, which are involved in the complex and multifactorial cascade leading to neurodegeneration [37]. Previous studies revealed that CysLT1R antagonist montelukast significantly attenuated striatal lesion, oxidative stress and mitochondrial dysfunction in rats caused by intrastriatal administration of QUIN [38]. Therefore, we have suggested QUIN is likely a major contributor to the impairment of synaptic plasticity and cognitive deficits in AD and blockade of CysLT1R might have neuroprotective potential. Our data indicated significant increases in QUIN content and more intensive colocalization with NR2A or NR2B in APP/PS1 mice, and the content and colocalization were attenuated by CysLT1R deletion. These data may explain why CysLT1R deficiency enhances hippocampal synaptic plasticity in APP/PS1 mice.
Overall, we propose that Aβ accumulation upregulates Cys-LTs inducing CysLT1R expression, which activates NF-κB pathway followed by increased release of proinflammatory cytokines. Consequently, proinflammatory cytokines induce neuronal apoptosis and KP dysregulation with increased IDO and KYUN and the synthesis of QUIN. Moreover, neuroinflammation accelerates amyloid deposition, forming a vicious circle. These lead to NMDARs overactivation and excitotoxicity correlated with synaptic dysfunction and cognitive deficits (Figure 7). However, CysLT1R gene knockout ameliorates amyloidogenesis and cognitive impairment AGING through inhibiting neuroinflammation and neurotoxic effects of QUIN on NMDARs. More importantly, CysLT1R gene knockdown using LV-shRNA in APP/PS1 mice confirms the results obtained from APP/PS1-CysLT1R -/mice. LV-CysLT1R shRNA generates stable knockdown in the hippocampus, which is regional specificity and selectivity. CysLT1R knockdown mediated by LV-CysLT1R shRNA also overcomes the possible compensation and avoids gene deletion-caused adverse effect on physiological function and lifetime. These experiments have been performed in a combination of CysLT1R knockdown and knockout, and the data suggest that downregulation of CysLT1R might be an effective therapeutic strategy for AD. The present study expends our understanding about the mechanism underlying CysLT1R-mediated AD pathology and suggests CysLT1R as a potential target for AD therapy.

Materials
A list of the antibodies used in this experiment is listed in the supplementary materials. Pyfidoxal-5'-phosphate, 3-hydroxyanthranilic acid and 3-hydroxykynurenine were purchased from Sigma-Aldrich Co. LLC.

Animals
Male APP/PS1ΔE9 mice (herein referred to as APP/PS1) and their wild-type (WT) littermates, CysLT1R -/mice and APP/PS1-CysLT1R -/mice were constructed and bred by Model Animal Research Center of Nanjing University (Nanjing, China). The constitutive CysLT1R knockout (CysLT1R -/-) mice were generated with the aid of CRISPR/Cas9mediated gene editing in mice of C57BL/6 background. CysLT1R -/mice developed normally and were fertile while they showed abnormal vascular permeability. CysLT1R -/mice were further bred with APP/PS1 mice to generate APP/PS1-CysLT1R -/mice. RT-PCR was used for genotyping and the following primers were used: CysLT1R forward 5'-GAATGGA ACTGAAAATCTGACGAC-3' and reverse 5'-ATAA TAGACCACACGGAGAGGCA-3', APP forward  expression, which activates NF-κB pathway followed by increased release of proinflammatory cytokines. Consequently, proinflammatory cytokines induce neuronal apoptosis and KP dysregulation with increased expression of IDO and KYUN and the synthesis of QUIN. Moreover, neuroinflammation accelerates amyloid deposition, forming a vicious circle. This leads to NMDARs overactivation and excitotoxicity correlated with synaptic dysfunction and cognitive deficits.

Human AD brain samples
Post-mortem brain sections were dissected from frozen brains of AD patients (about 80-87 years old, Chinese male patients) and age-matched controls (about 78-87 years old, Chinese male patients), which were provided by South Central University for Nationalities (Wuhan, China). Informed consents were obtained from the subjects' family members. Pathological and clinical criteria were matched for the diagnosis of AD. The post-mortem interval of the brain tissue collection was between 2 and 4 h. The use of brain tissue samples was approved by the ethics committee of China Pharmaceutical University and was in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki declaration.

Lentivirus generation
Lentiviral shRNA was used to silence CysLT1R gene according to a previous study [35]. The lentiviral vector construct expressing the short hairpin RNA (shRNA) was generated which was complementary to the coding exon of the mice CysLT1R gene and tagged with a fused enhanced green fluorescent protein (EGFP). The construct was labelled as LV-CysLT1R shRNA. A lentiviral vector expressing EGFP alone (LV-EGFP) was used as the control. The sequence for the CysLT1R miRNA (shRNA-mir hairpin structure) was 5′-TGCATCAAATTGTTGCTTTTTCAAGAGAAAAGC AACAATTTGATGCAttttttc-3′ (the antisense target sequence in bold, while the sense target sequence in italics). The normal control (NC) sequence from Genechem was 5′-TTCTCCGAACGTGTCACGTTTC AAGAGAACGTGACACGTTCGGAGAAtttttg-3′. The underlined 9 nt in the middle represents the hairpin loop in both the CysLT1R shRNA and NC sequences. PCR was used to amplify the coding sequence of the CysLT1R shRNA. The primer sequences were: 5′-GCCCCGGTTAATTTGCATAT-3′ (forward) and 5′-GAGGCCAGATCTTGGGTG-3′ (reverse). All of the lentiviral vectors contained EGFP as a reporter to track lentivirus-mediated expression.

Stereotaxic injection of lentivirus in mouse brain
Both WT and APP/PS1 mice at 9 months of age were anaesthetized with 350 mg/kg chloral hydrate. Bilateral hippocampal dentate gyrus (DG) injection of LV-CysLT1R shRNA-EGFP was performed stereotactically at coordinates 2.1 mm to posterior, 1.7 mm to lateral, and 2.1 mm to ventral relative to brain bregma of APP/PS1 mice. LV-EGFP served as the control vector. The viral suspension at a volume of 2.5 μl (2.5×10 6 vector genome) was injected into each injection site using a 10 μl glass syringe at a rate of 250 nl/min. The needle was left untouched for 10 min and then removed slowly over 2 minutes to avoid backflow of the suspension. The mice were kept warm until the surgical recovery. After 4 weeks, the mice underwent behavioral testing and biochemical examinations.

Western blot
Western blot was performed as described previously [10]. RIPA buffer was used to homogenize the isolated hippocampi and its formulation details have been listed in Supplementary Table 2. The total protein concentration was determined by a BCA assay kit from Beyotime (China). The details of antibodies have been listed in Supplementary Table 3.

RT-PCR assay
Total RNA was extracted according to the manufacturer's protocol. The mouse hippocampi were homogenized using Trizol reagents. Formulation details of reaction system has been listed in Supplementary Table 4. This reaction mixture was then incubated for 5 min at 25° C, for 60 min 42° C, and then at for 10 min 72° C to deactivate the reverse transcriptase. PCR mixture was prepared with cDNA template (10 μl) dissolved into 40 μl reaction mixture containing 10 × PCR buffer (4.0 μl), 25 mM MgCl2 (3.0 μl), primer (12.5 pmol), and Taq DNA polymerase (0.25 μl). The cycling parameters (Supplementary Tables 5, 6) and primer sequences (Supplementary Table 7) have been listed in supplementary materials. The amplified products were separated on an ethidium bromide-containing agarose gel (1.5 %) by electrophoresis and then photographed. The bands were checked by an image analysis system from Tanon Science and Technology Co. Ltd. and the PCR was performed on an Eppendorf Master Cycler from Eppendorf (Germany).

Immunohistochemistry and immunofluorescence
Following transcardial perfusion mice brains were fixed with 4 % paraformaldehyde at 4° C for 2 days, and then equilibrated in 30 % sucrose at 4° C for 1 day and 30 μm sections were made. For immunohistochemistry, the sections were washed with PBS (0.1 M, 3 × 5 min) and then treated with H2O2 (3 %, 10 min). They were washed again in PBS (0.1 M, 3 × 5 min) and treated with Triton X-100 (0.3 %, 30 min). After washing again with PBS (0.1 M, 3 × 5 min), the sections were blocked with the solution containing 0.3 % Triton X-100 and 5 % BSA for 1 h. Then the sections were incubated with primary antibodies (details of the antibodies listed in Supplementary Table 8) overnight at 4° C. Next morning, sections were incubated with biotinylated anti-rabbit lgG or anti-mouse lgG at 37° C for 20 min, and again after AGING washing with PBS, incubated with streptavidin-biotin complex at 37° C for 20 min. Diaminobenzidine was used to detect target proteins. Finally, the sections were gradient dehydrated and photographed and images were quantified using Image J and Image-Pro Plus software. The integrated optical density of CysLT1R expression or Aβ deposition (% area occupied) was calculated based on quantification. The mean values of 3 sections from each animal were analyzed.
For immunofluorescence, after an overnight blocking in 2 % serum and 0.3 % triton, sections were incubated for 48 h or 72 h in primary antibody (details of the antibodies listed in Supplementary Table 8). The primary antibodies were localized with corresponding affinity-purified IgG. The sections were examined using a fluorescence microscope from Leica. Three sections were randomly analyzed to calculate the % colocalization.

ELISA
Aβ detection by ELISA was performed as described previously [39]. Fractions of the hippocampus (100-120 mg) were homogenized in 5 × mass of ice-cold lysis buffer containing Tris-HCl (10 mM), EDTA (5 mM) and proteinase inhibitor cocktail in 320 mM sucrose (pH 7.4). Then the homogenate was lysed on ice for 15 min and then centrifugated at 4° C (14,000 g, 15 min), leaving the supernatants containing triton-soluble Aβ peptides. The pellets which remained insoluble were re-homogenized using 10 volumes of 5 M guanidine-HCl (diluted in 50 mM tris-HCl, pH 8.0) and shaken for 4 h at room temperature. Homogenates were centrifuged (8,000 g, 5 min). The triton-soluble and guanidine-HCl-soluble fractions were used for final detection of Aβ40 or Aβ42 according to the manufacturer's manual. For IL-6, and TNF-α detection, 100 mg hippocampal tissue was rinsed with PBS and homogenized in 1 ml PBS followed by overnight storage at -20° C. Cell membranes of the homogenate were broken through two freeze-thaw cycles and the homogenates were centrifuged at 2-8° C (5,000 g, 5 min). The supernatants were collected and assayed immediately. The ELISA kits for Aβ1-40 and Aβ1-42 were from Cusabio Biotech Co. Ltd. (China), and for IL-6 and TNF-α from Neobioscience Co. Ltd. (China).

Hippocampal slice preparation and electrophysiology
Transversal hippocampal slices from WT and APP/PS1 mice at 10 months of age were prepared and placed in a recording chamber with continuous ACSF perfusion at a rate of 3 ml/min with at 24° C. Cells were visualized with an upright microscope and Schaffer collaterals was stimulated by a tungsten monopolar electrode. The field excitatory postsynaptic potentials (fEPSPs) were recorded under current-clamp mode from the CA1 stratum radiatum by a glass microelectrode filled with ACSF (3)(4). A high-frequency stimulation protocol consisting of two one-second long 100 Hz trains was used to induce long term potentiation (LTP) [40].

Golgi staining
FD Rapid Golgi Stain Kit (FD Neuro Technologies, USA) was used for Golgi staining [41]. Coronal brain blocks were made from unfixed brain samples and were immersed in a solution containing solution A and B with the volume ratio of 1 to 1 at room temperature for 2 weeks followed by soaking in solution C at 4° C for 48 h. Then the brain samples were frozen with dryice powder and serially sectioned into 100 μm coronal slices (containing the hippocampus) with a freezing microtome [42,43]. These frozen sections were mounted with solution C on a 0.5 % gelatin-coated glass slide and allowed to dry naturally at room temperature. Then they were immersed in the solutions containing solution D, solution E, and distilled water at a ratio of 1:1:2 for 5 min, followed by washing with distilled water twice for 4 min each. The sections were cover slipped after dehydrating with graded alcohol solutions and clearing with xylene. The preparations were observed under a microscope. For morphological analysis of hippocampal DG neurons, 5 granule neurons from each mouse (4 mice/group, 20 neurons from each group) were calculated from the hippocampal DG. Only neurons with non-truncated dendrites, and consistent and dark staining along the dendrites were selected for analysis; they were also selected only if there was clear isolation from neighboring neurons in order to avoiding interference with analysis [44,45]. Dendritic spine densities were estimated on secondary dendritic branches of each imaged neuron in clearly evaluable areas, approximately 150-200 μm from the soma. Only branches over 20 μm in length were included in the study. The number of spines was quantified by ImageJ. Spine density was calculated per 10 μm of dendritic length.

Electron microscopy
Synaptic density was detected by electron microscopy [42]. Following transcardial perfusion with PBS containing 2 % glutaraldehyde and 3 % paraformaldehyde (under anesthesia), fixed brains were taken out and hippocampal slices were then prepared. The slices were fixed again with cold OsO4 (1 %, 1h). Ultrathin sections (90 nm) were made and stained with uranyl acetate and lead acetate, and viewed at 100 kV on a JEOL 200CX electron microscope. Synapses were identified by the presence of synaptic vesicles and postsynaptic densities.

HPLC for hippocampal KYNU activity
Isolated hippocampi were sonicated in 9 × mass of 0.1 M PBS and centrifuged (12,000 g, 30 min) at 4° C. Protein in supernatants was determined by BCA protein assay and was stored at -80° C until use. For total 3hydroxyanthranilic acid (3-HANA) content assay, 1 volume of pyfidoxal-5'-phosphate (PLP) solution (48 μM PLP in barbitone sodium-hydrochloric acid buffer, pH 8.4) was added into 5 volumes of hippocampal lysates, and 6 volumes of 3-hydroxykynurenine (3-HK, 2 mM), a substrate of KYNU, were immediately added into the mixture and incubated for 8 min at 30° C. 12 volumes of trichloroacetic acid were added to terminate the reaction. The mixture was thoroughly mixed and centrifuged (13,200 g, 20 min) at 40° C. The levels of 3-HANA, the product of 3-HK, in the supernatants were determined by HPLC. For basal 3-HANA content assay, 12 volumes of trichloroacetic acid were added into 5 volumes of hippocampal lysates. The mixture was thoroughly mixed and centrifuged (13,200 g, 20 min) at 40° C and then the supernatants were collected and determined by HPLC. The actual content of 3-HANA was calculated by subtracting the basal 3-HANA content from the total 3-HANA content. KYNU activity (pmol/min/mg) was expressed as actual content of 3-HANA divided by reaction time and total protein concentrations in the reaction system.
Chromatographic separation was carried out on C18 column (4.6 mm × 250 mm, I.D., 5 µm). The mobile phase was 0.1 M KH2PO4 buffer containing l % acetonitrile. The column temperature was set at 30° C with a flow rate of 1.0 mL/min, and the injection volume is 50 µl. The fluorescent intensity was detected at 322 nm (excitation wavelength) and 414 nm (emission wavelength).

LC-MS/MS for QUIN content
Liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) was performed to measurement of QUIN and 2,4-Pyridinedicarboxylic acid was used as internal standard. Hippocampi were sonicated in 9 × mass of 0.2 % acetic acid in aqueous solution containing internal standard and centrifuged (12,000 g, 15 min) at 4° C. After centrifugation (15000 g, 60 min) at 4° C, the supernatants were then filtered through a 3 kDa Amico Ultra filter. The LC-MS/MS system consisted of an Acquity UPLC system (Waters Corp, Inc., CA, USA) and a Waters Xevo TQ-S triple quadrupole mass spectrometer equipped with an electrospray ionization source interface operated in the positive multiple reaction monitoring mode. Ion source settings were: capillary voltage, 3.00 kV; desolation temperature, 450° C; source temperature, 150° C; nitrogen was used as the desolation gas with an API gas flow rate of 1000 L/h. Quantification was operated in MRM of transition m/z 168→78, and MassLynx V4.1 software (Waters Corp, Inc., CA, USA) was used to data acquisition and processing. Chromatography was performed on a Zorbax Eclipse XDB-C8 column (100 mm × 4.6 mm, i.d. 3.5 μm). Gradient methods with a total duration of 6 min each were used for chromatographic separation (Mobile A: 0.5 % formic acid in aqueous; Mobile B: 1 % formic acid in acetonitrile). The flow rate for detection of both analytes was 0.3 mL/min. Retention time for the analytes was 1.17 min. Average concentrations were based on quadruplicate measures. All samples and regents were of HPLC grade.

Behavioral tests
Spatial learning and memory were assessed by Morris water maze (MWM) [46], and working memory was assessed by Y-maze [47] and novel object recognition test (NORT) [48]. Open field test (OFT) was used for assessing the general locomotor activity [49]. These behavioral tests were performed and analyzed as described previously.

Statistical analyses
Repeated measure ANOVA was used to analyze group differences in MWM escape latency with "days" as the within-subject factor and "group" as the between-subject factor. Other data were analyzed using either Student's t-test (for two-group comparison) or one-way ANOVA (for comparison among more than two groups) followed by a Dunnett's post-hoc analysis, if necessary. Mean ± standard error of the mean (SEM) was used for representation of the descriptive data. All analyses were carried out using SPSS, version 20.0. Statistical significance was considered when P-value was < 0.05.