Dysfunction of EAAT3 Enhances LPS-induced Postoperative Cognitive Dysfunction

Background: Studies have shown that excitatory amino acid transporter 3 (EAAT3) function inhibition is related to several neurodegenerative diseases. Our previous studies also found that the EAAT3 function is intimately linked to learning and memory. In this study, we examined the role of EAAT3 in postoperative cognitive dysfunction (POCD) and explored the potential benet of riluzole against POCD. Methods: We measured EAAT3 protein expression in hippocampus of male mice at different ages. Next, we established a recombinant adeno-associated viral (rAAV)-mediated shRNA to knockdown EAAT3 expression in the hippocampus of adult male mice. And then the mice received 2μg of lipopolysaccharide (LPS) intracerebroventricular microinjection to construct the POCD model. In addition, we intraperitoneally injected 4mg/kg of riluzole 2 days before LPS microinjection for consecutive 3 days in elderly male mice. Cognitive function was assessed using a Morris water maze 24h after LPS microinjection. Animal behavioral tests, as well as pathological and biochemical assays, were performed to clarify the role of EAAT3 function in POCD and evaluate the effect of activation of EAAT3 function by riluzole. Results: We found that the expression of EAAT3 was signicantly decreased in old mice and EAAT3 knockdown in hippocampus aggravated LPS-induced learning and memory decits in adult male mice. LPS signicantly inhibited hippocampal EAAT3 membrane protein expression and GluA1 protein phosphorylation level in adult male mice. Moreover, riluzole pretreatment signicantly increased hippocampal EAAT3 membrane protein expression and ameliorated LPS-induced cognitive impairment in old male mice. Conclusions: Our results demonstrated that the dysfunction of EAAT3 is an important risk factor for POCD susceptibility and riluzole may be a promising strategy for prevention and treating of POCD in the elderly people.

They were fed food and water ad libitum. All animals were acclimatized to the environment for one week before the experiment and were xed cage mates throughout the acclimation and testing periods. Each experimental group consisted of 5-12 mice, and mice in the same cage were in the same treatment group.
The experimental animal procedures were approved by the Animal Care Committee of the Chinese People's Liberation Army General Hospital (Beijing, China). All animal experiments were carried out in accordance with the current laws of China and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
Riluzole (Cayman Chemical Company, USA) was rst dissolved in dimethylsulphoxide (DMSO; Fish Scienti c, NJ) to 100 mM (27 mg/ml) and then diluted in saline to 0.4 mg/ml with gentle warming. It (4 mg/kg) was then intraperitoneally injected 2h before Morris Water Maze (MWM) acquisition training or LPS microinjection. The same concentration of DMSO (1.6%) was used as solvent control.

Experimental design and groups
The experiments were carried out between 8:00 am and 6:00 pm. All mice were sacri ced by deep sodium pentobarbital anesthesia (100 mg/kg) for obtaining the tissue. The following sets of experiments were performed. Experiment 1: Detection EAAT3 expression in the hippocampus of adult and aged mice Six adult and six aged mice were randomly selected as Adult and Old groups. Animals were decapitated, and their hippocampi were removed for western blot analysis. Experiment 2: The establishment of mice with hippocampus EAAT3 knockdown Recombinant adeno-associated viral (rAAV) vectors were microinjected into the mouse hippocampus by stereotactic technique to construct the hippocampal EAAT3 knockdown model. After receiving bilateral hippocampal microinjection of RNA interference vector (RNAi), 36 adult mice were randomly divided into six groups (n=6 for each): CON (immediately after microinjection, Day 0), D1, D7, D14, D21, and D28. The mice were used to obtain hippocampal tissue for RT-qPCR and Western blot analysis.
Twenty adult mice were randomly separated into two groups according to receiving rAAV-shRNA-NC (NC) or rAAV-shRNA-mSLC1A1 (RNAi) (n=10): NC group, and RNAi group. The spontaneous activities of mice were observed by an Open eld test (OFT) 21 days after microinjection in the hippocampus. Five mice in each group were randomly selected to perform brain immuno uorescence after OFT. Experiment 3: Effect of LPS on hippocampal EAAT3 knockdown mice and mechanism Adult mice were randomly assigned to four groups according to whether they received rAAV-shRNA-NC (NC), rAAV-shRNA-mSLC1A1 (RNAi), ACSF/LPS (n=12): NC+ACSF group, NC+LPS group, RNAi+ACSF group, and RNAi+LPS group.
Sixteen days after microinjection in the hippocampus, four groups of mice received MWM training for 5 days. Twenty-one days after microinjection in hippocampus, all mice received intracerebroventricular microinjection of ACSF or LPS. After 24h, mice were subjected to the MWM probe test. Six mice in each group were used for harvesting the hippocampus for western blot analysis; in other mice brain tissues were dissected for Golgi-Cox Staining after the behavioral observation. Experiment 4: Effect of Riluzole on LPS-induced cognitive impairment in the old mice The old mice were randomly divided into four groups according to whether they received DMSO, Riluzole, or LPS treatment (n=8 in each group): Old+DMSO, Old+Riluzole, Old+DMSO+LPS, and Old+Riluzole+LPS.
The mice in Old+DMSO+LPS and Old+Riluzole+LPS received an intraperitoneal injection of DMSO or Riluzole 2 days before LPS microinjection for consecutive 3 days. The probe test for reference memory was conducted 1 day after LPS administration, and the hippocampus was obtained for western blot analysis after the test.

Construction of hippocampal EAAT3 knockdown mouse model mediated by shRNA
Four potential different shRNA sequences (shRNA-mSLC1A1-1 4) targeting mSLC1A1 and the negative control shRNA (shRNA-NC) were designed and synthesized to construct rAAV vectors respectively named rAAV-shRNA-SLC1A1-1 4 and rAAV-shRNA-NC by Gemma Gene Company (Suzhou, China). To identify the most effective shRNA sequence that could knockdown EAAT3 in the hippocampus, we screened 4 different sequences to infect the HT-22 cell line, nding that rAAV-shRNA-SLC1A1-2 displayed the lowest EAAT3 expression level by RT/PCR and western blot (Suppl. Fig. 1).
For stereotactic injection of rAAV vector into the bilateral hippocampus, mice were anesthetized with pentobarbital sodium (70mg/kg) and placed on brain stereotaxic apparatus (RWD Life Science, Shenzhen, China). After exposing the skull via an incision, two small holes were drilled for injection. The stereotactic coordinates were 2.1 posterior, ± 1.7 lateral, and 2.0 ventral from bregma. Injection speed was 50nl/min, and the needle was kept in place for an additional 15 minutes before it was slowly withdrawn. RNAi group and RNAi+LPS group received bilateral hippocampal microinjection of rAAV-RNAi at 1μL per side (1×10 13 TU/mL), while the NC group and NC+LPS group received an equal volume of negative control rAAV-NC.

Open eld test
To evaluate the effects of hippocampal injection of rAAV-RNAi on spontaneous activity in mice, an open eld test was carried out 21 days after bilateral hippocampal rAAV-RNAi injection. Mice were placed in the corner of an opaque plastic box (50×50×30cm) in which the base was equally divided into 16 parts (4×4). A camera was set up right above the box to record all the activities of the mice. The parameters such as total moving distance, moving speed, and times of grid crossing were recorded for 5 min and analyzed by the ANY-MAZE system. The open eld was cleaned with 5% ethyl alcohol and allowed to dry between tests.

Establishment of LPS-induced cognitive impairment model
The LPS-induced cognitive impairment mouse model was performed according to the previously described protocol [7]. LPS was administered via the intracerebroventricular (i.c.v.) route; the stereotactic coordinates were 0.5 posterior, ± 1.0 lateral, and 2.0 ventral from bregma. After anesthesia with pentobarbital sodium (65-70 mg/kg ip), mice received 2μL LPS microinjection.

MWM test
The MWM test, a hippocampal-dependent test of spatial learning and memory for rodents, was performed as previously described [25] with minor modi cations. The water maze was a stainless steel circular pool (diameter 125cm, high 50cm) with a white inner wall, lled with opaque water containing skimmed milk powder at 22 ± 1.0 ℃ (water depth 25cm).
The pool with automatic visual tracking cameras on the ceiling can record the whereabouts of mice. The pool was divided into four equal quadrants , , , and . An escape platform (diameter 10cm) xed in the rst quadrant (target quadrant) was submerged below the water surface 1cm. The spatial learning was evaluated through a 5-day repetitive trial. Mice were randomly released into the pool facing the wall and trained to nd the platform within 60s. When the mouse failed to nd the platform, it was guided to it. All mice were given four trials (once per quadrant; swim-start position randomized) each day and were allowed to stay on the platform for 10s. After the daily session, mice were dried under a heater and returned to the home cage.
Animals underwent LPS microinjection 24 h after the nal acquisition trial, and a probe trial was conducted by removing the hidden platform to assess spatial reference memory 24 h after the microinjection. Total swimming distance, average speed, platform-site crossings, the times of entering the target quadrant (original platform quadrant), and time in the target quadrant were recorded. Each mouse was placed in the pool for 60 s at a time, and the starting point of entry was the third quadrant (opposite to the rst quadrant).

Western blot
The total protein and membrane protein were extracted from hippocampal tissue of the mice using a

Fluorescence immunohistochemistry
The brain tissue of mice was xed with 4% paraformaldehyde for 2 days and then embedded in para n.
Coronal 3μm sections were prepared and stained with uorescence immunohistochemistry. First, para n sections were dewaxed and placed in the EDTA buffer (pH8.0) to repair the antigen. Second, sections were washed in 0.01% Triton X-100 in phosphate-buffered saline (PBS-T) and blocked with 3% hydrogen peroxide for 15 min at room temperature. Then, samples were incubated overnight at 4℃ in the appropriate primary antibody, anti-EAAT3 (1:400; Cell Signaling Technology). Next, the sections were incubated with the appropriate uorescent secondary antibody, anti-rabbit IgG (1:400; ZF-0513), for 30 min at 37℃. After washing out the secondary antibodies, sections were incubated with 4',6-diamidino-2phenylindole (DAPI) for nuclear staining. Immuno uorescence was captured with a scanning confocal microscope.

Golgi-Cox Staining
The Golgi-Cox method is one of the most effective techniques for studying the morphology of neuronal dendrites and dendritic spines [26]. The brains of mice were quickly removed and rinsed with double distilled water and were stained with the FD Rapid Golgi Stain™ kit (FD Neuro Technologies, Ellicott City, MD, USA) according to manufacturer's instructions. Coronal slices (100 µm thickness) were obtained by using a cryostat (Leica, Wetzlar, Germany), and then were placed on a gelatin-coated microscope slide, stained and dehydrated. Images were taken by using an Eclipse Ci-L microscope (Nikon, Japan) and Image-Pro Plus 6.0 software. Sholl analyses were performed using the ImageJ 1.51K Sholl plugin. Spine density was estimated as the number of spines per 10 µm of dendrite length. The number of dendrites was estimated by counting dendritic intersections with multiple circular regions of interest centered on the cell soma with a spacing of 10 μm.

Statistical analysis
All data were analyzed by an observer who was blinded to the experimental protocol. Fisher exact probability method was used for descriptive analyses. Statistical comparisons between and within groups were made by two-way ANOVA, followed by a Tukey test where necessary. For acquisition training, data were analyzed using two-way ANOVA (treatment × trial time) with repeated measures (trial days) followed by the Bonferroni post hoc test. For all other data, two-way ANOVA was used. The results of behavioral counting were expressed as median ( rst quartile, third quartile), and the other results were expressed as mean ± standard error (SE); P values <0.05 were considered as statistically signi cant.

Hippocampal EAAT3 protein expression was signi cantly decreased in old mice
To determine whether there was any difference in the expression of EAAT3 in the hippocampus of adult and old mice, we measured the expression of total protein and membrane protein in the hippocampus of mice in the Adult and Old group by western blot. The results showed that EAAT3 expression of total protein (Fig. 1A, ** P < 0.01) and membrane protein (Fig. 1B, ** P < 0.01) in the hippocampus of the Old group was signi cantly lower than that in the Adult group.
2. Expression of EAAT3 in adult mice hippocampus was signi cantly decreased 21 days after microinjection rAAV-RNAi targeting EAAT3 To verify the effect of RNAi mediated-knockdown of EAAT3 on adult mouse hippocampus, the hippocampus tissues were collected at different time points (0, 1, 7, 14, 21, and 28 days) after the microinjection of rAAV-RNAi for RT-qPCR and western blot analysis. The PCR results showed that the hippocampal EAAT3 mRNA expression was signi cantly decreased on Day 21 ( * P < 0.05) and Day 28 (*P < 0.05) compared with the CON group (Fig. 2B). Moreover, the western blot results showed that the hippocampal EAAT3 protein levels were signi cantly decreased on Day7 ( *** P<0.001) and remained downregulated on Day14 ( ** P<0.01), Day21 ( *** P<0.001), and Day28 ( *** P<0.001) compared with the CON group (Fig. 2C). In addition, the immuno uorescence results showed that the expression of EAAT3 in the hippocampal CA area of the RNAi group was signi cantly lower than that of the NC group after receiving hippocampal microinjection for 21 days (Fig. 2A).
3. EAAT3 knockdown in the hippocampus signi cantly increased the motility of adult mice Twenty-one days after hippocampal microinjection of rAAV-RNAi/NC, adult mice's autonomic activity was examined by open eld test. The results showed that the total moving distance ( ** P < 0.01), moving speed ( * P < 0.05), and times of grid crossing ( * P < 0.05) in RNAi group were signi cantly higher than those in the NC group (Fig. 3).

EAAT3 knockdown in hippocampus signi cantly aggravated LPS-induced learning and memory de cit in adult mice
The MWM test was conducted to examine the effects of EAAT3 knockdown in the hippocampus on the LPS-induced learning and memory de cit model. In the MWM test, all adult animals learned the original platform location on the fourth day during the acquisition phase. The escape latency in all groups shortened as the training times increased ( ** P < 0.01, * P < 0.05); yet, no difference was observed between the groups on the same day (Fig. 4A). During the probe test, it was observed that the times of entering the target quadrant of mice in RNAi + ACSF group and NC + LPS group were signi cantly reduced compared with that in NC + ACSF group ( * P < 0.05), and that in RNAi + LPS group were lower than that in NC + LPS group ( # P < 0.05) (Fig. 4B). Similarly, the platform-site crossings of mice in RNAi + ACSF group and NC + LPS group were signi cantly reduced compared with that in NC + ACSF group ( * P < 0.05), and the platform-site crossings of mice in RNAi + LPS group were lower than that in NC + LPS group ( # P < 0.05) (Fig. 4C).
5. LPS signi cantly decreased EAAT3 plasma membrane protein level in the adult mice To determine the effect of LPS on EAAT3 protein expression, western blot analyses were performed to detect both EAAT3 total protein levels and membrane protein levels in the hippocampus. As shown in Fig. 5A, we found that EAAT3 total protein level in RNAi + ACSF group was signi cantly lower than that in NC + ACSF group ( * P < 0.05), and that EAAT3 total protein level in RNAi + LPS group was also signi cantly lower than that in NC + LPS group ( # P < 0.05). There is no signi cant difference regarding EAAT3 total protein level between the RNAi + ACSF group and RNAi + LPS group.
As shown in Fig. 5B, EAAT3 plasma membrane protein level in NC + LPS group was signi cantly lower than that in NC + ACSF group ( ** P < 0.01), and EAAT3 plasma membrane protein level in RNAi + LPS group was signi cantly lower than that in NC + LPS group ( # P < 0.05) and RNAi + ACSF group ( && P < 0.01).
6. Expression of GluA1 proteins and GluA1 phosphorylation is inhibited by LPS in the hippocampus of EAAT3 knockdown mice Next, we explored the effect of EAAT3 knockdown and LPS microinjection on GluA1 expression. As shown in Fig. 6A, the GluA1 total protein expression level in the RNAi + ACSF group was signi cantly decreased compared to that in the NC + ACSF group ( * P < 0.05), and the GluA1 total protein expression level in the RNAi + LPS group was signi cantly decreased compared to that in NC + LPS group ( # P < 0.05).
As shown in Fig. 6B, the phosphorylation level of GluA1 protein at Serine 845 site (p-GluA1-Ser-845) in NC + LPS group was signi cantly lower than that in NC + ACSF group ( * P < 0.05), and p-GluA1-Ser-845 level in RNAi + LPS group was signi cantly lower than that in NC + LPS group ( # P < 0.05) and RNAi + ACSF group ( & P < 0.05).

Effects of LPS on neuronal dendritic morphology in hippocampus of EAAT3 knockdown mice
To investigate the effect of EAAT3 knockdown and LPS microinjection on neuronal dendritic morphology, we measured the density of dendritic spines and dendritic branch numbers of the hippocampus one day after LPS intracerebroventricular injection. As shown in Fig. 7A, the densities of dendritic spines were markedly decreased in the NC + LPS group ( *** P < 0.001) and RNAi + ACSF group ( * P < 0.05) compared with that in the NC + ACSF group, and that in the RNAi + LPS group was signi cantly lower than that in NC + LPS group ( # P < 0.05) and RNAi + ACSF group ( &&& P < 0.001).
As shown in Fig. 7B, the number of intersections between dendritic processes and ten concentric circles centered on the cell soma with a spacing of 10µm were counted to estimate the number of dendrites. We found that the number of intersections in the NC + LPS group was signi cantly decreased compared to that in the NC + ACSF group ( * P < 0.05), and the number of intersections in the RNAi + LPS group was signi cantly decreased compared to that in the NC + LPS group ( # P < 0.05) and RNAi + ACSF group ( & P < 0.05).

Riluzole improves learning and memory ability and LPSinduced cognitive impairment in old mice
Finally, we explored whether riluzole could improve the cognitive de cits induced by LPS in old mice.
Regarding the MWM test, during the training phase, all old mice were able to learn the original platform location on the fourth day, and the escape latency in all groups shortened as the training times increased compared with the rst day ( * P < 0.05); yet, no difference was observed between the Old + DMSO group and Old + Riluzole group on the same day (Fig. 8A). During the probe test, it was observed that compared with the Old + DMSO group, the time in the target quadrant of mice in the Old + Riluzole group was signi cantly increased ( * P < 0.05), whereas the time in the target quadrant of mice in the Old + DMSO + LPS group was signi cantly decreased ( * P < 0.05). In addition, the time in the target quadrant of mice in the Old + Riluzole + LPS group was signi cantly decreased compared with that in the Old + DMSO + LPS group ( # P < 0.05) (Fig. 8B). Similarly, compared with the Old + DMSO group, the platform-site crossings of mice in the Old + Riluzole group were signi cantly increased ( * P < 0.05), whereas the platform-site crossings of mice in the Old + DMSO + LPS group were signi cantly decreased ( * P < 0.05). Also, the platform-site crossings of mice in the Old + Riluzole + LPS group were signi cantly less than that in the Old + DMSO + LPS group ( # P < 0.05) (Fig. 8C).

Riluzole signi cantly increased EAAT3 membrane protein level in the hippocampus of the old mice
To explore a potential mechanism for riluzole treatment, improving learning and memory in old mice, we investigated the effect of riluzole on the expression levels of EAAT3 in the hippocampus by western blot. For EAAT3 total protein expression, there was no signi cant difference among different groups (Fig. 9A). Yet, EAAT3 plasma membrane protein expression in the Old + Riluzole group was signi cantly increased compared with that in the Old + DMSO group ( * P < 0.05). Moreover, the EAAT3 plasma membrane protein expression in the Old + DMSO + LPS group was signi cantly decreased compared with that in the Old + DMSO group ( * P < 0.05), and EAAT3 plasma membrane protein expression in the Old + Riluzole + LPS group was signi cantly lower than that in the Old + Riluzole group ( # P < 0.05) (Fig. 9B).

Discussion
Postoperative cognitive dysfunction (POCD) is commonly observed in perioperative care following surgery and general anesthesia in elderly individuals. Yet, its underlying mechanisms remain largely unknown. In addition, no preventive or interventional agents have been established so far. Previous studies suggested that extracellular glutamate increased in the brain with age and that its dysregulation is associated with impaired learning and memory [27][28][29].
EAAT3 is part of a family of Na + -dependent excitatory amino acid transporters that regulate extracellular glutamate homeostasis in the CNS, which is encoded by the SLC1A1 gene and is enriched in the hippocampus neurons [11,12].. EAAT3 has been found to play a critical role in learning and memory [30]. And previously, we found that lack of EAAT3 effects leads to impaired cognition after iso urane exposure in EAAT3(-/-) mice [31]. Accordingly, we detected EAAT3 expression in adult and old mice. Our results revealed that both expressions of EAAT3 total protein and membrane protein in the old mice hippocampus were signi cantly decreased compared to the adult mice. Combining with the higher incidence of POCD in the elderly, we inferred that the dysfunction and degradation of EAAT3 expression increased the susceptibility to POCD in the elderly. However, the underlying mechanism is still elusive.
Previous studies have used the EAAT3 gene knockout animal model to explore the function of EAAT3 in the CNS [32][33][34]. Considering that the animal models with reduced gene expression rather than deletion are of clinical relevance since they better re ect the impact of human polymorphisms affecting protein levels and rarely complete loss of expression [35], in this study, we constructed rAAV-mediated shRNA to knockdown hippocampal EAAT3 expression via hippocampal microinjection, which achieved the local interference and minimized the impact on the whole brain and the whole body [36]. The results of RT-qPCR and western blot con rmed that bilateral hippocampal injection of RNAi could signi cantly inhibit hippocampal EAAT3 expression and achieve a stable knockdown effect from 21 to 28 days after RNAi injection, which is consistent with the time point selected by other AAV vectors injection studies [37,38].
Also, we found that the autonomic activity (including total moving distance, moving speed, and times of grid crossing) in the RNAi group was signi cantly higher than those in the NC group, which may be due to the increased glutamate between synaptic space induced by the interference of EAAT3 expression.
Glutamate acts as the major excitatory neurotransmitter and mediates fast neurotransmission in neuronal networks [39]. The interference of EAAT3 expression can disrupt the absorption of extracellular glutamate (Glu) and lead to neuronal hyperexcitation and excitotoxicity injury [40]. These results indicated that the interference of EAAT3 expression could increase extracellular glutamate level and nally, brain excitability. However, the role of EAAT3 in POCD has been rarely explored.
We used the intracerebroventricular administration of LPS to construct the POCD model according to our previously described approach [7]. Our results revealed that LPS led to signi cant learning and memory de cits detected by the MWM task and that interference of EAAT3 expression by rAAV-RNAi signi cantly aggravated the learning and memory impairment induced by LPS. These results indicated that EAAT3 dysfunction increased the susceptibility to LPS-mediated learning and memory de cits. However, the underlying mechanism is still elusive.
Under basal conditions, EAAT3 is primarily sequestered in the intracellular compartment, with about 20% of the transporter localized at the cell surface [41], which is a key determinant of its buffering e ciency [42] and accounts for 40% of glutamate uptake in the hippocampus [10]. In this study, we found that LPS treatment did not signi cantly decrease the EAAT3 total protein level ether in NC group or in RNAi group but signi cantly decreased the EAAT3 plasma membrane protein level both in NC group and in RNAi group, thus indicating that LPS mainly disrupted the expression of EAAT3 plasma membrane protein.
Interestingly, we also found that, in ACSF group, rAAV-RNAi only signi cantly decreased EAAT3 total protein level, but did not signi cantly decrease EAAT3 membrane protein level; however, in LPS treatment group, rAAV-RNAi signi cantly decreased both EAAT3 total protein level and EAAT3 membrane protein level, which further demonstrated that LPS primarily disrupted EAAT3 plasma membrane protein. So far, no study has reported this nding. Yet, the expression of SLC1A1, encoding EAAT3, is impaired in diseases related to neuroin ammation, such as multiple sclerosis, schizophrenia, hypoxia/ischemia, and epilepsy [43], which suggests that neuroin ammation induced by LPS maybe lead to the disruption of EAAT3 plasma membrane protein. However, the underlying mechanisms need to be further explored.
Tra cking of AMPAR to the plasma membrane in the neurons is believed to be a fundamental cellular process for learning and memory [44][45][46], mediated via the phosphorylation of GluA1 (an AMPAR subunit) at serine 845 (P-GluA1-s845) and then promoting AMPAR tra cking to the plasma membrane [45][46][47]. In this study, we found that LPS treatment did not signi cantly decrease the GluA1 protein level in the NC group or in the RNAi group, while it signi cantly decreased the P-GluA1-s845 level, which indicated that LPS primarily interfered with the P-GluA1-s845 level. In addition, we found that interference of EAAT3 expression by rAAV-RNAi signi cantly decreased both the GluA1 protein level and the P-GluA1-s845 level, thus indicating that EAAT3 could affect the expression of GluA1 and P-GluA1-s845. Moreover, our previous research also found that EAAT3 could regulate GluA1 tra cking to the plasma membrane [48]. These results indicated that LPS disrupted the P-GluA1-s845 by interfering with EAAT3 expression, and EAAT3 may become a new therapeutic target for POCD.
Dendritic spines, de ned as small protrusions arising from dendrites, are used for communication by excitatory glutamatergic synapses of the CNS. They are remarkably dynamic structures that may undergo adaptive changes following different stimuli [49], and their structural plasticity is thought to underlie memory formation [50]. In this study, we found that LPS treatment signi cantly decreased the densities of dendritic spines and dendritic processes intersections. Many studies have shown that LPS could signi cantly reduce the dendritic spine density in hippocampal neurons [51,52] by activating microglia to hyperactively secret proin ammatory cytokines [53]. In addition, glutamate excitotoxicity, which could be induced by the interference of EAAT3 expression, is a mechanism that causes secondary damage to neurons, accompanied by loss of dendritic spines and changes of synaptic activity [54,55]. Our data indicated that interference of EAAT3 expression signi cantly decreased the dendritic spine density, and LPS treatment further aggravated this impairment. Thus, we inferred that LPS decreased the dendritic spine density by interfering with EAAT3 expression.
Previous studies have indicated that riluzole is bene cial for neurodegenerative diseases such as AD [24] and PD [22,23]. In this study, we found that, in old mice, riluzole pretreatment signi cantly promoted their learning and memory function and increased the EAAT3 membrane protein level, but not EAAT3 total protein level. This demonstrated that the riluzole could improve the cognitive function of old animals by promoting EAAT3 membrane protein level. In LPS treated old mice, riluzole pretreatment signi cantly alleviated LPS induced learning and memory de cits, which may be related to its up-regulation effect of EAAT3 membrane protein level. However, riluzole did not promote EAAT3 membrane protein level in LPS treated old mice, which may be due to the serious disruption of LPS on EAAT3 membrane protein in old mice. Thus, riluzole could be identi ed as a new strategy for POCD prevention in old surgery patients.
In conclusion, this study showed that the decrease of EAAT3 expression and function is associated with advanced age, and the decrease of EAAT3 expression can aggravate the cognitive impairment induced by LPS. In ammation-induced by LPS signi cantly decreases the plasma membrane expression of EAAT3 in the hippocampus, which in turn inhibits the phosphorylation of Glua1, the mechanism of AMPAR transport to the plasma membrane, and reduces synaptic density in hippocampal neurons. This may be the reason why EAAT3 knockdown aggravates the cognitive impairment induced by LPS. Availability of data and materials The datasets during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
The experimentation procedures were approved by the Animal Care Committee of Chinese PLA General Hospital (Beijing, China). The maintenance and handling of the mice were consistent with the guidelines of the National Institutes of Health, and adequate measures were taken to minimize animal discomfort.

Consent for publication
Not applicable.    Data are expressed as median ( rst quartile, third quartile) (n=12). *P <0.05, vs. NC+ACSF group; #P <0.05, vs. NC+LPS group. Protein bands on the gel and their relative intensities. The expression levels of EAAT3 protein in total protein and plasma membrane of the hippocampus in adult mice were normalized to that of β-actin as an internal control. Data are expressed as mean ± SE (n=6). *P <0.05, **P <0.01, vs. NC+ACSF group; #P <0.05, vs. NC+LPS group; &&P <0.01 vs. RNAi+ACSF group.

Figure 6
Expression of GluA1 proteins and GluA1 phosphorylation is inhibited by LPS-induced neuroin ammatory in the hippocampus of EAAT3 knockdown mice. (A, B) Protein bands on the gel and their relative intensities. The expression levels of GluA1 protein and Phospho-GluA1-Ser845 in the total protein of hippocampus were normalized to that of β-actin as an internal control. Data are expressed as mean ± SE (n=6). *P <0.05, vs. NC+ACSF group; #P <0.05, vs. NC+LPS group; &P <0.05, vs. RNAi+ACSF group.   Riluzole increases the expression of EAAT3 in hippocampal membrane protein of old mice. (A, B) Protein bands on the gel and their relative intensities. The expression levels of EAAT3 protein in the total protein and plasma membrane of the hippocampus in aged mice were normalized to that of β-actin as an