CXCR5 Down-regulation Alleviates Cognitive Dysfunction in a Mouse Model of Sepsis-associated Encephalopathy: Potential Role of Neuronal Autophagy and the p38MAPK/NF-κB/STAT3 Signaling Pathway

Background Cognitive decits are common in patients with sepsis. Previous studies in sepsis-associated encephalopathy (SAE) implicated the C-X-C chemokine receptor type (CXCR) 5. The present study used a mouse model of SAE to examine whether CXCR5 down-regulation could attenuate cognitive decits. Methods Sepsis was induced in adult male C57BL/6J and CXCR5 -/- mice by cecal ligation and puncture (CLP). At 14-18 days after surgery, animals were tested in a Morris water maze, followed by a fear conditioning test. Transmission electron microscopy of hippocampal sections was used to assess levels of autophagy. Primary microglial cultures challenged with lipopolysaccharide (LPS) were used to examine the effects of short interfering RNA targeting CXCR5, and to investigate the possible involvement of the p38MAPK/NF-κB/STAT3 signaling pathway. signaling to inhibit hippocampal autophagy during sepsis and thereby contribute to cognitive dysfunction. Down-regulating CXCR5 can restore autophagy and mitigate the proinammatory microenvironment in the hippocampus.

Mechanisms for the development SAE are not fully understood. A leading hypothesis is that sepsis induces neuroin ammation in the central nervous system (CNS), leading to brain damage and dysfunction [7,8]. Microglia, the major resident immune cells in the CNS are activated upon sepsis, and shift from a "surveillance" phenotype to a proin ammatory M1 phenotype to release in ammatory signals [9,10]. In this way, microglial activation exacerbates neuronal injury and impairs learning and memory [11]. In mouse models, SAE could be alleviated by blocking microglial activation via inhibiting the IL-17A/IL-17R in ammatory pathway [12], injecting attractylone to polarize microglia toward the M2 phenotype [13], or injecting minocycline [14] or the ginsenoside Rg1 [15] to inhibit neuroin ammation.
Autophagy, an evolutionarily conserved catabolic process to recycle damaged or senescent organelles and proteins [16], is elevated upon sepsis in hepatocytes [17], cardiomyocytes [18], as well as the CNS [10,19]. The process of autophagy includes lysosome activation, autophagosome formation, conversion of microtubule-associated protein 1 light chain 3 (LC3) from form I to form II, and a reduction in levels of Beclin-1, LAMP-1 and Rab7 [19]. Deletion of the genes that encode autophagy-related proteins, such as autophagy related gene-5 (Atg-5), exacerbates the production of pro-in ammatory cytokines in multiple tissues following sepsis [20], suggesting increased autophagy is a compensatory response that limits sepsis-induced tissue damage.
De ciency in the C-X-C motif chemokine receptor 5 (CXCR5) in retinal pigment epithelium cells has been linked to up-regulation of autophagy [21]. A previous study in a mouse model of sepsis from this laboratory showed that CXCR5 contributes to hippocampal neuroin ammation, subsequently leading to hippocampal neurogenesis disorder and cognitive impairment, and CXCR5 de ciency alleviates sepsisinduced de cits in hippocampal neurogenesis and cognitive function [22]. In the current study, we examined whether down-regulating CXCR5 could alleviate cognitive de cits induced by sepsis by regulating autophagy in hippocampus. Considering previous studies that linked CXCR5 to p38MAPK activation [23], and p38MAPK to autophagy and neuroin ammation [24], the mechanistic investigation in the current study focused on the p38MAPK/NF-κB/STAT3 pathway.

Animals
Study protocols involving animal subjects were approved by the Animal Ethics Committee of Nanjing  (Institute of Pain Medicine, Nantong University, China) [23]. Mice were housed in a pathogen-free facility in the Experimental Animal Center at Nanjing First Hospital, and maintained under a 12-hour light-dark cycle with ad libitum access to standard food and water.

Study design
This study consisted of both in vivo and in vitro experiments. The in vivo experiments were conducted in mice (C57BL/6J or CXCR5 −/− ) with cecal ligation and puncture (CLP; Fig. 1A-B). A group of C57BL/6J mice undergoing sham surgery was included as a control. Tests for cognitive functions included a standard 5-day Morris water maze (14 days after CLP) and then a fear conditioning test (19 days after CLP). Separate groups of mice were sacri ced 3, 7, and 14 days after CLP for Western blotting, transmission electron microscopy and immunohistochemistry analyses in the hippocampus. The in vitro experiments were conducted in primary microglial cells (Fig. 1C). Brie y, cells were seeded at a density of 3.5 × 10 5 cells/mL in serum-free DMEM (Thermo Fisher Scienti c, Waltham, MA, USA) into 6-well plates containing coverslips, incubated at 37 ℃ overnight, and then exposed to lipopolysaccharide (LPS; Thermo Fisher Scienti c) at a concentration of 200 ng/mL for 24 h [13]. In some experiments, cultures were pretreated with short interfering RNA (siRNA) targeting CXCR5 (5'-CUGGACAGAUUGGACAACU-3'; GenePharma, Shanghai, China) or with scrambled sequence (5'-UUCUCCGAACGUGUCACGU-3'; GenePharma) at 24 h before LPS challenge [22]. The siRNAs (20 pmol) were dissolved in 100 µL serumfree OptiMEM, mixed with 1 µL Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA), and added to the cultures. In some experiments involving CXCR5 siRNA, cells were pretreated with the p38MAPK inhibitor SB203580 (10 mmol/L; Sigma-Aldrich, St Louis, MO, USA), the p38MAPK agonist P79350 (50 mmol/L; Calbiochem, La Jolla, CA, USA) or vehicle for 1 h prior to adding CXCR5 siRNA administration. Cells were collected at 24 h after LPS treatment for Western blotting and immuno uorescence analysis. CLP CLP was conducted as previously described [25]. Brie y, mice were anesthetized with intraperitoneal injection of sodium pentobarbital (40 mg/kg). After disinfection, an incision 2-3 cm long was made at 1.5 cm below the xiphoid to expose the abdominal cavity. The cecum was isolated and ligated at half the distance between the distal pole and the base of the cecum, punctured with a 22-gauge needle, and gently squeezed to force the fecal contents into the peritoneal cavity. The cecum was returned to the peritoneal cavity, and the abdomen was closed using 3 − 0 silk sutures. In the case of sham-operated mice, the abdominal cavity was opened to expose the cecum without ligation or puncture.
Core body temperature was maintained at 37 ± 0.5 ℃ using a heating blanket during surgery. At the end of surgery, mice received 40 ml/kg sterile saline via subcutaneous injection and returned to home cage with a warm cotton pad and free access to food and water. No antibiotics were given.
containing 10% heat-inactivated fetal bovine serum (Sigma-Aldrich) and antibiotics. The cell suspension was ltered through a 100-µm cell sieve (Becton Dickinson AB, Stockholm, Sweden) and centrifuged at 864 g for 5 min at 4 ℃. The pellet was suspended in DMEM, then incubated in a 75-mL culture ask (Thermo Fisher Scienti c) for 14 days in a humidi ed incubator at 37 ℃ in an atmosphere containing 5% CO 2 . Next, the ask was shaken at 100 rpm at 37 ℃ for 6 h on a rotary shaker to harvesting microglia with centrifugation (288 g at 4 ℃ for 5 min).

Morris water maze
Learning and memory were examined using a standard 5-day Morris water maze test, as described previously [28]. A circular water pool (125-cm diameter and 40-cm height) was lled with opaque water to a depth of 30 cm and maintained at 23 ± 1 ℃. The escape platform (10-cm diameter) was placed 1 cm below the water surface in the target quadrant. Each mouse was given 4 daily trials with a 20-min intertrial interval for 4 consecutive days to nd the platform, and individually placed in the pool at one of 4 quadrant locations. Mice that failed to locate the escape platform within 60 s were manually guided to the platform, and allowed to stay for 20 s. The escape latency from four sessions on the same day was averaged. On the fth day, the platform was removed to allow for probe testing. During the 60-s session, the number of crossings over the target quadrant and the total time spent in target quadrant were recorded. Training and probe tests were analyzed using motion detection software (Actimetrics Software, Evanston, IL, USA).

Fear conditioning test
The fear conditioning test was conducted on the next day after the Morris water maze experiments were completed, using a standard fear conditioning chamber with metal grid oor [28]. Mice were allowed to the chamber for 2 min prior to 30-sec mono-frequency sound (2 kHZ, 80 dB); during the last 2 sec of the sound, foot shock (1 mA) was delivered. After 3 min, this process was repeated once. On the next day, mice were placed in the same chamber again, but without any stimulation. Freezing was de ned as a completely immobile posture except for respiration [29].

Immuno uorescence
Brains were post-xed in 4% paraformaldehyde overnight at room temperature, embedded in an optimal cutting temperature compound (Sakura Finetek, Torrance, CA, USA), and cut into 5-µm coronal sections.

Statistical analysis
Statistical analysis was performed using GraphPad Prism 9.0.0 (Graph Pad Software, San Diego, CA, USA). All continuous variables followed normal distribution (Shapiro-Wilk test; data not shown), and were reported as mean ± standard deviation (SD). Data from Morris water maze training were analyzed using two-way ANOVA for repeated-measures, followed by the Bonferroni post hoc test for multiple comparisons. All other variables were analyzed using Student's t test (for comparisons between two groups), or one-way ANOVA followed by Tukey's multiple test (for comparisons among at least three groups). P < 0.05 was considered statistically signi cant.

Sepsis induces cognitive de cits and up-regulates CXCR5
In the Morris water maze, escape latency progressively decreased over the rst four days of training in sham and CLP groups (Fig. 2a). The CLP group showed longer escape latencies than the sham group. The CLP group spent shorter time and had fewer crossings in the target quadrant on the test day than the sham control (Fig. 2b, c). CLP mice exhibited shorter freezing time than the sham control (Fig. 2d).
Knocking out CXCR5 ameliorates cognitive dysfunction in SAE mice CLP mice showed higher hippocampal CXCR5 expression than the sham control (Fig. 3a, b), as well as longer escape latencies during Morris water maze training (Fig. 3c). Escape latencies were shorter in CXCR5 −/− mice that underwent CLP than in WT mice with CLP. CLP mice spent shorter time and had fewer crossings in the target quadrant on the test day than the sham control, whereas CXCR5 −/− mice that underwent CLP spent longer time in the target quadrant and made more crossings than CLP mice (Fig. 3d, e). CLP mice showed shorter freezing time than the sham control, whereas CXCR5 −/− mice that underwent CLP showed longer freezing time than CLP mice (Fig. 3f).

Knocking out CXCR5 promotes hippocampal autophagy in SAE mice
Transmission electron microscopy demonstrated more autophagosomes in the hippocampus from CLP mice than the sham control (Fig. 4a). The number was even higher in CXCR5 −/− mice that underwent CLP. Similarly, the hippocampus of CLP mice had higher numbers of LC3-positive cells and a higher LC3-II/LC3-I ratio than the sham control; these parameters were even higher in CXCR5 −/− mice that underwent CLP (Fig. 4b-e). The number of LC3-expressing cells or the LC3-II/LC3-I ratio did not differ between WT and CXCR5 −/− mice that underwent sham surgery.
Regardless of the genotype, mice that underwent CLP showed signi cantly higher levels of beclin-1 and Atg-5 and lower levels of p62 in the hippocampus than the sham control (Fig. 4d, f-h). These CLP-induced changes were even greater in CXCR5 −/− mice.
Knocking out CXCR5 attenuates sepsis-induced upregulation of phosphorylated p38MAPK, IL-1β and IL-6 in hippocampus CLP increased the levels of phosphorylated p38MAPK, IL-1β and IL-6 in the hippocampus in WT mice; the effects were less pronounced in CXCR5 −/− mice (Fig. 6).
Knocking down CXCR5 activates p38MAPK-dependent autophagy in microglial cultures treated with LPS LPS challenge triggered a signi cant increase in levels of LC3, beclin-1 and Atg-5; an increase in the LC3-II/LC3-I ratio; and a decrease in p62 in primary microglial cultures (Fig. 7). Such effects were augmented by either CXCR5 knockdown or treatment with the p38MAPK inhibitor SB203580. Conversely, the p38MAPK agonist P79350 mitigated the effects of CXCR5 knockdown on autophagy.
Knocking down CXCR5 attenuates LPS-induced microglial polarization and in ammatory cytokine production in microglial cultures LPS triggered an increase in Iba-1, an increase in the M1 marker CD86 and a decrease in the M2 marker CD206 (Fig. 8a-f). CXCR5 knockdown or the p38MAPK inhibitor SB203580 mitigated these changes. Conversely, the p38MAPK agonist P79350 mitigated the effects of CXCR5 knockdown on microglial polarization.
LPS triggered increases in the levels of the in ammatory cytokines IL-1β and IL-6 ( Fig. 8c). Such effects were attenuated by CXCR5 knockdown or the p38MAPK inhibitor SB203580 (Fig. 8g, h). Conversely, the p38MAPK agonist P79350 mitigated the effects of CXCR5 knockdown on in ammatory cytokine production.

Discussion
Results from the present study suggest that CXCR5 contributes to cognitive impairment by enhancing p38MAPK/NF-κB/STAT3 signaling. We further showed that CXCR5 knockout could restore autophagy, promote the microglial polarization to the M2 phenotype, and inhibit both pro-in ammatory cytokine release and p38MAPK activation in the hippocampus of the SAE mice. These effects were associated with attenuated cognitive dysfunction after CLP. In primary microglial cultures, we showed that CXCR5 knockdown could partially reverse LPS-induced phosphorylation of p38MAPK as well as p38MAPKdependent up-regulation of NF-κB and STAT3, ultimately restoring autophagy activation and inhibiting proin ammatory cytokine production.
CLP resulted in prolonged latency as well as decreased time and fewer crossings in the target quadrant in the Morris maze test, and decreased freezing time in the conditioned fear test. In addition, sepsis induced microglial M1 polarization and production of proin ammatory cytokines in mouse hippocampus, as well as in primary microglia cultures. These ndings are consistent with our previous work [28]. Growing evidence indicates that microglia are rapidly activated in response to septic challenge, and these cells produce substantial amounts of in ammatory cytokines [10]. Microglia-mediated neuroin ammation has a major role in the development of long-term cognitive dysfunction after sepsis [31]. Sepsis patients with encephalopathy have higher IL-6 levels in cerebrospinal uid than non-septic controls without encephalopathy [32]. Patients who die during septic shock show strong, localized up-regulation of the microglial M1 polarization marker CD68 in the hippocampus [33]. Sepsis also up-regulates the in ammatory cytokines TNF-α, IL-6 and HMGB1 in BV-2 microglia cultures and animals [10]. The present experiments in vivo and in vitro showed that sepsis induced the activation of microglial M1 polarization and the production of pro-in ammatory cytokines. These results provide evidence for a critical role of microglia-mediated neuroin ammation in SAE. Promoting the conversion of microglia M1 polarization to the M2 phenotype [13] or inhibiting neuroin ammation [11,28] can alleviate cognitive de cits and functional decline in SAE animals.
Autophagy is activated by the systemic in ammatory response in the septic hippocampus [19]. When autophagy is activated, LC3-I in the cytoplasm is converted to LC3-II, which aggregates on the autophagosome membrane; p62/SQSTM1 is the adaptor protein linking ubiquitin, LC3 and the autophagosome, and these proteins form a complex, which is degraded by lysosomes as a substrate of autophagy [34]. In this study, we found that sepsis induced more double-membrane autophagosomes, the shift from LC3-I to LC3-II, increased levels of beclin-1 and Atg-5, and reduction in free p62 in the mouse hippocampus, indicating the activation of hippocampal autophagy. Clinical and preclinical studies suggest that a certain degree of autophagy can protect the body from sepsis-induced injury. For example, pyrrolidine dithiocarbamate can increase autophagy in the hippocampus, helping protect brain tissue from sepsis-induced injury in rats [19]. Suppression of autophagy has been linked to worse clinical outcomes in patients with severe sepsis [35]. Impairment in autophagosome-lysosome fusion, which stalls autophagy, may contribute to sepsis-induced brain injury [36]. This literature and our ndings suggest that increasing levels of autophagy may help mitigate sepsis-induced brain injury. However, a balance is needed: excessive autophagy can destroy cell homeostasis [37]. Further work should explore what autophagy levels may exert the best therapeutic effects. In any case, our ndings con rm that autophagy is deregulated in the hippocampus of sepsis animals.
The same upstream signals may trigger autophagy and subsequent microglia-mediated neuroin ammation in sepsis. CXCR5, the only known receptor for the chemokine CXCL13, is known to mediate neuroin ammation and thereby contribute to impairment of learning and memory in intractable temporal lobe epilepsy patients and pilocarpine-induced epileptic rats [38]. The CXCL13/CXCR5 signaling axis contributes to neurodegeneration in individuals with cognitive de cit disease [38], as well as in animal models with neurodegeneration [23], as we have con rmed in our previous work [22,39]. Infection of the CNS induces production of CXCL13 in microglia, macrophages, and endothelial cells there [40,41]. CXCR5 de ciency in retinal pigment epithelial cells has been shown to deregulate autophagy, as re ected in decreased LC3B-II, increased p62, abnormal autophagosomes and impaired lysosome enzymatic activity [21]. This dysregulation of autophagy contributes to age-related macular degeneration in mice [42]. Our results suggest that CXCR5-mediated dysregulation of autophagy may also contribute to cognitive de cits in SAE. Our results further suggest that CXCR5 de ciency restores autophagy and inhibits microglia-induced neuroin ammation, ameliorating sepsis-induced cognitive dysfunction. Similarly, CXCR5 de ciency in mice reduces in ammatory pain and decreases activation of spinal microglia and astrocytes in a mouse model of neuropathic pain induced by peripheral injection of complete Freund's adjuvant [43].
We found that sepsis induced the phosphorylation of p38MAPK in the hippocampus in vivo and in microglia cultures. CXCR5 de ciency partially reversed this phosphorylation, inhibiting downstream activation of NF-κB and STAT3. Our results are consistent with other studies involving the CXCR5/CXCL13 axis. In a mouse model of neuropathic pain, CXCL13 acts via CXCR5 to activate p38MAPK signaling, triggering orofacial neuropathic pain and promoting neuroin ammation [23]. Sepsis in neonatal rats has been shown to activate p38MAPK signaling in the brain, and the p38MAPK inhibitor SB203580 can protect against sepsis-associated cognitive de cits [44]. CXCL13 level was increased in the serum of patients with sepsis, and CXCL13 also acts via p38MAPK signaling to drive LPS-induced hyperpermeability of the endothelium in human umbilical vein endothelial cells, suggesting that targeting CXCL13 may alleviate sepsis [45]. Similarly, p38MAPK mediates LPS-induced morphological changes and production of IL-1β in primary microglial cultures and the brain [24], and it contributes to acute lung injury in a mouse model by stimulating autophagy, oxidative stress and in ammatory responses [46]. Our results add SAE to the list of diseases in which p38MAPK drives pathophysiology and may therefore be a useful therapeutic target. Analogously, inhibiting p38MAPK using SB203580 has been shown in vitro to reverse the dysregulation of NF-κB and downstream STAT3 and to slow pancreatic tumor growth [47]. We show here that knocking out or down-regulating CXCR5 reduces p38MAPK activation and, consequently, its downstream signaling, ultimately ameliorating sepsis-induced cognitive defects.

Conclusion
Our results in a mouse model of SAE and in primary microglial cultures suggest that sepsis up-regulates hippocampal CXCR5, which contributes to incomplete activation of autophagy, polarization of microglia toward the M1 phenotype, production of in ammatory cytokines and appearance of cognitive de cits.
Our results further suggest that down-regulating CXCR5 can restore autophagy, polarize microglia toward the M2 phenotype, and inhibit p38MAPK/NF-κB/STAT3 signaling, ultimately attenuating sepsis-induced neuroin ammation and cognitive dysfunction. In this way, our study provides the strongest evidence so far that targeting the CXCR5/p38MAPK axis may help treat or even prevent neurodegeneration in SAE.