Gut microbiota is causally associated with poststroke cognitive impairment through lipopolysaccharide and butyrate

The subjects were recruited from the Department of Neurology of Nanfang Hospital of Southern Medical University (Guangzhou, China) from September 2018 to December 2020. The inclusion criteria were as follows: (i) an age greater than 18 years (ii) a diagnosis within 4 days of stroke onset according to established guidelines [22], and (iii) a National Institutes of Health Stroke Scale (NIHSS) score ≤ 8. The exclusion criteria were as follows: (i) patients who presented significant neurological deficits such as drowsiness, aphasia, or limb weakness and were, therefore, unable to complete the cognitive function test (ii) patients with a history of seizures and obvious cognitive impairment (AD8 ≥ 2) before stroke [23], mental disorders or significant emotional problems; (iii) patients with infectious diseases, such as pneumonia or urinary system infection; (iv) patients administered antibiotics or probiotics within 1 month before admission or during follow-up; and (v) patients for whom stool samples could not be obtained within 4 days of admission or at the 3-month follow-up. The serum samples were isolated by centrifugation at 3000 rpm for 10 min and stored at − 80 °C until testing. All participants provided written informed consent in accordance with the Declaration of Helsinki. This study was approved by the Ethics Committee of Nanfang Hospital, Southern Medical University (NFEC-2020-169) and registered at http://​clinicaltrials.​gov (NCT04688138).

Cognitive impairment was assessed by the Montreal Cognitive Assessment (MoCA) and the Mini-Mental State Examination (MMSE) 3 months after stroke onset. MMSE and MoCA scores range from 0 to 30. The MoCA is currently the most widely recognized tool for assessing cognitive function, including visuospatial/executive function, naming, attention, abstraction, language, delayed recall, and orientation [24]. Patients with a MoCA score < 22 were considered to have PSCI [25]; higher scores indicate better cognitive performance.

The experiments were approved by the Ethics Committee for Animal Care and Research of Zhujiang Hospital of Southern Medical University (Guangdong, China) and were performed according to the ARRIVE guidelines. As female hormone such as estrogen and follicle-stimulating hormone can significantly affect cognition [26, 27], to exclude possible influence of female hormone, male mice were used in this study. Adult male C57BL/6J mice (8–10 weeks, 22–25 g) were purchased from Guangdong Medical Laboratory Animal Center (Guangzhou, China). Upon arrival, mice were randomly divided into three groups in FMT experiment and were randomly divided into two groups in intraperitoneal injection experiment. Mice that were allocated to the same group were raised in the same cage. All animals were housed under controlled temperature and humidity conditions on a 12-h:12-h light/dark cycle and were provided food and water ad libitum. After acclimatization for 1 week, fecal samples from mice were collected and stored in a − 80 °C freezer until analysis. For antibiotic treatment, broad-spectrum antibiotics (1 g ampicillin, 1 g neomycin sulfate, and 1 g metronidazole, Sigma-Aldrich, CA, USA) were dissolved in 1 L drinking water (replace every 3 days) and provided ad libitum to the mice for 14 consecutive days, after which fecal samples were collected and stored. For sodium butyrate (NaB) treatment, 11 g NaB (Aladdin) was dissolved in drinking water to a concentration of 0.1 mol/L and provided ad libitum to the mice. Supplementation with NaB by this dose and regime effectively shapes the gut microbiota of mice [28, 29]. For lipopolysaccharide (LPS) treatment, LPS serotype 0111:B4 (Sigma-Aldrich, St. Louis, MO, USA; dissolved in phosphate-buffered saline (PBS), 5 μg/mouse) was administered intraperitoneally to the mice every 3 days. For FMT, fecal samples were collected from patients at 3 months after stroke. All individual samples from the same group were pooled together. Fecal microbiota suspensions were prepared by diluting and mixing 10 g of fecal samples obtained from PSCI or non-PSCI (nPSCI) patients in 100 mL of sterile PBS and vortexed vigorously. Then the suspensions were centrifuged at 500g for 1 min to remove insolubilized material and the supernatants were collected and stored in a − 80 °C freezer until further use. Three days after stroke, each mouse was intragastrically administered 0.2 mL of the suspension once daily for 28 consecutive days.

Focal cerebral ischemia was induced by transient (30 min) middle cerebral artery occlusion (MCAO) using an intraluminal filament as previously described [28]. Surgical anesthesia was induced by intraperitoneal injection of 1.25% tribromoethanol (0.02 mL/g of body weight). Body temperature was maintained throughout the procedure with a feedback-controlled heating blanket. A filament was introduced into the external carotid artery and gently advanced into the internal carotid artery until it reached the middle cerebral artery. After 30 min of cerebral ischemia, the filament was withdrawn to establish reperfusion. Two Mice were excluded due to absence of deficits in neurological function (modified neurological severity score ≤ 1). Two mice died after surgery, one mouse was excluded due to subarachnoid hemorrhage during surgery.

Neurological function was assessed by the modified neurological severity score (mNSS) [28]. The mice were tested 3 days after experimental stroke. The mNSS is a composite of motor (muscle status and abnormal movement), sensory (visual, tactile, and proprioceptive sensations), and reflex tests. Neurological severity was graded on a scale of 0 to 18 points. A higher score indicated a more severe brain injury. Cognitive functions were assessed by the Morris water maze on the next day after FMT or intraperitoneal injection. The mice were subjected to four trials per day for 5 consecutive days. In all trials, the mice were introduced into a pool (120 cm in diameter and 50 cm high) containing a 10-cm-diameter platform submerged 1 cm below the surface of the water in quadrant III. In this test, the mice searched for the hidden platform by using memory of visual cues around the pool. Each mouse was put in the water facing the pool wall and given 60 s to find the hidden platform. If a mouse failed to find the platform within a limited time (60 s), it was guided to swim to the platform and kept there for 20 s. The time taken to reach the platform (escape latency) was measured. On the 6th day of trails, a 60-s probe trial was conducted to evaluate the memory of the mice. The number of times the mice crossed the area of platform and the percentage of time staying in the target quadrant (quadrant III) were recorded with a digital video camera.

Bacterial genomic DNA was extracted using a QIAamp PowerFecal Pro DNA Kit (QIAGEN, Valencia, CA, USA) according to the manufacturer’s instructions. The barcoded primers V4F (GTGYCAGCMGCCGCGGTAA) and V4R (GGACTACNVGGGTWTCTAAT) were used to amplify the V4 variable region of the 16S rRNA gene. PCR was performed according to a previously described method [15]. All PCR amplicons were mixed and sequenced using the Illumina iSeq 100 platform. The Shannon index, phylogenetic diversity (PD) whole-tree index, and Chao1 index were determined to assess α-diversity. UniFrac distances were used to analyze the β-diversity by illustrating the phylogenetic dissimilarity among samples. A smaller UniFrac distance between two samples indicates a higher similarity. As a dimensionality reduction method, principal-coordinate analysis (PCoA) was used to describe the relationships among samples based on the distance matrix and visualize the unsupervised grouping pattern of the complex data set, i.e., the microbiome. Linear discriminant analysis effect size (LEfSe) was used to compare the discriminative data between groups.

Approximately 0.2 g of feces was homogenized in 1 mL of ultrapure containing an internal standard of 2,2-dimethylbutyric acid. The homogenate was centrifuged at 12,000 rpm for 10 min at 4 °C. Then, the supernatant was transferred to another tube and mixed with 10 mL of 50% sulfuric acid, 2 mL of analytically pure diethyl ether, and 0.5 g of sodium sulfate (Macklin, China). The mixture was vigorously vortexed for 1 min and then centrifuged at 5000 rpm for 10 min at room temperature. The ether layer was collected for gas chromatography with mass selective detection (5977B GC-MSD system; Agilent Technologies, Santa Clara, CA, USA). An HP-free fatty acid phase capillary column was used for chromatographic separation, with helium as the carrier gas. The oven temperature was increased from 90 to 180 °C at a rate of 15 °C/min. Gas chromatography spectrometry (GC–MS) data were collected and analyzed with MassHunter Workstation software (Agilent Technologies). Final concentrations were calculated based on internal standards and are presented as micromoles per gram of wet feces (μmol/g).

After cognitive assessment, the mice were anesthetized and subjected to cardiac perfusion with saline and fixation with paraformaldehyde (PFA). The ileum (1 cm from the cecum) of each mouse was harvested, dissected in a length of 1 cm, fixed with 4% PFA for 24 h, and then embedded in paraffin. Next, 4-mm-thick sections were cut, dewaxed, and stained with hematoxylin and eosin using standard protocols. For Nissl staining, brain tissue was harvested carefully, fixed with 4% PFA for 24 h, and then cryoprotected with 30% sucrose for 2 days. A 4-mm-thick serial frozen coronal section was cut using a cryostat (Leica CM1950) at − 20 °C. Then, the sections were dried in air, washed twice with distilled water, and stained with 1% toluidine blue for 5 min. The sections were washed 3 times with distilled water, placed in 70% ethanol for 2 min, washed twice with 95% ethanol, and then washed with xylene for 5 min. The sections were observed with a microscope (DM2500 microscope; Leica). Crypt depth were measured from the bottom of the crypt to the crypt-villus junction and villus length were measured from the crypt-villus junction to the tip of the villus.

Coronal brain slices prepared as described above were blocked and incubated with an anti-Iba1 (1:500; Abcam, ab178846) or anti-β-amyloid (Aβ) (1:500; Abcam, ab32136) antibody overnight at 4 °C and then with Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:1000, Life Technologies) for 1 h at room temperature and counterstained with 4ʹ,6-diamidino-2-phenylindole (DAPI). Fluorescence signals were visualized with a laser scanning confocal microscope (Zeiss, Oberkochen, Germany). To measure hippocampal apoptosis, brain sections were subjected to TUNEL staining (Thermo Fisher) and counterstained with DAPI following the manufacturer’s instructions.

Colon and brain tissue samples were snap-frozen in cryotubes submerged in liquid nitrogen. Tissues were ground and lysed radioimmunoprecipitation assay lysis buffer (Beyotime Biotechnology, China) containing the protease inhibitor phenylmethylsulfonyl fluoride (PMSF; Beyotime Biotechnology, China) using a homogenizer and incubated on ice for 30 min. After centrifugation at 12,000 rpm for 30 min at 4 °C, the supernatants were collected. Protein concentration was measured by BCA protein assay kit (Thermo Fisher Scientific, Waltham, MA, USA), and a total of 40 μg protein was separated by 10% SDS–PAGE and subsequently electrophoretically transferred onto a polyvinylidene difluoride membrane (Millipore, Temecula, CA, USA). The membrane was blocked with 5% nonfat milk at room temperature for 1 h and then incubated with primary antibodies against ZO-1 (1:1000; Abcam, ab216880), Occludin (1:1000; Abcam, ab167161), Claudin-4 (1: 1000; Abcam, ab53156), Toll-like receptor 4 (TLR4) (1:1000; Abmart, TA7017S) and β-actin (1:1000; Abcam, ab8226) at 4 °C overnight. The membrane was washed with Tris-buffered saline containing 0.1% Tween 20 and incubated with a horseradish peroxidase-conjugated goat anti-rabbit or goat anti-mouse secondary antibody (1:10,000; Thermo Scientific). The membranes were then visualized with an enhanced chemiluminescence system (Thermo Scientific, Rockford, IL). The band intensity was assessed using Image J software. To measure the expression of each protein, the relative intensity was calculated by comparing the intensity of β-actin, the fold change of protein expression = relative intensity of each protein/mean relative intensity of nPSCI or PBS group.

The data are expressed as the mean ± SD. Statistical significance between the two groups was assessed using Student’s t test or nonparametric Mann–Whitney test. One-way ANOVA followed by the least significant difference (LSD) post hoc test or nonparametric Kruskal–Wallis test was used to compare three groups. The changes in the abundance of Enterobacteriaceae within patients was assessed by Wilcoxon matched-pairs signed rank test. Escape latency was compared by repeated-measure ANOVA. Correlations were analyzed by Spearman’s rank correlation. Multivariate logistic regression analyses were performed, and odds ratio (OR) and 95% confidence interval (CI) were calculated. P < 0.05 was considered statistically significant. Statistical analysis was performed using GraphPad Prism 8 software (GraphPad Software Inc).

In the present study, 83 patients were recruited; 34 of the subjects were PSCI patients (assessed using the MoCA score at 3 months after stroke onset) and 49 of the subjects were nPSCI patients. The demographic data of patients are listed in Additional file 1: Table S1, the PSCI patients were older, had higher proportion of female and smoking; we found no difference in drinking, body mass index, NIHSS score, Barthel index, and history of hypertension, hyperlipidemia, diabetes and stroke, and stroke cause, infarct location and diet types (Additional file 1: Table S1). The PSCI patients had significantly lower MoCA (17.25 ± 3.17 versus 25.44 ± 2.30) and MMSE (24.08 ± 4.03 versus 27.87 ± 1.99) scores than nPSCI patients (Fig. 1a). The PSCI patients had higher levels of peripheral inflammatory factors, such as IL-6, IL-1β, LPS, LBP and DLA, and a higher ESR than the nPSCI patients (Fig. 1b). We analyzed gut microbiota composition at the family level and found a higher abundance of Enterobacteriaceae in PSCI patients (Fig. 1c). This finding was further substantiated by the LEfSe analysis, which identified several taxa that show significant differences in abundance between the two groups (Fig. 1d). Then, we conducted a correlation analysis and discovered that the abundance of Enterobacteriaceae was positively associated with the levels of several inflammatory factors (Fig. 1e). We analyzed the abundance of Enterobacteriaceae at two timepoints, i.e., stroke onset and 3 months after stroke, and found that Enterobacteriaceae abundance tended to decline over time in nPSCI patients but tended to increase over time in PSCI patients (Fig. 1f). In addition, the abundance of Enterobacteriaceae did not differ between the two groups at stroke onset (Additional file 2: Fig. S1a), and PSCI patients had a significantly lower abundance of butyrate-producing bacteria and lower levels of butyrate than nPSCI patients at stroke onset (Additional file 2: Fig. S1b–d). Moreover, multivariate logistic regression analyses revealed that the abundance of Enterobacteriaceae at 3 m after stroke (OR, 1.476; 95% CI 1.089 to 1.690; P = 0.006) and the increase of its abundance is an independent risk factor of PSCI (OR, 1.313; 95% CI 1.069 to 1.612; P = 0.009) after adjusting for sex, age, history of diabetes, hypertension, hyperlipidemia, stroke, smoke and drink, NIHSS score, Barthel index, body mass index, stroke cause, infarct location and diet types (Additional file 1: Table S2).

To validate the role of the gut microbiota in the pathogenesis of PSCI, we transferred gut microbiota from PSCI patients into stroke mice. After acclimatization for 1 week, mice received antibiotics via drinking water for 2 weeks. The α- and β-diversity of the gut microbiota were drastically altered after antibiotic treatment (Additional file 2: Fig S2a, b). Then, the mice were subjected to experimental stroke. The mNSS was assessed 3 days after stroke, and then FMT from PSCI patients (PSCI mice) or nPSCI patients (nPSCI mice) was performed daily for 1 month (Fig. 2a). The PCoA plot of unweighted UniFrac distances revealed that the gut microbiota profile of PSCI mice was clearly separate from that of nPSCI mice (Fig. 2b). The average relative abundances of prevalent microbiota at the family level were used to determine the gut microbiota compositions of the mice (Fig. 2c), and LEfSe showed that the abundance of Enterobacteriaceae was significantly higher in PSCI mice than in nPSCI mice (Fig. 2d). In addition, the PSCI mice had a lower level of fecal butyrate than nPSCI mice (Fig. 2e). Correlation analysis revealed a negative association between the abundance of Enterobacteriaceae and the level of fecal butyrate (Fig. 2f). NaB treatment modulated the PSCI-associated gut microbiota towards the nPSCI-associated gut microbiota (Fig. 2b–d), decreased the abundance of Enterobacteriaceae and increased butyrate level (Fig. 2e).

Three days after experimental stroke, the mNSS was assessed to evaluate stroke injury. mNSS were not significantly different between the two groups (Fig. 3a). After FMT, the Morris water maze was performed to assess the learning and memory of the mice. Mice in each group showed a downward trend in escape latency from day 1 to day 5, with PSCI mice showing worse spatial learning performance than nPSCI mice on days 4 and 5 (Fig. 3b). In addition, PSCI mice spent significantly less time in quadrant III and crossed the platform location fewer times than nPSCI mice (Fig. 3c–e). NaB treatment was efficient to improved cognitive decline of PSCI mice (Fig. 3b–e).

To investigate the underlying mechanism of cognitive decline induced by PSCI-associated gut microbiota, we analyzed the morphology of the intestine. We found that PSCI mice exhibited epithelial disruption with a shorter villus height and crypt depth (Fig. 4a). We also found that the intestinal expression of TLR4, a receptor that recognizes LPS, was significantly higher in PSCI mice (Fig. 4b). Consistently, peripheral levels of LPS and LBP, along with the inflammatory factors IL-6, IL-1β and TNF-α were significantly higher in PSCI mice than in nPSCI mice (Fig. 4c). Furthermore, we found that the expression of tight-junction proteins associated with the blood–brain barrier (BBB), e.g., ZO-1, Occludin and Claudin-4 was significantly decreased in PSCI mice (Fig. 4d). Using immunofluorescence staining, we observed prominent microglial activation in the hippocampi of PSCI mice (Fig. 4e). Nissl staining showed significant neuronal loss and neuronal karyopyknosis and shrinkage of cell bodies in the hippocampal CA1 region in PSCI mice (Fig. 4f). TUNEL staining showed prominent neuronal apoptosis in the hippocampal CA1 region in PSCI mice (Fig. 4g). Notably, stroke mice did not exhibit significant Aβ deposition in the hippocampus (Additional file 2: Fig. S3a) but showed Aβ accumulation in the thalamus (Fig. 4h). Supplementation with NaB rescued these pathological changes in PSCI mice (Fig. 4a–h).

To further demonstrate the role of LPS in the pathogenesis of PSCI, stroke mice were intraperitoneally injected with LPS (LPS mice) or PBS (PBS mice) (Fig. 5a). mNSSs were not different between the two groups 3 days after stroke (Fig. 5b). However, LPS mice showed worse spatial learning performance than PBS mice on days 4 and 5 of the trials (Fig. 5c). Moreover, LPS mice spent less time in quadrant III and crossed the platform fewer times than PBS mice (Fig. 5d–f). LPS mice also exhibited intestinal morphology damage (Fig. 5g) and higher intestinal expression of TLR4 (Fig. 5h) than PBS mice. In addition, the peripheral levels of the inflammatory factors LPS, LBP, IL-6, TNF-α and IL-1β were also higher in LPS mice than in PBS mice (Fig. 5i). As expected, LPS mice exhibited lower expression of tight-junction proteins associated with the BBB and more neuronal loss than PBS mice (Fig. 5j, k). Furthermore, LPS mice showed more microglial activation, and neuronal apoptosis in the CA1 region of the hippocampus and more Aβ deposition in the thalamus (Fig. 5l, m). Taken together, these data suggest that intraperitoneal injection of LPS after stroke causes similar pathology to those seen in mice receiving PSCI-associated gut microbiota.