Dietary fructose-induced gut dysbiosis promotes mouse hippocampal neuroinflammation: a benefit of short-chain fatty acids

A hippocampal neuroinflammatory response, characterized by significant upregulation of interleukin-1 beta (IL-1β), tumor necrosis factor alpha (TNF-α), and IL-6 mRNA levels, was observed in C57BL/6N mice fed a high-fructose diet for eight weeks (Fig. 1a). Immunofluorescence staining showed that the numbers of Iba-1⁺ microglia were significantly increased in the whole hippocampus of high-fructose diet-fed mice compared with standard diet-fed animals (Fig. 1b and Additional file 1: Figure S1a). The numbers of NeuN⁺ neurons and doublecortin (DCX)⁺ newborn neurons were significantly reduced, whereas of GFAP⁺ astrocytes were increased, in the hippocampal dentate gyrus (DG) (Fig. 1b) but not in the cornu ammonis 1 (CA1) and CA3 areas (Additional file 1: Figure S1b) in fructose-fed mice. These results confirm that a high-fructose diet induces hippocampal neuroinflammatory response and neuronal loss in C57BL/6N mice.

To determine whether dietary fructose-induced hippocampal neuroinflammatory response required continuous input from the resident microbes, half of the control and fructose-fed animals were orally treated with broad-spectrum antibiotics (including ampicillin, vancomycin, neomycin, metronidazole, and amphotericin B) during the last four weeks of the experiment. Antibiotic treatment severely reduced microbial abundance and diversity, resulting in a blank microbial composition profile in both the control and fructose-fed groups (Additional file 1: Figure S1c and d). In control mice, antibiotics caused significant reduction in IL-1β and TNF-α mRNA levels and a decreasing trend in IL-6 mRNA levels (Fig. 1a) and the numbers of Iba-1⁺ microglia in the whole hippocampus (Fig. 1b and Additional file 1: Figure S1a). Notably, in fructose-fed mice, antibiotics inhibited the upregulation of hippocampal IL-1β, TNF-α, and IL-6 mRNA levels and the increase in the numbers of Iba-1⁺ microglia (Fig. 1a, b and Additional file 1: Figure S1a). Moreover, antibiotics significantly decreased the numbers of NeuN⁺ neurons and DCX⁺ newborn neurons without affecting the numbers of GFAP⁺ astrocytes in the hippocampal DG in control mice (Fig. 1b). Of note, antibiotics suppressed the increase in the numbers of GFAP⁺ astrocytes but failed to attenuate the fructose-induced decrease in the numbers of NeuN⁺ neurons and DCX⁺ newborn neurons (Fig. 1b) in mice. There were no significant differences in the numbers of GFAP⁺ astrocytes in the hippocampal CA1 and CA3 regions among the groups (Additional file 1: Figure S1b). These results indicate that gut dysbiosis plays an important role in a high-fructose diet-induced hippocampal neuroinflammatory response in C57BL/6N mice.

To assess the influence of dietary fructose on the gut microbiota in mice, the fecal microbiota was analyzed using 16S rRNA gene amplicon sequencing. A high-fructose diet feeding for eight weeks changed the microbial community structure but had no effect on microbial alpha-diversity in mice (Fig. 2a and b, Additional file 1: Figure S2a). The abundance of Bacteroidetes was significantly decreased and Proteobacteria was significantly increased in fructose-fed mice. Moreover, these animals showed an increasing trend in Firmicutes (Fig. 2c). Fructose feeding also led to marked enrichment of the pathogenic bacterial taxa Deferribacteraceae (Mucispirillum) and Helicobacteraceae (Helicobacter) in mice (Additional file 1: Figure S2b). Although main SCFAs-producing bacteria, such as Lachnospiraceae and Ruminococcaceae, were highly enriched in fructose-fed mice (Additional file 1: Figure S2c), the fecal concentrations of acetate, propionate, butyrate, and total SCFAs were significantly lower in fructose-fed mice than in control group (Fig. 2d). Antibiotics reduced the fecal concentrations of acetate, propionate, butyrate, and total SCFAs in both control and fructose-fed animals (Fig. 2d).

A high-fructose diet feeding caused histological changes characterized by thinning of the intestinal mucosa, epithelium, and muscularis mucosae; loss of crypts and glands, edema of the lamina propria; and discrete infiltration of inflammatory cells in mice (Fig. 2e and Additional file 1: Figure S2d). Consistently, serum endotoxin and FITC-dextran levels were increased significantly in fructose-fed mice compared with control animals (Fig. 2f). Electron microscopy also showed short and sparse microvilli, severe mitochondrial swelling, and vacuole change in epithelial cells as well as loss of TJ domains, irregular gaps, and increased distance between adjacent epithelial cells in the colon and distal ileum in fructose-fed mice (Fig. 2g and Additional file 1: Figure S2f). Moreover, colonic mucin 2 (Muc2), occludin, and zonula occludens-1 (ZO-1) protein levels were downregulated significantly in fructose-fed mice (Fig. 2h). After antibiotic treatment, both control and fructose-fed mice showed intestinal histological changes, but intestinal epithelial cells were arranged tightly and in an orderly manner with perfect mitochondrial structure (Fig. 2e, g, Additional file 1: Figure S2d and f). In addition, fructose-fed mice showed a significant decrease in serum endotoxin levels after antibiotic treatment (Additional file 1: Figure S2e), although there was still significantly decreased colonic Muc2, occludin, and ZO-1 expression in the fructose-fed group compared with the control group (Fig. 2h). In control mice, antibiotics caused a significant decrease in colonic Muc2 expression and a trend toward a decrease in colonic occludin and ZO-1 expression (Fig. 2h). Together, these results indicate that a high-fructose diet induces gut dysbiosis, reduces SCFAs levels, and causes intestinal epithelial barrier impairment in C57BL/6N mice.

Next, we tested whether oral administration of SCFAs and a PPAR-γ agonist prevented dietary fructose-induced intestinal epithelial barrier impairment in C57BL/6N mice. Although the microbial structures in fructose-fed mice were still different from those in control mice, to some extent, SCFAs and pioglitazone reduced the enriched diversity in fructose-fed mice (Fig. 3a, b). SCFAs and pioglitazone improved bacterial composition and structure (Fig. 3c and Additional file 1: Figure S3a) and prevented the increase in the abundance of the pathogenic bacterial taxon Deferribacteraceae (Mucispirillum), but the former failed to decrease the abundance of Helicobacteraceae (Helicobacter) (Additional file 1: Figure S3b) in fructose-fed mice. In addition, SCFAs and pioglitazone alleviated the pathological changes in the intestinal mucosa, epithelium, and muscularis mucosae and prevented substructural damage to epithelial cells in fructose-fed mice (Fig. 3d, f, Additional file 1: Figure S3c and d). Consistently, the increase in blood endotoxin levels and the reduction in Muc2, ZO-1, and occludin protein expression in fructose-fed mice were also reversed by SCFAs and pioglitazone (Fig. 3e, g). SCFAs-treated normal mice showed no significant change in intestinal structure or colonic occludin expression (Additional file 1: Figure S3e and f). These results support the notion that SCFAs and PPAR-γ agonist pioglitazone shaped gut dysbiosis and prevented intestinal epithelial barrier impairment in fructose-fed mice.

Notably, the abnormal production of colonic interferon gamma (IFN-γ), TNF-α, and IL-10 further demonstrated a gut immune imbalance in fructose-fed mice, which were ameliorated by SCFAs and pioglitazone (Additional file 1: Figure S4a). Colonic NLRP6 and cleaved caspase-1 P10 levels were significantly reduced, and colonic IL-18 secretion was decreased in fructose-fed mice (Fig. 4a). In the control group, antibiotics significantly decreased NLRP6 and cleaved caspase-1 P10 protein levels but showed no effect on colonic IL-18 secretion (Fig. 4a). In fructose-fed mice, antibiotics reversed the decrease in colonic IL-18 secretion and further suppressed colonic NLRP6 expression (Fig. 4a). SCFAs or pioglitazone also prevented the reduction in colonic NLRP6 levels and caspase-1 activation in fructose-fed mice, resulting in the recovery of IL-18 production (Fig. 4b). In comparison to that in control mice, colonic NLRP3 expression was not altered in fructose-fed mice treated with SCFAs and pioglitazone but was decreased significantly in antibiotic-treated control and fructose-fed mice (Fig. 4a, b). Moreover, colonic expression of PPAR-γ and its target gene (ANGPTL4) was downregulated in fructose-fed mice, which were restored by SCFAs and pioglitazone (Additional file 1: Figure S4b and c).

In the colons of normal mice, oral administration of SCFAs upregulated NLRP6 expression and increased IL-18 production (Additional file 1: Figure S4d). In an ex vivo system of cultured mouse colonic explants (ex vivo colonic explants), SCFAs also increased NLRP6 expression and cleaved caspase-1 P10 protein levels and promoted IL-18 production, and these effects were suppressed by histamine (a bacterial metabolite) (Fig. 4c). Additionally, NLRP6 protein levels and IL-18 production were increased by SCFAs but decreased by NLRP6 siRNAs in murine CT26 colon carcinoma cells (Fig. 4d). However, neither pioglitazone nor GW9662 (a PPAR-γ antagonist) affected basal or SCFAs-stimulated NLRP6 inflammasome activation in the ex vivo colonic explants (Additional file 1: Figure S4e). These results suggest that SCFAs may activate the NLRP6 inflammasome independently of PPAR-γ activation.

We confirmed that oral treatment with SCFAs significantly reduced IL-1β, TNF-α, and IL-6 mRNA levels and the numbers of Iba-1⁺ microglia in the hippocampus of fructose-fed mice (Fig. 5a, b and Additional file 1: Figure S5a). SCFAs clearly increased the numbers of DCX⁺ newborn neurons and NeuN⁺ neurons (Fig. 5b) but could not inhibit the increase in the numbers of GFAP⁺ astrocytes in the DG of the hippocampus in fructose-fed mice (Fig. 5b). More interestingly, SCFAs significantly increased the numbers of DCX⁺ newborn neurons in the DG of the hippocampus in control mice (Additional file 1: Figure S5c). Similarly, pioglitazone significantly decreased IL-1β, TNF-α, and IL-6 mRNA levels and the numbers of Iba-1⁺ microglia in the hippocampus of fructose-fed mice (Fig. 5a, b and Additional file 1: Figure S5a). Moreover, pioglitazone clearly increased the numbers of DCX⁺ newborn neurons and NeuN⁺ neurons in hippocampal DG in fructose-fed mice (Fig. 5b). The numbers of newborn neurons in fructose-fed mice treated with pioglitazone were even greater than that in control mice (Fig. 5b). Pioglitazone also failed to change the increase in the numbers of GFAP⁺ astrocytes in the hippocampus of fructose-fed mice (Fig. 5b). There were no significant differences in the numbers of GFAP⁺ astrocytes in the hippocampal CA1 and CA3 regions among fructose-fed mice treated or not treated with SCFAs and pioglitazone (Additional file 1: Figure S5b). These results suggest that SCFAs and pioglitazone can prevent neuronal injury by inhibiting neuroinflammation in fructose-fed mice.

Four-week-old male C57BL/6N mice (17-20 g) raised under SPF conditions were purchased from the Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China; production license: SCXK (Su) 2016-0003) and allowed to acclimatize to the animal facility environment for a week before being used for experimentation. They were housed in a specific pathogen-free, temperature- and humidity-controlled environment (22 ± 2 °C, 50 ± 5% humidity) with a standard 12 h light/dark cycle. These mice were given access to food and water ad libitum.

The mice were fed a standard diet (control group, 30 mice) or a high-fructose diet (fructose-fed group, 30 mice) with normal drinking water for four weeks firstly before any treatment in experiments I and II. The total calories were 3.4 kcal/g for the standard diet (rodent diet, XT002, Jiangsu Synergy Pharmaceutical Biological Engineering Co., Ltd., Nanjing, China) and 3.7 kcal/g for a high-fructose diet (35% fructose-derived calories). The high-fructose diet was made with standard mouse chow by adding approximately 30% fructose according to a modified formula based on Choi’s report [56]. Starting from the 4th week, the control and fructose-fed groups in experiments I and II were treated in a different way. In experiment I, half of the control and fructose-fed mice were treated orally with broad-spectrum antibiotics for the last four weeks according to Gacias’ report [57]. The mice were given ampicillin (1 g/L) in drinking water and a cocktail of vancomycin (50 mg/kg), neomycin (100 mg/kg), metronidazole (100 mg/kg), and amphotericin B (1 mg/kg) by gavage once daily. All antibiotics were obtained from Solarbio Science & Technology Co., Ltd. (Beijing, China). In experiment II, fructose-fed mice were divided into three subgroups (15 mice/group) in the fourth week. Two groups were fed the high-fructose diet and simultaneously treated with SCFAs or pioglitazone (a PPAR-γ agonist) for the last eight weeks according to our preliminary study, respectively. Another group continued to be fed the high-fructose diet for the last eight weeks. An admixture (3:1:1 ratio) of sodium acetate (S2889; Sigma-Aldrich Co., Ltd, Shanghai, China), sodium propionate (P1880; Sigma-Aldrich), and sodium butyrate (303410; Sigma-Aldrich) was incorporated into a high-fructose diet at a proportion of 5% (wt/wt). A dose of 30 mg/kg pioglitazone was administered orally to the mice by gavage once a day. Body weight was recorded once a week throughout the experiment.

Intestinal permeability was assessed with an in vivo FITC-dextran (FD4; Sigma-Aldrich) permeability assay, as described previously [21]. Mice fasted for 4 h were gavaged with 0.6 mg/g body weight FITC-dextran (4 kDa) in a 25 mg/mL solution, and blood was collected by submandibular bleeding after 3 h. The fluorescence intensity in serum was measured using a fluorescence spectrophotometer (Synergy 2, Biotek, Winooski, VT). FITC-dextran concentrations were determined from a standard curve generated with serial dilutions of FITC-dextran.

At the end of the experiments, the mice were fasted overnight and anesthetized with sodium pentobarbital (50 mg/kg, i.p.), and blood was obtained by cardiopuncture. These blood samples were centrifuged at 2500 rpm for 10 min to collect the serum samples, which were immediately frozen at − 80 °C for biochemical assays. Hippocampus and colon tissues were dissected for histopathology, transmission electron microscopy (TEM) analysis, immunofluorescence staining, quantitative polymerase chain reaction (qPCR), and Western blot analysis.

Fresh stool samples from mice were collected in disinfected tubes, immediately frozen in liquid nitrogen upon collection, and stored at − 80 °C until analysis. Stool DNA samples were isolated, and the microbiomes were analyzed at the Beijing Genomics Institute on an Illumina MiSeq platform. High-quality reads for bioinformatics analysis were selected, and all of the valid reads from all samples were clustered into operational taxonomic units (OTUs) based on 97% sequence similarity. α-Diversity was calculated based on the Chao1 diversity index. The variation between the experimental groups (β-diversity) was assessed with principal coordinate analysis (PCoA) plots. Linear discriminant analysis coupled with effect size (LEfSe) was performed with LEFSE software.

SCFAs in the frozen fecal samples were determined by gas chromatography. The method used was adapted from that of Frost et al. [58]. Briefly, 150–200 mg of feces was combined with 550 μL of PBS. These samples were vortexed thoroughly and centrifuged at 7200 rpm for 10 min, and the supernatant was collected. The SCFAs were extracted through the addition of 250 μL of concentrated hydrochloric acid and 1 mL of diethyl ether followed by vortex mixing for 1 min. The samples were centrifuged at 7200 rpm for 10 min after 12 h, and the ether layer was transferred to a separate capped vial. Gas chromatography was performed on an HP 6890 Plus gas chromatograph equipped with a flame ionization detector, split/splitless injector, and an HP-INNOWax column (30 m length × 0.25 mm inner diameter, 0.25 μm film thickness; Agilent Technologies, USA). The injector and detector temperatures were 300 °C, and the column temperature was programmed to increase from 80 °C for 1 min to 150 °C at 5 °C per min. Nitrogen served as the carrier gas, and injections (1 μL) were made in the split mode (10:1 split). The peak areas were recorded, and all subsequent data manipulation was completed using ChemStation Software (Agilent Technologies). Reference standards of acetic acid (A801295; Macklin), propionic acid (P816183; Macklin), and butyric acid (B802731; Macklin) were dissolved in diethyl ether to make the mixed stock solution. The concentrations of SCFAs were calculated by the standard curve method. The reported values were normalized according to the wet weight of the original fecal sample used.

An ex vivo system of cultured mouse colonic explants was constructed according to the methods in a previous report [13] (Additional file 2). Colons from eight-week-old normal C57BL/6N mice were longitudinally opened and washed thoroughly by flushing several times with cold HBSS containing 100 U/mL penicillin, 100 μg/mL streptomycin, and 50 μg/mL metronidazole. The colon tissues were cut into approximately 4-mm² fragments, placed on presoaked Gelfoam rafts, and cultured in 1 mL of RPMI 1640 medium containing 0.01% BSA, 200 U/mL penicillin, 200 μg/mL streptomycin, and 1% Fungizone. The colonic explants were incubated in the presence of a mixture (3:1:1 ratio) of sodium acetate, sodium propionate, and sodium butyrate with histamine (H7125; Sigma-Aldrich), pioglitazone, or GW9662 (a PPAR-γ antagonist) for 24 h according to the experimental design. The culture medium and the colonic explants were collected to test the activation of the NLRP6 inflammasome and the transcription of PPAR-γ.

Murine CT26 colon carcinoma cells were purchased from the Shanghai Institutes for Biological Sciences (Shanghai, China) and grown in 1640 medium supplemented with 10% fetal bovine serum in a humidified atmosphere containing 5% CO₂ at 37 °C. For the experiments, cells were plated in 6-well plates for 12 h and transfected with NLRP6 siRNA or negative control siRNA (GenePharma, Shanghai, China) using Lipofectamine 2000 for 6 h. These cells were then incubated in 1640 medium in the presence or absence of a 12.5 mM mixture (3:1:1 ratio) of sodium acetate, sodium propionate, and sodium butyrate for 48 h. The culture medium was collected, and the total cellular protein was extracted for biochemical and Western blot analysis.

Hippocampus and colon tissues were isolated from the experimental mice and homogenized in TRIzol Reagent (Invitrogen), and total RNA was isolated according to the manufacturer’s instructions. RNA (1 μg) was reverse transcribed into complementary DNA using HiScript II Select qRT SuperMix (Vazyme, Nanjing, China). Quantitative polymerase chain reaction was performed using gene-specific primer sets and SYBR Green (Vazyme, Nanjing, China) on a real-time PCR detection system (Bio-Rad). The primer sequences used for amplification were as follows: mTNF-αF (5′-CCC CTT TAT TGT CTA CTC CTC-3′), mTNF-αR (5′-CCC AGC ATC TTG TG TTT C-3′), mIL-1βF (5′-ATT GTG GCT GTG GAG AAG-3′), mIL-1βR (5′-AAG ATG AAG GAA AAG AAG GTG-3′), mIL-6F (5′-GCC TTC CCT ACT TCA CAA-3′), mIL-6R (5′-ACA ACT CTT TTC TCA TTT CCA C-3′), mPPAR-γF (5′-GTC TTG GAT GTC CTC GAT GGG-3′), mPPAR-γR (5′-TTA TGG AGC CTA AGT TTG AGT TTG C-3′), mAngplt4F (5′-ATC TCC GAA GCC ATC CTT GTA-3′), mAngplt4R (5′-CTC TGG GGT CTC CAC CAT TTT-3′), mβ-actinF (5′-CTC TCC CTC ACG CCA TC-3′) and mβ-actinR (5′-ACG CAC GAT TTC CCT CTC-3′). All the primers were provided by GENEray Biotechnology (Shanghai, China). The reaction conditions were 94 °C for 30 s followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s. Relative expression was calculated using the △△Ct method with β-actin serving as the reference housekeeping gene. The expression of each target gene was normalized to β-actin expression, and the normalized data were presented as the fold change in gene expression in treated mice compared with control mice.

Colon tissues from experimental mice or cultured ex vivo colonic explants from normal mice were homogenized in ice-cold RIPA buffer containing protease inhibitors. The cleared lysates were obtained by centrifugation at 10,000 rpm at 4 °C for 15 min. Protein quantification was carried out using a bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific, USA) with bovine serum albumin as a standard. Equivalent amounts of protein from each sample were separated by 10% SDS–PAGE and transferred onto polyvinylidene fluoride membranes (Millipore, USA). Subsequently, the membranes were blocked in 5% milk, probed overnight at 4 °C with primary antibodies and then incubated with HRP-conjugated secondary antibodies. The primary antibodies included rabbit anti-Muc2 (ab76774, Abcam), rabbit anti-occludin (ab168986, Abcam), rabbit anti-ZO-1 (H-300, Santa Cruz), rabbit anti-NLRP6 (PA5-21022, Thermo Fisher Scientific), rabbit anti-caspase-1 (sc-514, Santa Cruz), rabbit anti-NLRP3 (#15101, CST), rabbit anti-PPAR-γ (#2435, CST), and mouse anti-β-actin (#4970, CST). The signals were detected with an enhanced chemiluminescence system (Tanon, Shanghai, China). The immunoreactive bands were quantified via densitometry using ImageJ (Version 1.50b, National Institutes of Health, USA) and standardized to β-actin and were expressed as fold changes relative to the control value.

Excised ileum and colon tissues were fixed in 4% paraformaldehyde and embedded in paraffin. The blocks were serially cut into 5-μm-thick sections and stained with hematoxylin and eosin (H&E). Histological images were obtained using an Olympus IX53 microscope (Tokyo, Japan), an Olympus DP73 digital camera, and Olympus cellSens imaging software.

Excised ileum and colon tissues were extensively washed of fecal matter and fixed in 2.5% glutaraldehyde in PBS. For TEM, the tissue samples were processed with standard protocols at the Analysis Center of Nanjing Medical University (Nanjing, China) and examined using a JEOL JEM-1010 transmission electron microscope (Tokyo, Japan).

Three animals from each group were anesthetized and transcardially perfused with ice-cold 4% paraformaldehyde. The brains were excised, fixed in 4% paraformaldehyde overnight, incubated in 20% sucrose/PBS overnight at 4 °C, and incubated in 30% sucrose/PBS overnight at 4 °C. Then, the brain samples were frozen in optimum cutting temperature compound (Sakura Finetek, USA) and coronally cut with a cryostat into 30-μm-thick sections. The frozen sections were blocked with 10% fetal bovine serum in PBS for 1 h at room temperature. The slides were incubated with primary antibodies overnight at 4 °C and then incubated for 30 min at 37 °C with Alexa Fluor-conjugated secondary antibodies (Invitrogen). The primary antibodies included mouse anti-Iba1 (sc-32725, Santa Cruz), mouse anti-NeuN (ab104224; Abcam), rabbit anti-DCX (ab77450; Abcam), and rabbit anti-GFAP (Z0334; Dako). The sections were counterstained with 4,6-diamidino-2-phenylindole (DAPI) for nuclear staining. Images were captured using a Leica TCS SP8 confocal microscope. Six randomly selected brain sections from each animal were used to outline and count immunofluorescence-positive cells in the hippocampal DG, CA1, and CA3 regions with Image-Pro Plus software (Media Cybernetics, USA).

Data are expressed as the mean ± SEM. Statistical analyses were performed with GraphPad Prism Software 6.01 (GraphPad Software, San Diego, USA). Differences were analyzed by one-way analysis of variance (ANOVA) with post hoc tests for multiple group comparisons. p values < 0.05 were considered to indicate significance.

All 16S rRNA gene sequencing reads data has been deposited to the National Center for Biotechnology Information’s Sequence Read Archive under accession number PRJNA540110.