Ciprofloxacin changes gut bacteria and may increase risk of seizures through gut-brain communication

Sprague-Dawley rats (5–6 weeks old, 180 ± 20 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. The Institutional Ethical Committee for Animal Welfare of the Seventh Affiliated Hospital of Sun Yat-sen University(approval No. SYSU-IACUC-2022-B1339) approved the study. Rats were housed in a temperature-controlled room with a 12-h light/dark cycle under specific pathogen–free (SPF) conditions, with free access to food and water.

Rats were randomly assigned to two groups: Control (CTRL, n = 22) and Ciprofloxacin-treated (CPF, n = 42). Among the CTRL rats, 8 were designated as microbiota donors (CTRL-PBS, n = 8). To induce gut microbiota dysbiosis, CPF rats received ciprofloxacin monohydrochloride (TargetMol, USA; 10 mg/mL in sterile saline, Shuanghe Pharmaceutical Co., China) at 100 mg/kg via oral gavage for 14 consecutive days. CTRL and CTRL-PBS rats were gavaged with equivalent volumes of sterile saline. From Days 15 to 28, CTRL-PBS rats received PBS (Meilunbio, China) by oral gavage every other day and served as healthy microbiota donors for FMT in CPF-FMT rats.

After the 14-day ciprofloxacin treatment, 8 rats each from the CPF and CTRL groups underwent PTZ-induced seizure testing, while the remaining 6 rats per group were sacrificed for molecular and histological analyses of colonic and cortical tissues. The 28 remaining CPF rats were divided into two subgroups: CPF-FMT (n = 14) and CPF-PBS (n = 14). CPF-FMT rats received FMT every other day from Days 15 to 28 using Fecal microbial solution from CTRL-PBS donors, while CPF-PBS rats received PBS as a placebo control. After the intervention, 8 rats from each subgroup underwent PTZ-induced seizure testing, and the remaining 6 were sacrificed for analyses of the integrity of the colon and blood-brain barrier and neuroinflammatory markers in cortical tissues.

Gut microbiota dysbiosis was induced by oral gavage of ciprofloxacin monohydrochloride (TargetMol, USA; 10 mg/mL dissolved in sterile saline) at a dosage of 100 mg/kg body weight once daily for 14 consecutive days. Control rats were administered an equivalent volume of sterile saline (10 mL/kg) via oral gavage over the same period.

FMT was performed to reconstitute gut microbiota composition (28, 29). Fresh feces from healthy donor rats (CTRL-PBS group) were suspended in PBS (1 g feces/5 mL PBS), centrifuged at 2500 rpm for 10 min at 4 °C, and filtered through a nylon mesh (40-μm) to prepare the microbial suspension. CPF-FMT rats received 10 mL/kg of this suspension daily via oral gavage for 14 days, starting on Day 15 after ciprofloxacin treatment. CPF-PBS rats were gavaged with an equivalent volume of PBS as a control.

On Day 14 and Day 28 of the experiment, rats were implanted with cranial electrodes and allowed to recover for two days. And then PTZ (45 mg/kg) was administered intraperitoneally to induce seizures. Seizure activity was monitored using synchronized video-electroencephalography (VEEG) with LabChart software. Baseline EEG was recorded for 10 minutes before PTZ injection, followed by continuous recording for 90 minutes post-injection. Seizure severity was evaluated using the modified Racine scale: Score 0: No response; Score I: Rhythmic facial or mouth movements; Score II: Head nodding or tail jerking; Score III: Clonic movements of one limb; Score IV: Bilateral limb clonic movements or tonic extension; Score V: Loss of balance and falling.

Fecal samples were gathered on Day 0 (baseline), Day 14 (post ciprofloxacin), and Day 28 (post FMT or PBS treatment). 16S rRNA sequencing was conducted by Guangdong Magigene Biotechnology Co., Ltd.

Histological evaluation of colonic tissues was conducted using hematoxylin and eosin (H&E) staining. Immunohistochemical staining was performed to detect tight junction proteins, including ZO-1 (1:2000, 21773-1-AP, Proteintech) and Occludin (1:8000, 27260-1-AP, Proteintech), in colonic and brain tissues. Additionally, glial fibrillary acidic protein (GFAP; 1:2000, 80788, Cell Signaling Technology) staining was performed on brain tissues. Quantification of staining was carried out using 3–5 representative fields per sample.

Tight junction proteins (ZO-1, Occludin, and Claudin-5) in intestinal and brain tissues were assessed using Western blotting. Tissues were homogenized in RIPA lysis buffer (P0013B, Beyotime, China) with the addition of protease and phosphatase inhibitors (P1005, Beyotime, China; P1045, Beyotime, China). Enhanced BCA Protein Assay Kit (P0010S, Beyotime, China) was used to determine protein concentrations. Equal quantities of protein were separated via SDS-PAGE, transferred to NC membranes, and then blocked. Membranes were incubated with primary antibodies against ZO-1 (1:5000, 21773-1-AP, Proteintech), Occludin (1:8000, 27260-1-AP, Proteintech), Claudin-5 (1:5000, 29767-1-AP, Proteintech), and GAPDH (1:10000, 60004-1-Ig, Proteintech), followed by HRP-conjugated secondary antibodies (anti-rabbit IgG, 1:1000, A0208, Beyotime; anti-mouse IgG, 1:5000, ab205719, Abcam). Signals were visualized using chemiluminescence (Tanon 4800 Multi, China).

Cortical tissues from the CPF and CTRL groups (n = 3 per group) were subjected to RNA sequencing, performed by Lianchuan Biotechnology Co. (Hangzhou, China). Total RNA was extracted, and transcriptomic libraries were prepared and sequenced. The R language was used to identify differentially expressed genes (DEGs) and conduct the KEGG and GO enrichment analysis, with significance thresholds set at P < 0.05 and |log2FC| > 2 for differential expression analysis. PPI network was established using STRING website, Cytoscape software was used to identify hub genes, and NetworkAnalyst was used for transcription factor prediction and interaction network construction. Immunity-related genes were obtained from The Immunology Database and Analysis Portal (ImmPort).

Microglial activation was evaluated using immunofluorescence co-staining of Iba1 with CD14 and CD68. Cortical sections were applied with primary antibodies against Iba1 (1:100, ab283319, Abcam), CD14 (1:100, 17000-1-AP, Proteintech), and CD68 (1:100, 28058-1-AP, Proteintech), Fluorescent dye-conjugated secondary antibodies (CoraLite488-conjugated Goat Anti-Rabbit IgG (H+L) were used in the next step, 1:1000, SA00013-2, Proteintech; Alexa Fluor^® 594-conjugated Goat Anti-Mouse IgG (H+L) F(ab’)₂ Fragment, 1:1000, 8890, Cell Signaling Technology). Fluorescent images were captured using Mshot MF43-N. The number and percentage of Iba1+CD14+ and Iba1+CD68+ cells in cortical tissues were evaluated.

Serum concentrations of lipopolysaccharide-binding protein (LBP), interleukin-6 (IL-6), and tumor necrosis factor-alpha (TNF-α) were determined in serum and cortex, using enzyme-linked immunosorbent assay (ELISA) kits(Cusabio, China). Due to the rapid in vivo fluctuations of lipopolysaccharide (LPS) and the technical challenges of its direct measurement, we used LBP as a surrogate marker to reflect LPS levels and immune burden. LBP is a recognized biomarker of LPS exposure and systemic inflammation (30).

Data were analyzed with SPSS software (version 25.0; IBM, Somers, NY, USA). Continuous data were expressed as mean ± SD, and categorical variables were summarized as frequencies or percentages. Normally distributed numerical data were performed Student’s t-test, non-normally distributed data were analyzed by the Mann–Whitney U test. Kaplan–Meier survival curves were constructed to evaluate seizure probability. The data analysis of 16S rRNA sequencing was conducted using the Magigene cloud platform (http://cloud.magigene.com↗). P < 0.05 was considered statistically significant.

An overview of the experimental design is shown in Figure 1A. Kaplan-Meier survival analysis revealed that seizure latency was significantly shorter in the CPF group than in the CTRL group (1.13 min vs. 35.35 min, Log-rank P = 0.028; Figure 1B). This indicates that gut microbiota dysbiosis increased seizure susceptibility. Similarly, the CPF group exhibited a higher median Racine stage (4.50; IQR: 3.25–5.00) than the CTRL group (3.50; IQR: 0.75–4.00), not significant (P = 0.08; Figure 1D). However, the duration of the first seizure showed no significant difference between these two groups (P = 0.792; Figure 1C). Additionally, the latency to the first spike was significantly shorter in the CPF group (0.99 min vs. 1.71 min, P = 0.012; Figure 1E). These results suggest that gut microbiota dysbiosis alters seizure severity and onset.

To explore the impact of ciprofloxacin on gut microbiota and its potential relevance to seizure susceptibility, we analyzed fecal samples collected at baseline (Day 0) and Day 14 using 16S rRNA sequencing. At baseline, the gut microbiota alpha and beta diversity did not differ significantly between the CPF and CTRL groups. (P > 0.05, Supplementary Figures 1A, B). The microbial community composition at the genus level was nearly identical between the CPF.0d and CTRL.0d groups, indicating consistent baseline levels before the experiment (Supplementary Figure 1F). However, after 14 days of ciprofloxacin administration, alpha diversity was markedly decreased in the CPF group relative to the CTRL group (P = 0.001, Figure 1F). NMDS analysis indicated a unique gut microbiota clustering pattern in the CPF group compared to the CTRL group (Figure 1G), indicating substantial microbiota disruption. Significant differences in microbial composition at the genus level were observed between the CTRL.14d and CPF.14d groups, reflecting the impact of ciprofloxacin on gut microbiota. Ciprofloxacin treatment significantly altered the abundance of key gut microbial taxa (Figures 1H, I). Akkermansia and Bacteroides were notably enriched in the CPF group, whereas several taxa, including f_Muribaculaceae, f_Prevotellaceae, Lachnospiraceae_NK4A136_group, o_Clostridia_UCG-014, and Parasutterella, Parabacteroides, were markedly depleted. The reduction of these protective taxa highlights a potential link between microbiota dysbiosis and increased seizure susceptibility.

To explore the mechanism of gut microbiota dysbiosis affecting seizure susceptibility through central inflammatory and neuromodulatory mechanisms, we performed RNA sequencing of the cerebral cortex transcriptome of the CTRL and CPF groups. 1035 differentially expressed genes (DEGs, P < 0.05) were identified, of which 138 genes met the higher significance criterion (P < 0.05, |log2FC| > 2) (Figure 2A). Among the differentially expressed genes, LBP encodes a protein that binds to the bacterial outer membrane antigen LPS. Compared to the CTRL group, LBP was significantly upregulated in the CPF group. Additionally, the downstream inflammatory cytokine IL-6 exhibited increased expression in the CPF group. KEGG analysis revealed significant enrichment of these genes in the PI3K-Akt signaling pathway, cytokine-cytokine receptor interaction, ECM-receptor interaction, and Toll-like receptor (TLR) signaling pathways (Figures 2B–D). Heat map and GSEA analysis further revealed that pro-inflammation-related genes (including LBP, CD14, IL-6) were significantly up-regulated in the CPF group (Figures 2E, F), supporting the activation of central inflammation. In addition, cross-matching the DEGs with the ImmPort database revealed 93 overlapping genes (Figure 2G), which were significantly enriched in the MAPK signaling pathway (Figure 2H), suggesting that the TLR pathway may contribute to increased seizure susceptibility induced by gut microbiota dysbiosis through activation of downstream MAPK signaling. We created a heatmap showing the correlation between inflammation-related genes and gut microbiota (Figure 2I). The results showed that f_Prevotellaceae, Lachnospiraceae_NK4A136_group, Lactobacillus, Parasutterella, o_Clostridia_UCG-014 and Parabacteroides were associated with LBP, CD14, Casp12, and IL-6 gene expression negatively, and positively with CCL28, a gene that regulates mucosal immunity. In contrast, Bacteroides, ASF356, Eisenbergiella, Lachnoclostridium, and Lachnospiraceae_UCG-001 were positively correlated with LBP, CD14, Casp12, and IL-6 gene expression and negatively correlated with CCL28, and the gut microbiota may be involved in immune regulation by influencing the inflammatory response and participate in immune regulation. In addition, the PPI network of 93 inflammation-related genes (Figure 2J) showed that 10 hub genes (Il6, Mmp9, Icam1, Pparg, Thbs1, Pdgfrb, Cdh1, Edn1, Ccn2, Fgf17) were at the core of the network. Combined with the differentially expressed ion channel genes (SCN8A, KCNJ13, KCNV2, RYR1, GABRB2, CHRNA6, GJB1, CACNA1D, ANO6, KCNJ14, GRIN2B, RYR2, CLIC1, CLIC2, GABRP, PKD2, KCNMA1, BEST1) from transcriptome sequencing, we performed transcription factor prediction and interactions network analysis through the JASPAR database using the NetworkAnalyst tool. The results revealed that these genes shared 39 common transcription factors (Figure 2K), including key factors such as NF-κB1, STAT3, and CREB1, which are closely related to the inflammatory pathway and the regulation of neural excitatory/inhibitory balance.

To evaluate the impact of gut microbiota dysbiosis on barrier integrity and to determine the restorative potential of FMT, morphological and molecular changes in colonic and brain tissues were analyzed across CPF group and CTRL group.

Histological analysis of colonic tissue (HE staining) revealed significant morphological disruptions in CPF-treated rats (Figure 3A). Compared to the CTRL group, the CPF group displayed wider inter-crypts distance (P = 0.013, Figure 3B), reduced crypt depth (P = 0.018, Figure 3C). Inflammatory cell infiltration in the mucosa and submucosa and villus architectural damage were also observed in the CPF group (Supplementary Figure 2A). These morphological alterations indicate severe intestinal barrier damage caused by gut microbiota dysbiosis. Further analysis of tight junction proteins (TJPs) expression confirmed the loss of intestinal barrier integrity in the CPF group. Western blot (Figures 3D–G) and immunohistochemical staining (Figures 3H, I) showed significantly reduced ZO-1 levels in the CPF group compared to the CTRL group (WB P = 0.042, IHC P<0.0001). These findings demonstrate that gut microbiota dysbiosis compromises intestinal barrier function. Similarly, gut microbiota dysbiosis impaired the BBB. ZO-1 was reduced in the CPF group compared to the CTRL group (P = 0.017, Figures 3J–M). Immunohistochemical staining further confirmed the loss of BBB integrity, showing decreased ZO-1 expression (P<0.0001, Figures 3N, O). These findings imply that gut microbiota dysbiosis leads to greater BBB permeability.

To assess systemic inflammation, we measured levels of LBP, IL-6, and TNF-α both in serum and cortex. In the CPF group, these inflammatory markers were elevated compared to the CTRL group (P = 0.031, P = 0.004, P = 0.036, respectively; Figures 4A–C). Compared to the CTRL group, the levels of LBP, IL-6, and TNF-α were significantly elevated in the cerebral cortex of CPF group rats (P < 0.001, P < 0.0001, and P = 0.014, respectively; Figures 4D–F).

To investigate whether gut microbiota dysbiosis affects glial cell activation in the brain, we performed immunohistochemical staining of Iba1-labeled microglia (Supplementary Figure 3A) and immunofluorescence co-staining of Iba1 with CD14 (Figure 5A), as well as Iba1 with CD68 (Figure 5D), in cortical tissues. Immunohistochemical staining of Iba1-labeled microglia revealed a statistically significant increase in the number of cortical microglia in the CPF group compared to the CTRL group (Supplementary Figure 3B). Iba1 and CD14 co-staining showed a marked increase in Iba1+CD14+ microglia in the CPF group relative to the CTRL group (P = 0.0001; Figure 5B). Additionally, the proportion of Iba1+CD14+/Iba1+ cells was significantly more abundant in the CPF group (P = 0.0001; Figure 5C). Co-staining with Iba1 and CD68, a marker of activated microglia, showed a significant increase in the number of Iba1+CD68+ microglia in the CPF group compared to the CTRL group (P < 0.001; Figure 5E). The proportion of Iba1+CD68+/Iba1+ cells was also elevated in the CPF group (P = 0.002; Figure 5F).

We also evaluated astrocytic activation using GFAP immunohistochemistry(Figure 5G). In the CPF group, significant astrocytic proliferation was observed in the cortex, hippocampal CA3 and hippocampal DG region compared to the CTRL group (P < 0.0001, P = 0.0003, P = 0.0048, respectively; Figures 5H, J, K). In the hippocampal CA1 regions, GFAP staining intensity showed no significant variation between the CPF and CTRL groups (Figure 5I).

To validate the contribution of gut microbiota dysbiosis to increased seizure susceptibility, FMT was employed to modulate the gut microbiota in CPF group rats. An overview of the experimental design is shown in Figure 6A.

We compared seizure characteristics between CPF-FMT and CPF-PBS groups. Kaplan-Meier survival analysis demonstrated a significantly longer seizure latency in the CPF-FMT group (50.50 min vs. 1.28 min, Log-rank P = 0.037; Figure 6B). Moreover, the CPF-FMT group showed a shorter duration of first seizures (0.37 min vs. 1.48 min, P = 0.002; Figure 6C) and a lower median Racine stage (3.50; IQR: 0–4.00) than the CPF-PBS group (5.00; IQR: 4.25–5.00; P = 0.005; Figure 6D). Notably, the latency to the first spike was significantly longer in the CPF-FMT group (3.40 min vs. 1.11 min, P = 0.002; Figure 6E), indicating partial restoration of gut microbiota may alleviate seizure susceptibility.

On Day 28, the CPF-FMT group exhibited alpha and beta diversity fully restored to levels equivalent to those of the donor group (CTRL-PBS) (P > 0.05, Supplementary Figure 1C, D). The genus-level microbial composition in the CPF-FMT group closely resembled that of the CTRL-PBS group, confirming successful colonization following FMT (Figure 6H). The CPF-FMT group showed significantly higher alpha diversity than the CPF-PBS group (P<0.001, Figure 6F). NMDS analysis showed distinct clustering between CPF-FMT and CPF-PBS groups (Figure 6G), and significant differences in genus-level microbial composition persisted between the CPF-FMT and CPF-PBS groups, demonstrating that gut microbiota dysbiosis could not fully recover without FMT intervention (Figure 6H and Supplementary Figure 1E). Importantly, FMT reversed the depletion of f_Muribaculaceae, f_Prevotellaceae, Lachnospiraceae_NK4A136_group, o_Clostridia_UCG-014, and Parasutterella compared to the CPF-PBS group. Although the relative abundance of Bacteroides increased in the CPF-FMT group, no significant difference was identified compared to the CPF-PBS group (Figures 6H, I). These changes suggest partial restoration of gut microbial composition following FMT.

In the CPF-FMT group, inter-crypts distance and crypt depth were restored to near normal levels (P = 0.010, P<0.0001, Figures 7A–C). However, CPF-PBS rats exhibited persistent disruption of tight junction proteins, including reduced Occludin expression, as evidenced by Western blot and immunohistochemical analysis (P = 0.049, P<0.0001, Figures 7D–I). FMT effectively restored BBB function. In the CPF-FMT group, TJPs showed significantly higher expression levels relative to the CPF-PBS group (Figures 7J–M). Prolonged gut microbiota dysbiosis in the CPF-PBS group led to further reductions in Occludin and Claudin-5 expression (P<0.0001, P<0.0001, respectively), highlighting the necessity of gut microbiota restoration for BBB integrity. Immunohistochemical staining further confirmed the loss of BBB integrity, showing decreased Occludin expression (P = 0.014, Figures 7N, O).

FMT reduced systemic inflammation in CPF-treated rats. Serum LBP and IL-6 levels were markedly lower in the CPF-FMT group relative to the CPF-PBS group (P = 0.021, P = 0.002, respectively; Figures 8A, B), while TNF-α levels showed a decreasing trend but no significant difference between CPF-FMT and CPF-PBS groups (P = 0.419, Figure 8C). Compared to the CPF-PBS group, the levels of LBP, IL-6, and TNF-α were significantly lower in the cerebral cortex of CPF-FMT group rats (P = 0.012, P = 0.044 and P = 0.012, respectively; Figures 8D–F).

Immunohistochemical staining of Iba1-labeled microglia revealed a statistically significant decrease in the number of cortical microglia in the CPF-FMT group compared to the CPF-PBS group (Supplementary Figure 4A, B). CPF-FMT rats exhibited reduced numbers and proportions of Iba1+CD14+ microglia compared to the CPF-PBS group (P = 0.002, P < 0.001, respectively; Figures 9A–C), as well as numbers and proportions of Iba1+CD68+ microglia compared to CPF-PBS rats (P = 0.005, P = 0.0001, respectively; Figures 9D–F). CPF-FMT rats exhibited reduced astrocytic proliferation compared to the CPF-PBS group in both cortex and hippocampal CA3 regions (P < 0.001, P < 0.0001, respectively; Figures 9G, H, J). While FMT reduced astrocytic activation in the CA1 and DG region compared to the CPF-PBS group, the difference was not statistically significant (P = 0.169, P = 0.112; Figures 9I, K).

The RNA-seq data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject (https://dataview.ncbi.nlm.nih.gov/object/PRJNA1438453?reviewer=if7bh4b6cudj8ah7jpbpju09rc↗) accession number PRJNA1438453.