Gut dysbiosis leads to cognitive decline through CNTF-mediated activation of microglia in mice

Abstract

Background

The gut microbiota is essential for maintaining host homeostasis through its influence on metabolism, immunity, and neural signalling. Disruption of this microbial balance, known as gut dysbiosis, can alter gut-brain communication and has been associated with cognitive decline, with impairments in learning and memory. However, the cellular and molecular factors that lead to cognitive decline are not well understood. In this study we used an antibiotic-induced gut dysbiosis model.

Results

We observed that the animals with antibiotic-induced gut dysbiosis showed deficits in cognition, especially long-term memory consolidation. There was an increase in astrocytes and microglial activation in the CA1 subregion of the hippocampus. The microglia were observed to engage in synaptic pruning at the presynaptic terminals. This aberrant pruning might have disrupted synaptic plasticity and connectivity, contributing to the observed cognitive deficiency. CNTF was also observed to be elevated along with activation of the JAK/STAT3 pathway. CNTF can activate microglia. Our findings revealed that astrocytes, microglia, and CNTF form an inflammatory activation loop within the CA1 region of the hippocampus following antibiotic-induced gut dysbiosis.

Conclusions

In summary, our study demonstrates that antibiotic-induced gut dysbiosis triggers a cascade of neuroinflammatory events in the hippocampus, involving the elevation of CNTF, microglial pruning at presynaptic terminals, and reciprocal activation of glial cells, resulting in cognitive deficits. These findings highlight the critical role of gut-brain communication in maintaining neural homeostasis and identify CNTF as a potential therapeutic target for dysbiosis-associated cognitive disorders.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-025-02454-x.

Background

The gut microbiota encompasses a vast and diverse community of trillions of microorganisms residing in the gastrointestinal tract. This microbial consortium plays a crucial role in regulating gastrointestinal physiology and influences the functioning of distant organ systems [1]. Notably, the gut and brain engage in a bidirectional communication process, whereby disturbances in the function of either organ can significantly impact the other, a phenomenon collectively termed the gut-brain axis (GBA) [2–4]. Dysbiosis, defined as an imbalance in the gut microbiota (often induced by external factors), has been associated with various neurological disorders, including the exacerbation of mental health conditions, cognitive dysfunction, and developmental disabilities [5, 6]. While the link between the gut microbiota and brain function is well recognised, the specific mechanisms of communication remain elusive and are only beginning to be elucidated. Establishing robust models of gut dysbiosis is particularly challenging due to the host organism's complex biological mechanisms that work to restore homeostasis, which can complicate the detection and analysis of specific effects resulting from microbial disruptions. Continued research is essential to elucidate these interactions and to understand their implications for neurological health and disease.

In cases of gut dysbiosis, cognitive decline has been associated with various alterations in the brain resulting from dysregulation of the GBA. Previous studies have listed alterations in neurotransmitter levels [7–9], oxidative stress [10, 11], disruption of the blood–brain barrier (BBB) [12, 13], and reduced neurogenesis [14, 15] as plausible reasons for cognitive decline.

Within the intricate architecture of the brain, glial cells (particularly astrocytes and microglia) play dynamic roles that extend far beyond structural support. They actively modulate neuronal function and maintain homeostasis. Microglia, the resident immune cells of the brain, exhibit remarkable adaptability: they transition from a resting surveillance state to an active state in response to injury or disease. Activated microglia release inflammatory signals and clear threats via phagocytosis, aiding in repair [16]. Astrocytes are often regarded as caretakers of the brain and undergo a transformation known as astrogliosis when activated. This process alters the form and function of astrocytes to regulate blood flow, nourish neurons, and support synapses [17]. However, chronic activation of these glial cells can lead to neuroinflammation and exacerbate neurodegenerative and cognitive disorders. Few findings highlight the link between glial overactivation and cognitive impairment [18, 19]. While this relationship holds significant promise, the effects of gut dysbiosis on glial alterations remain largely unexplored.

Like glial cells, neurotrophic factors such as brain-derived neurotrophic factor (BDNF) and ciliary neurotrophic factor (CNTF) play vital roles in the regulation of various cell functions within the brain. Recent research suggests that these factors are key in the link between gut dysbiosis and glial activation [20–23]. Gut dysbiosis can disrupt the balance of gut microbes, leading to the production of metabolites and inflammatory molecules. These can travel throughout the body and may influence neurotrophic factors, which in turn activate microglia in the brain, influencing central nervous system (CNS) function [24]. While BDNF has been extensively studied in the context of dysbiosis [22, 25], CNTF has received less attention. CNTF has been reported to support neuronal connectivity and protect against cell loss in models of neurodegeneration [26]. CNTF can alter brain metabolism by increasing glycolysis and mitochondrial function, thereby supporting neuronal energy demands and viability under stress [27]. Hence, CNTF is said to play a neuroprotective role [28] and aids in neurogenesis [29]; however, the overproduction of CNTF has been shown to cause toxic effects [30]. Increased levels of CNTF have been shown to enhance microglial activation, promoting a proinflammatory state in the CNS [20]. This interaction may contribute to neuroinflammatory conditions and neurodegenerative diseases. However, understanding the role of CNTF in the context of gut dysbiosis is unexplored.

Using a mouse model of antibiotic-induced gut dysbiosis that was previously established [31], we studied cognitive decline via mechanisms involving oxidative stress, inflammation, blood–brain barrier (BBB) disruption, glial activation, and altered neurotrophic signalling. Here, we hypothesised that gut dysbiosis may contribute to cognitive decline through neuroinflammation, with a significant contribution from CNTF (Fig. 1).

Fig. 1.

Graphical abstract. Schematic representation summarizing the key findings of the study

Results

Antibiotic-induced gut dysbiosis led to cognitive impairment

Given the growing evidence linking the gut microbiota to cognitive processes, we conducted behavioural tests to assess the effects of antibiotic-induced gut dysbiosis on cognition. Our data revealed that recognition memory, assessed by the NORT, the test animals spent significantly less time with the novel object (Fig. 2A) (80.3 vs 59.54; p = 0.0489; n = 13), indicating a reduced ability to discriminate between novel and familiar objects and suggesting impaired recognition memory. In the MWM test, the test mice tended to have longer latencies to locate the hidden platform during the training phase, although the difference was not statistically significant (Fig. 2B). During the probe trial, the test animals spent significantly less time in the target quadrant than the controls did (Fig. 2C) (17.82 vs 8.05; p = 0.04; n = 12–13), indicating impaired spatial memory. However, spatial memory (assessed by the number of spontaneous alternations using the Y-maze) (Fig. 2D) (40.32 vs 39.51; p = 0.9312; n = 14), and spatial reference memory (assessed by the percentage of time spent by the animal in the novel arm of Y maze) (Fig. 2E) (37.56 vs 49.29; p = 0.7743; n = 12–14), were not significantly different between the test group and the control group. Dysbiosis appears to impair long-term memory, as evidenced by significant deficits in recognition (NORT) and spatial memory (probe trial, MWM)) observed in the test group. Both of these memory functions are linked to the hippocampus, with recognition memory also involving the medial prefrontal cortex. In contrast, short-term or working memory, primarily assessed by the Y-maze and dependent on the prefrontal cortex, remained unaffected. This evidence leads us to conclude that dysbiosis mainly impairs long-term memory formation and hippocampal function.

Fig. 2.

Antibiotic-induced gut dysbiosis impairs cognitive performance in mice. Cognitive function was assessed using various behavioural tests to evaluate the impact of gut dysbiosis. Spatial learning and memory were evaluated using (A) Novel Object Recognition Test (NORT), (B, C) Morris Water Maze (MWM), (D) Y maze- Spontaneous alternations and (E) Y maze- Reference memory test. n = 12–14 mice/group. The data are presented as mean ± SEM. Statistical analysis was performed using the Mann–Whitney U-test and two-way ANOVA. *p < 0.05, ns: no significance

Antibiotic-induced gut dysbiosis did not affect acetylcholinesterase levels

As cognitive decline was observed in the behavioural assays, we next investigated changes in acetylcholine, a key neurotransmitter. Acetylcholine plays a vital role in the development and consolidation of long-term memory by modulating processes involved in learning, memory formation, and retrieval. AChE estimation serves as an indirect method for analysing the levels of acetylcholine. The assay revealed no significant changes in AChE levels (Additional file 1: Fig. S1 A) in the cerebellum (35.32 vs 29.00; p = 0.9660; n = 7), prefrontal cortex (29.61 vs 28.69; p > 0.9999; n = 7) or hippocampus (32.60 vs 23.17; p = 0.4287; n = 7), suggesting no possible alteration in the levels of acetylcholine.

Proinflammatory mediators were not significantly altered despite antibiotic-induced gut dysbiosis

Inflammation is known to contribute to cognitive decline; therefore, major proinflammatory mediators, such as IL6, IL-1β and TNF-α, were measured in the cerebellum, prefrontal cortex and hippocampus. The analysis revealed no significant changes in the levels of IL-6 (Additional file 1: Fig. S1 B) in the cerebellum (170.60 vs 141.6; p = 0.4147; n = 6–7), prefrontal cortex (66.54 vs 58.39; p > 0.9999; n = 4–6) or hippocampus (54.41 vs 54.22; p > 0.9999; n = 6). However, IL-1β levels were significantly lower in the cerebellum (Additional file 1: Fig. S1 C) than in the control group (124.47 vs 90.19; p = 0.0418; n = 6–7), whereas IL-1β levels in the prefrontal cortex (48.02 vs 38.64; p > 0.9999; n = 4–7) or hippocampus (41.71 vs 29.55; p > 0.9999; n = 6–7) remained unchanged. Similar to IL-6, TNF-α levels also showed no significant changes (Additional file 1: Fig. S1 D) in the cerebellum (260.48 vs 186.99; p = 0.1825; n = 6–7), prefrontal cortex (148.05 vs 83.05; p > 0.9999; n = 5–6) or hippocampus (94.32 vs 53.051; p > 0.9999; n = 6).

Antibiotic-induced gut dysbiosis did not lead to oxidative stress

Another factor that can lead to cognitive decline is oxidative stress. The high metabolism and lipid-rich composition of the brain make it particularly susceptible to oxidative damage. Over time, this damage to DNA, proteins, and lipids can contribute to brain ageing and a spectrum of cognitive decline [32]. Here, the most commonly used oxidative stress markers, TBARS and GSH, were analysed.

Thiobarbituric acid reactive substances (TBARS)

Malonaldehyde levels across three brain regions (Additional file 1: Fig. S1 E), namely, the cerebellum (0.31 vs 0.35; p > 0.9999; n = 7), prefrontal cortex (0.81 vs 0.72; p > 0.9999; n = 7) and hippocampus (0.81 vs 0.82; p > 0.9999; n = 7), did not significantly differ.

Glutathione (GSH)

Similar to TBARS, GSH levels remained unchanged following induced dysbiosis (Additional file 1: Fig. S1 F). The GSH levels in the cerebellum (6.85 vs 7.56; p > 0.9999; n = 7), prefrontal cortex (59.27 vs 60.24; p > 0.9999; n = 7), and hippocampus (48.95 vs 43.07; p > 0.9999; n = 7) were not associated with oxidative stress.

Tight junction protein stability preserved blood‒brain barrier integrity despite antibiotic-induced gut dysbiosis

To understand whether the BBB is compromised by dysbiosis, the mRNA levels of tight junction proteins, which are the gatekeepers that maintain the integrity of the blood–brain barrier, were analysed. Claudin-5, a key transmembrane protein, forms the backbone of these tight junctions, directly contributing to the barrier's resistance to the passage of molecules. ZO-1, a cytoplasmic scaffolding protein, anchors claudins to the actin cytoskeleton, stabilising the tight junction structure. Occludin works in concert with claudins to regulate paracellular permeability. The relative mRNA expression of Claudin-5 (0.28; p = 0.9904; n = 7), Zo-1 (0.09; p = 0.9996; n = 7), and Occludin (2.20; p = 1666; n = 7) in the cerebellum (Additional file 1: Fig. S2 A) did not change significantly. Similar results were observed in the prefrontal cortex (Additional file 1: Fig. S2 B) for Claudin-5 (0.37; p = 0.8905; n = 7), Zo-1 (0.23; p = 0.9694; n = 7) and Occludin (1.30; p = 1301; n = 7), suggesting that no barrier leakage occurred in these areas. However, in the hippocampus (Additional file 1: Fig. S2 C), although the mRNA levels of Claudin-5 (0.48; p = 0.9588; n = 7) and Zo-1 (2.38; p = 0.1540; n = 7) did not change, a significant increase in the level of Occludin (3.26; p = 0.0374; n = 7) was observed.

Antibiotic-induced gut dysbiosis alters microglial function and enhances synaptic pruning

Neuroinflammatory processes represent another key mechanism by which dysbiosis may contribute to learning and memory deficits. Neuroinflammation is the process in which the resident immune cells in the central nervous system (astrocytes, microglia or peripheral immune cells) are activated by inflammatory mediators (including interleukins, chemokines, reactive oxygen species, or nitric oxide) [33]. Activated microglia and astrocytes can impair neuronal function when they adopt a reactive state in response to stress or injury [34, 35]. This transition is a defensive response designed to protect the brain and sustain homeostasis. GFAP (Glial Fibrillary Acidic Protein), a marker used to detect astrocytes, is a protein that acts as a structural component of the cytoskeleton in astrocytes, and IBA-1 (Ionised Calcium-binding Adaptor Molecule 1), also known as Allograft Inflammatory Factor 1 (AIF-1), is a protein primarily found in microglia and macrophages.

The RT‒PCR assay for determining the mRNA level of Gfap (Fig. 3A) revealed no significant alterations in the cerebellum (0.92; p = 0.8605; n = 7), prefrontal cortex (0.86; p = 0.8816; n = 7) or hippocampus (2.32; p = 0.2759; n = 7). Given the observed cognitive decline, further analysis focused on the hippocampus's Cornu Ammonis (CA) subregion, the area known for its role in long-term spatial and recognition memory [36]. The IHC results revealed a significant increase in the percentage of GFAP + cell density within the CA1 region of the hippocampus in the test mice compared with the controls (Fig. 3B) (142.36; p = 0.0010; n = 3). No significant change in GFAP fluorescence intensity was observed (Fig. 3C) (97.24; p = 0.7032; n = 3).

Fig. 3.

Antibiotic-induced gut dysbiosis increases glial cell populations in the hippocampus. mRNA expression of (A) Gfap and (D) Iba-1 (n = 7 mice/group). Immunohistochemical evaluation of (B, C) GFAP (E, F) and IBA-1 (n = 3 mice/group; ROIs per group = 15). Representative IHC images of the CA1 region showing GFAP-stained (G, H) astrocytes and (I, J) IBA-1-stained microglia, Scale = 50 μm. Data are presented as mean ± SEM. Data was analysed by Mann–Whitney U-test and by Ordinary 2-way ANOVA. **p < 0.01, ***p < 0.001, ns: no significance

The relative mRNA levels of Iba-1 (Fig. 3D) did not significantly change in the cerebellum (−1.33; p = 0.7121; n = 7) or prefrontal cortex (−1.69; p = 0.5557; n = 7), whereas a significant increase in the hippocampus (5.27; p = 0.0054; n = 7) was detected in test animals. To validate the RT‒PCR results, we performed IHC. Despite increased mRNA expression, no significant difference in cell density (Fig. 5E) (112.64; p = 0.1138; n = 3) or percentage fluorescence intensity (Fig. 3F) (88.79; p = 0.1883; n = 3) was detected in the CA1 subregion of the hippocampus. Representative images are shown (Fig. 3G- J).

Fig. 5.

Antibiotic-induced dysbiosis alters microglial function, enhancing synaptic pruning. A Estimation of the volume of lysosomal vesicles in microglia. n = 3 mice/group. n of randomly selected microglia in each animal = 24. SYNAPTOPHYSIN1 puncta in (B) LAMP-1⁺ lysosomal vesicles in IBA-1⁺ cells and (C) IBA-1⁺ cells. HOMER-1 puncta in (D) LAMP-1⁺ lysosomal vesicles in IBA-1⁺ cells and (E) IBA-1⁺ cells. n = 3 mice/group. n of randomly selected microglia in each animal = 12. Representative 3D reconstructions of images of an entire microglial cell (blue) with its lysosomal vesicle (red) (F, G), SYNAPTOPHYSIN1 puncta (yellow) and (H, I) HOMER-1 puncta (yellow). Scale = 5 μm. (J-M) Enlargement of the selected section of its corresponding image. Scale = 2 μm. n = 3 mice/group. The data are presented as mean ± SEM and analysed by Mann–Whitney U-test. ***p < 0.001, ****p < 0.0001, ns: no significance

To investigate the activation status of microglia, the mRNA level of Cd68, a marker of activated phagocytic microglia, was analysed, and significant upregulation of Cd68 was observed in the hippocampal region of the test animals (Fig. 4A) (2.76; p = 0.0051; n = 7). This prompted us to examine morphological changes in microglia. Under resting conditions, microglia exhibit a ramified shape with long, thin processes that monitor the environment. In response to neuroinflammatory signals, they transition to an activated state characterised by retraction of processes and cell body enlargement [37]. This morphological transformation facilitates their roles in phagocytosis, cytokine production, and the modulation of neuronal activity [38]. Therefore, microglial morphology serves as a reliable indicator of their activation state and functional response to pathological stimuli. Single-cell microglial analysis revealed no significant alterations in total microglial volume (Fig. 4B, G, H) (102.61; p = 0.7859; n = 3) or the soma volume of microglia in response to dysbiosis (Fig. 4C, G, H) (111.40; p = 0.0743; n = 3). However, filament analysis of microglia revealed a significant reduction in the length of microglial processes (Fig. 4D) (395.6 vs 328.6; p = 0.0039; n = 3) in the CA1 hippocampal region of mice subjected to dysbiosis. Additionally, there was a notable decrease in the percentage of branching points (Fig. 4E, I, J), (65.60; p < 0.0001; n = 3) in comparison with the control. These findings indicate that antibiotic-induced gut dysbiosis leads to a change in microglia, causing them to adopt an activated state. We applied Sholl analysis to 3D models of microglia to quantify subtle changes in microglial processes [39]. The microglia of the test animals displayed a significant loss of complexity (indicated by a significantly reduced average number of total Sholl intersections, especially in the vicinity of the cell soma) (Fig. 4F) (p < 0.001; n = 3).

Fig. 4.

Antibiotic-induced gut dysbiosis triggered microglial activation, leading to a transition from a resting to an activated phenotype. A mRNA expression levels of the Cd68 marker. n = 7 mice/group. B The overall change in the volume of microglia and (C) the volume of the microglial soma. Alterations in microglial filament (D) length and (E) branching points. F Sholl analysis revealed a reduction in ramification in test animals. n = 3 mice/group. n of randomly selected microglia in each animal = 24. Representative images showing 3D reconstructions of IBA-1 immunostained microglia, highlighting (G, H) cell volume and (I, J) soma volume (red) and branching points on filaments (red dots). Scale = 5 μm. The data are presented as mean ± SEM and analysed by Mann–Whitney U-test, two-way ANOVA. **p < 0.01, ***p < 0.001, ****p < 0.0001, ns: no significance

As a next step to further investigate the functionality of activated microglia, we analysed the volumetric changes in lysosomes within the microglia. Lysosomes are dynamic organelles that are integral to cellular catabolism. They exhibit adaptability in their number, size, and distribution according to the specific needs of the cell at any given moment [40]. Given that most phagocytosed materials are anticipated to undergo degradation within lysosomal vesicles, the total volume of lysosomes within microglia may be correlated with their phagocytic activity. Our analysis revealed that the lysosomal volume of IBA-1⁺ cells (Fig. 5A) (269.96; p < 0.0001; n = 3) in response to dysbiosis was significantly increased, confirming that microglia have active phagocytic function.

Given the phagocytic characteristics exhibited by microglia in the hippocampus of mice in response to dysbiosis, (determined through both morphological and lysosomal analyses) we then investigated whether these phagocytic microglia might contribute to synapse loss. Pruning synapses to refine neural circuits during development and maintain adult plasticity is one of the key functions of microglia [41, 42]. However, dysregulated microglia may disrupt homeostatic functions and can lead to excessive pruning, causing synaptic loss and, eventually, neuronal damage [43]. To investigate microglia-mediated synaptic pruning, we analysed the percentage of SYNAPTOPHYSIN1⁺ and HOMER-1⁺ puncta within microglia across both experimental groups. SYNAPTOPHYSIN1 is a crucial integral membrane protein predominantly found in synaptic vesicles in presynaptic terminals that plays a significant role in neurotransmitter release and synaptic function [44]. On the other hand, HOMER-1 is a synaptic scaffold protein located at postsynaptic terminals that plays a critical role in regulating glutamatergic synapses and the morphogenesis of dendritic spines [45]. The increased density of SYNAPTOPHYSIN1 or HOMER-1 puncta in microglia is directly associated with synaptic loss at presynaptic or postsynaptic terminals, respectively. Our data revealed that the density of SYNAPTOPHYSIN1 puncta within LAMP-1⁺ lysosomal vesicles of IBA-1⁺ microglia (Fig. 5B) (322.64; p < 0.0001; n = 3) and IBA-1⁺ microglia (Fig. 5C) (174.7; p = 0.0001; n = 3) was significantly increased in test subjects (Fig. 5F, G). These findings indicate that microglia-mediated synaptic loss occurs at the presynaptic terminals of neurons within the CA1 subregion of the hippocampus. However, the number of HOMER-1-positive puncta within LAMP-1⁺ lysosomal vesicles of IBA-1⁺ microglia (Fig. 5D) (113.25; p = 0.6806; n = 3) and IBA-1⁺ microglia (Fig. 5E) (101.81; p = 0.9234; n = 3) was not significantly altered (Fig. 5H, I). These findings suggest that microglia are not engaged in synaptic pruning at postsynaptic terminals.

Antibiotic-induced dysbiosis elevated CNTF levels

Given the observed glial alterations, we next examined changes in neurotrophic factors, as glial cells play a central role in the production, regulation, and signalling of these molecules, which are critical for neuronal survival, development, and repair [46]. Brain-derived neurotrophic factor (BDNF) has been elaborately studied and is directly linked to cognitive decline [47–49]. The mRNA expression of BDNF (Fig. 6A) in the hippocampus (−2.41; p = 0.0118; n = 6) was significantly downregulated, which aligns with the cognitive deficiency observed in the test animals during the behavioural assays.

Fig. 6.

Antibiotic-induced dysbiosis increased CNTF and its downstream signalling. mRNA expression of (A) Bdnf, (B) Cntf and (C) Cntfrα. Protein expression of (D, G) CNTF, (E, H) CNTFRα and (F, I) STAT3. n = 7 mice/group. Data from Western blot were normalised to β-actin and expressed relative to control (set as 1). The data are presented as mean ± SEM. Data analysed by Mann–Whitney U-test and Ordinary 2-way ANOVA. *p < 0.05, **p < 0.01, ns: no significance

The observed increase in the number of astrocytes (Fig. 3B), which are known to release CNTF, especially under inflammatory conditions (Kamiguchi et al., 1995), along with the impact of microglial pruning on synaptic plasticity prompted us to investigate the alterations in CNTF. This neurotrophic factor is recognised for its role in modulating synaptic plasticity and neurogenesis (Yang et al., 2008). The expression of the Cntf mRNA (Fig. 6B) in the hippocampus increased significantly (2.68; p = 0.0204; n = 7) in the test animals. The mRNA expression of Cntfrα, a receptor in the tripartite complex for CNTF binding, was subsequently analysed. Significant upregulation of Cntfrα mRNA expression was observed in the hippocampus (Fig. 6C) (3.28; p = 0.0011; n = 7), paralleling the increase in CNTF mRNA levels. Changes in protein levels in the whole hippocampus were assessed by Western blotting to validate the findings observed via RT‒PCR. The protein expression of CNTF (Fig. 6D, G) (2.35; p = 0.0169; n = 7) was significantly increased in the test animals, but the protein level of its receptor, CNTFRα (Fig. 6E, H) (0.93; p = 0.1801; n = 7), was not significantly altered. The biological effects of CNTF are primarily mediated through downstream activation of the JAK/STAT signalling cascade, with preferential activation of STAT3 over other STAT family members [50]. Consistent with these findings, a significant increase in STAT3 protein levels was observed (Fig. 6F, I) (1.15; p = 0.0169; n = 7), indicating activation of the JAK/STAT3 signalling pathway.

Discussion

Several approaches can be used to induce dysbiosis in animals, including different antibiotic cocktails, germ-free mice, faecal microbiota transplantation, and dietary interventions. Among these, antibiotic administration is considered efficient, controllable, and minimally invasive [51, 52]. Germ-free mice provide an extreme model of microbial depletion, whereas antibiotic-induced alterations are more suitable for dysbiosis studies [53]. Our previous work using the same antibiotic regimen demonstrated significant alterations in the gut microbiome, including a reduction of beneficial commensal taxa and an overrepresentation of opportunistic species [31]. In the present study, we show that these antibiotic-induced microbial shifts are associated not only with local gut changes but also with alterations in neural and cognitive parameters, highlighting the interconnectedness of the gut–brain axis. In our study, after antibiotic-induced gut dysbiosis, the animals were subjected to behavioural assays to assess their cognitive function. The Morris water maze (MWM) is used to assess spatial learning and memory, whereas the novel object recognition test (NORT) is used to evaluate recognition memory. Both tasks primarily assess long-term memory, which relies heavily on hippocampal function. Interestingly, the observed cognitive deficits appear limited to long-term memory, suggesting that short-term or working memory, which is governed primarily by prefrontal cortex activity, may be relatively preserved during the acute 14-day dysbiosis period examined.

This aligns with the study’s emphasis on the hippocampus as a central region impacted by induced gut dysbiosis. The hippocampus is particularly susceptible to changes in the gut microbiota because of its direct and indirect communication with the gut via neural, immune, and endocrine pathways [54]. One key neural route is the vagus nerve, and a significant portion of vagal afferent input projects to brain regions within the limbic system, including the hippocampus. Alterations in the composition of the gut microbiota can influence vagal tone, thereby modulating hippocampal activity and affecting processes such as synaptic plasticity, neurogenesis, and memory consolidation [55]. Additionally, microbial metabolites (e.g., short-chain fatty acids), inflammatory cytokines, and neuroactive compounds can cross the blood–brain barrier or act via peripheral signalling pathways to further impact hippocampal function, making this region especially sensitive to dysbiosis [56].

In the event of gut dysbiosis, cognitive decline can generally be attributed to several changes in the brain due to GBA dysregulation. Previous studies have shown that alterations in neurotransmitter levels [7–9], oxidative stress [10, 11], disrupted barriers in the brain [12, 13], and reduced neurogenesis [14, 15] lead to impairments in the process of learning and memory.

Neurotransmitters are essential for synaptic communication and directly influence neuronal signalling. Acetylcholine, a key excitatory neurotransmitter, plays a pivotal role in learning and memory [57, 58]. Acetylcholinesterase regulates acetylcholine levels by breaking it down, ensuring precise neural signalling. When this balance is disrupted, it can lead to cognitive impairments, prompting our investigation into acetylcholine levels in the context of observed cognitive deficits [59]. However, our study did not find significant results regarding these levels. While it is undeniable that reduced acetylcholine plays a role in cognitive decline, particularly in conditions such as Alzheimer's disease, it is important to recognise that other factors also contribute to this complex process [60]. Our results on the levels of proinflammatory cytokines and oxidative stress markers revealed no significant inflammation or oxidative stress. This could be because, under dysbiosis conditions, both inflammation and the oxidative stress response are based on specific microbial shifts [61], which are further shaped by host genetics [62], type (mild vs severe) [63] and period of dysbiosis (short vs chronic) [64]. In our study, the impact might have been localised within the gut without systemic effects, or the specific microbial changes involved might not have been strong inducers of these processes [65, 66]. Interestingly, there was a considerable reduction in IL-1β levels in the cerebellum compared with those in controls. This finding indicates that dysbiosis might affect the brain in a region-specific manner [67]. The gut microbiota produces metabolites and molecular signals that help maintain immune balance. In our study, the loss of these signals may have resulted in blunted immune activity and a suppressed cytokine response [68]. Additionally, IL-1β is an early response cytokine; hence, the initial inflammatory response might have already peaked and then entered a resolution phase, resulting in a lower than baseline level of IL-1β at the time of assessment [69].

The BBB relies heavily on tight junction proteins to maintain its integrity and selective permeability [70]. The findings of this study indicated that occludin, a gene essential for maintaining the BBB, was upregulated in the hippocampal region, whereas other tight junction proteins remained unchanged, suggesting that occludin may selectively respond to signals from the gut microbiota [12, 71, 72]. Since occludin is involved in paracellular permeability, its upregulation can change how molecules pass between endothelial cells. Upregulated mRNA expression potentially results in increased restrictive barrier function, thereby limiting the influx of potentially deleterious substances into the brain [73, 74]. Hence, this result suggests a region-specific response to gut dysbiosis, triggering an adaptive response to mitigate permeability changes and maintain homeostasis in response to altered gut-derived metabolites [75]. A key consideration in our study is that tight junction alterations were examined only at the mRNA level by qRT-PCR, without protein-level validation by Western blot or immunohistochemistry. Future studies using these approaches will yield more conclusive insights into BBB integrity, particularly in the context of region-specific responses.

Glial responses are common in gut dysbiosis studies [76]. Many studies have shown how glial cells act as soldiers in the brain and facilitate the maintenance of an ideal environment for neuronal function. Glial cells can adopt a destructive role when triggered by various factors, such as invading pathogens, neuronal injuries [77], or toxic compounds from microbial metabolites [78]. This response leads to proinflammatory signalling and alters microglial gene expression, releasing neurotoxic mediators that can ultimately contribute to neurodegeneration [79]. We observed that the mRNA expression of Iba-1, a panmicroglial marker, and CD68, a marker of phagocytic microglia, was significantly increased in the hippocampus of test mice, whereas a trend toward an increase in the expression of a Gfap marker in the hippocampus was also observed notable evidence of glial cell activation. The CA1 region of the hippocampus plays crucial roles in learning, memory consolidation, and long-term potentiation (LTP), which are essential for synaptic plasticity [80–82]. Therefore, our research focused on examining specific changes within this subregion of the hippocampus. Our studies demonstrated a significant increase in GFAP + cells through immunohistochemistry analysis. Previous studies have indicated that antibiotic-induced gut dysbiosis can increase GFAP levels in the brain by modifying gut–brain signalling via metabolites [83], which can subsequently lead to microglial activation [84]. Since phagocytic microglial gene expression is upregulated and the number of GFAP⁺ cells is increased, closely examining microglial function is important. When activated, ramified microglia adopt an ameboid shape characterised by increased cell and soma volume, along with thicker and less complex processes [85, 86]. Morphological analysis supported our hypothesis by demonstrating that microglia in the test animals were activated, resulting in significant reductions in process length and branching complexity. This finding suggests a shift towards a more activated state, which is characterised by fewer and shorter processes, indicating a change in their functional role. The composition of gut microbiota is a key determinant of microglial homeostasis. Loss of beneficial microbes through dysbiosis can provoke microglial overactivation, contributing to neuroinflammation [87, 88].

Interestingly, microglial activation triggers an increase in GFAP (astrocyte activation), which reciprocally amplifies microglial reactivity. This interaction creates a glial activation loop that emphasises their mutual signalling and synergistic effects [89, 90]. The lysosome vesicle volume also significantly increased, indicating enhanced cellular cleaning activity. Lysosomes function as the cell's garbage disposal unit, effectively clearing debris, cellular waste, and toxic substances to prevent proteinopathy [91, 92]. This increase in lysosomal volume suggests increased phagocytic activity and an active response to maintaining cellular health. Consistent with the established role of activated microglia in synaptic pruning, our data demonstrated a significant increase in SYNAPTOPHYSIN1 + puncta within microglia and their lysosomes, indicating enhanced engulfment of presynaptic terminals and suggesting that microglial activity under these conditions may contribute specifically to presynaptic terminal loss [93]. It is important to note that our findings demonstrate that microglia engulf synaptic elements, supporting the occurrence of synaptic remodelling under antibiotic-induced dysbiotic conditions. Nonetheless, we did not assess whether these changes result in a global reduction in synaptic density. Quantitative approaches such as Golgi staining or high-resolution imaging would be required to address this question, and future studies incorporating these methods will be essential to clarify the extent of synaptic loss.

While microglia can prune at both sides (pre- or postsynaptic boutons), preferential presynaptic pruning has been reported. This could be because of complement C1q/C3 tagging at the synapse, neuronal "eat-me" signals such as phosphatidylserine recognised by microglial TREM2/MERTK, or simply because presynaptic terminals (at the small axon boutons) are more accessible to microglia than the larger postsynaptic dendrites [94, 95]. Recent research has described a sub-phenotype of microglia, dark microglia (named aptly owing to their dense appearance under an electron microscope), which are abundantly observed in aging, stress and neurodegenerative conditions and prefer pruning at the presynaptic terminals of neurons [96, 97]. However, further investigation is needed to identify the specific trigger that initiated presynaptic pruning in our study.

In our study, we observed a significant upregulation of CD68 expression, indicative of macrophagic microglia [98], along with pronounced morphological alterations consistent with microglial activation [86], and an increase in astrocyte number [99]. Collectively, these glial changes are indicative of neuroinflammation, even though they occurred without a corresponding elevation in cytokine levels- a dissociation that has also been reported previously [100, 101].

The loss of synapses mediated by microglial pruning may have contributed to the cognitive deficiencies observed in the test animals. This occurrence of presynaptic loss led us to investigate the trigger for microglial activation or potential upstream neurotrophic mediators, such as BDNF and CNTF. We observed a significant reduction in Bdnf mRNA levels, and BDNF is critical for learning, memory, and synaptic plasticity. BDNF is a well-recognised marker of synaptic integrity and cognitive function [102, 103]. Notably, gut dysbiosis has been previously linked to decreased BDNF levels in the hippocampus and cortex, which may in turn compromise neuroplasticity and cognitive performance [104, 105]. Moreover, the mRNA expression of Cntf and its receptor was found to be upregulated. Western blot analysis further confirmed increased CNTF protein levels and increased STAT3 levels in the test animals, indicating activation of the JAK/STAT3 pathway. However, it should be emphasised that, although we observed an elevation of CNTF protein levels in the hippocampal region, we did not determine its precise cellular origin. Since CNTF is primarily expressed by astrocytes, the observed increase is likely glial in origin [106]. However, as CNTF is an intracellular, low-abundance protein that is difficult to localise by IHC, future cell-type–specific approaches will be required to confirm this.

CNTF is an injury molecule produced by astrocytes in response to trauma and plays a crucial role in their survival and activation. CNTF promotes the expression of GFAP, which helps reinforce astrocyte function, and their reciprocal relationship maintains central nervous system homeostasis [107]. The observed upregulation of GFAP in our findings could be attributed to CNTF signalling. Previous studies have shown that CNTF binds to CNTFRα to recruit LIFRβ and gp130 and can activate the JAK/STAT3 pathway to trigger microglia [20]. The effects of CNTF on microglia have been only partially studied. CNTF is primarily neuroprotective, but it can have proinflammatory or modulatory effects in certain contexts. Given that CNTF is structurally related to IL-6, it is plausible that CNTF can exert similar effects and stimulate the immune functions of microglia [106]. Some studies suggest that CNTF may indirectly contribute to microglial activation, especially during chronic inflammation or injury, but it is not a classical microglial activator, such as IL-1β, TNF-α, or LPS [106]. CNTF stimulates microglial phagocytosis via a calcium-dependent signalling pathway and increases the expression of α integrin, a receptor involved in phagocytic activity [108]. Together, our results suggest that dysbiosis may downregulate BDNF while promoting astrocytic CNTF expression, potentially contributing to an environment that facilitates microglial activation.

Conclusions

Our findings suggest that CNTF may play a modulatory role in microglial activity, with potential implications for understanding neuroinflammation and its contribution to neurodegenerative processes. In this study, antibiotic-induced gut dysbiosis led to increased CNTF expression, which was associated with microglial activation and subsequent synaptic loss at presynaptic terminals. This loss likely contributes to the memory deficits observed in dysbiotic animals, suggesting a mechanistic link between gut health and cognitive function via neuroinflammatory pathways. Dysbiosis-driven CNTF upregulation may trigger aberrant synaptic pruning by microglia within the CA1 subregion of the hippocampus. This highlights a potential pathway through which gut microbiota alterations can influence brain function. It should be noted that the present findings are specific to the antibiotic cocktail used in this study. Different antibiotic classes, doses, or treatment durations may induce distinct microbial and neurobehavioral outcomes. Additionally, the current study was designed to examine brain alterations during the period of antibiotic-induced dysbiosis; thus, post-treatment recovery was not assessed. Therefore, our conclusions should be interpreted within the context of the regimen employed here. We used only male mice in this study to minimise variability (estrous cycle can result in variations in cognitive performance among females); however, recognising sex as a key biological variable, we note that future studies should include both sexes to improve translational relevance.

In addition, our investigation was limited to the CA1 region, and the involvement of other brain areas remains unexplored. Additionally, the roles of other glial cells, such as astrocytes, in synaptic regulation have not been addressed. Future studies should aim to characterise the contributions of both microglia and astrocytes to synaptic remodelling in the context of gut dysbiosis and neuroinflammation. Such insights could deepen our understanding of gut–brain communication and pave the way for novel therapeutic approaches aimed at restoring synaptic integrity and cognitive function in neurological and psychiatric disorders.

Methods

Animals

Male C57BL/6 mice (adult, 13 weeks old) were housed in a temperature- and humidity-controlled room with a 12-h light/dark cycle with access to ad libitum water and standard chow at the small animal facility of Cochin University of Science and Technology (CUSAT), Kerala, India. The mice were caged in individually ventilated cages (IVCs) on the basis of their groups, with each cage containing 3–4 animals. Only male mice were used to avoid variability due to estrous cycle–driven hormonal fluctuations, which can interfere with hippocampal-dependent memory (Frick et al., 2018; Smejkalova and Woolley, 2010; Yagi and Galea, 2019).

Antibiotic treatment

One week after acclimatisation, the mice were randomly allocated to the control or test groups. Gut dysbiosis was induced according to a previously established protocol adopted directly from our previous work, in which the same antibiotic cocktail was administered to age-matched male mice and gut microbial alterations were characterised [31]. The present study was carried out at identical settings as the preceding study. The control group received autoclaved ultrapure water via oral gavage for 14 days. The test group was administered a cocktail of broad-spectrum antibiotics (200 mg/kg ampicillin sodium salt, 200 mg/kg neomycin sulphate, and 100 mg/kg vancomycin hydrochloride; CAS Nos. 69–52-3, 1405–10-3 and 1404–93-9, respectively) prepared in autoclaved ultrapure water for 14 days. To prevent gut microbiota repopulation from faeces due to coprophagic behaviour, bedding was changed every alternate day.

Experimental timeline and tissue harvesting

To minimise stress, the animals were carefully assigned to behavioural tests, ensuring that each animal participated in only one experiment (Fig. 7). The mice received antibiotic treatment for 8 days before the onset of behavioural training (a duration considered sufficient to induce gut dysbiosis, as supported by previous studies [109]), with testing conducted on day 14. On day 15, after the treatment period, the animals were sacrificed via deep anaesthesia with isoflurane USP. As learning and memory are regulated primarily by the cerebellum, prefrontal cortex, and hippocampus, we included these brain areas for analysis. The prefrontal cortex, cerebellum and hippocampus were separately isolated, promptly frozen at −80 °C and stored for future experiments, including oxidative stress analysis, quantitative RT‒PCR (qRT‒PCR), and ELISA. For the immunohistochemical experiments, the extracted brains were fixed in 4% paraformaldehyde (PFA) at 4 °C for 24 h. The samples were then transferred to a 30% sucrose solution in 1 × PBS and stored at 4 °C until sectioning.

Fig. 7.

Experimental timeline. Timeline followed for treatment and behavioural experiments

Behavioural experiments

Learning and memory-associated behavioural assays were performed to assess cognitive behaviour. Animal activity was recorded by an overhead camera connected to a computer, and testing sessions were analysed with ANY-maze software (Stoelting, Wood Dale, IL; v.7.3). All animals were acclimatised to the behavioural testing room for 1 h before testing. All equipment was wiped with 70% alcohol and air-dried both before and after each session to eliminate olfactory cues. Different groups of animals were used for each experiment to avoid repeated behavioural testing on the same subjects. For all behaviour assays, both the experimenter and the analyst were blinded to the group assignment of the mouse during testing and data analysis.

Novel object recognition test (NORT)

The novel object recognition test (NORT) was used to evaluate cognitive dysfunction, as multiple brain regions contribute to the recognition process. This assay leverages the natural tendency of an animal to explore novel objects over familiar objects. On the first day (habituation), the mouse was placed in the testing chamber without any objects to acclimate for 5 min. The mouse was then returned to its home cage. After 24 h, for the NORT familiarisation phase, the mouse was placed in the same testing chamber containing two identical objects and allowed to explore them for 5 min. The mouse was then returned to its home cage. After another 24 h, one object was replaced with a novel object that differed in colour, shape and texture during the testing phase. The percentage of time spent with the novel object was calculated by dividing the time the mouse spent exploring the novel object (O1) by the sum of the time spent exploring the novel object (O1) and the familiar object (O2) and then multiplying the result by 100 [110, 111]. % time spent = (O1/O1 + O2) × 100. Owing to the rodent’s innate preference for novelty, a mouse that remembers a familiar object is expected to spend more time exploring the novel object.

Morris water maze (MWM)

The MWM test assesses the ability of an animal to learn from cues and navigate via memory. The maze consisted of a circular black tank (150 cm diameter) filled with water (made opaque with titanium dioxide) and maintained at an ambient temperature of 19–22 ⁰C. The escape platform (30 cm in height and 10 cm in diameter) was submerged 1 cm below the water surface. The tank was arbitrarily divided into four quadrants, with the escape platform positioned in the centre of one quadrant. Distinct distal visual cues were placed on the walls surrounding the tank. During the learning phase, each animal received four trials per day for 5 days with semirandom start positions. If an animal failed to reach the platform within the allotted time (60 s), it was guided to the platform and placed on it for approximately 15 s. After each trial, the animals were removed from the water and dried with a towel. The probe trial (performed without the escape platform) was conducted 24 h after the last training (on day 6) to assess the reference memory of the animal [112]. All data, including escape latency (latency to reach the platform) and percentage of time spent in the four quadrants of the pool, were collected and analysed via ANY-maze behavioural tracking software.

Y maze

The Y maze was used to assess short-term spatial memory. Spontaneous alternations were analysed by placing the mice in the middle of the apparatus and allowing them to move freely for 8 min. One alternation was defined as the animal consecutively entering all arms. The percentage of relative alternation was calculated from the ratio of the number of alternations divided by the number of entries. To assess spatial reference memory, one arm of the maze was closed with a divider, and the animal was left to explore the maze freely for 15 min. After one hour, the animal was removed and placed back into the Y maze for 5 min to assess reference memory. The percentage of time spent in the novel arm is calculated to evaluate the animal’s ability to recognise and explore the novel arm, driven by its innate curiosity [113, 114].

Acetylcholine esterase (AChE) assay

The prefrontal cortex, cerebellum and hippocampus were homogenised in phosphate-buffered saline (PBS) with a protease inhibitor cocktail (Sigma‒Aldrich; Cat No: S8830) using a homogeniser (IKA T10 basic- ULTRA TURRAX®). The homogenised samples were sonicated at 30% amplitude with 5-s pulses on and 3-s pulses on ice. The homogenate was then centrifuged at 12,000 × g for 20 min at 4 °C, and the supernatant was collected in a fresh sterile 1.5 mL tube. The protein concentration was estimated using a Bradford protein assay kit (Himedia; Cat No: ML106), and the supernatant was stored at –80 ℃ until analysis. This supernatant was used to measure AChE activity, which was assessed by the addition of the substrate acetylthiocholine iodide (Sigma‒Aldrich; Cat No: A5751) (0.75 mM) and Ellman's reagent (1.5 mM DTNB; 5,5′-dithio-bis 2-nitrobenzoic acid (Sigma‒Aldrich; Cat No: D8130)) using Ellman's method. The change in absorbance was measured at 412 nm per minute [115, 116].

Enzyme-linked immunosorbent assay (ELISA)

The prefrontal cortex, cerebellum and hippocampus were homogenised in 1 × PBS containing a protease inhibitor cocktail. The samples were subjected to pulse sonication (for 4 min at 30% amplitude, with 5-s on and 2-s off pulses) on ice and then placed on a rocker for 30 min at room temperature (RT). The homogenate was centrifuged at 12,000 × g for 20 min at 4 °C, and the resulting supernatant was transferred into fresh tubes and stored at −80 °C. The protein concentration was estimated using a Bradford protein assay kit. Proinflammatory cytokine levels were analysed via Duoset ELISA kits (R&D Systems; Cat Nos: DY406, DY401 & DY410), following the manufacturer's instructions for the markers Interleukin-6 (IL-6), Interleukin 1 beta (IL-1β), and Tumour necrosis factor alpha (TNF-α). Unknown concentrations were calculated using a four-parameter logistic curve fit with OriginPro 2023 software.

Oxidative stress assay

To evaluate oxidative stress levels, we quantified the production of reactive oxygen species (ROS) and assessed antioxidant activity using the following biochemical assays.

Thiobarbituric acid reactive substances (TBARS) assay

Tissues from the prefrontal cortex, cerebellum, and hippocampus (20–25 mg) were homogenised in 250 μL of 1 × PBS via a homogeniser and sonicated for 15–20 s on ice. The homogenate was subsequently centrifuged at 2375 × g for 5 min at 4 ºC to collect the supernatant. Malonaldehyde (MDA) levels in the supernatant were analysed via a commercially available kit (EZAssay™ TBARS Estimation Kit; Cat No: CCK023) following the manufacturer's instructions.

Glutathione (GSH) assay

Tissues from the prefrontal cortex, cerebellum, and hippocampus were homogenised in PBS containing 10% trichloroacetic acid (TCA). The homogenate was centrifuged, and the supernatant was collected. 40 μL of this supernatant was added to a 96-well plate, followed by the addition of 80 μL of Ellman's buffer (0.3 M Na₂HPO₄) and 10 μL of Ellman's reagent (0.04% DTNB in 1% sodium citrate). The plate was then placed on a shaker for 60 s, and the absorbance was recorded within 10 min at 405 nm using a microplate reader (Tecan Spark multimode plate reader) [117].

Evaluation of mRNA expression (qRT‒PCR)

Total RNA was extracted from the prefrontal cortex, cerebellum, and hippocampus via RNAiso Plus (TaKaRa; Cat No: 9108) according to the manufacturer's protocol. cDNA was subsequently synthesised via reverse transcription of 1000 ng of total RNA using the Primescript™ 1 st strand cDNA synthesis kit (TaKaRa; Cat No: 6110 A). qRT-PCR was performed on a real-time thermocycler (Applied Biosystems 7300 Real-Time PCR System) using the TB Green kit (TaKaRa; Cat No: RR820A) following the manufacturer's instructions.

To assess the potential compromise of the blood‒brain barrier, the mRNA expression of the tight junction proteins- zonula occludens-1 (ZO-1), Claudin-5, and Occludin was evaluated. Glial alterations were assessed by analysing the gene expression of Iba-1 (ionised calcium-binding adapter molecule 1) (KiCqStart™; NM_001099327), Gfap (glial fibrillary acidic protein) (KiCqStart™; NM_001131020), and Cd68. The expression of Bdnf, Cntf, and Cntfrα was analysed to investigate neurotrophic gene regulation. The primers were designed via NCBI BLAST, with a final concentration of 10 μM for each primer (the sequences are given in Table 1). The fold change was calculated via the 2^−ΔΔCt method to compare the control and test groups [118]. Housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used as an internal control for normalisation [6, 119]. Gapdh being a glycolytic enzyme has a stable expression across tissues including brain [120], and was validated in our study as a consistent reference gene.

Table 1.

Primer sequences

Gene	Primer Sequence
	Forward (5′–3′)	Reverse (5′−3′)
Occludin	TTGAACTGTGGATTGGCAGC	CAAGATAAGCGAACCTTGGCG
Zo-1	CCCTCCTTACTCACCACAAGC	GATGAGGCTTCTGCTTTCTGTT
Claudin-5	TCAGCTTCCCGGTCAAGTACTC	CCGCCCTTAGACATAGTTCTTCTT
Cd68	GACCGCTTATAGCCCAAGGA	TCATCGTGAAGGATGGCAGG
Gapdh	ACCCAGAAGACTGTGGATGG	TTCAGCTCTGGGATGACCTT

Immunohistochemistry (IHC)

To investigate hippocampal microglia, brain hemispheres were cryoprotected in a 30% sucrose solution prepared in 1 × PBS for 24 h and subsequently embedded in Tissue-Tek optimal cutting temperature compound (Hartenstein Laborversand) at − 70 °C, following established protocols [121, 122]. The frozen hemispheres were sectioned into 20 µm-thick slices via a cryostat microtome (Leica 2800E Frigocut). Six consecutive sections from each mouse were selected and placed into 24-well plates for subsequent free-floating immunohistochemical staining. The sections were washed twice with 1 × PBS for 2 min each and then washed three times with 0.1% Triton X-100 for 5 min each while on a shaker. This was followed by a 1-h incubation at RT on a shaker with a blocking solution (composed of 10% goat serum + 5% bovine serum albumin (BSA) + 0.3% Triton X-100 in 1 × PBS for HOMER-1 and GFAP staining; 5% goat serum + 5% donkey serum + 0.3% Triton X-100 in 1 × PBS for SYNAPTOPHYSIN1 staining; 5% goat serum, 5% donkey serum, 5% BSA and 0.3% Triton X-100 in 1 × PBS for all other markers). The sections were then incubated overnight at 4 °C with primary antibodies prepared in blocking solution (Table 2) with continuous shaking.

Table 2.

Primary and secondary antibodies and their respective dilutions were used for immunohistochemistry (IHC) and western blot (WB) assays

Antibodies	Dilution (WB)	Dilution (IHC)	Source	Host species
Primary antibody
CNTFRα	1:1000	-	Thermo Fisher Scientific (Lot No: Ye3932784B; Cat No: PA5-119,975) RRID: AB_2913547	Rabbit
CNTF	1:1000	-	Abcam (Lot No: GR3392285-3; Cat No: ab270992)	Rabbit
STAT3	1:2000	-	Cell Signalling (Lot No: 7; Cat No: 79D7) UniProt ID: P40763	Rabbit
Β-ACTIN	1:2000	-	Sigma-Aldrich (Cat No: A2228) UniProt ID: P60709	Mouse
IBA-1	-	1:1000	Wako (Lot: PAM5010; Cat. No.: 019–19741)	Rabbit
CD107a (LAMP-1)	-	1:500	BD Pharmingen™ RRID: AB_2134499	Rat
HOMER-1	-	1:500	Synaptic Systems RRID: (Lot No: 2–18; Cat No: AB_2631222)	Chicken
SYNAPTOPHYSIN1	-	1:500	Synaptic Systems (Lot No: 8–59; Cat.No. 101 004)	Guinea pig
GFAP	-	1:1000	Sigma-Aldrich – (Lot No: 236421; RRID: AB_477010)	Mouse
S100β	-	1:500	Synaptic Systems – (Lot No: 1–13; Cat. No: 287 004)	Guinea pig
Secondary antibodies
Anti-rabbit IgG-HRP	1:5000	-	ThermoFisher Scientific (Cat No: 31460) RRID: AB_228341	Goat
Anti-mouse IgG-HRP	1:5000	-	ThermoFisher Scientific (Lot No: XG350778; Cat No: 31430) RRID: AB_228307	Goat
Cy™3 AffiniPure goat anti-rabbit IgG (H + L)	-	1:500	Jackson Immuno Research – (Lot No: 166394; RRID: AB_2338006)	Goat
Cy™5 AffiniPure goat anti-rat IgG (H + L)	-	1:500	Jackson Immuno Research (Lot: 163,322; RRID: AB_2338264)	Goat
Cy™3 AffiniPure goat anti-mouse IgG (H + L)	-	1:500	Jackson Immuno Research (Lot No: 156133; RRID: AB_2338686)	Goat
Alexa Fluor® 488 AffiniPure donkey anti-chicken IgY (IgG) (H + L)	-	1:500	Jackson Immuno Research (Lot No: 168728; RRID: AB_2340375)	Donkey
Alexa Fluor™ 488 Goat anti-Guinea Pig IgG (H + L)	-	1:500	Invitrogen Cat.No. A-11073, RRID: AB_2534117	Goat

The next day, after a 30-min incubation at RT on a shaker, the sections were washed three times with 1 × PBS for 10 min each. The samples were then incubated with secondary antibodies (diluted in 0.05% Triton X-100 in 1 × PBS) for 2 h at RT in the dark while shaking (Table 2). Afterwards, three additional 10-min washes with PBS were performed (Table 2). Finally, the sections were stained with 4′,6-diamidino-2-phenylindole (DAPI) (1:1,000, ThermoFisher Scientific; Cat No: D1306) for 5 min, followed by six 5-min washes with PBS. The stained sections were mounted onto glass slides using a Fluorogel embedding medium (Electron Microscopy Sciences, Hatfield, PA; Cat No: 17985).

Single-cell imaging and analysis of microglia

Single microglia were imaged from triple-stained sections (IBA-1/LAMP-1/HOMER-1 or IBA-1/LAMP-1/SYNAPTOPHYSIN1). Using a confocal laser scanning microscope (cLSM, Olympus; Model No: BX61WI), Z-stacks of individual microglia were obtained in 0.35 µm increments via a × 40 UPLFLN oil objective (numerical aperture 1.30) with a 6X zoom setting, resulting in a final pixel resolution of 0.103 µm × 0.103 µm. For each animal, Z-stacks were acquired from three randomly selected single microglia within the CA1 subregion of the hippocampus across three different sections. Prior to analysis in IMARIS (Bitplane), the images underwent blind 3D deconvolution via AutoQuantX (Adobe Systems GmbH, vX3.1) to enhance clarity. Within IMARIS, microglial cell surfaces were modelled via IBA-1 staining with a surface detail setting of 0.2 µm. Positive LAMP-1 signals were masked within these constructed surfaces to model vesicular structures at the same surface detail. Additionally, the HOMER-1 or SYNAPTOPHYSIN1 spots within the LAMP-1 vesicles were labelled and modelled via the software’s spot function, with the spot diameter set to 0.5 µm. The overall microglial structure was visualised by masking the IBA-1 signal onto the modelled IBA-1 cell surface. The complexity of microglial branching was assessed via the "Filament Analysis" module, which considers parameters such as the largest diameter (3.5 µm), thinnest diameter (0.2 µm), and sphere region diameter (3.5 µm).

Several metrics were meticulously recorded, including IBA-1 volume (in cubic micrometres), LAMP-1 volume within IBA-1, the number of synaptic terminal spots in LAMP-1, branching points, and the soma size of microglia (in cubic micrometres). All values were documented in a Microsoft Excel spreadsheet for subsequent analysis. All the data were normalised to the mean of the control group for each parameter.

Imaging and quantification of microglial cell density

Sections stained via immunohistochemistry were initially imaged via an Apotome microscope (Imager. M2 AXIO, ZEISS) with a 20X objective (N.A. 0.8). The imaging focused exclusively on the DAPI and Cy3 channels to identify IBA-1- or GFAP-positive cells. Z-stacks were acquired at 1 µm intervals across five to six hippocampal slices per animal. For each slice, the CA1 subregion of the hippocampus was imaged to assess microglial density.

Image analysis was conducted in a blinded manner via Fiji software (BioVoxxel; v2.9.0). Eight central slices from each Z-stack were selected and flattened into 2D projections using the “Z-Project” toolset to maximum intensity. For microglia, the DAPI and Cy3(IBA-1) channels were merged, and for astrocytes, the DAPI, Cy3(GFAP), and Cy2 (S100β) channels were merged via the “Colour” and “Merge Channels” tools. Microglia and astrocytes were manually counted using the “Multiple Points” tool. For astrocyte counts, only cells showing overlap in all three channels were included. The cell density, determined as the number of cells per square millimeter, was calculated using Microsoft Excel.

Additionally, Fiji software was used to quantify the fluorescence intensity of GFAP and IBA-1 staining as indicators of astrocytic and microglial activation in the CA1 subregion. All the data were normalised to the mean of the control group for each respective staining.

Western blot

Hippocampal tissue was lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Sigma‒Aldrich; Cat No: P5726). The protein concentration was determined via the Bradford assay. To detect CNTFR and STAT3, 30 μg of total protein lysate was used, whereas 100 μg of total protein lysate was used for detecting CNTF. The samples were heated to 95 ºC for 5 min in 2 × Laemmli buffer (Himedia; Cat No: ML021) and loaded onto a 10% SDS‒PAGE gel. Proteins were transferred with a semidry blot (Trans-Blot® SD Semi-Dry Transfer Cell) to nitrocellulose membranes (Himedia; Cat No: SF108A) and blocked with 5% BSA. The membranes were incubated overnight with primary antibodies (Table 2), followed by a 1-h incubation with HRP-conjugated secondary antibodies. The bands were visualised via an enhanced chemiluminescence (ECL) kit (TaKaRa; Cat. No: T710A). The membranes were subsequently stripped with mild stripping buffer and reprobed sequentially with the corresponding antibodies. β-actin was used as a loading control. The blots were imaged using a gel documentation system (ChemiDoc XRS⁺ System). Band intensities were quantified via ImageJ (NIH, Bethesda, MD; v1.51) and normalised to that of β-actin for comparison. Densitometry values were normalized to β-actin and then expressed relative to the mean of the control group, which was set to 1 for comparison [123]. Control values were normalized to 1 across all samples, and variability among controls is therefore not depicted in the graph.

Statistical analysis

All analyses were conducted by investigators blinded to the treatment groups wherever possible, ensuring an unbiased assessment. Robust regression and the Outlier Removal (ROUT) method with a value of 1% were used to identify and remove outliers from the dataset. The reported n represents the data after outlier removal. The statistical significance of differences between groups was analysed via either a two-tailed unpaired Student’s t test followed by a nonparametric test (Mann–Whitney U test) or ANOVA followed by post hoc tests (Bonferroni's or Dunnett’s multiple comparison test), as indicated. All the statistical analyses and data visualisations were performed via GraphPad PRISM (v.9.5.1). The data are presented as the mean ± SEMs. Statistical significance was indicated as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****).

Abbreviations

CNTF

Ciliary neurotrophic factor

GBA

Gut-brain axis

BDNF

Brain-derived neurotrophic factor

CNS

Central nervous system

PFA

Paraformaldehyde

NORT

Novel object recognition test

MWM

Morris water maze

ELISA

Enzyme-linked immunosorbent assay

ROS

Reactive oxygen species

TBARS

Thiobarbituric acid reactive substances

MDA

Malonaldehyde

GSH

Glutathione

TCA

Trichloroacetic acid

IHC

Immunohistochemistry

ECL

Chemiluminescence

BBB

Blood-brain barrier

GFAP

Glial Fibrillary Acidic Protein

IBA-1

Ionised Calcium-binding Adaptor Molecule 1

Authors’ contributions

K, R, D.K.K, I.H and S.H conceptualized, designed and conducted experiments and data analysis. A.D and T.S provided technical support. K.M supervised the project. U.S, M.K and B.C.P.S supervised the project and acquired funding. K wrote the manuscript with the input of all authors. D.K.K prepared the graphical abstract. All authors read and approved the final manuscript.

Funding

This work was supported by Indian Council of Medical Research (ICMR, No:5/4–5/180/Neuro/2019-NCD-1), Department of Health Research, Government of India -HRD grant (No. NRI/PIO/2020/000007, R.12015/01/2024-HR/E-Office:8292829), the Kerala State Council for Science, Technology and Environment (KSCSTE, No:316/2022/KSCSTE), YIPB- Kerala Biotechnology Commission Grant (No. 348/2023/KSCSTE), DST-DAAD (No: DST/INT/DAAD/P-04/2023 (G)) and Cochin University of Science and Technology fellowship (Ac.B3/UJRF/2019–20, CUSAT/AC(C).C1/3498/2022).

Data availability

The datasets generated and analysed, supporting the conclusions of this article, are available to download from https://doi.org/10.5281/zenodo.16280652.

Declarations

Ethics approval and consent to participate

All animal experiments were conducted in strict accordance with national and institutional guidelines for the care and use of laboratory animals. This study complies with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. All animal experiments were carried out with prior approval from the Institutional Animal Ethics Committee (approved protocol: 363/G0/Re/S/01/CCSEA/43) of CUSAT, which operates under the guidelines set by the Committee for the Control and Supervision of Experiments on Animals (CCSEA), Government of India. All efforts were made to minimise the number of animals used and their suffering.

Consent for publication

All authors consent for publication.

Competing interests

The authors declare no competing interests.