Gut microbial metabolite TMAO impairs cognitive function and induces hippocampal synaptic plasticity decline through modulation of GSK-3β activity

Abstract

Background and objectives

Growing evidence has suggested that elevated Trimethylamine N-oxide (TMAO) levels, a gut microbiota-dependent metabolite, are closely associated with brain aging and cognitive impairment. Glycogen synthase kinase-3 beta (GSK-3β) activity was depicted to be essential in regulating learning and memory. The current study examined the impact of TMAO on cognitive function in mild cognitive impairment (MCI) patients and rat models while exploring the mechanisms regulating the TMAO-induced GSK-3β signaling.

Methods

This study recruited 115 MCI patients and 128 healthy controls. All participants underwent neuropsychological assessments. Fasting plasma TMAO was measured using high-performance liquid chromatography with online electrospray ionization tandem mass spectrometry. The study also explored whether the GSK-3β signaling was involved in cognitive and function deficits linked with elevated TMAO in rat models.

Results

Our results indicated that TMAO plasma levels were elevated in MCI patients compared to healthy controls, depicting a significant association with potential MCI risk. Furthermore, chronic exposure to choline considerably impacted spatial cognitive performance in the Morris water maze task. This reduced the phosphorylation of Ser9 of GSK-3β and the synaptic plasticity-related proteins within the hippocampus, which could be restored by inhibiting TMAO with ABS. In addition, inhibition of GSK-3β by SB216763 significantly prevented the TMAO-induced synaptic damage while decreasing the membrane level of GluA1 and improving hippocampal learning and memory.

Discussion

These results indicate that TMAO can induce hippocampal-dependent learning and memory ability impairment with deficits in synaptic plasticity by regulating the GSK-3β activity.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-025-01846-z.

Introduction

The human gut houses trillions of microbes, such as bacteria, eukaryotes, and viruses, leading to a lifelong symbiotic relationship. The host gut and the microbiota living in the gut coproduce an extensive array of small molecules while metabolizing food and xenobiotics. These play critical roles in metabolism [1], immunity [2], emotion, and cognition [3]. Trimethylamine N-oxide (TMAO), a gut microbiota-dependent metabolite, is obtained from phosphatidylcholine, choline, or L-carnitine. Growing evidence has suggested that increased TMAO is strongly related to enhanced brain aging and cognitive impairment [4, 5, 6, 7]. Vogt et al. [8] reported that TMAO concentrations were higher in cerebral spinal fluid (CSF) in patients suffering from mild cognitive impairment (MCI) and Alzheimer’s disease (AD) compared to cognitively unimpaired individuals. However, the influence of TMAO on cognitive function and the elaborate molecular mechanisms remain unclear.

Glycogen synthase kinase-3 (GSK-3), a serine/threonine protein kinase, phosphorylates and inactivates glycogen synthase, thereby modulating neuronal function, including gene expression, neurogenesis, synaptic plasticity, and neuronal structure [9]. A well-known mechanism regulating the activity of the two isoforms of GSK-3, GSK-3α, and GSK-3β, is phosphorylating regulatory serine residues (Ser21 in GSK-3α and Ser9 in GSK-3β), which inhibits GSK-3 activity [10, 11]. It is now widely accepted that the GSK-3β signaling plays an important role in the regulation of learning and memory [12, 13, 14]. Inhibiting GSK-3β induces long-term potentiation, the best characterized molecular and cellular component of the plasticity that could underlie learning and memory [15]. Moreover, the activated GSK-3β immunoreactivity is enhanced in AD brains, co-localized with granulovacuolar degeneration and neurofibrillary tangles [16, 17]. These are significantly associated with the extent of memory deficit [18]. However, whether the GSK-3β signaling is involved in cognitive and function deficits linked with elevated TMAO remains unclear.

The present study investigated the impact of TMAO on cognitive function in MCI patients and rat models while exploring the mechanisms regulating the TMAO-induced GSK-3β signaling.

Materials and methods

Patients

This prospective cohort study was conducted in the Affiliated Wuxi People’s Hospital of Nanjing Medical University between June 2022 and May 2024. All the participants underwent a standardized clinical interview, such as a demographic inventory, followed by physical and mental health examinations. MCI was diagnosed based on Peterson’s criteria, which were operationalized in a similar manner to the Mayo Clinic Study of Aging [19]. The criteria for defining MCI included: (1) subjective cognitive concern in subjects or informants; (2) objective cognitive impairment involving at least one cognitive domain among the four; (3) essentially preserved functional activities; and (4) absence of dementia based on the DSM-IV criteria for diagnosis of dementia. The final judgment depended on neuropsychological test scores and a consensus agreement among neurologists specializing in cognitive disorders. We excluded patients with (1) history of severe stroke, alcoholism, head injury, Parkinson’s disease, epilepsy, major depression (determined using the Self-Rating Depression Scale), or other neurological or psychiatric illnesses (excluded by clinical assessment and case history); (2) major medical illnesses (e.g., cancer, immune system disorders, anemia, thyroid dysfunction and disease related to gut dysbiosis); (3) severe visual or hearing loss; (4) recent severe infections; (5) renal impairment (characterized by serum creatinine level > 2 mg/dl).

Moreover, age- and gender-matched controls were recruited from the general population who attended an annual health examination at the physical examination center. The controls did not have cognitive impairment and did not consume probiotics or antibiotics within one month of study enrollment.

Clinical characteristics

Basic demographics, lifestyle, and health-related information were gathered during the initial stages of the experiment. The body mass index (BMI) was calculated by dividing weight (kg) by squared height (m). Blood pressure was measured with a mercury sphygmomanometer. Hypertension was diagnosed based on previous treatment history involving antihypertensive drugs. Type 2 diabetes mellitus (T2DM) was diagnosed according to the 1999 World Health Organization (WHO) criteria [20]. The estimated glomerular filtration rate (eGFR) was determined using the Chronic Kidney Disease Epidemiology Collaboration equation. Venous blood samples were collected at 6 AM after a 10-hour overnight fast to evaluate biochemical indicators.

Neuropsychological assessments

All the participants were subjected to a comprehensive and standardized neuropsychological assessment involving global cognition and four cognitive domain tests. Global cognition was evaluated with the Montreal Cognitive Assessment (MoCA) and Mini-Mental State Examination (MMSE). The four cognitive domain tests included (1) Episodic memory: Auditory Verbal Learning Test involving immediate recall (AVLT-IR) and Auditory Verbal Learning Test having a 20-min delayed recall (AVLT-20 min DR); (2) Information processing speed: Trail Making Test A (TMT-A) and Stroop Color and Word Test A and B (Stroop-A, Stroop-B); (3) Executive functions: Trail Making Test B (TMT-B), Stroop Color and Word Test (Stroop-C), and Digital Span Test (DST-backward); (4) Visuospatial functions: Clock Drawing Test (CDT).

TMAO measurement

Fasting blood samples were obtained from all subjects using ethylene diamine tetra acetic acid tubes. The samples were immediately processed and stored at -80 °C until further analysis. As previously described [21], TMAO concentrations were quantified using a stable-isotope-dilution assay and high-performance liquid chromatography. The chromatography setup was an online electrospray ionization tandem mass spectrometry over an AB SCIEX QTRAP 5500 mass spectrometer. d9 (trimethyl) TMAO (d9-TMAO) served as the internal standard. The assay demonstrated good inter- and intra-day reproducibility (CVs < 7.5%), accuracy (> 98.0% across low, mid, and high values), and stability with repeated freeze-thaw cycles (≥ 5, intercycle CV% <10).

Animals

Adult male Sprague-Dawley rats (weighing 500–600 g and six months old) were subjected to the experiments. Animals were kept inside a 4/cage with food and water available ad libitum. All the rats were maintained on a 12 h light-dark cycle (lights on at 7 AM) within the same colony room, with constant temperature (21 ± 2 °C) and humidity (55 ± 5%). All the experiments followed the guidelines on animals set forth by the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

After habituation to the laboratory environment for 1 week, the rats were fed with a standard diet (control), a choline-supplemented diet (1.0% choline), or a TMAO supplementation (common diet + 0.12% v/v TMAO provided in sterile drinking water) for six months. A schematic diagram of the experimental design has been added to the supplementary materials (Figure S1). An oral broad-spectrum antibiotic (ABS) cocktail-containing vancomycin (0.5 g/L), neomycin sulfate (1 g/L), metronidazole (1 g/L), and ampicillin (1 g/L)-was administered ad libitum in drinking water. This regimen effectively depletes commensal bacteria and suppresses TMAO generation, as previously shown [21]. At 11 months, some rats received SB216763 (2 mg/kg, i.p.) or saline treatment every alternate day for a month. Previous reports indicate that SB216763 could cross the blood-brain barrier post-intraperitoneal injection [22]. The rest of the macronutrients and micronutrients were consistent with standard rodent chow, showing no differences in food intake across groups.

Morris water maze test

The Morris water maze test was conducted as previously described [23]. The water maze included a 160 cm diameter black circular pool filled with opaque water (30 cm depth) at 25 ± 1 °C. An escape platform (11 cm diameter) was kept inside the middle of one of the quadrants (1 cm below the water surface). The animal’s behavior was monitored using a video camera mounted on the ceiling above the pool’s center. The rats were trained for 120 s per trial and four daily trials in the acquisition trials. The trial was started at four positions with 30-minute intervals for four consecutive days. Each trial began with the rat in the pool facing the sidewalls. If the rat could not escape within 120 s, the experimenter guided it to the platform. When the rat escaped onto the platform, it was allowed to stay on the platform for 30 s before being returned to the home cage. On day 5, the hidden platform was removed, and memory retrieval was evaluated using a probe trial lasting 180 s. A computerized video tracking system recorded the escape latency during the acquisition trials, the number of crossings over the platform, and the time spent in the target quadrant during the probe test.

Western blotting

After the behavioral test, hippocampal tissues were frozen on dry ice post-dissection and stored at -80 °C. Membrane and soluble fractions were separated with a membrane extraction kit (Thermo, USA). Western blotting was performed based on a previous study [24]. For separation, equal protein quantities were loaded onto a 10% polyacrylamide gel with 0.2% SDS. The separated proteins were transferred onto a PVDF membrane (Millipore) and incubated overnight at 4 °C based on the following primary antibodies: GSK-3β (1:1000, Cell Signaling), phospho-Ser9-GSK-3β (1:1000, Cell Signaling), phospho-Tyr216-GSK-3β (1:1000, Abcam), α-tubulin (1:2000, Invitrogen), PSD95 (1:1000, Abcam), and Synapsin Ⅰ (1:1000, Abcam); GluA1 (1:1000, Abcam). After washing, the membranes were incubated using a secondary antibody solution (goat anti-mouse or goat anti-rabbit IgG-HRP, 1:5000, Santa Cruz) at room temperature for 2 h, then detected with the enhanced chemiluminescence (ECL) method.

Ethics

The Ethics Committee of the participating hospital approved all the study procedures. After explaining the study details, all participants or their legal guardians provided written informed consent.

Statistical analysis

Data are represented as frequencies and percentages for categorical variables and medians (IQRs), mean ± SD or mean ± SEM for continuous variables. The Kolmogorov-Smirnov test helped assess the normality of data distribution. The Student’s t-test and Mann-Whitney U test were performed for continuous variables, and the chi-square test helped analyze categorical variables. The prognostic value of TMAO for MCI was evaluated using the receiver operating characteristics (ROC) curve, followed by determining the area under ROC. The threshold was the best TMAO cutoff value, leading to the maximum summation of sensitivity and specificity. Spearman rank correlation test helped determine the association between plasma TMAO levels and neuropsychological assessment scores. Any significant differences within the baseline clinical characteristics and other selected confounders were adjusted while performing multiple linear regression analysis. Moreover, binary and multivariable logistic regression models were used after adjusting for other significant predictors. The regression results were expressed as odds ratios (OR) with 95% confidence intervals (CI). Since the acquisition trials of the Morris water maze test were performed for four consecutive days, repeated measure ANOVA was initially performed. In other cases, one-way or two-way ANOVA was incorporated. Post-hoc analyses were performed using Bonferroni’s test for selected or multiple comparisons when p < 0.05. Overall significance was considered at a 2-sided value of p < 0.05, and all the statistical analyses were conducted using the SPSS version 22 software.

Results

Patient characteristics and clinical variables

The study enrolled 263 consecutive patients. Among these, 12 patients were excluded from the final analysis due to incomplete cognitive assessment scores and 8 for missing blood plasma results, leading to a final sample size of 243 patients. The clinical features and laboratory results are represented in Table 1. Significant differences could be observed in age and cognitive assessment scores (i.e., MoCA, MMSE, AVLT-IR, AVLT-20 min DR, TMT-A, Stroop-A, Stroop-B, TMT-B, Stroop-C, DST, and CDT) between the MCI and non-MCI groups.

Table 1.

Difference of characteristics between the patients with non-MCI and MCI

Characteristics	non-MCI (n = 128)	MCI (n = 115)	p value
Age, years	65.0 ± 8.4	69.3 ± 7.0	< 0.001
Gender, male/female Education (years) BMI, kg/m2 Smoking Alcohol intake > 30 g/day	62/66 8.0 (8.0–11.0) 25.0 (22.9–26.4) 31 (24.2%) 28 (21.9%)	61/54 9.0 (8.0–11.0) 25.0 (23.5–26.7) 26 (22.6%) 20 (17.4%)	0.473 0.919 0.449 0.767 0.381
Hypertension Diabetes mellitus	89 (69.5%) 52 (40.6%)	75 (65.2%) 42 (36.5%)	0.474 0.512
Coronary artery disease	22 (17.2%)	15 (13.0%)	0.369
Atrial fibrillation	12 (9.4%)	8 (7.0%)	0.493
Initial SBP, mmHg	138.1 ± 16.3	137.6 ± 16.3	0.810
Initial DBP, mmHg	81.0 (74.0–90.0)	80.0 (72.0–90.0)	0.395
Hemoglobin, g/L	135.9 ± 10.2	134.5 ± 11.0	0.309
CRP, mg/L	0.7 (0.5–2.4)	0.9 (0.5–3.5)	0.352
Fibrinogen, g/L	3.0 ± 0.8	2.9 ± 0.9	0.386
ALT, U/L	19.0 (14.0–23.0)	21.0 (14.0-24.6)	0.228
AST, U/L	18.0 (14.0-22.3)	19.0 (14.0–23.0)	0.162
Initial glucose, mmol/L	6.7 (5.5–8.3)	6.5 (5.5–7.9)	0.518
TC, mmol/L	4.2 ± 1.2	4.5 ± 1.3	0.130
LDL-C, mmol/L	2.5 (1.8-3.0)	2.6 (2.2–3.2)	0.118
Creatinine, µmol/L BUN, mmol/L	62.0 (55.0–78.0) 5.6 (4.6–6.1)	66.0 (55.0–82.0) 5.6 (4.6–6.6)	0.207 0.531
eGFR, mL/min per 1.73 m2	90.5 (87.0-96.7)	90.9 (86.5–98.3)	0.411
MoCA score	26 (26–28)	23 (21–24)	< 0.001
MMSE score	28 (27–29)	27 (26–28)	< 0.001
AVLT-IR score	5.7 ± 1.4	4.7 ± 1.4	< 0.001
AVLT-20 min DR score	5 (4–6)	3 (2–5)	< 0.001
TMT-A, sec	62 (50–80)	89 (72–121)	< 0.001
Stroop-A, sec	31 (26–42)	32 (28–42)	0.002
Stroop-B, sec	51 (41–65)	60 (48–77)	< 0.001
TMT-B, sec	160 (135–196)	243 (208–320)	< 0.001
Stroop-C, sec	92 (76–112)	118 (97–180)	< 0.001
DST (backward) score	4 (4–5)	4 (3–4)	0.003
CDT score	9 (8–9)	8 (7–9)	< 0.001

Abbreviations: MCI, mild cognitive impairment; BMI, body mass index; SBP, systolic blood pessure; DBP, diastolic blood pressure; CRP, C-reactive protein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TC, total cholesterol; LDL-C, low density lipoprotein cholesterol; BUN, blood urea nitrogen; eGFR, estimated glomerular filtration rate; MoCA, montreal cognitive assessment; MMSE, mini-mental state examination; AVLT-IR, auditory verbal learning test-immediate recall; AVLT-20 min DR, auditory verbal learning test-20-min delayed recall; TMT-A, trail making tests A; Stroop-A, stroop color and word test A; Stroop-B, stroop color and word test B; TMT-B, trail making tests B; Stroop-C, stroop color and word test C; DST, digit span test; CDT, clock drawing test

TMAO levels between MCI and non-MCI groups

The MCI group exhibited significantly higher plasma TMAO levels than the non-MCI group (4.88 [IQR 2.93–7.47] vs. 2.87 [IQR 1.70–4.46] µmol/L; p < 0.001) (Fig. 1A). The area under the ROC curve for differentiating MCI patients from non-MCI controls was 0.704 (95% CI 0.639–0.769; p < 0.001) (Fig. 1B). The optimal cutoff value was 4.74 µmol/L, yielding a sensitivity of 52.2% and a specificity of 82.8%.

Fig. 1.

(A) Plasma concentrations of TMAO in different groups. (B) Receiver operating characteristic curves were used to assess the prognostic value of TMAO for distinguishing patients with MCI from non-MCI controls. AUC, area under the receiver operating curve; CI, confidence interval

Patients having plasma TMAO levels ≥ 4.74 µmol/L demonstrated a significantly higher MCI risk (OR 5.256, 95% CI 2.922–9.455; p < 0.001) than those possessing plasma TMAO levels < 4.74 µmol/L. This significance remained after adjusting for factors such as age, gender, education, body mass index, smoking, drinking, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, initial systolic blood pressure, initial glucose, low-density lipoprotein cholesterol, and eGFR (adjusted OR 5.745, 95% CI 2.913–11.328; p < 0.001) (Table 2). Additional analyses indicated the MCI risk in these patients depended on the TMAO levels as evaluated using quartiles. Being in the highest quartile (OR 6.563, 95% CI 2.967–14.517; p < 0.001), when the lowest quartile was used as a reference, the third quartile (OR 2.549, 95% CI 1.192–5.451; p = 0.016), and second quartile (OR 2.386, 95% CI 1.115–5.108; p = 0.025) had a higher risk association with MCI (Table 2). Patients in the second, third, and fourth quartile of plasma TMAO level exhibited a statistically significant increased MCI risk after adjusting for age, gender, education, body mass index, smoking, drinking, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, initial systolic blood pressure, initial glucose, low-density lipoprotein cholesterol, and eGFR in logistic regression analysis (Table 2).

Table 2.

Relationship between TMAO levels and MCI

TMAO level, µmol/L	OR (95% CI)
	Unadjusted	P value	Adjusteda	P value
Continuous TMAO level
< 4.74	1		1
≥ 4.74	5.256 (2.922–9.455)	< 0.001	5.745 (2.913–11.328)	< 0.001
Quartile of TMAO level
1 (< 2.10)	1		1
2 (2.10–3.66)	2.386 (1.115–5.108)	0.025	2.736 (1.156–6.478)	0.022
3 (3.67–5.53)	2.549 (1.192–5.451)	0.016	3.021 (1.227–7.435)	0.016
4 (> 5.53)	6.563 (2.967–14.517)	< 0.001	6.953 (2.741–17.638)	< 0.001

Abbreviations: TMAO, trimethylamine N-oxide; MCI, mild cognitive impairment; CI, confidence interval

^a Adjusted for age, gender, education, body mass index, smoking, drinking, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, initial systolic blood pressure, initial glucose, low density lipoprotein cholesterol and eGFR

Univariate analysis revealed that elevated TMAO (OR 1.338, 95% CI 1.195–1.499; p < 0.001) depicted a more significant risk association with MCI. The following factors were utilized in a multivariate logistic regression model: age, gender, education, body mass index, smoking, drinking, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, initial systolic blood pressure, initial glucose, low-density lipoprotein cholesterol, and eGFR. The results highlighted that TMAO (OR 1.378, 95% CI 1.205–1.576; p < 0.001) and age (OR 1.069, 95% CI 1.025–1.114; p = 0.002) were significantly related to an elevated risk of MCI.

TMAO levels and cognitive ability

The Spearman rank correlation results demonstrated a significant negative correlation between plasma TMAO levels and MoCA scores (r = -0.212, p = 0.001) (Fig. 2A). Additional linear regression analysis indicated a significant decline in MoCA score by 0.2 points for every 1-µmol/L elevation in plasma TMAO levels (95% CI 0.314 − 0.082; p < 0.001). Hence, there was an approximate median decline of 1 point in the MoCA score for every 5.0-µmol/L elevation in plasma TMAO level. After adjusting for age, gender, education, body mass index, smoking, drinking, hypertension, diabetes mellitus, coronary artery disease, atrial fibrillation, initial systolic blood pressure, initial glucose, low-density lipoprotein cholesterol, and eGFR, the results depicted a significant association between MoCA scores and plasma TMAO concentrations (β = -0.208, 95% CI 0.331 − 0.085; p = 0.001). Moreover, plasma TMAO levels were significantly correlated with the MMSE, TMT-A, and TMT-B scores (Figs. 2B-D).

Fig. 2.

Correlation analysis between neuropsychological assessments and TMAO levels. (A) Correlation between MoCA score and plasma TMAO level. (B) Correlation between MMSE score and plasma TMAO level. (C) Correlation between TMT-A test and plasma TMAO level. (D) Correlation between TMT-B test and plasma TMAO level. TMAO, trimethylamine N-oxide; MoCA, montreal cognitive assessment; MMSE, mini-mental state examination; TMT-A, trail making tests A; TMT-B, trail making tests B

Impairment of spatial cognitive performance caused by chronic TMAO exposure in rats

Rats were supplemented with choline (1%) in chow for 6 months, and spatial cognitive performance was evaluated in the Morris water maze test to examine the TMAO exposure effects on cognitive function. In the acquisition trials, there was a significant effect of the day [F(3, 80) = 251.920, p < 0.001], choline [F(1, 80) = 44.438, p < 0.001], and ABS [F(1, 80) = 18.818, p < 0.001] on latency to identify the platform. After additional day-by-day analysis, the choline + saline group latencies were significantly longer than the control group on day 2 (p < 0.05), day 3 (p < 0.01), and day 4 (p < 0.01). In contrast, administering ABS significantly reversed the longer latencies caused by choline supplements (Fig. 3A). In the probe trial, ANOVA depicted the primary effects of ABS and choline treatment on crossing times [F(1, 20) = 15.838, p < 0.001 for ABS; F(1, 20) = 50.329, p < 0.001 for choline]. Moreover, a significant effect of choline treatment could be observed on time in the target quadrant [F(1, 20) = 10.073, p = 0.004] but no significant impact for ABS [F(1, 20) = 3.795, p = 0.066]. Post-hoc tests revealed that administering ABS significantly reversed the reduced crossings (p < 0.01, Fig. 3B) and time swimming within the target quadrant (p < 0.05, Fig. 3C) induced by choline supplement. No significant differences between the groups could be observed in the swim distance and swim speed (data not shown).

Fig. 3.

Effects of chronic choline exposure on spatial cognitive performance and GSK-3β activity in rats. (A) In the acquisition trials of the Morris water maze test, chronic choline exposure rats showed longer escape latency during training days 2 to 4, while broad-spectrum antibiotics (ABS) administration significantly reversed the longer latencies induced by choline supplement. (B-C) In the probe trial, ABS treatment restored the choline-induced fewer crossing times over the platform position and less time spent in the target quadrant. (AD) Western blotting analysis showing the effects of choline treatment on hippocampal GSK-3β expression, along with plasma TMAO levels. (E) Quantification of western blotting signals of GSK-3β and α-tubulin proteins. (F) Western blotting analysis showing the effects of choline treatment on synaptic protein levels. (G) Quantification of western blotting signals of synapsin Ⅰ, PSD95, and α-tubulin proteins. Data are presented as mean ± SEM (n = 6 per group). * p < 0.05, ** p < 0.01 vs. control + saline group, # p < 0.05, ## p < 0.01 vs. choline + saline group

Effects of chronic TMAO exposure on GSK-3β and synapse-associated proteins

Figure 3D highlighted that the plasma TMAO levels of choline + saline animals were significantly elevated compared with control rats (p < 0.001), followed by their reversal after oral ABS administration (p < 0.001). Western blotting analysis of whole-hippocampal homogenates indicated that choline supplements exhibited no significant effect on the total protein level of GSK-3β. The phosphorylation state of GSK-3β was further examined. It was observed that phosphorylation on the Ser9 residue of GSK-3β was significantly reduced post-choline supplementation compared to the control group (p < 0.01, Figs. 3D and E). In contrast, phosphorylation on the Tyr216 of GSK-3β was not significantly altered (p > 0.05). Furthermore, compared to the choline + saline group, ABS significantly prevented the choline-induced reduction of phosphorylation on Ser9 of GSK-3β (p < 0.05, Figs. 3D and E).

Synaptic plasticity-associated proteins are a key component of the learning and memory machinery within the brain, particularly the post-synaptic density 95 (PSD95) protein, which GSK-3β can phosphorylate. Hippocampal synapse-associated protein analyses revealed a significant reversal of the reduction in the expression of PSD95 and synapsin Ⅰ, induced by choline supplement in the choline + ABS group (p < 0.01 for PSD95, p < 0.01 for synapsin Ⅰ, Figs. 3F and G).

TMAO-induced cognitive impairment and synaptic damage are reversed by GSK-3β inhibition

Rats were supplemented with TMAO and SB216763, a specific GSK-3β chemical inhibitor, to investigate whether GSK-3β is directly associated with the cognitive impairment caused by TMAO exposure. Morris water maze test analysis revealed a significant effect of the day [F(3, 80) = 384.385, p < 0.001], TMAO [F(1, 80) = 56.813, p < 0.001], and SB216763 [F(1, 80) = 29.660, p < 0.001] on latency in acquisition trials to identify the platform. The TMAO group latencies were significantly longer than the control group on day 2 (p < 0.05), day 3 (p < 0.01), and day 4 (p < 0.01). In contrast, SB216763 treatment significantly reversed the longer latencies in the day-by-day analysis due to TMAO (Fig. 4A). In the probe trial, ANOVA highlighted significant effects for TMAO and SB216763 treatment during both crossing times [F(1, 20) = 48.584, p < 0.001 for TMAO; F(1, 20) = 15.289, p < 0.001 for SB216763] and the time in target quadrant [F(1, 20) = 15.037, p = 0.001 for TMAO; F(1, 20) = 4.798, p = 0.041 for SB216763]. Post-hoc tests indicated that SB216763 administration significantly reversed the reduced crossings (p < 0.01, Fig. 4B) and time swimming within the target quadrant (p < 0.01, Fig. 4C) caused by TMAO. Moreover, western-blotting analysis indicated that SB216763 administration significantly suppressed the TMAO-induced decline of both PSD95 and synapsin Ⅰ levels compared with the TMAO group (p < 0.01 for PSD95; p < 0.05 for synapsin Ⅰ; Figs. 4D and E).

Fig. 4.

Influence of GSK-3β inhibition on TMAO-induced cognitive impairment and synaptic damage. (A) SB216763 treatment restored the TMAO-induced longer latencies in the acquisition trials of Morris water maze test. (B-C) In the probe trial, SB216763 treatment restored the TMAO-induced fewer crossing times over the platform position and less time spent in the target quadrant. (D) Western blotting analysis showing the effects of SB216763 treatment on synaptic protein expression, along with plasma TMAO levels. (E) Quantification of Western blotting signals of synapsin Ⅰ, PSD95, and α-tubulin proteins. (F) Western blotting analysis showing the effects of SB216763 treatment on membrane glutamate receptor 1 (M-GluA1) and total GluA1 (T-GluA1) expression. (G) Quantification of Western blotting signals of M-GluA1, T-GluA1, and α-tubulin proteins. Data are presented as mean ± SEM (n = 6 per group). *p < 0.05, ** p < 0.01 vs. control group, # p < 0.05, ## p < 0.01 vs. TMAO group

Previous studies showed that GSK-3β could phosphorylate and destabilize PSD95 to promote AMPA GluA1 receptor subunit internalization, a critical step for long-term depression. Therefore, the TMAO exposure effects on the total cellular and membrane levels of GluA1 were measured. Figure 4F revealed that TMAO administration significantly reduced the membrane level of GluA1 (p < 0.05). In contrast, compared to the control group, it had no significant effect on the total cellular level of GluA1. Furthermore, compared to the TMAO group, SB216763 significantly suppressed the TMAO-induced decline of the membrane level of GluA1 (p < 0.05, Figs. 4F and G).

Discussion

The present study indicated an association between plasma TMAO levels and cognitive function. Plasma TMAO levels were elevated in MCI patients compared to healthy controls, with a significant potential risk association with MCI. Furthermore, chronic exposure to choline revealed a considerable influence on spatial cognitive performance in the Morris water maze task. It reduced the Ser9 phosphorylation of GSK-3β and the synaptic plasticity-related proteins inside the hippocampus, which was restored by inhibiting TMAO using ABS. Notably, the results further revealed that inhibition of GSK-3β by SB216763 significantly suppressed the TMAO-induced synaptic damage, decreased the membrane level of GluA1, and improved hippocampal learning and memory. Therefore, TMAO can induce hippocampal-dependent learning and memory ability impairment with synaptic plasticity deficits by regulating the GSK-3β signaling pathway.

Recent studies have highlighted that plasma TMAO levels are higher in cognitively impaired patients than in cognitively healthy controls. A cross-sectional study involving 253 hospitalized T2DM patients and 150 healthy controls indicated that TMAO levels were higher in the MCI group than in the non-MCI group [25]. Another large-scale epidemiological study involving 1,535 participants identified elevated serum TMAO levels in MCI patients compared with cognitively normal controls [26]. In addition, a retrospective observational study involving 310 transient ischemic attack (TIA) participants indicated that plasma TMAO at baseline was independently related to cognitive impairment at the 3-month follow-up post-TIA [27]. Consistent with previous findings, MCI patients had significantly increased TMAO levels compared to cognitively healthy controls.

Furthermore, TMAO could be a biomarker for cognitive function. Xu R. et al. [28] used a data-driven, integrated computational approach to determine the relationship between microbial metabolites and Alzheimer’s disease development. The study identified TMAO as the metabolite most strongly associated with AD biomarkers among 56 metabolites connected to AD pathology. Vogt et al. [8] demonstrated that increased CSF TMAO correlated with AD pathology biomarkers, such as phosphorylated tau protein, the tau/amyloid-beta (Aβ)-42 ratio, and neuronal degeneration markers like total tau protein and neurofilament light chain. A cross-sectional study with 223 participants revealed that elevated plasma TMAO level was independently related to MCI in a high cardiovascular-risk population after adjusting for potential confounders, such as age, gender, health care service scheme, smoking history, and metabolic syndrome [29] The current study indicated that increased TMAO levels significantly enhanced the MCI risk, validating the hypothesis of plasma TMAO being a predictive biomarker for MCI risk.

On the other hand, several studies have identified a connection between increased TMAO levels and affected cognitive function in animal models. Li et al. [30] described that exogenous TMAO supplementation could aggravate aging-related cognitive dysfunction in SAMR1 and SAMP8 mice. Moreover, chronic TMAO supplements significantly reduced the novel object recognition test performance compared with controls, impairing memory and learning function [31]. Furthermore, Zhou et al. [32] used various hippocampal-dependent spatial references and working memory-related behavioral texts and observed that TMAO treatment induced long-term and short-term memory deficits in mice. The present study revealed that chronic choline supplementation significantly impaired learning and memory in adult rats via elevated TMAO levels. Complementarily, Wang et al. [33] showed that oral gavage of deuterium-labeled choline (d9-choline) in mice resulted in time-dependent detection of d9-TMAO isotopomers in plasma. Importantly, suppressing gut microbiota with oral antibiotics prior to d9-choline administration completely abolished plasma d9-TMAO detection, confirming intestinal flora’s essential role in converting dietary choline to TMAO. Consistent with this mechanism, antibiotic treatment in our study suppressed TMAO generation and rescued choline-induced cognitive deficits in the Morris water maze.

Learning and memory involve sophisticated mechanisms in which the structural basis is formed using synaptic plasticity. Many synaptic plasticities exist, including long-term potentiation formation (LTP), long-term depression (LTD), and homeostatic synaptic plasticity [34]. Growing evidence has depicted that TMAO could impair synapses. Govindarajulu et al. [35] observed that the intracranial injection of TMAO could induce LTP defects. In addition, Zhou et al. [32] described that TMAO treatment reduced the number of synapses and altered the structure of chemical synapses with swollen presynaptic membrane, fewer synaptic vesicles, disappeared or coupled synaptic cleft, and reduced thickness of postsynaptic density in the hippocampus. Herein, choline and TMAO treatments significantly reduced the expression of PSD95 and synapsin Ⅰ. This involves neuronal growth and synapse assembly and regulates learning and memory.

Many protein kinases and protein phosphatases, including protein kinase A (PKA), extracellular-signal-regulated kinase (ERK), c-Jun N-terminal kinase (JNK), phosphatidylinositol-3-kinase (PI3K), protein kinase B (PKB) and GSK-3β, are implicated in learning and memory because protein phosphorylation is the fundamental mechanism in modulating synaptic plasticity [36, 37, 38]. Abnormal GSK-3β signaling has been associated with the pathophysiology of learning and memory deficits 14]. Fortress et al. [39] reports that learning rapidly activates GSK-3β/β-catenin signaling inside the dorsal hippocampus. Moreover, regulating the canonical Wnt/GSK-3β signaling pathway in the hippocampus improves episodic memory and enhances LTP in adult wild-type mice. Furthermore, it also rescues memory loss and enhances synaptic dysfunction in APP/PS1-transgenic mice [40], an Alzheimer’s disease model. The present study results indicated that TMAO exposure affected spatial cognitive performance and reduced the phosphorylation of Ser9 of GSK-3β in the hippocampus. Simultaneously, inhibiting GSK-3β significantly ameliorated the TMAO-induced cognitive deficits, characterizing the essential function of GSK-3β signaling in the TMAO-induced cognitive impairment pathogenesis.

To further determine the potential mechanisms of GSK-3β signaling in regulating TMAO-induced learning and memory deficits, the results described that TMAO significantly reduced the membrane level of GluA1. Simultaneously, inhibiting GSK-3β with SB216763 significantly prevented the TMAO-induced reduction of the GluA1 membrane levels. Emerging evidence indicates that the regulated expression and trafficking of α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid subtype of ionotropic glutamate receptors (AMPARs), particularly GluA1-homomeric subtype of AMPARs, regulates diverse types of synaptic plasticity [41, 42, 43]. GluA1-mediated synaptic plasticity could be central to the early development of AD [44]. GSK-3β could phosphorylate PSD95 on T19 and destabilize PSD95 in the postsynaptic density. This is a critical step for AMPA receptor internalization and LTD [45]. Furthermore, Urbanska et al.‘s [46] study in GSK3β-S9A transgenic mice demonstrated that sustained neuronal GSK3β activation selectively modulated hippocampal GluA1 phosphorylation while maintaining normal expression of GluA2, GluN1, EAAT2, ErbB4, and TrkB. These findings indicate that GSK-3β exerts specific regulatory role over GluA1. According to the previous studies and the current study findings, TMAO reduces the expression of phosphorylated Ser9 GSK-3β in the hippocampus, destabilizes PSD95 in the postsynaptic density, and enhances the internalization of GluA1 receptors. This induces abnormal synaptic plasticity and regulates the occurrence of cognitive dysfunction ( Fig. 5).

Fig. 5.

Global schema illustrating the pathway linking dietary sources of choline in a western diet, gut microbiota and host hepatic FMOs, resultant TMAO production, and subsequent reduction of phosphorylated Ser9 GSK-3β in the hippocampus, destabilization PSD95 in the postsynaptic density and internalization of GluA1 receptors. TMA, trimethylamine; FMOs, flavin monooxygenases; TMAO trimethylamine N-oxide; GSK-3β, glycogen synthase kinase-3 beta; GluA1, glutamate receptor 1; PSD95, post-synaptic density 95

While the data from this study strongly support GSK-3β hyperactivation as an important mechanism for TMAO-induced cognitive deficits, complementary pathways-particularly neuroinflammation, oxidative stress, and altered trophic factors-may contribute to this pathology. Wang et al. [47] demonstrated that elevated TMAO levels increased degenerating Nissl-positive neurons, microglial activation, TNF-α/IL-1β levels, and TLR-4/NF-κB/MyD88 expression in the hippocampal CA3 region. Inhibition of TMAO synthesis also mitigated methamphetamine-induced neuronal damage and neuroinflammation. Furthermore, elevated TMAO levels caused upregulation of ROS, hydrogen peroxide, and lipid peroxidation, resulting in oxidative stress in brain [48, 49]. We expect to identify additional mechanisms underlying TMAO-induced cognitive impairment in future studies.

Some study limitations include a small sample size involving a single center, affecting generalization. Secondly, there was no follow-up information, which could hinder assessing the long-term TMAO effects on cognitive function. Thirdly, this study did not evaluate the ultrastructure of chemical synapses and electrophysiological research, which constrains our understanding of other forms of synaptic plasticity related to TMAO.

In conclusion, the results of the current study indicate that TMAO is elevated in MCI patients and has a significant association with the risk of MCI. Moreover, learning and memory deficits due to TMAO exposure are associated with GSK-3β signaling, linking the downregulation of synaptic plasticity-related proteins and internalizing GluA1 receptors. These preliminary findings may provide new insights into the impact of intestinal microbiota on cognitive dysfunction and facilitate the identification of new therapeutic targets through the modulation of intestinal flora metabolites.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, Guangjun Xi and Jiaojie Hui; methodology, Guangjun Xi, Yachen Shi, Feng Wang and Jiaojie Hui; investigation, Guangjun Xi, Yachen Shi, Pan Wang, Jingyu Deng, Yunuo Chen, Yan Han, Hui Wang and Yang Li; data curation, Guangjun Xi, Yachen Shi and Xiangming Fang; writing-original draft preparation, Guangjun Xi; writing-review and editing, Guangjun Xi and Jiaojie Hui; supervision, Feng Wang; project administration, Guangjun Xi and Yachen Shi; funding acquisition, Guangjun Xi and Feng Wang. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by Medical Young Talents Program of Jiangsu Province (No.QNRC2016191), Wuxi Taihu Lake Talent Plan, Supports for Leading Talents in Medical and Health Profession (2020THRC-DJ-SNW), and Top Talent Support Program for young and middle-aged people of Wuxi Health Committee (No.HB2020016).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All study procedures and experimental protocols were approved by the Ethics Committee of Wuxi People’s Hospital. All participants provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.