Similar brain protection from raw and ripened Pu-erh tea in aging mice through the gut-brain connection

To evaluate the impact of free access to RPT/FPT on the basic physiological signs of mice subjected to D-galactose modeling, we continuously monitored the voluntary water/tea consumption, body weight, and food intake of the mice. As shown in Fig. 1B, significant differences in voluntary fluid intake were observed among the groups. Mice in the NC and AG groups (provided with water) exhibited a higher average daily intake of approximately 7 mL per mouse. In contrast, voluntary intake was significantly lower in the RPT and FPT groups (provided with tea), averaging around 3.0 mL per mouse per day. This difference could be attributed to an inherent preference for water over tea in mice. Notably, the average daily tea consumption did not differ significantly between the RPT and FPT groups. This observation validates the subsequent comparative analysis of their respective anti-aging efficacies, as the potential differential effects can be attributed to the intrinsic properties of the teas rather than variations in intake volume. With respect to body weight and food intake (Fig. 1C, D), all groups showed an overall increase in body weight over the 8-week period, while food intake fluctuated slightly across groups. No statistically significant differences were detected in body weight among the NC, AG, RPT, and FPT groups. Collectively, these findings indicate that D-galactose modeling and free access to RPT or FPT tea did not exert a significant influence on the long-term body weight gain or food consumption patterns of the mice.

To investigate the effects of RPT/FPT on cognitive impairment in D-galactose-induced aging mice, the Y-maze test and Morris water maze test were employed to evaluate short-term spatial memory and long-term spatial learning and memory, respectively. Specifically, the Y-maze spontaneous alternation test assesses mice’s short-term memory by quantifying their tendency to explore novel arms based on their memory of previously visited ones. Thus, mice with cognitive impairment tend to make errors in exploring novel arms (Fig. 1E). The results of the Y-maze spontaneous alternation test are presented in Fig. 1F, G. As shown, there was no statistically significant difference in the total number of arm entries among the four groups, indicating that the experimental interventions did not affect the general locomotor activity or exploratory motivation of the mice. In contrast, significant differences were observed in the spontaneous alternation ratio across groups. Specifically, compared with the NC group, the alternation ratio of the AG group was significantly decreased, suggesting impaired short-term spatial memory in aging mice with cognitive deficits. Notably, intervention with RPT or FPT effectively reversed this reduction in the alternation ratio, with no statistically significant difference observed between the RPT and FPT groups. This suggests that both RPT and FPT exert potential protective effects on short-term spatial memory in D-galactose-induced aging mice, with comparable efficacy. For the Morris water maze test, to assess long-term spatial learning and memory, we analyzed key behavioral parameters, including the path length to the platform (Fig. 1H, M), average swimming speed (Fig. 1I), escape latency (Fig. 1J), times crossing the original platform (Fig. 1K), and time spent in the target quadrant (Fig. 1L). As shown, there was no significant difference in average swimming speed among the groups, indicating that experimental interventions didn’t affect mice’s swimming ability. Compared with the NC group, the AG group had a significantly longer path length to the platform and escape latency, and shorter time spent in the target quadrant and fewer times crossing the original platform, suggesting impaired long-term spatial learning and memory in D-galactose-induced aging mice. Notably, RPT and FPT interventions effectively ameliorated these deficits. Both groups showed reduced path length to the platform and escape latency, alongside increased time in the target quadrant and platform crossing times, with all indicators approaching the levels of the NC group. Surprisingly, there were no significant differences in any of the water maze indicators between the RPT group and the FPT group, indicating comparable efficacy of the two interventions in restoring long-term spatial cognitive function. Worthy of note, although previous studies linked increased target quadrant time to memory recovery, our study found decreased time in both the normal control and tea-treated groups¹³. We hypothesize this relates to optimized spatial search strategies—normal mice shift from “edge search” to “focused search” after platform removal, while aging mice show random search. RPT and FPT interventions restored efficient search, reflected by reduced target quadrant time. Visual evidence from the mice’s platform crossing frequency trajectories provides support for this hypothesis (Fig. 1N). Collectively, results from the Y-maze and Morris water maze tests demonstrate that both RPT and FPT effectively ameliorate short-term and long-term spatial cognitive impairments in D-galactose-induced aging mice, with no significant difference in their protective efficacy.

Building upon the behavioral findings that D-galactose and RPT/FPT interventions impacted mice’s spatial learning and memory, and given the intimate link between the brain (particularly the hippocampus) and spatial cognitive abilities¹⁴, as well as the established mechanism where D-galactose-induced modeling drives oxidative stress through heightened metabolism and excessive free radicals, thereby triggering pathological alterations¹⁵, we proceeded to explore the effects of RPT/FPT on brain oxidative stress and hippocampal histopathological changes in AG mice.

As shown in Fig. 2A, representative hematoxylin and eosin (H&E) staining images revealed distinct histopathological changes in the hippocampus across groups. In the NC group, the hippocampal tissue structure was intact, with neatly arranged cells and clear boundaries. In contrast, the AG group exhibited obvious hippocampal damage, characterized by blurred contours of the dentate gyrus, obscured serrated structures, disordered arrangement of granular cell layer cells with disrupted normal single-layered and orderly pattern, loss of cell polarity, and ectopic migration of some cells to the molecular layer. A substantial reduction in the number of neurons was observed in the polymorphic layer, along with cell body shrinkage, nuclear pyknosis, and hyperchromasia, which are typical morphological features of apoptosis¹⁶. Additionally, there was an increase in vacuolation of cells in the principal cell layer and enlarged intercellular spaces. Notably, RPT and FPT interventions alleviated these pathological alterations; the hippocampal tissue in both RPT and FPT groups showed improved cell arrangement and reduced degeneration, approaching the histological features of the NC group. Regarding brain oxidative stress indicators, as depicted in Fig.2B–D, no statistically significant difference in SOD activity was observed among the four groups, indicating that neither D-galactose-induced aging nor RPT/FPT intervention affected the activity of this key antioxidant enzyme. For GSH, D-galactose modeling did not alter its brain levels, while FPT intervention significantly elevated GSH concentration compared to the AG group. As for MDA concentration, a marker of lipid peroxidation, the AG group displayed a significant increase compared to the NC group (P < 0.05). Both RPT and FPT treatments reduced MDA concentration, with values closer to those of the NC group, though the regulatory effect of FPT on MDA was slightly more pronounced (no statistical difference from RPT). Collectively, these results demonstrated that RPT and FPT could alleviate D-galactose-induced brain oxidative damage in AG mice, and FPT specifically exerted a superior regulatory effect on GSH, which might contribute to the improvement of hippocampal histopathological damage.

To investigate the influence of tea-gut microbiota interactions on the gut microbiota features that underlie the backgrounds in AG mice, we performed 16S rRNA sequencing on collected mouse feces and compared the gut microbial characteristics in AG mice. The Shannon and PD-tree indexes are used to assess the species evenness and lineage diversity of the microbial communities in this study. As shown in Fig. 3A, B, the Shannon and PD-tree indices of intestinal microbiota were significantly reduced in D-galactose-induced (intraperitoneal injection) AG mice compared to normal mice (P < 0.05). Notably, these two diversity indices increased markedly following RPT and FPT intervention (P < 0.05), indicating effective restoration of gut dysbiosis. Since all mice received the same diet, dietary interference was excluded. These findings confirm that cognitive impairment pathology is connected with gut dysbiosis and that RPT/FPT intervention distinctly ameliorates this imbalance. To further visualize differences in the overall structure of intestinal microbial communities among groups, we performed UPGMA clustering analysis (Fig. 3C). Results showed that AG mice clustered tightly into a separate branch, whereas samples from the RPT and FPT intervention groups clustered interspersedly with normal mice. These findings indicate that the pathological state of cognitive impairment accompanies a characteristic shift in the intestinal microbial community, and RPT/FPT intervention effectively reverses this shift, restoring the overall microbial structure to a normal state. Principal coordinate analysis (PCoA) based on Bray-Curtis distances further confirmed the inter-group separation pattern of the gut microbial communities. Along the PCo1 axis, the AG group was significantly separated from all other groups, while the tea intervention groups (RPT and FPT) showed substantial overlap with the normal control group along both PCo1 and PCo2, with no distinct boundaries between them. This visual separation pattern was supported by statistical testing: permutational multivariate analysis of variance (PERMANOVA) confirmed highly significant differences in community structure among groups (R² = 0.313, P = 0.02), and a test for homogeneity of multivariate dispersions (PERMDISP) verified that within-group dispersion was homogeneous across groups (P = 0.0599), thereby ruling out differences in within-group variation as the driver of the observed pattern (Fig. S1A, B). Consistent with UPGMA clustering results, these findings verify that cognitive impairment accompanies characteristic differentiation of the microbial community, and that RPT/FPT intervention effectively reverses this trend.

To dissect the structural remodeling of gut microbiota induced by RPT/FPT intervention in AG mice, we analyzed differentially abundant bacterial taxa across experimental groups. Gut microbial profiles at the phylum, family, and genus taxonomic levels are illustrated in Fig., which revealed that mice with cognitive impairment exhibit taxonomic-specific dysbiosis. Amplicon Sequence Variants (ASVs), clustered with 100% sequence similarity, offer representation of unique, naturally occurring biological sequence variants. To avoid “information blurring” inherent to higher taxonomic levels, we performed LEfSe analysis at the ASV level. First, a Venn diagram identified 718 ASVs common to all groups—these stably occurring taxa allowed exclusion of random “presence-absence” variations stemming from technical bias or sample heterogeneity. Second, LEfSe analysis (LDA > 3) on these 718 shared ASVs identified 47 ASVs as differentially abundant across the NC, AG, RPT, and FPT groups (Fig.). However, some of the predominant ASVs were classified into the same cluster, suggesting that the scope of key microorganisms should be further defined. S1C–E S2

As previously noted, the AG group was significantly separated from all other groups along PCo1. We thus conducted correlation analysis between the 718 shared ASVs and PCo1 scores of all samples, identifying 141 ASVs that may contribute to the separation between the AG group and other intervention groups. To narrow down the scope of key microorganisms (as prompted by the clustering of some predominant ASVs), we took the intersection of these 47 differential ASVs and 141 separation-related ASVs, ultimately obtaining 43 common ASVs. LEfSe analysis results for these refined core ASVs are depicted in Fig. 3G, which clearly delineates taxonomic biomarkers that distinguish the NC, AG, RPT, and FPT groups. Specifically, the AG group was characterized by significant enrichment of Bacteroides acidifaciens, along with the genera Bacteroides, Thomasclavelia, Parabacteroides, and Massiliomicrobiota. In contrast, the RPT group showed a distinct taxonomic profile, with enrichment of Alistipes, Lachnospiraceae_NK4A136_group, Roseburia, and Eubacterium_siraeum_group. The FPT intervention drove enrichment of multiple taxa, including Lactobacillus_johnsonii, the genus Lactobacillus, Lachnospiraceae_NK4A136_group, Candidatus_Saccharimonas, Alistipes, and UCG-003. For the NC group, the enriched taxa comprised Lachnospiraceae_NK4A136_group, Odoribacter, Alistipes, Rikenella, Rikenellaceae_RC9_gut_group, and Candidatus_Saccharimonas. Notably, several core ASVs enriched in the RPT or FPT groups clustered within the same taxonomic lineages as those characteristics of the NC group, reflecting a clear convergence toward normal gut microbiota. For example, Lachnospiraceae_NK4A136_group, Alistipes were consistently enriched across the RPT, FPT, and NC groups (with significantly lower abundance in the AG group), forming a shared biomarker of microbial homeostasis. These LEfSe-derived taxonomic signatures directly support the findings from PD-tree and PCoA analyses. Collectively, this taxonomic and phylogenetic convergence reinforces the conclusion that RPT/FPT interventions effectively reverse AG-induced gut dysbiosis by reinstating NC-like microbial profiles.

The bidirectional communication between the gut microbiota and the central nervous system could be mediated by circulating metabolites. Based on this mechanism, the present study analyzed serum metabolites to systematically characterize the global alterations in metabolic profiles of aged mice following RPT and FPT intervention. Firstly, orthogonal partial least squares discriminant analysis (OPLS-DA) was employed to evaluate the overall separation of serum metabolite profiles between groups. As depicted in Fig. S3A–C, clear separations were observed in OPLS-DA score plots for NC vs AG, AG vs RPT, and AG vs FPT comparisons, with permutation tests validating model reliability. These results indicated that RPT and FPT interventions reshaped the serum metabolite landscape. Subsequently, volcano plots were generated to identify differentially abundant metabolites (DAMs) between groups with the criteria of P < 0.05 and VIP > 1¹⁷. As shown in Fig. 4A–C, compared with the NC group, the AG group had 15 downregulated and 74 upregulated DAMs. For the AG vs RPT comparison, there were 57 downregulated and 64 upregulated DAMs in the RPT group relative to the AG group. In the AG vs FPT comparison, the FPT group showed 37 downregulated and 49 upregulated DAMs compared to the AG group. Furthermore, the KEGG pathway classification histogram revealed that RPT and FPT interventions significantly increased the involvement in lipid metabolism pathways (Fig. S3D). To validate this, we analyzed the classification of DAMs across groups (Supplementary Tables 1 and 2). As shown in Fig. 4D–F, in the AG vs NC comparison, Lipids and lipid-like molecules accounted for 27.0% of DAMs. For AG vs RPT, this category made up 50.8% of DAMs, and in AG vs FPT, it constituted 64.7% of DAMs. These results indicated that RPT and FPT interventions primarily regulated serum lipid metabolites, which might serve as crucial mediators in gut-brain axis communication and cognitive function modulation. To further pinpoint underlying gut-brain axis regulation (and exclude confounding signals from exogenous metabolites, such as dietary residues or environmental contaminants that are not inherently involved in host-microbiota crosstalk), we first excluded non-endogenous differential metabolites by using the Human Metabolome Database. Then, we examined the KEGG MetPA results for AG vs NC (Fig. 4G), AG vs RPT (Fig. 4I), and AG vs FPT (Fig. 4J). Notably, sphingolipid metabolism (ko00600) was the only pathway consistently identified across all three comparisons. This convergence suggests that RPT/FPT may act by restoring sphingolipid metabolism, which in turn modulates host systemic metabolism.

Building on serum metabolomic findings, which highlighted the potential role of sphingolipid metabolism in RPT/FPT-induced cognitive improvement. We sought to gain functional insights at the microbial level that could complement the serum metabolomic findings implicating sphingolipid metabolism. A predictive analysis of gut microbiota functional potential using PICRUSt2 was performed. This in silico prediction based on taxonomic profiles offered a complementary perspective, indicating that pathways including microbially mediated sphingolipid metabolism could potentially be modulated by the RPT and FPT interventions (Fig. S4). However, critical knowledge gaps remain to solidify this mechanistic link. Firstly, it remains unclear whether the microbial taxa driving sphingolipid metabolism restoration are the 43 key ASVs previously identified via LEfSe analysis and PCo1 differentiation. While PICRUSt2 predicted an overall recovery of microbial sphingolipid metabolism capacity, it failed to pinpoint which core taxa—closely associated with RPT/FPT-induced microbiota structural normalization—functionally underpin this restoration. In addition, serum sphingolipid alterations (reflecting systemic metabolic status) do not directly indicate regulatory effects on brain function. Despite emerging evidence indicating that peripheral sphingolipids can traverse the blood–brain barrier, whereupon they contribute to the activity of amyloid-β, it’s not clear whether these circulating metabolites are assumed to directly cross the blood–brain barrier or to indirectly influence brain pathology. Thus, it is essential to further verify the link between peripheral sphingolipid metabolism changes and brain cognitive recovery mechanisms.

To address the first knowledge gap, we performed correlation analysis between 43 key ASVs and sphingolipid metabolism-related differential metabolites. As depicted in Fig.5, a total of 29 ASVs related to sphingolipid metabolism were identified through the correlations with the differential metabolites from AG vs RPT (N-palmitoyl-d-sphingosine, DL-serine, Phytosphingosine) and AG vs FPT (Sphingomyelin (d18:1,18:0), N-[1,3-dihydroxyoctadec-4-en-2-yl]tetracos-15-enamide, N-tetracosenoyl-4-sphingenine). Notably, three metabolites, N-palmitoyl-d-sphingosine, DL-serine, and Sphingomyelin (d18:1,18:0), were consistently correlated with key ASVs across all four groups. Importantly, their correlation patterns were aligned between the RPT/FPT intervention groups and the NC group, while showing an opposite trend in the AG group. Furthermore, we annotated the taxonomic clusters of these 29 sphingolipid-related ASVs and found that Lachnospiraceae_NK4A136_group and Alistipes were consistently enriched across the NC, RPT, and FPT groups, which aligned with our earlier observations. In the prior taxonomic profiling, these two taxa were identified as shared biomarkers of microbial homeostasis, being significantly depleted in the AG group but restored to NC-like levels following RPT/FPT intervention. This striking consistency between structural and functional connections strongly suggests that these two core taxa are the key microbial drivers underlying RPT/FPT-mediated sphingolipid metabolism recovery.

Previous studies have confirmed disrupted sphingolipid metabolism in Alzheimer’s disease (AD) patients, characterized by elevated serum ceramide levels accompanied by astrocyte activation. Activated astrocytes enhance neuronal susceptibility to Aβ, and both ceramides and Aβ have been demonstrated to participate in binding processes with these cells, collectively driving disease progression¹⁸. To directly validate whether this mechanism—mediated by peripheral sphingolipid metabolic changes—can influence core pathologies of the central nervous system (addressing the second knowledge gap), we subsequently performed Aβ immunofluorescence staining in the hippocampal region and quantified cerebral ceramide levels in brain tissues. The results indicated that the Aβ immunofluorescence signal was conspicuously enhanced in the brains (especially in the hippocampal region) of AG group mice compared to the NC group, while interventions with both RPT and FPT effectively restored the signal intensity to a level comparable to that of the NC group (Fig. 6A–C). Meanwhile, we observed that the brain ceramide level exhibited a parallel trend to the Aβ immunofluorescence signal. Specifically, the AG group showed a marked elevation in cerebral ceramide content relative to the NC group, whereas both RPT and FPT interventions significantly reduced ceramide accumulation to levels indistinguishable from those in normal controls (Fig. 6D).

Collectively, these results associated the restoration of specific gut microbiota (Lachnospiraceae_NK4A136_group and Alistipes) with concurrent amelioration of hippocampal sphingolipid-related pathologies (including Aβ deposition and ceramide levels) in AG mice. These coordinated changes suggest a potential link between the gut microbiota and hippocampal sphingolipid metabolism that might underpin the cognitive benefits of RPT and FPT interventions.

A comparison of the physicochemical compositions between raw Pu-erh tea and ripened Pu-erh tea is pivotal for revealing the link between microbial fermentation and tea bioactivity, and for advancing insights into tea’s functional properties. Therefore, we performed metabolic profiling of RPT and FPT infusions, coupled with quantitative analysis of major chemical constituents, to identify fermentation-mediated changes in their bioactive component profiles.

Firstly, a total of 102 metabolites were identified in the tea extracts via UPLC-Orbitrap-MS analysis, with identification based on comparisons of retention times, mass spectra, and RPT/FPT representative ion chromatograms, plus matching to metabolic databases. Applying the screening threshold (VIP > 1, |log₂FC| ≥ 1, P < 0.05)¹⁹, 28 differential compounds were filtered out, and their details are provided in Table 1. Overall, the pile fermentation triggered profound alterations in the metabolite profile. These 28 differential metabolites spanned multiple classes, including organic acids, phenol acids and their derivatives, amino acids and their derivatives, nucleobases, sugars and sugar derivatives, etc. A core trend of these changes was the marked upregulation of phenol acids, organic acids, and their derivatives, represented by 3,5-Dihydroxy-4-(sulfooxy)benzoic acid, Gallic acid, and D-Tartaric acid. This upregulation might be attributed to microbial enzyme-driven transformations during fermentation²⁰, such as the hydrolysis of ester-type catechins to release gallic acid²¹ and the biosynthesis of novel organic acids via microbial central metabolic pathways²². Given the well-documented antioxidant, anti-inflammatory, and neuroprotective properties of phenol acids and specific organic acids^23,24, their accumulation likely enhances the bioactive potential of FPT. Conversely, several metabolites showed significant downregulation, including amino acids (L-Theanine, L-Valine, Pipecolic acid, L-Pyroglutamic acid), central metabolic organic acids (Citric acid, Malic acid), and sugar acids (D-Gluconic acid). This decrease is presumably due to these compounds being utilized as carbon and nitrogen sources by fermenting microorganisms to support their growth and metabolic activities²⁵. It is worth noting that several of these downregulated metabolites, such as L-Theanine, are well-recognized neuroprotective constituents of tea²⁶, which might exert specific physiological functions linked to cognitive function regulation in RPT.

Secondly, quantitative analysis was performed on the bioactive substances of tea. As presented in Table 2, pile fermentation profoundly reshapes the functional component profiles of RPT and FPT. The core differences and underlying mechanisms are elaborated as follows. In terms of active components, multiple key substances in FPT exhibit significant reductions: free amino acids, tea polyphenols, and L-theanine all decrease to varying degrees. Among catechins, EGC, EGCG, GCG, and ECG are drastically depleted to levels below the instrument detection limit (nearly undetectable); EC remains detectable but with a marked content reduction, resulting in a substantial decline in total catechins. Regarding tea pigments, theaflavin content shows no significant fluctuation compared with RPT, while thearubigin content decreases remarkably, and theabrownin content increases sharply. Concurrently, the contents of theobromine, caffeine, and gallic acid in FPT display distinct upward trends. The variation trends of partial components are consistent with the aforementioned UPLC-Orbitrap-MS analysis results, verifying the reliability of the data. Collectively, these changes in functional components provide a basis for further exploring the pharmacological effects related to RPT and FPT.

To clarify the key bioactive components underlying the cognitive-improving effects of RPT and FPT in D-galactose-induced AG mice, we conducted a targeted correlation analysis between tea chemical constituents and cognitive function/mechanistic indicators. A critical limitation of relying solely on RPT and FPT is that constituent changes driven by fermentation may accidentally align with trends in physiological indicators (trend-based spurious correlations), rather than reflecting true functional relevance, which could lead to misidentifying non-core components.

To address this limitation and boost the reliability of core component screening, we incorporated data from two additional aged raw Pu-erh tea samples (RPT-2004 and RPT-2014, which, together with RPT, were processed and manufactured by the same company) processed in the same animal experimental batch. Our goal was not to evaluate the cognitive effects of tea aging itself. Instead, we used these aged samples as an independent validation cohort to strengthen our correlation analysis. By doing so, we effectively filtered out non-specific, trend-driven associations, allowing us to confidently identify and confirm the components truly linked to the cognitive benefits of RPT and FPT (the effective intake doses of these key components in the mouse experiment are detailed in Table 3). As shown in Fig. 7. Through Spearman’s correlation analysis, we found that several tea chemical constituents exhibited notable correlations with cognitive function and mechanistic indicators. For RPT, components like 6-O-Galloyl-glucose, L-Valine, Pipecolic acid, and total catechins showed significant correlations with cognition-related indicators, strongly suggesting their crucial roles in improving cognitive performance. Specifically, these RPT-related functional components were correlated with key ASVs, which were classified within the genus Lachnospiraceae_NK4A136_group and Alistipes — both of which are the key microbial taxa we previously identified as being involved in cognitive impairment improvement. In contrast, for FPT, compounds including thearubigins, theobromine, theaflavins, gallic acid, and theabrownins were significantly correlated with relevant cognitive and mechanistic markers, indicating their potential as functional components underlying its cognitive-improving effects. These FPT-associated functional compounds were also found to correlate with ASVs that fell into Alistipes, another key microbial genus we identified for alleviating cognitive impairment. Collectively, these findings highlight that specific tea components from RPT and FPT, by interacting with these previously confirmed key microbial taxa, may exert prominent effects on cognitive improvement.

It should be noted that total catechins—comprising individual monomers (EGC, EC, EGCG, GCG, and ECG)—showed a significant positive correlation with cognitive markers, whereas none of these individual catechins did so independently. However, this finding does not mean the individual monomers are inactive. Two plausible mechanisms might account for this discrepancy. Firstly, a bioactive concentration threshold effect is likely involved. Each catechin monomer in RPT exists at a concentration too low to reach the threshold required for a detectable correlation with cognitive markers²⁷. However, their cumulative content as total catechins surpasses this threshold, thereby enabling a significant association with cognitive improvement. Secondly, coordinative or synergistic interactions between the monomers may drive the outcome^28,29. Each catechin may exert a weak but real regulatory effect on cognitive-related pathways. When acting together, these weak individual effects are amplified to a magnitude that becomes statistically significant in correlation with cognitive markers.

In sum, RPT and FPT rely on distinct sets of core components to exert their cognitive-improving effects: RPT depends on 6-O-Galloyl-glucose, L-Valine, Pipecolic acid, and total catechins (with the latter’s efficacy driven by either concentration threshold or monomer synergy), while FPT relies on tea pigments (thearubigins, theaflavins, theabrownins), alkaloids (theobromine), and gallic acid. This finding not only clarifies the bioactive basis of RPT/FPT’s cognitive benefits but also highlights the importance of considering both individual components and their collective effects in natural product research.

D-galactose was purchased from Sigma-Aldrich (St. Louis, MO, USA). D-Malondialdehyde (MDA), glutathione (GSH), and superoxide dismutase (SOD) commercial kits were obtained from Nanjing Jiancheng Bioengineering Institute (Nanjing, China), and the mouse ceramide ELISA kit was from Shanghai Enzyme-linked Biotechnology Co., Ltd. (Shanghai, China). All kits are measured in strict accordance with the instructions.

The fresh tea leaves (Camellia sinensis var. assamica) were sourced from Xiaguan, Yunnan, China, and both raw Pu-erh tea (RPT) and ripened Pu-erh tea (FPT) were produced using sun-dried raw tea (SRT) as the base material, with manufacturing procedures referring to the Chinese national standard (GB/T 9833.5-2013) with slight modifications: the fresh leaves were first processed via spreading, enzyme inactivation, rolling, and ball-breaking, followed by sun-drying to produce the base SRT; for RPT, the SRT was directly steamed, compressed and dried into the final product, whereas for FPT, the SRT was subjected to pile fermentation (moistened with 30–40% water), after which the fermented tea was steamed and compressed then dried to obtain the final FPT product.

Based on previous analysis of tea concentration in daily scenarios, tea infusion concentration was selected for the intervention. The tea infusion was prepared with slight modifications according to the sensory evaluation method for tea (GB/T 23776—2018). Specifically, tea leaves were placed in a beaker and steeped in boiling water for 5 miutes at a ratio of tea to distilled water of 1 g: 50 ml. The tea infusion was collected after two rounds of filtration and stored at −80 °C until subsequent experiments. Before use, the tea infusion was thawed to ensure stability.

Thirty-two 8-week-old male specific-pathogen-free Institute of Cancer Research mice were purchased from SJA Laboratory Animal Co., Ltd. (Hunan, China) for this study. All mice were housed in a standardized environment maintained at 23 ± 2 °C with 50 ± 10% relative humidity and a 12 h light/dark cycle, and were given free access to an identical standard maintenance diet (SY10001F, Shuyu Biotechnology Co., Ltd., Shanghai, China) and drinking water throughout the experiment.

Following a 7-day acclimation period to minimize stress, the mice were stratified by body weight and then randomly assigned to four groups (n = 8 per group): normal control (NC), aged control (AG), RPT intervention (RPT), and FPT intervention (FPT). The NC group received a daily subcutaneous injection of 0.3 mL 0.9% physiological saline and had free access to distilled water. In contrast, the AG, RPT, and FPT groups were subcutaneously injected with D-galactose at a dose of 150 mg/kg body weight daily to induce an aging model. Among these three groups, the AG group was provided with distilled water, while the RPT and FPT groups were given RPT infusion and FPT infusion, respectively, as their exclusive drinking fluids (ad libitum).

During the 8-week experimental period, daily recordings were made of the mice’s voluntary intake of water or tea infusions, and weekly measurements were taken of their body weight and food intake. After 7 weeks of intervention, the Y-maze and Morris water maze test were conducted to evaluate spatial learning and memory. At the end of the 8th week, all mice were fasted overnight, anesthetized via intraperitoneal injection of pentobarbital sodium (50 mg/kg), and euthanized by cervical dislocation. Brain tissue, cecal contents, and serum samples were immediately collected from each mouse and stored at −80 °C for subsequent laboratory analyses (Fig. 1A). Data acquisition and subsequent analysis were conducted in a single-blind manner.

All animal experiments were conducted in strict accordance with the Chinese Laboratory Animal Management regulations, adhering to the guidelines for the Care and Use of Laboratory Animals, and with the approval granted by the Laboratory Animal Welfare and Ethics Committee of Hunan Agricultural University (Certificate No. 2021-2137: Changsha, Hunan, China).

Hippocampal tissues were first fixed in 4% paraformaldehyde infusion for 24 h to preserve their structural integrity. After fixation, the tissues underwent a standard histological processing workflow: sequential dehydration through graded ethanol infusions, clearing in xylene, and embedding in paraffin wax to form solid tissue blocks. These paraffin blocks were then sectioned into thin slices (3–4 μm thick) using a microtome to ensure optimal visualization of cellular details. The resulting paraffin sections were deparaffinized and rehydrated before being stained with H&E. Stained sections were finally observed and imaged under a light microscope (Nikon Eclipse Ci, Nikon, Tokyo, Japan) equipped with a digital camera, allowing for detailed examination of hippocampal tissue morphology.

Mouse brain paraffin sections were dewaxed, rehydrated, and subjected to antigen retrieval. In brief, after heat-induced antigen retrieval using citrate antigen retrieval buffer (pH 6.0), the sections were blocked with 10% goat serum (PN0038, Pinuofei Bio) at 37 °C for 30 min. Subsequently, the sections were incubated overnight at 4 °C with the primary antibody diluted in antibody diluent (PN1023, Pinuofei Bio). The primary antibody used in this study was rabbit anti-β-amyloid (Aβ) (GB111197-100, Saiveier Bio, 1:100 dilution). The following day, after washing with PBS (PN0031, Pinuofei Bio), the sections were incubated with Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody (PN0108, Pinuofei Bio, diluted according to the manufacturer’s instructions) at 37 °C for 1 h in the dark. Finally, cell nuclei were counterstained with DAPI (PN0015, Pinuofei Bio), and the sections were mounted with an anti-fade mounting medium (PN0024, Pinuofei Bio). All fluorescence images were captured using an inverted fluorescence microscope (Nikon, Tokyo, Japan). The fluorescence intensity of Aβ was quantified using Image-Pro Plus 6.0 software (Media Cybernetics, MD, USA).

The Y-maze spontaneous alternation test was employed to assess short-term spatial memory in mice, capitalizing on the inherent propensity of rodents to prioritize exploration of novel arms over those they had previously visited. The Y-maze apparatus comprises three identical arms intersecting at a 120° angle at the central hub. For the experimental procedure, each mouse was first placed at the center of the three-arm intersection and permitted 5 min of free exploration to acclimate to the maze environment. Following a 1-h interval, the formal test was initiated: the mouse was repositioned at the central intersection and allowed to explore the maze freely for 8 min. The spontaneous alternation ratio was quantified using the formula: The spontaneous alternation rate = Alternation number/(total number of arms entered − 2), where “alternations” refer to consecutive entries into three distinct arms during the test period.

The Morris water maze test was conducted to evaluate spatial learning and memory in mice, based on the principle that rodents avoid water and will actively learn to locate a hidden escape platform using spatial cues placed around the maze. The apparatus consisted of a circular pool, photosensors, and a computerized video-tracking module. One day before the formal test, all mice underwent acclimation: they were placed on a visible platform in an undyed pool and allowed to stay for 60 s to adapt to the water environment and recognize the platform. The formal test, conducted following the protocol described by published paper¹³, included two phases: the hidden platform phase (Day 6) and the probe test phase (Day 7). For the hidden platform phase, the pool water was dyed white to make the platform invisible; mice were randomly placed into the pool to search for the platform, with parameters including escape latency, average swimming speed, time spent in the target quadrant, and number of crossings over the original platform recorded. For the probe test phase, the platform was removed, and each mouse was allowed to search the pool freely for 60 s, with the same parameters (number of crossings over the former platform area, average swimming speed, time spent in the target quadrant) recorded to evaluate consolidated spatial memory.

The 16S rRNA gene sequencing for mice cecal contents was performed by Guangzhou Gene Denovo Biotechnology Co., LTD., with detailed procedures referred to the published study¹¹.

Serum nontargeted metabolomics analysis was conducted by Guangzhou Gene Denovo Biotechnology Co., LTD., with detailed procedures referred to the published study⁵².

To explore the regulatory effects of pile fermentation—the key differential process between raw and ripened Pu-erh tea—on the metabolic profile of tea infusions, we conducted quantitative analysis on the main functional components with neuroprotective effects in tea. Specifically, the contents of free amino acids and tea polyphenols were quantified according to the method described by Zhou et al.⁵³; for specific components—including theabrownins, theobromine, caffeine, epigallocatechin (EGC), epicatechin (EC), epigallocatechin gallate (EGCG), (−)-gallocatechin (GCG), and epicatechin gallate (ECG)—their quantitative determination followed the protocol reported by Gong et al.⁵⁴; Additionally, the contents of theabrownins, theaflavins, and thearubigins were quantified using the method established by Wang et al.⁵⁵. To further clarify how fermentation drives metabolic variations, an ultra-high-performance liquid chromatography-Orbitrap-mass spectrometry (UPLC-Orbitrap-MS) system was utilized for qualitative and semi-quantitative analysis of chemical components, with the entire experimental operation carried out in strict accordance with the technical protocol described in the published study⁵².

In the present study, one-way analysis of variance along with Fisher’s LSD test was used to undertake the statistical analysis of the column chart data with the aid of GraphPad Prism 9. Data were presented as the mean ± standard error of the mean (x®± SEM), and values with different letters represent statistically significant results. Online platforms OmicShare, Omicsmart (https://www.omicshare.com/tools↗, https://www.omicshare.com↗) were used to analyze the 16 rRNA gene sequencing and serum nontargeted metabolomics data. The correlations were analyzed using Spearman’s correlation method through IBM SPSS Statistics 25, and P < 0.05 was considered significant.

The datasets produced from the study are available from the corresponding authors on reasonable request.