Twelve-week male specific-pathogen free (SPF) homozygote db-db mice (C57BLKS/J-leprdb/leprdb, n = 20) and age-matched heterozygote male db-m mice with normal blood glucose (C57BLKS/J-leprdb/+, n = 10) were obtained from Changzhou Cavans Laboratory Animal CO.,Ltd (Jiangsu, China). Heterozygous db-m mice were used as controls for homozygous db-db mice. The justification of male rats was based on the relatively longer distance between the urethral orifice and genitals compared with female rats. The animals were housed individually in cages under a 12:12 h light/dark cycle, with temperature maintained at 22°C ± 1°C, and provided ad libitum access to food and water. All experimental procedures were carried out in compliance with the Guide for the Care and Use of Laboratory Animals. Approval for animal handling protocols was obtained from the Animal Ethics Committee of Hebei Medical University. A period of 2 weeks was allotted for the animals to acclimate to the laboratory environment before commencing the experiment.

To investigate the impact of MEL, db-db mice were randomly assigned to two groups: db-db (db-db + saline solution, n = 10) and db-dbMEL (db-db + melatonin (20 mg/kg), n = 10). MEL powder (MedChemExpress, USA) was dissolved in anhydrous ethanol and then diluted with physiological saline solution to the required concentration. The db-m mice (n = 10) received saline solution as the db-db group. The drugs were administered intraperitoneally once daily for a duration of 6 weeks. Prior to behavioral assessments, mouse fecal samples were collected in sterile cages, immediately frozen in liquid nitrogen, and stored at −80°C. Behavioral evaluations were conducted following the 6-week drug regimen. Subsequently, the mice were anesthetized using sodium pentobarbital. Hippocampal tissues and serum samples were extracted and preserved in a −80°C refrigerator.

Serum samples (n = 5 in each group) were collected for the detection of inflammatory factor concentrations using TNF-α ELISA kit (Proteintech, USA) and IL-18 ELISA kit (Elabscience, China). The test was performed according to the manual instructions.

The brain tissues (n = 5 in each group) were fixed in a 4% paraformaldehyde solution, then embedded in paraffin, and subsequently sliced into 5-μm thick sections of the hippocampus and cerebral cortex. Following this, the sections underwent hematoxylin and eosin staining (HE staining) for 5 min each (Servicebio, China) Nissl staining was carried out with 1% toluidine blue (Servicebio, China), for a duration of 30 min. Post-staining, the sections were washed, dehydrated, and sealed before being examined and imaged using an optical microscope from Zeiss, Germany.

Brain tissue sections embedded in paraffin were deparaffinized using xylene and then hydrated with descending concentrations of alcohol. Subsequently, the sections underwent antigen retrieval with citric acid buffer and were rinsed with PBS. Following treatment with 3% hydrogen peroxide, the sections were incubated overnight at 4°C with antibodies against caspase-3 and IL-1β. Secondary antibodies were then applied for 30 min at room temperature, and visualization was achieved using DAB. The sections were dehydrated with a series of increasing alcohol concentrations and mounted before being imaged at ×20 magnification. Optical density (OD) values were quantified using ImageJ software.

The Morris water maze test is widely recognized as the gold standard for assessing spatial learning and memory in rodents. The experiment was conducted in a circular metal pool measuring 150 cm in diameter and 60 cm in depth, filled with opaque water mixed with skimmed milk powder to obscure a platform placed 1.5 cm below the water surface. The water temperature was maintained at 22°C ± 2°C. The pool was surrounded by distinct visual cues, with the platform positioned at the center of the southwest quadrant, measuring 10 cm in diameter. Mice were released randomly from four starting points (east, west, south, and north) with their heads facing the pool wall. The time taken to find the hidden platform (escape latency) was recorded and analyzed over five consecutive days. Following the removal of the platform on day 6, the time spent in the target quadrant and the number of platform crossings were recorded and analyzed. Behavioral data were captured and analyzed using an automated image acquisition and processing system (SuperMaze software, Shanghai Xinruan Information Technology, Co., Ltd.). Since MWM path varies remarkably, 10 mice in each group were chosen to test avoiding the deviation of results.

The NOR test was performed in a plexiglass white box of dimensions 40 cm × 40 cm × 30 cm. The experiment process was divided into three stages, namely, adaptation, familiarization, and testing. The discrimination index (DI) was used to assess the learning and memory of mice. It was calculated as N/(N + F) × 100%, where N (new) refers to the time to explore the novel object and F (familiar) refers to the time to explore familiar objects.

The fecal bacterial DNA (n = 6 in each group) was extracted for gut 16S rRNA microbiome sequencing and data analysis using the CTAB/SDS method. Invalid data was filtered and excluded by using Flash (Version 1.2.11) and QIIME (Version 1.9.1) software. Then the Clean Tags were obtained for the subsequent analysis. The Clean Tags were clustered as OTU(Operational Taxonomic Units) of a 97% sequence similarity using USEARCH(Version 7.0.1090). The OTUs were classified into various taxonomic levels by comparison with a database. Alpha diversity was computed using Mothur software (Version 1.31.2). Beta diversity was assessed using QIIME (Version 1.80) and visualized through principal coordinate analysis (PCoA). MetaStat and T-tests were employed to detect significantly different gut microbiota species among the three groups at each taxonomic level. The linear discriminant analysis (LDA) effect size, with a LDA score threshold of 2, was utilized to discern significant differences among the groups via the LEfSe software (Version 1.0). According to previous study, the number of samples no less than six is needed to prevent the deviation of standard error.

Fecal samples were mixed with water, centrifuged, derivatized, and analyzed utilizing an ultraperformance liquid chromatography coupled with triplequadrupole tandem mass spectrometry (UPLC-TQ-MS) system (Acquity UPLC-XEVO TQS, Waters Corp., Milford, MA, United States). The results are quantified in micrograms per Gram. One scattered value was removed from the result.

For the metabolomics analysis, fecal samples were thawed slowly at 4°C in a solution containing methanol, acetonitrile, and water in a 2:2:1 ratio. The samples were then vortexed, mixed for 30 s, and incubated at −20°C for 10 min following sonication in ice water for 30 min. Subsequently, the samples were centrifuged at 4°C for 20 min at 14,000 rpm, and the resulting supernatant was collected for analysis.

Analysis was conducted using a UHPLC system (Agilent Technologies) coupled with a quadrupole time-of-flight (AB Sciex TripleTOF 6600). The mobile phase, comprising phase A (25 mM ammonium acetate and 25 mM ammonium hydroxide in water) and phase B (acetonitrile), was employed in both ESI positive and negative modes. The gradient started at 95% B for 0.5 min, decreased linearly to 65% over 6.5 min, further reduced to 40% in 1 min, maintained for 1 min.

The raw MS data were first converted to MzXML files using ProteoWizard MSConvert and subsequently imported into XCMS software for peak analysis. The processed data underwent analysis through principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA) using the R language. The variable importance in the projection (VIP) value of each variable within the OPLS-DA model was computed to assess differential metabolites, with a threshold set at VIP >1 and a p-value <0.05 to determine significance. Signal pathway enrichment analysis was performed using the KEGG pathway database.

The data were presented as the mean ± standard error and analyzed using GraphPad Prism version 8 (GraphPad Software, La Jolla, CA, USA). Two-way repeated measures analysis of variance (ANOVA) was employed to demonstrate the statistical differences of means within each treatment group at the specific time points for the MWM tests. Group variations were assessed using one-way ANOVA followed by Turkey’s multiple comparison tests. All p values <0.05 were considered statistically significant. Spearman’s method was applied to the data for correlation analysis.

The Morris water maze (MWM) tests were conducted to assess the impact of MEL on cognitive impairment in db-db mice. A notable difference in escape latency time was observed in the db-db group starting from day 2 (54.5 ± 3.0 s) to day 5 (49.0 ± 3.8 s) compared with the db-m group (day 2: 42.9 ± 1.7 s; day 5: 24.7 ± 3.2 s, both p < 0.001), which indicated impaired cognitive function in the diabetic mice. Notably, the db-dbMEL group exhibited a significantly decreased escape latency time starting from day 2 (49.2 ± 4.0 s) to day 5 (31.3 ± 1.3 s, p < 0.001) compared with the db-db group (Figure 1B). The statistical analysis comparing crossing and swimming times in the platform quadrant on day 5 is illustrated in Figures 1C, D. Mice in the db-db group exhibited significantly shorter platform crossing times (2.0 ± 0.4) compared to the db-m group (6.5 ± 0.3, p < 0.001). The db-dbMEL group exhibited a significantly higher number of crossing times compared to the db-db group (2.7 ± 0.3, p < 0.05). Likewise, the db-db group demonstrated a significantly lower percentage of time spent swimming in the platform quadrant (53.0% ± 1.0%) in comparison to the db-m group (25.4% ± 0.5%). The db-dbMEL group displayed a significant increase in the percentage of time spent swimming in the platform quadrant (36.0% ± 1.0%) as opposed to the db-db group (p < 0.001). Additionally, as illustrated in Figure 1E, a reduced discrimination index was evident in the db-db group (21.0% ± 1.9%) when contrasted with the db-m group (48.0% ± 2.0%, p < 0.001). Following MEL treatment, the discrimination index in the db-dbMEL group notably increased (38.7% ± 1.7%, p < 0.001). These experimental findings indicate that MEL treatment ameliorated cognitive function decline in db-db mice.

Pathological assessment of the CA3 region of hippocampus and the cerebral cortex was performed by hematoxylin and eosin (H&E) and Nissl staining. As shown in Figure 2, the neuronal morphology in the hippocampus and cortex region of db-m mice was clear and evenly distributed, with large and distinct nucleus. In contrast, the db-db group demonstrated pronounced nuclear pyknosis and indistinct nuclei in CA3 and cortex region, which were notably diminished by MEL treatment. Nissl staining indicated a reduction in Nissl bodies, disrupted cellular structure, and neural impairment in the db-db group in contrast to the db-m group. Following MEL intervention, the count of Nissl bodies was restored.

The study investigated the impact of MEL on inflammatory markers in peripheral blood of db/db mice. Serum IL-18 levels were significantly elevated in the db/db group compared to the db/m group (285.3 ± 62.76 vs. 55.52 ± 10.80 pg/mL, respectively, p < 0.001), but MEL intervention notably decreased IL-18 levels in the db-dbMEL group compared with db-db group (120.3 ± 19.31 pg/mL, p < 0.01). However, serum TNF-α levels demonstrated no difference between db-db group and db-m group (89.56 ± 23.68 vs. 55.33 ± 6.996 pg/mL, respectively, p = 0.0978), MEL significantly reduced TNF-α levels compared with db-db group (29.08 ± 14.57 pg/mL, p < 0.05) (Figures 3I, J).

Immunohistochemical analysis (Figures 3A–H) revealed that the optical density (OD) of caspase-3 protein-positive cells in the hippocampus was lower in db/m mice than db-m mice. In contrast, the db/db group exhibited significantly higher OD of caspase-3-positive cells compared to the db-m group (p < 0.05). Following MEL treatment, the OD of caspase-3-positive cells in the db-dbMEL group decreased. Similarly, the OD of IL-1β positive cells was notably higher in db-db mice than in the db-m group (p < 0.05), with a reduction observed in the db-dbMEL group.

PCA, PLS-DA, and OPLS-DA clustering analysis were performed to assess the fecal metabolites Profiling distinction among the three groups. The PCA score plot (Supplementary Material S1) illustrated that the samples of the QC group gathered tightly, which indicated that the results were credible. The PLS-DA scatter plots (Figures 5A, B) showed that both the db-m and db-dbMEL groups were well separated from the db-db group. All the metabolites identified were classified into 14 classes, and the proportion of each class is illustrated in Figure 5C. As shown in the Volcano plots (Figures 5D, E), compared with the db-m group, the db-db group showed 52 differential upregulated and 67 downregulated metabolites. The db-dbMEL group showed 29 differential upregulated and 16 downregulated metabolites compared with the db-db group. The differential metabolites in positive ion modes were shown in Figures 5F, G. The vast majority of these differential metabolites were lipids and lipid-like molecules, such as Suberic acid, Ganoderic acid, 2-cis-4-trans-abscisic acid, Octanoic acid, Pregnenolone sulfate. These differential metabolites showed a lower level in the db-db group compared to the db-m group and exhibited an elevated trend after MEL treatment.

KEGG pathway enrichment analysis revealed that the HIF-1 signaling pathway and Glycolysis/Gluconeogenesis pathway were significantly upregulated in the db-db group compared with the db-m group (Figure 5H). Four metabolic pathways were identified as significantly different between db-dbMEL and db-db groups, including Pyrimidine metabolism, Rheumatoid arthritis, Vitamin digestion and absorption, and Steroid hormone biosynthesis (Figure 5I).

Since lipids represented the majority of the differential metabolites, examination of the SCFAs levels in the fecal samples among the three groups was carried out (Figure 6A). Acetate, propionate and butyrate are the three dominant SCFAs in the samples. The level of these three SCFAs was reduced in the db-db group, and the MEL treatment reversed the reduction.

The correlations between the key gut microbiota and the biomarkers of DCI including SCFAs, inflammatory factors, glycolysis metabolites, dipeptide, and behavioral indicator were investigated in Figure 6B. Both Dorea and Bacteroides were significantly downregulated while Prevotella significantly upregulated after MEL treatment. Dorea and Bacteroides were negatively correlated with SCFAs, dipeptides, and platform crossing times. Dorea and Bacteroides were positively correlated with inflammatory factors and glycolysis metabolites. In Contrast, Prevotella was positively correlated with SCFAs, Trp-Trp, and crossing times. Prevotella was negatively correlated with TNF-α and pyruvate.

Spearman’s correlation analysis was performed between the genus level microbiota and differential intestinal metabolites. As shown in Figure 6C, Akkermansia, Lactobacillus, and Sutterella were most strongly associated with the vast majority of differential intestinal metabolites between db-db and db-m group. For the differential intestinal metabolites between db-db and db-dbMEL group, Bacteroides, Parabacteroides, and Dorea had a negative correlation with most differential intestinal metabolites (Figure 6D).