The complete YMP was collected and extracted using 95% EtOH in a reflux setup three times, lasting two hours each time. To create the 95% ethanolic extracts, the solvent was evaporated using a vacuum rotary evaporator. 50 mg of YMP powder were dissolved in 1.2 mL of 70% methanol in a centrifuge tube. Every 30 minutes, the mixture was vortexed six times for 30 seconds each. A 0.22 µm microporous membrane was used to filter the clear supernatant after the tube was centrifuged for three minutes at 12,000 rpm. After filtering, the solution was collected and put in an autosampler vial for a further qualitative UPLC-MS/MS analysis. The chromatographic separation was carried out using a SHIMADZU Nexera X2 UPLC system equipped with an Agilent SB C18 column (1.8 µm, 2.1 mm×100 mm). Ultra-pure water was used as mobile phase A, while chromatographic acetonitrile was used as mobile phase B. Both phases contained 0.1% formic acid. The gradient elution conditions were as follows: Over nine minutes, B increased from 5% to 95%, held for one minute, then dropped to 5% between 10 and 11 minutes, held for another minute. The column temperature was kept at 40°C, the injection volume was 4 μl, and the flow rate was set at 0.35 ml/min.

Fifty (20 ± 2 g) female mice were supplied by Beijing HFK Bioscience Co., Ltd. (Beijing, China) and kept in cages with free access to food and water, 12 hours of light and dark cycles, and 25 ± 1°C and 55 ± 5% relative humidity. The animal protocol was reviewed and approved by the Nantong University Animal Ethical and Welfare Committee (R240411744), and on April 25, 2024, the animal ethics and welfare were approved using Approve No. S20240425-006. After a week of acclimatization, 10 mice were randomly chosen to form the control group (CON), which was fed a regular diet, while the remaining mice were fed a high-fat (45%) diet (HFD). The ingredient ratio of the high-fat diet is: fat 45%, carbohydrates 20%, protein 20%, vitamins and minerals 15%. Two YMP-treated groups (40 and 160 mg/kg/d; known as the YMPL and YMPH groups, respectively) of HFD mice received YMP injections. The dosage of YMP was determined by our previous pre-experiment. After four weeks, the CON mice received injections of sodium citrate buffer, while the HFD mice received intraperitoneal injections of streptozotocin (STZ, 50 mg/kg, prechilled 0.1 M/l sodium citrate buffer, pH 4.2). Three days after STZ administration, the model group (MOD) and the metformin group (160 mg/kg/d, MET) were randomly assigned. Based on earlier studies, the dosages were determined with safety in mind. Changes in body weight and blood glucose levels were also regularly recorded. After receiving therapy, the animals were killed in conformity with laws protecting animal welfare, and mouse blood was taken for the evaluation of biochemical indicators.

Commercial assay kits (Nanjing Jiancheng Technology Co., Ltd.) were used to assess the following parameters in the serum: total protein, albumin content, globulin (GLB) content, urinary Microalbumin Creatinine Ratio (UmACR), malondialdehyde (MDA), superoxide dismutase (SOD), total antioxidant capacity (T-AOC), glutathione (GSH), interleukin-1β (IL-1β), and interleukin-6 (IL-6).

For staining subsequently, the renal samples were stored in a 10% formaldehyde solution. Hematoxylin-eosin (H&E) and Masson stains were applied to the paraffin slices. The Leica DM 1 inverted microscope was used to view and take pictures of the kidney tissues.

Following UPLC-MS/MS analysis, chemical components that were successfully identified were assessed for network pharmacology. For each of the previously described chemical components in YMP, the putative target was the analytical platform (https://prediction.charite.de/subpages/target_prediction.php). The targets were aggregated and combined using the Uniport module in Perl software, and the names of the selected targets were standardized to the official human gene to remove duplicates. Following this, we employed five databases: OMIM (https://omim.org/) (16), DrugBank (https://www.drugbank.ca/) (17), DisGeNet (https://www.disgenet.org/home/) (18), Genecards (https://www.genecards.org/) (19), and TTD (http://db.idrblab.net/ttd/) (20) to search for DN-related genes. The search results from many databases were combined to create a complete gene collection linked to DN. We eventually obtained a collection of genes connected to chemical components target genes and DN by taking the intersection of the sets of genes associated with DN and chemical components target genes. A protein-protein interaction (PPI) network with a set parameter of moderate confidence (0.400) was built using the STRING database, the target gene set for chemical components, and the gene set linked to DN (21). The PPI network was exported using STRING, and the important subnetwork was further examined using Cytoscape. Genes that had Betweenness, Closeness, and Degree values greater than the median were filtered based on CytoNca scores (22). A primary subnetwork was constructed using these filtered genes, and a final, vital subnetwork was created by further filtering the primary subnetwork. We constructed a compound-target network using the DN-related gene set and the chemical components target gene set in Cytoscape version 3.10.2. The underlying biological processes (BP), cellular components (CC), molecular functions (MF), and significant signaling pathways were then identified using enrichment analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and gene ontology (GO). This investigation was conducted using the Metascape database (https://metascape.org/) (23–25).

The urine and serum samples were mixed three times. Following three minutes of mixing, the mixture was centrifuged for twenty minutes at 10,000 rpm and 4°C. The supernatant was collected to conduct a focused metabolomics study. A centrifuge tube was filled with 120 µl of water, 190 µl of methanol, 380 µl of dichloromethane, and 100 µl of serum. The tube was vortexed for three minutes and then centrifuged for fifteen minutes at 12,000 rpm (4°C). QC (quality assurance) sample: Ten microliters of each urine and serum sample were combined, and the mixture was centrifuged to make serum quality control samples.

Using the Ultimate 3000 UHPLC technology, a metabolomics investigation with an untargeting focus was conducted. Mobile phase A was water containing 0.1% (v/v) formic acid, while mobile phase B was acetonitrile. The flow rate is 0.3 milliliters per minute. The gradient elution program used was 5% B for 0–2 minutes, 5–60% B for 2–5 minutes, 60–95% B for 5–12 minutes, 95% B for 12–20 minutes, 95–5% B for 20–20.1 minutes, and 5% B for 20.1–25 minutes. Every seven samples, a quality control sample was added to the analysis queue. The injection volume is 8 µl. The temperature of the column is 35°C. For testing, LTQ Orbitrap Velos Pro was utilized. They used an electrospray ionization (ESI) source. Positive and negative ion modes served as the foundation for this effort.

Orthogonal partial least squares discriminant analysis (OPLS-DA) and unsupervised principal component analysis (PCA) were the two main methods used for multivariate statistical studies. The load plot S-plot and variable importance in projection (VIP) plot were utilized in the OPLS-DA study to illustrate the magnitude of the metabolite contribution. The degree to which each metabolite influences the VIP value-based classification of samples from each group may be measured by this method. The t-test was used to determine whether there was a significant difference between the groups. Differential metabolites were identified using the following criteria: VIP >1, t-test (p < 0.05), and fold change (FC) values (FC > 1.5 or < 0.7).

Thorough preprocessing was performed on the raw sequencing data to provide clean reads that could be used for additional research. Initially, reads comprising more than 1% of unknown bases, reads polluted by adaptor sequences, and reads with a low-quality base ratio (quality score ≤ 15) exceeding 40% were filtered out using SOAPnuke (version 1.6.5). Following that, the clean reads were saved in the FASTQ format. Bowtie2 (version 2.4.5) was then used to map the cleaned data to the constructed unique gene sequences. After mapping, RSEM (version 1.3.1) was used to precisely quantify the gene expression levels. Public databases like Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) were used to annotate the genes. Important details regarding the biological roles of the genes were revealed by this annotation process. In order to remove differentially expressed genes (DEGs) with fragments per kilobase of transcript per million fragments mapped (FPKM) values less than 1 in both comparison groups and to make sure that the detection rate of DEGs is greater than or equal to 66.6% in at least one group, DESeq2 was used to identify genes that are differentially expressed between groups with fold change (FC) > 2 or < 0.5. The DEGs were functionally categorized using the KEGG and GO annotations. KEGG enrichment analysis was performed using R’s phyper function, and GO enrichment analysis was carried out using the TermFinder package to obtain more understanding of the enriched biological processes and pathways. Significantly enriched genes and pathways were identified using an AQ value threshold of < 0.05.

The program SPSS 22.0 was used to evaluate the data. The results were displayed using the mean plus or minus standard deviation. One-way ANOVA and the Tukey post-test were used to create group comparisons; a p-value of less than 0.05 was considered statistically significant.

The primary compounds in YMP were discovered using UPLC-MS/MS. The MS/MS fragment pattern served as the primary technique for compound identification in the evidence. By cross-referencing with database data and published literature, 43 compounds in YMP were found based on similarities in mass spectrometric behavior and retention time ( Supplementary Figure S1 ). Among the several kinds of compounds discovered are lignins and flavonoids, among others. A comprehensive list of these chemicals can be found in Table 1 .

Integrated multi-database analysis identified 101 high-confidence targets linking YMP components to DN pathogenesis. Initial screening revealed 369 YMP-associated targets (Super-pred) and 1,354 unique DN-related genes (aggregated from OMIM, DrugBank, GeneCards, PharmGKB, and TTD). Crucially, the intersection of these datasets yielded 101 shared targets ( Figure 1A ), suggesting convergent mechanisms for therapeutic intervention.

Protein-protein interaction analysis (STRING; Figure 1B ) revealed significant modularity among these targets, indicating functional cooperativity in DN-relevant processes. KEGG pathway enrichment identified 152 significantly enriched pathways, with top pathways implicating HIF-1 signaling, metabolic dysregulation, and AGE-RAGE signaling - all established drivers of DN progression ( Figure 1C ). Notably, enriched cancer pathways were deprioritized as likely context-independent artifacts.

GO enrichment analysis further refined functional insights: Targets were significantly associated with cation/small molecule binding (MF), extracellular vesicle regulation (CC), and cellular response to chemical stimuli (BP) ( Figure 1D ). These terms align mechanistically with DN pathophysiology, particularly extracellular matrix remodeling, exosomal communication, and hyperglycemia-induced cellular stress.

CON mice displayed normal agility and physiological parameters. STZ/HFD-induced diabetic MOD mice exhibited characteristic metabolic dysfunction, including significant weight loss (p < 0.001), lethargy, polyuria, and reduced body mass, confirming successful model establishment ( Figure 2A ). Notably, the liver index of the diabetic mice was significantly greater (p < 0.001) than that of the CON group, suggesting that the organ index could be used to assess a drug’s toxicity. YMP appears to have a protective effect against STZ damage, as evidenced by the statistically significant difference (p < 0.001) between YMP and the other groups in Figure 2B . YMP at 40 and 160 mg/kg may considerably lower the FBG (p < 0.001) in comparison to the MOD group ( Figure 2C ). The MOD and CON groups drank significantly different amounts of water (p < 0.001). The group that got a high dosage of YMP had lower water administration (p < 0.001), while the YMPL group did not differ from the administration groups in terms of water intake following administration (p > 0.05) ( Figure 2D ). YMP administration demonstrated significant therapeutic effects.

Diabetic (MOD) mice exhibited significantly elevated urine volume compared to controls (CON) (p < 0.001). YMP treatment displayed a downward trend and the impact was rose with concentration, with both YMPL and YMPH groups showing significantly lower urine output versus MOD (p < 0.001; Figure 2E ). Urinary total protein excretion was markedly increased in MOD mice (p < 0.001). The treatment group showed a reduction in proteinuria with increasing concentrations ( Figure 2F ). Serum albumin levels were substantially reduced in MOD mice (p < 0.001), while YMP administration significantly restored albumin concentrations (p < 0.01; Figure 2G ). No significant differences were observed in serum globulin (GLB) levels between MOD and CON groups (p > 0.05). YMP treatment did not alter GLB concentrations ( Figure 2H ). Similarly, the albumin/globulin ratio (A/G) showed no statistically significant changes across groups ( Figure 2I ). MOD mice displayed significantly elevated urinary microalbumin/creatinine ratio (UmACR) versus CON (p < 0.001). Both YMP doses effectively reduced UmACR levels compared to MOD (p < 0.001; Figure 2J ). YMP significantly ameliorates diabetic nephropathy by reducing polyuria, proteinuria, albuminuria, and restoring serum albumin levels.

To determine the effect of YMP on the inflammatory response and oxidative stress of the STZ-induced diabetic mice, proinflammatory cytokines and oxidative stress markers were evaluated. Important indicators of the body’s antioxidant status include the levels of MDA, SOD, T-AOC, and GSH. In contrast to the CON group, the MOD group had significantly lower levels of GSH, SOD, and T-AOC and higher levels of MDA. The MDA of the MET (p < 0.01) and YMPH (p < 0.001) groups was significantly lower than that of the MOD group. However, Figure 2K shows that YMPL did not significantly down-regulate MDA (p > 0.05). The MET and YMPL groups did not exhibit an increase in SOD when compared to the MOD group; however, there was evidence that the group administered a high dosage of YMP would change the SOD concentration (p < 0.05) ( Figure 2L ). The MOD group had a considerably greater YMPH than the CON group (p < 0.001). T-AOC decreased in each delivery group, and the impact was rose with concentration (p < 0.001) ( Figure 2M ). Notably, the GSH levels of the diabetic mice were significantly lower (p < 0.001) than those of the CON group, suggesting that GSH could be a helpful marker ( Figure 2N ). Compared to the CON group, the MOD group’s levels of IL-1β and IL-6 were considerably greater (p < 0.001). IL-1β levels were considerably lower in the YMPL and YMPH groups than in the MOD group (p < 0.001), and the impact was rose with concentration ( Figure 2O ). The IL-6 levels of the YMPL and YMPH groups did not exhibit a concentration-dependent effect, and they were significantly lower than those of the MOD group (p < 0.001) ( Figure 2P ). Based on the comprehensive findings, YMP ameliorates oxidative stress and inflammation in diabetic mice.

CON animals did not exhibit hyperplastic mesangial matrix, thickening of the capillary basement membrane, hypertrophy of the glomerulus as shown by H&E staining. The glomerulus exhibited mild hypertrophy and the mesangial area significantly increased in the MOD group. The glomerular hypertrophy in the YMPH group was notably less pronounced than in the other groups. The relevant data was shown in Figure 2Q . The collagen fibers appeared blue after Masson staining. Using a semiquantitative method, it was rated based on the degree of collagen fiber proliferation. The CON group’s tubular and glomerular structures were normal. The MOD group’s glomerular collagen hyperplasia was much higher than that of the CON group. The findings demonstrated that YMP was helpful in treating DN, with a higher intervention impact at high doses ( Figure 2R ).

Serum and urine samples were subjected to non-targeted metabolomic analysis in positive and negative modes. Each sample was clearly separated from the mice’s CON group. The results demonstrated that the metabolisms of the MOD and administration groups were abnormal. There were 256 and 273 metabolites in positive and negative modes, respectively, in the serum samples ( Supplementary Tables S1 , S2 ). In urine samples, 427 and 460 metabolites were found in positive and negative modes, respectively ( Supplementary Tables S3 , S4 ). Based on the global features of the raw data, OPLS-DA analysis was carried out to distinguish between the groups and show the metabolic differences. The various metabolite groups were obviously distinct from one another based on differences in physiological markers. OPLS-DA may eliminate negligible discrepancies and increase classification accuracy. The two groups were separated using the OPLS-DA model score plots ( Figures 3A-H ). In MOD groups, endogenous chemical metabolisms were changed. Metabolic profiles could be used to identify biomarkers. Supplementary Figures S2A-H displayed the R2Y and Q2 values for serum and urine samples in both positive and negative modes. The MOD groups’ and the CON mice’s OPLS-DA score graphs were clearly different from one another. The metabolic profiles of the serum and urine also showed discernible changes after YMP administration. It was shown that tissue samples were more effective in detecting metabolic fingerprints specific to individual organs than biofluids that depict the metabolic state of the complete body. Serum samples from the CON and MOD groups showed differences in the pathways of glycolysis/gluconeogenesis, aspartate, glutamate, and neomycin biosynthesis, gentamicin, kanamycin, and nitrogen metabolism, pentose and glucuronate interconversions, starch and sucrose metabolism, and selenium metabolism. Numerous metabolic pathways, such as the biosynthesis of valine, leucine, and isoleucine, the metabolism of arginine and proline, the metabolism of nitrogen, the metabolism of butanoates, and the metabolism of histidine, were disrupted in urine samples. The biosynthesis of arginine was also disturbed. The following pathways were affected in serum samples between the MOD and YMPH groups: pentose phosphate pathway, glycolysis/gluconeogenesis, alanine, aspartate, and glutamate metabolism, glyoxylate and dicarboxylate metabolism, arginine and proline metabolism, nitrogen metabolism, arginine biosynthesis, butanoate metabolism, histidine metabolism, pentose and glucuronate interconversions, and citrate cycle (TCA cycle). Urine samples showed disruptions to multiple metabolic pathways: Glycolysis/Gluconeogenesis; Alanine, aspartate, and glutamate metabolism; Pantothenate and CoA biosynthesis; Valine, leucine, and isoleucine biosynthesis; Arginine biosynthesis; D-Amino acid metabolism; Nicotinate and nicotinamide metabolism; Histidine metabolism. Urine positive modes showed complementary but less comprehensive pathway alterations ( Figure 3I ). YMP administration significantly reversed diabetes-associated metabolic disruptions, with YMPH showing pronounced efficacy.

There were 87 and 323 genes in up and down modes, respectively, between the CON and MOD groups ( Figure 4A , Supplementary Tables S5 , S6 ). Significantly increased were the genes GPI, GAPDH, G6PC, HK2, HK1, and HK3 linked to Glycolysis/Gluconeogenesis. The intricate relationships that these target genes’ proteins demonstrated are depicted in the protein-protein interaction network that was generated from the STRING database ( Figure 4B ). A KEGG enrichment analysis was performed to determine which pathways were impacted by the 380 target genes. 36 KEGG pathways were found to be significantly enriched ( Supplementary Table S7 ). These results suggest that the target genes are linked to several networks, including Complement and coagulation cascades, Staphylococcus aureus infection, Toll-like receptor signaling pathway, Cytokine-cytokine receptor interaction, Neomycin, kanamycin and gentamicin biosynthesis, Chemokine signaling pathway, Starch and sucrose metabolism, Hematopoietic cell lineage, Glycolysis/Gluconeogenesis, Viral protein interaction with cytokine and cytokine receptor. Figure 4C showed the bubble plot with the top 10 KEGG pathways. The 380 target genes’ MF, CC, and BP were identified using GO enrichment analysis. 1111 highly enriched GO keywords were discovered. Figure 4D and Supplementary Table S8 displayed a graphic representation of the top 10 terms. According to the analysis’s findings, these target genes are essential for biological functions like defense response, immune system process, immune effector process, cytoplasmic vesicle part, cytoplasmic vesicle, intracellular vesicle, CXCR3 chemokine receptor binding, signaling receptor binding, carbohydrate binding.

There were 958 and 605 genes in up and down modes, respectively, between the MOD and YMPH groups ( Figure 4E , Supplementary Tables S9 , S10 ). Significantly increased were the genes G6PD, PGLS, RPE, TALDO1, HXLB linked to Pentose phosphate pathway. The intricate relationships that these target genes’ proteins demonstrated are depicted in the protein-protein interaction network that was generated from the STRING database ( Figure 4F ). A KEGG enrichment analysis was performed to determine which pathways were impacted by the 1563 target genes. 88 KEGG pathways were found to be significantly enriched ( Supplementary Table S11 ). These results suggest that the target genes are linked to several networks, including Cytokine-cytokine receptor interaction, Viral protein interaction with cytokine and cytokine receptor, Leishmaniasis, Measles, Hematopoietic cell lineage, Malaria, Tuberculosis, Chemokine signaling pathway, Osteoclast differentiation, Cell adhesion molecules (CAMs). Figure 4G showed the bubble plot with the top 10 KEGG pathways. The 1563 target genes’ MF, CC, and BP were identified using GO enrichment analysis. 2861 highly enriched GO keywords were discovered. Figure 4H and Supplementary Table S12 displayed a graphic representation of the top 10 terms. According to the analysis’s findings, these target genes are essential for biological functions like immune system process, immune response, cell activation, cell surface, plasma membrane part, external side of plasma membrane, signaling receptor binding, signaling receptor activity, molecular transducer activity. When administering STZ, the model groups activate glycolysis and gluconeogenesis, which is the same outcome of transcription analysis as it is of untargeted metabolomics study. But when YMPH was given, the pentose phosphate pathway was triggered ( Figure 4I ).

To elucidate YMP’s therapeutic mechanism, an integrated multi-omics network was constructed by combining key targets from network pharmacology with significantly altered metabolites from metabolomics ( Figure 5 ). This synthesis revealed critical interactions between: Gene targets: Arginine metabolism regulators (ART1/3/4/5, DDAH1/2, NOS3), glucose phosphorylation enzymes (GCK, HK1/2/3), and neurotransmitter modulators (GAD1/2). Differential metabolites: Glycolytic intermediates (β-D-glucose, glucose-6-phosphate, fructose-6-phosphate), pentose phosphate pathway components (gluconolactone, ribulose-5-phosphate), and amino acids (alanine, aspartate, glutamate, citrulline, arginine, branched-chain amino acids). Pathway enrichment analysis identified five core disrupted pathways: glycolysis/gluconeogenesis, pentose phosphate pathway, arginine biosynthesis, alanine, aspartate and glutamate metabolism; valine, leucine and isoleucine biosynthesis. These findings suggest YMP restores metabolic homeostasis through coordinated regulation of central carbon metabolism and amino acid biosynthesis.