In this experiment, 32 9-week-old male BALB/c mice weighing 20 ± 1 g each were purchased from SPF (Beijing) biotechnology Co., LTD. (Beijing, China) for this experiment, the certification number is SCXK (JING) 2019–0010. The mice received sterile maintenance feed, with free access to drinking water. These animals were housed at 25 ± 0.5°C and relative humidity of 60%–70%, and the padding was changed twice a week. All animal experimental procedures were following the principles of the local animal ethics committee (The review number is BUCM-4-2,021,112,601-4,074).

After 3 days of acclimatization, at time point P1, the 32 mice were randomly assigned to the control group, the model group, the mesalazine group and the tuina group, 8 per group. Mice in the control group received distilled water. And experimental UC models were induced by freely drinking 3% dextran sulphate sodium (DSS) solution (w/v in distilled water; MPBIO, Canada) for 7 days in the other three groups (Figure 1).

After 7 days of DSS solution induction, at time point P8, all groups received distilled water. The mesalazine group was treated with a once daily mesalazine solution (intragastrically; 500 mg/kg; IPSEN, China) solution for 7 days. The tuina group received a 10-min tuina treatment per day for 7 days (Figure 1).

During acclimatization feeding, the manipulator performed daily handling to reduce animal stress. The tuina protocol consisted of two parts: rubbing the abdomen and pointing. The procedure was performed as follows (Figure 1): The manipulator held a mouse in the palm. First, set RN 4, and RN 15 as the operating area (the abdomen). The abdomen was exposed, and circular rubbed the abdomen with the thumb was in a clockwise direction for 3 min (100 times per minute), then operated counterclockwise for 3 min, 6 min in total, with the force of 4 N (Liu, 2006). Second, the index finger pointed bilateral ST 36 and ST 37, 1 min per acupoint, 4 min in total, with the force of 4 N (Huizhu et al., 2021). The manipulator had been trained prior to the experiment. Pressure and frequency are controlled by the fingerTPS II wireless pressure measurement system (Pressure Profile System, US).

At time point P15, all animals were assessed for the disease activity index (DAI) score according to the following criteria (Table 1). Then, all groups were sacrificed by IP with 2% pentobarbital. Immediately separated the colon from the rest of the abdominal tissue. Removed the entire colon between the anus to the cecum, and measured the colon length. Furthermore, 2 cm of the distal colon sample was taken for hematoxylin–eosin (HE) staining and another 2 cm for 16 S rDNA bioinformatic analysis. Each mouse collected the same segments. Meanwhile, the colon contents were rapidly removed and stored at −80°C for further analysis.

The colon tissues were fixed in a 4% paraformaldehyde fixing solution at 4°C for 48 h. The tissues were then embedded in paraffin and sectioned, the thickness was 4 μm. The glass slides were hydrated at 65°C for 1 h and stained with HE solution (Solarbio, China), then observed and photographed through the microscope (NIKON, Japan). Colon histological scores were graded according to the following standard of tissue injury (Table 2).

Gut microbiota analysis was performed by the 16 S rDNA bioinformatics approach. Briefly, the genomic DNA of the microbial community in fecal samples was extracted by a soil DNA kit (Omega Bio-Tek, US). After qualifying for the concentration and purity test, proceed to the next operation. The V3-V4 hypervariable region of bacterial 16S rRNA gene was amplified by forwarding primer 338 and reverse primer 806, and the primer sequences are 5′-ACTCCTACGGGAGGCAGCAG-3′ and 5′-GGACTACHVGGGTWTCTAAT-3′. Then triple PCR 16S rRNA gene amplification was performed. According to the standard protocol, the purified amplicons were pooled and end-sequenced using the MiSeq PE 300 (Illumina, US) platform. Then, using FastQ to quality filter the raw reads of 16S rRNA gene sequencing, and using Fast Length Adjustment of Short reads to merge (Magoč and Salzberg, 2011; Chen et al., 2018). The UPARSE was used to cluster the operational taxonomic units (OTUs) with 97% similarity, and the chimeric sequences were identified and deleted (Edgar, 2013). Using 0.7 as the confidence threshold, the 16S rRNA database was then analyzed by the RDP Classifier for each representative OTU sequence (Wang et al., 2007).

The DIA quantitative analysis process includes 5 steps: total protein extraction, protein digestion, high pH RP-UPLC separation, liquid chromatography-mass spectrometry analysis, and protein identification (Law and Lim, 2013). Briefly, the colon tissue samples were incubated on ice in protein lysis buffer (with 8 M urea, 1% SDS, and protease inhibitor) for 30 min, then, all samples were centrifuged for 30 min at 4°C, set the speed to 16,000 g, and determined the protein supernatant concentration using the bicinchoninic acid method. The protein digestion process includes 3 steps: resuspension was performed with triethylammonium bicarbonate buffer (100 mM, Sigma-Aldrich, and Germany). Then the reduction was performed with Tris (2-carboxyethyl) phosphine buffer (10 mM, Sigma-Aldrich, and Germany) for 60 min at 37°C. At last, alkylation was performed with iodoacetamide buffer (40 mM, Sigma-Aldrich, and Germany) for 40 min in darkness at room temperature. After the samples were centrifuged for 20 min at 4°C and set the speed to 10,000 g, the collected pellet was resuspension and incubation at 37°C overnight after adding trypsin. The trypsin-digested peptides were vacuum-dried and resuspended in UPLC loading buffer, then fractionated into fractions to increase proteome depth. Online analysis of the redissolved peptides using a nanoflow liquid chromatography–tandem mass spectrometry method with the EASY-nLC system (Thermo Fisher Scientific, USA). Ran the Q Exactive HF-X in DIA mode with variable isolation windows, set with 40 windows, each overlapping by 1 m/z. Last, use the default settings of Spectronaut (Version 14; Biognosys AG) for analysis of DIA data files. A cutoff value of 1% was used as the Q-value (FDR) at the precursor level and the protein level. The six highest intensity peptides were used for quantification.

Data from this study were statistically analyzed using GraphPad Prism (Version 9.3.1). Data were presented by mean ± standard error (SEM). A two-way ANOVA and Tukey’s multiple comparisons were used to test the significance differences of the data and p < 0.05 was considered statistical significance. Differential abundance testing between groups and post hoc tests were performed using the Benjamini-Hochberg method and Tukey’s multiple comparisons. The Chao diversity index was used to measure Alpha diversity. The binary-jaccard distance and principal coordinates analysis (PCoA) was used to analyze Beta diversity. The metabolic functional prediction analysis was performed using the PICRUSt analysis platform. In this study, the identification thresholds for DEPs were selected as fold change >1.2 or < 0.83 and p-value <0.05. Using the GO (http://geneontology.org/) and KEGG pathway (https://www.kegg.jp/kegg/) platforms for DEPs annotation and enrichment bioinformatics analysis (Mi et al., 2019).

The body weights of the mice were measured every 2 days during the experiment (Figure 2A). As the experiment proceeded, the control group gradually gained weight. The model, mesalazine, and tuina group gradually lost weight during model induction, and with statistical differences at P5 (^ap = 0.015). At time points P8 to P15, the model group maintained the trend of weight loss. In contrast, treatments with mesalazine or tuina showed effects (^bp = 0.010; ^cp = 0.007). Mice that received tuina showed a slightly lower weight but no statistical difference compared with the mesalazine group (p = 0.208). Along with pathological weight loss, changes in the properties of stool, fecal occult blood, or even bloody stool are also caused by DSS (Figure 2B). The stool properties in the control group were consistently normal, with a DAI score of zero. However, the model group showed persistent occult blood in the feces started on the 3rd day after model induction, and bloody stools appeared on the 9th day after modeling. After all, when the treatments were completed, the stool properties of the mesalazine group and the tuina group returned to normal and without occult blood (^bp < 0.000; ^cp < 0.000). Although the DAI score in the mesalazine group was better than the tuina group, there was no statistical between the two groups (p = 0.496).

DSS-induced DAI score changes can also be reflected by changes in colon lengths (Figure 2C). The colons of the model group were the shortest among these four groups, and the control group was the longest. A comparison of colon lengths showed that the mesalazine group was longer than the tuina group, but there was no statistical difference (p = 0.612). However, 2 groups were statistically significant different when compared with the model group (^bp < 0.000; ^cp < 0.000). The degree of change in the morphological structure of the colon before and after treatments was assessed using the colon histological score (Figures 2D,E). The model group had the highest colon histological score and the most severe colon morphological abnormalities, such as colon inflammation and destruction of the epithelial barrier. HE staining showed that the model group had crypt injury, goblet cells, and epithelial cell loss, and infiltration of transmural inflammatory cells in the mucosa, submucosa, or glands. After tuina treatment, the colon histological score was significantly reduced (^bp < 0.000), at the same time, reduced colon inflammation, preservation of structural integrity, and promotion of cell recovery can be observed. The results of the mesalazine group were similar to the tuina group (p = 0.101). Altogether, these results indicated that tuina could improve damages, reduce inflammation, and protect the structure of the colon.

Compared with the control group, there was a significant difference in the gut microbiome of the model group. At the phylum level, the predominant phyla of all groups are composed of Firmicutes and Bacteroidota, and the model mice showed an increased relative abundance of Verrucomicrobiota and a decreased relative abundance of Patescibacteria, Desulfobacterota, Actinobacteriota, and Campilobacterota (Figure 3A). At the family level, the abundance of Lachnospiraceae, Prevotellaceae, and Bacteroidaceae increased and Muribaculaceae, Staphylococcaceae and Rikenellaceae decreased (Figure 3B). At the genus level, the relative abundance of Staphylococcus, Desulfovibrio, and Rikenella decreased and Bacteroides, Lachnospiraceae_NK4A136_group, Prevotellaceae UCG-001, and Alloprevotella increased (Figure 3C). Overall, the alpha diversity of the model group was reduced (Figure 3D).

Compared with the model group, at the phylum level, the relative abundance of Patescibacteria and Desulfobacterota increased and Actinobacteriota and Verrucomicrobiota decreased in the tuina group. In the mesalazine group, Proteobacteria and Cyanobacteria increased, Desulfobacterota, Actinobacteriota, and Verrucomicrobiota decreased (Figure 3A). At the family level, Lachnospiraceae and Saccharimonadaceae increased, and Rikenellaceae and Prevotellaceae decreased in the tuina group. Prevotellaceae, Bacteroidaceae, and Oscillospiraceae increased and Lachnospiraceae and Rikenellaceae decreased in the mesalazine group (Figure 3B). At the genus level, Lachnospiraceae_NK4A136_group and Candidatus_Saccharimonas increased, Prevotellaceae UCG-001, Alloprevotella, and Parabacteroides decreased in the tuina group. The relative abundance of Bacteroides and Prevotellaceae UCG-001 increased, and Lactobacillus, Lachnospiraceae_NK4A136_group, Alloprevotella and Parabacteroides decreased in the mesalazine group (Figure 3C). Mesalazine and tuina treatment can help restore the alpha diversity index to the gut microbiota (Figure 3D).

The results of beta diversity showed that 4 groups were clustered separately in the PCoA, suggesting that modeling and intervention were the primary factors that affected community differences. There were very small areas of overlap between the model and tuina groups, this suggested that the tuina treatment may not be as effective as the mesalazine treatment, which is consistent with other results from this study (Figure 4A). However, on the PC 1 axis, the diversity of the tuina group was similar to that of the control group (Figure 4B). And on the PC 2 axis, the tuina group had good intragroup aggregation (Figure 4C). The merits of the results of the nonmetric multidimensional scaling (NMDS) analysis tested with the stress value, and a stress value of 0.137 represented a certain degree of explanatory significance, which indicated that the species composition was not similar between the 4 groups of samples (Figure 4D).

The functional abundance of metabolic COG in this community was deduced from the 16S compositional data utilizing the PICRUSt analysis platform. The metabolic COG functions of the colony were mainly focused on the transport and metabolism of carbohydrate, amino acid, and inorganic ion, and energy production and conversion (Figure 5). We suggested that the gut microbiota may play a positive role through the signaling pathways of carbohydrate, amino acid, inorganic ion, and signal transduction.

Quantitative proteomics analysis of DIA and bioinformatics analysis can further elucidate the potential regulatory mechanisms of tuina treatment on UC. The principal component analysis (PCA) showed excellent sample aggregation in each group, and the distance was long between the sample points in each group, indicating that the four groups were not similar to each other, the model was successfully established, and tuina or mesalazine treatment was effective (Figure 6). Proteomics analysis identified 6,294 quantifiable proteins, of which 987 DEPs in the model group vs. control group, 589 were down-regulated and 398 were up-regulated. Tuina treatment had 292 down-regulated DEPs and 578 up-regulated DEPs, for a total of 870 DEPs compared with the model group, then 843 DEPs (413 down-regulated and 430 up-regulated) were found in the mesalazine group vs. the model group.

To understand the unique mechanism of tuina, it is necessary to identify the DEPs that are regulated by tuina. There were 437 overlapping DEPs between the control group vs. model group and the model group vs. tuina group, of which 370 DEPs were regulated by tuina. And there were 379 overlapping DEPs between the control group vs. model group and the model group vs. mesalazine group, with 247 DEPs regulated by mesalazine treatment (Figure 7).

GO enrichment analysis revealed a total of 265 GO terms based on 370 DEPs regulated by tuina treatment, including glandular epithelial cell differentiation, oxygen carrier activity, regulation of cytokine production involved in immune response, etc. (Figure 8A). There were 323 GO terms based on 247 DEPs regulated by mesalazine treatment, including positive regulation of leukocyte mediated immunity, response to bacterium, and positive regulation of lymphocyte leukocyte mediated immunity (Figure 9A). KEGG annotation analysis showed that there were 304 pathways based on tuina-regulated DEPs, the level 1 pathway categories with the highest number of enriched proteins were human diseases (305 proteins), organismal systems (220 proteins), and metabolism (184 proteins), level 2 pathways including signal transduction, infectious disease: viral, transport and catabolism, endocrine system, and immune system (Figure 8B). KEGG enrichment analysis showed level 3 pathways, and tuina-regulated pathways included biotin metabolism, Notch signaling pathway, linoleic acid metabolism, and autophagy (Figure 8C). There were 290 pathways based on mesalazine-regulated DEPs, and the level 1 pathway categories with the highest enriched proteins were human diseases (199 proteins), metabolism (149 proteins), and organismal systems (122 proteins), level 2 pathways including the immune system, infectious disease: viral, transport and catabolism, and infectious disease: bacterial (Figure 9B). Level 3 pathways included the intestinal immune network for IgA production, vitamin digestion and absorption, arachidonic acid metabolism, and mucin type O-glycan biosynthesis (Figure 9C).