I3C was procured from Shanghai yuanye Bio-Technology Co., Ltd. (S49927); Orlistat enteric tablets were commercially available from Zhejiang Hisun Pharmaceutical Co., Ltd. (H20100089); Metformin was supplied by ChongQing KeRui NanHai Pharmaceuticals Ltd. (691,504).

Sixty 5-week-old male C57BL/6N mice weighting 18–22 g (Beijing Vital River Laboratory Animal Technology Co., Ltd.) were kept in the standard laboratory (25°C ± 2°C, 12-h light/12-h dark cycle) without food and water restriction. All experiments concerning animals were ratified by the Ethics Committee of Xinjiang Medical University (Ethic No.: IACUC-20211118–02).

Following a one-week acclimation period, the mice were assigned at randomization into 6 groups with 10 mice/group: NCD (mice on a normal diet), NCD-I3C (mice on a normal diet administrated with I3C), HFD (mice on an HFD), HFD-I3C (mice on an HFD administrated with I3C), HFD-Orlistat (HFD-O, mice on an HFD administrated with Orlistat), and HFD-Metformin (HFD-M, mice on an HFD administrated with Metformin) groups. The normal diet contained 10% kcal from fat (supplied by the Animal Center of Xinjiang Medical University), while the HFD comprised 60% kcal from fat (Beijing Vital River, D12492).

Mice in the NCD and HFD groups were administered intragastrically with an equivalent amount of normal saline. Mice in the NCD-I3C and HFD-I3C groups were orally administered with I3C, 40 mg/kg/day. Mice in the HFD-Orlistat were orally given Orlistat, 60 mg/kg/day, and the mice in the HFD-Metformin groups were orally given Metformin, 120 mg/kg/day. The weight and food intake of all mice were measured once a week. Fresh fecal samples from all mice were harvested and then preserved in a −80°C freezer. After intraperitoneal injection of pentobarbital sodium (35 mg/kg) for anesthesia, mouse eyeball vein blood was extracted and placed in a heparin-containing blood collection tube, followed by 15-min centrifugation (4°C, 1000 g). The plasma was separated and preserved at −80°C. The liver tissues and colon tissues were harvested, one portion of which was fixed in 4% paraformaldehyde for 24 h, while the other portion was placed in a −80°C freezer for later use.

Liver and colon tissues fixed in 4% paraformaldehyde were sheared, followed by graded dehydration in 50%, 75%, and 95% ethanol. The tissues were hyalinized after 2 h of immersion in xylene. After paraffin embedding, the tissues were sliced into 5-µm sections and then hydrated. The sections were dyed by hematoxylin-eosin for 5 min, rinsed in water for 10 min, and then washed in distilled water. Following 2-min dehydration in 95% ethanol, the sections were rinsed with distilled water 2 times. Next, the sections were hyalinized in xylene for 5 min, followed by 2 rinses in distilled water again. These sections were sealed with neutral resin. Ultimately, the pathological and morphological changes of mouse liver and colon tissues were visualized under the microscope and photographed by the electron microscope (Aximujiang et al., 2022).

Body fat compositions (fat, muscle, and body fluid contents) in all mice were analyzed on a small animal nuclear magnetic resonance (NMR) (Nanjing Xinfeida Oetech. co.,LTD., LF50) body fat analyzer.

The ELISA kits were adopted for the measurements blood lipid level of T-CHO (Nanjing Jiancheng Bioengineering Institute, A111-1–1), TG (Nanjing Jiancheng Bioengineering Institute, A110-1–1), HDL-C (Nanjing Jiancheng Bioengineering Institute, A112-1–1), LDL-C (Nanjing Jiancheng Bioengineering Institute, A113-1–1).

Inflammation factor in the serum were measured via ELISA kits, IL-6 (Quanzhou Ruixin Biotechnology Co., Ltd., RX203049M), IL-1β (Shanghai Jianglai Biotechnology Co., Ltd., JL18442), CXCL1 (Shanghai Jianglai Biotechnology Co., Ltd., JL20150), MPO (Shanghai Jianglai Biotechnology Co., Ltd., JL10367), IFN-γ (Shanghai Jianglai Biotechnology Co., Ltd., JL10967) and TNF-α (Quanzhou Ruixin Biotechnology Co., Ltd., RX202412M).

ALT and AST activities in the serum were measured via an ALT assay kit (Shanghai Jianglai Biotechnology Co., Ltd., JL12668) and a AST assay kit (Shanghai Jianglai Biotechnology Co., Ltd., JL13992).

Colon tissue sections were subjected to dewaxing, gradient ethanol dehydration, antigen retrieval, and PBS rinses. Sections were incubated with 10% goat serum, primary antibody (overnight, 4°C), and secondary antibody (37°C, 30 min). After DAB staining, the cell nucleus was dyed by hematoxylin and sealed with neutral resin, followed by microscopic image acquisition. The positive rate was analyzed based on the acquired images using ImagePro Plus 6.0 software.

After 11 weeks of dietary intervention, mouse insulin and fasting blood glucose (FBG) levels were measured. ITT and OGTT experiments were simultaneously carried out. All mice were fasted for 8 h after the last lose, without water limitation. FBG levels were measured in mouse tail-tip blood samples by a glucometer. Then, each mouse was administered intragastrically with a glucose solution of 2 g/kg, and tail-tip blood samples were taken at 0, 30, 60, and 120 min, respectively, for the measurement of the blood glucose levels. In the 12th week of the experiment, mice were subcutaneously injected with conventional insulin (0.5 U/kg). The 0-, 40-, 90-, and 120-min tail-tip blood glucose levels were again measured on the glucometer. The changes in blood glucose levels from t = 0 to t = 120 were analyzed, yielding the total area under the curve (AUC). An insulin ELISA kit was utilized for insulin level detection. The insulin resistance (HOMA-IR) was computed with the assistance of a steady-state model: fasting insulin (µUI/mL) × FBG (mM)/22.5.

DNA was extracted from fecal samples by the cetyltrimethylammonium Bromide (CTAB) method, the quality of which was determined by agarose gel electrophoresis, followed by DNA quantitation on an ultraviolet spectrophotometer. The V3-V4 region of the bacterial 16S rRNA gene was amplified in the polymerase chain reaction (PCR) system (Bio-Rad, United States): amplification at 98°C for 30 s; subsequently, amplification at 98°C for 10 s, 54°C for 30 s, and 72°C for 45 s, 35 cycles in total; finally, extension at 72°C for 10 min using 341F 5′-CCTACGGGNGGCNGCAG-3′ and 805R 5′-GACTACHVGGGTATCTAATCC-3'. The PCR amplification reaction system (25 µL) comprised 12.5 µL Phusion Hot start flex 2×Master Mix, 2.5 µL Forward Primer, 2.5 µL Reverse Primer, and 50 ng Template DNA. Purification of PCR products was made by AMPure XT beads (BeckmanCoulterGenomics, Danvers, MA, United States), followed by quantitative detection by Qubit (Invitrogen, United States). After 2% agarose gel electrophoresis, products were harvested using the AMPure XT beads recovery kit. A bioanalyzer (2100Agilent, United States) and an Illumina library quantification kit (Kapa Biosciences, Woburn, MA, United States) were applied to assess the purified PCR products. Lastly, a NovaSeq 6000 sequencer was used for sequencing (Wang et al., 2020).

The double-ended data obtained after sequencing were split based on barcode information, and the connectors and barcode sequences were eliminated. The data were subsequently assembled and filtered. Length filtering and denoising were implemented using the QIIME2 software package and DADA2 plugin, generating Amplicon Sequence Variants (ASV) feature sequences and ASV abundance tables. Also, singletons ASVs were eliminated. The α-diversity and β-diversity were analyzed subsequently. α-diversity analysis can evaluate the diversity according to parameters like observed species, shannon, simpson, chao, goods coverage, and pielou e. β-diversity calculates four distances (weighted unifrac, unweighted unifrac, jaccard, bray curtis) for diversity assessment. With the ASV sequence files, the species were annotated on SILVA (Release 138, https://www.arbsilva.de/documentation/release138/) and NT-16S databases and their abundance in each sample was statistically analyzed on the strength of the ASV abundance table. In this analysis, a confidence threshold of 0.7 was used for annotations.

Based on the obtained species abundance information, a differential analysis was conducted. Different statistical methods were selected depending on the sample conditions: the Fisher’s exact test was applied for comparisons of samples without biological replicates; The Mann-Whitney U test and Kruskal–Wallis test were utilized for comparisons of samples with biological replicates between two groups and among multiple groups, respectively. Screening threshold: P < 0.05.

At first, 5 mL of abdominal aortic blood was extracted from mice and maintained at ambient temperature for 30 min. Following that, 15-min centrifugation was done at 3000 g and 4°C. The supernatant (serum) was packaged into an EP test tube. Then, 200 µL serum was mixed with 400 µL PBS in the EP tube before another 10-min centrifugation (10,000 g, 4°C). Next, 550 µL filtered serum was placed into a 5 mm NMR tube. Plasma ¹H-NMR spectroscopy was performed with spin-echo pulse sequences on the Inova NMR spectrometer (Varian, United States). The water peak suppression was achieved using a pre-saturation method at 25°C, with a saturation time of 2 s, a sampling number of 32K, a scanning frequency of 128 times, and a spectral width of 1000Hz. The obtained ¹H-NMR spectrum was subjected to phase and baseline corrections and processed using an exponential window function with a broadening factor of 0.5Hz. The width of each ¹H-NMR spectrum (δ) was 9.0–0.5 ppm, and at the same time, the range for water signal removal was set as δH = 5.20–4.70 ppm.

Orthogonal partial least-squares-discriminant analysis (OPLS-DA) was implemented with the use of SIMCA-P11 software (Hong et al., 2022). The accuracy of OPLS-DA analysis results was assessed with the use of parameters such as R²X and Q², wherein Q² represents the validity of the statistical results. The multivariate statistical analysis of ¹H-NMR spectral data involved principal component analysis, PLS-DA, and OPLS-DA. The OPLS-DA approach was employed to determine the variable importance in the projection (VIP) and determine the differences in metabolites between the two groups. VIP > 1 and P(Corr) > 0.3 were assigned as the identification criteria for differentially expressed metabolites.

All result analyses were realized with the application of GraphPad Prism version 9. Continuous variables were presented in the form of mean ± standard deviation, and categorical variables were described as percentages. Experimental results between the two groups were analyzed using unpaired t-tests. For the results among more than two groups, a one-way analysis of variance (ANOVA) was implemented, with the criteria being α = 0.05. Figures were plotted with the assistance of GraphPad Prism version 9.

Relative to the NCD group, insignificant changes were detected in terms of obesity-associated parameters including body weight, weight gain, total fat content, white fat content, fat mass/body mass, lean mass/body mass, energy efficiency, body muscle content, and body fluid content in the NCD-I3C group. Except for body muscle content, other obesity-associated parameters mentioned above were elevated markedly in the HFD-fed mice (P < 0.05 or P < 0.01). The HFD-I3C, HFD-O, and HFD-M groups exhibited significant reductions in all the obesity-associated parameters mentioned above versus the HFD group (P < 0.05 or P < 0.01) (Figures 2A–I). Thus, I3C was suggested to be significantly effective for the treatment of HFD-associated obesity in mice.

FBG, insulin concentration, and HOMA-IR index showed no remarkable difference between NCD and NCD-I3C groups. However, these parameters were noticeably higher in the HFD group than in the NCD group (P < 0.05 or P < 0.01). The aforementioned parameters in the HFD-fed mice were markedly reduced in response to I3C, Orlistat, and Metformin treatment (P < 0.05 or P < 0.01) (Figures 3A–E). Furthermore, I3C exhibited a superior glucose-lowering effect to the positive control Orlistat but was less effective than the other positive control Metformin.

TG, TC, LDL-C, and HDL-C levels in NCD-fed mice displayed no significant changes after I3C treatment. Relative to the NCD mice, noticeable elevations were detected in TG, TC, and LDL-C levels with a remarkable reduction in HDL-C in the HFD-exposed mice (P < 0.05 or P < 0.01). After treatment with I3C, Orlistat, or Metformin, the TG, TC, and LDL-C levels in HFD-fed mice were markedly reduced, while that of HDL-C was noticeably raised (P < 0.05 or P < 0.01) (Figures 4A–D). Moreover, the regulatory impact of I3C on blood lipid levels was superior to the positive control Orlistat and almost equal to Metformin. Therefore, the intake of I3C could ameliorate blood lipid abnormalities triggered by HFD in mice and modulate the lipid metabolism disturbance of obese mice.

The IL-6, TNF-α, IL-1β, CXCL1, IFN-γ, and MPO levels in mouse plasma had no noticeable differences between NCD and NCD-I3C groups. The plasma levels of the aforementioned proteins were sharply upregulated in HFD-fed mice (P < 0.01). I3C, Orlistat, or Metformin treatment resulted in significant reductions in the plasma levels of these inflammatory proteins in HFD-fed mice (Figures 4E–J). The anti-inflammation action of I3C was superior to positive controls Orlistat and Metformin. It was illustrated that I3C could relieve HFD-elicited systemic inflammation.

The pathological results showed good overall structure and tight arrangement of hepatocytes with uniform cytoplasm but no edema in the NCD and NCD-I3C groups, and a small amount of lipid droplets were visible. The HFD group exhibited slight swelling in the cytoplasm of hepatocytes, swelling of some organelles, accumulation of massive lipid droplets, and severe inflammatory cell infiltration. I3C treatment reversed the hepatic steatosis induced by HFD. In the HFD-I3C groups, we observed uniform cytoplasm of hepatocytes without swelling, slight swelling of organelles, and a small number of lipid droplets without obvious fusion. The results of the HFD-O group, HFD-M group, and HFD-I3C group were consistent (Figure 5A). To sum up, long-term intake of HFD will significantly damage the integrity of mouse liver tissue structure and promote the generation of massive lipid droplets in mice, and intragastric administration of I3C can significantly reduce lipid droplets, which may be the primary reason for lipid reduction.

No significant differences existed in histological scores, TC content in liver, AST, and ALT levels between the NCD and NCD-I3C groups. The aforesaid indicators markedly increased in the HFD-fed mice (P < 0.05 or P < 0.01). Additional I3C, Orlistat, or Metformin sharply downregulated the above indicators in the HFD-fed mice (P < 0.05 or P < 0.01) (Figures 5B–E). Relative to the positive controls (Orlistat and Metformin), I3C displayed a higher anti-inflammatory activity. These data together demonstrated that I3C could alleviate hepatic steatosis resulting from HFD.

According to the pathological analysis results of colon tissue sections, the NCD-I3C group displayed no significant reduction in the colon length, with intact structure of intestinal mucosal epithelium and no infiltration of inflammatory cells, as compared to the NCD group. In comparison with the NCD-fed mice, HFD-induced mice had significantly shortened colon length, incomplete intestinal mucosal epithelium, large area defects, significant reduction of goblet cells, and infiltration of massive inflammatory cells deep into the submucosa. I3C therapy reversed the colon injury induced by HFD. I3C treatment led to a significantly prolonged colon length, an increased number of goblet cells, reduced inflammatory cell infiltration, and intact intestinal mucosal epithelium in the HFD-fed mice. The results of the HFD-O group, HFD-M group, and HFD-I3C group were consistent (Figure 6A).

No significant changes were noted regarding the levels of Claudin4, Occludin, and ZO-1 proteins in the NCD-fed mice after I3C administration. These proteins were noticeably downregulated in the HFD-fed mice as compared to the NCD mice (P < 0.05 or P < 0.01). Furthermore, I3C, Orlistat, or Metformin administration markedly augmented the aforementioned protein levels in the HFD-fed mice (P < 0.05 or P < 0.01) (Figures 6B–F). In comparison with the positive controls (Orlistat and Metformin), I3C exhibited superior function to improve the intestinal mucosal barrier function of mice. To sum up, I3C could attenuate HFD-evoked intestinal barrier damage.

This research collected 4,222,786 high-quality 16S rRNA V3-V4 region base sequences in total. After examination of quality and chimeras as well as elimination of low-quality base sequences, 17,059 ASVs were identified. The ASV numbers in the NCD, NCD-I3C, HFD, HFD-I3C, HFD-O, and HFD-M groups were 4,656, 1812, 4,528, 2,347, 1940, and 1778, respectively. It was suggested that long-term intake of HFD could reduce the diversity of gut microbiota, which could be further diminished upon I3C treatment (Supplementary Figures S1A–E).

The results of α-diversity exhibited noticeable decreases the in diversity of gut microbiota in the NCD-I3C and HFD groups than in the NCD group (P < 0.05 or P < 0.01) as well as more significant reductions in the HFD-I3C, HFD-O, and HFD-M groups than in the HFD group (P < 0.05 or P < 0.01) (Supplementary Figures S1F–I). Conclusively, long-term intake of I3C in healthy mice and obese mice could significantly reduce the diversity of gut microbiota.

For β-diversity, the results of principal coordinate analysis (PCoA) based on the Bray-Cruits distance matrix and NMDS analysis, the NCD, NCD-I3C, HFD-I3C, HFD-O, and HFD-M groups displayed significant separation of gut microbiota without intersection phenomenon. Among these groups, the HFD-I3C group showed the distribution of gut microbiota close to the NCD group. Accordingly, significant differences existed in microbiota composition between the HFD group and other groups (Supplementary Figures S1J–L). The obesity model was successfully established, and I3C could mediate the composition of gut microbiota.

Herein, the experiment identified 25 phyla and 465 genera. The bacterial phyla in all groups were dominated by Firmicutes, Bacteroidota, and Desulfobacterota. The relative abundance of Firmicutes was 49.83%, 85.28%, 72.00%, 58.18%, 47.66%, and 40.38% in the NCD, NCD-I3C, HFD, HFD-I3C, HFD-O, and HFD-M groups, respectively. Bacteroidota accounted for 29.73%, 15.29%, 17.26%, 28.28%, and 30.10% while Desulfobacterota constituted 0.57%, 6.84%, 9.65%, 6.84%, 2.21%, and 3.04% in the NCD, NCD-I3C, HFD, HFD-I3C, HFD-O, and HFD-M groups, respectively (Figure 7A).

In contrast to the NCD group, the abundance of Firmicutes, abundance of Desulfobacterota, and Firmicutes/Bacteroidota ratio were dramatically raised (P < 0.01), while the abundance of Bacteroidota markedly decreased (P < 0.01) in the NCD-I3C and HFD groups. The Firmicutes abundance, Desulfobacterota abundance, and Firmicutes/Bacteroidota ratios sharply decreased (P < 0.01) but the Bacteroidota abundance significantly increased (P < 0.01) in the HFD-I3C, HFD-O, and HFD-M groups as compared to the HFD group.

The Circos plot was adopted to analyze the distribution of dominant gut microbiota in different taxa, the top 10 species ranked at the genus level and their relative abundance were identified as follows: Ligilactobacillus (18.44%), Muribaculaceae_unclassified (8.79%), Clostridiales_unclassified (6.00%), Alistipes (5.71%), Eisenbergiella (4.28%), Desulfovibrionaceae_unclassified (3.25%), Lachnospiraceae_unclassified (2.88%), Helicobacter (2.88%), Rikenellaceae_RC9_gut_group (2.86%), and Akkermansia (2.80%) (Figure 7B).

NCD-I3C and HFD groups displayed remarkably higher abundance of Eisenbergiella and Rikenellaceae_RC9_gut_group (P < 0.01) but notably lower abundance of Akkermansia (P < 0.01) versus the NCD group. After I3C, Orlistat, or Metformin treatment, the relative abundance of Eisenbergiella and Rikenellaceae_RC9_gut_group was notably reduced (P < 0.01), and that of Akkermansia markedly raised (P < 0.01) in HFD-fed mice (Figures 8A–D).

Significant reductions were noted concerning the abundance of microbiota related to SCFA production (Lactococcus and Coprococcus) in the NCD-I3C and HFD groups than in the NCD group (P < 0.01). Reversely, Lactococcus and Coprococcus showed higher abundance in the HFD-I3C, HFD-O, and HFD-M groups than in the HFD group (P < 0.01) (Figures 8E,F).

LEfSe analysis with a logarithmic LDA score threshold setting as 4 was employed for the identification of discriminative features. The NCD group was featured by relatively high abundance of Muribaculaceae_unclassified, Lachnospiraceae_NK4A136_group, and Prevotellaceae_NK3B31_group. The NCD-I3C group was characterized by a relatively high abundance of Ligilactobacillus and Clostridiales_UCG_014_unclassified. The HFD group exhibited a sharp elevation in the abundance of Mucispirillum, Clostridiales_unclassified, and Eisenbergiella. The HFD-I3C group showed a notably elevated abundance of Helicobacter, Akkermansia, and Eubacterium (Figure 9).

Akkermansia and Ligilactobacillus were reversely correlated with Fat mass/body mass, weight gain, and white fat but positively associated with Lean mass/body mass. Additionally, Akkermansia and Ligilactobacillus shared negative associations with TC, TG, HOMR-IR, blood glucose, and LDL-C but a positive correlation with HDL-C. Also, these two bacteria were positively linked to claudin 4 and occludin (Figure 10). It was indicated that the aforementioned two bacteria might exert anti-obesity effects by controlling lipid metabolism and restoring intestinal barrier function.

Eisenbergiella and Rikenellaceae_RC9_gut_group were positively related to Fat mass/body mass, weight gain, and white fat but negatively correlated with Lean mass/body mass. Meanwhile, Eisenbergiella and Rikenellaceae_RC9_gut_group were positively linked to TC, TG, HOMR-IR, blood glucose, and LDL-C but inversely linked to HDL-C. These two bacteria also displayed negative associations with claudin 4 and occludin. The above-mentioned two bacteria were, therefore, suggested to potentially induce obesity by increasing white fat content, disrupting lipid metabolism, and disrupting the intestinal barrier.

Given the above-mentioned structural changes, we further analyzed the 16S rRNA data and performed PICRUSt functional predictions (Figure 11). On the strength of Welch’s t-test (P < 0.001), the metabolic characteristics of gut microbiota were analyzed. Compared with the NCD group, the HFD group exhibited a noticeable downregulation in the expression of 10 KEGG pathways, which was significantly upregulated after I3C administration: adenosine ribonucleotides de novo biosynthesis, aromatic biogenic amine degradation (bacteria), anhydromuropeptides recycling, adenosine deoxyribonucleotides de novo biosynthesis II, acetylene degradation, 4-deoxy-L-threo-hex-4-enopyranuronate degradation, 4-aminobutanoate degradation V, 2-methylcitrate cycle I, 2-methylcitrate cycle II, 1,4-dihydroxy-2-naphthoate biosynthesis I. In contrast to the NCD group, the HFD group showed significant upregulation of 5 KEGG pathways, including adenosylcobalamin salvage from cobinamide I, adenosylcobalamin salvage from cobinamide II, adenosylcobalamin biosynthesis from cobyrinate a,c-diamide I, acetyl-CoA fermentation to butanoate II, 1,4-dihydroxy-6-naphthoate biosynthesis I. These pathways were significantly downregulated after I3C administration.

PCA data yielded 5 principal components. The parameters R²X (cum) and R²Y (cum) in the NCD and HFD groups were 0.543 and 0.851, respectively. R ² was relatively large, indicating good model fit accuracy. Pattern recognition analysis was carried out pairwise on different groups to reflect differences in samples and overall metabolic differences. A significant separation among the HFD, NCD, and HFD-I3C groups was visible in a PCA score chart, suggestive of alterations in serum metabolites triggered by HFD. These metabolites were also changed after I3C administration (Figures 12A–C). The metabolic characteristics were compared among these three groups through H-NMR metabolomics analysis (Figures 12D,E). Relative to the NCD mice, the HFD mice displayed 9 differentially expressed metabolites (Table 1). The HFD and HFD-I3C groups exhibited 6 differentially expressed metabolites (Table 2). The intersection metabolites among the aforementioned three groups were argininosuccinic acid and galactose (Figure 12G).

We analyzed the potential correlation between gut microbiota abundance (genus level) and serum metabolites (Figure 12F; Table 3) and found a strong and broad correlation between bacterial species and serum metabolites. Akkermansia, Eisenbergiella, Ligilactobacillus, and Rikenellaceae_RC9_gut_group, were correlated with the changes in metabolites Argininosuccinic acid and D-(+)-Galactose in serum.