Indica rice was obtained from Guangxi Province, China. The moisture of the rice was determined by an OHAUS MB27 moisture meter and adjusted to 25% by adding pure water. The moisture-adjusted rice was put in a sealed glass container and heated in an oven (BGZ-70, Boxun Industry & Commerce Co., Ltd.) at 110°C for 2 h. The HMT conditions were referred to previous reports (9, 18). The obtained HMT-rice and the untreated indica rice (UT-rice) were milled and used for characterization and preparation of the mice diet.

Lipid, ash and protein contents were measured according to the AOAC Official Method 922.06, 923.03 and 988.05, respectively (19–21). Moisture content was measured using a moisture meter (OHAUS, America). Carbohydrate content was calculated as below (22):

X-ray diffraction (ADVANCE; Bruker, Germany) determined the crystal structures of HMT-rice and UT-rice, similar to a previous report (23). The divergent slit of CuK α Radiation (λ = 1.5405 nm) was set as 0.38 mm. The voltage and current were 40 kV and 40 mA, respectively. The diffraction angle (2θ) from 4° to 40° was scanned.

The procedure was like previous reports (24, 25) with minor modifications. Forty milliliter deionized water was used to dissolve 4.5 g pancreatin (Sigma-Aldrich, United States) and the solution was heated in 37°C water with shaking for 10 min. Afterward, centrifugation at 5,000 × g for 10 min was conducted to obtain the supernatant. A total of 3.2 mL amyloglucosidase (Sigma-Aldrich, United States) was blended with 27 mL supernatant. A 0.2 g (dry weight) sample of HMT-rice or UT-rice was blended with 4 mL deionized water. The mixture was thoroughly blended and heated in boiling water for half an hour. After cooling to 37°C, 4 mL sodium acetate buffer (0.5 M, pH 5.2) and 2 mL enzyme mixture were added. The mixture was thoroughly mixed and heated in 37°C water with shaking for further reaction. At time points of 0, 20, 60, 90, and 120 min, a 0.1 mL reaction solution was blended with 90% ethanol to inactivate the enzyme. Centrifugation at 10,000 × g for 5 min was performed to obtain the supernatant. The glucose oxidase-peroxidase kit (Beijing Lideman Biochemical Co., Ltd., China) was used to measure the glucose concentration. The RS content of the samples was calculated from the glucose content at 120 min (G120) as follows:

where 0.9 is the conversion factor, S is the sample weight.

After feeding for 11 weeks, oral glucose tolerance test (OGTT) was carried out. The mice were administered 2 g/kg body weight glucose solution by gavage after fasting overnight. At 0, 15, 30, 60, 90 and 120 min, an Omron blood glucose meter was used to record the blood glucose concentration.

The mice were euthanized via isoflurane anesthesia followed by cervical dislocation after fasting in the 12th week. Blood, fat (including epididymal, perirenal, and subcutaneous fat), liver, and intestinal tissues were collected and stored at −80°C. A 4% paraformaldehyde was used to preserve the liver and intestinal tissues. Centrifugation of the blood sample at 4,000 rpm and 4°C was able to separate the serum. An ELISA kit (Jingmei Biotech Co., Ltd., China) was used to measure blood insulin. Nanjing Jiancheng Bioengineering Institute kits were used to measure blood lipids including total cholesterol (CHO), high- and low-density lipoprotein cholesterol (HDL- and LDL-C), and total triglyceride (TG).

Oil red O staining was used for the liver tissues and the fat droplet area (estimating Oil red O staining score) was estimated by the Image J software (National Institutes of Health, America). Hematoxylin and eosin (H&E) staining was used for colon and liver tissues. Images were captured through an optical microscope.

Six mice in each group were randomly selected to perform the metabolomic analyses. The randomization was performed using a computer-generated random number list.

For detecting serum metabolites, serum and precooled acetonitrile were mixed (1:15 volume ratio) by vortex. Centrifugation of the mixture at 12,000 rpm and 4°C was able to separate the supernatant. After being filtered by a 0.22 μm membrane, the filtrate was injected into an Agilent 6530C liquid chromatography equipped with quadruple-time-of-flight mass spectroscopy. A Waters™ ACQUITY BEHC18 column (100 mm × 2.1 mm, 1.7 μm) was used with a gradient elution of acetonitrile and 0.1% formic acid (flow rate 0.4 mL/min). The range of 20–1,000 m/z was scanned. The flow rate and temperature of the dry gas were 7.9 L/min and 340°C, respectively. A capillary voltage of 4,500 V was used.

The fecal metabolites were measured similarly as the serum metabolites. Briefly, A 1:1 ratio of methanol and ultrapure water was precooled and used to extract 85–95 mg cecal content (5 μL extract/mg cecal content). The mixture was homogenized using zirconium beads and processed as the serum and acetonitrile mixture above.

ABf Converter (version 4.0.0) transformed the raw data to abf. MS-DIAL (version 4.38, RIKEN CSRS, University of California at Davis) was used to extract and align the peaks and reduce noise. Criteria for differentiating metabolites were P (false discovery rate-corrected) < 0.05 and fold change > 1.2 (or <0.83). The metabolites were determined by accurate mass using the Human Metabolome Database. Heatmap, metabolic function analyses, and principal component analyses graphs were generated by Metaboanalyst 5.0.

Data were shown as mean ± SD. Statistical analyses were carried out by SPSS software (Version 27.0). A one-way analysis of variance (ANOVA) test evaluated differences among the groups; and the subsequent pairwise comparison was carried out by the least square method (if the sample size was equal among the groups) or the Bonferroni method (if the sample size was not equal). For comparing blood glucose during OGTT and weight over time, a two-way repeated measured ANOVA was used. p < 0.05 suggested statistical significance. The association between differential metabolites and clinical indices was assessed by Spearman correlation analysis. Cytoscape (Version 3.9.0) was used to plot the interaction map between serum & fecal metabolites and blood glucose/lipid parameters.

The X-ray diffraction patterns of both UT-rice and HMT-rice (Figure 1A) displayed peaks at around 15°, 17°, 18°, 20°, and 23°, which indicated A-type crystallinity of rice starch (22, 26). The chemical composition of HMT-rice was not significantly different from that of UT-rice (Table 1).

In-vitro digestion showed that the composition of resistant starch in HMT-rice was significantly higher than that in UT-rice (Figures 1B; p < 0.05). Consistently, in-vivo digestibility experiments (Figures 1C,D) showed that the mean blood glucose level after the intake of HMT-rice was significantly reduced compared with that after the intake of UT-rice (p < 0.05); and the area under curve (AUC) after the intake of HMT-rice was also significantly reduced compared with that after the intake of UT-rice (p < 0.05), consistent with the literature (9).

The effects of HMT-rice on blood glucose/lipids were displayed in Figure 2. In the OGTT, the HFD group had a significantly larger AUC of the blood glucose curve than the control (p < 0.05); while the HMT-rice group had a significantly smaller AUC than the HFD group (p < 0.05). The differences in blood glucose and insulin among the three groups were not significant. The HFD group had significantly larger blood CHO and HDL-C than the control (p < 0.05). The three groups had similar blood TG and LDL-C.

The HFD-fed mice acquired weight much more rapidly than the control mice (p < 0.05); the HMT-rice-treated mice acquired weight slightly slower than the HFD-fed mice without statistical significance (Figure 3A). The HFD-fed mice had significantly heavier epididymal and subcutaneous fat than the control mice (p < 0.05); the HMT-rice-treated mice had slightly lighter epididymal and subcutaneous fat than the HFD-fed mice without statistical significance (Figures 3B–D). The differences were not significant in perirenal fat and liver weight among the groups (Figures 3C,E).

The liver and colon histopathology were displayed in Figures 3F–I. The HFD group had a significantly higher Oil Red O staining score than the control (Figures 3F,G, p < 0.05); and the HMT-rice group had a significantly lower score than the HFD group (p < 0.05). The H&E staining indicated that more fat vacuoles appeared in the liver of the HFD-fed mice than that of the control mice; while the HMT-rice-treated mice had less fat vacuoles in the liver than the HFD-fed mice (p < 0.05). The H&E staining of the colon tissue suggested that the goblet cell number declined and the integrity of the epithelial structure was destroyed in the HFD-fed mice (versus the control mice). These changes were alleviated in the HMT-rice-treated mice.

Principal Component Analysis (PCA; Figure 4) plots showed clear separated metabolic profiles among the three groups (p < 0.05) and the detailed differential metabolites among the three groups were summarized in Supplementary Tables A.2–A.7.

In hierarchical cluster analysis, the serum differential metabolites were classified into two clusters according to changing trends (Figure 5A). In Cluster One, the HFD and HMT-rice group had 23 significantly decreased serum metabolites versus the control group. These metabolites mainly included lysophospholipids, bile acid, and amino acid derivatives. In Cluster Two, the HMT-rice group had 13 significantly increased serum metabolites versus the control and the HFD groups including 4 lysophospholipids, lysoPE(24:0/0:0), lysoPC(P-18:0/0:0), lysoPE(22:5(4Z,7Z,10Z,13Z,16Z)/0:0) and lysoPC(22:5(4Z,7Z,10Z,13Z,16Z)/0:0).

The top 100 fecal differential metabolites were shown in Figure 5B, which were classified into three clusters. In Cluster One, the HFD group had 53 significantly enriched fecal metabolites versus the control group which were restored in the HMT-rice group including oxidized phospholipids, long-chain fatty acids, amino acid and bile acid derivatives. In Cluster Two, the HFD and HMT-rice groups had 17 significantly increased metabolites versus the control group. In Cluster Three, the HFD and HMT-rice groups had 30 significantly decreased metabolites than the control group.

Differential metabolic pathways between groups were determined by metabolic functional analysis based on Metabolite Functional Network (Figure 6; Supplementary Tables A.8–A.13). The differential pathways based on serum metabolites were represented in Figures 6A–C. Versus the control group, the HFD group had top distinct pathways including taurine and hypotaurine metabolism, biosynthesis of unsaturated fatty acid, and fatty acid biosynthesis. Versus the HFD group, the HMT-rice group had top distinct pathways including one carbon pool by folate, valine, leucine and isoleucine biosynthesis and degradation.

The pathways based on fecal metabolites were shown in Figures 6D–F. Versus the control group, the HFD group had top distinct pathways including biosynthesis of unsaturated fatty acids, steroid hormone biosynthesis, and sphingolipid metabolism. Versus the HFD group, the HMT-rice group had top distinct pathways including steroid hormone biosynthesis, fatty acid degradation, and sphingolipid metabolism.

Correlation between the differential serum/fecal metabolites (HFD vs. HMT-rice) and clinical indicators was assessed by a Spearman rank correlation analysis (Figure 7; Supplementary Tables A.14, A.15). Inverse correlations were observed between fasting blood glucose and three serum metabolites including lysoPE(24:0/0:0), N-acetyl-L-phenylalanyl-3,5-diiodo-L-tyrosine, and sagerinic acid. Seventeen fecal metabolites were positively correlated with the fasting blood glucose mainly including several bile acid and amino acid derivatives, tetracosahexaenoic acid, and PG(22:5(4Z,7Z,10Z,13Z,19Z)-O(16, 17)/i-18:0). Six fecal metabolites were positively related to the CHO level.