The animal procedures were approved by the Ethics Committee of Experimental Animals at Jiangnan University [JN.No20190930c0501205(256)], following Directive 2010/63/EU guidelines. Animals were allowed to adapt to the environment (21°C~23°C, 50%~60% of humidity, 12:12h light-dark cycle) for seven days before the experiment. The group and experimental schedule are shown in Figure 1A (N=8-10 per group).

The chronic restraint stress (CRS) paradigm was used to establish the depression-like phenotype (15). Mice were exposed to the restraint stress three h per day (from 09:00 to 15:00 every day, 1hour/each time, three times in total) for 14 consecutive days. Each mouse was put into a 50 mL plastic tube with a breathing hole and no food and water access. Mice in the control group fasted simultaneously during the CRS procedure. After the stress, all the mice were placed back in their home cages.

The protocols of open field test (OFT), tail suspension test (TST), and forced swim test (FST) were described in detail in our previously published papers (13). For the OFT, mice were assessed for their locomotor activity in a square arena (50 cm×50 cm) enclosed by continuous opaque walls made of plexiglass. Animals were monitored for 10 minutes, and the time spent in the center zone and active time during the whole period were measured. For the TST, animals were hanged using adhesive tape to fix the tails to a suspension bar (30 cm high from the floor). Locomotor activity is monitored for six minutes, and the duration of immobility was measured. For the FST, mice were allowed to swim in the water with a temperature of 23-25°C, and the immobility time was recorded during the ten-minute swimming. All mice are trained to swim for 15 minutes the day before the test. Before all the behavioral tests, animals were habituated to the room for one hour. All tests were performed under dim light (60 lux) and monitored by the video tracking system (Ethovision version 13, Noldus, Netherland).

SCFA-acylated starches were prepared as previously described (16, 17). Briefly, 40% (w/v) high amylose corn starch solution was continuously stirred in the water bath with a constant temperature of 40°. The pH was maintained at 8.00 by the addition of dilute NaOH (0.5 mol/L). Meanwhile, the acetic anhydride, propionic anhydride, butyric anhydride, and isobutyric anhydride were dripped into the starch solution, respectively, to reach a final amount of 20% (w/v). The solution was continued to be stirred for 2 hours, then adjusted to a pH of 5.70 using HCl (0.5 mol/L). The precipitation was collected by centrifuge (6000×g, 5min) and washed with water until the lotion was neutral. After a suction filtration at vacuum (-20 kPa, 10min), the precipitation was freeze-dried to get the SCFA-acylated starches. The starch solution without adding anhydride but processed as the same procedure above was used as control. The methods of measuring the degree of substitution of the SCFA-acylated starches and the results were described in Supplemental 1.1 .

SCFAs levels were measured as previously described with some modifications (12, 18). Fecal samples from the distal ileum and cecum are weighed (about 50 mg), homogenized, and lyophilized. The samples were soaked in 500 μl of saturated NaCl solution for 30 min. The mixture was homogenized, and 40 μL 10% (w/v) sulfuric acid was added. 1 mL ether was added to the mixture and vortexed following centrifugation (18,000g, 15 min, 4°C). 500 μL supernatant was filtered (0.22 μm) before further analysis. SCFA concentrations were determined in the TSQ 9000 GC-MS system (Thermo Scientific) equipped with an Rtx-WAX capillary column, using helium as carrier gas (flow rate of 0.89 mL/min). A gradient profile of the oven temperature was used, starting at 100°C and increased to 140°C (7.5°C/min), then increased to 200°C (6.0°C/min) and maintained for 3 min. The ion source and interface temperature in the mass spectrometer were 220°C and 250°C, respectively.

Serum corticosterone and inflammatory cytokines (TNF-α, IL-β, and IL-6) were measured using enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturer’s protocol (SenBeiJia Biological Technology Co., Ltd., Nanjing, China).

Total RNA of the colon tissues was extracted using Trizol reagent (Invitrogen, USA). Complementary DNA was prepared using the HiFiScript gDNA Removal cDNA Synthesis Kit (ComWin Biotech Co., Ltd., China). Transcription levels of the zonula occludens-1 (ZO-1), Claudin, and Occludin were determined by quantitative PCR. The gene expression was normalized to Gapdh based on the cycle threshold (Ct) values and the 2^−ΔΔCt method. All samples are measured in triplicate. Primer sets information are shown in Supplemental Table S1 .

16S rRNA gene amplicon sequencing of the fecal microbiome was performed using universal primers (341F/806R, for V3-V4 region) as previously described. Raw data were processed using the QIIME2 software package, and the specialized bioinformatic analysis was performed using the online software of MicrobiomeAnalyst (https://www.microbiomeanalyst.ca), including alpha diversity, beta diversity, and network analysis (19, 20). Gut metabolic modules (GMMs) analysis was performed using the R version of the Gomixer tool as previously described (21, 22). Detailed bioinformatical methods are described in Supplemental 1.2 .

Data are presented as means with SEM. All data were checked for normality by the Shapiro-Wilk test. Unpaired Student’s t-test was performed between the Control and Depressed groups. One-way ANOVA followed by Dunnett’s multiple comparisons test against the Depressed group was performed to compare the effects of SCFA-acylated starches in the CRS-treated animals. A criterion for significance was set to P<0.05 in all comparisons. The P-value of multiple comparisons was adjusted by family-wise significance and confidence levels of 0.05 (95% confidence interval).

A two-week CRS induced significantly depressive behaviors, such as the significantly reduced active time in the OFT (P=0.002; Figure 1B ) and increased immobility time in the FST (P=0.011; Figure 1D ). The immobility time in the TST was increased but without any statistical difference (P=0.308; Figure 1C ). Acetylated, butylated, and isobutylated starch reversed the animal’s behavioral abnormality in the OFT ( Figure 1B ; P<0.001, P=0.044, P<0.001, respectively) and TST (P=0.011, P=0.007, P=0.013, respectively; Figure 1C ). Butylated starch also reduced the animal’s immobility time in the FST, with a trend statistical difference (P=0.056, Figure 1D ). In addition, the stressed mice also showed a significantly reduced time spent in the center zone of the open field, but no SCFA-acylated starch reversed this abnormality ( Figure S1 ). The hyperactivity of the hypothalamus-pituitary-adrenal axis was also observed in the stressed mice, as reflected by the significantly increased serum corticosterone level (P=0.002; Figure 1E ). Acetylated, butylated, and isobutylated starch normalized the serum corticosterone level ( Figure 1E ; P=0.031, P=0.005, P=0.040, respectively).

In the distal ileum, the propionate (P=0.001), butyrate (P=0.009), and isobutyrate (P<0.001) levels were significantly increased when compared with the depressed mice, while the acetate level was not changed by the acetylated starch ( Figures 2A–D ). In the cecum, the acetate (P=0.044) and butyrate (P=0.024) levels were significantly decreased in the depressed mice, and the abnormalities were reversed by the acetylated (P=0.035) and butylated (P=0.001) starch treatment ( Figures 2E, G ). The cecum propionate (P=0.009) and isobutyrate (P<0.001) levels were also significantly increased when compared to the depressed mice ( Figures 2F, H ). All above data indicated that the SCFA-acylated starches resulted in a release of SCFAs in the gut, especially in the large intestine.

Gut barrier function was evaluated by the gene expression of tight junction protein levels. CRS induced a significant reduction of the transcriptional level of the ZO-1 gene P=0.049), the Claudin gene (P=0.038), and the Occludin gene (P=0.048). Intriguingly, the butylated starch showed a superior effect to others to restore the gut barrier ( Figures 3A–C ). The deficit of gut barrier resulted in increased serum lipopolysaccharide (LPS; Figure 3D ) and inflammatory cytokines, including serum IL-1β (P=0.001), IL-6 (P<0.001), and TNF-α (P=0.074) ( Figures 3E–G ). Butylated starch significantly reduced the serum IL-1β (P=0.025) and IL-6 (P=0.002) levels. The isobutyrate starch could also reduce the serum IL-6 level (P=0.036; Figure 3F ). In addition, butylated starch reduced the serum TNF-α level, with a trend statistical difference (P=0.054, Figure 3D ).

CRS induced a significantly increased gut microbial alpha-diversity, as reflected by the Shannon (P<0.001) and Chao1 (P=0.001) index ( Figures 4A, B ). The butylated starch treatment normalized the Chao1 changes (P=0.003; Figure 4B ). Besides, the beta diversity estimated using Aitchison distance and principal component analysis (PCA) showed that every two groups are significantly different (PERMANOVA results in Figure 4C ). The linear discriminant analysis (LDA) effect size algorithm (LEfSe) identified five genus-level biomarkers of the butylated starch treated gut microbiome, including Bacteroide, Ruminiclostridium, Odoribacter, and Oscillibacter ( Figure 4D ). The relative abundance of these four taxa was all significance decreased under stress, and the butylated starch treatment increased the Odoribacter (P=0.072) and Oscillibacter (P=0.048) abundances ( Figure 4E ). The microbial genomic functions were analyzed based on the gut-brain modules as previously described (22). A total of 17 pathways were identified, and four of them are significantly affected by the CRS and reversed by the butylated starch treatment, including striatal dopamine (DA) and metabolite (DOPAC) synthesis, Dopamine degradation, Inositol degradation, and Tryptophan synthesis ( Figure 4F ).