C57BL/6J male mice (Animal Center, Third Military Medical University) were bred, no more than six mice per cage in the animal facility of Third Military Medical University. Animals were housed in a controlled environment (12 h per day/night cycle and lights off at 19:00) with free access to food and water. All animal protocols were approved by the institute of Animal Care and Use Committee (Approval number: SYXC-2014-00203). The animal experiments were performed in accordance with the Guidelines for the Care and Use of Laboratory Animals at Third Military Medical University.

After 1 week of acclimation (week 0) on a normal chow diet, 5-week-old mice (n = 12) were randomly divided into two groups of six and fed the following diets (Luo et al., 2012): (1) a HFD containing 45% fat and (2) a HFD with 0.01% CAP (a gift from Prof. Zhiming Zhu, Daping Hospital, Third Military Medical University) treatment in food (HFD-CAP). Animals were given unrestricted access to food and water.

Body weight and average food intake were monitored every day in the 1st week (week 1) after treatment and once a week after week 1. During week 9, mice feces were collected and stored at −80°C until DNA extraction, and oral glucose tolerance test (OGTT) was assessed. At last, animals were anesthetized with amobarbital and then sacrificed by cardiac puncture. Gut tissues were collected for histological staining or tissue RNA extraction.

For RT-qPCR experiment, sixteen 6-week C57BL/6J male mice were randomly divided into four groups and all were fed orally with CAP (0.4 mg in 200 μL PBS and 0.4 mg is computed according to the daily consumption of 4 g of chow per mice multiplied by 0.01%), At the four time points (0, 30, 60, and 90 min), one group of mice (n = 4) were anesthetized with amobarbital, and the jejunum and proximal colon tissue samples were collected for RNA extraction.

Mice were fasted overnight (17:00–9:00), and an OGTT was performed after gavage with glucose (1 g/kg body weight, volume: ∼200 μL) (Anhê et al., 2015). Blood glucose concentrations were measured with glucometer (Johnson, USA) before (0 min) and after (15, 30, 60, 90, and 120 min) glucose challenge (Ayala et al., 2010; Anhê et al., 2015).

Fecal bacterial DNA was extracted using QIAamp DNA Stool Mini Kit (QIAGEN, German) from approximately 100 mg feces according to the manufacturer’s protocol. For each sample, the V4 hypervariable region of the bacterial 16S rRNA gene was amplified using primers 515F (5′-GTGCCAGCMGC CGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAA T-3′) (Caporaso et al., 2011), purified, and then sent to Novogene Corporation for the generation of sequencing libraries by using TruSeq DNA PCR-Free Sample Preparation Kit (Illumina, USA) and sequenced on the Illumina Hiseq 2500 platform.

After de-multiplexing of the paired-end 250 bp raw reads, Trimmomatic (v0.32) (Bolger et al., 2014) was used to clean low-quality reads and remove adapter sequences. Paired ends were merged using FLASH (v1.2.7) (Magoc and Salzberg, 2011) and filtered under specific filtering conditions to obtain high-quality clean tags according to the QIIME (v1.91) (Caporaso et al., 2010) quality control process. The OTU picking processing adopted an open-reference OTU picking protocol by searching reads against the Greengenes (gg_13_8) (DeSantis et al., 2006) database (similarity ≥ 0.97) according the QIIME tutorial¹ and the diversity analyses followed the same tutorial. LEFSe (Version 1.0) (Segata et al., 2011) was used to discover the metagenomic biomarker for both organisms and pathways. PICURSt (Version 1.0.0-dev) (Langille et al., 2013) and HUManN (Version 0.99) (Abubucker et al., 2012) were used to predict functional profiling of microbial communities using 16S rRNA amplicon sequencing result following the PICURSt tutorial^2,3 .

Determination of absolute abundance of A. muciniphila followed the method from Everard et al. (2013). Briefly, A. muciniphila specific primers (forward: 5′-CAGCACGTGAAG GTGGGGAC-3′, reverse: 5′-CCTTGCGGTTGGCTTCAGAT-3′) were used to detect A. muciniphila through real time PCR. A standard curve (performed in triplicate) was made by fivefold serial dilutions of genomic DNA of A. muciniphila (ATCC BAA-835). And the cycle threshold of each sample was compared with the standard curve.

Akkermansia muciniphila strain ATCC BAA-835 was purchased from the ATCC. Strains were cultured in BHI medium in tubes at 37°C in anaerobic chamber (Whitley A35 Workstation 2.5, Don Whitley Scientific, UK). To acquire the growth curve of A. muciniphila, 3 mL cell cultures were grown with different concentrations (2, 20, and 200 μg/mL) of CAP or with equal volume (2 μL for 1 mL medium) of ethanol. Mucin (from porcine stomach, Type II, Sigma, USA, #SLBH5886V) was added to the final concentration of 4 g/L when needed. CAP (Sigma, USA, #MKBS2249V) was dissolved in ethanol. Cell density was measured by serial 10-fold dilution and plating on BHI agar plates (1% agar). Each experiment was repeated three times.

Total RNA was extracted from the jejunum and proximal colon using TriPure (Roche, Switzerland). Genomic DNA removal and cDNA synthesis were performed using TransScript II One-Step gDNA Removal and cDNA Synthesis SuperMix (Transgen, China). Real-time PCR was performed using SYBR Premix Ex Taq II (Takara, Japan) with 1:5 diluted cDNA products.

Target genes including Muc2 and Klf4 (a goblet cell marker), and Reg3g (an antimicrobial marker), gene were detected with hypoxanthine guanine phosphoribosyl transferase (Hprt) as the housekeeping gene (Anhê et al., 2015). Data were calculated according to the 2^-ΔΔCt method.

The distal ileum and proximal colon were stained with PAS, (Sigma, USA) as previously described (Shin et al., 2014). The goblet cells were stained pink to red and nuclei were blue.

Weight, glycemia, goblet cell count and gene expression level data are expressed as mean ± SEM. For different abundance taxon analyses in STAMP (Parks et al., 2014), Welch’s t-test was used for the comparison of two groups. One way ANOVA followed by Newman–Keuls post hoc was used for real time PCR results tests. Student’s t-test was used for other cases.

To assess the effect of CAP on obesity, 5-week-old male C57BL/6 mice were fed either with HFD or HFD-CAP diet for 9 weeks (Figure 1A). The HFD-CAP-fed mice had a lower (P < 0.05) body weight gain (6.52 ± 0.45 g) than HFD-fed mice (7.7 ± 1.29 g) (Figure 1B; Supplementary Figure S1A), which was associated with reduction of food intake (Supplementary Figure S1B). The sharp reduction of weight at week 1 could be explained by the reduction of food intake during the early days after suddenly change of food containing spicy CAP. Next, we used a glucose tolerance test to examine glucose homeostasis in CAP-treated and non-treated mice. Compared with HFD-fed mice, HFD-CAP-fed mice showed a smaller (P < 0.05) AUC (Figure 1C; Supplementary Figure S1C), which indicated an improvement in glucose tolerance. These findings confirmed that dietary CAP helped prevent obesity and improved glucose homeostasis under HFD conditions.

Fecal DNA of all the 12 mice at week 9 were extracted and sequenced by 16S rRNA gene amplicons sequencing. The average effective tags per sample were 54 thousand (ranging from 47 thousand to 63 thousand tags/sample).

The NMDS plot of the microbial compositions of the 12 samples revealed that the HFD and HFD-CAP group have similar microbiome composition (Figure 2A). However, at the phylum level, among the top 10 abundant phyla, the average proportion of Acidobacteria, Bacteroidetes, and Firmicutes was increased in HFD-CAP group, while the abundance level of Deferribacteres and Proteobacteria was decreased (Figure 2B). According to Welch’s t-tests by STAMP (Parks et al., 2014), only the decrease (from 24.63 ± 4.18% to 16.92 ± 3.83%) of Proteobacteria abundance is significant (P = 0.013) (Figure 2D).

At the genus level, among the top 10 abundant genera, the average proportion of Helicobacter and Mucispirillum was decreased in the HFD-CAP group (Figure 2C). According to Welch’s t-tests by STAMP, the abundance levels of Bacteroides (from 0.56 ± 0.35% to 1.55 ± 0.66%), Coprococcus (from 0.88% ± 0.32% to 1.60 ± 0.57%), Prevotella (from 1.58 ± 0.84% to 3.07 ± 1.18%), and Akkermansia (from 0.03 ± 0.04% to 0.12 ± 0.07%) increased significantly (Figure 2E). Real-time PCR also confirmed the increase (p < 0.01) of absolute abundance of A. muciniphila (Figure 2F) in HFD-CAP group.

We compared the fecal microbiota in HFD and HFD-CAP groups using LEfSe (Segata et al., 2011) to identify the specific bacterial taxa associated with treatment of CAP. A cladogram representative of the structure of the fecal microbiota was shown in Figure 3B. The greatest differences in multiple levels of taxa between the two communities were displayed (Figure 3A). These data indicated the significantly decreased phylum Proteobacteria can be one of the biomarkers of HFD group. Family Bacteroidaceae and its genus Bacteroides can be the biomarkers of HFD-CAP group.

The changes in microbial taxa are associated with changes in functional gene abundances, as predicted from 16S rRNA data analysis using the PICRUSt (phylogenetic investigation of communities by reconstruction of unobserved state) software (Langille et al., 2013). LEFSe identified that three KEGG pathways, including lysine degradation, Glutathione metabolism, and Parkinsons disease, were enriched in HFD group, while two pathways, including cyanoamino acid metabolism and streptomycin biosynthesis, are enriched in HFD-CAP group (Figure 4A). At the module level (Figure 4B), the type IV secretion system modules and arginine/ornithine transport system are enriched in HFD group. Glycolysis (Embden-Meyerhof pathway), uridine monophosphate biosynthesis, and aminoethyl phosphonate transport system are enriched in HFD-CAP group.

To test whether CAP directly stimulated the growth of A. muciniphila in vitro, the growth curve of A. muciniphila strain ATCC BAA-835 was monitored in both BHI medium and BHI medium with mucin (from porcine stomach, Type II) (Figure 5). High concentration of CAP (20 or 200 μg/mL) inhibited the growth of A. muciniphila, whereas low concentration of CAP (2 μg/mL) did not promote the growth of A. muciniphila in vitro directly (Figures 5A,B). However, the addition of mucin directly promoted the growth of A. muciniphila in BHI medium (Figures 5B,C). When mucin was provided, A. muciniphila grew faster at log phase and plateaued at a much higher cell density. Thus, mucin could promote A. muciniphila growth.

The addition of mucin directly stimulated A. muciniphila growth, and mucin is secreted by goblet cells in vivo. Thus, we examined the number of goblet cells in jejunum and colon in HFD and HFD-CAP mice at week 9. PAS staining was used to stain goblet cells (Figures 6A,B), and PAS-positive goblet cells per villus were counted. An average of 13 goblet cells was seen in HFD-CAP mice, which did not show a significant difference from that in HFD mice (Figure 6C).

To further explore the effect of CAP on goblet cells, sixteen 6-week C57BL/6J male mice were randomly divided into four groups (n = 4) and all were fed orally with CAP (0.4 mg in 200 μL PBS, see Materials and Methods). Gene expressions at four time points (0, 30, 60, and 90 min) in jejunum and proximal colon tissue samples were detected by qPCR. We tested the Klf4 gene, a marker of goblet cells that regulated goblet cells’ differentiation. CAP did not directly increase Klf4 mRNA expression in jejunum or colon within 90 min after the oral administration of CAP (Figure 6D).

We next examined the expression of the major goblet cell mucin production gene (Muc2) after CAP treatment. Mucin 2 mRNA expression was increased twofold at 30 min after CAP administration and lasted for 1.5 h in the jejunum (P < 0.05).

Last, we measured the mRNA expression of Reg3g, an antimicrobial protein that is predominantly expressed in the gastrointestinal tract (Wang et al., 2016). Reg3g has bactericidal activity and restricts bacterial colonization of mucosal surfaces. As is shown in Figure 6D, according the mean expression abundance values, Reg3g mRNA expression is steadily up-regulated in the colon after CAP administration and increased to about threefold at 90 min. However, the differences were not statistically significant.