Resistant starch decreases intrahepatic triglycerides in patients with non-alcoholic fatty liver disease

Journal: Cell Metabolism

Key Finding: This landmark study found that participants who consumed 40g of resistant starch daily saw a nearly 40% reduction in liver fat (triglycerides). The study proved that fiber alters the gut microbiota, specifically reducing Bacteroides stercoris, which directly influences how the liver metabolizes fat.

Link: https://www.cell.com/cell-metabolism/fulltext/S1550-4131(23)00297-8

Resistant starch decreases intrahepatic triglycerides in patients with NAFLD via gut microbiome alterations

Yueqiong Ni1,2,3,12 ∙ Lingling Qian1,12 ∙ Sara Leal Siliceo2,12 ∙ Xiaoxue Long1,12 ∙ Emmanouil Nychas2 ∙ Yan Liu4,5 ∙ Marsena Jasiel Ismaiah6,7 ∙ Howell Leung2 ∙ Lei Zhang1 ∙ Qiongmei Gao1 ∙ Qian Wu1 ∙ Ying Zhang1 ∙ Xi Jia4,5 ∙ Shuangbo Liu8 ∙ Rui Yuan8 ∙ Lina Zhou9 ∙ Xiaolin Wang9 ∙ Qi Li9 ∙ Yueliang Zhao8 ∙ Hani El-Nezami6,7 ∙ Aimin Xu4,5,10 ∙ Guowang Xu9 Send email to xugw@dicp.ac.cn ∙ Huating Li1 Send email to huarting99@sjtu.edu.cn ∙ Gianni Panagiotou2,5,11 Weiping Jia 1,13

Highlights

RS intake for 4 months reduces IHTC effectively and independently of weight loss
Gut microbiome alteration mediates the effect of RS on alleviating NAFLD
Microbiome-driven LPS and BCAA production promotes NAFLD
B. stercoris is a key species in NAFLD progression

Summary

Non-alcoholic fatty liver disease (NAFLD) is a hepatic manifestation of metabolic dysfunction for which effective interventions are lacking. To investigate the effects of resistant starch (RS) as a microbiota-directed dietary supplement for NAFLD treatment, we coupled a 4-month randomized placebo-controlled clinical trial in individuals with NAFLD (ChiCTR-IOR-15007519) with metagenomics and metabolomics analysis. Relative to the control (n = 97), the RS intervention (n = 99) resulted in a 9.08% absolute reduction of intrahepatic triglyceride content (IHTC), which was 5.89% after adjusting for weight loss. Serum branched-chain amino acids (BCAAs) and gut microbial species, in particular Bacteroides stercoris, significantly correlated with IHTC and liver enzymes and were reduced by RS. Multi-omics integrative analyses revealed the interplay among gut microbiota changes, BCAA availability, and hepatic steatosis, with causality supported by fecal microbiota transplantation and monocolonization in mice. Thus, RS dietary supplementation might be a strategy for managing NAFLD by altering gut microbiota composition and functionality.

Introduction

An estimated 30% of the world’s population currently has non-alcoholic fatty liver disease (NAFLD), which has reached epidemic proportions globally.¹^,² It is a multisystem disease that may not only develop into severe chronic hepatic diseases but also contribute to extrahepatic diseases such as type 2 diabetes, cardiovascular disease, and chronic kidney disease, causing a tremendous clinical and economic burden.³^,⁴ A recent large nationwide cohort study with long-term follow-up showed significantly increased overall mortality with all NAFLD histological stages including steatosis,⁵ thus suggesting that steatosis can no longer be ignored as “benign and an incidental finding.”⁶ Although there have been clinical trials exploring drug candidates, no pharmacological treatments have been approved for NAFLD so far.⁷ Hence, effective intervention strategies are urgently needed to delay or halt its progression to related hepatic and extrahepatic diseases.

Accumulating evidence suggests that NAFLD is a disease closely related to gut microbiota via the gut-liver axis,⁸^,⁹ which stimulates the efforts to explore therapeutic interventions to improve NAFLD by modulating the gut microbiota.¹⁰ The safety and persistence needed for microbiota-targeted intervention in humans highlight the importance of exploring microbiota-directed foods (MDFs), which by definition can elicit a targeted metabolic response in specific indigenous microbiota that confer a health benefit on the host.¹¹ Prebiotics and synbiotics like oligofructose and yogurt, which can manipulate gut microbiota, were found to reduce insulin resistance (IR), intrahepatic lipids, liver enzymes, and histologically confirmed steatosis in patients with NAFLD/nonalcoholic steatohepatitis (NASH).¹²^,¹³^,¹⁴ Despite the promise of MDFs in patients with NAFLD, such studies are in an early stage.¹⁵ According to the NAFLD practice guidance from the American Association for the Study of Liver Diseases, rigorous, prospective, longer-term trials are required before making recommendations about specific diets.¹⁶ The complexity and heterogeneity of the NAFLD pathogenesis calls for deep and extensive phenotyping to evaluate the multiple effects of an intervention on NAFLD and their molecular mediators.¹⁷ Furthermore, subsequent causal investigations are needed to verify the specific microbial signatures identified in clinical studies and their metabolites, which represent the highest levels in the chain of evidence in microbiome-linked disease.¹⁸

In line with the above, we performed here a randomized, double-blinded, placebo-controlled clinical trial in individuals with NAFLD that lasted for 4 months to enable a relatively long-term observation. We used as MDF resistant starch (RS), a prebiotic of nondigestible fibers that are fermented in the large intestine,¹⁹ which has been shown to reduce adiposity and exert metabolic benefits in previous animal studies²⁰^,²¹^,²²^,²³; however, so far, no clinical study has investigated the therapeutic effect of RS on NAFLD. Comprehensive clinical measurements were conducted to evaluate the changes in metabolic phenotypes of NAFLD during the intervention. Multi-omics profiling was used to provide an integrated understanding of how RS and associated alterations in the gut microbiota or metabolites contributed to NAFLD improvement. In addition, the potential RS-targeted gut species and microbial metabolites revealed by multi-omics analysis were validated in mice and cell lines for causal insights.

Results

RS intervention for 4 months alleviates NAFLD in Chinese adults

To investigate the effects of RS on NAFLD, we conducted a randomized, double-blinded, placebo-controlled clinical trial in Shanghai, China, from 2016 March to 2017 October (ChiCTR-IOR-15007519). A total of 200 participants with NAFLD (145 males and 55 females) were recruited and randomized with a 1:1 allocation to the 4-month administration of RS type 2 from high-amylose maize (HAM-RS2, 40 g/day) or control starch (CS) with equal energy supply (Figures 1A and S1A). The average age of all the participants receiving randomization was 39.1 ± 9.1 years (mean ± SD), while the average intrahepatic triglyceride content (IHTC) was 24.12% ± 14.64% (mean ± SD). Both groups were counseled to manage their diet following the standard menu designed by nutritionists. We measured the IHTC by magnetic resonance spectroscopy (MRS) during interventions along with anthropometric parameters and biochemical indexes. Four participants (3 in the CS group and 1 in the RS group) did not receive the corresponding intervention after randomization and were therefore excluded from the primary analysis (Figure S1A). Baseline anthropometric and clinical characteristics of participants were balanced between the two groups (Table 1). During the 4-month (120-day) intervention, the median (lower/upper quartile) percentage of the meals for which participants adhered to starch intake was 90.0% (85.0%, 93.3%) in CS and 91.7% (86.7%, 96.7%) in the RS group, with no significant difference (Figure S1B). Similarly, no significant difference was found in the adherence to diet (Figure S1C; Table S1). Dietary intake of energy and macronutrients except fiber was not significantly different between the two groups (Table S1).

After the 4-month intervention, the primary outcome IHTC was significantly decreased in the RS group compared with the CS group (p < 0.0001) (Figure 1B). The net absolute and relative change of IHTC in the RS group relative to the CS group was −9.08% (95% CI: −11.91% to −6.26%) and −39.42% (95% CI: −56.13% to −22.72%), respectively (Table 1). Together with the alleviation of steatosis, we observed significant reduction of body weight and BMI in the RS group compared with the CS group (Figure 1C). The waist circumference, hip circumference, and waist-hip ratio (WHR) in the RS group were all lower compared with the CS group. Regarding the body composition, the reduction of fat percentage (FAT%) and fat mass (FM) were all significantly higher in the RS group compared with the CS group (Table 1). The reduction of visceral fat areas (VFAs) and subcutaneous fat areas (SFAs) evaluated by abdominal magnetic resonance imaging (MRI) was significantly higher after RS consumption compared with CS consumption (Figures 1D and 1E).

Furthermore, we observed significant reductions in alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transpeptidase (GGT) after RS intervention (Figures 1F–1H; Table 1), which indicate the improvements of liver injury. The dyslipidemia was also alleviated by the RS intervention as shown in the improvement of total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C), which was absent after CS intervention (Table 1). Notably, fibroblast growth factor 21 (FGF21), a generally acknowledged NAFLD biomarker,²⁴ was reduced after RS consumption (Figure 1I). The level of CK18 M65ED, which correlates with hepatocyte apoptosis and independently predicts the presence of NASH,²⁴ was also significantly lower after RS compared with CS consumption (Table 1). In addition, the circulating levels of lipopolysaccharides (LPSs) and other inflammatory markers including monocyte chemoattractant protein-1 (MCP-1), interleukin (IL)-1β, and tumor necrosis factor alpha (TNF-α), were all significantly reduced after RS intervention in comparison to CS intervention (Figures 1J–1M).

Although the fasting blood glucose level was reduced in both groups after the 4-month intervention, neither fasting nor postprandial glucose levels during the meal tolerance test demonstrated significant differences between the RS and CS interventions. Both the fasting and postprandial insulin levels were significantly decreased in the RS compared with CS group, as well as the IR evaluated by homeostasis model assessment (HOMA-IR) and IR index of adipose tissue (Adipo-IR) (Table 1). Besides, the 4-month RS consumption also resulted in cardiovascular improvements as the blood pressure was significantly decreased compared with the CS consumption (Table 1). Due to dietary management, significant reductions of adiposity and several metabolic parameters were also observed in the CS group, albeit significantly smaller than the RS group.

We also performed a secondary analysis to adjust for the effect of weight loss. The net absolute change of IHTC in the RS group relative to CS group after adjusting for weight loss was −5.89% (95% CI: −8.87% to −2.91%), corresponding to a relative change of −24.30% (95% CI: −42.42% to −6.18%), and remained statistically significant (p = 0.0001) (Table 1). A regression analysis associating absolute change of IHTC with weight loss showed an R² of 23%, suggesting only a small part of the RS effect was mediated by weight loss. Although some clinical parameters showed weight-loss-dependent changes, the changes of other parameters related to adiposity (VFA), glucose metabolism (insulin levels, HOMA-IR, Adipo-IR), lipid metabolism (TG, TC, HDL-C, LDL-C), hypertension (systolic blood pressure [SBP], diastolic blood pressure [DBP]), and ALT all remained significant (Table 1). Moreover, we included all randomized participants including four participants who did not receive the corresponding intervention in our sensitivity analysis, and the conclusion remains the same (Table S2). Collectively, a 4-month RS intervention reduced IHTC and improved liver injury and related metabolic disorders in patients with NAFLD, even after adjusting for weight loss.

RS intervention alters both fecal and serum metabolites in patients with NAFLD

To investigate how the 4-month RS intervention affected the metabolism of the human host and the commensal intestinal microbiota, we performed targeted metabolomics on serum and fecal samples of participants in both the RS and CS groups before and after the intervention. In total, we measured 30 amino acids (AAs) and 26 bile acids (BAs) in serum and 10 short-chain fatty acids (SCFAs) and 18 BAs in feces. The RS and CS interventions had different effects on the overall changes in measured metabolites (p < 0.05, PERMANOVA) (Figure 2A). Examination of the different categories of metabolites showed a small but significant change in both serum and fecal BA profiles (p < 0.05, PERMANOVA).

Figure 2 Fecal and serum metabolomic changes after 4 months of resistant starch (RS) and control starch (CS) interventions

At the level of individual metabolites, 13 metabolites among fecal BAs, serum BAs, and serum AAs were significantly changed (p < 0.05, Wilcoxon signed-rank test) by the RS intervention but not the CS intervention (Figure 2B). No differences were observed in fecal SCFA metabolites. All serum AAs with significant changes showed different directions of change between the RS and CS groups. Interestingly, the serum levels of all three branched-chain amino acids (BCAAs) (valine, leucine, and isoleucine) decreased after the RS intervention. The glutamate-serine-glycine (GSG) index, a possible marker of liver disease severity that is independent of BMI,²⁵ was also significantly reduced after the RS intervention. In addition, we found 10 metabolites (Figure S2A) showing no differences at baseline to be significantly different between the RS and CS groups at the end of the intervention (p < 0.05, Wilcoxon rank-sum test), including valine, phenylalanine, and alpha-aminobutyric acid.

Spearman’s correlation analyses showed multiple strong, significant correlations between the identified significant metabolites and patients’ clinical parameters (Figure S2A). To identify key metabolites and their possible relationships with NAFLD that were independent of body weight, we repeated the correlation analyses, adjusting for clinical parameters related to obesity, including BMI, waist circumference, VFA, SFA, and body FAT%. This revealed multiple metabolites that significantly positively (alanine, valine, leucine, and tyrosine) or negatively (aminobutyric acid) correlated with levels of human IHTC (false discovery rate [FDR]-corrected q < 0.1) (Figure 2C). BCAAs and some BAs including serum taurocholic acid (TCA) and serum glycocholic acid (GCA) were significantly correlated with three NAFLD-relevant liver enzymes ALT, AST, and GGT (FDR-corrected q < 0.1). The serum levels of alanine, α-aminobutyric acid, and valine (p = 0.062) also correlated with serum TGs (Figure 2C). Furthermore, the correlations between BCAAs and IHTC, the primary outcome in our trial, remained significant after controlling for obesity-related measures and IR (HOMA-IR).

In summary, the RS intervention may exert its beneficial effects on patients with NAFLD by altering the levels of microbial metabolic products, specifically the AA pool and BCAA levels available for the human host.

The changes of gut microbiota upon RS intervention are associated with NAFLD alleviation

To investigate changes in the gut microbiota, we performed shotgun metagenomic sequencing on fecal samples before and after the 4-month intervention for 50 participants randomly selected from each group (matched with the full analysis set), generating 6.1 Gbp of sequencing data on average (SD 1.3 Gbp per sample). Although similar at baseline in alpha (richness, Simpson index, and Faith’s phylogenetic diversity) and beta diversity (weighted or generalized UniFrac) based on MetaPhlAn2 taxonomic profiling, significant differences between the RS and CS groups were observed after the 4-month intervention (p < 0.05, Wilcoxon rank-sum test for alpha and PERMANOVA for beta diversity) (Figures 3A and 3B). This result suggested different effects of RS on the overall gut microbiota community compared with CS. Specifically, the RS group had lower alpha diversity than the CS group after the intervention. This is consistent with many human and animal studies into the effects of RS2 consumption, as reviewed before.²⁶ Bendiks et al. also suggested the enrichment of particular taxa, which can efficiently metabolize RS and its degradation products, as the possible reason of decreased alpha diversity. In addition to the MetaPhlAn2 profiling, we used an approach relying on co-abundance gene groups (CAGs) to quantify the gut microbiota composition. This led to the same findings for comparisons of microbiota alpha and beta diversity (Figures S2B and S2C).

To uncover the bacterial species that were potentially associated with the beneficial effects of the RS intervention, we adopted two approaches: a non-parametric Wilcoxon test and generalized linear models. We focused on species that either significantly changed their abundance after the RS treatment (but did not change after CS intervention) or became significantly different in abundance between the two groups after the intervention (with no differences at baseline). The non-parametric test revealed that the relative abundances of 31 species significantly changed compared with the baseline or control group (p < 0.05, Wilcoxon signed-rank test or Wilcoxon rank-sum test). The generalized linear model found microbiota species that were significantly associated with the intervention while controlling for the effect of obesity-related measures. This analysis led to the identification of species including Bacteroides stercoris, whose abundance was significantly lower after the RS compared with the CS intervention (FDR-corrected q < 0.2) (Figure 3C; Table S3). We correlated the abundances of all significant bacterial species with a panel of clinical parameters, adjusting for obesity-related measurements, to pinpoint the key species that were relevant to NAFLD (Figure 3C). We focused on the bacteria that significantly correlated with important clinical features in NAFLD (IHTC, ALT, AST, GGT, and FGF21) and found B. stercoris correlated positively with IHTC, ALT, and AST. Significance remained (except p = 0.054 for AST) after further adjusting for IR (HOMA-IR).

We next sought a deeper understanding of the bacterial-phenotype associations by integrating them with the metabolomic profiles. We observed significant correlations between the gut microbial community and the overall fecal BA and fecal SCFA profiles, as well as the serum AA profile (p < 0.05, Mantel test). Moreover, significant associations were found between microbiota composition and serum levels of valine, isoleucine, and leucine (p < 0.01, PERMANOVA). To further disentangle the interplay between gut microbiota taxonomy and serum or fecal metabolite pools, we used Spearman’s correlations to link microbial species and metabolites with significantly differential abundances. Parabacteroides merdae, whose abundance was significantly lower in RS than CS group after intervention, had the highest number of significant correlations (mostly positive) with multiple metabolites (Figure 3D). The RS-depleted intestinal microbe B. stercoris correlated positively with serum valine level (p < 0.05, Spearman’s correlation), which also showed significant positive correlations with IHTC, ALT, AST, GGT, and TG (Figure 2C).

Transplantation of RS-altered gut microbiota alleviates NAFLD in mice

To investigate the potential causality between RS-induced broad gut microbiota alteration and reduction of hepatic steatosis, we performed fecal microbiota transplantation (FMT) (Figure 4A) into conventional antibiotics-treated mice fed with high-fat, high-cholesterol (HFHC) diet, using samples from human donors after RS or CS intervention (whose changes in IHTC after the intervention were close to the corresponding group average; n = 2 per group). Compared with CS donors, FMT from RS donors led to significant reduction of the body weight and liver weight (Figures S3A–S3C). Serum level of FGF21 was significantly lower in the RS group, which was accompanied by the increased expression of FGF21 receptor, co-receptor, and adiponectin in the adipose tissue (Figures S3D–S3F). Improvement of glucose metabolism, especially a significantly increased insulin sensitivity, was also observed in mice receiving fecal microbiota from RS donors (Figures S3G and S3H). Histological assessments demonstrated significant decrease in hepatic steatosis, ballooning, inflammation, and NAFLD activity score after FMT from RS donors (Figures 4B and 4C). Moreover, the RS group had lower levels of liver enzymes ALT and AST, hepatic TG, and TC in the liver (Figures 4D–4G). At the molecular level, FMT from RS donors reduced the expression of marker genes in the liver related to inflammation, macrophage, and neutrophil recruitment (Figure 4H). It also reduced gene expression in the liver for lipogenesis and promoted the expression of genes related to lipolysis (Figures 4I and 4J). Moreover, we also observed the improvement of gut barrier integrity as reflected by the increased expression of genes encoding tight junction proteins (Figure 4K), together with a significant reduction of serum LPS suggesting a possible alleviation of systemic inflammation (Figure 4L). The levels of BCAAs in the colon content were also significantly reduced in mice receiving FMT from RS donors compared to CS donors (Figure 4M).

In addition to the wild-type mice, we also performed the experiment using a genetic model of NAFLD, where ApoE^−/− mice were fed the HFHC diet followed by the same FMT procedure. The causal effect of RS-mediated microbiome changes was successfully replicated in the ApoE^−/− mice, including changes in body weight, liver weight, histological scores, liver enzymes, and serum FGF21 (Figures S4A–S4H). Consistent with the wild-type mice, the ApoE^−/− mice receiving FMT from RS donors had decreased expression of lipogenesis-related genes and increased expression of lipolysis-related genes in the liver, as well as lower levels of colonic BCAAs (Figures S4I–S4K). Moreover, serum LPS was also significantly reduced in the RS compared with the CS group, coupled by increased expression of genes related to gut barrier integrity in the ileum (Figures S4L and S4M). In line with the histological changes in inflammation, the expression of inflammation-related genes in the liver were effectively reduced (Figure S4N).

Multi-omics integration analysis identifies key species associated with NAFLD alleviation

We profiled the functional potential of the gut microbiota and examined functional differences in the RS and CS intervention groups. We found using the MetaCyc database the relative abundances of 8 pathways to be significantly altered after the RS intervention (p < 0.05, Wilcoxon signed-rank test) (Figure S5A). The microbiota functional potential for starch degradation (MetaCyc PWY-6731) significantly increased after the RS intervention (p = 0.038, Wilcoxon signed-rank test), but not in the CS group (Figure S5B). In a particular category of gene families responsible for carbohydrate metabolism, we found that 14 carbohydrate-active enzymes (CAZy) families were significantly altered after both the RS (Figure 5A) and CS (Figure S5C) interventions (p < 0.05, Wilcoxon signed-rank test), with no common families between RS and CS. Interestingly, a significant decrease was observed only after the RS intervention in the abundances of the Kyoto Encyclopedia of Genes and Genomes (KEGG) functional modules M00060 (LPS biosynthesis, KDO2-lipid A, p = 0.024) and M00320 (LPS export system, p = 0.012, Wilcoxon signed-rank tests) (Figure 5B). Another two LPS-biosynthesis-related KEGG modules (M00063 and M00064) had significantly increased abundances only after the CS intervention (p = 0.017 and p = 0.023, respectively, Wilcoxon signed-rank test). As a pro-inflammatory bacterial compound, LPS can reduce intestinal barrier function and increase translocation and is demonstrated to accelerate hepatic steatosis in NAFLD development.²⁷ Moreover, B. stercoris-specific LPS biosynthesis potential (M00060) was also significantly lower in RS than CS group after intervention (p = 0.012, Wilcoxon rank-sum test).

The analyses above revealed several potential intestinal species/function markers and signature metabolites related to NAFLD improvement after the RS intervention. To uncover potential mechanistic links between changes in gut microbiota and NAFLD alleviation, we applied a recently developed computational framework to integrate various data types.²⁸ Initially, we used a three-tiered analysis to screen out microbiota functions (KEGG modules) that were significantly correlated with metabolites and important phenotypes (IHTC, ALT, AST, GGT, and FGF21), while adjusting for obesity-related parameters. These functions may serve as a bridge between the gut microbiota and host metabolism and thus could be potentially related to NAFLD progression, such as the biosynthesis and transport of various AAs (tryptophan, histidine, lysine), cobalamin (vitamin B12), and LPS (Figure S6). In the functional modules significantly correlated with IHTC (p < 0.05, Spearman’s correlation coefficient ≥ 0.2) in the RS group, we identified four KEGG modules related to BCAA biosynthesis (Figures 5C and 5D), emphasizing the strong relevance of BCAAs in NAFLD pathogenesis. Subsequently, gut microbiota driver species analysis was performed to determine which species were the main contributors to the function-phenotype associations. Overall, many more potential KEGG modules (correlated to FGF21 or IHTC) and driver species were observed in the RS than in the CS group, adding evidence that RS shaped the gut microbiome composition and activity in a directed way (Table S4). In particular, we found three highly contributing driver species involved in the correlations between the four BCAAs modules and IHTC (Figure 5D). B. stercoris, the abundance of which was correlated with IHTC, ALT, AST, and serum valine, had the strongest driving effect on average of the four modules, including M00019 for valine/isoleucine biosynthesis.

To validate the positive association between B. stercoris and NAFLD, we re-analyzed the metagenomic data from two published cohorts involving NAFLD (STAR Methods). In a Chinese cohort,²⁹ the abundance of B. stercoris was significantly higher in patients with NAFLD than in NAFLD-free participants (Figure 5E). In a European cohort, patients diagnosed by liver biopsy³⁰ with moderate or severe steatosis also had higher levels of B. stercoris compared with mild steatosis or control (Figure 5F).

B. stercoris promotes NAFLD progression partially through LPS and BCAA production

To examine a potential causal effect of B. stercoris on NAFLD progression, mice were given an HFHC diet for 8 weeks to induce NAFLD, together with daily oral gavage of live or heat-killed B. stercoris at 5 × 10⁹ CFU/day along with the HFHC feeding (Figure 6A). Real-time PCR showed a significantly increased amount of B. stercoris in the feces of the live bacteria group compared with mice on the HFHC diet only (Figure 6B). B. stercoris treatment showed no obvious effects on body weight or FM percentage (Figures S7A and S7B) but significantly increased the liver weight percentage compared with control (Figure S7C). Despite no obvious effects on either glucose or insulin levels in both fasting and fed status, B. stercoris intervention for 8 weeks led to an impaired ability of the mice to dispose glucose and decreased insulin sensitivity (Figures S7D–S7G). The serum level of ALT increased 1.8-fold in mice gavaged with live B. stercoris while AST did not change significantly (Figures 6C and 6D). By histological assessment, hepatic lipid accumulation, inflammatory cell infiltration, and fibrogenesis were all markedly enhanced in mice gavaged with live B. stercoris (Figures 6E–6J) compared with mice only fed with HFHC. Consistent with the histological observations, the level of hepatic TG was more than 2-fold higher in mice gavaged with live B. stercoris (Figure 6K). Serum level of LPS was also significantly increased after 8-week oral gavage (Figure 6L). To determine the specific impact of B. stercoris at the molecular level, we explored the transcription of inflammatory and fibrogenesis markers in liver tissues. In line with the higher histological scores, genes involved in pro-inflammatory response, inflammatory cell infiltration, and collagen formation were significantly higher in mice gavaged with live B. stercoris (Figures 6M and 6N). Collectively, these findings suggest that increased abundance of B. stercoris contributed to NAFLD progression.

Besides live B. stercoris, heat-killed B. stercoris also showed an ability to aggravate NAFLD and elicited similar effects on inflammation, as reflected in the histological score, ALT, serum LPS, and expression of genes involved in inflammation activation (Figures 6C, 6H, 6L, and 6M). Yet the content of hepatic TG in mice with heat-killed B. stercoris showed the trend to be lower than that in the live group (p = 0.054) (Figure 6K).

Given the strong driving effect of B. stercoris in the correlations between gut microbial BCAA biosynthesis and IHTC (Figures 5C and 5D), we then measured the fecal levels of BCAAs in mice. We found the 8-week oral administration of live B. stercoris significantly increased levels of fecal valine and isoleucine (Figure 6O). Unlike live B. stercoris, mice daily gavaged with heat-killed B. stercoris showed only a minimal effect on accumulation of fecal BCAAs, with similar pattern as steatosis score and hepatic TG (Figures 6F and 6K). To further demonstrate a direct metabolic production of BCAA by B. stercoris, we performed various cell cultures, with and without B. stercoris for different time periods, followed by targeted metabolomics analysis of the cultured supernatant. Compared with other groups, the cultured supernatant of live B. stercoris showed a remarkable accumulation of BCAAs in a time-dependent manner, especially for valine (Figures 6P and S7H).

The significant correlation identified in our clinical study between valine and IHTC after controlling for obesity-related measures and HOMA-IR suggested a possible direct influence of valine on liver fat accumulation and thus NAFLD pathogenesis. We therefore investigated the direct in vitro effect of valine, which can be derived from NAFLD-promoting B. stercoris, on lipid metabolism in HepG2 cells. Compared with incubation with only fatty acid (FA), we observed a significant increase in intracellular TG content (Figure S7I) and a dose-dependent increase in the expression of the transcription factor SREBP1 and lipogenic genes following incubation with valine (Figure S7J). Expression of FA transporters and their corresponding transcription factors demonstrated similar dose-dependent increases (Figure S7K). CPT1A, a gene involved in beta-oxidation and lipid catabolism, had lower expression following incubation with valine (Figure S7K).

Discussion

The vital role of gut microbiota in liver diseases has been demonstrated by studies involving FMT³¹ or single species such as Roseburia intestinalis³² and Klebsiella pneumoniae.³³ Such bidirectional relationship between the gut (and its resident microbiota) and the liver, i.e., the gut-liver axis, has gained attention in the last several years with the hope of developing microbiome-based strategies for diagnosis, prognosis, and therapeutics of liver diseases.⁹^,³⁴^,³⁵ However, the efficacy of most of the potential therapeutics for NAFLD needs confirmation in well-designed human studies.¹⁰ Previous clinical trials have demonstrated the ability of MDFs to modulate human immune status³⁶ and to contribute to healthier metabolic and growth profiles of undernourished children.³⁷ In our randomized clinical trial, we evaluated the effects of RS as an MDF for NAFLD treatment. To quantify changes in liver fat content, we used MRI, a highly reproducible and the most accurate non-invasive approach to detect hepatic steatosis.²⁴^,³⁸ Several studies have also confirmed the superiority of MRI over liver histology in assessing liver fat.³⁹^,⁴⁰ It is more sensitive than histological grading in detecting changes in liver fat over time.⁴¹ The 4-month intervention with this MDF was effective in reducing IHTC in patients with NAFLD by an absolute reduction of −5.89% and a relative reduction of −24.30% after adjusting for weight loss. Such effect was partly mediated by altered composition and metabolic profile of gut microbiota. Indeed, transfer of fecal microbiota from human donors receiving 4-month RS into mice fed with HFHC diet reduced hepatic steatosis, lobular inflammation, and expression of lipogenesis- and inflammation-related genes, suggesting a causal role of gut microbiota in alleviating NAFLD. Moreover, the expression of genes related to gut barrier integrity was enhanced while the level of serum LPS was reduced in mice receiving FMT from RS donors. This is consistent with decreased circulating level of LPS and the lower microbiota functional potential for LPS biosynthesis in human participants after RS intake.

AAs were also identified as possible molecular mediators of the RS beneficial effects. Perturbation in AA metabolism, especially aromatic AAs (AAAs), GSG index, and BCAAs, has been shown to be involved in NAFLD and NASH pathogenesis.²⁵^,³⁰ Serum levels of two AAAs, phenylalanine and tyrosine, were significantly lower after RS than CS intervention, and serum glutamic acid for GSG index calculation was significantly reduced after RS intake (Figure 2B). Serum BCAAs have been associated with gut microbiome alteration and IR,⁴² which represents an NAFLD pathophysiology. Here, we observed consistent correlations between BCAAs and IR, and the 4-month RS intervention in humans could significantly reduce the serum levels of BCAAs. Furthermore, serum BCAAs were positively correlated with IHTC, ALT, AST, and GGT. Importantly, the correlations between BCAAs and the primary outcome IHTC remain significant after adjusting for obesity-related parameters and IR, suggesting a direct influence of BCAAs on hepatic steatosis and thus NAFLD pathogenesis. In the FMT experiment where transfer of RS-altered microbiota into mice alleviated NAFLD, the colonic levels of BCAAs were also decreased, suggesting that the change of gut microbiota caused by RS led to the change in BCAAs. The role of BCAAs in hepatic steatosis was also supported by in vitro experiments investigating the direct effect of valine on intracellular TG levels, through the modulation of lipogenic transcription factors, increased lipogenesis, and decreased FA oxidation. AAs may modulate lipogenic transcription factors through participating in the processing of enzymes and transcriptional regulators as well as acting as substrates for lipid synthesis.⁴³^,⁴⁴^,⁴⁵ Elevated hepatic lipogenesis is intimately involved in pathological consequences.⁴⁵

Apart from the causality between RS-induced broad gut microbiota alteration and reduction of hepatic steatosis, we also attempted to pinpoint specific microbial species involved in NAFLD development though multi-omics integration analysis. Among them, we found RS reduced the abundance in the gut of B. stercoris, which is one of the species highly correlated with IHTC, ALT, and AST. These positive correlations remained significant after controlling for obesity-related parameters and HOMA-IR, suggesting a body weight- and IR-independent effect of B. stercoris on NAFLD aggravation. The positive association of B. stercoris in the gut with NAFLD was further validated in two independent external case-control cohorts from Asia and Europe (Figures 5E and 5F). In addition, B. stercoris was selected as a feature in a metagenome-based model for predicting advanced fibrosis in US patients with NAFLD.⁴⁶ Furthermore, we conducted a monocolonization study to confirm the NAFLD-promoting effect of B. stercoris and to explore the possible mechanisms involved. Oral gavage of both live and heat-killed B. stercoris into mice could lead to increased lobular inflammation and enhanced expression of genes involved in inflammation activation, which might be explained by the increased serum level of LPS in both groups. On the other hand, considerably higher hepatic lipid accumulation was only observed in the mice gavaged with live B. stercoris, which was accompanied by the significantly higher levels of fecal BCAAs. Notably, the abundance of B. stercoris in human participants was also found to positively correlate with BCAAs (statistically significant for valine), and targeted measures of BCAAs in the monoculture supernatant of live B. stercoris substantiated its BCAA-releasing activity. Altogether, it suggests that B. stercoris can promote NAFLD progression, at least partially through LPS and BCAA production.

The serum level of FGF21 was found to be significantly reduced after the 4-month RS intervention. A number of preclinical and clinical studies demonstrate the robust effects of FGF21 on alleviation of dyslipidemia and NAFLD.⁴⁷^,⁴⁸ Contrary to the multiple metabolic benefits of FGF21, circulating FGF21 is paradoxically elevated in individuals with NAFLD.⁴⁹^,⁵⁰ The concept of “FGF21 resistance” was proposed to explain the paradoxical changes of plasma FGF21 levels, in analogy to obesity-associated insulin and leptin resistance.⁵¹ Based on animal studies, aberrant FGF21 signaling has been suggested as a key pathological step in the development and progression of NAFLD.⁵² Notably, both circulating and hepatic levels of FGF21 in obese mice were markedly reduced by exercise training, where the FGF21 sensitivity in adipose tissue was enhanced.⁵³ Besides engineered human FGF21 analogs, the sensitization of the actions of FGF21 may represent an alternative strategy for the treatment of metabolic disorders.⁴⁸ In line with this, here, we observed decreased serum level of FGF21 in participants after RS intervention and in mice receiving feces from RS-fed donors, as well as increased expression of its receptor complex and downstream effector in adipose tissue. Our findings suggested that RS-induced microbiome changes might also lead to the sensitization of FGF21 actions.

Altogether, our study provides evidence that RS could be a novel, relatively simple, and inexpensive microbiota-targeted therapeutic option for NAFLD, which can reduce IHTC by 5.89% in a weight-loss-independent manner and decrease the liver enzymes indicative of liver injury and markers for systemic inflammation. The change of gut microbiota composition and functionality is an important mediator of the beneficial effect of RS on NAFLD amelioration, including one gut microbe B. stercoris that aggravates NAFLD at least partially through LPS and BCAA production. Our findings might contribute to further understanding of NAFLD pathogenesis and the development of innovative microbiome-based therapeutics or MDFs.

Limitations of the study

First, due to the lack of liver biopsy, we could not evaluate whether there were beneficial histological changes in the liver, such as biopsy-proven steatosis, NASH, or fibrosis. However, our primary outcome was the change of liver fat content (hepatic steatosis), and IHTC is considered to be more sensitive than the histological steatosis grades in quantifying such changes, which has been recommended for clinical trial usage⁵⁴ and adopted by other NAFLD intervention studies.⁵⁵^,⁵⁶ Notably, the limitations of liver biopsy, including invasiveness, sampling error, poor acceptability, and only moderate reproducibility, also constrain its use as a repeat measurement to investigate histological changes in intervention studies.⁵⁷ Therefore, liver biopsy is not suitable for widespread use to assess disease stage or determine progression or response to therapy.⁵⁸ Second, in our randomized clinical trial, dietary guidelines were offered to the enrolled patients, and information on their dietary intake was collected through questionnaires and was further compared between the two intervention groups. Similar studies in the future may use a standard identical diet to directly control for the effect of diet as a potential confounding factor. Further research may reveal other possible molecular mechanisms by which the RS-altered metabolites or gut microbes lead to the accumulation or reduction of liver fat, the change of inflammation, and fibrosis in the liver.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to the Lead Contact, Weiping Jia (wpjia@sjtu.edu.cn).

Materials availability

This study did not generate new unique reagents.

Experimental model and study participant details

Study participants

A total of 200 individuals who met all the following eligibility criteria were recruited in the Shanghai Jiao Tong University Affiliated Sixth People’s Hospital. Inclusion criteria were: (i) ethnic Chinese, (ii) liver steatosis diagnosed by ultrasonography, (iii) aged 18-70 years old, and (iv) written informed consent obtained.

Exclusion criteria were: (i) participants with diabetes mellitus; (ii) alcohol consumption history of more than 20 g per day for men and more than 10 g per day for women; (iii) acute or chronic gastrointestinal diseases (including diarrhea, gastrointestinal infection, inflammatory bowel disease), malignant tumor or severe renal dysfunction; (iv) pregnancy, breastfeeding or planning to get pregnant; (v) consuming antibiotics within the last 3 weeks or during the study; (vi) viral hepatitis, drug-induced liver disease, total parenteral nutrition, Wilson’s disease, autoimmune liver disease or other specific diseases that can lead to fatty liver; (vii) routine use of prescription medicines or adjuvant Chinese and Western medicines (except regular contraceptives); (viii) expected poor compliance; (ix) use of weight loss medication or participation in weight-loss program in the past 3 months; (x) mental disorder preventing cooperation; or (xi) wearing pacemaker or metallic implants, claustrophobia or other conditions that would be unable to undergo magnetic resonance examinations.

The study was approved by the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, following the principles of the declaration of Helsinki. All relevant ethical regulations were followed during the study. Written informed consent was obtained from all participants. Complete clinical trial registration is deposited in the WHO International Clinical Trials Registry Platform and Chinese Clinical Trial Registry (http://www.chictr.org.cn/showproj.aspx?proj=12353; ChiCTR-IOR-15007519). The primary indication was change in intrahepatic triglyceride content (IHTC), with changes in anthropometric indicators, body composition, glycemic control and insulin sensitivity, liver and renal function, lipid profiles, cytokines, multi-omic parameters, Single nucleotide polymorphisms (SNPs), NAFLD remission rate, and percent change in intrahepatic triglyceride content as secondary outcomes.

Animal model

All mice were housed in a specific-pathogen-free facility with a 12-h/12-h light/dark cycle and given free access to food and water. All protocols for mouse experiments were approved by the Committee on the Use of Live Animals for Teaching and Research of the University of Hong Kong (CULATR No. 4361-17). All relevant ethical regulations were complied. Mice involved in all experiments were given a HFHC diet (D12079B, Research Diets, New Brunswick, NJ, USA) to induce NAFLD. For FMT experiments, 5-week-old male C57BL/6J wild-type mice and 6-week-old male C57BL/6J ApoE^-/- mice were purchased from GemPharmatech (Nanjing, China). Mice were randomly divided into two groups: RS group and CS group (n =8 per group in the dietary model; n=6 per group in the genetic model). Both were fed the HFHC diet for 11 weeks before fecal microbiota transplantation (diet, water, and bedding were all sterilized). For B. stercoris gavage, eight-week-old male C57BL/6J mice were randomly divided into three groups: HFHC with PBS group, HFHC with live B. stercoris group, and HFHC with heat-killed B. stercoris group, and were treated for 8 weeks. Body composition was assessed with a Minispec LF90 body composition analyzer (Bruker, Billerica, MA, USA) after 8 weeks of feeding. Glucose and insulin levels in both fasting and fed status, intraperitoneal glucose tolerance test, and insulin tolerance tests were performed after 8 weeks of daily gavage.⁶²

Culture and administration of B. stercoris

B. stercoris (catalog No. 19555, DSMZ-German Collection of Microorganisms and Cell Cultures GmbH, Germany) was cultured in chopped meat medium (Hardy Diagnostics, USA) at 37 °C in an anaerobic workstation (Gene Science AG300, China) with a gas mix containing 10% hydrogen, 10% carbon dioxide and 80% nitrogen. The concentration of bacteria was calculated by measuring the absorbance at the wavelength of 600nm. A fresh culture containing 5×10⁹ cfu of B. stercoris in 200μL PBS was orally gavaged daily to C57BL/6J mice, with sterile PBS as control. For one experimental group, B. stercoris was heat killed at 121°C under 225-kPa pressure for 15 min.

Cell culture and treatment with valine

Human hepatocellular carcinoma cells from the HepG2 cell line (ATCC, Manassas, VA, USA) were cultured in Dulbecco’s Modified Eagle Medium (Gibco, NY, USA) with 10% fetal bovine serum (Gibco, NY, USA). This cell model demonstrated comparable results to primary human hepatocytes in terms of lipid accumulation,⁶³ which was the scope of our in vitro study. Cells were kept in a 37°C incubator with 5% CO₂. Culture medium was replaced every 2–3 days, and cells were sub-cultured upon reaching 80% confluence.

L-valine (TCI Chemicals, Portland, OR, USA) was dissolved in Milli-Q water to form a stock solution and diluted with serum-free medium to working concentration (30-750 μM). The concentration range for cell experiments was determined by the serum concentration range from human clinical samples and preliminary cytotoxicity assays. Sodium oleate (Sigma-Aldrich, St. Louis, MO, USA) was dissolved in water and sodium palmitate (Sigma-Aldrich, St. Louis, MO, USA) in methanol for stock solutions. The stock solution was further diluted with serum-free medium supplemented with 1% FA-free bovine serum albumin (Gibco, NY, USA) to a working concentration of 1000 μM. FA-BSA complex was prepared fresh before treatment.

For experiments, cells were plated on 6-well plates at 1 x 10⁶ cells/well and allowed to adhere overnight. Cells were incubated with valine for 24 hours followed by fatty acid incubation for another 24 hours. Cells were then collected for further assays.

Method details

Study design

The study procedure has been detailed in the study protocol (supplemental information). Briefly, the study was a randomized, double-blinded, placebo-controlled trial conducted at Shanghai Jiao Tong University Affiliated Sixth People’s Hospital from 2016-March to 2017-October. Participants were randomized into two groups with an allocation ratio of 1:1 and consumed either HAM-RS2 (Ingredion, Bridgewater, NJ, USA) at 255.4 kcal/day (2.8 kcal/g, 91.2 g, containing 40 g RS) or matched CS (Ingredion, USA) at 255.6 kcal/day (3.55 kcal/g, 72 g, containing 0 g RS) for 4 months (120 days). RS and CS were packaged in sealed bags that were identical in appearance. During the entire trial, participants received dietary and lifestyle counselling. All were engaged in light physical labor or had a sedentary lifestyle, and were advised to keep their usual physical activity habits. Dietary counseling was conducted by a trained dietitian. Standard menus with targeted dietary caloric restrictions and macronutrient intake designed by the dietitian from the Department of Clinical Nutrition, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital were provided to participants, as well as the oilcan and scale. Participants were asked to fill in three consecutive 24-hour dietary records (2 weekdays and 1 weekend day) at each visit period and were encouraged to weigh foods to ensure they accurately reported their caloric intake. In each visit, participants were met with a nutritionist individually for assessment of their adherence to both the diet and the starches (adherence to diet was evaluated as whether the total energy intake according to the 24-h dietary recalls met the requirement of diet management; adherence to starch was evaluated by counting the empty packaging bags of starch participants returned at each visit).

At each visit, participants came to the Department of Endocrinology and Metabolism in the morning for collection of blood, urine, and stool samples, and for the measurement of anthropometric and biochemical indexes. Abdominal magnetic resonance imaging (MRI) scan and MRS were conducted at V1 and V5, whereas meal tolerance tests were conducted at V1, V3, and V5. The primary outcome was the change in IHTC evaluated by MRS. Secondary outcomes were changes in anthropometric indexes, body composition, body fat analysis by MRI, glycemic control, insulin sensitivity, liver and renal function, lipid profiles, measurement of serum biomarkers, and other tests.

We recorded the combined medication during the follow-up visits and no gastrointestinal drugs such as antacids were used. No serious adverse events were reported throughout the study. Other potential intervention-related adverse events, including constipation (8 participants in RS and 15 participants in CS, P = 0.108) and flatulence (20 participants in RS and 19 participants in CS, P = 0.914), were equally distributed between the two groups, except the intestinal exhaust (35 participants in RS compared with 8 in CS, P < 0.001).

Anthropometric and biochemical measurements

Blood pressure, body weight, height, waist circumference, and biomedical indices were measured according to the study protocol (supplemental information). BMI (weight [kg]/ height² [m²]) was also calculated. Blood samples were collected from participants after an overnight fast of at least 10 hours and were used to measure serum ALT, AST, GGT, TG, TC, HDL-C, LDL-C, and non-esterified FA (NEFA). To assess the glucose metabolism, serial blood samples were taken in a fasting state and at postprandial time points for laboratory tests of plasma glucose, insulin, and c-peptide after a standardized meal tolerance test (85 g of non-fried instant noodles without soup: 376.98 kcal including 68.6 g carbohydrate, 9.4 g protein, and 6.8 g fat) (China Oil & Foodstuffs Corporation, China). Insulin resistance indexes were calculated as follows: HOMA-IR = FPG (mmol/L) × FINS (mU/L)/22.5; Adipo-IR = fasted insulin (mmol/L) × fasted NEFA (pmol/L).

MRS examination

Participants underwent liver MRS using the 3.0-T Philips Ingenia medical system (Philips Healthcare, The Netherlands). Sagittal, coronal, and axial slices through the right lobe of the liver were acquired, and regions of interest were selected by an experienced radiologist, who avoided visible blood vessels and bile ducts. IHTC was measured in a single voxel (2 × 2 × 2 cm³) and calculated by dividing the integral of the methylene groups in fatty acid chains of the hepatic triglyceride by the sum of methylene groups and water. The experienced radiologists who performed the test were blinded to the clinical data.

MRI examination

Levels of SFA and VFA were determined by MRI using a 3.0-T Philips Ingenia medical system (Philips Healthcare, The Netherlands) with spin echo sequences: 500/20 (TR/TE) and matrix size = 256 × 25,659. Scan time was approximately 180 seconds. MRI scans were obtained at the abdominal level between L4 and L5 vertebrae in the prone position. Analysis of images was performed on a workstation provided by the manufacturer. MRI was performed by experienced radiologists who were blinded to clinical presentation and laboratory findings. Acquired images underwent measurement of SFA and VFA using a semiautomated segmentation method. According to the signal intensity of adipose tissue, SFA and VFA outlines were manually traced with a graphic user interface. The area inside the outline was automatically labelled and calculated by the software SliceOmatic (Version 5.0, TomoVision, Canada).

Diagnostic criteria for NAFLD

We followed guidelines for the assessment and management of NAFLD in the Asia-Pacific region. For all participants, NAFLD was diagnosed by B ultrasonography (detailed in study protocol), ruling out secondary causes of hepatic fat accumulation including acute infectious disease, biliary obstructive diseases, alcohol abuse (more than 20 g per day for men and more than 10 g per day for women), acute or chronic cholecystitis, acute or chronic viral hepatitis.

Measurement of FGF21 and cytokeratin 18 M65ED

Concentration of FGF21 in human serum was quantified using an enzyme-linked immunosorbent assay (ELISA) kit from Antibody and Immunoassay Services, the University of Hong Kong (AIS, HKU, China). Human serum cytokeratin 18 (CK18) M65ED concentration was quantified with the M65 EpiDeath ELISA kit (Peviva AB, Bromma, Sweden). Intra-assay variations for the measurement of FGF21 and CK18 M65ED were 1.89% and 0.77%, respectively, and for inter-assay variations, these values were 4.08% and 8.23%.

Measurement of LPS and pro-inflammatory factors

Human serum LPS was measured by the Limulus Amebocyte Lysate assay (Hycult Biotech, The Netherlands). Concentrations of pro-inflammatory factors including IL6, IL1β, TNFα, and MCP1 were quantified with ELISA kit (Invitrogen, USA). Intra-assay variations and inter-assay variations for the measurements were all below 10%.

Targeted metabolomics analysis of human fecal bile acids, serum bile acids, and amino acids

Sample pre-treatment

For fecal samples, about 20-30 mg freeze-dried sample was added to 2 mL Eppendorf tubes. One mL ethanol solution containing internal standards (CA-d5 0.3 μg/mL, CDCA-d4 0.9 μg/mL, GCA-d5 0.6 μg/mL, GCDCA-d4 0.6 μg/mL, TCA-d5 0.3 μg/mL, TDCA-d5 0.3 μg/mL) was added and vortexed. Subsequently, samples were ground with zirconia beads (30 Hz, 1 min). After centrifugation (14,000 x g, 4°C for 10 min), 800 μL supernatant was transferred for freeze-drying. Samples were then dissolved in 800 μL aqueous solution containing 25% acetonitrile and filtered through a 0.22 μm filter membrane. For serum samples, 50 μL sample was fully mixed with 200 μL acetonitrile solution containing internal standards (CA-d5 0.1 μg/mL, CDCA-d4 0.3 μg/mL, GCA-d5 0.2 μg/mL, GCDCA-d4 0.2 μg/mL, GCDCS-d5 0.2 μg/mL, TCA-d5 0.1 μg/mL, TCDCA-d5 0.1 μg/mL, TDCA-d5 0.1 μg/mL, alanine-d3 3 μg/mL, phe-d5 3 μg/mL, histine-¹³C₆ 1 μg/mL) for protein precipitation and metabolite extraction. Supernatants were pipetted for freeze-drying. Finally, the powder was dissolved in 70 μL aqueous solution containing 25% acetonitrile at 70 μL and 50 μL redissolved solution was transferred into sample bottles. Another 10 μL was used for freeze-drying for AA analysis.

Using the above 10 μL freeze-dried sample, derivative reactions were performed with AccQTag derivatization kits (Waters, USA) before AA liquid chromatography (LC)-MS analysis. Derivative reactions were performed according to the protocol and briefly described as follows: 70 μL AccQ·Tag ultra-borate buffer (pH 8.8) was added to freeze-dried samples and mixed for 30 seconds and 20 μL AccQ·Tag derivative reagent was added after 10 seconds of vortex. The mixture was kept at room temperature for 1 min and heated for 10 min at 55°C for derivatization reaction.

LC-MS analysis

For both BA and AA profiling, a high-performance liquid chromatograph Nexera X2 (Shimadzu, Japan) and triple quadrupole mass spectrometer (MS) 8050 (Shimadzu, Japan) system equipped with electron spray ionization (ESI) ion source was employed. The main MS parameters were: nebulizing gas flow at 3 L/min, heating gas flow at 10 L/min, interface temperature at 300°C, DL temperature at 250°C, heat block temperature at 400°C, drying gas flow at 10 L/min. Multiple reaction monitoring (MRM) was used to detect BAs and AAs. ACQUITY UPLC C18 columns (100 mm × 2.1 mm, 1.7 μm) were used for chromatograph separation.

Elution conditions for BA analysis were: Mobile phase A was 10 mM ammonium bicarbonate aqueous solution and mobile phase B was pure acetonitrile. The gradient started from 25% B and was maintained for 0.5 minutes, then linearly increased to 40% B in 12.5 minutes and 90% B in another 1 minute. The gradient was maintained at 90% B for 3 minutes, returning to 25% B in 0.5 minutes. The initial pre-equilibrium time was 2.5 minutes. Column temperature was 35°C. Flow rate was 0.35 mL/min. Injection volume was 5 μL.

Elution conditions for AA analysis were: The gradient started from 1% B, was maintained for 1.08 min, and increased to 9.1% B in 10.4 min. At 16.3 min, the gradient was linearly increased to 21.2% B, then quickly to 59.6% B in 0.6 min, and maintained for 1.2 min. The gradient was returned to 1% B in 0.18 min and maintained for 3.72 min for initial pre-equilibrium. Column temperature was 55°C. Flow rate was 0.35 mL/min. Injection volume was 0.1 μL.

Targeted metabolomics analysis of fecal SCFAs

Sample processing

About 20 mg feces sample and 200 μL 50% acetonitrile/Milli Q water were mixed in an Eppendorf tube. Samples were ground twice with zirconia bead (30 Hz for 1 min). After centrifugation (14,000 x g, 4°C for 10 min), supernatants were collected and filtered, and an aliquot of 40 μL was transferred into 1.5 mL Eppendorf tubes following addition of 10 μL hexanoic acid -d11 (50 μg/mL in 50% acetonitrile/MilliQ water). After vortexing, 20 μL 3-nitrophenyl hydrazine (200 mM in 50% acetonitrile/MilliQ water) and 20 μL EDC (120 mM in 50% acetonitrile/MilliQ water containing 6% pyridine) were added. Tubes were incubated in a water bath (40^oC) for 30 min and placed on ice for 1 min to stop derivative reactions. Before LC-MS analysis, 910 μL 10% acetonitrile/MilliQ water was used for dilution.

LC-MS analysis

Quantitative analysis used an AB SCIEX ExionLC AD UPLC coupled with AB SCIEX triplequadrupole 6500 plus MS (AB SCIEX, Framingham, US). An ESI ion source was used. MRM scan was operated in negative ionization mode. Ion source parameters were capillary temperature 325^oC, capillary voltage 49V, and sheath gas 40 arb. An ACQUITY UPLC C18 column (100 mm × 2.1 mm, 1.7 μm) was used for separation. Mobile phase A was 0.1% formic acid in MilliQ water. Mobile phase B was 0.1% formic acid in acetonitrile. The total run time was 11 min per sample. The gradient started from 15% B and was increased to 27% B in 4 min, then to 42% B in 4 min, then 100% B in 0.5 min, maintained for 1 min before returning to 15% B in 0.5 min, and maintained for 1 min. Column temperature was 40^oC. Flow rate was 0.35 mL/min. Injection volume was 5 μL.

Fecal sample collection and DNA extraction

Fecal samples were collected using a commercial tube with DNA stabilizer (STRATEC Molecular, Berlin, Germany) and stored at -80°C. Stool DNA was extracted using PSP Spin Stool DNA Kits (STRATEC Molecular, Berlin, Germany) according to the manufacturer’s instructions. Fecal DNA extracts were used to construct shotgun metagenomic libraries using the KAPA soil kit following the standard protocol. The Novaseq 6000 platform was used for 150 bp paired-end sequencing at Novogene, China.

Quality control and taxonomic profiling

For quality control of raw reads, human DNA contamination was removed using BWA mem version 0.7.4⁵⁹ against human reference genome ucsc.hg19 and adaptors, low-quality reads, bases or PCR duplicates were filtered as previously described.³⁵ High-quality reads were taxonomically profiled at different taxonomic levels using MetaPhlAn2⁶⁰ version 2.7.7 with default settings, generating taxonomic relative abundances (total sum scaling normalization). For the CAGs-based approach, genes obtained from HUMAnN2 (“functional profiling” below) were clustered into CAGs and then metagenomic species (MGS, referring to CAGs with >700 genes) as described before⁶⁴ using default algorithm options. MGS were assigned a species-level annotation if more than 50% of genes were assigned the same species level taxonomy and if the second-most assigned taxonomy was <10% or unclassified.

Microbial community diversity analysis

The alpha diversity was calculated using the R package vegan⁶⁵ and picante. Statistical comparisons of alpha diversity between groups were by Wilcoxon rank-sum test or signed-rank test using R package stats. Beta diversity (Bray-Curtis dissimilarity, weighted UniFrac and generalized UniFrac) was calculated with the R packages phyloseq and GUniFrac. Statistical comparison between groups was by the function adonis to perform a permutational multivariate analysis using R package vegan with 999 permutations. P < 0.05 was considered significant.

Functional profiling

Microbial gene families and pathway abundances were determined using HUMAnN2 software⁶¹ version 0.11.2 and the UniRef90 and MetaCyc databases. Gene families were mapped to Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology database included in HUMAnN2 to obtain KEGG modules and KEGG pathway abundances. Gene families were also mapped to level-4 enzyme commission (EC) categories using the EC database included in HUMAnN2. Carbohydrate-Active enZYmes (CAZy)⁶⁶ were obtained by annotating ECs to the CAZy database. Tables of pathway and gene family abundance obtained using HUMAnN2 were normalized to copies per million, including unmapped and unintegrated reads.

Integrating microbiome, metabolome and phenotypes

The omics computational framework described before²⁸ was used to perform a three-way analysis to screen potential KEGG modules that significantly correlated with metabolites and phenotypes, and leave-one-species-out analysis to determine the driver species of KEGG modules and important phenotypes. In the three-way analysis, correlations between the functional potential and phenotypes were by partial Spearman’s correlation adjusting for obesity-related parameters. In cases of correlations between microbiota functional potential and metabolites, Spearman’s correlation was used. In the leave-one-species-out analysis, we checked for KEGG modules with strong correlations (absolute Spearman’s correlation ≥ 0.2) with IHTC or FGF21. Species with > 10% effect on the correlation after removal were deemed driver species.

Validation of B. stercoris in external cohorts

Two independent external cohorts of different ethnicity were used. The Chinese cohort included 100 patients with NAFLD (diagnosed with ultrasonography) and 90 NAFLD-free control.²⁹ The European cohort had different degrees of steatosis confirmed by liver biopsy controls,³⁰ including 32 participants with no or mild steatosis and 24 participants with moderate or severe steatosis. The metagenomic data were processed with the above pipeline for quality control and taxonomic profiling.

FMT Experiment

Mice were treated with antibiotics cocktail (ampicillin 1g/l, neomycin 1g/l, metronidazole 1g/l, vancomycin 0.5g/l) for 7 days for microbiota depletion, followed by a 4-day wash-out period to eliminate antibiotics before fecal microbial transplantation as described previously.³⁰^,⁶⁷ Two human donors from RS or CS group who were close to the average change of IHTC within RS and CS intervention were selected for fecal microbial transplantations. Approximately 500 mg fresh human stools from each donor were collected in the anaerobic workstation and suspended in 5 ml PBS buffer containing 0.2 g/l Na₂S and 0.5 g/l cysteine. The mixture was homogenized and centrifuged, and the supernatant was collected under a stream of nitrogen. Stool samples from participants were not pooled, and fecal slurry from each donor was transferred into 4 conventional antibiotic-treated mice housed in one cage. The mice were colonized by oral gavage with 200 μl of RS or CS fecal slurry. Mice were treated once daily for three consecutive days by gavage in the first week of colonization, then fecal slurries were introduced every other day to reinforce colonization during the remaining days of the 3 weeks.

Histological examination

Mouse liver samples were resected and fixed with 10% formaldehyde phosphate-buffered saline (pH 7.4), embedded in paraffin, sectioned, stained with hematoxylin/eosin (H&E, Sigma, USA) for morphology and Sirius Red (Abcam, UK) for fibrosis, followed by analysis with a Nikon DS-Ri2 microscope (Nikon Instruments, Melville, USA). For detection of neutral lipids, liver cryosections embedded in OCT were stained with Oil Red O (Sigma, USA). Histology was evaluated by two independent researchers who were blinded to the experimental design and treatment groups according to the NAFLD scoring system as previously reported.⁶⁸ In brief, 4 histological features were evaluated semi-quantitatively including steatosis (0-3), lobular inflammation (0-3), hepatocellular ballooning (0-2), and fibrosis (0-4). The unweighted sum of the first three features was defined as NAFLD activity score (NAS).

Biochemical assays in mice

Serum levels of AST, ALT, TC, TG, and glucose in mice were measured with commercial kits from Stanbio Laboratory (Boerne, TX, USA) or Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Serum lipopolysaccharide (LPS) was measured by the Limulus Amebocyte Lysate assay (Hycult Biotech, The Netherlands). Insulin level was determined by immunoassay from Immunodiagnostics (AIS, HKU, China). Serum FGF21 in mice was measured by ELISA (AIS, HKU, China).

Quantification of mice hepatic lipids

TG content of livers was determined by a modified Folch method.⁶⁹ Briefly, 50 mg of liver tissue was homogenized in chloroform/methanol (2/1; v/v). After extraction at room temperature overnight, the organic phase was used to measure hepatic TG with commercial kits from Stanbio Laboratory (Boerne, TX, USA) or Nanjing Jiancheng Bioengineering Institute (Nanjing, China).

RNA preparation and real-time quantitative polymerase chain reaction

Total RNA from livers or ileum was extracted with TRIzol reagent (Invitrogen, CA, USA), and total RNA from cells was extracted by RNAIso Plus reagent (Takara, Japan) according to the manufacturer’s manual. RNA concentration was determined using a NanoDrop ND-1000 Spectrophotometer (Nano-Drop Technologies, Wilmington, DE, USA) and RNA quality was determined by the A260/A280 ratio of 1.8-2.1. RNA integrity was also checked by 1% agarose gel electrophoresis to ensure 2 intact bands of 28S and 18S RNA. Genomic DNA digestion and reverse transcription yielded cDNA from 1 μg RNA template using the HiScript RT SuperMix for qPCR kit (Vazyme Biotech, Nanjing, China) or PrimeScript RT reagent Kit (Takara, Japan) according to the manufacturer’s manual. Quantitative real-time PCR was performed using SYBR Green master mix on StepOnePlus Real-Time PCR system (Applied Biosystems, Foster City, CA, USA) or Light Cycler 480 system (Roche, USA). The mouse glyceraldehyde-3-phosphate dehydrogenase gene and human beta-actin gene were the reference for tests in mice or cell line, respectively. Relative changes in gene expression were calculated using the 2^-▵▵CT method. Primers used for PCR are listed in Table S5.

In vitro testing of B. stercoris for valine-releasing activity

B. stercoris was grown in the same culture medium as mentioned above to stationary phase and was inoculated to a fresh medium that had been sterilized by an autoclave. Then the mixture was aliquoted and incubated under anaerobic conditions for 3, 5, 8, 12, and 24 hours, respectively. At each time point, an aliquot was removed for centrifuging at 4500rpm (4°C for 15 min) to obtain culture supernatants. Following filtration (pore size 0.22 μm; Millipore, USA), the samples were stored at -80 °C until use. In parallel, fresh medium was aliquoted and incubated under the same conditions for the same period of time as blank control. At each time point, the supernatant was obtained from aliquot centrifuged at 4500rpm (4°C for 15 min) as control.

Experiments seeking to test whether heat-killed B. stercoris could produce valine were carried out by first incubating B. stercoris in the medium at 37 °C under anaerobic conditions to stationary phase. Stationary phase culture was then treated at 121°C under a pressure of 225 kPa for 15 minutes. And the culture was added to fresh medium. The mixture was incubated under anaerobic condition for 8h (logarithmic phase of growth of live B. stercoris) and then centrifuged to obtain culture supernatants.

All cultures in each condition were performed in triplicate. Ultraperformance liquid chromatography coupled to tandem mass-spectrometry (UPLC-MS/MS) system (ACQUITY UPLC-Xevo TQ-S, Waters, USA) was used to quantitate all targeted metabolites by Metabo-Profile Biotechnology.⁷⁰

Intracellular TG assays

Intracellular TG assays were performed using Triglyceride Colorimetric Assay Kits (Cayman Chemicals, Ann Arbor, MI, USA) according to the manufacturer’s manual. Intracellular TG content was normalized to cellular protein quantified using Bradford Protein Assay (Bio-Rad, Hercules, CA, USA).

Data visualization

Circos plots were made using interactive Tree of Life (https://itol.embl.de/). All other figures were generated by R software 3.6.3, using ggplot2 and ComplexHeatmap packages, or by GraphPad Prism 9.0.

Quantification and statistical analysis

Clinical and experimental data

For clinical data, analyses were performed with SPSS 25.0 (Chicago, IL, USA) and SAS version 9.4 (SAS Institute, Cary, NC, USA) as 2-sided with a significance level of α = 0.05. Analyses were performed mainly in the full analysis set, which included all randomized patients who received at least one dose of study medication and had at least one post-intervention assessment of effectiveness. Numerical variables were expressed as mean (95% CIs). Categorical variables were expressed as percentages. Student’s unpaired t-tests and chi-square tests were used for comparison between two groups at baseline. Linear mixed model was used to assess within-group differences. Comparison between the RS and CS groups at each visit was through analysis of covariance with treatment group as a factor and baseline value as a covariate. Differences in outcomes between the RS and CS groups were assessed using a linear mixed model. Fixed effects included baseline values of the assessed variable; treatment group (RS vs. CS as a categorical variable); categorical time points represented by 5 visits at days 0, 30, 60, 90, 120; and the interaction term of visit × treatment group. Repeated measures were added as a random effect. Weight loss was used as an additional fixed effect when the weight loss-independent effect was assessed.

For animal and cell line experiments, all analyses were performed with GraphPad Prism 9.0 (GraphPad Prism, USA). Data were shown as mean ± SEM. Two-tailed Student’s unpaired t test (normally distributed) or non-parametric Wilcoxon rank-sum test (non-normally distributed) was used for comparisons between two groups. Comparisons among more than two groups were performed using one-way ANOVA (normally distributed) followed by Tukey’s post hoc test, or Kruskal-Wallis test (non-normally distributed) followed by Dunn’s test.

Metabolomics and metagenomics data

We performed partial least squares discriminant analysis using metabolite fold-change (log₂-transformed) profiles with the R package mixOmics.⁷¹ Statistical comparison between groups was based on Bray Curtis dissimilarity using the function “adonis” in R package vegan with 999 permutations. Taxonomic and metabolite variations were further calculated as the ratio between microbial relative abundance or metabolite abundance at week 16 against abundance at baseline. Log₂ transformation was applied to fold-changes. For taxonomic variation, before deriving fold-changes, zero values were additively smoothed by minimal nonzero abundance among all observed measurements. Differentially abundant species, functions, and metabolites were identified by two-sided Wilcoxon rank-sum test or Wilcoxon signed-rank test, when appropriate, using R package stats. Z scores for metabolites variation were calculated using R package rcompanion. Generalized linear models were used to obtain differentially abundant species after adjusting for obesity-related parameters (species ∼ group + VFA + SFA + BMI + Waist circumference + FAT%) with glm function from R package stats. To determine if metabolome and microbiome were associated, a mantel test was performed using the mantel function from the R package vegan. Bray-Curtis dissimilarity matrices based on the species and metabolite abundance tables were computed to perform this test.

Spearman’s correlation analysis was performed using R package stats. Partial Spearman correlation adjusting for obesity-related parameters (VFA + SFA + Waist circumference + BMI + FAT%) was performed between metabolites/species and clinical data using the R package ppcor. All statistical analyses were performed with R software 3.6.3 and P < 0.05 was deemed significant unless otherwise stated. P values were adjusted by an FDR method⁷² using R package stats.

Acknowledgments

We would like to thank all the medical staff and study participants who took part in the trial. The authors thank Prof. Jiarui Wu and Prof. Rong Zeng for the fruitful discussions and Dr. Ruben V. Uribe and Prof. Morten Sommer for their help with the design of qPCR primers. This work was supported by the National Natural Science Foundation of China (NSFC) major international (regional) joint research project (81220108006), Shanghai Municipal Key Clinical Specialty, Shanghai Research Center for Endocrine and Metabolic Diseases (2022ZZ01002), and the National Key Research and Development Program of China (2018YFA0800402) to W.J.; the Marie Skłodowska-Curie Actions (MSCA) and Innovative Training Networks (H2020-MSCA-ITN-2018 813781) “BestTreat” (G.P.‬, S.L.S., E.N., and H. Leung) and DFG under Germany’s Excellence Strategy – EXC 2051 – Project ID 390713860 (G.P. and Y.N.); Excellent Young Scholars of NSFC (82022012), National Key Research and Development Program of China (2022YFA1004804), General Program of NSFC (82270907), Two Hundred Program from Shanghai Jiao Tong University School of Medicine (20191830), and Innovative research team of high-level local universities in Shanghai (SHSMU-ZDCX20212700) to H. Li; and the Innovation Program (DICP I202019) of Science and Research from the Dalian Institute of Chemical Physics, Chinese Academy of Sciences, and Key Foundation (21934006) from the National Natural Science Foundation of China to G.X.

Author contributions

W.J., G.P., H. Li, G.X., and Y.N. conceived and designed the study. L.Q., X.L., L. Zhang, Q.G., Q.W., and H. Li. recruited participants and collected and analyzed clinical data. L.Q. collected serum and fecal samples and extracted DNA from feces. L. Zhou, X.W., and Q.L. generated the targeted metabolomics data. Y.N., S.L.S., E.N., and H. Leung performed bioinformatics analyses. L.Q., X.L., Y.L., X.J., S.L., R.Y., and Y.Z. conducted animal experiments. X.L., Q.W., and Y.Z. performed in vitro monoculture experiments. M.J.I. performed HepG2 cell line experiments. Y.N., L.Q., X.L., Y.L., and M.J.I. wrote the manuscript. Y.N., G.P., and H. Li coordinated and supervised the study. W.J., G.P., H. Li, G.X., A.X., and H.E.-N. reviewed and edited the manuscript. All authors made substantial contributions and approved the final version of the manuscript.

Declaration of interests

The authors declare no competing interests.

References

Le, M.H. ∙ Le, D.M. ∙ Baez, T.C. …

Global incidence of non-alcoholic fatty liver disease: a systematic review and meta-analysis of 63 studies and 1,201,807 persons

J. Hepatol. 2023; 79:287-295

Resistant starch decreases intrahepatic triglycerides in patients with NAFLD via gut microbiome alterations

Highlights

Summary

Introduction

Results

RS intervention for 4 months alleviates NAFLD in Chinese adults

RS intervention alters both fecal and serum metabolites in patients with NAFLD

The changes of gut microbiota upon RS intervention are associated with NAFLD alleviation

Transplantation of RS-altered gut microbiota alleviates NAFLD in mice

Multi-omics integration analysis identifies key species associated with NAFLD alleviation

B. stercoris promotes NAFLD progression partially through LPS and BCAA production

Discussion

Limitations of the study

Resource availability

Lead contact

Materials availability

Experimental model and study participant details

Study participants

Animal model

Culture and administration of B. stercoris

Cell culture and treatment with valine

Method details

Study design

Anthropometric and biochemical measurements

MRS examination

MRI examination

Diagnostic criteria for NAFLD

Measurement of FGF21 and cytokeratin 18 M65ED

Measurement of LPS and pro-inflammatory factors

Targeted metabolomics analysis of human fecal bile acids, serum bile acids, and amino acids

Sample pre-treatment

LC-MS analysis

Targeted metabolomics analysis of fecal SCFAs

Sample processing

LC-MS analysis

Fecal sample collection and DNA extraction

Quality control and taxonomic profiling

Microbial community diversity analysis

Functional profiling

Integrating microbiome, metabolome and phenotypes

Validation of B. stercoris in external cohorts

FMT Experiment

Histological examination

Biochemical assays in mice

Quantification of mice hepatic lipids

RNA preparation and real-time quantitative polymerase chain reaction

In vitro testing of B. stercoris for valine-releasing activity

Intracellular TG assays

Data visualization

Quantification and statistical analysis

Clinical and experimental data

Metabolomics and metagenomics data

Acknowledgments

Author contributions

Declaration of interests

References

Share this:

Related

you'll also love

Join the List

Leave a Reply Cancel reply

Follow Along

@thekitchendivision

Explore

Info

stay in the know