Hepatology

ORIGINAL RESEARCH Dietary cholesterol drives fatty liver-­associated liver Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from cancer by modulating gut microbiota and metabolites Xiang Zhang,1 Olabisi Oluwabukola Coker,1 Eagle SH Chu,1 Kaili Fu,1 Harry C H Lau ‍ ‍ ,1 Yi-Xiang­ Wang,2 Anthony W H Chan,3 Hong Wei,4,5 Xiaoyong Yang,6 Joseph J Y Sung,1 Jun Yu ‍ ‍ 1

►► Additional material is ABSTRACT published online only. To view Objective Non-­alcoholic fatty liver disease (NAFLD)-­ Significance of this study please visit the journal online associated hepatocellular carcinoma (HCC) is an (http://dx.​ ​doi.org/​ ​10.1136/​ ​ What is already known on this subject? gutjnl-2019-​ ​319664). increasing healthcare burden worldwide. We examined the role of dietary cholesterol in driving NAFLD–HCC ►► Non-­alcoholic fatty liver disease (NAFLD)-­ For numbered affiliations see associated hepatocellular carcinoma (HCC) is end of article. through modulating gut microbiota and its metabolites. Design High-fat/high-­ ­cholesterol (HFHC), high-fat/­ an increasing healthcare burden worldwide. ►► Cholesterol is a major lipotoxic molecule. Correspondence to low-­cholesterol or normal chow diet was fed to C57BL/6 Jun Yu, Institute of Digestive male littermates for 14 months. Cholesterol-lowering­ What are the new findings? Disease and The Department drug atorvastatin was administered to HFHC-­fed mice. ►► High-­fat high-­cholesterol diet (HFHC) of Medicine and Therapeutics, Germ-free­ mice were transplanted with stools from Chinese University of Hong spontaneously and sequentially induced fatty Kong, New Territories, Hong mice fed different diets to determine the direct role liver, steatohepatitis, fibrosis and NAFLD–HCC Kong; junyu@​ ​cuhk.edu.​ ​hk of cholesterol modulated-microbiota­ in NAFLD–HCC. development, while high-­fat low-­cholesterol Gut microbiota was analysed by 16S rRNA sequencing diet induced only hepatic steatosis in male XZ and OOC are co-first­ authors. and serum metabolites by liquid chromatography– mice. mass spectrometry (LC–MS) metabolomic analysis. Received 19 August 2019 ►► Dietary cholesterol-­induced NAFLD–HCC Revised 4 June 2020 Faecal microbial compositions were examined in 59 formation was associated with gut microbiota Accepted 15 June 2020 hypercholesterolemia patients and 39 healthy controls. dysbiosis. The microbiota composition changed Results High dietary cholesterol led to the sequential along stages of NAFLD–HCC formation: progression of steatosis, steatohepatitis, fibrosis and

Mucispirillum, Desulfovibrio, Anaerotruncus http://gut.bmj.com/ eventually HCC in mice, concomitant with insulin and Desulfovibrionaceae were sequentially resistance. Cholesterol-­induced NAFLD–HCC formation increased; while Bifidobacterium and was associated with gut microbiota dysbiosis. The Bacteroides were depleted in HFHC-­fed microbiota composition clustered distinctly along stages mice, which was corroborated in human of steatosis, steatohepatitis and HCC. Mucispirillum, hypercholesteremia patients. Desulfovibrio, Anaerotruncus and Desulfovibrionaceae ►► Gut bacterial metabolites alteration including increased sequentially; while Bifidobacterium and increased serum taurocholic acid (TCA) on September 24, 2021 by guest. Protected copyright. Bacteroides were depleted in HFHC-fed­ mice, which was and depleted 3-­indolepropionic acid (IPA) corroborated in human hypercholesteremia patients. was found in NAFLD–HCC. IPA inhibited Dietary cholesterol induced gut bacterial metabolites cholesterol-­induced lipid accumulation and cell alteration including increased taurocholic acid and proliferation, while TCA aggravated cholesterol-­ decreased 3-indolepropionic­ acid. Germ-free­ mice induced triglyceride accumulation in human gavaged with stools from mice fed HFHC manifested normal immortalised hepatocyte cell line. hepatic lipid accumulation, inflammation and cell ►► Germ-­free mice gavaged with stools from proliferation. Moreover, atorvastatin restored cholesterol-­ HFHC-­fed mice manifested hepatic lipid induced gut microbiota dysbiosis and completely accumulation, inflammation and cell prevented NAFLD–HCC development. proliferation, in consistence with donor mice Conclusions Dietary cholesterol drives NAFLD–HCC phenomes. formation by inducing alteration of gut microbiota and ►► Anticholesterol treatment restored dietary metabolites in mice. Cholesterol inhibitory therapy and © Author(s) (or their cholesterol-­induced gut microbiota dysbiosis employer(s)) 2020. Re-­use gut microbiota manipulation may be effective strategies and completely prevented NAFLD–HCC permitted under CC BY-­NC. No for NAFLD–HCC prevention. formation. commercial re-­use. See rights and permissions. Published How might it impact on clinical practice in the by BMJ. foreseeable future? INTRODUCTION ► Our findings indicate that anticholesterol drug To cite: Zhang X, Coker OO, Non-­alcoholic fatty liver disease (NAFLD) is the ► Chu ESH, et al. Gut Epub and manipulation of the gut microbiota might hepatic manifestation of metabolic syndrome, ahead of print: [please represent effective strategies in preventing include Day Month Year]. which encompasses a spectrum of liver pathologies NAFLD–HCC. doi:10.1136/ that range from simple steatosis to non-­alcoholic gutjnl-2019-319664 steatohepatitis (NASH).1 NASH can progress to

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 1 Hepatology hepatic cirrhosis, end-stage­ liver failure and hepatocellular carci- stored at −80°C for further experiments. Mice stool samples noma (HCC).2 Currently, NAFLD is a major cause of morbidity were collected for bacterial 16S rRNA gene sequencing and oral and a healthcare burden worldwide. Large population-based­ gavage. Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from cohort studies demonstrate that the prevalence of NAFLD–HCC C57BL/6 male germ-­free mice (7 weeks) were bred at the has increased by fourfold in the last decade compared with 2.5-­ Department of Laboratory Animal Science at the Third Military fold for hepatitis, making it the most rapidly growing indication Medical University in Chongqing, China. Adult mice (8 weeks for liver transplantation.3 Lipotoxicity drives the progression of old) were divided into three groups (11–28 mice per group), NASH, fibrosis/cirrhosis, and even HCC.4 Among hepatic lipid and gavaged twice at 0 and 7 months with stools obtained from species, cholesterol is considered a major lipotoxic molecule conventional mice fed with NC, HFLC, HFHC or HFHC mice in NASH development.4 The liver is central to the regulation treated with atorvastatin for 14 months. Briefly, 1 g of stool of systemic cholesterol homeostasis. Abnormalities in hepatic samples were homogenised in 5 mL of phosphate buffered saline cholesterol homeostasis have been demonstrated in both human (PBS). Recipient mice were then transplanted with 200 μL of and experimental models of NASH.5 6 We have revealed that the suspension by gastric gavage. Mice from each group were squalene epoxidase, a rate-limiting­ enzyme in cholesterol biosyn- randomly selected and sacrificed at 8, 10 or 14 months following thesis, drives NAFLD–HCC development.7 Dietary cholesterol transplantation. has an important impact on plasma and hepatic cholesterol Additional methods are provided in online supplementary file homeostasis.5 Although the contribution of dietary cholesterol 1. to NASH progression has been reported,5 8 the role and patho- genic basis of long-term­ cholesterol treatment on spontaneous RESULTS and progressive NAFLD–HCC development are unknown. Dietary cholesterol drives the development of NAFLD–HCC The intestinal microbiota has a symbiotic relationship with spontaneously its host and contributes nutrients and energy by metabolising To examine the role of dietary cholesterol in the evolution of dietary components including cholesterol in the large intes- steatosis, NASH, fibrosis and subsequent NAFLD–HCC, mice tine.9 Several studies have suggested that the gut microbiome were fed with HFHC,17 HFLC or NC. Serum level of alpha-­ represents an environmental factor contributing to the devel- fetoprotein (AFP, marker for liver cancer) was monitored at opment of NAFLD and its progression to NAFLD–HCC.10 11 different time points (3, 8, 10, 12 and 14 months). An elevated Microbe-­derived metabolites, such as bile acid, short-­chain AFP level was observed at month 10 (107.1±127.9 ng/mL), fatty acids and trimethylamine and the signalling pathways they which was further increased at month 12 (151.6±129.3 ng/mL) affect may contribute to NAFLD development.12 13 In partic- and month 14 (174.2±203.1 ng/mL) in mice fed with HFHC ular, 3-(4-­hydroxyphenyl) lactate, a metabolite derived from gut compared with HFLC (50.9±7.5 ng/mL in HFLC) or NC microbiome, has a shared gene‐effect with both hepatic steatosis (59.7±20.9 ng/mL) at month 14 (figure 1A). MRI scan showed and fibrosis.14 Moreover, the alteration of gut microbiota profile liver tumours in HFHC-fed­ mice but not in HFLC-fed­ or by dietary cholesterol has been reported.15 However, whether gut NC-fed­ mice at month 14 (figure 1B). Mice were then harvested microbiota dysbiosis is the cause or effect of dietary cholesterol-­ http://gut.bmj.com/ at month 14. Liver tumours were identified in 68% (13/19) of induced NASH and NAFLD–HCC progression remains unclear. mice fed with HFHC but not in mice fed with HFLC or NC diet The present study was performed to determine the role and (figure 1C). Histological examination of liver sections confirmed the associated molecular mechanisms of dietary cholesterol that all liver tumours were HCCs (figure 1C) with an average in the development of NAFLD–HCC. We found that dietary number of 2.7±2.6 HCCs per mouse and maximal tumour diam- cholesterol caused spontaneous NAFLD–HCC formation by eter of 4.1±5.0 mm. The liver sections of HFHC-fed­ mice have cholesterol-­induced gut microbiota changes and metabolomic significantly more Ki-67 positive cells compared with HFLC-­fed alterations. Cholesterol inhibition restored gut microbiota and on September 24, 2021 by guest. Protected copyright. mice, indicating increased cell proliferation in HFHC-­fed mice completely prevented NAFLD–HCC development. (figure 1C). These results demonstrate that dietary cholesterol can spontaneously induce NAFLD–HCC formation. MATERIALS AND METHODS Along with HCC formation, mice fed with HFHC showed Animals and diets significantly increased body weight, visceral fat, liver weight and Male C57BL/6 wild-type­ littermates (8 weeks old) were fed with liver-to-­ ­body weight ratio compared with NC-­fed mice at month normal chow (NC, 18% fat, 58% carbohydrate, 24% protein, 14 (figure 1D). HFLC-fed­ mice also displayed enhanced body 0% cholesterol), high-­fat/low-­cholesterol diet (HFLC, 43.7% weight, visceral fat and liver weight (figure 1D). Significantly fat, 36.6% carbohydrate, 19.7% protein, 0.013% cholesterol) or increased serum cholesterol, hepatic free cholesterol and choles- high-­fat/high-­cholesterol diet (HFHC, 43.7% fat, 36.6% carbo- terol ester, glucose intolerance and fasting insulin were demon- hydrate, 19.7% protein, 0.203% cholesterol) (Specialty Feeds, strated in HFHC-­fed mice compared with HFLC-­fed or NC-fed­ Glen Forrest, WA) ad libitum for 14 months. Mice were also fed mice (figure 1E). Enhanced glucose intolerance and fasting with HFLC or HFHC for 3, 8,10 and 12 months (n=8–19 per insulin were also observed in mice fed with HFLC compared group). In additional experiments, atorvastatin (20 mg/kg) was with NC (figure 1E). administered to HFHC-fed­ mice after 7 months of diet initiation and continued for additional 7 months (n=10 per group). Mice NASH and fibrosis are present in mice fed with high- were kept on a 12/12 hour light/dark cycle. At experimental end cholesterol diet for 14 months points, mice were fasted and serum/tissues were harvested.16 Further examination of liver sections revealed the presence of Body weights and visceral fat weights were recorded. Livers steatohepatitis characterised by steatosis and lobular inflamma- were rapidly excised and weighed. The presence and dimensions tion in HCC adjacent liver tissues and non-­HCC liver tissues of of surface nodules were evaluated. HCCs were confirmed histo- mice fed with HFHC for 14 months, while only steatosis was logically from both grossly and histologically evident nodules. observed in HFLC-­fed mice (figure 2A, online supplementary Liver tumours were isolated, snap frozen in liquid nitrogen and figure S1). Consistent with histological inflammation, serum

2 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 Hepatology

A 800 B MRI

Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from * NC 14 months HFLC 14 months HFHC 14 months 600 * 400

200 AFP (ng/mL) AFP

0 NC HFLC 3m 8m 10m 12m 14m H&E staining Ki-67 staining HFHC C NC * 2.0 ** 1.5 1.0 HFLC

-67 scores -67 0.5

Ki 0.0 NC HFLC HFHC

HFHC T T

D 80 ** 6 * 6 ** 0.15 ** ** ** * * 60 4 4 0.10 40 ** /body 2 2

20 0.05

Liver weight ratio Liver Liver weight (g) Body Body weight (g) 0 0 0 0.00

NC HFLC HFHC NC HFLC HFHC NC HFLC HFHC NC HFLC HFHC Visceral fat weight (g)

http://gut.bmj.com/ NC E *** 5 1.5 ** HFLC 500 /L) HFHC *** *** * 30 L) 4 *** 5 ** 400 * 1.0 20 ** 4 300 3 mmol * 3

(mg/dl) 200 2 0.5 10 * 2

100 liver) (mg/g *** 1 liver) (mg/g 1

0 Insulin (ng/m

on September 24, 2021 by guest. Protected by copyright. Serum cholesterol Serum cholesterol 0 0 0 Glucose ( 0.0 0 30 60 90 120 NC HFLC HFHC Hepatic free cholesterol NC HFLC HFHC NC HFLC HFHC NC HFLC HFHC Hepatic cholesterol cholesterol ester Hepatic Time (min)

Figure 1 Cholesterol induces spontaneous HCC formation in C57BL/6 mice. (A) Serum AFP levels in mice fed with HFHC diet for 3, 8, 10, 12 and 14 months as well as mice fed with NC or HFLC diet for 14 months; (B) liver MRI, (C) representative gross morphology, representative microscopic features and immunohistochemistry pictures of Ki67 staining in liver of mice fed with NC, HFLC and HFHC for 14 months, Ki-67 was scored according to the following criteria: 0 (<10% of cells staining), 1 (11%–30% of cells staining), 2 (30%–50% of cells staining) or 3 (>50% of cells staining); (D) body weight, visceral fat, liver weight, liver-­to-­body weight ratio, (E) serum cholesterol level, hepatic free cholesterol, hepatic cholesterol ester content, glucose tolerance test and fasting insulin levels in mice fed with NC, HFLC and HFHC for 14 months. *p<0.05, **p<0.01, ***p<0.001. AFP, alpha-­ fetoprotein; NC, normal chow; HFLC, high-­fat/low-cholesterol­ diet; HFHC, high-­fat/high-­cholesterol diet; H&E, hematoxylin and eosin. alanine aminotransferase (ALT) (p<0.01) and aspartate amino- presented severe fibrotic injury with significantly more collagen transferase (AST) (p<0.01) levels were significantly increased distribution areas (figure 2E), collagen content by hepatic in HFHC-fed­ mice compared with mice fed with HFLC or hydroxyproline assay (figure 2F) and hepatic stellate cells acti- NC (figure 2B). Serum and hepatic proinflammatory cytokines vation as evidenced by increased alpha-­smooth muscle actin including IL-6, IL-1 and IL-1 (figure 2C,D) and proinflamma- α β (α-SMA) mRNA and protein levels (figure 2E). Examination tory factors (online supplementary figure S2A-B)­ are increased of hepatic oxidative stress revealed that oxidised nicotinamide in HFHC-­fed mice as measured by cytokine profiling assay and adenine dinucleotide (NAD+) to NADH (reduced form of ELISA. The crucial NASH-related­ proinflammatory cytokines, including Cx3cl1, Mcp1, Cxcl10, Mip1β, Mip1α, Ccl5, Cxcl16 NAD)ratio and antioxidant superoxide dismutase (SOD) activity and Tnfα, are significantly upregulated in the liver tissues of 14 were significantly decreased in HFHC-fed­ mice (figure 2G), months HFHC-­fed mice compared with HFLC-fed­ mice as indi- suggesting the induction of hepatic oxidative stress by dietary cated by RNA sequencing analysis (figure 2D). HFHC-fed­ mice cholesterol. Collectively, these findings indicate that NASH and

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 3 Hepatology

NC HFLC HFHC A *** 3 *** from Downloaded 2020. July 21 on 10.1136/gutjnl-2019-319664 as published first Gut: *** 2 ** *

Scores 1 H&E staining H&E 0 Steatosis Inflammation

800 B 600 * C ** 60 * 60 1500 ** ** * * 600 **

(pg/mL) *

400 40 (pg/mL) 40 1000 1β

400 1α -

200 20 20 500 200

0 (U/L) AST activity 0 0 0

ALT activity (U/L) activity ALT 0

Serum IL Serum Serum IL-6Serum (pg/mL) NC HFLC HFHC NC HFLC HFHC NC HFLC HFHC IL- Serum NC HFLC HFHC NC HFLC HFHC

25 D 3000 * ** HFLC 800 1500 HFHC

** 20 1β

α **

- -6

600 -1 * 2000 1000 15 400

mRNA levels mRNA ** 10 **

1000 500 ** *

Hepatic IL /mg tissue) /mg

200 (pg *** /mg tissue) /mg (pg **

Hepatic IL 5

Hepatic IL

/mg tissue) (pg/mg

0 0 0 Relative 0 HFLC HFHC HFLC HFHC HFLC HFHC Cx3cl1 Mcp1 Cxcl10 Mip1β Mip1α Ccl5 Cxcl16 Tnfα

E NC HFLC HFHC *** 25000 ***

20000

/field) 15000

2 m

μ 10000 (

http://gut.bmj.com/ 5000

0 Sirius red Sirius staining Hepatic collagen area NC HFLC HFHC

30 *

20

on September 24, 2021 by guest. Protected by copyright. by Protected guest. by 2021 24, September on -SMA mRNA

10

-SMA staining -SMA α 0 Relative α Relative NC HFLC HFHC *** *** F G 300 * 0.6 *** 250 *** 200 * 0.4 250 150

100 *

μg/mg)

( 0.2 200

50

Hydroxyproline Liver NAD+/NADH Liver 0.0 0 150 NC HFLC HFHC NC HFLC HFHC pro) (U/mg SOD activity NC HFLC HFHC

Figure 2 Cholesterol causes NASH and fibrosis in non-­HCC liver tissues of mice fed with HFHC diet. (A) Representative H&E staining, histological scores of liver sections; (B) serum ALT and AST levels; (C) serum IL-6, IL-1α and IL-1β protein levels by cytokine profiling assay in mice fed with NC, HFLC and HFHC for 14 months; (D) hepatic proinflammatory cytokines IL-6, IL-1α and IL-1β protein levels by ELISA and Cx3cl1, Mcp1, Cxcl10, Mip1β, Mip1α, Ccl5, Cxcl16, Tnfα mRNA levels by RNA sequencing in mice fed with HFLC and HFHC for 14 months; (E) collagen deposition by Sirius Red staining, α-SMA protein and mRNA levels by immunohistochemistry staining and RT-PCR,­ respectively; (F) hepatic hydroxyproline content, (G) liver NAD+ to NADH ratio and SOD activity in mice fed with NC, HFLC and HFHC for 14 months. *p<0.05, **p<0.01, ***p<0.001. ALT, alanine aminotransferase; AST, aspartate aminotransferase; α-SMA, alpha-­smooth muscle actin; HCC, hepatocellular carcinoma; HFHC, high-fat/high-­ ­ cholesterol diet; HFLC, high-­fat/low-cholesterol­ diet; H&E, Hematoxylin and eosin; NAD, nicotinamide adenine dinucleotide; NASH, non-alcoholic­ steatohepatitis; NC, normal chow; RT-­PCR, reverse transcription polymerase chain reaction; SOD, superoxide dismutase.

4 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 Hepatology fibrosis were formed in non-­HCC liver tissues of HFHC-­fed Desulfovibrio_Otu047, Anaerotruncus_Otu107 and Desulfovib- mice. rionaceae_Otu073 were observed to have sequentially increased from 3 to 8, and to 14 months of HFHC feeding (figure 4D). Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from Also, among others, we observed the enrichment of Clostridium High-cholesterol diet causes fatty liver, steatohepatitis and OTUs such as Clostridium celatum_Otu070, C. ruminan- fibrosis sequentially in mice tium_Otu059, C. cocelatum_Otu036 and C. methylpentosum_ To clarify the progression of NAFLD prior to HCC formation by Otu053, and the depletion of Bifidobacterium_Otu026, dietary cholesterol, we monitored serum ALT, AST, cholesterol, Akkermansia municiphila_Otu034, Lactobacillus_Otu009, AFP and liver histological changes in mice fed with HFHC. Bacteroides acidifaciens_Otu032, Bacteroides_Otu012, B. Mice were harvested at 3, 8, 10 and 12 months following diet uniformis_Otu080 and B. eggerthii_Otu079 with high dietary feeding (figure 3A). We observed increases in body weight and cholesterol. Moreover, HFHC-­enriched Helicobacter ganmanii_ visceral fat weight (online supplementary figure S3A), accompa- Otu031 was more abundant in HFHC-­fed mice with tumour nied with increased liver weight and liver-to-­ ­body weight ratio in compared with HFHC-­fed mice without tumour, while HFHC-­ mice fed with HFHC compared with mice fed with HFLC at 3 depleted Bacteroides_Otu012 was reduced in HFHC-fed­ mice and 8 months (online supplementary figure S3B). The increased with tumour compared with HFHC-­fed mice without tumour profiles of serum ALT and AST (online supplementary figure (online supplementary figure S5C). These suggest that further S3C) were consistent with enhanced serum cholesterol levels enrichment of Helicobacter ganmanii_Otu031 and depletion of in HFHC-­fed mice (online supplementary figure S3D). Liver Bacteroides_Otu012 may be important for the role of gut micro- histology of HFHC-fed­ mice showed steatosis with mild inflam- biota in NAFLD–HCC. In addition, gut microbiota associated mation at 3 months, steatohepatitis with fibrosis at 8 months tryptophan metabolising capacity was reduced in HFHC-fed­ and HCC formation at 10, 12 and 14 months, while HFLC-­fed mice compared with HFLC-fed­ mice (figure 4E). These, taken mice showed only steatosis without further HCC development together, suggest that high-­cholesterol diet-­induced gut micro- at 3, 8, 10 and 14 months (figure 3B and online supplementary biota dysbiosis and impaired microbial tryptophan metabolism. figure S4A-­B). We also performed immunohistochemistry for Correlation analysis was performed to determine the poten- two HCC markers AFP and Golgi protein 73 (GP73). Positive tial association of bacterial abundance with mice phenomes. We staining was observed on liver tissues from HFHC-­fed mice but observed that M. schaedleri_Otu038, Desulfovibrio_Otu047, not from HFLC-­fed mice (figure 3B and online supplementary Anaerotruncus_Otu107, C celatum_Otu070, C. cocelatum_ figure S4A). Scores of hepatic steatosis, and lobular inflamma- Otu036 and C. methylpentosum_Otu053 which were enriched tion and hepatic collagen areas confirmed the increased severity in faecal samples of HFHC-­fed mice were positively correlated of liver histology across the disease stages (figure 3B). Moreover, (figure 4F) while Bifidobacterium_Otu026, B. acidifaciens_ 10% (1/10) of HFHC-­fed mice developed HCC at 10 months (tumour number=0.1 ± 0.32, maximum tumour diameter=0.55 Otu032, B. uniformis_Otu080, A. municiphila_Otu034 and ± 1.74 mm), 25% (3/12) developed HCC at 12 months (tumour Lactobacillus_Otu009, which were depleted in HFHC-­fed mice, number=0.25 ± 0.45, maximum tumour diameter=1.57 were negatively correlated with high-­cholesterol diet, serum and ± 3.10 mm) and 68% (13/19) developed HCC at 14 months liver cholesterol levels (figure 4F). These results suggest that gut http://gut.bmj.com/ (figure 3C), indicating that HFHC-fed­ mice progressively devel- microbiota dysbiosis in NAFLD–HCC correlates with choles- oped steatosis, steatohepatitis, fibrosis and NAFLD–HCC. terol levels. To corroborate our findings from animal experiments in human patients, we analysed the correlation of serum choles- Dietary cholesterol-induced gut microbiota dysbiosis in the terol and gut microbiota in 59 cases of hypercholesterolemia initiation and progression of NAFLD–HCC and 39 healthy subjects. The characteristics of these 98 subjects In order to explore the potential involvement of the gut micro- are shown in online supplementary table S6. Bifidobacterium on September 24, 2021 by guest. Protected copyright. biota in mediating dietary cholesterol-­induced NAFLD–HCC, we and Bacteroides were negatively correlated with serum total performed 16S rRNA gene sequencing on stools from HFLC-fed­ cholesterol and low-­density lipoprotein (LDL)-­cholesterol but and HFHC-fed­ mice at 14 months. Rarefaction analysis shows positively correlated with high-­density lipoprotein (HDL)-­ that observed number of operational taxonomic units (OTUs) cholesterol (figure 4G). These results were consistent with reached saturation (online supplementary figure S5A-­B). Gut observations in HFHC-­fed mice (online supplementary tables microbiota compositional discrimination, by weighted Unifrac S1-5), further inferring the involvement of the gut microbiota in principal component analysis (PCA), was observed between contributing to cholesterol-­induced disorder. HFLC-fed­ and HFHC-­fed mice at 14 months (p<0.001) (figure 4A). Additionally, lower bacterial diversity and increased bacterial richness were observed in HFHC-­fed mice with HCC Dietary cholesterol promotes NASH–HCC progression by than HFLC diet-fed­ mice with only simple steatosis at 14 months inducing metabolites alteration (figure 4A). Cholesterol associated bacterial taxa were determined To reveal metabolic phenotypes, related to the gut microbiome, by differential abundance analysis. Several bacterial OTUs were that are potentially involved in cholesterol-induced­ NAFLD– differentially abundant in mice fed with HFHC compared with HCC, we performed metabolic profiling of the serum from HFLC-fed­ mice (figure 4B, online supplementary tables S1-5). HFHC-­fed and HFLC-fed­ mice. Serum metabolites were signifi- Principal component and redundancy analyses also showed that cantly different according to dietary cholesterol content by PCA the microbiota composition clustered distinctly for mice fed with (figure 5A and online supplementary table S7). Bile acid biosyn- HFHC for 3 months (simple steatosis with mild inflammation), thesis was a key pathway altered in mice fed with high-cholesterol­ 8 months (steatohepatitis with fibrosis) and 14 months (HCC) diet (figure 5B). Primary bile acids including taurocholic acid (figure 4C) indicating compositional changes in gut microbiota (TCA), tauroursodeoxycholic acid (TUDCA), glycocholic acid along stages of NAFLD–HCC progression. Moreover, bacterial (GCA) and taurochenodeoxycholic acid (TCDCA) were outlier richness was sequentially increased with NAFLD–HCC progres- upregulated metabolites in HFHC-fed­ mice (figure 5C). On the sion (figure 4D). In particular, Mucispirillum schaedleri_Otu038, other hand, 3-indolepropionic­ acid (IPA), which is a product

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 5 Hepatology

A C57BL/6 Time points 8 wk old Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from 0 3m 8m 10m 12m 14m

HFHC diet

B HFHC HFLC 3m 8m 10m 12m 14m H&E (low) H&E (high) H&E AFP GP73 Sirius red

http://gut.bmj.com/ Histological scores of liver sections

HFLC HFHC_3m HFHC_8m HFHC_10m HFHC_12m HFHC_14m Steatosis 0.68±0.53 1.13±0.99 1.41±0.34** 1.74±0.70*** 1.69±0.42*** 1.41±0.55**

Inflammation 0.63±0.52 0.32±0.11 0.90±0.52** 1.47±0.44*** 1.39±0.48*** 2.39±0.46*** Hepatic collagen 3505±3051 2180±716.2 25017±14652* 20875±13290* 32741±22194* 20227±3267*** on September 24, 2021 by guest. Protected copyright. area (µm2/field) *p<0.05, **p<0.01, ***p<0.001 v.s. HFLC. C 80 10 10

60 8 8 6 6

incidence 40 (%) 4 4 20

2 2

diameter (mm) Maximum tumour Maximum Tumour 0 Tumour number 0 0 HFLC 3m 8m 10m 12m 14m HFLC 3m 8m 10m 12m 14m HFLC 3m 8m 10m 12m 14m HFHC HFHC HFHC

Figure 3 HFHC-­fed mice sequentially develop a fatty liver, steatohepatitis, fibrosis and HCC. (A) Schematic illustration of the treatment of C57BL/6 mice with HFHC diet. Mice were sacrificed at 3, 8, 10, and 12 months of age. (B) Representative gross morphology, H&E staining and IHC staining of tumour markers AFP and GP73 in the liver of mice fed with HFLC for 14 months and HFHC for 3, 8, 10, 12 and 14 months; histological scoring of steatosis and inflammation and the quantitation of Sirius Red staining were calculated. (C) Tumour incidence, tumour number and maximum tumour diameter of the largest tumour from mice fed with HFHC at different time points. *p<0.05, **p<0.01, ***p<0.001. AFP, alpha-fetoprotein;­ HCC, hepatocellular carcinoma; HFHC, high-­fat/high-cholesterol­ diet; HFLC, high-­fat/low-cholesterol­ diet; H&E, hematoxylin and eosin; IHC, immunohistochemistry.

6 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 Hepatology

HFHC_3m HFHC_3m HFHC_8m HFHC_8m A HFLC HFLC C HFHC_14m HFLC p=0.0014 HFHC_14m HFHC HFHC HFHC Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from p=0.018 Chao1 Shannon diversity Shannon

B D Mucispirillum_schaedleri Desulfovibrio HFLC HFHC _Otu038 _Otu047

Helicobacter_ganmani_Otu031 Odoribacter_Otu028 p=0.00004 Enterococcus_Otu027 HFHC_3m Adlercreutzia_Otu025 HFHC_8m Peptostreptococcaceae_Otu022 Alistipes_finegoldii_Otu019 HFHC_14m Bacteroidetes_Otu018 Rikenellaceae_Otu015 Ruminococcaceae_Otu011 Ruminococcus_gnavus_Otu006 Oscillospira_Otu004

Clostridiales_Otu003 Chao1 Lachnospiraceae_Otu001 Anaerovorax_Otu108 Anaerotruncus Desulfovibrionaceae Anaerotruncus_Otu107 Christensenellaceae_Otu092 _Otu107 _Otu073 Streptococcaceae_Otu088 Paraprevotella_Otu083 Desulfovibrionaceae_Otu073 Helicobacteracae_Otu072 Clostridium_celatum_Otu070 Mogibacteriaceae_Otu069 Dehalobacterium_Otu066 Clostridium_Otu063 Helicobacter_Otu062 Desulfovibrionales_Otu060 Roseburia_Otu056 Clostridium_rumniantium_Otu059 (square Abundance root) Relative Parabacteroides_Otu057 Clostridium_methylpentosum_Otu053 Dorea_Otu050 Gemella_Otu048 Desulfovibrio_Otu047 AF12_Otu046 E Tryptophan metabolism Rikenella_Otu045 G Human subjects Desulfovibrio_C21_c20_Otu042 Mucispirillum_schaedleri_Otu038 p=0.0082 Genus Bilophila_Otu037 Clostridium_cocleatum_Otu036 SMB53_Otu035 Lactococcus_Otu033 Clostridium_Otu021 YS2_Otu040 Bacteroides_acidifaciens_Otu032 Bacteroides_Otu012 RF32_Otu065 Parabacteroides_distasonis_Otu068 Olsenella_Otu055 Allobaculum_Otu005 Bifidobacterium_Otu026 Sutterella_Otu074 Bacteroides_uniformis_Otu080 Bacteroides_eggerthii_Otu079 Eubacterium_dolichum_Otu146 Prevotella_Otu167 Defluviitalea_sacchaophila_Otu313 Firmicutes_c_Otu007 Anaeroplasma_Otu122 Lactobacillus_Otu009 Tenericutes_o_RF39_Otu044 Erysipelotrichaceae_Otu043 (square Abundance root) Relative Akkermansia_muciniphila_Otu034 HFLC HFHC S247_Otu002

F

High cholesterol diet http://gut.bmj.com/ Liver to body weight ratio Serum AST Serum ALT Total liver cholesterol Liver weight Serum cholesterol Serum free cholesterol Serum triglyceride Serum AFP Fast glucose Body weight Adipose weight Liver triglyceride Low cholesterol diet 92 on September 24, 2021 by guest. Protected copyright. YS2_Otu040 S246_Otu002 AF12_Otu046 RF39_Otu044 RF32_Otu065 Bacilli_Otu010 Dorea_Otu050 Blautia_Otu087 SMB53_Otu035 Allstipes_Otu049 Gemelia_Otu048 Bilophlia_Otu037 Rikenelia_Otu045 Clostridia_Otu052 Olsenella_Otu055 Sutterella_Otu074 Bacillales_Otu041 Prevotella_Otu167 Roseburia_Otu056 Firmicutes_Otu007 Clostridium_Otu013 Clostridium_Otu102 Oscillospira_Otu004 Turicibacter_Otu024 Odoribacter_Otu028 Bacteroides_Otu012 Clostridiales_Otu003 Allobaculum_Otu005 Helicobacter_Otu062 Lactococcus_Otu033 Enterococcis_Otu027 Desulfovibrio_Otu047 Lactobacillus_Otu009 Coprococcus_Otu030 Adiercreutzia_Otu025 Sporosarcina_Otu090 Anaerotruncs_Otu107 Coprobacillus_Otu084 Bacteroidales_Otu020 Bacteroidetes_Otu018 Rikenellaceae_Otu015 Lactobacilales_Otu014 Streptococcus_Otu039 Paraprevotella_Otu083 Ruminococcus_Otu029 Bifidobacterium_Otu026 Staphylococcus_Otu071 Escherichia_coll_Otu168 Parabacteroides_Otu057 Lactobacillaceae_Otu016 Dehalobacterium_Otu066 Lachnospiraceae_Otu001 Mogibacteriaceae_Otu069 Desulfovibrionales_Otu060 Alistipes_finegoldli_Otu019 Ruminococcaceae_Otu011 Helicobacteraceae_Otu072 Christenseneliaceae_Otu Erysipelotrichaceae_Otu043 Clostridium_Celatum_Otu070 Alistipes_Indistinctus_Otu106 Bacteria_unclassified_Otu008 Bacterodes_eggerthll_Otu079 Bacteroides_uniformis_Otu080 Staphylococcus_sciuri_Otu094 Helicobacter_ganmani_Otu031 Clostridium_cocelatum_Otu036 Ruminococcus_gnavus_Otu006 Desulfovibrio_C21_c20_Otu042 Helicobacter_hepaticus_Otu061 Peptostreptococcaceae_Otu022 Candidatus Arthromitus_Otu081 Atopostipes_suicioacalis_Otu023 Mucispirillum_schaedleri_Otu038 Bacteroides_acidifaciens_Otu032 Akkermancia_muciniphila_Otu034 Parabacteroides_gordonlie_Otu076 Butyricoccus_pullicaecorum_Otu086 Parabacteroides_distasonis_Otu068 Clostridium_methylpentosum_Otu053

Figure 4 Dietary cholesterol-­induced gut microbiota dysbiosis. (A) Principal component ordination analysis (PcoA), Shannon diversity and the richness of gut microbiota between 14 months HFLC and HFHC-fed­ mice. (B) Heatmap plot of the bacteria in stool of mice fed with HFLC or HFHC for 14 months. (C) PCoA and redundancy analysis of gut microbiota in fed with HFHC for 3 months, 8 months and 14 months, respectively. (D) The microbiota richness measured by chao1 index and sequentially increased bacteria from 3, 8 to 14 months of HFHC feeding. (E) Gut microbiota tryptophan metabolism capacity in HFLC-­fed and HFHC-­fed mice. (F) Correlation of bacterial abundance with mice phenomes. (G) The association of bacteria by metagenomic sequencing and serum total cholesterol, triglyceride, LDL-cholesterol­ and HDL cholesterol in 59 human cases of hypercholesterolemia and 39 healthy subjects. *p<0.05, **p<0.01, ***p<0.001. AFP, alpha-fetoprotein;­ ALT, alanine aminotransferase; AST, aspartate aminotransferase; HDL, high-­density lipoprotein; HFHC, high-­fat/high-­cholesterol diet; HFLC, high-­fat/low-­cholesterol diet; LDL, low-­density lipoprotein; TCHO, total cholesterol; TG, triglyceride. of microbial tryptophan metabolism,18 was an outlier down-­ and metabolites. Consistently, correlation analysis revealed that regulated metabolite in HFHC-­fed mice (figure 5C). Moreover, HFHC-­enriched Mucispirillum schaedleri_Otu038 is positively serum lipopolysaccharides (LPS) concentration in portal vein correlated with TUDCA, TCDCA, TCA and GCA. HFHC-­ was elevated and loss of colonic E-cadherin­ was identified in enriched Roseburia_Otu056 and Helicobacter_ganmanii_ HFHC-­fed mice compared with HFLC-­fed mice (figure 5D), Otu031 are also positively associated with TUDCA and TCDCA. suggesting that dietary cholesterol could impair intestinal Moreover, HFHC-­depleted Akkermancia_muciniphila_Otu034 barrier function. Correlation analysis was performed to deter- is negatively associated with TCDCA and TUDCA while mine the potential association of the HFHC-­altered microbes

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 7 Hepatology

A HFLC B Metabolites Sets Enrichment Overview C Metabolomics: Conventional mice

HFHC from Downloaded 2020. July 21 on 10.1136/gutjnl-2019-319664 as published first Gut: Bile Acid Biosynthesis HFHC vs. HFLC 9 Metabolomics: Conventional mice Amino Sugar Metabolism p value 8 Galactose Metabolism 6e-02 7 Spermidine and Spermine Biosynthesis 6 IPA Vitamin B6 Metabolism 5 4e-01

PC2 4

Betaine Metabolism Log10 (p) TCDCA Starch and Sucrose Metabolism 3 8e-01 2 GCA Methionine Metabolism TCA 1 Glutamate Metabolism TUDCA 0 Glycine and Serine Metabolism -1 -5 -4 -3 -2 -1 0 1 2 3 4 5 0.0 0.5 1.0 1.5 2.0 2.5 Fold Enrichment Log2 (FC) F D 20 E-cadherin

** 15 2.5

HFLC 2.0 10

1.5

(ng/mL) portal vein portal 5 1.0

LPS concentration in 0.5 HFHC

0 levels mRNA Relative HFLC HFHC 0.0 Cyp7a1 Cyp8b1 Cyp27a1 Cyp7b1 E

IPA TCDCA TUDCA TCA

GCA

YS2_Otu040

S247_Otu002

RF39_Otu044

AF12_Otu046

RF32_Otu065

Dorea_Otu050

SMB53_Otu035

Gemella_Otu048

Rikenella_Otu045

Bilophila_Otu037

Olsenella_Otu055

Sutterella_Otu074

Firmicutes_Otu007

Roseburia_Otu056

Prevotella_Otu167

Clostridium_Otu021

Clostridium_Otu063

Lactococcus_Otu033

Odoribacter_Otu028

Bacteroides_Otu012

Allobaculum_Otu005

Clostridiales_Otu003

Oscillospira_Otu004

Enterococcus_Otu027

Lactobacillus_Otu009

Bacteroidetes_Otu018

Anaerovorax_Otu108

Adlercreutzia_Otu025

Helicobacter_Otu062

Desulfovibrio_Otu047

Rikenellaceae_Otu015

Aerococcaceae_Otu054

Anaerotruncus_Otu107

Paraprevotella_Otu083

Bifidobacterium_Otu026

Parabacteroides_Otu057

Lachnospiraceae_Otu001

Streptococcaceae_Otu088

Mogibacteriaceae_Otu069

Dehalobacterium_Otu066

Ruminococcaceae_Otu011

Desulfovibrionales_Otu060

Helicobacteraceae_Otu072

Alistipes_finegoldii_Otu019

Erysipelotrichaceae_Otu043

Christensenellaceae_Otu092

Clostridium_celatum_Otu070

Desulfovibrionaceae_Otu073

Bacteroides_eggerthii_Otu079

Bacteroides_uniformis_Otu080

Helicobacter_ganmani_Otu031

Peptostreptococcaceae_Otu022

Ruminococcus_gnavus_Otu006

Desulfovibrio_C21_c20_Otu042

Clostridium_cocleatum_Otu036

Bacteroides_acidifaciens_Otu032

Akkermansia_muciniphila_Otu034

Mucispirillum_schaedleri_Otu038

Clostridium_ruminantium_Otu059 Parabacteroides_distasonis_Otu068

http://gut.bmj.com/ Defluviitalea_saccharophila_Otu313 LO2 Clostridium_methylpentosum_Otu053 HKCI-2 G H 3 DMSO Cholesterol Cholesterol+0.1µM TCA Cholesterol+1µM TCA 1.2 ** DMSO *** IPA (10µM)

IPA (100µM) *** 1.1 *** 2

1.0 Cell viability Cell Relative Oil Relative Red O 1

0.9 Oil Red O staining O Red Oil Cholesterol - + + + TCA(µM) -- 0.1 1 copyright. by Protected guest. by 2021 24, September on HKCI-2 DMSO Cholesterol Cholesterol+10µM IPA Cholesterol+100µM IPA 0 0 24 48 72 96 2.5 ** 2.0 Time (h) 1.5 HKCI-10 1.0 DMSO 0.5 1.5 IPA (10µM)

Relative Oil Relative Red O 0.0 IPA (100µM) Cholesterol - + + +

IPA (µM) -- 10 100 1.0 *** HKCI-10 *** DMSO Cholesterol Cholesterol+10µM IPA Cholesterol+100µM IPA ***

2.0 *** *** Oil Red O staining O Red Oil 1.5 0.5

1.0 viability Cell

0.5 0.0 Relative Oil Relative Red O 0.0 0 24 48 72 96 Cholesterol - + + + IPA (µM) -- 10 100 Time (h)

Figure 5 Dietary cholesterol-­induced alteration of metabolic profiles in mice serum. (A) Serum metabolites were significantly different between mice fed with HFLC and HFHC diet by principal component ordination analysis. (B) Pathway analysis of differentially enriched metabolites in mice fed with HFHC diet. (C) Volcano plot of serum metabolomics of mice fed with HFLC and HFHC diet, the outliers of metabolites are indicated. (D) LPS concentration in portal vein blood of mice fed with HFLC and HFHC diet for 3 months and E-cadherin­ expression in the colon tissues of 14 months HFLC and HFHC diet fed mice. (E) Correlation analysis of the association of the HFHC-altered­ microbes and metabolites. (F) mRNA levels of Cyp7a1, Cyp8b1, Cyp27a1 and Cyp7b1 in the liver tissues of mice fed with HFLC and HFHC diet. (G) TCA aggravated cholesterol-induced­ triglyceride accumulation in human LO2 cell line while IPA inhibited cholesterol-induced­ triglyceride accumulation in NASH–HCC cell lines, HKCI-2 and HKCI-10 by Oil Red O staining. Cholesterol, 200 μg/mL; (H) IPA inhibited cell proliferation in NASH–HCC cell lines. *p<0.05, **p<0.01, ***p<0.001. DMSO, dimethyl sulfoxide; GCA, glycocholic acid; HCC, hepatocellular carcinoma; HFHC, high-fat/high-­ ­cholesterol diet; HFLC, high-­fat/low-cholesterol­ diet; IPA, 3-­indolepropionic acid; NASH, non-­alcoholic steatohepatitis; LPS, lipopolysaccharides; TCA, taurocholic acid; TCDCA, taurochenodeoxycholic acid; TUDCA, tauroursodeoxycholic acid.

8 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 Hepatology

HFHC-­enriched Anaerotruncus_Otu107 is negatively correlated inducing metabolite alteration, in contributing to cholesterol-­ with IPA (figure 5E). induced NAFLD–HCC formation. Further experiments revealed that high dietary cholesterol Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from could not change the mRNA expression of bile acid synthesis Anticholesterol treatment completely prevents NAFLD–HCC enzymes, including cytochrome P450 (Cyp)7a1, Cyp8b1, formation in HFHC-fed mice Cyp27a1 and Cyp7b1 in the liver (figure 5F). Additional in-vitro­ As dietary cholesterol drives the progression of NAFLD–HCC, experiments showed that TCA aggravated cholesterol-­induced we evaluated if anticholesterol drug could dampen NAFLD and triglyceride accumulation in human normal immortalised hepato- its progression to HCC. Atorvastatin (20 mg/kg), a cholesterol-­ cyte cell line LO2 (figure 5G), while IPA suppressed cholesterol-­ lowering drug, was administered to mice that had been fed with induced lipid accumulation (figure 5G), and cell proliferation HFHC for 7 months, and continued for additional 7 months (figure 5H) in NASH–HCC cell lines, HKCI-2 and HKCI-10. with HFHC and atorvastatin (figure 7A). At the end of experi- These results indicated that cholesterol promoted NASH–HCC ment (14 months), atorvastatin completely prevented NAFLD– progression through the modulation of host serum metabolites HCC formation induced by HFHC diet and improved severity and, at least in part, through increased TCA and decreased IPA. of NASH (figure 7B). This was concomitant with significantly reduced serum cholesterol, hepatic free cholesterol, serum AFP High cholesterol-modulated gut microbiota promotes (figure 7C), ALT, lowered serum pro-inflammatory­ cytokines steatohepatitis and hepatocyte proliferation in germ-free (IL-6, IL-1α, IL-1β, MCP-1, MIP-1α and MIP-1β) and oxida- mice tive stress (increased NAD+ to NADH ratio and SOD activity) We evaluated the contribution of cholesterol-modulated­ (figure 7D). Atorvastatin also ameliorated liver fibrosis as microbiota in NAFLD–HCC by faecal microbiota transplan- supported by significant reduction of hepatic collagen deposition tation (FMT). Stools of NC-­fed, HFLC-fed­ and HFHC-­fed and hydroxyproline content (figure 7E). Microbiota analysis by mice (14 months) were gavaged to germ-free­ mice (G-NC,­ 16S rRNA gene sequencing was performed on the stools of mice G-HFLC­ and G-HFHC)­ under NC diet (figure 6A). The liver-­ fed with HFHC under atorvastatin treatment (HFHC+At) and to-body­ weight ratio was significantly higher in G-­HFHC mice compared with mice fed with NC, HFLC and HFHC. Bacterial compared with G-­NC mice (online supplementary figure S6A). richness which increased in HFHC-­fed mice was significantly Moreover, hepatic lipid accumulation by triglyceride assay and restored by anticholesterol atorvastatin treatment (figure 7F). Oil Red O staining (figure 6B) and peroxidation by thiobar- Furthermore, among the dysregulated OTUs in HFHC-fed­ mice bituric acid reactive substances assay (figure 6B) were signifi- (figure 4B and D), the abundance of Mucispirillum schaedleri_ cantly increased accompanied with impaired liver histology by Otu038, Desulfovibrio_Otu047, Anaerotruncus_Otu107 and H&E staining (figure 6B) in G-HFHC­ mice at 8, 10 and 14 Desulfovibrionaceae_Otu073 were found to be reversed by anti- months following FMT. Increased inflammation was evidenced cholesterol treatment (figure 7F). To study the direct effect of by enhanced hepatic cytokines and chemokines including IL-6 the microbiota in anticholesterol drug-prevented­ NAFLD–HCC, we gavaged stool from HFHC+At mice to germ-free­ mice

(figure 6C) by cytokine profiling assay, Fos, Ccl12, Cxcr1, http://gut.bmj.com/ (G-HFHCA­ t, n=10). G-­HFHCAt mice showed improved liver Ccl1, Myd88, Il-1β, Cxcl10 and C3ar1 (online supplementary histology with decreased liver triglyceride and lipid peroxida- table S8) by cDNA expression assay and enhanced CD45+ tion compared with G-­HFHC group at 14 months after stool lymphocytes accumulation (online supplementary figure S6B) gavage (figure 7G). An induction of IPA and a reduction of TCA by flow cytometry in G-HFHC­ mice compared with G-HFLC­ were observed in HFHC mice treated with atorvastatin (online mice. supplementary figure S8). These results further show that gut Increased hepatocyte proliferation was observed in the liver microbiota plays an active role in mediating cholesterol-­induced of the G-HFHC­ mice at 14 months (p<0.05) (figure 6D) but on September 24, 2021 by guest. Protected copyright. NAFLD–HCC. not at 8 and 10 months after FMT. Cancer pathways PCR array revealed the upregulation of genes involved in oncogenic path- ways including cell proliferation (Cdc20), angiogenesis (Pgf), DISCUSSION invasion/metastasis (Serpinb2, Snai3) and downregulation Despite cholesterol being a known cytotoxic lipid in of genes involved in apoptosis (Fasl and Lpl) in liver tissues NASH,5 8 information about the role and the underline mecha- of G-HFHC­ mice compared with G-NC­ and G-HFLC­ mice nisms of cholesterol in HCC development from NASH is limited. (figure 6E1). Upregulation of CDC20 was validated by Western In this study, we demonstrated for the first time that prolonged blot (figure 6E2). One liver nodule was observed in G-­HFHC high dietary cholesterol feeding caused spontaneous NAFLD– mice at 14 months, but not at 8 and 10 months (figure 6A). HCC development in mice. High-­fat diet induces obesity, insulin Histology examination confirmed the nodule as dysplasia with resistance, glucose intolerance and steatosis. The progression globules and increased cell proliferation (figure 6A). We evalu- of steatohepatitis to fibrosis and HCC was, however, predom- ated the composition of gut microbiota in recipient germ-­free inantly associated with high dietary cholesterol. Germ-­free mice and their corresponding donor conventional mice. We mice transplanted with faecal microbiota from high-­cholesterol found that gut microbiota was significantly different among the diet-­fed mice substantially phenocopied the donor mice in lipid recipient germ-­free mice in relation to donors’ diet (figure 6F). accumulation, inflammation and cell proliferation, suggesting The microbial ecosystem after FMT remained stable over time in that gut microbiota dysbiosis contributes to dietary cholesterol-­ all groups of mice as shown by β-diversity and composition anal- induced NASH and potential development of NAFLD–HCC. ysis (online supplementary figure S7A-­D). Furthermore, analysis We investigated the underline mechanisms of HCC devel- of serum metabolites from germ-free­ mice revealed decreased opment from NAFLD by high dietary cholesterol. We found IPA in G-HFHC­ mice compared with G-­HFLC mice in consis- enhanced ROS accumulation in NAFLD–HCC in mice fed with tence with conventional mice fed with HFHC (figure 6G). Taken high-cholesterol­ diet. The accumulated ROS is a toxic mediator, together, these data indicated that dietary cholesterol-­modulated which can induce inflammatory response, insulin resistance and microbiota promote NAFLD and hepatocyte proliferation by oxidative damage.19 Indeed, our analysis revealed that dietary

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 9 Hepatology

Fecal microbiota Germ-free 8m 10m 14m G-HFHC at 14 months A transplantation mice

Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from Normal Normal chow chow Normal chow NC G-NC

Normal staining H&E Normal chow chow Normal chow HFLC G-HFLC

Normal Normal chow chow Normal chow HFHC

G-HFHC

67 staining 67 Ki-

B *** H&E staining Oil red O staining 40 *** 4 ** 30 3 ** G-NC G-HFLC G-HFHC G-NC G-HFLC G-HFHC

8m 20 2 /mg protein) /mg

10 TBARS 1

g/mg protein) g/mg

Triglyceride Triglyceride

μ

8m (

0 (nmol 0 G-NC G-HFLC G-HFHC G-NC G-HFLC G-HFHC

40 ** * 8 * 30 6

10m 20

4

/mg protein) /mg TBARS TBARS

g/mg protein) g/mg 10

Triglyceride Triglyceride 10m

μ 2 ( 0 G-NC G-HFLC G-HFHC (nmol 0 G-NC G-HFLC G-HFHC

40 * 8 *** * *

30 6 14m

20 4 14m

/mg protein) /mg TBARS TBARS

g/mg protein) g/mg 2

Triglyceride Triglyceride 10

μ (

0 (nmol 0 G-NC G-HFLC G-HFHC G-NC G-HFLC G-HFHC Ki-67 staining 70 * C * D G-NC G-HFLC G-HFHC ** 2.5 -6 * 60 2.0

positive positive 1.5 1.0

50 cells(%)

HepaticIL 0.5

14m

67Ki-

/mg tissue) pg/mg ( 0.0 40 G-NC G-HFLC G-HFHC G-NC G-HFLC G-HFHC

E1 Mouse cancer pathway Finder PCR array Angiogenesis E2

G-HFHC vs G-HFLC Pgf 6 G-NC G-HFLC G-HFHC Invasion/Metastasis 4 Apoptosis CDC20 2 Fasl Serpinb2 GAPDH NASH-HCC Stool

http://gut.bmj.com/ 0 Lpl Snai3

Foldchanges -2

-4 Cdc20 Pgf Cdc20 Serpinb2 Snai3 Angpt2 Snai1 Lpl Fasl Cell proliferation

F G-HFLC G-HFHC G G-NC Clostridium_Otu034 Ruminicoccaceae_Otu005 Metabolomics: Germ-free mice G-HFLC Helicobacter_Otu024 Lachnospiraceae_Otu001 G-HFHC vs. G-HFLC

on September 24, 2021 by guest. Protected by copyright. G-HFHC Desulfovibrionaceae_Otu046 4.0 Lactococcus_Otu040 Bacteroidetes_Otu006 Eubacterium_Otu078 3.5 IPA Mucispirillum_Otu045 Odoribacter_Otu033 3.0 Dorea_Otu019 Alistipes_Otu015 Peptostreoptococcaceae_Otu018 2.5 Rikenella_Otu049 Gemella_Otu042 2.0

Clostridium_sensu_stricto_Otu036 Log10(p) Enterorhabdus_Otu039 1.5 Roseburia_Otu027 Oscillibacter_Otu030 1.0 Desulfovibrionales_Otu044 Paraprevotella_Otu051 Clostridium_XIVb_Otu037 0.5 TCA Desulfovibrio_Otu087 Enterococcus_Otu050 0.0 Anaerotruncus_Otu100 Streptococcus_Otu055 Johnsenella_Otu053 -0.5 Clostridium_XVIII_Otu041 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Bacilli_Otu022 Enterococcaceae_Otu057 Log2 (FC) Erysipelotrichaceae_Otu010 Bifidobacterium_Otu025 Bacteroides_Otu004 Porphyromonadaceae_Otu002 Bacteroidales_Otu008 Lactobacillus_Otu013 Akkermansia_Otu026 p<0.001 Firmicutes_Otu007 Bacteria_unclassified_Otu009 Betaproteobacteria_Otu064 Barnesiella_Otu011 Alphaproteobacteria_Otu075 Clostridium_XIVa_Otu021 Turicibacter_Otu023 Lactobacillaceae_Otu029 Allobaculum_Otu017 Olsenella_Otu060 Blautia_Otu012 Parasutterella_Otu128

Figure 6 High cholesterol-­modulated microbiota promotes hepatocyte proliferation in germ-free­ mouse model. (A) Stools were transplanted from NC-­fed, HFLC-fed­ and HFHC-­fed mice (14 months) to germ-free­ mice (G-­NC, G-­HFLC and G-­HFHC) under NC. Gross morphology, histological examination and Ki-67 staining of livers in stool-gav­ aged germ-free­ mice in the G-NC,­ G-HFLC­ or GHFHC groups were shown. (B) Liver triglyceride content, lipid peroxidation and liver histology in recipient germ-free­ mice of the G-NC,­ G-HFLC­ or G-HFHC­ groups at 8, 10 and 14 months. (C) Hepatic IL-6 protein levels and (D) Ki-67 staining of liver sections in the G-NC,­ G-HFLC­ or G-HFHC­ groups at 14 months. (E1) Mouse cancer pathway Finder PCR array in the liver tissues of G-­NC, G-HFLC­ or G-HFHC­ mice. (E2) CDC20 protein level was validated. (F) Principal component ordination analysis and histogram of the linear discriminant analysis (LDA) scores of gut microbiota among G-NC,­ G-­HFLC or G-­HFHC mice. (G) Serum metabolomic analysis of G-­HFLC and G-­HFHC mice. *p<0.05, **p<0.01, ***p<0.001. HFHC, high-­fat/high-­cholesterol diet; HFLC, high-fat/low-­ cholesterol­ diet; NC, normal chow; PCA, principal component analysis.

10 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 Hepatology

Atorvastatin Sacrifice A B HFHC HFHC C57/BL6 20mg/kg Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from 8 wk old AN HFHC diet 0 7 m 14 m T 500 HFHC C HFHC+At 5 600 ** * HFHC+At HFHC+At 400 4 400 300 3 200 (mg/dl) * 2 200 100 liver) (mg/g 1

Serum cholesterol cholesterol Serum 0 0 0

Hepatic Hepatic cholesterol free 7 8 10 12 14 HFHC HFHC+At (ng/mL) AFP Serum HFHC HFHC+At Time (months) D HFHC Histological scores of liver sections HFHC+At 80 * 280 600 * 60 260 HFHCHFHC+At 400 240 Steatosis 1.71±0.82 1.45±0.51 40 220 Inflammation 2.19±0.62 0.78±0.52*** 200 20 * pro) (U/mg 200

SOD activity SOD 180

Liver NAD+/NADH Liver

Serum ALT (U/L) ALT Serum 0 0 7 8 10 12 14 HFHC HFHC+At HFHC HFHC+At Time (Months) 150 800 250 * 100 100 2000 ** * * * 200 * 600 80 80 /mL) pg 100

/mL) pg (

1500 /mL) pg (

/mL) pg ( 60 /mL) pg ( 60 150

β α 1000 400 40 40 100 50 500 200 20 20 50 0 0 0 0 0 0

Serum IL-6 Serum /mL) pg (

Serum IL-1 Serum

Serum IL-1 Serum

1 ( 1 MCP- Serum

Serum MIP- 1β Serum HFHC HFHC+AtHFHC HFHC+AtHFHC HFHC+At HFHC HFHC+At MIP- 1α Serum HFHC HFHC+At HFHC HFHC+At

25000 E HFHC HFHC+At 0.6 20000

area 15000 * 0.4 **

/field)

2 10000 g/mg) m 0.2

μ (

μ ( Red staining Red 5000

Collagen

Hydroxyproline 0 0

Sirius HFHC HFHC+At HFHC HFHC+At

NC HFLC HFHC HFHC+At F AF12_Otu046 Desulfovibrio_C21_c20_Otu042 * ** HFLC Desulfovibrio_Otu047 HFHC Dehalobacterium_Otu066 HFHC + At Mogibacteriaceae_Otu069 Streptococcaceae_Otu088 Butyricimonas_Otu098 Desulfovibrionaceae_Otu073 Anaerotruncus_Otu107

Chao1 Paraprevotella_Otu083 Helicobacteracae_Otu072 Mucispirillum_schaedleri_Otu038

Lactococcus_Otu033 http://gut.bmj.com/ Bilophila_Otu037 Ruminococcaceae_Otu011 Dorea_Otu050 Rikenellaceae_Otu015 Helicobacter_ganmani_Otu031

40 25 G G-HFHC G-HFHC G-HFHCAt G-HFHCAt 30 20 15 20 *

/mg protein) /mg TBARS TBARS 10

g/mg protein) g/mg Triglyceride Triglyceride 10

μ

( 5

(nmol *

0 0 on September 24, 2021 by guest. Protected by copyright. G-HFHC G-HFHCAt G-HFHC G-HFHCAt

H Intestine Liver

Atorvastatin Cholesterol Cholesterol ROS IL-6

CDC20 Dysbiosis Oxidative stress NAFLD- Cell proliferation IPA HCC Lipid accumulation Tryptophan metabolism- Bacteroides related bacteria Bifidobacterium TCA

IPA TCA

IPA TCA

Figure 7 Cholesterol inhibition abrogated NAFLD–HCC progression in HFHC-fed­ mice. (A) Schematic illustration of the atorvastatin treatment of C57BL/6 mice fed with HFHC diet. (B) Representative gross morphology of liver and microscopic features in H&E-stained­ HFHC-fed­ mice with or without the treatment of atorvastatin, histological scores of steatosis, inflammation and collagen were calculated. (C) Serum cholesterol levels, hepatic free cholesterol levels, serum AFP levels, (D) serum ALT levels, liver NAD+ to NADH ratio, liver SOD activity, serum IL-6, IL-1α, IL-1β, MCP-1, MIP-1α, MCP-1β protein levels, (E) collagen deposition and hydroxyproline content in HFHC-fed­ mice with or without the treatment of atorvastatin. (F) Bacterial richness and heatmap plot of the bacteria in stool of mice fed with NC, HFLC and HFHC with or without atorvastatin treatment. (G) Gross morphology, histological examination, triglyceride and lipid peroxidation of livers in HFHCAt stool gavaged germ-free­ mice (G-­HFHCAt). (H) Schematic diagram of the mechanism of cholesterol-induced­ NAFLD–HCC development. *p<0.05, **p<0.01, ***p<0.001. AFP, alpha-fetoprotein;­ At, atorvastatin, ALT, alanine aminotransferase; HCC, hepatocellular carcinoma; HFHC, high-fat/high-­ ­cholesterol diet; IPA, 3-indolepropionic­ acid; NAD, nicotinamide adenine dinucleotide; NAFLD, non-­alcoholic fatty liver disease; ROS, reactive oxygen species; SOD, superoxide dismutase; TCA, taurocholic acid.

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 11 Hepatology cholesterol induced both ROS and proinflammatory cytokines, bile acid metabolism was impaired with high-cholesterol­ diet. thereby promoting NASH and HCC development. In line with Primary bile acids including TCA, GCA, TCDCA and TUDCA our findings, Wolf et al have shown that long-­term choline-­ were enriched in HFHC-­fed mice, consistent with high plasma Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from deficient high-­fat diet feeding in mice-­induced hepatic ROS, TCA, GCA and TCDCA previously reported in human NASH intrahepatic CD8+ T cells, NKT cells and their secreted inflam- patients compared with healthy subjects.36 Disruption of bile acid matory cytokines to promote NASH and HCC transition.20 homeostasis alters metabolic homeostasis in the liver and can Accumulating evidence demonstrates that the human gut lead to hepatic inflammation and metabolic diseases including microbiota can influence various pathologic conditions including diabetes and NAFLD.37 38 TCA, GCA, TCDCA and TUDCA are liver cancer.21 We therefore explored the potential involvement key signalling molecules linking the gut and the liver to impact of the gut microbiota in mediating dietary cholesterol-induced­ hepatic lipid accumulation and inflammation.39 Bifidobacte- NAFLD–HCC. Along with sequentially increased bacterial rium and Bacteroides are the main bacterial genera of the gut richness, M. schaedleri_Otu038, Desulfovibrio_Otu047, Anaer- microbiota involved in bile acid metabolism.40 They can decon- otruncus_Otu107 and Desulfovibrionaceae_Otu073 increased jugate taurine-conjugated­ and glycine-conjugated­ bile acids to along NAFLD–HCC progression. This is consistent with the their unconjugated free forms through bile acid hydrolase and report that M. schaedleri and Desulfovibrionaceae are strongly convert the unconjugated primary bile acids into secondary bile correlated with obesity, metabolic syndrome and inflammation in acids.40 Therefore, the decreased Bifidobacterium and Bacte- mouse models.22–24 Desulfovibrionaceae and Desulfovibrio were roides in the stool of HFHC-­fed mice may account for the accu- described to be abundant in a swine NASH model.25 Enriched mulated taurine-­conjugated bile acids in HFHC-­fed mice. TCA Anaerotruncus was also reported in MCD diet-induced­ exper- aggravated cholesterol-­induced triglyceride accumulation in imental NASH mouse model.26 In addition, we observed that vitro in this study which further supports its contribution and several Clostridium OTUs were enriched with high-­cholesterol potentially, those of other enriched bile acids to the pathogen- diet and positively associated with high serum cholesterol. esis of cholesterol-induced­ NAFLD–HCC. Additionally, IPA, Previous reports have shown increased prevalence of Clostridia which is exclusively produced by commensal gut microbes,41 42 in NASH patients,27 supporting our observation in this study. was depleted in both HFHC-­fed conventional mice and germ-­ Moreover, the depletion of OTUs of Akkermansia, Lactobacillus, free mice transplanted with stools from HFHC-­fed mice. IPA Bifidobacterium and Bacteroides were demonstrated in NAFLD– is a specific metabolite whose production completely depends HCC mice. These OTUs including A. municiphila_Otu034, on tryptophan metabolism by the gut microbiota18 and charac- Lactobacillus_Otu009, Bifidobacterium_Otu026, B. uniformis_ terised as being anti-inflammatory­ and protective of intestinal Otu080 and B. acidifaciens_Otu032 consistently decreased barrier integrity.42 We identified significant reduction in micro- with high serum cholesterol. Bifidobacterium and Bacteroides bial tryptophan metabolism. Our in-vitro­ functional analysis were further shown to be negatively associated with total serum showed that IPA could inhibit cell proliferation and suppress cholesterol in human hypercholesteremia patients in this study. lipid accumulation in NASH–HCC cell lines. Collectively, our Depletion of Bifidobacterium and Bacteroides has been previ- findings suggest that cholesterol could impair bile acid metabo- 27 28 ously demonstrated in human NASH patients. There are lism and microbial tryptophan metabolism, leading to enhanced http://gut.bmj.com/ evidences that Bifidobacterium has cholesterol-lowering­ activity serum TCA, and reduced IPA, thereby promoting NAFLD–HCC through gut microbiota modulation. Additionally, B. acidifaciens development (figure 7H). is a potential probiotic bacteria as it possess the ability to prevent Gut microbiota transfer through faecal transplantation have obesity and improve insulin sensitivity in mice.29 A strain of B. been used to demonstrate active roles of gut microbiota in uniformis, namely B. uniformis CECT 7771 reportedly amelio- metabolic diseases including obesity and NAFLD.27 43 There- rated metabolic and immunological dysfunction induced by fore, we sought to determine whether the gut microbiota can 30 high-­fat diet in mice. Strains of Lactobacillus has also been play a direct causative role in dietary cholesterol-induced­ liver on September 24, 2021 by guest. Protected copyright. reported to be protective against NASH31 and beneficially influ- steatohepatitis and tumourigenesis. Compared with control ence liver cholesterol metabolism in a pattern similar to that groups, cholesterol-­modulated microbiota promoted steatohep- resulting from statin treatment in murine model.32 Moreover, atitis and hepatocyte proliferation/dysplasia in germ-free­ mice. decrease in A. muciniphila reportedly promotes NAFLD through Proinflammatory factor, and decreased IPA were associated with thinning of the intestinal mucus layer, thereby impairing gut microbiota-induced­ liver steatohepatitis and tumourigenesis in permeability barrier.33 These results support our observations in germ-free­ mice, in consistence with observations in conventional this study and further indicate that the suppression of protective mouse model. In addition to hepatic lipid accumulation, inflam- gut microbes might represent a means through which choles- mation, cell proliferation and dysplasia, HCC was not observed terol exerts its procarcinogenic effect in the liver. Our results in germ-free­ mice transplanted with faecal microbiota from provide new insights into the potential roles of gut dysbiosis in high-cholesterol­ diet-fed­ mice for 14 months. This suggested cholesterol-­induced NAFLD–HCC progression. that dietary cholesterol plays an important role in inducing In addition to gut microbiota dysbiosis, small molecule metab- NAFLD–HCC formation and that the change of gut microbiota olites produced by commensal bacteria could contribute to the by dietary cholesterol contributes to the disease progression. pathogenesis of NAFLD.13 Gut microbiota-related­ metabolites, If cholesterol plays a key role in the pathogenesis of NASH including short-­chain fatty acids, amino acid catabolites and and HCC, it would be important to establish that its functional bile acid, can both agonise and antagonise their cognate recep- blockade ameliorates the severity of steatohepatitis and NAFLD– tors to reduce or exacerbate liver steatosis and inflammation.34 HCC development. To test this, we used atorvastatin, a 3-hydro­ Recently, a new gut flora generated metabolite, N,N,N-trimethyl-­ xy-3-methyl-­glutaryl-coenzyme­ A (HMG-CoA)­ reductase inhib- 5-aminovaleric­ acid (TMAVA), is found to be elevated in plasma itor widely used to treat hypercholesterolemia, to inhibit choles- of human liver steatosis and exacerbates fatty liver progression terol biosynthesis in HFHC-fed­ mice. We found that no HCC in mice.35 We therefore attempted to unravel metabolites that was identified in atorvastatin-­treated mice and the severities of potentially mediate the NAFLD–HCC promoting effect of steatohepatitis and fibrosis were largely blunted. Additionally, high dietary cholesterol. Metabolomics profiling revealed that bacterial richness, Mucispirillum, Desulfovibrio, Anaerotruncus

12 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 Hepatology and Desulfovibrionaceae which increased along cholesterol-­ REFERENCES induced NAFLD–HCC progression were reversed with ator- 1 Romeo S. Notch and nonalcoholic fatty liver and fibrosis. N Engl J Med vastatin treatment. Restoration of the gut microbiota diversity 2019;380:681–3. Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from 2 Kanwal F, Kramer JR, Mapakshi S, et al. Risk of hepatocellular cancer in patients with could be one of the mechanisms through which atorvastatin non-alcoholic­ fatty liver disease. Gastroenterology 2018;155:1828–37. elicits its protective effect against NAFLD–HCC as shown in 3 Wong RJ, Cheung R, Ahmed A. Nonalcoholic steatohepatitis is the most rapidly this study. Moreover, unlike the mice gavaged with stools from growing indication for liver transplantation in patients with hepatocellular carcinoma mice fed with HFHC only, the stools from atorvastatin-­treated in the U.S. Hepatology 2014;59:2188–95. 4 Ioannou GN. The role of cholesterol in the pathogenesis of NASH. Trends Endocrinol HFHC-­fed mice did not promote hepatic cell proliferation in Metab 2016;27:84–95. recipient germ-free­ mice. These observations further highlight 5 Van Rooyen DM, Larter CZ, Haigh WG, et al. Hepatic free cholesterol accumulates the potential use of atorvastatin in preventing the development in obese, diabetic mice and causes nonalcoholic steatohepatitis. Gastroenterology of cholesterol-­induced NAFLD–HCC. 2011;141:1393–403. 6 Musso G, Gambino R, De Michieli F, et al. Dietary habits and their relations to insulin In conclusion, this study shows for the first time, that resistance and postprandial lipemia in nonalcoholic steatohepatitis. Hepatology prolonged high dietary cholesterol induces spontaneous and 2003;37:909–16. progressive development of NAFLD–HCC in male mice by 7 Liu D, Wong CC, Fu L, et al. Squalene epoxidase drives NAFLD-­induced hepatocellular modulating the gut microbiota. Cholesterol induces increased carcinoma and is a pharmaceutical target. Sci Transl Med 2018;10:eaap9840. TCA and decreased IPA through gut microbiota alteration, 8 McGettigan B, McMahan R, Orlicky D, et al. Dietary lipids differentially shape nonalcoholic steatohepatitis progression and the transcriptome of Kupffer cells and thereby promoting lipid accumulation, cell proliferation in the infiltrating macrophages. Hepatology 2019;70:67–83. liver, leading to NAFLD–HCC development (figure 7H). Anti- 9 Canfora EE, Meex RCR, Venema K, et al. Gut microbial metabolites in obesity, NAFLD cholesterol treatment completely abrogated dietary cholesterol-­ and T2DM. Nat Rev Endocrinol 2019;15:261–73. induced NAFLD–HCC formation. This study highlights that 10 Aron-­Wisnewsky J, Vigliotti C, Witjes J, et al. Gut microbiota and human NAFLD: disentangling microbial signatures from metabolic disorders. Nat Rev Gastroenterol cholesterol inhibition and manipulation of the gut microbiota Hepatol 2020;17:279–97. and its related metabolites might represent effective strategies in 11 Loo TM, Kamachi F, Watanabe Y, et al. Gut Microbiota Promotes Obesity-­Associated preventing NAFLD–HCC. Liver Cancer through PGE 2 -Mediated Suppression of Antitumor Immunity. Cancer Discov 2017;7:522–38. 12 Sayin SI, Wahlström A, Felin J, et al. Gut microbiota regulates bile acid metabolism Author affiliations by reducing the levels of tauro-­beta-­muricholic acid, a naturally occurring FXR 1State Key Laboratory of Digestive Disease, Institute of Digestive Disease and The antagonist. Cell Metab 2013;17:225–35. Department of Medicine and Therapeutics, Li Ka Shing Institute of Health Sciences, 13 Chu H, Duan Y, Yang L, et al. Small metabolites, possible big changes: a microbiota-­ CUHK Shenzhen Research Institute, The Chinese University of Hong Kong, Hong centered view of non-­alcoholic fatty liver disease. Gut 2019;68:359–70. Kong SAR, China 14 Caussy C, Hsu C, Lo M-­T, et al. Link between gut-­microbiome derived metabolite 2Department of Imaging and Interventional Radiology, The Chinese University of and shared gene-­effects with hepatic steatosis and fibrosis in NAFLD. Hepatology Hong Kong, Hong Kong SAR, China 3 2018;68:918–32. Department of Anatomical and Cellular Pathology, The Chinese University of Hong 15 Bo T, Shao S, Wu D, et al. Relative variations of gut microbiota in disordered Kong, Hong Kong SAR, China 4 cholesterol metabolism caused by high-­cholesterol diet and host genetics. Department of Precision Medicine, Sun Yat-­Sen University First Affiliated Hospital, Microbiologyopen 2017;6:e00491.

Guangzhou, Guangdong, China http://gut.bmj.com/ 5 16 Zhang X, Shen J, Man K, et al. CXCL10 plays a key role as an inflammatory Department of Laboratory Animal Science, College of Basic Medical Sciences, Third mediator and a non-­invasive biomarker of non-­alcoholic steatohepatitis. J Hepatol Military Medical University, Chongqing, China 6 2014;61:1365–75. Department of Comparative Medicine and Department of Cellular and Molecular 17 Liang JQ, Teoh N, Xu L, et al. Dietary cholesterol promotes steatohepatitis related Physiology, Yale University School of Medicine, New Haven, Connecticut, USA hepatocellular carcinoma through dysregulated metabolism and calcium signaling. Nat Commun 2018;9:4490. Contributors XZ and OOC were involved in study design, conducted the 18 Roager HM, Licht TR. Microbial tryptophan catabolites in health and disease. Nat experiments and drafted the paper; EC, KF, HCHL and Y-­XW performed the Commun 2018;9:3294. experiments; AWHC determined the histology; HW provided germ-­free animal; XY 19 Jaeschke H. Reactive oxygen and mechanisms of inflammatory liver injury: present on September 24, 2021 by guest. Protected copyright. and JJYS commented and revised the manuscript; JY designed, supervised the study concepts. J Gastroenterol Hepatol 2011;26:173–9. and wrote the paper. 20 Wolf MJ, Adili A, Piotrowitz K, et al. Metabolic activation of intrahepatic CD8+ T cells Funding This project was supported by research funds from Guangdong Natural and NKT cells causes nonalcoholic steatohepatitis and liver cancer via cross-talk­ with Science Foundation (2018B030312009), RGC Collaborative Research Fund (C4041-­ hepatocytes. Cancer Cell 2014;26:549–64. 17GF, C7026-­18G, C7065-18G),­ RGC Theme-­based Research Scheme Hong Kong 21 Ma C, Han M, Heinrich B, et al. Gut microbiome–mediated bile acid metabolism (T12-703/19 R), CUHK direct grant for research, Vice-­Chancellor’s Discretionary Fund regulates liver cancer via NKT cells. Science 2018;360:eaan5931. CUHK. 22 Ussar S, Griffin NW, Bezy O, et al. Interactions between gut microbiota, host genetics and diet modulate the predisposition to obesity and metabolic syndrome. Cell Metab Competing interests None declared. 2015;22:516–30. Patient consent for publication Not required. 23 Loy A, Pfann C, Steinberger M, et al. Lifestyle and Horizontal Gene Transfer-­Mediated Evolution of Mucispirillum schaedleri, a Core Member of the Murine Gut Microbiota. Ethics approval All animal studies were performed in accordance with guidelines mSystems 2017;2. doi:10.1128/mSystems.00171-16 approved by the Animal Experimentation Ethics Committee of the Chinese University 24 Zhang C, Zhang M, Pang X, et al. Structural resilience of the gut microbiota in adult of Hong Kong and the Third Military Medical University. mice under high-­fat dietary perturbations. Isme J 2012;6:1848–57. Provenance and peer review Not commissioned; externally peer reviewed. 25 Panasevich MR, Meers GM, Linden MA, et al. High-­Fat, high-­fructose, high-­cholesterol feeding causes severe NASH and cecal microbiota dysbiosis in juvenile Ossabaw Data availability statement All data relevant to the study are included in the swine. Am J Physiol Endocrinol Metab 2018;314:E78–92. article or uploaded as supplementary information. 26 Ye J-­Z, Li Y-T,­ Wu W-­R, et al. Dynamic alterations in the gut microbiota and Open access This is an open access article distributed in accordance with the metabolome during the development of methionine-­choline-­deficient diet-­induced Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which nonalcoholic steatohepatitis. World J Gastroenterol 2018;24:2468–81. permits others to distribute, remix, adapt, build upon this work non-­commercially, 27 Zhu L, Baker SS, Gill C, et al. Characterization of gut microbiomes in nonalcoholic and license their derivative works on different terms, provided the original work is steatohepatitis (NASH) patients: a connection between endogenous alcohol and properly cited, appropriate credit is given, any changes made indicated, and the use NASH. Hepatology 2013;57:601–9. is non-­commercial. See: http://​creativecommons.org/​ ​licenses/by-​ ​nc/4.​ ​0/. 28 Vernon G, Baranova A, Younossi ZM. Systematic review: the epidemiology and natural history of non-­alcoholic fatty liver disease and non-­alcoholic steatohepatitis in adults. ORCID iDs Aliment Pharmacol Ther 2011;34:274–85. Harry C H Lau http://orcid.​ ​org/0000-​ ​0003-3581-​ ​2909 29 Yang J-­Y, Lee Y-S,­ Kim Y, et al. Gut commensal Bacteroides acidifaciens prevents Jun Yu http://orcid.​ ​org/0000-​ ​0001-5008-​ ​2153 obesity and improves insulin sensitivity in mice. Mucosal Immunol 2017;10:104–16.

Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664 13 Hepatology

30 Gauffin Cano ,P Santacruz A, Moya Ángela, et al. Bacteroides uniformis CECT 7771 36 Puri P, Daita K, Joyce A, et al. The presence and severity of nonalcoholic ameliorates metabolic and immunological dysfunction in mice with high-­fat-­diet steatohepatitis is associated with specific changes in circulating bile acids. Hepatology

induced obesity. PLoS One 2012;7:e41079. 2018;67:534–48. Gut: first published as 10.1136/gutjnl-2019-319664 on 21 July 2020. Downloaded from 31 Okubo H, Sakoda H, Kushiyama A, et al. Lactobacillus casei strain Shirota protects 37 Horie Y, Suzuki A, Kataoka E, et al. Hepatocyte-­Specific PTEN deficiency results in against nonalcoholic steatohepatitis development in a rodent model. Am J Physiol steatohepatitis and hepatocellular carcinomas. J Clin Invest 2004;113:1774–83. Gastrointest Liver Physiol 2013;305:G911–8. 32 Naudin CR, Maner-­Smith K, Owens JA, et al. Lactococcus lactis subspecies cremoris 38 Chiang JYL, Ferrell JM. Bile acid metabolism in liver pathobiology. Gene Expr elicits protection against metabolic changes induced by a western-­style diet. 2018;18:71–87. Gastroenterology 2020. doi:10.1053/j.gastro.2020.03.010. [Epub ahead of print: 10 39 Perez MJ, Briz O. Bile-acid-­ ­induced cell injury and protection. WJG 2009;15:1677–89. Mar 2020]. 40 Jia W, Xie G, Jia W. Bile acid–microbiota crosstalk in gastrointestinal inflammation and 33 Miura K, Ohnishi H. Role of gut microbiota and Toll-lik­ e receptors in nonalcoholic fatty carcinogenesis. Nat Rev Gastroenterol Hepatol 2018;15:111–28. liver disease. WJG 2014;20:7381–91. 41 Dodd D, Spitzer MH, Van Treuren W, et al. A gut bacterial pathway metabolizes 34 Ding Y, Yanagi K, Cheng C, et al. Interactions between gut microbiota and non-­ aromatic amino acids into nine circulating metabolites. Nature 2017;551:648–52. alcoholic liver disease: the role of microbiota-­derived metabolites. Pharmacol Res 42 Venkatesh M, Mukherjee S, Wang H, et al. Symbiotic bacterial metabolites regulate 2019;141:521–9. 35 Zhao M, Zhao L, Xiong X, et al. TMAVA, a metabolite of intestinal microbes, is gastrointestinal barrier function via the xenobiotic sensor PXR and Toll-­like receptor 4. increased in plasma from patients with liver steatosis, inhibits γ-Butyrobetaine Immunity 2014;41:296–310. hydroxylase, and exacerbates fatty liver in mice. Gastroenterology 43 Boursier J, Diehl AM. Implication of gut microbiota in nonalcoholic fatty liver disease. 2020;158:2266–81. PLoS Pathog 2015;11:e1004559. http://gut.bmj.com/ on September 24, 2021 by guest. Protected copyright.

14 Zhang X, et al. Gut 2020;0:1–14. doi:10.1136/gutjnl-2019-319664