Interaction between dietary lipids and gut microbiota regulates hepatic cholesterol metabolism1[S]

The gut microbiota influences many aspects of host metabolism. We have previously shown that the presence of a gut microbiota remodels lipid composition. Here we investigated how interaction between gut microbiota and dietary lipids regulates lipid composition in the liver and plasma, and gene expression in the liver. Germ-free and conventionally raised mice were fed a lard or fish oil diet for 11 weeks. We performed lipidomics analysis of the liver and serum and microarray analysis of the liver. As expected, most of the variation in the lipidomics dataset was induced by the diet, and abundance of most lipid classes differed between mice fed lard and fish oil. However, the gut microbiota also affected lipid composition. The gut microbiota increased hepatic levels of cholesterol and cholesteryl esters in mice fed lard, but not in mice fed fish oil. Serum levels of cholesterol and cholesteryl esters were not affected by the gut microbiota. Genes encoding enzymes involved in cholesterol biosynthesis were downregulated by the gut microbiota in mice fed lard and were expressed at a low level in mice fed fish oil independent of microbial status. In summary, we show that gut microbiota-induced regulation of hepatic cholesterol metabolism is dependent on dietary lipid composition.

using the Folch method ( 18 ) and quantifi ed using straight-phase HPLC coupled to evaporative light scattering detection ( 19 ).

Microarray expression analysis
RNA was isolated from the left hepatic lobe using RNeasy lipid tissue mini kit (Qiagen, Hilden, Germany). cDNA synthesis, labeling, hybridization to GeneChip® ST arrays (GeneChip® Mouse Gene 1.0 ST array), washing, and scanning were performed as previously described ( 7 ). The raw data were normalized using the robust multi-array average method ( 20 ). Differential expression of microarray data was evaluated by Student's t -test followed by correction for false discovery rate (FDR). Analysis of interactions, using colonization status and diet as independent variables, was performed by two-way analysis of interaction followed by correction for FDR. Analysis of enrichment of regulated genes within functional categories [gene ontology (GO) categories] ( 21 ) was performed using the software DAVID (http://david.abcc.ncifcrf. gov/) ( 22 ). The results of the enrichment calculations were fi ltered for GO categories that were signifi cantly enriched ( P < 0.05) after Bonferroni correction. Redundancy within lists of GO terms was reduced by the Revigo software (http://revigo.irb.hr/) ( 23 ) with a similarity score set to 0.5. Principle component analysis (PCA) was performed in MultiExperiment Viewer. Microarray data have been uploaded to the Gene Expression Omnibus (GEO) database with accession number GSE73195. 2

Statistical analysis
Data are represented as mean ± SEM. Statistical comparison of two groups was performed by Student's t -test. Analysis of datasets with 2 × 2 factorial design (comparing bacterial status and diet) was performed by two-way ANOVA with Tukey's multiple comparison test. Statistical analysis was performed in GraphPad Prism 6 unless otherwise stated.

Interaction between dietary lipids and gut microbiota regulates the levels of cholesteryl esters in the liver
We have previously shown that the gut microbiota affects liver and serum lipid composition in mice fed a chow diet ( 24 ) and that interaction between gut microbiota and dietary lipids infl uences adiposity, adipose tissue infl ammation, and systemic glucose metabolism ( 7 ). Here we performed lipidomics analysis of liver and serum from conventionally raised (CONV-R) and GF mice fed a lard or fi sh oil diet for 11 weeks to study how interaction between Here, we aimed to determine how the gut microbiota affects hepatic lipid metabolism during the metabolic challenge of a high-fat diet and how interaction between dietary lipids and the gut microbiota infl uences lipid composition and regulation of metabolic pathways.

Mice
C57Bl/6 mice were maintained under standard specifi c-pathogen free or GF conditions, as described previously ( 10 ). All mice were males, age matched, and 12-14 weeks of age. Mice were fed irradiated isocaloric diets (45% kcal fat) of identical composition except for the source of fat, which was either menhaden fi sh oil (Research Diets, D05122102) or lard (Research Diets, D10011202) for 11 weeks ( 7 ). Fatty acid composition of the diets is displayed in supplementary Table 1. The mice were fasted for 4 h before they were euthanized. Blood samples and liver samples were harvested at the end of the experiment. All experiments were performed with protocols approved by the University of Gothenburg Animal Studies Committee.

Lipidomics analysis
Lipids were analyzed by ultra-performance LC coupled with TOFMS using a Waters Q-TOF Premier (Waters, Milford, MA) with a methodology described earlier ( 15 ). Briefl y, the samples were extracted with a chloroform:methanol mixture after addition of internal standards containing lysophosphatidylcholine was added. The data were processed using MZmine 2 ( 16 ), which included alignment of peaks, peak integration, isotopic grouping, normalization, and peak identifi cation. Lipids were identifi ed using an in-house lipid species database and based on their MS/MS spectra. The data were normalized using the internal standards representative of each class of lipid present in the samples, as previously described ( 15 ). Sphingomyelins were normalized to the phosphatidylcholine standard.
The serum cholesterol concentrations were obtained by a metabolomics method based on 2D GC-MS. Briefl y, each serum sample (30 l) was spiked with 10 l of internal standard solution (valine-d8, c = 37 mg/l; heptadecanoic acid, c = 187 mg/l; and succinic-d4 acid, c = 63 mg/l), extracted with 400 l of methanol, centrifuged, and evaporated to dryness. Cholesterol was converted into its methoxime and TMS derivative(s) by two-step derivatization using methoxime and n-methyl-n-trimethylsilyl trifl uoroacetamide reagents, and a retention index standard mixture (n-alkanes) in hexane was added . The analysis was performed as previously described ( 17 ) using a Leco Pegasus 4D GC×GC-TOFMS instrument (Leco Corp., St. Joseph, MI) equipped with a cryogenic modulator and with an Agilent 6890 gas chromatograph (Agilent Technologies, Palo Alto, CA) as the GC part of the instrument. The raw data processing was done using ChromaTOF version 4.22 (Leco Corp.) software. Finally, alignment and normalization of the data were done using an in-house developed program, Guineu ( 17 ), and concentrations of cholesterol were calculated using a calibration curve. Hepatic free cholesterol was extracted status in the second dimension ( Fig. 1A ), which represented approximately 1% of the total variation in the dataset. Serum samples did not separate on microbial status in either of the dietary groups ( Fig. 1B ).
Dietary lipids affected the abundance of the majority of lipid classes both in liver and in serum ( Fig. 1C, D ). However, the gut microbiota only induced increased hepatic levels of cholesterol and cholesteryl esters ( Fig. 1C ) and decreased serum levels of ether phosphatidylcholines and ether phosphatidylethanolamines in mice fed lard, but not in mice fed fi sh oil ( Fig. 1D ).
Despite the relatively small contribution by the microbiota to the overall variation in the lipidomics dataset, many individual lipid species in the liver and in serum were regulated by the gut microbiota (supplementary Tables 1, 2). To investigate whether the effect of the gut microbiota on individual lipid species in the liver and serum was similar, we plotted regulation induced by the gut microbiota in the the gut microbiota and dietary lipids affects hepatic and serum lipid composition.
GF mice gained approximately 30% less weight than CONV-R mice (data not shown). To investigate the infl uence of gut microbiota on lipid metabolism independent of differences in body weight, we used GF mice that were slightly heavier than their CONV-R counterparts at the beginning of the feeding period. At the end of the experiment the CONV-R and GF mice had matching body weights (supplementary Fig. 1A). Diet consumption was similar between CONV-R and GF mice, while mice fed lard consumed more food than mice fed fi sh oil (supplementary Fig. 1B). PCA of lipidomics data showed that most of the variation in the dataset was induced by diet. Samples separated by diet in the fi rst dimension, which accounted for 96% of the variation among liver samples ( Fig. 1A ) and for 91% of the variation among serum samples ( Fig. 1B ). Liver samples from mice fed lard separated on microbial indicating that many genes are regulated by the gut microbiota independently of dietary lipids.
GO analysis showed that the microbiota induced hepatic expression of genes involved in the immune response in mice fed fi sh oil ( Table 1 ). Most of these genes also showed a tendency toward being upregulated in mice fed lard, but did not reach signifi cance in these mice (data not shown). Gene expressions of the macrophage markers Cd68 , Emr1 (encoding F4/80), and Clec4f were upregulated by the gut microbiota on both diets (supplementary Fig. 2A-C) indicating that the microbiota induces infi ltration of macrophages to the liver. However, the expression of genes encoding interleukins was not affected by either dietary lipids or microbial status (data not shown). Genes in functional categories associated with cholesterol biosynthesis were downregulated by the gut microbiota in mice fed lard ( Table 1 , Fig. 3C ). Two-way ANOVA also showed that genes regulated by the interaction between dietary lipids and the gut microbiota were highly enriched in processes associated with cholesterol biosynthesis ( Table 1 , Fig. 3C ).
We found that the gut microbiota reduced the hepatic expression of many genes encoding enzymes in: 1 ) the mevalonate pathway, including the rate-limiting enzyme HMG-CoA reductase encoded by Hmgcr ( Fig. 4A, B ); liver against regulation induced by the gut microbiota in serum and observed a positive linear relationship in both dietary groups ( Fig. 2A, B ). An exception to this trend was observed for cholesteryl ester species, which were upregulated by the gut microbiota in the liver, but not in serum (except for a small increase in 18:1) of mice fed lard ( Fig.  2A, C, D ), but were not infl uenced by the gut microbiota in either the liver or serum of mice fed fi sh oil ( Fig. 2B-D ).

Interaction between dietary lipids and gut microbiota regulates hepatic cholesterol biosynthesis
To study how the gut microbiota and dietary lipids regulate liver lipid metabolism, we performed microarray analysis of liver samples from CONV-R and GF mice fed lard or fi sh oil for 11 weeks.
PCA of hepatic gene expression data revealed that mice separated on diet in the fi rst dimension and on microbial status in the second dimension ( Fig. 3A ). To compare how the gut microbiota affects hepatic gene expression in mice fed lard or fi sh oil, we plotted differences in hepatic gene expression level that were signifi cant between CONV-R and GF mice, with values for mice fed fi sh oil on the y axis and values for mice fed lard on the x axis ( Fig. 3B ). We found a linear relationship between microbiota-induced gene regulation in liver from mice fed lard or fi sh oil ( Fig. 3B ),   Fig. 2. Cholesteryl esters in the liver are increased by the gut microbiota in mice fed lard for 11 weeks. Microbial regulation of liver and serum lipid species in mice fed lard (A) and fi sh oil (B). Fold changes for lipid species signifi cantly regulated by the microbiota in serum and liver are plotted on the y axis and x axis, respectively. ChoE, cholesteryl ester. Abundance of cholesteryl ester species in the liver (C) and in serum (D) from CONV-R and GF mice fed lard or fi sh oil. Liver samples: n = 11 (CONV-R lard); 10 (GF lard); 11 (CONVR fi sh oil); 18 (GF fi sh oil). Serum samples: n = 11 (CONV-R lard); 10 (GF lard); 10 (CONVR fi sh oil); 15 (GF fi sh oil). Mean values ± SEM are plotted. Variation induced by the diet: † † P < 0.01, † † † P < 0.001. Variation induced by the gut microbiota: § P < 0.05, § § P < 0.01, § § § P < 0.001. Post hoc multiple comparison analysis: * P < 0.05, *** P < 0.001. the gut microbiota both in mice fed lard and in mice fed fi sh oil ( Fig. 4G ).
Taken together, we found that genes involved with cholesterol biosynthesis were downregulated: 1 ) by the gut microbiota in mice fed lard; and 2 ) by fi sh oil independent of microbial status.

DISCUSSION
In the present study, we demonstrate that the interaction between gut microbiota and dietary lipids regulates cholesterol metabolism in the liver. The gut microbiota increased hepatic levels of cholesterol and cholesteryl esters and decreased the expression of genes involved in cholesterol biosynthesis in mice fed lard, but not in mice fed fi sh oil. and 2 ) the steroid biosynthesis pathway ( Fig. 4A, C ) in mice fed lard. The expression levels of these genes were low in mice fed fi sh oil independent of microbial status ( Fig. 4B, C ).
To study the mechanisms underlying the regulation of cholesterol biosynthesis by gut microbiota and dietary lipids, we analyzed gene expression and protein cleavage of sterol regulatory element-binding protein 2 (SREBP2), a key regulator of cholesterol homeostasis. Gene expression of Srebf2 (encoding SREBP2) was increased in mice fed lard compared with mice fed fi sh oil ( Fig. 4D ). The bioactive cleaved form of SREBP2 was higher in mice fed fi sh oil than in mice fed lard, and there was a trend toward a reduction by the microbiota in mice fed lard ( Fig. 4E, F ).
The rate-limiting enzyme of bile acid synthesis, CYP7A1, which uses cholesterol as substrate, was downregulated by Fig. 3. Gene expression in the liver of CONV-R and GF mice fed lard or fi sh oil for 11 weeks. A: PCA of global hepatic gene expression in CONV-R and GF mice fed lard or fi sh oil (n = 6 mice per group). B: Microbial regulation of hepatic genes in mice fed lard (x axis) or fi sh oil (y axis). Fold changes (FC) for genes signifi cantly regulated ( P < 0.05, corrected for FDR) are displayed (n = 6 mice per group). C: Microbial regulation of hepatic genes within GO category GO:016126 (sterol biosynthesis) in mice fed lard (x axis) or fi sh oil (y axis). Fold changes for genes signifi cantly regulated ( P < 0.05, corrected for FDR) are displayed (n = 6 mice per group). lipids strongly affected the abundance and composition of lipids in the liver and in serum. We have previously shown that the presence of a gut microbiota in mice fed a chow Fatty acids are absorbed from the intestine and utilized as building-blocks in the biosynthesis of lipids. In accordance with previous reports ( 25 ), we found that dietary diet regulates lipid composition in serum and liver and decreases abundance of hepatic cholesteryl esters ( 24 ). The present study demonstrates that the gut microbiota also infl uences lipid composition in mice on a high-fat diet . The gut microbiota regulated the abundance of many lipid species, but had a modest effect on the abundance of lipid classes with the exception of cholesterol and cholesteryl esters, which were upregulated by the gut microbiota in the liver of mice fed lard. The genes encoding sterol O -acyltransferase ( Soat1/2 ), carboxyl ester lipase ( Cel ), and lipase A ( Lipa ), which convert cholesterol into cholesteryl esters, were not regulated by the gut microbiota (data not shown). The increased levels of cholesteryl esters in CONV-R mice fed lard may instead result from the increased availability of hepatic free cholesterol.
Depending on dietary composition, GF mice may be protected against obesity and associated conditions such as increased levels of triglycerides in the liver ( 9,12 ). However, this is not always the case and Fleissner et al. ( 26 ) have shown that CONV-R and GF mice differ in adiposity and hepatic triglyceride levels on a Western diet, but not on a high-fat diet. Here we show that weightmatched GF and CONV-R mice on a high-fat diet have similar levels of lipids in the liver.
In accordance with Rabot et al. ( 12 ), we found that genes involved in cholesterol biosynthesis were downregulated by the gut microbiota in liver from mice fed lard. Interestingly, we also found that expression of genes involved in cholesterol biosynthesis was low in mice fed fi sh oil, independent of microbial status. Cholesterol biosynthesis is controlled by negative feedback through inhibition of SREBP2 processing. We found that the increased levels of hepatic cholesterol induced by the gut microbiota and by a lard diet were inversely related to the levels of activated SREBP2. Microbiota-induced inhibition of SREBP2 was paralleled by decreased expression of genes involved in cholesterol biosynthesis in mice on a lard diet. Surprisingly, however, mice fed fi sh oil had low expression of genes encoding cholesterol biosynthesis enzymes despite high levels of activated SREBP2. Saturated lipids have been shown to induce cholesterol biosynthesis ( 27 ), while lipids rich in polyunsaturated fatty acids have been shown to reduce cholesterol biosynthesis in the liver ( 28 ). Moreover, in vitro experiments have previously demonstrated that the polyunsaturated fatty acid, DHA, activates SREBP2 without increasing cholesterol biosynthesis ( 29,30 ), suggesting that polyunsaturated fatty acids regulate cholesterol biosynthesis through alternative mechanisms.
Cholesterol is the precursor for bile acids, and hence bile acid production is a potential sink for the hepatic cholesterol pool. In accordance with previous reports on mice fed a chow diet ( 31 ), we found that expression of CYP7A1, the rate-limiting enzyme in the bile acid biosynthesis pathway, was downregulated by the gut microbiota in mice fed lard and in mice fed fi sh oil. The decreased levels of hepatic cholesterol in GF mice may therefore be related to increased bile acid production associated with suppression of the FXR-mediated negative feedback loop due to increased levels of the FXR antagonist, ␤ -muricholic acid ( 31 ). Previous investigators have observed increased expression of genes encoding the membrane transporters ABCG5 and ABCG8 in the liver of GF mice ( 12 ) and suggested that these proteins may mediate increased cholesterol secretion and decreased hepatic cholesterol levels in GF mice. However, we did not fi nd any differences in the expression of Abcg5 or Abcg8 between GF and CONV-R mice (data not shown). The dissimilarity in regulation of ABCG5 and ABCG8 may be related to differences in experimental design between the studies. In the present study, diets with 45% energy from fat were used and CONV-R and GF mice were matched for body weight, while in the previous study diet, 60% energy from fat was used and GF mice were leaner than their CONV-R counterparts ( 12 ).
In summary, we show that interaction between the gut microbiota and dietary lipids regulates hepatic cholesterol biosynthesis and that the infl uence of gut microbiota on hepatic cholesterol metabolism is greater in mice fed a lard diet compared with those on a fi sh oil diet.