A Lipidomic Readout of Disease Progression in A Diet-Induced Mouse Model of Nonalcoholic Fatty Liver Disease

Abstract

Multiple changes in lipid metabolism occur in nonalcoholic fatty liver disease. However, it is not known which of these contribute to disease progression. The objective of this study was to define changes in hepatic lipid composition over time in a diet-induced model of nonalcoholic fatty liver disease to identify changes associated with disease progression. A lipidomic approach was used to quantify individual lipid species with lipid classes of interest including diacylglycerols (DAG), cholesterol, phospholipids, plasmalogens, sphingolipids, and eicosanoids. C57b/S129J mice fed a high-fat, high-cholesterol diet developed fatty liver, inflammation, and ballooning by 16 weeks and extensive fibrosis by week 52. There was a marked increase in monounsaturated fatty acid containing DAGs and cholesterol esters by week 16 which decreased by week 52. The changes in DAG were associated with a 500- to 600-fold increase in phosphatidic acid (< 0.001) and its downstream product phosphatidylglycerol (P <0.01) whereas phosphatidylethanolamine, phosphatidylcholine, and phsophatidylserine all decreased. Disease progression was associated with a significant further decrease in phosphatidylcholine and phosphatidylethanolamine while several lysolecithin species increased. Disease progression was associated with a significant increase in the plasmalogen PC-P 16:0/16:1. Saturated fatty acid (16:0 and 18:0) containing ceramides, sphingosine, sphingosine-1-phosphate, dihydrosphingosine, and dihydrophingosine-1-phosphate increased by week 16 after high-fat high-cholesterol diet. Globotrioseacylceramide (GB3) also increased significantly by week 16 and increased further with disease progression. 12-hydroxyeicosatetranoic acid decreased at week 16 but increased with disease progression. In conclusion, multiple lipids were associated with disease progression and provide clues regarding lipid drivers of nonalcoholic steatohepatitis.

INTRODUCTION

Nonalcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver disease in the Western world and is a growing problem in other parts of the world (1,2). The clinical-histological phenotype of NAFLD extends from nonalcoholic fatty liver (NAFL) to nonalcoholic steatohepatitis (NASH). NAFL and NASH can progress to cirrhosis in 2% to 3% and 15% to 20% of cases, respectively (3,4), and contribute substantially to the burden of end-stage liver disease and hepatocellular cancer (5,6). The mechanisms underlying disease progression are incompletely defined.

The histological hallmark of NAFLD is the accumulation of triglyceride-rich lipid droplets within hepatocytes (7,8). The majority of the literature on the role of lipids in the genesis of NASH has focused on the role of free fatty acids in triglyceride accumulation and as agents of liver injury (9–14). However, it is well known that the liver plays a central role in many aspects of lipid metabolism across various classes of lipids. Published literature on the role of other classes of lipids has shown changes in several such classes in NAFLD (15–18). Many altered lipid classes include lipid species that are biologically active and affect insulin signaling, lipogenesis, cell injury, and repair pathways. Therefore, it is both possible and plausible that there are complex disturbances across multiple classes of lipids that contribute to development of a specific NAFLD phenotype (fatty liver versus steatohepatitis) and disease progression.

Although several abnormalities in various lipid classes have been identified in NAFLD (18–21), the relationship of such changes to disease progression to advanced fibrosis has been limited by the cross-sectional nature of the studies. Another limitation of many studies which have focused on individual specific lipid species is that concurrent changes in other lipid moieties were not captured. To our knowledge, there are no published studies specifically relating changes across all major lipid classes to disease progression in a longitudinal manner.

The use of lipidomics allows an unbiased assessment of changes across all the major classes of lipid simultaneously. Using such an approach, we and others have previously documented altered fatty acid composition across many classes of lipids in NAFLD (18,21). However, these early studies were limited by the technology available which simply involved hydrolysis of fatty acids from individual lipid classes and their enumeration. Current methodology allows simultaneous identification of individual lipid species across classes of lipid, thereby providing much greater insight in to the metabolic perturbations in the systems being studied. It also allows identification of specific species that may play an important role in disease pathogenesis and thus be a target for intervention in future studies.

The goal of this study was to define the changes in hepatic lipid composition over time in a diet-induced mouse model of NAFLD and to identify specific lipids and lipid pathways that were related to disease progression. For this we used a mouse model (obtained from Dr. Sandra Erickson, University of California, San Francisco, CA) that was fed a high-fat, high-cholesterol diet (HFHCD) which developed fat and scattered inflammation by weeks 8 through 16 and increasing fibrosis from week 32 onwards. We used an unbiased lipidomic approach to quantify 1) the amount of hepatic cholesterol esters, phosphatidic acid, diacylglycerols, phospholipids, plasmalogens, and sphingolipids following an HFHCD administered for 16 and 52 weeks; 2) the fatty acids at specific carbon atoms within these lipids to define the amounts of specific molecular species of a given lipid; and 3) the hepatic eicosanoids in these high-fat dietary groups and to compare them from chow-fed controls.

MATERIALS AND METHODS

Mice and Diets

Female 129S1/SvlmJ;C57Bl/6J mice were obtained from the University of California, San Francisco (UCSF, CA) and kept on a 12:12-hour light-dark cycle. At 10 weeks of age, the mice were randomly assigned to two sets of groups: chow (n = 5) or high-fat diet (HFD; n = 5) given for 16 weeks; and chow (n = 5) or HFD (n = 5) given for 52 weeks. The chow-fed mice were fed a standard chow diet (Teklad 7012, Harlan) with 17% of energy derived from fat, 58% from carbohydrates, and 25% from protein. HFD-fed mice received a western diet (Teklad 88137, Harlan) with 42% of energy derived from fat, 43% from carbohydrate, and 15% from protein. Food and water were provided ad libitum until the end of the study period. At the end of the feeding period, the animals were fasted for 12 hours, euthanized, and body weights determined. Blood was collected through heart puncture, allowed to clot, and serum obtained by centrifugation at 3,000 rpm for 15 minutes at 4°C. Livers were obtained, weighed, and dissected; a portion of fresh tissue was fixed in 10% buffered formalin and remaining tissues were snap-frozen in liquid nitrogen. Serum and liver samples were stored at -80°C until lipidomic or biochemical analysis. All animal experiments were approved by Institutional Animal Care and Use Committee of Virginia Commonwealth University.

Serum Biochemistry Profile

Serum measurements of aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma-glutamyltransferase, bilirubin, cholesterol, high-density lipoprotein, low-density lipoprotein (LDL), and triglycerides were performed in the clinical chemistry laboratories of the author's institution using established commercially available methods.

Lipid Analysis

The liver tissue samples were weighed, pulverized with CP02 CryoPrep Dry Pulverization System (Covaris, Brighton, UK), and resuspended in ice-cold methanol containing 0.1% butyl-hydroxy-toluene in a concentration of 100 mg/mL. The homogenized samples were stored at -80°C before lipid extraction and analysis. For lipidomics analysis, lipids were extracted using a modified Folch lipid extraction (2) performed on a Hamilton Microlab Star robot (Reno, NV, USA). Samples were spiked with known amounts of non-endogeneous synthetic internal standards. After lipid extraction, samples were reconstituted in chloroform:methanol (1:2, v/v) and a synthetic external standard was post-extract spiked to the extracts. The extracts were stored at -20°C before mass spectrometry (MS) analysis. The lipid extracts were analyzed on a hybrid triple quadrupole/linear ion trap mass spectrometer (QTRAP 5500, Sciex, Singapore) equipped with a robotic nanoflow ion source (NanoMate HD, Ithaca, NY, USA) according to Ståhlman et al (22). Molecular lipids were analyzed in both positive and negative ion modes using methods based on multiple precursor ion scanning and neutral loss (23,24). Lipid class-specific internal standards were used for quantifying endogenous lipid species. The MS data obtained from MS instruments were exported as .wiff or .txt files that represented the basic raw files of lipidomic analysis. These files contained information on masses of identified molecules and their counts (intensities and areas). Masses and counts of detected peaks were converted into a list of corresponding lipid names and concentrations. Calibration lines were generated to determine the dynamic quantification range for each lipid class monitored, for example, the quantification limits. As the internal standards used behave in the same way as endogenous lipids, they were used for quantifying endogenous lipid species. Calibration lines were not used to normalize the endogenous lipids due to lack of appropriate external standards, for example, lipid class-specific standards were not available. However, the calibration lines determined the quantification limits of the method and were based on the same internal standards that were used for quantification of the endogenous lipids. The calibration lines consisted of a minimum of four accepted standard points covering the linear quantification range.

Quality control was performed based on the ratio of synthetic internal standards (IS) to corresponding post-extract spiked external standards (ES), and MS analysis of extracted matrix and solvents served as quality controls of the analysis. In addition, extracted reference liver samples were analyzed for monitoring the instruments' performance. The analysis acceptance standards were based on the linearity of the calibration lines. The linear regression had to exceed 0.95 based on at least four of six non-zero standards. The analysis was accepted based on the identification of sample-specific IS and ES. The coefficient of variation of an area ratio (cps) of internal to external standards (IS/ES) was used to identify potential technical outliers per analysis platform.

Eicosanoid Analysis

Lipids were extracted using solid-phase extraction. Known amounts of isotope-labeled standards were included as synthetic internal and external standards. The solid-phase extraction plate wells were washed with 35% methanol and eluted in 100% acetonitrile. The samples were then evaporated until dry and reconstituted in methanol followed by addition of external standard mixture. Lipids were analyzed on a hybrid triple quadrupole/linear ion trap mass spectrometer (5500 QTRAP) equipped with an ultra-high−pressure liquid chromatography system (CTC HTC PAL autosampler and Rheos Allegro pump) using a multiple reaction monitoring–based method in negative ion mode.

Histology

Liver sections were fixed in 10% formalin. Tissues were embedded in paraffin blocks and stained with hematoxylin and eosin and Masson's trichrome stain using standard commercially available methods. Histology was read by a single independent pathologist blinded to the experimental design and dietary groups. Steatosis, inflammation, and fibrosis were quantified separately.

Statistical Analysis

The levels of lipid metabolites in HFHCD-fed mice for 16 weeks versus 52 weeks were compared to chow-fed mice and to each other using unpaired Student's t-test or Kruskal Wallis analysis of variance as appropriate. Data are presented as the mean ± standard error of the mean. A P value of <0.05 was considered to be significant.

RESULTS

HFHCD Leads to Obesity, Dyslipidemia, and Insulin Resistance

A total of 5 to 6 mice per group were given either chow diet or HFHCD for 16 or 52 weeks. The biochemical profile of the animal groups is shown in Table 1. Mice fed an HFHCD gained more weight and had higher aminotransferases, cholesterol, and LDL-cholesterol compared to chow-fed mice at both 16 and 52 weeks. Mice fed an HFHCD also had continuous weight gain and were significantly heavier at 52 weeks compared to 16 weeks (mean 50 g versus 16 g, P <0.05). The liver weight was also significantly higher in HFHCD fed mice and greater after 52 weeks of treatment than at 16 weeks. Both the fasting blood sugar and insulin levels were higher in HFHCD-fed mice indicating the development of insulin resistance. HFHCD-fed mice for 52 weeks also had significantly higher serum aminotransferases, cholesterol, and LDL compared to HFHCD-fed mice for 16 weeks.

TABLE 1.

Biochemical Profile of the Animal Groups

Item	CD 16	WD 16	CD 52	WD 52
Weight gain (g)	6.06 ± 0.65	15.55 ± 1.97		50.5 ± 3
Liver weight (g)	1.05 ± 0.14	1.66 ± 0.87		2.4 ± 0.4
AST (U/L)	154.21 ± 16.91	285.55 ± 73.55		628 ± 83
ALT (U/L)	48.21 ± 6.66	166.30 ± 58.71		586 ± 127
ALP (U/L)	81.06 ± 3.81	81.35 ± 18.33		132 ± 26
Bilirubin (mg/dL)	<0.1	<0.1		< 0.1
Glucose (mg/dL)	360.50 ± 16.54	345.93 ± 25.54		382 ± 72
Cholesterol (mg/dL)	130.40 ± 3.84	297.00 ± 31.83		337 ± 40
TG (mg/dL)	86.00 ± 7.30	103.75 ± 3.50		137 ± 37
LDL (mg/dL)	36.60 ± 6.39	231.000 ± 39.84		209 ± 33
Insulin (pg/mL)	532.33 ± 85.87	1311.67 ± 582.20		1291 ± 375

Abbreviations: CD, chow diet; WD, western diet; AST, aspartate transaminase; ALT, alanine transaminase; ALP, alkaline phosphatase; TG, triglycerides; LDL, low-density lipoprotein.

HFHCD Leads to Progressive NAFL Disease

The gross morphology and histology of the liver in chow-fed mice was normal at both 16 and 52 weeks (Figures 1A and 1B). In contrast, the liver of all HFHCD-fed mice was a light tan color and developed some nodularity by week 52 (Figures 1C and 1D). HFHCD-fed mice developed macrovesicular steatosis, foci of small droplet steatosis, and scattered inflammation at 16 weeks (Figure 1E) along with focal cytological ballooning and pericellular fibrosis (Figures 1E and 1F). There were also areas of small droplet steatosis mostly in centrilobular regions. By 52 weeks, macrovesicular and centrilobular small droplet steatosis remained prominent along with extensive pericellular fibrosis and areas of bridging fibrosis (Figures 1G and 1H). Two of the five mice fed an HFHCD up to 52 weeks developed foci of hepatocellular cancer, whereas none of mice fed a chow diet or an HFHCD up to 16 weeks developed hepatocellular cancer.

Fig. 1.

Histology of mice fed chow or high fat high calorie diet (HFHCD) for 16 and 52 weeks. (A,B) In chow-fed mice, the liver histology was normal at both 16 and 52 weeks, respectively. (C,D) In HFHCD-fed mice, the liver developed a light tan color and some nodularity by week 52. HFHCD-fed mice also developed macrovesicular steatosis, foci of small droplet steatosis, and scattered inflammation at 16 weeks (E) along with focal cytological ballooning (arrow) and pericellular fibrosis (E–F). By 52 weeks, there was prominence of macrovesicular and centrilobular small droplet steatosis along with extensive pericellular fibrosis and areas of bridging fibrosis (G–H).

Changes in Diacylglycerols

Total diacylglycerols (DAGs) increased significantly by week 16 (P <0.01) in HFHCD-fed mice compared to chow-fed mice, but this difference was no longer significant by week 52. However, there was a significant increase in monounsaturated fatty acid (MUFA) − and saturated fatty acid (SFA) −containing DAGs at both 16 and 52 weeks (Figure 2A). The principal MUFA enrichment of the DAG pool was due to palmitoleic (16:1) acid and oleic (18:1) acid. This was accompanied by a significant decrease in polyunsaturated fatty acid (PUFA) −containing DAGs at both times. However, as the disease progressed over time from 16 to 52 weeks, there was a significant decrease in DAGs especially (18:2/20:2), (18:0/18:2), and (16:1/18:0).

Fig. 2.

Changes in diacylglycerols, cholesterol esters, and phospholipids. Total diacylglycerols (DAGs) increased significantly by week 16 of high-fat, high-cholesterol diet (HFHCD) (P <0.05) compared to chow-fed mice (A). Monounsaturated fatty acid (MUFA) −and saturated fatty acid (SFA) −containing DAGs increased significantly at both 16 and 52 weeks, whereas polyunsaturated fatty acid (PUFA) −containing DAGs decreased significantly at both time points. In mice fed an HFHCD, there was a significant increase in cholesterol esters (CEs) both at 16 weeks and 52 weeks (P <0.001) compared to chow-fed mice (B). CE was enriched with SFA and MUFA both at week 16 and week 52 in the former group. The levels of CEs decreased in HFHCD-fed mice from week 16 to week 52 (P <0.01), accompanied by a decrease in SFA-, MUFA-, and PUFA-containing CEs. There was a significant increase in phosphatidic acid (PA) by week 16 (P <0.001) in HFHCD-fed mice compared to chow-fed mice (C). This was accompanied by an increase in its downstream product phosphatidylglycerol (PG) (P <0.001). Disease progression in HFHCD-fed mice from week 16 to week 52 was accompanied by continued increase in PG and PA (non-significant) and a significant decrease in PC and PE (P <0.01 for both) (D).

Changes in Cholesterol Esters

HFHCD for 16 and 52 weeks led to 1070% and 500% increase in cholesterol esters (CEs) (P <0.001 for both), respectively, compared to chow-fed mice (Figure 2B). CE was enriched with SFA (14:0, 15:0, 16:0, 17:0, and 18:0) and MUFA (16:1, 17:1, 18:1, 19:1, and 20:1) both at weeks 16 and 52 in HFHCD-fed mice. The levels of CEs decreased in HFHCD-fed mice from week 16 to 52 (P <0.01). This was accompanied by a decrease in mainly in SFA- (14:0, 15:0, 16:0, 17:0, and 18:0) and MUFA- (14:1, 15:1, 16:1, 17:1, 18:1) containing cholesterol esters.

Changes in Phospholipids

There was a significant (P <0.001) 500- to 600-fold increase in phosphatidic acid by week 16 (Figure 2C). This was accompanied by an increase in its downstream product phosphatidylglycerol (PG) (P <0.01) while phosphatidylinositol (PI), phosphatidylcholine (PC), and phosphatidylethanolamine (PE) decreased (P =ns). There was a specific enrichment of PG (18:1/18:1 and 16:0/18:1), PI (18:1/18:1), PC (16:1, 18:1), PE (16:1, 18:1), and phosphatidylserine (18:1) by MUFAs. PC and PE were also enriched with SFA (16:0, 17:0, and 18:0). PI was also enriched with SFA (18:0) and PUFAs (20:3, 20:4). Specific plasmalogens (PE-P 18:1/18:2, PE-P 16:0/20:3, PC-P 18:0/16:0) increased by week 16, although total plasmalogens did not change significantly. By week 52, PA and PG continued to increase. PC and PE decreased further but remained enriched with MUFA. PUFA containing PC, PE, and PI were decreased by week 16 and decreased further by week 52.

Disease progression in HFHCD-fed mice from week 16 to week 52 was accompanied by a further significant decrease in PC and PE, although they remained enriched with MUFA (Figure 2D). This was accompanied by an increase in several lyso-PC (lysolecithin) and lyso-PE species. The levels of PG (18:1/18:1 and 18:1/18:2) increased further with disease progression but did not reach significance (from week 16 to 52). Whereas other PI species declined from week 16 to week 52, PI (18:1/18:1) increased. Disease progression was additionally associated with increased plasmalogen PC-P 16:0/16:1.

Changes in Sphingolipids

Ceramides containing SFA (16:0 and 18:0) were increased at week 16 in HFHCD-fed mice compared to chow-fed mice (Figure 3A). This was accompanied by an increase in sphingosine, sphingosine-1-phosphate as well as dihydrosphingosine and dihydrosphingosine-1-phosphate (Figures 3B and 3C). This was accompanied by an increase in several species of sphingomyelins, although only one reached statistical significance (18:1/15:0–18:1/14:1) (Figure 3D).

Fig. 3.

Changes in sphingolipids. Ceramides containing saturated fatty acid (SFA) (16:0 and 18:0) increased at week 16 in high-fat, high-cholesterol diet (HFHCD)−fed mice compared to chow-fed mice (A). This was accompanied by a significant increase in sphingosine, sphingosine-1-phosphate as well as dihydrosphingosine and dihydrosphingosine-1-phosphate (B–C). From week 16 to week 52 in HFHCD-fed mice, sphingosine remained relatively unchanged while dihydrophingosine and sphingosine-1-phosphate declined significantly (P <0.01 and <0.001, respectively) and dihydrosphingosine-1-phosphate increased significantly (P <0.05). Several species of sphingomyelins also increased, although only one reached statistical significance (18:1/15:0–18:1/14:1) (3). Both galactosylceramide and glucosylceramide decreased whereas lactosylceramide and globotrioseacylceramide (GB3) increased by week 16 (P =ns) in HFHCD-fed mice compared to chow-fed mice (E).

Another sphingolipid metabolic pathway relates to formation of globotrioseacylceramide (GB3) and gangliosides. Both galactosylceramide and glucosylceramide decreased while lactosylceramide and GB3 increased by week 16 in HFHCD-fed mice compared to chow-fed mice (Figure 3E). This was mainly due to enrichment of MUFA containing species of these compounds.

With disease progression at week 52, Cer (d18:1/24:1 and 18:1/16:0) increased significantly (P <0.01 and P <0.05, respectively) compared to week 16 in HFHCD-fed mice (Figure 4A). Sphingosine-1-phosphate was still higher in HFHCD-fed mice at week 52 compared to chow-fed mice (Figures 3B and 3C). The concentration of sphingosine-1-phosphate declined significantly from week 16 to week 52 (P <0.001) in HFHCD-fed mice but remained higher than in chow-fed controls while dihydrosphingosine-1-phosphate increased significantly (P <0.05). GB3 increased further by week 52 in HFHCD-fed mice (Figure 4B) and was significantly higher compared to both chow-fed mice as well as mice fed an HFHCD for 16 weeks (Figure 4C). Of the various species of GB3, (18:1/22:1 and 18:1/16:0) increased from week 16 to week 52 (P <0.001 and P <0.01, respectively).

Fig. 4.

Changes in ceramide and globotrioseacylceramide (GB3) with disease progression from 16 weeks to 52 weeks of high-fat, high-cholesterol diet (HFHCD) −fed mice. Cer d18:1/24:1 and 18:1/16:0 increased significantly (P <0.01 and P <0.05, respectively) at week 52 compared to week 16 in HFHCD-fed mice (A). GB3 increased further by week 52 in HFHCD-fed mice compared to chow-fed mice and also relative to HFHCD-fed mice at 16 weeks (B–C). Specifically, 18:1/22:1 and 18:1/16:0 increased significantly at week 52 (HFHCD versus chow diet). 18:1/24:1 increased significantly in mice fed HFHCD for 52 weeks compared to HFHCD for 16 weeks.

Changes in Eicosanoids

Relative to chow-fed mice, there was a decrease in most measured eicosanoids in HFHCD-fed mice at 16 weeks with the exception of thromboxane B2 and PGF2α which trended up (Figure 5A). Specifically, there was a significant decrease in PGE2, PGD2, 5-hydroxy eicosatetraenoic acid (HETE), 8-HETE, 11-HETE, 15-HETE, 5–6 dihydroxyeicosatetraenoic acid (DHET), 8,9-DHET, 11,12-DHET, and 14,15-DHET. By week 52, compared to chow-fed mice, there was a further decrease in multiple eicosanoids, many of which were statistically significant (Figure 5B). Whereas 12-HETE, the metabolic product of 12/15 lipoxygenase (mouse Alox 15 gene) from arachidonic acid, decreased compared to chow-fed mice at week 16, this trend reversed and it was higher compared to chow-fed mice by week 52. When eicosanoid levels were compared in mice fed an HFHCD for 52 weeks versus 16 weeks, arachidonic acid increased (P <0.01) as did 12-HETE and 15-HETE (P =ns for these) (Figure 5C).

Fig. 5.

Changes in eicosanoids. Most measured eicosanoids in high-fat, high-cholesterol diet (HFHCD) −fed mice at 16 weeks, with the exception of thromboxane B2 and PGF2 alpha, decreased relative to chow-fed mice (A). By week 52, compared to chow-fed mice, there was a further decrease in multiple eicosanoids, many of which were statistically significant (B). 12-HETE, the metabolic product of 12/15 lipoxygenase (mouse Alox 15 gene) from arachidonic acid, decreased in HFHCD-fed mice at week 16 compared to chow-fed mice. This trend, however, reversed and was higher compared to chow-fed mice by week 52. When eicosanoid levels were compared in mice fed an HFHCD for 52 weeks versus 16 weeks, total eicosanoids (P <0.05) and arachidonic acid increased significantly (P <0.01). Multiple eicosanoids also trended upwards including 12-HETE and 15-HETE (C).

DISUSSION

Lipidomics allow an unbiased simultaneous assessment of the functional state of various lipid metabolic pathways. They provide novel insight on the status of lipid metabolism in an organ but are primarily a discovery tool which generates hypotheses that require validation in specifically designed studies. Previous cross-sectional studies have shown an increase in triglycerides, DAG, free cholesterol, and depletion of PC in NAFLD (18,20,21). The current study provides a lipidomic signature of disease progression in a diet-induced mouse model of NAFLD that progressed to advanced fibrosis. Importantly, many of these changes were revealed by identification of specific lipid species and would have been missed by measurement of total lipids or fatty acids alone.

An important finding was an increase in MUFAs across several lipid classes including DAGs, CEs, and several types of phospholipids. MUFAs are produced by the activity of steroyl Coenzyme A (CoA) desaturase, a sterol response element binding protein-1 (SREBP-1)−dependent enzyme. Previous studies have established increased steroyl CoA desaturase and SREBP-1 activation in NAFLD (25,26). The data in this study are compatible with and likely to reflect hyperinsulinemia-driven SREBP-1 activation and downstream increase in desaturase activity (27). The regression of this enrichment with disease progression may reflect increasing resistance to insulin functions to include insulin-mediated SREBP-1 activation. However, this hypothesis requires experimental validation.

A novel finding is the highly significant increase in PA in HFHCD-fed mice relative to chow-fed mice. PA is derived from DAG and this increase likely reflects increased DAG. PA- and MUFA-enriched DAGs both promote insulin resistance and gluconeogenesis (28,29). These two previously unsuspected metabolites may thus provide a potential metabolic basis for perpetuation of insulin resistance in NAFLD and the relationship between NASH and type 2 diabetes. It is also noteworthy that despite a decline in total DAG, the PA level remained increased with disease progression suggesting increased DAG kinase activity.

There were also several other findings particularly associated with disease progression. Several ceramides, especially SFA-containing ceramides, were increased at both week 16 and week 52. This was accompanied by an increase in sphingosine and sphingosine-1-phosphate. Both ceramides and sphingosine-1-phosphate have important biological properties affecting cell viability, proliferation, insulin signaling, and metabolism (30–33). The findings from this study provide a strong rationale for further studies to better elucidate the role of these lipids in disease progression in NAFLD and their potential utility as targets for therapy.

Another novel finding is the progressive increase in GB3 (18:1/24:1) with disease progression at week 52 in the NAFLD mice. GB3 functions as a receptor for shiga toxin and also is a potent activator of the innate immune system (34,35). Activation of the innate immune system plays an important role in the inflammation and disease progression in NAFLD (36,37). Most of the previous literature has focused on the role of intestinal bacterial products, for example, endotoxin and SFAs in driving the innate immune system in NAFLD (38,39). The data from this study raise the exciting possibility that GB3 is an as yet unknown but important driver of inflammation and disease progression in NAFLD and a potential target for therapeutics.

Previous studies have found a specific increase in several lipoxygenase products such as 5-HETE, 8-HETE, 12-HETE, and 15-HETE in the plasma of patients with NAFLD (21). Before this study, there were no published data on the levels of these metabolites in the liver itself. Unlike what was previously noted in circulation, most measured eicosanoids were decreased in this mouse model of NAFLD. However, the pro-inflammatory thromboxane B2 and 12-HETE were increased and 12-HETE was specifically associated with disease progression. This suggests that the previously published changes in circulation albeit in humans only, with the exception of 12-HETE, largely reflect systemic changes associated with NAFLD and its associated insulin resistant state. Knockdown of 12–15 lipoxygenase has previously been shown to reduce the development of NAFLD following a high-fat diet (40). Together with our observation that 12-HETE increases are associated with disease progression, 12-HETE emerges as the most potentially relevant eicosanoid target for therapeutics in NAFLD.

A potential limitation of this study is the relevance of mouse models of NAFLD to human disease. None of the mouse models truly reflect human disease. The leptin deficient ob/ob mouse and the methionine-choline−deficient diet models are two of the most common models used. The former usually does not develop progressive disease whereas the methionine-choline−deficient mouse loses weight and remains insulin sensitive (41,42). The mouse model used in the current study is diet-induced and associated with weight gain, insulin resistance, and increased abdominal fat stores. Although there were some differences in the histology with human disease, the disease did progress to advanced fibrosis. Thus, of the models available, this model comes close to reflecting human disease.

In summary, the current longitudinal study provides a lipidomic signature of disease progression of NAFLD. Numerous hitherto unknown findings associated with disease progression were noted. It is likely that one or more of these are important as drivers of disease progression and the findings from this study provide direction and a rationale for the study of several specific and novel lipid targets for the treatment of this condition for which there is currently no approved therapy.

DISCUSSION

Nettleman, South Dakota: So do you apply this in your clinical setting? Do you talk to your patients who have fatty liver disease and appear to have obesity and say to them that, in addition to all the things we would like you not to do, we would like you not to drink, or if you drink at least don't binge?

Sanyal, Richmond: So, what got us involved in this area was the initial observation that in our clinics at least a third — and probably now more like 50% of the patients we see — consume alcohol that is more than what the classic definition of a nonalcoholic is. So, they cannot get into clinical trials. But they do not drink enough to actually be called alcoholic liver disease. We call them BASH because they have both, it is an overlap syndrome and this is a substantial population. What we know is from the two extremes — of alcohol and nonalcohol — and we extrapolate and we say everything is the same. We don't know that for a fact. So, we certainly tell them not to engage in risky drinking behavior. With early-stage disease, there is evidence that their insulin resistance might get better and there might be some improvement in high-density lipoprotein and some other parameters. So, I think the data is a little mixed. It is hard to be very evangelical. But for people who have advanced disease, who have a lot of fibrosis, there is evidence that even modest amounts of alcohol increases the risk of hepatocellular carcinoma. So, for them we certainly would say to not drink.

Vierling, Houston: In alcoholic liver disease there is an inflammatory component including an adaptive immune component to adducts formed through the metabolism of alcohol. There is now interest in their role in the inflammasome and associated molecular pattern signaling. I'm curious whether you had a chance to look in your normal alcohol, regular alcohol, and your binge for any adduct formation that might contribute to pathology?

Sanyal, Richmond: No, we have not actually. Our foray into the alcohol world is relatively recent because all my past work has been on the nonalcoholic side. So, I think trying to elucidate the metabolic basis for the inflammatory response is important and all the long-term consequences driven by inflammation. Inflammation drives the fibrogenesis, the risk of hepatocellular carcinoma, and decompensation. I think that is an important role, but we just have not gotten around to it yet.

Vierling, Houston: The second question relates to nonalcoholic fatty liver disease and the protective effect in women of estrogen or, in post-menopausal women, of estrogen replacement therapy. I noticed that your studies were in male mice. I wonder if you have comparable data in females.

Sanyal, Richmond: We are doing the alcohol obesity interaction studies in female mice right now. We have a lot of lipidomic data in female mice, and actually the lipidome looks remarkably similar. But it's very interesting that the female mice get much less liver cancer. The fibrosis readout is about very similar between the female and the male mice. The big difference that we see — and also this recapitulates the human situation where hepatocellular carcinoma as you know is much more male predominant — is either the androgens are bad actors or the estrogens are protective or something in the middle.

Schuster, New York: Two of the signals that came were two of the epoxyeicosatrienoic acids, and then they sort of fell off? They are anti-inflammatory. So what are your thoughts about the mix of the eicosanoid mediators, and are there resolvins that are coming up?

Sanyal, Richmond: We have recent data that there is a decrease in resolvins. So there is both an increase in the pro-inflammatory eicosanoids and there is also a decrease in the anti-inflammatory eicosanoids. Furthermore, there is a decrease in the DHETs which you know are important for regulating the microcirculation.

Bodenheimer, New York: You know alcohol and alcoholic liver disease is where the gut and the liver connect. I wondered if there are any data that exist, or if you have noticed changes in your animal model, on bowel function. Does binge drinking affect the microbiome differently than just amount of alcohol per week?

Sanyal, Richmond: We actually have that data. The data shows interaction between the microbiome. You have alcohol-related changes in the microbiome and there are also obesity-related changes in the microbiome. The big change with obesity is that you get an increase in proteobacteria, and there is an alteration in the firmicutes and the bacteriodes. We have some very recent data — and it took a long time for us to actually get the methodology worked out because there is so much endotoxin in stool that it is very hard to measure that it swamps out everything and after a lot of dilution and other stuff — that nonalcoholic steatohepatitis patients seem to have a remarkable increase in both intestinal luminal endotoxin as well as circulating endotoxin. Intestinal alkaline phosphatase activity in stool is much higher, and its primary role is to dephosphorylate endotoxin and reduce the endotoxin burden in the body. There are clearly changes both with alcohol and obesity in the microbiome and the microbiome-associated metabolome.

Boxer, Ann Arbor: Have you had an opportunity to look at sphingolipid inhibitors?

Sanyal, Richmond: The mice have been sacrificed this week. We will have the histology in. So, we used the commercially available S1P receptor 1 modulator that is available for multiple sclerosis and we will have that data very soon.