Dietary modification dampens liver inflammation and fibrosis in obesity-related fatty liver disease


  • Disclosure: The authors declared no conflict of interest.

  • Funding agencies: This research was supported by Australian National Health and Medical Research Council (NHMRC) project grant 418101, NHMRC Fellowship 525473 and NHMRC Scholarship 585539


Background: Alms1 mutant (foz/foz) mice develop hyperphagic obesity, diabetes, metabolic syndrome, and fatty liver (steatosis). High-fat (HF) feeding converts pathology from bland steatosis to nonalcoholic steatohepatitis (NASH) with fibrosis, which leads to cirrhosis in humans.

Objective: We sought to establish how dietary composition contributes to NASH pathogenesis.

Design and Methods: foz/foz mice were fed HF diet or chow 24 weeks, or switched HF to chow after 12 weeks. Serum ALT, NAFLD activity score (NAS), fibrosis severity, neutrophil, macrophage and apoptosis immunohistochemistry, uncoupling protein (UCP)2, ATP, NF-κB activation/expression of chemokines/adhesion molecules/fibrogenic pathways were determined.

Result: HF intake upregulated liver fatty acid and cholesterol transporter, CD36. Dietary switch expanded adipose tissue and decreased hepatomegaly by lowering triglyceride, cholesterol ester, free cholesterol and diacylglyceride content of liver. There was no change in lipogenesis or fatty acid oxidation pathways; instead, CD36 was suppressed. These diet-induced changes in hepatic lipids improved NAS, reduced neutrophil infiltration, normalized UCP2 and increased ATP; this facilitated apoptosis with a change in macrophage phenotype favoring M2 cells. Dietary switch also abrogated NF-κB activation and chemokine/adhesion molecule expression, and arrested fibrosis by dampening stellate cell activation.

Conclusion: Reversion to a physiological dietary composition after HF feeding in foz/foz mice alters body weight distribution but not obesity. This attenuates NASH severity and fibrotic progression by suppressing NF-κB activation and reducing neutrophil and macrophage activation. However, adipose inflammation persists and is associated with continuing apoptosis in the residual fatty liver disease. Taken together, these findings indicate that other measures, such as weight reduction, may be required to fully reverse obesity-related NASH.


Nonalcoholic fatty liver disease (NAFLD) affects 15-45% of most societies in the Asia-Pacific region, North and South America and Europe, and prevalence is 70-80% in obesity. However, only 10-25% of people with NAFLD develop nonalcoholic steatohepatitis (NASH). Unlike simple steatosis which has few adverse liver outcomes, NASH is characterized by hepatocyte injury and inflammation, now conceptualized as a manifestation of lipotoxicity [1, 2], which can lead to hepatic fibrosis and eventually cirrhosis and hepatocellular carcinoma (HCC) [3]. Over-nutrition and genetic factors resulting in central obesity and insulin resistance, intake of a high saturated fat or high simple carbohydrate diet, particularly with excessive fructose and/or cholesterol content, diets low in antioxidant micronutrients, and lower levels of physical activity have all been reported in patients with NAFLD [1], but the relative importance of these components and how they impact on disease pathogenesis remains unclear. Patients with NASH tend to be older than other NAFLD patients, more obese, and they have a greater number of components of metabolic syndrome, glucose intolerance, or established type 2 diabetes [3, 4]. Despite these compelling relationships, the ways in which dietary and others factors causing steatosis lead to hepatocellular injury, inflammatory recruitment and liver fibrosis are poorly understood.

An older concept of NASH pathogenesis was that the metabolic abnormalities linked to NAFLD lead only to steatosis without hepatocyte injury, while development of NASH requires a second tier of injury-inducing mechanisms, such as oxidative stress or cytokines, particularly tumor necrosis factor-alpha (TNFα). However, the contemporary concept of lipotoxicity acknowledges that patients with NASH have more profound metabolic abnormalities than those with steatosis alone [1, 2]. Further, serum TNFα levels are usually elevated in patients with obesity irrespective of whether they have NASH or steatosis without injury, and steatohepatitis can be produced experimentally in mice lacking a functional TNF receptor [5, 6]. Conversely, metabolic factors such as low serum adiponectin, hyperinsulinemia (a consequence of insulin resistance in prediabetes), diabetes, and number of components of metabolic syndrome correlate with disease severity [4, 7]. A possible role for diet in the transition from steatosis to steatohepatitis has been proposed, particularly with roles of increased saturated fat intake at the expense of polyunsaturated fats identified in a number of studies, and the relationship of fibrosis severity in Americans with NASH to high fructose intake, but evidence to indicate that individual macronutrients could be involved is conflicting [1]. On the other hand, there is abundant experimental evidence that saturated fatty acids can activate nuclear factor-kappa B (NF-κB) in cell systems and human livers showing NASH [8]. NF-κB is activated in lipotoxicity, being linked to both apoptotic and necrotic forms of cell death, and has been implicated in NASH pathogenesis [5], providing a potential link between dietary factors, lipotoxicity and phenotypic expression of NAFLD as NASH.

As recently reviewed, among numerous models used to study NAFLD, few faithfully exhibit NASH according to current pathological criteria and metabolic profiling [9]. We previously reported the temporal evolution of liver lesions between 2 and 24 weeks of feeding an atherogenic HF diet to hyperphagic foz/foz (Alms1 mutant) mice in relation to lipid partitioning, hyperinsulinemia, hypoadiponectinemia, hypercholesterolemia and hyperglycemia [10]. Chow-fed foz/foz mice and high-fat (HF)-fed wild type (WT) mice exhibit only steatosis at 12 and 24 weeks. Conversely, intake of the HF diet in foz/foz mice accentuates metabolic and liver abnormalities so that at 12 weeks livers show early NASH (mean NAS score 4.9) with upregulation of the key matrix-forming gene, collagen-α1. By 24 weeks of HF dietary intake, severe NASH is present with fibrosis stage 1 (pericellular distribution) or 2 (dense fibrotic bands) being manifest.

In the present studies, we used this model to test the hypothesis that HF-feeding in foz/foz mice contributes directly to hepatic lipid accumulation by mechanisms additional to those leading only to steatosis, thereby promoting hepatocellular injury and inciting pro-inflammatory pathways. Particular attention was given to hepatocellular uptake of free fatty acids (FFA) in addition to their de novo synthesis; uptake of long chain fatty acids from the periphery is quantitatively more important in human NASH than hepatic lipogenesis [11]. Further, FFA uptake is more likely to lead to hepatic FFA accumulation and contribute to lipotoxicity than lipogenesis, which is a tightly coordinated pathway leading to formation of triglycerides, that are nontoxic in the liver [2]. We then used a dietary intervention, “switching” from HF diet back to chow, to test relationships between hepatic lipid turnover, severity and patterns of hepatocellular injury (necrosis versus apoptosis) and inflammatory recruitment (including macrophage phenotype). We finally assessed the implication of these changes for fibrogenesis, which was attenuated by this intervention.


Animals and diets

Breeding, genetic characterization, and genotyping of foz/foz (Alms1 mutant) mice are reported [12]. All experiments were approved by the ANU Animal Experimentation Ethics Committee (protocol F.MS.07.05). From 6 weeks of age, female foz/foz mice were fed HF-diet (23% fat, 45% carbohydrate, 20% protein, 0.19% cholesterol w/w [Specialty Feeds, Glenn Forrest, Western Australia]) (n = 7) or chow (n = 5) for 24 weeks (4.5% fat w/w). WT controls (n = 4) were fed chow 24 weeks (we studied only a small group here because we have previously reported the effects of chow and HF feeding in these WT mice [10, 13]). To study the effects of dietary intervention, another group of foz/foz mice were fed HF-diet for 12 weeks, then fed only chow for a further 12 weeks (“switch”, n = 8). At the end of experiments, mice were fasted 4 h, serum, liver, and peri-ovarian adipose tissue were collected, weighed, and frozen in liquid nitrogen for later analyses.

Assessment of metabolic and liver phenotype

Blood glucose, serum ALT, and lipids were measured using automated techniques at The Canberra Hospital. Liver histology was assessed by a blinded pathologist (MMY) according to the NAFLD Activity Score (NAS) [14]. Sirius red staining was used to demonstrate fibrosis, and morphometric analysis using Image-J was performed to quantify fibrosis area.

Serum and hepatic protein measurements and gene expression

Serum insulin (Millipore, Billerica, MA) adiponectin and leptin levels were measured by ELISA (both RnD Systems, Minneapolis, MN). Hepatic proteins were measured by western blotting, as reported [15], using Hsp90 as loading control. Nuclear protein was extracted using the NE-PER nuclear and cytoplasmic extraction kit (Pierce, Waltham, MA). Immunoblots were visualized by enhanced chemiluminescence and quantified using MultiGauge software (FujiFilm, Japan). Gene expression analyses were performed by real-time PCR with previously described primer sets [13, 16, 17] and those in Supporting Information Table 1.

Hepatic ATP levels

Hepatic ATP levels were measured in liver homogenates using the Molecular Probes ATP Determination Kit (Invitrogen, Carlsbad, CA). Frozen liver tissue was homogenized in 9× volume 150 mM NaCl, 50 mM Tris (pH 7.5), 1% Triton X 100, 0.1% sodium dodecyl sulfate, and 1% sodium deoxycholate. Homogenates were boiled for 3 min, then centrifuged at 16,000g for 5 min. Supernatants were diluted 1:1 in 0.1 mM tris acetate and then used in the ATP determination kit according to the Manufacturer's instructions.

Hepatic lipids and fatty acid synthase activity

Hepatic triglyceride, free cholesterol, cholesterol ester, diacylglyceride (DAG), free fatty acid (FFA) content, and fatty acid synthase activity were measured as previously described [18].


Following antigen retrieval on formalin-fixed liver, macrophages were stained using F4/80 (AbD Serotech, Oxford, UK), neutrophils by myeloperoxidase (Abcam, Cambridge, UK), and apoptotic cells using cleaved cytokeratin 18 (M30 antigen, Roche Applied Science, Indianapolis, IN). The number of positively stained cells was counted in at least four photomicrographs taken at 400× magnification. For F4/80 and MPO, data were normalized to the number of hepatocytes per field and are presented as number of cells per 300 hepatocytes counted. M30 data are presented as the percentage of cytoplasmic stained hepatocytes compared with total hepatocyte numbers.

Statistical analyses

Data are presented as mean ± SEM. Using SPSS v15.0, continuous data were analyzed by ANOVA, with Tukey post hoc testing. Histology scores were analyzed by the Kruskal Wallis test, and individual groups compared using Mann Whitney U test. For all comparisons, P < 0.05 was considered significant.


Reducing dietary fat and cholesterol intake alters pathological phenotype of fatty liver disease in foz/foz mice

Consistent with our earlier report, foz/foz mice fed HF-diet for 24 weeks developed steatohepatitis, with high serum ALT, ballooning degeneration of hepatocytes, moderate liver inflammation, and fibrosis (Table 1, Figure 1A and 1B). In subsequent intervention experiments, these animals form the positive controls of “fibrosing steatohepatitis.” To determine whether progression of liver pathology following initiation of NASH at 12 weeks (reported in [10]) depends on continued dietary factors, we discontinued HF-feeding in some mice, providing chow instead (dietary “switch”). At 24 weeks, serum ALT levels tended to be lower in mice with dietary restitution than in positive controls (Table 1), and there was a major reduction in steatosis (P < 0.005), and trend (not significant) in lobular inflammatory score, which contributed to a significantly lower NAFLD activity score (NAS) compared with HF-fed foz/foz mice at 24 weeks (P < 0.05; Table 1, Figure 1A).

Table 1. Effects of reduced dietary fat intake on steatohepatitis development in foz/foz mice
Genotype DietWT Chowfoz/foz ChowHigh fatHigh fat/chow
  1. Fibrosis area was determined in HF-fed and dietary switch foz/foz mice by morphometric analysis of Sirius red staining. n.d., not determined.
  2. aP < 0.05 compared with WT chow.
  3. bP < 0.05 compared with foz/foz chow
  4. cP < 0.05 compared with foz/foz HF.
Serum ALT (U/l)26 ± 1145 ± 38397 ± 43a285 ± 70a
Steatosis score00.8 ± 0.42.9 ± 0.1a, b1.6 ± 0.2a, c
Inflammation score00.8 ± 0.41.9 ± 0.3a1.4 ± 0.3a
Ballooning score01.0 ± 0.41.6 ± 0.2a1.4 ± 0.3a
NAFLD activity score02.6 ± 1.26.3 ± 0.5a, b4.4 ± 0.4a, c
Fibrosis arean.d.n.d.0.22 ± 0.070.07 ± 0.02c
Hepatic triglyceride (mg/mg protein)0.2 ± 0.010.5 ± 0.11.0 ± 0.1a, b0.7 ± 0.04a, c
Figure 1.

Fibrosing steatohepatitis develops in HF-fed foz/foz mice, but severity is attenuated by dietary switch. A: Hematoxylin and eosin (H&E) stained liver sections from foz/foz mice fed HF diet (23% fat, 0.19% cholesterol) for 24 weeks (left), or for 12 weeks followed by 12 weeks standard rodent chow (4.5% fat) (right). B: Sirius red stained liver sections (red color indicates collagen). Fibrosis severity is decreased in foz/foz mice switched from HF diet back to chow. 100× magnification. [Color figure can be viewed in the online issue, which is available at]

Reducing dietary fat and cholesterol intake improves metabolic abnormalities in foz/foz mice

In earlier work, we showed that HF-feeding is associated with early saturation of adipose tissue lipid storage in foz/foz mice, causing lipid to partition into liver; this process was associated with hyperinsulinemia, hyperglycemia, and hypoadiponectinemia [10]. Here, we examined whether the improvements in liver pathology achieved by dietary correction were related to effects on obesity or lipid partitioning. After 24 weeks, all foz/foz mice were heavier than WT controls (P < 0.05), and there was no significant difference in body weight between foz/foz mice fed the three dietary regimens (Table 2). Thus, reversion to a physiological diet had not effect on obesity in foz/foz mice. However, whereas marked hepatomegaly developed in HF-fed foz/foz mice, dietary switch after 12 weeks reduced liver mass by nearly 50% (P < 0.001), and this beneficial effect was associated with a nearly 50% increase in adiposity as determined by peri-ovarian adipose pad weight (P < 0.05, Table 2). Commensurate with this more physiological bodily lipid storage, fasting blood glucose decreased (P < 0.01) in association with apparently lower (but not normal) serum insulin (Table 2; the major difference in mean values was not significant, most likely because of the high variance of values in the HF-fed group). Likewise, hypercholesterolemia in HF-fed mice was corrected by dietary reversion to chow (P < 0.001, Table 2). Consistent with the expansion of adipose tissue, hyperleptinemia persisted after dietary switch, but serum adiponectin levels, which are markedly reduced in HF-fed foz/foz mice, increased twofold (P = 0.052, Table 2). Likewise, adipose tissue expression of adiponectin mRNA, which was suppressed by HF feeding in foz/foz mice, tended to increase after dietary switch, albeit changes were not significant (Table 2).

Table 2. Metabolic characteristics in foz/foz mice
DietChowChowHigh fatHigh fat/chow
  1. aP < 0.05 compared with WT chow.
  2. bP < 0.05 compared with foz/foz chow.
  3. cP < 0.05 compared with foz/foz HF.
  4. dP = 0.052 compared with foz/foz HF.
Body weight (g)30 ± 1.749 ± 5.9a63 ± 4.9a56 ± 1.6a
Liver weight (% body weight)4.7 ± 0.36.3 ± 1.213.3 ± 1.1a, b7.7 ± 0.7c
Peri-ovarian adipose weight (right side, % body weight)1.3 ± 0.33.2 ± 0.1a2.0 ± 0.3b2.9 ± 0.2a, c
Blood glucose (mM)5.2 ± 0.38.3 ± 1.0a10.3 ± 0.4a7.5 ± 0.3a, c
Serum insulin (nmol/l)0.08 ± 0.011.6 ± 1.17.2 ± 2.7a2.9 ± 1.2a
Serum adiponectin (μg/ml)6.5 ± 0.64.5 ± 0.81.6 ± 0.2a3.2 ± 0.4a, d
Adipose adiponectin mRNA (arbitrary units)1.0 ± 0.20.5 ± 0.20.1 ± 0.03a0.3 ± 0.1a
Adipose CD11b mRNA (arbitrary units)1.0 ± 0.24.4 ± 2.55.6 ± 1.6a6.9 ± 1.5a
Adipose CD11c mRNA (arbitrary units)1.0 ± 0.356 ± 40a179 ± 58a198 ± 54a, b
Adipose CD68 mRNA (arbitrary units)1.0 ± 0.316 ± 8.5a15 ± 4.8a11 ± 2.7a
Serum leptin (ng/ml)4 ± 183 ± 19a114 ± 12a96 ± 8a
Serum cholesterol (mM)1.6 ± 0.12.7 ± 0.56.8 ± 0.3a, b3.4 ± 0.4a, c
Serum triglyceride (mM)1.2 ± 0.11.2 ± 0.20.7 ± 0.10.9 ± 0.1

Effects of dietary fat and cholesterol on hepatic lipid accumulation in foz/foz mice

Consistent with the improvement in steatosis score, dietary switch affected ∼30% reduction in hepatic triglyceride content (P < 0.05, Table 1). Hepatic cholesterol esters, free cholesterol, and DAG levels also decreased significantly in mice switched from HF-diet to chow compared with those continued on HF diet, whereas hepatic FFA content was unaltered (Figure 2A). To clarify mechanisms of lipid accumulation in foz/foz mice with NASH versus steatosis alone, we assessed pathways of fatty acid uptake, as well as synthesis (lipogenesis) and catabolism (fatty acid oxidation). Hepatic levels of the fatty acid and cholesterol transport protein, CD36, increased markedly in HF-fed foz/foz mice compared with controls. Further, CD36 expression depended on dietary factors or the related humoral changes (in serum insulin and/or adiponectin), as levels were reduced in the switch group (P < 0.01, Figure 2B). Conversely, nuclear accumulation of the lipogenic transcription factors, sterol regulatory element binding protein-1 (SREBP1) and carbohydrate response element binding protein (ChREBP), was not altered in foz/foz mice by dietary changes (Figure 2C). Despite there being no changes in lipogenic transcription factor activation, fatty acid synthase (FAS) activity, the rate-limiting enzyme in de novo fatty acid synthesis, was increased in foz/foz mice (P < 0.005) and tended to be highest in mice switched from HF diet back to chow (Figure 2D). Transcripts for stearoyl Co-A desaturase (SCD1), which is rate-limiting for triglyceride formation, showed an expression pattern similar to FAS activity (Figure 2E), whereas hepatic mRNA levels of several fatty acid oxidation enzymes (mitochondrial carnitine palmitoyl transferase (CPT)-1, peroxisomal acyl CoA oxidase (ACOX)-1, microsomal Cyp4a14) were similar (or tended to be lower) between dietary switch and HF-fed foz/foz mice (Figure 2E). In summary, the dietary change which reduced hepatic triglyceride content appears to operate by suppressing a major fatty acid uptake pathway (CD36) without suppression of substrate-dependent lipogenesis or activation of fatty acid oxidation.

Figure 2.

Lipid turnover in steatohepatitis. A: Hepatic content of cholesterol esters, free cholesterol and diacylglyceride (DAG) are lowered switch from HF diet to chow. Hepatic content of B: Hepatic CD36 protein is increased in HF-fed foz/foz mice, but expression is attenuated with dietary switch. C: Nuclear protein levels of the lipogenic transcription factors sterol regulatory element binding protein (SREBP)1 and carbohydrate response element binding protein (ChREBP) were not significantly different between groups. D: FAS activity appeared higher in foz/foz compared with wild type mice, with highest activity in dietary switch group. E: Hepatic gene expression of stearoyl CoA desaturase (SCD)1, acyl CoA oxidase (ACOX)1, cytochrome P450 (Cyp)4a14, and carnitine palmitoyl transferase (CPT)1. *P < 0.05 compared with wild type chow; #P < 0.05 compared with foz/foz chow, P < 0.05 compared with foz/foz HF.

Dietary composition influences liver injury, ATP levels and necrotic cell death in foz/foz mice

Hepatocellular injury is the defining difference between steatohepatitis and steatosis [19], and injured hepatocytes are the source of serum ALT in liver disease. Thus, in the present experiments, we interpret the reduction in serum ALT following dietary restitution as reflecting reduced hepatocellular injury. ALT release is particularly prominent when cell death is by necrosis [20]. One factor contributing to cell death in steatohepatitis is mitochondrial energy; when cellular ATP levels fall, as has been observed in human livers with NASH [21], cell death signaling is likely to terminate in necrosis rather than in apoptosis [20]. In HF-fed foz/foz mice with NASH, hepatic ATP levels were considerably lower than in dietary and genetic controls (P < 0.005, Figure 3A). The return to chow feeding after 12 weeks of HF intake restored ATP levels almost completely, as compared with values in chow-fed foz/foz control liver (P < 0.05). A ready explanation for these changes in energy status of liver cells was evident in hepatic expression of uncoupling protein 2 (UCP-2) mRNA, which increased in HF-fed foz/foz mice, but whose expression was normalized by dietary switch (P < 0.05, Figure 3B).

Figure 3.

Effects of diet on cellular ATP and pattern of liver inflammation and injury in foz/foz mice. A: Hepatic ATP levels and B: mRNA for uncoupling protein (UCP)2. C: Quantification of immunohistochemistry staining of (C) neutrophils (see Figure 4A), (D) for hepatocellular apoptosis (M30) (see Figure 4B), and D: neutrophils (see Figure 4A) (E) macrophages (see Figure 4C). F: Hepatic gene expression of CD68, CD11b, and CD11c. *P < 0.05 compared with wild type chow; #P < 0.05 compared with foz/foz chow, P < 0.05 compared with foz/foz HF.

The contention that diet-induced expression of UCP-2 and the resultant decrease in cellular ATP levels contributes to necrotic cell death in steatohepatitis is further supported by our observations of neutrophil infiltration (myeloperoxidase positive cells) in HF-fed foz/foz mice with NASH (P < 0.001, Figures 3D and 4A). After dietary reversion to chow, neutrophil infiltration (a common feature of necrosis) was substantially suppressed (P < 0.005), despite the minimal change in liver inflammatory score (which is determined by the total number of inflammatory cells, not their phenotype; Table 1).

Figure 4.

Hepatic inflammation and apoptosis. Immunohistochemistry staining of liver sections from HF-fed foz/foz (left) and dietary switch (right) mice. A: Neutrophils (arrows) using myeloperoxidase. B: Hepatocyte apoptosis using M30 antibody for cleaved cytokeratin 18. Apoptotic cells show cytoplasmic staining (arrows). C: Macrophages using F4/80 (arrows). 400× magnification. [Color figure can be viewed in the online issue, which is available at]

Dietary composition influences regulation of apoptosis and pattern of liver inflammation in foz/foz mice, but not adipose inflammation

Apoptosis is also a feature of steatohepatitis [22]. In HF-fed foz/foz mice, hepatocyte apoptosis was greatly increased, as demonstrated by approximately tenfold increase in M30 immunohistochemistry (IHC)-positive cells; M30 is the caspase 3-cleaved product of cytokeratin 18, which is only found in hepatocytes (P < 0.05, Figures 3C and 4B). Given that apoptosis is a more organized form of cell death but requires energy, one might expect that when ATP levels are restored in fatty liver disease with incomplete removal of the inciting injury-inducing molecules (lipotoxic molecules and/or cytokines) [23, 24], apoptosis might actually be facilitated. This was indeed the case. Thus, hepatocellular apoptosis was further increased in foz/foz mice that underwent dietary switch (P < 0.05, Figures 3C and 4B).

As already intimated, necrotic cell death is associated neutrophil infiltration, which declined in livers of HF-fed foz/foz mice after dietary switch. Conversely, when ATP levels are sufficient and the predominant cell death process is apoptosis rather than necrosis, this is often accompanied by recruitment of macrophages to engulf apoptotic bodies released from dying cells. As shown in Figure 3E, macrophage numbers increased following the switch in diet from HF to chow, as indicated by accumulation of F4/80-positive cells (Figures 3E and 4C). However, mRNA analyses demonstrated a shift from a pro-inflammatory macrophage phenotype characterized by high expression of CD68, CD11b, and CD11c transcripts in HF-fed foz/foz mice, to a more “quiescent” phenotype in mice switched from HF diet back to chow (Figure 3F). In contrast to the diet-induced changes in liver macrophages, adipose tissue macrophage phenotype was pro-inflammatory in all dietary groups of foz/foz mice (Table 2).

Relationships between dietary lipids and hepatic NF-κB activation, expression of chemokines and other pro-inflammatory molecules

In addition to hepatocellular injury, inflammatory recruitment is a key difference between simple steatosis and NASH, leading in the latter to fibrosis, cirrhosis, and HCC. To clarify the regulators of liver inflammation in these mice, we analyzed NF-κB activation, as well as elaboration of chemokines and other NF-κB-regulated pro-inflammatory molecules. As previously reported [15], serum TNFα levels were not elevated in HF-fed foz/foz mice (data not shown), and neither were either hepatic or adipose tissue TNFα mRNA levels different between dietary groups (Figure 5A). On the other hand, we confirmed increased serum levels of monocyte chemoattractant protein (MCP)1 in all foz/foz mice, and dietary intervention failed to change MCP-1 serum levels (Figure 5B). In contrast, diet strongly influenced hepatic MCP1 mRNA levels, which were induced tenfold in HF-fed foz/foz mice, and markedly suppressed by dietary switch (P < 0.001, Figure 5C). An explanation for the discrepancy between serum and hepatic MCP1 levels was apparent from determination of MCP1 mRNA in adipose tissue, levels of which were significantly increased in HF-fed foz/foz mice; in contrast to the changes in liver, dietary reversion to chow failed to “switch off” this source of MCP1 (P < 0.01, Figure 5C). Thus, high serum MCP1 levels appear to be more directly related to adipose tissue inflammation than to changes in the liver.

Figure 5.

Mediators of inflammation. A: mRNA levels for TNFα in liver and adipose tissue. B: Serum MCP1 levels, and C: MCP-1 gene expression in liver and adipose tissue. D: Hepatic gene expression for neutrophil cytosolic factor (NCF)2, NCF4, intercellular adhesion molecule (ICAM), and vascular cell adhesion molecule (VCAM). E: Hepatic nuclear levels of nuclear factor (NF)-κB p65 protein. F: Hepatic protein levels of ICAM and VCAM. *P < 0.05 compared with wild type chow; #P < 0.05 compared with foz/foz chow, P < 0.05 compared with foz/foz HF.

In addition to MCP1, hepatic neutrophil cytosolic factor (NCF)2, NCF4, intercellular adhesion molecule (ICAM), and vascular adhesion molecule (VCAM) mRNA levels were all elevated in HF-fed foz/foz mice with steatohepatitis, and their expression was substantially attenuated by dietary switch (P < 0.001, Figure 5D). Consistent with this expression pattern, nuclear translocation of NF-κB p65 protein was increased in HF-fed foz/foz mice, and normalized with dietary switch (P < 0.005, Figure 5E). By western blot analyses, ICAM and VCAM protein levels in foz/foz mice switched from HF-diet back to chow were significantly lower than their HF-fed counterparts (P < 0.05, Figure 5F).

Dietary composition influences stellate cell activation, matrix metalloproteinase-9 expression, and development of hepatic fibrosis in foz/foz mice with NASH

After 24 weeks HF-feeding, foz/foz mice developed appreciable zone 3 perisinusoidal and periportal fibrosis. Switching foz/foz mice from HF diet to chow markedly attenuated hepatic fibrosis, as demonstrated by Sirius red staining and its quantification (P < 0.05, Figure 1B, Table 1). In addition, while HF-fed foz/foz mice exhibited centrizonal accentuation of collagen deposition or both central (zone 3) and portal (zone 1) fibrosis distribution, which typifies the most common pattern of fibrosis in human liver with NASH, foz/foz mice switched from HF-diet to chow showed less zone 3 accentuated fibrosis distribution with either equal zone 1 and zone 3 or portal accentuation (Figure 1B). Consistent with this attenuation of fibrosis, hepatic collagen α1 mRNA levels were markedly reduced in mice switched from HF-diet to chow in comparison with those continued on HF-diet (P < 0.05, Figure 6A). Similarly, hepatic α-smooth muscle actin protein, which reflects stellate cell activation, increased in HF-fed foz/foz mice and returned to normal expression levels after dietary restitution (P < 0.001, Figure 6B).

Figure 6.

Fibrosis severity is attenuated by dietary switch. A: HF-fed foz/foz mice have increased hepatic collagen α1 mRNA, and B: α-smooth muscle actin protein levels. C: Hepatic gene expression of transforming growth factor (TGF)β, platelet-derived growth factor (PDGF)-A and PDGF-B are also increased. D: Hepatic protein levels of matrix metalloproteinase (MMP)2, MMP9, tissue inhibitor of MMP (TIMP)1, and TIMP2 are also altered in HF-fed foz/foz mice with fibrosis. *P < 0.05 compared with wild type chow; #P < 0.05 compared with foz/foz chow, P < 0.05 compared with foz/foz HF.

To explore mechanisms by which dietary switch attenuated fibrogenesis in foz/foz mice, we first measured expression of growth factors implicated in stellate cell activation and matrix deposition. Hepatic mRNA levels of transforming growth factor (TGF)β1 increased approximately twofold in HF-fed foz/foz mice and were significantly reduced by reversion to chow (P < 0.001, Figure 6C). Similarly, hepatic expression of platelet-derived growth factor (PDGF)-A and PDGF-B, which activate stellate cells, increased with HF-feeding (P < 0.01), and dietary intervention reduced values similar to chow-fed foz/foz mice (P < 0.05, Figure 6C). While hepatic MMP2 protein did not increase significantly with HF-feeding in foz/foz mice, protein expression of MMP9 was greatly enhanced (Figure 6D). Importantly, expression levels of both proteins were suppressed, returning to normal in the case of MMP9, in foz/foz mice switched from HF diet back to chow compared with HF-fed mice (P < 0.05, Figure 6D). In contrast, TIMP1 and TIMP2 proteins were significantly suppressed in HF-fed foz/foz mice (P < 0.05, Figure 6E), but less so in mice switched from HF diet to chow.


Important unanswered questions remain about the relative importance of obesity and the factors that contribute to it, such as dietary composition and physical activity, and particularly their pathogenic relevance to the phenotype of fatty liver disease that occurs in most obese humans. In the present experiments, we specifically addressed the issue of dietary composition by first feeding foz/foz mice a HF-diet for 12 weeks to initiate onset of steatohepatitis, then either continuing the HF-diet or reverting (“switching” back) to rodent chow for a further 12 weeks. The results show clear relationships between diet and hepatic/adipose lipid partitioning, an important connection between hepatic energy status and type of cell death, pathways of inflammatory recruitment and resultant cellular phenotype in steatohepatitis, and provide novel insights into how a simple intervention of “nutritional restitution” might alter fibrosis progression and other elements of NASH pathology without resolving either adipose inflammation or on-going (but less fibrotic) liver injury.

Previous work showed that hepatic triglyceride accumulation occurs within 2 weeks in HF-fed foz/foz mice in association with increased FAS activity, whereas enhanced fatty acid uptake (via CD36) occurs later. The present results clearly demonstrate that pathways of de novo lipid synthesis or disposal (fatty acid oxidation) do not account for the substantial decrease in hepatic lipid content achieved by dietary correction. In contrast, CD36 expression was decreased by switching HF diet to chow. This important observation provides a molecular basis for the dynamic role of hepatic lipid uptake in NASH, which had earlier been inferred by tracer studies in obese humans with NASH [11]. To clarify the role of CD36 in NASH cellular localization studies should be performed, as CD36 can contribute to steatohepatitis pathogenesis not only by facilitating lipid-uptake by hepatocytes (both FFA and cholesterol), but also by mediating inflammation through its expression on macrophages [25]. Dietary restitution did not alter body weight in these obese mice, but did expanded adipose tissue, and as anticipated by our earlier observation of adipose restriction and the seminal study by Kim et al., such adipose expansion reversed diabetes and partly restored serum adiponectin levels [26]. Interestingly, adipose tissue retained a pro-inflammatory phenotype despite the apparent restoration of physiological whole-body lipid partitioning, and as suggested by other work, this clearly comprises one source of continuing inflammation in insulin resistance, metabolic syndrome, and NASH [15, 27, 28].

The contribution of individual hepatic lipid molecules to liver injury and inflammation is now a central focus of NASH research [1, 2, 23]. In vivo studies have highlighted a possible mechanistic role of FFA and/or lysophosphatidyl choline in hepatocyte lipotoxicity [29, 30]. Other work has emphasized the importance of hepatic cholesterol accumulation in both human NASH [31] and animal models [32, 33]. In the present studies, reversion to chow decreased hepatic cholesterol esters and free cholesterol, as well as DAG levels, but failed to alter total FFA levels. In another dietary intervention which improved NASH pathology, we showed removal of cholesterol alone from HF diet lowers cholesterol fractions but similarly failed to alter hepatic saturated or total FFA levels, and DAG levels did not correlate with steatohepatitis severity [32]. These findings indicate that removal of excessive dietary lipids and/or simple sugars decreases hepatic lipotoxic stimuli, and the lipidomic analyses provide further evidence that the lipid species which are toxic in NASH are unlikely to be FFA (total levels of which remained unaltered after switch) but more likely related to one or more cholesterol metabolites (free cholesterol or oxysterols) or DAG.

One of the most novel findings of the present studies is that switching animals from an atherogenic diet back to rodent chow causes a marked difference in the molecular signature of inflammation in the liver. HF-feeding of foz/foz mice increases nuclear translocation of NF-κB p65 protein, with resultant increased expression of NF-κB target genes. This activation of NF-κB appeared to be independent of TNFα (which did not change after dietary switch). Thus, NF-κB appears to be a key mediator of hepatic chemokine expression and inflammatory recruitment in HF-fed foz/foz mice, as shown previously in the MCD model [5]. On the other hand, while the role of TNFα in steatohepatitis has been controversial [5, 6, 34], its continued presence after dietary restitution provides one possible mechanism for operation of hepatocyte apoptosis in this group. Others have shown that hepatocytes loaded with cholesterol are particularly sensitive to cytokine-mediated apoptosis by a mitochondrial pathway [34].

HF-feeding induced UCP2 and resultant hepatic ATP depletion were associated with higher neutrophil numbers. Conversely, dietary switch decreased neutrophil infiltration in association with lowering serum ALT. An apparent paradox was the associated increase in hepatocyte apoptosis. The shift from necrotic cell death to apoptosis can be at least partly explained by the restitution of hepatic ATP levels, a critical arbiter of whether cell death signaling terminates in apoptosis (which requires energy for its execution) or necrosis [20]. In addition, macrophages are often recruited to the liver during recovery from injury and are known to play a role in engulfing apoptotic bodies [35] and in resolution of extracellular matrix [36]. Therefore, persistent macrophage infiltration during improvement of steatosis and reduction of ALT level may not represent a tissue-damaging process. Rather, it may represent an appropriate tissue response to limit the amount of pro-inflammatory material within the liver. Further evidence to support macrophages playing a remodeling role in livers of mice switched from HF-diet to chow is the change in macrophage phenotype. In liver, active (recruited/pro-inflammatory) macrophages express high levels of CD11b, CD11c, and CD68, and this expression profile was observed in HF-fed foz/foz mice. Switching HF-fed mice back to chow reduced mRNA levels of CD11b, CD11c, and CD68 despite persistently high numbers of F4/80 positive cells in the liver. This indicates a change in macrophage phenotype (CD11c and CD68) as well as a decrease in neutrophils (CD11b), as additionally shown by the lower number of myeloperoxidase-positive cells.

Collagen-α1 expression is already elevated after 12 weeks HF-feeding in foz/foz mice and values are further increased by 24 weeks HF-feeding [10]. In the dietary switch group, minimal fibrosis was observed, and collagen-α1 mRNA levels were similar to those after only 12 weeks HF-feeding (C. Larter and G. Farrell, unpublished observation). The enhanced fibrogenesis at 24 weeks was also associated with increased expression of fibrogenic mediators, and dietary correction reduced expression of these growth factors, as well as NF-κB activation. There was also a shift from central (zone 3) accentuated to zone 1 attenuated fibrosis in the dietary switch group, and this is similar to observations made following surgical intervention in humans with NASH, which similarly reduces liver fibrosis [37]. These findings indicate that diet may influence fibrogenesis, either directly or through its beneficial effects on a necrotic pattern of hepatocellular injury and the phenotype of hepatic inflammation (M2 versus activated macrophages, suppression of neutrophil infiltration). Lastly, as adiponectin antagonizes the pro-fibrogenic effects of leptin [38], the partial restoration of adiponectin levels may also contribute to the attenuation of fibrosis achieved by dietary restitution.

In summary, switching from a HF diet to chow in foz/foz mice with metabolic syndrome and NASH alters whole-body lipid partitioning, favoring triglyceride storage in adipose tissue rather than liver. This redistribution of lipid, coupled with decreased hepatic fatty acid uptake, reduced steatosis, and hepatic cholesterol stores, changed the pattern of liver injury, with decreased neutrophil numbers, increased apoptosis, and a decrease in macrophage activation. The reversal of NF-κB activation and decreased chemokine and growth factor expression in these mice was associated with attenuation of fibrogenesis. Increased apoptosis and altered macrophage phenotype may represent a remodeling of damaged liver leading to its recovery by regeneration of nonsteatotic hepatocytes, suppression of matrix deposition, and stimulation of matrix resolution. However, a key observation was that correction of dietary composition does not alone reduce obesity (at least in this line of mice), and that adipose inflammation, with possible secondary effects of cytokines on the liver, remains. In conclusion, this study supports an interactive role for dietary and metabolic factors in promoting liver injury, inflammation, and fibrogenesis in obesity-related NAFLD that has mechanistic implications for therapeutic intervention in NASH.


The authors thank Jacqueline Williams, Matthew Clyne, and Leah Bala for their technical assistance. The research was funded by Australian National Health and Medical Research Council (NHMRC) project grant 418101, CZL by NHMRC Fellowship 525473, and DVR by NHMRC Scholarship 585539.