Rosiglitazone Requires Hepatocyte PPARγ Expression to Promote Steatosis in Male Mice With Diet-Induced Obesity

Abstract

Thiazolidinediones (TZD) are peroxisome proliferator-activated receptor γ (PPARγ) agonists that may reduce hepatic steatosis through their effects in adipose tissue and therefore have been assessed as potential therapies to treat nonalcoholic fatty liver disease (NAFLD) in humans. However, some studies suggest that expression and activation of hepatocyte PPARγ promotes steatosis and that would limit the benefits of TZD as a NAFLD therapy. To further explore this possibility, we examined the impact of short-term rosiglitazone maleate treatment after the development of moderate or severe diet-induced obesity, in both control and adult-onset hepatocyte-specific PPARγ knockout (Pparg^ΔHep) mice. Independent of the level of obesity and hepatic PPARγ expression, the TZD treatment enhanced insulin sensitivity, associated with an increase in white adipose tissue (WAT) fat accumulation, consistent with clinical observations. However, TZD treatment increased hepatic triglyceride content only in control mice with severe obesity. Under these conditions, Pparg^ΔHep reduced diet-induced steatosis and prevented the steatogenic effects of short-term TZD treatment. In these mice, subcutaneous WAT was enlarged and associated with increased levels of adiponectin, while hepatic levels of phosphorylated adenosine 5′-monophosphate–activated protein kinase were also increased. In addition, in mice with severe obesity, the expression of hepatic Cd36, Cidea, Cidec, Fabp4, Fasn, and Scd-1 was increased by TZD in a PPARγ-dependent manner. Taken together, these results demonstrate that hepatocyte PPARγ expression offsets the antisteatogenic actions of TZD in mice with severe obesity. Therefore, in obese and insulin resistant humans, TZD-mediated activation of hepatocyte PPARγ may limit the therapeutic potential of TZD to treat NAFLD.

Nonalcoholic fatty liver disease (NAFLD) is the main cause of chronic liver disease, with an estimated prevalence of 25% in the general population (1). The main feature of NAFLD is hepatic steatosis, which is the accumulation of triglycerides (TG) in more than 5% of the parenchyma. Although steatosis is considered a benign condition, it can progress into nonalcoholic steatohepatitis (NASH), which is an advanced stage of NAFLD associated with liver injury, inflammation, and hepatocyte ballooning with or without fibrosis (2). Of note, the onset of steatosis and the subsequent progression to NASH is closely related with obesity and type 2 diabetes in humans. Since the rate of obesity continues increasing in the Western population, NAFLD cases are expected to rise and become a serious burden on our healthcare system (1,3).

While the etiology of NAFLD is multifactorial (2), chronic consumption of fat and carbohydrate-rich foods, and the lack of physical activity in individuals with obesity and type 2 diabetes can significantly contribute to the development of NAFLD. Therefore, lifestyle modifications to reduce caloric intake and increase physical activity, combined with drugs that reduce insulin resistance and improve peripheral metabolism are considered potential NAFLD therapies (4-6). Specifically, thiazolidinediones (TZD), a class of heterocyclic compounds that selectively activate peroxisome proliferator-activated receptor γ (PPARγ) and are approved to treat type 2 diabetes, may also be effective in treating NAFLD (7). PPARγ belongs to a family of nuclear receptors that are expressed in multiple tissues where they regulate different metabolic processes (8). There are 2 isoforms of PPARγ, PPARγ1 and PPARγ2. PPARγ1 has 30 amino acids less in the N-terminal domain than PPARγ2, but both isoforms have the same DNA binding domain, and both can be activated by rosiglitazone and pioglitazone (Food and Drug Administration–approved TZD to treat type 2 diabetes). The activation of PPARγ with TZD in muscle and adipose tissue improves insulin sensitivity, glucose disposal, and reduces the levels of lipids in the circulation (7-9). Based on these positive actions of TZD, long-term trials were conducted to examine the efficacy of TZD in NAFLD patients (5,6). Briefly, the meta-analysis of trials <24 months indicate that TZD reduce steatosis in NAFLD patients (10-12). However, the extension of the TZD treatment >24 months does not correlate with a progressive reduction of steatosis (5,13) and is speculated to be due to the “exhaustion” of the effects of TZD on insulin sensitivity (5). In general, although TZD have been used in different trials with promising effects, the limited efficacy, coupled with potential side effects, dampens the use of TZD to treat NAFLD.

The therapeutic effects of TZD to reduce steatosis may be offset by their actions on hepatocyte function. To date, the direct and indirect effects of TZD on the liver remain controversial (8,14,15). In fact, it remains to be determined if TZD affect liver steatosis directly by activating hepatocyte PPARγ in vivo in a model of established diet-induced obesity, which is an experimental condition that mimics NAFLD, insulin resistance, and obesity. This is a relevant question because the expression of hepatic PPARγ is increased in NAFLD patients (16,17), and in mice with diet-induced obesity (18,19). To better understand the specific role that hepatocyte PPARγ plays in the response to TZDs, we have used an adult-onset, hepatocyte-specific PPARγ knockout mouse model (Pparg^ΔHep) to determine whether the direct effects of TZD on hepatocytes reduce their therapeutic effects on the liver, in a PPARγ-dependent fashion, after diet-induced obesity and insulin resistance are established.

Methods

Mouse Studies

These studies were approved by the Institutional Animal Care and Use Committee of the University of Illinois at Chicago. PPARγ floxed breeders in a C57Bl/6J background were originally purchased from Jackson Laboratories [Pparg^fl/fl, Stock # 004584, Jackson Laboratories, Bar Harbor, ME, USA (20)] and maintained as a homozygous breeding colony, with standard chow diet and a 14-h light/10-h dark cycle (lights on at 6:00 am). At 4 to 6 weeks of age, a group of chow-fed Pparg^fl/fl mice were switched to a low-fat diet (LFD) containing 10% kcal from fat (D12450H, D12450J, Research Diets, New Brunswick, NJ, USA) or a high-fat diet (HFD) containing 45% or 60% kcal from fat (D12451 (HFD 45%) or D12492 (HFD 60%) (Research Diets) and fed for 7 weeks. Another subgroup of Pparg^fl/fl male mice were switched to the LFD or HFD 60% at 10 weeks of age and fed for 16 weeks. After this period of feeding, adult-onset hepatocyte-specific PPARγ knockout (Pparg^ΔHep) mice were generated by injecting adeno-associated virus serotype 8 (AAV8) vectors that contain a thyroxine-binding globulin-driven (TBGp) Cre recombinase (AAV8-TBGp-Cre, Penn Vector Core, University of Pennsylvania, Philadelphia, PA, USA, and Addgene, Watertown, MA, USA) as previously described (19,21,22). Pparg^fl/fl littermates were injected with AAV8-TBGp-null to generate control mice. One week later, half of the control and Pparg^ΔHep mice fed a HFD were then switched to a HFD with the same percentage of kcal from fat and supplemented with 70 mg of rosiglitazone maleate/kg of diet (HFD/TZD, Cat# D1711502, D18061406, Research Diets) for 5 or 6 additional weeks. Rosiglitazone maleate was purchased from AdipoGen Life Sciences (San Diego, CA, USA). Diet composition is provided in Table 1.

Table 1.

Composition of the diets used in this study

	LFD		HFD (45% Kcal from fat), %		HFD/TZD (45% Kcal from fat), %		LFD		HFD (60% Kcal from fat), %		HFD/TZD (60% Kcal from fat), %
	D12450H		D12451		D1711502		D12450J		D12492		D18061406
	g	kcal	g	kcal	g	kcal	g	kcal	g	kcal	g	kcal
Protein	19	20	24	20	24	20	19	20	26	20	26	20
Carbohydrate	67	70	41	35	41	35	67	70	26	20	26	20
Fat	4	10	24	45	24	45	4	10	35	60	35	60
Total		100		100		100		100		100		100
kcal/gm	3.8		4.7		4.7		3.8		5.2		5.2
Ingredients
Casein	200	800	200	800	200	800	200	800	200	800	200	800
L-cystine	3	12	3	12	3	12	3	12	3	12	3	12
Corn starch	452.2	1809	72.8	291	72.8	291	506.2	2025	0	0	0	0
Maltodextrin 10	75	300	100	400	100	400	125	500	125	500	125	500
Sucrose	172.8	691	172.8	691	172.8	691	68.8	275	68.8	275	68.8	275
Cellulose, BW200	50	0	50	0	50	0	50	0	50	0	50	0
Soybean oil	25	225	25	225	25	225	25	225	25	225	25	225
Lard	20	180	177.5	1598	177.5	1598	20	180	245	2205	245	2205
Mineral mix S10026	10	0	10	0	10	0	10	0	10	0	10	0
DiCalcium phosphate	13	0	13	0	13	0	13	0	13	0	13	0
Calcium carbonate	5.5	0	5.5	0	5.5	0	5.5	0	5.5	0	5.5	0
Potassium citrate 1H2O	16.5	0	16.5	0	16.5	0	16.5	0	16.5	0	16.5	0
Vit Mix V100001	10	40	10	40	10	40	10	40	10	40	10	40
Choline bitartrate	2	0	2	0	2	0	2	0	2	0	2	0
Cholesterol	0.054		0.167		0.167		0.054		0.216		0.216
Rosiglitazone maleate	0		0		0.0601		0		0		0.0542
Total	1055.05	4057	858.15	4057	858.21	4057	1055.05	4057	773.85	4057	773.85	4057

Abbreviations: HFD, high-fat diet; LFD, low-fat diet; TZD, thiazolidinedione.

One week after injection of AAV vectors in the groups of mice that were fed a LFD or HFD 60% for 16 weeks, whole body nuclear magnetic resonance–based fat and lean mass was measured with a minispec LF50 Body Composition Analyzer (Bruker, Billerica, MA, USA) every 2 weeks. In all groups of mice, an insulin tolerance test (ITT; 1-1.5 mU insulin ip/g) was performed 4 h after food withdrawal at 7:00 am, 5-6 weeks after injection of AAV vectors. Blood glucose was measured through lateral tail vein bleeds with a glucometer (Accu-check, Roche) over a 2-h timespan, as described previously (23). Mice were euthanized by decapitation, after a 4-h food withdrawal at 7:00 am, and blood was collected into EDTA-coated microtainers to obtain plasma after centrifugation. Tissues were weighed, fixed in formalin, or snap-frozen in liquid nitrogen for further analysis.

Plasma Endpoints

Plasma TG, nonesterified fatty acids, cholesterol, ketone bodies (Fujifilm Wako Diagnostics, Richmond, VA, USA), alanine transaminase (ALT) and aspartate aminotransferase (AST) (Pointe Scientific, Canton, MI, USA) were measured with colorimetric assays. Plasma insulin (Mercodia, Uppsala, Sweden) (24) and adiponectin (Abcam, Cambridge, MA) (25) were measured with commercial enzyme-linked immunosorbent assay kits. All kits were used according to the manufacturer’s instructions and measured using a Synergy H1 Hybrid plate Reader (BioTek, Winooski, VT, USA).

Liver TG Content

Small pieces of frozen livers were homogenized in isopropanol, incubated at room temperature while shaken for 45 min, followed by a 10-min centrifugation to collect cellular debris. The supernatant was then used to measure TG levels as described previously (26).

Hepatic Gene Expression

RNA was isolated with Trizol Reagent (Life Technologies, Carlsbad, CA, USA). Total RNA was treated with RNAse-free DNAse (Promega, Madison, WI, USA), and retrotranscribed to complimentary (cDNA) with First Strand cDNA Synthesis Kit (Thermo Scientific, Waltham, MA, USA). cDNA samples were used to measure gene expression using gene-specific primers provided in Table 2, and Brilliant III Ultra-Fast QPCR Master mix (Agilent Technologies, Santa Clara, CA, USA), as described previously (26).

Table 2.

Sequence of primers used for qPCR in this study

Name	NM_ID	Sense	Antisense
Pparg	NM_001127330	AGACCACTCGCATTCCTTTG	CCTGTTGTAGAGCTGGGTCTTT
Cd36	NM_001159558	GGAGCCATCTTTGAGCCTTC	TGGATCTTTGTAACCCCACAAG
Cidea	NM_007702	GCAGCCTGCAGGAACTTATC	TCATGAAATGCGTGTTGTCC
Cidec	NM_178373	AAGATGGCACAATCGTGGAG	TTAGTTGGCTTCTGGGAAAGG
Fabp4	NM_024406	AAGAAGTGGGAGTGGGCTTT	GCTCTTCACCTTCCTGTCGT
Mogat1	NM_026713	TCTGGTTCTGTTTCCCGTTG	ACATTGCCACCTCCATCCTT
Acc1	NM_133360	ATCCTGCGAACCTGGATTCT	CCCACCAGAGAAACCTCTCC
Fasn	NM_007988	TGAGCACACTGCTGGTGAAC	CAGGTTCGGAATGCTATCCA
Scd1	NM_009127	ATCGCCCCTACGACAAGAAC	GTTGATGTGCCAGCGGTACT
Cpt1a	NM_013495	ACAACAACGGCAGAGCAGAG	GGACACCACATAGAGGCAGAAG
Hmgsc2	NM_008256	TCAGGGGTCTAAAGCTGGAA	TAAGCCTGAGCCGTAGGAGA
Beta actin	NM_007393	CTGGGACGACATGGAGAAGA	ACCAGAGGCATACAGGGACA
Peptidylprolyl isomerase A	NM_008907	TGGTCTTTGGGAAGGTGAAAG	TGTCCACAGTCGGAAATGGT

Abbreviation: qPCR, quantitative polymerase chain reaction.

Western Blots

Proteins were obtained from frozen livers, separated in 26-well Criterion TGX gels and transferred onto nitrocellulose membranes as previously described (21,23). The membranes were incubated overnight at 4°C with either adenosine monophosphate–activated protein kinase C (AMPK) #D5A2 (27) or phosphorylated AMPK T172 #D4D6D (28), and PPARγ #C26H12 (29) (Cell Signalling Technologies, Danvers, MA, USA) at a dilution 1/1000. Goat antirabbit immunoglobulin G (H+L)-horseradish peroxidase conjugate [#172-1019, 1/2000 for 1 h (30); Bio-Rad Laboratories, Hercules, CA, USA] was used as secondary antibody, and the specific signal was detected and quantified as previously reported (21,23).

Histology and Immunohistochemistry

The Research Histology Core at the University of Illinois at Chicago paraffin-embedded and stained (hematoxylin and eosin) 5-µm liver sections. The Human Tissue Resource Center of the University of Chicago paraffin-embedded and generated 5-µm subcutaneous adipose tissue sections. Deparaffined sections of the adipose tissue were blocked with 10% normal donkey serum in a humidity chamber for 20 min at room temperature and incubated with caveolin-1 antibody [#ab2910, 1:200, Abcam (31)] at 4°C. Then, the sections were incubated with 488 goat antirabbit [#4412S, 1:1000 (32); Cell Signaling Technologies] for 1 h. After multiple wash steps with phosphate buffered saline, slides were counterstained with 4′,6-diamidino-2-phenylindole and mounted with DAKO fluorescent mounting medium. Images of the stained sections were taken with a Leica DMi8 microscope with a DFC3000 (fluorescent images) or DMC5400 (hematoxylin and eosin) camera (Leica Microsystems, Wetzlar, Germany). To measure cell frequency and cell size of the subcutaneous adipose tissue (33), images were processed with Cellprofiler image analysis software [Broad Institute (34)].

Statistical Analysis

Data are represented as means ± standard error of the mean. Data were analyzed by 2-way analyses of variance followed by Tukey or Bonferroni post-hoc test. In each group, 2 different analysis were performed to test the effect of HFD and Pparg^ΔHep between LFD- and HFD-fed mice and the effect of TZD and Pparg^ΔHep in HFD-fed mice. The statistical analysis was performed with GraphPad Prism 8 (GraphPad Software, La Jolla, CA, USA). In the tables and figures, daggers (†) and double daggers (‡) indicate significant differences between LFD- and HFD-fed control and Pparg^ΔHep mice, respectively. Letters a to d indicate significant differences induced by TZD in HFD-fed mice. Asterisks (*) indicated significant differences induced by Pparg^ΔHep. P-value < 0.05 is considered significant.

Results

The Obesogenic Effects of Rosiglitazone Are Dependent on the Level of Obesity and Independent of Hepatocyte PPARγ Expression in Male Mice

We tested whether TZD (rosiglitazone maleate) increased insulin sensitivity and obesity in a hepatocyte-specific PPARγ-dependent manner in different groups of diet-induced obese mice. In 2 groups of mice in which HFD feeding was started at 4 to 6 weeks of age, 7 weeks of HFD modestly increased body weight above LFD: 4 g (HFD 45%) and 3.1 g (HFD 60%). Whereas 10 weeks of HFD 60%, starting at 10 weeks of age, increased body weight above LFD by 11.25 g (Fig. 1A-1C). After these periods of HFD feeding, control and Pparg^ΔHep mice were generated (Fig. 1A-1C, AAV arrows), and 1 week later half of the mice in each group were maintained on the HFD or switched to the HFD/TZD diet for an additional 5 to 6 weeks (Fig. 1A-1C, TZD arrows). Pparg^ΔHep did not affect weight gain in all diet groups (Fig. 1D-1F). Consistent with the actions of TZD in humans (7), TZD had an overall obesogenic effect [Effect of TZD, P-value = 0.0024; HFD 45% Fig. 1D)], 0.033 [HFD 60% (Fig. 1E)], and <0.0001 [HFD 60% (Fig. 1F)]. Similarly, TZD increased subcutaneous fat weight [Effect of TZD, P-value = 0.001; HFD 45% (Fig. 1D)], 0.075 (HFD 60%) (Fig. 1E), and <0.0001 (HFD 60%) (Fig. 1F)] and brown adipose tissue (BAT) in all groups (Table 3). However, the most dramatic impact of TZD was observed in mice that were fed HFD 60% for 16 weeks before the TZD treatment (Fig. 1F). In this group, TZD had an overall significant impact on the sum of WAT (<0.0001) (Table 3) and increased nuclear magnetic resonance–based fat mass but not lean mass (Fig. 1G). The obesogenic effect of TZD may be due in part to an increase of food intake (not measured in this study) as previously reported in mice with established diet-induced obesity (35). Of note, Pparg^ΔHep enhanced the impact of HFD/TZD on subcutaneous fat weight only in the group of mice in which HFD was started at 10 weeks of age (Fig. 1F), indicative of a stronger adipogenic effect of TZD on this group of mice.

Figure 1.

Loss of hepatocyte peroxisome proliferator-activated receptor γ (PPARγ) expression did not have a major effect in diet-induced obesity or thiazolidinedione (TZD)-mediated adiposity. Body weight curves of 4- to 6-week-old Pparg^fl/fl mice were fed a HFD containing 45% kcal from fat (A) or 60% kcal from fat (B) for 7 weeks or a high-fat diet (HFD) containing 60% kcal from fat (C) for 16 weeks before being treated with adeno-associated virus (AAV; AAV arrow) to generate control (C) or Pparg^ΔHep knockout (KO) mice. A week later, half of the mice fed a HFD in (A) and (B) were switched to a HFD/TZD for 5 weeks (TZD arrow), and mice in (C) were switched to a HFD/TZD for 6 weeks (TZD arrow). Body weight and subcutaneous fat (SC-fat) weight at the end of the study in mice fed a HFD containing 45% kcal from fat for 13 weeks (D), 60% kcal from fat for 13 weeks (E), or 60% kcal from fat for 23 weeks (F). Dotted lines in (D)-(F) show average weight of mice fed a HFD or HFD/TZD at the end of the study. Nuclear magnetic resonance–based fat mass and lean mass (G) in mice fed a HFD containing 60% kcal from fat. † and ‡ indicate significant differences induced by HFD as compared as low-fat diet (LFD)-fed mice within group: †differences in control (C) mice; ‡differences in KO mice. Letters a to d indicate significant differences induced by HFD/TZD as compared to HFD-fed mice within group. Asterisks indicate significant differences induced by Pparg^ΔHep as compared to control mice within diet. a,†,‡ P < 0.05; **, P < 0.01; c, †††,‡‡‡ P < 0.001; d, ††††,‡‡‡‡, P < 0.0001. n = 6-10 mice/group.

Table 3.

Weight of white adipose tissue and brown adipose tissue in the different groups of mice used in this study

	LFD		HFD		HFD/TZD
WAT (g)	Control	PpargΔHep	Control	PpargΔHep	Control	PpargΔHep
Cohort 1 (HFD 45%)	1.50 ± 0.15	1.50 ± 0.16	2.59† ± 0.39	2.75‡‡ ± 0.19	2.71 ± 0.25	3.83 ± 0.51
Cohort 2 (HFD 60%)	2.07 ± 0.4	2.55 ± 0.35	4.95† ± 0.78	4.93‡ ± 0.7	6.16 ± 0.28	5.71 ± 0.82
Cohort 3 (HFD 60%)	1.88 ± 0.23	1.57 ± 0.23	4.73††† ± 0.22	5.01‡‡‡‡ ± 0.26	6.59c ± 0.22	7.33d ± 0.35
BAT (g)	Control	Pparg ΔHep	Control	Pparg ΔHep	Control	Pparg ΔHep
Cohort 1 (HFD 45%)	0.10 ± 0.01	0.10 ± 0.01	0.11 ± 0.01	0.10 ± 0.01	0.18b ± 0.01	0.21d ± 0.01
Cohort 2 (HFD 60%)	0.17 ± 0.02	0.15 ± 0.02	0.22 ± 0.03	0.21 ± 0.03	0.34 ± 0.04	0.39a ± 0.04
Cohort 3 (HFD 60%)	0.16 ± 0.02	0.15 ± 0.01	0.38†††† ± 0.03	0.345‡‡‡‡ ± 0.03	0.70b ± 0.05	0.92d ± 0.11

WAT is the sum of urogenital, mesenteric, perirenal and inguinal (subcutaneous) adipose tissue. Cohort 1: mice fed a LFD or HFD containing 45% kcal from fat for 13 weeks and half switched to HFD/TZD in the last 5 weeks. Cohort 2: mice fed a LFD or HFD 60% kcal from fat for 13 weeks and half switched to HFD/TZD in the last 5 weeks. Cohort 3: mice fed a LFD or HFD 60% kcal from fat for 23 weeks and half switched to HFD/TZD in the last 6 weeks. † and ‡ indicate significant differences induced by HFD as compared as LFD-fed mice within group: †Differences in control mice; ‡Differences in knockout mice. Letters a to d indicate significant differences induced by HFD/TZD as compared to HFD-fed mice within group. a,†,‡ P < 0.05; **, P < 0.01; c,†††,‡‡‡ P < 0.001; d,††††,‡‡‡‡, P < 0.0001 (n = 6-10 mice/group).

Abbreviations: BAT, brown adipose tissue; HFD, high-fat diet; LFD, low-fat diet; TZD, thiazolidinedione; WAT, white adipose tissue.

HFD impaired insulin sensitivity in all diet groups (Fig. 2). Consistent with the well-known actions of TZD as insulin sensitizers (7), there was an overall effect of HFD/TZD on the response to ITT and the levels of insulin (Fig. 2A, 2C, 2F, and 2H). However, there was minimal effects of TZD on basal blood glucose and plasma lipids in control mice (Table 4). Interestingly, the TZD-mediated reduction in plasma insulin in Pparg^ΔHep mice, with prolonged 60% HFD feeding, was not associated with an improved response to the ITT (Fig. 2H, and 2I). This suggests that hepatocyte PPARγ may contribute to the insulin-sensitizing effects of TZD in obese mice. In fact, this negative effect of Pparg^ΔHep on insulin sensitivity was previously reported in muscle and adipose tissue of ob/ob mice with hepatocyte PPARγ knockout (36).

Figure 2.

Loss of hepatocyte peroxisome proliferator-activated receptor γ (PPARγ) expression did not have a major effect in the insulin sensitizing effects of thiazolidinedione (TZD). Insulin tolerance tests (ITT) performed in control (C) or Pparg^ΔHep knockout (KO) mice fed a high-fat diet (HFD) containing 45% kcal from fat (A), 60% kcal from fat (B) for 11 weeks, or a HFD containing 60% kcal from fat (C) for 22 weeks. Area under the curve (AUC) of the ITT and plasma insulin of mice fed a low-fat diet (LFD), a HFD containing 45% kcal from fat (D and E), 60% kcal from fat for 13 weeks (F and G), or a HFD containing 60% kcal from fat for 23 weeks (H and I). † and ‡ indicate significant differences induced by HFD as compared as LFD-fed mice within group: †differences in control (C) mice; ‡differences in KO mice. Letters a to d indicate significant differences induced by HFD/TZD as compared to HFD-fed mice within group. Asterisks indicate significant differences induced by Pparg^ΔHep as compared to control mice within diet. *,a,†,‡ P < 0.05; b, P < 0.01; c,†††,‡‡‡ P < 0.001; ‡‡‡‡, P < 0.0001. (n = 6-10 mice/group).

Table 4.

Plasma endpoints analyzed in this study

	LFD		HFD		HFD/TZD
	Control	PpargΔHep	Control	PpargΔHep	Control	PpargΔHep
Glucose, mg/dL
Cohort 1 (HFD 45%)	192.0 ± 9.8	172.0 ± 17.0	185.5 ± 11.1	188.5 ± 8.0	158.3 ± 8.9	180.3 ± 10.1
Cohort 2 (HFD 60%)	181.3 ± 2.7	194.7 ± 10.4	213.5 ± 18.0	194.7 ± 8.4	202.1 ± 3.2	204.4 ± 4.6
Cohort 3 (HFD 60%)	149.1 ± 9.6	141.2 ± 11.6	189.8† ± 6.5	177.2‡ ± 7.8	159.8a ± 5.3	159.1 ± 7.6
NEFA, mEq/L
Cohort 1 (HFD 45%)	1.26 ± 0.08	1.17 ± 0.09	0.98 ± 0.15	0.98 ± 0.05	0.95 ± 0.14	0.79 ± 0.06
Cohort 2 (HFD 60%)	0.83 ± 0.04	0.86 ± 0.07	0.51†† ± 0.05	0.49‡‡‡ ± 0.04	0.41 ± 0.01	0.46 ± 0.03
Cohort 3 (HFD 60%)	0.82 ± 0.09	0.71 ± 0.07	0.60 ± 0.04	0.51 ± 0.03	0.63 ± 0.08	0.71 ± 0.07
TG, mg/dL
Cohort 1 (HFD 45%)	25.8 ± 2.4	40.1 ± 4.9	21.6† ± 3.2	29.4 ± 4.7	29.6 ± 4.3	26.9 ± 0.9
Cohort 2 (HFD 60%)	37.8 ± 6.0	35.9 ± 3.3	21.4 ± 2.9	33.6 ± 2.7	22.0 ± 2.0	28.0 ± 2.1
Cohort 3 (HFD 60%)	44.6 ± 7.0	47.2 ± 4.8	38.8 ± 4.4	38.9 ± 3.9	28.3 ± 2.8	39.2 ± 4.1
Cholesterol, mg/dL
Cohort 1 (HFD 45%)	87.1 ± 4.0	71.1 ± 11.4	111.8 ± 12.8	106.6 ± 8.4	83.5 ± 10.6	88.5 ± 5.7
Cohort 2 (HFD 60%)	102.8 ± 16.4	117.4 ± 10.5	152.0 ± 14.9	139.3 ± 7.0	137.2a ± 5.3	90.9* ± 11.9
Cohort 3 (HFD 60%)	177.1 ± 29.1	254.2 ± 33.4	202.3 ± 17.8	257.5 ± 20.5	193.8 ± 20.6	198.2 ± 17.0

Cohort 1: Mice fed a LFD or HFD containing 45% kcal from fat for 13 weeks and half switched to HFD/TZD in the last 5 weeks. Cohort 2: mice fed a LFD or HFD 60% kcal from Fat for 13 weeks and half switched to HFD/TZD in the last 5 weeks. Cohort 3: mice fed a LFD or HFD 60% kcal from fat for 23 weeks and half switched to HFD/TZD in the last 6 weeks. † and ‡ indicate significant differences induced by HFD as compared as LFD-fed mice within group: †differences in control mice; ‡differences in knockout mice. Letters a to d indicate significant differences induced by HFD/TZD as compared to HFD-fed mice within group. asterisks indicate significant differences induced by Pparg^ΔHep as compared to control mice within diet. a,†,‡ P < 0.05; **, P < 0.01; c,†††,‡‡‡ P < 0.001; d,††††,‡‡‡‡, P < 0.0001 (n = 6-10 mice/group).

Abbreviations: HFD, high-fat diet; LFD, low-fat diet; NEFA, nonesterified fatty acids; TG, triglycerides; TZD, thiazolidinedione.

Expression of Hepatocyte PPARγ Serves to Enhance Liver Steatosis in Severely Obese Mice Treated With Rosiglitazone for 6 Weeks

A HFD did not increase liver weight in young mice after 13 weeks of feeding, but it increased liver weight in adult mice after 23 weeks of feeding (Fig. 3A-3C). Of note, HFD 45% modestly increased hepatic TG content in young mice, which was reduced by TZD only in Pparg^ΔHep mice (Fig. 3D and 3E). Interestingly, the increase in liver weight induced by prolonged HFD 60% was associated with a dramatic increase in liver TG content, which was even further enhanced by the TZD treatment (Fig. 3F). The loss of hepatocyte PPARγ expression in this group (mice fed a HFD 60% for 23 weeks) reduced liver weight and TG content (Fig. 3C and 3F), as reflected by representative micrographs of hematoxylin and eosin–stained liver sections (Fig. 3G). Also in this diet group, we measured circulating markers of liver injury: ALT and AST. Although these markers were increased in control mice by HFD, this TZD treatment did not reduce ALT or AST in control mice (Fig. 3H and 3I). However, TZD did reduce AST levels in Pparg^ΔHep mice (Fig. 3I).

Figure 3.

Loss of hepatocyte peroxisome proliferator-activated receptor γ (PPARγ) expression reduces liver weight and steatosis in obese mice treated with rosiglitazone. Liver weight of control (C) or Pparg^ΔHep knockout (KO) mice fed a high-fat diet (HFD) containing 45% kcal from fat (A), 60% kcal from fat (B) for 13 weeks, or a HFD containing 60% kcal from fat (C) for 23 weeks. Liver triglycerides (TG) of control (C) or Pparg^ΔHep (KO) mice fed a HFD containing 45% kcal from fat (D), 60% kcal from fat (E) for 13 weeks, or a HFD containing 60% kcal from fat (F) for 23 weeks. Representative pictures of hematoxylin and eosin-stained liver sections (G), and plasma alanine transaminase (ALT) (H) and aspartate aminotransferase (AST) (I) levels of mice fed a HFD for 23 weeks. † and ‡indicate significant differences induced by HFD as compared as low-fat diet (LFD)-fed mice within group: †differences in control (C) mice; ‡differences in KO mice. Letters a to d indicate significant differences induced by HFD/thiazolidinediones (TZD) as compared to HFD-fed mice within group. Asterisks indicate significant differences induced by Pparg^ΔHep as compared to control mice within diet. *,‡, P < 0.05; **,b,††,‡‡ P < 0.01; ***,c,†††,‡‡‡ P < 0.001 (n = 8-9 mice/group).

PPARγ is a transcription factor known to regulate the expression of genes associated with lipid metabolism and the development of lipid droplets in hepatocytes (37). Consistent with previous reports (18), prolonged HFD 60% feeding in PPARγ-intact control mice dramatically increased hepatic PPARγ messenger RNA levels, compared to control LFD-fed mice (Fig. 4A) while 6 weeks of HFD/TZD did not increase PPARγ expression further. Since the polymerase chain reaction primers used could not differentiate between the PPARγ isoforms, western blot analysis was performed and revealed the PPARγ2 isoform was undetectable in LFD-fed control mice but dramatically increase in response to HFD, while TZD had no further stimulatory actions (Fig. 4B). In contrast, the PPARγ1 isoform was detectable in LFD-fed control mice and was reduced by HFD feeding. Since PPARγ2 was completely eliminated in Pparg^ΔHep mice, but PPARγ1 expression remained low but detectable, we can conclude PPARγ2 is the primary isoform expressed in the hepatocyte and responsive to HFD feeding, while PPARγ1 isoform is largely expressed in nonparenchymal cells (Fig. 4B). Interestingly, Pparg^ΔHep reduced HFD-induced expression of the PPARγ target genes related to fatty acid uptake and processing: fatty acid translocase (Cd36), cell death activator A and C (Cidea and Cidec), fatty acid–binding protein 4 (Fabp4), and monoacylglycerol O-acyltransferase 1 (Mogat1) (Fig. 4C). Also, Pparg^ΔHep blocked the positive effect of TZD on the expression of Cd36, Cidea, Cidec, and Fabp4, indicating that TZD directly activates these steatogenic factors and contributes to the increase of hepatic TG levels through the direct activation of hepatocyte PPARγ. Since steatosis can also be promoted by enhanced de novo lipogenesis (DNL) and/or reduced fatty acid oxidation, we examined the expression levels of genes key in these processes. The expression of genes critical for DNL—acetyl-CoA carboxylase (Acc1), fatty acid synthase (Fasn), and stearoyl-CoA desaturase (Scd1)—was not upregulated by HFD (Fig. 4D), consistent with a previous report showing long-term HFD feeding reduces DNL (38). However, TZD did increase the expression of Fasn and Scd-1 in a hepatocyte PPARγ-dependent manner. This suggests that TZD-mediated activation of PPARγ in hepatocytes could increase DNL in obese mice and contribute to increased steatosis. The expression of genes critical for fatty acid oxidation—carnitine palmitoyltransferase 1A (Cpt1a) and 3-hydroxy-3-methylglutaryl-CoA synthase 2 (Hmgcs2)—was not significantly altered by HFD or TZD. However, there was an overall positive effect (P < 0.0001) of Pparg^ΔHep on the expression of hepatic Cpt1a, which was significantly increased in HFD fed mice (Fig. 4E). This result suggests that enhanced fatty acid oxidation in Pparg^ΔHep mice could contribute to reduce steatosis.

Figure 4.

Expression of hepatic genes in mice fed a high-fat diet (HFD) for 23 weeks. (A and B) Hepatic expression of peroxisome proliferator-activated receptor γ (PPARγ) and the isoforms PPARγ1 and 2. (C) Hepatic expression of PPARγ-target genes: Cd36, Cidea, Cidec, Fabp4, and Mogat1. (D) Hepatic expression of de novo lipogenesis (DNL) genes: Acc1, Fasn, and Scd1. (E) Hepatic expression of genes involved in fatty acid oxidation: Cpt1a and Hmgsc2. † indicates significant differences induced by HFD as compared as low-fat diet (LFD)-fed control mice. Letters a to d indicate significant differences induced by HFD/thiazolidinediones (TZD as compared to HFD-fed mice within group. Asterisks indicate significant differences induced by Pparg^ΔHep as compared to control mice within diet. a,*,† P < 0.05; b,**,†† P < 0.01; c,***,††† P < 0.001; ****,††††, P < 0.0001 [n = 3-4 mice/group in (B); n = 8-9 mice/group in (A, C-E).

Loss of Hepatocyte PPARγ Increased Phosphorylation of Hepatic AMPK and Enhanced the Effects of Rosiglitazone in Adipose Tissue

We previously reported that phosphorylation of AMPK was reduced in mice with NASH but restored by Pparg^ΔHep (21). In the current study, Pparg^ΔHep increased the hepatic levels of phosphorylated (activated) AMPK in HFD-fed mice in the absence or presence of TZD (Fig. 5A and 5B). Phosphorylated AMPK (activated) increases fatty acid oxidation and reduces DNL (39); the increased activity of AMPK in Pparg^ΔHep mice may explain the reduction of steatosis observed in these groups (Fig. 3F). Plasma ketones are markers of hepatic fatty acid oxidation and they were not increased in Pparg^ΔHep mice under basal (not fasted) conditions. Interestingly, TZD reduced plasma ketones (Fig. 5C) and increased the expression of lipogenic genes Fasn and Scd1 (Fig. 4C) in obese control mice, which may be indicative of reduced AMPK activity due to activation of PPARγ in hepatocytes of control mice.

Figure 5.

Loss of hepatocyte peroxisome proliferator-activated receptor γ (PPARγ) expression is associated with increased hepatic phosphorylated adenosine 5′-monophosphate–activated protein kinase (pAMPK) and adiponectin levels in mice fed a high-fat diet (HFD) for 23 weeks. (A) Western blots of AMPK (adenosine 5′-monophosphate–activated protein kinase; top) and pAMPK (bottom). Ponceau S stained membranes were used to normalize AMPK and pAMPK levels. (B) Hepatic pAMPK/AMPK ratio. (C) Plasma ketone levels. (D) Plasma adiponectin levels. (E) Relative frequency of the size of adipocytes of subcutaneous adipose tissue. Letters a to d indicate significant differences induced by HFD/TZD as compared to HFD-fed mice within group. Asterisks indicate significant differences induced by Pparg^ΔHep as compared to control mice within diet. a,* P < 0.05; b, P < 0.01; d, P < 0.0001 (n = 3-9 mice/group).

Adiponectin is a key adipokine that activates hepatic AMPK to reduce TG content in the liver (40). Plasma adiponectin levels are increased with TZD treatments (41). The short-term and low dose of TZD used in this study increased plasma adiponectin in obese mice, but that rise was more pronounced in Pparg^ΔHep mice (Fig. 5D). Interestingly, this effect was associated with an increase in adipocyte size (Fig. 5E) and adipose tissue mass (Fig. 1F), suggesting that the adipogenic effects of TZD were increased in Pparg^ΔHep mice and that could contribute to reduce hepatic TG via adiponectin/AMPK pathway. Taken together, the expression of PPARγ in hepatocytes of control mice may suppress AMPK-mediated regulation of hepatic metabolism and contribute to sustain liver steatosis, as well as alter the effects of TZD on adipose tissue. However, this latter effect requires further investigation.

Discussion

To date, the hepatocyte-specific actions of TZD in NAFLD are not completely understood and remain controversial because clinical reports show variable effects (8,14,15). Some studies suggest that PPARγ, the primary effector of TZD, is not significantly expressed in the human liver and TZD should reduce NAFLD by increasing insulin sensitivity in peripheral tissues without activating PPARγ in hepatocytes (14). However, the expression of PPARγ is increased in liver biopsies of NAFLD patients and in the steatotic livers of diet-induced obese mice (16-19). Similarly, hepatic PPARγ expression is increased in mice with leptin deficiency (ob/ob mice), lipodystrophy (A-ZIP mice), and diet-induced obesity that have steatosis (36,42,43). In these preclinical models, TZD treatment further increases hepatic TG content, leading the authors to conclude that the steatogenic actions of TZD were due to its direct actions on PPARγ in liver hepatocytes (36,42). These direct actions are supported by the fact that hepatic PPARγ2 is increased by HFD (44) in a hepatocyte-specific manner, TZD promotes lipid accumulation and enhances PPARγ-target gene expression, indicative of lipid uptake and storage, in isolated mouse hepatocytes (21). Therefore, although TZD increases insulin sensitivity in peripheral tissues in patients with NAFLD, TZD-mediated activation of hepatocyte PPARγ reduces the positive effect of increased insulin sensitivity in NAFLD. This may represent a critical limitation for the use of TZD as potential drugs to treat NAFLD patients.

To further study the interrelationship between TZD and PPARγ on hepatocyte function, our group used a mouse model with inducible hepatocyte-specific knockout of PPARγ: Pparg^ΔHep. This mouse model allows us to avoid the confounding effects of congenital loss of hepatocyte PPARγ (albumin-driven Cre recombinase-mediated inactivation of PPARγ) and to assess the specific contribution of TZD-mediated activation of hepatocyte PPARγ after the liver has developed NAFLD. Also, we have used diet-induced obese mice, which is a more translational model than A-ZIP or ob/ob mice, to study the development and/or regression of NAFLD associated with diet-induced insulin resistance. It should be noted that in the current study, the effect of HFD paradigm on liver steatosis and obesity was dependent of the age at which diets were introduced: 5 to 6 weeks of age (peripubertal) vs 10 to 12 weeks of age (postpubertal). In fact, we previously reported that HFD-induced metabolic dysfunction was less severe in mice in which HFD was started at peripubertal age, compared to HFD started in adult mice (10-12 weeks), and was attributed to higher levels of Insulin-like growth factor 1 and lean mass (45). Also, 14 weeks of HFD 60% initiated in adult mice (19) induced similar steatosis and obesity as the prolonged (23 weeks) HFD 60% of this study. Therefore, in the current study, comparing the impact of hepatocyte PPARγ loss and TZD treatment after different diet regimens show for the first time that the deleterious effects of activating hepatocyte PPARγ are dependent on the severity of diet-induced obesity/steatosis.

Our data complement previous work showing short-term, low-dose TZD treatment enhances insulin sensitivity in mice with moderate and severe obesity. However, under severe obesogenic conditions induced only by prolonged HFD 60%, TZD promotes steatosis in a hepatocyte-specific PPARγ-dependent fashion. These findings further support the hypothesis that the direct actions of TZD in hepatocytes reduce their potential as effective therapies to treat NAFLD, and the potential negative actions of TZD/PPARγ on hepatocyte function should be weighted in the design of therapies for NAFLD. Based on our overall findings, our main question was how hepatocyte PPARγ sustains and TZD enhances established diet-induced steatosis to contribute to the development of NAFLD.

Hepatocyte PPARγ is strongly associated with DNL and steatosis in multiple models (37,46). However, in this and previous reports (19,47), adult-onset loss of hepatocyte PPARγ had minimal impact on expression of DNL genes. Nonetheless, TZD enhanced the expression of a subset of DNL genes (Fasn and Scd1) in PPARγ-intact mice despite the reduction in insulin levels (direct insulin-sensitizing effect of TZD). Furthermore, our study neither challenged the mice to increase hepatic DNL nor measured the rate of DNL in vivo, and so hepatocyte PPARγ may contribute to sustain DNL and promote steatosis in the obese mice of this study. Hepatocyte PPARγ also promotes fatty acid uptake and re-esterification. We and others reported that hepatocyte PPARγ directly increases hepatic Cd36 and Mogat1 expression, among other genes, to increase steatosis in diet-induced obese mice (19,48). Overexpression of PPARγ in the liver increases hepatic Cd36, Cidea, Cidec, Fabp4, and Mogat1, among other genes, to increase fatty acid uptake and storage in hepatocytes (37,48,49) and is correlated with steatosis in male and female mice (50). Loss of hepatocyte PPARγ expression reduced the expression of these genes and steatosis, while TZD upregulated them in a hepatocyte PPARγ-dependent manner to promote steatosis despite the improvement of insulin sensitivity. Therefore, our results indicate that fatty acid uptake and storage are key mechanisms regulated by PPARγ in hepatocytes, which sustains steatosis in obese mice and is enhanced with TZD treatments.

Hepatic DNL and fatty acid uptake are key mechanisms that promote NAFLD in humans. The direct effects of TZD on DNL and fatty acid uptake in hepatocytes of obese individuals may reduce the therapeutic benefits of TZD on NAFLD. Our results are in agreement with other studies that show the steatogenic effect of TZD treatments on diet-induced obese mice (49), and diabetic KKAy obese mice (51), as well as those in AZIP and ob/ob mice (36,42). However, these results are in stark contrast with the antisteatogenic effect of TZD in humans (5,6), and lean mice (49). Here, it should be mentioned lifestyle modifications adopted by humans treated with TZD could contribute to the different hepatic responses (steatogenic vs antisteatogenic) of TZD in humans and preclinical models. In humans, TZD are provided along with reduced caloric intake and increased physical activity, whereas in mice TZD are provided with HFD ad libitum. Also, the different hepatic responses to TZD treatments may be due to the degree of obesity and the cell-specific level of PPARγ expression: PPARγ2 in hepatocytes (16-19,43,44). In this sense, mouse studies showed that loss of hepatocyte PPARγ expression (21,52) or the constitutive deacetylation of PPARγ (53) enhanced the antisteatogenic properties of TZD in HFD-fed obese mice. Therefore, if TZD cannot promote DNL and fatty acid uptake in hepatocytes (as occurs in Pparg^ΔHep mice), the insulin-sensitizing effects of TZD in peripheral tissues will reduce liver steatosis efficiently. The lack of steatogenic effects of TZD in the liver of Pparg^ΔHep mice is consistent with RNAseq results from our group as well as others, showing rosiglitazone and pioglitazone upregulates gene ontology terms related to fatty acid metabolism in control mice but not in hepatocyte-specific PPARγ knockout mice (GEO: GSE162276, GSE140607) (21,52). Therefore, despite the fact that TZD improves insulin sensitivity, the collective work of ourselves and others suggest that the therapeutic potential of TZD in treating NAFLD patients should be considered with caution due to its ability to activate hepatocyte PPARγ and exacerbate steatosis.

In conclusion, our results support that hepatocyte PPARγ is an important regulator of steatosis in a translational mouse model of NAFLD with diet-induced obesity and insulin resistance, and HFD- and TZD-mediated activation of hepatocyte PPARγ regulates similar mechanisms that contribute to increase the storage of fat in hepatocytes in vivo. Importantly, our study provides additional evidence related to the design of the HFD feeding paradigm selected to induce obesity and to assess the effect of TZD on liver physiology.

Additional Information

Disclosures: Authors do not have any conflict of interest.

Data Availability

Data and analysis generated in this study are not publicly available but are available from the corresponding author on reasonable request.