Hepatocyte‐specific deletion of adipose triglyceride lipase (adipose triglyceride lipase/patatin‐like phospholipase domain containing 2) ameliorates dietary induced steatohepatitis in mice

Abstract Background and Aims Increased fatty acid (FA) flux from adipose tissue to the liver contributes to the development of NAFLD. Because free FAs are key lipotoxic triggers accelerating disease progression, inhibiting adipose triglyceride lipase (ATGL)/patatin‐like phospholipase domain containing 2 (PNPLA2), the main enzyme driving lipolysis, may attenuate steatohepatitis. Approach and Results Hepatocyte‐specific ATGL knockout (ATGL LKO) mice were challenged with methionine‐choline–deficient (MCD) or high‐fat high‐carbohydrate (HFHC) diet. Serum biochemistry, hepatic lipid content and liver histology were assessed. Mechanistically, hepatic gene and protein expression of lipid metabolism, inflammation, fibrosis, apoptosis, and endoplasmic reticulum (ER) stress markers were investigated. DNA binding activity for peroxisome proliferator‐activated receptor (PPAR) α and PPARδ was measured. After short hairpin RNA–mediated ATGL knockdown, HepG2 cells were treated with lipopolysaccharide (LPS) or oleic acid:palmitic acid 2:1 (OP21) to explore the direct role of ATGL in inflammation in vitro. On MCD and HFHC challenge, ATGL LKO mice showed reduced PPARα and increased PPARδ DNA binding activity when compared with challenged wild‐type (WT) mice. Despite histologically and biochemically pronounced hepatic steatosis, dietary‐challenged ATGL LKO mice showed lower hepatic inflammation, reflected by the reduced number of Galectin3/MAC‐2 and myeloperoxidase‐positive cells and low mRNA expression levels of inflammatory markers (such as IL‐1β and F4/80) when compared with WT mice. In line with this, protein levels of the ER stress markers protein kinase R–like endoplasmic reticulum kinase and inositol‐requiring enzyme 1α were reduced in ATGL LKO mice fed with MCD diet. Accordingly, pretreatment of LPS‐treated HepG2 cells with the PPARδ agonist GW0742 suppressed mRNA expression of inflammatory markers. Additionally, ATGL knockdown in HepG2 cells attenuated LPS/OP21‐induced expression of proinflammatory cytokines and chemokines such as chemokine (C‐X‐C motif) ligand 5, chemokine (C‐C motif) ligand (Ccl) 2, and Ccl5. Conclusions Low hepatic lipolysis and increased PPARδ activity in ATGL/PNPLA2 deficiency may counteract hepatic inflammation and ER stress despite increased steatosis. Therefore, lowering hepatocyte lipolysis through ATGL inhibition represents a promising therapeutic strategy for the treatment of steatohepatitis.


INTRODUCTION
As a consequence of the pandemic of obesity and diabetes, NAFLD has become a leading cause of liver disease in the Western world. [1,2] As such, more than 20% of the general population [3] and 75% of individuals with obesity [4] develop NAFLD. NAFLD is characterized by benign hepatic fat accumulation (i.e., steatosis) which can progress to NASH, advanced fibrosis, cirrhosis, and cancer, [5,6] with inflammation as a central feature in the disease progression from benign steatosis to the more severe stages of the disease spectrum. [7] Increased free fatty acids (FAs) are key lipotoxic triggers driving hepatocyte injury and inflammation and the progression from NAFLD to NASH. [8] Increased FA flux from adipose tissue to the liver due to insulin resistance as well as their release from hepatocellular triglyceride (TG) stores may critically determine hepatic FA concentrations and thereby lipotoxicity. [9] However, FAs also have other powerful roles and are directly involved in cellular signaling pathways and regulation of gene transcription. [10] Specifically, FAs are known to activate potential anti-inflammatory peroxisome proliferator-activated receptor (PPAR) signaling. [11] Adipose TG lipase (ATGL) as the major enzyme in TG breakdown may provide ligands for FAregulated transcription factors, such as PPARα and PPARδ, [12][13][14][15][16] thereby modulating hepatic inflammation and disease progression. ATGL-catalyzed TG hydrolysis is a prerequisite for functional PPARα signaling, suggesting that FAs deriving from TG stores activate PPARα and consequently the expression of genes for mitochondrial FA oxidation. [13] This aspect is discussed controversially for PPARδ. [16,17] Despite the difference in their way of activation, both PPARα and PPARδ seem to have a crucial role in the protection against NAFLD/ NASH development. [18] Accordingly, PPARα knockout (KO) mice that are fed a methionine-choline-deficient (MCD) diet develop more severe steatohepatitis than wild-type (WT) mice, [19,20] whereas treatment of human apolipoprotein E2 knock in/PPARα KO mouse with a PPARα/δ dual agonist significantly improved MCD diet-induced liver injury. [21] Interestingly, PPARδ can compensate for the absence of PPARα in regulating FA homeostasis in skeletal muscle. [22] To test whether this appealing concept is also relevant in liver, this study was designed to investigate whether PPARδ signaling may compensate for impaired PPARα signaling in the liver, thereby attenuating hepatic inflammation and subsequent progression from steatosis to steatohepatitis. For this purpose, hepatocyte-specific ATGL KO mice (with impaired PPARα signaling) were subjected to an MCD diet, a well-established model of steatohepatitis. Despite several disadvantages of the MCD diet, such as absence of obesity and insulin resistance, it is a suitable model to initiate profound lipolysis in adipose tissue, resulting in increased FA flux from the periphery to the liver. [23] Additionally, and to also obtain insights into the metabolic state of the disease, hepatocyte-specific ATGL KO mice were fed a high-fat high-carbohydrate (HFHC)-rich diet. Using this experimental approach, we observed that absence of ATGL/patatin-like phospholipase domain containing 2 (PNPLA2) in hepatocytes results in steatosis but reduced inflammation, which was most likely through activation of hepatic PPARδ signaling as an adaptation to loss of PPARα signaling.

Animal experiments
C57/BL6 ATGL f/f were crossed with C57/BL6 Alb-Cre mice to generate hepatocyte-specific ATGL KO mice. ATGL f/f mice were used as control animals. ATGL f/f were kindly provided by Erin E. Kershaw [24] (JAX stock number of ATGL f/f mice: 024278

Human liver sections
Human liver paraffin sections of patients with NAFLD with different stages of fibrosis (F0-F4) were kindly provided by the Department of Pathology, Medical University of Vienna. Patients gave informed consent at the time of recruitment, and their records were anonymized and deidentified. Patients' data and pathologic evaluations were approved by Ethics Committee of the Medical University of Vienna (EC: 747/2011). No donor organs were obtained from executed prisoners or other institutionalized persons. Detailed clinical data of the patient cohort are provided in Jha et al. [25] Liver histology Liver tissue was fixed in 4% neutral buffered formaldehyde solution for 24 hours, embedded in paraffin, and stained with hematoxylin and eosin (H&E) or sirius red. To quantify and characterize the hepatic inflammatory cell infiltrate, immunohistochemistry (IHC) for Galectin3/MAC-2 (MAC-2) + cells and immunofluorescence of myeloperoxidase (MPO) + cells was performed. Oil Red O (ORO) staining was performed on 7-µM-thick cryosections as described. [26] To obtain the degree of steatosis and inflammation, computational quantification was done with ImageJ 1.51j8. The NAFLD grading shown in Supporting Tables S1 and S2 is based on computational analysis of H&E (macrovesicular and microvesicular steatosis) and MAC-2 + (inflammatory cells) sections. The scoring is in accordance with Liang et al. [27]

Serum analysis
At the end of the experiment, blood was collected from the vena cava and centrifuged for 15 minutes at 2150 g. Serum was stored at −80°C until analysis.

DNA binding activity of PPARα and PPARδ
PPARα as well as PPARδ activity was assayed using an enzyme-linked immunosorbent assay-based PPARα Transcription Factor Assay Kit (Abcam) according to the manufacturer's instructions. Nuclear proteins were isolated from liver using a NE-PER nuclear and cytoplasmic extraction kit (Abcam) according to the manufacturer's instructions. Binding activity was measured at 450 nm (minus the blank) and calculated as relative activity to the control.

Western blotting
Protein isolation and western blotting was performed as described. [28]

RNA isolation and qRT-PCR analysis
Tissues were snap frozen in prechilled 2-methylbutane and stored in liquid nitrogen. RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The 1.5 μg of RNA was used for complementary DNA synthesis using a random hexamer primer (Applied Biosystems) and Superscript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. A 1:20 dilution of the complementary DNA was used for qRT-PCR using SYBR Green Master Mix (Applied Biosystems) and was performed using an AB7900 Real-Time PCR System (Applied Biosystems). Reactions were performed in duplicates, and relative mRNA levels were quantified using a calibration dilution curve normalized to the housekeeping genes. mRNA levels were normalized to 36b4 as a housekeeping gene that did not vary between groups.

Liver TGs, nonesterified fatty acid, and fatty Acyl chain profiling
Hepatic lipids were extracted and analyzed by gas chromatography as described. [29] Cell culture HepG2 and Hepa1c1c7 cells (American Type Culture Collection, Manassas, VA) were grown in Dulbecco's modified Eagle's medium and knocked down for ATGL (ATGL knockdown [KD]) using a lentivirus containing a short hairpin RNA against ATGL as described. [26] WT and ATGL KD cells were treated with lipopolysaccharide (LPS) for 6 hours or oleic acid (OA):palmitic acid (PA) in a ratio 2:1 (OP21). OA concentration: 1.2 mM; PA concentration: 0.6 mM. Furthermore, in a separate experiment, HepG2 WT cells were pretreated for 18 hours with the PPARδ agonist GW0742 before 6 hours of LPS treatment. THP1 and RAW 264.7 cells (American Type Culture Collection) were grown in Dulbecco's modified Eagle's medium and incubated for 6 hours with a medium taken from WT and ATGL KD cells with and without LPS or incubated 24 hours with medium taken from WT and ATGL KD cells with and without OP21 treatment.

Cytokine analysis
Proteome profiler mouse XL cytokine array from R&D Systems was assessed according to the manufacturer's instructions.

Statistical analysis
Results were evaluated using SPSS V.27.0. Statistical analysis was performed using Student unpaired twotailed t test. In the MCD setting, data are reported as means of WT Ctrl n = 5; ATGL LKO Ctrl n = 7; WT MCD n = 7; and ATGL LKO MCD n = 9 animals per group ± SD. In the HFHC setting, data are reported as means of WT Ctrl n = 7; ATGL LKO Ctrl n = 5; WT HFHC n = 7; and ATGL LKO HFHC n = 5 animals per group ± SD. A p value ≤0.05 was considered statistically significant.

Absence of hepatic ATGL reduces PPARα activity in MCD-induced steatohepatitis
Studies have demonstrated that global ATGL KO mice challenged with an MCD diet develop more severe steatohepatitis because of impaired PPARα signaling due to lack of free FAs serving as PPARα ligands. [28] To investigate the specific role of hepatic ATGL in development of steatohepatitis, hepatocyte-specific ATGL KO (ATGL LKO) and floxed control mice (WT) were fed an MCD diet for 5 weeks. In accordance with the findings in global ATGL deficiency, ATGL LKO mice also showed reduced PPARα DNA binding activity (Supporting Figure S1A) when compared with MCD-fed WT mice. Accordingly, mRNA expression levels of established PPARα target genes were significantly lower in MCD-fed ATGL LKO compared with MCD-fed WT animals (Supporting Figure S1B).

MCD feeding aggravates development of hepatic steatosis
Because impaired PPARα signaling and consequently defective FA oxidation potentially contribute to development of steatosis, the next hepatic TG content was assessed. In line with the observations at the histological level, quantification of hepatic TG significantly increased TG levels in ATGL LKO mice already at chow diet, and levels were further boosted by the MCD challenge ( Figure 1A,B, Supporting Table S1). This observation was confirmed by ORO staining ( Figure 1C). Serum levels of liver aminotransferases alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were elevated on MCD independent of the genotype, whereas alkaline phosphatase (AP) was solely increased in ATGL LKO mice on the dietary challenge ( Figure 1D). Serum levels of total cholesterol and TG as well as nonesterified FA (NEFA) were reduced on MCD feeding in both WT and ATGL LKO mice ( Figure 1D).

Absence of hepatic ATGL ameliorates hepatic inflammation in MCD diet-induced steatohepatitis
Besides steatosis, inflammation is a major hallmark of hepatic steatohepatitis. Therefore, we investigated inflammatory parameters such as MAC-2, MPO, IL-1β, F4/80, and TNFα in mice kept on MCD diet. Surprisingly, IHC revealed fewer MAC-2 and MPO positive cells in livers of ATGL LKO than in WT mice under MCD challenge ( Figure 2A,B). Accordingly, mRNA levels of inflammatory markers were significantly decreased in ATGL LKO animals under MCD condition in comparison with challenged WT animals (Supporting Figure S2A). Of particular interest, sirius red and collagen 1 staining (as a marker for fibrosis) were similar in ATGL LKO compared with WT mice, indicating no differences in collagen deposition. However, mRNA levels of Col1a1, Col1a2, and Tgfβ were increased in both MCD-fed WT and ATGL LKO mice (Supporting Figure S3).

Hepatic ATGL deficiency mitigates endoplasmic reticulum stress in MCD diet-induced steatohepatitis
Because hepatic inflammation may trigger hepatic endoplasmic reticulum (ER) stress, we assessed protein and mRNA levels of ER stress relevant genes. The ER stress sensors protein kinase R-like endoplasmic reticulum kinase (PERK) and inositol-requiring enzyme 1α (IRE1α) showed markedly lower protein expression levels in ATGL LKO mice compared with WT mice on MCD feeding ( Figure 2C). In line with this, mRNA expression of ErDj4 and Grp78 followed the same pattern (Supporting Figure S2B). Of note, mRNA levels of the apoptotic markers Chop and B cell lymphoma 2-associated death promoter (Bad) as well as protein levels of cleaved caspase 3 were increased on MCD feeding independent of the genotype (Supporting Figure S2C,D).

Absence of hepatic ATGL induces PPARδ binding activity in MCD diet-induced steatohepatitis
It has been shown that PPARδ (but not PPARα) can be activated through FAs directly coming from the serum (not being hydrolyzed from endogenous TG stores). [16] Moreover, PPARδ was also identified to have antiinflammatory effects under NAFLD conditions. [18,30] Therefore, we considered augmented PPARδ activation as a possible compensatory mechanism of compromised (anti-inflammatory) PPARα signaling (Supporting Figure S1) and thus a mechanistic explanation for attenuated hepatic inflammation in MCD-challenged ATGL LKO mice. DNA binding activity assay revealed increased PPARδ binding activity solely in MCD-fed ATGL LKO mice ( Figure 3A). Consequently, mRNA expression of PPARδ target pyruvate dehydrogenase lipoamide kinase isozyme 4 (Pdk4) was boosted in MCD-fed ATGL LKO animals ( Figure 3B). Accordingly, PDK4 protein expression was significantly increased in MCD-fed ATGL LKO mice compared with challenged WT mice ( Figure 3C). Moreover, intrahepatic FA profiling elucidated enhanced levels of α-linolenic and γ-linolenic acid as well as OA, which are established ligands for PPARδ ( Figure 3D). Interestingly, the level of vaccenic acid, which has been shown to maintain PPARδ in its activated conformation, [31,32] was reduced in WT mice on MCD feeding, whereas in ATGL LKO mice, levels stayed normal (Supporting Figure S4). Of note, intrahepatic concentration of anti-inflammatory palmitoleic acid, which is not related to PPAR signaling, [33] was exclusively reduced in WT mice on MCD diet and accordingly normal in ATGL LKO mice on MCD diet (Supporting Figure S4).

Absence of hepatic ATGL reduces PPARα activity in HFHC-induced steatohepatitis
To investigate whether PPARα signaling is disturbed in ATGL LKO mice in a second metabolic model, PPARα DNA binding activity was assessed under HFHC conditions. Accordingly, ATGL LKO mice showed reduced PPARα DNA binding activity (Supporting Figure S5A) when compared with HFHC-fed WT mice. In line with this, mRNA expression levels of established PPARα target genes were significantly lower in HFHC-fed ATGL LKO compared with HFHC-fed WT animals (Supporting Figure S5B).

HFHC feeding aggravates development of hepatic steatohepatitis
In accordance with the observations of increased hepatic lipid load, at the histological level, quantification of hepatic TG revealed an already significant increase in ATGL LKO mice at chow diet that was further elevated by HFHC application (Figure 4A,B, Supporting Table S2). This finding was confirmed by ORO staining ( Figure 4C). Serum levels of liver aminotransferases ALT and AST were already elevated in ATGL LKO mice at baseline. On the other hand, in WT mice, the HFHC challenge led to a further increase of ALT and AST. In HFHC-fed ATGL LKO mice, ALT and AST levels remained unchanged compared with control mice. AP and TG concentration were neither affected by genotype nor by feeding ( Figure 4D). Total cholesterol as well as NEFA concentration were increased because of HFHC feeding independent of the genotype ( Figure 4D).

Absence of hepatic ATGL ameliorates hepatic inflammation in HFHC diet-induced steatohepatitis
Investigation of inflammatory parameters such as MAC-2, MPO, IL-1β, F4/80, and Tnfα revealed that loss of ATGL in the liver attenuated development of hepatic inflammation induced by HFHC diet. IHC conceded fewer MAC-2 and MPO positive cells in livers of ATGL LKO than in WT mice under HFHC challenge ( Figure 5A,B). Accordingly, mRNA levels of inflammatory markers were significantly decreased in ATGL LKO animals under HFHC condition in comparison with challenged WT animals (Supporting Figure S6A). Fibrosis markers, such as sirius red and collagen 1 staining, were similar in ATGL LKO compared with WT mice, indicating no differences in collagen deposition. However, mRNA levels of Col1a1, Col1a2, and Tgfβ tended to be decreased in ATGL LKO compared to WT mice challenged with HFHC diet (Supporting Figure S7).

Hepatic ATGL deficiency mitigates ER stress in HFHC diet-induced steatohepatitis
To evaluate whether HFHC feeding induced hepatic ER stress, we assessed protein and mRNA levels of ER stress relevant genes. The ER stress sensors PERK and IRE1α were increased in both WT and ATGL LKO mice fed with HFHC diet (Figure 5C), whereas mRNA expression of the downstream target Grp78 was only significantly increased in WT mice upon HFHC challenge, and expression levels of ErDj4 remained unchanged in the different groups (Supporting Figure  S6B). mRNA levels of the apoptotic marker Chop were increased in HFHC-fed WT mice only. mRNA levels of Bad as well as protein levels of cleaved caspase 3 were unaffected on HFHC feeding independent of the genotype (Supporting Figure S6C,D).

Absence of hepatic ATGL induces PPARδ binding activity In HFHC diet-induced steatohepatitis
DNA binding activity assay revealed increased PPARδ binding activity in HFHC-fed ATGL LKO mice ( Figure 6A). Accordingly, mRNA expression of PPARδ downstream target Pdk4 was elevated in HFHC-fed ATGL LKO animals ( Figure 6B). In line with this, PDK4 protein expression was significantly increased in HFHC challenged ATGL LKO mice compared with control mice ( Figure 6C). Intrahepatic FA profiling elucidated unchanged levels of α-linolenic and γ-linolenic acid between the groups, whereas OA levels were already increased in ATGL LKO mice at baseline. HFHC feeding boosted OA concentration in WT but not in ATGL LKO mice ( Figure 6D). Concentration of vaccenic acid, F I G U R E 3 MCD feeding increases PPARδ activity in ATGL LKO mice. (A) Semiquantitative PPARδ DNA binding activity assay. MCD feeding induced PPARδ signaling in ATGL LKO (but not in WT) mice. (B) mRNA expression levels of PPARδ downstream target Pdk4. mRNA expression of Pdk4 was increased because of MCD feeding in both genotypes, but the increase was more pronounced in ATGL LKO mice. Data represent means ± SD and are shown relative to the expression levels of unchallenged WT animals. (C) Representative picture of PPARδ downstream target PDK4 immunoblot. Protein expression levels of PDK4 were elevated in MCD-fed WT mice and stayed at baseline levels in MCD-fed ATGL LKO mice. Data are normalized to total protein and represent means ± SD and are shown relative to the expression levels of unchallenged WT animals. (D) Intrahepatic free FA profile. α-linolenic and γ-linolenic acid (known to be PPARδ ligands) were most prominent in MCD-fed ATGL LKO animals. * indicates a significant difference from untreated WT controls; $ indicates a significant difference from MCD-fed WT mice; # indicates a significant difference from ATGL LKO controls; p < 0.05 known to maintain PPARδ in its activated conformation, [31,32] was increased in ATGL LKO mice already at baseline (Supporting Figure S4B). Of note, intrahepatic concentration of anti-inflammatory palmitoleic acid, which is not related to PPAR signaling, [33] was also induced at baseline in ATGL LKO mice (Supporting Figure S4B).

ATGL KD protects from LPS and FA-induced inflammation in vitro
To investigate whether absence of ATGL in hepatocytes per se may contribute to attenuation of endotoxininduced inflammation (seen in MCD-fed animals [34] ), ATGL was silenced with short hairpin RNA in the human hepatoma cell line HepG2 (Supporting Figure S8A,B). Cells were cultured in a medium without FBS 16 hours before the treatment with LPS as proinflammatory stimuli inducing the activation of TLR4-related pathways. After 6 hours of LPS treatment, mRNA expression of inflammatory markers monocyte chemoattractant protein 1 (Mcp1), chemokine (C-X-C motif) ligand (Cxcl) 2, and Tnfα stayed at basal levels in ATGL KD cells, whereas all makers were increased in WT cells ( Figure 7A). From another batch of cells, LPS-containing medium was removed after 6 hours and exchanged with an empty medium. Twenty-four hours later, the medium was used for protein profiling. Medium taken from WT LPSchallenged cells contained the chemokines osteopontin, IL-8, Macrophage inhibitory cytokine-1, and VEGF. Notably, signals of these proteins were much lower in medium from ATGL KD cells compared with WT cells ( Figure 7B). Because chemokines are involved in recruitment and activation of inflammatory cells, cytokine cocktails produced from HepG2 WT and ATGL KD cells after LPS treatment were used to treat the human macrophage cell line THP1 for 6 hours. THP1 cells treated with medium from LPS-treated HepG2 WT cells displayed increased mRNA expression levels of the inflammatory markers IL-1β, IL-6, and Mcp1. In strong contrast, mRNA levels of these genes stayed low in THP1 cells treated with medium from LPS-treated HepG2 ATGL KD cells ( Figure 7C), arguing for a direct anti-inflammatory effect of reduced ATGL activity in hepatocytes. Of note, to validate the translational effect between human and mouse, the same experiment was performed in a hepatocyte cell line derived from F I G U R E 4 HFHC feeding induced liver injury in WT and ATGL LKO mice. WT and ATGL KO mice received HFHC diet for 17 weeks. (A) H&E staining of liver sections of control and HFHC-fed WT and ATGL LKO mice. Hepatic steatosis was evident in ATGL LKO mice at baseline and even more pronounced after HFHC challenge (×20 magnification). (B) Biochemical quantification of intrahepatic TG. Intrahepatic TG levels were increased in ATGL LKO mice at baseline as well as in HFHC-treated WT and ATGL LKO mice. (C) ORO staining. ORO staining and subsequent computational quantification displayed elevated intrahepatic lipid accumulation in ATGL LKO mice at baseline. HFHC feeding increased lipid load in both genotypes with a more pronounced rise in ATGL LKO mice. (D) Serum markers of liver injury. Twenty-one-week-old ATGL LKO mice showed increased levels of liver enzymes (AST, ALT) already at baseline. Although HFHC feeding led to a further increase of aminotransferases in WT mice, in ATGL LKO mice, the levels remained unaffected. AP and TG levels remained constant in WT and ATGL LKO mice. Serum total cholesterol and NEFAs were significantly increased after HFHC feeding. * indicates a significant difference from untreated WT controls (WT); # indicates a significant difference from ATGL LKO control mice; p < 0.05 [Color figure can be viewed at wileyonlinelibrary.com] mouse (Hepa1c1c7). Like in the human cell line, loss of ATGL protected murine hepatocytes from LPS-induced inflammation (Supporting Figures S8 and S9).
Moreover, to investigate whether loss of ATGL also protects from FA-induced inflammation, cells were cultured in medium containing OP21 as proinflammatory stimuli. After 24 hours of OP21 treatment, mRNA expression of inflammatory markers Mcp1, IL-1b, IL-6, and Cxcl2 stayed at basal levels in ATGL KD cells, whereas all markers were increased in WT cells (Supporting Figure S10A).
Following the same experimental procedure as for the LPS setting, OP21-containing medium was removed after 24 hours and exchanged with medium without OP21. Twenty-four hours later, the medium consisting of secreted cytokines was used for protein profiling. Although the medium taken from WT cells challenged with OP21 contained the chemokines osteopontin and IL-8, these signal proteins were not detected in medium from ATGL KD cells (Supporting Figure S10B). The cytokine cocktails produced from HepG2 WT and ATGL KD cells after OP21 treatment were used to treat the human macrophage cell line THP1 for 6 hours. THP1 cells treated with medium from OP21-treated HepG2 WT cells displayed increased mRNA expression levels of the inflammatory markers IL-1β, IL-6, and Mcp1. In strong contrast, mRNA levels of these genes stayed low in THP1 cells treated with medium from OP21-treated HepG2 ATGL KD cells (Supporting Figure S10C), arguing for a direct anti-inflammatory effect of reduced ATGL activity in hepatocytes.

Activation of PPARδ reduces LPS-induced inflammation in vitro
To investigate whether the LPS-mediated inflammation seen in HepG2 WT cells can be counteracted by PPRAδ activation, HepG2 WT cells were pretreated with the PPARδ agonist GW0742 for 18 hours before 6 hours of LPS incubation (Figure 8). Treatment of WT cells with GW0742 result in increased protein expression of PPARδ downstream target PDK4 ( Figure 8A). The PPARδ anti-inflammatory effect has been linked to inhibition of NFκB signaling, which prompted us to measure protein expression of NFκB subunits. Although expression of p50 and p65 subunits was increased in the nuclei of LPS-treated HepG2 WT

F I G U R E 5 HFHC feeding-induced hepatic inflammation and ER stress is attenuated in ATGL LKO mice. (A) MAC-2 IHC.
Representative IHC staining for MAC-2 + cells in liver specimens of control and HFHC-fed WT and ATGL LKO mice (×20 magnification) as well as computational quantification show increased inflammation in HFHC-fed WT and ATGL LKO mice. HFHC-fed ATGL LKO mice show significantly fewer MAC-2 + cells compared with WT mice on HFHC treatment. (B) MPO immunofluorescence. Representative immunofluorescence staining for MPO + cells in liver specimens of control and HFHC-fed WT and ATGL LKO mice (×20 magnification) as well as computational quantification show increased inflammation in HFHC-fed WT and ATGL LKO mice. HFHC-fed ATGL LKO mice show significantly fewer MPO + cells compared with WT mice on HFHC treatment. (C) Representative picture of ER stress sensor immunoblots. Protein expression levels of PERK and IRE1α were elevated in both HFHC-fed WT and ATGL LKO mice. Data are normalized to total protein and represent means ± SD and are shown relative to the expression levels of unchallenged WT animals. * indicates a significant difference from untreated WT controls; $ indicates a significant difference from HFHC-fed WT mice; # indicates a significant difference from ATGL LKO controls; p < 0.05 [Color figure can be viewed at wileyonlinelibrary.com] cells, pretreatment with GW0742 prevented elevated expression levels of both NFκB subunits ( Figure 8B). Consequently, expression levels of the proinflammatory cytokines Mcp1, Cxcl2, and Tnfα were increased in LPS-treated cells, whereas expression levels remained low and comparable with control cells on GW0742 pretreatment cells ( Figure 8C).

PPARδ expression correlates negatively with disease progression
To clarify whether the murine findings that PPARδ may have a beneficial role in counteracting NAFLD development, PPARδ protein expression was investigated in a patient cohort with NAFLD consisting of patients aged ≥18 years with biopsy-proven NASH. [25] PPARδ expression was found to be reduced in patients with NASH with fibrosis stage 3 (NASH F3) (Supporting Figure S11), arguing for a crucial role of PPARδ in counteracting disease progression.

DISCUSSION
In this study, we investigated the hepatocyte-specific role of ATGL in the development of steatohepatitis by challenging hepatocyte-specific ATGL KO (ATGL LKO) mice with an MCD as well as HFHC diet. Both diets are common models to induce steatohepatitis, leading to endotoxin translocation, steatosis, and inflammation. However, despite the MCD diet having some limitations, such as absence of obesity and insulin resistance, it reflects several NASH key features, such as steatosis and inflammation. [35][36][37] Importantly, the MCD diet induces massive FA flux from adipose tissue to the liver, thus reflecting a key pathogenic step in NASH development because an increased peripheral FA flux has been linked to a 60% increase of hepatic TG content in patients with NAFLD. [38] Therefore, the two complementary models used in this study-ATGL LKO mice fed either with MCD or HFHC-offer the opportunity to investigate whether a lack of hepatic ATGL (and hence impaired hepatic lipolysis) may increase the susceptibility F I G U R E 6 HFHC feeding increases PPARδ activity in ATGL LKO mice. (A) Semiquantitative PPARδ DNA binding activity assay. HFHC feeding induced PPARδ signaling in ATGL LKO (but not in WT) mice. (B) mRNA expression levels of PPARδ downstream target Pdk4. mRNA expression of Pdk4 was increased due to HFHC feeding in ATGL LKO mice. Data represent means ± SD and are shown relative to the expression levels of unchallenged WT animals. (C) Representative picture of PPARδ downstream target PDK4 immunoblot. Protein expression levels of PDK4 were elevated in HFHC-fed WT and ATGL LKO mice. Data are normalized to total protein and represent means ± SD and are shown relative to the expression levels of unchallenged WT animals. (D) Intrahepatic free FA profile. α-linolenic and γ-linolenic acid (known to be PPARδ ligands) concentration remained unchanged among the different groups, whereas OA concentration was increased in ATGL LKO mice at baseline and further increased by HFHC feeding in both WT and ATGL LKO animals. * indicates a significant difference from untreated WT controls; $ indicates a significant difference from HFHC-fed WT mice; # indicates a significant difference from ATGL LKO controls; p < 0.05 to hepatic inflammation through altered lipid partitioning and impaired PPARα signaling, as demonstrated in global ATGL KO mice. [28] Alternatively, the FA spillover to the liver could result in activation of compensatory anti-inflammatory mechanism(s), such as increased PPARδ signaling, which was shown to countervail detracted PPARα signaling in skeletal muscle. [22] Our data demonstrate that despite reduced hepatic PPARα signaling, loss of ATGL in the liver attenuates development of hepatic inflammation and ER stress as major events progressing fatty liver disease from simple steatosis to more severe stages of the disease, such as steatohepatitis. Although serum markers for liver damage (AST and ALT) were increased at comparable levels in WT and ATGL LKO mice on dietary challenge, hepatic steatosis was moderately increased in ATGL LKO mice fed either with MCD or HFHC diet compared with challenged WT animals, whereas inflammatory as well as ER stress markers (in the MCD setting) were significantly reduced. These findings indicate that counteracting hepatic lipolysis and consequently increasing hepatic TG formation may have a protective role against inflammation and ER stress development.
Interestingly, mRNA expression of fibrosis markers Col1a1, Col1a2, and Tgfβ tended to be increased because of MCD as well as HFHC feeding independent of the genotype. However, computational quantification of SR and collagen 1 IHC did not show differences between challenged and control animals. This controversy between mRNA and protein expression may indicate that mRNA expression levels do not reflect the polymerization and formation of collagen fibers. Furthermore, these data indicate that to investigate the role of ATGL in development of hepatic fibrosis, the feeding period of 5 weeks MCD and 17 weeks HFHC diet has to be extended (or other experimental models such as carbon tetrachloride challenge should be taken into account).
Of note, it was also demonstrated that global ATGL KO mice are protected from acute ER stress development because of OA enrichment in the liver. [26] Increased hepatic OA concentration in MCD-challenged liver-specific ATGL KO mice suggests that reduced ER stress development is linked to a similar mechanism.

F I G U R E 7 Knockdown of ATGL protects against LPS-induced inflammation and prevents LPS-induced cytokine secretion in vitro. (A)
Six hours after LPS challenge, gene expression of inflammatory markers Mcp1, Cxcl2, and Tnfα were significantly increased in HepG2 WT cells and did not change in HepG2 ATGL KD cells. Data represent means ± SD and are shown relative to the expression levels of untreated HepG2 WT cells. * indicates a significant difference from untreated HepG2 WT cells; $ indicates a significant difference from LPS-treated HepG2 WT cells; # indicates a significant difference from untreated HepG2 ATGL KD cells; p < 0.05. (B) Proteome profiling revealed that HepG2 ATGL KD cells secrete fewer cytokines/chemokines upon LPS challenge. Cells were treated with LPS for 6 hours. Thereafter, LPS-containing medium was exchanged with an empty medium. After 24 hours, the medium was used for proteome profiling. LPS-treated HepG2 cells secreted more osteopontin, IL-8, Macrophage inhibitory cytokine-1, and VEGF than HepG2 LPS-treated ATGL KD cells. (C) Human macrophage cell line THP1 was cultured in a medium taken from HepG2 WT cells and HepG2 ATGL KD cells treated with LPS for 6 hours. mRNA expression levels of IL-1b, IL-6, and Mcp1 were increased in THP1 cells treated with a medium from LPS-treated HepG2 WT cells but not in THP1 cells treated with a medium from LPS-treated HepG2 ATGL KD cells. Data represent means ± SD and are shown relative to the expression levels of THP1 cells cultured in medium from HepG2 WT cells. * indicates a significant difference from THP1 cells cultured in medium from HepG2 WT cells; # indicates a significant difference from THP1 cells cultured in medium from HepG2 ATGL KD cells; $ indicates a significant difference from THP1 cells cultured in medium from LPS-treated HepG2 WT cells; p < 0.05 In addition to OA, α-linolenic and γ-linolenic acid are also among the most abundant FAs found in livers of MCD-fed ATGL LKO mice ( Figure 5C). These three FA species are capable of activating PPARδ. [39] Accordingly, PPARδ DNA binding activity was also increased in this group of animals. Anti-inflammatory effects of PPARδ have been demonstrated in several preclinical NAFLD studies. [40][41][42] Although PPARα activation depends on FAs derived from intracellular TG hydrolysis, PPARδ activation does not appear to be dependent on intracellular TG catabolism. [16] Therefore, augmented PPARδ activation in ATGL LKO mice may compensate for impaired PPARα signaling, thereby attenuating hepatic inflammation and subsequent progression from NAFLD to NASH. This observation is in line with the finding that PPARδ directly compensates for the absence of PPARα in regulating FA homeostasis in human and murine skeletal muscle cells. [22] Impaired PPARα signaling (seen in ATGL global [28] as well as the ATGL LKO mice) was also found to be present in patients with NASH, where its expression correlates negatively with progression of the disease. [43] Longitudinal analysis showed that increase in expression of PPARα and its target genes was associated with histological improvement, underlining the use of PPARα agonists in the treatment of NASH. [43] Also, PPARδ expression in patients with NAFLD was investigated in several studies. [43][44][45] Interestingly, the findings are Representative picture of NFkB subunits p50 and p65 immunoblots. Six hours after LPS challenge, protein expression of the NFkB subunits p50 and p65 was significantly decreased in LPS-challenged cells pretreated with GW0742. (C) Gene expression of NFkB downstream targets. Six hours after LPS challenge, gene expression of inflammatory markers Mcp1, Cxcl2, and Tnfα were significantly decreased in LPS-challenged cells pretreated with GW0742. Data represent means ± SD and are shown relative to the expression levels of unchallenged HepG2 cells. * indicates a significant difference from untreated control cells (Ctrl); $ indicates a significant difference from LPS-treated HepG2 cells; p < 0.05 somewhat controversial. Although in a small study it was shown that PPARδ expression was increased in patients with NASH, [44] in a larger cohort of patients with NAFLD, PPARδ expression remained unchanged. [43] In line with this, we did not see changes in PPARδ protein expression (data not shown) in a cohort consisting of patients aged ≥18 years with biopsy-proven NASH, grouped into steatosis (F0), NASH early stage (F1-2), and NASH advanced stage (F3-4). [25] However, if the patients are grouped based on the stage of fibrosis (NASH F1, NASH F2, and NASH F3), PPARδ protein expression tended to be reduced (Supporting Figure S11), arguing for a potential role of PPARδ as a pharmacological target to counteract NAFLD progression. Thus, a clinical study investigating the effect of the PPARδ agonist seladelpar in patients with NASH (F1-F3) showed an improvement of liver enzymes after 52 weeks of treatment. [46] Also, pharmacological inhibition of ATGL in adipose tissue with the specific inhibitor Atglistatin was proven to be beneficial in counteracting high-fat diet-induced fatty liver development. [47] Inhibition of adipose tissue lipolysis may thus be favorable in patients who are strongly obese through reduction of FA flux to skeletal muscle and liver, thereby preventing ectopic lipid accumulation and insulin resistance. Here, we show that loss of ATGL function also has beneficial effects in a lean mouse model of steatohepatitis. MCD diet promotes intrahepatic lipid accumulation mainly through increased hepatic FA uptake and decreased VLDL secretion, and, apparently, lipotoxic effects can be reduced by inhibition of lipolysis in hepatocytes promoting FA storage and neutralization in the form of inert TG.
Interestingly, despite its close homology with adiponutrin/PNPLA3 (which has been found to be associated with development and progression of NAFLD/NASH and many other liver diseases [48] ), for ATGL/PNPLA2 polymorphisms, no convincing evidence for a role in fatty liver development was found. However, it has been shown that PNPLA3 effectively competes with ATGL for α/β hydrolase domain containing 5 (ABHD5/CGI-58), an essential coactivator of ATGL and that PNPLA3 I148M is more effective in this regard, [49] thus weakening lipolytic function of ATGL, favoring lipid accumulation. In contrast to PNPLA3 variants, PNPLA2 variants were associated with anthropometric and metabolic parameters such as fat mass and subcutaneous adipose tissue. [50] In vitro experiments revealed that loss of ATGL in hepatocytes has a direct anti-inflammatory effect, reducing secretion of proinflammatory cytokines and thereby ameliorating immune cell recruitment/activation, corroborating findings in liver ATGL deficiency. In line with this, ATGL-deficient macrophages have also been identified to have reduced inflammatory activity [51] and increased lipolysis, induces macrophage migration. [52] In conclusion, our study provides profound evidence that inhibition of ATGL exclusively in the liver decelerates the progression of NAFLD from simple steatosis to more severe disease stages, such as steatohepatitis. Therefore, ATGL inhibitors may be considered as useful strategy to combat NASH development.