PNPLA3(I148M) Inhibits Lipolysis by Perilipin-5-Dependent Competition with ATGL

The single nucleotide polymorphism I148M of the lipase patatin-like phospholipase domain containing 3 (PNPLA3) is associated with an unfavorable prognosis in alcoholic and non-alcoholic steatohepatitis (ASH, NASH), with progression to liver cirrhosis and development of hepatocellular carcinoma. In this study, we investigated the mechanistic interaction of PNPLA3 with lipid droplet (LD)-associated proteins of the perilipin family, which serve as gatekeepers for LD degradation. In a collective of 106 NASH, ASH and control liver samples, immunohistochemical analyses revealed increased ballooning, inflammation and fibrosis, as well as an accumulation of PNPLA3–perilipin 5 complexes on larger LDs in patients homo- and heterozygous for PNPLA3(I148M). Co-immunoprecipitation demonstrated an interaction of PNPLA3 with perilipin 5 and the key enzyme of lipolysis, adipose triglyceride lipase (ATGL). Localization studies in cell cultures and human liver showed colocalization of perilipin 5, ATGL and PNPLA3. Moreover, the lipolytic activity of ATGL was negatively regulated by PNPLA3 and perilipin 5, whereas perilipin 1 displaced PNPLA3 from the ATGL complex. Furthermore, ballooned hepatocytes, the hallmark of steatohepatitis, were positive for PNPLA3 and perilipins 2 and 5, but showed decreased perilipin 1 expression with respect to neighboured hepatocytes. In summary, PNPLA3- and ATGL-driven lipolysis is significantly regulated by perilipin 1 and 5 in steatohepatitis.


Introduction
Fatty liver disease (FLD, novel synonym: steatotic liver disease/SLD), the accumulation of lipid droplets (LDs) in hepatocytes, is a steadily increasing health risk affecting more than one-third of the population in the western world. Chronic hepatitis C, alcohol abuse and metabolic syndrome are the main causes of hepatocyte steatosis (alcoholic/nonalcoholic fatty liver disease: AFLD/NAFLD) [1]. NAFLD is the most common liver disease in the western world, with a prevalence of 25% [2]. In some patients, bland steatosis progresses to steatohepatitis (NASH, in analogy to the term ASH) [3] and liver cirrhosis, which bears an increased risk of developing hepatocellular carcinoma (HCC). The histological criteria that distinguish steatohepatitis from bland steatosis include hepatocyte ballooning, commonly regarded as the hallmark of steatohepatitis, as well as lobular inflammation [4,5].
For immunohistochemistry, 59 formalin-fixed, paraffin-embedded liver biopsies of 23 patients with NASH, 19 patients with ASH, including 4 with alcoholic and metabolic (BASH) as well as 21 liver biopsies from control patients (normal liver, HCV; unclear cases, see Table 1) were used and compared to a collective of 47 ASH cases previously described [32]. These samples exhibited the PNPLA3 genotypes 23× C/C, 17× C/G, and 7× G/G, respectively (study in accordance with the ethics committee of the University of Heidelberg (no: S280-2011). 3.00 ± 0.00 1.00 ± 0.00 5.00 1.00 ± 0.00 0.0 Summary of the clinical parameters of the human liver samples analyzed. Values are given as mean ± SD.

Mouse Models
ATGL knockout mice were received from H. Haemmerle [19] and were kept in accordance with the regulations of the local ethics committee of Heidelberg (AZ 35-9185.81/ G-72/16).
Cryopreserved human tissue sections with a thickness of 5 µm were placed in tissue homogenizing kit CKMix tubes (Precellys) and then lysed with a Precellys 24 homogeniser 2 times at 6500 rpm for 20 s.

Antibodies
The antibodies and conditions are described in the Supplementary Materials, Materials and Methods.

Immunofluorescence Microscopy
HEK293T, Huh7, and HepG2 cells were seeded on 13 mm (∅) coverslips in 12-well plates, transfected the next day and fixed after an additional incubation period of 24 h or 48 h. After washing with PBS, cells were fixed for 10 min with 3.7% formaldehyde in PBS. Subsequently, cells were washed with PBS, permeabilized, and blocked with 10% FBS, 0.1% Triton X-100 in PBS at 37 • C. Cells were then incubated with the primary antibody in blocking solution at RT for 1 h. After a brief wash with 0.1% Triton X-100 in PBS, cells were either incubated with the secondary antibodies (Alexa Fluor ® 546 goat anti-mouse IgG (H + L), Alexa Fluor ® 635 goat anti-rabbit IgG (H + L), Alexa Fluor ® 635 goat anti-mouse IgG (H + L) or without (EGFP-and mRFP fusion proteins), the nuclear dye DAPI and the LD-dye BODIPY TM 493/503 in 1% FBS, 0.1% Triton X-100 in PBS at RT for 1 h. After a final washing step with 0.1% Triton X-100 in PBS, cells were mounted with MOWIOL and images were taken with a Leica SP8 confocal microscope using either a 40× 1.30 NA Oil CS2 HC Plan Apo or a 63× 1.40 NA Oil CS2 HC Plan Apo objective operating at 25 • C. Both objectives were operated with Type F immersion liquid (Leica Microsystems). Leica Application Suite X was used to acquire the images. LD-size was then analysed with ImageJ (v.1.52n; https://imagej.net/downloads, accessed on 1 December 2022). For detailed instructions, please see the Supplementary Materials, Materials and Methods.

Image Analysis
Stained sections were digitised with a NanoZoomer 2.0 HT (Hamamatsu) at a magnification of 40× and the intensity of the staining was quantified using the digital image analysis software QuPath (v.0.1.2) [34]. To determine the area occupied by LDs in biopsies, the digitised images were analysed with the software Aperio ImageScope (v.12.2.2.5015, Leica Biosystems, Wetzlar, Germany). For detailed instructions, please see the Supplementary Materials, Materials and Methods.

Transmission Electron Microscopy
Transmission electron microscopy of glutaraldehyde fixed liver biopsies was performed as previously described [28,29]. Images were taken on a JEOL JEM 1400 electron microscopy with a 4 k camera.

CRISPR/Cas
HEK293T cells did not harbour the PNPLA3(I148M) missense variant on either of the two alleles. Therefore, a previously described protocol [36] was used to generate PN-PLA3(I148M) knockin cell lines with two designed guides and oligonucleotides ligated into the plasmid pSpCas9(BB)-2A-Puro (PX459) V2.0. Sequences of the guide oligonucleotides used for cloning were PNPLA3_G2_for 5 -CACCGCCTTCAGAGGCGTGGTAAGT-3 , PN-PLA3_G2_rev 5 -AAACACTTACCACGCCTCTGAAGGC-3 , PNPLA3_G6_for 5 -CACCGG GGATAAGGCCACTGTAGAA-3 , PNPLA3_G6_rev 5 -AAACTTCTACAGTGGCCTTAT CCCC-3 . Cells were transfected, genomic DNA was isolated and subjected to mutation analysis with HRMA and sequencing. Guide 2 (G2) and 6 (G6) showed the highest efficiencies and were therefore used to generate single cell clones. Selection of positively transfected cells was performed with 1 µg/mL puromycin for one week and cells were further seeded at a density of one cell per well in a 96-well plate. Single-cell clones were continuously expanded and screened for successful insertion of the single-stranded donor oligonucleotide (ssODN) in a high throughput procedure with the HindIII test digest when a 6-well plate size was reached and finally genotyped by sequencing.

Lipolysis/Steatogenesis
Lipolysis was induced by a combination of 10 µM forskolin with 500 µM IBMX for 1-3 h and compared with the vehicle (DMSO). Steatosis was induced with 240 µM oleic acid BSA complex for up to one week and compared with BSA treatment alone.

Quantitative PCR
Cryopreserved samples were collected in tissue homogenizing kit CKMix tubes (Precellys). A total of 1 mL TRI Reagent TM Solution (Invitrogen) was added, and lysis was performed with a Precellys 24 homogeniser twice for 20 sec at 6500 rpm. RNA was isolated following the manufacturer's instructions. RNA was reverse-transcribed into cDNA using the high-capacity cDNA reverse transcription kit (Applied Biosystems™), diluted, and used for real-time PCR (QuantStudio 3 Real-Time PCR System, Applied Biosystems). For sequences of real-time primers, please see the Supplementary Materials, Materials and Methods.

Statistical Analysis
Data were visualized as bar diagrams or scatter diagrams, each bar representing mean ± standard deviation. The statistical analysis was performed with GraphPad Prism v.9.4.0 (673; version 9.4.0 (673) for Windows, GraphPad Software, San Diego, California USA, www.graphpad.com. First, the data were tested for normal distribution using the Shapiro-Wilk normality test. If the data were normally distributed, a t-test was performed. If the standard deviations of the mean values of two groups differed significantly, Welch's correction was also applied. If, on the other hand, the data were not normally distributed, the non-parametric Mann-Whitney test was used. p-values < 0.05 (*), p ≤ 0.001 (***), p ≤ 0.0001 (****) were considered significant, whereas p-values > 0.05 not significant (n.s.).

Increased Inflammation and Fibrosis in Patients with PNPLA3(I148M)
Since the PNPLA3(I148M) polymorphism is associated with steatohepatitis and increased fibrogenesis in ASH and NASH patients, we analysed the histology as well as the expression pattern of the LD-associated proteins of the perilipin-family in liver biopsies of patients with steatotic liver disease with respect to PNPLA3 status. Immunohistochemistry against perilipins and in part together with PNPLA3 was performed with a total of 106 liver biopsies. In total, 23 NASH, 19 ASH (including 4 BASH-cases), 9 HCV and 8 control patients (Table 1) were analysed and compared with 47 ASH liver biopsies from a previous clinicopathological study [32].
We could thereby recapitulate a positive correlation of PNPLA3(I148M) with steatosis, inflammation, the amount of microgranulomas, ballooning and fibrogenesis [32]. Generally, steatosis was strongly highlighted by perilipin 1 and 2 staining at LDs, whereas perilipins 3, 4 and 5 showed predominant cytoplasmic and only faint LD-staining. There was no significant difference in the general intensity of perilipin stainings between liver biopsies of different PNPLA3 genotypes (Figures 1, A1 and A2, Table 2.   However, strikingly, heterozygous PNPLA3(I148M)-carriers showed a weak, and homozygous carriers an even stronger localization of PNPLA3 at LDs in zones 1 and 2 of the liver which were negative or only weakly positive for perilipin 1 ( Figure 1A,C), whereas in patients with PNPLA3(I148), PNPLA3 appeared almost exclusively in granular structures. In liver biopsies of PNPLA3(I148M)-carriers, perilipin 5 was more frequently localized at LDs and less in the cytoplasm ( Figure 1D). The degree of steatosis was slightly higher in homozygous PNPLA3(I148M)-carriers ( Figure 1E), accompanied by a significant increase in LD-size ( Figure 1F). Ballooned hepatocytes were more frequently found in PNPLA3(I148M) carriers ( Figure A1), and were strongly positive for perilipin 2 in both NASH and ASH livers irrespective of the PNPLA3 genotype. These hepatocytes were also positive for PNPLA3 and perilipin 5; however, negative or only mildly positive for perilipin 1 (Figure 2A-C). In contrast, adjacent non-ballooned hepatocytes showed a significantly higher perilipin 1 expression (

Reduced Perilipin 1 Expression in PNPLA3(I148M)-Hepatocytes In Situ
In order to investigate the underlying molecular mechanisms of PNPLA3 action, we examined the expression of PNPLA3 and perilipins at the protein and transcript level. Western blot analysis of 15 non-neoplastic livers of patients resected for colorectal metastases showed no obvious association between PNPLA3-genotype and the amount of PNPLA3protein ( Figure 2F). Perilipin 5 was slightly and perilipin 1 strongly reduced in livers of hetero-and homozygous carriers of the PNPLA3-polymorphism ( Figure 2F). Real-time PCR analysis of the same samples showed reduced levels of PLIN1 and PNPLA3 mRNA in heteroand homozygous carriers of the PNPLA3-polymorphism ( Figure 2G,H). Interestingly, in the collective of 47 ASH patients, PLIN5 mRNA was significantly downregulated, whereas PLIN1 was only mildly downregulated in liver biopsies of patients with PNPLA3(I148M) without reaching significance, PLIN2 and PLIN 3 mRNA levels were not significantly altered [32]. Similar changes were observed on a protein level using immunohistochemistry.
Our findings confirm our histological data and suggest a regulatory link between PNPLA3 and perilipin 1 and 5 on the protein and mRNA level, respectively, in human livers in situ.

Opposing Effects of Perilipin 1 and 5 on the Recruitment of PNPLA3 to LDs
Since perilipin 5 physically interacts with ATGL, the protein sharing the highest similarity with PNPLA3 [37], we hypothesized that perilipin 5 may form complexes with ATGL and PNPLA3. Therefore, PNPLA3 was overexpressed either in HEK293T or Huh7 cells alone and together with perilipin 1, 2, 3, 5, as well as ABHD5. In coimmunoprecipitation analyses, only perilipin 5 coprecipitated together with PNPLA3 in cell culture (Figures 3A and A3) and in human liver ( Figure A4) indicative of a direct or indirect interaction.
Repetition of the in vitro studies with ATGL revealed interactions of ATGL with perilipin 5 and ABHD5 ( Figure 3B) as already shown for skeletal muscle [22]. In immunofluorescence microscopy of transfected HepG2 cells ( Figure 3C,D), perilipin 5 exhibited cytoplasmic as well as weak LD-localization and colocalized with PNPLA3 which was almost entirely LD-associated ( Figure 3C). Co-expression of PNPLA3 and perilipin 5 resulted in a translocation of perilipin 5 from cytoplasmic to LD-bound localization and a strong colocalization of both proteins at LDs ( Figure 3C) as observed for perilipin 5 with ATGL ( Figure A5). This was confirmed by the colocalization of PNPLA3 with perilipin 5 at the same LDs in human liver cryosections in situ ( Figure A4B) but not with other perilipins.  with PNPLA3 or ATGL alone or together with FLAG-tagged PLIN1, 2, 3, 5 or ABHD5. PNPLA3 co-immunoprecipitates only with perilipin 5 (PLIN5), but not with other perilipins, whereas ATGL co-immunoprecipitates with perilipin 5 and ABHD5. The asterisk indicates a weak ABHD5-signal shifted in height. Molecular weight in kDa is given on the left. (C,D) Immunofluorescence microscopy of HepG2 cells transfected with FLAG-tagged perilipin 5 (green) or perilipin 1 (green) alone or with PNPLA3 (red) (LDs: BODIPY, violet; nuclei: DAPI, blue). Cells co-expressing PNPLA3 and perilipin 5 show strong colocalization at LDs. Co-expression of PNPLA3 and perilipin 1 shows granular cytoplasmic localization of PNPLA3 apart from LDs. Scale bars: 15 µm, and 5 µm for magnified images.
As shown for constitutive perilipins, perilipin 1 localized to LDs when overexpressed in cells ( Figure 3D). Surprisingly, co-expression of PNPLA3 and perilipin 1 led to a complete displacement of PNPLA3 from LDs to granular cytoplasmic structures ( Figure 3D). This mechanism appears to be specific for PNPLA3 since perilipin-1-induced displacement from LDs was not observed for ATGL ( Figure A5). Perilipin 1 may thus provide a mechanism to regulate PNPLA3-binding to LDs.

PNPLA3 Interacts with ATGL, the Rate-Limiting Enzyme in Lipolysis, in a Perilipin 5-Dependent Manner
Since PNPLA3 and ATGL both localized to LDs ( Figures 3C and A5), we carried out studies to determine whether there was a possible interaction. In addition to a coprecipitation of ATGL with PNPLA3 ( Figure 4A), both proteins also strongly colocalize to LDs ( Figure 4B), suggesting the role of PNPLA3 in regulating ATGL.
As perilipin 5 was the only perilipin shown to co-precipitate with PNPLA3 and ATGL ( Figure 4A), the impact of perilipin 5 on the interaction between PNPLA3 and ATGL was examined. In the absence of perilipin 5, PNPLA3 and ATGL interacted only weakly, whereas the presence of perilipin 5 led to a strong increase in PNPLA3 and ATGL interaction in a dose-dependent manner ( Figure 4C). Phylogenetically, ATGL and PNPLA3 are the most closely related proteins within the PNPLA-family having 37% identical and a high degree of similar amino acids ( Figure 4D). Therefore, we hypothesized that ATGL may also form homodimers. This was confirmed by ATGL-ATGL-complexes detected by co-immunoprecipitation using differently tagged ATGL-constructs ( Figure 4E). However, small amounts of PNPLA3 were able to displace ATGL from the ATGL-ATGL-complexes ( Figure 4F). Therefore, PNPLA3 may compete with ATGL providing a key mechanism to how PNPLA3 may interfere with the lipid metabolism.
When compared to wild-type PNPLA3, PNPLA3(I148M) did not show an altered interaction with perilipin 5 and ATGL or an altered localization in vitro ( Figure A6). PN-PLA3(I148M) leads to the reduced hydrolysis of fatty acids and accumulation of TAGs [11], yet this does not sufficiently explain the increased fat deposition and steatohepatitis in livers of I148M-carriers [17]. Therefore, we investigated whether ATGL-activity may be modulated by the physical interaction with PNPLA3. Overexpression of ATGL resulted in a significant reduction in the average LD size (see also Figure A7). In contrast to overexpression of PNPLA3, overexpression of PNPLA3(I148M) led to a significant increase in the average LD-size ( Figure 5A,B). In contrast, lack of ATGL as in ATGL-deficient mice of 8 weeks of age demonstrated lack of ATGL, but not yet significant changes in expression of perilipins 2-5, PNPLA3 and ABHD5 as measured by immunoblot. ATGL-deficient mice at 12-13 weeks of age demonstrated microvesicular steatosis of the liver shortly before cardiac death, with increased expression of perilipin 2 and slight downregulation of perilipin 5, as measured by immunohistochemistry. Prominent microvesicular steatosis in ATGl-deficient mice of that age was also confirmed by transmission electron microscopy, which was not detected in wild-type mice of the same age. Due to early cardiac death of ATGL-deficient mice at about 12-13 weeks, older ages could not be analysed ( Figure A9).   Co-expression of ATGL and PNPLA3 led to a significant increase in LD-size compared to cells expressing ATGL alone, which was even higher in the presence of PNPLA3(I148M) suggesting a negative effect on the lipolytic activity of ATGL.
To further investigate the negative effect of PNPLA3(I148M) on lipolysis, PNPLA3(I148M) knockin cells were generated ( Figure 5C). Unmodified PNPLA3-control cells formed very few small LDs when seeded at a low density ( Figure 5D,E). In PNPLA3(I148M)heterozygous cells however, or cells carrying PNPLA3(I148M) and a knockout of one PN-PLA3 allele, LDs were significantly larger ( Figure 5E). Following steatogenic treatment with oleic acid, all three cell lines showed comparable increases in LD size ( Figures 5E and A8). Lipolysis induced by forskolin and IBMX reduced the LD-size differently among cell lines, which was highest in control cells, and less pronounced in heterozygous I148/I148M cells. No significant difference was observed in I148M/-cells treated with vehicle or lipolytic stimuli, suggesting that the missense variant had the highest negative impact on cAMP-mediated lipolytic activity ( Figure 5D,E).

Perilipin 1 Displaces PNPLA3 from the ATGL Complex and Drives ATGL-Mediated Lipolysis
To investigate the impact of perilipin 1 on the stability of the PNPLA3-ATGL-complex, co-IP and subsequent immunoblot were carried out. Increasing amounts of perilipin 1 led to a dose-dependent release of up to 80% of PNPLA3 from the PNPLA3-ATGL complex ( Figure 6A). Cells co-expressing perilipin 1 and PNPLA3 exhibited granular, cytoplasmic localization for PNPLA3, which was negative for ATGL, whereas LDs were positive for perilipin 1 but not PNPLA3 ( Figure 6B). In addition, the number of ATGL-positive foci that represented the small residual LDs were strongly reduced ( Figure A7). In contrast, when the lipolytic activity of ATGL was inhibited, the number of such foci increased. With normal ATGL-activity, no ATGL-positive foci were detectable and only diffused cytoplasmic localization was visible ( Figure 5A). The perilipin 1-mediated displacement of PNPLA3 from the PNPLA3-ATGL complex appeared to reduce the negative impact of PNPLA3 on ATGL and, therefore, significantly increased ATGL-mediated lipolysis.  In the absence of PNPLA3 as part of the lipolytically active ATGL-complex, the activity is significantly higher, so that the degradation of triacylglycerides and sterol esters is elevated, and LDs become smaller. In carriers of the PNPLA3(I148M) polymorphism, the perilipin 1-mediated regulatory mechanism that displaces PNPLA3 from LDs appears to be disturbed.
Finally, the overexpression of PNPLA3 or PNPLA3(I148M) was examined to determine the effects on the expression of perilipin 1-5, the lipases ATGL, HSL, and PNPLA3, as well as the co-activator of ATGL, ABHD5. Overexpression of either PNPLA3 or PNPLA3(I148M) strongly induced perilipin 1 irrespective of cell type ( Figure 6C), pointing to a possible feedback mechanism in which PNPLA3 induces perilipin 1 that replaces PNPLA3 from LDs ( Figure 6D,E).

Discussion
Our comprehensive in situ and in vitro data provide evidence for the pivotal role of deregulated basal lipolysis in the progression of steatosis to steatohepatitis. In this process, we could demonstrate that PNPLA3 competes with ATGL for perilipin 5 binding within lipolytic perilipin 5-ATGL complexes. We are the first to show that PNPLA3 physically interacts with perilipin 5 and ATGL at larger LDs found in carriers of the PNPLA3(I148M) polymorphism. In addition, we are the first to be able to localize PNPLA3 in a large collective of patients with steatohepatitis in human liver biopsies in situ with respect to different PNPLA3-polymorphisms. We demonstrated that PNPLA3 competitively displaces ATGL from LDs and thereby strongly reduced the lipolytic activity, thereby favouring LDaccumulation. Both perilipins 1 and 5 have been shown to be key regulators of lipolysis in different tissues. In adipocytes, perilipin 1 regulates lipolysis via HSL and ABHD5, whereas in cell types using oxidative energy supply, perilipin 5 [31] controls lipolysis by interacting with ATGL and ABHD5 [41]. Strikingly, co-expression of PNPLA3 and perilipin 5 led to an increased association of both proteins with LDs. Although perilipin 1 did not interact with PNPLA3, co-expression resulted in an almost complete displacement of PNPLA3 from LDs. We hypothesize that perilipin 1 may not function as a LD-barrier protecting against degradation of TAGs [42], but rather as a factor displacing PNPLA3 from the PNPLA3-ATGL-complex, thus reducing the negative impact of PNPLA3 on ATGL-mediated lipolysis and switching to an alternative type of lipolysis.
We have demonstrated a possible pathophysiologic mechanism of how the polymorphism I148M of the lipase PNPLA3 (for epidemiologic studies see [6][7][8]) may trigger NAFLD-progression on the molecular level. We and others have shown that PN-PLA3(I148M) leads to a reduction in the lipolytic activity and an accumulation of TAGs in LDs of increased sizes via an indirect effect of PNPLA3 on ATGL-mediated lipolytic activity [9,10,13,43]. Pnpla3 knockout mice did not develop steatosis [11,12]. In contrast, a liver specific knockin or overexpression of PNPLA3(I148M) in mice induced FLD, not observed for PNPLA3 [13,14]. Consistent with these findings, the presence of PNPLA3 led to a marked reduction in ATGL-mediated lipolytic activity with significantly increased LDs, an effect even stronger for PNPLA3(I148M). We have shown that the mechanism for the reduced amount of lipolytically active ATGL at LDs may be due to the disruption of homodimeric or -multimeric complexes formed by ATGL by displacing ATGL in a perilipin 5-dependent manner. This displacement, mediated by an interaction of PNPLA3 with ATGL, was also observed by Wang et al. [18]. Strikingly, this suggests that this weak interaction [18] could be dramatically increased in the presence of perilipin 5 and decreased by perilipin 1, indicating opposing roles of perilipin 1 and 5 under these conditions. Sequestration of ABHD5 by PNPLA3 has been shown to prevent activation of ATGL, yet only a very weak interaction could be demonstrated [18,44]. Both the weak negative influence of PNPLA3 and the strong negative influence of PNPLA3(I148M) on ATGL-mediated lipolytic activity have been recently demonstrated in Huh7 cells [18]. However, this reduction in total lipolytic activity could not be explained by the loss of lipolytic activity of PNPLA3 caused by the missense mutation [10] since PNPLA3 exhibits only low lipolytic activity compared to ATGL and the loss of this low activity does not provide a sufficient explanation for the strong phenotype in PNPLA3(I148M)-carriers. Two possible mechanisms may lead to a reduction in total lipolytic activity by PNPLA3: First, PNPLA3 may stoichiometrically compete with ATGL for LD-binding and thus easily block potential binding sites for ATGL without the interaction of both proteins. Second, PNPLA3 may interact directly with ATGL, thus affecting the activity of ATGL. An interaction between PNPLA3 and ABHD5, was proposed as a model for brown adipocytes, in which PNPLA3 sequesters ABHD5 and therefore prevents binding and activation of ATGL, thereby indirectly suppressing lipolysis [44]. A similar mechanism in suppressing lipolysis by sequestration of ABHD5 was shown for perilipins 1 and 5. In livers of untreated mice with ATGL deletion, we could only observe slightly increased perilipin 5, ABHD5, and PNPLA3 levels, with respect to the control mice. Interestingly, in the livers of hepatocyte-specific perilipin 2 knockout mice treated with an HFD and CDAA diet, leading to steatosis and steatohepatitis, upregulation of perilipin 5 was observed concomitant to a decrease in inflammation, cell death and fibrosis (parallel manuscript in preparation).
Our findings demonstrate important implications for the mechanism by which PN-PLA3(I148M) promotes progression to steatohepatitis in FLD, irrespective of the causative agent. We hypothesize that cell stress caused by supernutrition, alcohol or HCV together with a block of basal lipolysis caused by PNPLA3(148M) leads to hepatocyte ballooning and inflammation. In ballooned hepatocytes, the hallmark of steatohepatitis, perilipin 1 was markedly reduced, whereas perilipin 2 was strongly expressed but localized diffusely in the cytoplasm. Cytoplasmic localization of perilipin 2 in ballooned hepatocytes is a phenomenon that has not been observed in any other cell type under physiological conditions and can therefore be interpreted as a sign of a severely damaged/dying cell.
Our study identified several key players that are dysregulated in steatohepatitis and therefore may have both a diagnostic and a therapeutic potential. Generally, the presence of ballooned hepatocytes in non-heavy drinkers indicates a heterozygous or homozygous PNPLA3(I148M)-mutation [35]. The antibody staining of accumulated PNPLA3 at LDs in situ [9] in PNPLA3(I148M) carriers may be used diagnostically. The identification of altered ATGL activity makes this mechanism a potential drug target and the therapeutic intervention may be guided based on the PNPLA3-status of the patient, thus making this method applicable for personalized therapy. Since perilipin 1 is strongly downregulated in severe ASH and at the cellular level in ballooned hepatocytes, and since we could unravel under in vitro conditions that perilipin 1 is able to replace PNPLA3 from the ATGL-complex, it is possible that perilipin 1 may also be a possible therapeutic target in steatohepatitis in patients with PNPLA3(I148M).

Conclusions
With this study involving human, mouse and cell culture, we could unravel the LDassociated protein perilipin 5 as a binding partner of PNPLA3 and ATGL in vitro and in situ, and the functional role of perilipin 5 ATGL/PNPLA3 complex formation during steatohepatitis, especially in the frequent PNPLA3 polymorphism I148M. Our data point towards an important role of deregulated lipolysis during the formation of ballooned cells in steatohepatitis, triggering the progression of inflammation and fibrosis.