p62/Sequestosome-1 Is Indispensable for Maturation and Stabilization of Mallory-Denk Bodies

Mallory-Denk bodies (MDBs) are hepatocytic protein aggregates found in steatohepatitis and several other chronic liver diseases as well as hepatocellular carcinoma. MDBs are mainly composed of phosphorylated keratins and stress protein p62/Sequestosome-1 (p62), which is a common component of cytoplasmic aggregates in a variety of protein aggregation diseases. In contrast to the well-established role of keratins, the role of p62 in MDB pathogenesis is still elusive. We have generated total and hepatocyte-specific p62 knockout mice, fed them with 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) to induce MDBs and allowed the mice to recover from DDC intoxication on a standard diet to investigate the role of p62 in MDB formation and elimination. In the absence of p62, smaller, granular and less distinct MDBs appeared, which failed to mature to larger and compact inclusions. Moreover, p62 deficiency impaired the binding of other proteins such as NBR1 and Hsp25 to MDBs and altered the cellular defense mechanism by downregulation of Nrf2 target genes. Upon recovery from DDC intoxication on a standard diet, there was an enhanced reduction of p62-deficient MDBs, which was accompanied by a pronounced decrease in ubiquitinated proteins. Our data provide strong evidence that keratin aggregation is the initial step in MDB formation in steatohepatitis-related mouse models. Interaction of p62 with keratin aggregates then leads to maturation i.e., enlargement and stabilization of the MDBs as well as recruitment of other MDB-associated proteins.


Introduction
The fate of misfolded proteins gained significant relevance in a variety of diseases characterized by the accumulation of abnormal proteins, collectively known as "protein aggregation diseases" [1,2]. The folding of proteins into their functionally active three-dimensional structures is among the most fundamental roles of a living cell [2]. Misfolding on genetic or toxic basis may lead to aggregation-prone proteins, being deposited in the cytoplasm, nucleus or extracellular compartments [1,3]. These inclusion bodies can serve as morphological hallmarks of a variety of diseases. They include among others, neuronal Lewy bodies in Parkinson's disease, neurofibrillary tangles in Alzheimer's disease, TDP-43 aggregates in amyotrophic lateral sclerosis, and Mallory-Denk bodies (MDBs) in steatohepatitis as well as other chronic liver disorders [3][4][5]. As a reflection of a common pathogenesis, these inclusions display, in addition to their diseasespecific protein "backbone", a common molecular composition, since they collectively contain the multifunctional stress and adaptor protein p62/Sequestosome-1 (p62) and ubiquitin (Ub) as constant constituents.
In this study, we focused on the function of p62 in the formation of MDBs in vivo. MDBs are hepatocytic protein inclusions found in several chronic liver diseases, especially alcoholic (ASH) or non-alcoholic (NASH) steatohepatitis, idiopathic copper toxicosis, chronic cholestasis, and hepatocellular carcinoma (HCC) [5]. MDBs can be induced in mice by chronic treatment with chemicals such as griseofulvin or 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) [5,6]. Of note, both human and murine MDBs are reversible; they disappear in ASH after alcohol cessation or in mice upon recovery on a standard diet after prolonged DDC intoxication [7].
MDBs mainly consist of cytoskeletal intermediate filament proteins keratin 8 (K8) and keratin 18 (K18), p62 and ubiquitin [5,8,9]. Furthermore, they contain a high molecular weight component that is recognized by the M M 120-1 antibody. This component has so far been exclusively detected in human and murine MDBs and, therefore, serves as a specific marker of MDBs [10]. In addition to these major MDB constituents, several proteins have been found as facultative MDB components, such as heat shock proteins 25 and 70 (Hsp25 and Hsp70) and the structural homolog of p62 called Neighbor of BRCA1 (NBR1) [5,11]. In MDBs, keratins are hyperphosphorylated, transamidated, and partially degraded [12][13][14][15]. Moreover, keratins (K8 in particular) exhibit a conformational change from predominantly α-helical to cross βsheet (amyloid-like) structure [14,16]. Earlier in vitro studies on MDB formation showed that p62 binds to ubiquitinated, misfolded keratins [9,17,18]. It has been previously reported that p62 binds polyubiquitinated proteins via its UBA (ubiquitin associated) domain, shapes them into sequestosomes and shuttles them to degradation pathways, such as the ubiquitin proteasomal system (UPS) for ubiquitinated and soluble proteins or the autophagic machinery, which is responsible for bulk degradation of larger structures, such as protein aggregates or organelles [19][20][21]. Also, p62 itself can form aggregates such as intracellular hyaline bodies (IHBs) in livers of patients with idiopathic copper toxicosis and HCC [22]. In these liver diseases, transition stages between keratin-negative IHBs and keratin-positive MDBs have been observed [22,23]. This led to the conclusion that aggregates of p62 might serve as a matrix for the incorporation of "abnormal" keratins eventually leading to MDB formation [18]. Whether this mechanism is also relevant to MDB pathogenesis in ASH, NASH or related animal models is not clear [24,25]. It has been previously argued that soluble oligomeric proteins are detrimental to cells [1], and p62 may play a protective role by packing them into insoluble and less toxic aggresomes or sequestosome-like aggregates [26]. In addition to its role in MDB formation, p62 may also be involved in the elimination of MDBs. It has been suggested that MDBs are degraded by the UPS and autophagy, and p62 may play a role in both processes [27]. Although aggregation properties of p62 have been intensively investigated in cell culture experiments, no data exist on p62's in vivo role in disease-related protein aggregates. Since p62 is expressed in most, if not all, cell types and has been shown to play a role in a variety of cellular processes, such as fat metabolism, autophagy, and cell proliferation [28,29], we had to distinguish between cell autonomous roles of p62 in MDB formation in hepatocytes from possible indirect effects mediated by other cells (e.g., by bile duct epithelia, macrophages, inflammatory cells) or systemic metabolic effects. Therefore, we have generated p62 total (p62 -/-)-and hepatocyte-specific (p62hep -/-)-knockout mice and intoxicated them with DDC to induce MDBs to delve into the in vivo role of p62 in MDB formation.

Materials and Methods
Generation of Total (p62 -/-) and Hepatocyte-Specific (p62hep -/-) p62-Knockout Mice We generated p62-knockout mice using the Cre/loxP strategy. 5' fragment of p62/Sqstm1 gene (NC_000077) containing exons 1-5 was PCR-amplified and inserted via SacI into the pBluescript vector (kindly provided by C. Birchmeier; Max-Delbrueck-Centre for Molecular Medicine, Berlin, Germany) upstream of the loxP-neomycin-loxP cassette. An additional loxP site was introduced before exon 5 using the QuickChange Site-Directed Mutagenesis protocol (Stratagene, La Jolla, CA). To complete the construct, the PCR-amplified 3' fragment of p62/ Sqstm1 gene was ligated via SalI into the vector and the BamHI-SalI-cleaved product was introduced into a wildtype ES (embryonic stem) cell line with a 129S2/SvPas background by electroporation. Neomycin resistant colonies were screened for homologous recombination by longrange PCR and the positive clones with a single integration event were microinjected into C57BL/6 blastocysts. Chimeric males were mated with C57BL/6-Tg(Meu-Cre40) animals [30,31] to obtain mice which either lacked the neo-selection cassette but contained floxed exons 1 to 4 of the Sqstm1 gene (Sqstm1 tm1Biat , designated in the text as p62 floxed/floxed or p62 f/f ) or lacked the complete floxed region (Sqstm1 tm1.1Biat , designated in the text as p62 -/-) (S1 Fig). To generate a hepatocyte-specific p62-deficient mouse line (p62hep -/-), p62 f/f mice were mated with mice that expressed Cre-recombinase under the control of the albumin promoter (Alb-Cre) (S1 Fig) [30]. To obtain mice that were more susceptible to MDB formation, animals were crossed for several generations with wildtype 129S2/SvPas mice.
Genotyping was performed by PCR using genomic tail tip DNA and a combination of three sets of primers as shown in S1B and S1C Fig and S1 and S2 Tables. Littermate controls of the p62 -/and p62hep -/mouse strains were included in all experiments. To test for the presence of the Alb-Cre transgene, an additional Cre-specific PCR was performed (S1C Fig; S1 Table, PCR C). The PCR conditions maintained were (i) initial denaturation at 95°C for 5 minutes, (ii) denaturation at 95°C for 30 sec (iii) annealing at 60°C for 30 seconds, (iv) elongation at 72°C for 30 seconds for 35 cycles and (v) Final elongation at 72°C for 10 minutes.
All animals used in this study were generated in Institute of Laboratory Animal Science, University of Veterinary Medicine Vienna, Vienna, Austria.

Animal Experiments
MDBs were induced in two months old male p62 -/-, p62hep -/and their corresponding controls p62 f/f and p62hep -/mice by feeding DDC (0.1% DDC wt/wt in standard diet; Sigma Aldrich, Steinheim, Germany) for 8 weeks [18]. To study MDB disappearance, DDC-fed mice were allowed to recover for 4 weeks on standard diet. At least five male mice per genotype and treatment condition were chosen for the experiments. Mice were sacrificed by cervical dislocation and blood was collected by cardiac puncture. To test for the serum parameters of liver damage, serum was separated by centrifuging the blood sample for 15 minutes at 2000 x g. Then, the serum was diluted in saline (1:3) to obtain values of seven different serum parameters from a single run of automatic chemistry analyzer (Hitachi, Roche, Switzerland). Livers were immediately removed and cut into pieces which were (i) snap frozen in methyl butane precooled with liquid nitrogen for immunofluorescence staining, (ii) fixed in 4% formaldehyde solution for histological/immunohistochemical analysis or (iii) placed in liquid nitrogen for biochemical/ RNA expression analyses. Animal experiments were approved by the Austrian Federal Ministry of Science, Research and Economy in compliance with the Austrian Law for Welfare of Laboratory Animals with the license number: BMWF-66.010/0114-II/3b/2012 and BMWF-p62 and Mallory-Denk Bodies 66.010/0114-ii/3b/2013. All animals received humane care according to the criteria outlined in the "Guide for the Care and Use of Laboratory Animals" prepared by the National Academy of Sciences and published by the National Institutes of Health (NIH publication 86-23 revised 1985).

Extraction of RNA and Polymerase Chain Reactions
RNA extraction and cDNA preparation from livers were performed as described previously [18]. To check the efficiency of p62 deletion, reverse transcriptase PCR (RT-PCR) was performed (S1 Table). To determine the expression of selected genes, quantitative real-time PCR (qPCR) was performed as described previously [18]. Samples were analyzed in triplicates using specific primers (S2 Table). 18S rRNA or β-actin was used as internal control. The amplification efficiencies for the genes of interest and the internal control were approximately equal, the quantification was expressed as a ratio of 2 ΔCt/gene to 2 ΔCt/internal control , in which ΔCt/gene and ΔCt/internal control represent the difference between the threshold cycle of amplification for the gene of interest and internal control, respectively.

Immunofluorescence and Immunoelectron Microscopy
Immunofluorescence staining was performed as described previously [18] with antibodies to target proteins (S3 Table). Proteins with cross β-sheet structure were detected with the luminescent conjugated oligothiophene h-HTAA [16]. Liver sections were analyzed with a LSM 510 META confocal laser scanning microscope (Carl Zeiss, Oberkoechen, Germany) with Plan-Neofluar 40/0.75 objective and a 32-element photomultiplier tube array detector (META detector accessory). Images were processed with Zeiss LSM Image Browser 4.2 software (Carl Zeiss, Oberkoechen, Germany). Immunoelectron microscopy was performed as described previously using acetone-fixed cryosections and M M 120-1 antibody [17].

Protein Isolation and Western Blot
Whole liver tissue extracts were prepared using 3% sodium dodecyl sulphate (SDS)-containing buffer. Alternatively, soluble proteins were extracted with Triton-X-100 buffer and keratinenriched fractions were obtained as described previously taking advantage of their relative insolubility in Triton-X-100 and high-salt buffers [33]. Nuclear fractions were extracted using subcellular protein fractionation kit (Thermo Scientific, Vienna, Austria). All western blots were performed on whole tissue extracts unless mentioned otherwise. Protein concentration was quantified by Bradford assay. Equal amounts of proteins were separated using Nu-PAGE Bis-Tris Gels (Life Technologies, Vienna, Austria) and transferred onto a PVDF membrane for immunoblotting with antibodies for target proteins (S3 Table). The antigen-antibody complexes were visualized using Pierce ECL Western Blotting Substrate (Thermo Scientific, Vienna, Austria). than 0.05 were considered statistically significant. Densitometric quantification of protein bands was performed with ImageJ software. Morphometric quantification of M M 120-1 immunofluorescence images and the colocalization efficiency of M M 120-1 with other antibodies was carried out with ImageJ interface (NIH, USA) using the particle counter by keeping the threshold value equal for all images. Photoshop CS5 (Adobe, San Jose, USA) was used to prepare the immunofluorescence figures whereas schematics were assembled with CC Illustrator software (Adobe, San Jose, USA).

Characterization of p62 -/and p62hep-/-Mice
To study the role of p62 in MDB formation in vivo, we generated p62 total (p62 -/-) and hepatocyte-specific (p62hep -/-) knockout mice by Cre-loxP technology (S1A Fig). The genomic deletion of p62 was confirmed by genotyping PCR (S1B Fig, S1 Table) and reverse transcriptase PCR (RT-PCR) (S1C Fig). Both p62 -/and p62hep -/and their corresponding littermate controls p62 f/f and p62hep +/+ mice were born at the expected Mendelian ratio and had a normal life span. To induce MDB formation, mice were fed a diet containing 0.1% DDC starting at the age of 2 months for 8 weeks. Among DDC-exposed animals, p62 -/mice displayed a significantly higher body weight than p62 f/f mice whereas no difference was seen between p62hep +/+ and p62hep -/mice (S2A Fig). As reported previously, DDC treatment induced a significant increase in liver-related serum enzymes and liver/body weight ratios [18]. However, these parameters did not differ significantly among genotypes (S2C and S2D Fig).

p62 Deficiency Prevents the Maturation of MDBs
Histological analysis of livers from p62 f/f , p62 -/-, p62hep +/+ and p62hep -/mice fed DDC for 8 weeks (Fig 1A, panel 2), in contrast to animals fed a normal diet ( Fig 1A, panel 1), revealed the occurrence of characteristic features of DDC-induced liver injury, such as porphyria (brown pigment deposition), bridging fibrosis, inflammation and ballooning of hepatocytes (compare panels 1 and 2). No differences in these morphological features were found between the different genotypes. Interestingly, the typical dense and distinct eosinophilic "classical" MDBs regarded as "mature" MDBs were only detected in H&E-stained liver sections of DDC-intoxicated p62 f/f and p62hep +/+ mice ( Fig 1A, panel 2) whereas in p62-deficient livers, MDBs were smaller, granular and less distinct ( Fig 1A, panel 2). Immunohistochemical staining for K8/K18 confirmed the presence of enlarged "empty" hepatocytic cells lacking keratin cytoskeleton in all genotypes of DDC-fed mice ( Fig 1A, panel 4). Keratin and p62 immunohistochemistry and immunoblotting confirmed the genotypes of the animals as well as the formation of MDBs (Fig 1A, Fig 1B).

panels 3-4 and
To further monitor the formation of MDBs, we relied on the detection of M M 120-1 antigen that constitutes a constant component of MDBs [10,18]. Double immunofluorescence staining with M M 120-1 antibody and antibodies to K8/K18 revealed predominantly smaller, granular and less distinct aggregates in DDC-intoxicated p62-deficient livers whereas both small granular as well as large MDBs were seen in p62 f/f and p62hep +/+ livers (Fig 2, panels 2-4). The morphometrical analysis confirmed smaller but more numerous aggregates in DDC-intoxicated p62-deficient livers than in corresponding p62 f/f and p62hep +/+ livers ( Fig 3A). No alterations of the keratin cytoskeleton were observed in livers under a normal diet (Fig 2, panel 1). The absence of p62-positive aggregates in DDC-intoxicated p62-deficient animals was confirmed by performing various combinations of double immunofluorescence stainings with antibodies against MDB-specific components (Fig 2, panels 5-7). Moreover, we observed a rather weak ubiquitin staining of p62-deficient MDBs when compared to the pronounced colocalization of ubiquitin in MDBs of p62 f/f and p62hep +/+ livers (Fig 2, panel 7). Immunoelectron microscopy with M M 120-1 antibody revealed that p62-deficient livers contained smaller aggregates when compared to larger MDBs in control livers, which however resembled the classical filamentous type II MDBs, occasionally, with electron-dense amorphous material as described for type III MDBs [5] (Fig 3B). Therefore, on the ultrastructural level, no differences existed between the filamentous material of MDBs in p62-deficient and corresponding control animals.

Loss of p62 Does Not Affect Keratin Modifications but Impairs the Binding of NBR1 and Other Proteins to MDBs
Keratins are the primary components of MDBs. To analyze molecular structural features of MDBs, we used the luminescent conjugated oligothiophene h-HTAA, which labels cross βsheet structure of proteins [16]. Double staining with h-HTAA and the M M 120-1 antibody showed no positivity for cross β-sheet structures in hepatocytes of p62-deficient and corresponding p62 f/f and p62help +/+ control mice fed a normal diet (Fig 4A, panel 1). On the other hand, in DDC-fed mice both p62-deficient and "mature" MDBs were positive for cross β-sheet conformation indicating that DDC-induced structural changes of K8 to cross β-sheet conformation in MDBs were independent of p62 ( Fig 4A, panel 2-4). The formation of high molecular weight cross-linked K8 species, increased amount of keratin expression (Fig 4B), and altered K8>K18 mRNA ratio (S3A Fig) observed after DDC exposure also did not differ among the genotypes. However, the protein levels of transglutaminase 2 (Tg2), which was reported to form glutamine-lysine crosslinks of keratins in MDBs [13], were decreased in DDC-intoxicated p62 -/livers in comparison to p62 f/f livers (S3B Fig).
Since NBR1 can also be detected in large "classical" MDBs and shares both structural and functional similarities with p62 ( Fig 5A) [34], we investigated whether NBR1 is able to compensate for the loss of p62 in MDB formation. To this aim, we compared the mRNA and protein levels of p62 and NBR1 in p62 f/f and p62 -/livers. Mice fed a normal diet showed similar mRNA levels of NBR1and p62 (Fig 5B). Of note, an increased expression of NBR1 protein was observed in in western blots of total liver extracts of p62-deficient mouse livers when compared to p62 f/f livers (Fig 5C, left panel). The minimal levels of p62 seen in the western blot are due to very low concentrations of p62 in mice fed a normal diet (non stressed situation).
DDC-intoxication led to a dramatic increase in both mRNA and protein levels of p62 ( Fig  5B). In contrast to p62, there was only a slight increase of NBR1 in DDC-treated p62 f/f livers (Fig 5B and 5C, right panel). Surprisingly, NBR1 mRNA and protein expression was markedly reduced in DDC-intoxicated p62-deficient livers as compared to p62-deficient mice fed a normal diet (Fig 5B and 5C). Moreover, double immunofluorescence staining for NBR1 and M M 120-1 antigen demonstrated an almost perfect overlap of both antigens in DDC-treated p62 f/f and p62hep +/+ livers whereas a much weaker colocalization was seen in p62-deficient livers ( Fig 5D, Table 1). Triple staining for M M 120-1 antigen, NBR1 and ubiquitin revealed that only MDBs containing ubiquitin were also positive for NBR1 (S4 Fig). This demonstrates that p62 enhances the binding of NBR1 to MDBs, but the latter does not compensate for p62's role in the maturation of MDBs. morphological signs of liver injury and no MDBs were found in livers under normal diet (panel 1). Large and distinct MDBs (arrows) were only visible in DDC-fed p62 f/f and p62hep +/+ mice in H&E stained sections (panel 2). These MDBs were also decorated by p62 (green; panel 3) and K8/K18 antibodies (red; panel 4). In p62-deficient mice, MDBs are small, less distinct and only decorated by K8/K18 antibodies (arrows; panel 4). MDB-containing hepatocytes display diminished or lacking keratin cytoskeleton (empty cells). Brown pigment in the livers indicates protoporphyrin accumulation. (Scale bar = 50 μm). (B) Immunoblotting revealed abundant p62 expression in livers of DDC-fed p62 f/f and p62hep +/+ mice upon DDC-intoxication while no p62 signal was detected in p62 -/liver. The minimal p62 signal detected in p62hep -/livers reflects preserved p62 expression in non-hepatocytic cells. β-actin was used as a loading control.  In the context of characterizing possible impacts of the absence of p62 on the composition of MDBs we have also investigated the presence of the phosphoepitope SMI 31 in MDBs since it has been described as common feature of MDBs and neurofibrillary tangles (NFTs) [11]. We observed that the SMI 31 epitope was not present in p62-deficient MDBs which demonstrates that the SMI 31 epitope in MDBs is located on p62 and not on tau protein as originally reported for NFTs (Table 1,

p62 Provides Stability to MDBs
To determine whether p62 alters MDB dynamics, we analyzed the amount of inclusions in mice recovered from DDC intoxication on a standard diet for four weeks. As reported previously [7,18], the recovery from DDC intoxication resulted in an almost complete disappearance of large MDBs (Fig 6A) while small aggregates occasionally persisted at the cell periphery in proximity to desmosomes (S5 Fig). The numerical reduction of MDBs was much more pronounced in p62-deficient than in control livers as confirmed by morphometry (Fig 6B and 6C). In contrast, the reduction in size of MDBs was more prominent in controls (Fig 6C). In addition, insoluble ubiquitinated proteins and the high molecular weight insoluble K8 species were found at similar levels in DDC-fed p62 f/f and p62 -/mice, but disappeared more quickly in p62deficient animals upon recovery (Fig 6D, see stacking gels). Moreover, no changes in the overexpression of K8 (52 kD and 97 kD) was observed in DDC-treated livers irrespective of the genotype (Fig 6D, see resolving gel). Ubiquitin levels were slightly decreased in DDC-treated livers of p62-deficient mice (Fig 6D, resolving gel).
Further, we measured the levels of p62's binding partner microtubule-associated light chain 3-II to the light chain 3-I (LC3-II/LC3-I) ratio in the untreated control, DDC-treated and recovered mouse livers (S6 Fig). In chronic DDC intoxication, we observed an increased LC3-II to LC3-I ratio in all genotypes when compared to untreated controls indicating autophagic activity. LC3-II to LC3-I ratios decreased in the recovery period to levels comparable to the untreated controls. There was no significant difference in LC3-II to LC3-I ratios between the genotypes (S6 Fig).

p62 and the Cellular Defense Mechanisms
Previous studies with the DDC model suggested that the increased accumulation of p62 and MDB formation is, at least in part, a response of the liver to oxidative stress [35]. Nrf2 (Nuclear Whole liver tissue extracts of p62 f/f and p62 -/mice fed a normal diet or 8weeks DDC were analyzed by western blotting with a K8-specific antibody and densitometric quantification was performed (n = 3). DDC-treatment resulted in an increased amount of K8 (52 kD; bands with lower molecular mass reflect degradation products) and formation of crosslinked K8 dimers (ca 97 kD). However, the extent of dimer formation did not differ significantly between p62 -/and p62 f/f mice. β-actin was used as a loading control. *p<0.05, **p<0.01. factor (erythroid-derived 2)-like 2)-antioxidant pathway is a key signalling pathway determining the cellular defense against oxidative stress [36]. It regulates the expression of cytoprotective genes involved in the detoxication and elimination of reactive oxidants by enhancing cellular antioxidant capacity. Upon DDC intoxication, Nrf2 protein was increased in liver nuclear extracts of p62 f/f and p62 -/- (Fig 7A) mice suggesting dissociation of Nrf2 from Keap1 complex and nuclear translocation of Nrf2, followed by an increase in transcript levels of Nrf2 target genes, such as Nqo1 and Gst ( Fig 7C). Interestingly, in addition to the translocation of Nrf2 into the nucleus there was also a marked overexpression of the Nrf2 mRNA in DDCintoxicated livers (Fig 7A) suggesting additional transcriptional mechanisms regulating Nrf2 expression level, which to some extent were influenced by p62 since Nrf2 mRNA expression and the nuclear translocation were slightly reduced in p62-deficient mice when compared to the p62 f/f livers. This reduced Nrf2 expression was accompanied by a reduction in Nqo1 and Gst mRNA levels in p62-deficient livers when compared to p62 f/f livers (Fig 7C). No significant difference in the transcript and protein levels of Nrf2-interacting partners Keap1 (Kelch-like ECH-associated protein 1) and Nrf2 target gene Hmox-1 (hemeoxygenase 1) was observed between DDC-intoxicated p62 f/f and p62 -/livers (Fig 7B). Taken together, these data indicate an impaired cellular defense mechanism in DDC-intoxicated p62-deficient liver leading to more pronounced injury. Although, this was not readily apparent at light microscopy (Fig 1A), the significant increase in the number of Ki-67-positive mitotic cells and the slightly enhanced ductular reaction in p62-deficient livers support this assumption (Fig 7D).

Discussion
MDBs developed in DDC-intoxicated p62-deficient mice were smaller, granular and less distinct in their outlines when compared to that of the control mice (Fig 2). Thus, MDBs in p62-NBR1 and p62 interact via their PB1 domains [34]. (B) qPCR analyses of NBR1 and p62 mRNA expression in livers of p62 f/f and p62 -/mice under normal diet or 8 weeks DDC. 18S rRNA was used as a normalization control (n = 5). (C) Whole liver tissue extracts of p62 f/f and p62 -/mice fed a normal diet or 8 weeks DDC were immunoblotted for NBR1 (99 kD) and p62 (~55 kD). A sample from DDC-treated p62 f/f liver was run on the same gel with liver samples from mice fed a normal diet (left panel) as common reference (right panel). Densitometric analyses of p62 and NBR1 are also shown (n = 3). *p<0.05, **p<0.01, ***p<0.001(D) Double immunofluorescence staining with M M 120-1 (green) and NBR1 antibodies (red) visualized the distribution of both proteins in livers of mice treated with DDC for 8 weeks. An extensive co-localization between NBR1 and M M 120-1 (C. E = colocalization efficiency) was seen in animals with intact p62 production (p62 f/f C.E = 97%, and p62hep +/+ C.E = 99%) but not in the total (p62 -/-C.E = 21%) and liver-specific (p62hep -/-C.E = 16%) p62-knockout animals (scale bar = 20μm). NBR1-positive and -negative MDBs are highlighted by arrows and circles, respectively.  deficient livers resembled early stages of MDB development suggesting that the coalescence to larger aggregates is impaired. However, these "early" p62-deficient MDBs already displayed the principal features of classical MDBs, i.e., cross β-sheet conformation, type II filamentous ultrastructure and M M 120-1 positivity (Figs 2-4) [6,10,16]. Considering p62's ubiquitin binding nature, the fact that ubiquitinated-keratin-rich "early" MDBs were formed in p62-deficient animals was surprising and suggestive of the presence of other ubiquitin adaptors such as NBR1. However, there were only a few NBR1-positive MDBs in p62-deficient livers, the majority of which lacked NBR1 (Fig 4B, Table 1), suggesting that MDB formation is initiated by keratin aggregation independent of p62 or NBR1. This is also supported by previous immunofluorescence microscopic observations [18] and recent proximity ligation studies demonstrating the occurrence of MDBs devoid of p62 [37]. We noticed a (compensatory) increase in the protein and mRNA levels of NBR1 in livers of p62-deficient mice fed a normal diet. However, this increase was reversed upon DDC-intoxication, with significant decrease of NBR1 in the absence of p62 (Fig 5). This suggests a transcriptional and post-transcriptional regulation of NBR1 by p62 during chronic stress. This role of p62 can be extended to other protein aggregation disease models such as α-synucleinopathy, where interdependency of p62 and NBR1 expressions in aggregate formation process was reported [38]. During chronic stress, small heat shock proteins, such as Hsp25 in mouse or Hsp27 in humans act as molecular chaperones by keeping the newly formed misfolded proteins in a folding-competent state until the physiological situation is improved allowing ATP-dependent chaperones like Hsp70 to refold misfolded proteins [32,39]. The presence of Hsps in protein aggregates implies a futile attempt to prevent protein misfolding and aggregation [39]. Ubiquitinated misfolded aggregates with attached chaperones are then destined to bind cargo adaptor proteins, such as p62 or NBR1 via UBA domain for removal by autophagy [40]. Molecular chaperones such as Hsp25 and Hsp70 were previously described as component of MDBs [41]. In a DDC-primed mouse model, heat shock led to the expression of both Hsp25 and Hsp70 before MDB formation suggesting a functional role of Hsps in MDB pathogenesis [42]. In our study, MDBs in p62-deficient livers were negative for Hsp25 (Table 1, S3 Fig) suggesting a novel cooperative mechanism between p62 and Hsp25 in response to aggregates of misfolded proteins, which needs to be further characterized.
In contrast to the lack of involvement of p62 in the initiation of MDB formation, p62 stabilizes MDBs against degradation as indicated by the pronounced reduction of MDBs and the rapid disappearance of high molecular mass K8 and ubiquitinated proteins in p62 -/livers in the recovery phase after DDC intoxication (Fig 6). This marked decrease of high molecular weight crosslinked proteins could also be attributed to the observed decrease in Tg2 levels in p62-deficient livers (S3 Fig) [13]. According to earlier reports, stimulated autophagy counteracts MDB formation [27]. However, whether p62 modulates autophagy for the clearance of MDBs after recovery from DDC intoxication is still unclear. In our study, the presence or absence of p62 had no effect on the alterations of LC3-II/ LC3-I ratio in the different treatment groups (S6 Fig). Thus, it is unlikely that p62-dependent impairment of autophagy is responsible for reduced stability of MDBs in p62-deficient mice.
MDB size was observed after recovery in the different genotypes, which confirmed the observations of immunofluorescence staining (n = 5).
(D) Immunoblotting of insoluble liver protein extracts with antibodies against keratin 8 (K8) and ubiquitin and corresponding densitometric quantification (n = 5) showed that loss of p62 did not affect the accumulation of insoluble ubiquitinated proteins or formation of cross-linked high molecular mass K8 species during DDC intoxication. However, the disappearance of ubiquitinated proteins and K8 cross-linked proteins during the recovery period was accelerated in p62-deficient mice suggesting that the inclusions were less stable. A Coomassie-stained gel for indicated genotypes and treatment shows the protein loading used for western blotting. Values are expressed as mean ± SEM. (n = 5). *p<0.05, ***p< 0.001).
doi:10.1371/journal.pone.0161083.g006 p62 and Mallory-Denk Bodies We found no major difference in liver morphology or liver-damage-related serum enzyme levels between DDC-intoxicated control and p62-deficient mice, despite differences in MDB formation (S2 Fig, Fig 1). This was surprising since p62 exerts important functions in the cell by affecting NFkB signaling, regulating apoptosis through caspase 8 aggregation and modulating oxidative stress signaling and defense mechanisms [43,44]. It was previously reported that p62 transcription is regulated by Nrf2 during chronic liver inflammation and oxidative stress [45]. This is particularly pertinent in our model since oxidative stress is considered to play an important role in steatohepatitis of alcoholic and non-alcoholic etiology [46]. In the nonstressed situation, Nrf2 is sequestered in the cytoplasm by binding to its interaction partner Keap1, followed by ubiquitination and proteasomal degradation by E3 ubiquitin ligase Cul3 [47]. Oxidative stress leads to a conformation change in keap1-Cul3 association, inhibiting the enzymatic activity of Cul3, leading to Nrf2 dissociation. Free Nrf2 translocates to the nucleus and initiates the transcription of various antioxidant genes, such as Nqo1 and Gst [48,49]. Alternately, p62 competitively inhibits the Keap1-Nrf2 interaction, leading to the stabilization of Nrf2 allowing nuclear translocation and expression of cytoprotective genes [43]. DDC intoxication led to nuclear translocation of Nrf2 that was less pronounced in p62-deficient livers than in p62 f/f controls, which, at least in part, can be attributed to the known function of p62 as competitive inhibitor of Keap1 [50]. Interestingly, DDC intoxication led to marked upregulation of Nrf2 mRNA level. This effect was weaker in p62 deficiency when compared to p62 f/f livers. The decrease in transcript and protein levels of Nrf2 in p62-deficient mice upon DDCstress can be attributed to the role of p62 as an endogenous modulator of Nrf2 expression during chronic stress situations (Fig 7). This is also mirrored by a further reduction in Nqo1 and Gst mRNA in p62-deficient livers (Fig 7). Similar observations were recently made in a mouse model of hepatocellular carcinoma where chronic stress (Diethylnitrosamine treatment) led to reduced expression of Nrf2 target genes including Gst and Nqo1 mRNA in p62-deficient livers [51]. The impaired activation of proteins involved in cellular defense mechanisms in p62 deficiency could also explain the increased cell turnover as demonstrated by increased Ki-67 positive mitotic cells and slightly enhanced ductular reaction suggesting more pronounced liver injury (Fig 7). These data further strengthen the purported role of p62 in activation of Nrf2 and promoting cell survival by creating a positive feedback loop resulting in enhanced expression of cytoprotective genes.
On the basis of morphology, ultrastructural and chemical composition, MDBs are related to, but not identical with, IHBs [22]. IHBs can be found in a few cases of HCC [22] and livers of infants with copper toxicosis [23] occasionally in coexistence with MDBs [16,22]. IHBs contain p62 and ubiquitin (although not constantly) but differ from MDBs by lacking keratins [6]. IHBs can thus be regarded as in vivo equivalents of "p62 bodies" found in stressed tissue culture cells caused by increased expression of p62 [19,24,34]. It can be speculated that p62 bodies can serve distinct cellular functions [21]. For example, p62 may reversibly recruit specific target proteins from their functionally relevant site(s) in the cell. By doing that, the proteins Absence of p62 alters cellular defense mechanisms. (A) Liver nuclear extracts and whole tissue extracts of p62 f/f and p62 -/mice fed a normal diet and of mice DDC-intoxicated for 8 weeks were analyzed by immunoblotting with antibody against Nrf2 (65 kD). Furthermore, quantitative qPCR analyses of Nrf2 mRNA (n = 5) was performed from the same mice. Corresponding densitometric quantification is shown for nuclear extracts (n = 3). Lamin B1 (62 kD) was used as nuclear loading control. (B) Liver whole tissue extract from p62 f/f and p62 -/mice fed a normal diet or 8 weeks DDC were immunoblotted for Keap1 (70 kD) and Hmox1 (32 kD) and qPCR analyses were performed for Keap1 and Hmox-1 mRNA (n = 5). Densitometric quantification for Keap1 and Hmox-1 are shown (n = 3). β-actin was used as a loading control. (C) qPCR analyses of Nqo1 and Gst mRNA levels were performed in livers of mice livers (n = 5) fed normal diet or after 8 weeks of DDC-intoxication. 18S rRNA was used as a normalization control. (D) Morphometric assessment of Ki-67-positive hepatocytes (Ki-67 score) and K19-positive bililary cells (ductular reaction score) was made by counting at least ten high power fields (400X) of five DDC-intoxicated livers of indicated genotypes stained with Ki-67 and K19 antibodies. At least five animals per genotype were used for qPCR, Ki67 and ductular reaction scoring. Values are expressed as mean ± SEM. **p< 0.01, *p<0.05.
doi:10.1371/journal.pone.0161083.g007 p62 and Mallory-Denk Bodies may be stored until the appropriate signal causes a conformational change leading to their release [20]. Our data from this study and previous reports [22] suggest that MDB formation may occur via two different pathways (Fig 8). In the first, which seems to be typical for human p62 and Mallory-Denk Bodies steatohepatitis and related animal models (e.g., DDC intoxication), small keratin-containing "early" MDB granules are generated in response to stress-induced conformational changes and ubiquitination of keratin; the subsequent incorporation of p62 and NBR1 results in the formation of large "mature" MDBs. In the second pathway, which can be proposed on the basis of previous histological and ultrastructural analyses of IHBs in HCCs and idiopathic copper toxicosis, overexpression and accumulation of p62 precede keratin aggregation [5,51]. The later incorporation of "abnormal" keratins into a matrix of aggregated p62 causes the transformation of p62-containing IHBs/aggregates into classical MDBs [5,22]. liver-to-body weight ratio, (C) serum alanine aminotransferase (ALT) (D) alkaline phosphatase (AP) levels were measured in 8 weeks DDC-fed 4 months old total (p62 -/-) and hepatocyte-specific (p62hep -/-) p62-knockout mice as well as in their littermates with unaffected hepatic p62 expression (p62 f/f and p62hep +/+ ). Among DDC-fed mice, p62 -/mice displayed significantly higher body weight than p62 f/f animals. DDC induced a substantial increase in liver size and liver enzymes. However, none of these parameters differed between p62-deficient animals and their respective controls. Values are expressed as mean ± SEM. n = 5, Ã p<0.05. (TIF) S3 Fig. Characterization of MDB components. (A) qPCR for K8 and K18 was performed on untreated and DDC-treated livers of p62 f/f and p62 -/animals. β-actin was used as reference for normalization. The transcript levels of both K8 and K18 were upregulated in DDC-intoxicated livers of p62 f/f and p62 -/mice when compared with untreated mice of both genotypes. Moreover, the increase in transcript expression ratio of K8/K18 was observed in both DDC-intoxicated p62-deficient and wildtype mice when compared with untreated controls. (B) Whole tissue extract from 8 weeks DDC-treated livers of p62 f/f and p62 -/mice were immunoblotted for MDB components Hsp70, Hsp25 and Tg2. The expression of Hsp70 did not differ between p62 f/f and p62 -/livers. However, the expression of Hsp25 and Tg2 was markedly decreased in p62 -/livers when compared to wildtypes (C) Double immunofluorescence staining with M M 120-1 (green) and SMI-31 (red) antibodies was performed on liver sections of DDC-intoxicated p62 f/f and p62 -/mice. SMI-31 colocalized with M M 120-1 in p62 f/f livers wheras no colocalization of SMI-31 was observed in p62-deficient MDBs. (TIF) S4 Fig. Loss of p62 impairs recruitment of NBR1 and ubiquitin to MDBs. Triple immunofluorescence staining with antibodies against M M 120-1 (MDB marker; blue), NBR1 (green) and ubiquitin (red) visualized the distribution of the respective antigens in p62 f/f and p62 -/livers intoxicated with DDC for 8 weeks. NBR1-positive and -negative MDBs are highlighted by arrows and dotted circles, respectively. An extensive co-localization of NBR1 and the MDB markers M M 120-1 and ubiquitin was seen in animals with intact p62 production (p62 f/f and p62hep +/+ ) but less in the complete (p62 -/-) and liver-specific (p62hep -/-) p62-knockouts. Whole tissue extracts from untreated mice, mice fed with DDC for 8 weeks (8w DDC) and from mice recovered on standard diet for four weeks (4w R) after DDC exposure were analyzed by western blotting with an antibody against LC3 as a marker of autophagy activation. Densitometric analysis depicted the intensity of both LC3-I and -II isoforms (dark/light bar) normalized to β-actin. DDC-intoxicated mice showed. higher LC3-II/I ratio but recovered mice acquired attenuation of autophagy (i.e. lower LC3-II/LC3-I ratio). The extent of autophagy did not differ between total p62 -/and p62 f/f mice of the same treatment regimen. (TIF) S1 Table. Genotyping PCR. PCR A/B was performed to distinguish between p62 non-transgenic (p62NT), p62 floxed (p62f) and p62 Δexon1-4 (p62-or p62hep-) while PCR C was performed to detect the presence of Cre-recombinase. (PDF) S2 Table. List of primers used for genotyping, qPCR and RT-PCR. (PDF) S3 Table. List of Antibodies used for immunofluorescence, immunohistochemistry and western blot. (PDF) contribution in designing the construct for p62-knockout mouse, Dr. Dagmar Kolb-Lenz, Andrea Koschell, and Dominique Pernitsch for electron microscopy, Dr. Karl Kashofer for assisting in laser confocal microscopy, Miss Gintare Siaulyte for the maintenance of mouse strains used in this study and Dr. Penelope Kungl for proof reading the manuscript. We would also like to thank Dr. Ruth Birner-Gruenberger and Ms. Laura Liesinger for their help in characterization of the MDB components.

Author Contributions
Conceived and designed the experiments: PL KZ LF PS TR HD.
Performed the experiments: PL CS IK VS TR.