Protection of a Ceramide Synthase 2 Null Mouse from Drug-induced Liver Injury

Background: Ceramide synthase 2 null mice cannot synthesize very long acyl chain ceramides and display severe hepatopathy. Results: Ceramide synthase 2 null mice are protected from drug- and chemical-induced liver injury and display impaired gap junction function. Conclusion: Altering sphingolipid levels modulates gap junction function. Significance: Sphingolipids may play a key role in regulating drug-induced liver injury. Very long chain (C22-C24) ceramides are synthesized by ceramide synthase 2 (CerS2). A CerS2 null mouse displays hepatopathy because of depletion of C22-C24 ceramides, elevation of C16-ceramide, and/or elevation of sphinganine. Unexpectedly, CerS2 null mice were resistant to acetaminophen-induced hepatotoxicity. Although there were a number of biochemical changes in the liver, such as increased levels of glutathione and multiple drug-resistant protein 4, these effects are unlikely to account for the lack of acetaminophen toxicity. A number of other hepatotoxic agents, such as d-galactosamine, CCl4, and thioacetamide, were also ineffective in inducing liver damage. All of these drugs and chemicals require connexin (Cx) 32, a key gap junction protein, to induce hepatotoxicity. Cx32 was mislocalized to an intracellular location in hepatocytes from CerS2 null mice, which resulted in accelerated rates of its lysosomal degradation. This mislocalization resulted from the altered membrane properties of the CerS2 null mice, which was exemplified by the disruption of detergent-resistant membranes. The lack of acetaminophen toxicity and Cx32 mislocalization were reversed upon infection with recombinant adeno-associated virus expressing CerS2. We establish that Gap junction function is compromised upon altering the sphingolipid acyl chain length composition, which is of relevance for understanding the regulation of drug-induced liver injury.

ceramide). The N-acylation reaction is governed by the ceramide synthases (CerS), of which six are known in mammals. Each CerS synthesizes ceramide with a distinct acyl chain length (6,7). CerS2, which synthesizes C22-C24 ceramides (8), is the most ubiquitous CerS and is found at particularly high levels in the liver, kidney, and lung (8). Recently, we generated a CerS2 null mouse (9). This mouse lacks very long acyl chain (i.e. C22-C24) ceramides and SLs but contains elevated levels of C16-ceramide and sphinganine in the liver. CerS2 null mice display increased rates of hepatocyte death and proliferation, which results in the formation of multiple hepatic nodules and hepatocellular carcinoma (10). In addition, they have chronic oxidative stress because of disruption of the mitochondrial respiratory chain (11) and display hepatic insulin resistance because of altered detergent-resistant membranes (DRMs) (12), which appears to be related to changes in membrane biophysical properties (13,14).
Because of these various liver pathologies, we have now examined hepatotoxicity induced by a number of drugs and chemicals. Surprisingly, CerS2 null mice were largely resistant to the hepatotoxic effects of APAP, D-galactosamine, CCl 4 , and thioacetamide (TAA), mainly because of altered gap junction function as a result of mislocalization of connexin (Cx) 32, a key player in gap junctions. We suggest that the SL pathway might be a novel drug target for alleviating drug-induced liver damage and possibly other forms of acute liver failure.
Mice-CerS2 null mice were generated as described (9,10). Mice were maintained under special pathogen-free conditions and treated in accordance with the Animal Care Guidelines of the Weizmann Institute of Science. Eight-to ten-week-old male mice were injected intraperitoneally with drugs as follows: 300 mg/kg APAP, 800 mg/kg D-galactosamine, 2 ml/kg CCl 4 , and 200 mg/kg TAA. For inhibition of MRP, 25 mg/kg MK571 was injected intraperitoneally 1 h before and 1 h after APAP treatment (15). In some cases, FB1 (1.5 mg/kg/day) was injected intraperitoneally for 5 days prior to APAP treatment (16,17). For lysosomal inhibition and GSH depletion, 50 mg/kg/day chloroquine and 3 mmol/kg/ day L-buthionine sulfoximine were injected intraperitoneally for 5 days before APAP treatment (18,19).
Serum Aspartate Aminotransferase (AST) and Alanine Aminotransferase (ALT)-Serum AST and ALT were measured by Pathovet Veterinary Diagnostic Services (Kfar Bilu, Israel) using a Roche Applied Science/Hitachi 917 system.
Histology-Liver tissues were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned at 4 m, and stained with H&E.
For immunohistochemistry, paraffin sections were deparaffinized, and endogenous peroxidase activity was quenched by 0.3% (v/v) hydrogen peroxide and 0.5% (v/v) HCl in methanol. Antigen retrieval was performed using 10 mM citric acid (pH 6.0). Peroxidase activity was measured using 3,3Ј-diaminobenzidine as the chromogen. Sections were counterstained with hematoxylin. For immunofluorescence of tissue sections, deparaffinized liver sections were incubated with an anti-Cx32 antibody followed by a DyLight 488-or 549-conjugated secondary antibody.
GSH Analysis-Liver GSH levels were measured using a glutathione assay kit (Biovision, Mountain View, CA).
Real-time PCR-mRNA expression was determined by realtime PCR (9). Primers are listed in supplemental Table 1.
Tissue Scrape-and-Load Assay for Gap Junction Function-A dye solution containing 0.5% (v/v) Lucifer yellow (Invitrogen) and 0.5% (v/v) 10-kDa dextran-Texas Red (Invitrogen) was placed onto the surface of freshly prepared liver tissues, and incisions were made with a scalpel, followed by addition of an excess of dye onto the incisions (1,22). After incubation with the dye for 5 min, liver slices were washed with phosphate-buffered saline and then fixed with 4% paraformaldehyde. Samples were frozen in optimal cutting temperature compound (Sakura Fineteck), cryosectioned at 7 m, and mounted on slides.
Pulse-Chase-[ 35 S]methionine labeling and immunoprecipitation were performed as described (23,24) with some modifications. Hepatocytes were starved for 1 h at 37°C in DMEM lacking methionine and supplemented with 10% FBS and 2 mM glutamine. The medium was then replaced with fresh medium containing [ 35 S]methionine (0.1 mCi/35-mm dish). The radioactive medium was removed after 30 min, and hepatocytes were washed three times and then incubated with DMEM containing 10% FBS and 2 mM glutamine for various times (2, 4, and 8 h). At the end of the incubation period, hepatocytes were rinsed once with sterile phosphate-buffered saline at 4°C and resuspended in lysis buffer (5 mM Tris, 5 mM EDTA, 5 mM EGTA, 10 mM iodoacetamide, 2 mM phenylmethylsulfonyl fluoride (pH 8.0), 0.6% SDS, and protease inhibitors (Sigma Aldrich)). Lysates were heated at 100°C for 3 min to minimize aggregation of Cx32 and immunoprecipitated with an anti-Cx32 antibody and protein A-agarose (Sigma Aldrich) overnight at 4°C. Cx32 immunoprecipitates were separated by 10% SDS-polyacrylamide gel electrophoresis, and gels were dried. The dried gels were exposed to a BioMax Transcreen LE-intensifying screen (Kodak) at Ϫ80°C.
Mouse Liver Mitochondria and Cytosol-Fresh liver was minced on ice and homogenized using a glass-Teflon homogenizer in 250 mM sucrose, 2 mM EDTA, 10 mM Tris (pH 7.4) and 50 IU heparin (11). Homogenates were centrifuged at 1000 ϫ g for 10 min at 4°C. The supernatant was transferred to clean tubes and centrifuged again at 14,000 ϫ g for 15 min at 4°C. The supernatant was used as the cytosolic fraction to detect cytochrome c release.

Gap Junction Dysfunction in the Liver of a CerS2 Null Mouse
Statistics-Values are given as means Ϯ S.E. Unless stated otherwise, statistical significance was calculated using Student's t test, and p Ͻ0.05 was considered statistically significant.

Lack of APAP Hepatotoxicity in CerS2
Null Mice-CerS2 null mice were completely protected from the hepatotoxic effects of APAP (Fig. 1). Although WT mice displayed huge elevations in serum AST (Fig. 1A) and ALT (B) levels 6 h after APAP injection, no changes were detected in CerS2 null mice, indicating a lack of APAP-induced liver damage, which was confirmed by morphological analyses (H&E staining, C) and analysis of formation of nitrotyrosine protein adducts (D), which demonstrated significant damage in WT mouse liver but not in CerS2 null mouse liver. We also examined a number of signaling pathways involved in APAP-induced liver injury. Consistent with lower levels of reactive oxygen species generated from mitochondria by treatment with APAP (3454 Ϯ 476 in WT versus 1453 Ϯ 243 of relative fluorescence unit in CerS2 null mice), MKK4 (31) was not activated upon APAP injection in CerS2 null mice (Fig. 1E). Likewise, JNKs were only transiently activated (Fig. 1E), and, consequently, cytochrome c was not released into the cytosol in CerS2 null mouse liver (F), indicating that mitochondria were not damaged by APAP treatment.
In addition to the reduction in very long chain (C22-C24) and elevation in C16-ceramide levels, CerS2 null mouse liver displays a significant elevation in sphinganine levels (9). To determine whether elevated sphinganine might be responsible for the lack of hepatotoxicity, WT mice were injected with the CerS inhibitor FB1 (16, 17) for 5 days prior to APAP injection. As  expected, FB1 injection reduced ceramide and elevated sphinganine levels and caused some liver damage similar to that reported previously (16,17) (Fig. 2, A and B). However, FB1 administration did not alleviate APAP-induced centrilobular liver damage or APAP-induced elevation of liver enzymes in the serum of WT mice (Fig. 2, C and D). Likewise, FB1 administration to CerS2 null mice did not affect APAP toxicity (Fig. 2, E  and F). Therefore, we conclude that elevated sphinganine and increased C16-ceramides are not likely to be directly involved in the protective mechanism of APAP-induced hepatotoxicity.
Elevated GSH May Be Partially Responsible for Lack of APAP Hepatotoxicity-We first examined APAP metabolism. APAP is metabolized to the active metabolite N-acetyl-p-benzoquinone imine by Cyp2E1. However, Cyp2E1 activity and expression were unaltered in CerS2 null mice both before (821.67 Ϯ 13 pmol/min/mg) and after (692 Ϯ 97.2 pmol/min/mg) APAP injection. We next examined levels of GSH because GSH deple-tion is one of the main mechanisms of APAP-induced liver damage (3,32). GSH levels were elevated 1.5-fold in CerS2 null liver prior to APAP injection and decreased to a similar extent as in WT mice after APAP injection (Fig. 3). However, levels of unmetabolized APAP were lower in liver (3146.7 Ϯ 461.6 in WT versus 1100.7 Ϯ 302.3 nmol/g in CerS2 null mice) and in plasma (873.1 Ϯ 112.3 versus 428.7 Ϯ 63.9 nmol/ml) with a concomitant increase in APAP-GSH in liver (1567.6 Ϯ 170.7 versus 3172.7 Ϯ 578.8 nmol/g), demonstrating that APAP detoxification by GSH is increased in CerS2 null liver. We conclude that changes in APAP metabolism may be partly responsible for the lack of APAP toxicity in CerS2 null mice, although it is unlikely to be the major mechanism because CerS2 null mice were also resistant to other hepatotoxins (see below).
Multidrug-resistant pumps are responsible for the removal of toxic drugs from the liver (33), and, specifically, MRP4 may  OCTOBER 25, 2013 • VOLUME 288 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 30909
protect against APAP-induced liver injury (34 -36). Because plasma acetaminophen-glucuronide levels (472.2 Ϯ 39.7 in WT versus 687.9 Ϯ 48.5 nmol/ml in CerS2 null mice) and acetaminophen sulfate (77.6 Ϯ 6.6 in WT versus 109.2 Ϯ 5.4 nmol/ml in CerS2 null mice) were elevated somewhat in CerS2 null mice, we examined whether multidrug-resistant pumps are altered in CerS2 null mouse liver. Levels of MRP4 mRNA and MRP4 protein were significantly elevated in CerS2 null mouse liver (Fig. 4,  A and B). However, treatment with the MRP inhibitor MK571 (15) had no effect on APAP-induced hepatotoxicity in CerS2 null mice, whereas it induced ALT elevation in WT mice (Fig.  4C), suggesting that elevated levels of MRP4 are not responsible for the lack of APAP-mediated hepatotoxicity in CerS2 null mice. In summary, although the above mechanisms might contribute, at least partially, to the lack of APAP hepatotoxicity in CerS2 null mice, it seems likely that additional mechanisms are involved.
Impaired Gap Junction Function in CerS2 Null Mouse Liver-We next treated CerS2 null mice with a variety of other hepatotoxic agents that each cause liver damage by a different mechanism (37,38). Remarkably, three other chemicals, namely CCl 4, D-galactosamine, and TAA were also ineffective in inducing hepatotoxicity, although the extent of protection differed for each (Fig. 5). Image magnification is ϫ100. D, tissue scrape-and-load assay for gap junction function using Lucifer yellow (which is gap junction-permeable) and 10 kDa dextran-Texas Red (which is gap junction-impermeable). Image magnification is ϫ10. E, quantification of the distance of dye spread of Lucifer yellow and 10 kDa dextran. Data are mean Ϯ S.E. n ϭ 4. *, p Ͻ 0.05. F, Western blot analysis of Cx43 levels at various times after APAP treatment (300 mg/kg, intraperitoneally). This experiment was repeated three times with similar results.
Gap junctions have recently been implicated in the hepatotoxicity of all four of the drugs and chemicals used in this study by the demonstration that ablation of Cx32 (a key protein in hepatic gap junctions) completely protects against DILI (1,39,40). Thus, we examined Cx levels and gap junction function in CerS2 null liver. mRNA levels of Cx26, 32, and 43 were not altered (Fig. 6A), but levels of the Cx32 protein were decreased significantly in CerS2 null mouse liver as detected by Western blotting (B) and immunofluorescence (C). Gap junction function, assayed using a tissue scrape-load assay, was also compromised severely (Fig. 6, D and E). Similar to results obtained in a Cx32-deficient rat (41), Cx43 expression did not increase in CerS2 null mouse liver after APAP injection, in contrast to a significant elevation in WT mice (which indicates hepatocyte damage (3, 41)) (Fig. 6F).
Because Cx32 mRNA levels were unaltered in CerS2 null mouse liver (Fig. 6A), we analyzed the rate of degradation of Cx32 by pulse-chase using [ 35 S]methionine. Cx32 degradation was significantly faster in hepatocytes obtained from CerS2 null mice (half-life, 4.4 Ϯ 0.2 h in WT versus 2.4 Ϯ 0.4 h in CerS2 null mice) (Fig. 7A). Interestingly, Cx32 could be restored to levels similar to those found in the WT upon incubation with the lysosomal protease inhibitors leupeptin or chloroquine (19) but not upon incubation with the proteasomal inhibitor MG132 (Fig. 7B). Lysosomal inhibition in CerS2 null hepatocytes also slowed down the rate of Cx32 degradation so that it was similar to the rate of degradation in the WT (half-life of 4.6 Ϯ 0.5 h in WT versus 4.2 Ϯ 0.2 h in CerS2 null mice) (Fig. 7C).  Cx32 localizes to DRMs (42). We demonstrated recently that the density and composition of DRMs differ in CerS2 null mice (12), along with changes in the biophysical properties of membranes (13,14). Although Cx32 localized to DRMs isolated from the liver of both WT (Fig. 8A) and CerS2 null mice (B), the density of the DRMs differed between the two, and the extent of co-localization was significantly lower in CerS2 null mice (B). Disruption of DRMs isolated from hepatocytes using methyl-␤-cyclodextrin decreased Cx32 levels in a dose- (Fig. 8C) and time-dependent (D) manner, consistent with the idea that Cx32 degradation is accelerated upon disruption of DRMs, as occurs upon changing the SL acyl chain composition in the CerS2 null mouse.
To determine the role of enhanced Cx32 degradation in the protection against APAP-induced liver injury, mice were injected with chloroquine together with BSO (an inhibitor of GSH synthesis) for 5 days prior to APAP injection. As expected, CX32 levels increased in the CerS2 null mouse to a similar level to that of the WT (Fig. 9A), and there was a significant decrease in GSH levels (B). However, CerS2 null mice were still protected from APAP-induced liver injury upon pretreatment with chloroquine and BSO, as demonstrated by the lack of change in liver function (AST and ALT levels, Fig. 9C) and gap junction function (tissue scrape-load assay, D and E). Rather, the intracellular transport of Cx32 is defective in CerS2 null mouse hepatocytes irrespective of its intracellular levels. Thus, although Cx32 is localized to the plasma membrane in hepatocytes from WT mice, it is localized intracellularly in hepatocytes from CerS2 null mice irrespective of whether its levels are increased by inhi- bition of lysosomal degradation (Fig. 10). The inability of Cx32 to be transported to the plasma membrane provides a mechanistic explanation of the protection of CerS2 null mice against DILI.
To attempt to reverse the lack of APAP hepatotoxicity in CerS2 null mice, we generated rAAV expressing CerS2. CerS2 null mice were infected via tail vein injection of 2 ϫ 10 12 viral particles. After 5 weeks, CerS2 protein levels were increased significantly with a concomitant elevation of Cx32 levels (Fig.  11A). When 300 mg/kg APAP was injected into CerS2 null mice that had been infected with rAAV-CerS2, serum AST and ALT levels were elevated, demonstrating significant hepatotoxicity (Fig. 11B), which was confirmed by H&E staining (C). Finally, although Cx32 levels were localized intracellularly in CerS2 null mice, rAAV-CerS2 infection resulted in its retargeting to the plasma membrane (Fig.  11D). Together, the rAAV-CerS2 infection experiments demonstrate a direct mechanistic link between levels of CerS2 and APAP-induced hepatotoxicity.

DISCUSSION
In this study we demonstrate that, despite the severe hepatopathy observed in CerS2 null mice, these mice are resistant to drug-and chemical-induced liver injury. The molecular mechanisms underlying this resistance may involve a combination of factors, but the most likely mechanism is the mislocalization and, hence, an enhanced rate of degradation, of a major player in gap junction function, namely Cx32. This is strongly supported by the ability to reverse the mislocalization of Cx32 using rAAV-CerS2.
Gap junctions have recently emerged as important players in regulating DILI (1,39). Gap junction channels consist of Cx proteins and are involved in intercellular communication and in the amplification of liver inflammation and injury (1,39,41,43). Ablation of Cx32, a major Cx in liver, abrogates liver injury induced by the same four hepatotoxic agents used in this study (1, 39 -41), and APAP-induced-Cx43 expression was inhibited in Cx32 dominant-negative transgenic rats (41) similar to that observed in the CerS2 null mouse.
Our data are consistent with the following model. Changes in the acyl chain length of sphingolipids leads to altered biophysical properties of membranes (14) that are exemplified by changes in the properties of DRMs. Although the use of DRMs as a measure of membrane properties is clouded in controversy (44), they do nevertheless provide a quick and simple way to measure some changes in membrane characteristics. We previously demonstrated altered DRMs in the CerS2 null mouse, which results in defective insulin receptor signaling (12) and in the internalization of the TNF␣ receptor. 6 We now suggest that the altered biophysical properties of hepatocyte membranes results in the mislocalization of Cx32. Whether the mislocalization is due to defective assembly, transport along the secretory pathway, or defective internalization remains to be established. In CerS2 null mice, the rate of Cx32 degradation is increased, perhaps because of its instability in the altered membrane environment of CerS2 null mouse hepatocytes. The increased rate of Cx32 degradation is directly related to lysosomal proteolytic degradation, but even inhibition of lysosomal degradation does not result in the correct localization of Cx32 to the plasma membrane, even though it does restore its intracellular levels. Finally, reintroduction of CerS2 into CerS2 null liver using rAAV-CerS2 corrected the Cx32 mislocalization, supporting the idea that altered membrane properties caused by lack of CerS2 are a major reason for Cx32 mislocalization. Thus, we suggest that the increased rate of lysosomal degradation is the result of the instability of Cx32 rather than the cause of reduced Cx32 levels. Previous studies have shown that misfolded connexins are more vulnerable to proteasomal degradation (23), and we now suggest that Cx32 is also more vulnerable to lysosomal degradation.
A number of other pathways are altered in the CerS2 null mouse. CerS2 null mice display chronic oxidative stress caused by high sphinganine and C16-ceramide (11), which may also contribute to the augmented Cx32 degradation (45). In addi-tion, altered drug metabolism, higher GSH levels, and other downstream signaling pathways may all play a part in the lack of APAP-induced liver damage, but we surmise that these pathways are likely to play minor roles because of the lack of commonality of these pathways in the effects of the four hepatotoxic agents used in this study. For example, CYP2E1 is required to transform many drugs but is not involved in D-galactosamine metabolism. Apoptosis signal-regulating kinase 1 knockout mice are protected from APAP-and TNF␣-induced hepatotoxicity but not from CCl 4 -or Fas-induced liver injury (38), and treatment with a JNK inhibitor alleviates APAP-and TNF␣induced liver injury but does not affect CCl 4 -or Fas-induced liver injury (46). In contrast, oxidative stress, mitochondrial dysfunction, and GSH depletion are all commonly involved in DILI (3) and are all altered in the CerS2 null mouse. However, as discussed above, these factors are likely to play a minor role in the protection against DILI in the CerS2 null mouse.
Although the CerS2 null mouse was protected from chemically induced liver injury, the extent of protection differed depending on the drug. For example, the CerS2 null mouse was only partially protected from CCl 4 -induced hepatotoxicity in contrast to complete protection from APAP-induced liver injury. Along with impaired gap junction function, mRNA levels of dual specificity phosphatases (including dual specificity phosphatases 8 and 10) were elevated (27.1 Ϯ 5.4-and 7.5 Ϯ 1.1-fold, respectively). Because a JNK inhibitor alleviates APAP-induced hepatotoxicity but not CCl 4 -induced hepatotoxicity (46), the different extent of protection against DILI in the CerS2 null mouse could be explained by an elevation of dual specificity phosphatases because of their role as JNK phosphatases (47).
SLs are now known to be important mediators in many cellular processes. This study suggests that SLs are also involved in regulating the function of proteins and pathways involved in DILI, perhaps because these proteins need a specific membrane environment to function properly. This is exemplified by the altered localization of Cx32 in CerS2 null mice. It is to be expected that other membrane proteins, involved in a number of other physiological and pathophysiological pathways, will be modulated upon alteration of the acyl chain composition of membrane SLs.