Increased liver AGEs induce hepatic injury mediated through an OST48 pathway

The protein oligosaccharyltransferase-48 (OST48) is integral to protein N-glycosylation in the endoplasmic reticulum (ER) but is also postulated to act as a membrane localised clearance receptor for advanced glycation end-products (AGE). Hepatic ER stress and AGE accumulation are each implicated in liver injury. Hence the objective of this study was to increase the expression of OST48 and examine the effects on hepatic function and structure. Groups of 8 week old male mice (n = 10–12/group) over-expressing the gene for OST48, dolichyl-diphosphooligosaccharide-protein glycosyltransferase (DDOST+/−), were followed for 24 weeks, while randomised to diets either low or high in AGE content. By week 24 of the study, either increasing OST48 expression or consumption of high AGE diet impaired liver function and modestly increased hepatic fibrosis, but their combination significantly exacerbated liver injury in the absence of steatosis. DDOST+/− mice had increased both portal delivery and accumulation of hepatic AGEs leading to central adiposity, insulin secretory defects, shifted fuel usage to fatty and ketoacids, as well as hepatic glycogen accumulation causing hepatomegaly along with hepatic ER and oxidative stress. This study revealed a novel role of the OST48 and AGE axis in hepatic injury through ER stress, changes in fuel utilisation and glucose intolerance.

The liver, however, is not only a site for the clearance of AGEs, but also a target organ and expresses various AGE receptors including the receptor for advanced glycation end-products (RAGE) 9 , advanced glycation end product-receptor 1 (OST48 also known as AGE-R1) 10 and galectin-3 (AGE-R3) 11 . There is accumulating evidence which suggests that in NAFLD, AGEs are a 'second hit' that triggers progression from steatosis to NASH 12 . Galectin-3 (AGE-R3) has also been associated with inflammation and liver fibrosis 11 and a Phase 2a multi-centre trial is underway in individuals with portal hypertension and NASH cirrhosis (NCT02462967), using galectin-3 inhibitors. Although the studies to date in galectin-3 have focused on its immunomodulatory roles, it is feasible that AGE binding and signalling via galectin-3 is also prevented by galectin-3 inhibition contributing to the efficacy of these agents in preventing liver pathology.
The effects of AGEs are mediated by their receptors, which the effects can be broadly contrasting depending on the type of receptor. For example RAGE facilitates oxidative stress, cell growth and inflammation 13 . Specifically, in chronic liver injury, hepatic expression of RAGE is significantly increased 14 and several studies in acute liver injury have identified that blockade of RAGE can ameliorate toxic, ischemic and cholestatic liver damage 15 . Alternatively, another AGE receptor, OST48, is thought to be responsible for the detoxification and clearance of AGEs and negative regulation of AGE pro-inflammatory signalling 16,17 , and cellular oxidative stress 18 . A decline in the expression of OST48 associated with increases in AGEs has been demonstrated both in murine models 19 and in individuals with diabetes 17 . Although OST48 is postulated as an AGE clearance receptor, its primary role is within the endoplasmic reticulum (ER) lumen 20 where as a subunit of the multiprotein oligosaccharyltransferase complex it facilitates enzymatic N-linked glycosylation of selected asparagine residues during protein translation 21 .
The present studies have demonstrated that OST48, a protein that was previously thought to mediate its effects via N-glycosylation or detoxification and clearance of AGEs, is rather a facilitator of increased AGE deposition in the liver. Specifically, a novel OST48-AGE pathway that leads to the onset of liver injury in combination with central adiposity and glucose intolerance, despite ample physical activity. These results establish a previously unappreciated physiological trafficking role of OST48 in the gastrointestinal tract and liver, which when dysregulated result in liver injury.

Results
Generation of a ubiquitous OST48 knock-in mouse model. OST48 knock-in mice (ROSA26 tm1(DDOST)Jfo hereon termed as DDOST+/−) were heterozygous in their expression of human DDOST at the ROSA26 locus under the control of the ubiquitin promoter (Fig. S1A). There was a significant increase in hepatic OST48 gene (DDOST) expression (Fig. S1B), whilst endogenous gene expression (Ddost) was unaffected (Fig. S1C). Targeted proteomics identified that 32 week old DDOST+/− mice did not show a significant change in total OST48 protein in the gut using the major peptides detected (Fig. S1D). However, there is a substantial increase in OST48 protein in each major section of the gut than compared to an average of 5 differently tissue locations (Fig. S1D). Moreover, in fractionated liver tissue taken from DDOST+/− mice at the same time point, there was a significant increase in plasma membrane localisation of DDOST+/− fed on a high AGE diet (Fig. S1E). This increase in plasma membrane OST48 protein localisation and content in liver taken from DDOST+/− high AGE fed mice was not seen in the cytosol (data not shown). The small intestine (duodenum, jejunum and ileum) had a high relative abundance of OST48 protein when compared with other tissues (Fig. S1D).
In the absence of steatosis, hepatic injury was evident in DDOST+/− mice and this was exacerbated by a high AGE diet. Hepatomegaly was evident in DDOST+/− mice irrespective of the diet (Fig. 1A). H&E staining indicated that wild-type mice on a high AGE content diet had some hepatocellular ballooning, with modest rarefication of the cytoplasm (Fig. 1B). All DDOST+/− mice exhibited hepatocellular enlargement and ballooning, clusters of lobular plasma cells, increased inflammatory infiltration and an abundance of rarefied cytoplasm in hepatocytes (Fig. 1B), as well as increases in hepatic fibrosis exhibited by increased Sirius Red staining and positive α-SMA staining, respectively, ( Fig. 1C and D). Although there was no substantial differences in Sirius Red staining between high AGE fed WT mice and DDOST+/− mice (Fig. 1C). We could however observe a gradual exacerbation of α-SMA staining in high AGE fed DDOST+/− mice (Fig. 1D). Plasma concentrations of the liver enzymes ALT and AST were also increased in DDOST+/− mice (Fig. 1E), and plasma ALP concentrations were the highest in DDOST+/− mice fed a high AGE diet, suggesting the presence of hepatocellular damage.
Hepatocellular ballooning is often indicative of lipid droplet accumulation in the liver. Despite the presence of vascular fibrosis, there were no differences among groups in hepatic Oil Red O staining for lipid droplets ( Fig. 1F; right). Further, there was a significant reduction in both hepatic DAGs ( Fig. 1F; top; P = 0.0101) and ceramides ( Fig. 1F; bottom; P = 0.0246) associated with high AGE dietary intake. There was also a modest decrease (−30.9% low AGE diet and −15.9% high AGE diet, P = 0.0516, 2-way ANOVA) in hepatic ceramide content when comparing the DDOST+/− genotype with the WT mice however, this did not reach statistical significance.

DDOST+/− mice exhibited increased absorption of AGEs in the liver.
The two-hit model of a high AGE diet and DDOST+/− mice exhibited increased deposition of both CML and OST48 in hepatic tissue sections ( Fig. 2A). Oral administration of near-infrared labelled AGEs using IVIS/MR imaging, suggested prolonged exposure of AGEs in the liver of DDOST+/− mice (Fig. 2B). We further identified that hepatic AGE deposition was greater in mice fed a high AGE diet and in DDOST+/− mice irrespective of dietary alterations (Fig. 2C). Although WT mice fed a high AGE diet had similar deposition of AGEs in the liver, it was identified that circulating AGE concentrations were reduced in all DDOST+/− mice as compared with high AGE fed littermate WT mice (Fig. 2D) indicating an increase in trafficking and exposure of AGEs from the circulation in the liver. Despite the absence of steatosis DDOST+/− mice had increased adiposity in addition to increased physical activity. There were no differences in body weight or lean body mass at either the beginning ( Fig. S2A and B) or the end of the study ( Fig. S2D and E) among mouse groups. Body fat mass was increased in DDOST+/− mice as compared to their respective WT group ( Fig. 3A; left), and this was more pronounced with high AGE feeding at the study end. Indeed, adiposity, measured as the intra-abdominal fat pads (mesenteric and omental fat pads), was increased in DDOST+/− mice as compared to WT littermates ( Fig. 3A; right). At the beginning of the study, there were no significant differences in adiposity between WT and DDOST+/− mice (Fig. S2C). Increased adiposity in DDOST+/− mice was not a consequence of reduced physical activity ( Fig. 3B; left), nor of increased food intake (Fig. S2F). In fact, DDOST+/− mice exhibited increased locomotor activity during both dark/awake and light/sleep phases ( Fig. 3C; right), irrespective of diet. Hepatic injury associates with ER stress, oxidative stress and activated inflammasome. Confocal microscopy, showed co-localisation of GRP-78 with OST48 (Fig. 4A). Unbiased proteomic profiling of hepatic tissue using SWATH-MS, identified increases in key proteins involved in ER stress, specifically GRP78 ( Fig. 4B; top) and EIF3A ( Fig. 4B; bottom) in our two-hit model of OST48 over-expression and high AGE feeding. The anti-oxidant enzymes SOD1/SODC ( Fig. 4C; top) and SOD2/SODM ( Fig. 4C; bottom) were increased in DDOST+/− mice irrespective of the diet consumed. DDOST+/− mice fed a low AGE diet also had increases in the redox responsive GPX1 ( Fig. S3A; left) and aconitase/ACON ( Fig. S3A; right) when compared to low AGE fed WT mice. Furthermore, we also observed a significant (P = 0.034) effect of a high AGE diet on increased 8-isoprostane levels in the urine (Fig. 4D). DDOST+/− mice fed a high AGE diet also had significant increases in the inflammatory proteins CD47 ( Fig. S3B; left) and HA1D ( Fig. S3B; right).

Defects in N-glycosylation do not explain the hepatic injury seen in DDOST+/− mice fed on a high AGE diet.
Prolonged ER stress culminating in liver injury is often associated with impaired ER function leading to changes in glycosylation occupancy on proteins. We therefore investigated whether DDOST+/− mice altered hepatic protein synthesis N-glycosylation. Fifty-six glycosylation sites were identified on mouse plasma proteins ( Fig. S4A) with six indicating partial modification by N-glycosylation (Fig. S4B), but there were no differences among groups in glycosylation occupancy on plasma proteins (Fig. S4A), nor in plasma proteins specifically glycosylated and secreted by the liver (Fig. S4C).
Hepatic injury was associated with changes in fatty acid metabolism. Using continuous indirect calorimetry, we determined that DDOST+/− mice had decreased respiratory exchange ratios during both the sleep/light ( Fig. 5A; left) and active/dark ( Fig. 5A; right) phases. This downward shift in RER suggested a proportional shift towards the use of free fatty acids and/or ketones for ATP generation as compared to wild-type littermates consuming either a low/high AGE diet. In support of this, an RER of 0.9, seen in DDOST+/− mice on a high AGE diet during their active cycle, demonstrates a ~65:35 of carbohydrate:fat/ketone oxidation as compared with WT mice fed a low AGE diet, (RER of 0.98; ~93:7 carbohydrate:fat/ketone oxidation) 22  Using SWATH-MS proteomics, significant increases in proteins associated with fatty acid oxidation, peroxisome proliferator-activated receptor (PPAR) protein signaling and fatty acid transport/export were demonstrated in DDOST+/− mice fed a high AGE diet when compared with WT low AGE fed mice (Fig. 5D). Specifically, the fatty acid oxidation enzymes ACOX1 and ACBP were increased, as well as the transport proteins FABP5 and FABPL, and the HDL core protein APOA1. The other proteins associated with lipid metabolism examined were not significantly altered. DDOST+/− mice exhibit impaired glucose tolerance at a young age. Young (8 week old) DDOST+/− mice had increases in fasted blood glucose concentrations (Table 1 and Fig. S6A), lower fasting insulin concentrations and decreases in first phase insulin secretion (Table 1 and Fig. S6B) during ipGTT as compared with WT littermates. There were no differences in fed blood glucose concentrations (Table 1), or in insulin tolerance tests (Fig. S6C) among young mice.
Adult DDOST+/− mice exhibited glucose intolerance which is augmented by high AGE dietary intake. By the end of the study, all DDOST+/− mice had elevated glycated haemoglobin (Fig. 6A) and lower fasted plasma insulin concentrations (Fig. 6B), without differences in fasted or fed plasma glucose concentrations as compared with WT mice (Table 1). However, high AGE fed DDOST+/− mice also had increased 2 hour plasma glucose concentrations following a 2 g/kg glucose bolus (ipGTT), as compared with other groups (Fig. S7A and B). While on a high AGE diet, WT mice were less glucose tolerant as compared with low AGE fed mice (Fig. S7C).
In response to an insulin bolus (ipITT), DDOST+/− mice fed a high AGE diet had higher plasma glucose concentrations than other groups (Fig. 6C). Fasting plasma glucagon concentrations were not increased (Table 1), but hepatic glycogen storage was increased by 30-40% in DDOST+/− mice (Fig. 6D). When compared with the WT low AGE fed group, all mice also had lower protein intensities of 1,4-alpha-glucan-branching enzyme (GLGB), a facilitator of glycogenolysis (Fig. 6E), consistent with greater levels of liver glycogen described above.
The gene expression of the gluconeogenic enzyme, pyruvate carboxylase kinase (Pck2; Fig. 6F) was increased by high AGE dietary feeding. Pathway analysis of proteins involved in GNG showed that hepatic proteins involved in oxidative phosphorylation were elevated in all mouse groups when compared with WT low AGE fed mice (Fig. 6G). DDOST+/− mice on a high AGE diet exhibit increased amino acid metabolism. Increased gluconeogenesis and hepatic glucose release to potentially sustain an associated increased physical activity can be further substantiated by the evident increased circulating concentrations of the amino acids alanine (Fig. 7A) and serine (Fig. 7B) and the loss of glucogenic glycine in DDOST+/− mice (Fig. 7A). Plasma concentrations of the essential amino acids lysine and histidine (Fig. 7C) were also increased in DDOST+/− mice. Plasma concentrations of both tyrosine and tryptophan were also increased in the high AGE fed DDOST+/− mice when compared with other mouse groups (Fig. 7D). Pathway analyses showed an increase in protein associated with amino acid metabolism in high AGE fed DDOST+/− mice (Fig. 7E). There were no other changes observed in circulating concentration of other amino acids (Fig. S8).

Discussion
Our discovery of the accumulation of AGEs in the liver through an OST48 mediated pathway of uptake and clearance is novel. Despite the unconventional pathway of liver injury (normally the presence of steatosis leading to the onset of fibrosis) in our DDOST+/− mice, this allowed us to identify other metabolic factors that may contribute an important role in the development of liver injury 23 . A recent study in children with NAFLD demonstrated that the presence of early portal inflammation on biopsy was independently associated with severity of liver fibrosis and metabolic risk factors for type 2 diabetes, in particular waist circumference 24 . Indeed in that study, central adiposity was the only non-invasive variable associated with biopsy proven portal inflammation. These findings are consistent with the present study, where central adiposity was seen in concert with early liver injury and portal fibrosis in DDOST+/− mice fed a high AGE diet. Whether DDOST+/− mice bypass the development of    steatosis and proceed directly into fibrogenic pathways would require further investigation into a high fat diet model. Despite this caveat what was particularly interesting was that central adiposity and liver injury persisted in DDOST+/− mice fed a high AGE diet, despite them partaking in greater levels of physical activity. This interesting finding reiterates the complexity of the relationship between metabolic factors and liver injury, given that exercise is a feasible intervention to improve liver fibrosis in patients with NAFLD 25 and relates to adiposity in NAFLD 26 . The liver is a node which controls systemic glucose and lipid fluxes in the body, balancing these against its own energy requirements 27 . These processes are influenced heavily by pancreatic islet function via the endocrine hormones insulin and glucagon 28 as well as other hormones. In the present study, abnormalities in insulin secretion and elevations in glycated haemoglobin were seen in the context of hepatic injury and increased fuel oxidation, particularly of fatty acids in high AGE fed DDOST+/− mice. This is in agreement with previous evidence that a shift in hepatic fuel utilisation towards fatty acids results in liver damage 29 . The increased utilisation of fatty acids for β-oxidation, and likely increased fatty acid export/transport proteins, may also explain the lack of hepatic fat accumulation seen in DDOST+/− mice fed a high AGE diet. Surprisingly, this led to an increase in central adiposity, despite increased physical activity and the absence of changes in caloric intake and lean body mass. There also appeared to be an excess supply of circulating glucose generated from hepatic gluconeogenesis which may be required by the skeletal muscle to facilitate the increases in physical activity seen in DDOST+/− mice, rather than for thermogenesis. This is also supported by the decreases in heat generation seen in AGE-R1 mice fed a high AGE diet. A futile cycle of energy generation in adipocytes has been linked to reduced body fat storage 30 , however it has not been previously linked to the development of liver fibrosis.
In summary, this group of studies revealed a previously unappreciated role of the OST48-AGE axis within the body. Indeed, the data suggest that the "trafficking role" of OST48 in normal hepatic physiology is not solely involved in protein N-glycosylation or elimination (detoxification/clearance) of AGEs but could also alter gastrointestinal uptake through the small intestine of AGEs leading to changes in glucose homeostasis, hepatic function and fuel utilisation. The physiological significance of this should form the basis of future studies. In the context of a high AGE diet, increases in OST48 expression exacerbated liver injury leading to fibrosis, likely via known pathways of injury including oxidative and ER stress. Further studies into the role of hepatic and gastrointestinal OST48 mediated AGE uptake could lead to a novel approach to minimise liver injury. Plasma AGE measurement. CML concentrations in plasma were measured by an in-house indirect ELISA as previously described 33 . Human and mouse OST48 ELISAs. Mouse and human OST48 concentrations were determined in hepatic cortical membrane and cytosolic fractions by commercial sandwich ELISAs (Cloud-Clone Corp., Houston, United States), according to the manufacturer's specifications.

Materials and Methods
Liquid Chromatography-Mass Spectrometry (LC-MS/MS). As previously described 34 , proteins were extracted from whole liver tissue samples using guanidine denaturing buffer (6 M guanidinium, 10 mM DTT and 50 mM Tris-HCl). Reduced cysteines were alkylated with acrylamide, and quenched with excess DTT. Proteins were precipitated in 4 volumes of 1:1 methanol:acetone and digested with trypsin. Peptides were desalted and analysed by Information Dependent Acquisition LC-MS/MS as described 35 using a Prominence nanoLC system (Shimadzu, NSW, Australia) and Triple TOF 5600 mass spectrometer with a Nanospray III interface (SCIEX). SWATH-MS analysis was performed 36 and analysed with MSstats as previously described. Differentially abundant proteins were analysed using DAVID 37 .

Serum and urine biochemistry and analysis of hepatic lipids.
The plasma concentrations of alanine transaminase (ALT), aspartate transaminase (AST) and alkaline phosphatase (ALP) were measured by an auto-analyzer (Beckman Instruments, USA). Hepatic diacylglycerols (DAGs) and ceramides were extracted and quantified by thin layer chromatography and a liquid scintillation analyzer (LS6500; Beckmann Coulter Inc., CA, USA) as previously described 39 . Lipid peroxidation was examined by urinary 8-isoprostane in a competitive ELISA (Oxford Biomedical Research, United States), according to the manufacturer's specifications.
Glucose and insulin tolerance tests. Intraperitoneal glucose tolerance tests (ipGTT) were performed following a 2 g/kg D-glucose bolus as previously described 40 . Insulin was determined by rat/mouse insulin ELISA (RnD systems, MN, United States). Intraperitoneal insulin tolerance test (ipITT), to determine glucose output was performed using a 1 IU/kg bolus of Humulog fast-acting insulin (Eli Lilly, IN, United States). Area under the curve was calculated using the trapezoidal rule (GraphPad Software, CA, United States).
Indirect calorimetry and assessment of body composition. Whole body composition was measured at 32-weeks of age in conscious, but physically restrained mice using an EchoMRI ™ 3-in-1 body composition analyzer (EchoMRI, TX, USA). A cohort of mice were allocated for repeated measures of energy expenditure by indirect calorimetry (VO 2 , VCO 2 ) normalised to lean body mass, locomotor/physical activity and rate of energy expenditure (heat generation) using a Comprehensive Lab Animal Monitoring System (CLAMS; Columbus Instruments, OH, United States). Data was collected over a 24-hour cycle period, post-acclimatisation to the cages for a 12-hour period.
Liver glycogen content. As previously described 41 , total liver glycogen content was determined using a glucose oxidase/peroxidase assay procedure and absorbance was analysed on a UV-1700 PharmaSpec UV-vis spectrophotometer (Shimadzu, NSW, Australia).
Quantitative real-time PCR. Messenger RNA was purified and was used (1 µg) to synthesize cDNA using SuperScript first-strand synthesis system (Invitrogen). Quantitative real-time PCR was performed using pre-designed TaqMan Gene Expression Assays ® for Acadm, Acadvl, Acox1, ADRB2, Ccl2, Col1a1, Col3a1, Ddost, DDOST, G6pc, Gcgr, Got1, Gyk, Kcnma1, Lepr, Pck1, Pck2, Ppara, Slc27a4 and Slc37a4 (Supporting Experimental Procedure 2; Life Technologies, Mulgrave, VIC, Australia) in ViiA ™ 7 real-time PCR system (Applied Biosystem, Darmstadt, Germany). Gene expression levels were calculated after normalisation to an endogenous multiplexed control (18S) using the ΔΔCT method as previously described 32 and expressed as relative fold change compared to the wild-type mice fed a high AGE content diet. The results are represented as mean ± %CV. Amino acid measurements. Approximately 40 µl of serum was mixed with 160 µl of an extraction solution (3:1 ratio of analytical grade methanol to ddH 2 O), centrifuged at 16,000 × g for 10 minutes and the supernatant removed for processing. Amino acids were measured by HPLC as described in Chacko et al. 42 .

Statistical analysis.
Results are expressed as mean ± SD (standard deviation), and analyzed by 2-way ANOVA followed by post hoc testing for multiple comparisons using the Bonferroni method unless otherwise specified. α (genotype effect) P < 0.05, β (diet effect) P < 0.05, δ (interaction effect) P < 0.05 were reported. For comparison between groups as required, a two-tailed unpaired Student's t-test was used. For SWATH-MS, MSstatsV3.5.1 was used to detect differentially abundant proteins estimating the log-fold changes between compared conditions of the chosen experimental group and with the WT mice fed a low AGE diet group. For all calculations a P < 0.05 was considered as statistically significant.