Diabetes Care Volume 41, June 2018 1235

fi Maria Apostolopoulou,1,2,3 Ruth Gordillo,4 Speci c Hepatic Chrysi Koliaki,2,3 Sofia Gancheva,1,2,3 Tomas Jelenik,2,3 Elisabetta De Filippo,2,3 Relate to Insulin Resistance, Christian Herder,2,3 Daniel Markgraf,2,3 Frank Jankowiak,5 Irene Esposito,5 Oxidative Stress, and Matthias Schlensak,6 Philipp E. Scherer,4 Inflammation in Nonalcoholic and Michael Roden1,2,3 Steatohepatitis Diabetes Care 2018;41:1235–1243 | https://doi.org/10.2337/dc17-1318

OBJECTIVE Insulin resistance and nonalcoholic fatty liver disease have been linked to several metabolites inanimals, but theirrole inhumans remains unclear. Thisstudy examined the relationship of sphingolipids with hepaticand peripheral metabolism in21 insulin- resistant obese patients without (NAFL2) or with (NAFL+) nonalcoholic fatty liver and nonalcoholic steatohepatitis (NASH) and 7 healthy lean individuals undergoing tissue biopsies during bariatric or elective abdominal surgery.

RESEARCH DESIGN AND METHODS PATHOPHYSIOLOGY/COMPLICATIONS 2 Hyperinsulinemic-euglycemic clamps with D-[6,6- H2]glucose were performed to 1 fi Division of Endocrinology and Diabetology, quantify tissue-speci c insulin sensitivity. Hepatic oxidative capacity, lipid perox- Medical Faculty, Heinrich-Heine University, idation, and the phosphorylated-to-total c-Jun N-terminal kinase (pJNK-to-tJNK) ratio Dusseldorf,¨ Germany were measured to assess mitochondrial function, oxidative stress, and inflammatory 2Institute for Clinical Diabetology, German Di- activity. abetes Center, Leibniz Center for Diabetes Re- search at Heinrich-Heine University, Dusseldorf,¨ RESULTS Germany 3German Center for Diabetes Research, Munchen-¨ Hepatic total were higher by 50% and 33% in NASH compared with Neuherberg, Germany NAFL+ and NAFL2, respectively. Only in NASH were hepatic dihydroceramides 4Touchstone Diabetes Center, Department of (16:0, 22:0, and 24:1) and lactosylceramides increased. Serum total ceramides and Internal Medicine, The University of Texas Southwestern Medical Center, Dallas, TX dihydroceramides (hepatic dihydroceramides 22:0 and 24:1) correlated negatively 5Institute of Pathology, Medical Faculty, Heinrich- with whole-body but not with hepatic insulin sensitivity. Hepatic maximal respiration Heine University, Dusseldorf,¨ Germany related positively to serum lactosylceramide subspecies, hepatic sphinganine, and 6General Surgery Department, Schon¨ Clinic, lactosylceramide 14:0. Liver lipid peroxides (total ceramides, 22:0) and Dusseldorf,¨ Germany the pJNK-to-tJNK ratio ( 24:0; hexosylceramides 22:0, 24:0, and 24:1) all Corresponding author: Michael Roden, michael. [email protected]. positively correlated with the respective hepatic sphingolipids. Received 1 July 2017 and accepted 6 March 2018. CONCLUSIONS Clinical trial reg. no. NCT01477957, clinicaltrials Sphingolipidspeciesare not only increased ininsulin-resistant humans with NASH but .gov. also correlate with hepatic oxidative stress and inflammation, suggesting that these This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/suppl/ may play a role during progression of simple steatosis to NASH in humans. doi:10.2337/dc17-1318/-/DC1. © 2018 by the American Diabetes Association. Obesity, insulin resistance, and type 2 diabetes (T2D) are frequently associated with Readers may use this article as long as the work is properly cited, the use is educational and not nonalcoholic fatty liver disease (NAFLD), which in turn worsens the prognosis of T2D (1). for profit, and the work is not altered. More infor- Accumulation not only of triglycerides but even more of biologically active lipid mation is available at http://www.diabetesjournals intermediates, such as diacylglycerols and sphingolipids, has been frequently observed .org/content/license. 1236 Sphingolipids in the Progression of NAFLD Diabetes Care Volume 41, June 2018

in animal models of T2D and/or NAFLD. In lean individuals serving as control (CON) Insulin (Actrapid; Novo Nordisk) was ad- rodent models, diacylglycerol and cer- undergoing bariatric or abdominal surgery. ministered as a primed-continuous in- amide species were related to whole- fusion (40 mU/[m2 body surface area 3 bodyand hepatic insulin resistance and to min]) from 0 to 180 min. Blood glucose ectopic lipid deposition, including hepatic RESEARCH DESIGN AND METHODS measurements were performed every 5 min, steatosis(1,2).Inhumans,theroleof Study Participants and a 20% dextrose infusion labeled with The volunteers participated in a prospec- 2 circulating and tissue-specific concentra- D-[6,6- H2]glucose (2% enriched) was ad- tions of lipid intermediates for insulin tive study on obese humans undergoing justed to maintain normoglycemia (5 mmol/L). resistance and ectopic lipid deposition bariatric surgery and on healthy humans Whole-body insulin sensitivity was measured is less clear. In skeletal muscle, specific undergoing elective abdominal surgery, from whole-body mean glucose infusion diacylglycerol, but not ceramide, species such as cholecystectomy (n =6)orher- rates (M value) with glucose space correction predict the onset of insulin resistance (3). niotomy (n = 1), for nonmalignant dis- (19). The hepatic insulin sensitivity index In the human liver, some studies reported eases (clinical trialreg. no.NCT01477957, (HIS) was calculated as the quotient of a negative relationship between hepatic clinicaltrials.gov)(10).Thecurrentanalysis 100/(fasting EGP 3 fasting insulin concen- 2 total diacylglycerols, but not ceramides, comprised 21 obese (BMI .30 kg/m ) tration), and insulin-mediated EGP sup- and insulin sensitivity (4–6) and an patients and 7 CON subjects. The classi- pressionwascalculatedas1003 clamp elevation of hepatic diacylglycerol con- fication of obese patients into NAFL2, EGP/fasting EGP (20). centrations in patients with NAFLD (7). NAFL+, and NASH was based on liver Interestingly, recent data suggest that histology using hematoxylin-eosin and Tissue Biopsies hepatic ceramide levels may dissociate special stainings (Van Gieson trichrome Samples were obtained from the lower liver steatosis from insulin resistance in stain,Perlsironstain,periodicacid–Schiff– part of the right liver lobe, rectus abdom- NAFLD of different origin (8). diastase, and Gomori reticulin stain), inis muscle, the superficial layer of sub- Patients at risk for or with overt T2D according to standard routine proce- cutaneous abdominal adipose tissue lying also frequently feature abnormal mito- dures, aspreviouslydescribed(10).NAFL+ directly beneath the skin, and visceral chondrial function linked to insulin re- was defined by the presence of .5% abdominal adiposetissuesurroundingintra- sistance and NAFLD (9). Of note, obese steatotic hepatocytes in a liver tissue abdominal organs, respectively (amounts people have increased hepatic mitochon- section (17) and NASH by a NAFLD activity ranging between 200 and 1,000 mg). drial capacity (10,11), which declines in score of $5 (18). The NAFL+ group Serum samples for sphingolipids quanti- nonalcoholicsteatohepatitis(NASH)along included one participant with T2D. All tation were also obtained during surgery. with increased hepatic production of re- participants maintained stable body All specimens were taken by the surgeons active oxygen species and oxidative DNA weight for at least 2 weeks before sur- at 30 min after induction of anesthesia damage (10,12,13). Lactosylceramides gery. They gave written informed consent according to standardized protocols (10). have been implicated in the development before inclusion in the study, which was Aliquotsof100 mg fresh liver samplewere of mitochondrial defects in the hearts approved by the Heinrich-Heine-University fixed in 1% formaldehyde for histological of diabetic mice (14). In patients with DusseldorfInstitutionalReview¨ Board and examination, 50 mg was transferred into coronaryheartdisease,plasmaceramides performed according to the World Med- ice-cold biopsy preservation solution for correlatednotonly withinsulinresistance ical Association Declaration of Helsinki. high-resolution respirometry, and 200 mg but also with systemic inflammation (15). was transferred into isolation buffer for iso- Recent studies identified certain dihydro- Hyperinsulinemic-Euglycemic Clamp lation of mitochondria. Samples from liver ceramide species, particularly dihydro- Test and other tissues were all rapidly snap frozen ceramide 18:0, as biomarkers of diabetes The experimental procedures have been in liquid nitrogen and stored at 280°C until onset in mice and humans (16). However, described in detail (1). Briefly,participants further analysis. it remains unclear which pathways of were asked to refrain from physical activity metabolism are related to for 3 days before the clamp test. The patient Hepatic Oxidative Capacity the adaptation of hepatic mitochondrial with T2D withdrew oral glucose-lowering High-resolution respirometry was applied function and to oxidative stress and in- medication for 3 days before the test (19). in liver tissue anddin case of sufficient flammation during progression of NAFLD, On the day of the test and 1 week before amount of sampledalso in isolated mito- from steatosis or nonalcoholic fatty liver surgery, participants arrived at the clinical chondria upon exposure to various sub- (NAFL) to NASH. research center at 0800 h, where they strates for b-oxidation, tricarboxylic acid We thus aimed to examine the re- received two venous catheters in the (TCA) cycle, and ADP titration, as pre- lationship of total and specific sphingo- antecubital veins of both arms for blood viously described, after mechanical per- lipid species in various tissues and in the sampling and infusions of glucose and meabilization of the liver sample (10,21). circulation with hepatic and peripheral insulin. A primed-continuous infusion for Mechanical permeabilization was per- insulin sensitivity, hepatic mitochondrial 10 min (3.6 mg fasting glucose [mg/dL]/ formed to avoid potential cell damage function, oxidative stress, and inflamma- 90 [mg/dL]/[min 3 kg body weight]) and and effects due to incubation with deter- tionin insulin-resistantpeoplewith NAFLD. infusion for 360 min (0.036 mg/[min 3 gents. Because of the limited sample size To this end, we used comprehensive kg body weight]) of 98% enriched of isolated mitochondria, we performed 2 metabolic phenotyping to compare obese D-[6,6- H2]glucose were performed until correlation analysis only for the liver tissue patients without steatosis (NAFL2)orwith the end of the clamp test to measure data. Of note, analysis of liver specimens steatosis(NAFL+)orwithNASHwithhealthy endogenous glucose production (EGP). yielding sufficient material for both care.diabetesjournals.org Apostolopoulou and Associates 1237

measurementsoflivertissueandisolated added. Two-phase liquid extraction was CON group compared with NAFL2 and mitochondria revealed a tight correlation performed, the supernatant was trans- NAFL+. LDL-cholesterol was higher in the of maximal uncoupled respiration be- ferred to a new tube, and the pellet was NAFL2 andNAFL+ groupsthan inthe CON tween permeabilized tissue and isolat- reextracted. Supernatants were com- group. Peripheral insulin sensitivity was ed mitochondria (Supplementary Fig. 1). binedandevaporatedundernitrogen.The comparable in the three obese groups but Maximal uncoupled (state u) respiration dried residue was reconstituted in 200 mL markedly lower than in the CON group. was measured after exposure to the HPLC solvent B (methanol/formic acid MeanHISwasnominallyhigherintheCON mitochondrial uncoupler (carbonyl cyanide 99:1;v:v,containing5mmol/Lammonium group but not significantly different from p-trifluoromethoxyphenylhydrazone)and formate) for LC-MS/MS analysis. Lipid that of the obese groups. The NAFLD adjusted for individual citrate synthase separation was achieved on a 2.1 (internal score was higher in NAFL+ and NASH than activity (CSA) and mitochondrial DNA as diameter) 3 150 mm Kinetex C8, 2.6-mm in both NAFL2 and CON. By definition of measures of mitochondrial content (10). core-shell particle column (Phenom- the groups, the level of hepatocellular enex, Torrance,CA).Plasmasphingolipids lipids was similar in CON and NAFL2 and Hepatic Lipid Peroxidation and were quantified using a similar method- lowerthan in NAFL+ andNASH,withNASH Systemic Inflammation ology requiring 50 mL serum. Sphingoli- having more profound steatosis. As a marker of lipid peroxidation, thiobar- pids were quantified using a Nexera X2 bituric acid–reactive substances (TBARS) UHPLC system coupled to a Shimadzu Tissue-SpecificSphingolipid weremeasuredinserumandlivertissueas LCMS-8050 triple quadrupole mass spec- Concentrations previouslydescribed(22).Catalaseactivity trometer operating the dual ion source in We measured total ceramide species and and 8-oxo-guanosine in hepatic tissue electrospraypositive mode. Sphingolipid dihydroceramides in serum and liver, lysates were measured with ELISA (10). species were identified based on exact which reflect de novo sphingolipid syn- Interleukin6(IL-6),tumornecrosisfactor-a mass and fragmentation patterns and thesis. Lactosylceramide and hexosylcer- (TNF-a) and IL-1 receptor antagonist (IL- verified by lipid standards. The concen- amide species result from breakdown of 1RA) were quantified using Quantikine HS tration of each metabolite was determined more complex sphingolipids and there- (IL-6, TNF-a) or Quantikine (IL-1RA) ELISA according to calibration curves using peak- fore allow examining activation of the kits (R&D Systems, Wiesbaden, Germany) to-area ratio of the analyte versus the salvage pathway, which reuses these spe- (23). Total c-Jun N-terminal kinase (tJNK) corresponding internal standard. Calibra- cies. Figure 1 provides an overview of and Thr183/Tyr185-phosphorylated (p)JNK tion curves were generated using serial sphingolipid metabolism. were quantified using specific antibodies dilutions of each target analyte. Sphingo- In serum, total ceramides were com- (Cell Signaling Technology) (10). lipid true standards were purchased from parable, whereas total dihydroceramides Avanti Polar Lipids. were higher in NASH than in CON and Laboratory Measurements NAFL2 (Fig. 2A and B), a difference which fi Blood glucose was measured with Cobas Statistical Analysis remained signi cant after adjustment c311 (Roche Diagnostics, Mannheim, Ger- Data are presented as mean 6 SD or for BMI (P = 0.01 and 0.03, respectively). many), serum concentrations of insulin median (25th, 75th percentiles) as ap- NASHhadhigherdihydroceramidespecies and C-peptide were measured with radio- propriate. Differences in metabolic char- 16:0 compared with all other groups, immunoassay (Millipore, St. Charles, MI), acteristics between the four groups of species 22:0 compared with CON and and free fatty acids (FFAs) were measured patients were calculated with ANOVA-like NAFL2, and species 24:1 compared with microfluorometrically (24). Total adipo- least-square linear regression using the CON (Fig. 2C). After BMI adjustment, nectinwas measuredinserumandplasma Tukey-Kramer method to adjust for mul- dihydroceramides 22:0 and 24:1 were samples with the Human Adiponectin tiplecomparisonsofthefourgroups. Linear increased in NASH compared with CON ELISA (Millipore). regression models were used to calculate (P = 0.01 for both) and NAFL2 (dihydro- t values (estimate/SEM) and P values of ceramide 22:0: P = 0.01; 24:1: P =0.03). Sphingolipid Measurements associations between metabolic parame- Liver total ceramides were higher in Sphingolipids were quantified using liquid ters and sphingolipid species with and NASH than in other groups (Fig. 2D). After chromatography–tandemmass spectrom- withoutadjustmentforBMI.Pvaluesfrom adjustment for BMI, total ceramides were etry (LC-MS/MS) methods (25). Briefly, two-sided tests #5% were considered to 33% and 50% higher in NASH compared flash-frozen tissue samples (40 mg) were indicate significant differences. Analyses with NAFL+ (P = 0.01) and NAFL2 (P = homogenized in 2.0 mL organic extraction were performed using SAS version 9.4 0.001). The increase in ceramide 16:0 in solvent (isopropanol:water:ethyl acetate, software (SAS Institute, Cary, NC). NAFL2 compared with CON disappeared 25:10:65; volume [v]:v:v). Immediately after adjustment for BMI. Independent afterward, 20 mL internal standard solu- of BMI, ceramide 24:0 was increased in tion was added (AL Ceramide/Sphingoid RESULTS NASHcomparedwithallothergroups(P, Internal Standard Mixture II diluted 1:10 Participants’ Characteristics 0.001). Overall, dihydroceramides and in ethanol; Avanti Polar Lipids, Alabaster, The groups had comparable age, sex, and species 16:0, 22:0, and 24:1, as well as AL). The mixture was vortexed and soni- levels of circulating triglycerides and FFAs total lactosylceramides and species 24:1, catedinanultrasonicbathfor40minat (Table 1). BMI was similar across all obese were higher in NASH than in CON (Fig. 2E 40°C. The samples were allowed to reach groups (NAFL2, NAFL+, NASH) but was and F), only before adjustment for BMI. room temperature, and 1.5 mL high- higher in all obese groups than in the CON Hexosylceramides 22:0 and 24:0 were also performance LC (HPLC)–grade water was group. HDL-cholesterol was higher in the increased in NASH (Fig. 2E), but only the 1238 Sphingolipids in the Progression of NAFLD Diabetes Care Volume 41, June 2018

Table 1—Participant characteristics CON NAFL2 NAFL+ NASH N (females) 7 (5) 7 (6) 7 (4) 7 (6) Age (years) 40 6 13 43 6 7466 12 42 6 8 BMI (kg/m2) 25.2 6 3.3¶¶¶,###,*** 49.5 6 8.3 56.1 6 7.0 51.4 6 7.1 Waist circumference (cm) 82.8 6 13.0¶¶,###,** 125.7 6 19.8‡ 144.5 6 16.2 129.2 6 17.0 Triglycerides (mg/dL) 75 (71, 94) 138 (74, 166) 135 (94, 286) 99.5 (94, 120) FFA (mmol/L) 635 (473, 761) 819 (634, 947) 509 (261, 730) 599 (577, 671) HDL-cholesterol (mg/dL) 64 (57, 83)¶,# 44 (35, 54) 35 (33,45) 52 (43, 59) LDL-cholesterol (mg/dL) 89 (72, 99)¶,# 128 (105, 153) 128 (100, 151) 118 (107, 118) Alanine aminotransferase (units/L) 25 (19, 34) 25 (14, 25) 22 (17, 25) 32 (28, 40) Aspartate aminotransferase (units/L) 19 (19, 20) 20 (17, 23) 25 (19, 27) 25.5 (18, 31) g-Glutamyltransferase (units/L) 24 (16, 35) 16 (10, 35) 22 (15, 35) 32 (28, 40) NAFLDactivityscore 0(0,2)##,*** 0(0,1)‡‡,§§§ 2 (2, 4)†† 5(4,6) Peripheral insulin sensitivity (mg/kg/min) 7.4 6 2.2¶¶,## 3.1 6 1.7 1.8 6 0.3 2.8 6 0.6 HIS index ([mg/kg/min]/[mU/mL]) 15.1 6 14.0 4.8 6 2.1 1.7 6 0.1 4.8 6 4.4 Hepatocellularlipids(%) 1(0,5)##,*** 2(0,5)‡‡,§§§ 40 (10, 40)† 45 (40, 65) Data are presented as mean 6 SD or median (quartile 1, quartile 3) unless otherwise indicated **P # 0.01, ***P , 0.001 CON vs. NASH. ¶P # 0.05, ¶¶P # 0.01,¶¶¶P,0.001CONvs.NAFL2.#P#0.05,##P# 0.01, ###P,0.001CONvs.NAFL+.‡P# 0.05, ‡‡P# 0.01,NAFL2vs.NAFL+.§§§P,0.001NAFL2vs. NASH. †P # 0.05, ††P # 0.01 NAFL+ vs. NASH.

22:0speciesremainedincreasedcompared 22:0, and 24:1) correlated positively with visceral tissue, and subcutaneous tis- with NAFL2 and NAFL+ after BMI adjust- the respective hepatic concentrations (Sup- sue. In skeletal muscle, total ceramides ment (P = 0.03 and 0.01, respectively). plementary Fig. 3A–C). Serum sphingo- were higher in NASH than in all other Furthermore, we tested the associa- myelins18:0,24:0(Supplementary Fig. 3D), groups(SupplementaryFig.2A). After BMI tions of serum ceramides with the re- and 18:1 also correlated positively with the adjustment, NASH patients had still 26% spective hepatic species to identify which respective hepatic sphingomyelin species. and 24% higher ceramides than NAFL+ possible circulating biomarkers reflect the Because sphingolipids can be derived (P = 0.002) and NAFL2 (P =0.001),respec- progression of NAFLD. Relationships were from other tissues and have been impli- tively. In subcutaneous and visceral ad- found between certain serum and hepatic cated in their metabolism, we performed ipose tissue, total ceramides were lower ceramide species. Serum concentrations a comprehensive analysis of these meta- in NASH than in NAFL2 and CON (Supple- of dihydroceramides (species 18:0, 20:0, bolites in samples for skeletal muscle, mentary Fig. 2B). After BMI adjustment, NASH had 16% lower visceral adipose tissue ceramides than NAFL+ as well as 19% and 27% lower subcutaneous adi- pose tissue ceramides than NAFL+ (P = 0.04) and NAFL2 (P = 0.003), respectively (Supplementary Fig. 2C). The complete data set of the various sphingolipids in all analyzed tissues is summarized in Supplementary Table 1.

Sphingolipids and Whole-Body and Hepatic Insulin Sensitivity Due to the evidence for a role of sphingo- lipids in insulin resistance from rodent models, we also aimed at identifying relationships between sphingolipid spe- cies and peripheral and/or hepatic insulin sensitivity. We accounted for the overall body fat mass by testing these associa- tions before and after adjustments for BMI. Serum ceramide species 16:0 (Fig. 3A), 18:0 (t = 22.2, P = 0.04), and 20:0 Figure 1—Overview figure of pathways involved in ceramide synthesis (de novo, sphingomyelinase, (t = 23.7, P = 0.001) correlated nega- catabolic, salvage) and degradation to and sphingosine-1-phosphate (P). CDase, ceramidase;CS, ceramide synthetase;DES,dihydroceramide desaturase;GCase, glycosylceramidase; tively with whole-body insulin sensitiv- GCS, glycosylceramide synthase; SK, sphingosine kinase; SMase, sphingomyelinase; SMS, sphingo- ity, as assessed by the M value. Total myelin synthetase; SPT, serine palmitoyltransferase; 3 KSR, 3-ketosphinganine reductase. serum dihydroceramides (Fig. 3B)and care.diabetesjournals.org Apostolopoulou and Associates 1239

Figure 2—Total ceramides, dihydroceramides, and dihydroceramide species in serum (A–C)andliver(D–F) in lean individuals (CON) and obese NAFL2 and NAFL+participantsandpatientswith NASH.Dataarepresentedasmean6SEM.Cer, ceramide;Hexosyl,hexoxylceramide;Lactosyl,lactosylceramide.¶P# 0.05 CON vs. NAFL2.#P # 0.05 CON vs. NAFL+. *P # 0.05, **P # 0.01, ***P , 0.001 CON vs. NASH. §P # 0.05, §§P # 0.01, §§§P , 0.001 NAFL2 vs. NASH. +P # 0.05, ++P # 0.01 NAFL+ vs. NASH. dihydroceramide 20:0 (t = 22.7, P =0.01) negatively with whole-body insulin sensitiv- sensitivity (Fig. 3C and D), which re- also correlated negatively with whole-body ity. Of note, these correlations disappeared mained only for the visceral adipose insulin sensitivity. In the liver, only dihydro- after BMI adjustment. Deoxysphinganine tissue compartment after BMI adjust- ceramides 22:0 (t = 22.31, P = 0.04) and in visceralandsubcutaneousadiposetissue ment (t = 22.51, P = 0.02). 24:1 (t = 22.3, P = 0.03) correlated related negatively to whole-body insulin In serum, specific lactosylceramide spe- cies (18:0, 20:0, and 22:0) and hexosylcer- amide 18:0 related negatively with basal EGP (t = 22.5, P = 0.03; t = 22.2, P = 0.04; t = 22.4, P = 0.03; and t = 23.1, P =0.007, respectively) after adjustment for BMI. Hepatic sphingosine related negatively with insulin-mediated EGP suppression (t = 22.3, P = 0.03).

Sphingolipids and Adiponectin Concentrations Because of a possible interactionbetween sphingolipids with adiponectin and insulin resistance, we also examined the relation- ships of adiponectin with insulin sensi- tivity and ceramide levels in our cohort. Serum adiponectin correlated positively with whole-body insulin sensitivity (M value: t =3.14,P =0.006)andtheHIS index (t =2.16,P = 0.01). In visceral adi- pose tissue, total ceramides (t = 6.42, P , 0.0001) and lactosylceramide species 14:0, 16:0, 24:1, and 24:0 (all t . 2.1, P , 0.04) related positively to circulating adiponectin concentrations.

Sphingolipids and Hepatic Oxidative Figure 3—Correlations between ceramide species in serum (A and B), visceral (visc.) fat (C), and Capacity subcutaneous (subcut.) fat (D) and whole-body insulin sensitivity (M value) across all groups (CON: Given the recently reported dynamic blue circles, NAFL2: yellow circles, NAFL+: orange circles, NASH: red circles). changes of mitochondrial function in 1240 Sphingolipids in the Progression of NAFLD Diabetes Care Volume 41, June 2018

different stages of NAFLD, we tested from TCA cycle substrates, after CSA involved in the pathogenesis of NASH, associations of sphingolipid species with correction and adjustment for BMI. Lac- such as oxidative stress and inflamma- hepatic oxidative capacity and with a tosylceramides 18:0 (t =2.33,P =0.03), tion. Liver TBARS correlated positively surrogate of mitochondrial density. Se- 20:0 (t =2.46,P = 0.02), and 14:0 (t =2.97, with liver total ceramides (Fig. 4C)and rum total lactosylceramides (Fig. 4E) P = 0.008) in serum and 22:0 (t =2.42,P = sphingomyelin 22:0. Further positive cor- and species 14:0 (t =3.6,P =0.001), 0.03) in the liver related to mitochondrial relations of TBARS with hepatic ceramide 16:0 (t =3.1,P = 0.007), and 24:1 (t = 2.3, DNA content; however, these species did 24:0 (Fig. 4D) and 24:1 disappeared after P = 0.03) were associated with increased not associate with maximal respiration in adjustment for BMI. Of note, hepatic maximal uncoupled respiration from the liver. ceramidesdidnotrelatetoserumTBARS,he- TCA cycle substrates after correction paticcatalaseactivity,and8-oxo-guanosine for CSA and adjustment for BMI. Hepatic Hepatic Ceramide Species, Oxidative (data not shown). sphinganine (t = 2.3, P = 0.03), sphingo- Stress, and Inflammation Overall, the hepatic pJNK-to-tJNK ratio sine-1-phosphate (t =2.17,P = 0.04), and The increase of sphingolipid species in correlated positively with serum and lactosylceramide 14:0 (Fig. 4F)related NASH promptedustofurtherexaminethe hepatic total ceramides (Fig. 4A and B) positively to rates of state u respiration relationship with markers of mechanisms after adjustment for BMI. In serum, specific ceramide species 14:0 (t = 3.3, P = 0.003), 16:0 (t =2.5,P = 0.02), 18:0 (t = 2.6, P = 0.01), and 24:0 (t =2.7,P =0.01), dihydroceramides species 16:0 (t = 4.8, P = 0.0001) and 22:0 (t = 4.1, P = 0.0006), hexosylceramide species 22:0 (t =3.5,P = 0.002), and lactosylceramide species 24:0 (t = 3.1, P = 0.006) all correlated positively with the pJNK-to-tJNK ratio after BMI adjustment. In the liver, ceramide 24:0 and hexosylceramides 22:0 (t =5.2,P , 0.0001), 24:0 (t =3.1,P = 0.006), and 24:1 (t =2.4,P = 0.02) also related to increased pJNK-to-tJNK ratio. The positive associa- tions of liver pJNK with hepatic total lactosylceramides and its species 16:0 and 24:1 were abolished after adjust- ment for BMI. We performed additional exploratory linear regression analyses within the four patient groups to identify possible differ- ences between the obese patients with or without NASH. Lactosylceramide 14:0 inserum(CON:t=20.14,P=0.88;NAFL2: t = 2.42, P = 0.03; NAFL+: t =2.35,P = 0.03; NASH: t = 20.87, P = 0.39), serum total lactosylceramides (CON: t =0.07,P = 0.94; NAFL2: t = 2.71, P =0.01;NAFL+:t =1.81, P = 0.09; NASH: t =0.52,P = 0.60), hepatic sphinganine, andhepatic lactosylceramide 14:0 (Supplementary Table 2) showed increased slopes in the obese NAFL2 group and decreased or even inverse slopes within the NASH group. Further- more, we found greater slopes and signif- icant correlations of liver total ceramides with TBARS in the NASH group (Supple- mentary Table 2). In addition, we assessed the role of sphingolipids not only for hepatic but also for circulating markers of inflam- mation. Although we found some cor- Figure4—Correlationsbetween liver Thr183/Tyr185–pJNK-to-tJNKratio(AandB), TBARSinthe liver(C andD),maximaluncoupledrespiration(stateu)ofmitochondriainlivertissuecorrectedforCSA(Eand relationsofserumceramidespecies14:0 F), and ceramide species in the liver (CON: blue circles, NAFL2: yellow circles, NAFL+: orange circles, (t =2.75,P = 0.01) and serum sphingosine NASH: red circles). AU, arbitrary units. (t =2.70,P = 0.01) with circulating TNF-a care.diabetesjournals.org Apostolopoulou and Associates 1241

concentrations even after adjustment for dihydroceramide species reflect activa- secondary changes occurring with obesity- BMI, neither serum nor hepatic sphingo- tion of de novo ceramide synthetic path- associated NAFLD. Importantly, we found lipidscorrelatedwith circulating IL-6 levels way, resulting from condensation of that the NASH group featured decreased (data not shown). palmitoyl-CoA with serine and mediated hepatic mitochondrial respiration along byserinepalmitoyltransferase.Targeted with increased lipid peroxidation, which inhibition ofthisenzyme in mice showed in turn correlated with hepatic ceramides. CONCLUSIONS reversalofglucose intoleranceandwhole- This extends our previous findings on the The main findings of this study are that 1) body insulin resistance in diet-induced lost adaptation of hepatic mitochondria insulin-resistant patients with NASH ex- obesity (28). Hepatic dihydroceramides due to increased oxidative stress in NASH hibit higher hepatic concentrations of 22:0 and 24:0 correlated negatively with livers (10). total ceramides,certaindihydroceramides, the M value, in line with findings in insulin- Increased adiponectin has been shown and lactosylceramides, 2) hepatic dihydro- resistant patients, underlining the role of de toplay a centralroleinNASH byimproving ceramide species 22:0 and 24:1 correlate novo ceramide synthesis in human NAFLD liver histology in response to pioglitazone negatively with whole-body insulin sensi- (8). Liver total ceramides did not relate to treatment (33) and to be associated with tivity, 3) total hepatic dihydroceramides whole-body insulin sensitivity, extending reduced tissue ceramides (29,30). Most and species and hexosylceramides and previous findings using HOMA-insulin re- recently, adiponectin receptors were at- lactosylceramides associate with hepatic sistanceasasurrogateofinsulinresistance tributed to have an inherent ceramidase oxidative stress and activation of hepatic (5). Similarly, total ceramides were also ele- activity, hydrolyzing ceramides to sphin- inflammatory pathways, and 4) higher vated in skeletal muscle of NASH patients, in gosine and FFA, which may be a novel lipid peroxidation along with lower he- linewithstudiesinsevereobesity(29,30), but mechanism underlying the beneficial patic mitochondrial respiration relates to did not relate to peripheral insulin resistance. metabolic effects of adiponectin (34). increased hepatic ceramides in obese patients Serum total lactosylceramides and sub- The current study confirmed that serum with NASH. species, as well as hepatic sphinganine adiponectin levels relate to improved Rodent studies using a high-fat diet or and sphingosine-1-phosphate, correlated hepatic and peripheralinsulinsensitivity in leptin-deficient ob/ob mice have shown with increased maximal uncoupled respi- but did not detect associations with controversialresultsfortherelationshipof ration of hepatic mitochondria. Of note, hepatic or circulating ceramides. How- hepatic accumulation of ceramides with increased lipid oxidation and oxidative ever, total ceramides and certain lacto- hepatic insulin sensitivity (26,27). Here we capacity has been identified as an im- sylceramides in visceral adipose tissue reportthatobesepatientswithNASHhave portant mechanism contributing to the positively related to serum adiponectin elevated liver total ceramides, dihydro- progression of NAFLD in obese patients levels, likely reflecting the increased total ceramides, and lactosylceramides, even (10,11). Hepatic ceramides, diacylglycer- ceramide levels in visceral adipose tissue after adjustment for BMI. Furthermore, ols, and acylcarnitines have been pre- among lean, insulin-sensitive individuals specific ceramide species (i.e., serum viously found increased, along with compared with obese individuals. dihydroceramides 22:0 and 24:1 as well induced TCA cycle, in mice with NASH Increasing hepatic ceramides and ce- as hepatic ceramide 24:0) are character- (31). Lipid oversupply with subsequent ramide 24:0, as well as hexosylceramide istic for the presence of NASH, whereas high rates of lipid oxidation leads to species 22:0, 24:0, and 24:1, paralleled other species were already elevated in oxidative stress and damage of the re- raising hepatic inflammation as assessed obesity without NAFLD. This extends spiratorychain,resultingindecreasedATP from the hepatic pJNK-to-tJNK ratio. findings showing higher liver total ceram- production (20) and insulin resistance (9). Activation of TLR4 by saturated fatty ides and dihydroceramides in insulin- Hepatic total ceramides and sphingo- acids leads to transcriptional activation resistant people (8). In the current study, myelin 22:0 associated positively with of ceramide synthases and subsequently hepaticceramidespeciesdidnotcorrelate hepatic lipid peroxidation as assessed to ceramide elevation, inhibition of Akt, with basal EGP, whereas hepatic sphingo- from TBARS. Ceramides have been sug- and insulin resistance (35). In the current sine related negatively with hepatic insu- gested to induce reactive oxygen species study, circulating TNF-a levels also corre- lin sensitivity. Hepatic insulin sensitivity production with subsequent opening lated withceramide14:0andsphingosine, alsodidnotcorrelatewithlivertotal of the mitochondrial permeability pore indicating that sphingolipids relate not ceramides among obese patients in pre- and cytochrome c release in isolated mito- only to intrahepatic but also to systemic vious studies (4). More recently, hepatic chondria of rat liver (32). More recently, inflammation in NAFLD. Of note, one insulin sensitivity was related to cytosolic lactosylceramides have been indicated previous study reported similar associa- diacylglycerol but not to total hepatic as mediators of oxidative stress by inhibi- tions with the ceramides 18:0 and 18:1 in ceramides (6). tion of the respiratory chain and increased patients with overt T2D (36). We found that serum ceramide spe- sensitivity of the mitochondrial perme- Finally, this study comprehensively cies 16:0, 18:0, and 20:0 and total and ability transition pore to Ca2+ in heart analyzed sphingolipid levels in multiple dihydroceramide species 20:0 showed in- mitochondria of diabetic mice (14). Lac- tissues, which allowed us to identify that verse associations with whole-body insulin tosylceramides and hexosylceramides ceramide subsets, specifically, and spe- sensitivity, as assessed with the gold originate from degradation of more com- cies as well as certain in standard clamp technique, but not after plex in lysosomes and serum reflect the respective species in the adjustment for BMI. This suggests that can be further reused in the “salvage liver. Plasma dihydroceramide species and, the observed correlations with M value pathway” for ceramide synthesis. Thus, specifically, 18:0 were also recently iden- weremediatedsolelybyobesity.Increased the increase in these species could reflect tified as markers for diabetes development 1242 Sphingolipids in the Progression of NAFLD Diabetes Care Volume 41, June 2018

in the following 9 years (16). In the current Funding. This study was supported in part by the 8. Luukkonen PK, Zhou Y, Sadevirta¨ S, et al. study, serum dihydroceramide 18:0 cor- Ministry of Science and Research of the State of Hepatic ceramides dissociate steatosis and insulin North Rhine-Westfalia and the German Federal resistance in patients with non-alcoholic fatty liver related with the respective concentra- Ministry of Health (BMG); by a grant from the – fi disease. J Hepatol 2016;64:1167 1175 tions in the liver. The present nding Federal Ministry for Research (BMBF) to the 9. Szendroedi J, Phielix E, Roden M. The role of might therefore be helpful for the future German Center for Diabetes Research (DZD Grant mitochondria in insulin resistance and type 2 di- use of certain serum sphingolipids as 2012); by grants from the Helmholtz portfolio abetes mellitus. Nat Rev Endocrinol 2011;8:92–103 “ biomarkers for the detection and pro- theme Metabolic Dysfunction and Common 10. KoliakiC,SzendroediJ,KaulK,etal.Adaptation Disease,” the Helmholtz Alliance to Universities gressionofNAFLDaswellasalinkbetween of hepatic mitochondrial function in humans with “Imaging and Curing Environmental Metabolic non-alcoholic fatty liver is lost in steatohepatitis. NAFLD and diabetes onset. Diseases (ICEMED),” the German Research Foun- Cell Metab 2015;21:739–746 Strengths of our study are the simulta- dation (SFB 1116), the German Diabetes Associ- 11. Sunny NE, Parks EJ, Browning JD, Burgess SC. neous analysis ofbiopsysamplesobtained ation, and the Schmutzler Stiftung; and by funding Excessive hepatic mitochondrial TCA cycle and from various tissues relevant to whole- from the Touchstone Diabetes Center at the gluconeogenesis in humans with nonalcoholic University of Texas Southwestern Medical Center – body and hepatic metabolism from obese fatty liver disease. Cell Metab 2011;14:804 810 (to P.E.S.). 12. Gornicka A, Morris-Stiff G, Thapaliya S, patients with and without steatosis and Duality of Interest. No potential conflicts of Papouchado BG, Berk M, Feldstein AE. Transcrip- NASH but also from healthy lean volunteers. interest relevant to this article were reported. tional profile of genes involved in oxidative stress All participants were thoroughly pheno- Author Contributions. M.A. and R.G. obtained and antioxidantdefenseinadietary murinemodel typed using gold-standard techniques. In- and analyzed the data and wrote, edited, and of steatohepatitis. Antioxid Redox Signal 2011;15: reviewed the manuscript. C.K. and S.G. performed – sulinsensitivity,mitochondrialfunction,lipid 437 445 the clinical experiments and edited and reviewed 13. Seki S, Kitada T, Yamada T, Sakaguchi H, peroxidation, and inflammation were all the manuscript. T.J., E.D.F., C.H., D.M., F.J., and I.E. Nakatani K, Wakasa K. In situ detection of lipid also measured at the hepatic level. performed laboratory analyses and edited and peroxidation and oxidative DNA damage in non- The main limitationresides in the cross- reviewed the manuscript. M.S. obtained tissue alcoholic fatty liver diseases. J Hepatol 2002;37: samples during surgery and edited and reviewed 56–62 sectional study design, which does not themanuscript.P.E.S.ledthesphingolipidmeasure- allow conclusions about the causality of 14. Novgorodov SA, Riley CL, Yu J, et al. Lacto- ments and wrote, reviewed, and edited the sylceramide contributes to mitochondrial dysfunc- the observed associations. Six of seven manuscript. M.R. initiated the investigation, de- tion in diabetes. J Lipid Res 2016;57:546–562 of our lean CON subjects underwent signed and ledthe clinical experiments, and wrote, 15. deMelloVD,LankinenM,SchwabU,etal.Link cholecystectomy. However, although in- reviewed, and edited the manuscript. All authors between plasma ceramides, inflammation and gave final approval of the version to be published. insulin resistance: association with serum IL-6 creased ceramide concentrations in se- M.R. is the guarantor of this work and, as such, had rumandinbilehavebeenreportedinmice concentration in patients with coronary heart full access to all the data in the study and takes disease. Diabetologia 2009;52:2612–2615 fed a lithogenic diet (37), whether chole- responsibility for the integrity of the data and the 16. Wigger L, Cruciani-Guglielmacci C, Nicolas A, cystectomy affects sphingolipid metabo- accuracy of the data analysis. et al. Plasma dihydroceramides are diabetes lism in healthy, lean humans is unknown. Prior Presentation. Parts of this study were susceptibility biomarker candidates in mice and presented in abstract form at the 77th Scientific In conclusion, several sphingolipids are humans. Cell Reports 2017;18:2269–2279 Sessions of the American Diabetes Association, 17. Brunt EM, Tiniakos DG. Histopathology of elevated in insulin-resistant people with – San Diego, CA, 9 13 June 2017. nonalcoholic fatty liver disease. World J Gastro- NASH, and certain hepatic sphingolipids, enterol 2010;16:5286–5296 such as dihydroceramides and lactosyl- References 18. Kleiner DE, Brunt EM, Van Natta M, et al.; ceramides, specifically relate to hepatic 1. Tilg H, Moschen AR, Roden M. NAFLD and Nonalcoholic Steatohepatitis Clinical Research oxidative stress and inflammation, sug- diabetes mellitus. Nat Rev Gastroenterol Hepatol Network. Design and validation of a histological – scoring system for nonalcoholic fatty liver disease. gesting that these lipid metabolites 2017;14:32 42 2. ChaurasiaB,SummersSA.Ceramidesdlipotoxic Hepatology 2005;41:1313–1321 contribute to the progression of simple inducers of metabolic disorders [published cor- 19. Apostolopoulou M, Strassburger K, Herder C, fatty liver to NASH in humans. These rection appears in Trends Endocrinol Metab et al.; GDS group. Metabolic flexibility and findings further imply different contribu- 2018;29:66–67]. Trends Endocrinol Metab 2015; oxidative capacity independently associate with tionsofdenovoceramidesynthesisand 26:538–550 insulin sensitivity in individuals with newly di- 3. Szendroedi J, Yoshimura T, Phielix E, et al. Role agnosed type 2 diabetes. Diabetologia 2016;59: the salvage pathway for peripheral insulin ofdiacylglycerolactivationofPKCuinlipid-induced 2203–2207 resistance and hepatic oxidative stress/ muscle insulin resistance in humans. Proc Natl 20. Szendroedi J, Chmelik M, Schmid AI, et al. inflammation, respectively. They further Acad Sci U S A 2014;111:9597–9602 Abnormal hepatic energy homeostasis in type 2 highlight the importance of focusing on 4. Magkos F, Su X, Bradley D, et al. Intrahepatic diabetes. Hepatology 2009;50:1079–1086 specific sphingolipid subspecies rather diacylglycerol content is associated with hepatic 21. Kuznetsov AV, Kehrer I, Kozlov AV, et al. insulin resistance in obese subjects. Gastroenter- Mitochondrial ROS production under cellular than on total levels of the different cer- ology 2012;142:1444–1446.e1442 stress: comparison of different detection meth- amides and when analyz- 5. Kumashiro N, Erion DM, Zhang D, et al. Cellular ods. Anal Bioanal Chem 2011;400:2383–2390 ing their role for hepatic inflammation mechanism of insulin resistance in nonalcoholic 22. Jelenik T, Sequaris´ G, Kaul K, et al. Tissue- and abnormal mitochondrial function. fatty liver disease. Proc Natl Acad Sci U S A 2011; specific differences in the development of insulin 108:16381–16385 resistance in a mouse model for type 1 diabetes. 6. Ter Horst KW, Gilijamse PW, Versteeg RI, et al. Diabetes 2014;63:3856–3867 Hepatic diacylglycerol-associated protein kinase 23. Herder C, Bongaerts BWC, Rathmann W, et al. Acknowledgments.TheauthorsthankDr.Julia Ce translocation links hepatic steatosis to hepatic Association of subclinical inflammation with poly- Szendroedi for her contribution to the clinical insulinresistanceinhumans.CellReports2017;19: neuropathy in the older population: KORA F4 experiments and Kai Tinnes, Myrko Esser, and 1997–2004 study. Diabetes Care 2013;36:3663–3670 Ulrike Partke (all from the Institute for Clinical 7. Kotronen A, Seppanen-Laakso¨ T, Westerbacka 24. Szendroedi J, Saxena A, Weber KS, et al.; Diabetology, German Diabetes Center, Leibniz J, et al. Hepatic stearoyl-CoA desaturase (SCD)-1 GDS Group. Cohort profile: the German Diabetes Center for Diabetes Research at Heinrich-Heine activity and diacylglycerol but not ceramide Study (GDS). Cardiovasc Diabetol 2016;15:59 University, Dusseldorf,¨ Germany) for their excel- concentrations are increased in the nonalcoholic 25. Holland WL, Adams AC, Brozinick JT, et al. An lent technical support. human fatty liver. Diabetes 2009;58:203–208 FGF21-adiponectin-ceramide axis controls energy care.diabetesjournals.org Apostolopoulou and Associates 1243

expenditure and insulin action in mice. Cell Metab 30. Warshauer JT, Lopez X, Gordillo R, et al. Effect roleof adiponectin.AlimentPharmacol Ther2010; 2013;17:790–797 of pioglitazone on plasma ceramidesinadults with 32:769–775 26. Holland WL, Miller RA, Wang ZV, et al. metabolic syndrome. Diabetes Metab Res Rev 34. Holland WL, Scherer PE. Structural biology: Receptor-mediated activation of ceramidase ac- 2015;31:734–744 receptors grease the metabolic wheels. Nature tivity initiates the pleiotropic actions of adiponectin. 31. Patterson RE, Kalavalapalli S, Williams CM, 2017;544:42–44 Nat Med 2011;17:55–63 et al. Lipotoxicity in steatohepatitis occurs de- 35. Maceyka M, Spiegel S. Sphingolipid metabo- 27. Trevino MB, Mazur-Hart D, Machida Y, et al. spite an increase in tricarboxylic acid cycle lites in inflammatory disease. Nature 2014;510: Liver perilipin 5 expression worsens hepatostea- activity. Am J Physiol EndocrinolMetab2016;310: 58–67 tosis but not insulinresistanceinhighfat-fed mice. E484–E494 36. Haus JM, Kashyap SR, Kasumov T, et al. Mol Endocrinol 2015;29:1414–1425 32. Garc´ıa-Ruiz C, Colell A, Par´ısR,Fernandez-´ Plasma ceramides are elevated in obese subjects 28. UssherJR,KovesTR,CadeteVJ,etal.Inhibition Checa JC. Direct interaction of GD3 with type 2 diabetes and correlate with the of de novo ceramide synthesis reverses diet-induced with mitochondria generates reactive oxygen species severity of insulin resistance. Diabetes2009;58: insulin resistance and enhances whole-body oxygen followed by mitochondrial permeability transition, 337–343 consumption. Diabetes 2010;59:2453–2464 cytochromecrelease,andcaspaseactivation. FASEBJ 37. Lee BJ, Kim JS, Kim BK, et al. Effects of 29. Xia JY, Morley TS, Scherer PE. The adipokine/ 2000;14:847–858 sphingolipid synthesis inhibition on cholesterol ceramide axis: key aspects of insulin sensitization. 33. Gastaldelli A, Harrison S, Belfort-Aguiar R, gallstone formation in C57BL/6J mice. J Gastro- Biochimie 2014;96:130–139 et al. Pioglitazone in the treatment of NASH: the enterol Hepatol 2010;25:1105–1110