Impaired Hepatic Mitochondrial Capacity in Nonalcoholic Steatohepatitis Associated With Type 2 Diabetes

Recent findings challenge the paradigm of diabetes classification by proposing diabetes subtypes (1). Specifically, the severe insulin-resistant diabetes subtype features increased prevalence of hepatic steatosis at diagnosis and hepatic fibrosis at 5 years of follow-up (2). Of note, hepatic fibrosis affects at least one of six individuals with type 2 diabetes (3). These data suggest tight links between insulin resistance and nonalcoholic fatty liver disease (NAFLD) progression. Abnormal adipose tissue function, lipotoxicity (4), and glucose toxicity, as well as abnormal mitochondrial function (5), have been implicated as possible mechanisms underlying insulin resistance. Notably, rapid increases in liver fat content during the early course of type 2 diabetes have been related to deterioration of hepatic energy metabolism (6).

Indeed, type 2 diabetes has been associated with altered hepatic energy metabolism, including reduced hepatic ATP concentrations (7). Individuals with nonalcoholic steatohepatitis (NASH) also exhibit hepatic mitochondrial structural defects and impaired ATP repletion upon fructose challenge (8,9). However, recent studies have shown that mitochondrial function is not uniformly impaired in insulin resistance or NAFLD, with obese individuals with or without hepatic steatosis exhibiting up to fivefold higher maximal oxidative capacity than lean individuals, indicating hepatic mitochondrial adaptation to augmented lipid availability (10). Whether type 2 diabetes triggers loss of hepatic mitochondrial adaptation in states of severe obesity and insulin resistance is currently unclear.

Individuals with NASH also feature abnormal hepatic mitochondrial redox homeostasis, with higher generation of reactive oxygen species (ROS) paralleled by reduced antioxidant defense (10), possibly because of lipotoxicity and glucotoxicity (11). Increased lipid availability enhances oxidative metabolism and amplifies anaplerosis and cataplerosis, which is required for gluconeogenesis but also causes oxidative and endoplasmic reticulum (ER) stress (12,13). Type 2 diabetes leads not only to elevated lipid peroxidation but also to hyperglycemia-mediated increases in advanced glycation end products (AGE) (14,15). However, data on a contribution of oxidative stress and glyoxal- and methylglyoxal-induced cytotoxicity to liver injury in individuals with NASH and type 2 diabetes have not yet been reported.

Therefore, we combined in vivo measurement of hepatic insulin sensitivity with direct ex vivo analysis of mitochondrial functionality and oxidative stress from intraoperative liver samples to test the hypothesis that oxidative stress and AGEs from hyperglycemia drive the loss of hepatic mitochondrial adaptation in obese individuals with type 2 diabetes and NASH.

The prospective cohort study, BARIA-DDZ, investigates obese Caucasians before and at timed intervals for at least 1 year after bariatric (metabolic) surgery (Clinical trial reg. no. NCT01477957, ClinicalTrials.gov). All participants receive information about all procedures and risks before providing their written consent to the experimental protocol, which was approved by the ethics board of Heinrich Heine University and University Hospital Düsseldorf (Düsseldorf, Germany) and the ethics board of the North Rhine regional physicians’ association (Dsseldorf, Germany).

This cross-sectional analysis comprised 45 obese individuals with NASH (30 without [OBE] and 15 with type 2 diabetes [T2D]) as well as 14 nonobese individuals undergoing elective surgery, such as cholecystectomy or herniotomy, serving as controls (CON). Data of four OBE, three T2D, and 12 CON had been reported previously (10,16). All participants were nonsmokers and engaged only in light physical activity. T2D had no antihyperglycemic medication (n = 10), metformin only (n = 1), metformin and glucagon-like peptide 1 receptor agonist (GLP-1ra) (n = 1), DPP4 inhibitor (DPP4i) and GLP-1ra (n = 1), metformin, GLP-1ra, and short-acting insulin (n = 1), or short-acting insulin (n = 1). Metformin and DPP4i were discontinued for 3 days, GLP-1ra for 1 week, and insulin for at least 10 h prior to the tests. The participants underwent 3-h hyperinsulinemic-euglycemic clamps (primed continuous insulin infusion: 80 mU ∗ m⁻² ∗ min⁻¹ for 10 min, followed by 40 mU ∗ m⁻² ∗ min⁻¹) (Insuman Rapid; Sanofi, Frankfurt am Main, Germany) combined with d-[6,6-²H₂]glucose for measuring whole-body or peripheral (M value, calculated from steady-state glucose infusion rate with glucose space correction) and hepatic insulin sensitivities (hepatic insulin sensitivity index, calculated as 100× the inverse of the product of fasting endogenous glucose production [EGP] and fasting serum insulin concentration) (17). Rates of glucose appearance (Ra) were calculated by multiplying the tracer infusion rate by tracer enrichment, dividing by percentage of tracer enrichment in plasma, and subtracting the tracer infusion rate (17). Clamp-Ra was calculated using Steele steady-state equations. EGP was calculated from the difference between Ra and mean glucose infusion rate. Adipose tissue insulin resistance was assessed by the adipose tissue insulin resistance index, as calculated from the product of fasting serum free fatty acids (FFA) and insulin concentrations (18). Blood samples were collected before and during the clamp for measuring hormones and metabolites. NASH was defined by the steatosis, activity, and fibrosis score (19). Because T2D were older than OBE, all analyses were age adjusted in order to avoid the confounding effect of age.

Respiration rates and maximum oxidative capacity were assessed using ex vivo high-resolution respirometry in liver tissue and isolated hepatic mitochondria (Oxygraph 2k; Oroboros Instruments, Innsbruck, Austria) upon sequential titration of substrates and normalization for mitochondrial citrate synthase activity (CSA) as a marker of mitochondrial content (10,13). The protocol for tricarboxylic acid (TCA) cycle–linked respiration included addition of malate, glutamate, ATP, succinate, cytochrome C, and carbonylcyanide p-triflouromethoxyphenylhydrazone for assessment of maximum uncoupled resipration. Additionally, for testing glycolysis-linked TCA respiration and β-oxidation–linked TCA respiration, pyruvate and octanoyl-carnitine were added to the protocols, respectively. Hepatic H₂O₂ emission in isolated mitochondria was measured fluorimetrically using Amplex Red, reflecting ROS production mainly from complexes I and III as reported previously (10). Catalase activity was measured colorimetrically in hepatic lysates (Cayman Chemical Company, Ann Arbor, MI). Thiobarbituric acid reactive species (TBARS) were measured fluorimetrically (BioTek Instruments, Bad Friedrichshall, Germany) in liver homogenates (13).

DNA was extracted from liver tissue using the DNeasy Blood & Tissue Kit (Qiagen, Düsseldorf, Germany) following the manufacturer’s instructions. Briefly, liver tissue was lysed using ATL plus buffer and proteinase K at 56°C for 2 h. Later, DNA was purified using a DNeasy Mini Spin Column and dissolved in 50 μL AE buffer. DNA concentration and purity were assessed by a nanoplate reader (Tecan, Männedorf, Switzerland) and diluted to 5 ng/μL using PCR-grade H₂O. Mitochondrial DNA (mtDNA) copy number was quantified with the StepOne Plus PCR System (Applied Biosystems, Foster City, CA) using primers for nuclear gene lipoprotein lipase (LPL) (forward primer CGAGTCGTCTTTCTCCTGATGAT; reverse primer TTCTGGATTCCAATGCTTCGA) and mitochondrial gene NADH dehydrogenase subunit 1 (ND1) (forward primer CCCTAAAACCCGCCACATCT; reverse primer GAGCGATGGTGAGAGCTAAGGT). A melting curve was created to ensure primer specificity. Each sample was measured in duplicate, and the ratio of mtDNA to nuclear DNA was calculated as described (20).

Metabolites, insulin, and hs-CRP were assessed as described previously (10). FFA were assayed microfluorimetrically (Wako Chem USA Inc. Osaka, Japan) after collecting blood into ice-cold orlistat-containing vials (21).

Protein of mitofusin 1 (MFN1) and MFN2, mitochondrial fission 1 protein (FIS1), mitochondrial fission factor (MFF), DRP1, phosphorylated Ser616-DRP1 (phospho-Ser616-DRP1), PINK1, phospho-Thr257-PINK1, ubiquitin E3 ligase parkin (PARKIN), phospho-Ser65-PARKIN, protein kinase RNA-like endoplasmic reticulum (ER) kinase (PERK), activating transcription factor 4 (ATF4), eukaryotic initiation factor 2 (eIF2), and electron transport chain (ETC) complexes I–V (NDUFB8 [NADH:ubiquinone oxidoreductase subunit B8], SDHB [succinate dehydrogenase complex iron sulfur subunit B], UQCRC2 [ubiquinol-cytochrome C reductase core protein 2], COXIV [cytochrome c oxidase subunit IV], and ATP5A [ATP synthase F1 subunit α]) were quantified in total liver protein lysates (22) and normalized to GAPDH as loading control (23). Antibodies were obtained from Abcam (Cambridge, U.K.) (MFN1, MFN2, PINK1, and total OXPHOS rodent antibody cocktail, including NDUFB8, SDHB, UQCRC2, COXIV, and ATP5A), Cell Signaling Technology (Frankfurt, Germany) (DRP1, phospho-DRP1, GAPDH, COX IV, PARKIN, PERK, ATF4, and eIF2), Ubiquigent (Dundee, U.K.) (phospho-PINK and phospho-PARKIN), and Merck (FIS1).

For the quantification of optic atrophy protein (OPA1) long and short isoforms, tissue lysates (equivalent of 40 μg protein) were mixed with 4× Laemmli buffer and heated at 95°C for 5 min, loaded onto a 10% SDS gel (24). SDS-PAGE was run for 2.5 h at 40 mA per gel (20 mA per gel in the stacking gel) to achieve maximum separation of OPA1 bands. Western blotting was performed in the semidry procedure for 2.5 h at 60 mA per gel. Gels were ponceau stained, imaged, cut into stripes at 70 kDa, blocked with 5% milk in Tris-buffered saline, and incubated with their respective antibodies OPA1 (1:1,000) (self-purified antibody produced in rabbit by Pineda, Berlin, Germany) and Tubulin (1:2,000) (RRID:AB_2210370) overnight at 4°C. After second antibody incubation, blots were imaged with the Peqlab Fusion Imaging System. Blots were quantified manually in Fiji (based on ImageJ 1.53f51) by selecting bands and appropriate adjacent areas for background correction.

Plasma free AGE and hepatic protein–bound AGE were measured by isotope dilution tandem mass spectroscopy as described (25). Briefly, for protein-bound AGE, total protein extracts from the liver (∼10 mg) were prepared by homogenization in 10 mmol/L Na-phosphate buffer (pH 7.4). The soluble protein fraction was retained and then concentrated by microspin ultrafiltration (10-kDa cutoff) at 14,000 rpm for 30 min at 4°C and then washed by five cycles of concentration. An aliquot of the washed protein (100 μg/20 μL) was then hydrolyzed by serial enzymatic digestion using pepsin, pronase E, aminopeptidase, and prolidase (25). For free AGE, 20 μL plasma was diluted to 500 μL with H₂O and filtered by microspin ultrafiltration (10-kDa cutoff) at 14,000 rpm for 30 min at 4°C. The ultrafiltrate was then retained for the free adduct analysis. Approximately 30 μL sample was spiked with an equal volume of 0.2% trifluoroacetic acid in H₂O containing the isotopic standards. Samples were analyzed by liquid chromatography–tandem mass spectrometry using an ACQUITY ultrahigh-performance liquid chromatography system with a Xevo-TQS liquid chromatography–tandem mass spectrometry spectrometer (Waters Corporation, Milford, MA). AGE, including oxidation and nitration markers, were detected by electrospray positive ionization with multiple reaction monitoring. Molecular ion and fragment ion masses, as well as cone voltage and collision energies, were optimized to ±0.1 Da and ±1 eV for multiple reaction monitoring detection of the analytes. Acquisition and quantification were completed with MassLynx 4.1 and TargetLynx 2.7 (Waters Corporation).

Liver samples were fixed immediately after preparation for 2 h at room temperature by immersion in 2.5% glutaraldehyde in 190 mmol/L Na-cacodylate buffer (pH 7.4). Subsequently, they were postfixed in 1% reduced osmium tetroxide in double-distilled H₂O for 60 min and afterward stained with 2% uranyl acetate in maleate buffer (pH 4.7). The specimens were dehydrated in graded ethanol and embedded in epoxy resin. Ultrathin sections (70–80 nm) were picked up onto formvar carbon–coated grids, stained with lead citrate, and viewed in a 910 transmission electron microscope (Zeiss Elektronenmikroskopie, Oberkochen, Germany). Images of one CON, three OBE, and three T2D were included in this analysis.

Normally distributed data are given as mean and SD or SEM or otherwise as median (interquartile range). Nonnormally distributed data were log_e transformed to achieve near-normal distribution. To enhance power and avoid type I errors, residual variances were allowed to be different between groups. Statistical analyses using one-way ANCOVA adjusted for age with Tukey correction for multiple comparisons were performed using SAS (version 9.4; SAS Institute, Cary, NC). Covariate-adjusted Spearman correlation analyses in obese patients with NASH with and without type 2 diabetes were adjusted for age, sex, and BMI. Significance of differences was set at P values ≤0.05.

Despite good glycemic control, T2D were more hyperglycemic and older than OBE, so all analyses were adjusted for age. Both T2D and OBE had similar degrees of obesity (class III), fasting hyperinsulinemia, whole-body (M value) and adipose tissue insulin resistance, and low-grade inflammation (hs-CRP) (Table 1). Hepatic insulin sensitivity index was decreased in T2D and tended to be lower in OBE (P = 0.08) than in CON. In T2D, serum AST was higher than in CON and OBE, whereas FFA, triglycerides, γ-glutamyl transferase, ALT, and alkaline phosphatase were not different between all groups.

OBE and T2D had ∼20-fold greater liver fat content. Lobular inflammation was present in 90% and 100% of OBE and T2D, ballooning grade ≥1 in all OBE and T2D, and fibrosis stage ≥1 in 63% and 53% of OBE and T2D, respectively (Supplementary Table 1).

In liver tissue, maximum uncoupled mitochondrial respiration (i.e., maximum oxidative capacity) linked to TCA cycle was increased by 34% in OBE compared with CON (Fig. 1A). An increase of 59% was seen after normalization by CSA (Fig. 1C) but not by mtDNA (Fig. 1E and F). Of note, isolated mitochondria showed 182% and 170% higher oxidative capacity linked to TCA cycle and β-oxidation, respectively, in OBE than in CON (Fig. 1G and H). In contrast, T2D exhibited comparable respiration rates as CON and even a trend toward 30% lower TCA-linked oxidative capacity than OBE (Fig. 1C) (P = 0.14), while ETC complex II–linked respiration was 33% reduced versus OBE (Supplementary Fig. 1A).

Biomarkers of hepatic mitochondrial content were assessed from CSA, mtDNA copy number, and ETC complex. CSA (Fig. 2A) and ETC complexes I, III, IV, and V were comparable between groups (Supplementary Figure 1B–E), while mtDNA and complex II were lower in OBE than in CON (Fig. 2B and C).

T2D presented with 96% greater hepatic H₂O₂ emission than CON (Fig. 3A) and gradually higher H₂O₂ emission than OBE (P = 0.15). Hepatic catalase activity (Fig. 3B) and TBARS (data not shown) were similar between all groups.

Hepatic content of the mitochondrial fusion marker MFN2 was lower in T2D and OBE compared with CON (Fig. 3E). There were differences neither in total OPA1 (Fig. 3C) nor in ratios of long OPA1 isoforms to total OPA1 (Fig. 3D) or of short OPA1 isoforms to total OPA1 (data not shown) between all groups. The marker of mitochondrial fission MFF was also comparable between groups (Fig. 3G), and no differences were found after normalization to mtDNA (data not shown).

The ER stress markers PERK and ATF4 were lower in the livers of OBE and T2D compared with those of CON (all P < 0.001), while eIF2 was higher in OBE and T2D (P = 0.001 and P < 0.001 vs. CON, respectively), and phospho-eIF2 did not differ between groups (data not shown).

Transmission electron microscopy revealed mitochondrial swelling and megamitochondria in both OBE and T2D but not in CON (Fig. 2D–F). Also, autophagosomes and inhomogeneous mitochondrial degeneration were found in both NASH groups, while only OBE also presented with paracrystalline inclusions.

Hepatic methylglyoxal-derived hydroimidazolone isomer 1 (MG-1H) was higher in OBE compared with CON but not T2D, whereas fructosyl-lysine or fructosamine was higher in T2D compared with CON (Supplementary Table 2). Glyoxal-derived hydroimidazolone isomer 1, carboxyethyl-lysine, carboxymethyl-lysine, 3-nitro-tyrosine, and argpyrimidine were not different between groups (Supplementary Table 2).

Plasma fructosyl-lysine was lower in OBE compared with CON, whereas plasma concentrations of other AGE were not different between groups (Supplementary Table 2).

Of note, Spearman correlation analysis across all patients with NASH with or without type 2 diabetes revealed a positive association between plasma MG-1H and fasting blood glucose and a negative association between hepatic carboxymethyl-lysine and hepatic oxidative capacity (Supplementary Table 3).

Covariate-adjusted Spearman correlation analysis, adjusted for age, sex, and BMI, revealed negative relationships of hepatic oxidative capacity with fasting blood glucose, fasting EGP, and hepatic TBARS (Supplementary Tables 3 and 4). Lower hepatic mitochondrial content, as assessed from ETC complexes II, III, and V, was related to increased hepatic H₂O₂ emission (Supplementary Tables 3 and 4).

To further analyze the impact of fibrosis itself, the NASH cohort was stratified into those without any histological fibrosis (F0) and those with any fibrosis (F1+) and then compared with CON. Despite similar ages, F1+ exhibited severe peripheral and adipose tissue insulin resistance when compared with CON and F0 (Supplementary Table 5). Mitochondrial content from hepatic CSA and mtDNA was increased in F1+ compared with F0 (Supplementary Fig. 2A and B). Hepatic mitochondrial oxidative capacity, with and without correction for CSA, was 67% and 28% lower in F1+ than in F0, respectively (Supplementary Fig. 2C and D). Hepatic content of ETC complexes I, III, IV, and V was lower in F1+ and F0 compared with CON (Supplementary Fig. 3).

Catalase activity in F1+ was increased compared with CON but reduced compared with F0 (Supplementary Fig. 4). Hepatic MG-1H was higher in F1+ than in CON, while other hepatic AGE were similar between groups (Supplementary Fig. 5). F1+ presented with lower MFN1, MFN2, FIS1, and MFF (Supplementary Fig. 6), as well as lower PARKIN and phospho-PARKIN, than CON (Supplementary Fig. 6).

This study provides direct evidence from intraoperative human liver samples that 1) hepatic oxidative capacity is not upregulated in obese individuals with NASH and type 2 diabetes when compared to obese nondiabetic individuals with NASH with type 2 diabetes, which is possibly related to hyperglycemia, and 2) early hepatic fibrosis is also linked to lower hepatic mitochondrial oxidative capacity in the setting of enhanced AGE formation. These mechanisms may contribute to accelerated NAFLD progression and thereby determine worsening of its prognosis.

The strikingly higher mitochondrial oxidative capacity in obese individuals with NASH without type 2 diabetes, observed both in liver tissue and isolated liver mitochondria, is not explained by higher mitochondrial content because mtDNA and ETC complex II were also reduced in this group, in agreement with our previous findings of upregulated oxidative capacity in obesity and hepatic steatosis (10). While mouse model studies also support the concept of increased mitochondrial respiration and TCA cycle activity in the early stages of NAFLD (26,27), in NASH models oxidative capacity and content are diminished (28), and with the development of hyperglycemia, ketogenesis is blunted (29). Thus, alterations in hepatic mitochondrial function represent a promising treatment target in models of NASH and hepatic fibrosis (30,31). Here, our findings suggest that loss of adaptation of hepatic mitochondrial function can be possibly explained by the presence of type 2 diabetes in individuals with NASH.

Of note, the decline of mitochondrial function does not seem to be related to insulin sensitivity, because both NASH groups exhibited comparable peripheral, hepatic, and adipose tissue insulin resistance. Surprisingly, oxidative capacity was not uniformly downregulated in all NASH livers. Higher lipid peroxidation and hyperglycemia independent of age, sex, and BMI were related to the reduction of hepatic oxidative capacity in NASH and type 2 diabetes. Hepatic mitochondrial H₂O₂ release was increased in NASH with type 2 diabetes, which might reflect higher mitochondrial respiration, but not necessarily augmented ROS production, because TBARS and 3-nitro-tyrosine were not increased. Thus, our findings do not clearly support a contribution of hepatic oxidative stress to altered mitochondrial function. These findings are in contrast to evidence of unaltered muscle mitochondrial function in obesity (32). In addition, no changes in skeletal muscle mitochondrial respiration were found in the settings of lipid-induced ROS production (12) or of reduced lipid availability and lower ROS (33), suggesting organ-specific mitochondrial differences.

Mitochondrial dynamics have been shown to play a vital role in NAFLD development (34). The current study found reductions for the hepatic mitochondrial fusion marker MFN2 but not for the long OPA1 isoforms. Consequently, changes in hepatic mitochondrial fusion cannot explain the changes in oxidative phosphorylation. Similarly, mitochondrial ultrastructural alterations, such as swelling, autophagosome activity, and matrix degeneration, were found in both NASH groups, in line with previous reports (8), independent of the presence of diabetes. Although activation of ER stress represents a possible link between mitochondrial dynamics and oxidative stress (35), we found no consistent evidence of elevated ER stress, because hepatic ATF4 and PERK were not increased in the NASH groups.

Aging may lead to impaired mitochondrial functionality in skeletal muscle (36,37) or altered mitochondrial morphology in the liver (38). However, the present findings were observed to be independent of age, because age adjustment did not affect the differences in the various measures of oxidative capacity. The finding of reduced mitochondrial functionality exclusively in combined NASH and type 2 diabetes, is in agreement with previous in vivo studies showing lower hepatic ATP concentrations in individuals with type 2 diabetes compared with obese individuals with a comparable degree of steatosis (7). Of note, recent longitudinal findings show doubling of hepatocellular lipids over the first 5 years after type 2 diabetes diagnosis, which is, however, paralleled by constant hepatic ATP content (6). These prospective data suggest failure of hepatic mitochondria to adapt to increased lipid availability in the setting of hyperglycemia and act as a bridge to the present cross-sectional analysis demonstrating the key role of type 2 diabetes in alterations in mitochondrial function in metabolic liver disease.

Of note, one other study using high-resolution respirometry in intraoperative liver biopsies found no differences in hepatic oxidative phosphorylation capacity between lean and obese individuals with or without type 2 diabetes (39). This is likely due to the very small cohort size, broad variation in liver fat content, and lower severity of disease (based on lower BMI and better glucometabolic control) compared with the current study. Likewise, the unchanged hepatic mitochondrial oxidation, derived from in vivo ¹³C magnetic resonance spectroscopy (40) or plasma 3-hydroxybutyrate measurements (41), may be the result of the inclusion of normal to overweight individuals without histological NAFLD staging or of the distinct methods assessing other features of mitochondrial function.

In type 2 diabetes, hyperglycemia leads to formation of reactive dicarbonyls, such as glyoxal and methylglyoxal, which can cause protein and nucleic carbonylation and oxidative stress in hepatocytes, among other cells (42,43). This study showed that fasting blood glucose are negatively related to hepatic oxidative capacity, even after adjustments for age, sex, and BMI, underlining a role of hyperglycemia and glucotoxicity in impaired hepatic energy metabolism in NAFLD, as suggested previously for type 1 diabetes (44). Still, liver MG-1H was only increased in individuals with NASH without type 2 diabetes, whereas plasma MG-1H levels were higher in T2D versus OBE. Furthermore, a majority of hepatic and plasma AGE were comparable between groups. Thus, these findings do not support the concept that hepatic accumulation of AGE significantly contributes to enhanced oxidative stress or loss of mitochondrial adaptation in combined NASH and type 2 diabetes. Moreover, plasma MG-1H levels were closely related to fasting glycemia but not to findings on hepatic metabolism.

Hepatic fibrosis rather than inflammation is the key histological feature of NAFLD, defining the progression of disease, especially in insulin-resistant cohorts (45,46). Our findings suggest that the reduction of hepatic oxidative capacity is also key for hepatic fibrosis and loss of mitochondrial adaptation in NASH, because individuals exhibiting any degree of hepatic fibrosis had lower mitochondrial respiration rates than those showing no signs of fibrosis. These data further corroborate the relevance of hepatic mitochondria in the progression of NAFLD as well as in metabolic liver injury. Along these lines, individuals with any hepatic fibrosis (F1+) showed lower hepatic oxidative capacity as well as reduced hepatic catalase activity, possibly reflecting declining antioxidant defense capacity. Of note, hepatic MG-1H was increased in patients with fibrosis, suggesting that AGE formation related to failing mitochondria along with hyperglycemia might be associated with the progression of fibrosis. This is in agreement with the key role of ROS in the development of liver fibrosis (47). A possible contribution of hyperglycemia could not be excluded here, because F1+ patients exhibited slightly increased HbA_1c.

Several limitations must be mentioned, such as the low availability of intraoperative liver samples from healthy individuals without liver disease, which was due to ethical considerations. Of note, the small number of samples from this control group may explain some of the inconsistencies between expression data and functional data. In addition, the design of a cross-sectional study does not allow us to draw conclusions as to the time course of changes in the different groups. Nevertheless, the comprehensive phenotyping provides novel insights into pathophysiological alterations. Finally, the absence of a uniformly accepted reference method impedes the assessment of mitochondrial content.

Taken together, this study shows evidence of mitochondrial function decline as a common feature of patients with NASH who also have type 2 diabetes or hepatic fibrosis, representing different axes of hepatic metabolic disease. In type 2 diabetes, increased ROS formation along with hyperglycemia might be linked to derangements of mitochondrial respiration. Liver fibrosis is characterized by reductions in the content of ETC complexes and increased AGE formation. Thus, loss of mitochondrial adaptation seems to be closely related to metabolic liver disease and is possibly linked to its progression.

Acknowledgments. The authors thank all volunteers for their participation in this study and also thank Drs. Tomas Jelenik, Chrysi Koliaki, Daniel Markgraf, and Julia Szendroedi for contributions to clinical examination and/or laboratory assessments as well as Kai Tinnes, Myrko Esser, David Höhn, Andrea Sparla, Kerstin Förster, Fariba Zivehe, Jan-Marc Leonhard, Michelle Reina Do Fundo, Gisela Pansegrau, and Sandra Schmidt for excellent technical assistance.

Funding. This study was supported in part by the German Diabetes Center, which is funded by the Ministry of Culture and Science of the State of North Rhine-Westphalia and the German Federal Ministry of Health, through grants of the Federal Ministry for Research to the German Center for Diabetes Research (grant 2016). Parts of the study were also supported by grants from the European Funds for Regional Development (EFRE-0400191), EUREKA Eurostars-2 (E! 113230 DIA-PEP), the German Research Foundation (SFB 1116/2 and GRK 2576), the German Diabetes Association, and Schmutzler Stiftung.

The funding sources had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Duality of Interest. M.R. is on scientific advisory boards of Allergan, AstraZeneca, Bristol-Myers Squibb, Eli Lilly, Gilead Sciences, Inventiva, Intercept Pharma, Novartis, Novo Nordisk, Servier Laboratories, Target NRW, and Terra Firma and has received support for investigator-initiated studies from Boehringer Ingelheim, Nutricia/Danone, and Sanofi. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.G. performed the clinical experiments, collected and analyzed data, and wrote, edited, and reviewed the manuscript. S.K. and T.S. performed clinical experiments and edited and reviewed the manuscript. D.P., L.M., B.D., M.W., M.Z., and A.S.R. performed laboratory analyses and edited and reviewed the manuscript. K.S. performed statistical analyses and edited and reviewed the manuscript. E.S. and M.S. performed bariatric surgery procedures and edited and reviewed the manuscript. I.E. performed liver histology and edited and reviewed the manuscript. J.W. performed transmission electron microscopic measurements and edited and reviewed the manuscript. T.F and P.N. performed analysis of AGE and edited and reviewed the manuscript. M.R. initiated the investigation, designed and led the clinical experiments, and wrote, reviewed, and edited the manuscript. All authors gave final approval of the version to be published. M.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.