Enhanced liver but not muscle OXPHOS in diabetes and reduced glucose output by complex I inhibition

Abstract Mitochondrial function is critical in energy metabolism. To fully capture how the mitochondrial function changes in metabolic disorders, we investigated mitochondrial function in liver and muscle of animal models mimicking different types and stages of diabetes. Type 1 diabetic mice were induced by streptozotocin (STZ) injection. The db/db mice were used as type 2 diabetic model. High‐fat diet‐induced obese mice represented pre‐diabetic stage of type 2 diabetes. Oxidative phosphorylation (OXPHOS) of isolated mitochondria was measured with Clark‐type oxygen electrode. Both in early and late stages of type 1 diabetes, liver mitochondrial OXPHOS increased markedly with complex IV‐dependent OXPHOS being the most prominent. However, ATP, ADP and AMP contents in the tissue did not change. In pre‐diabetes and early stage of type 2 diabetes, liver mitochondrial complex I and II‐dependent OXPHOS increased greatly then declined to almost normal at late stage of type 2 diabetes, among which alteration of complex I‐dependent OXPHOS was the most significant. In contrast, muscle mitochondrial OXPHOS in HFD, early‐stage type 1 and 2 diabetic mice, did not change. In vitro, among inhibitors to each complex, only complex I inhibitor rotenone decreased glucose output in primary hepatocytes without cytotoxicity both in the absence and presence of oleic acid (OA). Rotenone affected cellular energy state and had no effects on cellular and mitochondrial reactive oxygen species production. Taken together, the mitochondrial OXPHOS of liver but not muscle increased in obesity and diabetes, and only complex I inhibition may ameliorate hyperglycaemia via lowering hepatic glucose production.


| 5759
ALIMUJIANG et AL. type 2 diabetic patients. [1][2][3] Following reports indicated that mitochondrial dysfunction may be a consequence rather than cause of insulin resistance and type 2 diabetes. [4][5][6] In contrast, some studies showed that patients or rats with type 2 diabetes had normal muscle mitochondrial function. [7][8][9] Muscle mitochondrial function is also widely investigated in high-fat diet (HFD)-induced obese and pre-diabetic conditions, and the findings are also contradictory, as mitochondrial function has been reported unchanged, 10 increased 11,12 or deceased. 13,14 Additionally, it is noteworthy that data on muscle mitochondrial function/dysfunction in type 1 diabetes are sparse.
Compared to numerous studies focusing on the activity of mitochondrial respiration in skeletal muscle, reports about liver mitochondrial function in diabetes were limited and the findings were also inconsistent, with various results indicating the function decreased, 15 unchanged 16 or increased. 17,18 As different studies used different experimental paradigms and approaches to address the issue, it is difficult to integrate all the inconsistent findings on this subject.
The aim of this study was to get a clear picture of mitochondrial functional alteration in diabetes mellitus. Therefore, the disease models in our study include pre-diabetic obese mice induced with HFD, early stage and late stage of both type 1 and type 2 diabetic mice. The oxidative phosphorylation (OXPHOS) of both liver and skeletal muscle, two of the most important organs in whole energy metabolism, was examined.
Moreover, several lines of evidence proved that insulin sensitizers, including metformin, berberine and thiazolidinediones, were able to exert their anti-diabetic effects through inhibition of mitochondrial respiratory chain at complex I. [19][20][21] However, whether inhibition of other complexes than complex I in respiratory chain had beneficial effects on glucose metabolism remained unknown. As our results showed that mitochondrial OXPHOS of liver but not skeletal muscle altered dramatically in the pre-diabetic and diabetic mice, inhibitors to each complex were tested in the hepatocytes with or without oleic acid (OA) pre-treatment. Our results indicated only complex I inhibitor rotenone decreased glucose output in the primary hepatocytes without cytotoxicity through changing cellular energy state.

| Animals
All mice were obtained from SLACCAS (Shanghai, China) and were

| Type 1 diabetic mice
Male C57BL/6J mice (8 weeks) were weighed and fasted overnight prior to STZ injection. STZ (150 mg/kg) was administrated to induce type 1 diabetes. Briefly, STZ was dissolved in freshly prepared sodium citrate buffer (pH 4.5), and STZ solution was injected intraperitoneally immediately after preparation. For the early stage of type 1 diabetes, the following experiments were conducted 3 weeks after STZ injection. For the late stage of type 1 diabetes, the following experiments were conducted 3 months after STZ injection.

| Type 2 diabetic mice
For early-stage type 2 diabetic animal model, 12-week-old male C57BL/Ks db/db (db/db) mice, genetically obese and diabetic mice with leptin receptor deficiency, were used. For late stage of type 2 diabetic animal model, 42-week-old male db/db mice were used.
Corresponding control mice were male C57BL/6J mice with same age.

| Pre-diabetic obese mice
C57BL/6J male mice of 7 weeks old were randomized and fed with HFD (59% of its calories derived from fat, 15% from protein and 26% from carbohydrate, TP24220, Trophic Diet, China) or regular chow diet (CD, 1010041, Jiangsu Xietong Pharmaceutical Bioengineering Co., Ltd.) for 6 months. Bodyweight and random blood glucose were monitored once a week or every 2 weeks.

| The intraperitoneal glucose tolerance test (IPGTT) and insulin tolerance test (ITT)
Mice were fasted overnight (12 hours) before glucose injection (1 g/ kg) for IPGTT and were fasted for 6 hours prior to insulin injection (1 U/kg for type 1 diabetic mice, and 1.5 U/kg for type 2 diabetic and pre-diabetic obese mice) for ITT. Blood glucose levels were measured using tail vein blood with Roche glucometer at 0, 15, 30, 60 and 120 minutes after the injection.

| Isolation of liver and muscle mitochondria
Animals were killed by cervical dislocation, and liver and skeletal muscle of hind legs (quadriceps femoris, biceps femoris, soleus, gastrocnemius) were removed rapidly. Intact mitochondria were isolated by differential centrifugation, as previously described. 22,23 Mitochondrial protein concentration was measured using the Bradford Protein Assay Kit (Beyotime) following the manufacturer's instruction.
State 3 respiration was initiated by adding referred substrates and 0.2 mmol/L ADP (final concentration). OXPHOS capacity, indicating maximum ADP-unlimited rate of state 3, was determined by adding 2 mmol/L ADP for evaluation of O 2 consumption exerted by the maximal ATP synthesis (ATP synthase coupled with a transmembrane proton transport). Electron transport chain (ETC) capacity was measured in the presence of 0.2 mmol/L DNP (uncoupler) to evaluate the uncoupled respiration rate, which was not limited by the capacity of ATP synthesis. OCRs were calculated using strathkelvin 782 system and expressed as nmol/L atoms of O 2 /minute/mg of mitochondrial protein.

| Citrate synthase activity
Citrate synthase (CS) activity of liver and muscle mitochondria was measured as described previously 25,26 with minor modifications.
Mitochondria were treated with 5% cholate (pH 7.0) before adding to the reagent cocktail. CS activity was expressed as nmol/min/mg of mitochondrial protein.

| Measurement of adenine nucleotide levels
The early-stage STZ-induced type 1 diabetic mice and the corresponding controls were fasted overnight, and the specimens were obtained following anaesthesia with 1% amobarbital intraperitoneally injected. After adequate anaesthesia, the mice were placed supine on an operation table, and a celiotomy was performed to excise liver segments sharply. The liver segment was put into microtube and immersed into liquid nitrogen immediately.
For minimizing the degradation of adenine nucleotides (especially ATP), this procedure was completed within 4 seconds. Then, the muscle segment was taken from hind legs. After being frozen in liquid nitrogen for more than 4h, specimens were stored at −80°C.
The supernatant was neutralized with NaHCO 3 and filtered prior to determination by high-performance liquid chromatography (HPLC).

| Oil Red O staining
When the mice were killed, liver sections of each mouse were fixed using 4% paraformaldehyde, and then sent to Wuhan Goodbio technology CO., LTD, where they were sectioned and stained with haematoxylin and eosin (H&E) or frozen and stained with oil red O. The slides were observed and photographed under the microscope right after being stained.

| Primary hepatocytes isolation
Primary hepatocytes were isolated from male C57BL/6J mice of 8-10 weeks old by a two-step perfusion method as previously described with some modifications. 27,28 Briefly, the liver was perfused was at least 85% (as determined by trypan blue exclusion before plating cells). After 4-6 hours of attachment, the medium was replaced by M199 medium (Gibco) supplemented with antibiotics and 100 nmol/L dexamethasone (Sigma-Aldrich).

| Glucose output assay and Lactate dehydrogenase (LDH) cytotoxicity assay
Primary hepatocytes on 12-well plates were pre-treated with M199 medium containing antibiotics and 100 nmol/L dexamethasone for 16h overnight before the glucose output assay. After 16 hours, the cells were rinsed with PBS once and then treated with phenol redfree, glucose-free DMEM containing substrates needed for gluconeogenesis (20 mmol/L lactate, 2 mmol/L pyruvate) and inhibitors of each mitochondrial complex at different concentrations indicated in the figure. After the cells were treated for 6 hours, 50 μL culture medium was taken to measure the glucose concentration. LDH contents in the medium and cells were detected simultaneously using the culture medium with a LDH-Cytotoxicity Assay Kit (Beyotime) as instructed.

| Fatty acid treatment
OA was dissolved in H 2 O via being incubated at 75°C thermostatic water bath; then, OA was bound with fatty acid-free BSA to gain 40 mmol/L OA (with 10% BSA) stock solution. OA was diluted in DMEM, and the cells were incubated in the presence or absence of OA at the concentration of 1 mmol/L (with 0.25% BSA) for 20 hours.

| Measurement of ATP/ADP ratio in primary hepatocytes
Primary hepatocytes were cultured and treated as described above.
The ATP content and ATP/ADP ratio were measured using bioluminescence assay (Abcam catalog #65313, USA) following manufacturer's instruction.

| Detection of cellular and mitochondrial ROS
Prior to 20 μmol/L 2′,7′-dichlorofluorescin diacetate (DCFDA, Abcam catalog #113851) treatment for 45 minutes, the primary hepatocytes were treated with indicated inhibitors in the medium for glucose output determination. The cellular ROS was measured with excitation at 485 nm (Ex bandwidth: 9 nm) and emission at 535 nm as instructed. The mitochondrial ROS was detected after the hepatocytes were incubated with 2.5 μmol/L MitoSOX Red (Thermo Fisher Scientific, catalog #M36008) for 10 minutes, and fluorescence was detected at 510 nm excitation (Ex bandwidth: 9 nm) and 595 nm emission wavelengths as manufacturer's instruction.

| Western blot analysis
After treatment with indicated inhibitors of the complexes, the primary hepatocytes were washed twice with ice-cold PBS and lysed with RIPA (Beyotime) supplemented with protease and phosphatase inhibitor cocktail (Roche). The lysates were boiled at 95°C for 10 minutes and separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). Then, the separated proteins were transferred onto a nitrocellulose membrane (GE Healthcare). After blocked with 5% skim milk in the Tris-Buffered Saline Tween-20 (TBST) buffer for 1 hour at room temperature, the membrane was probed overnight at 4°C with primary antibody as listed below: phospho-AMPKα Technology 5832) and β-actin (Cell Signaling Technology 3700).
Next day, the membrane was washed and re-blotted with HRPconjugated secondary antibody (Cell Signaling Technology) at room temperature for 1h. Chemiluminescent HRP substrate (Millipore) was used to visualize protein bands by electrochemoluminescence (ImageQuant LAS4000). ImageJ was used to quantify the Western signals. Similar procedure was performed on frozen isolated liver mitochondria to detect mitochondrial complex I-V expression using total OXPHOS rodent WB cocktail (Abcam, ab110413) and VDAC1 (Abcam, ab154856) primary antibody.

| Statistical analysis
Data are expressed as means ± SEM. Two-tailed Student's t test and one-way ANOVA (SPSS 20.0, Dunnett's multiple comparisons test for the post hoc test) were used in statistical analysis. A level of P < .05 was considered as statistically significant.

| The OCR of liver mitochondria increased at both early stage and late stage of type 1 diabetes
The quality of isolated mitochondria was evaluated with mitotraker green staining and transmission electron microscopy (TEM). As shown in Figure S1, the isolated liver and muscle mitochondria were stained by mitotraker green, and the TEM images displayed intact mitochondria with ultrastructure. Then, we assessed main functional changes in liver and muscle mitochondria isolated from early-stage type 1 diabetic mice and found out the liver mitochondrial OCR of type 1 diabetes increased significantly compared with that of control ( Figure 1A-C, each representing complex I-, II-and IV-dependent respiration. A representative trace of recording complex I-dependent OCR was shown in Figure S2. However, the skeletal muscle mitochondrial OCR from early-stage type 1 diabetic mice did not change on all tested substrate conditions, except TMPD/ Asc (complex IV substrate)-stimulating condition ( Figure 1D-F).
Besides, the complex II-dependent calcium retention capacity (CRC) and H 2 O 2 production of liver mitochondria increased significantly at the early stage of type 1 diabetes ( Figure S3F,J), yet complex I-dependent membrane potential and CS activity in liver mitochondria were not affected ( Figure S3H,L), and there were no significant differences of liver lipid accumulation in the early-stage type 1 diabetic mice ( Figure S3N,O). Furthermore, we measured liver mitochondrial function of type 1 diabetic mice at their late stage of disease progress (3 months after STZ injection). The liver mitochondrial complex I-, II-dependent state 3 OCR and overall complex IV-dependent OCR of late-stage type 1 diabetic mice increased markedly, but the complex I-and II-dependent state 2, OXPHOS capacity and ETC capacity of late-stage type 1 diabetic mice only had an increasing tendency without statistic differences ( Figure 1G-I). Above all, the increase of complex IV-dependent OCR was the most significant and steadiest during disease progression of type 1 diabetes. However, no significant differences in ATP, ADP and AMP contents of liver and skeletal muscle were observed between early-stage type 1 diabetic and control mice, although the abundance of ATP content in muscle was higher than that in liver ( Figure 1J and Figure S4A,C,E,G). We further assessed the protein expression of complex I to V in isolated liver mitochondria from early-stage type 1 diabetic and control mice, and no significant difference was observed between the two groups ( Figure 1K,L), indicating the liver mitochondrial OXPHOS function enhanced without elevated expression of mitochondrial complexes. In addition, there was mild liver lipid accumulation in the late-stage type 1 diabetic mice compared to control mice ( Figure S5J,K).

| The liver mitochondrial OCR increased at early stage rather than late stage of type 2 diabetes
To examine the change of mitochondrial function during type 2 diabetes, we conducted the same experiments applied to type 1 diabetic mice on early stage and late stage of type 2 diabetic db/ db mice and the corresponding controls. Impaired glucose tolerance and insulin resistance were observed in both early-and latestage type 2 diabetes (Figures 2A,B and 3A,B). Notably, the oil red O staining of liver from both early-stage and late-stage type 2 diabetic db/db mice presented significant lipid accumulation ( Figures 2C,D and 3C,D). At the early stage of type 2 diabetes, the liver mitochondrial complex I-and II-dependent OCRs increased significantly ( Figure 2E,F). However, the skeletal muscle mitochondrial OCR from the same animals did not alter ( Figure 2I-K).
Meanwhile, the CRC, mitochondrial membrane potential and H 2 O 2 production of liver and muscle mitochondria from early-stage type 2 diabetic db/db mice did not alter ( Figure S6A-G). At the F I G U R E 1 The oxygen consumption rate (OCR) of liver but not muscle mitochondria increased in early-stage STZ-induced type 1 diabetic mice. A-C, The OCR of liver mitochondria isolated from early-stage type 1 diabetic and control mice. A, Complex I-dependent OCR. B, Complex II-dependent OCR. C, Complex IV-dependent OCR. D-F, The OCR of muscle mitochondria isolated from early-stage type 1 diabetic and control mice. D, Complex I-dependent OCR. E, Complex II-dependent OCR. F, Complex IV-dependent OCR. G-I, The OCR of liver mitochondria isolated from late-stage type 1 diabetic and control mice. G, Complex I-dependent OCR. H, Complex II-dependent OCR. I, Complex IV-dependent OCR. J, The ATP content of liver and muscle tissue from early-stage type 1 diabetic and control mice. K, L, Immunoblotting analysis and quantification of protein levels of isolated liver mitochondrial complex components (CI, CII, CIII, CIV, CV) from control and early-stage STZ-induced type 1 diabetic mice. Data are expressed as means ± SEM (n = 8). * P < .05, ** P < .01, *** P < .001 vs control late stage of type 2 diabetes, the change of liver mitochondrial OCR was not significant as at the early stage, for only the complex II-dependent OCR remained statistically significant ( Figure 3E-G). In addition, the liver mitochondrial complex II-dependent CRC of late-stage type 2 diabetic db/db mice increased, and the complex I-dependent membrane potential decreased significantly ( Figure S7B,C).

| The liver mitochondrial OCR increased after long-term HFD feeding
We obtained a pre-diabetic obese model by feeding C57BL/6J male mice with HFD for 6 months. At the end-point, the average bodyweight of HFD and control mice was 48.1 ± 10.7 vs 29.0 ± 10.5g (P < .001, Figure 4A). HFD mice developed insulin resistance and impaired glucose tolerance without elevation of fasting blood glucose levels ( Figure 4B,C). Long-term HFD exposure resulted in significant hepatic lipid accumulation as well ( Figure 4D,E), whilst the CS activity in both liver and muscle mitochondria isolated from HFD mice was higher than that from control mice (P < .001 and P = .0362, Figure 4F,G). In addition, long-term HFD increased liver complex I-dependent OCR and complex II-dependent ETC capacity ( Figure 4H,I), whilst had no effects on liver complex IV-dependent OCR and overall muscle mitochondrial OCR ( Figure 4J-M). The CRC, mitochondrial membrane potential of liver and muscle mitochondria from HFD mice did not changed, but H 2 O 2 production decreased ( Figure S8B-H).

| The mitochondrial complex I inhibitor reduced hepatic glucose output
As enhanced gluconeogenesis was indicated to play an important role in hyperglycaemia of diabetes, and our above findings showed just liver mitochondrial OXPHOS function increased in both type 1 and type 2 diabetes, we explored the effects of inhibitors to each complex in OXPHOS system on hepatic glucose output.
Meanwhile, the cytotoxicity of each inhibitor in hepatocytes was determined with LDH release. After treatment with rotenone, a specific complex I inhibitor, glucose production in the primary mouse hepatocytes, was strongly suppressed ( Figure 5B). In

F I G U R E 2
The OCR of liver mitochondria increased in early-stage db/db mice. A, B, The IPGTT and ITT for early-stage db/db and control (con) mice. C, D, Representative images showing oil red O staining of liver sections from control (C) and early-stage db/db (D) mice. E-G, The OCR of liver mitochondria from early-stage db/db and control mice. E, Complex I-dependent OCR. F, Complex II-dependent OCR. G, Complex IV-dependent OCR. H, The CS activity of liver mitochondria isolated from early-stage db/db and control mice. (I-K) The OCR of muscle mitochondria from early-stage db/db and control mice. I, Complex I-dependent OCR. J, Complex II-dependent OCR. K, Complex IV-dependent OCR. L, The CS activity of muscle mitochondria isolated from early-stage db/db and control mice. Data are expressed as means ± SEM (n = 8). * P < .05, ** P < .01, *** P < .001 vs control addition, LDH release did not change in the hepatocytes exposed to rotenone for 6 hours with the same dose range ( Figure 5A). But

| The mitochondrial complex I inhibitor reduced hepatic glucose output through affecting hepatocyte energy state
We further assessed the energetic state, cellular and mitochondrial ROS of primary hepatocytes following treatment with complex I, III and V inhibitors (representative images were shown in Figure S11A,B). The ATP content and ATP/ADP ratio decreased with increasing doses of rotenone and reached a statistically significant difference at 5 μmol/L dosage ( Figure 7A,B). Meanwhile, the cellular and mitochondrial ROS production were not affected by same

| D ISCUSS I ON
One of our significant findings was that the liver mitochondrial Buchner et al proved that the amount of mitochondria isolated per gram of liver from obese mice was less than that from lean mice. 30 Thus, when using same amount of tissue weight to evaluate the respiratory function, the results might be influenced by the different content of mitochondria not by its intrinsic function. In this study, induced OXPHOS was observed in isolated mitochondria from liver in both pre-diabetic and diabetic mice. We postulate that fuel overload in the liver may account for this alteration. It is known that GLUT2 is the major glucose transporter of hepatocytes in rodents and humans. 31 Unlike GLUT1, GLUT2 has a high Km value (15-20 mmol/L), which allows it to transport much more glucose into hepatocytes in the presence of hyperglycaemia. 32 Furthermore, hyperglycaemia is able to increase GLUT2 mRNA and protein expression in liver. 33 In addition, high level of free fatty acids in obesity and diabetes provide another type of excess fuel to liver, 34 which might explain the enhanced mitochondrial OXPHOS function in pre-diabetic obese mice without hyperglycaemia.
In contrast to liver, we observed that the skeletal muscle mitochondrial OXPHOS function of early-stage type 1, type 2 and pre-diabetic mice did not change, which is in accordance with some other studies. 8,10 We assume that the unchanged fuel load in myocytes F I G U R E 4 The OCR of liver mitochondria of HFD mice partially increased compared with CD mice. A, The bodyweight of HFD and CD mice. B, C, The IPGTT and ITT for HFD mice and CD controls. D, E, Representative images showing oil red O staining of liver sections from CD (D) and HFD mice (E). F, G, The CS activity of liver and muscle mitochondria isolated from HFD and CD mice. H-J, The OCR of liver mitochondria from HFD and CD mice. H, Complex I-dependent OCR. I, Complex II-dependent OCR. J, Complex IV-dependent OCR. K-M, The OCR of muscle mitochondria from HFD and CD mice. K, Complex I-dependent OCR. L, Complex II-dependent OCR. M, Complex IVdependent OCR. Data are expressed as means ± SEM (n = 10-15). * P < .05, ** P < .01, *** P < .001 vs control might explain this result. It is established that GLUT4 mediates insulin-stimulated glucose transport in muscle and adipose tissue. 35 In the development of insulin resistance and type 2 diabetes, GLUT4 fails to translocate to plasma membrane in response to insulin. 36 Absolute deficiency of insulin in type 1 diabetes or insulin resistance in type 2 diabetes could diminish GLUT4-mediated uptake of glucose from bloodstream in skeletal muscle, leading to reduced glucose load in myocytes. However, in aforementioned conditions, skeletal muscle shows higher lipid oxidation during post-absorptive state. 37 Reduced glucose uptake in the myotubes is compensated by excess triglycerides or free fatty acids in obesity and diabetes. Taken together, it suggests that skeletal muscle is able to keep energy Previous studies reported that both muscle-specific insulin receptor knockout mice and GLUT4-deficient mice displayed normal glucose tolerance. 38,39 Holmström et al found that in obese diabetic (db/db) mice, the OXPHOS function of glycolytic muscle increased and that of oxidative muscle decreased. 15 Since they investigated fresh tissue rather than isolated mitochondrial, and red muscles were rich in mitochondria, myoglobin and oxidative enzymes compared with white muscles, 40 it was hard to assess the exact mitochondrial function.
The isolated mitochondria in this study are from the skeletal muscle of hind legs including glycolytic and oxidative muscle. Thus, the results of our study represent the overall mitochondrial OXPHOS  4,41,42 In the present study, we did observe that CS activity increased in the muscle mitochondria of HFDinduced pre-diabetic mice, indicating the Krebs cycle was enhanced. The ATP content in both liver and muscle did not change in early-stage type 1 diabetic mice. It is known that the dephosphorylation of ATP and rephosphorylation of ADP and AMP occur repeatedly in the course of aerobic metabolism. Our data may suggest a tight balance between ATP synthesis and expenditure in diabetic liver. More ATP produced by overheated OXPHOS leads to more ATP consumption. It is known that hepatic gluconeogenesis, an ATP consuming pathway, dramatically enhances in diabetes. 36 This study suggests that, in metabolic disorders, glucose and fatty acids rush into liver to over-activate mitochondrial OXPHOS. We assume that liver mitochondrial OXPHOS directs much more ATP to gluconeogenesis than anabolic pathways like glycogen, triglyceride and protein synthesis in insulin resistance or insulin deficiency compared with healthy condition. Our data suggest induced liver OXPHOS, as a characteristic of diabetes, in return enhances hepatic gluconeogenesis and deteriorates the hyperglycaemia, becoming a pivotal process in the vicious circle of hyperglycaemia and gluconeogenesis ( Figure 8B).
Our study also revealed that the changing pattern of liver OXPHOS function was quite different between type 1 and type 2 diabetes. The liver complex IV-dependent respiration increased the most in type 1 diabetic mice. However, in type 2 diabetes, the alteration of liver OXPHOS in complex I was the most significant. Furthermore, OXPHOS function in liver steadily inclined during type 1 diabetes. In contrast, enhanced OXPHOS reached the peak at the onset of type 2 diabetes then declined to nearly normal level in the late stage of the disease ( Figure 8A, which is based on the ratio of liver mitochondrial state 3 respiration of disease models to corresponding controls shown as Table S1). The different pattern of liver OXPHOS between type 1 F I G U R E 8 A, The patterns of liver mitochondrial OXPHOS change during type 1 and type 2 diabetes development. The schematic diagram was based on the ratio of liver mitochondrial state 3 respiration of disease models to corresponding controls (shown as Table S1). Briefly, the liver mitochondrial OXPHOS function was increasing continuously during the course of type 1 diabetes, and especially, the increase of complex IV-dependent OXPHOS function is the most significant. The liver OXPHOS function kept a high level at pre-diabetes and early stage of type 2 diabetes. Then it declined slowly and returned to nearly normal level at the late stage of type 2 diabetes, among which the change of complex I-dependent OXPHOS function was the most dramatic, and the complex IV-dependent OXPHOS function was the least affected. B, The liver OXPHOS contributes differently to anabolism and gluconeogenesis pathways under different energy homeostasis. In healthy condition, majority of ATP produced by OXPHOS fuels anabolism pathway rather than gluconeogenesis (presented by the thick and thin arrows). However, in diabetic condition, liver OXPHOS directs much more ATP to gluconeogenesis than anabolic pathways like glycogen, triglyceride and protein synthesis. Thus, induced liver OXPOHS plays a pivotal role in the vicious circle that hyperglycaemia and gluconeogenesis interacts as both cause and effect and type 2 diabetes also reflects two distinct mechanisms. Burst of hyperglycaemia resulted from islet destruction fuels liver respiratory chain to work overload after type 1 diabetes onset. Parabola-shaped alteration of liver OXPHOS in type 2 diabetes indicates the respiratory chain becomes overburdened long before disease onset. This study revealed the difference in liver mitochondrial function between type 1 and type 2 diabetes, and we demonstrated liver mitochondrial complex IV may be involved in the progression of type 1 diabetes.
However, in most of published studies the complex IV-dependent respiration was rarely investigated in metabolic disorders, so the clinical implication of this finding needs further exploration.
According to our in vitro data, only the inhibition of mitochondrial complex I rather than other complexes in the primary hepatocytes under normal and lipid accumulated condition markedly decreased glucose production without cytotoxicity, suggesting inhibitors of complex II, III, IV or V may not be safe or feasible for diabetes treatment. Our previous studies demonstrated that berberine, metformin and rotenone were able to lower glucose through inhibition of mitochondrial complex I. 44,45 Coincidentally, this study showed that the liver mitochondrial complex I-dependent respiration was up-regulated greatly in the pre-diabetes and early stage of type 2 diabetes.
Our findings provide pathophysiological evidence for the usage of complex I inhibitors as anti-hyperglycaemic agents. In addition, our results warn that it should be prudent to prescribe these agents to aged patients with long duration of type 2 diabetes, because the induced liver mitochondrial complex I-dependent respiration returned to nearly normal level as the disease progressed to the late stage.
Last but not least, the limitations of our study are as follows: (a) C57BL/6J mice instead of wt/db or wt/wt littermates were used as controls of db/db mice, because the littermates were not commercially available then; (b) the adenine nucleotide levels were only measured in early-stage type 1 diabetic mice under fasted and anaesthetic condition, which might have an impact on the ratio of ATP/ ADP; and (c) our experiment design included measuring state I, II, III respiration, OXPHOS capacity and ETC capacity, but mitochondrial respiratory control ratio (RCR), an important indicator of coupling, was not measured. We will include the RCR in our future study.

| CON CLUS IONS
Our results demonstrated that the liver mitochondria were more sensitive and adaptive during metabolic disorders than skeletal muscle mitochondria. Over-active complex IV and I of liver might be an important feature during the development of type 1 and type 2 diabetes, respectively. Moreover, our in vitro results indicated that inhibition of liver complex I rather than other complexes was able to ameliorate hyperglycaemia through suppressing hepatic glucose production through reducing cellular energy state.

ACK N OWLED G EM ENTS
This study was mainly conducted at the Shanghai Key Laboratory of Diabetes Mellitus, Shanghai Diabetes Institute. We acknowledge all the staff at the platform for their excellent cooperation, and we specially thank Mrs Junxi Lu for her generous research support. The content of adenine nucleotides was determined at the Department of Chemistry, Shanghai Jiao Tong University School of Medicine, and we are thankful to Mrs Ruolin Yang for her generous support regarding the HPLC techniques.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest associated with this manuscript.

AUTH O R S' CO NTR I B UTI O N S
JY 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

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analysed during this study are included in this published article.