Mitochondrial Uncoupling Coordinated With PDH Activation Safely Ameliorates Hyperglycemia via Promoting Glucose Oxidation

Type 2 diabetes (T2D), which is characterized by hyperglycemia and insulin resistance, has become a global epidemic. Although several antidiabetic drugs have been applied in the clinic, few of them are effective in correcting the underlying pathology of insulin resistance. Therefore, the identification of novel drugs with new mechanisms to improve insulin resistance is urgently needed.

Obesity is a highly intricate disease and is one of the strongest risk factors for T2D through obesity-driven insulin resistance in liver, adipose tissue, and skeletal muscle, combined with insufficient insulin secretion to overcome this resistance (1). It has been reported that excessive ectopic lipid accumulation is an important contributor to insulin resistance through the development of lipid-derived lipotoxicity in insulin-sensitive tissue (2). How lipid accumulates in peripheral tissue has not been precisely clarified, but several mechanisms involved in the development of insulin resistance have been well documented, including increased circulating cytokines, inflammatory cell infiltration, and intracellular lipid metabolites (3–7). Therefore, reducing ectopic lipid accumulation in peripheral tissue might be an attractive strategy for treating T2D. Mitochondrial uncoupling, which promotes the proton gradient across the mitochondrial inner member through non-ATP synthase and creates a futile cycle of glucose and lipid oxidation (8–10), is an energy-wasting strategy to combat ectopic lipid accumulation.

There is well-known evidence for the effectiveness of the mitochondrial uncoupler 2,4-dinitrophenol (DNP), which was approved for weight loss in the clinic. However, DNP therapy ceased in the 1930s due to its severe adverse effects, such as hyperlactacidemia, hyperthermia, and even death (11–15), which limited the further development of uncoupling agents. The efficacy of DNP administered in vivo was largely attributable to the expense of fat as fuel, and fat combustion resulted in weight loss but also heat overproduction by uncoupling actions. Therefore, stimulation of lipid oxidation by uncoupling activity seems to be a double-edged sword (16). In addition, ATP production was dramatically decreased through the mitochondrial electronic chain by uncoupling actions, and then glucose was forced to undergo anaerobic glycolysis for ATP production, leading to lactate overproduction (17,18). The toxicities of DNP were mostly on-target effects related to uncoupling actions (19). Thus, it is crucial to dissociate the therapeutic benefits of uncouplers from their adverse effects.

To relieve the main adverse effects derived from fat combustion and anaerobic glycolysis, we hypothesized that switching anaerobic glycolysis to glucose oxidation would remit fat burning and lactate overproduction and maintain hypoglycemic actions. Dichloroacetic acid (DCA), a pyruvate dehydrogenase (PDH) activator via the inhibition of PDH kinases (PDKs) (20–22), was identified as meeting our expectations. We verified that DCA combined with DNP in vivo ameliorated hyperglycemia by preferentially inducing the oxidization of glucose over fatty acids and protected mice from hyperlactacidemia, hyperthermia, and death. Encouraged by the excellent profile of DCA combined with DNP, a novel compound possessing both mitochondrial uncoupling and PDH activation effects was identified and confirmed to improve hyperglycemia without associated systemic toxicities in vivo. Therefore, our work opens the possibility of exploring a new class of mitochondrial uncouplers for treating T2D.

DNP was obtained from Sinopharm Chemical Reagent. Sodium dichloroacetate and streptozocin (STZ) were obtain from Sigma-Aldrich. 9,10-[³H]-palmitic acid, 1-[¹⁴C]-glucose, and 1,2-[³H]-deoxy-d-glucose were obtained from PerkinElmer. The primary antibodies of PDH, protein kinase C (PKC)-θ, PKC-ε, Na⁺-K⁺-ATPase, pSer473-Akt, Akt, and GAPDH were from Cell Signaling Technology. The pSer473-PDH antibody was from Novus Biologicals.

L6 myoblasts were cultured in DMEM supplemented with 10% FBS (Invitrogen). For the differentiation of L6 myotubes, the concentration of FBS was reduced to 2% for 6 days. Mouse primary hepatocytes were isolated using Selgen’s two-step perfusion method (23) and were maintained in low-glucose DMEM supplemented with 10% FBS. The 3T3-L1 cells were cultured in DMEM supplemented with 10% CS. The differentiation of 3T3-L1 cells was conducted as described previously (24).

L6 myotubes, hepatocytes, and 3T3-L1 cells were plated in an XF96 plate (Seahorse Biosciences). Cells were equilibrated in DMEM containing 25 mmol/L glucose and 2 mmol/L sodium pyruvate. The baseline oxygen consumption rate (OCR) was recorded, and then oligomycin, test compounds, and carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone were injected, in turn, during continuous oxygen measurements. For combination OCR and extracellular acidification rate (ECAR) assay with DNP, the baseline OCR and ECAR values were recorded, and then test compounds and DNP were injected, in turn, during continuous oxygen and pH measurements.

L6 myotubes, hepatocytes, and 3T3-L1 cells in six-well cell culture plates were scraped into 100 μL of 4% (v/v) ice-cold HClO₄. The acidic supernatants were neutralized and used for high-performance liquid chromatography determination as described previously (25).

Cells were cultured in a 96-well plate and treated with compounds in serum-free cell culture medium containing 0.5% BSA for the indicated time. Lactate in the medium was measured by the lactate assay kit.

The 2-deoxyglucose uptake was measured as described previously (26). Briefly, L6 myotubes were incubated in serum-free medium containing the respective treatments in 24-well plate for 2.5 h, followed by the addition of 100 μmol/L 2-deoxyglucose and 0.5 μCi per well of 2-deoxy-[³H]-d-glucose in HBS buffer. After 10-min incubation, cells were washed with ice-cold PBS and lysed using 0.1% Triton X-100. Radioactivity was quantitated by a scintillation counter.

L6 myotubes were incubated in serum-free medium containing 1,2-[¹⁴C]-glucose with the appropriate treatments in a 24-well plate for 4 h. Each well was then covered with a piece of Whatman paper. Reactions were stopped following the injection of 70% perchloric acid into the medium. In the NaOH filter paper, ¹⁴C-labeled CO₂ formed from glucose oxidation was then collected. The filter paper traps were counted in a scintillation counter.

Fatty acid oxidation was measured as reported previously (25), with some modifications. Cells were exposed to DMEM containing 0.25 mmol/L palmitate and 1.5 μCi of 9,10-[³H]-palmitic acid for 4 h. The nonmetabolized palmitate in the medium was absorbed with charcoal slurry (0.1 g/mL charcoal in 20 mmol/L Tris-HCl pH 7.5) for 30 min and then removed by centrifugation. The radioactivity of the medium was measured by a scintillation counter.

All animal experiments were approved by the Animal Ethics Committee of the Shanghai Institute of Materia Medica. All animals were housed in a temperature-controlled room (22°C ± 2°C) with a 12-h light/dark cycle. The hyperglycemic mouse model was implemented with C57BL/6J mice injected intraperitoneally daily with 60 mg/kg STZ for 6 days. For high-fat diet (HFD) mice, 6-week-old male C57BL/6J mice were fed with an HFD (D12492; Research Diets). The 14-week-old mice were randomly assigned and gavaged with 300 mg/kg DCA and 10 mg/kg DNP individually or together for 4 weeks. For db/db mice, 7-week-old mice were assigned randomly and gavaged with either vehicle or 6j for 7 weeks. Body weight and food intake were recorded daily. Blood glucose levels were measured following 6-h starvation. Glucose tolerance test and insulin tolerance test were performed in 6-h or 4-h starvation mice. Finally, HFD mice and db/db mice were killed after injection with insulin for 15 min, and tissues were freeze clamped for further experiments.

The energy expenditure (EE), OCR, respiratory exchange rate (RER), and activity of the mice were determined by the Oxymax system (TSE PhenoMaster). Rectal core temperature was detected by BAT-12 microprobe digital thermometer and RET-3 mouse rectal probe (Physitemp Instruments, Clifton, NJ).

For the quantification of intrahepatic and intramuscular triglyceride (TG), tissue was homogenized with 0.5 mL PBS, and then total TG was extracted with 1.5 mL chloroform and methanol (2:1 v/v) overnight. The samples were centrifuged, and the bottom liquid phase was transferred to a clean tube and air dried. TG was then dissolved in 1 mL of ethanol containing 1% Triton X-100. TG contents were determined using the TG determination kit and normalized to the tissue weight. Tissue diacylglycerol (DAG) was detected by a DAG assay kit (EIAab, Wuhan, China).

Pharmacokinetic analyses of oral DCA and DNP were performed in C57BL/6J mice. Briefly, 1 g/kg DCA and 30 mg/kg DNP individually or together were dissolved in 0.5% methyl cellulose and given to male C57BL/6J mice, and blood samples and tissue were collected at the indicated time. Tissue distribution of oral 6j was determined in db/db mice. Briefly, 40 mg/kg 6j was dissolved in 0.5% methyl cellulose and given to db/db mice. Tissue was collected 2 h after administration. All samples were subjected to liquid chromatography–tandem mass spectrometry analyses.

For liver histological studies, mice were euthanized by decapitation, and liver tissues were fixed with neutral buffered formalin 10% and embedded in paraffin. Tissue sections were prepared and stained with hematoxylin-eosin.

L6 myotubes and hepatocytes were seeded in a 24-well plate and transfected with synthesized siRNA by Lipofectamine RNAiMAX Reagent (Thermo Fisher) for 48 h for follow-up experiments. The sequences of oligonucleotide used are described in Supplementary Table 1.

Total RNA was extracted from mouse muscle, liver, or inguinal white adipose tissue (iWAT) lysates using RNAiso Plus reagent (Takara, Kusatsu, Japan) and standard protocol. Total RNA from each sample was then used to generate high-fidelity cDNA by PrimeScript RT reagent kit (Takara) for quantitative PCR analyses. The relative expression levels of each gene were normalized against the expression of 18s. The primers used are described in Supplementary Table 2.

Cell and tissue lysates were subjected to electrophoresis through SDS-PAGE and blotted with antibodies. For the translocation of PKC-θ and PKC-ε in muscle and liver, protein was isolated from the membrane and cytosol by a membrane and cytosol protein extraction kit (Yeasen, Shanghai, China). Then protein was electrophoresed through SDS-PAGE and blotted with antibodies. The immunoblots were visualized by chemiluminescence using the enhanced chemiluminescence Western Blotting System. Quantification was determined by measuring band intensities using ImageJ software.

Results represented means ± SEM. Differences between two groups were examined using the unpaired two-tailed Student’s t test. Differences between four groups were assessed by one-way ANOVA. P < 0.05 was regarded as significant.

All data needed to evaluate the conclusions in the study are present in the article or Supplementary Data. Additional data related to this article may be requested from the authors.

The adverse effects of DNP are mostly on-target effects related to uncoupling actions (19); thus, we examined the uncoupling effects on energy metabolism in vitro. In myotubes, hepatocytes, and 3T3-L1 cells, DNP stimulated the OCR in the presence of oligomycin and increased the ADP/ATP ratio, meeting the criteria for uncoupling agents (Supplementary Fig. 1A and B). As expected, lactate production and fatty acid oxidation were robustly stimulated by DNP in cells (Supplementary Fig. 1C and D). However, the increase in glucose oxidation induced by DNP in myotubes was weaker than its effect on fatty acid oxidation (Supplementary Fig. 1E). Increased glucose uptake by DNP was also observed in myotubes, but it was induced at lower concentrations of DNP than those required for glucose oxidation, suggesting that part of the glucose uptake was contributed by lactate production (Supplementary Fig. 1F).

On the basis of the striking adverse effects of DNP derived from fat combustion and anaerobic glycolysis, we assumed that inhibition of lipid oxidation or lactate production would achieve our toxicity-reducing goal. We tried to combine DNP with etomoxir (carnitine palmitoyl-transferase I inhibitor), and we found that the combination treatment inhibited the increased OCR induced by DNP without affecting the ECAR in myotubes (Supplementary Fig. 2A and B). Otherwise, DNP combined with galloflavin (27) (lactate dehydrogenase inhibitor) decreased the ECAR and increased the OCR without affecting DNP-induced lipid oxidation upon etomoxir challenge (Supplementary Fig. 2C and D). Thus, inhibition of lipid oxidation or lactate production derived from DNP might not achieve the toxicity-reducing goal.

Next, we hypothesized that switching anaerobic glycolysis to glucose oxidation might remit fat burning and lactate overproduction. We designed a combination assay with DNP based on the OCR and ECAR to test diverse compounds and determine the different characteristics of metabolic profiles. On the basis of the ratio of combination treatment to DNP treatment alone in myotubes, DCA and amiloride were shown to meet our expectations, each decreasing the ECAR without affecting the OCR upon DNP uncoupling (Fig. 1A and B and Supplementary Table 3). However, amiloride treatment was reported to cause hyperkalemia in the clinic (28,29), which is similar to DNP. Thus, DCA, a classical PDH activator, acts as a switch from anaerobic glycolysis and lipid oxidation to glucose oxidation for cellular energy homeostasis (20,22,30) and was considered further as a candidate. In myotubes, hepatocytes, and 3T3-L1 cells, DCA dose-dependently inhibited the ECAR without affecting the OCR (Fig. 1C). Moreover, these phenomena also occurred in cells treated with different uncouplers and PDH activators (Supplementary Fig. 2E).

Next, we explored the effects of DNP combined with DCA (combination treatment) on metabolic profiles in vitro. PDH is regulated by four PDKs, which phosphorylate and inhibit PDH activity (31). Reduced phosphorylated Ser293-PDH was observed with DCA treatment and combination treatment in vitro, while DNP alone had no effects (Fig. 1D and Supplementary Fig. 2F). On the basis of the inhibition of anaerobic glycolysis and fatty acid oxidation by DCA, the stimulation of lactate production and fatty acid oxidation by DNP decreased after combination treatment in cells. Furthermore, fatty acid oxidation was adjusted to control levels by combination treatment (Fig. 1E and F). Because of the distinct mechanisms exerted by DCA/DNP alone on glucose oxidation, combination treatment further promoted glucose oxidation in myotubes (Fig. 1G). These data implied that combination treatment resulted in a preference for glucose as an oxidative substrate in vitro.

Then, we examined the metabolic effects of DNP/DCA treatment and combination treatment in mice through indirect calorimetry (IDC). IDC showed that DNP or combination treatment, but not DCA, increased the EE and the OCR without affecting activity for 2 h after oral administration. Meanwhile, DCA and combination treatment exhibited a higher RER than that of the DNP group, which indicated that a larger proportion of RER occurring with combination treatment was derived from glucose oxidation by DCA. In the next 3–5 h, the effect on RER induced by DCA disappeared, while the OCR and EE were still elevated in the DNP and combination treatment groups (Fig. 1H). Consistently, IDC performance after combination treatment was evidenced by lower phosphorylated Ser293-PDH levels in skeletal muscle and liver but not in iWAT (Supplementary Fig. 3). Previous research reported that glucose-derived oxidation was not significantly affected by DCA-induced PDH activation in adipose tissue (32), suggesting that PDH activation in adipose tissue might contribute little to EE. Our data implied that combination treatment drives metabolic cells to prefer glucose over fatty acids as an oxidative substrate in vivo.

The effects of combination treatment on the preferential promotion of glucose oxidation prompted us to investigate the effect on glycemic control in vivo. DNP and DCA individually enhanced glucose clearance in normal and hyperglycemic mice based on glucose challenge. Moreover, combination treatment further promoted glucose clearance in both mouse models compared with DNP alone (Fig. 2A and B). Thus, these data suggested that DNP combined with DCA enhanced the hypoglycemic effect in vivo.

Then, we further examined whether combination treatment achieved the toxicity-reducing goal. The metabolic effects of PDH activation in mice were demonstrated with a single dose of DCA applied for a short time (Supplementary Fig. 3). With 2 weeks of daily DCA pretreatment (multidose DCA), continuous activation of PDH was observed in skeletal muscle, liver, and iWAT, accompanied by downregulated PDK1, 2, 3, and 4 (Fig. 2C and D). Following gavage with multidose DCA, the death caused by 60 mg/kg DNP was avoided (Fig. 2E). In addition, lactate dehydrogenase release was significantly reduced by DCA protection, suggesting that tissue injury might be relieved (Fig. 2F). Mice gavaged with 30 mg/kg DNP did not die, but hyperthermia emerged. Inhibition of lipid oxidation by PDH activation in PDK2/PDK4 double-deficiency mice was reported to induce hypothermia (33). Consistently, hypothermia occurred following DCA treatment. However, this aberrant rectal temperature was adjusted by combination treatment (Fig. 2G). Also, the increase in plasma lactate by DNP was reduced by combination treatment (Fig. 2H). Additionally, we observed that the pharmacokinetic profiles and tissue distribution of DNP and DCA were unchanged with combination treatment compared with DCA/DNP–treated mice (Supplementary Fig. 4). These data implied that combination treatment protected against adverse effects due to uncoupling by DNP.

We identified an effect-enhancing and toxicity-reducing method to harness PDH activation with DNP to improve hyperglycemia. However, from the phenotype of muscle-specific PDK2/PDK4 double-knockout mice, the hypoglycemic actions of PDH activation driving HFD mice to switch their oxidative substrates from fatty acids to glucose led to increased re-esterification of fatty acids into DAG and TG (34). DAG has been confirmed to be closely related to insulin resistance through activation of PKC isoform and decreased insulin signaling in vivo (35–38). Thus, we tested whether combination treatment could ameliorate insulin resistance in HFD mice relative to DCA treatment. Although DCA treatment decreased blood glucose, insulin resistance was aggravated (Fig. 3A and B). Additionally, DCA treatment increased the organ index of the liver over body weight (Fig. 3C). Furthermore, histological analyses showed that mice treated with DCA showed exaggerative lipid accumulation in the liver, which was confirmed by quantification of the TG and DAG contents (Fig. 3D, E, and H). Additionally, we assessed intramuscular lipid contents and found that DCA treatment induced intramuscular TG and DAG accumulation (Fig. 3F and G). Next, increase in the membrane translocation of PKC-θ and PKC-ε and reduction in insulin-stimulated phosphorylated Ser473-Akt occurred in skeletal muscle and liver (Fig. 3I–K). With combination treatment, the exaggerated hepatic steatosis induced by DCA was relieved according to histological analyses. Combination treatment reduced intramuscular and intrahepatic lipid accumulation and membrane translocation of PKC-θ in skeletal muscle and PKC-ε in liver, leading to the restoration of insulin sensitivity in skeletal muscle and liver (Fig. 3). Taken together, these data suggested that DNP restored the sensitivity of metabolic tissues to insulin impaired by DCA, which might be partially attributed to the rectification of DAG accumulation and activation of PKC-ε and PKC-θ.

Through the aforementioned evidence, we demonstrated that mitochondrial uncoupling combined with PDH activation might be a potentially safe and effective strategy for hyperglycemia therapy (Fig. 4A). Then, we assumed that a small molecule possessing both the above dual actions might safely ameliorate hyperglycemia. On the basis of our previous research on high-throughput screening for modulators of mitochondrial membrane potential in myotubes (39), LGH00277 was selected as an active hit, and its weak uncoupling activity was also confirmed (Supplementary Fig. 5A–C). Then, a series of derivatives were synthesized (Medicinal Chemistry Section in the Supplementary Data), the OCR was determined to reflect uncoupling activity, and ECAR was determined to indirectly reflect lactate production upon uncoupling challenge (Fig. 4B). Then, the phosphorylated Ser293-PDH was confirmed. With the increase in OCR, a decrease in phosphorylated Ser293-PDH, and no increase in ECAR, lactate content was examined following 12-h treatment in myotubes. By analyzing uncoupling activity, Ser293-PDH phosphorylation, and lactate production, 6j was identified, in myotubes, as a potential compound for validation of our novel concept (Fig. 4B and Supplementary Fig. 5D–F).

We then examined the uncoupling activity and PDH activation effects of 6j in hepatocytes and 3T3-L1 cells and found that 6j stimulated the OCR in hepatocytes but not in 3T3-L1 cells. Increased ADP/ATP ratios and reduced Ser293-PDH phosphorylation by 6j were also observed in myotubes and hepatocytes (Fig. 4C–E). These results suggested that 6j was a myotube- and hepatocyte-specific mitochondrial uncoupler and PDH activator. To address the uncoupling action of 6j, glucose uptake and glucose oxidation was stimulated by 6j in myotubes (Fig. 4F and G). However, fatty acid oxidation and lactate production were not induced by 6j in myotubes and hepatocytes (Fig. 4H and I). Then, we examined the metabolic effects of orally administered 6j in vivo. A tissue distribution analysis showed that the 6j concentration in liver and skeletal muscle might reach the working concentration for myotubes or hepatocytes for PDH activation and uncoupling in db/db mice (Supplementary Fig. 6A). The reduction in phosphorylated Ser293-PDH was observed in skeletal muscle and liver but not in iWAT (Supplementary Fig. 6B). Meanwhile, IDC studies showed that mice gavaged with 6j had increased EE, OCR, and RER without effects on activity, indicating a preferential proportion of EE was obtained from glucose oxidation (Supplementary Fig. 6C). These data implied that 6j, with the ability to induce uncoupling and PDH activation effects, preferentially promoted glucose oxidation in vitro and in vivo.

We further examined whether 6j was effective in improving glycemic control and insulin resistance in HFD mice. Daily treatment with 6j for 4 weeks led to a reduction in blood glucose. Also, the capacities of glucose tolerance and sensitivity to insulin were improved in mice that received 6j chronically (Supplementary Fig. 7). Then, we further examined the similar effects in db/db mice. The reduced blood glucose and enhanced glucose tolerance were observed during 6j treatment chronically (Fig. 5A–C). Furthermore, improved sensitivity to insulin and increased insulin-stimulated phosphorylation of Ser473-Akt in skeletal muscle and liver were observed in the 6j-treated mice (Fig. 5D–F). Then, we determined whether 6j improved insulin sensitivity via reducing ectopic lipid accumulation. Histological analysis of livers showed that 6j-treated mice had little intracellular lipid accumulation, which was evidenced by the quantification of TG and DAG contents (Fig. 5G and H). Also, a decrease in intramuscular TG and DAG was also observed in the 6j-treated group (Fig. 5I). On the basis of the reduction of intrahepatic and intramuscular DAG, membrane translocation of PKC-θ in skeletal muscle and PKC-ε in liver was significantly decreased (Fig. 5J and K). We also analyzed the uncoupling effects of 6j on body temperature and lactate production during chronic treatment and did not observe any differences between vehicle- and 6j-treated mice (Fig. 5L and M). Collectively, the in vivo results supported that 6j could improve hyperglycemia and insulin resistance without lactate and heat overproduction.

The in vitro and in vivo results support the hypothesis that 6j improves glycemia and insulin resistance through mitochondrial uncoupling combined with PDH activation and consequently preferentially promotes glucose oxidation. To investigate the dependency of preferential glucose oxidation on PDH activation, we examined the effect of 6j on PDH in vitro. The decrease in phosphorylated Ser293-PDH by 6j in myotubes and hepatocytes disappeared upon PDH knockdown by the application of siRNA (Fig. 6A). As expected, glucose oxidation induced by 6j declined with deficient PDH activation in myotubes, but 6j still had a slight effect on promoting glucose oxidation because of its uncoupling action (Fig. 6B). Meanwhile, fatty acid oxidation and lactate production were dramatically enhanced by 6j under deficient PDH activation in myotubes and hepatocytes (Fig. 6C–F). These results indicate that the effects of 6j on preferential glucose oxidation depend on PDH activation.

The promotion of mitochondrial oxidative metabolism by uncoupling actions represents a conceptual approach for the treatment of diabetes, nonalcoholic fatty liver disease, or other metabolic diseases. In view of this, a pharmacological uncoupling agent, DNP, was applied clinically in the 1930s. However, DNP was taken off the market because of its narrow therapeutic index, evidenced by the emergence of severe adverse effects, including hyperlactacidemia, hyperthermia, and even death (11–15). Thus, it is a challenge to discover safe mitochondrial uncouplers for practical use. In our study, we demonstrated that PDH activation combined with mitochondrial uncoupling might be an alternative safe and effective approach for treating T2D.

It is critical and challenging to separate adverse effects from uncoupling activity. Several studies focused on the development of new classes of chemical mitochondrial uncouplers to safely treat metabolic diseases. On the basis of the excellent effects and large therapeutic indexes of liver-specific mitochondrial uncouplers, including liver-targeted derivatives of DNP, controlled-release mitochondrial protonophore, and niclosamide ethanolamine, on hyperglycemia and hyperlipidemia (40–43), it might be possible that tissue-specific chemical mitochondrial uncouplers have potential safety and efficacy. In our study, we focused on exploring a novel strategy of developing a new class of mitochondrial uncouplers. Previous research showed that the adverse effects of DNP mainly resulted from uncoupling-induced excessive lipid oxidation and anaerobic glycolysis. However, we found that inhibition of lipid oxidation alone by etomoxir or lactate production alone by galloflavin did not affect lactate production or DNP-derived lipid oxidation, suggesting that DNP-induced adverse effects could not be relieved by combination with etomoxir or galloflavin. Thus, we explored whether other targets combined with DNP could differentiate its hypoglycemic actions from its adverse effects by switching anaerobic glycolysis to oxidation to decrease fat burning and lactate overproduction. DCA, as a fat-to-glucose switcher, was selected to prove our hypothesis. DNP combined with amiloride had an effect similar to its effects with DCA in myotubes, but hyperkalemia was observed with amiloride treatment in the clinic (28,29,44–46), similar to the adverse effects of DNP. However, linking other safe potassium-retaining diuretic drugs to DNP might provide another effect-enhancing and toxicity-reducing option for T2D therapy.

Heat production from fat combustion is far higher than that produced from glucose (47). Inhibition of lipolysis abolished the mobilization of lipids and cold-induced brown adipose tissue thermogenesis, suggesting that fat is a standby source of heat for adaption to environmental changes (48). Once excessive fat mobilization by uncoupling actions occurs, hyperthermia develops. Conversely, inhibition of fatty acid oxidation by PDH activation in a double-knockout PDK2/PDK4 mouse model resulted in hypothermia (33). In our study with mice treated with DNP and DCA together, normal rectal temperature was maintained, suggesting that the abnormal status of fatty acid oxidation was corrected. Furthermore, a normal rectal temperature might partially contribute to protecting mice from death. In addition, increased plasma LDH, which is released from tissue injured by the toxic effects of DNP, declined with DCA treatment, suggesting that decreased LDH might be an indicator of survival against DNP toxicity. However, neither normal rectal temperature nor LDH release solely accounts for the survival and its underlying mechanism against DNP-induced mortality, and the mechanism associated with combination treatment required further investigation.

The nutrient storage pathways evolved to maximize efficient energy utilization that exposed to chronic energy surplus causes obesity and obesity-derived insulin resistance. Several previous studies clarified that obesity-derived ectopic lipid accumulation or obesity-induced tissue inflammation was closely associated with the progression of insulin resistance. Samuel and Shulman (49) proved that DAG directly activated PKC-θ in skeletal muscle and PKC-ε in liver, which decreased downstream insulin signaling. Meanwhile, it was demonstrated that an increase in glucose utilization and a decrease in fatty acid oxidation by PDH activation led to increased intramuscular DAG and an impaired insulin pathway (34). In our study, the lowered blood glucose and DAG accumulation–induced exaggerative insulin resistance were observed by DCA treatment. Interestingly, these phenomena were corrected by combination treatment. These results might be explained by the fact that the DNP contributed to relieve the inhibition of fatty acid oxidation by PDH activation. According to histological analyses of mouse livers under combination treatment, hepatic steatosis was markedly alleviated compared with that in the DCA group, but it was still severe compared with that in the DNP group. However, sensitivity to insulin was similar between the DNP and combination treatment group. Apart from the DAG/PKC pathway, hepatic lipid overload might increase the expression of proinflammatory cytokines, such as tumor necrosis factor-α (TNF-α) and interleukin (IL)–6, or the accumulation of specific lipid metabolites, such as ceramides, which decrease insulin signaling (6,50). Thus, it is possible that the decreased expression of TNF-α and IL-6 or ceramide accumulation in liver in the combination treatment group partially contributed to insulin sensitivity, which still requires further investigation.

On the basis of the metabolic benefits of DNP combined with DCA, the compound 6j was identified and confirmed to possess mitochondrial uncoupling and PDH activation in vitro and in vivo. Chronic administration of 6j significantly improved glycemic control in db/db mice. Correlated with the reduction of DAG and TG accumulation in liver and skeletal muscle, insulin resistance was restored by chronic 6j treatment. Apart from DAG accumulation, the decrease in the expression of proinflammatory cytokines was observed (Supplementary Fig. 8), which also might have contributed to the improvement of insulin sensitivity. In this study, 6j was a classical chemical mitochondrial uncoupler, promoting the proton across mitochondrial inner membrane through non-ATP synthase. However, the way that 6j affected the proton leakage through mitochondrial transporters such as uncoupling proteins, adenine nucleotide translocators, or the other proton transporters still needs to be further explored. Unlike classical mitochondrial uncouplers, the effects of 6j on PDH activation altered the uncoupling actions and stimulated oxidative substrate flux from fatty acids to glucose, which might have partially contributed to the reduction in adverse effects. Also, PDH activity increased by 6j still requires further investigation.

In summary, our research demonstrated that PDH activation combined with mitochondrial uncoupling promoted hypoglycemic effects and partially reduced adverse effects through preferential glucose oxidative flux. A novel compound, 6j, was identified and exhibited hypoglycemic effects in vivo. Thus, our research provides the possibility of further exploring a new class of mitochondrial uncouplers for treating T2D.

Funding. This work was supported by National Natural Science Foundation of China grants 81125023 and 81673493, Strategic Priority Research Program of Chinese Academy of Sciences grant XDA12040301, Shanghai Science and Technology Commission grant 16JC1405000, and K.C. Wong Education Foundation.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.J. contributed to the design and conducted the experiments, collected the research data, provided analyses and discussion, and wrote the manuscript. J.J. and H.L. contributed to the synthesis compounds. Y.D., Z.X., Y.L., X.Z., C.P., C.X., and T.D. assisted with the animal experiments. Y.D. and A.G. assisted with the cellular experiments. M.G. contributed to the measurement of adenine nucleotide levels. L.Y., J.T., and F.Y. contributed to design, analysis, and manuscript drafting for the chemistry. Jin. Li and Jia Li contributed to the research design, analysis, discussion, and drafting of the manuscript. F.Y., Jin. Li, and Jia Li are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 79th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 7–11 June 2019.