**4. Discussion**

Previous studies showed that severe hypoglycemia can cause seizure, unconsciousness, and neuronal death as extreme end-points [8,38,39]. The mechanisms of hypoglycemia-induced neuronal cell death are still unclear. Our previous study demonstrated that neuronal cell death induced by hypoglycemia is caused not only by the low glucose level itself but also by glucose reperfusion [37]. Hypoglycemia causes excessive vesicular zinc release, which leads to NADPH oxidase activation and causes neurotoxicity [4,9]. Zinc influx into neurons after hypoglycemia can lead to mitochondrial dysfunction, resulting in reduced ATP levels in the brain [61]. Therefore, we investigated whether treatment with DCA, an activator of ATP formation, has neuroprotective e ffects in the hippocampus after severe hypoglycemia.

The ATP required for neurological function in the brain is predominantly generated by glucose oxidation (GO) and pyruvate oxidation (PO) in mitochondria [62] and it is closely related to hypoglycemic state. Under physiological conditions, formation of pyruvate increase rates of glycolysis and enhances glucose oxidation by way of activation of PDH, which converts pyruvate into Acetyl-CoA. However, under pathophysiological conditions such as hypoglycemia (HG) and ischemia-reperfusion (IR), it decreases the rates of glucose oxidation due to mitochondrial dysfunction [62,63]. Depressed glucose oxidation can lead to neuronal cell death [49,64].

Dysfunction of mitochondrial metabolism is central to the pathological results following brain injury, such as traumatic brain injury or ischemia [65,66]. Previous studies reported that PDH is important in altered brain energy metabolism in diverse brain injuries [67–69]. Traumatic brain injury is reported to enhance the expression of PDK and phosphorylate PDH, leading to inhibition of PDH-regulated glucose metabolism [45]. Bowker-Kinley et al. mentioned that when they looked at levels of PDK mRNA, PDK1 (heart), PDK3 (testis), PDK4 (skeletal muscle and heart) was found in certain tissues, whereas PDK2 was found in all tissues. Also, they indicated that in the brain, for example, PDK activity corresponds primarily to the isoenzyme PDK2. In addition, they showed that DCA, a structural analog of pyruvate, attaches to the pyruvate-binding site, resulting in inhibition of PDKs with the order of inhibition being PDK2 > PDK1~PDK4 >> PDK3 [23]. In the present study, we confirmed that hypoglycemic insult increased the level of PDK2 in the hippocampus and thus reduced PDH, causing inhibition of glucose metabolism. However, administration of DCA decreased PDK2 levels in the hippocampus.

After several brain insults, mitochondrial damage induces excessive PDK activation, which restricts ATP formation and results in more severe neuronal death [26,45]. In the case of severe hypoglycemic animal models, neuronal death occurs, and given the result, mitochondrial dysfunction occurs in both the hypoglycemic state itself and also during the hyperglycemic state when glucose is reperfused [51,70–72]. DCA is known to be neurotoxic when high doses, above 500 mg/kg, are used, or if 300 mg/kg is administered for more than 10 weeks [73,74]. However, many previous animal studies have confirmed the beneficial e ffects of administering DCA at dose of 50 to 200 mg/kg [75–79], and our previous study also confirmed that treatment with a dose of 100 mg/kg after ischemia reduced neuronal death [26]. According to the paper by Stacpoole, it was noted that DCA was rapidly absorbed following treatment, crosses the blood-brain barrier and activates PDH within a few minutes [80].

In the present study, we found that DCA, a PDK inhibitor, significantly reduces the number of degenerating neurons after hypoglycemia. FJB staining demonstrated a decline in the number of degenerating neurons in the hippocampus in the DCA-treated group compared to the vehicle-treated group after hypoglycemia. This trend correlated with the NeuN staining, which indicated an increase of surviving neurons in the DCA group. This result suggests that DCA increases PDH levels by inhibiting PDK2, which may result in increased ATP synthesis, reducing neuronal cell death in the subiculum, CA1 and dentate gyrus of hippocampus compared with the vehicle-treated group, after hypoglycemia (Figure 6).

**Figure 6.** Possible association of PDK2, PDH, DCA, and neuronal death under hypoglycemic conditions. This schematic illustration shows the effects of DCA on the process of hypoglycemia-induced hippocampal neuronal death. (**A**) Experimental timeline. (**B**) Neuronal cell death mechanism caused by hypoglycemia: (1) After hypoglycemia, mitochondria disruption occurs and then PDK2 increase abnormally (2) An abnormally increased PDK2 inhibits the activity of PDH by phosphating PDH (3) The conversion from pyruvate to acetyl-CoA is unsuccessful due to the inhibition of PDH (4) The rate of synthesis of ATP through the TCA cycle significantly reduces. When these conditions dominate, neuronal death is more likely to occur. (**C**) Effects of DCA on hypoglycemia-induced neuronal cell death: administration of DCA can inhibit PDK2 and thus prevent hippocampal neuronal death after hypoglycemia.

D'Alessandro et al. stated that mitochondrial impairment induced by the reduction of the glucose-derived pyruvate results in decreased glutathione synthesis [39,81]. Glutathione plays an important role in the cells, which reduces the oxidative stress by acting as an antioxidant. This synthesis of glutathione requires glycine, glutamate and cysteine, and consumes ATP [82]. Mitochondrial dysfunction after hypoglycemic insult contributes to oxidative stress by activating NADPH oxidase, leading to ROS formation, which is heavily associated with neurodegenerative diseases [4,9,83]. Therefore, we conducted 4HNE staining to confirm whether DCA decreases oxidative stress in the hippocampus. We found that DCA administration successfully reduced oxidative injury in the hippocampus compared to the vehicle-treated group after hypoglycemia. Therefore, we assumed that the administration of DCA increases the level of ATP, and this has resulted in the reduction of the oxidative stress due to the e fficient supply of ATP for glutathione formation.

Neural or immune cells participating in neuroinflammation experience metabolic changes including a glycolytic metabolic shift. The altered glycolytic metabolism leads to promotion or inhibition of neuroinflammation [84,85]. PDKs have been reported to control functional polarization of macrophages [86]. Macrophage polarization toward the pro-inflammatory M1 phenotype is usually followed by a cellular metabolic change from oxidative phosphorylation to aerobic glycolysis, as well as nitric oxide (NO) production [86]. Likewise, NO produced by brain diseases suppresses enzyme activity in PDH [87]. Jha et al. suggested that PDK2/4 play vital roles in the inflammatory infiltration of immune cells, in the induction of the pro-inflammatory macrophages, and in inhibition of the anti-inflammatory phenotype of peripheral macrophages [88]. Microglia and astrocyte can be activated immediately during brain injury and induce pro-inflammatory molecules such as tumor necrosis factorα (TNFα) or NO [89,90]. Therefore, brain inflammation has been considered to be a potential target in treating brain diseases for several years, and various approaches have been implemented to suppress disease-induced brain inflammation [91–93]. The selective vulnerability of CA1 hippocampus is very well studied in rodents and CA1 is known to have the highest susceptibility to damaging conditions [94–96]. We performed CD11b and GFAP staining, which is an inflammatory marker, to confirm the neuroprotective e ffects of DCA on inflammation induced by microglia and astrocyte activation after hypoglycemia. Thus, we found that DCA reduced microglia and astrocyte activation by inhibiting PDK2 in the hippocampal CA1 region after hypoglycemia.

BBB disruption was assessed on the basis of IgG extravasation [43] and is related to an influx of neutrophils after brain insult [97]. Generally, IgG concentration and an infiltration of neutrophils in the brain is rare because it exists in the blood vessels when undamaged. However, under pathological conditions, IgG leakage and neutrophils influx occur as the BBB is disrupted by brain injuries such as seizure, ischemia, or traumatic brain injury [55,98–100]. Doll et al. reported that mitochondria play an important role in the opening of the BBB [101]. If mitochondria are destroyed by lack of oxygen or glucose, ATP production decreases in the endothelial cells surrounding the blood vessels in the brain, resulting in the disruption of the BBB, which exacerbates brain injury [102]. Mitochondrial dysfunction also causes the collapse of the BBB because it further reduces the rate of ATP synthesis [101,103]. Therefore, we conducted IgG and MPO staining to confirm whether DCA prevents the BBB breakdown by increasing the formation of ATP. Here, we found that DCA decreased IgG leakage in the whole brain and the number of MPO (+) cells in the hippocampus after hypoglycemia. This result demonstrates that DCA has the capability to prevent BBB disruption by enhancing ATP production in brain endothelial cells after hypoglycemia.

In conclusion, we found that DCA treatment can alleviate hippocampal neuronal death by inhibiting PDK activity following hypoglycemia. It is unclear exactly how PDK2 is increased by hypoglycemia/glucose reperfusion. We speculate that a complex signaling pathway may be generated by the hypoglycemic insult inside mitochondria. Therefore, our findings sugges<sup>t</sup> that DCA may be a vital therapeutic tool for preventing neuronal death induced by hypoglycemia.

**Author Contributions:** A.R.K. researched data and reviewed and edited the manuscript. B.Y.C., K.-H.P. and J.B.P. reviewed and edited the manuscript. S.H.L., D.K.H., J.H.J., B.S.K. and D.H.K. researched the data. S.W.S. contributed to the discussion and wrote/reviewed and edited the manuscript. S.W.S. is the person who takes full responsibility for the manuscript and its originality. All authors read and approved the final manuscript.

**Funding:** This research was supported by the Brain Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Science, Information and Communication Technology and Future Planning (NRF-2017M3C7A1028937 and 2018R1A4A1020922) to Sang Won Suh.

**Conflicts of Interest:** The authors declare no conflict of interest.
