β-Lapachone Prevents Diet-Induced Obesity by Increasing Energy Expenditure and Stimulating the Browning of White Adipose Tissue via Downregulation of miR-382 Expression

Obesity is a global epidemic that results from energy imbalance, which is characterized as chronic caloric intake in excess of energy expenditure; this imbalance is a common occurrence in the Western lifestyle (1). Excess energy is stored as triglycerides in white adipose tissue (WAT) and other organs. Recently, obesity has become a significant contributor to many chronic diseases, including type 2 diabetes, cardiovascular disease, and cancer (2). Therefore, new approaches to reduce adiposity are needed. Exercise is an effective intervention for the prevention and treatment of obesity (3) because it increases the expression of peroxisome proliferator-activated receptor γ coactivator 1 α (PGC1α), which then increases mitochondrial biogenesis and fatty acid β-oxidation (4). Concomitantly, enhanced PGC1α induces uncoupling protein 1 (UCP1) in adipose tissue, which results in an increase in energy expenditure and prevents obesity.

Recently, there has been great interest in brown adipose tissue (BAT), which is specialized for dissipation of chemical energy in the form of heat. The dissipation of energy depends on high mitochondrial content and UCP1 expression (2). UCP1 is highly expressed in the mitochondria of BAT and catalyzes a proton leak across the inner mitochondrial membrane, thereby uncoupling respiration from ATP synthase. In BAT, there are high numbers of mitochondria and high expression levels of UCP1, which enhance heat production and thermogenesis to lead to antiobesity effects (5,6). In contrast, WAT is mainly composed of subcutaneous and visceral fat. Although metabolic disease is caused by the excessive accumulation of visceral fat in the abdomen, subcutaneous WAT (scWAT) does not cause the same metabolic risks (7). Studies have suggested that cells expressing UCP1 and other brown adipocyte–specific genes in response to various stimuli, including chronic cold exposure and β-adrenergic agonists, have been found in WAT, especially in the scWAT of adult rodents and humans. These cells are referred to as adaptive brown fat cells, brown-in-white (brite) cells, or beige cells (3,7,8). Although beige adipocytes originate from WAT, their morphological and molecular phenotypes are more similar to BAT, with multilocular lipid droplets and high UCP1 expression, mitochondrial biogenesis, and respiration rates (6). Because beige adipocytes that are found in the scWAT and brown adipocytes have similar functional characteristics, many studies are focused on the browning of WAT to beige adipocytes in the therapeutic targeting of beige cells for obesity treatment (9).

MicroRNAs (miRNA) are small, noncoding RNA molecules that regulate target gene expression to repress the translational process and RNA degradation (10) and to mediate many cellular processes such as energy homeostasis (11,12). Recent studies have demonstrated that miRNAs, such as miR-27, -133, -155, -193b, -196, and -203, are deeply associated with brown adipocyte and beige adipocyte differentiation (8,13–15). However, the roles of only a few miRNAs in fat browning have been verified.

β-Lapachone (BLC) is a naphthoquinone that was originally isolated from a Bignoniaceae tree Tabebuia avellanedae Lorentz ex Griseb found in South America’s rainforest. Many studies have reported that NAD⁺ and NADH regulate energy metabolism and that the increased intracellular NAD⁺-to-NADH ratio activates sirtuin 1 (Sirt1) (16–18). Previous studies suggested that BLC facilitates the NADH:quinone oxidoreductase 1–dependent oxidation of NADH to NAD⁺. Moreover, stimulated NADH oxidation by BLC ameliorates obesity by increasing mitochondrial biogenesis (16). We therefore hypothesized that the antiobesity effect of BLC may be associated with the upregulation of energy expenditure and the stimulation of fat browning. In this study, we demonstrated the positive role of BLC in fat browning and thermogenic capacity in vivo. We also sought to elucidate the possible mechanism underlying this action of BLC in vitro.

All procedures were conducted in accordance with the Guidelines for Institutional Animal Care and Use Committee of the Korea Food Research Institute (KFRI-IACUC, KFRI-M-13021). Male C57BL/6 mice (4 weeks old) were maintained at a temperature of 21–25°C and humidity of 50–60% and kept on a 12-h light/12-h dark cycle with free access to food and water. After 1 week of adaptation, mice were divided into three groups and fed a high-fat diet (HFD; 25% w/w, 45% calories as fat). The HFD group (vehicle control) received 0.5% carboxyl methyl cellulose-Na in distilled water via oral gavage. The low-dose BLC (LBLC) or high-dose (HBLC) group received each 20 mg or 40 mg/kg body weight/day BLC, respectively, which was dissolved in 0.5% carboxyl methyl cellulose-Na in distilled water for 11 weeks. At the end of experiments, adipose tissue quantification was performed using the eXplore CT-120 microcomputed tomography (micro-CT) system (GE Healthcare, Waukesha, WI). Serum insulin and leptin levels were analyzed by ELISA (ALPCO Diagnostics, Salem, NH). A commercially available kit was used for the assay of citrate synthase activity (Sigma-Aldrich, St. Louis, MO) and mitochondria complex II enzyme activity (Abcam, Cambridge, MA) in the tissues.

For glucose tolerance testing (OGTT), mice were fasted for 12 h and given d-glucose (2 g/kg body weight) orally. For insulin intraperitoneal tolerance testing (IPITT), mice were fasted for 4 h and injected with recombinant human insulin (0.75 units/kg body weight).

Each mouse was allocated into a calorimetry chamber (Oxymax OPTO-M3 system; Columbus Instrument, Columbus, OH) to measure VO₂ and energy expenditure during 12 consecutive hours (dark time from 7:00 p.m. to 7:00 a.m.). Mice were monitored at 22°C or 30°C to compare energy metabolism at standard room temperature or under thermoneutral conditions, respectively.

Gastrocnemius muscle sections were fixed in 1% paraformaldehyde with 4% glutaraldehyde solution, and changes in mitochondrial morphology were observed under a LEO 912AB EF-transmission electron microscopy (Carl Zeiss, München-Hallbergmoos, Germany) at the Korean Basic Science Institute in Chuncheon.

Total RNA was extracted using NucleoSpin RNA II (Macherey-Nagel, Düren, Germany), and cDNA was generated with the iScript cDNA Synthesis Kit (BioRad, Hercules, CA). Quantitative (q)RT-PCR was performed with the SYBR Green Master Mix (Toyobo, Tokyo, Japan) and specific primers (Supplementary Table 1) in a StepOnePlus PCR System (Applied Biosystems, Foster City, CA).

Protein was extracted with radioimmunoprecipitation assay buffer, and Western blotting was performed with primary antibodies (Supplementary Table 2).

Stromal vascular fractions (SVFs) from the inguinal adipose tissue of male C57BL/6 mice were isolated using a previously described method (19). Collected SVFs were cultured and differentiated into mature white adipocytes or beige adipocytes with BLC according to previously described methods (20,21).

SVFs were differentiated into mature white adipocytes in the presence and absence of BLC. Oxygen consumption rate (OCR) was measured by the XF24 extracellular Flux Analyzer (Seahorse Bioscience, Billerica, MA). Basal OCR was measured in the medium containing 25 mmol/L glucose and 1 mmol/L pyruvate. Next, 1 μmol/L oligomycin A was injected to inhibit ATP synthase and measure ATP turnover, and 1 μmol/L carbonyl cyanide-p-trifluoromethoxy phenylhydrazone was injected to dissipate the proton gradient across the mitochondrial membrane. Finally, 1 μmol/L rotenone was injected to block mitochondrial respiration.

TaqMan miRNA low-density arrays (TLDAs; TaqMan Rodent MicroRNA A Array v2.0, Applied Biosystems) were used for miRNA profiling. TLDAs were performed on the ViiA7-Real-Time PCR System (Applied Biosystems), and quantitative miRNA expression data were acquired and analyzed using the ABI 7900HT SDS software (Applied Biosystems).

To specifically induce miR-382 expression, miRIDIAN miR-382 mimics (Thermo Scientific, Lafayette, CO) were used. Oligonucleotide sequences for negative control mimics were based on Caenorhabditis elegans miR-67. Transfections with 50 nmol/L miR-382 mimics or 20 nmol/L Dio2 small interfering RNA (siRNA; Thermo Scientific) were performed for 2 days with Lipofectamine RNAiMAX before the induction of differentiation.

The pMIR-REPORTER System (Ambion, Austin, TX) was used to identify the Dio2 3′ untranslated region (UTR) as an miR-382 binding site. The mouse Dio2 3′UTR fragment was amplified by RT-PCR and cloned into the pMIR-REPORTER luciferase reporter vector. A point mutation was generated using a site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). Lipofectamine 2000 (Invitrogen) was used to contransfect HEK-293T with 100 ng of reporters and 50 nmol/L miR-382 mimic. At 24 h after transfection, luciferase activity was measured using the Dual-Light System (Applied Biosystems).

Data are expressed as mean ± SEM. Statistical analyses were performed with the GraphPad Prism 6.0 software (GraphPad Software, Inc., La Jolla, CA). One‐way ANOVA was used to compare more than two groups, followed by the Bonferroni post hoc test to detect difference between-groups analysis. The Student t test was used to compare the differences between two groups. Comparison with a P value <0.05 was considered significant.

After 11 weeks of the experimental diet, the final mean body weight of HBLC-fed mice was significantly lower than that of HFD- or LBLC-fed mice (HFD, 36.3 ± 1.1 g; LBLC, 30.3 ± 0.9 g; HBLC, 27.2 ± 0.5 g) (Fig. 1A). The average weekly food intake was not significantly different among the groups (Fig. 1B), suggesting that the reduction in body weight gain in LBLC and HBLC groups was not due to decreases in food consumption. To determine whether the weight-reducing effect of BLC was due to decreases in adipose tissue, the individual adipose tissue depot was scanned and quantified by micro-CT scanning, which was then used to generate a series of cross-sectional X-ray images. Consistent with the decrease in body weight, micro-CT images showed dramatic decreases in fat accumulation with BLC treatment (n = 3; fat mass/total body mass, HFD = 45.8 ± 1.6%; LBLC = 30.3 ± 2.7%; HBLC = 22.2 ± 1.9%) (Fig. 1C). Moreover, the weight of WAT in BLC-treated mice was significantly decreased, and BLC supplementation significantly decreased the size of the adipocytes (Fig. 1D). BLC also decreased lipid accumulation and increased the expression of thermogenic genes in the liver (Supplementary Fig. 1). To investigate whether BLC elicits a beneficial metabolic effect, we conducted the OGTT and IPITT. Compared with HFD mice, BLC-treated mice had lower blood glucose concentrations after glucose or insulin administration, showing a significantly low area under the curve (Fig. 1E and F). In particular, the HBLC group showed insulin and leptin levels that were markedly reduced by 73.6% and 80.7%, respectively, compared with the HFD group (Table 1). In addition, BLC treatment did not cause any adverse side effects or growth (Supplementary Fig. 2).

To determine whether BLC reduces body weight gain and adiposity by modulating energy expenditure and to understand the mechanism of BLC-induced thermogenesis, we performed indirect calorimetry at room temperature (22°C) or at thermoneutrality (30°C). At 22°C, BLC-treated mice showed significantly higher Vo₂ and Vco₂ throughout the 12-h dark time compared with HFD mice (Fig. 2A and B). HBLC mice also showed significant increases in energy expenditure (Fig. 2C). Although energy expenditure was much lower at 30°C than at 22°C, BLC treatment also elicited a significant increase in the metabolic rate at 30°C (Fig. 2D–F).

We analyzed the expression of brown adipocyte–specific genes and proteins in the WAT of HFD-induced mice. A significant increase in brown adipocyte–specific genes, including UCP1, PRDM16, PGC1α, Cidea, and Elovl3, was observed in the LBLC and HBLC groups (Fig. 3A). In particular, the expression of UCP1 and Elovl3 mRNA in the scWAT of the HBLC group was 62.5- and 24.7-fold higher than that of the HFD group. This increase in expression was also significantly higher in the epididymal WAT (epWAT) of the BLC groups than that of the HFD group (Fig. 3C). However, the expression of these genes was much higher in the scWAT than in the epWAT. These results are consistent with the findings of Wu et al. (22). In addition, brown adipocyte–specific proteins were increased with BLC administration (Fig. 3B and D). These increases in the scWAT of HBLC mice were accompanied by profound morphological changes toward a BAT-like phenotype, which was characterized by numerous UCP1-expressing clusters. However, UCP1-expressing adipocytes were not detected in the scWAT of HFD mice. Notably, the scWAT of HBLC mice appeared browner than that of HFD mice (Fig. 3E). In addition, the size and number of lipid droplets were clearly reduced in the interscapular BAT of the HBLC group (Fig. 3F), and key thermogenic factors were stimulated (Fig. 3G and H).

We further evaluated the effect of BLC on mitochondrial function. The morphometric analysis of mitochondria in the gastrocnemius of HBLC mice showed the aggregation of larger mitochondria between adjacent myofibrils and well-organized cristae (Fig. 4A). HBLC mice also exhibited significantly increased complex II subunit activity and citrate synthase activity in the muscle, resulting in oxidative phosphorylation and higher expression levels of genes related to mitochondrial function (Fig. 4B, C, and E). Furthermore, BLC induced mitochondrial expansion, as evidenced by increases in mitochondrial DNA (mtDNA) content in HBLC mice (Fig. 4D). Previous studies have shown that activated mitochondrial function can be induced by cold stress or β-adrenergic stimuli as well as by the expression of brown adipocyte–specific genes (23–25). Exposure of mice to cold temperatures in this study resulted in an ∼15-fold increase in UCP1 mRNA levels and a significant increase in UCP1 protein levels in the scWAT of mice compared with mice at room temperature (Supplementary Fig. 4A and B). Genes related to mitochondrial function and UCPs in the skeletal muscle were also significantly induced (Supplementary Fig. 4C and D). Overall, we show that cold exposure induced fat browning and activated mitochondrial function; these results are consistent with previous studies (23–25). Therefore, we also suggest that BLC shows comparable effects to cold stimulation on fat browning.

BLC induced the expression of brown adipocyte–specific genes in the WAT of HFD-induced obese mice in vivo. To confirm whether BLC affects the brown response during adipocyte differentiation, we isolated SVFs and induced differentiation into mature white adipocytes in the presence and absence of BLC (Fig. 5A). BLC treatment had no effect on adipocyte differentiation (Fig. 5B). Treatment with BLC led to an upregulation of brown-specific genes (PRDM16, UCP1, PGC1α, and Cidea) (Fig. 5C and Supplementary Fig. 5), and proteins (UCP1 and PGC1α) (Fig. 5D). BLC also strongly induced the expression of genes related to mitochondrial function and increased mtDNA content (Fig. 5E and F). Furthermore, basal and maximal OCRs were higher in adipocytes from BLC-treated SVFs, although these increases were not statistically significant (Fig. 5G). Higher degrees of UCP1-stained adipocytes were observed in the presence of BLC compared with control adipocytes, although BLC treatment did not show any effect on lipid droplet formation (Fig. 5H).

TLDAs were performed to understand the mechanism by which BLC exerts its fat-browning effect, and miRNAs that exhibit altered expression levels in the inguinal WAT after BLC treatment were identified. Expression clusters of miRNAs that were related to brown fat differentiation or adipocyte browning were obtained by hierarchical clustering analysis (Fig. 6A). miR-382, -133a, -133b, -27b, and -155 were downregulated in the scWAT of BLC-fed mice and thus identified as negative regulators of fat browning (Fig. 6A). The qRT-PCR results showed that miR-382 expression was significantly downregulated in the scWAT of BLC-fed mice (Fig. 6B). We therefore hypothesized that miR-382 might be involved as a negative regulator in the expression of brown fat–specific genes. Next, we investigated the potential target genes of miR-382 that may stimulate fat browning. Bioinformatics analyses that were generated using the miRWalk, TargetScan, and miRBase databases revealed that Dio2 was a putative target gene of miR-382, and the miRNA 5′-seed sequence shared homology with Dio2 (Fig. 6C). Dio2 activates the conversion of thyroxin (T4) to 3,3′,5-triiodothyronine (T3), which accelerates the adrenergic response and leads to energy expenditure (26–28). We observed a significant increase in Dio2 mRNA in the scWAT of BLC-fed mice (Fig. 6D). To confirm whether BLC also downregulates the expression of miR-382 in vitro, we induced the differentiation of mature white adipocytes from SVFs treated with BLC. BLC significantly downregulated miR-382 and upregulated Dio2 mRNA expression in white adipocytes (Fig. 6E and F).

To investigate the effect of miR-382 on Dio2 expression, we activated miR-382 expression by transfecting SVFs with mimic–miR-382 (50 nmol/L) for 2 days. Transfection with mimic–miR-382 dramatically increased miR-382 expression (Fig. 6G) and downregulated Dio2 mRNA expression (Fig. 6H). We performed luciferase reporter assays to determine whether miR-382 specifically targets Dio2 by directly binding to its 3′UTR in HEK-293T cells (Fig. 6I). The introduction of mimic–miR-382 significantly repressed the Dio2 3′UTR construct compared with that observed in control construct-transfected cells. Repressed luciferase binding activity was abrogated when the seed sequences in the 3′UTR were mutated (uAACcc). Furthermore, we investigated whether Dio2 is responsible for mediating the browning effect of BLC. Transfection with Dio2 siRNA successfully inhibited the expression of brown-specific UCP1 and Cidea. Moreover, the browning effect of BLC was blunted by inhibition of Dio2 (Fig. 6J and K). These findings suggest that BLC increases the browning effect by mediating Dio2.

To investigate the effect of miR-382 on beige adipocyte differentiation, we transfected SVF cells with mimic–miR-382 or mimic-control for 48 h. The brown adipocyte differentiation cocktail was used to differentiate transfected cells into beige adipocytes. During the 8 days of differentiation, the cells were treated with or without 1 μmol/L BLC. In accordance with the marked activation of miR-382 by mimic–miR-382 (Fig. 7A), adipocytes transfected by mimic–miR-382 exhibited significant downregulation of Dio2, UCP1, and peroxisome proliferator-activated receptor γ (PPARγ) mRNA and protein levels (Fig. 7B–E). These data strongly suggest that miR-382 is a negative regulator of beige adipocyte differentiation. However, treatment with 1 μmol/L BLC recovered the miR-382–induced depression of beige adipocyte differentiation, and the expression of Dio2, UCP1, and PPARγ was upregulated significantly (Fig. 7B–E). Furthermore, treatment with 1 μmol/L BLC significantly downregulated the expression of miR-382 (Fig. 7A). These data suggest that BLC stimulates fat browning by modulating the expression of miR-382, which is a negative regulator of beige adipocyte differentiation.

Recent studies have suggested that the enhancement of browning of white fat and BAT function can stimulate thermogenesis and energy expenditure, thereby decreasing adiposity (29). In addition, the risks of obesity and metabolic syndrome in humans are reduced by the expansion of brown-specific adipocytes in WAT (30,31). Thus, fat browning and BAT development are a target of antiobesity therapeutics, and many efforts have been conducted to identify the natural compound that can stimulate the browning process and BAT development. Furthermore, identifying miRNAs that can regulate the browning of adipocytes has become popular in the antiobesity field.

Our study demonstrated BLC as a likely inducer of white adipocytes that have unique features of brown adipocytes, which are termed beige (22) or brite (32) adipocytes. Our findings also reveal a critical role of BLC in stimulating fat browning in the WAT and thermogenesis in BAT in vivo. Furthermore, the fat-browning effect of BLC was partially caused by the regulation of miR-382, which targets Dio2.

Importantly, we provided compelling evidence to demonstrate the antiobesity and fat-browning effects of BLC. BLC significantly decreased body weight gain, which may be due to decreases in body fat accumulation. In addition, BLC increased the expression of brown adipocyte–specific genes and induced morphological changes toward a BAT-like phenotype in the scWAT. Numerous clusters of UCP1-expressing cells, which are typical features of BAT (33,34), were observed. These findings strongly suggest that BLC has a fat-browning effect. Thermogenesis predominantly involves classical BAT, which expresses much higher levels of UCP1 than the scWAT (5). In this study, we also found that UCP1 mRNA expression was much higher in BAT than in scWAT (Supplementary Fig. 3). However, BLC induced a robust increase in UCP1 expression in scWAT (a more than 60-fold increase in scWAT vs. a 2.3-fold increase in BAT). In addition, the BLC-mediated induction of thermogenic capacity was also observed in BAT. Although the tissue source of total main thermogenic capacity remains unclear, our results suggest that BLC synergistically induces increases in energy expenditure in scWAT and BAT.

Furthermore, enhanced mitochondrial function is closely related to brown fat function (35). Mitochondrial function can affect whole-body metabolism, and this is most evident in the muscle (18). In the current study, BLC treatment induced morphological changes in the mitochondria, as analyzed by transmission electron microscopy in the gastrocnemius muscle. The combination of increased mitochondrial size and well-organized cristae in the skeletal muscle, as well as increases in mtDNA contents, citrate synthase activities, and oxidative capacity in BAT and scWAT, indicate that BLC has a widespread effect on mitochondrial function. BLC-induced increases in mtDNA contents in SVF-derived adipocytes in vitro also suggest that BLC has an effect on mitochondrial function.

BLC has been shown to activate Sirt1 and AMPK1, thereby ameliorating obesity (16). Indeed, BLC increased the expression of phosphorylated AMPK and UCP1 in this study. However, this increase by BLC was inhibited by AMPK siRNA in SVF-derived adipocytes (Supplementary Fig. 6). Lee et al. (36) also reported that BLC increases metabolic rates partly by increasing UCPs. These results suggest that the BLC-mediated increase in mitochondrial function and fat browning might contribute considerably to favorable metabolic profiles against obesity.

Moreover, BLC activated the thermogenic capacity because an increase in energy expenditure was observed at both room temperature (22°C) and thermoneutral conditions (30°C). This raises the possibility that significant increases in energy expenditure by BLC at 30°C might be caused by a Dio2-mediated elevation. These results help us to further understand the mechanisms of BLC on thermogenic activity.

Through the targeted study of miRNAs, we identified miR-382 as a negative regulator of fat browning, and Dio2 was identified as the direct target of miR-382. Typically, Dio2 converts T4 to its active metabolite, T3 (37). The Dio2-mediated increase in T3 levels can enhance the cyclic AMP (cAMP) response. Increases in T3 levels and cAMP responses can also stimulate Dio2 and UCP1 mRNA expression, leading to thermogenic and fat browning function (26,38,39). Although whether Dio2 directly upregulates UCP1 expression is not clear, the absence of Dio2 in BAT generates less cAMP, reduces the metabolic rate, and lowers the expression of UCP1 and PGC1α in response to adrenergic stimulation (26,40). In our study, significant decreases in the expression of UCP1, Cidea, and Dio2 mRNA, as well as UCP1 protein, were observed during the differentiation process in beige adipocytes under T3-deleted conditions. However, treatment with BLC recovered the expression of these brown adipocyte–specific genes, including Dio2, in SVF-differentiated adipocytes (Supplementary Fig. 7A–D). Because the actions of Dio2 and T3 appeared to be important for the browning effect of BLC, we investigated T3 levels in the serum and tissues. Although T3 levels in the serum were not affected by BLC treatment, T3 was increased significantly in the scWAT and BAT (Supplementary Fig. 7E). This may be due to the high tissue specificity of Dio2 (41). We also confirmed that the browning effect of BLC was blunted by the thyroid hormone receptor antagonist, 1-850 (Supplementary Fig. 7F). On the basis of the importance of T3 for UCP1 gene expression, we might suggest that T3-mediated transcriptional regulation is involved in the browning effect of BLC.

In addition, BLC downregulated the expression of miR-382, which is a negative regulator of fat browning, and upregulated the expression of Dio2. Therefore, BLC may enhance fat browning and energy expenditure, at least in part, by controlling the targeting of Dio2 by miR-382. Phosphatase and tensin homolog (PTEN) has also been reported to be a functional target gene of miR-382 (42); PTEN increases brown adipose function and energy expenditure (43). We observed increases in PTEN protein levels in the adipose tissue of BLC-supplemented mice (Fig. 2). Therefore, further studies investigating possible mechanisms involving miR-382 and PTEN are warranted.

Overall, our results show that BLC stimulated browning and thermogenesis in WAT and BAT, respectively. These led to increases in energy expenditure. The effect of BLC on the browning of WAT was partially mediated via the regulation of miR-382, which targeted Dio2 (Fig. 7F). These findings suggest novel roles for BLC in the prevention of obesity; moreover, these results extend our focus on dietary therapeutic compounds and fat browning and suggest a new approach for the treatment of obesity and metabolic disorders by beige adipocyte induction.

Funding. This work was supported by Korea Food Research Institute grant E0143073645.

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

Author Contributions. W.H.C. and T.Y.H. designed research and wrote the manuscript. W.H.C. performed the experimental work and acquired and analyzed the data. J.A., C.H.J., and Y.J.J. contributed to discussion and analyzed the data. All authors reviewed and approved the manuscript. T.Y.H. 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.