Emodin, as a mitochondrial uncoupler, induces strong decreases in adenosine triphosphate (ATP) levels and proliferation of B16F10 cells, owing to their poor glycolytic reserve

Many types of cancer cells show a characteristic increase in glycolysis, which is called the “Warburg effect.” By screening plant extracts, we identified one that decreases cellular adenosine triphosphate (ATP) levels and suppresses proliferation of malignant melanoma B16F10 cells, but not of noncancerous MEF cells. We showed that its active ingredient is emodin, which showed strong antiproliferative effects on B16F10 cells both in vitro and in vivo. Moreover, we also found that emodin can function as a mitochondrial uncoupler. Consistently, three known mitochondrial uncouplers also displayed potent antiproliferative effects and preferential cellular ATP reduction in B16F10 cells, but not in MEF cells. These uncouplers provoked comparable mitochondrial uncoupling in both cell types, but they manifested dramatically different cellular effects. Namely in MEF cells, these uncouplers induced three to fivefold increases in glycolysis from the basal state, and this compensatory activation appeared to be responsible for the maintenance of cellular ATP levels. In contrast, B16F10 cells treated with the uncouplers showed less than a twofold enhancement of glycolysis, which was not sufficient to compensate for the decrease of ATP production. Together, these results raise the possibility that uncouplers could be effective therapeutic agents specifically for cancer cells with prominent “Warburg effect.”


| INTRODUCTION
Even with the impressive advances in medical care over the last several decades, about one out of two Japanese suffers from cancer over their lifetime, and one third of Japanese die of cancer. One of the most prominent characteristics of cancer cells, regardless of type or stage, is their uncontrolled, aggressive proliferation. Such aggressive proliferation requires a vast amount of cellular energy in the form of adenosine triphosphate (ATP), as compared with noncancerous cells, as cell proliferation entails many energy-consuming steps, for example, synthesis of proteins and RNAs (rRNA, mRNA and tRNA), DNA replication and cytokinesis (Buttgereit & Brand, 1995;Gibbons & Rowe, 1965;Summers & Gibbons, 1971). As a consequence, cancer cells should be sensitive to ATP depletion by various means.
It is well known that many types of cancer cells show a characteristic increase in glycolysis, even in aerobic conditions, a phenomenon is called "Warburg effect" (Warburg, 1956;Warburg, Posener, & Negelein, 1924). Thus, glycolysis has garnered considerable interest as a possible therapeutic target. From this point of view, 2-deoxy-D-glucose (2-DG), a nonmetabolizable glucose analogue that inhibits glycolytic hexokinase, had been assessed as an anticancer drug in clinical trials. Disappointingly, however, patients treated with 2-DG showed severe side effects, so this drug is now under evaluation in combined therapies with other chemotherapeutic compounds (Dwarakanath & Jain, 2009;Mohanti et al., 1996;Singh et al., 2005;Vander Heiden, 2011).
Recent epidemiologic studies have shown that diabetic patients treated with metformin manifest lower incidence of various types of cancers, such as breast cancer, colorectal cancer, pancreatic cancer and lung cancer (Bodmer, Meier, Krähenbühl, Jick, & Meier, 2010;Evans, Donnelly, Emslie-Smith, Alessi, & Morris, 2005;Vernieri et al., 2016;Yin, Zhou, Gorak, & Quddus, 2013;Zhang et al., 2014). In 2000, it was reported that a target of metformin in cancer cells was mitochondrial respiratory chain complex Ⅰ (El-Mir et al., 2000;Owen, Doran, & Halestrap, 2000), and this drug is now being tested in clinical trials for different types of cancers (Vernieri et al., 2016). In preclinical research, there are two types of compounds that are expected to be efficacious as new anticancer drugs, which also target the mitochondrial energy metabolism (Weinberg & Chandel, 2015). The first one, tigecycline, is an inhibitor of translation of electron transport chain (ETC) proteins, (Skrtić et al., 2011), and the second one is Gamitrinib which inhibits heat shock protein-90 (HSP90) and tumor necrosis factor receptor-associated protein-1 (TRAP-1) ATPase. This inhibitor lead to a decrease of ATP production from oxidative phosphorylation (OXPHOS) (Chae et al., 2012).
Malignant melanoma has the dubious distinction of having one of the worst prognoses, because of its aggressive metastatic nature from early stages. Melanoma is responsible for over 75% of skin cancer death (Corrie, Hategan, Fife, & Parkinson, 2014), and its incidence is rapidly increasing (Siegel, Miller, & Jemal, 2018). The two-year overall survival rate of malignant melanoma patients at stage Ⅳ is only 10.7% (Sandru, Voinea, Panaitescu, & Blidaru, 2014). Recently, some molecularly targeted therapies have achieved extensive attention as new melanoma treatments, like vemurafenib and nivolumab (Schadendorf et al., 2018). However, the treatments by vemurafenib, a specific inhibitor of BRAF V600E, have allowed responsive tumors to be resistant to this chemotherapy (Flaherty et al., 2010). On 2014, a PD-1 blocking antibody drug, "nivolumab," was approved as a new drug, but the objective response rate to treatment of metastatic melanoma without a BRAF mutation was not satisfying, 40% (Robert et al., 2015). Thus, melanoma remains one of the most refractory neoplasms in spite of these energetic studies, and it is an urgent need to develop a more effective strategy for the treatment.
B16F10 cells, which were established from a mouse malignant melanoma (Fidler, 1973), serve as a model to examine the efficacies of experimental drugs in vitro and in vivo. In this study, we evaluated compounds that reduce ATP levels in B16F10 cells and found a plant extract from rhizomes of Polygonum cuspidatum (RPC). This plant extract has long been administrated to human body as a component of a traditional medicine, and various types of pharmacological effects are known like treating cough, hepatitis, jaundice, arthralgia and skin burns. Although recent studies reported that this extract has antiviral, antimicrobial, anti-inflammatory, neuroprotective and cardioprotective activities (Peng, Qin, Li, & Zhou, 2013;Zhang, Li, Kwok, Zhang, & Chan, 2013), there are no reports about the ATP reduction in cancer cells. Here, we show an active ingredient in RPC, emodin, manifests antiproliferative effects on B16F10 cells, and acts as a "mitochondrial uncoupler" following ATP down-regulation. We thus propose mitochondrial uncoupling as a potential therapeutic strategy for cancer treatment, especially for cancers with enhanced energetic demands and reduced glycolytic reserves, like melanoma. F I G U R E 1 A plant extract from rhizomes of Polygonum cuspidatum preferentially decreases cellular adenosine triphosphate (ATP) levels in B16F10 cells but not MEFs. (a) The proliferative propensities of B16F10 and MEF cell lines were compared by mean values of total cell numbers at the respective time points. n = 3 at each point. (b and c) Analyses of energy metabolism. Extra cellular acidification rate (ECAR) (b) and oxygen consumption rate (OCR) (c) were measured using Seahorse XF96 Analyzer (Agilent). ECAR of glycolysis: (ECAR of basic state) − (ECAR after treatment with 100 mM of 2-deoxyglucose). OCR of mitochondrial respiration: (OCR of basic state) − (OCR after the treatment with respiration inhibitors, rotenone and antimycin (3 μM each)). B16F10: n = 5, MEF: n = 3. (d) Two cell lines were stained with tetramethylrhodamine methyl ester (TMRM), and the mitochondrial membrane potential (MMP) in the basic state for each cell line was compared by measuring the intensity of fluorescence using microscopic analysis. Fluorescence intensities are presented as mean values of respective cells in each cell line; B16F10, n = 22. MEF, n = 28. (e) ATPase activities of whole cell lysates from B16F10 and MEF cells were measured by a modified molybdate assay. Each cell lysate was diluted to the same concentration of total protein, and ATP hydrolysis reactions were carried out. The results are presented as means of three ATP hydrolysis reactions. (a-e) Error bars indicate standard deviations. (b-e) *p < 0.05, ***p < 0.005, by Student's t test. (f) The two cell lines, B16F10 and MEF, were cultured with each plant extract at 100 μg/ml for 6 hr (A: Rhizoma Polygonum cuspidatum, B: Rhizoma Cortex periplocae, C: Fructu Cucurbitae moschatae, D: Flos Chrysanthem, E: Semen Lepidii) and subjected to luciferase-based ATP quantifications. The results were normalized by the protein amount in each well and showed as averages and standard deviations. *p < 0.05, ***p < 0.005, (B16F10, vs. DMSO), †p < 0.05, † †p < 0.01 (MEF, vs. DMSO), by Dunnett's test. N.S., not significant SUGIYAMA et Al.

Polygonum cuspidatum preferentially decreases cellular ATP levels in B16F10 cells but not MEFs
We compared growth rates and metabolic activities of two mouse cell lines, malignant melanoma B16F10 and mouse embryonic fibroblasts (MEF). Proliferation rates of B16F10 cells were much higher than those of MEF cells (Figure 1a). Assuming that highly proliferating cancer cells display enhanced energy metabolism, we compared several metabolic indicators between the two cell lines. The extracellular acidification rate (ECAR), which indicates the strength of glycolysis, was significantly higher in B16F10 cells than in MEF cells ( Figure 1b); moreover, the ECAR in B16F10 cells was also higher than in HeLa (derived from cervical cancer) and A549 cells (derived from lung cancer) ( Figure S1a), consistent with the notion that B16F10 cells manifest a prominent Warburg effect. The oxygen consumption rate (OCR), which indicates the strength of mitochondrial respiration, was also much higher in B16F10 cells than in MEF, HeLa and A549 cells (Figure 1c, Figure S1b). The mitochondrial membrane potential was also much higher in B16F10 cells than in MEF cells ( Figure 1d). These results imply that B16F10 cells rely on active energy metabolism through not only glycolysis but also mitochondrial respiration, which likely contribute to the aggressive proliferation of these cells. We also measured ATPase activities of whole cell lysates (Manno, Noguchi, Fukushi, Motohashi, & Kakizuka, 2010) and found that the rate of ATP hydrolysis of B16F10 cell extracts was about twice the rate in MEF cell extracts ( Figure 1e). These results were consistent with the notion that B16F10 cells produce and consume cellular ATP more actively than MEF cells.
On the assumption that some plant extracts used in Chinese medicine might reduce ATP levels in highly proliferating cells, we screened more than 1,000 plant extracts and found one (extract A) that was able to decrease cellular ATP levels in B16F10 cells, but not in MEF cells (Figure 1f). Extract A was derived from rhizomes of Polygonum cuspidatum (RPC).

| Emodin is the principle active ingredient in the RPC extract
In order to identify active ingredients in the RPC extract, we carried out polarity-based fractionations (Figure 2a). For this purpose, we used the second lot (#2) of the RPC extract, which was prepared in a larger amount, although its activity was lower than the first lot (#1) (Figure 2b). First, we separated the extract into two fractions, water-soluble (F1) and organic-soluble (F2), by the Bligh-Dyer method. We treated B16F10 and MEF cells with each of the two fractions at 100 μg/ml for 6 hr and measured ATP levels of the cells. The results showed that the F2 fraction drastically decreased cellular ATP levels in B16F10 cells, and this effect was clearly much weaker in MEF cells ( Figure 2b). Next, F2 was dissolved in 100% CHCl 3 , followed by solid phase extraction (SPE), and we separated it into three fractions, F2-1 (eluted by 100% CHCl 3 ), F2-2 (eluted by CHCl 3 :MeOH = 4:1) and F2-3 (eluted by CHCl 3 :MeOH = 1:1). We then treated cells with each fraction (F2, F2-1, F2-2 or F2-3), at 30 μg/ml each for 6 hr and measured cellular ATP levels. Two fractions, F2 and F2-1, showed significant reductions of cellular ATP levels in B16F10 cells, and the effect was much more pronounced with the F2-1 treatment than the F2 treatment ( Figure 2c).
of emodin (11.53 min) and physcion (19.04 min), respectively ( Figure 2d). Furthermore, optical spectra of each peak (A or B) in F2-1 were almost identical with that of emodin or physcion, respectively (Figure 2e). To confirm the structure of peak A and B, we analyzed F2-1, emodin and physcion, by 1 H-NMR. The chemical shifts of each peak (A or B) were accurately matched with that of emodin or physcion, respectively, and molar ratio of A and B was approximately 1:0.28 (Figure 2f,g) (Danielsen, Aksnes, & Francis, 1992). To confirm this ratio, we compared peak areas between F2-1 and a mixed sample of emodin and physcion, which was prepared at 1:0.28 molar ratio. The result showed that the area ratio of the two peaks in F2-1 and the mixed sample were very close ( Figure S3). From these results, we concluded that F2-1 was composed of two anthraquinones, emodin and physcion (Figure 2h), at a molar ratio of approximately 1:0.28.
We then treated two cell lines with 8 μg/ml of emodin or 8 μg/ml of emodin + 2.24 μg/ml of physcion (emodin: physion = 1:0.28 in molar ratio), examined the ATP levels of the cells, and found that both had similar capacities to decrease cellular ATP levels in B16F10 cells (Figure 2i). Physcion did not affect emodin's action in B16F10 cells. However, the mixure of emodin and physcion appeared to reduce ATP levels in MEF cells, although this effect was not significant. This tendency was also observed in MEF cells treated with 30 μg/ml of F2-1 (Figure 2c). From these results, we concluded that an effective ingredient in the RPC extract, which reduces ATP levels preferentially in B16F10 cells, is emodin.

| Emodin suppresses cell proliferation of B16F10 cells in vitro and in vivo
We next examined the suppressive effects of emodin on B16F10 cell proliferation. As expected, 4, 8 and 16 μg/ml of emodin showed significant suppression of proliferation of B16F10 cells (Figure 3a). In contrast, only 16 μg/ml of emodin showed significant suppression on the proliferation of MEF cells. Similar suppressive effects on the cellular ATP levels were observed (Figure 3b, Figure S4). In these experiments, we used cisplatin as a control, which is well known to manifest severe side effects in patients, and found that cisplatin showed similar antiproliferative effects in both B16F10 and MEF cells ( Figure S5a). Because its mode of action is DNA interstrand crosslinking rather than perturbation of energy homeostasis per se, cisplatin did not reduce ATP levels in both B16F10 and MEF cells ( Figure S5b). Following up on emodin's ability to preferentially decrease cellular ATP levels, much more in B16F10 cells than in MEF cells, we evaluated phosphorylation (Thr172) of AMP-activated protein kinase (AMPK). Western blot analyses showed a clear increase in phosphorylation of AMPK at 8 and 16 μg/ml of emodin in B16F10 cells, compared with MEF cells ( Figure  3c).
We next examined the antiproliferative effects of emodin up to 3 days in B16F10 and MEF cells (Figure 3d). 4 and 8 μg/ml emodin produced significant effects in B16F10 cells (p < 0.005 and p < 0.005, respectively), and marginal or less significant effects in MEF cells (p = 0.571 and p = 0.059, respectively). It is noteworthy that cisplatin suppressed proliferation much more effectively in MEF cells than B16F10 cells. In order to evaluate emodin's antiproliferative effects on B16F10 cells in vivo, we treated mice bearing B16F10 cell-derived tumors with emodin (50 mg kg −1 day −1 ), as outlined in Figure 3e. An emodin-treated group (n = 9) showed significant suppressions of tumor growth at 8, 10 and 12 days after the injection of B16F10 cells, as compared to a vehicle-treated group (n = 9) (Figure 3f).

| Emodin functions as a mitochondrial uncoupler
Because emodin decreased cellular ATP levels in B16F10 cells, we next evaluated mitochondrial functions after emodin treatment. We first evaluated the mitochondrial membrane potential (MMP) by using tetramethylrhodamine methyl ester

F I G U R E 3 Emodin suppresses cell proliferation of B16F10 cells in vitro and in vivo. (a and b) B16F10 cells and MEF cells were treated
with or without the indicated concentrations of emodin for 24 hr followed by a cell proliferation assay using a trypan blue dye-exclusion test (left), and 6-hr treatment was carried out for an evaluation of adenosine triphosphate (ATP) levels using a luciferase assay (right). Each result shows the average of three independent trials, and error bars indicate standard errors. Statistical analyses were carried out using total cell numbers.  (TMRM) and we found that the addition of emodin rapidly reduced the MMP, which dropped to levels approximately 30% of those before its addition, within only 30 min ( Figure  4a) (Qu et al., 2013). This effect continued up to 24 hr after its addition (Figure 4b). Notably, similar phenotypes were also observed in MEF cells. We considered two possibilities for the decrease of the MMP: inhibition of OXPHOS or proton leakage from the inner membrane space to the matrix, which would tend to decrease or enhance the oxygen consumption rate (OCR), respectively. Thus, we monitored the OCR before and following the addition of emodin; we also used carbonyl cyanide 3-chlorophenylhydrazone (CCCP), a mitochondrial uncoupler, or rotenone, a respiration inhibitor, as controls. Treatment with emodin or CCCP similarly stimulated the OCR in both cell lines, and the stimulated OCR was completely abrogated by the inhibition of the mitochondrial respiratory chain with rotenone and antimycin (Figure 4c). It is known that cells increase OXPHOS when faced with proton leakage in order to compensate for the lack of the proton gradient between the inner membrane space and the matrix. This notion and our results support the idea that proton leakage, but not inhibition of OXPHOS, is the underlying mechanism. ATP synthetase is a well-known proton leaker. However, the enhancement of the OCR by emodin or CCCP was also induced in the presence of oligomcin, an inhibitor of ATP synthetase, indicating that ATP synthetase is not involved in this increased respiration (Figure 4d, Figure S6). It is also known that proton leakage can also be chemically induced by an uncoupler, such as CCCP. Indeed, emodin and CCCP acted very similarly to decrease the MMP and increase the OCR. CCCP combines protons in a proton-rich environment and is able to pass through the cellular membrane and then provoke the release of protons in a neutral environment. For both of these to occur, the uncoupler requires two basic features, lipophilicity and weak acidity. It is notable that two parameters, pKa and logP, are very similar between emodin and CCCP (Figure 4e) (Scifinder, 2017).
We next examined the relationship between mitochondrial uncoupling and cellular phenotypes induced by emodin. We treated both B16F10 and MEF cells with three different mitochondrial uncouplers and evaluated their effects. Treatment with CCCP, carbonyl cyanide 4-(trifluoromethoxy)phenylhydazone (FCCP) and 2,4-dinitrophenol (DNP) all decreased MMP similarly to what we observed with the emodin treatment (Figure 4f, Figure S7). FCCP and DNP marginally and emodin and CCCP significantly reduced ATP levels ( Figure  4g), and emodin and all three uncouplers preferentially showed antiproliferative effects on B16F10 cells, opposed to MEF cells (Figure 4h). Furthermore, inhibition of mitochondrial respiratory chain complex Ⅰ with rotenone or metformin did not show preferential ATP reduction and antiproliferation effects in B16F10 cells rather than MEF cells (Figure 4i,j). It is notable that both inhibitors reduced the ATP levels more potently in MEF cells than in B16F10 cells (Figure 4i). From these lines of evidence, we concluded that emodin functions as a mitochondrial uncoupler (Attene- Ramos et al., 2013;Betina & Kuzela, 1987;Ubbink-Kok, Anderson, & Konings, 1986).

F I G U R E 4 Emodin functions as a mitochondrial uncoupler. (a) Cell lines were cultured on collagen-coated glass-bottom dishes for
approximately 24 hr, and medium was replaced by fluorescent dye-containing media (50 nM of TMRM, 1 μg/ml of Hoechst 33342, phenol red(−)) followed by incubation at 37℃ for 30 min. The fluorescence images representing MMPs, before and after the addition of emodin (8 μg/ml, final), were captured, and each fluorescence intensity per cell was quantified using MetaMorph (Molecular Devices). ***p < 0.005 (B16F10, vs. 0 min), † † †p < 0.005 (MEF, vs. 0 min), by Dunnett's test. (b) In parallel, dishes with each respective cell line were subjected to a preliminary treatment with DMSO or emodin (8 μg/ml) for 24 hr, and the respective intensity of TMRM was compared using the same analysis as in (a) to yield the result at 24 hr. The results are presented as box-and-whisker plots. ***p < 0.005 (vs. DMSO) by Student's t test. (c and d) The effect of emodin on mitochondrial respiration was evaluated by Seahorse XF96 Analyzer. Cells were cultured for approximately 24 hr, and growth medium was replaced by XF RPMI 1640-based medium (bicarbonate-free, Seahorse). In (c), measurements of the OCR were carried out with the following initial injections: emodin (8 μg/ml, final), CCCP (5 μM, final) or rotenone (0.3 μM: final). The second injection was a mixture of complex Ⅰ inhibitors (rotenone + antimycin, 3 μM each, final). Respective results represent the average of five wells, and error bars indicate standard deviations. In (d), the first injection was oligomycin (3 μM: final), which was followed by the second injections: emodin (8 μg/ml, final) or CCCP (5 μM, final); a third injection consisted of a mixture of complex Ⅰ inhibitors (rotenone + antimycin, 3 μM each, final). Each result represents the average of 5 wells, and error bars indicate standard deviations. (e) Calculated values from the database (SciFinder), of pKa and logP. (f) MMPs of two cell lines, before and after the addition of emodin, uncouplers (CCCP, FCCP or DNP) or CDDP, were evaluated using TMRM (50 nM) and Hoechst 33342 (1 μg/ml). Scale bar, 50 μm. In this figure, the MPP brightness of MEF cells was enhanced to clarify the differences before and after the injection of uncouplers. The unenhanced images are shown in Figure S7. (g and h) Cell lines were treated with DMSO, emodin, uncouplers (CCCP, FCCP or DNP) or CDDP, for 6 hr followed by luciferase-based ATP quantification (g), and in the case of 24-hr treatments, cells were subjected to a trypan blue dye-exclusion test (h). Each vertical bar indicates the average of three wells, and error bars indicate standard deviations. *p < 0.05, ***p < 0.005 (B16F10, vs. DMSO), †p < 0.05, † † †p < 0.005 (MEF, vs. DMSO), by Dunnett's test. (i and j) Cell lines were treated with DMSO or complex Ⅰ inhibitors (rotenone or metformin) for 6 hr, followed by luciferase-based ATP quantification (i), and in case of the 24-hr assay, total and living cell numbers were assessed by a trypan blue dye-exclusion test (j). Results indicate mean values, and error bars indicate standard deviations. ***p < 0.005 (B16F10, vs. DMSO), † † †p < 0.005. (MEF, vs. DMSO), by Dunnett's test. Statistical analyses were carried out using total cell numbers in (h, j) SUGIYAMA et Al.

| Glycolytic reserve of B16F10 cells is poor compared with MEF cells
Given the enhanced sensitivity of B16F10 cells to emodin, we next focused on another energy-producing system, glycolysis, in cells treated with uncouplers. First, we measured glucose consumption in media and found that emodin and three mitochondrial uncouplers drastically enhanced glucose consumption in MEF cells, whereas only a weak increase was observed with B16F10 cells (Figure 5a). Next, we monitored the extracellular acidification rate (ECAR) followed by the addition of mitochondrial uncouplers and 2-deoxy-D-glucose (2-DG). Because of the autofluorescence of emodin, we could not obtain reliable ECAR data for cells treated with emodin in this system. Nevertheless, CCCP, FCCP and DNP dramatically increased the ECAR of MEF cells, up to three to five times higher than the baseline points before their addition, indicating an obvious increase in glycolysis because of uncoupling (Figure 5b,c right). This compensatory activation of glycolysis in the face of mitochondrial dysfunctions, termed the "glycolytic reserve," could potentially maintain cellular ATP levels. By contrast, with B16F10 cells, the enhancements of ECAR were only 50%-80% compared with baseline levels (Figure 5b,c, left). It is notable that HeLa and A549 cells appeared to have more glycolytic reserve than B16F10 cells ( Figure S8a). Consistently, emodin and all three uncouplers produced smaller effects on HeLa and A549 cells than on B16F10 cells ( Figure S8b).
Next, we treated B16F10 and MEF cells with emodin in various concentrations of glucose and measured ATP levels by a luciferase-based assay. The result showed that, in all concentrations of glucose (200, 100, 50 and 0 mg/dl), cellular ATP levels of B16F10 cells treated with 8 μg/ml of emodin for 6 hr clearly decreased (Figure 5d, left). In contrast, MEF cells could not keep their ATP levels only under the glucosefree conditions (in correct, this medium included FBS-contained glucose) by treatment of 8 μg/ml of emodin for 6 hr (Figure 5d, right). And this reduction of ATP levels could not be rescued by treatment of pyruvate, a metabolite for TCA cycle ( Figure S9a). Furthermore, treatment with 8 μg/ ml emodin suppressed the proliferation of MEF cells under glucose-free condition ( Figure S9b,c). These results suggest that the resistance of MEF cells to emodin depends on the glycolytic reserve.

| DISCUSSION
Cancer cells use higher amounts of energy for their enhanced proliferation, as compared with nonmalignant normal cells, indicating the presence of cancer-specific energy metabolisms, for example, enhanced glycolysis or Warburg effects. The corresponding cancer-specific energy metabolisms have long been considered as potential targets for cancer therapy. In this regard, the down-regulation of ATP production might be a simple and effective strategy for limiting the proliferation of cancer cells. Indeed, "metformin," which functions as an inhibitor of mitochondrial respiratory chain complex I, has been evaluated in clinical trials for cancer therapy.
We have collected more than 1,000 plant extracts and used them as sources of chemicals with potential benefits for human health; we have previously identified "Garcinielliptone HC" from hop flower extracts as a potential prophylactic for Alzheimer's Disease (Sasaoka et al., 2014). We then hypothesized that some plant extracts might contain chemicals that can reduce ATP levels in a cancer cell-specific manner. We then screened the extracts and found that one extract, which was derived from rhizomes of Polygonum cuspidatum (RPC),

F I G U R E 5 Glycolytic reserve of B16F10 cells is poor compared with that of MEF cells. (a) B16F10 and MEF cells were cultured in 24-
well plates for approximately 24 hr, and the growth medium was replaced by fresh medium containing DMSO, emodin (8 μg/ml), CCCP (5 μM), FCCP (3.5 nM), DNP (150 μM) or CDDP (30 μM), respectively. After incubation for 4 hr, respective culture media were collected, and glucose concentrations were measured by the glucose CII test (Wako). The results were normalized to the attached cell number for each well. The data represent averages of four wells, and error bars indicate standard deviations. ***p < 0.005 (B16F10, vs. DMSO), † † †p < 0.005 (MEF, vs. DMSO), by Dunnett's test. N.S., not significant. (b) CCCP's effect on glycolysis was evaluated by monitoring ECAR using Seahorse XF96 Analyzer. Approximately 24 hr after plating, cells were treated with the indicated concentrations of CCCP (2.5 or 5 μM, final) for the initial injections, followed by a second injection of 2-DG (100 mM, final). Results are shown as ECAR per 1 × 10 4 cells, which represent the mean values of four wells. Error bars show standard deviations. (c) Uncouplers' effects on glycolysis were evaluated by monitoring ECAR using Seahorse XF96 Analyzer. Two cell lines were treated with the indicated concentrations of mitochondrial uncouplers (CCCP, FCCP, DNP), CDDP (30 μM) or DMSO for the initial injections, followed by a second injection of 2-DG (100 mM, final). Results are shown as ECAR per 1 × 10 4 cells, which represent the mean values. Error bars show standard deviations. (d) B16F10 and MEF cells were treated with DMSO or emodin (8 μg/ml) for 6 hr, with the indicated concentrations of glucose (0, 50, 100 or 200 mg/dl) (in correct, this medium included FBS-contained glucose), and were subjected to cellular adenosine triphosphate (ATP) quantification using the luciferase assay. Each value for the luciferase assay was normalized by the protein amount in each well. The data represent mean values, and error bars indicate standard deviations. *p < 0.05, ***p < 0.005, by Student's t test (vs. DMSO). N.S., not significant. (e) A model of energy balance in each cell line, under the two different conditions. In basic conditions, B16F10 cells produce and consume a large amount of ATP. In the presence of a mitochondrial uncoupler, this cell line cannot maintain its energy state by relying on glycolysis, because of the poor glycolytic reserve. In contrast, because the turnover of ATP in MEF cells is slow and basal glycolysis is very low, their glycolytic reserve can compensate for the energy depletion brought about by mitochondrial uncoupling SUGIYAMA et Al. had such activities; the extract decreased cellular ATP levels in rapidly proliferating melanoma B16F10 cells but not normal mouse embryonic fibroblast (MEF) cells. Using several analytical techniques, we identified the principle active ingredient as "emodin," an anthraquinone. "Emodin" has been repeatedly purified from many different plant extracts and been reported to elicit many cellular effects, for example, anticancer, hepatoprotective, anti-inflammatory, antimicrobial and antioxidant activities (Dong et al., 2016). These observations surprised us, because only one plant extract, RPC, effectively lowered ATP levels in B16F10 cells. These results indicate that RPC extract contains much more emodin than any other plant extracts tested.
Emodin as well as RPC extract showed very promising characteristics as potential agents for cancer therapy. Both dramatically reduced ATP levels in B16F10 cells but not in MEF cells. Concomitantly, both inhibited proliferation of B16F10 cells more potently than MEF cells. These differential inhibitory effects present a sharp contrast to cisplatin, an anticancer drug widely used in current clinical treatments. Cisplatin did not reduce ATP levels in both B16F10 cells and MEF cells; moreover, it inhibited proliferation of both cells. It is notable that the antiproliferative effects of cisplatin were even greater on MEF cells than on B16F10 cells, which may explain some of the severe side effects of cisplatin (Tsang, Al-Fayea, & Au, 2009). With this in mind, emodin and RPC extract might produce milder side effects than cisplatin. The antiproliferative effects of emodin were also confirmed in an in vivo assay, using a mouse melanoma model in which B16F10 cells were subcutaneously implanted. Administrations of emodin (50 mg kg −1 day −1 ) to the mice significantly suppressed the tumor growth of B16F10 cells. These results support the idea that ATP down-regulation in cancer cells is a potential therapeutic strategy to suppress cancer cell proliferation.
What is molecular basis of the antiproliferative effect of emodin? Emodin has been shown to produce oxidative stress in cells (Qu et al., 2013;Su, Chang, Shyue, & Hsu, 2005). We thus examined the effect of N-acetyl-L-cysteine (NAC), a well-known antioxidant, on the antiproliferative effects of emodin. Treatment of 1 mM NAC, which clearly suppressed the cytotoxicity induced by H 2 O 2 , did not diminish the antiproliferative effects of emodin at all ( Figure S10). We further observed that emodin treatment led to stimulation of the OCR, which is usually accompanied by an elevation of the MMP. However, emodin reduced the MMP. These results indicate that emodin functions as a mitochondrial uncoupler. Consistently, other mitochondrial uncouplers, CCCP, FCCP and DNP, also showed similar effects on the OCR and MMP. These uncouplers also reduced cellular ATP levels and showed antiproliferative activities, and both were prominent on B16F10 cells, but not on MEF cells. Furthermore, the chemical profiles of emodin fit well with those of the uncouplers or the ionophore. Taken together, we conclude that the antiproliferative effects of emodin are due to function as a mitochondrial uncoupler. Some of the studies reported that emodin provokes an antiproliferative activity or cell death in cancer cell lines and also mentioned various molecule behaviors in cancer cells treated with emodin. For example, an activation of mitochondrial apoptotic pathway, stimulated FAS ligand pathway, decreased gene expression of C-MYC, lowered signals related to the cell stemness or reduced protein level of ERα (Dong et al., 2016). Nevertheless, in the cancer cells treated with emodin, the initial event at molecular level had been unclear. Furthermore, no studies focused on emodin's noteworthy actions which we observed in this study, mitochondrial uncoupling and ATP down-regulation, as a potential target to overcome cancer malignancy.
The ability of uncouplers to lower ATP levels directed our attention to the preferentially suppression of growth of cancer cells such as B16F10, as compared with MEF cells. Emodin and other uncouplers were able to enhance glucose consumptions in both B16F10 cells and MEF cells, but the effects were much stronger in MEF cells (300% to 450% enhancements) than in B16F10 cells (170% to 190% enhancements). Consistently, CCCP, FCCP and DNP were able to enhance ECAR much more strongly in MEF cells than in B16F10 cells (Recall that we were unable to conduct the ECAR assay with emodin because of its autofluorescence). Furthermore, MEF cells could not sustain their ATP levels after the emodin stimulation under the very low concentrations of glucose. These results indicated that B16F10 cells have much less "glycolytic reserve" than MEF cells and thus B16F10 cells were much more sensitive to emodin or other uncouplers than MEF cells, resulting in a more pronounced reduction in ATP levels and suppression of the growth of B16F10 cells than MEF cells. From these results, we concluded that the "glycolytic reserve" is a major determinant in maintaining cellular ATP levels in response to mitochondrial uncouplers. It is notable that many tumors grow in hypoxic conditions, indicating that such hypoxic cancer cells maximally use glycolysis and thereby maintain less glycolytic reserve, which would further enhance the Warburg effect. We thus propose ATP down-regulation by mitochondrial uncoupling as a challenging strategy for the cancer therapy, alone or in combination with other cancer therapeutics already in use.

| 581
Genes to Cells SUGIYAMA et Al. TESQUE). HeLa (also kindly provided by Professor Shin Yonehara (Kyoto University)) and A549 (obtained from Riken Bioresource Center Cell Bank) cell lines were maintained in Dulbecco Modified Eagle medium (DMEM-high glucose, NACALAI TESQUE). Cell culture was carried out with 10% fetal bovine serum (Sigma) in a humidified incubator with 5% carbon dioxide (CO 2 ) at 37℃. All cells were routinely tested for Mycoplasma using MycoAlert (Lonza).

| Extracellular flux analysis
Extracellular acidification rate (ECAR) and the oxygen consumption rate (OCR) were measured by Seahorse Extracellular Flux Analyzer 96 (Seahorse XF96 Analyzer, Agilent) according to the manufacture's protocol. Briefly, cells were plated at 1x10 4 cells per well in a Seahorse XF96 Cell Culture Microplate and were cultured for approximately 24 hr. Then, media were changed to media with lower pH buffering capacity (Agilent 103676-100) followed by placing the plate in Seahorse XF96 Analyzer. Emodin or other reagents were injected at defined time points automatically. After the measurements, cells were fixed (20% formaldehyde, 2% glutaraldehyde, PBS (without Ca 2+ and Mg 2+ ) and stained with Hoechst 33342 (Invitrogen). Cell numbers were counted by counting the number of nuclei using an ArrayScan VTI High Content Platform (Thermo Fisher Scientific); then, ECAR and OCR were normalized by the cell number in each well.

| ATPase activity assay
Cells cultured in 100-mm dishes were washed with ice-cold PBS and then harvested in 500 μl of ice-cold ATPase buffer (50 mM Tris-HCl, 250 mM sucrose, 5 mM MgCl 2 , 0.5 mM EDTA, 0.3 mM DTT) using a scraper. Cells were lysed by repeated passage through a 27 G needle attached to a 1cc syringe on ice, followed by centrifugation at 13,000 g, 4℃ for 10 min. The supernatants were transferred to new tubes, and aliquots were subjected to a bicinchoninic acid assay (BCA, NACALAI TESQUE) to quantify protein concentrations.
The cell lysates were diluted with ATPase buffer to 1 μg of total protein in 20 μl and incubated at 37℃ for 15 min. After the incubations, 100 μM [γ-32 P] ATP (18.5 GBq/mmol) in 20 μl of ATPase assay buffer (1 mM HEPES[pH 7.4], 2.5 mM KCl, 5 mM MgCl 2 , 50 μM ATP, 15 mM DTT) was added to monitor ATP hydrolysis reactions for 20 min. Then, 200 μl of ice-cold 8% TCA was added to quench the reactions, followed by the additions of 50 μl of solution A (3.75% ammonium molybdate, 0.02 M silicotungstic acid in 3 N H 2 SO 4 ) and 300 μl of n-butyl acetate to the reactions. The samples were thoroughly mixed and centrifuged at 15,000 g for 5 min at room temperature. After the phase separation, 200 μl aliquots of the upper layer were mixed with 2 ml of clear-sol Ⅱ, and radioactivity, corresponding to 32 P release, was quantitated with a liquid scintillation counter (PerkinElmer). The relative radioactivities were shown as results.

| Evaluations of mitochondrial membrane potential
Cells plated on a collagen-coated glass-bottom 35-mm dish (Mat-Tek) were treated with 50 nM of tetramethylrhodamine methyl ester (TMRM) (Invitrogen) in phenol red-free RPMI 1640 (NACALAI TESQUE) with 10% FBS for 30 min at 37℃. Imaging was carried out with a Nikon Ti-E inverted microscope using a 60× objective (Nikon; CFI Plan Apo λ 60× oil: NA 1.40) which was controlled with NIS-Elements (Nikon). The following filter sets (Semrock) were used: for imaging of TMRM, 562/40 excitation filter-dichroic mirror FF593-641/75 emission filter; and for imaging of Hoechst 33342, 405/10 excitation filter-dichroic mirror FF495-520/35 emission filter. Fluorescence emissions from samples were captured with a Zyla4.2 sCMOS camera (ANDOR). Throughout the imaging, cells were maintained at 37℃ with a continuous supply of a mixture of 95% air and 5% CO₂ using a stage-top incubator (Tokai Hit). The imaging results were quantified by MetaMorph (Molecular Devices).

| Luciferase-based ATP quantification
Cells were plated in a 24-well plate and were cultured for approximately 24 hr before drug treatment. At each period (6 or 24 hr), culture media were removed and cells were gently washed with PBS. 200 μl of Glo Lysis Buffer (1×, Promega) was added to each well, and cells were incubated for 5 min. Then, the plate was agitated with a plate-shaker for 1 min and the supernatants were transferred to 96-well plate. Bioluminescence of each well was measured by an ARVO multilabel counter (PerkinElmer) using ATP assay reagent for cells (ToyoB-net). Each value of bioluminescence was normalized by the protein content, respectively, using a Protein Assay Bicinchoninate Kit (NACALAI TESQUE).