Metformin activates KDM2A to reduce rRNA transcription and cell proliferation by dual regulation of AMPK activity and intracellular succinate level

Metformin is used to treat type 2 diabetes. Metformin activates AMP-activated kinase (AMPK), which may contribute to the action of metformin. Metformin also shows anti-proliferation activity. However, the mechanism is remained unknown. We found that treatment of MCF-7 cells with metformin induced the demethylase activity of KDM2A in the rDNA promoter, which resulted in reductions of rRNA transcription and cell proliferation. AMPK activity was required for activation of KDM2A by metformin. Because demethylase activities of JmjC-type enzymes require a side reaction converting α-ketoglutarate to succinate, these organic acids may affect their demethylase activities. We found that metformin did not induce KDM2A demethylase activity in conditions of a reduced level of α-ketoglutarate. A four-hour treatment of metformin specifically reduced succinate, and the replenishment of succinate inhibited the activation of KDM2A by metformin, but did not inhibit the activation of AMPK. Metformin reduced succinate even in the conditions suppressing AMPK activity. These results indicate that metformin activates AMPK and reduces the intracellular succinate level, both of which are required for the activation of KDM2A to reduce rRNA transcription. The results presented here uncover a novel factor of metformin actions, reduction of the intracellular succinate, which contributes to the anti-proliferation activity of metformin.

succinate can affect the activities of JmjC enzymes 28,[34][35][36][37] . For example, a defective mutation of succinate dehydrogenase, a TCA cycle enzyme, contributes to tumorigenesis in a subset of human cancers. This could be due to changes in the succinate level and subsequent epigenetic reprograming by JmjC enzymes 35,38 . We previously reported that a cell-permeable succinate, dimethyl-succinate (DMS), inhibited the demethylation activity of KDM2A and the reduction of rRNA transcription induced by serum and glucose starvation 28 .
Here, we investigated whether metformin controlled the KDM2A activity to regulate rRNA transcription. The effects of α-KG and succinate on the activation of KDM2A were also examined. We found that metformin activated KDM2A to reduce rRNA transcription and cell proliferation. Interestingly, a four-hour treatment with metformin not only activated AMPK, but also independently decreased the intracellular succinate level; metformin activated AMPK in conditions without reduction of the succinate level and reduced the level of intracellular succinate in conditions inhibiting AMPK activity. The activation of KDM2A required both the AMPK activity and the decrease of the intracellular succinate level. Our results indicate that reduction of the intracellular succinate level is a novel mechanism by which metformin activates KDM2A and exerts its anti-proliferation activity.

Results
Metformin induces KDM2A activity to reduce rRNA transcription. Breast cancer cell line MCF-7 cells were treated with metformin to examine whether metformin activated KDM2A to reduce rRNA transcription. Treatment of cells with metformin for four hours reduced rRNA transcription in a dose-dependent manner (Fig. 1A). A KDM2A knockdown impaired the reductions when cells were treated with 2.5 and 5 mM metformin (Fig. 1A). The KDM2A knockdown did not impair the activation of AMPK by metformin (Fig. 1B). We previously reported that KDM2A reduces the levels of H3K36me2 specifically in the rDNA promoter under starvation conditions, but does not reduce them in the transcribed region 28,30 . The levels of H3K36me3, which is not a direct substrate for KDM2A 32 , are not reduced under starvation conditions in both the rDNA promoter and the transcribed regions 28 . The treatment with 2.5 mM metformin reduced the level of H3K36me2 but not H3K36me3 in the rDNA promoter (Fig. 1C). A knockdown of KDM2A impaired the reduction of H3K36me2 by metformin (Fig. 1C). In the transcribed region, reductions of H3K36me2 and H3K36me3 marks by metformin were not observed (Fig. S1). We previously showed that the KDM2A-dependent reduction of rRNA transcription by starvation reduces proliferation of breast cancer cells 30 . It was found that treatment with 2.5 mM metformin reduced cell proliferation, and the KDM2A knockdown suppressed the reduction (Fig. 1D).Together these results suggest that metformin activates KDM2A to reduce H3K36me2 in the rDNA promoter and rRNA transcription, which results in a reduction of cell proliferation.
Previously, it was shown that the AMPK activation is required for the activation of KDM2A to reduce rRNA transcription in response to glucose starvation 30 . The AMPK inhibitor compound C and a new AMPK inhibitor SBI-0206965 39 , which shows lower kinase promiscuity than compound C, inhibited AMPK activity and the reduction of rRNA transcription induced by metformin ( Fig. S2A-D). A knockdown of AMPKα also inhibited the reduction of rRNA transcription by metformin (Fig. S2E,F). Compound C and SBI-0206965 also impaired the metformin-induced reduction of H3K36me2 in the rDNA promoter (Fig. S2G,H). These results suggest that AMPK activity is required for the metformin-induced reduction of rRNA transcription and H3K36me2 level in the rDNA promoter through KDM2A. treatments that decrease α-KG level inhibit the activation of KDM2A by metformin. The reaction of the JmjC-type demethylases proceeds with a side reaction that produces succinate from α-KG 32 . Therefore, α-KG and succinate may affect their demethylase activities. It was reported that treatment with a cell-permeable α-KG, dimethyl-alpha ketoglutarate (DMαKG), induces JmjC-enzyme activities 36,40 . We found that DMαKG treatment induced the KDM2A demethylase activity in rDNA promoter and reduced rRNA transcription (Fig. S3).
Next, we tried to decrease the levels of intracellular α-KG and evaluated the contribution of α-KG to the activation of KDM2A by metformin. α-KG is generated through two production systems, glycolysis followed by the TCA cycle and glutaminolysis from glutamine. Generally, in cancer cells, the production of α-KG by glutaminolysis is elevated, because cancer cells favor aerobic glycolysis and glucose is often diverted to lactate production [41][42][43] . When MCF-7 cells were cultured in glutamine-free medium for four hours, the α-KG level was decreased to half ( Fig. 2A), and the succinate and fumarate levels tended to be decreased ( Fig. 2A). The four-hour metformin treatment did not reduce rRNA transcription in the glutamine-free medium (Fig. 2B, −Gln), but it reduced the rRNA transcription culturing in glutamine-containing medium (Fig. 2B +Gln) and the glutamine-free medium that had been added glutamine (Fig. 2B, −Gln/+Gln). The reduction of rRNA transcription during culturing in the glutamine-free medium that had been added glutamine was inhibited by KDM2A knockdown (Fig. 2C). Pretreatment with an inhibitor of the glutaminase 44 , which inhibits the production of α-KG from glutamine, impaired the reduction of rRNA transcription by metformin (Fig. 2D). The effect of this inhibitor on the metformin-mediated regulation of rRNA transcription was not observed in the glutamine-free medium (Fig. S4). Therefore, two treatments that reduced intracellular α-KG levels inhibited KDM2A activity and the results are continuity of the story that the α-KG level in cells affects the KDM2A activity to reduce rRNA transcription induced by metformin.
Metformin reduced the H3K36me2 levels in the rDNA promoter in glutamine-containing medium but not in glutamine-free medium (Fig. 2E, −Gln). The addition of glutamine to the glutamine-free medium restored the reduction of the H3K36me2 levels induced by metformin (Fig. 2F, −Gln/+Gln). A KDM2A knockdown impaired the reduction of H3K36me2 levels (Fig. 2F). These results suggest that glutamine, which is converted to α-KG, is a critical factor for KDM2A activated by metformin to demethylate H3K36me2 marks in the rDNA promoter.
The rRNA transcription tended to be reduced by culture in glutamine-free medium for four hours without metformin (Fig. 2B). The glutamine-depletion activated AMPK (Fig. S5A) but did not reduce the level of H3K36me2 in www.nature.com/scientificreports www.nature.com/scientificreports/ the promoter (Fig. 2E) and in the transcribed region (Fig. S5B) of rDNA. These results suggest that the tendency of glutamine-free medium to reduce rRNA transcription occurred independently of the demethylase activity of KDM2A. These results suggest that AMPK activation under glutamine-free medium is not enough to induce the demethylase activity of KDM2A in the rDNA promoter and that the intracellular level of α-KG may be a factor to activate KDM2A.
Metformin decreases cellular succinate. Recently, it was reported that metformin treatment for more than 24 h decreased the intracellular levels of several metabolites in the TCA cycle [45][46][47] , including α-KG and succinate. However, the changes of metabolites soon after the metformin treatment have not been reported yet. Thus, we detected the amounts of intracellular metabolites after the metformin treatment in our conditions. After cells were treated with 0, 2.5, or 10 mM metformin for four hours, the levels of 65 metabolites were quantitated by LC-MS/MS (Supplementary Table 1 Table 2. Four metabolites were decreased, and two metabolites were increased in cells treated with 2.5 mM metformin, while the seven metabolites were decreased and 22 metabolites were increased in cells treated with 10 mM metformin (Supplementary Table 2).
Interestingly, the level of succinate was apparently most affected by metformin among the metabolites (Figs. 3A and S6A). The TCA cycle intermediates measured in this study are shown in Figs. 3B and S6B. Succinate was decreased to half by 2.5 mM metformin compared to that in non-treated cells (Fig. 3B). The levels of fumarate and malate tended to be slightly increased, while the levels of α-KG, citrate, and aconitic acid were not affected (Fig. 3B). These changes were clearer in cells treated with 10 mM metformin (Fig. S6B). The decrease of succinate as well as the increases of fumarate and malate by metformin were confirmed by another LC-MS/MS measurement with a different LC method (Fig. S6C).
It should be noted that α-KG, which was suggested to activate the KDM2A activity (Figs. 2 and S3), did not increase, and fumarate which was reported to inhibit JmjC enzymes 48 did not decrease with the metformin treatment (Figs. 3 and S6). Together, these results suggest that succinate, which can work as an inhibitor of JmjC enzymes, was specifically reduced by the four-hour treatment of metformin among TCA cycle intermediates.
Because KDM2A converts α-KG to succinate, the activity of KDM2A would be inhibited by succinate produced by KDM2A. Therefore, the negative feedback loop may exit. On the metformin treatment, KDM2A was activated and succinate was reduced. Therefore, metformin would suppress the effect of the negative feedback loop. One possible way to suppress this loop is that the reduction of succinate by metformin is too quick and large compared to the effect of the negative feedback loop at this time point. The fact that metformin reduced succinate level without change in α-KG level (Fig. 3B) suggest that KDM2A activity on the metformin treatment did not have much effect on the overall levels of the intracellular organic acids, and supported the above possibility.

Reduction of intracellular succinate is required for KDM2A activation by metformin.
Our results raised the possibility that the decrease of the succinate level by metformin could contribute to the activation of KDM2A to reduce rRNA transcription. To test this possibility, cells were treated with a cell-permeable succinate, dimethyl succinate (DMS), to increase the intracellular succinate level. The increase of the intracellular succinate was confirmed at four hours in metformin-treated cells with the treatment of 5 mM DMS (Fig. 4A), but it did not abolish the activation of AMPK by metformin (Fig. 4B). The DMS treatment inhibited the reduction of rRNA transcription induced by 2.5 mM metformin for four hours (Fig. 4C). The reduction of the H3K36me2 level in the rDNA promoter induced by metformin was also inhibited by the DMS treatment (Fig. 4D). The levels of H3K36me2 in the rDNA transcribed region (Fig. S7) and H3K36me3 in both the rDNA promoter (Fig. 4D) and the rDNA transcribed regions (Fig. S7) were not changed by metformin or DMS, suggesting the specific effect of DMS on the KDM2A activation in these conditions. The reduction of cell proliferation by metformin was canceled by addition of DMS (Fig. 4E). Together, these results suggest that the reduction of succinate is required for the activation of KDM2A by metformin.

AMPK activity is dispensable for reduction of succinate induced by metformin.
Our results suggest that the activation of KDM2A by metformin requires AMPK activation (Figs. 1 and S2) and the reduction of succinate (Fig. 4). Finally, we investigated whether the AMPK activation induced by metformin was required for the decrease of succinate. A knockdown of AMPKα or the AMPK inhibitor compound C (Fig. S2) hardly affected the decrease of succinate by metformin (Fig. 5A,B). These results suggest that the AMPK activity is dispensable for the reduction of the succinate level.  30 . Metformin did not induce the KDM2A activity in glutamine-free medium where the intracellular α-KG level was reduced. Metformin treatment for four hours reduced succinate, and the replenishment of succinate with DMS blocked the metformin-mediated KDM2A activation despite AMPK activation (Figs. [2][3][4]. From these results, we propose a mechanism in which the relative amount of α-KG against succinate constitutes a switch inducing the KDM2A activity in the rDNA promoter. Together, our results suggest that metformin induces the activation of AMPK and the reduction of succinate, both of which are required for activation of KDM2A to reduce rRNA transcription (Fig. 5C).
Recently, it was reported that metformin exerted tumor suppressive effects through up-regulation of the ten-eleven translocation 2 (TET2) protein 49 . Wu et al. reported that AMPK phosphorylated and stabilized TET2. Because TET2 is a JmjC-type DNA-hydroxymethylase, it is possible that the reduction of succinate by metformin contributes to the activation of TET2. The reduction of succinate by metformin may be widely involved in the anti-proliferation effects of metformin. The metformin dose used in the above tumor model and our study here was higher than that in blood concentration in metformin-treated diabetes patients. Therefore, it is not clear whether the rationale found in the in vitro experiments can be applied to the anti-cancer activity of metformin in diabetes patients treated with metformin. Further studies are required to clarify this point.

Possible mechanisms by which metformin reduces intracellular succinate.
Our study is the first report showing the specific reduction of the intracellular succinate level without concomitant reductions of other TCA cycle intermediates including α-KG, fumarate, and malate (Figs. 3 and S6). The reduction of the succinate level occurred even under conditions that suppress AMPK activity (Fig. 5). It had been reported that metformin inhibits complex I activity 5,6 . In addition, recently it was reported that metformin inhibits the redox shuttle enzyme mitochondrial glycerophosphate dehydrogenase (mGPD) 50 . Both complex I and mGPD supply electrons to coenzyme Q (CoQ) through oxidation of NADH or FADH 2 51 . Therefore, metformin reduces the number of electrons in the electron transfer system. Meanwhile, complex II produces electrons using its succinate dehydrogenase (SDH) activity, which catalyzes the conversion of succinate to fumarate. These electrons are transferred complex III and IV in the electron transport chain to generate ATP. The reduction of electrons by metformin may enforce SDH activity to produce electrons and reduce succinate. According to this hypothesis, an increase of the fumarate level accompanying a decrease of the succinate level would occur, which is consistent to our observation in Figs. 3 and S6. www.nature.com/scientificreports www.nature.com/scientificreports/ Our results suggest that the changes in the succinate level in mitochondria control the enzyme activities in the nucleus. There are precedents in which the amount of mitochondrial succinate affects the activities of nuclear factors. A defective mutation of SDH, which increased the succinate level, stabilized hypoxia inducible factor 1 (HIF-1) through inhibition of a JmjC type enzyme HIFα prolyl hydroxylases (PHDs) 34,52 . The mutations in the catalytic sites of SDH also influenced the oxidation of 5-methylcytosine by TET 48 . Recently a mitochondrial dicarboxylate carrier (DIC) SLC25A10 in the inner membrane was suggested to mediate the equilibration of mitochondrial and cytosolic succinate pools in brown adipocytes and macrophage cells [53][54][55][56] . These observations suggest the presence of inter-organelle communication between mitochondria and the nucleus, using succinate as a messenger molecule to modulate JmjC enzyme-activities in the nucleus.
Alternatively, it is also possible that the level of succinate is initially decreased in the cytoplasm and/or nuclei by metformin. Recently, numerous metformin-binding proteins were predicted 57 . A JmjC protein KDM6A/UTX was predicted to be a metformin-binding protein, and it was suggested that metformin inhibited its demethylase activity 57 . Because the JmjC enzymes produce succinate in the demethylation process 32 , the inhibition of KDM6A/UTX demethylase activity may reduce the succinate level in the nucleus.
We demonstrated here that the levels of α-KG and succinate are pivotal factors in the regulation of the KDM2A demethylase activity by metformin. Observations of succinate levels inside cells in each organelle would further clarify the regulation mechanism of nuclear enzymes.
In the experiments for culturing in glutamine-free medium (−Gln), MCF-7 cells were cultured in RPMI-1640 medium without L-glutamine (RPMI 1640 without L−Gln, Nakalai Tesque; # 05176-25) supplemented with 10% fetal calf serum (FCS), 100 units/ ml penicillin G, and 100 µg/ml streptomycin sulfate. Glutamine (Nakalai Tesque; #16948-04; 200mM-L-Glutamine Stock Solution) was added to glutamine-free medium with the same concentration to that in the standard RPMI-1640 medium (RPMI, Nakalai Tesque; # 30264) to produce glutamine-free medium with added glutamine (−Gln/+Gln). siRNAs. Cells were transfected with siRNAs using Lipofectamine RNAiMAX (Life Technologies; # 13778-150) according to the manufacturer's instructions with suitable amounts of siRNAs for this work (150 pmol per 10 cm dis'). The siRNA specific for KDM2A was 5′-GAACCCGAAGAAGAAAGGAUUCGUU-3′, which was previously described 28 . The control siRNA, Stealth ™ RNAi Negative Control Medium GC Duplex (Life Technologies), was purchased. Three days after transfection, these cells were used for each experiment. For knockdown of AMPKα1/2, the siRNA for AMPKα1/2 (Santa Cruz Biotechnology; #sc-45312) and control siRNA-A (Santa Cruz Biotechnology; #sc-37007) were used as previously described 30 . Total RNA extraction and quantitative reverse transcription-polymerase chain reaction (qRT-PCR). Total RNA was isolated from cells using NucleoSpin RNA II kit (Takara Bio; #U0955C) according to the manufacturer's instructions. Single-strand cDNA was synthesized from total RNA (0.4 µg) by a Superscript III First-strand Synthesis system (Life Technology; # 18080-051) using random hexamers according to the manufacturer's instructions. The products were diluted up to 100 µl with distilled water, and 2.5 µl of the resultant single-strand cDNA was used as the template for qRT-PCR using a KAPA SYBR FAST qPCR Master Mix Kit (KAPA Biosystems; # KR0389) with an Mx3000P QPCR system (Agilent Technologies) or CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.). The values measured by qRT-PCR were normalized by the amounts of β-actin mRNA.

Agents. Metformin
To evaluate levels of rRNA transcription, the amounts of pre-rRNA were measured. The sets of PCR primers for amplification of the pre-rRNA were 5′-GCTGACACGCTGTCCTCTG-3′ and rDNA promoter were analyzed by ChIP assay. The results are expressed as fold changes to the values in various conditions to those without DMS and metformin. (E) DMS inhibits the reduction of cell proliferation induced by metformin. MCF-7 cells were cultured with or without 2.5 mM metformin in the presence or absence of 5 mM DMS for 2 days. The fold changes of cell numbers counted at the indicated days compared to those at day 0 are shown. All experiments were performed three times (n = 3), and the mean values with standard deviations are indicated. *P < 0.05. Immunoblotting. Cells were harvested after treatment in each experiment and extracted using an SDS-PAGE sampling buffer (4% SDS solution containing 100 mM Tris, pH 6.8, 50 mM DTT, and 20% glycerol) of suitable volume to adjust extracts to the same concentration. Cell extracts were separated by SDS-PAGE and transferred to a PVDF membrane (Millipore, Burlington, MA; #IPVH00010). After treatment with antibodies, bands were detected using an Immobilon Western system (Millipore; #WBKLS0100) as described previously 30 .

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).