[1] 9w Cholinergic and metabolic effects of metformin in mouse brain

[1] 166w Metformin is widely used as a first-line treatment for type 2 diabetes, but central effects of metformin have received little attention. When metformin (200 mg/kg i.p.) was administered to C57Bl6 mice, metformin concentration in cerebrospinal fluid peaked at 29 μM after 30 min but dropped quickly and was low at 90 min. In mouse hypothalamus sampled by microdialysis, systemically administered metformin caused minor and transient increases of acetylcholine, glucose and lactate while choline levels decreased. When metformin (0.2-10 mM) was locally infused via retrodialysis, there was a short-lasting increase of acetylcholine in the hypothalamus. Extracellular lactate levels in hypothalamus showed a massive increase upon metformin infusion while glucose levels decreased. In isolated mitochondria of mouse brain, metformin inhibited oxygen consumption and the activity of complex I. Inhibition of mitochondrial respiration likely explains lactate formation in the brain during metformin infusion which may cause lactic acidosis during metformin intoxication. The changes of cholinergic activity in the hypothalamus may be associated with appetite suppression observed during metformin treatment.

[1] 184w Metformin is the world ′ s most popular drug for the treatment of type 2 diabetes (Bailey et al., 1996). It lowers blood glucose mainly by reducing hepatic gluconeogenesis and, to some degree, by increasing sensitivity to insulin. Several molecular mechanisms have been suggested for metformin ′ s action (An and He, 2016). The drug was shown to cause a partial inhibition of complex I of the mitochondrial electron transport chain (El-Mir et al., 2000;Owen et al., 2000). This effect also causes a slowing down of the tricarboxylic acid cycle and an increased formation of lactic acid which, at high metformin concentrations, can cause lactic acidosis (van Berlo-van de Laar et al., 2011). Metformin also increases AMP kinase activity, thereby reducing lipogenesis (Zhou et al., 2001), and it inhibits mitochondrial glycerol-3-phosphate dehydrogenase, thereby shifting the mitochondrial and cytosolic redox potentials and favoring lactate formation (Madiraju et al., 2014(Madiraju et al., , 2018)). In addition to its beneficial effects in diabetes, metformin improves cardiovascular disease and it is considered as a potential agent to suppress tumor growth and to increase longevity (Viollet et al., 2012).

[2] 246w The pharmacokinetics of metformin are well known. In humans, metformin has an oral bioavailability of 50-60 %; at therapeutic doses, it reaches plasma concentrations of 10-40 μM with a half-life of 1.5-5 hours (Graham et al., 2011). As a cationic drug, its distribution is strongly influenced by the presence of organic cation transporters OCT 1-3 (SLC22A1-A3) (Shu et al., 2008). The high expression of OCT1 in small intestine and liver is responsible for the accumulation of metformin in these tissues whereas metformin is excreted via the kidney which contains both OCT1 and 2. Central effects of metformin have been much less investigated. Metformin is known to cross the blood-brain barrier to a small degree. Lv et al. (2012) reported that CSF levels of metformin are approx. 4% of the plasma levels. Łabuzek et al. (2010) measured CSF levels of metformin in the 30-40 μM range when rats were dosed orally with 150 mg/kg. While plasma levels were initially higher than CSF levels, both concentrations were similar six hours after oral dosing. Within in the brain, some differences in metformin concentrations were reported (Łabuzek et al., 2010). These may be due to the distribution of the OCTs, with OCT1 widely distributed in neurons and OCT3 in glial cells whereas OCT2 has a more restricted expression (Koepsell et al., 2007). In brain endothelia, OCTs are expressed at low levels, and this lack of OCTs likely explains the slow entry of metformin into the brain (Chaves et al., 2020).

[3] 243w Understanding the brain permeability of metformin is important because the drug has been proposed to exert central actions, e.g. in cerebral ischemia (Leech et al., 2019) and epilepsy (Nandini et al., 2019). Metformin was shown to activate Nrf2-mediated pathways in neurons and to reduce inflammatory responses induced by ischemia or smoking (Ashabi et al., 2015;Kaisar et al., 2017). Some of its antidiabetic actions may also involve central mechanisms (Ruegsegger et al., 2019;Docrat et al., 2020). Importantly, metformin ′ s action in the brain may involve cholinergic mechanisms. Metformin exerts an appetite-suppressing effect which leads to weight loss in diabetic patients (Bailey and Turner, 1996). In rodents, metformin was shown to reduce appetite after i.c.v. administration by a hypothalamic mechanism (Kim et al., 2013). Cholinergic neurons regulate food intake in the dorsomedial hypothalamus via muscarinic receptors (Jeong et al., 2017), while nicotine suppresses appetite by activating nictonic cholinergic receptors. We have previously shown that acetylcholine levels in the hypothalamus increase during the anticipation of food in fasted but not sated mice (Thinnes and Klein, 2018). Moreover, a cholinergic mechanism, namely inhibition of acetylcholinesterase, has been implicated in metformin ′ s effects on cognition in experimental models of neurodegeneration (reviewed by Markowicz-Piasecka et al., 2017). In light of the vivid discussion of metformin ′ s central effects, we have designed the present study to investigate the effects of peripherally and centrally administered metformin on acetylcholine levels and on glucose and lactate levels in mouse hypothalamus.

[1] 75w After i.p. injection of 200 mg/kg metformin, we measured blood plasma levels of 882 μM after 6 min (not illustrated) and 173 μM after 15 min (Fig. 1A). CSF was withdrawn after 30 min, and the average metformin concentration of six pooled CSF samples was 29.0 μM (Fig. 1A). In microdialysates, the highest value was 2.47 μM after 30 min (Fig. 1B) which suggests that metformin recovery in the dialysis probe was approx. 8.5 %.

[2] 116w In mice that were fasted for 12 h prior to the experiment, the extracellular concentration of acetylcholine (ACh) in the hypothalamus was approx. 4 nM (Fig. 2A); it increased briefly after metformin administration, followed by a long-lasting decrease. Choline levels (Fig. 2B) were around 700 nM; after metformin, they decreased to approx. 600 nM. When food was offered after 90 min, ACh levels increased whereas choline levels decreased significantly (Fig. 2). In a separate experiment in non-fasted mice, basal ACh and choline levels were slightly higher than in fasted mice (about 6 nM and 900 nM, resp.); here, metformin administration caused a significant decrease of ACh levels and a concomitant increase of choline (Suppl. Fig. 1).

[3] 46w Energy metabolites in the hypothalamus showed minor responses to peripheral metformin administration (Fig. 3). Both glucose and lactate levels showed a brief and transient increase that did not reach significance; pyruvate was unchanged. Supply with food caused a significant increase of hypothalamic glucose levels (Fig. 3).

[4] 113w Infusion by retrodialysis of metformin caused brief but significant increases of hypothalamic ACh levels whereas extracellular choline concentrations decreased (Fig. 4). At 1 mM metformin, ACh increased 2-3fold whereas 10 mM metformin caused a stronger but highly variable increase of ACh levels (Fig. 4A). Hypothalamic glucose levels were unresponsive to metformin infusion at 0.2-1 mM but decreased significantly at 2-10 mM (Fig. 5). Lactate and (to a lesser degree) pyruvate levels were increased by metformin in a concentration-dependent manner: they were doubled by 1 mM metformin, tripled by 2 mM (Fig. 5A) and rose ten-fold during 10 mM metformin infusion (Fig. 5B). The lactate-pyruvate ratio increased at 2-10 mM metformin (Suppl. Fig. 2).

[5] 69w Mitochondrial oxygen consumption was measured in mitochondria isolated from frontal mouse brain. We did not see significant changes when mitochondira were incubated with 0.1 mM metformin for 15 min, or with 1 mM metformin for 6 min (data not shown), but 1 mM metformin incubated for 15 min caused a significant inhibition of complex I activity (Fig. 6). Complex IV was also affected, but not complex II (Fig. 6).

[1] 27w The present study shows small but measurable effects of metformin on the cholinergic system in the hypothalamus, and a strong effect on lactate levels after intracerebral application.

[2] 210w With respect to pharmacokinetics and brain permeability, we measured plasma and CSF levels of metformin and, for the first time, present microdialysis data. In agreement with previous studies (Łabuzek et al., 2010;Lv et al., 2012), we observed a short half-life of metformin in plasma and a moderate brain permeability of the drug which reached a CSF concentration of 29 μM after application of the high dose of 200 mg/kg metformin i.p. Our microdialysis data show that metformin levels in the brain drop quickly after 60 min, however, we do not know whether this is due to cellular uptake of metformin or to removal from the brain. Metformin is a cationic drug, and the mode of entry into the brain is unknown; it may involve cation transporter localized at the blood-brain barrier (BBB) (Inazu, 2019). Behind the BBB, neurons and glial cells broadly express OCT1 and OCT3, cation transporters that accept metformin as a substrate. OCT2, another metformin transporter, is expressed in the brain at lower density (Koepsell et al., 2007). Łabuzek et al. ( 2010) have shown that maximum levels of metformin in the brain occur later than plasma peaks. They also reported a non-homogeneous distribution of metformin in the brain that may be due to variations of OCT1/2 expression.

[3] 137w As summarized by Markowicz-Piasecka et al. (2017), metformin is suggested to have an influence on the cholinergic system in the brain. We here show that peripherally applied metformin has but a small effect on ACh levels in the hypothalamus; there was a significant decrease, however, when the mice were not fasted. In fasted mice, ACh was decreased to a minor degree, but presentation of food increased ACh (see also Thinnes and Klein, 2018). An increase of ACh was seen when metformin was given by retrodialysis; it is not known whether this increase represents increased ACh release or inhibition of acetylcholinesterase (AChE). A cholinergic pathway that controls appetite has been identified in the hypothalamus (Jeong et al., 2017) and it is possible that metformin ′ s appetite-suppressing effects are partly mediated by an action on hypothalamic cholinergic pathways.

[4] 183w Systemically applied metformin, even at the high dose of 200 mg/kg, did not appreciably affect glucose or lactate levels in the brain. This may be due to slow entry into the brain since local administration of metformin by retrodialysis caused a massive, up to ten-fold increase of lactate levels in a dose-dependent manner. This increase islikely due to lactate formation secondary to mitochondrial complex I inhibition, an effect of metformin that we here demonstrate for isolated forebrain mitochondria. Inhibition of complex I by metformin was previously shown in liver mitochondria (El-Mir et al., 2000;Owen et al., 2000), and increased formation of lactate is plausible due to reduction of pyruvate caused by increased cellular NADH levels. Formation of lactate upon exposure to metformin, albeit at high doses, was also observed in cultured neurons (Blumrich and Dringen, 2017) and astrocytes (Westhaus et al., 2017;Hohnholt et al., 2017). It should be noted that release of lactate by astrocytes is a typical feature of the astroglial-neuronal lactate shuttle (Pellerin and Magistretti, 2012), and the increase of lactate is likely due to astroglial release (Westhaus et al., 2017).

[5] 383w It is very difficult to consider the potential relevance of central lactate formation by metformin for human health. For systemic actions of metformin, we have used a dose that causes a plasma level of metformin around 1 mM which causes a CSF level of up to 30 μM. Therapeutic drug levels in humans are 15-40 μM (Graham et al., 2011) but plasma levels of up to 300 μM have been observed in metformin-associated lactic acidossis (van Berlo-van de Laar et al., 2011;Vecchio et al., 2014). Metformin levels in the brain of humans are not known. For central administration, we chose 0.2-10 mM metformin in the present study. While these are high concentrations, it must be kept in mind that only a fraction of the drug leaves the microdialysis probe (probably around 8.5 %). The metformin released into the extracellular space can be taken up into cells (depending on OCT transporters) and may reach high levels intracellularly, or it may be removed from the brain into venous blood. Within the cell, metformin (as a cationic drug) likely accumulates within mitochondria which have a very negative membrane potential, but this may be a delayed process. In our hands, 15 min of incubation with 1 mM metformin were required to elicit a significant complex I inhibition. Other groups reported pre-incubation times of metformin with tissues of up to 24 h (e.g., Owen et al., 2000). No information is available about mitochondrial transporters for metformin (Vial et al., 2019). In hepatocytes, Wilcock et al. (1991) observed that about 10 % of cellular metformin was associated with mitochondria whereas more than 80 % were present in cytosol. From the limited information available for metformin concentrations in the brain, we estimate that during infusion of 10 mM metformin (which causes massive lactate formation), high micromolar levels will be present in the extracellular space of hypothalamus, from which it is slowly taken up especially into astrocytes to increase aerobic glycolysis, causing a reduction of glucose and formation of lactate (Westhaus et al., 2017). During metformin overdose in humans, when plasma levels of 300 μM can be reached (Vecchio et al., 2014), metformin may slowly accumulate in brain cells and mitohondria and possibly cause lactic acidosis in brain, a state that is known to affect synaptic function (Walz and Harold, 1990).

[1] 79w The present study shows that metformin is brain-permeable and that it causes a measurable effect on the cholinergic system in the hypothalamus. We speculate that metformin may exert an influence on appetite control via cholinergic fibers in the hypothalamus. Moreover, metformin inhibits respiration in mitochondria of the brain and may cause lactate accumulation under conditions of prolonged increases of plasma metformin levels, e.g. in patients with renal insufficiency. Central lactic acidosis in situations of metformin overdose should be considered.

[1] 10w Chemicals were purchased from Sigma/Merck (Darmstadt, Germany) unless otherwise stated.

[2] 89w We used 9-15 week old male C57Bl/6JRj mice (25-30 g) from Janvier Labs (Saint Berthevin, France). Mice were housed (five per cage) in a temperature (22 • ±3 • C) and humidity (55 % ± 10 %) controlled room with a 12/12-h reversed light-dark schedule. Altromin chow (Type 1320, Lage, Germany) and water were available ad libitum. All procedures were conducted to minimize animal suffering in accordance with German and Eurpean law (EU directive 2010/63/EU). The study was registered and approved by the responsible government agency (Regierungspräsidium Darmstadt, Germany).

[3] 115w For surgery, animals were anesthetized with isoflurane (induction dose 5 %, maintenance dose 2 % v/v) in synthetic air (Air Liquide, Düsseldorf, Germany). Self-constructed, Y-shaped, concentric dialysis probes with a molecular weight cut-off of 10 kDa were stereotaxically implanted into the hypothalamus with the following coordinates (from bregma): AP -1.5 mm, L + 0.5 mm, DV -5.6 mm according to Franklin and Paxinos (Franklin and Paxinos, 1997). Glass ionomer eluting cement (PermaCem Smartmix Dual, Dental Milestone, Hamburg, Germany) was used to fix the probe on the skull (for further details, see Stein et al., 2019). Probes were implanted at least 18 h before each experiment to allow neurotransmitter recovery to stabilize (Sumbria et al., 2011).

[4] 122w On the next day, microdialysis was performed with a perfusion fluid (aCSF containing 147 mM NaCl, 4 mM KCl, 1.2 mM CaCl 2 and 1.2 mM MgCl 2 ). Instead of an outlet tube we used a tip (Ultratip, Greiner bioone, Kremsmünster, Austria) to collect the dialysate. After 20 min equilibration time, the dialysate collection was started. The perfusion rate of the microinjection pump was 2 μL/min. The collection intervals were 7.5 min. After sampling for 30-60 min mice received either an intraperitoneal injection (i.p.) of metformin or a local metformin infusion via the dialysis probe ("retrodialysis"). Subsequently, acetylcholine (ACh), choline and the energy metabolites glucose, lactate and pyruvate were measured. Data are given as absolute levels not adjusted for probe recovery.

[5] 79w For retrodialysis (Figs. 4 and 5), dialysate was first collected for the measurement of basal values. After 45 min, the perfusion solution was replaced by a perfusion solution containing 0.2 mM metformin for 75 min. The perfusate was then changed again to a 2 mM metformin perfusion solution for further 75 min. On the following day, the experiment was repeated with a 1 mM and a 10 mM metformin perfusion solution. The estimated recovery rate of metformin is 8%.

[6] 77w At the end ofthe experiment, the animals were briefly anesthetized (isoflurane) and were decapitated, trunk blood was dripped into icecooled EDTA tubes (Microvette®) and immediately centrifuged (4 • C, 20 min, 1.900 g). The supernatant (plasma) was collected and snapfrozen in liquid nitrogen. The plasma was stored at -80 • C until further processing. Cerebrospinal fluid was withdrawn from deeply anesthetized animals as described by Liu and Duff (2010) and stored at -80 • C until measurement.

[7] 92w Acetylcholine and choline concentrations in dialysates were determined by HPLC-ECD using an Eicom HTEC-500 system (Kyoto, Japan) consisting of degasser, low-speed pump, pre-and separation column, enzyme reactor and electrochemical detector with a platinum electrode operating at 0.5 V. The mobile phase contained 50 mM KHCO 3 , 134.3 mM EDTA-2Na and 1.64 mM sodium decane-1-sulfonate in Roti-solV® HPLC gradient grade water and was set to a flow rate of 150 μl/ min. The enzyme reactor carried immobilized acetylcholine esterase and choline oxidase. The detection limit of this system was 5 fmol/10 μl.

[8] 44w Glucose, lactate and pyruvate concentrations in microdialysates were determined by a colorimetric method (530 nm) using an ISCUSflex Microdialysis Analyzer (M dialysis AB, Solna, Sweden). Samples were vortexed and centrifuged at 9000 rpm for 30 s. before placing them in the ISCUSflex Microdialysis analyzer.

[9] 151w For the determination of metformin, microdialysis samples were obtained as described above. Cerebrospinal fluid (CSF) was obtained as described by Liu and Duff (2010); samples of six animals were pooled, frozen in nitrogen and stored at -80 • C until analysis. Samples were diluted in 40 ml of pure water (Milli Q®) and internal standard (Buformin) was added. SPE cartridges (Strata-X-CW 6 ml) were conditioned 3 times with 3 ml of methanol and 3 times with 3 ml of Milli Q®. The diluted samples were then placed over the SPE cartridges and propelled away with nitrogen. Elution was carried out with 10 ml of a solvent mixture consisting of acetonitrile and methanol (50:50, v/V) + 2% formic acid. The eluates were taken to dryness and taken up with 20 μl MBTFA and 200 μl acetonitrile (2 h at 60 • C.). 1 μl of the sample was injected for GC-MS analysis.

[10] 44w Data are represented as means with standard error of the mean (SEM) of N experiments. Statistical significance was tested by repeated measures ANOVA with Dunnett's post-test (Figs. 2345) or by unpaired Student's t-test (Fig. 6). Statistical analyses were performed with GraphPad Prism (version 5.0).

[1] 113w After decapitation, the brain was immediately dissected from the skull, the cerebellum was removed and the brain divided into hemispheres. From each hemisphere the frontal part of the brain (≈100 mg) was separated and homogenized in 2 mL MiR05. In addition, a protease inhibitor cocktail (PI) was added to the medium (cOmplete Tablets EASY pack, Roche, Mannheim, Germany). The homogenate was centrifuged twice to remove all cell debris (1.400 g, 7 min, 4 • C). The purified supernatant was then centrifuged again (10.000 g, 5 min, 4 • C), the resulting pellet containing the mitochondria was resuspended in 1000 μL MiR05+PI and centrifuged once again (1.400 x g, 3 min, 4 • C).

[2] 48w Finally, the supernatant was centrifuged one more time (10.000 g, 5 min, 4 • C) and the pellet resuspended in 250 μL MiR05+PI. To investigate the acute effect of metformin on mitochondria, an aliquot of the mitochondrial suspension was incubated with a metformin solution (end concentration 1 mM).

[3] 319w Untreated and metformin-exposed mitochondria were run in parallel chambers of the respirometer. Each chamber was filled with 2.4 mL MiR05 medium according to manufacturer ′ s instructions and kept at 37 • C with constant stirring (750 r.p.m.). After 30 min equilibration and subsequent air calibration, 80 μL of the mitochondrial suspensions were injected into the closed chamber. The remaining mitochondria were frozen in liquid nitrogen for later citrate synthase determination according to Kuznetsov et al. (2002). After equilibration, a solution containing pyruvate (5 mM) and malate (1 mM), two substrates linked to complex I (CI), was injected into the chamber (LEAK-state, non-phosphorylating resting state). Then, ADP (2 mM) was added to stimulate the oxidative phosphorylation (OXPHOS; CI OXPHOS ; ADP-stimulated and CI-linked respiration). To induce the full ADP-stimulated respiration, succinate (10 mM), a CII-linked substrate, was added (OXPHOS capacity). To verify the integrity of the outer mitochondrial membrane, cytochrome c (10 μM) was added; mitochondria whose respiration increased by more than 15 % upon cytochrome c addition were discarded. The maximum capacity of the electron transfer system (ETS) was determined by the stepwise titration of the uncoupler FCCP (state E). To see the isolated CII respiration, the complex I inhibitor rotenone (2.5 μM) was added (CII-linked substrate state, uncoupled). After inhibition of complex III by antimycin A (2.5 μM), the residual oxygen consumption (ROX; ROX-state) remains, which is used to correct the mitochondrial respiration states. Ascorbate (2 mM) and tetramethyl-phenylendiamine (TMPD, 0.5 mM) are artificial electron donors that induce maximum cytochrome c-oxidase (CIV) respiration by reducing cytochrome c. Ascorbate regenerates TMPD and is therefore injected first. At the end of the experimental run CIV is inhibited by a high concentration of sodium azide (120 mM). The chemical background as well as ROX remains. To obtain the CIV activity this value has to be subtracted from the total measured oxygen flux (for further details, see Schwarzkopf et al., 2015).