Assessment of Metformin-Induced Changes in Cardiac and Hepatic Redox State Using Hyperpolarized[1-¹³C]Pyruvate

Metformin, a biguanide, is the most commonly prescribed antihyperglycemic drug for patients with type 2 diabetes and is associated with a lower risk of weight gain or hypoglycemia than other oral antihyperglycemic therapies (1). Despite long experience with this drug, the mechanisms by which metformin improves hyperglycemia have been uncertain but have been thought to involve the suppression of hepatic gluconeogenesis either by activation of the AMPK signaling pathway (2) or by inhibition of mitochondrial complex 1, leading to a decreased cellular energy charge (3). More recently, metformin was also shown to inhibit the mitochondrial form of the redox shuttle glycerophosphate dehydrogenase (mGPD) in the liver (4). One effect of the inhibition of mGPD is to increase the cytosolic redox state, thus inhibiting the redox coupled gluconeogenic conversion of glycerol and lactate into glucose and lowering plasma glucose concentration.

mGPD is one of two major mitochondrial redox shuttles (the other being the malate-aspartate shuttle) and is expressed systemically, including in the heart (5). The cardiac effects of metformin are incompletely understood but are of interest because metformin is more effective at reducing the risk of cardiovascular complications from type 2 diabetes than other antihyperglycemic therapies (6) and is considered an investigational treatment for heart failure even in the absence of diabetes (7). Whether metformin alters cardiac redox state is unknown, in part because conventional imaging techniques cannot assess redox shifts in vivo.

The lactate dehydrogenase (LDH)-catalyzed interconversion of pyruvate and lactate is an important NAD-coupled reaction, which can be assessed noninvasively using hyperpolarized [¹³C]pyruvate magnetic resonance (MR) technology (8). Following injection, enzymatic conversion of hyperpolarized [¹³C]pyruvate into lactate results in the incorporation of the hyperpolarized ¹³C label into the [¹³C]lactate pool. The other major fate of hyperpolarized [1-¹³C]pyruvate is decarboxylation via the pyruvate dehydrogenase (PDH) enzyme complex leading to the production of ¹³CO₂ and [¹³C]bicarbonate, which provides a novel index of overall carbohydrate metabolism (9). Although hyperpolarized MR technology has been used to study redox state both in cancer cell lines in vitro (10) and tumors in vivo (11,12), its sensitivity to changes in cardiac and hepatic redox state is unknown. Hyperpolarized MR shows promise as a clinical tool in diabetes-related heart disease in humans, as pharmacological restoration of PDH flux was associated with improved cardiac function in a rodent model (13).

In this study, we used hyperpolarized [¹³C]pyruvate MR spectroscopy (MRS) to determine whether metformin caused changes in pyruvate metabolism in rodents using a sequential spectroscopic acquisition sequence to acquire information from both the heart and liver.

Male Wistar rats were used in this study and all procedures were carried out in accordance with the U.K. Animals (Scientific Procedures) Act 1986. Rats were provided with either 250 mg/kg/day metformin in drinking water (22 mmol/L) or water without metformin for 4 weeks. Both metformin-treated and untreated water were flavored with the same berry-flavored concentrate (SodaStream Waters Zero Cranberry Raspberry; SodaStream, Lod, Israel) to mask any flavor of the metformin, and water consumption was not different between the two groups (control 23.9 ± 1 mL/rat/day vs. metformin 26.3 ± 0.9 mL/rat/day, P = ns by unpaired t test). For subsequent studies of the acute effects of metformin, an intravenous infusion of either 50 mg metformin dissolved in 0.9% saline or a saline infusion of the same volume was administered via a tail-vein catheter 45 min prior to the administration of hyperpolarized pyruvate.

Animals were anesthetized with vaporized 3.5% isoflurane (using 2.5 L/min O₂ as a carrier gas), reduced to 2% isoflurane for anesthesia maintenance. A custom-built MR cradle incorporating a warm air blower system maintained body temperature at 37°C throughout the experiments; electrocardiographic monitoring was performed throughout to allow observation of stable heart and respiration rates during the experiments.

[1-¹³C]pyruvic acid was hyperpolarized in a prototype hyperpolarizer as described previously (8). Prepolarized pyruvate was administered as a bolus of 1 mL of an 80 mmol/L solution over 10 s. Slice-selective spectra were acquired interleaved over the following 2 min from two axial slabs covering the heart and the liver (350 µs sinc pulse, 1-cm-thick excitation, 8 kHz bandwidth, 15° flip angle, repetition time 1 s, both slices electrocardiogram gated). Experiments were performed on an Agilent 7T preclinical horizontal bore scanner with a volume transmit/two-channel surface receive array. For imaging experiments, a specifically designed MR sequence with spectral-spatial radiofrequency pulses with echo planar imaging gradients was used, as previously described (14).

Under light isoflurane anesthesia, systolic function was assessed from parasternal M-mode views, while diastolic function was assessed using pulsed wave Doppler examination directed to the mitral valve inflow (E wave) and tissue Doppler imaging of the medial mitral annulus (to derive E′).

Heart and liver tissue was harvested under isoflurane anesthesia and rapidly snap frozen using aluminum tongs cooled to the temperature of liquid nitrogen. Blood samples were taken and placed in heparinized tubes before centrifugation at 13,000 rpm for 10 min at 4°C. The plasma was then frozen in liquid nitrogen for later biochemical analyses. In order to avoid any potentially confounding effect of the hyperpolarized pyruvate administration on systemic redox state, all samples were obtained at least 2 days following the administration of hyperpolarized pyruvate (with either continued administration of oral metformin or a second metformin infusion depending on the study group). Because plasma levels of pyruvate have been shown to normalize within 10 min of administration of 1 mL 80 mmol/L hyperpolarized pyruvate (9), it is highly unlikely that any residual effect on systemic redox state would be present at this point.

Approximately 100 mg frozen tissue was homogenized using a rotor stator homogenizer and metabolites extracted in 6% perchloric acid. To maximize the stability of metabolites, the perchloric acid extract supernatants were neutralized prior to deproteination using 10 kDa molecular weight cutoff filtration and samples were maintained on ice whenever possible.

Lactate and β-hydroxybutyrate concentrations were determined using an automated assay system (ABX Pentra 400; Horiba Medical), pyruvate concentration was determined using a fluorescent assay kit (Cambridge Bioscience Ltd.), free [NAD⁺]:[NADH] ratio was determined using a colorimetric assay based on the enzyme cycling method (Cambridge Bioscience Ltd.), and acetoacetate concentration was derived using a colorimetric assay kit (Abcam, Cambridge, U.K.). Plasma lactate levels were determined using an automated assay system (ABX Pentra 400; Horiba Medical).

Approximately 30 mg frozen heart or liver tissue was homogenized in lysis buffer and RNA extracted using RNeasy mini columns (Qiagen). The integrity of the extracted RNA was confirmed by NanoDrop spectrophotometry. RNA was reverse transcribed using a High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher). TaqMan primers for rodent Slc22a1 (Rn00562250_m1), Ldha (Rn00820751_g1), and GPD2 (Rn00562472_m1) genes were used in 20 μL quantitative PCR (qPCR) reactions run in duplicate using a StepOnePlus qPCR system. Gene expression was normalized to expression of 45 s preribosomal RNA (Rn03928990_g1) and expressed according to the 2^-ΔΔCT method.

Data are reported as mean ± SEM. Statistical significance was determined as P ≤ 0.05, and as all differences were compared with independent control groups, statistical significance was calculated using unpaired two-tailed Student t tests.

Treatment with oral metformin for 4 weeks had no effect on either cardiac or hepatic PDH flux (cardiac [1-¹³C]bicarbonate + ¹³CO₂:[1-¹³C]pyruvate ratio 0.04 ± 0.01 vs. 0.057 ± 0.006, P = 0.19 and hepatic [1-¹³C]bicarbonate + ¹³CO₂:[1-¹³C]pyruvate ratio 0.054 ± 0.007 vs. 0.045 ± 0.006, P = 0.29) (Fig. 1C). However metformin increased the [1-¹³C]lactate:[1-¹³C]pyruvate ratio in both the heart (0.27 ± 0.06 vs. 0.10 ± 0.01, P = 0.02) (Fig. 1B) and liver (0.87 ± 0.21 vs. 0.36 ± 0.04, P = 0.04) and also modestly increased the plasma lactate concentration without leading to lactic acidosis (4.1 ± 0.3 vs. 2.4 ± 0.3 mmol/L, P = 0.002). Despite the change in lactate metabolism, metformin had no effect on either systolic or diastolic cardiac function (left ventricular ejection fraction 75 ± 2% vs. 74 ± 3%, P = 0.74 and E/E′ ratio 15 ± 1 vs. 16 ± 1, P = 0.47) (Fig. 1D).

In order to determine whether the same effect of metformin on the cardiac and hepatic [1-¹³C]lactate:[1-¹³C]pyruvate ratio could be recapitulated acutely, we next performed hyperpolarized [1-¹³C]pyruvate spectroscopy 45 min following a single intravenous infusion of either metformin (50 mg) or an equal volume of saline. Metformin again increased the [1-¹³C]lactate:[1-¹³C]pyruvate ratio in both the heart (0.30 ± 0.04 vs. 0.14 ± 0.02, P = 0.002) (Fig. 2A) and liver (0.65 ± 0.06 vs. 0.40 ± 0.04, P = 0.003) and again increased the plasma lactate concentration (2.9 ± 0.6 vs. 1.0 ± 0.2 mmol/L, P = 0.02). Acute metformin infusion also had no effect on either cardiac or hepatic PDH flux (cardiac [1-¹³C]bicarbonate + ¹³CO₂:[1-¹³C]pyruvate ratio 0.043 ± 0.009 vs. 0.058 ± 0.005, P = 0.15 and hepatic [1-¹³C]bicarbonate + ¹³CO₂:[1-¹³C]pyruvate ratio 0.044 ± 0.006 vs. 0.056 ± 0.006, P = 0.18).

In order to exclude the possibility that the increase in cardiac [1-¹³C]lactate signal measured by spectroscopy might reflect metabolic shifts in the adjacent blood pool or skeletal muscle, we next investigated the effects of metformin using a hyperpolarized MRI sequence to produce a metabolic “map.” Almost all of the lactate signal localized to the left ventricular myocardium, whereas the large majority of the pyruvate signal localized to the ventricular cavity, supporting a direct effect of metformin on the heart (Fig. 2C).

In the liver (15) and kidney (16), organic cation transporters (OCTs) are the major mediators of cellular uptake of metformin. We confirmed that genes encoding both OCT1 (Slc22a1) and mGPD (Gpd2) were expressed in the heart as well as the liver by qPCR and that expression of these genes was unchanged by metformin infusion (Fig. 2D). Similarly, expression of the Ldha gene encoding LDH was also unaffected by metformin infusion. These findings, together with the rapidity of the effect, suggest that nontranscriptional mechanisms explain the increase in hyperpolarized [1-¹³C]lactate signal.

Because redox shuttle inhibition is a mechanism by which metformin reduces hepatic gluconeogenesis, we next questioned whether the increase in heart and liver [1-¹³C]lactate:[1-¹³C]pyruvate ratio caused by metformin might reflect a shift in cellular redox state (4). The [NAD⁺]:[NADH] ratios in both heart and liver whole-cell lysates were unchanged by either longer-term or acute metformin treatment (cardiac and hepatic [NAD⁺]:[NADH] ratios 5.7 ± 0.6 vs. 8 ± 2, P = 0.18 and 0.8 ± 0.1 vs. 0.7 ± 0.1, P = 0.64, respectively, following 4-week treatment and 6.3 ± 0.1 vs. 7.5 ± 0.8, P = 0.14 and 1.9 ± 0.2 vs. 2.7 ± 0.6, P = 0.14 following single infusion) (Fig. 3A). However, metformin can cause a shift in the cytosolic redox state with either an increase (17) or decrease (4) in the mitochondrial redox state depending on the dosing strategy, and the total cellular [NAD⁺]:[NADH] ratio may not be a reliable measure of compartmental redox shift. We therefore measured intracellular [lactate]:[pyruvate] ratios as a surrogate for cytoplasmic redox state (18) and identified that metformin increased the intracellular [lactate]:[pyruvate] ratio, consistent with a shift in heart and liver cytosolic redox state and paralleling the changes in the hyperpolarized [1-¹³C]lactate:[1-¹³C]pyruvate ratio (cardiac and hepatic [lactate]:[pyruvate] ratios 46 ± 6 vs. 30 ± 6, P = 0.04 and 60 ± 9 vs. 27 ± 3, P = 0.002, respectively, following 4-week treatment and 50 ± 8 vs. 25 ± 7, P = 0.02 and 72 ± 12 vs. 16 ± 8, P = 0.002 following single infusion) (Fig. 3B). Metformin in this study caused no change in the [β-hydroxybutyrate]:[acetoacetate] ratio, an index of mitochondrial redox state, in cardiac or hepatic tissue following a single dose or in cardiac tissue following longer-term dosing, though it did increase the [β-hydroxybutyrate]:[acetoacetate] ratio in liver tissue following longer-term dosing (Supplementary Data).

Cellular redox state is an important determinant of metabolism and function as numerous enzyme-catalyzed reactions exist in redox couples. This study demonstrated that metformin treatment significantly increased the production of [1-¹³C]lactate following a hyperpolarized [1-¹³C]pyruvate infusion in both the heart and liver, a finding that was paralleled by an increase in the cytosolic, but not whole-cell, redox state. In addition to demonstrating the ability of hyperpolarized MR to detect cytosolic redox shifts in the heart and liver, the subcellular compartmentalization of metformin-induced redox shifts is consistent with the view that the interconversion of hyperpolarized [1-¹³C]pyruvate and [1-¹³C]lactate primarily reflects a cytosolic reaction, despite the recognition of mitochondrial capacity for lactate and pyruvate oxidation (19).

Metformin caused changes in the [lactate]:[pyruvate] ratio within 45 min of administration without upregulating LDH, supporting a rapid, redox-dependent mechanism (4). However, metformin has a complex pharmacokinetic profile and crosses plasma membranes slowly due to a net positive charge (20). As a result, metformin achieves higher hepatic concentrations following longer-term oral treatment than are possible with a single intravenous treatment at physiological doses (21) and leads to higher hepatic concentrations than are reached in the heart (16,22). The magnitude of the increase in the hyperpolarized [¹³C]lactate:[¹³C]pyruvate ratio in the liver was greater with longer-term treatment, and it is tempting to speculate that this may reflect higher bioavailability, although the magnitude of the increase in liver [lactate]:[pyruvate] ratio was similar for both routes of metformin administration.

Although previously described in the liver, a metformin-induced increase in cardiac cytosolic redox state has not been previously recognized. Metformin has conventionally been thought to improve cardiovascular outcomes in type 2 diabetes primarily by systemic reduction of hyperglycemia, although the elucidation of direct cardiac effects is of interest as the risks and benefits of metformin treatment in heart failure with or without diabetes remain contentious (23). Although the cardiac redox shift was not associated with a change in either systolic or diastolic function in this study, future work should be directed to understanding the effect of metformin in disease states, including models of type 2 diabetes.

Our finding that metformin does not alter either cardiac or hepatic PDH flux is consistent with the finding that the major mechanism by which metformin reduces hyperglycemia is through the inhibition of gluconeogenesis rather than potentiation of peripheral glucose utilization (24). The finding is also consistent with data from [¹⁸F]-2-fluoro-2-deoxy-d-glucose positron emission tomography studies demonstrating that metformin does not increase myocardial (25) or skeletal muscle (26) glucose uptake. PDH is a mitochondrial enzyme complex, which is partially regulated by redox state, though a phosphorylation/dephosphorylation cycle is also important (27). With the exception of livers following longer-term dosing, mitochondrial redox states were generally unaffected by metformin in this study, which is consistent with the understanding that both redox-dependent and redox-independent mechanisms determine PDH flux in vivo.

Hyperpolarized MR has high translational potential with early human studies already under way (28). The ability to noninvasively assess metformin-induced redox shifts in vivo in humans would be an important advance, as animal studies using metformin have been limited by differences in interspecies dosing and bioavailability profiles (21). It remains unclear to what extent inhibition of mGPD would be sufficient to cause cellular redox shifts in humans, where the malate-aspartate shuttle is the quantitatively dominant shuttle mechanism (29). Such an approach would also allow an assessment of cellular energy charge by combining ³¹P spectroscopy with hyperpolarized [1-¹³C]pyruvate.

In summary, metformin treatment led to significant increases in the hyperpolarized [¹³C]lactate:[¹³C]pyruvate ratios in both the heart and liver, a finding that appears to reflect a shift in the cytosolic redox state in these organs. These findings suggest that hyperpolarized pyruvate MRS is sensitive to metformin-induced changes in redox biology, suggest that metformin has a previously unrecognized effect on cardiac redox state, and may influence the design of future human studies using hyperpolarized ¹³C MR.

Acknowledgments. The authors are grateful to Vicky Ball (Department of Physiology, Anatomy and Genetics, University of Oxford) for excellent technical assistance.

Funding. A.J.M.L. and S.N. acknowledge funding from the British Heart Foundation Oxford Centre for Research Excellence. J.J.J.M. acknowledges funding from Engineering and Physical Sciences Research Council Doctoral Training Centre and Prize Fellowship (EP/M508111/1) grants. L.C.H. acknowledges funding from Diabetes UK (grant 11/0004175). D.J.T. acknowledges funding from the British Heart Foundation (grants FS/14/17/30634 and RG/11/9/28921).

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

Author Contributions. A.J.M.L., J.J.J.M., and C.M. researched data. A.J.M.L. wrote the manuscript. O.J.R., S.N., and L.C.H. reviewed and edited the manuscript and contributed to discussions. D.J.T. conceived of and designed the study and reviewed and edited the manuscript. D.J.T. 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 analysis.