Metformin is the drug of choice for adults with type 2 diabetes 1 . It is a safe drug 2 although lactic acidosis rarely develops as a side effect 3 4 5 . In most of the cases, drug use is only coincidental and lactic acidosis is due to concomitant hypoxia, tissue hypoperfusion or liver failure (metformin-associated lactic acidosis) 6 . Less frequently, no other major risk factor can be identified and drug accumulation, prompted by renal failure for instance, is the most probable cause of the syndrome (metformin-induced lactic acidosis) 7 .

The pathogenesis of metformin-induced lactic acidosis remains unclear. Some authors even question whether metformin per se can cause lactic acidosis 8 . Some others mainly attribute it to changes in liver lactate metabolism 9 . Metformin readily accumulates in hepatocytes, that express the organic cation transporter 1 10 , and inhibits their mitochondrial respiration in a dose-dependent manner 11 12 13 . As a possible result, liver lactate output increases 11 12 13 and uptake decreases (along with gluconeogenesis) 12 14 . However, critically ill metformin-intoxicated patients may have a 30 to 60% decrease in their global oxygen extraction and consumption 7 , as for generalized (and not solely hepatic) mitochondrial dysfunction. Lactate overproduction from tissues other than the liver may then occur 15 .

The aim of the present study was to clarify whether metformin intoxication inhibits whole-body respiration and alters mitochondrial function in the liver and other tissues.

The study complied with international recommendations 16 and was approved by the Italian Ministry of Health (protocol number 212).

All animals survived until end of the study. In sham controls, none of the variables of interest changed over time (Figure 1, Additional file 1). Conversely, pigs infused with metformin (6 ± 2 g; final serum drug concentration 77 ± 45 (range 3 to 146) mg/L) had, on average, severe lactic acidosis (final arterial pH 6.93 ± 0.24 (range 6.80 to 7.57); lactate 18 ± 7 (range 1 to 24) mmol/L) (Figure 1). Arterial pH decreased and lactatemia increased proportionally to final serum metformin level (R² 0.54, P = 0.015 and R² 0.40, P = 0.049, respectively). Hypoglycaemia usually developed (P = 0.002): more glucose was infused, on average, in intoxicated than in control animals (15 ± 7 vs. 3 ± 5 ml/hour of a 33% solution; P = 0.003).

Metformin intoxication progressively inhibited global oxygen extraction (P = 0.008) and consumption (P < 0.001) and increased mixed venous oxygen saturation (from 65 ± 6 up to 78 ± 7%, P < 0.001), despite no change in oxygen delivery (P = 0.58) (Figure 1). The overall rise in blood lactate levels was proportional to the decrease in global oxygen use (R² 0.70, P = 0.002). Body temperature fell from 38.0 ± 2.0 to 36.6 ± 1.2°C (P < 0.001).

Hemodynamic changes are shown in Additional file 1. Pigs infused with metformin became hypotensive (P < 0.001) and therefore received much more fluids (final water balance 1172 ± 834 vs. -139 ± 242 ml, P = 0.011) and norepinephrine (< 0.001) than controls. Cardiac output (not measured in one case) never significantly changed (P = 1.0).

Infusion of lactic acid decreased arterial pH (down to 6.86 ± 0.16) and increased lactatemia (up to 22 ± 3 mmol/L) just as metformin did. Nonetheless, global oxygen use and extraction augmented (P < 0.001 for both), mixed venous oxygen saturation diminished (from 65 ± 11 down to 39 ± 14%, P < 0.001) and delivery slightly, spontaneously, increased (P = 0.017) (Figure 2). Hemodynamic changes are reported in Additional file 2.

This work demonstrates that metformin intoxication: 1) causes lactic acidosis; 2) inhibits oxygen consumption (while lactic acidosis per se does not) and 3) impairs mitochondrial function in liver and other tissues.

Lactic acidosis is commonly due to cellular hypoxia 20 . Whenever oxygen delivery diminishes relative to demand, extraction increases to maintain a normal aerobic (mitochondrial) energy production. When this compensatory mechanism is exhausted, anaerobic (extra-mitochondrial) glycolysis accelerates, lactate production increases 21 and acidosis develops. Lactate released from skeletal muscle into systemic circulation to be taken up as an energy substrate by distal organs may further increase blood lactate levels 22 .

Lactic acidosis can also occur independently from cellular hypoxia 23 24 , as in these experiments. In fact, following metformin infusion in healthy pigs, oxygen use declined and lactic acidosis arose only because oxygen extraction, but not delivery, diminished. Accordingly, mixed venous oxygen saturation increased, suggesting that delivery even exceeded demand. The decline in body temperature further indicates hypometabolism. These data confirm our previous clinical findings: patients with lactic acidosis due to metformin accumulation have abnormally low oxygen extraction and consumption, high venous oxygen tension and are usually hypothermic 7 . Severe metformin intoxication therefore resembles cyanide poisoning 25 , when cells become unable to use otherwise available oxygen and energy production largely depends on glycolysis. The fact that oxygen consumption never diminished in animals infused with lactic acid (and no metformin) demonstrates that lactic acidosis per se cannot explain our present findings.

Serum metformin levels in intoxicated pigs (77 ± 45 mg/L) were similar to those we have previously recorded in humans (61 ± 25 mg/L; therapeutic level is < 4 mg/L) 7 . Severity of intoxication in pigs correlated with that of lactic acidosis. One single animal had a normal final serum drug concentration (possibly due to efficient drug urinary excretion) with only mild and transient metabolic alterations. All other pigs had clearly toxic serum drug levels with overt and persistent lactic acidosis.

Metformin variably accumulates in virtually every tissue 26 . In vitro, it dose-dependently inhibits mitochondrial function even in cells other than hepatocytes 27 28 29 30 31 32 . We have now shown that, in vivo and at toxic dose, it causes mitochondrial dysfunction not just in liver but also in heart, kidney, skeletal muscle and platelets (with differences in severity possibly explained by uneven drug distribution). Lower global oxygen use may thus reflect a generalized inhibition of mitochondrial respiration, which may lead to hepatic and extra-hepatic lactate overproduction. The large increase in blood lactate-to-pyruvate ratios may consistently signal accelerated glycolysis and lactate generation in face of diminished mitochondrial energy production 33 .

Previous reports have shown that chronic therapeutic levels of metformin can stimulate mitochondrial biogenesis 34 , the process by which cells form new mitochondria. Increased citrate synthase activity in kidney and liver is consistent with this finding. But still, in our present model of acute intoxication, mitochondrial function was mainly, largely, depressed. Other molecules variably modulate mitochondrial metabolism depending on dose and time of administration. For instance, low levels of nitric oxide chronically stimulate mitochondrial biogenesis 35 whereas toxic levels acutely inhibit respiratory chain enzymes 36 .

How metformin affects mitochondrial function remains controversial 11 12 and our study did not specifically address this issue. It may increase the formation of reactive oxygen species 37 that may ultimately disrupt mitochondrial integrity 38 . However, we could not demonstrate any clear change in serum markers of oxidative stress following metformin intoxication. Mitochondrial dysfunction in platelets, that are anucleated cells, suggests that de novo, nuclear-encoded, messengers may not be required.

Some of the limitations of this study deserve a comment. First, we used healthy pigs whereas patients with accidental metformin intoxication typically suffer from diabetes, renal failure and other comorbidities. Doing so, we aimed at investigating the pure effect of drug accumulation on global oxygen use and metabolism. Second, pigs infused with metformin also received much more fluids and norepinephrine (to treat hypovolemia and systemic vasodilatation) and glucose (to correct hypoglycaemia) than controls. None of these other factors, however, can explain our present findings. In fact, respiratory chain enzymes activities were expressed relative to citrate synthase activity, to correct for any difference in mitochondrial density (possibly due to edema, for instance). Norepinephrine increases oxygen consumption in vivo 39 and does not grossly alter porcine hepatic and skeletal muscle mitochondrial respiration in vitro 40 41 . Although (intracellular) hypoglycaemia can diminish oxygen consumption, it can hardly explain lactate overproduction, since lactate is generated from glucose. Notably, changes in oxygen consumption and blood lactate levels persisted even when hypoglycaemia was reversed. Third, cellular respiration and lactate generation were not directly measured.

The development of lactic acidosis during metformin intoxication is associated with diminished global oxygen consumption and clear evidence of mitochondrial dysfunction in liver and other tissues. Lactic acidosis per se does not inhibit whole-body respiration.

• Severe metformin intoxication largely decreases global oxygen consumption and impairs mitochondrial function in heart, kidney, liver, skeletal muscle and platelets of otherwise healthy pigs.

• Lactic acidosis per se does not decrease whole-body respiration.

• Diffuse inhibition of cellular respiration and secondary lactate overproduction may contribute to the development of metformin-induced lactic acidosis.

CO: cardiac output; DO₂: global oxygen delivery; Hb: haemoglobin; HClO: hypochlorous acid; IM: intramuscular; IV: intravenous; O₂: oxygen; SaO₂: arterial oxygen saturation; SD: standard deviation; VO₂: global oxygen consumption.

The authors declare that they have no competing interests.

AP conceived the study, ran the experiments, performed the analysis and wrote the manuscript. FF measured mitochondrial enzyme activities and participated in data analysis. MM ran the experiments. SV participated in data analysis and helped to draft the manuscript. SG participated in the study design and ran the experiments. GPC participated in the study design and analysis. RDG measured markers of oxidative stress. LG participated in the study design and helped to draft the manuscript. All authors read and approved the final version of the manuscript.