Metformin is among the most widely prescribed drugs in the world, and increasingly tested in cancer on the assumption that throttling a tumour's mitochondria is survivable. A study in The Innovation suggests why tumours tolerate it, and how to remove that tolerance. Blocking the respiratory chain leaves NADH, the electron courier of metabolism, piled up in cytosol and mitochondria; tumours drain it through a two-branch shuttle ending in NADPH. Remove the shuttle's first enzyme, pyruvate carboxylase, and metformin turns lethal.
Reductive stress is the reason. When the respiratory chain stalls, NADH/NAD+ climbs and biosynthesis shuts down. Lactate fermentation regenerates NAD+ but leaves cytosolic reducing pressure untouched, and the one transhydrogenase that could help sits in the mitochondrial matrix, dependent on a membrane potential a failing chain cannot hold.
The circuit spans both compartments. Pyruvate carboxylase (PC) makes oxaloacetate inside mitochondria. Mitochondrial malate dehydrogenase (MDH2) reduces it to malate while consuming mitochondrial NADH; malate is exported and oxidised by malic enzyme 1 (ME1) to pyruvate and NADPH. Alternatively oxaloacetate becomes aspartate, is exported, and is rebuilt in the cytosol by GOT1 and MDH1, which spends cytosolic NADH before ME1 finishes. Either way, NADH becomes cytosolic NADPH.
Which way flux runs matters more than the product. If the circuit supplied NADPH for antioxidant defence, destroying NADPH should hurt. The authors expressed TPNOX, an engineered oxidase that drains NADPH to NADP+. Under respiratory blockade it restored proliferation, and the rescue vanished when ME1, MDH1, MDH2 or PC was knocked down. NADPH consumption is the pull that keeps malate moving.
Then came the anomaly. Glucose-6-phosphate dehydrogenase (G6PD) is the textbook NADPH source, and its loss made cells hypersensitive to respiratory inhibitors — yet a catalytically dead mutant, R198P, rescued growth as well as the wild-type protein. G6PD binds ME1 directly, and purified G6PD raised purified ME1's activity in a tube. As a scaffold it speeds malate oxidation while withholding its own NADPH output.
Targeting the loop worked. The PC inhibitor ZY444 sensitised HeLa, SK-Hep1 and 143B cells to metformin; adding CB839, which blocks glutamine-fuelled metabolism, suppressed growth more than either pair. In nude mice carrying HeLa or SK-Hep1 xenografts the single agents did little, while the triple combination nearly abolished tumour progression. PC also had to sit inside mitochondria: stripping its targeting sequence abolished the rescue until one was restored — at odds with a report placing PC in the cytosol.
The limits are those of cell culture. The mechanism was mapped in four lines under pharmacological blockade, not spontaneous tumours, and ZY444 is a tool compound, not a drug. What it establishes is narrower: the PC-ME1 axis is not a NADPH factory but a drain for electrons that would otherwise choke the cell — and drains can be plugged.
The Innovation
Experimental study
Cells
Metabolic circuitry of NADH buffering reveals a vulnerability in redox-adaptive cancer cells
8-Sep-2026
The authors declare no competing interests.