Fifteen different models of overflow metabolism convey a single message: overflow metabolism is caused by two growth-limiting constraints.

We recently published an article in which we reviewed the different approaches that describe a puzzling metabolic phenomenon called overflow metabolism: (

Many different organisms secrete ‘overflow products’ in conditions where they grow and reproduce fast. For example, many yeasts produce ethanol, Escherichia coli produces acetate, and cancer cells produce lactate. This behaviour seems counter-intuitive because the production of overflow products is inefficient in terms of energy. The cells do not fully use the energy stored in their growth substrates at high growth rates, while they do fully use this energy at lower growth rates.

To study this phenomenon, many mathematical/computational models have been proposed, based on different biological fundaments and using different assumptions. However, since the models can all reproduce overflow metabolism, no final conclusion could be drawn about which model is closest to reality.

We therefore decided to analyse all constraint-based optimization approaches that describe overflow metabolism. Although these models appear very different: the approaches range from relatively simple flux balance analyses to self-fabricating metabolism and expression models, we could rewrite all of them in one standard form. We then looked for a common feature, for this could unveil the cause of overflow metabolism, and for differences, for these can be used to design experiments that test the different models and their assumptions.

We found that all models predict overflow metabolism when two growth-limiting constraints are hit, while the specific nature of the constraints differed between the models. This is in line with recent theory: ( We concluded that finding the true cause of overflow metabolism in a specific organism thus amounts to proving which two constraints limit the growth of that organism. However, we did not find such decisive proofs for any constraint. Therefore, we decided to list all the proposed constraints, such that these can hopefully all be tested by specific perturbation experiments in the future.