We pioneered analyses of the subcellular tasks on which the brain uses energy, establishing the first energy budgets for the grey and white matter of the brain (Attwell & Laughlin, 2001; Harris & Attwell, 2012), which were calculated from the bottom up, from the measured properties of ion channels, synapses, cell anatomy and action potential rates. We showed that brain energy use imposes profound constraints on the speed with which the brain can process information, dictating that energy-saving coding and information transmission strategies must be used. We also demonstrated that this implies that the energy supply to the brain is unexpectedly linked to parameters determining the speed with which it operates, including the affinity of glutamate receptors, the diameter of synaptic boutons and the rate of binding of glutamate to its transporters (Attwell & Gibb, 2005).
The energy budget analysis showed that most brain energy is used postsynaptically at synapses. This requires that ATP be made postsynaptically at active synapses. In collaboration with Josef Kittler, we have studied how the mitochondria which make this ATP are located at active synapses. This occurs by virtue of calcium entry through postsynaptic NMDA receptors uncoupling the mitochondrial adaptor protein Miro from the motor protein which moves mitochondria along microtubules.
Analysis of how information transfer through synapses depends on the number of postsynaptic receptors (performed experimentally using dynamic clamp: Harris et al., 2015) and on presynaptic release probability (Harris et al., 2012) revealed that both of these parameters are optimised to maximise the rate at which information is transmitted per energy used (rather than optimised purely to maximise information transmission). The remarkably low release probability of central synapses was thus predicted to result from the observed degree of convergence of synapses from a single presynaptic axon onto a postsynaptic cell.