Already in 2001 research cast substantial doubt on the underlying mechanism for the generation of a BOLD response in fMRI. Logothetis and collegues (Logothetis et al. Nature 2001) found that the BOLD response more accurately reflected synaptic input into an area than the actual spike output from the area. Viswanathan and Freeman took this argument further and showed that there didn’t need to be spikes to generate a BOLD response (Viswanathan and Freeman Nature Neuroscience, 2007). But this poses a problem. According to the Hodgkin-Huxley model of spike generation there is a substantial energy demand for each spike.
Now this apparent discrepancy has been resolved by Henrik Alle and colleagues (Alle et al. Science, 2009). In hippocampal mossy fibers they have found that the overlap in sodium and potassium flux is minimally overlapping. This enables the axons to generate action potentials with only 1.3 times the theoretical minimum energy (charging a perfect capacitor). This stands in stark contrast to the previously estimated 4.0 times the theoretical minimum.
The researchers came to this conclusion by applying previously recorded action potential waveforms as a voltage command to outside-out patches. The recorded sodium and potassium ion-fluxes were isolated and recorded. Interestingly, the percentage of overlap between the two ion-fluxes was as low as 20% of the total charge. Additionally, it seems that the calcium influx at the presynaptic membrane costs approximately 6 times as much as a spike traveling down the axon. This stresses the relative high energy consumption of synaptic transmission as opposed to action potential firing.
Carter and Bean (Carter and Bean Neuron, 2009) took these observations even further and investigated action potential energy consumption across several neuron types. Fast spiking neurons, like GABA-ergic interneurons, have much narrower spikes than slower spiking neurons, like neocortical pyramidal cells. They found that fast spiking comes at a cost. For narrower action potentials the sodium and potassium need to overlap; and consequently, the energy consumption per action potential for these fast spiking neurons is considerably higher than the 1.3 times the theoretical minimum Alle and collegues found.
Taken together it seems that the energy consumption per action potential, even for fast spiking neurons, is considerably smaller than previously assumed. This has implications for the interpretation of fMRI data. Action potential firing is probably not the underlying neuronal mechanism for the fMRI BOLD response. Most likely, the BOLD response is caused by synaptic communication. Thus projections to a brain area might be contributing to the BOLD response just as much as processing in that brain area.