04.01 BCA McCormick_Wolf eps.indd NS.indd

Axon initial segment Action potentials in cortical neurons show a variable threshold and a sudden rise in membrane potential at initiation. Naundorf et al. fail to explain these features using singleor double-compartment Hodgkin–Huxleystyle models, suggesting instead that they could arise from cooperative opening of Na channels, although there is no direct biological evidence to support this. Here we show that these so-called unique features are to be expected from Hodgkin–Huxley models if the spatial geometry and spike initiation properties of cortical neurons are taken into account — it is therefore unnecessary to invoke exotic channel-gating properties as an explanation. Cortical pyramidal cells initiate spikes in the axon initial segment (AIS) about 30–60 μm from their soma. These spikes then propagate antidromically through the soma and dendrites. A well known feature of antidromic spikes is their sudden rise from baseline. These critical properties were not considered by Naundorf et al.. We made simultaneous whole-cell recordings from the AIS by patching the cut end of the axon (Fig. 1, legend) and the soma of layer5 pyramidal neurons in vitro during spontaneous spike generation (Fig. 1). Somatic spikes showed a rapid rise, or ‘kink’, at initiation (Fig. 1a, b) and the slope of the phase plot of spike dV/dt versus V at dV/dt = 15 mV ms was 25 ± 6.8 ms (mean ± s.d.; n = 32). The phase plots of dV/dt versus V typically revealed a biphasic rise, which was suggestive of two underlying components (Fig. 1b; n = 30/32), as observed in many cell types. This biphasic component was not evident in the recordings of Naundorf et al., although the low peak dV/dt of their recordings indicates that their spikes may not have been fully represented. Intrasomatic injection of a noisy conductance that mimics the arrival of excitatory and inhibitory synaptic activity resulted in significant variation in the apparent spike threshold (n = 6; Fig. 1c, green lines), as observed in the recordings of Naundorf et al.. In contrast to somatic spikes, those recorded at the site of spike initiation, the AIS, showed a slower rise (n = 10; Fig. 1d, e). The slope of the phase plot of spike dV/dt versus V at dV/dt = 15 mV ms was much lower for the AIS (3.8 ± 1.7 ms; n = 6; P<0.01; Fig. 1d, e) than it was for the soma (Fig. 1a, b). The slow rise at spike initiation in the AIS is not an artefact of our method of axonal recording (Fig. 1, legend). On intrasomatic injection of a noisy conductance that mimics synaptic activity, the apparent spike threshold was less variable for the AIS (n = 6; Fig. 1f, green lines) than it was for the soma (Fig. 1c). Spike initiation in the AIS is mediated by either a high Na-channel density in the AIS, as indicated by immunocytochemistry, or by a lesser density of Na channels, which have a low threshold for activation. Using a previous model of spike initiation in a layer-5 cortical pyramidal cell, we adjusted the axonal and somatic densities of Na and K channels until the spike waveform and its derivative were similar to those of our actual recordings (compare Figs 1 and 2). Our Hodgkin–Huxley model initiated spikes in the AIS that then propagated antidromically through the soma and dendrites, as in real pyramidal cells. At the soma, these spikes showed a rapid rise at initiation (Fig. 2a, b), and the slope of the phase plot for spike initiation at dV/dt = 15 mV ms was 21 ms. Intrasomatic injection of artificial synaptic barrages into the modelled neuron revealed a high variability of apparent spike threshold in the soma (Fig. 2c). As in the whole-cell recordings, the rise in the model spike at initiation was smoother at the AIS (Fig. 2d, e) than at the soma (Fig. 2a, b). The slope of the phase plot for spike initiation at dV/dt = 15 mV ms was considerably lower for the model AIS (4 ms) than for the soma, and both were in the range observed in normal cells. Intrasomatic injection of artificial synaptic barrages showed a less variable threshold in NEUROPHYSIOLOGY