Nelson Spruston, Hae-Yoon Jung, and Tim Mickus
Dept. Neurobiology & Physiology, Institute for Neuroscience,
Experiments using simultaneous somatic and dendritic patch-pipette recording from hippocampal CA1 pyramidal neurons in slices have shown that action potentials are initiated near the axon and backpropagate into the dendritic tree in an activity dependent manner. At low firing rates action potentials invade much of the dendritic tree quite effectively, but at higher firing rates, action potentials invade the dendrites progressively less effectively, eventually leading to apparent failures of active propagation at some dendritic branch points (Spruston et al., Science 268:297-300, 1995). We have tested two major hypotheses as to how this activity-dependent spike propagation may arise: 1) that cumulative sodium channel inactivation develops during trains of action potentials, and 2) that a generalized shunt develops during trains of action potentials. Each of these factors would be expected to result in a decreased ability of a low density of dendritic sodium channels to support active action potential backpropagation. These expectations have been confirmed in a computational model (Migliore, Biophys. J. 71:2394-2403, 1996).
The first hypothesis was tested by examining the amplitude of sodium currents (I) activated by successive depolarizing current pulses. Repetitive depolarizations (50 mV, 15-20 pulses at 20 Hz) applied to either cell-attached (somatic or dendritic) or nucleated patches revealed that I became progressively smaller to successive depolarizations during each train of pulses. This suggests that voltage-gated sodium currents in hippocampal pyramidal neurons undergo a form of inactivation that is slow to recover. We examined the time course of this recovery following the end of the train of action potentials; was 39% at 500 ms, 50% at one second, and 89% two seconds after the train. Recovery could be accelerated by hyperpolarization during the time between the end of the train and the recovery test pulse. These properties of cumulative sodium current inactivation are consistent with the properties of activity-dependent action potential attenuation in CA1 pyramidal neurons.
The second hypothesis was tested by determining whether a dendritic shunt could be detected 100-500 ms after a train of action potentials - a time when attenuation is still considerable in dendritic recordings (Spruston et al., ibid. 1995). The presence of a shunt was tested by monitoring input resistance or EPSP amplitude before and after a train of action potentials. No change in either of these parameters was detected after a train of action potentials in either somatic or dendritic recordings, suggesting that a global dendritic shunt does not contribute to the activity dependence of action potential backpropagation. While a global shunt ought to be detectable by the methods we employed, we cannot rule out the possibility that localized shunts constitute an additional factor directing the invasion of action potentials into specific dendritic branches. Our data indicate, however, that cumulative sodium channel inactivation is likely to play an important role in the spread of action potentials into all CA1 dendrites.