Neurons generate their output signal — the action potential — in a distinct region of the axon called the initial segment. The location and extent of this trigger zone can be modified by neural activity to control excitability.
The brain changes with experience. But where are these changes, which ultimately result in altered patterns of activity, stored in neural circuits? Although the plasticity of the synaptic connections between neurons has received much attention, the intrinsic excitability of a neuron — its responsiveness to synaptic input — can also be markedly altered by experience. In this issue, two groups (Grubb and Burrone1 and Kuba et al.2) identify a new target of intrinsic plasticity in the axon, the output structure of the neuron. They show that the proximal region of the axon known as the axon initial segment, the initiation site of the action potential (the neuron's output signal), can be modified to make the cell more or less responsive to inputs.
Activity-dependent changes in neural circuits are crucial for brain development and memory formation. Deprivation of a sensory modality such as visual input causes long-term changes in sensory circuits3. The conventional view is that such activity-dependent plasticity involves changes in the strength of synapses. This idea has tremendous appeal, given that synaptic changes can be highly specific and show a wide dynamic range, and that, given their vast numbers, synapses offer enormous capacity for information storage.
Changing the overall intrinsic excitability of a neuron is an alternative, complementary mechanism for information storage. This widespread strategy, which is not specific to individual synapses, is of immense value in neural circuits. It allows multi-purpose circuits to switch between activity modes4, enables adjustments to be made in the firing rate of neurons during sensorimotor learning5, and provides a powerful mechanism for the homeostatic regulation of activity6. Intrinsic plasticity can come about through the altered composition, density and distribution of ion channels in the neuronal membrane that are activated by changes in voltage (voltage-gated ion channels). Which channels are altered, and where, is crucial for determining the selectivity and impact of the excitability change.
Over the past decade, activity-dependent changes in the density and function of many voltage-gated channels have been reported7,8,9,10. By being restricted to subneuronal compartments, such as individual segments of dendrites11,12 — short, branching neuronal processes that receive synaptic input — such changes can provide some degree of specificity. But typically they are widespread, being found mainly in the neuron's cell body and dendrites.
Grubb and Burrone1 (page 1070) and Kuba et al.2 (page 1075) implicate a specific axonal structure in intrinsic plasticity — the axon initial segment (AIS), a highly organized, ion-channel-enriched matrix of proteins that is situated close to the cell body. Significantly, the fact that the action potential is initiated in the AIS13 would allow this form of plasticity to regulate the final site of integration of the synaptic input directly.
To investigate how the location of the AIS might depend on activity, Grubb and Burrone use cultures of neurons from the hippocampus. They find that when extracellular potassium levels are chronically elevated, mimicking increased neuronal activity, the AIS shifts away from the cell body. This movement involves wholesale translocation of several types of AIS-specific protein, thereby creating a non-excitable 'spacer' region — 21 micrometres long — between the AIS and the cell body. This shift further isolates the action-potential trigger site from the synaptic input to the dendrites and thereby reduces the ability of the input to trigger action potentials. Consequently, neurons with a more distal AIS are less excitable and require stronger stimulation to fire.
To achieve more precise control over neuronal activity, Grubb and Burrone manipulated their cultures so that the neurons expressed a membrane protein called channelrhodopsin-2 (ChR2), which is a light-activated ion channel14 (Fig. 1a). They could then use light stimuli to directly trigger spiking with precise temporal control. Long-term, regular, low-frequency light stimulation at 1 hertz had little effect on AIS location. But when the neurons were activated by high-frequency bursts of light, the AIS shifted significantly.
What mechanisms underlie these activity-dependent changes? Many forms of neuronal plasticity are triggered by increases in intracellular calcium-ion concentration. Indeed, Grubb and Burrone1 show that blocking T- and L-type calcium channels prevents the AIS from moving. This suggests that activity-dependent calcium signals provide a read-out of the pattern of neuronal activity, which triggers structural plasticity at the AIS. But these experiments were carried out on neurons in cell culture, raising the question of whether this form of homeostatic axonal plasticity occurs in the intact brain.
Kuba et al.2 neatly answer this question. They had previously found that, in brainstem neurons responsible for encoding sound, the precise location and length of the AIS depend on the characteristic sound frequency that each neuron processes15. Now they have tested the effects of hearing loss on the AIS location (Fig. 1b). Removing the cochlea, the auditory part of the inner ear, from one-day-old chicks causes loss of synaptic input to neurons in the nucleus magnocellularis — an essential relay station in the auditory pathway. Kuba and co-workers find that such input deprivation has a dramatic effect on the AIS of these neurons: following hearing loss, the AIS is elongated by roughly 70%. These nucleus magnocellularis neurons lack dendrites, confirming that the excitability changes are indeed restricted to the axon.
Intriguingly, although Kuba et al. did not detect changes in the density and subtype composition of sodium channels in the AIS, hearing deprivation increased total sodium currents in the axons, and this could be attributed to the expansion of the AIS. Consequently, smaller current injections were sufficient to trigger action potentials after hearing deprivation, indicating that neuronal excitability had increased to compensate for the reduced synaptic drive.
Together, these papers1,2 show that the AIS is a powerful target for homeostatic plasticity mechanisms. Notably, the changes in the AIS are bidirectional and reversible — properties crucial for the proposed homeostatic role of such mechanisms.
The two studies also open the door to exploring how AIS plasticity occurs in different types of neuron under various conditions that affect neural circuits. For instance, both teams used relatively crude methods to alter neuronal activity: long-term increases in neuronal activity or sensory deprivation. More subtle and controlled manipulations, over shorter timescales, could reveal whether AIS plasticity is restricted to developmental and pathological conditions, or whether it is a normal physiological mechanism that could dynamically regulate excitability.
The studies identify distinct mechanisms for modulating neuronal excitability — either displacement or extension of the AIS (Fig. 1a,b). It will therefore be necessary to determine which prevails in different neuronal network states and brain areas. Neither group directly addressed how the changes in the AIS alter the integration of synaptic input by the neurons. This is particularly relevant for inputs mediated by the neurotransmitter γ-aminobutyric acid (GABAergic inputs), which cluster at the AIS16. Do these GABAergic synapses move with the AIS, and, if not, how does this affect their ability to influence the generation of action potentials? Finally, if the molecular mechanisms underlying this form of plasticity are identified, they could provide targets for manipulating neuronal excitability in various disease states that involve altered excitability, particularly epilepsy.