Inhibitory synaptic release properties are topographically distributed in auditory circuitry

Synaptic physiology has benefited from investigation of the circuitry of the brainstem auditory system where specializations in neural computation are readily related to function. This includes the widely known exploration of large synapses such as the Calyx of Held (Forsythe, 1994; Schneggenburger & Forsythe, 2006), as well as biophysical and morphological specialization in neurones that compare input from the two ears. In birds, neurones of nucleus laminaris (NL) compute the difference in a sound stimulus's arrival time at the two ears. These interaural time disparities (ITDs) are an acoustic cue that varies systematically with a sound source's position in space. Computation of ITD allows animals to accurately localize sound sources. ITDs are very small, on the order of tens to hundreds of microseconds, and depend primarily on head width. An NL neurone is capable of varying its firing rate over its entire dynamic range within this narrow physiological range of ITDs. Thus, the NL circuitry operates with precision despite exceedingly restrictive temporal constraints.

Another temporal constraint faced by NL arises from stimulus frequency. Neurones in the NL pathway exhibit phase-locked responses to the acoustic waveform in order to encode the temporal features of the signal. Phase-locked neurones respond precisely at a particular phase of the stimulus waveform for signals from 10 to about 2000 Hz. Thus, the computational demand on a particular NL neurone depends both on the ITD given by stimulus location, and by the stimulus frequency to which the NL neurone is tuned. These computational demands have driven the evolution of many physiological specializations in NL circuitry.

Auditory brainstem neurones are arranged topographically according to their characteristic frequency (CF), the frequency to which they are most sensitive. This topographic arrangement is referred to as ‘tonotopy’. Several biophysical properties have been identified that are arranged systematically along tonotopic axes in this system, suggesting that these neurones’ characteristics are ‘tuned’ for their computational function. Some examples of these features include a tonotopic gradient of dendrite length in NL (Fig. 1) (Smith & Rubel, 1979) and of voltage gated ion channel expression (Kuba & Ohmori, 2009). Additionally, the input neurons to NL show a systematic variation in the number and strength of excitatory synapses (Fukui & Ohmori, 2004).

Figure 1. Inhibitory kinetics are tonotopically distributed in nucleus laminaris.

Illustration shows the single layer of NL neurons with its characteristic dendrite length gradient along its tonotopic axis. GABAergic inhibitory innervation from SON synapses on a low frequency tuned neurone (left) and a high frequency neurone (right). Postsynaptic GABAergic currents are inward with kinetics showing a five-fold range of decay kinetics along the tonotopic axis.

Recently, Tang et al. (2011) showed that inhibitory inputs appear to vary along the NL tonotopy as well. Inhibition to NL as well as to its presynaptic input neurones has been shown to be primarily GABAergic, kinetically slow, and depolarizing. The source of GABAergic input is predominantly from the superior olivary nucleus (SON). These authors showed that middle and high CF NL neurons receive a small but sustained tonic depolarization arising from extrasynaptic GABA_A receptors, a result that was not observed for their low frequency counterparts.

Now, in a study appearing in this issue of The Journal of Physiology, Tang & Lu (2012) extend on their earlier work and reveal a novel tonotopically distributed feature of inhibition in NL. They show that the kinetics of synaptically derived input mediated by GABA_A receptors varies along NL's tonotopic axis as well. In high CF neurons, the decay of the inhibition is very slow, leading to prominent summation of synaptic current at physiologically relevant input rates. In contrast, the low CF synapses had a fast decay and exhibited relatively little summation. The τ_decay values varied fivefold along the tonotopic axis (τ_decay: 10.4 and 54.8 ms for low and middle/high CF synapses, respectively). The authors then provide compelling evidence that this variation results from differences in the presynaptic release properties of inhibitory input terminals along the tonotopic axis. The slow decay of high CF IPSCs arose largely from asynchronous vesicle release, especially during high frequency stimulation. Accordingly, manipulation of presynaptic intracellular Ca²⁺ altered the decay kinetics in high, but not low CF neurones. Complementary to their earlier study, the authors show that the sustained release in the high CF region was coupled to spillover of GABA to putative extrasynaptic receptor, which contributed to this persistent period of inhibition. Finally, Tang and Lu demonstrate that this sustained inhibition is capable of suppressing responses in high CF neurones for a period that persists well beyond the cessation of driven activity of the inhibitory neurones themselves.

The fivefold difference in inhibitory kinetics along the tonotopic axis is compelling in light of recent data that the major source of inhibition to NL is also phase-locked. This suggests inhibitory postsynaptic current in NL may be temporally patterned in low CF neurones (Coleman et al. 2011). It is conceivable then that phase-locked inhibition with relatively rapid kinetics may influence ITD computation on a cycle-by-cycle basis in low CF neurones, whereas middle and high CF neurones make use of a sustained plateau of inhibition. A cycle-by-cycle inhibition model has been experimentally supported in the analogous mammalian circuit (Brand et al. 2002), but several important differences between the two systems make it unlikely that the computational process is functionally identical. A full understanding of the computational advantage gained by the gradient of inhibitory kinetics in NL awaits further investigation. However, the present findings suggest that even within one organism, a single model of inhibitory function will be insufficient to explain the full complement of its contributions to the remarkable precision of ITD computation.