Transformation in the neural code for whisker deflection direction along the lemniscal pathway

Abstract

A prominent characteristic of neurons in the whisker system is their selectivity to the direction in which a whisker is deflected. The aim of this study was to determine how information about whisker direction is encoded at successive levels of the lemniscal pathway. We made extracellular recordings under identical conditions from the trigeminal ganglion, ventro-posterior medial thalamus (VPM), and barrel cortex while varying the direction of whisker deflection. We found a marked increase in the variability of single unit responses along the pathway. To study the consequences of this for information processing, we quantified the responses using mutual information. VPM units conveyed 48% of the mutual information conveyed by ganglion units, and cortical units conveyed 12%. The fraction of neuronal bandwidth used for transmitting direction information decreased from 40% in the ganglion to 24% in VPM and 5% in barrel cortex. To test whether, in cortex, population coding might compensate for this information loss, we made simultaneous recordings. We found that cortical neuron pairs conveyed 2.1 times the mutual information conveyed by single neurons. Overall, these findings indicate a marked transformation from a subcortical neural code based on small numbers of reliable neurons to a cortical code based on populations of unreliable neurons. However, the basic form of the neural code in ganglion, thalamus, and cortex was similar-at each stage, the first poststimulus spike carried the majority of the information.

Fig. 1.

Transformation in directional tuning along the whisker pathway. A: tuning curves aligned by preferred direction (PD) and averaged over units. B: aligned tuning curves normalized by firing rate in the PD and averaged over units.

Fig. 2.

Response of single units to whisker direction at different stages of the whisker pathway. A1–A4: spikes fired in response to deflection of the principal whisker in 430 different directions for a slowly adapting (SA) ganglion unit, a rapidly adapting (RA) ganglion unit, a ventro-posterior medial thalamus (VPM) unit, and a primary somatosensory cortex (S1) unit, respectively. B1–B4: mean spikes per trial after a deflection in a given direction, for units corresponding to A1–A4. Bars denote SD. C1–C4: mutual information conveyed by the spike count about direction (8 categories) in time windows starting at stimulus onset, ending 1–100 ms later, for units corresponding to A1–A4.

Fig. 3.

Transformation in reliability of responses along the whisker pathway. A: variance vs. mean of response of each ganglion unit to deflection in each of 8 direction categories. B and C: corresponding data for VPM and S1. D: Fano factor in the preferred direction, averaged over all units recorded from each brain area. Bars denote SE.

Fig. 4.

Transformation in direction information along the whisker pathway. A: mutual information conveyed by spike count in windows of progressively increasing duration, starting from deflection onset, averaged over all single units recorded from each brain area. Bars denote SE. B–D: distribution of mutual information values (100 ms time window) for ganglion, VPM, and S1, respectively.

Fig. 5.

Transformation in coding efficiency along the pathway. A: response entropy of the spike count response (Eq. 2), plotted as in Fig. 4A. B and C: analogous data for noise entropy (Eq. 3) and coding efficiency (Eq. 4). Bars denote SE.

Fig. 6.

Cortical population coding of whisker direction. Mutual information conveyed in spike counts by pairs of simultaneously recorded units and their constituent single unit responses in time windows increasing from deflection onset.

Fig. 7.

Role of first poststimulus spike. A–D: mutual information conveyed by the complete response compared with that conveyed by 1st spike alone, 1st and 2nd spike together, and 1st, 2nd, and 3rd spikes together. Data averaged over all SA ganglion, RA ganglion, VPM, and S1 responses, respectively.