Estimation of premotor synaptic drives to simulated abducens motoneurons for control of eye position

Abstract

The firing rate of an abducens motoneuron (AbMN) is linearly related to eye position with slope K, above recruitment threshold θ. Within the AbMN population K increases as θ increases. It is possible that these properties depend on the synaptic drives generated by the major premotor inputs to AbMNs, namely position-vestibular-pause (PVP) cells and eye and head velocity (EHV) cells in the medial vestibular nucleus, and eye-position and burst-position cells in the nucleus prepositus hypoglossi (NPH). Premotor inputs to AbMNs were therefore modelled by a two-layer net, in which the output nodes represented the AbMNs (with fixed intrinsic properties) and the input nodes the three classes of premotor units (n=20/class). Conjugate eye-position commands were used to generate the firing rates in premotor units found experimentally. The output of the net was compared with observed AbMN firing rates, and the resultant error used to adjust the magnitude and sign of the connection weights between premotor units and AbMNs. To provide additional constraints on permitted weights, the net was also trained under simulated smooth pursuit, cancellation of the vestibular-ocular reflex, and the vestibulo-ocular reflex itself (all at 0.5 Hz). Since the projections of EHV cells have not been clearly characterized, two versions of the model were trained, corresponding to different assumptions about these projections. In both versions, position-related AbMN firing rates were derived mainly from an excitatory drive from PVP cells and an inhibitory drive from NPH cells with the opposite ON direction. Variation in AbMN threshold θ and position sensitivity K depended on the strength of the drive from the NPH: the stronger the drive, the higher both K and θ. This arrangement was observed in six variants of the basic model with different parameter values, and in a simplified form (constant PVP drive and varying NPH drive) was able to generate qualitatively the observed relationship between K and θ even in the absence of input from EHV cells. It appears to be a robust mechanism for producing the experimentally observed variation in position-related firing of AbMNs, even without a contribution from their intrinsic properties, and predicts that local blocking of the inhibitory drive from cells in the NPH should lower both the position threshold and sensitivity of an individual AbMN. The model also indicates that if EHV cells have ipsilateral inhibitory projections, as has been proposed on the basis of their similarity with cells receiving input from the flocculus, then their role in eye-position control would reinforce that of cells in the NPH.