Visual-auditory interactions modulate saccade-related activity in monkey superior colliculus

Abstract

This paper reports on single-unit activity of saccade-related burst neurons (SRBNs) in the intermediate and deep layers of the monkey superior colliculus (SC), evoked by bimodal sensory stimulation. Monkeys were trained to generate saccadic eye movements towards visual stimuli, in either a unimodal visual saccade task, or in a bimodal visual-auditory task. In the latter task, the monkeys were required to make an accurate saccade towards a visual target, while ignoring an auditory stimulus. The presentation of an auditory stimulus in temporal and spatial proximity of the visual target influenced neither the accuracy nor the kinematic properties of the evoked saccades. However, it had a significant effect on the activity of 90% (45/50) of the SRBNs. The motor-related burst increased significantly in some neurons, but was suppressed in others. In visual-movement cells, comparable bimodal interactions were observed in both the visually evoked burst and the movement-related burst. The large differences observed in the movement-related activity of SRBNs for identical saccades under different sensory conditions do not support the hypothesis that such cells encode dynamic motor error. The only behavioral parameter that was affected by the presentation of the auditory stimulus was saccade latency. Auditory stimulation caused saccade latency changes in the majority of the experiments. Meanwhile, the timing of peak collicular motor activity and saccade onset remained tightly coupled for all stimulus configurations. In addition, saccade latency varied as function of the distance between the stimuli in 36% of the recordings. Interestingly, the occurrence of a spatial latency effect covaried significantly with a similar spatial influence on the SRBNs firing rate. These cells were always most active in the bimodal task when both stimuli were in spatial register, but activity decreased with increasing stimulus separation.

Introduction

The intermediate and deep layers of the midbrain superior colliculus (SC) are of crucial importance for the generation of saccadic eye movements [42]. Both electrical stimulation [38] and recording studies [4] have revealed the existence of topographically organized maps of motor space and sensory space. These maps have roughly orthogonal representations of saccade amplitude (or retinal eccentricity) and direction. The rostral half of the primate SC is dedicated to foveal and parafoveal visual space (and saccade amplitudes below 10°). The caudal half of the SC encodes peripheral visual space and is involved in the generation of large-amplitude saccades [34]. Prior to a saccade, a large population of so-called saccade-related burst neurons (SRBNs), located at a specific site within the motor map, is recruited in tight synchrony with the saccade onset 40., 58., and it has been hypothesized that the weighted activity of the entire population determines the amplitude and direction of the saccade vector 20., 22., 34., 49..

A substantial fraction of the SRBNs displays not only motor-related activity, but also produces a sensory evoked burst, that is triggered by the onset of a target in the cell’s receptive field [43]. Although this sensory activity is usually of visual origin, some cells can be activated by auditory or somatosensory stimuli (cat: [25]; monkey: 14., 15.). It has been shown that at the level of single SC cells, the visual, auditory, and somatosensory representations are approximately in register, despite the fundamental differences in initial reference frames (i.e., oculocentric, craniocentric, and body centered, respectively) and encoding formats (retinotopic, tonotopic, and somatotopic, respectively; in cat: 45., 46., 47., 56.; in monkey: [57]).

So far, the neural basis of saccadic eye movements has typically been investigated in orienting tasks involving the rapid foveation of a visual target. Nevertheless, accurate eye movements can be evoked by stimuli of different sensory modalities, and in daily life they are often made in response to peripheral stimuli that emit multimodal sensory cues. Therefore, at the motor programming level, the different sensory modalities need to be merged into a common, oculocentric frame of reference. The neural processing that subserves these multisensory remapping stages is, however, still poorly understood.

Due to both the multisensory convergence onto single cells, and its involvement in the generation of orienting movements of not only the eyes, but also head and ears, the SC is an excellent candidate to study the processes related to multimodal integration. It has been shown in the anesthetized cat that many SC neurons are capable of integrating multimodal stimuli [25]. Often, neurons displayed a strong response enhancement during appropriate bimodal stimulation that well exceeded linear addition of the unimodal responses. In a smaller number of cells, the multimodal interaction resulted in a response suppression [25]. Both multisensory response modes, enhancement, and suppression, have also been reported for SC cells of anesthetized guinea pigs [18] and, more recently, anesthetized monkeys [57].

Because the receptive fields for the various sensory modalities of a multimodally sensitive neuron usually overlap to a considerable degree, the strongest interaction effects were obtained for sensory stimuli presented in both spatial and temporal proximity. Changes in the relative temporal or spatial configurations of auditory and visual stimuli resulted in a systematic decrease of the interaction effect 23., 26., 27., 57..

Multisensory responsive neurons in the SC are thought to play an important role in the programming of goal-directed orienting responses, because the vast majority of these neurons was shown to possess efferent connections to brain stem and spinal motor circuitry 24., 28., 55.. Yet, most experiments on multisensory integration have so far been performed on anesthetized and paralyzed preparations. Due to the absence of orienting movements, research in anesthetized animals is necessarily directed at the sensory activity component of recorded units. Therefore, it is not a priori obvious how to extrapolate these important findings to the fully awake preparation. In particular, conclusions about the relationship between SC motor activity and multisensory-evoked orienting movements of the behaving animal cannot be drawn.

To our knowledge, only a limited amount of multisensory-evoked data is obtained from the behaving animal. A behavioral study with cats [45] has indicated strong effects of auditory-visual integration on the accuracy of visual stimulus localization. Also, a recent study on the human saccadic system reported systematic effects on saccade latency as a function of the visual-auditory stimulus configuration [12]. In addition, the accuracy of eye-head gaze shifts to either visual or auditory targets was shown to be affected by the task requirement, as well as by the temporal and spatial target configuration [3].

Electrophysiological results from the SC of the behaving cat have indicated a visual–auditory interaction effect on the responses of single units [35]. It was reported that some neurons, having saccade-related activity following the onset of a visually evoked eye movement, became presaccadic for saccades toward combined visual–auditory stimuli. Therefore, the timing of the saccade-related burst changed relative to saccade onset as function of the sensory conditions.

Recently, more experimental data on multimodal integration in the SC of the awake cat have become available 29., 36.. In line with earlier findings from anesthetized preparations (described earlier), it was found that a considerable portion of multisensory activated neurons display an enhancement of the sensory-evoked response for bimodal stimuli in the receptive field. The latter group also reported cells that showed a response suppression under these conditions. However, the multisensory effects on movement-related activity were not part of either study. Also, the possible influence of the spatial/temporal stimulus configuration on the multimodal interaction effects were not investigated.

To our knowledge, no data exist on visual–auditory interactions in SRBNs of the SC of the behaving monkey. Therefore, one of the goals of this article is to describe the influence of an auditory stimulus on the response properties of single units in the SC, when the monkey is required to make visually guided saccadic eye movements to various bimodal stimulus configurations.

The present study also addresses recent ideas that assign a prominent role to the collicular output in the precise encoding of the saccade trajectory and kinematics 5., 32., 52., 54.. Note that these theories necessarily predict unique activity patterns for saccades with identical trajectories and kinematics, regardless of the sensory conditions that evoked them. To test this prediction, we have also compared the observed activity changes of combined visual/auditory stimulation to the kinematics and timing of the bimodally evoked saccadic eye movements.

A preliminary account of our findings has been presented in abstract form [9] and in a succinct review [50].