Spatial receptive field organization of multisensory neurons and its impact on multisensory interactions

Abstract

Previous work has established that the spatial receptive fields (SRFs) of multisensory neurons in the cerebral cortex are strikingly heterogeneous, and that SRF architecture plays an important deterministic role in sensory responsiveness and multisensory integrative capacities. The initial part of this contribution serves to review these findings detailing the key features of SRF organization in cortical multisensory populations by highlighting work from the cat anterior ectosylvian sulcus (AES). In addition, we have recently conducted parallel studies designed to examine SRF architecture in the classic model for multisensory studies, the cat superior colliculus (SC), and we present some of the preliminary observations from the SC here. An examination of individual SC neurons revealed marked similarities between their unisensory (i.e., visual and auditory) SRFs, as well as between these unisensory SRFs and the multisensory SRF. Despite these similarities within individual neurons, different SC neurons had SRFs that ranged from a single area of greatest activation (hot spot) to multiple and spatially discrete hot spots. Similar to cortical multisensory neurons, the interactive profile of SC neurons was correlated strongly to SRF architecture, closely following the principle of inverse effectiveness. Thus, large and often superadditive multisensory response enhancements were typically seen at SRF locations where visual and auditory stimuli were weakly effective. Conversely, subadditive interactions were seen at SRF locations where stimuli were highly effective. Despite the unique functions characteristic of cortical and subcortical multisensory circuits, our results suggest a strong mechanistic interrelationship between SRF microarchitecture and integrative capacity.

Figure 1

Method for constructing a spatial receptive field (SRF) for an individual SC neuron. Green circles represent the tested stimulus locations within the classical excitatory receptive field. Responses at each of these locations are then assembled into the single unit activity (SUA) plots. A single SUA plot at one location is enlarged to show how the spike density function (SDF) is derived. The SUA/SDF data are then transformed into the pseudocolor spatial receptive field (SRF) plot. In this plot the normalized evoked response (scaled to the maximal response) is shown as a function of azimuth (x-axis) and elevation (y-axis), with warmer colors representing higher firing rates.

Figure 2

Representative example of an AES neuron exhibiting substantial changes of response and multisensory interaction as a function of changes in stimulus location. A. Visual, auditory, and multisensory SRFs plotted with each of the 3 representations being normalized to the greatest evoked response and the pseudocolor plots showing the relative activity scaled to this maxima. Symbols relate to the spatial locations of the stimulus pairings represented on the right (B). B. Rasters and spike density functions show the details of this neuron’s responses to the visual stimulus alone (top row), auditory stimulus alone (middle row), and the combined visual—auditory stimulus (bottom row) presented at 3 different azimuthal locations (circle, square, and star on the receptive field plots in A show the stimulus locations; columns show the evoked responses at these 3 different locations). Note the pairing of effective visual and auditory stimuli resulted in no interaction (B, circle and square columns), pairings at a location in which the visual and auditory stimuli were less effective resulted in significant response enhancement (B, star column). (**P<0.01).

Figure 3

Spatiotemporal receptive fields (STRFs) produced from a visual-auditory multisensory neuron recorded from cortical area AES. A. Visual (top), auditory (middle), and multisensory (bottom) STRFs aligned such that the relative timing of the stimuli depicted in the multisensory condition is preserved across panels. B. The difference STRF generated by subtracting the predictive multisensory STRF (linear sum of the visual and auditory STRFs) from the true multisensory STRF. Warmer colors reflect areas where the actual multisensory response exceeds the predicted multisensory response. The curve shown in the top panel represents the magnitude of multisensory integration (%) as the response evolves over time. C. Scatterplot highlights the relationship between response latency and response discharge duration plotted as a function of the stimulus condition. Plus signs represent the mean values for each stimulus condition. Note the leftward and upward shift in the multisensory response relative to the auditory and visual responses, reflecting the speeded and longer lasting responses, respectively.

Figure 4

A. Representative example of the spatial receptive field (SRF) architecture of an individual multisensory superior colliculus neuron. Visual, auditory, and multisensory SRFs are shown at the top, along with the predicted SRF derived by simple addition of the visual and auditory SRFs (V+A) and a contrast plot showing the difference between the actual multisensory response and this predicted response (M - [V + A]). In the pseudocolor plots on the bottom panel, warmer colors represent superadditive interactions and cooler colors represent subadditive interactions. Note the difference in the actual multisensory response when compared with the additive prediction. B. A second example of the spatial receptive field (SRF) architecture of a multisensory superior colliculus neuron. Conventions are the same as in A. Again, note the differences between the predicted and actual multisensory responses, which are best captured in the contrast plot (M — [V + A]). Note also in this example the near absence of an evoked visual response, but the clear modulation apparent in the multisensory profiles.

Figure 5

An example of a SRF analysis in an SC neuron illustrating the relationship between stimulus effectiveness and multisensory integration as a function of space. A. Spatial receptive fields (SRF) for this visual-auditory neuron. Warmer colors indicate higher evoked firing rates. The borders of the multisensory SRF are outlined with a dotted black line in all three panels. B. Stimulus locations for two spatially-coincident multisensory pairings (circle and square) within the SRF. C. Evoked responses for these two locations for the visual, auditory and multisensory conditions. Note that whereas the pairing at a weakly effective location (square) results in a large superadditive multisensory interaction, the pairing at a location in which a vigorous auditory response could be evoked (circle) results in a clear subadditive interaction.

Figure 6

Population analysis for the well-characterized multisensory SC neurons. A. Polar plots depict the mean spatial response distributions for all three conditions as well as for the multisensory interaction. Note the fairly uniform distributions for each of the conditions, yet the clear bias along the horizontal meridian for the integration profile. B. Bar graph shows the magnitude and sign (i.e., > 0 = enhancement; < 0 = depression) for all tested multisensory interactions in this sample of SC neurons. Note that each neuron contributes multiple observations to this distribution because of the large number of sampled locations. C. Mean firing rate distributions across the different conditions. Note the significant gain under multisensory conditions.