Frontal Eye Field Inactivation Diminishes Superior Colliculus Activity, But Delayed Saccadic Accumulation Governs Reaction Time Increases

Abstract

Stochastic accumulator models provide a comprehensive framework for how neural activity could produce behavior. Neural activity within the frontal eye fields (FEFs) and intermediate layers of the superior colliculus (iSC) support such models for saccade initiation by relating variations in saccade reaction time (SRT) to variations in such parameters as baseline, rate of accumulation of activity, and threshold. Here, by recording iSC activity during reversible cryogenic inactivation of the FEF in four male nonhuman primates, we causally tested which parameter(s) best explains concomitant increases in SRT. While FEF inactivation decreased all aspects of ipsilesional iSC activity, decreases in accumulation rate and threshold poorly predicted accompanying increases in SRT. Instead, SRT increases best correlated with delays in the onset of saccade-related accumulation. We conclude that FEF signals govern the onset of saccade-related accumulation within the iSC, and that the onset of accumulation is a relevant parameter for stochastic accumulation models of saccade initiation.SIGNIFICANCE STATEMENT The superior colliculus (SC) and frontal eye fields (FEFs) are two of the best-studied areas in the primate brain. Surprisingly, little is known about what happens in the SC when the FEF is temporarily inactivated. Here, we show that temporary FEF inactivation decreases all aspects of functionally related activity in the SC. This combination of techniques also enabled us to relate changes in SC activity to concomitant increases in saccadic reaction time (SRT). Although stochastic accumulator models relate SRT increases to reduced rates of accumulation or increases in threshold, such changes were not observed in the SC. Instead, FEF inactivation delayed the onset of saccade-related accumulation, emphasizing the importance of this parameter in biologically plausible models of saccade initiation.

Keywords: computational models; frontal eye fields; reversible inactivation; saccade; superior colliculus.

Figure 1.

Materials and methods. A, Cryoloops were inserted into the inferior (IA) and superior (SA) arm of the arcuate sulcus. B, Response field centers for iSC neurons recorded in this study, plotted on the SC map of Hafed and Chen (2016). C, FEF inactivation increased SRTs for contraversive and occasionally ipsiversive saccades. Each line connects mean SRT (±SE) across precooling, pericooling, and postcooling sessions for each of the four monkeys; solid lines indicate significant differences (p < 0.025, Wilcoxon signed-rank test). D, Functional classification of recorded neurons (see Materials and Methods). E, F, The matched-saccade analysis compared saccades of very similar eye position and velocity profiles but different SRTs across the FEF warm or FEF cool conditions (all saccades came from the same session; see Materials and Methods). E shows one example of a saccade match; F shows characteristics of the 3762 matched saccade pairs (pooled across both ipsiversive and contraversive saccades).

Figure 2.

FEF inactivation decreased visual activity of ipsilesional iSC neurons. A, Spike rasters (bottom) and mean spike-density functions (top) showing reduced visual response to peripheral cue onset on ipsilesional iSC neuron O1 with FEF inactivation (FP, fixation point; T, target). B, C, FEF inactivation decreased activity in the 50 ms interval following the start of the visual response in the ipsilesional (black circles) but not contralesional (green circles) iSC (B), without altering visual response latency in either iSC (C; line represents line of unity; p value shows results of Wilcoxon signed-rank test). D, FEF inactivation decreased visual responses of neurons exhibiting other functional responses (left axis, black circles, percentage change ± SE), but did not alter visual response latency (right axis, gray squares; difference ± SE). Filled symbols represent significant effects using Wilcoxon signed-rank test (p < 0.05).

Figure 3.

FEF inactivation decreased delay-period activity of ipsilesional iSC neurons. A, B, FEF inactivation nearly abolished modest delay-period activity in ipsilesional iSC neuron D1 during a memory-guided saccade (A), and reduced delay-period activity in ipsilesional iSC neuron O2 in both the visually and memory-guided tasks (B). C, D, FEF inactivation consistently decreased delay-period activity in the last 100 ms before peripheral cue offset for both visually and memory-guided tasks in ipsilesional (C) but not contralesional (D) iSC neurons (squares or circles denote neurons also displaying build-up activity or not, respectively; same general format as Fig. 2).

Figure 4.

FEF inactivation decreased saccade-related activity of ipsilesional iSC neurons. A, FEF inactivation decreased saccade-related activity in ipsilesional iSC neuron O3 (inset shows position and velocity profiles for matched visually guided saccades). B, FEF inactivation consistently decreased saccade-related activity (8 ms before saccade onset to 8 ms before saccade offset; see schematic) for ipsilesional iSC for both visually and memory-guided saccades (same format as Fig. 2B; neurons included only if they had ≥5 matched saccades). C, Direct comparison of saccade-related activity for all matched contralesional saccades. FEF inactivation generally decreased saccade-related activity (top row, blue). As a control, we also matched saccades from FEF warm trials (bottom row, red) and did not find consistently decreased saccade-related activity. D, FEF inactivation did not consistently influence saccade-related activity in the contralesional iSC.

Figure 5.

FEF inactivation reduced all aspects of activity in the ipsilesional iSC. For both visually and memory-guided saccade tasks, FEF inactivation reduced the firing rate (mean ± SE) across ipsilesional iSC neurons possessing visual, delay-period, and/or saccade-related activity.

Figure 6.

Inactivation-induced changes in SRT correlated with delays in the onset of saccade-related activity in the ipsilesional iSC. A, Depiction of how onset time is detected using a piecewise two-piece linear-regression approach for two trials (see Materials and Methods). The onset time (dotted line) coincides with the inflection point that minimizes the summed squared error (gray curve, plotted against right axis) between convolved iSC activity (top, spike train shown below) and the two linear regressions (green lines). Note how the first linear regression captured any delay-period or build-up activity before the saccade (SRT is represented by the circle above raster plot). B, For example ipsilesional neuron O4, FEF inactivation delayed the onset of saccade-related activity (ticks) and SRT (circles) for matched memory-guided saccades. Top part shows spike-density function and kinematics (inset) of one matched pair; bottom part shows rasters for all matched saccades. C, FEF inactivation generally delayed the onset of saccade-related activity in ipsilesional iSC neurons. Each point shows the average change per neuron (across trials with matched saccades), with filled or open circles denoting SRT increases or decreases, respectively. D, Changes in the onset of saccade-related activity strongly correlated with concomitant changes in SRT.

Figure 7.

FEF inactivation changed other aspects of iSC activity, but these changes poorly predicted changes in SRT. A, In example neuron O4, FEF inactivation decreased both the accumulation rate and threshold activities (see Materials and Methods for how these parameters were measured) for matched memory-guided saccades. Same format as Figure 6B. B–D, Across our sample, FEF inactivation decreased baseline activity (B), threshold activity (C), and accumulation rate (D) of ipsilesional iSC neurons, particularly for memory-guided saccades. E, F, To analyze how changes in these parameters related to SRT differences in a one-to-one manner, we computed the time-to-reach-threshold as the difference of threshold and baseline activities divided by the accumulation rate. Note how this parameter does not directly incorporate the onset of accumulation, so it assumes that activity starts to accumulate at an arbitrary point in time after the go-cue. While such changes did increase the time-to-reach-threshold in the ipsilesional iSC (E), such changes did not fully account for the concomitant changes in SRT (F).

Figure 8.

Across all matched trials, changes in the onset of accumulation best reflected changes in SRT, even without FEF inactivation. A, Across all matched trials extracted with (top row, blue lines) and without FEF inactivation (bottom row, red lines), differences in the onset of accumulation (left column) in the ipsilesional iSC related better to associated changes in SRT, compared with differences in the time-to-reach-threshold (right column). B, Amount of SRT variance explained by different combinations of individual or grouped parameters extracted from a rise-to-threshold model. Regardless of whether matched pairs were extracted across FEF inactivation or not, consideration of the change in the onset of accumulation greatly increased how well changes in iSC activity predicted concomitant changes in SRT. C, Across trials matched with or without FEF inactivation, the onset of accumulation best correlated with the remaining residual error following a multiple linear regression of the other individual parameters and SRT.