Of the many functions ascribed to the dorsolateral prefrontal cortex (DLPFC), the ability to override automatic stimulus-driven behavior is one of the most prominent. This ability has been investigated extensively with the antisaccade task, which requires suppression of saccades toward suddenly appearing visual stimuli. Convergent lines of evidence have supported a model in which the DLPFC suppresses unwanted saccades by inhibiting saccade-related activity in the ipsilateral superior colliculus (SC), a midbrain oculomotor structure. Here, we carried out a direct test of this inhibitory model using unilateral cryogenic deactivation of the DLPFC within the caudal principal sulcus (cPS) and simultaneous single-neuron recording of SC saccade-related neurons in monkeys performing saccades and antisaccades. Contrary to the inhibition model, which predicts that attenuation of inhibition effected by unilateral cPS deactivation should result in activity increases in ipsilateral and decreases in contralateral SC, we observed a delayed onset of saccade-related activity in the ipsilateral SC, and activity increases in the contralateral SC. These effects were mirrored by increased error rates of ipsiversive antisaccades, and reaction times of contraversive saccades. These data challenge the inhibitory model and suggest instead that the primary influence of the DLPFC on the SC is excitatory.
Our ability to carry out appropriate behavior in the face of a rapidly and constantly changing environment depends on the ability to control responses on the basis of internal states rather than the immediate sensory characteristics of external stimuli. The cognitive substrates of such behavior are collectively referred to as cognitive or “executive” control. A well-established aspect of such control is the ability to withhold automatic stimulus-driven actions in favor of less potent, but more advantageous responses. Multiple lines of evidence have established the prefrontal cortex (PFC) as a critical element in the neural substrate of response inhibition (see for review Aron et al. 2004; Schall and Boucher 2007).
In the oculomotor domain, response inhibition has been extensively investigated using the antisaccade task, which requires subjects to override the automatic tendency to look toward a suddenly appearing peripheral visual stimulus and look instead in the opposite direction (Hallett 1978; Munoz and Everling 2004). The classic conceptual model of antisaccade performance holds that the PFC facilitates correct responses by inhibiting task-inappropriate prosaccades toward the visual stimulus (Pierrot-Deseilligny et al. 1991). This model has been supported by human lesion (Guitton et al. 1985; Pierrot-Deseilligny et al. 1991, 2003; Walker et al. 1998; Ploner et al. 2005) and transcranial magnetic stimulation studies (Nyffeler et al. 2007), which have shown that PFC damage or deactivation leads to increased numbers of erroneous prosaccades on antisaccade trials. Findings of human functional magnetic resonance imaging (fMRI) studies are also broadly consistent with this idea; greater blood oxygen level-dependent (BOLD) responses are observed in the DLPFC on anti- when compared with prosaccade trials (Sweeney et al. 1996; Desouza et al. 2002; McDowell et al. 2002; Ford et al. 2005; Brown et al. 2007; Ettinger et al. 2008).
According to the inhibition model, the neural mechanism underlying suppression of unwanted saccades is a PFC-mediated attenuation of activity in neurons in cortical and subcortical oculomotor areas such as the frontal eye fields (FEFs) and midbrain superior colliculus (SC; Munoz and Everling 2004). Single-neuron recordings in rhesus macaques have revealed signals consistent with this notion. DLPFC neurons exhibit task-selective activity on pro- and antisaccade trials (Funahashi et al. 1993; Everling and Desouza 2005; Johnston and Everling 2006a, 2006b; Johnston et al. 2009), activity of neurons in the FEF (Everling and Munoz 2000) and SC (Everling et al. 1999) is lower on anti- than prosaccade trials, and antidromic identification of corticotectal neurons has revealed that task-selective signals are sent from the DLPFC to the SC (Johnston and Everling 2006a). Based on these findings, together with the fact that DLPFC projection neurons target the ipsilateral SC (Goldman and Nauta 1976; Leichnetz et al. 1981; Fries, 1984), and that corticotectal projections are excitatory (Tsumoto 1990; Dori et al. 1992). Johnston and Everling (2006a) proposed that the neural mechanism for DLPFC-mediated suppression of saccades must be via excitation of either fixation neurons in the rostral SC, or inhibitory interneurons in the ipsilateral SC, which suppress the activity of ipsilateral SC saccade neurons (Fig. 1A). This hypothetical model proposes that the enhanced DLPFC signal observed on antisaccade trials acts to suppress inappropriate contraversive saccades and to facilitate task-appropriate ipsiversive saccades.
Two studies using different causal manipulations of DLPFC activity in the macaque have produced results inconsistent with this inhibition model. First, unilateral deactivations effected by injections of the γ-aminobutyric acid agonist muscimol into the ventral or dorsal bank of the caudal principal sulcus (cPS) resulted in a failure to suppress task-inappropriate ipsiversive saccades, while the ability to suppress inappropriate contraversive saccades improved (Condy et al. 2007). Secondly, electrical microstimulation of the DLPFC impaired the ability to suppress task-inappropriate contraversive saccades (Wegener et al. 2008). Both results are opposite to those predicted by the inhibition model. In contrast to evidence from single-neuron recordings, these studies suggest a net excitatory effect of DLPFC inputs on the ipsilateral SC—that DLPFC deactivation decreases, and microstimulation increases activity of ipsilateral SC saccade-related neurons. This suggests that the DLPFC may not act to suppress inappropriate saccades by inhibiting activity of ipsilateral SC saccade neurons as suggested by the classical inhibition model, but might instead facilitate task-appropriate saccades via a direct excitatory effect on ipsilateral SC saccade-related neurons (Fig. 1B).
To investigate directly the influence of DLPFC signals on the SC, we implanted cryoloops to unilaterally deactivate the cortex of the cPS and simultaneously recorded the activity of SC saccade-related neurons, while rhesus macaques performed pro- and antisaccades. In a previous report focused on the investigation of the effects of bilateral DLPFC deactivation on SC activity during rule visible and rule memorized tasks, we reported results from an earlier data set that showed that unilateral DLPFC deactivation resulted in a decrease in activity in the ipsilateral SC and an increase in the contralateral SC (Koval et al. 2011). Here, we report an analysis of a new data set collected during randomly interleaved overlap and gap trials. The gap condition, in which the fixation spot is extinguished prior to visual stimulus appearance, was included to increase task difficulty (Bell et al. 2000). We show that prestimulus and stimulus-related activities are increased in the SC contralateral to the deactivated hemisphere, that these increases are persistent, and that they are accompanied by increases in the onset latency of saccade-related activity in the SC ipsilateral to the deactivated hemisphere. These changes are mirrored by increased error rates of ipsiversive antisaccades and increased reaction times of contraversive saccades. Altogether, these data support a model in which the DLPFC controls oculomotor behavior via an excitatory influence on the SC.
Materials and Methods
Three male macaque monkeys (Macaca mulatta, 9–16 kg) were prepared for chronic cPS deactivation experiments and single-neuron recordings in the SC using previously described techniques (Johnston and Everling 2006b). All procedures were carried out in accordance with the guidelines of the Canadian Council of Animal Care Policy on the Use of Laboratory Animals, and a protocol approved by the Animal Use Subcommittee of the University of Western Ontario Council on Animal Care. Briefly, monkeys underwent 2 aseptic surgical procedures. Animals received analgesics and antibiotics postoperatively and were closely monitored by a university veterinarian. In the first surgery, a plastic head restraint and a recording chamber were implanted. The recording chamber was centered on the midline and tilted 38° posterior of vertical to allow recordings from neurons in the SC. Monkeys were then trained to perform randomly interleaved pro- and antisaccade trials. This design was chosen due to the fact that previous neurophysiological studies using this approach have shown robust activity differences in PFC, FEF, and SC activities between pro- and antisaccade trials (Everling et al. 1999; Everling and Munoz 2000; Everling and DeSouza 2005; Johnston and Everling 2006b), and because the technical challenges of simultaneous cooling and SC recording and necessity of separate precooling, cooling, and postcooling periods placed a premium on the ability to efficiently collect a maximal roughly equivalent number of pro- and antisaccade trials. Once the animals achieved stable task performance, anatomical MR images were obtained to visualize the location of the implanted recording chambers and the shape of the principal sulci. Following this, animals underwent a second surgery in which stainless steel cryoloops (6 × 3 mm) were implanted bilaterally into the cPS in each animal (Fig. 2A). We targeted the cPS for deactivation, as this area has been shown to be the likely macaque homolog of the human middle frontal gyrus (MFG; Hutchison et al. 2012). Technical details of the cryloop technique have been previously described (Lomber et al. 1999).
During each experiment, the response field (RF) of a single SC neuron was mapped, and the animals commenced a task in which they were required to perform randomly interleaved pro- and antisaccade trials. Each trial began with the presentation of a colored, central fixation point (FP), which instructed the animal which of the 2 types of responses was required (Fig. 2B). For 2 monkeys (A and C), a green FP signaled a prosaccade trial and a red FP signaled an antisaccade trial. Color instructions were reversed for monkey B. Monkeys were required to fixate the central FP within a 0.5° × 0.5° window for a variable period of between 700 and 900 ms at the beginning of each trial. On 50% of trials, an FP was extinguished 200 ms prior to stimulus presentation (gap condition). On the remaining 50%, the FP remained present for the duration of the trial (overlap condition). Following this, a visual stimulus (0.15°) was pseudorandomly presented with equal probability either into the neuron's RF or at the mirror location on the opposite side of the vertical and horizontal meridian. Animals were required to maintain fixation throughout the fixation and gap periods and to generate a saccade either toward the visual stimulus (prosaccade trials) or to a location diametrically opposite to the stimulus (antisaccade trials) within 500 ms to obtain a liquid reward. Saccade endpoints were required to fall within a 5° × 5° window. Both gap and overlap conditions were included to manipulate of the inhibitory demand of the task. It has been previously shown that saccadic reaction times are reduced, and direction errors increased on gap when compared with overlap antisaccade trials (Bell et al. 2000), which has been interpreted as reflecting an increase in “inhibitory load” for antisaccades in the gap condition (Curtis et al. 2001). Horizontal and vertical eye movements were recorded at 500 Hz using a high-speed video eye tracker (EyeLink II, Kanata, ON, Canada). The task, behavior monitoring, and reward delivery were controlled using CORTEX (NIMH, Bethesda, MA, USA) running on two Pentium PCs. Monkeys received water until satiation, after which they were returned to their home cages. Daily records were kept of the weight and health status of the monkeys, and additional water and fruit were provided as needed.
Reversible Cryogenic Deactivation
Cryoloops were constructed from 23-gauge hypodermic stainless steel tubing, designed to deactivate the banks of the cPS, and implanted bilaterally in the principal sulci. The cPS was deactivated by pumping methanol through Teflon tubing connected to the cryoloops, which passed through a methanol ice bath that was reduced to subzero temperatures by the addition of dry ice. After passing through the cryoloop, methanol was returned to the same reservoir from which it came (Fig. 2A). Chilled methanol pumped through a cryoloop inactivates adjacent cortical tissue by disrupting synaptic activity. Given that cortical temperature increases rapidly with a distance from a cryoloop (10°C/mm), and evoked neural activity is absent in cortical tissue cooled below 20°C, we chose to maintain the cryoloop temperature at 1–3°C to inactivate as large an area of cortical tissue as possible, while avoiding potentially harmful subzero temperatures at the cortical surface (Lomber et al. 1999). Previous 2DG experiments have shown that the effective spread of cooling is restricted to roughly 2 mm on either side of the loop (Payne and Lomber 1999). Each of our cryoloops measured 6 mm by 3 mm. Because cryoloops were inserted in the principal sulcus, the volume of cortex deactivated by the loops can be approximated as the volume of a box with dimensions of 10 × 7 × 4 mm, or 280 mm3. Based on this estimate, the deactivated region included the cortex of the dorsal and ventral banks to the depth of the principal sulcus over the caudal half of its length.
Each cooling session consisted of precool, cooling, and postcool periods that ranged from 10 to 15 min in duration. A cooling session started with a precool period, after which the pump was turned on. It took an average of 85 s for the cryoloop to reach the desired temperature range of 1–3°C. Temperature was monitored with an attached thermocouple and maintained by adjusting the pump flow rate. We excluded the first 3 min after the pumps were turned on to ensure that the cortical tissue adjacent to the cryoloop was cooled to a temperature below 20°C, at which neurons are deactivated (Adey 1974). At the end of the cooling period, the pump was turned off and cryoloop temperature returned rapidly to normal. Cryoloop temperature reached 30°C within approximately 40 s. Data collected during a rewarming period consisting of the first 3 min after the pumps were turned off were excluded from all data analysis.
Extracellular single-unit activity was recorded in the intermediate layers of the caudal SC (Fig. 2A) with a hydraulically operated microdrive (Narishige International USA, Inc., East Meadow, NY, USA), which guided a tungsten microelectrode (UEWLFELMNN1E, FHC Inst., Bowdoinham, Maine) through a 23-gauge stainless steel guide tube that was positioned inside a plastic grid with 1-mm spacing between adjacent locations (Crist Instrument, Inc., Hagerstown, MD, USA). The intermediate layers were identified using single-neuron recordings and microstimulation (Everling et al. 1999). Neural activity was amplified, filtered, and stored by a Plexon multichannel acquisition processor system (Plexon, Inc., Dallas, TX, USA) for offline cluster separation using principal component analysis.
We examined the effects of unilateral cPS deactivation on the activity of SC saccade-related neurons. To be classified as a saccade-related neuron, a neuron had to be located 1–3 mm below the dorsal surface of the SC (determined as the electrode depth where visual background activity was first detected) and discharge above 100 spikes/s for prosaccades (gap and overlap conditions combined) into the neuron's RF during a 50-ms epoch centered around saccade onset. Saccade-related neurons were classified as buildup neurons if, in addition, they exhibited low-frequency prestimulus activity at the end of the gap epoch (50 ms before to 50 ms after stimulus presentation) that was significantly greater than during the visual fixation epoch (100-ms epoch before FP disappearance, paired t-test, P < 0.01; Munoz and Wurtz 1995; Dorris et al. 1997) on prosaccade trials in the gap condition.
Eye Movement Analysis
All analyses were performed offline using custom-written software in Matlab (Mathworks, Natick, MA, USA). For prosaccade trials, trials on which the animals made a saccade toward the location of the visual stimulus were classified as correct, while those made to the opposite location were classified as errors. For antisaccade trials, saccades to the location opposite the visual stimulus were classified as correct, while those toward the visual stimulus were classified as errors. Saccade onset was defined as the time at which the radial eye velocity exceeded 30 °/s, and the endpoint was taken to be the time at which this parameter fell below this value. Accurate saccade onset and offset on each trial and correct and error trial categorization by CORTEX was verified offline by visually examining the eye traces from each session and manually correcting or removing any erroneously categorized saccades.
Spike Density Function
To evaluate the relationship between neural activity and stimulus onset and saccade onset, continuous spike density functions were constructed. The activation waveform was obtained by convolving each spike with an asymmetric function that resembled a postsynaptic potential (Hanes and Schall 1996; Thompson et al. 1996). The advantage of this function over a standard Gaussian function (Richmond and Optican 1987) is that it accounts for the fact that spikes exert an effect forward, but not backward in time. We computed within-subject standard errors of the mean discharge rate at millisecond resolution for each function using the method of Loftus and Masson (1994).
Timecourse of cPS Deactivation Effects
To determine the timecourse of the effects of cPS deactivation on the discharge rate of the population of SC neurons, we performed sliding receiver operating characteristics (ROC) analyses comparing discharge rates observed in the cPS+ and cPS− conditions. For analyses of the timecourse of cPS deactivation on SC activity relative to visual stimulus onset, a series of ROC values quantifying the difference in discharge rates between the cPS+ and cPS− conditions was computed using the discharge rates obtained from the convolved spike trains. The values were calculated within a series of 10-ms epochs, slid in 1-ms increments, for the time interval beginning 200 ms prior to and ending 300 ms after peripheral stimulus onset. A single ROC timecourse, reflecting the difference in discharge rates between the cPS+ and cPS− conditions, was constructed separately for each neuron. These multiple single-neuron timecourses were then averaged to obtain a single ROC timecourse for the population of SC neurons. The ROC values in this averaged timecourse reflect the probability that an ideal observer could correctly categorize a discharge rate from the population of SC neurons as belonging to the cPS+ or cPS− condition at a given point in time. Statistical significance of ROC values was evaluated using a bootstrap analysis, which proceeded as follows: For each neuron, the 2 activation conditions (cPS+ and cPS−) were either exchanged, or left unchanged, with 50% probability, and a single average ROC timecourse was computed. This procedure was repeated 1000 times. The 95th and 5th percentile values of the distribution of 1000 average ROC values for each time point were used to indicate the 5% significance criterion. Both were plotted together with the average ROC timecourse of the nonrandomized data.
Onset of Motor Activity
The onset of the motor burst (P < 0.01) was determined using a Poisson spike train analysis (Hanes et al. 1995), implemented using Matlab code developed by the Schall laboratory (http://www.psy.vanderbilt.edu/faculty/schall/scientific-tools/).
Unilateral cPS Deactivation Affects Reaction Times and Error Rates
We recorded from 52 neurons (27 saccade-related neurons, 14 of which showed buildup activity) in the SC of 3 monkeys, while we deactivated the ipsilateral cPS and from 43 neurons (26 saccade-related neurons, 19 of which showed buildup activity), while we deactivated the contralateral cPS. In each of the 95 experimental sessions, we recorded the activity of a single SC neuron, while monkeys performed the randomly interleaved pro-/antisaccade task. After a 10- to 15-min precool period, the cPS region was then deactivated unilaterally for 10–15 min (Fig. 2), while the task continued. Following cooling, we also continued to record neural activity for at least 10 min during a postcooling period.
We observed effects of cPS cooling on reaction times and error rates of pro- and antisaccades in both gap and overlap conditions (Fig. 3). Consistent with previous studies, we observed decreased reaction times of pro- and antisaccades in the gap compared with the overlap condition (Everling et al. 1999; Bell et al. 2000; Everling and Munoz 2000). More importantly for the present study, unilateral cPS deactivation significantly increased reaction times of saccades directed to the hemifield contralateral to the deactivated cPS for both pro- and antisaccade trials, in both the gap and overlap conditions [Fig. 3A; P < 0.00001, analysis of variance (ANOVA)]. For saccades directed to the hemifield ipsilateral to the deactivated cPS, we found a significant decrease in prosaccade reaction times during the deactivation period in the gap and overlap conditions (P < 0.00001, ANOVA), whereas there was no effect on antisaccade reaction times in the gap condition and a marginal increase in reaction times in the overlap condition (P < 0.05, ANOVA).
We also calculated the effects of unilateral cPS deactivation on error rates (Fig. 3B). For this measure, we included only trials in which the monkeys commenced central fixation, maintained fixation throughout the fixation and gap periods, and made a saccade, either toward or away from the peripheral stimulus to the opposite direction. Consistent with the previous reports in monkeys and humans, we observed more direction errors on the anti- than prosaccade task and more errors in the gap than overlap condition (Everling et al. 1998, 1999; Bell et al. 2000; Everling and Munoz 2000). Unilateral deactivation of the cPS increased error rates in the gap condition on prosaccade trials and in both the gap and overlap conditions on antisaccade trials (F < 0.00001, ANOVA), such that monkeys often made erroneous ipsiversive saccades. Consistent with this increase in errors, the animals' performance actually improved during cPS deactivation on trials on which an ipsiversive saccade was the correct response. This effect was very pronounced in the gap condition for pro- and antisaccades (P < 0.00001, ANOVA) and also significant for antisaccades in the overlap condition (P < 0.05, ANOVA). While unilateral cPS deactivation had significant effects on reaction times and error rates, more general aspects of task performance remained unaffected (percentage of skipped trials, fixation breaks, or no response trials), indicating that the motivation and vigilance of the animals was not impaired (Fig. 4).
Together, these behavioral results show that unilateral deactivation of the cPS increased reaction times of contraversive pro- and antisaccades and decreased reaction times of ipsiversive prosaccades. Monkeys also made more erroneous ipsiversive saccades during deactivation. These results suggest that temporary unilateral deactivation of the cortex in the cPS decreased neural activity in ipsilateral and increased activity in contralateral saccade-related areas. In the following sections, we present an analysis of the effects of unilateral cPS deactivation on the activity of neurons in the SC intermediate layers. We compared the pooled pre- and postcooling data (cPS+ period) with the cooling period data, during which the cPS was deactivated (cPS− period). Population neural activity for the pre-, cooling, and postcooling epochs is presented in Supplementary Figure 1.
cPS Deactivation Increases Preparatory Activity in the Contralateral SC
To investigate the effects of unilateral cPS deactivation on SC preparatory activity, we compared the levels of activity of buildup neurons in the period from 50 before to 50 ms after stimulus presentation (see Materials and Methods) during the cPS+ and cPS− periods for pro- and antisaccade trials. SC buildup neurons were selected for this analysis because these saccade-related neurons exhibit prominent preparatory activity, that is, negatively correlated with saccadic reaction times (Dorris et al. 1997; Everling et al. 1999). For this analysis, we combined trials in which the subsequent stimulus appeared into or opposite to the neuron's RF. We also included both correct and error trials to capture the full effects of unilateral cPS deactivation on preparatory activity. Figure 5A shows the activity of 14 buildup neurons on prosaccade trials in sessions in which the ipsilateral cPS was deactivated. Consistent with previous reports (Dorris et al. 1997; Everling et al. 1999), the increase in activity of buildup neurons was greater in the gap than overlap condition. Although deactivation of the ipsilateral cPS had no significant effect on the overall activity of the sample of buildup neurons in the gap (37.1 ± 6.44 vs. 33.45 ± 6.03 spikes/s, P = 0.15; Wilcoxon signed-rank test) or overlap condition (25.79 ± 5.67 vs. 25.25 ± 5.37 spikes/s, P = 0.77; Wilcoxon signed-rank test), 7 of 14 (50%) neurons showed the significantly reduced levels of prestimulus activity during ipsilateral cPS deactivation in the gap condition (P < 0.05, Wilcoxon rank-sum test). There were no effects of ipsilateral cPS deactivation on the activity of SC buildup neurons on antisaccade trials in the gap (31.24 ± 5.55 vs. 30.50 ± 5.47 spikes/s, P = 0.70; Wilcoxon signed-rank test) or overlap condition (19.87 ± 4.63 vs. 21.16 ± 4.72 spikes/s, P = 0.29 Wilcoxon signed-rank test; Fig. 5B).
On sessions in which we deactivated the cPS contralateral to the recorded SC buildup neurons, we observed significant increases in the levels of prestimulus activity in the sample of 19 buildup neurons on prosaccade trials in the gap (39.15 ± 4.92 vs. 43.51 ± 5.63 spikes/s, P < 0.05) and overlap conditions (25.7 ± 4.27 vs. 29.43 ± 5.02 spikes/s, P < 0.05, Wilcoxon signed-rank test; Fig. 5C). We found the same effect on antisaccade trials for the gap (27.85 ± 4.39 vs. 31.84 ± 4.53 spikes/s, P < 0.05, Wilcoxon signed-rank test) and overlap conditions (17.73 ± 3.82 vs. 20.81 ± 3.85, P < 0.005, Wilcoxon signed-rank test; Fig. 5D).
These results show that unilateral cPS deactivation increased prestimulus activity in the contralateral SC on both pro- and antisaccade trials, and that there was a trend toward the decreased levels of prestimulus activity in the ipsilateral SC on gap prosaccade trials.
Preparatory Activity in the Contralateral SC Persists During Visual Stimulation Following cPS Deactivation
The previous analysis showed that cPS inactivation elevated the level of preparatory activity of contralateral SC buildup neurons. To test whether these differences extended into the visual stimulation period, we measured the level of activity in the period from 50 to 150 ms after stimulus onset on correct prosaccade trials on which the stimulus was presented opposite to the neurons' RF. Figure 6A shows the activity for the sample of 14 SC buildup neurons when the ipsilateral cPS was deactivated. Preparatory neural activity declined approximately 50 ms after stimulus onset, which coincided with the onset of stimulus-related activity of SC neurons (Everling et al. 1999). We found no differences between the cPS+ (red lines) and cPS− conditions (blue lines) on gap (30.02 ± 5.65 vs. 29.03 spikes/s, P = 0.75, Wilcoxon signed-rank test) or overlap trials (23.9 ± 5.8 vs. 23.25 ± 5.19 spikes/s, P = 0.72, Wilcoxon signed-rank test). On sessions in which the contralateral cPS was deactivated (Fig. 6B), the activity of buildup neurons remained higher during the cPS− (blue lines) than the cPS+ (red lines) period on gap (39.96 ± 4.67 vs. 28.15 ± 4.41 spikes/s, P < 0.002, Wilcoxon signed-rank test) and overlap trials ( 26.52 ± 4.84 vs. 22.1 ± 4.21 spikes/s, P < 0.02, Wilcoxon signed-rank test). These findings show that the elevated preparatory activity in the SC contralateral to the deactivated cPS persisted into the visual stimulation period on prosaccade trials. This finding resembles the nature of cooling-induced changes in stimulus-related activity we observed in the contralateral SC and indicates that silencing of inputs causes generally sustained changes in SC activity.
cPS Deactivation Increases Stimulus-Related Activity in the Contralateral SC on Antisaccade Trials
We next investigated whether unilateral cPS deactivation affected the activity of saccade-related SC neurons when a visual stimulus was presented in their RF on antisaccade trials. We included both correct and error trials in this analysis to determine whether cooling-induced changes in stimulus-related activity of SC neurons could account for the increased rates of erroneous saccades toward the cooled hemisphere we observed. Figure 7A depicts the average activity of 27 saccade-related SC neurons on both the cPS+ (red lines) and cPS− trials (blue lines) on sessions during which the ipsilateral cPS was deactivated. To determine the timecourse of deactivation effects on SC stimulus-related activity, we conducted ROC analyses and tested the significance of ROC values using bootstrap tests (see Materials and Methods). Consistent with our previous analysis of preparatory activity, cPS deactivation caused only brief reductions of stimulus-related activity in the ipsilateral SC on gap trials. Otherwise, unilateral cPS deactivation did not alter stimulus-related neural activity in ipsilateral SC saccade-related neurons.
In contrast, when we examined the effects of cPS deactivation on the stimulus-related activity of saccade-related neurons in the contralateral SC (Fig. 7B), we observed increases in neural activity between 80 and 200 ms after stimulus onset. These results demonstrate that unilateral deactivation of the cPS increased stimulus-related activity in the contralateral SC on antisaccade trials.
Previous studies have demonstrated that increased prestimulus activity and increases in stimulus-related responses of SC neurons are correlated with errors on antisaccade trials (Everling et al. 1998). To evaluate whether antisaccade errors observed during contralateral cPS deactivation were also a result of cooling-induced increases in prestimulus and stimulus-related activities, we compared neural activity for correct and error trials during noncooling and cooling trials using ROC analysis. For this analysis, we used data for gap trials and included only SC saccade-related neurons meeting the criterion of at least 5 correct and 5 error trials. Figure 8 shows the population activity of 10 SC neurons on correct trials (blue line) and error trials (red line). On correct noncooling and cooling trials, these neurons displayed a stimulus-related response that was quickly suppressed. On erroneous noncooled trials, activity increased during the prestimulus period, as well as during and after the initial stimulus-related response. On erroneous cooled trials, activity increased during the stimulus-related response. It is important to note that the increase in activity beginning roughly 150 ms following stimulus onset corresponds to a saccade-related increase in activity since, on erroneous antisaccade trials, a prosaccade is generated into the RF. Together, these results show that erroneous responses were associated with increased levels of activity in SC saccade-related neurons.
cPS Deactivation Shifts the Onset Times of SC Saccade-Related Activity
Our behavioral analysis showed that temporary deactivation of the cPS increased contraversive and decreased ipsiversive reaction times on prosaccade trials (see above). To determine whether these differences in reaction times had neural correlates in the activity of SC saccade-related neurons, we compared the onset times of saccade-related activity in SC neurons between the cPS+ and cPS− conditions using a Poisson spike train analysis (see Materials and Methods). We included only SC saccade-related neurons with no or little stimulus-related activity in this analysis (see Materials and Methods) to ensure that the algorithm detected the onset of the saccadic motor burst and not stimulus-related activity. In sessions in which the ipsilateral cPS was deactivated, the pooled activity of this subset of 16 SC saccade-related neurons increased later in the cPS− (blue lines) than cPS+ (red lines) condition in both gap (thin lines) and overlap conditions (thick lines; Fig. 9A). Consistent with this, the Poisson spike train analysis showed that the saccade burst began later in the cPS− than cPS+ condition on both gap (Fig. 9B; 192 ± 19.24 vs. 169.34 ± 17.86 ms; P < 0.05, Wilcoxon signed-rank test) and overlap trials (Fig. 9C; 253.5 ± 20.29 vs. 220.34 ± 13.84 ms, P < 0.01, Wilcoxon signed-rank test). These differences reached statistical significance in 4 of 16 (25%) neurons in the gap task and in 7 of the 16 (44%) neurons in the overlap task. Conversely, activity seemed to begin earlier in the contralateral SC on cPS− trials (Fig. 9D). These differences were significant in 2 of 11 (18%) SC neurons in the gap condition and in 3 of the 11 (27%) SC neurons in the overlap condition (P < 0.05, Wilcoxon rank-sum test), but did not reach significance for the population in the gap (Fig. 9E, 115.99 ± 15.59 vs. 133.17 ± 14.73 spikes/s; P = 0.46, Wilcoxon signed-rank test) or overlap condition (Fig. 9F, 174.65 ± 21.5 2 vs. 193.64 ± 17.13 spikes/s, P = .10, Wilcoxon signed-rank test). These data show that PS deactivation delayed the onset of saccade-related activity in the ipsilateral SC.
Saccade Threshold is Unaffected by cPS Deactivation
Our previous analyses showed that cPS deactivation increases preparatory activity of saccade-related neurons in the SC contralateral to the deactivated hemisphere and decreases in that of SC neurons to the ipsilateral hemisphere. We further showed that the onset of the motor signal was delayed in the SC ipsilateral to cPS deactivation. Here, we tested whether unilateral cooling of the cPS affected the level of presaccadic motor activity. This analysis was carried out for all SC saccade-related neurons on correct overlap prosaccade trials in which the saccade was directed into the neurons' RF. Based on previous physiological and anatomical studies, the latest time interval at which saccade initiation can be influenced by a neural signal from the SC is 18–8 ms prior to saccade onset (Segraves and Goldberg 1987; Munoz and Wurtz 1993; Miyashita and Hikosaka 1996; Munoz et al. 1996). Figure 10 shows that the activity in this time window did not differ between the cPS+ and cPS− trials for the population of neurons in the ipsilateral (Fig. 10A) or contralateral (Fig. 10B) SC (P > 0.05, Wilcoxon signed-rank test). This result indicates that the saccade motor threshold was not altered by unilateral cPS deactivation.
The dominant model of voluntary saccade control postulates that excitatory signals from the DLPFC act to inhibit reflexive visually-guided saccades via projections to the ipsilateral SC. Such projections are hypothesized to terminate on either rostral fixation or caudal inhibitory interneurons (Isa and Hall 2009), which in turn attenuate the activity of SC saccade-related neurons (Pierrot-Deseilligny et al. 1991, 2002; Gaymard et al. 1998, 2003; Munoz and Everling 2004; Ploner et al. 2005; Condy et al. 2007; Nyffeler et al. 2007). This model predicts that DLPFC deactivation should lead to an increase in activity of saccade-related neurons in the ipsilateral SC and a concomitant decrease in activity of saccade-related neurons in the contralateral SC as a result of an increase in intercollicular inhibition (Munoz and Istvan 1998; Takahashi et al. 2005). Our findings here were contradictory to this model. Deactivation of the cortex lining the cPS did not result in increases in activity in the ipsilateral SC, but instead revealed both a trend toward decreases in prestimulus activity and a delayed onset of saccade-related activity. In the SC contralateral to the deactivated cortex, deactivation resulted not only in activity decreases, but rather also increases in prestimulus activity, increases in the duration over which prestimulus activity was sustained, increases in stimulus-related activity, and a trend toward a reduced onset time of saccade-related activity. These changes in SC activity were matched by predictable changes in oculomotor behavior. Reaction times of contraversive saccades were increased, while those of ipsiversive saccades were decreased, and an increase in the rate of erroneous antisaccades was observed on trials in which the visual stimulus was presented to the hemifield ipsilateral to the cooled cPS.
In an earlier report describing the effects of unilateral DLPFC cooling on SC neurons (Koval et al. 2011), unilateral DLPFC deactivation was shown to lead to increases in prestimulus activity in the contralateral SC, and bilateral increases in stimulus-related activity. In contrast, we found here that prestimulus activity was reduced in the ipsilateral SC and increased in the contralateral SC, that changes in stimulus-related activity were confined to the SC contralateral to the deactivated cortex, and that DLPFC-induced changes in SC activity were linked directly to changes in oculomotor behavior; changes in the onset time of saccade-related activity were linked to shifts in reaction times, and enhanced prestimulus and stimulus-related activities in the contralateral SC were linked to erroneous responses on antisaccade trials. We showed also that cooling-induced enhancements in prestimulus activity in the contralateral SC were prolonged, similar to the effects observed on stimulus-related activity.
While the findings outlined above are not consistent with a model in which the DLPFC acts to inhibit saccade-related activity in the ipsilateral SC via inhibitory interneurons or rostral fixation neurons (Fig. 1A), they are broadly consistent with one in which the DLPFC provides direct excitatory inputs to SC saccade-related neurons (Fig. 1, right panel). According to an inhibition scheme, the removal of excitatory DFLPC outputs would result in a decrease in activity of SC fixation neurons and a decrease in SC inhibitory interneurons, resulting in a net increase in activity of SC saccade-related neurons in the ipsilateral SC. This would result in a decrease in activity in the contralalteral SC, due to increased intercollicular inhibition, and a net bias toward contraversive saccades due to the effects of the altered balance of SC outputs on downstream premotor burst neurons. The alternative excitatory scheme suggests that the DLPFC exerts an excitatory effect on saccade-related neurons in the ipsilateral SC. This model predicts that DLPFC deactivation would result in a reduction of direct excitatory drive to ipsilateral SC saccade neurons, causing a decrease in their activity. Such a decrease would reduce intercollicular inhibition and lead to increases in the activity of SC saccade neurons in the contralateral SC. The result of these changes would be a bias toward ipsiversive saccades, effected by changes in the balance of SC outputs to premotor burst neurons in the brainstem. Our data here were more consistent with this excitation model. Although we observed only a trend toward decreased activity of ipsilateral SC saccade-related neurons, the onset time of saccade-related activity was significantly delayed. We also found robust activity increases and reductions of the onset time of saccade-related activity in SC saccade-related neurons in the SC contralateral to the deactivated cPS.
Historically, support for the inhibition model has been provided by findings in patients, which have shown that DLPFC lesions involving the MFG result in significant increases in error rates in the antisaccade task (Ploner et al. 2005). Neurophysiological data have been more equivocal. Johnston and Everling (2006a) employed antidromic identification to investigate the response properties of DLPFC neurons sending a direct projection to the SC, while monkeys performed pro- and antisaccades. They observed enhanced activity on antisaccade trials, and based on the known excitatory nature of cortical outputs (Tsumoto 1990; Dori et al. 1992), proposed an excitatory effect of DLPFC outputs on SC fixation or inhibitory interneurons. This conceptualization is consistent with the inhibition model. They also observed, however, that activity of DLPFC neurons on antisaccade trials was negatively correlated with reaction times—increases in DLPFC activity were associated with reaction time decreases. Based on this, they proposed that DLPFC outputs must facilitate antisaccade task performance. This finding is at odds with the notion of an excitatory influence of DLPFC outputs on SC fixation neurons, since the activity of SC fixation neurons does not correlate with reaction time (Dorris et al. 1997), and microstimulation of these neurons delays saccade initiation (Munoz and Wurtz 1993). In addition, although this effect was more pronounced for ipsiversive antisaccades, it was also present for contraversive antisaccades, which suggests that the DLPFC could contribute a general excitatory signal responsible for preparation for antisaccades, and not necessarily act in an inhibitory capacity. It is worth noting that the increases in DLPFC activity observed by Johnston and Everling (2006a) resemble the increases in DLPFC BOLD activation observed on antisaccade trials in human imaging studies (Sweeney et al. 1996; Desouza et al. 2002; McDowell et al. 2002; Ford et al. 2005; Brown et al. 2007; Ettinger et al. 2008). This suggests that while an inhibitory role of such DLPFC activation is plausible, it is not, mechanistically speaking, obligatory. Similarly, the failure of patients with damage to the MFG to suppress prosaccades in the antisaccade may not represent a failure of an inhibitory mechanism per se, but rather an impairment of the more general cognitive function of establishing and maintaining appropriate task sets, though it should be noted that errors on prosaccade trials are not frequently reported in such patients (Ploner et al. 2005). Such generality is suggested by the robust activation of this region during numerous cognitive tasks (Duncan and Owen 2000), and it has been proposed that the activity of DLPFC neurons evolves to represent the demands of nearly any cognitive task (Duncan 2001).
The excitatory scheme we have proposed here is consistent with other studies using causal methods to manipulate DLPFC activity during performance of antisaccades. Our findings resemble those obtained by Condy et al. (2007) following muscimol injections in the lower bank of the cPS. They too observed increased rates of erroneous saccades to ipsilateral stimuli on antisaccade trials and increased reaction times for contralateral antisaccades. The primary difference between our results and those of Condy et al. is that we additionally found effects on prosaccade trials. This may be a result of the considerably larger spatial extent of the cortex deactivated by cryoloops compared with that of single muscimol injections. Condy et al. interpreted their results as being consistent with the DLPFC inhibition model by proposing that muscimol might have lead to a paradoxical increase in the activity of DLPFC output neurons. The fact that we have obtained such similar effects with unilateral cortical cooling of the cPS rules out this explanation. Moreover, the behavioral effects on pro- and antisaccades after unilateral sPS cooling are opposite to those effected by electrical microstimulation of the cPS (Wegener et al. 2008). The combined weight of evidence derived from the compatibility of our current findings with those of pharmacological inactivation and microstimulation studies (Condy et al. 2007; Wegener et al. 2008) suggests that the DLPFC does not inhibit the ipsilateral SC and provides strong evidence against the inhibition model.
Our data have potential for providing insights into the manner by which saccade generation is controlled by the SC. The role of SC fixation neurons has been debated for some time. According to one view, a push–pull interaction between fixation neurons in the rostral SC and saccade neurons in the caudal SC determines whether a saccade will be generated or fixation maintained (Munoz and Wurtz 1993). An alternative view is proposed by the equilibrium hypothesis, which posits that rostral SC neurons do not directly engage fixation, but that it naturally results from a balance of excitation between rostral regions of the left and right SCs (Goffart et al. 2012). Here, DLPFC deactivation induced an imbalance in SC activity, which leads to lateralized changes in reaction times and error rates of pro- and antisaccades. These findings are broadly consistent with the equilibrium hypothesis as they suggest that balanced excitation between the SCs determines whether appropriate saccades will be made.
The excitatory model suggests that the DLPFC could participate in either saccade preparation or attentional shifts toward the contralateral visual field. A role for the DLPFC in directing covert attention has been demonstrated by several single-neuron recording studies in monkeys (Everling et al. 2002; Lebedev et al. 2004; Kaping et al. 2012). Everling et al. (2002) recorded neural activity in the DLPFC, while monkeys viewed simultaneous streams of pictures presented the left and right of central fixation. The monkeys' task was to maintain fixation and to generate a saccade toward a target stimulus when it appeared in the stream of images at a previously cued side. Many DLPFC neurons exhibited selectively elevated responses to target stimuli presented at the attended side. The majority of neurons in that study preferred stimuli presented in the contralateral field. More recently, Kaping et al. (2012) have shown that neurons in and around the principal sulcus area respond to covert attentional shifts toward contralateral targets.
A well-known property of DLPFC neurons is that they exhibit persistent delay activity in spatial oculomotor delayed response tasks (Fuster and Alexander 1971; Funahashi et al. 1989). Such activity is thought to carry a retrospective representation of stimulus location in some neurons and a prospective signal for the forthcoming saccade in others (Funahashi et al. 1993; Takeda and Funahashi 2002). fMRI studies in humans have reported persistent activation in the MFG, the putative human homolog of the principal sulcus region in monkeys, during the delay period of oculomotor delayed response tasks (Leung et al. 2002; Curtis and D'Esposito 2003; Brown et al. 2004; Curtis et al. 2004; Yoon et al. 2006; Kastner et al. 2007). Ikkai and Curtis (2011) reported increased BOLD activation in the MFG when subjects are required to maintain covert attention to a stimulus during a delay period. Interestingly, this activation was contralaterally biased. They concluded that DLPFC, like posterior parietal cortex, contains a prioritized map of space that can be used to guide attention allocation, spatial memory, and motor planning. This interpretation is consistent with the notion of a dorsal frontoparietal network that underlies spatial attention, stimulus salience, and saccades (Corbetta and Shulman 2011). As in posterior parietal regions and the frontal and supplementary eye fields (Corbetta et al. 1998; Kastner et al. 1999; Hopfinger et al. 2000), the MFG shows enhanced activation when subjects respond to symbolic cues to voluntarily shift attention to a location (Hopfinger et al. 2000). Interestingly, the same regions are activated in both humans (Hon et al. 2006) and monkeys (Stoewer et al. 2010) when viewing a stream of visual stimuli in the absence of any task instruction, suggesting that updating of an attended visual representation may be sufficient to activate the frontoparietal network.
Taken together, our data are inconsistent with the predictions of the classic model of DLPFC-mediated inhibition of saccade-related activity in the SC. They suggest instead that the area in the cPS provides a direct or indirect excitatory drive to the ipsilateral SC, which when unilaterally attenuated decreases contraversive and increases ipsiversive saccade preparation. While these findings do not provide support for a circumscribed inhibitory role of the DLPFC, they are consistent with the notion that DLPFC outputs act to bias the activity of target structures to support goal-directed behavior, as proposed by prominent models of PFC function (Miller and Cohen 2001). An intriguing further test of the model would be to investigate the effects of DLPFC cooling on responses in the FEFs. FEF neurons display similar activity on antisaccade trials as SC neurons (Everling and Munoz 2000), and given that the majority of corticocortical projections are between pyramidal neurons, and thus excitatory (Bunce and Barbas 2011), it seems likely that the effects of DLPFC deactivation would be similar for its cortical and subcortical projection targets. Such a finding would potentially establish general principles of DLPFC-mediated modulation of target structures and illuminate its mechanistic role in implementing bias signals.
This work was supported by grants from the Canadian Institutes of Health Research to S.E. and S.G.L.
The authors thank B. Soper, S. Hughes, and A. McMillan for expert technical and surgical assistance. We are also grateful to J. Gati and R. Menon for assistance in MRI imaging. Conflict of Interest: None declared.