The anatomical connections of the pregeniculate complex (PrGC) with components of the visual-ocular motor system suggested its contribution to ocular motor behavior. Subsequent studies reported saccade-related activity in the primate PrGC. To determine its contribution, we characterized pregeniculate units (n = 128) in alert macaques during ocular motor tasks and visual stimulation. We found that 36/109 saccade-related units exhibited postsaccadic bursts or pauses in tonic discharge for saccades of any amplitude or direction. In contrast to previous results, 46/109 responses preceded or coincided with the saccade, while 47/109 responses were directionally tuned. Pregeniculate units were modulated not only in association with saccades (109/128) but also with smooth eye movements and visual motion (20/128) or eye position (23/128). Multiple ocular motor signals were recorded from 19% of the units, indicating signal convergence on individual neurons. Visual responses were demonstrated in 51% of PrGC units: visual field illumination modulated the resting discharge of 33 units; the responses of 37 saccade-related units and all 23 position-dependent units were modulated by visual stimulation. Early saccadic activity in the PrGC suggests that it contributes more to gaze than postsaccadic modulation of visual or ocular motor activity. The patterns of saccadic responses and the modulation of PrGC activity in association with a variety of visual-ocular motor behaviors suggest its potential role as a relay between the parietal cortex and elements of the brain stem ocular motor pathways, such as the superior colliculus and pretectal nucleus of the optic tract.
The primate pregeniculate complex (PrGC) is a retinorecipient component of the ventral thalamus that lies dorsal and medial to the lateral geniculate nucleus (LGNd) (Livingston and Mustari 2000; Niimi et al. 1963; Polyak 1957). The prominent anatomical connections of the PrGC with cortical and brain stem areas that are critical components of the ocular motor system suggest that it contributes significantly to visualocular motor behavior. Notably these anatomical connections include projections from the occipital and parietal visual cortices and reciprocal connections with the pretectal nucleus of the optic tract (NOT) and the superior colliculus (for review, see Harrington 1997; Livingston and Mustari 2000). The contribution of the PrGC to visual-ocular motor behavior is suggested also by studies of the homologous nucleus in nonprimates, the ventral lateral geniculate nucleus (LGNv) (Graybiel 1974; Harrington 1997; Jones 1985; Niimi et al. 1963). The LGNv is also retinorecipient and is extensively interconnected with critical visual and ocular motor structures; its proposed role as an integrative component of the visual and ocular motor systems is well supported (Agarwala et al. 1989; Conley and Friederich-Ecsy 1993; Crossland and Uchwat 1979; Graybiel 1974; Hada et al. 1985, 1986; Harrington 1997; Mathers and Mascetti 1975; Swanson et al. 1974).
In primates, pregeniculate involvement in ocular motor behavior is suggested further by the saccade-related (SR) discharge of PrGC neurons (Büttner and Fuchs 1973; Livingston 2000; Magnin and Fuchs 1977). Fuchs and colleagues reported that some PrGC neurons exhibit a postsaccadic burst in discharge, while other tonically active neurons exhibit a postsaccadic pause in activity (Büttner and Fuchs 1973).1 Similarly the discharge of many neurons in the LGNv of cats is modulated in association with saccades and the quick phases of vestibular or optokinetic nystagmus (Magnin and Putkonen 1978; Magnin et al. 1974; Putkonen et al. 1973). In cats, LGNv neurons are modulated also during oscillatory head motion in complete darkness. These results suggest the possibility that the LGNv plays a role in vestibular responses, and this is supported by studies demonstrating an anatomical association of the LGNv with the vestibular nuclei (Magnin and Kennedy 1979; Matsuo et al. 1994). This contrasts with the results of related experiments in primates, in which the responses of pregeniculate units were recorded during horizontal angular acceleration or during smooth pursuit of a visual target: no vestibular responses or alteration of pregeniculate discharge in association with the smooth eye movements were observed (Magnin and Fuchs 1977). However, these authors restricted their recordings to those pregeniculate units that exhibited the distinctive post-SR activity as originally reported. It is likely that use of this restrictive criterion for identification and testing of pregeniculate units resulted in an incomplete characterization of the PrGC.
In our studies to determine the contribution of the PrGC to saccadic behavior in primates, we found patterns of saccadic activity that are significantly different from those previously reported. We report here early (pre- to peri-) saccadic activity in many units as well as directionally selective SR responses similar to those reported for the cat LGNv (Magnin et al. 1974). We present evidence that neurons in the primate PrGC are modulated in association with both saccades and smooth eye movements. Eye position-dependent modulation of PrGC activity is also demonstrated. These results are more consistent with those obtained in other species (Hada et al. 1986; Harrington 1997; Hayashi and Nagata 1981; Hughes and Ater 1977; Hughes and Chi 1983; Magnin and Putkonen 1978; Mathers and Mascetti 1975; Pateromichelakis 1979; Spear et al. 1977; Sumitome et al. 1979). We confirm the presence of visual responses in PrGC units and show that the discharge of PrGC units may be modulated in response to movement of the visual field. Our results indicate that the PrGC exhibits a variety of visual and ocular motor responses and that earlier studies in which recordings were restricted to neurons exhibiting classical postsaccadic responses may have been biased due to selective recording of only a subset of pregeniculate neurons. Our characterization of pregeniculate activity raises several issues. First, the presence of early and late saccadic activity in the PrGC suggests that it contributes to additional aspects of saccadic or gaze behavior than were proposed based on reports of only postsaccadic activity. Second, the multiple visual and ocular motor signals shown to converge in the PrGC are suggestive of activity observed in the parietal cortical areas that are known to project to the PrGC. Our current results in conjunction with previous demonstrations of reciprocal anatomical connections of the PrGC with brain stem structures such as the pretectal nucleus of the optic tract and superior colliculus, suggest that the PrGC acts as a relay between the cortex and brain stem pathways. Visual modulation of pregeniculate activity suggests that one function of the PrGC is the integration of visual input with ocular motor signals that are relayed through this ventral thalamic structure.
Experiments were performed in alert, behaviorally trained nonhuman primates following procedures approved by the Institutional Animal Care and Use Committee of the University of Texas Medical Branch, Galveston, TX, and conforming to the National Institutes of Health “Guide for the Care and Use of Laboratory Animals.” All neuronal data were obtained from animals with chronically implanted recording chambers allowing access to the CNS, head-stabilization lugs, and scleral eye coils for simultaneous recording of eye position. Neurons (n = 128) were recorded in three rhesus (Macaca mulatta, 86 units) and one cynomolgous (M. fascicularis, 42 units) macaque. No differences or trends were found in the discharge properties of units, or in the percentage of unit types recorded in these two species; therefore the unit data from all four animals is reported together.
All surgical procedures were performed under general anesthesia (5:1 ketamine, 4 mg/kg: xylazine, 0.8 mg/kg) using appropriate sterile techniques. Postsurgical analgesics (aspirin, Stadol or Buprenex) were provided. Cylindrical stainless steel recording chambers, ∼19 or ∼24.5 mm in diameter, were implanted on the bone over a hole drilled in the skull. The chambers were centered above the PrGC using stereotaxic coordinates, 4.7–5.1 mm anterior, and 5.0–10.0 mm lateral to ear bar zero. The bottom surfaces of the chambers were beveled, allowing for electrode tracking at an angle of 0, 20, or 25° from the sagittal plane. The chambers were attached to the skull with dental acrylic, small stainless steel screws, and titanium struts/screws (TiMesh, Medtronic, Memphis, TN). The interiors of the recording chambers were sealed with medical-grade silastic (Dow Corning, Midland, MI) and closed with removable stainless steel caps bolted in place between recording sessions. The stainless steel lug for placement of the head stabilization post (Crist Instrument, Hagerstown, MD) was attached to the skull with dental acrylic. Eye coils (insulated stainless steel, Cooner Wire, Chatsworth, CA) for accurate measurement of eye position were implanted underneath the conjunctiva and stitched to the sclera using techniques similar to those described previously (Fuchs and Robinson 1966; Judge et al. 1980).
Eye-position monitoring and behavioral training
During recording, animals were seated comfortably in a standard primate chair with their heads immobilized. The chair was located within a sound-attenuated, light-proof booth. Eye position was measured using a scleral search coil (Fuchs and Robinson 1966) and a 3-ft coil frame and detector system (CNC-Engineering, Seattle, WA). The eye-position signal using this system is linear within 20° around the primary direction of gaze and accurate to 15 min of arc. Visual stimuli consisting of a discrete (diameter, ∼0.25°) red laser spot or a large (∼70 × 60°) patterned (random dot) image were rear-projected onto a ground glass tangent screen located ∼57 cm in front of the animal. Stimuli were projected onto this tangent screen from an optic bench equipped with two light paths for presentation and control of the laser spot and large field, patterned stimulus. Both light paths were under either manual or computer control. The visual stimuli were moved independently via x, y mirror galvanometers (General Scanning, Watertown, MA) located along each light path. Each path was equipped with a fast rise time (5 ms) shutter for discrete presentation of the stimuli.
Animals were trained to target fixation or pursuit by electronically comparing laser target and eye positions; when the animals fixated the target, they received an ∼0.2-ml applesauce reward. During training, the time required for continuous fixation of the target was gradually increased, and the allowable fixation error for reward receipt was reduced. Eye-movement signals were calibrated accurately by rewarding the monkey during fixation of the target stepped to known locations in the animal's visual field. In addition to visual tracking of discrete target steps via saccadic movements, the animals were trained to track smoothly moving targets (typically, sinusoidal oscillation along 4 visual axes: horizontal, vertical, and 2 intermediate).
Single units were recorded using glass-insulated, tungsten electrodes. The electrodes were advanced through stainless steel cannulae using an hydraulic microdrive (Trent-Wells, Coulterville, CA). The PrGC was identified by characteristic saccadic or smooth eye-movement-related activity, recorded 150-1350 μm above the dorsal surface of the LGNd. The activity of single PrGC units was isolated with a time-amplitude window discriminator (BAK Electronics, German-town, MD) and recorded concurrently with the eye position, laser target, and large field stimulus positions. Isolated single-unit spikes were converted to TTL pulses and sampled at 1 kHz. Analog signals were digitized to computer [Macintosh Quadra 650 with MIO-16X board (National Instruments, Austin, TX)] through BNC 2080 and custom-built interfaces at 1 kHz. Software for experimental control, data acquisition, and analysis were provided by Dr. Albert Fuchs at the University of Washington Regional Primate Research Center, Seattle, WA. Data were digitized concurrently using a 1401-plus CED interface and dedicated software (Cambridge Electronic Design, CED, Cambridge, UK) onto a PC (Optiplex GX1p, Pentium II, Dell Computer, Austin, TX) at a sampling rate of 1 kHz for all signals except the unit recording that was digitized at 60 kHz. Software protocols for experimental control, data acquisition, and analysis using the CED hardware were written by Dr. Andrew Burrows.
Ocular motor testing
Pregeniculate units were recorded during saccades, smooth eye movements, and fixation of the visual target during motion of the large field stimulus. Saccadic testing was performed under three conditions. 1) Single units were recorded during saccades evoked by steps (randomly variable amplitude and direction) of the laser target across the tangent screen while the animals were seated in the light-proof recording booth in the dark. 2) To determine the interaction of visual stimulation and SR responses, this test was repeated for saccades across the illuminated large field stimulus. 3) To verify that SR unit activity was not due to visual input, spontaneous saccades performed in complete darkness were also recorded. The activity of PrGC units during smooth eye movements was recorded similarly: to determine if PrGC units responded during smooth pursuit and to test for directional sensitivity, unit activity was recorded during pursuit of a target moving sinusoidally (±10°, 0.1–2.0 Hz, along each of 4 visual axes) in the dark or across the illuminated patterned visual field. The responses of some units were recorded during synchronized sinusoidal oscillation of the laser target and the illuminated large field stimulus. To test for sensitivity to visual motion, animals fixated the stationary target during constant velocity oscillation of the large field visual stimulus along each of the four visual axes.
To determine the response of PrGC units to visual stimulation, the large field stimulus was sometimes flashed off and on to determine its effect on unit discharge. In some instances, the large field stimulus was illuminated during saccadic tracking of the laser target to determine how visual input affected the SR discharge of PrGC units.
Data analysis was performed off-line. Saccade-related neurons were identified by averaging the unit discharge associated with 25–200 saccadic events. The discharge of units during saccades were averaged by aligning each response to saccade onset, defined typically by a 75°/s increase in eye velocity. Cumulative average firing rate histograms were generated to determine if unit discharge was modulated in association with saccades. Responses (pauses or bursts) were classified as pre-SR (onset preceded saccade initiation), peri-SR (coincided with saccade), or post-SR (onset after saccade completion). The average duration and latency of responses were calculated for those units with discharge rates that were regular enough to allow reliable measurements of latency and duration. For units exhibiting a post-SR burst, the peak and average burst frequency were measured. To determine if SR responses were directionally tuned, saccadic data for each unit was sorted into eight groups: saccades made to the right, left, up, down, up/right, up/left, down/left, and down/right. Average firing rate histograms for saccades in each group were generated and compared. To determine if the SR responses of PrGC units were modulated by visual input, the firing rate histograms and average response durations for saccades made under each of the three viewing conditions were compared.
Pregeniculate unit discharge during sinusoidal tracking was averaged over 8–30 cycles. Average eye-position and associated firing-rate histograms were generated after removing intermittent saccades. For neurons that also exhibited SR changes in discharge, saccadic unit responses were removed prior to averaging the unit data. Average firing-rate histograms for pursuit behavior in the dark or across the patterned background were compared. For units that responded more vigorously to pursuit across the patterned background, their response to visual motion was tested explicitly by requiring the animals to fixate the target while the large patterned stimulus was oscillated at a constant velocity. Firing rate histograms were averaged over 10–30 cycles of the patterned visual stimulus. The average firing rates for pursuit (or visual field motion) in each of eight directions were determined and fit by a Gaussian function to characterize the directional tuning and the depth of direction-dependent modulation for each unit (Mustari and Fuchs 1990; Wallman and Velez 1985). Subsequently the preferred directions (most vigorous discharge) of those units in which directional tuning was fully characterized, were normalized for direction toward (ipsiversive) or away (contraversive) from the side of recording, and plotted together.
To evaluate eye-position-dependent modulation of PrGC unit activity, the average firing rate of units during 50–150 inter-saccadic intervals were measured. Inter-saccadic intervals were defined as being outside of the time course for the SR response of each unit determined by latency and duration measurements. During these intersaccadic intervals, fixation on a visual target was maintained. For each interval, the firing rate was plotted as a function of the horizontal or vertical component of eye position. Data points were fit using linear regression analysis. If the correlation coefficient was >0.50, the unit was classified as carrying an eye-position-dependent signal. For some units, this analysis was performed for target fixation in the dark and with the patterned background illuminated.
Visual modulation of PrGC unit activity was evaluated in two ways. For neurons in which the overall firing rate was modulated by illumination of the large patterned, visual field stimulus, unit discharge was recorded during alternate periods of illumination and darkness. However, in other PrGC neurons, illumination of the large field visual stimulus did not affect overall firing rate, but did modulate SR responses. In these units the SR response amplitude indicated by average duration or peak burst frequency was determined for saccades performed under each of the three viewing conditions.
To determine if one or both eyes mediated visual modulation of PrGC activity, the visual modulation paradigms were repeated for 22 units during binocular, left-eye (LEV), and right-eye (REV) viewing. The response amplitudes were compared under each of these three conditions, and units were classified as binocular or left- or right-eye dominant.
To evaluate the average firing rates of PrGC units, we measured the inter-spike intervals during inter-saccadic periods, while the animals viewed the laser target in the dark. These data were used to calculate the average firing rates for 51 units and to demonstrate the regularity/irregularity of PrGC unit discharge. Average inter-spike intervals for each unit were calculated to demonstrate the range of firing rates, and a coefficient of variation (CV, SD/mean inter-spike interval) for each unit was calculated (Bialek and Rieke 1992; Goldberg and Fernandez 1971; Goldberg et al. 1984).
To verify that recorded units were located within the PrGC, we noted the location of each relative to the dorsal surface of the LGNd. The dorsal surface of the LGNd was identified by recording neurons with visual responses dominated from the contralateral eye (LGNd layer 6). Visual responses were recorded during alternating monocular eye patching as the animals fixated the laser target and the large patterned background stimulus was illuminated. Frequently electrode tracks continued into deeper layers of the LGNd to verify alternation of ocular dominance. Subsequent histological recovery of the electrode tracks provided a clear indication for each animal of the range of distances from the LGNd surface, in which PrGC units were recorded. All units reported were located 150-1350 μm above the dorsal surface of the LGNd. This range was determined based on tracking angle and the medial-lateral location of electrode tracks (see Fig. 1). Requirement that the location of each unit relative to the surface of the LGNd be verified was stringent and automatically excluded some unit data. However, verification of each unit's location within the PrGC was imperative.
The PrGC is a thin multi-laminar plate of cells that curves over the dorsal and medial aspect of the dorsal LGNd (Fig. 1, A and B). It is separated from the LGNd by the geniculate capsule. The locations of electrode tracks through the PrGC are shown in Fig. 1. The trajectory of tracks at an angle of 25° from the sagittal plane is shown in C, while vertical tracks from another animal are illustrated in D. We recorded the activity of 128 pregeniculate neurons, along tracks comparable to these.
Confirming previous results that the PrGC contributes to saccadic behavior, we found that 109 of the 128 (85%) pregeniculate neurons were saccade-related. Thirty-six of the SR units (33%) exhibited postsaccadic modulation similar to that reported previously (Büttner and Fuchs 1973; Livingston 2000): 22 were classified as postsaccadic pausers, whereas 14 were postsaccadic bursters (Fig. 2, A and B). Pregeniculate units that exhibit this classical pattern of activity, respond vigorously for saccades of any amplitude or direction. The responses occur late, after completion of the saccadic movement, are prolonged, and are variable in duration. Typical responses are illustrated, but it should be noted that for some pausers, the firing rate might require hundreds of milliseconds to return to its presaccadic level. For this class of saccadic units, the average pause duration relative to saccade initiation was 186 ± 115 (SD) ms (range: 67–542 ms) for saccades to target in dark, with an average latency of 79 ± 20 ms (range: 46–122 ms). Similarly, the post-SR discharge of some bursters was prolonged with the firing rate diminishing gradually over hundreds of milliseconds (burst durations: 83–329 ms). The average duration for classical SR bursts associated with saccades to target in the dark was 181 ± 104 ms with burst latencies averaging 108 ± 11 ms (range: 93–124 ms).
Although 36 SR units in the PrGC exhibited classical postsaccadic responses, we found that in contrast to previous reports, 46/109 (42%) SR units exhibited pre- (n = 14) or perisaccadic (n = 32) responses. These early SR responses were seen in both pausers and bursters (Fig. 3). For the unit shown in Fig. 3A, the irregularity of the firing rate made it difficult to define the latency of the pause. However, it is clear that the pause typically begins prior to saccade initiation. The SR-burst illustrated in Fig. 3B is initiated before or during the saccade, reaching its peak frequency during the saccade (average latency: –331.3 ms, average duration: 422.2 ms). In addition to early activity, these two units illustrate that the SR-response in many pregeniculate units could not be classified as a pure pause or burst. The pause illustrated in Fig. 3A is followed by a distinctive increase in the firing rate that exceeds presaccadic levels. The average latency of this burst is 105.9 ms with a duration of 169.0 ms. Similarly, the perisaccadic burst shown in Fig. 3B is followed by a distinct pause of variable duration (average latency, 90.9 ms). Various patterns of phasic responses such as these were observed. Many units classified as having postsaccadic bursts due to the prominence of this response also exhibited brief periods of inhibition that preceded the burst and occurred during the saccade. These results demonstrate that significant changes in pregeniculate SR activity are initiated earlier than previously reported, again indicating that the PrGC is associated with more than postsaccadic visual or motor processing.
In addition to early SR activity, we found that 47/109 PrGC units tested, exhibited directionally tuned SR responses. A typical directionally tuned SR response is illustrated (Fig. 4). The SR pause in this unit was most prominent for down/left saccades. There was no response for up/right saccades. The tuning of this response is evident if the pause duration is plotted as a function of saccade direction (Fig. 4, C and D). Although many of the PrGC units with directionally tuned saccadic responses exhibited no response for the opposite direction, some exhibited inverted responses for saccades made in the opposite direction. For example, a unit that burst for saccades made towards the right, paused for saccades made towards the left (e.g., Fig. 12A). Of the 47 SR units that were directionally tuned, 25 responded most strongly for saccades made towards the side of recording (ipsiversive), whereas 22 responded for saccades in the contraversive direction. An additional nine units had responses tuned for two directions, with the responses inverted for saccades made in opposite directions. Unlike the responses seen in collicular neurons, we found no correlation of the responses with saccade amplitude in pregeniculate neurons. Rather the pattern of directionally tuned saccadic activity seen in the PrGC is more comparable to that reported for some parietal cortical neurons, many of which are only broadly tuned for saccade amplitude and direction (Barash et al. 1991a,b).
The timing of saccadic activity in the various types of SR units is summarized in Table 1. The timing relationship of activity associated with saccades is illustrated for 49 PrGC neurons, including all types of SR units (Fig. 5). The average response latencies and durations are plotted paired with their respective average saccade duration. Although most activity is postsaccadic, it is evident that the discharge of many PrGC units is modulated prior to or during saccade initiation. Also illustrated is the wide range of response durations that were observed in PrGC units. On the basis of previous work and due to the postsaccadic modulation of activity in many PrGC units, we did not expect to find tight correlations between the pregeniculate responses (latencies, durations) and saccade metrics (initiation, duration). We evaluated the pregeniculate response characteristics and found no correlation with saccade metrics (Fig. 6). No correlation was found between the response duration or peak firing frequency (for bursters) and saccade amplitude, duration, or peak velocity. The relationship of response latency and duration is shown as a function of saccade onset and duration in Fig. 6. All data are referenced to saccade initiation. While the saccade duration for each unit tested (n = 49) remains relatively constant within a given range, the response latency (or onset) and end time vary widely. The lack of an association between the timing characteristics of the response and saccade suggests that the PrGC is associated with an aspect of visual processing or ocular motor behavior other than saccadic control.
Smooth eye movements and visual motion
Although sensitivity to head motion, smooth eye movements, or visual motion have never been demonstrated in the primate PrGC, responses to these stimuli are documented for the LGNv in a variety of nonprimates. In testing, we found that 20 of 128 (∼16%) pregeniculate units were modulated during smooth eye movements or in response to visual motion of the large field visual stimulus. Of the 20 units, 18 PrGC units responded during smooth eye movements; 9 units responded for visual motion during fixation. The neuron shown in Fig. 7A responded for smooth pursuit of the target towards the down/left but exhibited its most vigorous response for pursuit of the laser target towards the down/left across the patterned background stimulus (Fig. 7B). This unit responded in a directionally dependent way for visual motion towards the up/right (Fig. 7C). Under this condition, the increase in the unit's firing rate correlated with the visual slip of the large field stimulus. However, correlation of the unit response during smooth pursuit in the dark does not appear to correlate well with visual slip of the laser target.
Complete directional tuning curves were generated for 11 pregeniculate units that responded during smooth eye movements. Six additional units were tested for direction-selective responses. Most (13/17) were tuned for pursuit toward the side of recording (ipsiversive). The directional tuning curve for a unit recorded in the left PrGC that responded most vigorously for pursuit toward the down/left is shown (Fig. 8). The width of the curve about the preferred directional axis (highest predicted firing rate) demonstrates the broad tuning typical of pregeniculate units. The preferred direction of response for 11 PrGC units were normalized for the side of recording and plotted together (Fig. 8B). The predominance of units tuned for the ipsiversive direction (9/11 units) is evident. Also evident in the plot is the variable amplitude of the directionally tuned responses observed in pregeniculate neurons.
We recorded the pregeniculate responses during smooth eye movements at velocities of 4–80°/s. Although we did not rigorously analyze the velocity tuning of pregeniculate units, for those in which multiple velocities were tested, the most vigorous responses were recorded for pursuit at 14–30°/s.
Eye position-dependent signals
In 23 PrGC units, eye-position-dependent modulation of the firing rate was evident. A unit that responded with a clear increase in firing rate for fixation towards the left is illustrated (Fig. 9). Steps toward the right result in intermittent drops in the firing rate. Data taken from the target stepping task shown for this position-dependent unit are plotted in Fig. 9B. While a valid correlation cannot be generated from two clusters of fixation points, Fig. 9, C and D, illustrates valid correlations for two other position-dependent units. One unit is biased for rightward and one for leftward gaze. (Both units were recorded in the right PrGC.) The data for each of the three units shown were taken for conditions in which the animals fixated in the dark. For 14/23 eye position-dependent PrGC units, firing rate was modulated only in the dark. In contrast, the firing rate of 9/23 pregeniculate units was correlated with eye position only during illumination of the patterned background. The dependence of the eye-position signals on the viewing conditions illustrates the integration of visual and ocular motor signals by the PrGC.
Visual modulation of firing rate
Direct retinal input suggests that a population of PrGC units should exhibit visual responses. This in conjunction with previously documented visual responses in primates and nonprimates prompted visual testing in our experiments. In 33 pregeniculate units, the firing rate was modulated by illumination of the patterned background stimulus. This phenomenon is illustrated by two neurons that were recorded along the same electrode track, 50 μm away from one another (Fig. 10). The overall firing rate of one unit is inhibited dramatically by illumination of the patterned background (Fig. 10A). This inhibition is tonic, continuing for as long as the visual field is illuminated. It is independent of eye position or accuracy of fixation. Conversely, the firing rate of the second unit illustrated is driven by illumination of the visual field (Fig. 10B). This visual effect is independent of eye position, visual task, or target fixation. These units were recorded just dorsal to the geniculate capsule and were likely to be located within the most ventral (retinorecipient) sublayer of the PrGC. However, the difficulty of identifying the boundary between the dorsal and ventral sublayers precluded localization of visually responsive PrGC units within an identified subnucleus of the PrGC. Of 33 PrGC units in which tonic visual modulation was observed, the firing rates of 16 were decreased, whereas the firing rates of 17 were increased by illumination of the visual field. This suggests that at least some retinal input acts via an inhibitory system, possibly the GABAergic neuronal population within the retinorecipient sublayer.
In addition to the tonic visual modulation described in the preceding text, in 37 PrGC units, visual modulation of activity was apparent not simply as a change in general firing rate. Rather, the SR responses of these units were modulated by illumination of the visual field. In 25/37 units, responses were less vigorous for saccades across the illuminated patterned background, as compared with saccades made in the dark. This is effectively illustrated by comparing the durations of SR pauses or bursts under different viewing conditions. The average pause duration was 273 ms (SD 217) for spontaneous saccades in complete darkness, 169 ms (SD 116) for saccades to target in the dark, and 151 ms (SD 164) for saccades across the illuminated patterned background. The average duration for bursts associated with spontaneous saccades in complete darkness was 269 ± 172 ms, 174 ± 155 ms for saccades to target in the dark, and 88 ± 91 ms for saccades across the illuminated background. Such visual modulation of SR responses is similar to that reported by Büttner and Fuchs (1973), who reported a reduction in the amplitude of SR responses subsequent to brief flashes of diffuse light. To verify that the duration of SR responses in complete darkness was not associated with the increased saccade duration characteristic of saccades performed in complete darkness, we compared the response durations for large saccades of equivalent durations made with visual input to those made in complete darkness. We found that the prolonged response duration for saccades in complete darkness was not a function of the increased saccade duration (not illustrated). This finding is consistent with the observation that the timing of SR responses was not correlated with any aspect of saccade metrics.
To determine if visual modulation of PrGC unit activity is mediated binocularly, ocular dominance tests were performed for 22 units. In almost all units tested, the visual modulation of discharge is mediated by input from both eyes. This is clearly demonstrated by changes in the duration of SR responses in a unit during LEV, REV, or binocular viewing (Fig. 11). This unit exhibited a vigorous post-SR burst for saccades made to target in the dark (response duration, 150 ms). For saccades made across the illuminated patterned background, the average burst duration is reduced to 36 ms. The burst duration associated with saccades across the illuminated visual field while LEV is 81 ms, whereas REV is 79 ms. Reduction of the response duration due to illumination of the visual field is mediated equally by visual input from both eyes. This was typical of the results obtained for most units tested (18/22). Four units exhibited binocular responses that were dominated by the contralateral eye.
Convergence of visual and ocular motor signals
Clearly multiple types of visual and ocular motor signals converge within the PrGC. While most of the units illustrated carry a single prominent visual or ocular motor signal, 24/128 (19%) pregeniculate units carry multiple signals. The unit shown in Fig. 12 is an example in which saccadic, visual slip and eye-position-dependent signals converge. A SR burst for rightward saccades and a longer latency pause for leftward saccades is shown. This unit shows no response for smooth pursuit of a small target across a dark background; however, it does respond to visual slip of the large field stimulus toward the left (Fig. 12B). This visual response is apparent during rightward pursuit across the illuminated patterned background, during which the unit responds vigorously. In addition, the firing rate of this unit is increased for rightward gaze (Fig. 12C). This position-dependent signal can be seen superimposed on the response of the unit during pursuit. As shown in the inset (bottom right, Fig. 12B), the response associated with the leftward slip of the patterned background extends partly into the rightward slip phase of the cycle; the response continues while gaze is directed toward the right. This unit was recorded from the right PrGC. The combined effect of the signals carried by this unit could collectively contribute to output appropriate for driving the eye during rightward pursuit, interrupted by catch-up saccades.
Irregularity of firing rate
A prominent feature of PrGC activity was the wide range of firing rates observed in different neurons. We analyzed the discharge of 51 PrGC units during 5–15 s of intersaccadic intervals, during which the animals performed stable fixation of the stationary target spot in the dark. (Units that carried eye-position-dependent signals were excluded from this analysis.) Under these viewing conditions, the average firing rate was 30.44 ± 19.0 Hz; the range of resting discharge rates for pregeniculate units was 4.09–78.18 Hz. This value contrasts with peak discharge rates ranging from 78.00 to 443.15 Hz observed during SR bursts (average, 128.36 Hz for saccades to target in the dark). We also noted a striking irregularity in the discharge rate of individual pregeniculate neurons. A coefficient of variation (CV) for the interspike intervals was calculated for each unit. The average CV during intersaccadic intervals was 0.97 (range, 0.35–1.99), demonstrating the extremely irregular firing rates of pregeniculate neurons. This irregularity may reflect the convergence of visual and motor signals within the PrGC, and the relay of multiple signals by individual PrGC neurons.
The distribution of responses that we found among PrGC neurons is summarized in Fig. 13. It must be noted that the results presented here may not represent the entire population of PrGC units. We restricted our analyses to those units whose location relative to the surface of the LGNd could be demonstrated to verify their location within the PrGC. A significant number of pregeniculate units are located medial to the LGNd and so were excluded from this initial study. Further, the difficulty of recording each neuron long enough to test for responses during multiple eye movements and visual stimuli precluded full testing of every unit. Therefore it is likely that the percentages of neurons with sensitivity to smooth eye movements, visual sensitivity, or convergent signals for example are under-reported. We also noted that numerous units encountered during tracking in the PrGC did not respond during the ocular motor tasks or visual stimulation described here; these units are not included in this report. Similarly, Büttner and Fuchs also identified a subset of pregeniculate neurons that did not respond during saccadic testing. Therefore the response and discharge characteristics of a subset of PrGC neurons are still unknown.
The discharge patterns of most PrGC units are different than previously reported. We found that almost 39% of pregeniculate units are associated with nonsaccadic visual-ocular motor behaviors, including smooth eye movements (14%), visual slip (7%), and eye position (18%). Although the majority (85%) of PrGC units described here are saccade-related, 67% of these units exhibit timing and discharge characteristics that are significantly different than expected (Fig. 13, A and B). Notably, 42% of SR responses are not postsaccadic but occur before or during the saccade. Unlike the original description of saccadic responses in the PrGC, 43% of the responses are directionally tuned. Our findings dictate that previous hypotheses concerning the functional role of the PrGC be modified, and suggest that future experiments evaluate the PrGC in the context of gaze shifts and other aspects of ocular motor behavior. We suggest that one role of the PrGC is the relay of signals from the cortex to the brain stem.
Its potential role as an integrative relay between visual cortical areas and brain stem motor pathways is suggested by the prominence of cortical projections to the PrGC and its reciprocal connections with ocular motor sites in the brain stem (Fig. 14). Our results demonstrate similarities in the discharge of PrGC units to those in parietal cortex. We also demonstrate convergence of multiple ocular motor signals within the PrGC: 19% of the units carried multiple ocular motor signals. For example, of the SR units that we tested, 9% also exhibited smooth eye movement or visual motion responses and 13% eye position-dependent signals (Fig. 13C). A small percentage of SR units (3%) exhibited both smooth eye-movement and eye position-dependent signals. The signals recorded in the PrGC are suggestive of activity recorded in those parietal areas that project to the PrGC. Visual modulation of pregeniculate activity indicates that the PrGC integrates these ocular motor signals with visual input. Although limited study of the PrGC to date precludes detailed speculation regarding its functional role, the evidence that is currently available directs that future experiments address its role as an integrative visual-ocular motor relay between the cortex and brain stem. Evidence to support this proposal is discussed further.
Functional role of the pregeniculate complex
Our results confirm and extend previous observations regarding neuronal activity in the PrGC. The PrGC is associated primarily with saccadic behavior, and the modulation of many units is postsaccadic. In the absence of any demonstrable correlation between the responses of pregeniculate neurons and saccade metrics (Fig. 6), it is unlikely that the PrGC contributes to saccadic control. Based on these observations, the PrGC might be proposed to play a role in postsaccadic processing of visual or ocular motor signals (Büttner and Fuchs 1973; Livingston 2000). However, the PrGC does not project to the LGNd, a primary site for saccadic modulation of visual input (Ramcharan et al. 2001; Ross et al. 1996; Ross et al. 2001; Wilson et al. 1995). Our results force reconsideration of the pregeniculate's role in visual-ocular motor behavior. Modulation of activity in 42% of SR pregeniculate units before and during saccadic movement suggests that it contributes to more than postsaccadic gating of visual or motor signals.
The range of initiation times and the duration of saccadic unit activity in the PrGC as illustrated in Fig. 5 do not support a role for the PrGC in the control of saccades. Recent studies of other brain stem areas in which the timing of saccadic activity is broadly distributed suggest that late saccadic activity may be associated with aspects of gaze other than saccadic control. Notably, the discharge of many units in the central mesencephalic reticular formation is postsaccadic and corresponds with the timing of postsaccadic discharge observed in PrGC units. The discharge of these mesencephalic units is associated with head movements (Pathmanathan et al. 2002; Waitzman et al. 2002). Because our recordings were performed in a head-fixed preparation, we cannot address this issue at present. Further behavioral and recording experiments should address the relationship of pregeniculate activity to the later phases of gaze shifts.
Currently there is only weak evidence in primates suggesting that the PrGC may be associated anatomically with the mesencephalic reticular formation (Wilson et al. 1995). However, in a study of the thalamic connections of the LGNv in cats, a reciprocal connection of the LGNv with the MRF was demonstrated (Nakamura and Kawamura 1988). Given the similarities in the timing of activity in the PrGC and the mesencephalic reticular formation, the potential anatomical association between these areas necessitates further study.
Our results extend earlier descriptions of the pregeniculate by demonstrating modulation of PrGC unit activity in association with additional types of eye movements. The presence of vestibular, smooth eye-movement, and eye-position signals in the homologous LGNv of nonprimates prompted studies to determine if similar activity might be present in the primate PrGC. In contrast to results in chick, rat, rabbit, and cat (Hada et al. 1986; Harrington 1997; Hayashi and Nagata 1981; Hughes and Ater 1977; Hughes and Chi 1983; Magnin and Putkonen 1978; Magnin et al. 1974; Mathers and Mascetti 1975; Pateromichelakis 1979; Spear et al. 1977; Sumitome et al. 1979), neither smooth eye-movement nor vestibular signals were identified in the primate pregeniculate complex (Magnin and Fuchs 1977). This discrepancy may have been due to the identification of primate pregeniculate units based in part on the presence of classical post SR bursts. Our results demonstrate that a minority (14/128; 11%) of PrGC units exhibits classical post SR bursts, and only a small percentage of these might be expected to carry multiple ocular motor signals. It is likely that many pregeniculate neurons were not identified as such in the earlier study, and were not tested for smooth eye movement or vestibular signals. Therefore the possibility of vestibular signals in the PrGC cannot be excluded. Given the similarities between the macaque PrGC and the nonprimate LGNv, it is fair to speculate that vestibular signals will be identified in the PrGC.
Relationship of the pregeniculate complex to visual cortical areas and the brain stem
Our results and previous observations indicate that the PrGC may relay information from visual cortical areas to the brain stem motor pathways (Fig. 14): the PrGC receives afferents from several areas of the parietal cortex, including areas 7a, LIP (lateral intraparietal), and MT/MST (middle temporal/medial superior temporal) and the polysensory areas of STS (superior temporal sulcus) (Graham et al. 1979; Leichnetz 1990; Maioli et al. 1984; Maunsell and van Essen 1983b; Norden et al. 1978; Spatz and Tigges 1973; Ungerleider et al. 1984; Updyke 1981). The PrGC is the source of prominent efferent projections to the pretectal nucleus of the optic tract and superior colliculus (Büttner-Ennever et al. 1996; Livingston and Mustari 2000; Mustari et al. 1994).
Although the saccadic signals recorded in the PrGC could be derived from the superior colliculus, pretectum, or cortex, the activity of many saccadic units in the PrGC are comparable to those in the parietal cortical areas that project to it (Andersen et al. 1990; Andersen and Gnadt 1989; Barash et al. 1991a,b; Brotchie et al. 1995; Colby et al. 1993; Kawano et al. 1994). Pregeniculate saccadic activity is not comparable to that of collicular units as it is postsaccadic and is not correlated with saccade metrics. The SR discharge of only the postsaccadic pausers in the PrGC is similar to that of the pretectal omnipause neurons (Mustari et al. 1994, 1997). However, there are several similarities in the discharge of parietal and PrGC units. Most notably, responses in many parietal and pregeniculate units are postsaccadic and prolonged. The responses of many SR units in the PrGC are either not tuned for direction or amplitude or are very broadly tuned; similarly the saccadic responses of many parietal units are only broadly tuned (Barash et al. 1991a,b). Further, just as the saccadic responses in PrGC units do not correlate with saccade metrics, the SR responses of many parietal units are correlated with other aspects of saccadic behavior (Anderson and Buneo 2002; Cohen and Andersen 2002; Colby and Goldberg 1999; Goldberg et al. 2002; Gottlieb et al. 1998; Xing and Andersen 2000). These similarities suggest that saccadic activity in the PrGC is derived from parietal cortical input.
The presence of visual motion, smooth eye-movement and eye-position signals further suggest that ocular motor signals in the PrGC are derived from cortical input. Cortical projections to the ipsilateral PrGC originate from those parietal areas that play a role in visual motion detection and saccadic behavior and provide signals for smooth eye movements and saccades (Andersen and Gnadt 1989; Brotchie et al. 1995; Colby et al. 1995; Kawano et al. 1992; Kawano et al. 1994; Komatsu and Wurtz 1988a,b; Newsome and Pare 1988; Newsome et al. 1988; Pierrot-Deseilligny et al. 1995). These parietal areas also process eye- and head-position signals. Eye position modulates the discharge of cortical units in the lateral intraparietal sulcus and area 7a (Andersen et al. 1990).
In summary, activity in parietal areas MT, MST, and STS during visual motion and smooth eye movements is comparable to that of PrGC units. Eye-position signals in the LIP and parietal area 7a may also contribute to patterns of unit discharge in the PrGC. Each of these parietal areas is a potential source for the ocular motor signals recorded in the PrGC, supporting our hypothesis that cortical input contributes significantly to pregeniculate unit behavior.
Prominent reciprocal connections of the PrGC with the pretectal nuclei and the superior colliculus provide an anatomical basis by which signals can be relayed from the PrGC to motor pathways in the brain stem (Büttner-Ennever et al. 1996; Livingston and Mustari 2000; Mustari et al. 1994). The distinctive response of the postsaccadic pausers in the PrGC can be compared closely with that of the pretectal omnidirectional pausers (Mustari et al. 1994, 1997). The discharge characteristics of the pretectal pausers are essentially identical to those of the PrGC units in terms of latency and duration. Similarly the timing of activity in the distinctive postSR bursters of the PrGC is comparable to that of the pretectal pausers.
The directional tuning of smooth pursuit and visual slip responses in the PrGC is comparable to that reported for the pretectal nucleus of the optic tract (Hoffmann et al. 1988; Mustari and Fuchs 1990). The majority of PrGC units for which complete directional tuning curves were generated are tuned predominantly for ipsiversive pursuit. Additional units, for which complete tuning curves were not generated, are also tuned predominantly for ipsiversive pursuit. This is comparable to the predominantly ipsiversive tuning of neurons in the pretectal nucleus of the optic tract. Although we did not systematically test the velocity responses in PrGC neurons during pursuit, we did observe vigorous responses at velocities of 4–80°/s. This is within the range of velocities at which pretectal units respond vigorously (Hoffman et al. 1992; Mustari and Fuchs 1990). The response properties of NOT units are thought to derive in part from afferents originating in cortical areas MT and MST (Distler and Hoffmann 2001; Distler et al. 2002; Hoffmann et al. 1992). The PrGC was identified previously as an indirect source of cortical input to the pretectal nucleus of the optic tract (Büttner-Ennever et al. 1996). Our observations provide additional evidence to support this hypothesis.
Visual input to the pregeniculate complex
One possible function of the PrGC is the integration of ocular motor signals with visual input (Fig. 14). Ocular motor signals in the PrGC are not dependent on visual input as demonstrated by the presence of robust SR activity in the absence of visual stimulation (Büttner and Fuchs 1973; Livingston 2000). However, these signals may be modulated by visual stimuli. Visual input directly from the retina, and indirectly from areas 17, 18 and 19 of the occipital cortex (Hendrickson et al. 1970; Livingston and Mustari 2000; Ogren and Hendrickson 1976), suggests that neurons responsive to visual stimulation should be present in the PrGC, notably within the retinorecipient sublayer of the PrGC. Büttner and Fuchs (1973) reported that 10-μs flashes of diffuse light modulated the SR responses of some PrGC units. Similarly, phasic and tonic visual responses were noted for LGNv neurons in cat, rat, hamster, and rabbit (Harrington and Rusak 1991; Hayashi and Nagata 1981; Hughes and Ater 1977; Hughes and Chi 1983; Mathers and Mascetti 1975; Spear et al. 1977; Sumitome et al. 1979). Using prolonged visual stimulation similar to that during natural viewing, we confirmed visual modulation of SR responses in primate PrGC neurons. We also found two classes of units, with firing rates that were either tonically increased or decreased by visual stimulation.
The retinorecipient sublayer of the PrGC is largely GABAergic (Livingston and Mustari 2000). These GABAergic neurons may be excited by visual input as illustrated in Fig. 10B and subsequently inhibit other PrGC neurons. For example, the tonic inhibition of PrGC neurons such as the one shown in Fig. 10A may be mediated by GABAergic neurons intrinsic to the PrGC. GABAergic neurons of the retinorecipient sublayer in conjunction with visual input could also play a role in the modulation of the ocular motor signals relayed through the PrGC (Fig. 14).
The firing rates of pregeniculate neurons are extremely variable with some units exhibiting high tonic firing rates, whereas others are relatively inactive, exhibiting extremely low firing rates. Further, the firing rates of individual neurons are extremely irregular, as demonstrated by their high coefficients of variation (CV). Although the significance of this discharge irregularity is not clear, spike trains with high information content are predicted to exhibit highly irregular discharge, and to include high-frequency responses (Bialek and Rieke 1992; Rieke et al. 1999). This is the case for pregeniculate neurons, in which we found CVs of 0.35–1.99, and peak discharge rates of 78–442 spikes/s. The extremely irregular discharge rates of many PrGC units may reflect modulation of their activity by multiple convergent signals. In addition, there may be sources of input to the PrGC not yet identified contributing further to firing rate irregularity. Subsequent studies of the PrGC will require more rigorous analysis of how isolated signals modulate PrGC unit behavior and how the integration of multiple signals impacts pregeniculate discharge.
We have demonstrated that the timing of saccade-related activity in the PrGC is broadly distributed in terms of its initiation and duration. We interpret these results in conjunction with other observations, to suggest that the PrGC may play a role in gaze behavior other than saccadic control. The similarity of ocular motor signals in the PrGC to those recorded in parietal areas that project to it, suggests that the PrGC may represent a relay between cortex and brain stem pathways. In support of this hypothesis, projections from visual cortical areas to the PrGC and in turn from the PrGC to brain stem structures such as the pretectal nuclei and the superior colliculus have been documented. We have also shown that multiple ocular motor signals converge within the PrGC and on single PrGC units. These signals may be modulated by visual input directly from the retina or indirectly from the primary visual cortex. Integration of these visual and ocular motor signals is evidenced by the modulation of PrGC discharge rates and SR responses by visual stimulation.
The authors gratefully acknowledge the outstanding professional and technical support of B. Clough and the excellent technical assistance of J. Dawson. We thank Dr. Albert Fuchs of the University of Washington Regional Primate Research Center for the use of data acquisition and analysis software and Dr. Andrew Burrows for development of the CED software scripts. We also thank Drs. Galen Kaufman, Golda Leonard, and Michael Mustari for the thoughtful critique of this manuscript.
This work was supported by Development Grant 2548-01 from the John Sealy Memorial Endowment Fund for Biomedical Research and an National Eye Institute of Health Grant R29 EY-11546 to C. A. Livingston. S. R. Fedder was supported by a Summer Undergraduate Research Fellowship from the Marine Biomedical Institute, University of Texas Medical Branch.
↵1 Büttner and Fuchs (1973) originally used the term PGN in reference to the pregeniculate nucleus. To avoid confusion due to the more common use of PGN in reference to the perigeniculate nucleus, we use PrGC in reference to the pregeniculate complex, acknowledging its multiple subnuclei.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked ``advertisement'' in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2003 by the American Physiological Society