| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Physiological Optics and Vision Science Research Center, University of Alabama at Birmingham, Birmingham, Alabama 35294
Submitted 25 September 2002; accepted in final form 16 August 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
) are a critical component of the primate saccadic eye movement system. These neurons fire tonically in alert animals during fixation and pause just before and during saccades or quick phase eye movements of any size and direction (Everling et al. 1998; Raybourn and Keller 1977). OPNs must be silent for a saccade or quick phase to occur or to continue. Electrical microstimulation of the raphe interpositus delays the execution of a saccade and interrupts a saccade already started (Becker et al. 1981; Keller 1977; Keller et al. 1996; King and Fuchs 1977). OPNs exert this effect on saccades because they strongly inhibit pontine and midbrain saccadic medium lead burst (MLB) neurons. These burst neurons provide motoneurons with the eye velocity signals required for saccades.
The relationship between OPN activity and conjugate (i.e., without changes in depth) saccades of different sizes and durations has been investigated in cats (Evinger 1982; Paré and Guitton 1998) and in monkeys (Fuchs 1991). For the monkey, there is general agreement that there is a good correlation between saccadic start and pause start. OPNs resume firing around the saccadic end (Everling et al. 1998; Fuchs 1991), although the timing correlation with the end of the saccade is not as strong as that between pause onset and saccadic onset and varies among cells.
The observation that OPNs must pause for saccades to occur, together with the finding that they receive inputs from saccadic-related brain areas, places them in a crucial role in models of the saccadic system. Surprisingly, damage to OPNs does not cause the devastating effects such as opsoclonus that might be expected by the loss of burst neuron inhibition, as an early model (Zee and Robinson 1979) and some clinical studies (Ashe et al. 1991; Averbuch-Heller and Remler 1996) have suggested. The most evident effect of experimental damage to OPNs is slower but otherwise normal saccades (Kaneko 1996; Soetedjo et al. 2002). Although the precise mechanism is unknown, this result has been successfully simulated by the reduction of OPN activity within contemporary models of the saccadic system (Moschovakis 1994; Scudder 1988). In their models, the execution of a saccade is preceded by a gradual charging of the long lead burst neurons (LLBNs). At saccadic onset the LLBN signal starts to decrease. If a weakened OPN inhibition causes a premature release of the MLB cells during the integrating phase and therefore a premature start of the saccade, the local circuits will start to inhibit the LLBN integration before it reaches its optimal value. Consequently, the peak firing rate of the MLB cells will be lower. Behaviorally, this translates into a slower and longer saccade.
Several studies (Collewijn et al. 1995; Erkelens et al. 1989; van Leeuwen et al. 1998) have noted that horizontal saccades are often slowed when combined with vergence eye movements. Such mixed saccadic-vergence movements are typical of refixations in depth. We have noted that some OPNs show significant decreases in firing rate during vergence movements in the absence of saccades (Busettini and Mays 1999). Additional evidence for modification of OPN activity during vergence is suggested by the observation of Ramat et al. (1999) that random postsaccadic conjugate ocular oscillations occur during combined vergence/saccadic eye movements in humans. The authors interpreted these oscillations as evidence that the postpause resumption of OPN activity was delayed by vergence, allowing irregular firing of the MLBs after the end of the saccade proper. The observations that reduced OPN activity attributed to lesions results in slowed saccades and that vergence reduces OPN activity offer a possible explanation for slowed saccades during vergence. To test this hypothesis, we recorded the activity of OPNs associated with conjugate saccades and with saccades during vergence eye movements. OPN activity related to vergence without saccades was also measured. If reduced OPN activity were responsible for the slowed saccades seen during vergence, we would expect to see a significantly lower OPN activity level just before these saccades and a positive correlation between presaccadic activity and saccadic slowing.
Brief reports of some of these results were previously published (Busettini and Mays 1999; Reusser et al. 1995).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
Surgical procedures
Animals were trained to enter a primate chair before a sequence of aseptic surgical procedures. At the time of the surgery they were given an intramuscular injection of ketamine (Ketalar), and then intubated and maintained on isoflurane for the duration of the surgery. Heart rate, respiratory rate, blood pressure, O2 saturation, and body temperature were monitored continuously. During the postsurgical period, animals were given analgesics as needed to alleviate any discomfort, and training or experiments were started only after full recovery from each surgery.
In the initial surgery, stainless steel plates (Synthes) were attached to the skull with bone screws. A metal post was affixed to these strips with dental acrylic cement for head fixation. Following a protocol similar to that of Judge et al. (1980), a coil of fine wire (Biomed Wire AS633) was implanted under the conjunctiva of the right eye, so that eye position could be monitored using the magnetic search coil technique (Fuchs and Robinson 1966). After the animals reached a satisfactory level of training on simple saccadic and tracking trials, a 2nd eye coil was implanted on the left eye, allowing the monitoring of binocular eye movements. When the monkeys easily performed the more complex in-depth eye movements and preliminary behavioral recordings were completed, 2 recording cylinders were implanted with acrylic cement over 15-mm holes trephined into the skull. The 2 chambers were stereotaxically positioned bilaterally over the brain stem at a 20° angle to the sagittal plane, 14.5 mm lateral from the midline, and 1 or 2 mm anterior to ear-bar zero.
Behavioral task and visual display
The animal was placed in a primate chair with its head immobilized and was required to make transfers of fixation in depth and/or direction in response to target motion generated by a mirror haploscope with a system of lenses that matched the accommodative and vergence demands (Walton and Mays 2003). A pupillometer (ISCAN) was aligned with the right eye and functioned as a blink detector. The visual subtense for each monitor was ±20° (horizontal) and ±15° (vertical) and the range of vergence demand, limited by the corresponding maximum amount of accommodation demand obtainable with the system of lenses, was 13°. The monitors generating the visual stimuli had a vertical refresh rate of 90 Hz and the transition from the 1st to the 2nd target was completed within 2 video frames (22 ms). The targets were white Maltese crosses 1.2° wide on a black background. A Badal optical system kept the stimulus subtense on the retina constant. Thus changes in apparent depth were not matched by corresponding changes in stimulus size.
During behavioral training and recording, a brief auditory signal alerted the animal that a new trial was beginning. After 500 ms, an initial target appeared and the animal had 2,000 ms to look at it. The animal was required to maintain its fixation for 1,000–1,500 ms inside a ±4° window. Subsequently, the 1st target was turned off and the 2nd target was turned on. The animal was required to transfer fixation within 1,000 ms and maintain fixation of the new target for 600–1,500 ms. The selection of target steps induced the animals to make: 1) purely symmetrical vergence movements (i.e., slow convergence and divergence) by means of symmetric disparity steps; 2) horizontal and vertical saccades between targets at optical infinity (i.e., conjugate saccades between 2 far targets); 3) horizontal and vertical saccades with convergence (i.e., moving from a far to a near target); 4) horizontal and vertical saccades while converged (i.e., between 2 near targets); and 5) horizontal and vertical saccades during divergence (i.e., moving from a near to a far target). Although pure disparity steps enhanced the probability of saccadic-free smooth vergence movements, small saccadic intrusions were common and contaminated trials were analyzed as combined saccadic/vergence trials. Horizontal, oblique, and vertical versional target steps from 1 to 25° were used, along with vergence demand changes from 0 to 13°, although only subsets of these values were employed in a given session. Usually the 1st target started at the center of the display, but in some sessions its position was randomly offset to allow larger target steps. All trials were pseudo-randomly intermixed to avoid anticipation by the animal.
Data acquisition and single-unit recording
Presentation of the stimuli, reward administration, and acquisition of the eye and unit signals were made by a computer. The horizontal and vertical eye positions of both eyes were acquired at either 500 or 1,000 Hz. Asynchronous interrupt-driven events with a time resolution of 0.1 ms marked the occurrence of neuronal spikes as detected by a window discriminator (BAK Electronics). An additional channel, sampled at 20 kHz, was used to store the analog signal from the electrode. Low-impedance (0.1–0.3 M
) parylene-insulated tungsten microelectrodes (MicroProbe) with additional insulation were mounted in 26-gauge steel tubing. These electrodes were advanced through a 21-gauge hypodermic needle (used to penetrate the dura and as a guide tube) by a Kopf microdrive. Unit activity was filtered above 5 kHz to eliminate the high-frequency (28 kHz) interference produced by the magnetic field coils and was also high-pass filtered at 10 or 100 Hz depending on the amount of low frequency noise. The criteria used to determine that the electrode was located in the omnipause area were 1) the presence of neurons that displayed a constant firing rate during fixation and paused during saccades in all directions and for blinks; 2) the ability of microstimulation (maximum current 40 µA at 250 Hz) to block the occurrence of a saccade or stop an ongoing saccade in midflight; and 3) by visual observation of small marking lesions and/or electrode tracks in the omnipause area during histological reconstruction.
Data analysis
EYE MOVEMENTS. Right and left horizontal and vertical eye position traces (HR, VR, HL, VL) were linearized with 3rd-order polynomials and fit with a cubic spline with weight 1 x 108 (0.1 with the time expressed in ms). Velocity signals were obtained with a 2-point backward differentiation. Horizontal version (H) was computed as average (HR + HL)/2 of the 2 horizontal eye positions. Horizontal vergence position (VG) was calculated as HL – HR. For the vertical eye movements, the vertical eye positions were used to compute an average vertical version signal (VR + VL)/2 and all vertical measures were made on the vertical version (V) or the vertical version velocity (
). Rightward, upward, and convergence movements are represented by positive values. Pythagorean position (PY) was defined as 
and saccadic identification was performed using the Pythagorean velocity (P
). A preliminary analysis indicated minimal dependency of OPN activity on saccadic direction. Therefore we decided to characterize saccades on the basis of their Pythagorean amplitude, peak velocity, and duration without regard to saccadic direction. Saccades were defined as conjugate if the stimulus did not require a change in depth. Saccades were defined as divergent if the stimulus required a change in depth from near to far, and convergent if the stimulus required a change in depth from far to near. Using a cyclopean peak velocity of 40°/s as the minimum threshold for our automatic saccadic detection, all detected conjugate eye movements had a cyclopean amplitude–peak velocity relationship that followed the saccadic main sequence (Becker 1989). Saccades during vergence often had peak velocities lower than the conjugate main sequence but the deviations from the conjugate values were never large and were continuous with them, strongly suggesting that these movements were also saccades, albeit more or less slowed. We restricted the analysis to the primary saccades associated with the stimulus transition from the 1st to the 2nd target and to correctives saccades, if any, when the 2nd target was still on.
Initially, a velocity threshold of 20°/s applied to the Pythagorean velocity was used to mark saccadic onset and offset. This procedure proved unsuitable because the onsets of smaller, slower saccades were detected too late and their ends were detected too early. This problem was solved by the use of small timing corrections based on averaged acceleration values. All saccadic onset, end, and duration data are reported using adjusted values.
OPN ACTIVITY. The interval between the last prepause spike and saccadic onset determined the pause lead (Plead), whereas the interval between saccadic onset and the 1st resuming spike, indicating the end of the pause, was defined as resuming lag (Rlag).
Because a preliminary analysis showed that some OPNs are modulated with eye position, the firing rate of each cell when the animal was looking straight ahead with the eyes aligned at distance (vergence angle <2°) was used as a measure of the tonic activity during fixation. For each trial, baseline firing rate was defined as the average firing rate of the cell in the interval 0–40 ms after the visual target stepped from the center "far" position. The position of the 2nd target was irrelevant, given that no visual or motor responses to the target step occurred within 40 ms of target movement. An average baseline firing rate was calculated for each cell.
We determined whether the tonic OPN firing rate was sensitive to horizontal eye position of the left or right eyes, horizontal vergence, horizontal version, or vertical version by measuring the OPN firing rates during periods of fixation at various combinations of vergence and version values. This was done by displaying each trial so that average measures of eye position and firing rate over 100-ms intervals could be manually selected. Data were accepted if: 1) one of the 2 targets were present (i.e., not in a period of darkness); 2) there were no residual effects attributed to visual responses, eye movements, or blinks; 3) there were no changes associated with impending movements; 4) there was steady fixation of the target; and 5) the firing rate was constant. During long periods of fixation, measures were repeated every 200 ms, if possible. Only data sets with 
8 trials (i.e., 8 measures while fixating the 1st target and 8 measures while fixating the 2nd target) are reported. Visual responses were measured on a subset of the conjugate trials used to measure the baseline firing rate. In addition to the restrictions used earlier for the baseline measures, trials were selected for analysis if: 1) the 2nd target stepped to a different location with no change in vergence demand, and 2) the saccadic-related pause was 130 ms after the stimulus change, so that OPN activity in the 1st 100 ms after the target step would be unaffected by any presaccadic slowdown in activity. In preliminary analyses, we noted that visual responses, when present, started about 50–60 ms after stimulus onset. To quantify these responses, 2 measures were taken in each of the trials meeting these criteria. The 1st was the average firing rate in the 0- to 40-ms interval after target movement, and the 2nd measure was the average firing rate in the period 60–100 ms after target movement. For each neuron with 8 valid trials, a paired t-test compared the activity level in the 2nd period to the 1st. Unless specified otherwise, the statistical level for significance throughout the study is P < 0.01. For each cell, the onset and offset of the visual response were determined using an objective 2-segment linear curve fitting algorithm applied to the average firing rate profile. Before averaging, the individual firing profiles were interrupted 30 ms before the 1st saccadic onset. The average profile was stopped when the average was computed on fewer than 8 remaining observations.
The stimuli used to obtain smooth vergence responses were symmetric steps, inducing an abrupt change in disparity but not in (versional) eccentricity. The goal was to collect, for each cell and each of the 2 vergence directions, 8 trials of slow vergence without saccades or blinks. The mean firing rate of each trial in a 40-ms interval centered on the trial's peak vergence velocity was compared with the estimated baseline firing rate using a paired t-test. Both measures were adjusted to compensate for positional modulation, using the average eye positions inside the intervals and the previously determined position sensitivity coefficients. Using a measure centered on the peak vergence velocity, which occurred around 200 ms after stimulus onset, largely eliminated the possibility of contamination from the transient visual responses elicited by the stimulus onset. Unfortunately, some animals had an idiosyncratic tendency to associate an early, small saccade with all vergence movements in one of the 2 vergence directions even for pure symmetric disparity steps, forcing the elimination of the data set in that direction for that animal. However, trials containing saccades were included if saccadic onset, as defined by a Pythagorean velocity >10°/s, occurred 30 ms after the 40-ms measurement interval centered on the peak in horizontal vergence velocity. This was done because a preliminary analysis showed that OPN activity 30 ms or more before a saccade was unaffected by the occurrence of a saccade.
ASSESSMENT OF SACCADIC SLOWING. The peak velocity of saccades (Svpk) and saccadic duration (Sdur) are known (Becker 1989) to be monotonic functions of saccadic size (Ssize). To assess the degree of saccadic slowing associated with vergence, it was necessary to first control for the effects of saccadic size on peak Pythagorean velocity and duration. This was done by evaluating the following expressions for the conjugate trials for each cell
 | (1) |
 | (2) |
). Only cells with 8 conjugate observations were included in this analysis. These equations allowed us to assess, for each cell, the differences (DIFF) for peak saccadic velocity (DIFFSvpk) and duration (DIFFSdur) between saccades accompanied by either convergence or divergence and the conjugate saccades. The dependency of DIFF on the 2 variables thought likely to influence saccadic dynamics, vergence velocity at the start of the saccade (GONS) and saccadic disconjugacy (intrasaccadic vergence change or 
VG), were evaluated using linear regressions
 | (3) |
 | (4) |
VG is identified as DIFFSdur(VG). Measured values and associated averages are indicated in italics, whereas estimated parameters and associated averages are indicated in bold.
The relationship between OPN activity and saccades has usually (Everling 1998; Evinger 1982; Fuchs 1991; Paré and Guitton 1998) been expressed either in terms of pause lead or pause duration versus saccadic duration. We assessed pause lead as a linear function of saccadic size, peak velocity, and duration for the conjugate saccades of each cell. Of the 54 cells for which we had sufficient data, only 8 cells showed a significant modulation with one or more saccadic parameters. These relationships were small and in both directions and there was no significant average change in pause lead for the overall population with any of the 3 parameters. We therefore used a single variable to characterize pause lead for conjugate saccades
 | (5) |
GONS and VG using Eqs. 3 and 4. If there is a correlation between the deviation of pause lead from conjugate saccades and the perturbation in saccadic dynamics, we expect it to be mirrored by a correlation between DIFFSvpk and DIFFPlead and between DIFFSdur and DIFFPlead.
Previous reports in the monkey (Fuchs 1991) have indicated that there is a modest positive correlation between saccadic duration and pause duration. However, pause lead is a component of pause duration, given that pause duration is the sum of pause lead and resuming lag. The pause lead values for a given cell should be somewhat variable because the presaccadic activity of an OPN is likely to be relatively uncorrelated with the occurrence of the saccade and so its inclusion in pause duration should add a significant amount of uncorrelated noise to the relationship between pause duration and saccadic duration. If this is the case, it is more useful to evaluate the relationship between saccadic duration and the resuming lag. Accordingly, we evaluated the relationships between pause duration and saccadic duration as well as between resuming lag and saccadic duration for conjugate saccades using the following linear regression models
 | (6) |
 | (7) |
As anticipated, resuming lag proved to be more highly correlated with saccadic duration than did pause duration. Consequently, we examined the differences in resuming lag (DIFFRlag) for saccades with convergence and divergence relative to conjugate saccades of similar duration as a function of GONS and VG using Eqs. 3 and 4.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
Sixty-four OPNs were recorded from the nucleus raphe interpositus of 7 rhesus monkeys. As previously reported (Everling et al. 1998; Raybourn and Keller 1977), these cells had tonic activity during fixation but paused for saccades in all directions. Figure 1 shows data for one OPN. This cell shows clear evidence of a transient visual response with a peak around 75 ms from stimulus onset. The activity of this cell is clearly diminished for convergence (Fig. 1A) but not for divergence (Fig. 1C). As a consequence, the presaccadic pause lead can be much greater for a saccade with convergence (Fig. 1B) than for a saccade with divergence (Fig. 1D).
|
Modulation of firing rate with eye position
Overall, the baseline firing rate had a roughly Gaussian distribution with an average firing rate of 125 spikes/s and SD of 31 spikes/s (n = 64). The range was from 50 to 188 spikes/s. These values are very close to those reported by Everling et al. (1998).
We measured the firing rate of the cells when the animal was steadily fixating at different versional and vergence angles. Two different multiple regression analyses were used to determine whether the modulation fits either a conjugate + vergence ("Hering") model or a left eye/right eye ("Helmholtz") model. The Helmholtz model uses HR, HL, and V as independent variables. The Hering model uses H, VG, and V as independent variables. Using the Hering model, we found a very small and nonsystematic positional dependency with horizontal version and vertical version, as shown in the "H" and "V" histograms in Fig. 2. The average modulation over the entire population was 0.03 ± 0.48 spikes/° (±SD if not specified otherwise) for horizontal version on the 61 cells for which we had data, and 0.13 ± 0.65 spikes/° for vertical version on the 48 cells for which we had data. Overall, these averages were not significantly different from zero (P > 0.05). In contrast, 35 cells showed a robust, albeit small, modulation with vergence angle. This is evident looking at the histogram labeled "VG" in Fig. 2. The comparison of this histogram with the "HR," "HL" (Helmholtz model), and "H" histograms also suggests that this is a true vergence modulation with balanced contributions by the 2 eyes. We always observed opposite modulations for "HR" and "HL" components, and a much smaller and inconsistent "H" modulation. This would be expected if the modulation were linked to a positional contribution associated with vergence and not to monocular right or monocular left eye signals. Twenty-five cells showed a significant decrease of their activity for increased convergence, whereas 10 showed a significant increase, with a continuous distribution of values ranging from –7.27 spikes/° (SE = ±0.26 spikes/°) to 2.38 spikes/° (SE = ±0.08 spikes/°). Overall, the modulation with vergence position was –0.50 ± 1.51 spikes/°, which was significantly different from zero. Not surprisingly, the modulation with horizontal right and left eye positions, taken separately, had opposite signs. For HR the average over the entire population was 0.52 ± 1.52 spikes/° and significant. For HL the average value was –0.47 ± 1.55 spikes/° (P < 0.02).
|
Visual response
Responses of monkey OPNs to visual stimuli were assessed by synchronizing neuronal activity rasters either with the target step (Fig. 3A) or with saccadic onset (Fig. 3B). Figure 3 shows that a presaccadic transient increase in firing rate was associated with the stimulus transition (Fig. 3A) and not with the occurrence of a saccade (Fig. 3B). For most cells, like the one shown in Fig. 3, the visual response ended before the onset of a saccade. Even though vergence movements tended to have shorter latencies than saccades, some visual responses ended before vergence onset as well. This can be seen for 2 of the cells shown in Fig. 4. Trials were synchronized on stimulus onset before averaging. Figure 4A shows averaged vergence velocity (G) and firing rate (FR) data for 3 trial types (conjugate "far" saccade, convergence, divergence). The visual response ends before any eye movement. This cell shows a decrease in firing rate related to static convergence (indicated by the asterisk) and a much more pronounced suppression of the firing rate for dynamic convergence than for dynamic divergence. Figure 4B shows data for 2 levels of convergence, one level of divergence and for conjugate "far" saccades. The visual response is highly stereotyped and identical for all 4 conditions, and the firing rate decreases for both convergence and divergence. The delayed peak in divergence velocity is matched by a similar delay in the peak of the firing rate suppression, again suggesting that the main source for the dynamical OPN modulation is a signal related to vergence velocity. Figure 4C shows data for averaged conjugate "far" saccades and convergence for a cell that had no detectable visual response. The robust modulation of OPN activity associated with convergence indicates that this change in activity is unrelated to the presence of visual activation. Finally, data are shown for conjugate "far" saccades, 2 levels of convergence and one level of divergence for one of only 3 cells that appear to show a continuation of the visual response throughout the convergence movement (Fig. 4D). It is more likely, on the basis of the conjugate average profile and the lack of a similar prolonged increase of firing for divergence, that a decaying visual response was then masked by an actual increase in firing rate associated with the ongoing convergence.
|
|
Quantitative measures of the visual responses were performed on the conjugate "far" data sets, as described in METHODS. An inspection of average firing data for all cells showed that no visual response occurred earlier than 45–50 ms after target step, and that the main visual response was seen after 60 ms. The average activity in the 60- to 100-ms visual response period was compared with the 40-ms period of baseline activity. Fifty-three of the 64 OPNs tested (83%) showed statistically reliable visual responses in the visual response period. On average, OPNs increased their firing rate 17.6% (±13.3% SD; P < 0.000002) over baseline in the visual response period (range –3.6 to 76.3%). No cell showed a significant decrease in activity during the visual response period. The average latency of the visual response, measured on the 53 cells with significant visually related activity, was 57 ± 5 ms (range 47 to 71 ms). On average, the visual response for the same cells ended 140 ± 33 ms (range 77 to 240 ms) after target onset.
OPN modulation with slow vergence
Figure 4 indicates that some OPNs transiently decreased their firing rate for convergence, divergence, or both, in the absence of saccades. Figure 5 shows examples of these responses, together with trial-by-trial activity rasters, synchronized on the peak of the vergence velocity. The horizontal dotted lines indicate the baseline firing rate of each cell. It is important to note from the raster displays that the decreases in firing rate associated with vergence were always gradual, unlike the abrupt pauses associated with saccades. As Fig. 5 shows, small but robust visual responses (elevated discharge before vergence onset) were also observed for the symmetric disparity stimuli used to elicit the smooth vergence responses. For some cells, the visual response was so large as to possibly mask the earliest changes in activity associated with the vergence (Fig. 5C), and indeed, the presence of the visual responses on so many OPNs made it impossible to estimate the overall latency of these vergence-related changes in activity. No OPN showed a significant increase in activity associated with smooth divergence and only 4 OPNs showed a significant increase in activity associated with smooth convergence. One of these 4 cases is shown in Fig. 5D. Sufficient data were available for a quantitative analysis of 29 OPNs for convergence transients without saccades (Table 1) and 15 cells for divergence transients without saccades (Table 2). Note that there is only partial overlap in the cells shown in these 2 tables. The measures reported are the averages of the single trial average measures taken in 40-ms intervals centered on peak vergence velocity. With the vergence velocity peaks occurring around 200 ms from stimulus onset, it is very unlikely that the visual responses, with an average offset time of 140 ms from stimulus onset, influenced the measures. For convergence, 17 of the 29 OPNs (59%) showed a significant decrease in firing rate, and 4 (14%) showed a significant increase. For divergence, 7 OPNs (47%) showed a significant decrease. For divergence, because of the negative sign of the vergence velocity, a decrease in firing rate is associated with a positive value of the modulation coefficient Dp. The average depth of OPN activity modulation for convergence was –0.37%/(°/s) [SD ±0.64%/(°/s); range –2.51 to 0.70%/(°/s)]. Overall, this modulation was significantly different from zero. The average depth of modulation for divergence was 0.38%/(°/s) [SD ±0.36%/(°/s); range –0.12 to 1.27%/(°/s)], which was also significantly different from zero. For both convergence and divergence the average firing rate of the OPN population decreased, in agreement with our initial hypothesis of a suppressive effect of slow vergence on the OPNs. For our highest vergence velocities, around 70°/s, a slowdown linked to vergence velocity of –0.37%/(°/s) translates to a 26% reduction in overall activity of the OPN population.
|
|
|
There was a significant trend for the cells with the strongest dynamical convergence modulations to also show the strongest positional vergence modulations. Indeed, for convergence (Table 1) the dynamic modulation (Dp) could be estimated as 0.47 x Staticmod with R2 value of 0.64 (t-value = 8.5). However, a similar correlation was not observed for divergence.
Effects of static vergence on pause lead
We determined whether the OPNs show modifications in the pause lead values between conjugate saccades executed at different (static) vergence angles. Forty-six cells had 8 conjugate saccades executed at static vergence angles covering a range of values >5°. Of these cells, 43 cells (93%) did not show any significant change in pause lead with static vergence angle. Only 2 cells showed significant increases in pause lead with static vergence angle (0.40 and 0.44 ms/°) and one cell showed a significant decrease (–0.66 ms/°). The average value for the population was 0.07 ms/° (SD ±0.35 ms/°; range –0.72 to 0.89 ms/°) and was not significantly different from zero (P > 0.05). This result strongly suggests that the changes in baseline firing rates associated with static vergence are too small to modify the average pause lead. A comparison of the saccadic dynamics for similar saccadic sizes between "near'" and "far" saccades also failed to show any statistical difference and therefore to compute the conjugate estimates (see following text), we pooled both "near" and "far" conjugate saccades.
Quantitative analysis of conjugate saccades
The averages of the conjugate estimates for the 54 cells for which we had sufficient data are illustrated in Table 3. The values of the conjugate peak-size main sequence varied only slightly among sessions and animals. The quality of the estimates, expressed by the R2, was extremely high, with an average of 0.975 and was always above 0.90. As expected (Becker 1989), the estimates of the conjugate duration-size main sequence had lower R2, with an average of 0.84 and a range from 0.33 to 0.98. Nonetheless, the ranges of the estimated parameters were still relatively small.
|
The pause lead values were, on the contrary, highly variable. As suggested in METHODS, this would be expected if the presaccadic firing of the cell were largely unaffected by the forthcoming saccade and yet abruptly stopped by a trigger signal. The plots in Fig. 6 show that this is indeed the case. Figure 6A shows that the average duration of the last prepause interspike interval (ISI) was only slightly longer than the average baseline ISI after correction for the positional modulation. The average lengthening was 27% of the corrected baseline ISI. This finding is consistent with the one reported by Everling et al. (1998). Functionally, this means that the OPNs are firing near baseline up to the arrival of the trigger signal. Pause lead, which can be thought of as the sum of the trigger lead and of the ISI cut short by the trigger, is therefore expected to have, for each cell, an average and SD value linearly related to the presaccadic firing rate, the best available estimate of which is the average duration of the last prepause ISI of the cell. Figure 6, B and C confirm these predictions. Both average pause lead (Fig. 6B) and SD of pause lead (Fig. 6C) are positively correlated with the duration of the last prepause ISI. The scatter in the conjugate estimates of pause lead is therefore a secondary effect of the variability of the baseline firing rates among the OPN cells. Theoretically, the intercept of the regression line in Fig. 6B (4.1 ms) is the pause lead of a cell with 0-ms duration of the prepause ISI and therefore the value of the trigger lead itself. We expect the actual trigger signal to start a few milliseconds earlier, given that such a signal requires some time to drive the OPNs to a full stop from their presaccadic firing rates.
|
The inherent random scatter of pause lead is expected to cause a significant reduction in the correlation between pause duration and saccadic duration. This effect can be eliminated by using resuming lag instead of pause duration when comparing neuronal activity to saccadic duration. Figure 7A shows, for the conjugate saccades from cell 01023.713, the relationship between pause duration and saccadic duration, with R2 value of only 0.26. Figure 7B shows the same data set, but now as the relationship between resuming lag and saccadic duration, with a dramatic increase of the R2 to 0.80. This was a consistent finding for all cells, as it can be seen comparing the conjugate estimate averages of pause duration and resuming lag in Table 3, where the average R2 increased from 0.48 to 0.72. Of the 54 cells of the data set, 43 cells (80%) had R2 values higher than 0.6 between resuming lag and saccadic duration. For pause duration versus saccadic duration, only 21 cells (39%) had R2 values higher than 0.6. Even though the use of resuming lag greatly improved the relationship between the activity of the cell and the behavior, the variability among cells in both the intercept and slope of the resuming lag versus saccadic duration for conjugate saccades remained high. This can be seen in Fig. 7C, which shows the linear regressions of the 54 resuming lag versus saccadic duration comparisons, which have an average slope of 0.85.
|
Differences between conjugate and disconjugate saccades
The analyses of saccadic peak velocity, duration, pause lead, and resuming lag for conjugate saccades allowed us to estimate their variations for saccades executed during vergence movements. Table 4 shows the deviations from the conjugate averages for the 54 cells for which we had sufficient data for saccades with convergence and Table 5 shows the deviations from the conjugate averages for the 19 cells for which we had sufficient data for saccades with divergence.
|
|
For both convergence and divergence, saccades executed during changes in depth were significantly slower and longer than conjugate saccades of similar size. The slowing and lengthening were significantly larger for higher GONS and larger VG values. For convergence, the average deviation in peak saccadic velocity (DIFFSvpk), estimated as an average of the average deviation from the conjugate peak-size main sequence model for each cell, was –15.3°/s (±13.3 SD, range –64.9 to 7.6°/s), and was significant for 41 of the 54 data sets (76%). Both the DIFFSvpk(GONS) and DIFFSvpk(VG) average slopes were significantly different from zero, indicating that the slowing was related to both of these variables. Although these relationships were robust, being significant for VG in 48 of the 54 data sets (89%), the R2 values were usually poor, with an average of 0.18 for DIFFSvpk(GONS) and of 0.39 for DIFFSvpk(VG), suggesting that other factors might play a role in determining the slowdown. Very similar results were obtained for divergence (Table 5), suggesting that the degree of slowing and the robustness of the effects, on average, are similar for convergence and divergence.
For a given saccadic size, a slower saccade should take longer to reach its goal. Therefore we expect that data sets with larger negative average DIFFSvpk values will also show larger positive DIFFSdur averages. This is clearly the case, as can be seen by comparing the Svpk-related parameters and Sdur-related parameters in Tables 4 and 5. This consistency was even more evident when the comparisons were made for each cell. Figure 8A shows the expected negative relationship (R2 = 0.57) between the DIFFSdur and DIFFSvpk averages for the 54 convergent data sets (in black) and the 19 divergence data sets (in gray). The covariation between the slopes of the regression terms DIFFSdur and DIFFSvpk with VG was even stronger. Figure 8B shows the correlation between the DIFFSdur(VG) and DIFFSvpk(VG) slope values (R2 = 0.83). Similar values were seen for slope parameters related to GONS. The large range in values for DIFFSvpk and, consequently, DIFFSdur, as well as for DIFFSvpk(VG) and, consequently, for DIFFSdur(VG), suggests that slowing and lengthening of saccades during vergence varies significantly between animals and probably also between sessions for the same animal. Nonetheless, the covariations in Fig. 8 indicate that slowing and lengthening are interchangeable in describing the behavioral effects and that they are both closely related to vergence-related variables VG and GONS.
|
A central hypothesis is that the behavioral changes in saccadic dynamics are mirrored by modifications in pause lead. For convergence (Table 4) there was a significant lengthening of pause lead in 22 data sets (41%) and a significant shortening in 4 data sets (7%), consistent with what was observed for the convergence smooth vergence data, resulting in an overall average DIFFPlead of 2.6 ms, an increase that was significantly different from zero. The modulation of the firing of the OPNs with vergence velocity, DIFFPlead(GONS), was also significantly positive for 15 cells (24%) and significantly negative for 3 cells (6%), with an overall (significantly positive) average of 0.075 ms/(°/s). Although the results regarding DIFFPlead(VG) were similar, the overall average was not significantly different from zero. For divergence (Table 5), although the pattern of the results was consistent with the divergence smooth vergence data, the overall averages failed to reach significance, even though the behavioral effects on saccadic dynamics were as strong as for convergence. Four cells (21%) showed a significantly longer DIFFPlead for divergence and 5 cells (26%) showed a significant DIFFPlead(GONS) but their combined effects were not sufficient to generate an overall significant effect. As with the divergence smooth vergence data, for divergence no cell showed significantly shorter pause lead values or positive slopes.
If the modifications in pause lead are correlated with the changes in saccadic dynamics, we expect to find significant regression slopes DIFFSvpk(DIFFPlead) and DIFFSdur (DIFFPlead). The results, shown on the last 2 lines of Tables 4 and 5, suggest, from the average R2 values, very little, if any, trial-by-trial correlation. Furthermore, the overall average slopes were not significantly different from zero, and indeed for divergence, only 3 cells (11%) reached significance for DIFFSvpk(DIFFPlead). Nonetheless, 15 cells (28%) reached significance for DIFFSdur(DIFFPlead) for convergence, 14 of which had positive values and one was negative. Although the trend was in the right direction for both convergence and divergence, if the average R2 can be taken as measure of a functional linkage between DIFFPlead and DIFFSvpk, and consequently with DIFFSdur, their values in the range 0.05 to 0.08 argue against such a linkage.
With regard to the suggestion of Ramat et al. (1999) that the postsaccadic versional oscillations during vergence in humans are attributed to a delayed resumption of the OPN activity, we determined whether there is any significant change in resuming lag (DIFFRlag) for saccades with vergence. If saccadic lengthening is matched by a parallel lengthening of resuming lag, we would expect to find no significant DIFFRlag, which would argue against Ramat's hypothesis. Interestingly, many cells (29 or 54% for convergence and 9 or 47% for divergence) showed a significantly negative DIFFRlag, suggesting that, for these cells, the saccadic lengthening was greater than the pause increase or, in other words, the OPNs resumed their activity earlier with respect to saccadic end for slower and longer saccades, which is the opposite of what was suggested. Only 11 cells (20%) in the convergence data set and one cell (5%) in the divergence data set showed a significant positive DIFFRlag. This is illustrated in Fig. 9A, which shows that these cells were also the cells with the largest (positive) DIFFPlead values. This is consistent with the vergence effects acting similarly on both the beginning and end of the OPN pause. Further evidence of this can be seen in Fig. 9B, where the resuming lag slope value [DIFFRlag(VG)] is plotted as a function of the pause lead slope value [DIFFPlead(VG)]. The linear regression of the combined convergent and divergent data had R2 value of 0.71. However, the positive DIFFRlag contribution associated with the cells suppressed by the ongoing vergence was cancelled by the negative contribution associated with the cells with no modulation, with, as a result, no overall significant change in resuming lag between conjugate and disconjugate saccades.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
Although the effects of saccades on vergence velocity have been extensively reported (Enright 1984; Kenyon et al. 1980; Ono and Nakamizo 1978; Zee et al. 1992), the possibility that vergence might have some effect on saccades has attracted much less attention. In studies with humans, horizontal saccades during vergence often showed slower velocities and resulted in longer durations than conjugate saccades of comparable size (Collewijn et al. 1995; Erkelens et al. 1989) but the effects of vergence on vertical saccades were less clear (van Leeuwen et al. 1998). In monkeys, superior colliculus (tectal) long lead bursters (TLLBs) show reduced activity for both horizontal and vertical saccades during vergence (Walton and Mays 2003) and we have preliminary evidence (unpublished observations) that this is also seen in medium lead burst neurons (MLBs). No precise behavioral quantification of the saccadic slowing during an ongoing vergence movement is available for monkeys. Nonetheless, the finding of important vergence-related changes in the firing of saccadic-related bursters is a clear indication that vergence affects the saccadic system. Furthermore, those neuronal changes rule out that the saccadic slowing could be attributed to nonlinearities at the level of the oculomotor plant during combined vergence/saccadic eye movements. The omnidirectionality of this effect and its neuronal origin suggest an involvement of the OPNs. Furthermore, partial lesions of the nucleus raphe interpositus, where the OPNs are located, result in abnormally long and slow saccades (Kaneko 1996; Soetedjo et al. 2002). Optimal saccadic dynamics, as pointed out by Scudder (1988) and Moschovakis (1994) in explaining such OPN lesion-related saccadic slowing, require a precisely timed release of the MLBs by the OPNs. Sparks et al. (1987) found that stimulation in some locations of the pons can cause premature saccadic triggering, an indication of the presence of saccadic-related activity in the pons long before the actual execution of the movement. They estimated that, for normal visually directed saccades, there is a gradual presaccadic signal buildup lasting 100 ms or more within the premotor pontine structures before saccadic initiation. The OPN inhibition of the MLBs would block any early firing until the buildup is completed. A lesion of the OPN area would decrease the inhibitory action of the OPNs on the MLBs with a partial, early release of the bursters before the completion of the buildup, with abnormal saccadic profiles.
OPN behavior during static version, static vergence, and saccadic-free slow vergence
The first question addressed was whether OPN firing is affected by static eye position and vergence angle. With regard to versional eye position, our static data are in agreement with previous reports (Büttner-Ennever et al. 1988; Luschei and Fuchs 1972) stating that OPN firing rates during fixation do not change with static eye position. These previous studies did not manipulate viewing distance and therefore the static vergence angle of the animal was always the same, so that the positional changes were restricted to the versional components alone. The manipulation of the static vergence angle, on the contrary, showed a much stronger modulation for many cells. This modulation presented a continuous distribution of values across cells with a bias for a decrease in activity for increasing convergence angle. The result was an overall small, but significant, suppression of the average firing rate of the population sample by –0.50 spikes/°. The almost perfect symmetry of the contributions to the modulation from the 2 eyes clearly suggests that this modulation was from the vergence system, where the contributions from the 2 eyes are presumed to be symmetrically distributed (Hering 1868).
A subset of OPNs also showed a significant transient modulation associated with ongoing saccadic-free slow vergence after accounting for positional effects. Its time course followed, to a 1st approximation, the vergence velocity profile. This modulation was always smooth, and not the abrupt pause seen for saccades, indicating that OPNs do not act as gates for vergence commands. Furthermore, no cell showed a modulation strong enough to drive the cell to complete inhibition during the smooth vergence response. Similarly to the positional modulation, the modulation with slow convergence dynamics had a continuous distribution of values from a significant decrease in firing rate to a significant increase. For divergence, no cell showed a significant increase in firing rate. Unlike the positional data, we did not explicitly test the hypothesis that these dynamic effects are related to vergence per se, as opposed to the slow movements of either eye. Indeed, Missal and Keller (2002) recently reported a modulation of OPNs with smooth pursuit velocity (i.e., with versional velocity contributions). However, the fact that OPNs were modulated for vergence position suggests that the dynamic effects may well be related to vergence. Furthermore, a monocular modulation would be incompatible with the fact that the overall modulation was a slowdown for both convergence and divergence. A hypothetical "left eye" cell would accelerate (decelerate) its firing rate for convergence and decelerate (accelerate) its firing rate for divergence, whereas the most common behavior was to have the same sign of the modulation for both convergence or divergence or no modulation for one of the 2 vergence directions.
Effects of vergence on saccadic dynamics
Because we examined saccades of different sizes, it was necessary to first account for the effects of saccadic size on peak velocity and duration for conjugate movements (Table 3) and then assess the differences seen when saccades were executed during vergence movements (Tables 4 and 5). The result was a small but significant slowing (–15.3°/s) and lengthening (+3.0 ms) of saccades during convergence (Table 4), and a similar effect (–17.6°/s; +3.0 ms) for divergence (Table 5). Although some data sets did not show significant slowing and lengthening, significant effects were seen in most of the data sets for all animals. The failure to find a significant effect on some data sets may well have been attributable to a smaller number of trials for some cells, or to a particular combination of target steps. We suspect that there may be other variables that affect saccadic dynamics in these situations, such as variations in the timing of the saccades relative to the vergence movements, but these considerations are beyond the scope of the present investigation. In any event, the behavioral effect was quite reliable, albeit relatively small in magnitude.
We examined the effects of 2 vergence-related variables on saccadic dynamics (GONS and VG). The former was selected because it provides an index of the magnitude of the vergence velocity at the onset of the saccade, whereas the latter was selected because it indicates the overall disconjugacy of the saccade. Both GONS and VG were significantly correlated with saccadic slowing and lengthening, although VG appears to be the more potent variable, judging from its consistently larger R2 values.
Pause lead
There was no consistent relationship between saccadic parameters such as size, peak velocity or duration, and pause lead for conjugate saccades. This is in agreement with the report of a lack of dependency of pause lead with saccadic size (peak velocity and duration were not tested) in alert head-fixed cats (Paré and Guitton 1998). Pause lead was significantly longer before saccades during convergence, but not for divergence. However, even the effect associated with convergence was small, and was seen on fewer than one-half of the cells. Not surprisingly, many of these cells also showed decreases in firing rate for smooth convergence. Indeed, of the 16 cells in Table 1 that decreased their firing rate for smooth convergence for which we had corresponding data, 11 had a significantly longer average pause lead for saccades with convergence than for saccades without vergence. As expected, GONS, which is associated with reduced activity in these cells for smooth convergence, was also associated with longer pause leads for saccades with convergence. There was no consistent effect on pause lead associated with VG for either convergence or divergence.
Duration of the OPN pause
The strong reciprocal inhibition between OPNs and horizontal
