| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Aerospace Medical Research Unit, Department of Physiology, McGill University, Montreal, Quebec H3G 1Y6, Canada
Submitted 10 February 2003; accepted in final form 26 March 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
) elegantly demonstrated
that to generate saccades, motor commands proportional to the velocity of the
eyes must be generated to overcome the viscous properties of the oculomotor
plant. Furthermore, to keep the eyes at eccentric positions in the orbit, the
brain must provide an additional and sustained motor command to the
extraocular muscles to offset the restoring elastic forces of the oculomotor
plant. When combined, these two motor commands would generate discharge
patterns on motoneurons that are often referred to as
"burst-tonic," or "pulse-step." It has been
experimentally confirmed that burst-tonic signals are indeed carried by
extraocular motoneurons during conjugate saccadic eye movements
(Goldstein 1983;
Keller 1973;
Robinson 1970;
Robinson and Keller 1972;
Sylvestre and Cullen 1999a;
Van Gisbergen et al. 1981).
Conjugate saccades reorient the two visual axes between targets located at a
same depth and produce equal amplitude rotations of the two eyeballs.
During conjugate saccades, there is general agreement that the
"burst" discharge of brain stem saccadic burst neurons, which is
roughly proportional to the velocity of the eyes
(Cullen and Guitton 1997), is
integrated (in the mathematical sense) to produce the eye-position command
carried by motoneurons (reviewed in
Scudder et al. 2002). This
process is believed to occur through a distributed network known as the
"oculomotor neural integrator" (NI). A number of studies (reviewed
in Fukushima et al. 1992;
McCrea 1988) have implicated
the nucleus prepositus hypoglossi (NPH) and the adjacent medial vestibular
nucleus (MVN) in the process of neural integration for horizontal eye
movements. First, lesions made to these neural structures were shown to
dramatically impair gaze-holding abilities
(Arnold et al. 1999;
Kaneko 1997;
Mettens et al. 1994) and to a
lesser extent perturb eye-movement dynamics (Kaneko
1997,
1999;
Mettens et al. 1994). Second,
it was shown that different classes of neurons that are distributed across the
NPH/MVN have discharge properties consistent with those predicted for NI
neurons. Among these types of neurons are burst-tonic (BT) and tonic (T) units
(also termed burst-position and position neurons, respectively). These neurons
all carry an eye-position-related signal (monkey:
Cullen et al. 1993;
McFarland and Fuchs 1992;
Scudder and Fuchs 1992; cat:
Delgado-Garcia et al. 1989;
Escudero et al. 1992;
Lopez-Barneo et al. 1982) and
do not respond to vestibular stimulation during cancellation of the VOR
(Cullen et al. 1993;
McFarland and Fuchs 1992).
Furthermore, in monkeys, the majority carry an eye-velocity-related signal
during saccades that gives them BT discharge characteristics
(Cullen et al. 1993;
McFarland and Fuchs 1992;
Scudder and Fuchs 1992). There
is strong evidence that primate BT neurons project directly to the abducens
nucleus (McFarland and Fuchs
1992; Scudder and Fuchs
1992).
The primary objective of this study was to determine the role of BT neurons in the binocular control of disjunctive saccades. These eye movements, which we frequently utilize to reorient our visual axes between targets located at different eccentricities and at different depths relative to our eyes, are characterized by the two eyes rotating by different angles and with markedly different dynamics. We also compare the signals carried by individual neurons during conjugate saccades, disjunctive saccades and fixation.
To date, the premotor control of binocular eye movements has been primarily
studied during nonsaccadic disjunctive eye movements. Most of these
prior studies that have characterized the neurophysiology underlying the
binocular control of eye movements (e.g.,
King and Zhou 2000;
Mays 1998) have interpreted
their results with respect to theoretical frameworks inspired by the original
work of Hering (1868) and
Helmholtz (1910). Simplified
schemas illustrating the general concepts of each of these theories are shown
in Fig. 1, A and
B, respectively. The first schema
(Fig. 1A) highlights
that although the horizontal neural integrator is likely shared across many
oculomotor behaviors, such as saccades, smooth pursuit, and the
vestibuloocular reflex (reviewed by
Fukushima et al. 1992;
Moschovakis 1997), it should
be exclusive to conjugate eye movements. In this view, a separate NI would
exist for the vergence eye movements that ensure binocular alignment between
targets located at different depths (e.g., Mays and Gamlin
1995a,b;
Zee et al. 1992). A clear
prediction of such a model structure is that NPH/MVN neurons are the substrate
for a conjugate integrator, and therefore should solely encode the conjugate
movements of the eyes (Fig.
1A) (Mays and Gamlin 1995, 1996).
|
There is experimental evidence that both supports and contradicts this
concept. In support of this concept, a population of neurons in the
mesencephalon, termed near-response neurons, have been identified that encode
disparity and vergence position and velocity signals during symmetric vergence
shifts (Mays 1984; Mays and
Gamlin
1995a,b)
and project to medial rectus motoneurons (Zhang et al.
1991,
1992). Vergence-velocity
neurons, which are thought to project to near-response neurons, have also been
identified (Mays et al. 1986;
Mays and Gamlin 1995b). These
neurons could form the vergence subsystem required in this schema. However,
there is no evidence to date indicating that near-response neurons, or
vergence-velocity neurons, project to lateral rectus motoneurons in the
abducens nuclei (Fig.
1A, - - -) (Gamlin
1999). Moreover, other routes through which vergence-related
signals could reach the abducens nuclei, for example via oculomotor
internuclear neurons, have failed to demonstrate appropriate vergence
modulations (Clendaniel and Mays
1994). Thus while near-response and vergence-velocity neurons most
certainly play a role in shaping the discharge patterns of medial rectus
motoneurons during vergence eye movements, their involvement in modulating the
discharge patterns of abducens nucleus neurons (ABN) remains questionable.
This is an important concern given that abducens nucleus neurons' firing rates
are well modulated by vergence position and velocity-related signals during
disjunctive fixation and disjunctive saccades
(Sylvestre and Cullen
2002).
The second model structure, shown in
Fig. 1B, is more
consistent with Helmholtz's hypothesis (e.g.,
Zee et al. 1992; Zhou and King
1998 2000). Here, rather than
using conjugate and vergence coordinates to control binocular movements (as in
Fig. 1A), the brain
utilizes right and left eye coordinates. Consequently, a prediction of this
scheme is that neurons in the saccadic circuitry, including BT neurons, should
be monocular. In agreement with this concept, it has been shown that BT
neurons encode the position and velocity of a single eye during disjunctive
smooth pursuit (Zhou and King
1996). Moreover, during disjunctive fixation, Chen-Huang and
McCrea (1999) have noted that
BT neurons in the MVN appear to preferentially encode the position of a single
eye but provided no quantitative analysis of this observation. It should be
noted, however, that there is strong experimental evidence from abducens
neurons, during nonsaccadic disjunctive eye movements
(Gamlin and Mays 1992;
Gamlin et al. 1989;
Keller 1973;
Keller and Robinson 1972;
King et al. 1994;
Mays and Porter 1984;
Sylvestre and Cullen 2002;
Zhou and King 1996,
1998), which indicate that
this model's predictions clearly do not hold. Specifically, it has been shown
that abducens internuclear neurons and motoneurons encode similar signals
during disjunctive eye movements (Fig.
1B). Finally, this model would offer no explanation for
the existence of near-response and vergence-velocity neurons.
Most recently, a very limited number of studies have begun to address which
of these model structures best describes the premotor control of disjunctive
saccades. On the one hand, we have reported that the discharge dynamics of all
abducens neurons are similar during disjunctive saccades, implying that
motoneurons and internuclear neurons do not carry different information
(Sylvestre and Cullen 2002).
Thus as during nonsaccadic disjunctive eye movements, experimental data are
not completely consistent with the predictions of the model structure shown in
Fig. 1B. On the other
hand, it has been shown that the number of spikes generated by excitatory
burst neurons (EBNs) are generally related to the displacement of a single eye
(Zhou and King 1998). Thus it
has been argued that EBNs encode monocular signals, consistent with
Fig. 1B. Given that
EBNs provide a significant drive to BT neurons during disjunctive saccades,
these results suggest that BT neurons would also encode disjunctive signals
during this behavior. In the present study, we test this prediction.
Taken together, the results from prior single-unit recording experiments
cannot be accurately represented by either of these straightforward conceptual
frameworks. Accordingly, more physiologically realistic frameworks have been
proposed (Cova and Galiana
1994,
1995,
1996; King and Zhou
2000,
2002;
Sylvestre et al. 2002). These
more recent models have in common that abducens internuclear neurons and
motoneurons encode similar signals, near-response neurons project to
oculomotor motoneurons but not to abducens neurons, the vergence signals
carried by near-response neurons correct for the "inappropriate"
signals carried by abducens internuclear neurons, and the activity of
near-response neurons is shaped by bilateral projections from monocular NI
neurons. The main difference among these models is that during disjunctive
saccades, the Cova and Galiana
(1996; see also
Sylvestre et al. 2002) model
includes that premotor neurons encode the movements of the ipsilateral eye,
whereas in the King and Zhou
(2000,
2002) model, brain stem
premotor neurons are organized in two separate channels, one for each eye.
In the present study, we demonstrate that neither of the conceptual
frameworks presented in Fig. 1, A
and B, can fully account for the population discharges of
BT neurons during disjunctive saccades and fixation. For each BT neuron we
recorded, our analysis approach first involved quantifying its discharge
dynamics during the simpler case of conjugate saccades. Then during
disjunctive saccades, we determined whether the activity patterns of that
neuron could be predicted based on its discharge dynamics during conjugate
saccades. We also directly quantified the discharge dynamics of this same
neuron during disjunctive saccades using a model that accounted separately for
the movements of both eyes. We argue in DISCUSSION that our data
are more consistent with the predictions of the more recent models by Cova and
Galiana (1994,
1995,
1996) and King and Zhou
(2000,
2002).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
). A stainless steel post that allowed the complete
immobilization of the animal's head and two stainless steel recording chambers
oriented stereotaxically toward the right and left abducens nuclei,
respectively, were fastened to the animal's skull using cranial screws and
dental acrylic. An eye coil (3 loops of Teflon-coated stainless steel wire,
18- to 19-mm diam) was implanted in each eye
(Judge et al. 1980). All
procedures were approved by the McGill University Animal Care Committee and
were in compliance with the guidelines of the Canadian Council on Animal
Care. Data-acquisition procedures
During the experiment, the head-restrained monkey was seated in a primate
chair that rested on a vestibular turntable. Targets, rewards, on-line data
displays, and data acquisition were controlled on-line using REX (Real-Time
Experimentation System) (Hayes et al.
1982). When a neuron was properly isolated, its activity was
recorded on a digital audio tape together with the horizontal and vertical
positions of the right and left eyes, the velocity of the vestibular turntable
and the position of the target. Off-line analysis was performed using custom
algorithms (Matlab, The MathWorks).
The magnetic search-coil technique was utilized to record the horizontal
and vertical positions of both eyes (Fuchs
and Robinson 1966) (CNC Engineering). Each eye coil signal was
calibrated independently by having the monkey fixate monocularly (i.e., 1 eye
masked) on a variety of targets at different eccentricities and depths. Note
that only eye movements restricted to the horizontal plane are discussed in
the present report. Off-line, right eye, left eye and target position signals
were first low-pass filtered at 250 Hz (analog 8 pole Bessel filter) and
sampled at 1 kHz. Recorded eye position signals were next digitally filtered
with a 51st-order finite-impulse-response (FIR) filter with a Hamming window,
using a cut-off at 125 Hz. The position signals were differentiated to produce
eye velocity profiles. Zero-phase forward and reverse digital filtering
prevented phase distortion.
Extracellular single-unit activity was recorded using enamel-insulated
tungsten microelectrodes (7- to 10-M
impedance, Frederick Haer). A
neuron was considered to be adequately isolated only when, on playback,
individual action potential waveforms could be discriminated during saccades,
smooth pursuit, and fixation, using a windowing circuit (BAK) (for example,
see Fig. 1 in Sylvestre and Cullen
1999a). Neuronal discharges were represented as a spike density
function in which a Gaussian function (SD of 5 ms for saccades, and 10 ms for
smooth pursuit, cancellation of the VOR, and fixation) was convolved with the
spike train (Cullen and Guitton
1997; Cullen et al.
1996; Sylvestre and Cullen
1999a,b,
2002).
To determine the location of the nucleus prepositus, the location of the
abducens nucleus was first identified based on its stereotypical discharge
patterns during eye movements (Cullen et
al. 1993; Sylvestre and Cullen
1999a). Previous studies had shown that burst tonic neurons are
distributed between the vestibular nuclei and nucleus prepositus hypoglossi
(Cullen at al. 1993;
McFarland and Fuchs 1992). In
the present study, single-unit recordings were for the most part limited to a
small region of the brain stem extending 0.5–1.5 mm caudal to the
abducens nucleus and 0.5–1.5 mm lateral of the midline, corresponding to
the nucleus prepositus (Brodal
1983; McCrea et al.
1987). Reconstructions of recording locations indicated that most
neurons (85%) were located within this area. The remaining small percentage of
neurons (15%) were located in the most rostromedial aspect of the adjacent
rostral-medial vestibular nucleus. Note that in addition to these anatomical
criteria, we were careful to prevent, as much as possible, the inclusion of
abducens neurons in our sample by restricting our analysis to BT neurons for
which no BT activity could be recorded in the background (i.e., no
"beehive" sound characteristic of the abducens nucleus) (see
Robinson 1970; Sylvestre and
Cullen 1999a,
2002) or in the neuron's
immediate vicinity.
Behavioral paradigms
Monkeys were trained to fixate a target light in a dimly lit room for a
juice reward. The paradigms utilized in the present study were identical to
those used by Sylvestre and Cullen
(2002).
CONJUGATE PARADIGMS. A red HeNe laser target projected onto a
cylindrical screen located 55 cm away from the monkey's eyes (isovergent,
3.5° convergence) was utilized for all conjugate paradigms. To elicit
saccades, the target light was stepped between positions
±5–30°, in 5° increments and in predictable or
nonpredictable sequences. Fixation intervals were obtained by keeping the
target stationary for 2–3 s at each position. Smooth pursuit eye
movements were generated by moving the same laser target sinusoidally
(40°/s peak velocity, 0.5 Hz). Cancellation of the vestibuloocular reflex
(VORc) was also utilized to verify that neurons were not sensitive to
vestibular stimulation. In this paradigm, the monkey fixated a laser target
that moved with the primate chair during sinusoidal whole-body rotations
(40°/s peak velocity, 0.5 Hz).
DISJUNCTIVE PARADIGMS. We utilized an array of 16
computer-controlled red light-emitting diodes (LEDs; with intensities
comparable to that of the laser target) placed between the cylindrical screen
and the monkey's eyes to elicit vergence eye movements. First, symmetric
(pure) vergence eye movements were obtained using targets that were aligned
with the monkey's midsaggital plane (convergence angles: LEDs: 17, 12, 8, and
6° and laser: 3.5°). Second, disjunctive saccades were generated by
stepping the target from one of the 16 LEDs to one of the eccentric laser
target positions (described in the preceding text for saccades). Predictable
and nonpredictable target sequences were utilized (see
Sylvestre and Cullen 2002 for
details). This approach yielded a rich variety of disjunctive saccades with
conjugate components of amplitudes ±5–30° and vergence
components of amplitudes ±4–13°.
Data analysis
In this report, the eyes are referred to as either ipsilateral or contralateral based on their location relative to the recording site. For both the ipsilateral and contralateral eye, positive and negative values correspond to positions right and left of the mid-saggital plane, respectively. We also describe eye movements in terms of conjugate {conjugate = (left eye + right eye)/2} and vergence (vergence = left eye – right eye) coordinates.
CONJUGATE AND DISJUNCTIVE FIXATION. A neuron's sensitivity to
eye position during conjugate fixation was measured as the slope of the
relationship between the mean conjugate eye position and mean firing rate
measured during such intervals. Conjugate fixation periods (n >
40) were defined as time intervals 
200 ms in duration, during which the
vergence angle was <3.5° and the peak conjugate and vergence velocities
were <10°/s. To avoid fitting neuronal response as cells were driven
into cut-off, only data for which the firing rate was >20 spikes/s were
included in the optimization. A similar analysis was performed during
disjunctive fixation (defined as having vergence angle > 4°; n
> 40 segments) using a multiple regression model that included the mean
position of each eye. Standard statistical tests were performed on the model
parameters to determine 95% confidence intervals.
CONJUGATE SACCADES. The dynamic sensitivity of a neuron to eye
movements during conjugate saccades (n > 40) was estimated using
linear optimization techniques that have been described in detail elsewhere
(Sylvestre and Cullen 1999a).
By definition, horizontal conjugate saccades had vertical amplitudes <10%
of their horizontal amplitudes, and changes in vergence angles <2.5°.
The onset and offset of these saccades was determined using a typical
20°/s velocity criterion. The specific model structures utilized are
described in Table 2. The
goodness-of-fit to the data of each model was quantified using the
variance-accounted-for {VAF =1 – [var(mod – fr)/var(fr)], where
mod represents the modeled firing rate and fr represents the actual firing
rate}. The VAF is equivalent to the R2 coefficient for
linear models and can be easily utilized to evaluate the goodness-of-fit of
model predictions. A Bayesian Information Criteria (BIC) was also computed for
each model (Cullen et al.
1996). This criteria serves as a "cost index" that
indicates whether increasing the complexity of the model is justified by the
accompanying increase in VAF (Schwartz
1978). The dynamic lead time of individual neurons
(td) was determined during conjugate saccades as described
in Sylvestre and Cullen
(1999a).
|
DISJUNCTIVE SACCADES. The activity of BT neurons during
disjunctive saccades was first quantified using model Est-all, and
then using model Est-cv (see RESULTS). The analysis of
neuronal responses was limited to the saccadic interval. Note that 80% of
the vergence shift is executed during this interval for disjunctive saccades
like those utilized here, for which the conjugate component is larger than the
vergence component (see Maxwell and King
1992). Specifically, the analysis utilized disjunctive saccades
during which both eyes moved in the same direction to limit the analysis to
ON-direction responses only, one eye moved more than the other to
generate vergence velocities >100°/s (mean intra-saccadic vergence
shift: 6.5 ± 1.1°), and the onset and offset were marked using a
20°/s conjugate velocity criterion. Similar to our previous study of ABN
neurons (Sylvestre and Cullen
2002), we estimated the probability distribution of the model
parameters in Est-all (or Est-cv) using a nonparametric
bootstrap approach (Carpenter and Bithell
2000; Press et al.
1997; Richmond et al.
1987; Sokal and Rohlf
1995). Briefly, for each neuron 1999 "new data sets"
of n > 40 saccades were obtained by randomly re-sampling with
replacement from an original data set (with N/2 converging saccades
and N/2 diverging saccades; such balance is important to avoid
biasing the parameter estimates). The model parameters were then estimated on
each of the new data sets. The 1999 parameter values obtained with this
approach were then utilized to compute 95% confidence intervals (BCa method)
(Carpenter and Bithell 2000).
Parameters with 95% confidence intervals that overlapped with zero were not
statistically significant, and parameters with 95% confidence intervals that
overlapped with one another were statistically identical. To prevent the
biasing of meaningful parameter values due to the inclusion of inappropriate
parameters in the original model, these inappropriate parameters were removed
(if nonsignificant) or replaced (if identical) one at a time from the original
model, and the parameters of the reduced model were estimated after each
removal.
For each neuron, Ratiofix (for the eye position sensitivities during fixation), Ratiopos (for the eye position sensitivities during disjunctive saccades) and Ratiovel (for the eye velocity sensitivities during disjunctive saccades) indexes were computed using the reduced binocular model parameters (Est-red, see RESULTS) to quantify a unit's relative preference for one eye versus the other. For each sensitivity, a ratio index was computed using Ratio = [smaller parameter value]/[larger parameter value], where the smaller and larger parameter values are yielded by the nonpreferred and preferred eyes, respectively. To indicate which eye provided the larger parameter value (i.e., the neuron's "preferred eye"), each Ratio index was assigned an "i" or a "c", for the ipsilateral or contralateral eye, respectively. Using their ratio indexes, neurons were assigned to one of five categories based the criteria described in Table 1.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
;
McFarland and Fuchs 1992;
Scudder and Fuchs 1992). In
the following sections, we first describe the discharge properties of the
recorded neurons during conjugate and disjunctive fixation. We next provide
the first quantitative description of the discharge dynamics of BT neurons
during conjugate saccades. After this characterization, we demonstrate our
analysis approach of the discharge dynamics of the same neurons during
disjunctive saccades using two example neurons. Finally, we present the
distribution of discharge properties measured during disjunctive saccades
across our sample of neurons. Conjugate and disjunctive fixation
The eye-position sensitivities of BT neurons were first evaluated during
conjugate fixation intervals. In agreement with previous studies
(Cullen et al. 1993;
McFarland and Fuchs 1992), all
neurons significantly encoded the position of the eyes in the orbit. The
firing rate of two example units (units B61_2 and B43_1)
during intervals of conjugate fixation are shown in
Fig. 2, A and
B ( 
). Eye position traces are illustrated below (see
legend). Each neuron's eye position sensitivity was quantified using a
standard model
(
,
where bfix and kfix are the bias and
eye position sensitivity coefficients, respectively, and 
and 
are mean firing rate and conjugate eye position,
respectively). This simple model provided good fits to the data (VAF = 0.71
± 0.20, mean ± SD). The average bfix for our
sample was 66 ± 48 spikes/s, and the average kfix
was 3.2 ± 2.0 spikes · s–1 ·
°–1.
|
During disjunctive fixation, most units' firing rates were modulated differentially by the position of each eye. This is clearly seen in Fig. 2C, where unit B61_2's discharges were not modulated when the position of the ipsilateral eye (re the recording site) was kept relatively constant but the position of the contralateral eye was changed markedly. Hence this unit appeared to only encode the position of the ipsilateral eye. In contrast, other neurons' firing rates, like those of unit B43_1, were modulated by the position of both eyes (Fig. 2D). Clearly, this neuron's discharges were modulated even when the position of one of the eyes, or that of the other eye, was kept constant. Because this neuron obviously encoded the position of both eyes, possibly with different sensitivities, a more detailed quantitative analysis was warranted. In this case, as shown below, the quantitative analysis of this neuron's discharges has revealed that it was equally sensitive to the position of both eyes (i.e., it encoded the conjugate position of the eyes).
We quantified this observation using a rich data set of disjunctive
fixation intervals (see METHODS) and a second model that included
separate terms to estimate a neuron's sensitivity to the position of each eye
(
,
where IE and CE represent the ipsilateral and contralateral eyes,
respectively). This model also provided good fits to the data for our sample
of neurons (mean VAF = 0.72 ± 0.19). For the example neuron shown in
Fig. 2, A and
C, the model parameter ki-fix was
3.47 spikes · s1 ·
°–1 and the parameter
kc-fix was not significantly different from zero
(P > 0.05), indicating that this neuron only encoded the position
of the ipsilateral eye. In contrast, for the example neuron shown in
Fig. 2, B and
D, the model parameters ki-fix and
kc-fix were statistically identical (1.6 spikes ·
s1 · °–1; P >
0.05), suggesting that this neuron equally encoded the position of both eyes
(i.e., encoded the conjugate position of the eyes). Hence the analysis results
confirmed the qualitative observations made in the preceding text for both
neurons.
On average, across our sample of neurons, k-i-fix was 0.9 ± 2.9 spikes · s1 · °–1, and kc-fix was 1.8 ± 2.4 spikes · s1 · °–1. To better describe the signals encoded by BT neurons during disjunctive fixation at the sample level, we quantified the relative preference of a neuron for one eye versus the other by computing a Ratiofix index for (see METHODS). This index was used to assign each neuron in our sample to one of five categories based the criteria described in Table 1: a neuron could be monocular with a preference for the ipsilateral eye, monocular with a preference for the contralateral eye, unequal binocular with a preference for the ipsilateral eye, unequal binocular with a preference for the contralateral eye, or conjugate (i.e., equal binocular sensitivities). Figure 3 shows the distribution of BT neurons, with respect to their eye position sensitivities during disjunctive fixation, obtained using this classification. Note that each category was assigned a different color (see legend). As shown in Fig. 3, an important fraction of BT neurons (40%) were monocular (gray bars), with roughly equal number of neurons preferring the ipsilateral (pale gray bar) or the contralateral (dark gray bar) eye. Most of the remaining neurons unequally encoded the binocular position of the eyes (stripped bars), with a slight preference for the contralateral eye. Only few neurons (20%) equally encoded the position of both eyes (i.e., were conjugate, black bar). We conclude that most BT neurons (80%) differently encode right and left eye position signals during disjunctive fixation.
|
These results can also be interpreted in the conjugate/vergence coordinate
system. From the definitions of conjugate and vergence (see
METHODS), it can be derived that
 |
It follows that only those neurons that equally encode the position of the right and left eyes (e.g., unit B43_1, where kleft eye = kright eye) have no vergence sensitivities. Because only 20% of the neurons in our sample matched this criteria, we conclude that the majority (i.e., 80%) of BT neurons in our sample encoded both vergence and conjugate position signals during disjunctive fixation.
Conjugate saccades
Before quantifying the discharge patterns of BT neurons during disjunctive saccades, we determined, for the first time, what signals are carried by these neurons during the simpler situation of conjugate saccades (where both eyes rotate in the same direction, by the same amplitude, and with very similar dynamics). To do so, the goodness-of-fit of the different eye movement-based models shown in Table 2 were compared to determine which one provided the most appropriate description of BT neuron discharges during these eye movements. Note that we tested an extensive series of model structures on our sample of BT neurons but that for the sake of simplicity, only the most relevant models are reported in Table 2. Model fits obtained with three of the models in Table 2 are illustrated in Fig. 4 for two example BT neurons, units B87_1 (A) and B43_1 (B). For each neuron, three conjugate saccades of increasing amplitudes are shown. For clarity, the neurons' firing rates during these saccades are replicated three times (gray shaded curves, top 3 rows), and a different model fit (thick black curve) is shown in each row. Note that both units had clear bursts during conjugate saccades.
|
The first model we tested was solely based on a bias term (the firing rate when the monkey was fixating straight ahead) and the conjugate position of the eyes in the orbit (model M1, Table 2). This model provided a poor fit to the data (see top row, Fig. 4; mean VAF = 0.24 ± 0.25, Table 2). A second model, which was based on a bias term and the conjugate velocity of the eyes, only marginally improved our ability to describe the neuronal discharges (model M2; mean VAF = 0.28 ± 0.26, Table 2; not shown in Fig. 4).
The simplest model that provided a good description of BT neuron discharges
during conjugate saccades contained both eye position- and
eye-velocity-related terms (model M3, Table
2; 2nd row, Fig.
4). The average VAF values generated with this model were
virtually twice larger than those obtained with the two simpler models (133
and 100% relative improvements, model M1 and M2, respectively). This striking
increase in goodness-of-fit can be visualized by comparing the model fits
shown in the first and second rows of
Fig. 4. The average b
coefficient for this model was 143 ± 82 spikes/s, the average
k was 3.9 ± 2.5 spikes · s1 ·
°–1, and the average r was 0.34
± 0.31 spikes/°. When compared with fixation, the bias coefficients
were significantly larger during saccades (P < 0.01, paired
t-test), but the k coefficients were not statistically
different (P > 0.05). Models that included higher-order
derivatives of eye position (e.g., eye acceleration and jerk) generally did
not better describe the activity of BT neurons than model M3 (not shown). One
exception was when a term proportional to the derivative of the firing rate (a
"slide" term) was added to model M3 together with an eye
acceleration term (model M4, Table
2; 3rd row, Fig.
4); these terms provided an important increase in VAF across our
sample of neurons (21% mean improvement relative to model M3). On average, the
u coefficient (0.002 ± 0.005 spikes ·
s–1 · °–1)
and the c coefficients (0.018 ± 0.018) were small. Hence, as
for neurons in the abducens nucleus
(Sylvestre and Cullen 1999a),
we conclude that a first order model of eye position (model M3) is the
simplest model we can utilize to describe BT neuron discharges during
conjugate saccades.
Disjunctive saccades: example neurons
The approach that we utilized to characterize the activity of BT neurons
during disjunctive saccades was identical to that described in Sylvestre and
Cullen (2002). First, for each
neuron, we determined whether a model estimated during conjugate saccades
(model M3, Table 2) could
predict its activity during disjunctive saccades. Second, we estimated the
parameters of the following model on the data set of disjunctive saccades
collected for that neuron
 |
and
are instantaneous
ipsilateral and contralateral eye positions and velocities, respectively. This
model is the binocular expansion of model M3. Model M3, rather than model M4,
was chosen for this analysis to limit the number of free parameters in model
Est-all.
For each parameter in model Est-all, bootstrap confidence intervals (see
METHODS) were utilized to reduce the model to its simplest form
(reduced model labeled Est-red; can be different for each neuron). This
approach was necessary given that some basic assumptions (e.g., normally
distributed residuals) inherent to standard statistical tests on linear
regression parameters were invalid (see
Sylvestre and Cullen 2002 for
more details). The bootstrap approach utilized here provided an alternate
objective technique to determine which model parameters significantly
contributed to the goodness-of-fit of the model to the neuronal firing
rates.
Example monocular BT neuron, unit B87_1
Figure 5 shows the results of our analysis of disjunctive saccades for unit B87_1, the same unit that was shown in Fig. 4A during conjugate saccades. The unit's discharge during converging (Fig. 5A) and diverging (Fig. 5B) saccades are shown. The model predictions, shown in the top row (Pred-CS) and generated using the parameters of model M3 (Table 2) and the conjugate traces shown at the bottom of Fig. 5, clearly could not describe the activity of this neuron during disjunctive saccades. This prediction-based analysis hence suggested that unit B87_1 did not encode conjugate signals during these movements.
|
The black model fits in the second row of Fig. 5 were obtained by estimating the parameters of model Est-all on the data set of disjunctive saccades gathered for this neuron. In contrast to the model predictions, this estimated model could adequately describe the neuronal activity during both converging and diverging saccades. To determine whether all model parameters were significant, the bootstrap confidence intervals illustrated in Fig. 6 were analyzed. As shown in Fig. 6, both the contralateral eye position (kc-DS) and the contralateral eye velocity terms (rc-DS) in model Est-all had confidence intervals that markedly overlapped with zero (white horizontal bars, Fig. 6). This result indicates that these parameters did not play a significant role in describing the activity of unit B87_1 during disjunctive saccades (i.e., this unit was monocular with a preference for the ipsilateral eye).
|
To further validate this conclusion, we removed both terms related to the
movements of the contralateral eye from model Est-all and estimated the
parameter values of the reduced model [model Est-red, see
Fig. 5; for this neuron,
].
The obtained model fits were quantitatively as good as those obtained with
model Est-all (i.e., VAFEst-all = VAFEst-red = 0.78).
This is graphically shown in the second row of
Fig. 5, where the gray model
fit (model Est-red) is perfectly superimposed on the black model fit (model
Est-all). In summary, both the prediction-based and the estimation /
bootstrap-based analyzes indicated that unit B87_1 did not encode
conjugate signals during disjunctive saccades. Rather it solely encoded the
position and the velocity of the eye ipsilateral to the recording site.
Example conjugate BT neuron, unit B43_1
Figures 7 and 8 show the results of this analysis approach when applied to conjugate unit B43_1 (also shown in Fig. 4B during conjugate saccades). One obvious difference with the data shown for unit B87_1 was that the predictions computed using model M3 were very good for both converging and diverging saccades (Fig. 7, top, A and B, respectively). This observation strongly suggested that unit B43_1 encoded conjugate eye position and velocity signals during all saccades.
|
|
This conclusion was strengthened by the estimation-based analysis. As is
shown in Fig. 7, model Est-all
provided a very good fit to the data (2nd row, black model fits).
Furthermore, Fig. 8 clearly
shows that for both the eye position (kc-DS and
ki-DS) and the eye velocity (rc-DS and
ri-DS) related terms in model Est-all, the parameter
values for the ipsilateral and contralateral eyes were statistically identical
(i.e., there was extensive overlap of the confidence intervals). Also note
that no term had its confidence interval overlapping with zero (i.e., all
terms were significant). These results confirmed that unit B43_1
encoded conjugate eye-position and -velocity signals. Indeed, when the
parameters of model Est-all were replaced by conjugate parameters [model
Est-red, see Fig. 7; for this
neuron,
],
the obtained model fits were identical to those of the full binocular model
(2nd row, gray model fit, Fig.
7), and so were the VAF values. In summary, our analysis indicated
that unit B43_1 was equally sensitive to the position and velocity of
both eyes.
Disjunctive saccades: population results
The population signals carried by BT neurons during disjunctive saccades were quantified using an approach identical to that utilized during disjunctive fixation. For a given neuron, the eye position coefficients estimated in model Est-red for each eye were utilized to calculate a Ratiopos index, and the eye-velocity coefficients to calculate a Ratiovel index. As for fixation, these indexes were used to assign individual neurons to one of the five categories of ocular preferences described in Table 1. With respect to the eye-position sensitivity of BT neurons during disjunctive saccades, most units (70%) encoded the monocular position of one of the eyes (gray bars, Fig. 9A). Of these neurons, 50% encoded the position of each eye (pale and dark gray bars, for the ipsilateral and contralateral eyes, respectively). The remaining 30% of BT neurons encoded the conjugate position of the eyes (black bar). Note that in contrast to disjunctive fixation, no neuron unequally encoded binocular signals. A generally similar distribution was observed for the eye velocity sensitivity of BT neurons (Fig. 9B). Overall, many units (35%) encoded the velocity of the ipsilateral eye only, and 20% encoded the conjugate velocity of the eyes. Only 25% of the units unequally encoded binocular signals (stripped bars).
|
Because each neuron in our sample had a "preferred eye"
(defined as the eye that yielded the largest parameter value) for its position
and for its velocity sensitivities, we next asked whether these preferred eyes
were matched on a neuron-by-neuron basis. To simplify this analysis, the
neurons (in the preceding text separated in 5 categories) were regrouped under
three general categories: ipsilateral eye preference category
(grouping the monocular with ipsilateral eye preference and the unequal
binocular with ipsilateral eye preference cell types), contralateral eye
preference category (grouping the monocular with contralateral eye
preference and unequal binocular with contralateral eye preference cell
types), or conjugate category. The fraction of neurons that fell
within each of the nine possible permutations among these three categories are
illustrated in Fig.
10A for the eye-position and -velocity sensitivities
estimated during disjunctive saccades. In general, for 55% of the neurons in
our sample, there was coherence between their preferred eye for the two
sensitivities (Fig.
10A, 
). For the remaining neurons that did not
exhibit coherence of their preferred eye, no specific pattern could be
identified (Fig.
10A, ). Hence, we conclude that a majority of
individual BT neurons encoded the position and the velocity of the same eye
during disjunctive saccades and that no trend was identified for the neurons
that did not exhibit such coherence.
|
Disjunctive saccades versus disjunctive fixation
A similar approach was utilized to compare the "preferred" eye
position sensitivities of individual neurons during disjunctive fixation and
disjunctive saccades. A neuron's preferred eye for the position sensitivity
during disjunctive saccades was the same as during disjunctive fixation for
the majority of units (75% coherence; Fig.
10B, ). With respect to the noncoherent units
(Fig. 10B, ),
they were fairly uniformly distributed. However, it should be noted that even
if they preferred the same eye, slightly more units encoded unequal binocular
eye position signals during disjunctive fixation (40 vs. 0%), while a
comparable number of units encoded the conjugate position of the eyes during
both behaviors (20 vs. 20%). Hence, BT neurons generally encoded the position
of the same eye during disjunctive fixation and disjunctive saccades but
sometimes encoded the position of the nonpreferred eye with different
strengths across these behaviors.
Disjunctive saccades: alternative analysis model
Finally, for the sake of completeness, we also analyzed our sample of BT
neurons using a model based on conjugate/vergence coordinates
 |
).
Hence, we properly evaluated the parameters of model Est-cv on all the neurons
in our sample by independently computing bootstrap confidence intervals for
each parameter. Similar to our findings with abducens neurons, we occasionally
obtained slightly different results using models Est-all and Est-cv. However,
for 75% of the neurons in our sample, the reduced models obtained with model
Est-cv provided worse or equivalent goodness-of-fits as those obtained with
model Est-all (mean difference in VAF = –0.04 ± 0.07); for the
remaining units, the VAF values for the reduced conjugate/vergence-based
models were only 0.01 ± 0.003 larger than for the reduced binocular
models. Hence as for abducens neurons
(Sylvestre and Cullen 2002),
we conclude that the binocular approach (i.e., using model Est-all) is more
appropriate to describe the modulation of BT neurons during disjunctive
saccades. |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
Conjugate fixation and general response characteristics
During conjugate fixation, 90% of the BT neurons in our sample increased
their discharges with increasingly ipsilateral eye positions, which compares
well with the previous findings (86%) of McFarland and Fuchs
(1992). Moreover, the average
bias and eye position sensitivities reported here during fixation
(bfix =66 ± 48 spikes/s and
kfix =3.2 ± 2.0 spikes ·
s–1 · °–1,
respectively), were almost identical to those reported in previous reports
[mean kfix = 3.2 ± 1.3 spikes ·
s–1 · °–1
(McFarland and Fuchs 1992);
mean kfix = 3.6 spikes ·
s–1 · °–1,
(Scudder and Fuchs 1992)].
With respect to conjugate saccadic behavior, 95% of the units described in
this report had BT discharge properties, and a single neuron had tonic
discharge properties. Previous studies that recorded from nonvestibular
neurons in the MVN/NPH have reported slightly lower proportions of BT neurons
(73% across NPH/MVN, McFarland and Fuchs
1992; 53% only in MVN, Scudder
and Fuchs 1992). This is not surprising, given that for the sake
of the analysis utilized in this report, we preferentially recorded from
neurons that exhibited significant saccade-related behaviors.
In the monkey, many if not most BT neurons provide significant premotor
drives to the abducens nucleus. Both spike-triggered averaging
(Scudder and Fuchs 1992) and
intracellular recording/staining (McCrea
et al. 1987) experiments have identified neurons with BT discharge
characteristics that project to the abducens nucleus. In fact, Scudder and
Fuchs (1992) have noted that
more BT than tonic-only neurons project to the abducens nucleus. Moreover,
McFarland and Fuchs (1992)
have found that 91% of neurons in the marginal zone, where the highest NPH/MVN
projections to the abducens are located
(Langer et al. 1986), have
burst-tonic discharge characteristics. In this study, we found that, on
average, BT neurons had a dynamic lead time that was comparable to that
previously reported for premotor medium-lead inhibitory burst neurons (12.2
± 5.3 vs. 11.8 ± 2.7 ms, for BT vs. inhibitory burst neurons)
(Cullen and Guitton 1997).
Abducens nucleus neurons, on average, had shorter dynamic lead times (9.4
± 1.9 ms) (Sylvestre and Cullen
1999a). Thus our findings are consistent with previous evidence
showing that in the monkey, BT neurons are output neurons of the NI and are
therefore premotor neurons.
Signals encoded by BT neurons during conjugate saccades
In the monkey, it has been shown that BT neurons discharge bursts of action
potentials during saccades whose durations and number of spikes are well
correlated with the duration and amplitude of the concurring saccade,
respectively (McFarland and Fuchs
1992). Our analysis confirmed and extended these findings by
providing the first detailed description of the relationship between the
dynamics of the saccadic burst of BT neurons and the ongoing eye movement. We
found that, as for ABN neurons (Sylvestre
and Cullen 1999a), both the instantaneous eye position
and instantaneous eye velocity are required to describe BT neuron
firing rate dynamics during conjugate saccades (see model M3,
Table 2). Moreover, the average
parameter values for this model were generally similar for BT neurons and ABN
neurons (b: 143 ± 82 vs. 156 ± 89 spikes/s; k:
3.9 ± 2.5 vs. 4.2 ± 2.3 spikes ·
s–1 · °–1;
r: 0.34 ± 0.31 vs. 0.42 ± 0.26 sp/°; BT vs. ABN
neurons, respectively). Interestingly, the largest difference in parameter
values was for the eye-velocity-sensitivity parameters, which were on average
smaller for BT neurons than ABN neurons. These findings are consistent, during
saccades, with BT neurons being a primary source of eye-position-related
signals to the abducens nucleus, and with saccadic excitatory burst neurons
combining their eye velocity-related drives with those of BT neurons at the
level of the abducens nucleus.
We also found that BT neuron discharges carry a "slide" term
[i.e., 
, model M4 in
Table 2] during saccades that
is similar to that of ABN neurons (0.018 ± 0.018 vs. 0.015 ±
0.014, BT neurons vs. ABN neurons, respectively) (Sylvestre and Cullen 1999).
Postsaccadic slide terms have been described previously for BT neurons
(Lopez-Barneo et al. 1982;
McFarland and Fuchs 1992).
This exponentially decaying term is generally believed to offset the restoring
forces of the oculomotor plant following a saccade
(Goldstein 1983;
Robinson 1964;
Sylvestre and Cullen 1999a).
Aksay et al. (2001) have
further shown that BT neurons, in the goldfish, have the intrinsic properties
required to generate a "slide" term. Using in vivo intracellular
recordings in alert goldfishes, they have shown that area I neurons (the fish
equivalent of NPH/MVN) generate a burst of action potentials that is quickly
followed by an exponential decay in firing rate in response to an injected
depolarizing pulse. This decay has a time constant (50 ms) that is
similar to that observed on abducens nucleus neurons following saccades
(Goldstein 1983;
Sylvestre and Cullen 1999a).
Taken together, these results are consistent with the proposal that BT neurons
contribute to generating the "slide" signal present on ABNs during
and following saccades.
Signals encoded by BT neurons during disjunctive fixation
During disjunctive fixation, we have found that only 20% of BT neurons
encoded the conjugate position of the eyes (i.e., were equally sensitive to
the position of both eyes; Fig.
3), whereas 40% encoded the position of a single eye. To date, the
only other study that has quantitatively characterized the activity of BT
neurons during disjunctive eye movements is the study by Zhou and King
(1996). This study, however,
focused on disjunctive smooth pursuit, and there is no a priori reason to
assume that BT neurons should generate similar responses during both eye
movements. Nevertheless, there are important similarities between their
results and ours. For example, during monocular smooth pursuit, most NPH
neurons (70%) are monocular (Zhou and
King 1996). The general tendency to observe very few conjugate
neurons was therefore common to both samples. However, all the monocular
neurons recorded during smooth pursuit preferred the ipsilateral eye, while we
recorded equal numbers of neurons encoding the position of each eye during
fixation. These latter results are more consistent with the results obtained
for the only other premotor neurons studied quantitatively during disjunctive
fixation, position-vestibular-pause (PVP) neurons for which 50% encoded the
position of either eye (McConville et al.
1994). Further studies will be required to determine whether the
results obtained during disjunctive smooth pursuit reflect the likely distinct
properties of smooth pursuit premotor pathways or whether they are simply due
to sampling biases.
Our result that most BT neurons did not encode conjugate signals during
disjunctive fixation suggests a potentially important role for these neurons
in generating unequal movements of the eyes (i.e., vergence). To determine the
functional relevance of the vergence-related terms carried by BT neurons
during disjunctive fixation, we decided to compare the population results
obtained for BT neurons during disjunctive fixation to those obtained for ABN
neurons during a similar paradigm
(Sylvestre and Cullen 2002).
This comparison addresses whether BT neurons carry sufficient
eye-position-related information to shape the activity of ABN neurons during
disjunctive fixation or whether there is a need for additional
vergence-related inputs. Strikingly, as shown in
Fig. 11A, both
population distributions were fairly similar. The most notable difference was
that fewer BT neurons than ABN neurons encoded the position of a single eye.
One explanation for this small discrepancy is that during disjunctive
fixation, other premotor neurons such as PVP neurons were shown to send
monocular eye-position-related drives to the abducens nucleus
(McConville et al. 1994).
Moreover, eye-head neurons, which are also thought to be premotor, have been
qualitatively described as preferentially encoding the position of a single
eye (Chen-Huang and McCrea
1999). Hence, the summation of the drives carried by BT neurons,
PVP neurons, and possibly eye-head neurons, at the level of ABN neurons would
account for the signals carried by the latter neurons during disjunctive
fixation.
