|
|
||||||||
Aerospace Medical Research Unit, Department of Physiology, McGill University, Montreal, Quebec H3G 1Y6, Canada
Submitted 27 November 2002; accepted in final form 6 March 2003
| ABSTRACT |
|---|
|
|
|---|
| INTRODUCTION |
|---|
|
|
|---|
The overall goal of the present study was to determine what signals are
carried by the brain stem premotor pursuit pathway during head-restrained and
combined eye-head pursuit. The brain stem neurons in the rostral-medial and
ventral-lateral vestibular nuclei, which receive direct projections from the
floccular lobe, have been termed flocculus target neurons (FTN)
(Broussard and Lisberger 1992
;
Lisberger and Pavelko 1988
;
Lisberger et al.
1994a
,b
).
The responses of these brain stem neurons largely correspond with those of a
distinct physiological subclass of cells, termed eye-head (EH) neurons, which
have been well characterized during eye and head movements in the
head-restrained monkey (Chen-Huang and
McCrea 1999
; Cullen at al.
1993
; Gdowski and McCrea
1999
,
2000
;
Gdowski et al. 2001
;
McCrea et al. 1996
;
McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
;
Tomlinson and Robinson 1984
).
Accordingly, for the sake of simplicity, both FTN and EH neurons will be
referred to from here on as EH neurons (although these 2 populations of
neurons may not be strictly equivalent). EH neurons are thought to be the most
significant premotor input to the extraocular motoneurons of the abducens
nucleus during smooth pursuit eye movements
(Cullen at al. 1993
; Lisberger
et al.
1994a
,b
;
McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
).
To date, much is known about the signals carried by cerebellar neurons that
project to EH neurons (i.e., Purkinje cells of the floccular lobe) during
head-restrained smooth pursuit. The discharges of these cells can be
considered with respect to their two sources of input, namely climbing fibers
and mossy fibers. Inputs from the climbing fibers, which originate from the
inferior olive (Eccles et al. 1966; Thach 1967), result in the complex spikes
in the Purkinje cells (for review, see
Bloedel and Courville 1981
).
The role of the climbing fiber input is still unclear (reviewed in
Simpson et al. 1996
), but many
assume that it functions to modify the efficacy of the synapse between mossy
fibers and Purkinje cell. Moreover there is evidence that in rabbit
(Frens et al. 2001
;
Graf et al. 1988
) and monkey
(Kahlon and Lisberger 2000
;
Stone and Lisberger 1990b
)
complex spike trains encode performance errors (i.e., retinal slip).
Mossy fiber inputs are responsible for the simple spike activity of
Purkinje cells (for review, see Stone and
Lisberger 1990a
). Because simple spikes occur much more frequently
(discharges rates reaching
300 spikes/s) as compared with complex spike
activity (
1 spikes/s), it seems likely that the information carried to
the brain stem by simple spikes would govern, for the most part, the response
profiles of EH neurons. During smooth pursuit, simple spike trains, in
contrast to complex spikes, chiefly encode eye position, eye velocity, and, to
a much smaller degree, eye acceleration
(Leung et al. 2000
;
Suh et al. 2000
). Whether
simple spike trains also encode visual error signals remains controversial. On
the one hand, it has been proposed that simple spike activity in monkey can
encode relatively small, albeit significant, retinal velocity and acceleration
error information during sinusoidal optokinetic stimulation
(Hirata and Highstein 2001
) as
well as velocity error information when unpredictable changes in target
direction are applied during pursuit (Suh
et al. 2000
). On the other hand, Kahlon and Lisberger
(2000
) have suggested that
transient responses of simple spikes during pursuit initiation reflect the
influence of feed-forward image motion.
It is not yet known if EH neuron responses encode error-related information
(position, velocity, and acceleration) or eye-acceleration signals during
smooth pursuit. For example, EH neurons could potentially receive visual error
signals from either Purkinje cells within the floccular lobe or via direct
projections from midbrain structures such as the accessory optic system and/or
the nucleus of the optic tract that encode visual-slip information
(Kato et al. 1995
;
Wylie and Linkenhoker 1996
).
Prior characterizations of brain stem EH neurons during head-restrained
pursuit have focused on only the eye position and eye-velocity-related
response of these neurons during sinusoidal smooth pursuit
(Cullen et al. 1993
; Lisberger
et al.
1994a
,b
;
McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
).
Hence, the first specific goal of this study was to determine which
eye-movement-based and/or error-based model best describes the discharge
dynamics of EH neurons during smooth-pursuit eye movements made in the
head-restrained condition.
In the head-unrestrained condition, when coordinated eye and head movements
are made to pursue a target, at least three inputs could function to modify
the responses of EH neurons. First, as noted in the preceding text, floccular
lobe Purkinje cells send inhibitory projections to EH neurons. In the rhesus
monkey, Purkinje cells encode head velocity during passive
whole-body rotation where the monkey
"cancels" its VOR by tracking a target that moves with
the head (pWBRc) (Fukushima et al.
1999
; Kahlon and Lisberger
2000
; Lisberger and Fuchs
1978
; Miles et al.
1980
; Stone and Lisberger
1990a
). These neurons have been termed gaze velocity Purkinje
cells because they respond similarly to changes in the axis of gaze relative
to space during pWBRc and head-restrained smooth pursuit. Accordingly, it has
been proposed that they would send a gaze motor command rather than an eye
motor command to the brain stem during combined eye-head gaze pursuit
(Barnes 1993
). In contrast,
most floccular lobe Purkinje cells and presumably EH neurons in squirrel
monkey, encode eye-rather than gaze-related signals during smooth pursuit,
pWBRc, and head-unrestrained pursuit (Belton and McCrea
1999
,
2000b
). Thus it appears that
rhesus monkeys have a much greater proportion of gaze velocity Purkinje cells
as compared with squirrel monkeys
(Fukushima et al. 1999
;
Lisberger and Fuchs 1978
;
Miles et al. 1980
; Stone and
Lisberger 1990). To date, no study has recorded the responses of rhesus
floccular lobe Purkinje cells or EH neurons (in either species) during
coordinated eye-head gaze pursuit. Accordingly, the second specific goal of
the present study was to determine whether EH neurons encode an eye- or
gaze-related motor command during coordinated eye-head gaze pursuit.
In addition to the floccular lobe projection, vestibular and proprioceptive
inputs to the vestibular nuclei could further modify the responses of EH
neurons during the head movements made during gaze pursuit. EH neurons are
known to receive direct monosynaptic projections from the ipsilateral
vestibular nerve (Broussard and Lisberger
1992
; Chen-Huang and McCrea
1999
; Gdowski and McCrea
1999
,
2000
;
Scudder and Fuchs 1992
) and
polysynaptic projections from the contralateral vestibular nerve
(Broussard and Lisberger
1992
). Indeed, most EH neurons carry head-velocity-related signals
during passive whole-body rotations in the dark. Moreover, neurons in regions
of the vestibular nuclei that contain EH neurons receive inputs from neck
muscle proprioceptors via a disynaptic pathway
(Sato et al. 1997
). The active
head movements made during gaze pursuit would also activate neck
proprioceptors and could in turn modulate EH neuron responses.
Prior work in decerebrate and/or anesthetized cat has demonstrated that
passive activation of neck muscle proprioceptors can influence responses of
neurons in the medial vestibular nuclei (Anastasopoulos and Megner 1982;
Boyle and Pompeiano 1981
;
Wilson et al. 1990
). Studies
in alert squirrel monkey have suggested that most second-order neurons in the
medial vestibular nuclei including EH neurons are influenced by passive
activation of neck proprioceptors (Gdowski and McCrea
1999
,
2000
;
Gdowski et al. 2001
). In
contrast, our recent studies in rhesus monkey have found no evidence that
second-order neurons within the medial vestibular nuclei are influenced by
neck proprioceptive inputs (Roy and Cullen
2001
,
2002
). These latter studies
focused on two distinct classes of neurons in the medial vestibular nuclei,
namely position-vestibular-pause and vestibular-only neurons, and did not
consider EH neurons. Thus the third specific goal of the present study was to
address whether and how vestibular and/or neck proprioceptors inputs influence
EH neuron responses in rhesus monkey during the head-on neck movements made
during gaze pursuit.
| METHODS |
|---|
|
|
|---|
Surgical procedures
The surgical techniques were similar to those previously described by Roy
and Cullen (2001
,
2002
). Briefly, an 18- to
19-mm diam eye coil (3 loops of Teflon-coated stainless steel wire) was
implanted on the right eye behind the conjunctiva. In addition, a dental
acrylic implant was fastened to each animal's skull using stainless steel
screws. The implant held in place a stainless steel post that was used to
restrain the animal's head and a stainless steel recording chamber that was
positioned to access the medial vestibular nucleus (posterior and lateral
angles of 30°). During the surgery, isoflurane gas was utilized to
initiate (23%) and maintain (0.81.5%) anesthesia. After the
surgery, buprenorphine (0.01 mg/kg im) was utilized for postoperative
analgesia, and monkeys were allowed to recover for 2 wk before commencing
experimental sessions.
Data acquisition
During each experiment, the monkey sat comfortably in a primate chair,
which was placed on a vestibular turntable. With the monkey initially
head-restrained, extracellular single-unit activity was recorded using
enamel-insulated tungsten microelectrodes (710 M
impedance,
Frederick-Haer) as has been described elsewhere (Roy and Cullen
2001
,
2002
). To determine the
location of the medial and lateral vestibular nuclei, 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
1999
). Previous studies had shown that EH 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 0.51.25 mm caudal to the
abducens nucleus and 1.252.5 lateral of the midline, corresponding to
the rostral-medial and ventral-lateral vestibular nuclei
(McCrea et al. 1987
;
Tomlinson and Robinson 1984
).
Reconstructions of recording locations indicated that most neurons (38/42)
were located within this area. Consistent with prior studies in rhesus
(McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
), the
anatomical distribution of these cell demonstrated considerable overlap with
position-vestibular-pause and vestibular-only neurons. The remaining small
percentage of neurons (i.e., n = 4, <10%) were located in the most
lateral aspect of the adjacent nucleus prepositus hypoglossi.
Gaze and head position were measured using the magnetic search-coil
technique (Fuchs and Robinson
1966
), and turntable velocity was measured using an angular
velocity sensor (Watson). Unit activity, horizontal and vertical gaze and head
positions, target position, and table velocity were recorded on DAT tape for
later playback. Action potentials were discriminated off-line using a
windowing circuit (BAK) that was manually set to generate a pulse coincident
with the rising phase of each action potential. Gaze position, head position,
target position, and table velocity signals were low-pass filtered at 250 Hz
(8 pole Bessel filter) and sampled at 1,000 Hz.
Behavioral paradigms
For a juice reward, monkeys were trained to follow a target light (HeNe
laser) that was projected onto a cylindrical screen located 60 cm away from
the monkey's head. Target and turntable motion, and on-line data displays were
controlled by a UNIX-based real-time data-acquisition system (REX)
(Hayes et al. 1982
). The
discharges of EH neurons were first characterized during a series of
head-restrained paradigms. Neuronal responses during saccades and ocular
fixation were recorded while the monkey attended to a target that stepped
between horizontal positions over a range of ±30°. Neuronal
sensitivities to smooth pursuit eye movements were determined using two
different tasks: pursuit of sinusoidal (0.5 Hz, 80°/s peak velocity)
target motion in the horizontal plane and pursuit of step-ramp target motion
(Rashbass 1961
). In this
latter task, the monkey initially fixated a stationary target, which stepped
to an eccentric position after a random fixation period (7503,000 ms)
and then began to move at a constant velocity of either 40 or 80°/s in the
direction opposite to the step (Dubrovsky
and Cullen 2002
; Wellenius and
Cullen 2000
). A step size was chosen for each target velocity,
which provided initial smooth eye movements that were not preceded by
corrective saccades. Neuronal sensitivities to head movement during passive
whole-body rotation (0.5 Hz, 40 and 80°/s peak velocity) were tested by
rotating monkeys about an earth vertical axis in the dark (pWBRd) and while
they cancelled their VOR by fixating a target that moved with the vestibular
turntable (pWBRc).
After a neuron was fully characterized in the head-restrained condition, the monkey's head was slowly and carefully released allowing the monkey to rotate its head through the natural range of motion in the yaw (horizontal), pitch (vertical), and roll (torsional) axes. The waveform of the neuron was monitored to ensure that isolation was maintained. The response of the same neuron was then recorded during active head movements made during combined eye-head gaze shifts (1565° in amplitude) and combined eye-head gaze pursuit of a sinusoidal target (0.5 Hz, 80°/s peak velocity) and a step-ramp target moving at a constant velocity of either 40 or 80°/s.
To determine whether the activation of neck proprioceptive input modulated
neuronal discharges, EH cells were then recorded during two additional
paradigms. First, the experimenter manually rotated the monkey's head to
induce rapid motion of the head relative to a stationary body. Second, the
monkey's head was held stationary relative to the earth while its body was
passively rotated at 0.5 Hz at 40°/s peak velocity. Responses to rapid
unexpected perturbations of the head were also recorded for a subset of
neurons where either a short-duration (
40 ms), high-acceleration
(>10,000°/s2), and high-velocity (
100°/s)
perturbation was applied to the head via a precision torque motor
(Huterer and Cullen 2002
) or
the head was momentarily (
500 ms) braked using a magnetic clutch during
step-ramp pursuit (Cullen et al.
1993
).
Analysis of neuron discharges
Analyses of neuronal discharges were performed using custom algorithms
(Matlab, Mathworks). Recorded gaze- and head-position signals were digitally
low-pass filtered using a 51st-order finite-impulse-response (FIR) filter with
a Hamming window and cut-off frequency set to 125 Hz. Eye position was
calculated from the difference between gaze- and head-position signals. Gaze-,
eye-, and head-position signals were digitally differentiated to produce
velocity signals. Neuronal firing rate was represented using a spike density
function in which a Gaussian function (SD of 5 ms for saccades and gaze shifts
and 10 ms for remainder of the paradigms) was convolved with the spike train
(Cullen et al. 1996
). Saccade
and gaze shift onsets and offsets were defined using a ±20°/s gaze
velocity criterion.
To quantify a neuron's response to eye position, a regression analysis was used to determine the relationship between mean eye position and mean neuronal firing rate during periods of steady fixation. This analysis yielded a resting discharge (biasx, spikes/s) and an eye position sensitivity [kx, (spikes/s)/°]. A least-squared regression analysis was also applied to neuronal discharges during saccades, smooth pursuit, passive whole body rotation in the dark and while fixating a target that moved with the animal, passive body-under-head (BUH) rotations, passive head-on-body rotations (PHBR), gaze shifts, gaze pursuit, and high-frequency perturbations of the head applied during steady eye fixation. The model formulations used to estimate the eye-, head-, and/or neck-movement sensitivities in each condition are described in RESULTS. To avoid fitting neuronal response as cells were driven into cut-off, only data for which the firing rate was >10 spikes/s was included in the optimization. For all behavioral paradigms, except for saccades and gaze shifts, only unit data from intervals between quick phases of vestibular nystagmus and/or gaze shifts and saccades were included in the analysis. During sinusoidal passive whole body rotation paradigms, neuronal phase relative to head velocity was calculated from the estimated head velocity and acceleration sensitivities [phase = arctan (acceleration coefficient / velocity coefficient) * 180/Pi].
To quantify the ability of the linear regression analysis to model neuronal discharges, the variance-accounted-for (VAF) provided by each regression equation was determined. The VAF was computed as {1 [var(est fr)/var(fr)]}, where est represents the modeled firing rate (i.e., regression equation estimate) and fr represents the actual firing rate. The VAF provided a normalized measure of each model's goodness of fit that allowed comparisons across models and neurons. For example, a VAF of 0.5 would indicate that 50% of the variability in a unit's discharge is explained by the model, corresponding to a correlation coefficient (R) of 0.71 in a bivariate linear regression. Statistical significance was determined using paired Student's t-test.
| RESULTS |
|---|
|
|
|---|
Head-restrained characterization
EYE SENSITIVITY DURING FIXATION AND SINUSOIDAL SMOOTH PURSUIT.
Neuronal responses of EH neurons were first quantified for a standard series
of head-restrained paradigms. Figure
1A shows an example type I EH neuron during saccades and
ocular fixation. This neuron's firing rate increased for ipsilaterally
directed eye positions during intra-saccadic periods of fixation. For each
cell, the relationship between mean eye position and neuronal firing rate was
described using a regression analysis (Fig.
1A, inset). The eye-position sensitivity (slope
= kx) of our example neuron was 1.68 (spikes/s)/° and
the resting discharge rate (y intercept = biasx) was 68
spikes/s. Prior studies have shown that this relationship is only linear over
a limited range of eye positions for some EH neurons, typically spanning
<25° (McFarland and Fuchs
1992
). For such neurons, the analysis included only eye positions
within this linear range.
|
Table 1 provides the average (mean ± SD) biasx and kx for our sample of type I and II EH neurons. Recall that type I EH neurons are responsive to ipsilaterally directed eye movements and head movements during pWBRc and that type II neurons are responsive to eye and head movements in the opposite direction during these paradigms. Therefore to calculate the combined population coefficient averages for eye position, the values estimated for type II neurons were first multiplied by 1 and then averaged with the type I values. A comparable procedure was used for the calculation of average eye and head movement sensitivities across type I and II EH neurons for each of the behavioral tasks in this study.
|
To quantify each neuron's response during sinusoidal smooth pursuit, the
following model was used: FR(t) = biassp +
ksp * eye position(t) + rsp *
eye velocity(t) (pursuit model), where FR is the firing rate,
ksp is the eye-position sensitivity,
rsp is the eye-velocity sensitivity, and biassp
is the bias discharge. Overall, this model provided a good description of the
discharge activity of EH neurons during smooth pursuit (mean sample VAF = 0.40
± 0.05). For our example neuron, the estimated biassp was 65
spikes/s, the ksp was 1.1 (spikes/s)/°, and the
rsp was 1.13 (spikes/s)/(°/s) during pursuit of a
target with a peak velocity of 40°/s
(Fig. 1B). The average
biassp, ksp, and rsp for
type I and type II neurons are listed in
Table 1. For both type I and
II, the mean ksp was significantly smaller than the
kx estimated for the same neurons during fixation
(P < 0.05). Mean phase lag with respect to eye velocity was
similar for both types of neurons: 56.7 ± 20.6 and 42.8 ±
24.7°, type I and II respectively and combined mean = 52.7 ±
22.2°. Overall, our results are consistent with those of previous studies
in that neuronal eye-position sensitivities were significantly larger during
fixation than during pursuit and EH neuron firing rate consistently lagged eye
velocity during smooth pursuit (Cullen et
al. 1993
; Lisberger et al.
1994a
; McFarland and Fuchs
1992
).
HEAD-SENSITIVITY PASSIVE WHOLE-BODY ROTATION. The head-velocity sensitivity of each neuron was quantified during two passive whole-body rotation paradigms using the following model: FR(t) = bias + k * eye position(t) + g * head velocity(t) + a * head acceleration(t) (pWBR model).
First we determined each neuron's bias discharge (biaspWBRc),
sensitivity to eye position (kpWBRc), sensitivity to head
velocity (gpWBRc), and sensitivity to head acceleration
(apWBRc) during compensatory eye movements made during
WBRc. This allowed us to quantify the neuron's modulation with respect to head
rotation in the absence of eye motion. The model fit for our example neuron is
shown in Fig. 2A for
pWBRc at 40°/s peak velocity (thick solid trace, pWBRc estimate; sample
mean VAF = 0.37 ± 0.23). The example neuron had a biaspWBRc
of 74 spikes/s, a kpWBRc of 2.2 (spikes/s)/°, a
gpWBRc of 0.24 (spikes/s)/(°/s), and an
apWBRc of 0.08 (spikes/s)/(°/s2) during
this paradigm. Table 1 provides
the mean values of coefficients estimated for the entire population of type I
and II neurons. To determine if the neuronal responses to head velocity were
linear, responses elicited passive whole-body rotations of 40 and 80°/s
peak velocity were analyzed separately. We found that the estimated
coefficients were comparable in the two conditions for both type I and II EH
neurons, verifying that EH neurons respond linearly over this range
(Table 1). The average phase
lead of our sample EH neurons was 2.0 ± 5.8 and 5.0 ± 8.8°
for 40 and 80°/s peak velocity rotations, respectively. This phase lead is
comparable to that reported in previous studies that have characterized these
neurons during pWBRc (Cullen at al.
1993
; McFarland and Fuchs
1992
; Scudder and Fuchs
1992
).
|
Second, the pWBR model was used to quantify neuronal responses during pWBRd, in which a bias (biaspWBRd), eye-position sensitivity (kpWBRd), head-velocity sensitivity (gpWBRd), and head-acceleration sensitivity (apWBRd) were estimated for each neuron. Note that during pWBRd, eye and head velocities are not independentthey are equal in amplitude and opposite in direction. Accordingly these terms are redundant, and as a result, it is not possible to estimate them separately.
The example neuron had a biaspWBRd of 72 spikes/s, a kpWBRd of 1.97 (spikes/s)/°, a gpWBRd of 0.47 (spikes/s)/(°/s), and an apWBRd of 0.06 (spikes/s)/(°/s2) (Fig. 2B, pWBRd estimate; sample mean VAF = 0.33 ± 0.25). The mean parameter values estimated for our sample of neurons are listed in Table 1. All of the neurons in the population had larger head-velocity sensitivities during pWBRc as compared with pWBRd (P < 0.05), and for eight type I and five type II neurons, the sensitivities were in the opposite directions during the two paradigms. Indeed, our example type I neuron's responses increased for ipsilaterally directed head rotations during pWBRc and increased for contralaterally directed head rotations during pWBRd. Accordingly, a model based on the neuron's response during pWBRd (pWBRd model) provided a strikingly poor prediction of neuronal discharge during pWBRc (Fig. 2A, compare pWBRd prediction with pWBRc estimate). However, the bias and eye-position sensitivities (k) estimated during pWBRc, pWBRd, and sinusoidal smooth pursuit were comparable (P > 0.05).
Can a linear summation of pursuit and pWBRc sensitivities predict pWBRd responses?
Prior studies have shown that the eye- and head-velocity-related signals
generated by Purkinje cells in the floccular lobe are correlated with
smooth-pursuit eye movements and with head movements during pWBRc
(Lisberger and Fuchs 1978
;
Miles et al. 1980
;
Stone and Lisberger 1990a
).
Furthermore, because these eye- and head-velocity sensitivities are nearly
identical, it has been argued that these neurons encode the velocity of the
axis of gaze relative to space (i.e., eye velocity during smooth pursuit and
head velocity during pWBRc). Indeed, these neurons are not modulated during
vestibular stimulation when gaze is stable (i.e., pWBRd) as can be predicted
by summing their sensitivities to eye and head velocity during smooth pursuit
and pWBRc, respectively. Similarly, the eye and head sensitivities of EH
neurons are in the same direction. However, unlike gaze velocity Purkinje
neurons, the eye- and head-velocity sensitivities of EH neurons are usually
unequal with, on average, the eye favoring head by a ratio of 1.2:1. This
result is in agreement with prior studies of EH neurons
(Cullen at al. 1993
; Lisberger
et al.
1994a
,b
;
McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
).
Several prior studies (Cullen at al.
1993
; McFarland and Fuchs
1992
; Scudder and Fuchs
1992
) have investigated whether the responses of EH neurons during
pWBRd can also be predicted by adding a neuron's eye- and head-velocity
sensitivities during smooth pursuit and pWBRc, respectively. We carried out a
comparable analysis for our sample of EH cells in which the following model
was used to predict neuronal activity during pWBRd: FR(t) =
biassp + ksp * eye position(t) +
rsp * eye velocity(t) +
gpWBRc * head velocity(t) +
apWBRc * head acceleration(t) (pWBRd prediction),
where ksp, rsp, and biassp
are the eye-position sensitivity, eye-velocity sensitivity, and bias discharge
that were estimated during sinusoidal pursuit (peak velocity of 40°/s),
respectively, and gpWBRc is the head-velocity sensitivity
estimated during pWBRc (peak table of 40°/s). Recall from the preceding
text that the bias estimated during smooth pursuit (biassp) and
pWBRc (biaspWBRc) were comparable. Each neuron's response
modulation during pWBRd is compared with that predicted by summing eye- and
head-velocity sensitivities estimated during smooth pursuit and pWBRc,
respectively (Fig.
3A). Across neurons, predictions based on summing
coefficients were well correlated with coefficients estimated during pWBRd
(R2 = 0.64). This finding is consistent with the results
of previous characterizations of EH neurons
(Cullen et al. 1993
;
McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
).
However, as noted by Scudder and Fuchs
(1992
) because two large
signals (of opposite signs) were added to produce a smaller one, measurement
errors become more significant. Thus we also compared on a neuron-by-neuron
basis the difference between head-velocity sensitivities estimated during
pWBRc and pWBRd to the eye-velocity sensitivity estimated during sinusoidal
pursuit and found that this relationship was even more robust
(R2 = 0.88) and that the slope was 1.0
(Fig. 3B). A
comparable finding was obtained when the eye/head-velocity sensitivities
estimated for pursuit, pWBRd, and pWBRc at peak velocities of 80°/s were
compared (data not shown). Moreover, type I and II EH neurons behaved
similarly in this analysis, and in fact, the only notable difference between
the two neuron subclasses during the head-restrained characterizations was
that type II neurons had on average a significantly smaller eye-position
sensitivity during fixation (P < 0.05, unpaired t-test).
Because they encode similar signals during each head-restrained behavioral
task, we consider type I and II EH neurons collectively in the following
text.
|
Role of error terms during sinusoidal smooth pursuit
Prior studies have shown that Purkinje cells of the cerebellar flocculus
and ventral paraflocculus can encode relatively small but significant retinal
slip signals (Hirata and Highstein
2001
; Suh et al.
2000
). Thus we tested whether EH neurons might also encode these
retinal error signals by first using the following model to describe neuronal
activity during sinusoidal smooth pursuit: FR(t) =
biaserr1 + kerr1 * eye position(t) +
rerr1 * eye velocity(t) +
cerr1 * eye position error(t-lat) +
derr1 * eye velocity error(t-lat) (pursuit error
model 1), where FR is the firing rate, biaserr1 is the bias
discharge, kerr1 is the eye-position sensitivity,
rerr1 is the eye-velocity sensitivity,
cerr1 is the eye-position-error sensitivity, and
derr1 is the eye-velocity-error sensitivity.
Eye-acceleration and eye-acceleration-error terms were not included because
these terms would be redundant with the position terms during sinusoidal
tracking.
Eye-position error was calculated as the difference between target position
and eye position at a specified latency (lat). Likewise, eye-velocity error
was calculated as the difference between target velocity and eye velocity at
the same latency. A latency of 100 ms was initially chosen to approximate the
delay of visual input to these neurons
(Stone and Lisberger 1990b
;
Suh et al. 2000
). Our example
neuron was typical in that during pursuit of a target with a peak velocity of
80°/s, the fit of pursuit error model 1
(Fig. 4A, thick trace,
bottom) was comparable to the fit of the pursuit model
(Fig. 4A, thick trace,
middle bottom; VAF = 0.79 vs. 0.78, respectively). Indeed, when the
goodness of fit of the models was compared on a neuron-by-neuron basis, the
resulting regression slope was 0.95 (not different from 1, P >
0.05; Fig. 4B). For
the population of neurons (n = 42), the addition of error terms at
latencies of 0, 50, 75, or 100 ms only slightly improved our ability to fit
the discharge of EH neurons during sinusoidal pursuit at either 40 or
80°/s (Fig. 4C,
compare across solid columns and gray-shaded columns). For both type I and II
neurons, eye-position and -velocity-error terms estimated at both velocities
and across all latencies were quite small, but were nevertheless significantly
different from zero (see Table
2; P < 0.05).
|
|
It could be argued that an individual EH neuron receives information from the Purkinje cells at a latency that was not necessarily one of the four tested. To address this possibility, the error model was estimated with latencies ranging from 0 to 130 ms during 80°/s pursuit. The model was optimized for each increment of 1 ms (i.e., a total of 130 optimizations) over this interval. For each neuron, the VAF of the model fit was comparable at all latencies. The results of this analysis for the sample of neurons (n = 42) are summarized in Fig. 4D. Based on this analysis we conclude that retinal eye-position-error or -velocity-error inputs had little influence on EH neuron responses during sinusoidal smooth pursuit.
Role of error term in step-ramp smooth pursuit
It has been shown in behavioral experiments that subjects make anticipatory
or predictive eye movements when tracking repetitive target trajectories such
as sinusoids (Barnes and Asselman
1991
; Barnes and Grealy
1992
; Barnes et al.
1995
,
1997
;
Collins and Barnes 1999
). To
minimize the influences of such predictive eye movements a constant-velocity
step-ramp paradigm was utilized (Rashbass
1961
). This type of stimuli has two advantages as compared with
sinusoidal target trajectories. First, the use of step-ramp target
trajectories enabled a more comprehensive characterization of the signals
carried by EH neurons because eye position and acceleration do not co-vary
during this paradigm as they do during sinusoidal pursuit. Second, the image
slip occurring in the first 100 ms of the paradigm was much larger than that
that occurred during the tracking of sinusoidal target motion. For example,
retinal slip velocities were on average 4050 versus 23° for
80°/s step ramps and sinusoidal target, respectively.
Figure 5A illustrates
three example trials of step-ramp pursuit for our example neuron. Note that
the step component of the stimulus has been removed to simplify the
presentation. The data set analyzed comprised the time period spanning from
pursuit onset to 150 ms into the pursuit movement, which encompassed the
acceleration phase and the beginning of the steady-state phase of pursuit.
During this time period, there was significant retinal slip and neuronal
discharges were quantified using the following model: FR(t) =
biaserr2 + kerr2 * eye position(t) +
rerr2 * eye velocity(t) +
aerr2 * eye acceleration(t) +
cerr2 * eye position error(t-lat) +
derr2* eye velocity error(t-lat) +
eerr2* eye acceleration error(t-lat) (pursuit
error model 2), where FR is the firing rate, biaserr2 is the bias
discharge, kerr2 is the eye-position sensitivity,
rerr2 is the eye-velocity sensitivity,
aerr2 is the eye-acceleration sensitivity,
cerr2 is the position-error sensitivity,
derr2 is the velocity-error sensitivity, and
eerr2 is the acceleration-error sensitivity.
|
This model construct is identical to that used by Suh et al.
(2000
) in their analysis of
floccular lobe Purkinje cell activity during smooth pursuit. Again, a latency
(lat) of 100 ms was initially chosen to approximate the delay of visual input
to these neurons. For the sample of neurons tested (n = 14), the
estimated biaserr2 (68 ± 8 spikes/s),
kerr2 [1.3 ± 1.12 (spikes/s)/°], and
rerr2 [1.12 ± 1.1 (spikes/s)/(°/s)] were
comparable to those obtained with our original "pursuit model,"
which did not contain any error terms (P > 0.05). Indeed, as is
illustrated for our example neuron, the addition of the three error terms and
eye acceleration term (Fig.
5A, bottom, thick trace, VAF = 0.89) resulted in
a fit comparable to the fit of the pursuit model
(Fig. 5A, middle
bottom, thick trace, VAF = 0.89; sample mean VAF = 0.37 ± 0.36 vs.
0.36 ± 0.24, respectively).
When considered separately, the estimated acceleration error coefficients of both type I and II neurons were not significant, and estimated position and velocity error coefficients were small but were significantly different from zero (Table 2; P < 0.05). Moreover, the estimated eye acceleration coefficients (aerr2) was small [0.0002 ± 0.0002 (spikes/s)/(°/s2)] and not different from zero (P > 0.05). Finally, the eye-position and -velocity coefficients estimated during step-ramp pursuit were comparable to those estimated during sinusoidal pursuit (Fig. 5B, left and right, respectively).
The ability of pursuit error model 2 to fit neuronal firing rates was comparable across all latencies (0, 50, 75, or 100 ms) that were used in the model optimization (Fig. 5C, compare gray shaded columns). Similar results were obtained when the acceleration phase (pursuit onset to 100 ms) and the steady-state phase (100300 ms) were analyzed separately (data not shown). Taken together, these results provide evidence that retinal error information played a negligible role in shaping EH neuron discharges.
EH neuron responses during combined eye-head gaze pursuit
Once a neuron had been characterized in the head-fixed condition, the
monkey's head was released from its restraint, and the same neuron was
recorded during voluntary combined eye-head pursuit. Thirty-three (33) neurons
remained isolated after the transition from the head-restrained to the
head-unrestrained condition, and each of these neurons was analyzed during
gaze shifts (see following text). Analysis of neuronal discharges during
pursuit was limited to neurons recorded when the monkey generated voluntary
head velocities >20°/s during combined eye-head tracking (n =
24). Figure 6 shows the
discharge of our example neuron (Figs.
1,
2,
4,
5, and
10) during three cycles of
sinusoidal pursuit (Fig.
6A) and during step-ramp pursuit
(Fig. 6B). As a first
step, we determined whether neuronal activity during gaze pursuit could be
predicted using the neuron's responses during head-restrained smooth
pursuit and/or passive whole-body rotation. Three specific predictions were
tested. First, we attempted to predict the neuron's response based on its
eye-position and -velocity sensitivities during smooth pursuit. This
prediction (prediction 1) is illustrated in
Fig. 6, A (VAF = 0.54)
and B (VAF = 0.68), for sinusoidal and step-ramp targets,
respectively. For our sample of neurons, the VAFs provided by this prediction
are summarized in Fig. 7, A and
B (
).
|
|
|
Second, head-movement-related terms, for which the head-velocity and
-acceleration-sensitivity coefficients taken from pWBRd, were added to the
model prediction (prediction 2; not shown). The simple addition of this term,
however, did not significantly improve our ability to predict the firing rate
during pursuit of either sinusoidal or step-ramp targets
(Fig. 7, A and
B, compare
and
, P > 0.05). It was
not surprising that the best prediction (prediction 3) was with a model that
summed the eye-position and -velocity sensitivities estimated during smooth
pursuit and head-velocity and -acceleration sensitivities taken from our
analysis of the pWBRc condition: FR(t) = biassp +
ksp * eye position(t) + rsp *
eye velocity(t) + gpWBRc * head
velocity(t) + apWBRc * head
acceleration(t) (prediction 3).
This model provided a good prediction of neuronal discharge during pursuit
of both sinusoidally moving (VAF = 0.77, prediction 3,
Fig. 6A, middle
bottom, thick trace) and step-ramp targets (VAF = 0.83, prediction 3,
Fig. 6B, middle
bottom, thick trace). The predictions based on this model were
significantly better than those based on the previous prediction models
(prediction 3, Fig. 7, A and
B,
; P < 0.05).
To further quantify the responses of EH neurons during gaze pursuit, we next estimated the coefficients of the eye- and head-related signals carried by EH neurons during gaze pursuit. A model with eye movement terms (e.g., bias, eye position, velocity, and acceleration), was first used and was found to provide a good fit to the firing rate during pursuit of sinusoidal (sample mean VAF = 0.49 ± 0.30; not shown) and step-ramp pursuit (sample mean VAF = 0.26 ± 0.28; not shown).
The addition of head velocity and head acceleration terms to the model:
FR(t) = biassp + ksp * eye
position(t) + rsp * eye velocity(t) +
gpWBRc * head velocity(t) +
apWBRc * head acceleration(t) (eye-head
estimate), improved our ability to fit the responses during sinusoidal
combined eye-head pursuit (sample mean VAF = 0.54 ± 0.28;
Fig. 7A,
). The
model fit is shown in Fig.
6A for our example neuron (bottom, thick trace,
eye-head estimate; VAF = 0.84). Furthermore, on a neuron-by-neuron basis, the
goodness of fit (VAF) provided by prediction 3 was well correlated with, and
only marginally worse than, that provided by the same model when the
parameters were optimized for neuronal response during sinusoidal pursuit
(slope = 0.94, not different from 1, P > 0.05;
Fig. 7B).
Similar results were obtained for the analysis of combined eye-head pursuit
of step-ramp targets where the eye-head estimate provided a better fit of the
example neuron's discharge activity (Fig.
6B, bottom, thick trace, VAF = 0.83) than a model with
just eye movement terms (VAF =0.68). For the subset of neurons tested during
step-ramp gaze pursuit (n = 9), the sample mean VAF was 0.53 ±
0.25 (Fig. 7B,
). Taken together, our results show that EH neurons encode head as well
as eye movement related signals during combined eye-head gaze pursuit.
A comparison of the eye- and head-movement-related responses of EH neurons
during head-restrained and -unrestrained pursuit paradigms was then made to
determine whether they differed in these two conditions. First, to facilitate
comparison, estimated coefficients were normalized relative to those
estimated during head-restrained pursuit. We found that average estimated bias
values (Fig. 8A),
eye-position sensitivity (Fig.
8B), and eye-velocity sensitivity
(Fig. 8C) were
comparable across all pursuit tasks (P > 0.05). Second, a
neuron-by-neuron comparison of the head-velocity sensitivities estimated
during combined eye-head gaze pursuit and pWBRc revealed that they were
comparable (slope = 0.93, not different from 1, P > 0.05;
Fig. 9A). Moreover,
head-velocity sensitivities of the EH neurons tested were comparable during
pWBRc, sinusoidal gaze pursuit (n = 24,
Fig. 9B), and
step-ramp pursuit (n = 9, Fig.
9C). Thus EH neurons encode similar head-movement-related
signals during these two different behavioral tasks. This is an important
observation because these two head movements differ in that the head was
passively rotated during pWBRc while it was voluntarily moved during gaze
pursuit. In addition, head-velocity sensitivity coefficients estimated during
pWBRc and gaze pursuit were significantly larger than those estimated during
pWBRd (Fig. 9, B and
C, compare
and
, P < 0.05). The
implications of these findings are considered in the
DISCUSSION.
|
|
EH neuron responses during rapid gaze redirection: saccades and eye-head gaze shifts
The responses of EH neurons were characterized during head-restrained
saccades and -unrestrained gaze shifts in which the monkeys rapidly reoriented
their axis of gaze in space. Nearly one-half of the EH neurons (19/42) showed
a burst in discharge activity during saccades in the neuron's "on
direction" during pursuit (Fig.
10A, left). The remainder were either
unresponsive (n = 14, Fig.
10A, right) or paused (n = 9) in
activity. These results are consistent with previous studies
(Chen-Huang and McCrea 1999
;
Cullen at al. 1993
;
Gdowski and McCrea 1999
;
McFarland and Fuchs 1992
;
Scudder and Fuchs 1992
). For
the neurons that burst, the bias discharge (biassac), eye-position
sensitivity (ksac), and eye-velocity sensitivity
(rsac) were estimated for each neuron using the model,
FR(t) = biassac + ksac * eye
position(t-lat) + rsac * eye
velocity(t-lat) (saccade model).
The mean biassac was 106 ± 56 spikes/s, the mean
ksac was 1.97 ± 1.10 (spikes/s)/°, and the mean
rsac was 0.22 ± 0.17 (spikes/s)/(°/s). In
addition, the mean value estimated for dynamic latency (lat) (see
Cullen et al. 1996
and
Sylvestre and Cullen 1999
for
details) was 10 ± 9 ms, indicating that the burst lead was comparable
to that which has been reported for saccadic premotor burst neurons
(Cullen and Guitton 1997
).
EH neurons responded in a similar manner during gaze shifts and saccades. For example, all neurons that did not burst during saccades (n = 14) also did not burst during gaze shifts. Similarly, all neurons that burst during on-direction saccades (n = 19) also burst during on-direction gaze shifts. This is illustrated in Fig. 10B where responses during gaze shifts of the same two neurons as in Fig. 10A are shown. For these latter neurons, we found that a model based on their activity during saccades was a poor predictor of their neuronal activity during gaze shifts (sample mean VAF = 0.19 ± 0.29) suggesting that the discharge was under-modeled. A model that included a head-velocity sensitivity (ggs), as well as a bias discharge, eye-position sensitivity, and eye-velocity sensitivity: FR(t) = biasgs + kgs * eye position(t-lat) + rgs * eye velocity(t-lat) + ggs * head velocity(t-lat) (gaze shift model), provided a much better fit on neuronal response (sample mean VAF = 0.41 ± 0.17, P < 0.05; Fig. 10B). The estimated biasgs and kgs were not significantly different from those during saccades (mean = 118 ± 56 spikes/s and 1.98 ± 1.51 (spikes/s)/°, respectively, P > 0.05) and the eye-velocity sensitivity was estimated to be smaller than during saccades [mean rgs = 0.16 ± 0.14 (spikes/s)/(°/s), P < 0.05]. The estimated head-velocity sensitivity of the neurons was 0.66 ± 0.63 (spikes/s)/(°/s), which was greater than that estimated for the same neurons during pWBRd (P < 0.05) and comparable to that estimated during pWBRc (P > 0.05). In contrast, the head-velocity sensitivity of EH neurons that did not burst were comparable to those estimated during pWBRd (P > 0.05). We also estimated the head-velocity sensitivity of EH neurons in the post gaze shift interval where gaze was stable, but the head continued to move, and found that the head velocity sensitivities of the neurons were comparable to those estimated for the same neurons during pWBRd (P < 0.05). These findings suggest that EH neurons encode head-motion-related information similarly during active and passive head rotations when gaze is stable.
Influence of neck proprioceptive inputs
Our finding that EH neurons encoded similar head-movement-related signals during pWBRc and gaze pursuit strongly suggests that the activation of neck proprioceptors does not play a role in modulating neuronal activity during gaze pursuit. To further test this proposal, two different paradigms were used. First we passively rotated the monkey's body while holding its head earth stationary (Fig. 11A). The neuron shown in Fig. 11A was typical in that its discharge was not significantly affected by the passive neck rotations. Its activity could be well predicted by a model based on the neuron's bias and eye-position sensitivity (Fig. 11A, thick trace, prediction). The lack of influence was even more apparent after the firing rate was corrected for the neuron's eye-position sensitivity (Fig. 11A, FRcorr). The corrected response of each neuron was fit using the following model: FRcorr = biasBUH + nBUH * neck velocity (BUH model), where biasBUH is the bias discharge, neck velocity is the velocity of the body rotation, and nBUH is neck-velocity sensitivity. For all the neurons tested (n = 11), the mean biasBUH of 90 ± 49 spikes/s was comparable to that measured during fixation (P > 0.05), and the response to passive neck proprioceptor activation was negligible [mean nBUH = 0.07 ± 0.17 (spikes/s)/(°/s)].
|
To further investigate the influence of neck proprioceptor inputs on EH neuron discharge activity, the monkey's head was passively rotated on its earth stationary body. The rotations elicited head velocities and trajectories comparable to those observed during natural head movements (Fig. 11B). Each neuron's discharge was characterized during this passive head-on-body rotation (PHBR) paradigm using the following model: FR = biasPHBR+ kPHBR * eye position + gPHBR * head velocity (PHBR model) in which the bias discharge (biasPHBR), the eye-position sensitivity (kPHBR), and the head-velocity sensitivity (gPHBR) were estimated during segments when gaze was stable. Overall, for the neurons tested (n = 14), biasPHBR, the kPHBR, and gPHBR were comparable to those values estimated during pWBRd (P > 0.05). Furthermore the neuron illustrated in Fig. 11B was typical in that when gaze was stable, a model based on the neuron's activity during pWBRd provided a good prediction of neuronal discharge (pWBRd prediction, thick trace; VAF = 0.59; sample mean VAF = 0.29 ± 0.19). Based on the results from these two approaches (BUH and PHBR paradigms), we conclude