Spatial properties of second-order vestibulo-ocular relay neurons in the alert cat K. Fukushima t, S.I. Perlmutter 2, J.F. Baker 2, and B.W. Peterson 2 1 Department of Physiology,Hokkaido University School of Medicine, West 7, North 15, Sapporo 060, Japan 2 Department of Physiology,Ward 5, Northwestern University Medical School, 303 E. Chicago Ave., Chicago, IL 60611, USA Received September 11, 1989 / Accepted March 30, 1990
Summary. Second-order vestibular nucleus neurons which were antidromically activated from the region of the oculomotor nucleus (second-order vestibuloocular relay neurons) were studied in alert cats during wholebody rotations in many horizontal and vertical planes. Sinusoidal rotation elicited sinusoidal modulation of firing rates except during rotation in a clearly defined null plane. Response gain (spikes/s/deg/s) varied as a cosine function of the orientation of the cat with respect to a horizontal rotation axis, and phases were near that of head velocity, suggesting linear summation of canal inputs. A maximum activation direction (MAD) was calculated for each cell to represent the axis of rotation in three-dimensional space for which the cell responded maximally. Second-order vestibuloocular neurons divided into 3 non-overlapping populations of MADs, indicating primary canal input from either anterior, posterior, or horizontal semicircular canal (AC, PC, HC cells). 80/84 neurons received primary canal input from ipsilateral vertical canals. Of these, at least 6 received input from more than one vertical canal, suggested by MAD azimuths which were sufficiently misaligned with their primary canal. In addition, 21/80 received convergent input from a horizontal canal, with about equal number of type I and type II yaw responses. 4/84 neurons were HC cells; all of them received convergent input from at least one vertical canal. Activity of many vertical second-order vestibuloocular neurons was also related to vertical and/or horizontal eye position. All AC and PC cells that had vertical eye position sensitivity had upward and downward on-directions, respectively. A number of PC cells had MADs centered around the MAD of the superior oblique muscle, and 2/3 AC cells recorded in the superior vestibular nucleus had MADs near that of the inferior oblique. Thus, signals with spatial properties appropriate to activate oblique eye muscles are present at the second-order vestibular neuron level. In contrast, none of the second-order vestibuloocular neurons had MADs near those of the superior or inferior rectus muscles. Signals appropriate to activate these eye muscles Offprint requests to." S.I. Perlmutter (address see above)
might be produced by combining signals from ipsilateral and contralateral AC neurons (for superior rectus) or PC neurons (for inferior rectus). Alternatively, less direct pathways such as those involving third or higher order vestibular or interstitial nucleus of Cajal neurons might play a crucial role in the spatial transformations between semicircular canals and vertical rectus eye muscles.
Introduction The vestibulo-ocular reflex (VOR) produces compensatory eye movement in the orbit to keep vision stable during head rotation in any plane of space. This is achieved by spatial and temporal conversions of the input signals arriving from the semicircular canals and otolith organs to produce the appropriate output signals to activate extraocular motoneurons. Take the VOR during vertical rotation in the sagittal (pitch) plane, for example. Head movement is detected by the vertical semicircular canals that are located approximately 45 ~ from the vertical pitch plane (Blanks et al. 1972), and resultant compensatory vertical eye movement is produced by the vertical recti and oblique muscles that are positioned about 25 ~ and 53 ~ from the pitch plane (Robinson 1982; Ezure and Graf 1984; Peterson et al. 1986). A large portion of the vestibulo-ocular signal reaching extraocular motoneurons is carried by secondorder vestibuloocular neurons, which form the middle leg of a 3 neuron arc. Therefore, in order to understand the neural mechanisms of signal conversions in the VOR, it is essential to study in detail in alert animals the signals carried by these neurons. King et al. (1976) and Pola and Robinson (1978) have shown how these neurons participate in the dynamic transformations within the VOR. There has been a controversy concerning whether or not significant convergence occurs at the second-order
463 v e s t i b u l a r n e u r o n ; c a n a l specificity w a s r e p o r t e d in a n e s t h e t i z e d a n i m a l s ( W i l s o n a n d F e l p e l 1972 in t h e p i g e o n ; K a s a h a r a a n d U c h i n o 1974 in t h e cat), w h e r e a s c a n a l otolith and even canal-canal convergence were reported in u n a n e s t h e t i z e d p r e p a r a t i o n s ( C u r t h o y s a n d M a r k h a m 1971 ; M a r k h a m a n d C u r t h o y s 1972). P r e v i o u s s t u d i e s in this l a b o r a t o r y a l s o r e v e a l e d t h a t a m a j o r i t y o f s e c o n d a r y v e s t i b u l a r n e u r o n s o f a l e r t c a t s r e c e i v e d c o n v e r g e n t inputs from more than one pair of canals, and that many s h o w e d c a n a l - o t o l i t h c o n v e r g e n c e ( B a k e r et al. 1984a, b). W h i l e t h e c o n v e r g e n t n e u r o n s s t u d i e d in t h e s e r e p o r t s w e r e n o t i d e n t i f i e d as s e c o n d - o r d e r v e s t i b u l o o c u l a r n e u r o n s , t h e p r e s e n c e o f c o n v e r g e n c e at t h e s e c o n d - o r d e r n e u r o n level raises t h e p o s s i b i l i t y t h a t a p a r t o f s p a t i a l t r a n s f o r m a t i o n in t h e V O R m a y o c c u r as a r e s u l t o f convergent action of multiple labyrinthine receptors upon second-order vestibuloocular neurons. Indeed, Peters o n et al. (1987) s h o w e d in d e c e r e b r a t e c a t s t h a t a significant spatial transformation occurred already at secondorder vestibuloocular neurons. The present experiments analyze the properties of second-order vestibuloocular n e u r o n s in a l e r t cats. P r e l i m i n a r y r e s u l t s o n p a r t s o f this w o r k h a v e b e e n p r e s e n t e d e l s e w h e r e ( P e r l m u t t e r et al. 1988).
Animal preparations Three cats were used in this study. All surgery was done under halothane-nitrous oxide anesthesia (halothane 1%, N20 80%, O2 20%) and under aseptic conditions. Demerol was given after surgery to reduce pain, and penicillin was also given 1:o prevent infection. The cat's head was fixed in the stereotaxic frame, and a head holder and recording chamber were installed over the skull. The head holder had a rectangular hollow through which a bar could be inserted and tightly clamped. During recording the bar was securely clamped to the frame so that the cat's head was fixed in the stereotaxic frame without the use of ear bars or a palate clamp. The anterior-posterior orientation of the frame was reversed so that it did not obstruct the cat's vision. This approach allowed us to make neuronal recordings stereotaxically in the left vestibular nuclei. Ag-AgC1 electrodes were implanted in the skull around the eyes to record horizontal and vertical eye movements with DC electrooculography (EOG). Vertical EOG was recorded from the left eye and horizontal eye movements were recorded bitemporally. Silver ball electrodes were implanted against the oval and round windows via an opening made in the left tympanic bulla after which the bulla was sealed with dental cement. Bipolar single pulses of 0.1 ms, 0.1-1.0 mA were delivered to the electrodes to stimulate the left vestibular nerve. Stimulus strength was kept below or just above the threshold for twitch of facial musculature, and was lowered if the animal became perturbed. Stimulus intensities were referred to the threshold current for evoking a vestibular N 1 potential (Shimazu and Precht 1965) in the left vestibular nuclei. In order to identify second-order vestibuloocular neurons, a 0.5 mm diameter concentric electrode (inter-electrode distance 0.5 ram, Rhodes Medical Instruments, SNEX 100) was stereotaxically implanted in the region of the midbrain medial longitudinal fasciculus (MLF)-oculomotor complex area at A0-3.0, L0.5, H~0.5. Bipolar or monopolar 0.1 msec cathodal pulses of 10~500 gA were applied to this electrode to activate vestibuloocular neurons antidromically. At no time was there any behavioral indication that the animal was aware of the midbrain stimulation. In a fourth animal the stimulating electrode was misplaced in the deep reticular formation (Fig. 8I, cat A) and no vestibuloocular neurons were identified.
This suggests that antidromic activation of cells with the well-placed electrodes in other animals can be localized to stimulation in the region of the oculomotor nucleus.
Recordin 9 procedures All recordings were done under dim room light, and care was taken to maintain alertness of the cats by providing a variety of auditory stimuli. In all recording sessions, the cat's body was restrained in an Ace bandage and comfortably placed in a box fixed to an apparatus that had two (horizontal and vertical) servo motor powered axes of rotation. The interaural midpoint of the animal's head was brought close to the axes of rotation. The head holder was used to attach the head to a stereotaxic frame which was fixed to the apparatus in an inclined position so that the cat's head was pitched 28 ~ nose down from the stereotaxic horizontal plane to bring the vertical canals near a vertical position. Horizontal (yaw) rotations at this head orientation provided minimal vertical canal stimulation and a near maximal horizontal canal stimulus (Blanks et al. 1972). The cat box sat on a turntable which could be manually rotated about a vertical axis and locked in a fixed position with respect to the apparatus. As this positioning was independent of the horizontal servo axis, rotations were possible in many vertical planes relative to the cat (see cartoons in Fig. 1A). For example, if the turntable was positioned so that the cat's longitudinal (head-to-tail) axis was perpendicuiar to the horizontal servo axis (0 ~ orientation angle in our coordinate system), rotation about that servo axis was in the cat's pitch plane. If the turntable was positioned at a 45 ~ clockwise (as seen from above) angle (+ 45 ~ orientation angle), rotation was in the cat's left anterior/right posterior canal plane. Orientation at 135 ~ resulted in rotations in the left posterior/right anterior canal plane. Sinusoidal rotations at 0.5 Hz with typical peak-to-peak amplitude of 10~ were used for second-order vestibuioocular neuron characterization. In some cases, sum-of-sinusoids rotations containing 10 different frequencies, which were relatively prime harmonics of a common base frequency (Wilson et al. 1979), were used for vertical and horizontal rotations. Recordings were made extracellularly from neurons in the left vestibular nuclei where large N1 field potentials were recorded (Shimazu and Precht 1965), using Epoxilyte insulated-tungsten electrodes with an exposed tip of 10 20 lain and impedances of 0.8-1.2 Mf~ (tested at 1 kHz). Thresholds for evoking N1 field potentials (N1T) ranged from less than 100 ~tA to 300 ~tA. Tracking was done systematically from P6.0 to PI0.0 and from 2.0-4.5 mm left of the mid-sagittal plane. In each animal 50-120 recording tracks were made over a period of 1-3 months. None of the cats exhibited observable behavioral deficits at any time during this period. Repeated tracking in the same location was done occasionally, with the second penetration more than 2 weeks after the first. Combined yaw and pitch rotations were applied as the electrode was advanced. Action potentials of isolated cells were detected with a window discriminator, and neuronal responses to rotation in 2-15 (typically 8) vertical planes and in the earth horizontal plane were averaged over 20-40 cycles in each plane. Null points were determined in many cells from response phase reversal as in previous studies (Estes et al. 1975). To examine eye movement related responses of vestibular neurons, eye movements were induced by attracting the animals' attention to a variety of points in space while the turntable was stationary. Twenty to 40 sec of single neuron discharges and eye position records were stored on the computer for later analysis. EOG signals were calibrated by sinusoidal whole body rotation at 0.1-0.25 Hz, 5-10 ~ peak-to-peak amplitude of the cat in dim room light while it fixated a stationary interesting object.
Data collection and analysis Methods of data acquisition and analysis were similar to those described previously (Bilotto et al. 1982; Baker et al. 1985). Single neuron responses to 20-40 cycles of sinusoidal rotation were collect-
464 ed as cumulative spike-occurrence histograms (8 ms bin-width) representing the cell's average activity to a single cycle of rotation (e.g. Fig. 1A). This histogram was fitted with a sine wave using a least squares method to obtain gain (peak firing rate in spikes/s divided by peak head velocity in ~ and phase (with respect to head position) of the neuron's response. During sum-of-sinusoids rotation, cumulative spike-occurrence histograms with 64 bins/cycle were collected at each of the frequencies present in the stimulus. A least squares sinusoid at the stimulus frequency was fitted to each histogram to obtain gain and phase values. Gain and phase for vertical rotations were plotted against the preset orientation angle of the body with respect to the horizontal servo axis (see above, cf. Blanks et al. 1975). All neurons displayed a sinusoidal variation of response gain as a function of orientation angle, and a sine wave was fitted to these curves using a least squares method (e.g. Fig. 1B). Maximal vertical gain, and estimated gain values of responses to pitch and roll rotations were calculated from the sinusoidal fits for each cell. The fitting algorithm also calculated a standard error (SE) of estimation of the orientation angle eliciting the maximal response. Gain values for yaw, pitch and roll were used to construct a three dimensional (yaw, pitch, roll) response vector for each cell according to Baker et al. (1985). This vector was then normalized to have a length of 1.0. The resulting unit maximal activation direction (MAD) indicated the axis and direction of rotation that elicited the maximal response of that neuron (e.g. Fig. 1C). MADs were compared with the average semicircular canal coordinates obtained by Blanks et al. (1972, also Robinson 1982 for functional canal planes produced by coplanar canal inputs). To examine eye movement related responses, digitized eye position signals and single neuron spike-occurrence histograms were recorded for trials of approximately 60 s with the cat stationary. Saccades were removed by computer-assisted manual editing of EOG records. Average neuron firing rate during each fixation period (no. of spikes occurring between two saccades/time between the two saccades) was plotted as a function of vertical and horizontal eye position between saccades. A least squares method was used to detect any linear correlation between firing rate and eye position.
dromically elicited or s p o n t a n e o u s discharges (e.g. Fig. 6A). S e c o n d - o r d e r vestibular n e u r o n s that were antidromically activated f r o m the m i d b r a i n M L F area were classified as second-order vestibuloocular neurons. A total o f 108 second-order vestibuloocular neurons were encountered in 3 cats. L o c a t i o n o f the tip o f the midbrain stimulating electrodes is s u m m a r i z e d in Fig. 8 I - K . The electrode was in the M L F close to the trochlear nucleus in cat T, in the close vicinity o f the oculom o t o r nuclear complex in cat S, and in the periaqueductal grey a b o u t 0.7 m m a w a y f r o m the interstitial nucleus o f Cajal ( I N C ) and a b o u t 2 m m rostral to the oculom o t o r nuclei in cat K. Threshold currents for a n t i d r o m i c activation o f vestibular n e u r o n s were lowest in cats T and S, and r a n g e d f r o m less than 20 to 200 with a m e d i a n o f 50 pA, whereas the values in cat K r a n g e d f r o m 70 to 200 with a m e d i a n o f 120 gA. A n t i d r o m i c latencies were similar in these three cats and r a n g e d f r o m 0.3 to 1.0 with a m e d i a n o f 0.5 ms. O f the 108 second-order vestibuloocular neurons, 84 were recorded during rotations in m a n y vertical and horizontal planes and their spatial properties were analyzed. O f these 84 neurons, 80 responded m o r e strongly to vertical rotations ( m a x i m u m vertical gain at least twice y a w gain) a n d are classified as vertical secondorder vestibuloocular neurons. The remaining 4 n e u r o n s had y a w gains greater than the m a x i m u m vertical gain and are classified as horizontal second-order vestibuloocular neurons.
Spatial properties of vertical second-order vestibuloocular neurons Response veetors. Typical responses to sinusoidal rota-
Histological procedures Near the conclusion of the recording period in each cat, the recording sites of some cells were marked with electrolytic lesions by passing DC current through the microelectrode (electrode tip negative, 20-30 gA for 20 s). At the end of each experiment, the position of the tip of the stimulating electrode was also marked electrolytically. The animals were then deeply anesthetized with pentobarbital (50 mg/kg, i.p.), and perfused with 10% formalin. After fixation, the brainstem was sectioned at 50 Ilm on a freezing microtome, and sections were stained by the Kluver-Barrera method. The stimulating position and, where possible, locations at which neurons had been recorded were verified and reconstructed. There was no evidence of any unusual tissue damage in any of the animals.
Identification of secon&order vestibulo-ocular relay neurons Vestibular n e u r o n s that were activated within 1.3 ms o f the ipsilateral vestibular nerve stimulation (at less t h a n 5 times the threshold for eliciting an N1 field potential) were classified as s e c o n d - o r d e r vestibular nuclear neurons (cf Wilson et al. 1979). A n t i d r o m i c responses to stimulation o f the m i d b r a i n electrode were identified by short, c o n s t a n t latency, sharply defined stimulus threshold, and collision block o f a n t i d r o m i c spikes by o r t h o -
tions in m a n y vertical and horizontal planes are presented in Fig. 1A for three cells. S e c o n d - o r d e r vestibuloocular n e u r o n firing rate was m o d u l a t e d sinusoidally in all planes except for a clear null response in one orientation. The cell s h o w n in the left c o l u m n had a m a x i m a l response at the 135 ~ orientation angle w h e n the cat m o v e d noseand-right-ear up. It h a d a near null response at the 45 ~ angle, and a reversal (i.e. sign change) o f response phase at that null orientation. This behavior is w h a t one w o u l d expect o f a n e u r o n receiving its p r i m a r y excitatory input f r o m the left posterior semicircular canal. I n the 135 ~ orientation the r o t a t i o n plane is nearly c o p l a n a r with the canal, maximally activating p r i m a r y afferents. In the 45 ~ orientation the r o t a t i o n plane is nearly perpendicular to that o f the canal, and activation is minimal. A t the null plane, the phase o f posterior canal activation reverses. T h a t is, on one side o f the null plane (e.g. 50~ semicircular canal fluid in the left posterior canal flows in the a m p u l l o f u g a l (excitatory) direction w h e n the cat moves nose-and-left ear down. O n the other side o f the null plane (e.g. 40~ fluid flows in the ampullopetal (inhibitory) direction for the same m o v e m e n t . The relation o f these responses to semicircular canal inputs is perhaps better seen in Fig. 1B, which plots gain and phase values as a function o f orientation angle for vertical plane rotations. Positive gains and negative phases are associated with m a x i m u m response in the
PC C E L L
CONV C E L L
AC C E L L
7" 25 sp/s
0 -6.5 180
§ § § § t" § § t"
C 9 [-1
positive r o t a t i o n directions ( n o s e - d o w n pitch, left sided o w n roll, or nose-left yaw). The gain curves are reasonably sinusoidal with calculated m a x i m u m responses at 135 ~, 58 ~, a n d 51 ~ for the cells in the left, middle, and right columns, respectively (SEs r a n g i n g f r o m 2 ~ to 4~ Phases are m o s t l y c o n s t a n t a r o u n d 90 ~ relative to head position indicating a h e a d velocity signal. This sinusoidal variation in gain and c o n s t a n t phase b e h a v i o r can be explained as a simple linear c o m b i n a t i o n o f vertical canal inputs. The directional properties o f these responses are summarized in Fig. 1C by plotting normalized M A D vectors for each cell (see M e t h o d s , Baker et al. 1985). As described above, the response o f the cell in the left c o l u m n suggests that it receives its p r i m a r y input f r o m the left
Fig. 1. A Responses of three second-order vestibuloocular neurons during sinusoidal rotations (at 0.5 Hz) in the vertical planes at different head orientations (0-135) and in the horizontal plane (yaw). TAB indicates table position. Positive table position corresponds to the direction of rotation indicated by the + sign in the cartoons of the cat and TAB calibration bar. Vertical rotations at 0 ~ and 90 ~ indicate pitch and roll rotations, respectively. B Gain (relative to velocity) and phase (relative to turntable position) of responses to rotations at 0.5 Hz in different vertical planes plotted for each cell. Positive gains and negative phases are associated with maximum responses in the positive rotation directions (nose-down pitch, left side-down roll, or nose-left yaw). Curves show a least squares sinusoid fitted to gain data. C Normalized response axis vectors (MADs) for each cell. Top view shows the vectors in pitch (abscissae) and roll (ordinates) plane, and front view shows the vectors in pitch (abscissae) and yaw (ordinates) plane. Use the righthand rule to determine the plane of rotation represented by a vector
posterior canal. It is therefore classified as a P C - t y p e neuron. The cell s h o w n in the right c o l u m n behaves as if its p r i m a r y input were f r o m the left anterior canal, a n d is classified as an A C - t y p e neuron. The cell in the middle c o l u m n shows responses which reflect input primarily f r o m the left anterior canal with convergent input f r o m the left posterior canal sufficient to shift its M A D vector in the pitch-roll plane 19 ~ f r o m that o f the left anterior canal. All three cells also responded to leftward horizontal rotations (Fig. 1A, Y A W ) , as indicated by the front view in Fig. 1C. N o r m a l i z e d M A D vectors o f all vertical secondorder vestibuloocular n e u r o n s are s h o w n for each cat (K, S, T) in Fig. 2A C. A l t h o u g h there are some variations in M A D s a m o n g the three cats, the distributions are
n=35 -1 -1 -1
n =25 -1
similar and the response vectors for all 80 neurons are shown together in Fig. 2D, along with the maximal activation directions for the semicircular canals. All cells responded as if their p r i m a r y input was from the ipsilateral anterior or posterior canal (AC- or PC-type), with a percentage (see below) exhibiting convergent behavior. AC-type responses have M A D vectors which m a k e an angle of,,, 45 ~ with the pitch axis in the pitchroll plane (angle 13 in Fig. 2D) as does the left anterior canal. PC-type responses m a k e an angle o f ~ 135 ~ with the pitch axis as does the left posterior canal. There were no second-order vestibuloocular neurons whose behavior suggested they received p r i m a r y input f r o m a contralateral (right) semicircular canal (i.e. all cells have M A D vectors above the pitch axis in T o p Views of Fig. 2). Moreover, m a n y cells responded to horizontal rotation, being excited by either ipsilateral yaw (type I response of Duensing and Schaefer 1958) or contralateral yaw (type II).
Comparison of MAD vectors. Figure 3A, C summarize the M A D azimuth angles of 32 A C and 48 PC cells,
n = 80
Fig. 2A-D. Normalized MAD vectors for individual cats (A-C) and pooled data for all 3 cats (D). Top view and front view show the vectors in pitch and roll, and pitch and yaw planes, respectively, as in Fig. 1 C. /~ indicates convention for measuring MAD azimuth angle in pitch-roll plane. Arrows represent vectors of semicircular canals (data from Blanks et al. 1972 rotated into our coordinate frame), e.g. lpc = left posterior canal, lpc/rac = paired left posterior and right anterior canals
obtained by computing the arc tangent of the ratio of the roll to pitch components of the cell's M A D vector (equivalent to angle 13 in Fig. 2D). The mean values for all AC and PC cells were 42.8 ~ and 132.2 ~ respectively. The SEs of the estimated M A D azimuths from the sinusoidal gain fits for each cell ranged f r o m 0.5 ~ to 11.3 ~ with a median of 3.4 ~ (mean 4.4~ To compare these data with average semicircular canal coordinates measured by Blanks et al. (1972), we determined the n u m b e r of cells for which the M A D azimuth was either 5~ ~ (light shading) or > 10 ~ (dark shading) f r o m that of the corresponding canal, or of the corresponding canal when paired with the coplanar canal of the opposite labyrinth. Robinson (1982) argues that opposite canals may be considered as functional pairs with respect to the inputs converging on second order neurons due to commissural inhibition. His model of the VOR collapses the left anterior and right posterior (also the left posterior and right anterior) canals into a single functional canal which lies midway between the two. It is nonetheless important to compare second-order vestibuloocular neuron MADs to individual canal planes as well: Uchino et al. (1986) reported that some vertical second-order vestibuloocular neurons did not receive commissural
467 |i:i:~:i:};i:3it::!:~:~:7:i:i:i: l
Convergent at 5~ criterion
Convergent at 10~ criterion
2 1 1 0 25
!i i111 i
PC CELLS ca
L 4 Z
0 2 1 0
L tlH I I
VECTOR ANGLE (DEG FROM PITCH) Fig. 3 A - D . H i s t o g r a m s of M A D a z i m u t h s (A, C) and elevations (B,
D) for AC second-order vestibuloocular neurons (A, B) and PC second-order vestibuloocular neurons (C, D). RIO and LSO indicate vector angles for the right inferior oblique and left superior
inhibition in anesthetized cats. It is possible, therefore, that some of the neurons summarized in Fig. 3A, C received inputs from only one vertical canal. Blanks et al. (1972) made their measurements with cats positioned in the stereotaxic plane. Our recordings were done with the animals pitched 28 ~ nose down from stereotaxic horizontal. To allow direct comparison of our neuronal MAD vectors with the geometry of the semicircular canals, we performed a simple rotatory coordinate transformation on the data from Blanks et al. (1972). By rotating their canal coordinates into the 28 ~ nose down coordinate frame, we obtained the maximum activation direction vectors of the canals in our coordinate system. Angle 13 can be calculated from Table 2 of Blanks et al. to be 38 ~ and 133 ~ for the MADs of the anterior and posterior canals for the cat in average stereotaxic plane. The corresponding paired functional canal plane vectors are 43 ~ and 137~. The coordinate transformation to the 28 ~ nose down position yields values of 13 in our frame of reference (solid bars in Fig. 3A, C) of 41 ~ for the left anterior canal and 128 ~ for the left posterior canal. For the paired canal planes 13s are 46.5 ~ and 133.5 ~ M A D a z i m u t h s o f six cells (3 A C a n d 3 PC, 6/ 80 = 7.5%, d a r k s h a d e d a r e a s in Fig. 3A, C) were shifted m o r e t h a n 10 ~ f r o m b o t h the single a n d p a i r e d c a n a l a z i m u t h s . Since their r e s p o n s e s c a n n o t be e x p l a i n e d b y
1 1011 I
VECTOR ANGLE (DEG FROM HORIZONTAL PLANE)
oblique muscles obtained in decerebrate cats by Peterson et al. (1984). Solid bars indicate vector angles for semicircular canals as in Fig. 2
the a n t e r i o r o r p o s t e r i o r c a n a l i n p u t a l o n e , it seems t h a t t h e y received c o n v e r g e n t i n p u t s f r o m b o t h p a i r s o f vertical c a n a l s (see Discussion). A n e x a m p l e o f such a cell has been s h o w n in Fig. 1, m i d d l e c o l u m n ( c o n v e r g e n t cell). A n a d d i t i o n a l 4 A C a n d 13 P C cells w h o s e M A D azim u t h s lay f r o m 5~ ~ f r o m the n e a r e s t c a n a l p l a n e m a y also have received some c o n v e r g e n t i n p u t f r o m the o r t h o g o n a l vertical canals (light s h a d e d areas in Fig. 3A, C). Y a w r e s p o n s e s are s u m m a r i z e d in Fig. 3B, D for all the A C a n d P C cells b y c a l c u l a t i n g their M A D e l e v a t i o n angles, i.e. the angle b e t w e e n the M A D a n d the h o r i z o n t al (pitch-roll) plane. These values c o r r e s p o n d to the arcsine o f the n o r m a l i z e d y a w g a i n values for e a c h cell. T h e m e a n e l e v a t i o n s for all the A C a n d P C s e c o n d - o r d e r v e s t i b u l o o c u l a r n e u r o n s were 2.2 ~ a n d 5.3 ~ respectively. T h e e l e v a t i o n s o f single a n d p a i r e d v e r t i c a l c a n a l s f r o m the h o r i z o n t a l p l a n e a l o n g the d i r e c t i o n o f m a x i m a l vertical a c t i v a t i o n were also c a l c u l a t e d b y r o t a t i n g the B l a n k s et al. (1972) d a t a i n t o o u r c o o r d i n a t e f r a m e . T h e y were - 3 ~ 2.6 ~ a n d 0 ~ respectively (solid b a r s in Fig. 3B, D). I f we use the e r r o r a l l o w a n c e o f • 10 ~ as in the cases o f
468 Table 1. Number of second-order vestibuloocular neurons that did or did not receive convergent input from other canals, and mean (a: SD) maximal vertical gain (V) and yaw gain (Y). V conv: MAD azimuth deviated by more than 10~ from the planes of both the primary vertical canal, and the paired primary canal/contralateral pair. H cony: MAD elevation deviated by more than 10~ from the planes of both the left horizontal canal, and the paired left/right horizontal canals. 3 canal conv: Both V conv& H cony (i.e. neurons which receive input from all 3 ipsilateral canals or canal pairs). Non conv: cells that did not meet this 10 deg criterion for convergence. Gains are expressed as mean • standard deviation. SD values were not calculated for groups of less than 5 cells Non conv V conv H conv
3 canal Total conv
AC-type No of cells Max V gain Max Y gain
16 3.0 4- 0.9 0.3•
3 2.4 0.2
13 3.2 • 1.4 1.1 •
PC-type No of cells Max V gain Max Y gain
39 2.7 + 1.4 0.4+0.3
1 1.9 0
6 2.34- 1.0 0.8+0.4
HC-type No of cells Max V gain Max Y gain Total no.
M A D azimuths, elevation angles of 21 AC and 25 PC cells shifted m o r e than 10 ~ f r o m the paired canal plane (46/ 8 0 = 5 8 % , light shaded areas in Fig. 3B, D). Furthermore, elevations of 13 A C and 8 PC cells (21/80 = 26%, dark shaded areas in Fig. 3B, D) shifted m o r e than 10 ~ both f r o m the single and paired canal planes, suggesting convergence of horizontal canal inputs (see Discussion). Type I (positive angle) and II (negative angle) yaw responses were approximately equally prevalent in A C cells (B), but type I responses seemed m o r e prevalent for PC cells (D). The convergent properties o f second-order vestibuloocular neurons are summarized in Table 1. The percentage o f convergence f r o m the horizontal canal is significantly higher in A C neurons (13/32= 41%) than in PC neurons (8/48 = 17 %, chi-square test, p < 0.02). There is no significant difference in percentage of convergence from the other vertical canal between the two groups (3/32 = 9% for A C cells, 3/48--6% for PC cells).
Comparison of response 9ains. Maximal vertical gain and horizontal (yaw) gain with respect to turntable velocity are summarized in Fig. 4A, B for the 80 vertical secondorder vestibuloocular neurons together with the 4 horizontal second-order vestibuloocular neurons. The mean vertical gain for the 80 vertical cells is 2.7 (4- 1.4 SD) spikes/s/deg/s (1.2 for the 4 horizontal neurons). The mean yaw gain is 0.5 (4- 0.5 SD) spikes/s/deg/s for the 80 neurons and 2.2 spikes/s/deg/s for the 4 horizontal neurons. Vertical second-order vestibuloocular neurons that
showed type I and II yaw responses distributed similarly. All horizontal second-order vestibuloocular neurons showed a type I response. Second-order vestibuloocular neurons do not fall into distinct categories as a function of response gain. There was no significant difference in m e a n vertical gain between AC cells (3.04- 1.4 SD spikes/s/deg/s) and PC cells (2.64- 1.3 SD spikes/s/deg/s). N o r was there any significant difference in mean yaw gain between A C cells (0.64-0.6 SD spikes/s/deg/s) and PC cells (0.4:t:0.3 SD spikes/s/deg/s). G a i n values of second-order vestibuloocular neurons with M A D azimuths within the error ranges of individual vertical canals and those with shifted vectors distributed similarly. Table 1 summarizes the mean values of the maximal (4- SD) vertical gain and yaw gain relative to turntable velocity for each second-order vestibuloocular neuron group. There was no significant difference in vertical gain a m o n g cells that did or did not receive horizontal canal convergence.
Response phases. Phase values were mostly constant except near null points as already illustrated in Fig. lB. Since all vertical second-order vestibuloocular neurons were activated during left ear down roll rotation, it is convenient to examine population dynamics during roll rotation. F o r the 80 vertical second-order vestibuloocular neurons the m e a n phase lead relative to left ear down roll was 72 ~ (4- 14 SD). This corresponds to a lag of 108 ~ relative to left ear down angular acceleration. F o r yaw rotations phases of type I responses were measured relative to m a x i m u m leftward excursions, and for type II responses relative to m a x i m u m rightward excursions. The mean p h a s e lead was 73 ~ ( t 3 6 ~ SD) relative to m a x i m u m excursion, corresponding to a lag of 107 ~ relative to acceleration. There was no difference in phase values between A C and PC second-order vestibuloocular neurons during vertical or yaw rotations. The input responsible for shifting the responses of convergent second-order vestibuloocular neurons could reflect otolith inputs. To examine this, response phases (leads relative to head position) were c o m p a r e d during vertical rotation in a plane close to that of each neuron's p r i m a r y canal pair (45 ~ for A C cells and 135 ~ for PC cells) and in the plane orthogonal to that plane (135 ~ for A C cells and 45 ~ for PC cells). Responses produced by the primary canal should be m i n i m u m in the orthogonal plane, thus allowing us to determine whether the dynamics of the convergent signal resembled those of canal or otolith afferents. Figure 4C summarizes the results for second-order vestibuloocular neurons whose M A D azimuths were shifted more than 5 ~ f r o m the individual canal planes. Twelve of the 17 cells had phases in the orthogonal plane lying between 4-45 ~ and 4- 135 ~ like those in the optimal (primary) plane, and gain of these responses was low (not shown). Five cells (filled squares in Fig. 4C) exhibited a position response (phase near 0 ~ or 180 ~ with a significant gain, indicating an otolith input. These results suggest that the majority of the convergence observed in second-order vestibuloocular neurons reflects input f r o m orthogonal canals rather than otolith afferents.
V TYPE I
V TYPE II
HC TYPE I
1 2 3 4 5 6 7 MAX VERT GAIN (sp/s/deg/s)
1 2 3 YAW GAIN (sp/s/deg/s)
PHASE, O R T H O G C A N A L P L A N E
1.5i 1 .5"
-.5 r~= 0.78
-2 -2.5 -3
r ~: .002
~, ~ -.5
0 5 Pitch Gain (sp/sec/deg)
-125 -130 -10
-5 0 5 Yaw Gain (spikes/sec/deg)
Fig. 4. A, B Histograms of maximal vertical (A) and yaw (B) gain during sinusoidal rotation at 0.5 Hz for PC, AC and HC secondorder vestibuloocular neurons. In B, VTYPE I, VTYPE II refer to primarily vertical cells (AC and PC combined) with horizontal type I and type II responses, respectively. HC TYPE I refers to the 4 primarily horizontal cells (all type I responses). C Response phases (relative to peak excitatory position for each neuron, i.e. nose down and to the left for AC cells, nose up and to the left for PC cells) in the optimal (primary) canal plane plotted against phases in the plane orthogonal to the optimal plane for vertical second-order vestibuloocular neurons whose maximal response angles were shifted more than 5 ~ from the primary canal plane. Filled squares are cells with a position response (phase near 0 ~ or 180~ indicating an
.15 .2 ,25 ,3 .35 .4 .45 .5 Vertical Eye Position Sensitivity Pitch Gain otolith input. D Vertical eye position sensitivity plotted against pitch gain for each vertical second-order vestibuloocular neuron which had a significant vertical eye position sensitivity. Positive values indicate excitation to nose-down rotation (abscissa) and upward eye position (ordinate). E Horizontal eye position sensitivity plotted against yaw gain for each vertical second-order vestibuloocular neuron which had a significant horizontal eye position sensitivity and yaw gain. Positive values indicate excitation to noseleft rotation (abscissa) and leftward eye position (ordinate). F Response phase to pitch rotation plotted as a function of the ratio of vertical eye position sensitivity to pitch gain (expressed in spikes/ sec/deg). Fitted regression lines with indicated slope and correlation coefficients are shown in D, F
150 slope = 0.2 r = .07
slope = 1.05 r = .50
: ..... -
"- . " ~ -
slope = 0.01 r = .0t
slope -- -1.63 r = -.84
9 9 "l
0 Fixation Position (deg)
Fig. 5A-F. Eye movement related responses of vertical second-order vestibuloocular neurons. Horizontal and vertical eye position records and cell discharge rates are shown in A, D. B, C and E, F Show discharge rate of each cell plotted against horizontal (B, E) and vertical (C, F) eye position. Lines are fitted linear regressions with slope and correlation coefficient indicated
. . . .
Fixation Position (deg)
Eye movement related responses of vertical second-order vestibulooeular neurons As illustrated in Fig. 5A, activity of many second-order vestibuloocular neurons was related to vertical eye position change induced by visual stimuli while the turntable was stationary. This PC cell's firing rate increased when the eyes moved downward (downward on-direction). Responses were mostly tonic. Discharge rate of this cell is plotted as a function of horizontal (Fig. 5B) and vertical (Fig. 5C) eye position while the animal was fixating a visual object (see Methods). A linear correlation is seen with vertical (C), but not with horizontal (B) eye position with a slope (i.e. eye position sensitivity) of - 1.63 spikes/
s/deg (- indicates discharge rate decreased as the eyes moved downward). Rate position curves of 38 vertical second-order vestibuloocular neurons (11 AC, 27 PC) were examined as in Fig. 5. Nineteen cells (6 AC, 13 PC) showed a significant relationship (/9<0.05) with vertical eye position with either upward or downward on-direction. Linear regressions fitted the data in most cells. Absolute values of vertical eye position sensitivity ranged from 0.1 to 2.4 with a mean of 1.2 (=L 0.7 SD) spikes/s/deg (correlation coefficient of the linear regression= 0.44-0.85). Except for eye position related tonic discharges, no clear burst or pause was observed during saccades. Sixteen of the 38 vertical second-order vestibulo-
r~ < +4-*1. + + § 2 4 7 +
-180 -180 I
"0 9~. -5 -10 -180
. . . .
180Z [9 1~,
Fig. 6A-D. Responses of a horizontal second-order vestibuloocular neuron. A Top trace shows monosynaptic activation by stimulation of the ipsilateral vestibular nerve (first stimulus) and antidromic response to stimulation of the midbrain MLF area (second stimulus). Lower trace shows collision block of antidromic spike by a synaptically evoked preceding spike. B Responses during sinusoidal rotation at 0.5 Hz (TAB) in the horizontal plane (yaw) and vertical planes at different head orientations (0~176 C Phase (re head position) and gain (re head velocity) of responses to rotations at 0.5 Hz in different vertical planes are summarized for this cell. A least squares sinusoid was fitted to the gain curve. D Phase and gain curves during yaw rotations and vertical rotations at 30 ~ head orientation over a wide frequency range (0.2-4.0 Hz)
Table 2. Number of vertical second-order vestibuloocular neurons that had significant eye position sensitivity (p < 0.05) in either vertical or horizontal direction, or in both directions No of neurons
Vert eye pos sensitivity Horiz eye pos sensitivity Vert + Horiz eye pos sens No eye pos sensitivity Total no. tested
3 3 3 2 11
9 6 4 8 27
12 9 7 10 38
o c u l a r n e u r o n s s h o w e d h o r i z o n t a l eye p o s i t i o n sensitivity. T h e cell s h o w n in Fig. 5D i n c r e a s e d firing rate w h e n eyes m o v e d l e f t w a r d (Fig. 5E), a n d no clear c o r r e l a t i o n was o b s e r v e d with vertical eye p o s i t i o n c h a n g e (Fig. 5F). This cell r e s p o n d e d m o s t s t r o n g l y to r o t a t i o n s in the p l a n e o f the left a n t e r i o r canal, b u t also h a d a significant r e s p o n s e to r i g h t w a r d (type II) h o r i z o n t a l r o t a t i o n s . A b s o l u t e values o f h o r i z o n t a l eye p o s i t i o n sensitivity r a n g e d f r o m 0.3 to 2.6 with a m e a n o f 1.1 (4-0.6 S D ) spikes/s/ deg ( c o r r e l a t i o n coefficient, 0.31-0.72). T a b l e 2 s u m m a r i z e s the n u m b e r o f vertical second-
472 TOP VIEW
A HORIZONTAL, MONOSYNAPTIC, ANTIDROMIC n=4 -1
C ANTmROMIC, NOT MONOSYNAPTIC Fig. 7A-C. Normalized M A D vectors of three cell groups (A-C). Top view and front view show the vectors in pitch and roll, and pitch and yaw planes, respectively, as in Fig. 1C
order vestibuloocular neurons that showed eye position sensitivity. Of the 38 cells examined, 12 (3 AC, 9 PC) had eye position sensitivity only in the vertical direction, 9 (3 AC, 6 PC) only in the horizontal direction, and 7 (3 AC, 4 PC) in both directions, suggesting an oblique preferred direction. The elevation orientation angles for eye position sensitivity in these 7 neurons were estimated by calculating the arctangent of horizontal eye position sensitivity divided by vertical eye position sensitivity. The angles ranged from 15~ to 54 ~ with a median of 29 ~ from the vertical direction. There was no significant difference in the percentage of cells that had eye position sensitivity in the vertical, horizontal or oblique direction between AC and PC cells. Firing rate of the remaining 10 cells showed no significant correlation either with vertical or horizontal eye position (Table 2). Eye position sensitivity was not tested for the 4 horizontal second-order vestibuloocular neurons.
Compar&on of eye position sensitivity and other response properties. A significant correlation was observed between eye position sensitivity and MAD direction (Fig. 4D, E). The on-direction for all AC cells that had significant vertical eye position sensitivity was upward (mean eye position sensitivity of 0.71 • 0.20 SD spikes/s/deg), and the on-direction for all vertical eye position sensitive PC cells was downward (mean eye position sensitivity of -1.49• 0.64 SD spikes/s/deg). Twelve/16 neurons which had horizontal eye position sensitivity had significant responses to horizontal rotations. Of these, all three cells
which responded to rightward rotations (type II) had a leftward on-direction for eye position sensitivity. Eight of 9 cells which responded to leftward rotations (type I) had a rightward on-direction. It should be noted that the large correlation coefficient seen in Fig. 4D arises only because of the relationship of the signs of vertical eye position sensitivity and pitch responses. No correlation was observed between absolute values of vertical eye position sensitivity and maximal vertical gain or pitch gain, or between absolute values of horizontal eye position sensitivity and yaw gain. Maximum vertical vestibular gain, when expressed relative to head position, was usually 3-4 times larger than vertical eye position sensitivity for individual second-order vestibuloocular neurons. This suggests that during rotation the cells' responses were dominated by the vestibular input. To further test this hypothesis, phases of the response to pitch rotation (re head acceleration) were plotted as a function of vertical eye position sensitivity divided by pitch gain for all cells with a significant vertical eye position sensitivity (Fig. 4F). This ratio reflects the size of the eye position sensitivity relative to the vestibular sensitivity ( = 0 for pure vestibular response, = 1 for pure eye position response), assuming that during rotation the eye position signal carried by a neuron is equal to that measured while the cat fixates a visual object with no vestibular stimulus. If the eye position signal made a significant contribution to a neuron's response to rotation, it is likely that cells with larger eye position sensitivities would have response phases closer
1 mm Fig. 8A-K. Recording locations and stimulating positions. A - H Summarize reconstructed recording locations of second-order vestibuloocular neurons (large symbols) in 3 cats (K, S, T). The location of some neurons could not be reconstructed confidently and these are excluded. Each cell is represented by a symbol which indicates its response (inset). Filled and open circles indicate anterior canal and posterior canal cells, respectively. Vertical line through a circle indicates cell that received convergent inputs from the orthogonal vertical canal. Horizontal arrow through a circle indicates cell with convergent input from a horizontal canal. Triangles with arrows indicate horizontal neurons with convergent inputs from anterior canal. Also shown with small dots are cells that were
lmm activated monosynaptically from the ipsiIateral vestibular nerve but that were not fired antidromically from the midbrain electrodes in these 3 cats. The location of tip of stimulating electrode in each cat is shown in l-K, along with the location in cat A in which the electrode was misplaced. Abbreviations: BC-brachium eonjunctivum; D-descending vestibular nucleus; INC-interstitial nucleus of Cajal; L-lateral vestibular nucleus; M-medial vestibular nucleus; MLF-medial longitudinal fasciculus; R-red nucleus; RF-restiform body; S in A and B-superior vestibular nucleus; IIIn-oculomotor nucleus; IVn-trochlear nucleus; V-spinal trigeminal tract; VInabducens nucleus; VII facial nerve
474 to eye position. Figure 4F shows that this is not the case - there was no tendency for ceils with large eye position sensitivities to have phases shifted towards - 1 8 0 ~
Miscellaneous cells Properties of horizontal second-order vestibuloocular neurons. As illustrated in Fig. 6, four second-order vestibuloocular neurons responded more strongly to yaw than to vertical rotation, although two exhibited quite strong vertical responses as well (0~ and 45 ~ in B). D summarizes gain (relative to turntable velocity) and phase (re turntable position) curves of responses during yaw rotations and vertical rotations at 30 ~ head orientation over a wide frequency range. Gain increased sharply as the stimulus frequency increased, whereas the phase values showed relatively constant values of about 90 ~ over the wide frequency range. These responses are typical of a canal response (Fernandez and Goldberg 1971) suggesting that this cell received both horizontal and vertical semicircular canal inputs. The MAD azimuth for this cell was 31 ~ (B, C). This response angle cannot be produced by an anterior canal input alone, which suggests that this cell received input from all 3 semicircular canal pairs (see Discussion). MAD vectors of the 4 horizontal second-order vestibuloocular neurons make an angle of 560-76 ~ with the pitch axis in the pitch-yaw plane (front view, Fig. 7A). Second-order, but not antidromic vestibular neurons. During the course of recording, we encountered many vestibular neurons that responded monosynaptically to stimulation of the ipsilateral vestibular nerve but that were not activated antidromically from the midbrain electrode. Spatial properties of a total of 55 such neurons were analyzed in the 3 cats from which second-order vestibuloocular neurons were identified (Fig. 8I-K, cats T, S, K). In addition, responses of 22 neurons were analyzed in one cat in which the stimulating electrode was misplaced (see Methods). Normalized MADs of all these neurons are summarized together in Fig. 7B. The vectors are seen all around the cat, but the major responses were observed near the vectors of the left anterior, posterior (top view), and horizontal (front view) canal. Some responses were seen near the right anterior and posterior canal vectors (type II responses, top view). Some cells also responded to static tilts indicating otolith inputs. Higher-order antidromic cells. Neurons that were antidromically activated from the midbrain electrode, but that were not monosynaptically activated from the ipsilateral vestibular nerve, were also encountered in regions of the vestibular nuclei where large N1 field potentials were recorded. Many of these neurons were recorded at similar depths to other second-order vestibular neurons, and were polysynaptically activated from the vestibular nerve, indicating that they were higher-order vestibular neurons. Normalized MADs of 12 such cells are summarized in Fig. 7C. Some had MADs near that
of the left posterior, anterior, and horizontal canals. Four cells had vertical type II response (top view, i.e. 2 cells near the right posterior and 2 others near the right anterior canal). In two cats (cats S and T, Figs. 2, 10I-J), we systematically searched for all antidromically activated cells in the vestibular nuclei and examined whether or not they were second-order vestibular neurons. A total of 88 antidromically activated cells were encountered, and 66 of them (75 %) were monosynaptically activated from the vestibular nerve. The ratios of such cells were the same in the two cats (30/40 = 75%, 36/48 = 75%). This suggests that second-order vestibuloocular neurons constitute the majority of output neurons sending signals from vestibular nuclei directly to the trochlear or oculomotor nuclei.
Location of recorded cells Reconstructed locations of second-order vestibuloocular neurons recorded in 3 cats (K, S, T) are summarized in Fig. 8 together with their responses. Only those neurons which could be located with confidence are included in the histological reconstruction. Many second-order vestibuloocular neurons were found in the ventrolateral part of the rostral medial nucleus and medial part of the descending nucleus (D, E, F). Some were found in the lateral and superior vestibular nuclei (A-C). We did not explore the superior vestibular nucleus in cat K. Although AC- and PC-type second-order vestibuloocular neurons were intermingled, there was a tendency for neurons with similar canal inputs to be segregated. Cells with convergent canal inputs were found in all the major vestibular nuclei.
The primary goal of this study was to determine the role of second-order vestibuloocular neurons in the spatial transformation that occurs in the VOR between semicircular canal input and extraocular muscle activation. The first issue that must be discussed is whether the neurons we studied were in fact vestibuloocular relay neurons.
Identification of second-order vestibuloocular neurons and their dynamic properties and location in vestibular nuclei The second-order vestibuloocular neurons identified in cats S and T were activated at a median stimulus intensity of 50 gA from antidromic stimulating electrodes located close to the MLF at the levels of the trochlear and oculomotor nuclei. Studies employing intracellular labeling of axonal arbors indicate that virtually all secondorder vestibular neurons coursing through these sites are vestibuloocular relay neurons (Graf et al. 1983 ; Graf and Ezure 1986; McCrea et al. 1987). The same studies indicate that the great majority of these vertical vestibulo-
475 ocular relay neurons also send axon collaterals anteriorly to the region of the interstitial nucleus of Cajal (INC), where our stimulating electrode was located in cat K. Conversely, there has been no report of second-order vestibular neurons that project to the INC without terminating in the extraocular motor nuclei. It is therefore likely that all of the neurons which we identified as second-order vestibuloocular neurons terminated on extraocular motoneurons. The dynamic properties of the signals carried by our second-order vestibuloocular neurons are also consistent with their identification as vestibulo-ocular relay neurons. Second-order, morphologically identified vestibuloocular relay neurons in the cat have vertical eye position sensitivity but do not burst or pause during saccades (Yoshida et al. 1981). This closely resembles the behavior of our second-order vestibuloocular neurons, except eye position sensitivities were larger in Yoshida et al.'s study. Responses of vertical vestibulo-ocular relay neurons found in the ascending MLF in the rhesus monkey (TVP cells) also resemble that of our second-order vestibuloocular neurons. The TVP cells carry a head velocity signal with an average strength of 1 spike/s/~ and a signal related to vertical eye position with an average strength of 2.5 spikes/sec/~ (King et al.. 1976; Pola and Robinson 1978; Tomlinson and Robinson 1984; Chubb et al. 1984; McCrea et al. 1987). Second-order vestibuloocular neuron behavior in our study is similar, except that eye position sensitivity in our population averaged 1.2 spikes/s/~ with a range from 0 to 2.4 spikes/s/~ and neurons rarely exhibited a change in discharge related to saccadic eye movements. Thus the dynamic behavior of our second-order vestibuloocular neurons is quite similar to that of vestibulo-ocular relay neurons identified by other investigators using a variety of techniques. We have also been able to relate vertical eye position sensitivity to head velocity sensitivity and demonstrate that the two are always synergistic so that the eye position signal will enhance the response of the neuron during the VOR. What is unique in our data is the variable amount and direction of eye position sensitivity in vertical vestibuloocular relay neurons, such as horizontal eye position sensitivity in vertical second-order vestibuloocular neurons. We do not have at present a functional interpretation of the presence of horizontal eye position sensitivity on vertical second-order vestibuloocular neurons. Although many vertical second-order vestibuloocular neurons exhibited vertical eye position related discharges (Fig. 5), their responses to rotation are characterized by phases which are closer to head (and therefore eye, since the phase of the VOR is nearly compensatory) velocity than position (phase leads head position by 72 ~ see Results). Delgado-Garcia et al. (1986) described the responses ofabducens motoneurons during sinusoidal rotations in alert cats. At 0.5 Hz, neuron firing rate is closer to eye position than eye velocity (neuron leads eye position by approximately 25 ~ their Fig. 12). Our findings using multi-unit recordings from the abducens nucleus (Baker et al. 1988) are quantitatively similar for rotations at 0.8 Hz. Vestibuloocular reflex pathways convert the head velocity signal carried by
primary vestibular afferents (Fernandez and Goldberg 1971) into this near eye position signal carried on extraocular motoneurons. Second-order vestibuloocular neurons do not contain the appropriate temporal signals to account for those present on oculomotoneurons during 0.5 Hz rotations. In addition, midbrain lesions which spare the direct vestibular nucleus-oculomotor nucleus projection produce a clear impairment in VOR dynamics (Anderson et al. 1979; Fukushima 1987). These results indicate that an additional temporal conversion is necessary for the VOR beyond that which has occurred at the level of second-order vestibular nucleus neurons. Although we cannot tell from the present study where such a temporal conversion takes place, our results that vertical second-order vestibuloocular neurons constitute the majority of vestibular neurons projecting to the trochlear or oculomotor nuclei suggest that structures outside the vestibular nuclei possibly even the oculomotoneurons themselves - are involved in a temporal conversion of vertical semicircular canal signals. Intracellular HRP studies in cats and monkeys (Graf et al. 1983; Graf and Ezure 1986; McCrea et al. 1987) have shown that vertical canal-related second-order vestibular neurons project not only to the trochlear and oculomotor nuclei but also to the INC and the surrounding midbrain reticular formation. Single unit studies in alert animals have shown that neurons in the INC region carry a vertical eye position signal (King et al. 1981; Kaneko 1986; Fukushima et al. 1990) Bilateral lesions made in the INC region resulted in a decrease of the normal phase lag (with respect to head velocity) of the VOR and in an inability to maintain vertical eye position after saccades (Anderson et al. 1979; Fukushima 1987). These observations suggest that the midbrain reticular formation in the INC region is involved in a temporal conversion of vertical semicircular canal signals. Neurons in these areas are reported to project not only to the oculomotor nuclear complex but also to the vestibular nuclei (e.g. Nakao et al. 1987; Fukushima 1987, for review). A majority of our vertical second-order vestibuloocular neurons were located in the ventrolateral part of the medial vestibular nucleus and the medial part of the descending vestibular nucleus (Fig. 8A-H). These are the areas which are reported to contain second-order vestibular neurons projecting to the oculomotor nuclei (Uchino et al. 1981, 1982; McCrea et al. 1987). All these results, therefore, indicate that the antidromically identified second-order vestibular neurons we recorded were vestibulo-ocular relay neurons. Although some vertical canal-responding cells were recorded in the superior vestibular nucleus in the present study, our tracks did not cover this nucleus systematically enough to make a quantitative description of the properties of vertical secondorder vestibuloocular neurons in this nucleus (see below).
Spatial properties of responses of second-order vestibuloocular neurons to rotation Our multiplanar vestibular testing typically allowed us to characterize the MAD vectors of second-order vestibu-
476 loocular neurons within a standard error of + 3.4~ As illustrated in Figs. 2 and 9A, the second-order vestibuloocular neuron population divided into 3 non-overlapping populations, which received their primary input from anterior, posterior and horizontal semicircular canals. Within each population the MADs of individual neurons varied quite widely, suggesting that some of the neurons may have received convergent input from multiple semicircular canals. Except for inhibitory commissural input from the coplanar canal of the contralateral labyrinth, which is common (Kasahara and Uchino 1974), the presence of convergence of canal input upon secondorder vestibular neurons is a matter of some controversy. Electrophysiological studies in anesthetized animals (Wilson and Felpel 1972 in the pigeon; Kasahara and Uchino 1974 in the cat) found little evidence for such convergence whereas significant convergence has been described in alert cats using electrical (Markham and Curthoys 1972) or natural vestibular stimuli (Baker et al. 1984a, b; Curthoys and Markham 1971). None of these studies specifically identified vestibulo-ocular relay neurons, leaving open the question of whether such neurons receive significant convergent input. Our identification of second-order vestibuloocular neurons that received convergent input depends upon comparison of the MADs of those neurons with the MADs of semicircular canals estimated from data of Blanks et al. (1972). Their data indicate that the orientation of canals in different cats is very consistent, with standard errors of the angles between canal planes and stereotaxic planes less than 1.4~ in every case. The azimuth angles (13 in Fig. 2) of the MADs of our secondorder vestibuloocular neurons had a typical standard error of 3.4 ~. This suggests that angles greater than 5~ between canal planes and MADs are due to canal convergence rather than to measurement errors or variations in canal geometry. Since we cannot be sure how much commissural input each neuron received from the coplanar canal, it was necessary to consider differences from MADs of either the ipsilateral canal or of the coplanar canal pair, which widened the limits somewhat. We therefore settled on the conservative criterion of requiring a neuron's MAD to lie more than 10~ from the nearest primary (single or coplanar) canal plane, as measured from azimuth angles and inclinations from the horizontal plane. With this criterion 3/48 PC cells, 3/32 AC cells and 3/4 HC cells received convergent input from both pairs of vertical canals. Convergence between horizontal and vertical canals was more common: 8/48 PC and 13/32 AC cells received convergent horizontal canal input while all 4 HC cells received vertical canal input. One might still argue that apparent convergence resulted because the canal orientations in one of our cats lay far from the population mean described by Blanks et al. (1972). If this were the case, however, all MADs, in that animal should be shifted in the direction of this canal divergence. No such effect is seen in Fig. 2, which indicates that the distribution of MADs in the three cats was similar. Furthermore, neurons in the same cat exhibited MADs that were shifted more than 10~ from the estimated canal MADs in both the positive and negative
direction whereas a misorientation of canal planes could shift the MADs of neurons in an individual animal in one direction only. Thus our data establish that a significant amount of convergence between non-coplanar canals occurs in the population of second-order vestibuloocular relay neurons (overall, 26/84 second-order vestibuloocular neurons received input from two canal pairs and 5/84 from all three canal pairs). As shown by Fig. 7B and data of Baker et al. (1984a, b), however, the probability of such convergence in this population is less than in the second-order neuron population as a whole. Presumably more convergence is required in other systems, such as the vestibulo-collic reflex where muscle MADs lie much further from canal planes than do the MADs of extraocular muscles (Baker et al. 1985). The multi-canal convergence described in this study is qualitatively consistent with results obtained previously using intraaxonal recording in decerebrate cats (Peterson et al. 1987). Neither study indicates the neural substrates of this convergence. Given the lack of direct convergent input from multiple canals upon second-order neurons in electrophysiological studies (Wilson and Felpel 1972; Kasahara and Uchino 1974; Uchino et al. 1986), it is likely that input from the secondary canal(s) reaches second-order vestibuloocular neurons via polysynaptic pathways. We cannot make a quantitative description of convergence of otolith inputs upon second-order vestibuloocular neurons in the present study. We feel that a majority of second-order vestibuloocular neurons received little otolith input, since many cells had response phases near head velocity and only a few cells showed phases near position during rotations in the vertical plane orthogonal to the canal from which the cell received its strongest input (Fig. 4C). However, this result does not eliminate the possibility that the cells may have received an otolith input which was aligned spatially with their canal input. Kasper et al. (1988) have recently reported that many vestibular nucleus neurons with responses indicating both canal and otolith inputs have spatial properties which do not change with frequencies of rotation between 0.02-1.0 Hz. This suggests that vestibular nucleus neurons do have spatially aligned canal and otolith inputs.
Role of second-order vestibuloocular neurons in spatial transformation within the VOR To appreciate the role of second-order vestibuloocular neurons in the spatial transformation that occurs between MADs of semicircular canals and those of extraocular muscles, one must compare the MADs of these neurons to both canal and muscle MADs as in Fig. 3. The MADs of oblique extraocular muscles, indicated there by vertical arrows, are from our measurements of EMG responses of the 12 extraocular muscles in the decerebrate cat to a series of rotations very like that employed in this study (Peterson et al. 1984). It is clear in the figure that a significant number of PC cells have MADs centered around the MAD of the superior oblique
477 muscle. Not all of these cells fell in the group classified as receiving convergent input from more than one orthogonal vertical semicircular canal pair, because the plane o f the left posterior canal (as opposed to the plane of the left posterior-right anterior canal pair) is shifted in the direction of the superior oblique. Thus PC cells with MADs near that of the superior oblique could either receive convergent input from both pairs of vertical canals or selective input from the left posterior canal without commissural inhibition from the right anterior canal. In either case a signal appropriate to drive the superior oblique is being generated i n the vestibular nuclei by selective action of vertical semicircular canal(s) on second-order vestibuloocular neurons. In the case of AC cells, the plane o f the unpaired left anterior canal is shifted away from the M A D of the inferior oblique muscle so that signals appropriate to drive this muscle cannot arise from selective input from the unpaired canal. While we recorded only 3 AC second-order vestibuloocular neurons in the superior vestibular nucleus, 2 o f them had MADs near that of the inferior oblique. This suggests that second-order vestibuloocular neurons with the convergent input required to generate signals for the inferior oblique may be concentrated in this nucleus. Such a possibility would be consistent with data of McCrea et al. (1987) who reported that presumed excitatory AC second-order vestibuloocular neurons located in the squirrel monkey's medial vestibular nucleus projected only to the bilateral superior rectus m o t o r pools whereas presumed inhibitory AC second-order vestibuloocular neurons located in the superior vestibular nucleus projected to ipsilateral inferior rectus and inferior and superior oblique nuclei. Taken as a whole, our data indicate that signals with spatial properties appropriate to activate oblique muscles are present at the lvel of second-order vestibulo-ocular relay neurons. In contrast, none of our second-order vestibuloocular neurons had signals shifted far enough from vertical canal planes to overlap the MADs of the superior and inferior rectus muscles, which would lie at 20 ~ and 157 ~ on Fig. 3, respectively (Peterson et al. 1984). How, then, might the signals required to activate vertical rectus muscles arise? One possibility is that they could arise because some second-order vestibuloocular neurons project bilaterally to vertical rectus m o t o r pools (Uchino et al. 1980; G r a f et al. 1983; G r a f and Ezure 1986; McCrea et al. 1987). Considering the average M A D vector directions of our AC and PC cells in the pitch-roll plane, a signal appropriate to activate vertical rectus motoneurons could be produced if the secondary projections of these neurons to ipsilaterally located superior and inferior rectus motoneurons had 44% the strength of their primary projections to contralaterally located motoneurons. The alternative is that this transformation requires the action of less direct pathways such as those involving third or higher order vestibular or interstitial nucleus of Cajal neurons, which have been shown to exhibit MADs close to those of vertical rectus motoneurons (Fukushima et al. 1990). There is little doubt that these neurons participate in the VOR and that they are necessary to explain the large eye position sensitivity of
oculomotor neurons (Pola and Robinson 1978). The question of whether they are necessary to explain spatial properties of the VOR as well cannot be answered until further quantitative information on the effect of diverging projections of second-order vestibuloocular neurons is available. Many vertical second-order vestibuloocular neurons also received horizontal canal inputs. However, since type I and II responses were almost equally prevalent (Fig. 2D), it seems that these horizontal responses would cancel at the extraocular motoneurons during the normal vertical VOR. The responsiveness of AC and PC cells to horizontal rotation may function in different behavioral conditions such as cross axis VOR adaptation (Schultheis and Robinson 1982; Peterson et al. 1986; Perlmutter et al. 1988). When alert cats are trained by combining horizontal rotation with a vertical optokinetic stimulus, a vertical VOR response to horizontal rotation develops within one hour of training. While recording from vestibular neurons during this plastic, adaptive change in VOR direction, Perlmutter et al. (1988) observed a significant increase in the response of 2/4 vertical second-order vestibuloocular neurons to horizontal rotation. The direction of the change in each case was appropriate to contribute to the plastic change in VOR direction. This suggests an involvement of these neurons in transformation of horizontal semicircular canal signals into an adaptively learned vertical VOR component. This finding is also consistent with our hypothesis that second-order vestibuloocular neurons play a primary role in determining the direction of vestibulo-ocular eye movements.
Acknowledgement. Supported by NIH grants EY05289, EY06485, and EY07342.
References Anderson JH, Precht W, Pappas C (1979) Changes in the vertical vestibulo-ocular reflex due to kainic acid lesions of the interstitial nucleus of Cajal. Neurosci Lett 14:259-264 Baker J, Goldberg J, Hermann G, Peterson B (1984a) Optimal response planes and canal convergence in secondary neurons in vestibular nuclei of alert cats. Brain Res 294:133-137 Baker J, Goldberg J, Herman G, Peterson B (1984b) Spatial and temporal response properties of secondary neurons receiving convergent input in vestibular nuclei of alert cats. Brain Res 294:138-143 Baker J, Goldberg J, Peterson B (1985) Spatial and temporal response properties of the vestibulocollic reflex in decerebrate cats. J Neurophysiol 54:735-756 Baker JF, Wickland C, Goldberg J, Peterson B (1988) Motor output to lateral rectus in cats during the vestibulo-ocular reflex in three-dimensional space. Neuroscience 25 : 1-12 Billoto G, Goidberg J, Peterson BW, Wilson VJ (1982) Dynamic properties of vestibular reflexesin the decerebrate cat. Exp Brain Res 47:343-352 Blanks RHI, Curthoys IS, Markham CH (1972) Planar relationships of semicircular canals in the cat. Am J Physiol 223 : 55-62 Blanks RHI, Estes MS, Markham CH (1975) Physiologic characteristics of vestibular first-order canal neurons in the cat. II. Response to constant angular acceleration. J Neurophysiol 38:1250 1267
478 Chubb MC, Fuchs AF, Scudder CA (1984) Neuron activity in monkey vestibular nuclei during vertical vestibular stimulation and eye movements. J Neurophysiol 52:724-742 Curthoys IS, Markham CH (1971) Convergence of labyrinthine influences on units in the vestibular nuclei of the cat. I. Natural stimulation. Brain Res 35:469-490 Delgado-Garcia JM, Del Pozo F, Baker R (1986) Behavior of neurons in the abducens nucleus of the alert cat. 1. Motoneurons. Neuroscience 17:929-952 Duensing F, Schaefer K-P (1958) Die Aktivitfit einzelner Neurone im Bereich der Vestibulariskerne bei Horizontal-Beschleunigungen unter besonderer Beriicksichtigung des vestibularen Nystagmus. Arch Psychiat Nervenkr 198:225 252 Estes MS, Blanks RHI, Markham H (1975) Physiologic characteristics of vestibular first-order canal neurons in the cat. I. Response plane determination and resting discharge characteristics. J Neurophysiot 38:1232-1249 Ezure K, Graf W (1984) A quantitative analysis of the spatial organization of the vestibulo-ocular reflexes in lateral- and frontal-eyed animals. I. Orientation of semicircular canals and extraocular muscles. Neuroscience 12:85-93 Fernandez C, Goldberg JM (1971) Physiology of peripheral neurons innervating semicircular canals of the squirrel monkey. II. Response to sinusoidal stimulation and dynamics of peripheral vestibular system. J Neurophysiol 34:661-675 Fukushima K (1987) The interstitial nucleus of Cajal and its role in the control of movement of head and eyes. Prog Neurobiol 29:107-192 Fukushima K, Harada C, Fukushima J, Suzuki Y (1990) Spatial properties of vertical eye movement-related neurons in the region of the interstitial nucleus of Cajal in awake cats. Exp Brain Res 79 : 25-42 Graf W, Ezure K (1986) Morphology of vertical canal related second order vestibular neurons in the cat. Exp Brain Res 63 : 35-48 Graf W, McCrea RA, Baker R (1983) Morphology of posterior canal related secondary vestibular neurons in rabbit and cat. Exp Brain Res 52:125-138 Kaneko CRS (1986) Location and discharge of saccade-related neurons in alert cat and monkey: developments in oculomotor research. Satellite Meeting of Internat Union of Physiological Sciences (Abstr) Kasahara M, Uchino Y (1974) Bilateral semicircular canal inputs to neurons in cat vestibular nuclei. Exp Brain Res 20 : 285-296 Kasper J, Schor RH, Wilson VJ (1988) Response of vestibular neurons to head rotations in vertical planes. I. Response to vestibular stimulation. J Neurophysiol 60:1753-1764 King WM, Fuchs AF, Magnin M (1981) Vertical eye movementrelated responses of neurons in midbrain near interstitial nucleus of Cajal. J Neurophysiol 46:549-562 King WM, Lisberger SG, Fuchs AF (1976) Responses of fibers in medial longitudinal fasciculus (MLF) of alert monkeys during horizontal and vertical conjugate eye movements evoked by vestibular or visual stimuli. J Neurophysiol 39:1135 1149 Markham CH, Curthoys IS (1972) Convergence of labyrinthine influences on units in the vestibular nuclei of the cat. II. Electrical stimulation. Brain Res 43:383-396 McCrea RA, Strassman A, Highstein SM (1987) Anatomical and physiological characteristics of vestibular neurons mediating the vertical vestibulo-ocular reflexes of the squirrel monkey. J Comp Neurol 264:571-594
Nakao S, Shiraishi Y, Oikawa T (1987) Vertical eye movementrelated neurons in the cat mesodiencephalic junction: synaptic input from and output to the vestibular nuclei. Neurosci Res Suppl Abstr 5 : $73 Perlmutter SI, Fukushima K, Peterson BW, Baker JF (1988) Spatial properties of second order vestibuloocular relay neurons in the alert cat. Soc Neurosci Abstr 14:331 Peterson BW, Baker JF, Goldberg J, Wickland C (1984) Kinematic organization of cat vestibulo-ocular reflex (VOR). Soc Neurosci Abstr 10:162 Peterson BW, Baker JF, Wickland C (1986) Plastic changes in cervicoocular and vestibuloocular reflexes elicited by labyrinthine lesions or altered visual feedback. In: Keller EL, Zee DS (eds) Adaptive processes in visual and oculomotor systems: advances in the biosciences, Vol 57. Pergamon Press, Oxford, pp 399-408 Peterson BW, GrafW, Baker JF (1987) Spatial properties of signals carried by second order vestibuloocular relay neurons in the cat. Soc Neurosci Abstr 13:1093 Pola J, Robinson DA (1978) Oculomotor signals in medial longitudinal fasciculus of the monkey. J Neurophysiol 41:245-259 Robinson DA (1970) Oculomotor unit behavior in the monkey. J Neurophysiol 33 : 393-404 Robinson DA (1982) The use of matrices in analyzing the threedimensional behavior of the vestibulo-ocular reflex. Biol Cybern 46 : 53-66 Schultheis LW, Robinson DA (1982) Directional plasticity of the vestibulo-ocular reflex in the cat. In: Cohen B (ed) Vestibular and oculomotor physiology. Internat meeting Barany Society. New York Academy of Sciences, pp 20-30 Shimazu H, Precht W (1965) Tonic and kinetic responses of cat's vestibular neurons to horizontal angular acceleration. J Neurophysiol 28 : 991-1013 Tomlinson RD, Robinson DA (1984) Signals in vestibular nucleus mediating vertical eye movements in the monkey. J Neurophysiol 51:1121-1136 Uchino Y, Hirai N, Suzuki S, Watanabe S (1981) Properties of secondary vestibular neurons fired by stimulation of ampullary nerve of the vertical, anterior or posterior, semicircular canals in the cat. Brain Res 223:273-286 Uchino Y, Hirai N, Suzuki S (1982) Branching pattern and properties of vertical- and horizontal-related excitatory vestibuloocular neurons in the cat. J Neurophysiol 48:891-903 Uchino Y, Ichikawa T, Isu N, Nakashima H, Watanabe S (1986) The commissural inhibition on secondary vestibulo-ocular neurons in the vertical semicircular canal systems in the cat. Neurosci Lett 70:210-216 Uchino Y, Suzuki S, Watanabe S (1980) Vertical semicircular canal inputs to cat extraocular motoneurons. Exp Brain Res 41 : 45 53 Wilson VJ, Felpel LP (1972) Specificity of semicircular canal input to neurons in the pigeon vestibular nuclei. J Neurophysiol 35 : 253-264 Wilson VJ, Peterson BW, Fukushima K, Hirai N, Uchino Y (1979) Analysis of vestibulocollic reflexes by sinusoidal polarization of vestibular afferent fibers. J Neurophysiol 42:331-346 Yoshida K, Berthoz A, Vidal PP, McCrea R (1981) Eye-movementrelated activity of identified second order vestibular neurons in the cat. In: Fuchs AF, Becket W (eds) Progress in oculomotor research: developments in neuroscience, Vol 12. Elsevier, North-Holland, pp 371-378