INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Vestibular reflexes originate in hair cells located either in the cristae of semicircular canals that encode angular head acceleration or in the maculae of otolith organs that encode gravity and linear head acceleration. During locomotion, angular and translational accelerations of the head activate vestibular afferent inputs that initiate, e.g., compensatory ocular reflexes. Semicircular canal-related inputs generate angular vestibuloocular reflexes (AVORs) and otolith-related inputs generate linear vestibuloocular reflexes (LVORs). Both reflexes are often co-activated, and their spatial and dynamic properties can be assumed to be co-adapted. However, the neural organization of the AVOR is still far better understood than that of the LVOR. Semicircular canal inputs are mediated by excitatory and inhibitory second-order vestibular neurons and by motoneurons to those extraocular muscles that are roughly aligned with their respective semicircular canal. These principal connections of the classic three-neuron reflex arc (Lorente de Nó 1933; Szentagothai 1943, 1950) are supplemented by auxiliary connections from other semicircular canal organs. The combination of these inputs is spatially adequate to correct for geometric misalignments between the semicircular canals and the corresponding extraocular muscles, as predicted by theory (Ezure and Graf 1984; Pellionisz 1985; Robinson 1982) and demonstrated with natural semicircular canal stimulation (Baker and Peterson 1991; Graf et al. 1993; Pantle and Dieringer 1998). In essence, the central organization of the AVOR takes the pulling directions of extraocular muscles into account by species-specific convergence patterns of semicircular canal signals (Pantle and Dieringer 1998).

The neural organizations of the LVOR and of the AVOR appear to differ in anatomical and functional terms. In cats a three-neuron otolith-ocular reflex arc, a basic feature of the AVOR, was neither detected for utriculoocular (Uchino et al. 1996) nor for sacculoocular pathways (Isu et al. 2000). In addition, compensatory eye movements evoked by horizontal or vertical linear accelerations (Borel et al. 1988; Cheung et al. 1998; Hess and Dieringer 1991; Hess et al. 1984; Paige 1989) indicate a relatively small contribution by sacculoocular pathways when compared with the contribution by utriculoocular pathways. However, so far no studies with natural stimulation protocols exist that relate these LVOR responses with the local origin on the utricular macula, except for otolith-related abducens nerve responses in the frog during horizontal linear acceleration (Wadan and Dieringer 1994). In that study a functional cluster of utricular hair cells located medially with respect to the striola in a fan-like sector on the contralateral utricle was characterized. Similar fan-like sectors for the vertical extraocular muscles were predicted to be oriented perpendicularly with respect to the planes of the vertical semicircular canals (Hess and Dieringer 1991; Szentagothai 1952). But so far, no data exist that support or falsify this assumption.

Another organizational difference between the LVOR and the AVOR concerns the sensorimotor coordinate transformation. In analogy to the AVOR, such a coordinate transformation would require in the case of the LVOR a convergence of principal and auxiliary connections activated by horizontal and vertical otolith afferent inputs. The small LVOR evoked by vertical linear acceleration (see above) suggests that such a convergence is functionally less important for the LVOR than for the AVOR. In fact, these small LVOR responses could have even originated from the utricle and not from the saccule (or from the lagena in the case of anurans), for instance if the plane of the utricular macula were pitched or rolled with respect to the gravity vector. To investigate these possibilities, we studied the spatial response properties of three different extraocular motor nerves. Dynamic linear and dynamic angular accelerations were applied separately. For linear acceleration we used horizontal, vertical, or combinations of horizontal and vertical linear accelerations delivered in a ramp-like manner. For angular acceleration the frog was oscillated about an earth-vertical rotation axis. The vector directions of maximal sensitivity of the selected motor nerves were determined by altering the frog's static head position systematically in yaw, pitch, and roll before each test. Preliminary results had been published in abstract form (Rohregger and Dieringer 1999a,b).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animal preparation

For these experiments 25 adult grass frogs (Rana temporaria) were used. All surgical procedures were performed under general anesthesia (0.1% MS-222, 3-aminobenzoic acid ethyl ester dissolved in water). The day before the experiment the forebrain and parts of the diencephalon were disconnected from the rest of the brain by electrocoagulation through an opening drilled in the palatine bone. The inferior rectus and the inferior obliquus muscle nerves on the left side were dissected free through the mouth by opening the soft orbita and splitting the levator bulbi muscle. To gain access to the abducens nerve a small hole was drilled in the palatine bone. Two bone screws were attached to the same bone for a later fixation of the electrodes with dental cement. Details of the procedures of surgery and of electrode fixation for mechanically stable long-term recordings had been described in detail by Pantle et al. (1995) and by Pantle and Dieringer (1998). The day of the recording session the frog was immobilized by an intralymphatic injection of tubocurarine (about 0.03 mg). At the end of an experiment the frog was killed by an overdose of MS-222. Permission for these experiments was granted by Regierung von Oberbayern (211-2531-98/99).

Stimulation

The frog was placed in a small box with the maxilla oriented parallel to the ground plate of the box. Body and eyes were covered with moist gauze to prevent desiccation and vision. The box was mounted on a two-axes gimbal system that allowed the free positioning of the frog in pitch and roll with the head in the center of rotation (Fig. 1A). The gimbal system was mounted either on a two-axes angular accelerator (Acutronic) or on a platform that could be turned on a linear sled (Tönnies; Fig. 1B). To avoid dynamic stimulation of otolith organs during angular acceleration, the axis of rotation was always earth-vertical. If not stated otherwise, the position of the frog's head was pitched up by 20° during angular acceleration to bring the horizontal semicircular canals parallel to the earth-horizontal plane (standard head position according to Blanks and Precht 1976). During linear acceleration, however, the position of the frog's head was pitched up by 15° to produce data that were compatible with the results from earlier studies (Wadan and Dieringer 1994). At the beginning of each experiment, we searched for a stimulus intensity that evoked strong but unsaturated multiunit responses in the recorded nerve. For linear acceleration, usually a frequency of 0.33 or 0.5 Hz and amplitudes between ±4.95 and ±9.85 cm (corresponding to accelerations between 0.043 g and 0.075 g) were selected. For angular acceleration we used a frequency of 0.2 Hz and velocities between ±5 and ±15°/s.

View larger version (57K): [in this window] [in a new window]   Fig. 1. Gimbal system supporting the frog box that was mounted on the platform of a sled for linear or on a turntable for angular acceleration. A: the gimbal system allowed a static positioning of the frog (covered by moist gauze) about the pitch and/or the roll axis. B: the platform on the sled allowed to orient the frog's head with respect to the direction of sled motion over 360° in the horizontal plane (curved arrows). C: the sled could be inclined 90° with respect to earth horizontal. The angle of sled inclination (30° in C) was compensated by an inclination of the platform together with the frog for combinations of horizontal and vertical accelerations. Thereby, the standard head position in space was maintained at all angles of sled inclination.

For a given extraocular muscle nerve the maximal activation direction (MAD) was determined for angular as well as for linear head accelerations in the same individuals. To this end we employed the null-point technique (Estes et al. 1975) by altering systematically the orientation of the static head position before each test and calculated the vectors of the angular or translational MADs and their spatial orientation in head or canal coordinates as described by Pantle and Dieringer (1998). The stimulus orientation for minimal responses in a given eye muscle nerve was determined with different protocols for linear and for angular acceleration. In a first series of experiments sinusoidal linear acceleration in the horizontal plane was employed. Before each test the head was reoriented in a 10 or 15° step on the sled (Fig. 1B) over a range of 180° (0° corresponded to an acceleration along the body length axis, ±90° to an acceleration along the interaural axis). In the vicinity of head positions with minimal responses, the step size was reduced to 5° (see Fig. 4A). At least 1 min intervened between a change in head orientation and the onset of stimulation.

Whereas the exact orientation of the MAD for horizontal linear acceleration was determined after the experiment, an approximation (up to about 15° precise) was made during the ongoing experiment. This approximation was required to optimize the orientation of the frog on the sled for a second and a third series of experiments in which we searched for response components evoked by vertical linear acceleration. In the second series of experiments the frog's head was pitched before each vertical linear acceleration test in 10° steps about an axis that was perpendicular to the approximated MAD in the horizontal plane. Static head positions ranged between 90° (horizontal MAD pointing upward) and 90° (horizontal MAD pointing downward). The interpretation of the results obtained with this stimulus protocol remained ambiguous and made a third series of experiments necessary. In this third series, the head was oriented such that the direction of the acceleration was again along the approximated horizontal MAD, but now the track of the sled was inclined stepwise (by 10°) from horizontal to vertical (Fig. 1C). Thereby, combinations of horizontal and vertical linear accelerations were delivered in a ramp-like manner. With a tilting platform on the sled, it was possible to compensate the inclined static position of the frog in space such that the head remained in standard position during each test (see Fig. 1C). It is important to note that the static orientation of the head (and of the otolith organs) with respect to gravity changed in the second but not in the third series of experiments. Since the relative orientation of the acceleration with respect to head position remained the same, the dynamic stimulation of the otolith organs was the same in both series of experiments.

Data processing

Multiunit nerve recordings were amplified, filtered (band-pass 300-1,500 Hz, Krohn Hite 3550; Fig. 2B), rectified (Fig. 2C), and displayed on an oscilloscope together with the upper and lower trigger levels of a spike amplitude window discriminator (Mentor, N-750). The lower level was set above the background noise and below the peak of small, spontaneously occurring action potentials in the nerve. Normalized action potentials (Fig. 2D) were stored on the computer together with the position of the sled or of the velocity of the turntable (CED 1401 and Spike 2, Cambridge Electronics Design; Fig. 2, A-D). Further data analysis was performed off-line. As a measure of nerve activity, the instantaneous firing rate was calculated from reciprocal interspike intervals and smoothed with a uniform average filter (0.3 s corresponding to a corner frequency of 1.66 Hz; Fig. 2E). The responses to several (4-20, not necessarily consecutive) cycles of oscillations were averaged after the elimination of blink-related bursts. If the response exhibited one peak, a sine wave [a · sin ( · t  ) + b] was fitted to the modulated part after having removed electronically the unmodulated part. The parameter  was fixed to the stimulus frequency. The parameters a and  were fit values representing the response magnitude and the phase value, respectively. The parameter b represented the offset value. If the response exhibited two peaks per stimulus cycle (see Fig. 4A), a sine wave as described above was fitted, however, to each of the two response half cycles separately. In the presence of two response peaks per angular oscillation cycle, we subtracted the amplitude of the first fitted sine wave from the amplitude of the second fitted sine wave to determine the null-point. The two response peaks per horizontal linear oscillation cycle were fitted separately by a sine wave as described above, but the amplitudes of the fitted sine waves were not subtracted. The range of head positions over which two response peaks occurred during horizontal linear acceleration allowed us to define the width of the opening of a functional sector on the utricular macula, as introduced by Wadan and Dieringer (1994). The phase of the local maximum of the response was related in case of linear acceleration to the phase of maximal sled acceleration (see right side in Fig. 2E) and to peak table acceleration in case of angular acceleration. Positive phase values indicate that the response was lagging sled or table acceleration. The depth of modulation of the responses varied between different preparations, mainly because of the different numbers of axons recorded from. To allow a comparison of data from different individuals, we normalized the responses by taking the largest response of a given nerve in a series of experiments as 100%. Averages from a group of animals were expressed by mean values and SDs.

View larger version (26K): [in this window] [in a new window]   Fig. 2. Multiunit responses in the left inferior obliquus eye muscle nerve during sinusoidal horizontal linear acceleration on a sled. A: orientation of the frog on the sled and sled position (frequency of oscillation 0.5 Hz; R and L for right and left). B-D: action potentials before (B) and after rectification (C) and normalization (D). E: smoothed instantaneous discharge rate. The phase difference between peak sled acceleration (long vertical dashed lines) and peak nerve discharge rate (short vertical dashed line in E) is indicated by a double arrow and .

MAD directions were calculated by the use of the null-point technique: minimal responses at certain head orientations were characterized by the presence of two response peaks per stimulus cycle with equal amplitudes (e.g., at 25° in Fig. 4A). The difference between these two peaks was null, and the corresponding head position characterized the so-called null-point. Null-points in the pitch and in the roll planes were determined from sine waves fitted to the averaged population values according to the least-squares method. The cross products of the two vectors of these null-points allowed the calculation of the MAD vector for angular accelerations in head-fixed coordinates, since the MAD and the null-point vectors are orthogonal to each other. In a second step we then calculated the corresponding values in semicircular canal coordinates (based on results by Blanks and Precht 1976, see Table 2; for details see Appendix in Pantle and Dieringer 1998). The MAD for horizontal linear acceleration was defined by the stimulus direction orthogonal to the minimal activation direction in the horizontal plane. Ramp-like stimulation produced only one response maximum per stimulus cycle, which decreased as a cosine function with the angle of sled inclination. Sine waves were fitted to these responses evoked by combinations of horizontal and vertical linear accelerations to characterize the maxima and minima of the responses.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The average resting discharge rate in the investigated inferior obliquus (IO), inferior rectus (IR), and lateral rectus (LR) muscle nerves varied between different preparations, ranged typically between 40 and 100 imp/s (n = 11), but was below 20 imp/s or absent in other preparations (n = 6). In most preparations these resting rates remained stable over several hours, and saccade-related spontaneous bursts in the resting discharge rate were absent as in earlier experiments (Pantle and Dieringer 1998; Pantle et al. 1995; Wadan and Dieringer 1994). Blink-related bursts of activity occurred spontaneously and were removed electronically before response cycles were averaged.

Responses to horizontal linear acceleration

Sinusoidal linear acceleration in the horizontal plane modulated the spontaneous discharge rate of the recorded extraocular muscle nerves periodically (e.g., Fig. 2). The responses consisted mainly of an excitatory half cycle. In preparations with a resting discharge rate higher than in Fig. 2, only a relatively small disfacilitation of the spontaneous nerve discharge rate during the other half cycle of stimulation could be observed (Fig. 3A). In the same abducens nerve, however, a strong suppression of the spontaneous discharge was regularly observed during sinusoidal angular acceleration (Fig. 3C). We therefore investigated a possible inhibitory contribution from uncrossed utricular connections by recording abducens nerve responses during horizontal linear and angular acceleration before and acutely after the section of the ramus anterior of the contralateral VIIIth nerve. This section eliminated the inputs from the utricle, the horizontal and anterior vertical semicircular canals. After the lesion, all abducens nerve responses to horizontal linear acceleration had disappeared (Fig. 3B). Uncrossed inhibitory responses to horizontal angular acceleration, however, were still present, whereas the crossed excitatory responses had disappeared as expected (Fig. 3D). Qualitatively the same results were observed in other frogs after a section of the ramus anterior of N. VIII.

View larger version (28K): [in this window] [in a new window]   Fig. 3. Comparison of abducens nerve responses evoked by horizontal linear or angular acceleration before and acutely after a nerve section. A and B: responses to sinusoidal linear acceleration (0.2 Hz ±20 cm) before (A) and after (B) elimination of afferent inputs from the anterior and horizontal canal and from the utricle on the contralateral side. The inset shows the orientation of the frog's head with respect to the direction of sled motion. C and D: responses to sinusoidal angular acceleration (0.2 Hz ±10°/s) before (C) and after (D) nerve section. All data from the same individual. Dashed horizontal lines represent average resting activities prior to stimulation. R and L for right and left.

The response pattern, the depth of modulation, and the phase of the responses of a given eye muscle nerve evoked by horizontal linear acceleration depended on the orientation of the static head position in the yaw plane (Fig. 4A). The IO muscle nerve exhibited large responses during oscillations along or close to the interaural axis (90° in Figs. 4B and 5A). Clockwise reorientation of the frog's static head position in yaw toward 0° (longitudinal oscillation) resulted in a decrease in the amplitude of this response (Figs. 4, A and B, and 5A) and in the emergence of a small secondary response component (Figs. 4, A and B, and 5A) that was phase-shifted by about 180° with respect to the initial response maximum (Figs. 4C and 5B). The phase values of these small responses were difficult to measure reliably and were slightly biased toward the preceding or following larger response component during the same stimulus cycle. This new response increased and the initial response decreased in magnitude the more the static head position was reoriented toward an interaural oscillation (see Figs. 4, A and B, and 5A). The IO muscle nerve exhibited two response components per stimulus cycle over a range of static head positions of 45° (n = 11; Fig. 5A). The amplitudes of both response components were on an average equal (null-point) at a head orientation of 23° (±10°; n = 11). Since the null-point and the MAD of a given response differ by 90°, the functional response sector of the IO muscle nerve on the contralateral utricular macula was characterized by an axis of symmetry that was 67° lateral with respect to the body length axis and by a width of opening of about 45° (Figs. 7B and 10B1).

View larger version (43K): [in this window] [in a new window]   Fig. 4. Responses of the left inferior obliquus eye muscle nerve as a function of the orientation of the head in the horizontal plane with respect to the horizontal linear acceleration of the direction of the sled. A: the instantaneous discharge rate (average from 6-15 cycles) was sinusoidally modulated and exhibited 1 or 2 response peaks per stimulus cycle. B and C: changes in the depth of modulation (B) and in the phase of averaged responses (C; data from responses shown in part in A) as a function of the static head position in the horizontal plane with respect to the direction of sled motion. Triangles and squares indicate in A the peaks and in B and C the fit values of the 1st and 2nd response component, respectively. Gray tone outlines in B the range of head positions for which 2 response peaks per stimulus cycle were observed. Zero degree head position in A-C refers to a linear acceleration along the body length axis. The frequency of oscillation was 0.5 Hz.

View larger version (33K): [in this window] [in a new window]   Fig. 5. Population values of the inferior obliquus and inferior rectus eye muscle nerve responses as a function of the orientation of the head in the horizontal plane with respect to the direction of sinusoidal horizontal linear acceleration of the sled. A and B: depth of modulation (A) and phase values (B) of responses in the left inferior obliquus (IO) eye muscle nerve in a population of 11 frogs. C and D: same parameters as in A and B for the left inferior rectus (IR) eye muscle nerve in a population of 5 frogs. Triangles and squares represent mean values of fitted responses present during the 1st and/or during the 2nd stimulus half cycle. Gray tone in A and C outline the range of head positions for which 2 response peaks per stimulus cycle were observed. For better clarity part of the SDs (vertical bars) were shown in A only in one direction.

The IR muscle nerve showed a qualitatively similar response pattern during horizontal linear oscillation. The depth of modulation depended again on the orientation of the head in yaw. Sled oscillation evoked maximal responses if the static head position was about 60° lateral (Fig. 5C). Two response components per stimulus cycle, however, were present over a much smaller range of head orientations than in the responses of the IO muscle nerve. In two (of 5) animals the IR responses exhibited only one peak. A phase shift of about 180° occurred at a head orientation where only minimal responses were seen (null-point). On an average two response peaks per stimulus cycle were observed over a range of 5° (Fig. 5C), and the null-point was measured at a head orientation of +44° (±2°, n = 5; Fig. 5, C and D). Thus the functional IR response sector on the utricular macula was characterized by an axis of symmetry of 136° lateral with respect to the body length axis and a width of opening of 5° (Figs. 7B and 10A1).

The LR muscle sector of the abducens nerve that was described by Wadan and Dieringer (1994) overlapped partly with the IO muscle sector on the utricular macula. Since no data were available concerning abducens nerve responses during vertical linear acceleration, we analyzed in three animals the responses of this nerve to vertical as well as to horizontal linear acceleration. The results obtained for horizontal linear acceleration were in complete agreement with those obtained earlier, and we therefore combined the former results (n = 8) with our new results (n = 3). Abducens nerve responses exhibited an average null-point at a head orientation of 20° (±8°, n = 11). Two response components per stimulus cycle were present over a range of 60°. Accordingly, the axis of symmetry of the LR response sector was 70° lateral with respect to the body length axis, and the width of the LR response sector was 60° (Figs. 7B and 10C1).

Responses to vertical linear acceleration

Vertical linear oscillation of the frog with the head in standard position evoked very small or no responses in the IO muscle nerve. Often a small residual modulation of the discharge rate was only detected after several response cycles had been averaged. The depth of this modulation increased if the static head position was changed before the test about an axis along which horizontal linear acceleration evoked minimal responses (minimal activation direction). In addition, this increase in the modulation was asymmetric. If the recorded IO muscle nerve was facing downward the evoked responses were about one-half as large (left part in Fig. 6A) as the responses of the same nerve if the recording side was facing upward (right part in Fig. 6A). For none of these different static head positions was the depth of modulation in the IO muscle nerve as large as during horizontal linear acceleration along the MAD (100% reference in Fig. 6A). The phases of the responses evoked with the head inclined to the one or to the other side differed by about 180° (Fig. 6B). Given that practically no responses were seen during vertical linear acceleration (Fig. 6A), static utricular inputs might have disfacilitated the evoked dynamic utricular response components. A possible contribution from vertical otolith organs (lagena or sacculus), however, could have been masked by these static utricular signals.

View larger version (29K): [in this window] [in a new window]   Fig. 6. Response properties of the left inferior obliquus eye muscle nerve during sinusoidal linear acceleration along or at different inclinations with respect to the gravity vector. A and B: the response magnitude evoked in the inferior obliquus (IO) eye muscle nerve by vertical linear acceleration changed asymmetrically as a function of the static head position. C and D: responses in the IO eye muscle nerve evoked by ramp-like sled oscillation and their changes as a function of the inclination of the sled motion with respect to the gravity vector. E and F: similar responses as in C and D were evoked by ramp-like sled oscillation in the lateral rectus (LR) and in the inferior rectus (IR) eye muscle nerves. Note that standard head position was maintained throughout all tests in C-F and that the insets represent the direction of combined horizontal and vertical acceleration (arrows) but not the actual head orientation in the yaw plane. In A-F the horizontal head orientation was aligned with the maximal activation direction of responses during horizontal linear sled acceleration.

We therefore investigated in some of these (n = 4) and in other frogs (n = 2) the responses of the IO muscle nerve in addition with combinations of horizontal and vertical accelerations delivered in a ramp-like manner. Thereby, changes in static utricular stimulation were excluded (see Fig. 1C). The frog was oscillated along the approximated MAD of its IO muscle nerve, and the sled motion direction was inclined in 10° steps from horizontal to vertical or vice versa before each test run. The evoked responses in the IO muscle nerve changed in relation with the inclination of the sled motion direction (Fig. 6C). Vertical linear acceleration evoked practically no response in the IO muscle nerve (Fig. 6C). Responses evoked with the frog oriented in opposite directions with respect to the MAD in the horizontal plane were phase shifted by 180° (Fig. 6D).

The responses of the IR eye muscle nerve and of the abducens nerve were investigated with the same ramp-like stimulus protocol. The response characteristics of the IR eye muscle nerve (Fig. 6, E and F, ; n = 5) and of the abducens nerve (Fig. 6, E and F, ; n = 3) were very similar to those determined for the IO eye muscle nerve (Fig. 6, C and D; n = 6). More importantly, the changes in the response magnitude in each of these eye muscle nerves could be fitted with a sine wave (Fig. 6, C and E). The correlation coefficients of these fit curves (R²) were 0.99 for the LR, 0.94 for the IO, and 0.98 for the IR muscle nerves. Therefore these changes were proportional to the cosine of the angle between the inclination of the sled acceleration and the plane of the utricle (see DISCUSSION). The phases of the fitted sine waves allowed us to determine a possible local elevation of the utricular surface along the axes of symmetry of the investigated muscle nerve sectors. If the phase of a fitted sine wave deviates more from the direction of vertical acceleration (±90° in Fig. 6, C and E), the elevation of the utricular sector, re horizontal canal plane, is larger. The utricular sectors of the IO eye muscle nerve and of the abducens nerve overlapped to some extent (Fig. 7B), and both sectors exhibited an elevation of zero degrees (±5° for the IO muscle nerve; Fig. 6C; ±1° for the abducens nerve; Fig. 6E, ); i.e., these utricular sectors were coplanar with the horizontal semicircular canals. The utricular sector of the IR eye muscle nerve, however, was pitched caudally up by 6° (±3; Fig. 6E, ). This angle of elevation was statistically significantly different from the horizontal plane (P  0.01).

View larger version (24K): [in this window] [in a new window]   Fig. 7. Diagram of the frog's utricle in head coordinates with the orientations of the hair cell polarization vectors and the fan-like eye muscle sectors. A: outline of the surface of the right utricle and a schematic representation of the polarization vectors of hair cells located medially or laterally with respect to the striola (dotted lines). Modified from Baird and Schuff (1994). B: sectors located medially to the striola on the right utricle from which contralateral eye muscles are activated. The orientation of the axis of a given sector in head coordinates (closed arrows) and of the width of each sector (curved arrows) are given in degrees. IO, inferior obliquus; LR, lateral rectus; IR, inferior rectus eye muscle.

Responses to angular acceleration

The convergence of signals from different semicircular canals onto the motoneurons of the IO and of the IR muscles was analyzed first qualitatively and then quantitatively as in an earlier study for the abducens nerve (Pantle and Dieringer 1998). A qualitative estimation of the convergence was obtained by an oscillation of the frog in the planes of the functional canal pairs (left-right horizontal canals: LH-RH; left anterior-right posterior vertical canals: LA-RP; right anterior-left posterior vertical canals: RA-LP). The axis of rotation was earth-vertical and the head was oriented such that the common plane of a functional canal pair was aligned with the horizontal plane of the turntable (see METHODS) to avoid a simultaneous dynamic stimulation of the otolith organs.

OSCILLATION IN THE PLANES OF FUNCTIONAL CANAL PAIRS. As shown by Pantle and Dieringer (1998) the resting activity of the left abducens nerve was modulated during oscillations in the LH-RH and in the RA-LP but not in the LA-RP planes (Fig. 8A). The average phase lead of responses evoked in the plane of the horizontal semicircular canals was 67 ± 7° (mean ± SD, n = 4). The depth of modulation was smaller during oscillation in the RA-LP plane (40 ± 9% with the nose diagonally up and 33 ± 8% with the nose diagonally down, n = 4). The phase of these responses was 128 ± 6° (n = 4) and 58 ± 1° (n = 4), respectively. The left IR muscle nerve responded exclusively during oscillation in the LA-RP plane but not in the planes of the other functional canal pairs (Fig. 8B). The phase of these responses was about 110°.

View larger version (32K): [in this window] [in a new window]   Fig. 8. Responses of the lateral rectus, inferior rectus, and inferior obliquus eye muscle nerves on the left side evoked by angular acceleration in the planes of functional canal pairs: LH-RH, left-right horizontal canal plane; RA-LP, right anterior-left posterior vertical; LA-RP, left anterior-right posterior vertical semicircular canal planes. The responses in A-D were recorded in different animals. Records in D were obtained acutely after the ipsilateral horizontal canal (HC) nerve had been sectioned. Frequency of oscillation was 0.2 Hz, R and L for right and left.

NULL-POINT OF THE INFERIOR OBLIQUE EYE MUSCLE NERVE. We employed the null-point technique for a quantitative analysis of the convergence of signals from different semicircular canals onto eye muscle motoneurons. The null-point of the LR muscle nerve was known from a former study (Pantle and Dieringer 1998), and compatible data from three additional animals obtained in this study were added (for combined results see Fig. 10C2). A null-point analysis of the IR muscle responses was unnecessary, given that this eye muscle received its excitatory canal input exclusively from the contralateral posterior vertical semicircular canal (see Fig. 8B). Thus only the null-points of the IO muscle nerve remained to be determined for a comparison with the best response orientations of this nerve during linear acceleration.

View larger version (33K): [in this window] [in a new window]   Fig. 9. Responses of the left inferior obliquus eye muscle nerve during sinusoidal angular acceleration about an earth-vertical axis as a function of the static head position in pitch or roll. A and B: depth of modulation (A) and phase (B) as a function the static head position in the pitch plane. C and D: same parameters as a function of the static head position in the roll plane. Zero depth of modulation indicates that the two response components per stimulus cycle were equal in magnitude. Data in A and C were fitted with sine waves. The intersection of the fit curves with zero depth of modulation (dotted horizontal lines) indicates the null-point of the responses in the pitch (A) and in the roll plane (C). Horizontal lines in B and D represent mean values. The frequency of oscillation was 0.2 Hz. Insets indicate head positions in the pitch (A) and in the roll plane (C).

Spatial relations between MADs of responses evoked by linear or angular accelerations

From the null-points of the IO muscle nerve responses in static pitch and roll head positions, we calculated the MAD vector of these responses during angular acceleration in head coordinates (Table 1A), transformed these coordinates in canal-plane-related coordinates (Table 1B), and estimated the relative contributions of the converging excitatory semicircular canal signals (Table 1C). The numerical data for these canal-evoked response vectors together with those for the abducens nerve (Pantle and Dieringer 1998) and for the IR muscle nerve are presented in Table 2 together with the macula-evoked response vectors of the same eye muscle nerves. Both sets of data are illustrated in Fig. 10 to facilitate a comparison between the directions of these MAD vectors. The directions of maximal activation during horizontal linear acceleration (blue arrows in Fig. 10, A1-C1) and the rotation axes for maximal responses during angular acceleration (red arrows in Fig. 10, A2-C2) are shown for each of the investigated eye muscle nerves separately in canal coordinates. The spatial relations between these angular rotation axes and translational acceleration directions are summarized in Fig. 10, A3-C3. As a major result, the preferred axis of angular rotation and the best direction of linear accelerations of a given eye muscle nerve were by and large perpendicular with respect to each other (Fig. 10, A3-C3). The angular deviation from perpendicularity was 2.9° for the IR muscle nerve, 17.5° for the IO muscle nerve, and 9.8° for the abducens nerve (see Table 2).

View larger version (92K): [in this window] [in a new window]   Fig. 10. Spatial relations between the directions of horizontal linear accelerations and the axes of angular accelerations for maximal activation of each of the 3 eye muscle nerves in semicircular canal coordinates. The maximal activation direction (MAD) for inferior rectus muscle nerve responses during horizontal linear acceleration is shown by the blue arrow (A1). Best responses during angular acceleration are activated in the same eye muscle nerve by rotation around an axis shown by the red arrow (A2). The blue and the red arrows specifying the best inferior rectus muscle nerve responses during linear and angular accelerations are oriented almost perpendicularly with respect to each other (A3). Therefore angular acceleration around the red axis (A3) provides the strongest utricular stimulation to the macula along the MAD of the inferior rectus muscle nerve. B and C: corresponding relations for the responses of the inferior obliquus (B) and for the lateralis rectus muscle nerves (C). The sectors shaded in gray tone in B1 and C1 show the opening angles of the utricular sectors of the two eye muscle nerves. The percentages given in B2 and in C2 indicate the weighing factors of converging semicircular canal inputs. AC, HC, and PC, anterior vertical, horizontal and posterior vertical semicircular canal, respectively.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Linear acceleration evoked responses in extraocular muscle nerves that originated from the utricle but not from the lagena or from the saccule. Each eye muscle nerve received its excitatory input from a fan-like sector on the contralateral utricle. The axes of symmetry of these sectors had different orientations. The width of a given sector was determined by an opening angle. We specified for the same extraocular muscle nerves also the convergence patterns and quantified the weighing factors of converging semicircular canal-related inputs. These results allowed us to correlate the spatial properties of best responses evoked by linear or angular acceleration in each of the three investigated extraocular muscle nerves. In essence, in a given extraocular muscle nerve the vectors of best response orientations of utricular- and of semicircular canal-related inputs were arranged about orthogonally with respect to each other.

LVOR of frogs originates in the utricle

Utricular afferent signals, but not signals from vertically oriented otolith organs (lagena or saccule) contributed to the recorded maculoocular nerve responses. Since inputs from lagena and sacculus did not contribute to maculoocular responses in frogs during vertical linear acceleration, the response sensitivity during ramp-like accelerations was proportional to the cosine of the angle between the direction of the stimulus and the plane of the utricle (cosine rule). As a consequence, these data could be fitted with a sine wave (Fig. 6, C and E). Response sensitivities to linear (Fig. 6A) or angular accelerations were increased, whenever the head was tilted such that the utricular sector of the recorded eye muscle nerve was pointing toward gravity. A very similar facilitatory influence of static otolithic inputs onto canal-evoked dynamic abducens nerve responses had been observed earlier by Pantle and Dieringer (1998).

The absence of signals from vertical otolith organs is consistent with the absence of compensatory eye movements in frogs during vertical linear acceleration (Hess et al. 1984), the persistence of normal maculoocular reflexes in this species after selective bilateral lesions of the saccular or lagenar nerve branches of N. VIII (Hess and Precht 1984), and the presence of only very few second-order lagenar neurons with ascending axons (Holler and Straka 2000). The LVOR of mammals is dominated by utricular inputs as well. Vertical linear acceleration evokes only very small responses at relatively high levels of acceleration in cat and rat (Hess and Dieringer 1991; Xerri et al. 1988), and static ocular reflexes of rabbits survive a bilateral destruction of the saccular maculae unaltered (Versteegh 1927). Electrophysiological experiments showed that the sacculoocular reflex connectivity is relatively weak in the cat in comparison to utriculoocular or sacculocollic pathways (Isu et al. 2000). Very similar data for lagenar and utricular connections were obtained in frogs (Goto and Straka 2001).

Since extraocular muscle nerve responses exhibited exclusively utricular inputs during linear acceleration, no or only minimal responses should be present during earth-vertical linear acceleration, provided that the plane of a given utricular sector did not deviate from the plane of the horizontal semicircular canal. In fact, sine waves fitted to population data from the IO (Fig. 6C) and from the LR muscle nerves (Fig. 6E, ) indicated that minimal responses were present whenever the sled was inclined by an angle of 90° re earth horizontal. Therefore the LR and IO muscle sectors on the utricular surface were coplanar with the horizontal semicircular canal. Minimal responses of the IR muscle nerve (Fig. 6E, ), however, occurred when the sled was inclined by 6° with respect to verticality, suggesting that the IR muscle sector on the utricle was slightly pitched up. A similar inhomogeneity in the plane of the utricle was described for the guinea pig. There, the utricular surface is curved, slightly pitched down with respect to the plane of the horizontal canal, and exhibits an upturn of its rostral portion (Curthoys et al. 1999). The functional significance of these peculiarities is so far unknown but could be related to the natural head position at rest. Frogs as guinea pigs hold their head at rest such that the plane of the horizontal semicircular canals is slightly pitched up by about 5° (Blanks and Precht 1976; Schneider 1954; Vidal et al. 1986). At this resting head position the otoconia rest in a trough formed by the curvature in the utricular surface and static stimulation of hair cells will be minimal, since otolithic receptors are not sensitive to compression forces (Shotwell et al. 1981).

Eye muscle sectors on the utricle

Fan-like utricular sectors connected with particular sets of extraocular muscles were first demonstrated for the abducens nerve in the frog (Wadan and Dieringer 1994). The results of this study corroborate this notion of particular utricular sectors. Two response maxima per stimulus cycle in a particular range of stimulus directions, as first described by Wadan and Dieringer (1994), are explained by the convergence of otolith signals with different polarization vectors. These response characteristics were used in this study to analyze the spatial dimensions of these sectors. Similar sectors are expected to exist also in mammals, and their existence had been assumed to explain the spatial properties of the LVOR in the rat (Hess and Dieringer 1991). However, so far two response maxima per stimulus cycle had not been observed during linear acceleration in experiments with mammals (Angelaki et al. 1993; Bush et al. 1993; Si et al. 1997). Several reasons account for this difference. First, small modulations in the resting discharge rates are more difficult to detect in the records from single units than in those from multiunits. The irregularities in the resting discharge of single units tend to average out in records from multiunits because of their stochastic nature, whereas the small modulations in the discharge of each of the recorded units add up to a more easily detectable signal. Second, two converging excitatory inputs with response maxima 180° apart from each other will produce two response maxima per stimulus cycle in their target neuron, provided the excitatory input is larger than the simultaneously occurring disfacilitation from the other input. Such a difference develops automatically with an increase in stimulus intensity because of the limited magnitude of disfacilitatory responses due to clipping at 0 imp/s. Low resting rates, as they prevail in frog central vestibular or in abducens motoneurons (0-20 imp/s) (Dieringer and Precht 1986), facilitate the occurrence of response clipping and thereby the emergence of two response maxima per stimulus cycle. In mammals, higher resting rates (about 30-90 imp/s, depending on species) expand the range of response modulation by disfacilitation. Therefore two response maxima per stimulus cycle can be expected to be present also in mammals, however, only at higher stimulus intensities and over a smaller range of head orientations than in frogs.

It is important to note, that the above explanation of two response maxima per stimulus cycle and the predictions for mammals implies the absence of inhibitory maculoocular reflexes as demonstrated here with lesion experiments (see Fig. 3). The absence of reciprocal inhibition had been predicted earlier for the horizontal LVOR of the rat (Hess and Dieringer 1991) to explain the occurrence of conjugate and disconjugate compensatory eye movements in response to transverse and longitudinal acceleration in darkness, respectively. Very similar vergence movements of the eyes were observed in frogs during longitudinal acceleration in darkness as well (Dieringer, unpublished data). To allow the expression of separate vergence and version signals, disynaptic LVOR and AVOR pathways to abducens and medial rectus motoneurons may differ not only in the absence of reciprocal inhibition but also in the presence of parallel and independent excitatory connections.

Convergence between horizontal semicircular canal and utricular signals in central vestibular neurons is more frequently observed in rat (Angelaki et al. 1993), frog (Goto and Straka 2001), and cat (Zhang et al. 2001) than any other canal-utricular signal combination. The utricular inputs of vertical canal neurons had response vectors that were aligned with the anterior and posterior vertical canal, respectively. The utricular inputs of horizontal canal neurons, however, had response vectors that were distributed throughout the horizontal plane (Angelaki et al. 1993). While the majority of vestibular neurons with combined horizontal canal and utricular inputs projects to the spinal cord in frog (Goto and Straka 2001) and cat (Zhang et al. 2001), some of them may project to the oculomotor system and may contribute to the width of the utricular sector of this particular eye muscle. Such a possible correlation is suggested by the fact that the width of the utricular sector of extraocular motoneurons appears to increase with the percentage of horizontal canal inputs. Inferior rectus motoneurons receive no horizontal canal inputs, and their utricular sector is particularly small (Figs. 7B and 10A). Abducens and inferior obliquus motoneurons receive a strong contribution from the horizontal canal (80 and 50% of the excitatory inputs, respectively), and the width of the utricular sectors of these two groups of motoneurons is proportionally larger than that of inferior rectus motoneurons (Figs. 7B and 10, B and C).

Spatial organization of AVOR and LVOR: a comparison

Three neuron vestibular reflex arcs (Lorente de Nó 1933; Szentagothai 1943, 1950) are composed of vestibular nerve afferent fibers, second-order vestibular neurons, and extraocular motoneurons. The principal pathways of these three neuron reflex arcs link canal inputs to those eye muscles that pull the eye approximately in the plane of this canal. Very similar principal connections were also encountered in frogs and may be common, at least among tetrapods. However, the geometric alignment between the semicircular canals and the corresponding eye muscles alters during ontogeny, is less than perfect in adults, and differs between species (Graf and Simpson 1981; Grobstein and Comer 1977; Pantle and Dieringer 1998; Simpson and Graf 1981). Principal connections were therefore predicted by theory to be supplemented by additional noncoplanar canal signals (Ezure and Graf 1984; Pellionisz 1985; Robinson 1982) to compensate for these misalignments. Thereby, the vestibular sensory signals are transformed into motor signals such that the pulling directions of the eye muscles are taken into account. This organizational principle had been demonstrated most convincingly in a comparative study (Pantle and Dieringer 1998) in which two closely related species (grass and water frogs) with known differences in the orientation of their optic axes (same laterality but different angles of elevation) were investigated. As a result, the convergence pattern of horizontal and vertical semicircular canal signals in abducens motoneurons differed in these two species quantitatively as expected for their proper alignment with the different pulling directions of the lateral rectus muscles. Compatible results from the cat (Baker and Peterson 1991; Graf et al. 1993) support the generality of this conclusion. In essence, the AVOR consists of principal connections that link canal inputs to spatially appropriate eye muscles. In most vertebrates, including frogs (results of this study), these principal semicircular canal connections (Cohen et al. 1964) consist of excitatory and inhibitory projections that are organized in a push-pull mode (Graf and Simpson 1981). Supplementary canal signals are combined species-specifically to transform sensory signals into spatially adequate motor signals.

The organization of the LVOR is similar to but not identical with the above summarized organization of the AVOR. At variance with the AVOR are the apparent absence of a push-pull organization and the absence of a combination of horizontal and vertical otolith signals for the LVOR. Principal inhibitory utriculoocular connections comparable to second-order horizontal canal neurons inhibiting ipsilateral abducens motoneurons monosynaptically (Baker and Highstein 1975; Straka and Dieringer 1993) are absent in frogs (see Fig. 3B). Thus inhibitory second-order horizontal canal neurons receive either no or only very weak utricular afferent inputs.

Another difference between the AVOR and the LVOR concerns supplementary connections. Vertical otolith signals are apparently not used in the LVOR to transform utricular sensory into spatially appropriate extraocular motor signals to an extent as semicircular canal signals are used in the AVOR. However, one has to realize that pure linear head accelerations are small and rare in a biological context. Most often, i.e., whenever the AVOR is activated about an axis that is not earth-vertical, linear and angular acceleration components are co-activated. Earlier behavioral studies had shown that the combination of LVOR plus AVOR signals improves the performance of compensatory eye movements dynamically, particularly in the low-frequency range (cat: Rude and Baker 1988; Tomko et al. 1988; rat: Brettler et al. 2000; monkey: Angelaki and Hess 1996). A necessary prerequisite for such a mutual supplementation of AVOR and LVOR signals, however, is that the spatial organizations of both reflexes are in register. That means that the MAD vectors for best responses during linear acceleration and the rotation axes for best responses during angular acceleration should be orthogonal with respect to each other. Such a spatial tuning between canal- and otolith-related ocular motor signals had been predicted for vertical eye muscles by Szentagothai (1952) and by Hess and Dieringer (1991) and has been demonstrated quantitatively for the first time in this study (see Fig. 10). The linear MAD axis and the angular acceleration axis for best responses of each of the three investigated eye muscle nerves were almost orthogonally oriented with respect to each other (Fig. 10, A3, B3, and C3). Therefore angular head rotation about any not earth-vertical axis will co-activate utricular responses. The strongest utricular responses will occur whenever the rotation axis is earth-horizontal as for instance for the inferior rectus muscle nerve (red axis in Fig. 10A3). In this case, the strongest utricular responses are evoked along the linear MAD axis of this eye muscle nerve (blue axis in Fig. 10A3). Angular acceleration about inclined axes (between earth-horizontal and earth-vertical; see red axes in Fig. 10, B3 and C3) evoke utricular responses that are again strongest along the linear MAD axis of a given eye muscle nerve (blue axes in Fig. 10, B3 and C3) but that depend in their magnitude on the angle of inclination. Maximal responses are present, if the head is tilted such that the angular rotation axes (red axes in Fig. 10, B3 and C3) become earth-horizontal. This matched spatial organization of semicircular canal and otolith signals for a given eye muscle nerve has to be considered in the larger context of a common reference frame for head motion sensors. The preferred directions of neurons of the accessory optic system are matched to optic flow directions that occur during head rotations about the best-response axes of the semicircular canals, suggesting that the accessory optic system is organized in vestibular canal coordinates as well (Simpson et al. 1988). Such a common reference frame for otolith-, canal-, and optic flow-related signals facilitates the convergence of multimodal sensory signals indicating self-motion for the construction of an integrated motor response that takes the pulling directions of the different eye muscles into account.