INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recent anatomic and physiologic studies have demonstrated direct connections between the vestibular nuclei and brain stem regions that influence sympathetic and parasympathetic outflow (review: Balaban 1996). These pathways originate from a region within the vestibular nuclei that includes the dorsal aspect of the superior vestibular nucleus, the caudoventral aspect (pars ) of the lateral vestibular nucleus, and the caudal half of the medial vestibular nucleus and the inferior vestibular nucleus (Balaban 1996; Balaban and Beryozkin 1994; Porter and Balaban 1997; Ruggiero et al. 1996; Yates et al. 1994, 1995). The caudal medial vestibular nucleus and the inferior vestibular nucleus can influence parasympathetic and sympathetic outflow, either directly or indirectly, via ascending projections to the nucleus of the solitary tract, dorsal motor vagal nucleus, nucleus ambiguus, and rostral ventrolateral medullary reticular formation. An ascending pathway also originates from the dorsal aspect of the superior vestibular nucleus, pars alpha of the lateral vestibular nucleus, and caudal half of the medial vestibular nucleus and the inferior vestibular nucleus. This ascending projection terminates densely in the caudal third of the parabrachial nucleus, which has reciprocal connections with the amygdala, hypothalamus, and prefrontal cortex. These pathways have been suggested as substrates for vestibular contributions to cardiovascular control, respiratory patterns, gastrointestinal function, and anxiety disorders (Balaban 1999; Balaban and Thayer 2001; Balaban and Yates 2002). In particular, this pathway may participate in the anxiety and panic associated with losing one's balance, optic-flow-induced illusions of drifting or falling backward, and vestibular system dysfunction.

The parabrachial nucleus is generally recognized as a major relay for ascending visceral (including gustatory) and nociceptive information in central autonomic pathways (Bernard et al. 1993, 1995; Feil and Herbert 1995; Fulweiler and Saper 1984; Jasmin et al. 1997; Pritchard et al. 2000) with modality-specific regions such as a "gustatory region" (Grigson et al. 1998; Nishijo and Norgren 1997; Spector et al. 1992). The anatomical distribution of vestibulo-parabrachial fibers raises the further hypothesis that there may be a distinct "vestibular" or "body/head motion"-sensitive region within the parabrachial nucleus and the Kölliker-Fuse nucleus. Although anatomical evidence indicates that the vestibular nuclear projections are confined within the caudal aspect of the lateral, medial, and external lateral parabrachial nucleus and the Kölliker-Fuse nucleus, there is no physiological information regarding response properties of neurons in these regions to head and/or whole-body rotation. This study provides the first demonstration that a discrete region in the caudal parabrachial nucleus contains neurons that show complex responses to rotations in three dimensions in alert primates.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Surgical procedures

All surgical procedures were conducted under aseptic conditions in an animal surgical suite at the Central Animal Facility of the University of Pittsburgh. Three female macaque monkeys (Macaca nemestrina) were premedicated with atropine (0.05 mg/kg im) and ketamine (12 mg/kg im). After endotracheal intubation, anesthesia was maintained by inhalation of a 1-1.5% halothane-nitrous oxide-oxygen mixture. Three dental acrylic lugs were implanted for secure but painless head stabilization during recording sessions. One lug, positioned centrally on the top of the skull, served as a pedestal for electrical connectors; the other two lugs were positioned behind the ears. At the site of each lug, a 15 × 20-mm patch of skin and periosteum was removed, and small holes were drilled in the skull with a dental burr. Small stainless steel screws were tapped into these holes, and the lug was constructed by applying layers of dental acrylic around the screws to a height of approximately 9 mm.

A "search coil" was implanted on the right eye to measure eye movements, based on the technique of Judge et al. (1980). The conjunctiva was cut at the limbus, and a preformed 16-mm-diam coil (3 turns of Teflon-insulated stainless steel wire) was sutured to the sclera. Lead wires were passed subcutaneously to a connector on top of the skull. The conjunctiva was sutured with 7-0 vicryl to cover the coil.

A 20-mm-diam, 10-mm-high stainless steel recording chamber was implanted surgically over a hole that was trephined in the parietal bone to permit the chamber to contact the intact dura mater. The chamber was centered at 1 mm left of the midline and +1 to +3.5 mm anterior to the ear bars in different monkeys, angled 15° posteriorly. This approach permitted complete exploration of the left parabrachial nucleus and access to the medial edge of the right parabrachial nucleus. Stainless steel screws were anchored within the surrounding bone through small burr holes, and dental acrylic applied to fix the chamber to the skull. The chamber was filled with "triple antibiotic ointment" and covered with a tightly fitting metal cap.

Recording sessions

Recording sessions began after at least a 2-wk recovery period. The monkeys were seated in a primate chair with their heads fixed to the chair in the stereotaxic plane. They were placed in a two-axis rotation device enclosed in a soundproof, lightproof, and shielded booth. Horizontal rotation about a (vertical) yaw axis was driven by an 80 ft-lb servo-controlled motor (Contraves), which reliably produced waveforms ranging from velocity trapezoids of unlimited duration to sine waves exceeding 10 Hz. The stimulator had an inner axis for producing oscillations in a vertical plane or static deviations up to 90°. With the monkey facing forward, the oscillations were oriented in the pitch plane. Rotating the primate chair 90° in the apparatus allowed oscillations in the roll plane. Intermediate angles produced oscillations in the planes of the vertical semicircular canals.

Eye movements were measured with the magnetic search coil technique (CNC Engineering). Because the transmitting coils for generating the magnetic fields do not rotate, search-coil signals are demodulated with the "phase angle" method (Collewijn 1977).

Extracellular single-unit recordings were obtained with standard 0.005-in tungsten electrodes purchased from Microprobe and were positioned with a Trent Wells X-Y stage on top of the chamber and Trent-Wells microdrive. A guide tube protected the electrode as it was lowered through the dura mater. Signals from the electrode were amplified conventionally, filtered, and monitored. Unit, eye-movement, and vestibular data were encoded and recorded on videotape with a Vetter PCM recorder for off-line analysis.

The left abducens nucleus and left trochlear nerve root were first identified as landmarks by their characteristic burst-tonic properties during eye movements (Fuchs and Luschei 1970). Single units in the parabrachial nucleus were identified initially in the dimly illuminated booth using a search stimulus of 0.7-Hz oscillation in the pitch plane (±12 or 53°/s peak velocity) and 0.7-Hz oscillation about the yaw axis (±60°/s peak velocity). Units that showed any modulation with the stimulus were then tested during 0.3-Hz sinusoidal (±12°), 0.7-Hz sinusoidal (±12°), and 0.3 Hz position trapezoid (±9-13°, peak velocity 95-100°/s) rotation in the dark booth in the following planes: 1) pitch plane; 2) an approximate left anterior-right posterior semicircular canal plane (LARP, animal rotated 45° rightward from pitch axis); 3) an approximate right anterior-left posterior semicircular canal plane (RALP, animal rotated 45° leftward from pitch axis); 4) roll plane (animal rotated 90° rightward from the pitch axis); 5) yaw plane with head level (i.e., utricular and horizontal semicircular canal oriented about 25° upward); and 6) yaw plane with the monkey rotated 25° nose-down (utricular and horizontal semicircular canal plane).

The spike times were identified with 0.1-ms precision and the firing rates were digitized at 5 ms/sample. Each stimulus cycle was divided into 64 equally spaced time bins, and the number of neuronal spikes in each time bin was averaged to compute the firing rate across an average stimulus cycle. These 64-bin average responses were used for further analyses. The body position signal relative to gravity was obtained directly from the averaged position signal from the vertical plane rotation device; horizontal rotational velocity, and vertical and horizontal eye position signals were also saved.

The approach to analysis of the responses is described in detail in the presentation of the results. Briefly, least-squares methods (Marquardt-Levenberg algorithm) were used to estimate the baseline firing rate and unit sensitivities to angular velocity, angular position, and a leaky integration of angular velocity (with 250-ms exponential decay). As described in RESULTS, the inclusion of an empirically determined leaky integration of angular velocity in the analysis was sufficient to represent the responses of all neurons in all directions tested. This was not true if an acceleration or jerk component (peak velocity occurred 250-300 ms after stimulus onset, whereas peak acceleration occurred within 100-150 ms after stimulus onset) was substituted for integrated velocity in the analyses. Least-squares methods were then used to estimate the spatial tuning sensitivity of velocity, integrated velocity, and position sensitivity parameter for each unit during vertical plane and yaw rotations. These sensitivities were then used to reconstruct three-dimensional plots of velocity and integrated velocity sensitivity as a function of the orientation of axes of rotation based on the assumption of a linear summation of vertical plane- and yaw-related responses. Statistical tests such as ANOVA, least-significant differences (LSD), and Bonferrroni post hoc tests and probability plots were performed with SYSTAT (SPSS, Evanston, IL). Directional statistics (Mardia and Jupp 2000) were employed to perform tests of significance on angular measurements that avoid analytical problems that are inherent in "cutting" data distributed around a circle at an arbitrary point (e.g., at 0°) and performing linear statistics. The algorithms for these methods were taken from Mardia and Jupp (2000) and implemented in MATLAB. Finally, multivariate analyses for vectorial data (e.g., Hotelling's T² test) (e.g., Anderson 1958) were implemented in MATLAB.

Rotation axes for unitary responses were calculated using a right hand rule convention, with  designating azimuth ("longitude") and  (referenced to the horizontal semicircular canal plane) designating elevation, based on the observed cosine-like spatial tuning of the units (see RESULTS). Because all recordings were performed from the left parabrachial complex, nose-up pitch was produced by right-handed rotation about an axis pointing out the right (contralateral) ear, defined as (,) = (0.5 rad, 0 rad). Nose-down pitch was defined accordingly as right-handed rotation about the axis (,) = (0.5 rad, 0 rad), directed out of the ipsilateral (left) ear. Left ear down (LED) rotation was defined accordingly as right-handed rotation about the axis (,) = (± rad, 0 rad); right ear down (RED) right-handed rotation as (,) = (0 rad, 0 rad). Therefore nose-up and nose-down right-handed rotation in the RALP plane are defined right-handed rotation about the axes (,) = (0.75, 0) and (0.25, 0) radians, respectively, while nose-up and nose-down rotation in the LARP plane are defined as right-handed rotation about the axes defined by (,) = (0.25, 0) and (0.75, 0) radians, respectively. Excitation to leftward yaw was defined as (,) = (0 rad, 0.5 rad); excitatory responses to rightward yaw were represented as (,) = (0 rad, 0.5 rad).

After completion of PBN recording sessions, a tungsten electrode was used to verify the location and depth of the superior vestibular nucleus. A quartz-glass pipette [10-15 µm (ID) tip] containing biotinylated dextran-amine (BDA, 7.5% in 10 mM phosphate buffer containing 0.5 M NaCl, pH 7.0) was introduced into the superior vestibular nucleus, and the BDA was ejected iontophoretically (4 µA tip positive square wave, 15-s duty cycle, 30 min). During the ensuing survival time of 17 days, a series of small electrolytic lesions (25-30 µA, 20-30 s) was placed above and below the borders of the PBN region that contained responsive units. At the conclusion of the survival period, the monkeys were killed with a pentobarbital overdose and perfused transcardially with 50 mM phosphate-buffered saline (PBS), followed by sodium metaperiodate-lysine-paraformaldehyde (PLP) fixative (McLean and Nakane 1974). The brains were infiltrated with a 30% sucrose-4% paraformaldehyde solution overnight at 4°C, then cryoprotected in 30% sucrose in PBS until they sank. Frozen sections (40 µm, transverse plane) were cut on a sliding microtome in a transverse plane at the same orientation as the electrode tracks. Sections were either placed in 50 mM phosphate buffered saline (PBS, pH 7.2-7.4) or stored at 20°C in a solution of 30% sucrose-30% phosphate buffer and 30% ethylene glycol.

For visualizing BDA transport, the sections were rinsed successively in distilled water (3 × 10 min), 0.9% H₂O₂ and distilled water to suppress endogenous peroxidase activity, followed by a preincubation for 2 h at room temperature in 0.5% Triton X-100 in PBS. After rinsing in PBS, the sections were incubated for 1 h in ABC reagent, rinsed in buffer, and reacted for visualizing sites of peroxidase activity with either a nickel-enhanced DAB or a standard DAB (2 mg DAB, 8.3 µl H₂O₂ in 10 ml 500 mM sodium acetate buffer, pH 6.0) chromogen. After extensive rinsing, sections were mounted on subbed slides, cleared in xylene and coverslipped with a nonfluorescent DPX mounting medium.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Location of parabrachial nucleus units and confirmation of afferent projections from the superior vestibular nucleus

The physiological data were obtained from 85 units that were confirmed histologically to be within the borders of the lateral or medial parabrachial nucleus or the caudal aspect of the Kölliker-Fuse nucleus (Fig. 1). Fifty-four of these units (64%) were tested in multiple rotation conditions for characterization of the spatial tuning of the responses. The tracks were reconstructed from histological sections on the basis of marking lesions and the location of physiologically mapped borders of the abducens nucleus, spinal trigeminal nucleus, and trochlear nerve. The units were located between the caudal border of the trochlear nerve root and the caudal border of the parabrachial nuclear complex.

View larger version (22K): [in this window] [in a new window]   Fig. 1. Location of units in the parabrachial nucleus. The location of each unit is plotted on a series of transverse sections [from rostral (1) to caudal (4)]. The mesencephalic trigeminal nucleus (Mes V) and brachium conjunctivum (BC) are also indicated. These units are all located in the caudalmost 2 mm of the parabrachial and Kölliker-Fuse nuclei, which lies immediately caudal to the level of the trochlear nerve exit.

Anatomical tracing evidence confirmed that axons from the superior vestibular nucleus terminate within these regions of the parabrachial nucleus. Figure 2 shows a biotinylated dextran amine-labeled terminal in the lateral parabrachial nucleus at the level of section 3 from Fig. 1. Terminals were distributed throughout the region containing rotation responsive units.

View larger version (125K): [in this window] [in a new window]   Fig. 2. Photomicrographs of biotinylated dextran amine (BDA) tracing of vestibular nucleus projections to the parabrachial nucleus. A: low-magnification photomicrograph of a BDA injection site in the superior vestibular nucleus (SVN). The injection did not encroach on the inferior cerebellar peduncle (ICP), lateral vestibular nucleus (LVN), the rostral aspect of the medial vestibular nucleus (rMVN), or the angular bundle (ang). B: high-magnification photomicrograph illustrates a BDA-labeled terminal in the ventrolateral aspect of the parabrachial nucleus, reflecting anterograde transport from the site in A. A coarse caliber axon forms a basket of varicosities around a cell body and en passage varicosities in the neuropil. This type of terminal was found in the region ventrolateral to the superior cerebellar peduncle at the levels shown in Fig. 1. C: smaller-caliber axons gave rise to varicosities en passage in the medial parabrachial nucleus, immediately ventral to the superior cerebellar peduncle. These varicosities were observed in contact with somata and within the neuropil.

Temporal response properties

RESPONSES TO POSITION TRAPEZOID STIMULATION. The responses to position trapezoid stimulation have provided the greatest insight into temporal characteristics of the responses of parabrachial nucleus neurons to whole body rotation in vertical planes and the horizontal plane. The parabrachial nucleus units in this study were unresponsive to eye position and eye velocity during spontaneous saccades interspersed with periods of fixation. Subsequent data from another trained monkey indicate that they are unresponsive during sinusoidal smooth pursuit (unpublished observations). Four examples of single-unit responses to vertical position trapezoidal wave stimulation are shown in Fig. 3 and responses of two units to vertical and horizontal position trapezoidal stimuli are shown in Figs. 5 and 6. All units responded vigorously during the dynamic phase of whole-body position trapezoid rotation (Figs. 3-6). Responses during vertical rotation could be either unrectified (increased discharges in 1 direction, decreased discharges in the opposite direction), fully rectified (increased discharges in both directions), or biphasic (increased followed by decreased discharges in 1 direction, decreased followed by increased discharges in the opposite direction). Further, the discharges of the same unit could vary both from a biphasic response to a monophasic response pattern and in degree of rectification as a function of the stimulus direction (e.g., Fig. 6). As a result, a simple modeling approach was taken to investigate the relationship of these discharge patterns to whole-body acceleration, whole-body velocity, and whole-body position.

View larger version (33K): [in this window] [in a new window]   Fig. 3. Examples of responses of parabrachial nucleus neurons to whole-body rotation. The averaged responses of 4 units to position trapezoid stimulation represent the salient features of parabrachial nucleus responses to rotation in pitch, roll, or vertical semicircular canal (SC) planes. Firing rate, the average rate during a cycle of stimulation, and position, the average head displacement during a stimulus cycle, are shown for each unit. The firing rate during stimulation was modeled as a sum of 4 components: baseline firing rate, angular position sensitivity, angular velocity sensitivity, and a leaky integrator of angular velocity.  (firing rate), the fit of the model to the data. Note that units g1901 and a5901 have position sensitivity, while units a5603 and a7102 are insensitive to vertical position.

View larger version (21K): [in this window] [in a new window]   Fig. 4. Representative performance of models of unit responses. This figure compares the ability of a model incorporation whole-body angular acceleration and a model incorporating a leaky integration of whole-body angular velocity to explain the position trapezoid responses of 2 units from Fig. 3. Both models represented the firing rate of the cell as the sum of baseline firing rate, angular position sensitivity, whole-body angular velocity sensitivity, and either acceleration (thin line) or a leaky integration of velocity (thick line). Note that the latter model was capable of describing both responses, but that the acceleration-based model could not adequately describe the monophasic response of unit g1901. Thus the leaky integrator model was used as a parsimonious description of the cellular responses for further analyses.

View larger version (31K): [in this window] [in a new window]   Fig. 5. Summary of whole-body rotation responses of unit a5902. Top: the average firing rate during a cycle of position wave rotation in the right anterior-left posterior (RALP) SC, pitch, left anterior-right posterior (LARP) SC roll, and yaw (horizontal SC) plane. The durations of the upward and downward movements are shown by solid bars at the top of each graph. RAU, right anterior canal up; RAD, right anterior canal down; NU, nose up; ND, nose down; LAU, left anterior canal up; LAD, left anterior canal down; LEU, left ear up; and LED, left ear down. The solid trace represents the model fit of each response. Bottom: the directional tuning of the sensitivity of the position, angular velocity, and leaky integration of angular velocity response components to vertical rotation. The solid line in each panel represents the fit of an asymmetric cosine function (see text) to characterize the directional tuning.

View larger version (30K): [in this window] [in a new window]   Fig. 6. Summary of whole-body rotation responses of unit a6301. Top: the average firing rate during a cycle of position trapezoid rotation in the RALP SC, pitch, LARP SC roll, and yaw (horizontal SC) plane. The duration of the upward and downward movements are shown by solid bars at the top of each graph. The solid trace represents the model fit of each response. Botom: the directional tuning of the sensitivity of the position, angular velocity, and leaky integration of angular velocity response components to vertical rotation. The solid line in each panel represents the fit of an asymmetric cosine function (see text) to characterize the directional tuning.

BASELINE FIRING RATE. The baseline firing rates of parabrachial nucleus neurons ranged from 1.3 to 87.0 spikes/s and their distribution was positively skewed (Fig. 7A). The baseline firing rate did not appear to differ as a function of either the magnitude or best orientation of the velocity (Fig. 7B), integrated velocity (Fig. 7C), or position sensitivity of individual neurons or with the symmetry or rectification of the responses. Distributions with this degree of positive skew are typically described well by gamma functions. Maximum likelihood estimation methods indicated that the baseline firing rates are consistent with a random sample from a gamma distribution with a shape parameter of 2 and a scale parameter of 16.7 (Fig. 7A, ).

View larger version (43K): [in this window] [in a new window]   Fig. 7. Background discharge properties of parabrachial nucleus neurons. A: a histogram of the distribution of the background firing rates is positively skewed. , a gamma distribution with a shape parameter equal to 2 and a scale parameter equal to 16.7. B: the background discharge rate was not correlated with the velocity sensitivity for rotation about the axis of maximum response. C: the background discharge rate was not correlated with the integrated velocity sensitivity for rotation about the axis of maximum response.

VELOCITY SENSITIVITY FROM SINUSOIDAL RESPONSES. The analysis of sinusoidal responses was complicated by the fact that approximately 60% of the neurons had position sensitivity. For cells with both position and integrated velocity sensitivity, the relative contributions of the position and integrated velocity components of the sinusoidal responses cannot be separated uniquely. Therefore the unit responses were first classified as position-sensitive or -insensitive on the basis of responses to position trapezoids. The responses of position-insensitive units were then analyzed in the same manner as the responses to position trapezoids but with the position gain fixed at zero. Least-squares methods were used to estimate upward and downward gains for velocity and integrated velocity sensitivity of only the position-insensitive neurons during sinusoidal oscillation.

Three-dimensional organization of velocity sensitivity

The three-dimensional rotational sensitivity of both the yaw-responsive and -insensitive units was estimated from relative magnitudes of the unit sensitivity to yaw and to vertical plane rotation. The spatial tuning of the velocity response of each unit was expressed as an axis of rotation in modified spherical coordinates using a right-hand rule for both the excitatory (or, for rectified cells, preferred) and inhibitory (or for rectified cells, nonpreferred) directions. In this coordinate system, r designates the magnitude of the rotation vector. The azimuth (_vel or _int) designates the orientation of the vertical rotation component with respect to the head. The elevation (_vel or _int) designates the orientation of the yaw-sensitive component [0 indicates no yaw sensitivity, /2 indicates a strictly ipsilateral (left) yaw response and -/2 designates a purely contralateral (right) yaw response]. The responses were also represented graphically as response surfaces, which plot the rotational velocity or integrated velocity sensitivity (the projection of r) as a function of azimuth and elevation (Fig. 10). The plots assume a linear addition of yaw and vertical axis rotational sensitivity across all azimuths and elevations, which permit a direct comparison of rotational spatial tuning properties of velocity and integrated velocity response component.

The elevation (_vel) of three-dimensional angular velocity sensitivity axes and the presence of inhibitory (nonrectified) or excitatory (rectified) responses in the nonpreferred direction of rotation were sufficient to distinguish five populations of parabrachial nucleus neurons (Fig. 8). These categories are described in the sections that follow.

View larger version (18K): [in this window] [in a new window]   Fig. 8. Three-dimensional best response axis for excitatory responses of parabrachial nucleus neurons. The axis of rotation for the greatest excitatory response of each parabrachial nucleus neuron is expressed in modified spherical coordinates using a right hand rule. In this coordinate system, r designates the magnitude of the rotation vector. The azimuth () designates the orientation of the vertical rotation component with respect to the head (0 indicates nose-up pitch, /2 designates ipsilateral (left) ear-down roll, /2 indicates contralateral (right) ear-down roll and  =  indicates nose-down pitch). The elevation () designates the orientation of the yaw-sensitive component (0 indicates no yaw sensitivity, /2 indicates a strictly left yaw response and -/2 designates a purely right yaw response). The neurons could be classified by the elevations () of their 3-dimensional angular velocity sensitivity axes as: vertical-rotation-only cells ( = 0), vertical rotation plus ipsilateral yaw cells (or positive elevation axis neurons,  > 0) with unrectified responses, vertical rotation plus ipsilateral yaw cells (or negative elevation axis neurons,  < 0) with unrectified responses, and || > 0 cells with rectified responses.

VERTICAL-PLANE-ONLY RESPONSE NEURONS. Thirty-two neurons responded to whole body angular velocity in a vertical plane (i.e., _vel = 0). Twenty-one cells showed a nonrectified response pattern with a maximum excitatory response during rotation in one direction in a vertical plane and an inhibitory response during rotation in the opposite direction. Cells with rectified responses (11 units), on the other hand, showed excitatory responses for rotation in either direction within a best plane.

View larger version (40K): [in this window] [in a new window]   Fig. 9. Properties of angular velocity and "leaky integrator" responses of neurons with  = 0. A: 3 classes of neurons could be differentiated by rotational velocity sensitivity independent of rectification of responses. However, velocity gain was unrelated to integrated velocity responsiveness (B). C: polar plot of the vertical angular position sensitivity of parabrachial nucleus neurons with  = 0. The axis of rotation is represented by a right-hand rule; the length of the vector is the sensitivity of the cell in spikes/°. D: distribution of the best orientations for position sensitivity of  = 0 neurons. E: distribution of the best orientations for angular velocity sensitivity of position-insensitive  = 0 neurons. F: the differences between the preferred (excitatory) response directions for velocity sensitivity (vel) and for position sensitivity (pos) were distributed bimodally for position-sensitive units with  = 0. The estimated underlying von Mises distributions for the bimodal data are also plotted.

1

1

1

1

VERTICAL ROTATION PLUS YAW VELOCITY-SENSITIVE NEURONS. Three populations of neurons had preferred response axes with nonzero elevation (|_vel|>0). These populations (Fig. 8) have been designated positive elevation axis neurons, negative elevation axis neurons, and rectified neurons. The velocity and integrated velocity sensitivity of these units to rotation about axes with different azimuths and elevations is plotted for one positive elevation axis neuron (a6301), one negative elevation axis neuron (a5902), and two rectified neurons (a5801 and a5901) in Fig. 10.

View larger version (60K): [in this window] [in a new window]   Fig. 10. Static spatial tuning characteristics of units with non-0 elevation rotation axes for excitatory velocity responses. Top: the responses of the units shown in Figs. 5 and 6 are represented as response surfaces. Graphs plot the rotational velocity or (leaky) integrated velocity sensitivity as a function of azimuth and elevation of the rotation axis. The plots assume a linear addition of yaw and vertical rotational sensitivity across all azimuths and elevations and permit a direct comparison of rotational spatial tuning properties of velocity and integrated velocity response components. These response surfaces are termed static tuning characteristics because they show the response for rotation about any fixed axis. The velocity and integrated velocity sensitivity of these units to rotation about axes with different azimuths and elevations is plotted for 1 positive elevation axis neuron (a6301), 1 negative elevation axis neuron (a5902), and 2 rectified neurons (a5801 and a5901).

View larger version (37K): [in this window] [in a new window]   Fig. 11. Properties of angular velocity and vertical position responses of neurons with   0. A: positive elevation axis neurons (vel >0, ), negative elevation axis neurons (vel <0, ), and rectified elevation axis neurons () varied in the symmetry of velocity sensitivity during rotation about the preferred axis. B: polar plot of the magnitude of the 3-dimensional velocity response vector (r) as a function of its azimuth (vel). The population of negative elevation neurons () had a preferred vel (azimuth) centered near nose-down ipsilateral (left) anterior canal-contralateral posterior canal plane to nose-down pitch rotation (Rayleigh test statistic: 9.35, P < 0.01). By contrast, the population of positive elevation axis neurons () showed no significant preferred direction. C: polar plot of the magnitude and best excitatory response direction (pos) for position-sensitive positive elevation axis neurons (vel >0, ), negative elevation axis neurons (vel <0, ), and rectified elevation axis neurons (). D: the position-sensitive nonzero elevation axis neurons had significantly lower velocity sensitivity than the position-insensitive cells in both excitatory and inhibitory [or nonpreferred] directions of rotation [repeated-measures ANOVA, F(1,20) = 5.02, P < 0.05).

1

1

1

1

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study provides the first electrophysiological demonstration that neurons the parabrachial nucleus respond to whole-body rotation. The responses were restricted to neurons in the caudal aspect of the lateral parabrachial, medial parabrachial, and Kölliker-Fuse nuclei of these head-restrained monkeys. The responses were all confined to the region caudal to the root of the trochlear nerve. Because these neurons were circumscribed in a region that receives terminals from the superior vestibular nucleus, it seems likely that these responses reflect inputs from vestibulo-parabrachial relay neurons. More rostral electrode penetrations yielded only units that were unresponsive to whole-body rotation. These data, then, support the existence of a functional "vestibulo-receptive" region of the parabrachial nuclear complex.

The limited tracing data from this study in monkeys are consistent with previous anatomical evidence from rabbits (Balaban 1996) and rats (Porter and Balaban 1997) that demonstrated that the caudal aspect of the lateral parabrachial, medial parabrachial, and Kölliker-Fuse nuclei receive projections from the vestibular nuclei. However, vestibular nuclear afferents are only one of the known sources of input to this region. Anatomical tracing studies in rats indicate that these caudal regions of the parabrachial and Kölliker-Fuse nuclei also receive presumptive body and head nociceptive and thermal input from contralateral spinal cord laminae 1-3 and the spinal trigeminal nucleus (Feil and Herbert 1995), somatosensory input from the oral cavity, upper alimentary tract, and upper respiratory tract from the paratrigeminal nucleus (Feil and Herbert 1995), and visceral sensory inputs from the rostral and ventrolateral nucleus of the solitary tract (Herbert et al. 1990). Data from rats indicate that descending projections to this region originate in the central amygdaloid nucleus and the prefrontal, insular and infralimbic cortex (Moga et al. 1990); a central amygdaloid nucleus projection to the parabrachial nucleus was also reported earlier in primates (Price and Amaral 1981). Hence, it is important to recognize that the responses to rotation of these neurons may to be altered by other sensory information and descending limbic projections under other experimental conditions.

The data from this study are consistent with the hypothesis that the caudal region of the primate parabrachial nucleus is a source of information about whole-body motion to limbic pathways (Balaban and Thayer 2001). Anatomical studies have shown that the caudal parabrachial nuclei project to the central amygdaloid nucleus (Norita and Kawamura 1980; Pritchard et al. 2000), the lateral bed nucleus of the stria terminalis ("extended amygdala") (Pritchard et al. 2000), the frontal pole and anterior cingulate cortex (Porrino and Goldman-Rakic 1982), and the hypothalamus (Pritchard et al. 2000). Thus these neurons may convey information about body (and/or head) rotation to pathways that produce homeostatic and affective responses to body motion.

Three response components of parabrachial nucleus neurons were identified unambiguously from discharges during position trapezoid, whole-body rotation. First, the firing rates of all neurons were responsive to rotational velocity in a plane with a prominent vertical component. Second, all cells displayed a signal that could be modeled parsimoniously as a leaky integration of rotational velocity in a plane with a vertical component. The leaky integrator component also showed cosine-like or rectified cosine-like directional tuning, but the spatial tuning was not correlated with the velocity sensitivity of the unit. Finally, approximately 60% of the cells also displayed vertical position sensitivity with cosine-like tuning as a function of the orientation of the rotation plane relative to the head. These tuning properties are consistent with inputs from the vestibular nuclei, which show cosine-like spatial tuning of responses to otolith, semicircular canal, and proprioceptive neck afferents (Graf et al. 1993; Kasper et al. 1988; Schor and Angelaki 1992). The rotational velocity sensitivity of these units was consistent with the reported velocity sensitivity of units from the vestibular nuclei in alert primates (e.g., Gdowski and McCrea 1999; Roy and Cullen 2001; Scudder and Fuchs 1992), and all units displayed either cosine-like or rectified cosine-like tuning as a function of the orientation of the rotation axis. The leaky integrator response pattern has not been described in previous studies of the vestibular nuclei in primates, possibly as a consequence of the predominant use of sinusoidal velocity stimulation. The position sensitivity of the responsive units (range: 10.4-103.9 spikes · s¹ · G¹), though, was similar to reported position sensitivity of vestibular nucleus units in alert monkeys (Angelaki and Dickman 2000; Yakushin et al. 1999). Because parabrachial nucleus units were insensitive to eye movement, it is logical to suggest that they reflect input from eye-movement-insensitive vestibular nucleus neurons. Despite these similarities to the properties of vestibular nucleus neurons, it is important to note that neither the sources of the velocity, integrated velocity, and position signals nor relative contributions of semicircular canals, otolith organs, proprioceptors, or enteroceptors to these responses can be resolved unambiguously by the experimental paradigm.

There are three basic orientations for excitatory rotational velocity responses of the sampled parabrachial nucleus neurons. The first population of neurons is excited by rotation about an axis with zero elevation; this is equivalent to a best response in a purely vertical rotation plane. The second group of cells showed maximum excitation to rotational velocity about an axis that produced both vertical and ipsilateral yaw rotation components. The best axes were centered at 0.97 rad of elevation but showed no preference in azimuth. Finally, a third population of neurons was excited best by rotation about an axis that included a contralateral yaw component. These axes were focused at an elevation of 0.42 rad and showed a preferred azimuth that produced best responses for nose-down rotation in ipsilateral anterior canal-contralateral posterior canal plane to pitch planes. In addition, nearly half of the neurons showed position sensitivity with maximal excitation for nose-up rotation between the vertical semicircular canal planes. It is suggested that these signals may contribute to detection of changes in body trajectory relative to gravity during loss of balance, which may trigger homeostatic and affective (e.g., panic responses). For example, an unexpected fall will produce a change in vertical angular velocity and in the position of the head and body relative to gravity. Similarly, generation of a righting response will produce a yaw axis rotation of the body toward a nose-down orientation, resulting in changes in both angular velocity (i.e., velocity and direction of rotation about a body or head axis) and orientation of the head and body relative to gravity. It is suggested that these types of behavioral contingencies can be discriminated by a comparison of signals from position-sensitive and -insensitive neurons with similar dynamic responses to rotation. These comparisons would presumable occur at sites receiving parabrachial nucleus input, such as the central amygdaloid nucleus and limbic regions of neocortex.

The common feature of all sampled PBN neurons was prominent vertical angular velocity sensitivity and a response resembling a leaky integration of velocity sensitivity during whole-body rotation. The dynamic pattern of responses is determined as a summation of spatially independent often rectified velocity and integrated velocity signals. As a result, the responses of a single cell may vary according to the stimulus orientation from a monophasic response to a biphasic response. A monophasic response may reflect velocity alone, integrated velocity alone, or a sum of velocity and integrated velocity components of the same polarity. A biphasic response is produced by velocity and integrated velocity components of opposite polarity. The net result of this phenomenon is that the temporal profile of the response to a position trapezoid often varies with the axis of rotation (e.g., Figs. 5 and 6).

The different patterns of summation of velocity and leaky integrated velocity signals displayed by parabrachial nucleus units produce different temporal filtering characteristics during movements. The responses of simulated units with the either the same or an opposite polarity of velocity and (leaky) integrated velocity responses are shown in Fig. 12A, which simulates responses to rotation about a best axis. The excitatory and inhibitory responses were assumed to be symmetric. Based on the properties of units with a zero elevation best rotation axis, the velocity and integrated velocity gains were assumed to be equal. The simulated unit responses (Fig. 12A, top and middle) are shown for Gaussian velocity profiles ( = 0.04, 0.12, and 0.24 s) with equal peak velocities (normalized to 1). For biphasic unit response (opposite polarity velocity and integrated velocity signals; Fig. 12A, top), the simulated response closely parallels the velocity of the displacement for narrow pulses but becomes attenuated and shifted in phase toward acceleration for wider pulses (associated with larger displacements). Thus cells with opposite polarities of velocity and integrated velocity responses display a quasi-high pass behavior relative to peak acceleration (Fig. 12B). By contrast, the simulated unit with velocity and integrated velocity responses of the same polarity (Fig. 12A, middle) shows a monophasic response that is broader than the velocity of the displacement, with a magnitude that increases for wider pulses (and greater displacements). Hence, it shows a quasi-low-pass behavior relative to peak accelerations (Fig. 12B) but yields transient information about the magnitude of the displacement. Both types of component interactions, though, show equivalent velocity-related responses for rapid, brief perturbations.

View larger version (21K): [in this window] [in a new window]   Fig. 12. Simulation of interactions between signals reflecting angular velocity and leaky integration of angular velocity on parabrachial nucleus neurons. The responses to rotation in the excitatory and inhibitory directions were assumed to be symmetric. The angular velocity and leaky integrator of angular velocity gains were equal. A: the simulated unit responses are shown angular velocity and integrated angular velocity responses of opposite polarity (top) and the same polarity (middle) for Gaussian velocity profiles ( = 0.04, 0.12, and 0.24 s) with equal peak velocities (normalized to 1). The peak excitatory response is plotted as a function of peak angular acceleration in B.

It is important to note that velocity and leaky velocity integrator responses displayed by PBN neurons are frequently clipped (or even fully rectified) in the inhibitory (or nonpreferred) direction of rotation. Further, because the spatial tuning differs between these components, the population of neurons displays a wide diversity of patterns of variation in temporal filtering properties as a