graf

Highlights of Previous Work

My research regarding intrinsic reference frame organization and its expressions within the vestibulo-motor systems are presented in the following. The numbers in parenthesis in the text correspond to particular references in the bibliography (Werner Graf) following this exposé.

1) Vestibulo-oculomotor Relationship
The physical orientation of the sensory and motor periphery of the vestibulo-oculomotor system, the semicircular canals and the extraocular muscles was measured semi-quantitatively in a number of lateral- and frontal-eyed animals (16, 22), and quantified exactly in rabbit and cat (31). It turned out that particular eye muscle pairs (one muscle in each eye) are aligned with that canal from which they receive their principal input. This was true for all vertebrate species (except flatfish).

From the quantitative data (31), the necessary connectivity of the neuronal network underlying compensatory eye movements was calculated utilizing a matrix and a tensor analysis (32, 54). Good agreement between the mathematical analysis and experimental results was found with constancy of the so-called principal vestibulo-oculomotor connections to yoke muscles ("classical Szentagothai" three-neuron reflex arc) and species-specific so-called accessory or secondary connections. These accessory connections are weaker than the principal connections but are necessary, nevertheless, to equalize the slight incongruences between the vestibular sensory and the extraocular muscle effector planes (32, 54).

Simultaneously, the neuronal networks underlying vestibulo-ocular reflexes were investigated with the intracellular horseradish peroxidase (HRP) method at the level of identified second-order vestibular neurons in cat and rabbit, revealing the internal anatomical representation of intrinsic coordinate systems within the vestibulo-oculomotor system and confirming the abovementioned modeling approaches (27, 47, 103).

Sensory-motor transformation in the vestibulo-ocular reflex system was further analyzed using natural vestibular stimulation in all dimensions of physical space to determine the signals in identified second-order vestibulo-oculomotor neurons as well as their morphology. It could be shown that neurons were either tuned in canal planes or in particular eye muscle planes (50, 62, 86, 102). We conclude that vestibulo-oculomotor coordinate transformation takes place in two steps within the three-neuron reflex arc and is carried both by physiological and anatomical means (102).

A theoretical and conceptual treatment on the vestibulo-oculomotor system of vertebrates was offered as a follow-up on the above observations. The significance of the vestibular system as an intrinsic reference frame, and its functional and phylogenetic development was elaborated. In general, self-motion detection systems, such as the semicircular canals of the labyrinth seem to have emerged from evolution according to principles of bilateral symmetry, orthogonality and push-pull operation. These principles have been established in vertebrates and could also be demonstrated for some invertebrates (38, 66, 75, 79).

2) Oculomotor Organization
Intrinsic aspects of oculomotor organization in different vertebrates including fish (elasmobranchs, goldfish, flatfish) were obtained with extra- and intracellular neuroanatomical and electrophysiological methods (33, 39, 40, 60). All vertebrates displayed the familiar oculomotor pattern with four ipsilaterally projecting (lateral rectus, medial rectus, inferior rectus and inferior oblique) and two contralaterally projecting (superior rectus and superior oblique) motoneuron pools, except for elasmobranchs (33). In the elasmobranchs, medial rectus motoneurons also project contralaterally. The latter organization has far reaching consequences for the interpretation of the ontogenetic and phylogenetic development of intrinsic vestibulo-oculomotor circuitry (33, 66).

Individual oculomotor neurons in fish and rabbit were found to lack axon collaterals, in contrast to the situation present in cat and guinea pig. Apart from this difference, the morphology of rabbit oculomotor neurons closely resembles that of cat and guinea pig with regard to soma-dendritic organization (spherical), axonal trajectory, etc., whereas fish motoneurons are distinctly different (polarized dendritic tree). It could be shown that the dendritic complexity of extraocular motoneurons increases in a phylogenetic hierarchical order (60).

One important feature of vestibulo-oculomotor organization concerns its excitatory/inhibitory innervation. In this context, the previously denied existence of inhibition in the vestibulo-oculomotor system of teleosts was demonstrated in goldfish and flatfish by classical electrophysiological techniques (36, 45, 49, 90), and corroborated by electronmicroscopic and immunohistochemical material (45, 49, 53, 55). Electronmicroscopy revealed inhibitory terminals with chemical synapses and pleiomorphic vesicles. Excitatory terminals had spherical vesicles and chemical and electrical synapses. Immunohistochemistry revealed GABA as the transmitter in the vertical inhibitory vestibulo-ocular reflex pathway in fish, and glycine in the horizontal inhibitory loop.

3) Organization of Visual Movement Space
he intrinsic reference frame of the "visual space" involved in optokinetic and visual-vestibular eye movements was studied at the level of the inferior olive and the flocculus in rabbits. These studies demonstrated a close geometrical relationship of the "optokinetic reference frame" with the vestibular and extraocular muscle coordinate systems (17, 71, 72, 87).
Furthermore, the spatial tuning of second-order vestibular neurons in response to optokinetic stimulation was determined. These neurons also had preferred axes in and around the canal and extraocular muscle axes (66, 86).

The accessory optic system feeding optokinetic signals to the vestibular nuclei and the cerebellum was investigated in elasmobranchs and teleost fish with neuroanatomical and electrohysiological methods (37). Unlike the situation in the rabbit or the cat, cerebellar Purkinje cell responses to optokinetic stimulation are not tuned to spatial axes which resemble the semicircular canal orientations but are scattered over a wide range of azimuthal orientations. This orientation tuning is more reminiscent of an otolith or body movement space rather than a semicircular canal or extraocular muscle space (81).

4) Ontogeny of Visual Movement Space
Using dark-reared rabbits, the relationship between complex and simple spike discharge during development was studied by recording from visually driven Purkinje-cells of the cerebellar flocculus. It could be shown that the visual responsiveness of these cells does not depend on prior visual experience. This finding suggests a hard wiring of the visuo-cerebello-vestibulo-oculomotor loop as a postural control system rather than the cerebellum being a learning device. Since visual experience is not necessary for the ontogenetic establishment of the optokinetic reference frame, other intrinsic reference frames are being contemplated to provide the appropriate coordinate axes. Candidates are the semicircular canals and the extraocular muscle proprioceptors (70).

5) Eye Movement Recordings Regarding Velocity Storage
Optokinetically elicited eye movements were studied in a number of teleost and elasmobranch fishes. It could be demonstrated that only the goldfish has responses comparable to higher vertebrates. This finding is important in reference to the question about the general existence of a velocity storage integrator and its function (83).

6) Head-Neck Posture of Vertebrates
The head-neck anatomy of vertebrates was investigated with X-ray photography. Higher vertebrates adopt a stereotyped head-neck posture when at rest: the cervical vertebral column is held oriented vertically and the horizontal semicircular canals are tilted upward from earth horizontal (46). The cervical vertebral column is part of the S-shaped structure of the entire vertebral column. One inflection appears at the cervico-thoracic junction, the other one in the thoracic-lumbar region. Throughout all head-neck movements, the cervical vertebral column retains its original intrinsic configuration. The erect posture and stiffness of the cervical spine are the end product of distinct mechanical constraints within the head-neck ensemble that result in a reduction in degrees of freedom and in a functional compartmentalization of the head-neck movement system.

Subsequently, the biomechanics of the head-neck movement system were studied in man, monkey, cat, guinea pig, rat and rabbit. Mechanical testing following dissection of the CVC in cadavers, and X-ray photography during passive bending in anesthetized animals demonstrated a limited range of lateral flexion, but a large rigidity against dorso- and, in particular, ventro-flexion. The atlanto- occipital joint and the cervico-thoracic junction allow excursions in the pitch plane with little tolerance for movement in the transverse plane. In the resting position, when the cervical column is oriented vertically, the head is held in the extreme pitched-down position of the atlanto-occipital articulation. In this particular position the horizontal semicircular canals are not oriented earth horizontally, but are rather pitched upwards. Head movements result from motion about a limited number of joints. Movements in the horizontal plane occur from rotation about the odontoid process and by rotation of cervical vertebrae relative to each other. Pitch movements result from rotation in the atlanto-occipital articulation and at the cervico-thoracic junction. Side tilts are largely produced by tilting the entire head-neck arrangement at C7/Th1 (67, 80, 93, 94, 106).

The mechanical segmentation of the head-neck ensemble corresponds to a segregation of vestibular inputs. Otolithic input acting at the level of the cervico-thoracic junction (C6-Th2) maintains the vertical orientation of the neck at resting position, whereas the horizontal canals provide the straight-forward orientation of the head (84).

Movement strategy analysis by cineradiography showed a gradual change in the utilization of head-neck joints for head movement control taking place from rabbit to monkey. Rabbits stabilize their heads independently from cervical vertebral column orientation, whereas monkeys (and man), as the other extreme, utilize the cervical column exclusively for head orientation in space. During locomotion, the neck becomes extended. This particular posture brings forward the center of gravity of the animal. As soon as walking ceases the head is brought back into the resting position of the vertically oriented cervical column (93, 94, 107).

We conclude that distinct constraints in the head-neck joints (allowing only 4 degrees of freedom out of possibly 70 or more) are also facilitating and simplifying head movement control.

7) Flatfish Neurobiology
Flatfish offer a natural paradigm for studying adaptive changes in the vestibulo-ocular reflex (VOR). During metamorphosis, flatfish tilt 90 deg to one side or the other to become bottom-adapted adult animals. In this position, the labyrinths are rotated 90 deg relative to their premetamorphic orientation in space. Structurally, this arrangement requires a neuronal pathway from the horizontal semicircular canals to muscles that move the eye vertically. The morphological substrate subserving adaptation of the vestibulo-ocular reflex in post-metamorphic (adult) flatfish was obtained with a number of morphological and electrophysiological methods (28, 41, 90). Single cell morphology of second-order vestibular neurons linked to the right or left horizontal canal, visualized with the intracellular HRP injection method, revealed a qualitatively bilaterally symmetric distribution of neurons with terminals in vertical extraocular muscle motoneuron pools, in either both the superior rectus and inferior oblique subpopulations or with the antagonists to these muscles, the inferior recti and superior obliques (trochlear nuclei). Surprisingly, all second-order neurons had contralaterally ascending main axons. Vertical canal related second-order neurons were not different from that described in other vertebrates. Visualized by extra- and intracellular HRP methods, primary afferents from the semicircular canals terminated in the five subnuclei of the vestibular complex (anterior, magnocellular, descending, tangential, posterior), and also in the eminentia granularis and the medial reticular formation, as described in other teleost fishes (52, 63, 90).

In summary, our data indicate a thoroughly bilateral symmetric distribution of post-metamorphic second-order vestibulo-ocular reflex neurons for both excitation and inhibition linking the horizontal canals to vertical eye muscles. In light of normal primary vestibular afferent, second-order vertical vestibular neuron, and oculomotor organization, the species-specific horizontal vestibular neurons are both necessary and sufficient for adaptation of the vestibulo-ocular reflex in the adult flatfish.

A peculiar reaction to hemilabyrinthectomy was found in the adult flatfish. While lesions of the down-side labyrinth make the animal perform the classical hemilabyrinthectomy symptoms with circling towards the side of the lesion, extirpation of the up-side labyrinth made the animal circle away from the lesion. The neuronal mechanisms and possible consequences for the ontogeny and evolution of the flatfish are presently being investigated. We interpret the peculiar bilaterally asymmetric labyrinthine organization of the adult flatfish to be of fundamental importance for maintaining the flatfish as a flatfish (73, 82, 90).

8) Neuronal Correlates of Space Perception in Monkey Neocortex
The ventral intraparietal area (VIP) of macaque monkeys is located in the fundus of the intraparietal sulcus It bridges the gap between the dorsal visual processing stream (LIP) and the somatosensory and premotor system (MIP). The VIP area in macaques contains neurons responding in a directional selective manner to visual stimuli. Furthermore, VIP neurons also respond to somatosensory stimuli. Receptive field locations for visual and tactile stimuli correspond, and directional preferences for both kinds of stimuli are identical. Area VIP is thus considered to play an important role in the analysis of self-motion and multimodal representation of the three-dimensional movement space. Our recordings revealed directionally selective responses for vestibular stimulation, and to visual as well as tactile stimulation. There was, however, one peculiar characteristic of the visual vestibular interactive processing in these neurons that requires further study. Visual and vestibular on-directions were always in the same direction (not opposite) which, in fact, would generate a visual-vestibular conflict situation. Despite such an a priori mutually inhibitory interaction, the neuronal responses were actually amplified, and not diminished. The same neurons often responded also to optic flow stimuli (expansion and contraction) that simulate self-motion. Many neurons in area VIP also revealed an influence of eye position during active fixation in darkness similar to findings in areas LIP and 7a (125, 137), and also play a role in eye- versus head-centered spatial coding (132). This was shown by mapping visual receptive field structures that could remain retinocentric, or remain stationary in space during eye movements.

9) Transneuronal Labeling of Eye Movement Circuits with Rabies Virus
Retrograde transneuronal tracing with rabies virus (CVS strain) was used to visualize the neuronal circuitry involved in horizontal eye movement control (120, 131). After unilateral injection (1 µl) into the medial rectus muscle (MR) of guinea pigs, viral transfer was studied immunohistochemically at 6 hour intervals from 1.5 to 5 days post-inoculation. Specificity of uptake was demonstrated by double immuno-fluorescence for rabies and choline acetyl transferase (CAT). The kinetics of retrograde transneuronal transfer of the tracer show an exponential increase in categories of labeled neurons over time. The results signal an ever increasing complexity of afferent pathways and control circuits converging onto the different layers of the investigated networks and onto the final motoneuronal pathway in order to produce a desired motor output, i.e., a behavior, in our case horizontal eye movements.

Transneuronal transfer was time-dependent (see enclosed summary diagram). Initially (2 days), only the motoneuronal pathway appears (first-order neurons). At a later stage (2.5-3 days), the transneuronally labeled second-order neurons largely comprise the vestibulo-ocular reflex (VOR) control circuits. However, at this level, additional cell groups are interconnected with this circuitry whose involvement cannot be interpreted in a straight-forward fashion. For example, labeling of the Y group, a nucleus generally associated with vertical eye movements, has to be seen in a recently described context of adaptive plasticity.

At longer times (3.5 days), third-order transneuronal labeling already seems to involve the entire machinery underlying the eye movement repertoire such as saccade generation, optokinetic reflex and motor learning mechanisms, up to and including precursors of cortical perceptive functions (if a guinea-pig possesses such a capacity).

With regard to the olivo-cerebellar system, labeled Purkinje cells (PCs) appeared at 3.5 days in the ipsilateral flocculus (FL) in a single band that ran diagonally from caudomedial to rostrolateral in an intermediate position. This band corresponds to the so-called "horizontal zone". In some cases, labeling continued laterally across the posterolateral fissure into the ventral paraflocculus. At longer survival times (between 3.5 and 4 days), the initial band became slightly broader and additional separate bands of labeled PCs appeared in the rostromedial and caudolateral FL. At these times, an intermediately positioned band also appeared in the contralateral FL that mirrored the one initially labeled in the ipsilateral FL. The time difference in the appearance between these two bands reflects a similar time difference in labeling in ipsi- versus contralateral magnocellular medial vestibular nucleus neurons. After 4 days, the empty areas adjacent to the "horizontal zone" became filled with labeled PCs that were at first scattered, but at 5 days covered basically all areas. This increase suggests involvement of PCs belonging to the "vertical zones" in horizontal eye movement circuits via neurons in the vestibular nuclei or the deep cerebellar nuclei.

The results show the expected and the unexpected. Basically, all neuronal relay stations involved in horizontal eye movement control have been labeled. Quite unexpectedly, labeling involves also a number of nuclei which are thought to be only involved in vertical eye movement control (Y group, interstitial nucleus of Cajal). In general, horizontal and vertical systems comprise different functional entities, although cross-activation has been demonstrated and interpreted as a necessity for spatial coordination of three-dimensional eye movements and for vertical/ horizontal adaptive plasticity. Other labeled regions have not as yet been associated with an eye movement control function, such as the dorsal tegmental nucleus or certain limbic structures. Their role clearly remains to be determined. For example, labeling of the hypothalamus may be explained by its known projections to the horizontal saccade related nucleus reticularis cuneiformis (RCf).
Another unexpected finding is the absence of labeling at 3 days in areas that could be expected to contain it, such as the visual relay nuclei mediating optokinetic reflexes (nucleus of the optic tract, NOT; accessory optic system, AOS). For some of these nuclei, direct projections to oculomotor motor neurons have been postulated. However, the labeling of these nuclei at 3.5 days would suggest the existence of more polysynaptic relays. Noticeably, the only structure of the AOS labeled at 3.5 days is the contralateral dorsal terminal nucleus (DTN), that only contains horizontal visually direction-selective neurons. The lateral terminal nucleus (LTN) and medial terminal nucleus (MTN) are related to vertical directions and are not labeled.

All in all, the transneuronal labeling technique with rabies virus is a powerful tool to determine entire functional neuronal networks involved in a specific behavior.

Summary diagram