THE NEURAL INTEGRATORS OF THE MAMMALIAN SACCADIC SYSTEM

A.K. Moschovakis

Division of Computational Neuroscience,Institute of Applied and Computational Mathematics, FO.R.T.H., and Dept. Basic Sciences, Faculty of Medicine, University of Crete, P.O. Box 1393, Crete, Greece

Received 8/11/97 Accepted 9/29/97

5. INPUT-OUTPUT CONNECTIONS

To play the role expected of them, the neural integrators of the saccadic system must receive input from medium lead burst neurons (MLBs) and send their output to motoneurons. The discharge pattern of integrator cells is compared to that of MLBs in Section 5.1 and to that of motoneurons in Section 5.3. Evidence that supports the existence of connections between the vertical (and horizontal) MLBs and the neural integrators is presented in Section 5.2 while connections between the neural integrators and extraocular motoneurons are described in Section 5.4. Because some of the motoneuronal position sensitivity could be due to signals they receive from secondary vestibular neurons, projections of secondary vestibular neurons to motoneurons are described in Section 5.5. The NIC, the NPH and the vestibular nuclei could collectively form a distributed network that carries out the process of integration; interactions between these nuclei are described in section 5.6.

5.1. Comparison with MLBs

Medium lead burst neurons (MLBs) are instrumental for providing the neural integrators with the pulse of activity they need to integrate. In alert animals, MLBs discharge before and during rapid eye movements (typically, latencies are less than 12 ms). MLBs can be divided into several subclasses on the basis of their preferred direction, the location of their somata, the trajectory of their axons and the excitatory or inhibitory influence they exert on their targets. On the basis of their preferred directions, MLBs of each half of the brain can be divided into three separate groups: 1) horizontal ipsilateral, 2) upward and 3) downward. The horizontal MLBs encode the size of horizontal saccadic components and are located in the PPRF near the abducens nucleus (3, 4, 80-83). The upward and downward MLBs (UMLBs and DMLBs) encode the size of vertical saccadic components and most of them are located in the riMLF of the rostral mesencephalon (5-9, 54, 84, 85). Each one of the three groups of MLBs (horizontal, up and down) is further subdivided into excitatory and inhibitory neurons. In the case of horizontal cells, these are referred to as EBNs and IBNs, respectively.

There are strong projections of vertical MLBs (VMLBs) onto vertical motoneurons (#1, figure 4) as well as of horizontal MLBs (HMLBs) onto horizontal motoneurons (#3, figure 4). These will not be considered here as they have been considered in recent reviews (1, 86, 87). Suffice it to say that excitatory MLBs are one of the strongest sources of saccade related input to motoneurons. Accordingly, their discharge pattern carries enough information to specify parameters of saccades. The strongest correlation between a parameter of discharge and a saccadic parameter is the one between the number of spikes in the burst (Nb) and the amplitude of the horizontal or vertical component of saccades (average values in different samples of MLBs range from 0.65 6 to 0.99 88). On the average, the slope of this relationship can be as high as 2.3 (89) or as low as 0.68 spikes per degree of eye displacement (88). Values obtained in other laboratories are summarized in table 4. Secondly, there is a roughly 1:1 relationship between MLB burst duration (Bd) and saccade duration (Sd), whatever the animal species or the cell type (rhesus EBN, ref. 2; rhesus IBN, ref. 90; cat IBN, ref. 83; squirrel monkey EBN, ref. 4). Finally, the average firing rate of MLBs is well correlated with the peak horizontal or vertical velocity of the eyes. Typical average slopes of these relationships equal 0.5 spikes/s per degree/s for monkey EBNs (89) and 0.36±0.24 spikes/s per degree/s for cat EBNs (91). These relationships are due to the connections that these cells establish with extraocular motoneurons and reflect properties of the burst generating networks in which they participate.

Figure 4. Schematic diagram of the input-output connections (arrows) of the neural integrators of the mammalian saccadic system. Evidence to support their existence is summarized separately for each one of them (identified by numbers in circles) in Section 5. Abbreviations: DE, eye displacement; , eye velocity; , head velocity; HMLB, horizontal medium lead burst neurons; Vest, vestibular nuclei; VMLB, vertical medium lead burst neurons. Other abbreviations as in figure 1.

5.2. MLB projections to the neural integrators

The existence of connections between vertical MLBs and the vertical velocity to position integrator (#2, figure 4) is supported by evidence indicating that intraaxonally HRP injected axons of single functionally identified vertical MLBs ramify extensively within the NIC, on occasion bilaterally (7, 8). Similarly, the existence of connections between horizontal MLBs and the horizontal velocity to position integrator (#4, figure 4) is supported by the disclosure of ipsilateral PPRF projections to the regions that house the horizontal neural integrator (the VN/NPH complex) in retrograde (92, 93, 94, 95) and anterograde (96, 97, 98, 99, 100) tract tracing studies. The pattern of termination of individual intraaxonally HRP injected primate EBN axons in the VN/NPH complex also supports this notion (3). Figure 4 does not include connections between burst generators and neural integrators with opposite on-directions. Such inhibitory connections could account for the bi-directional modulation of the neural integrators. The existence of a connection between the IBNs of one side and the HNI of the opposite side is supported by the disclosure of projections from the IBN area of the reticular formation of the medulla to the contralateral VN/NPH complex with the help of retrograde (94, 95) and anterograde (97) tract tracing techniques. It is also supported by the disclosure of antidromic IBN responses to the electrical stimulation of the VN/NPH complex in the cat (82). Also, peri-spike time histograms of prepositus and vestibular nucleus neurons (specifically, type II neurons) aligned on single spikes of IBNs, demonstrate that the activity of the former is depressed when the latter discharge (82). Finally, it is supported by the fact that the terminal fields of single functionally and morphologically identified IBNs are distributed in the contralateral medial vestibular nucleus and the nucleus prepositus in both the monkey (3) and the cat (83). Apparently, the circumscribed region in the squirrel monkey VN/NPH complex that receives a heavy projection from ipsilateral EBNs, namely the region of the medial vestibular nucleus immediately caudal to the abducens nucleus, also receives input from contralateral IBNs (3).

5.3. Comparison with motoneurons

Extraocular motoneurons emit bursts of spikes for saccades in the pulling direction of the muscle they innervate and discharge tonically in proportion to the eye position reached at the end of the saccade. In general, their discharge can be described by the expression,

FR = F0 + kE + r (4)

where F0, k and r have the same meaning as in Eqs. 1 and 2. The values they obtain for different motoneurons have been estimated in several labs for both the cat and the monkey (summarized in table 5).

Abducens motoneurons have been the object of particular interest. Their average position sensitivity (column 4, table 5) is somewhat higher in the cat (range: 4.4 - 8.7 spikes/s/deg) than in the monkey (range: 3.5 - 6.2 spikes/s/deg). It is also higher than the average position sensitivity of burst-tonic NPH neurons (cat: 2.1 - 8.2 spikes/s/deg; monkey: 2.4 - 3.2 spikes/s/deg; column 4, table 2). In the cat, recruitment thresholds of abducens motoneurons are close to the primary position (mean±SD: -3.3 ± 5.2 deg, range: -19 to 7; ref. 101). Their relatively low firing rate at primary position (column 5, table 5) is due to the fact that several cells start firing at positions more lateral than the primary position (these occupy the range between -78 and 0 spikes of table 5). The discharge of primate abducens motoneurons at primary position (36-108 spikes/s; column 5, table 5) is higher than that of NPH cells in the same species (47-57 spikes/s; column 5, table 2) as well as that of abducens motoneurons in the cat (about 23 spikes/s on the average; ref. 101). This could be due to the contralateral eye positions occupied by the recruitment thresholds of primate abducens motoneurons. Early estimates of the recruitment threshold of primate abducens motoneurons ranged from extreme medial (40 - 45 deg in the off-direction) till between 10 - 25 deg lateral (102, 40). More than half of the units had been recruited by the time the eyes reached 20 deg in the off-direction and all of them were recruited at primary position (threshold range: -60 deg to -5 deg) in a more recent detailed study of 39 abducens nucleus neurons (103). Similar is the picture emerging from yet another study of abducens motoneurons (104); here, thirty per cent of the abducens motoneurons were already active while the eyes were adduced by 45 deg and all units but one were active at primary position. Consistent with the notion that bigger abducens motoneurons are progressively recruited as the eyes occupy more abducting positions (size principle), the slope of the rate position curve has been found to increase with the recruitment threshold (correlation coefficient: cat = 0.63, ref. 101; monkey = 0.81, ref. 104; 0.78, ref. 103) at a rate of 0.32 spikes/s/deg/deg in the cat (101) or 0.11 spikes/s/deg/deg in the monkey (103) as well as the conduction velocity of the cell (correlation coefficient: 0.46; ref. 101). A reasonable correlation (0.67) has also been found between the slope of the rate-velocity curve of primate abducens motoneurons and their recruitment threshold (104).

The average position sensitivity of vertical extraocular motoneurons (4.2 spikes/s/deg; ref. 54) is also higher than the average position sensitivities of tonic and burst-tonic NIC neurons (2.0 - 2.6 spikes/s/deg; column 4, table 1). Since the pulling directions of vertical muscles are not strictly vertical, a correlation between the discharge of vertical motoneurons and horizontal eye position should come as no surprise. Indeed, trochlear motoneurons modulate their discharge in proportion to the lateral angular deviation of the eyes as well as with downward eye position. The relationship however is variable and, with the exception of one cell (with a slope of 5 spikes/s/deg), the slope of the rate-horizontal-eye-position curve is much more shallow (<1.3 spikes/s/deg (105). Vertical motoneurons have a higher discharge rate at primary position (105 spikes/s; ref. 54) than the NIC cells (about 80 spikes/s; ref. 54). As in the horizontal system, the high rate of motoneurons at primary position could be due to their low thresholds. Early estimates of recruitment thresholds in a sample of 35 primate oculomotoneurons, ranged from 62 deg in the off-direction till 21 deg in the on-direction (10), while cells were recruited at a rate of 10% of the motoneuron pool for every 10 deg that the threshold moved in the pulling direction of the muscle they innervate. Similarly, 8/27 trochlear axons discharged for upward eye positions beyond 30 deg, 2 units were recruited at about 10 deg down, while the remaining 17 units were recruited somewhere in between (at a rate of 18% of the motoneuron pool for every 10 deg that the threshold moved downward; 105). A more recent study of 78 cells, has demonstrated that primate vertical motoneurons start being recruited at 97 deg in the off-direction and virtually all are active by the time the eyes reach primary position (mean threshold ± SD: -25 ± 18 deg; ref. 54). Further excursions of the eye in the on-direction are apparently due to an increase of firing rate alone. As with abducens motoneurons, the slope of the rate-position curve of both oculomotor (10) and trochlear motoneurons (105) has been found to increase with recruitment threshold.

The saccade related bursts of abducens motoneurons can reach up to a frequency of 200 - 830 spikes/s, (mean: 400 spikes/s; ref. 102) and precede saccades by 8.9±2.8 ms in the cat (101) and 5.4 ms in the monkey (102). There is a positive correlation between the peak frequency and the latency of abducens MN saccade related bursts. The higher the frequency reached by a unit the earlier it starts firing, so that abducens motoneurons that emit the most intense bursts can precede sluggish ones by as much as 3 ms (104). For saccades in the off-direction, activity decreases or pauses with a latency of 14.8 ± 4.05 ms in the cat (101) and 6.3 ms in the monkey (102).

The latency of bursts emitted by primate oculomotoneurons is somewhat shorter. On the average, they precede saccades by 4.4 ± 2.2 ms (mean±SD) when they are recorded inside the nucleus (54) or by 3 ms (range: 0 - 5.5 ms) when they are recorded from fibers of the trochlear nerve (105). These values do not differ from the mean latency of the bursts emitted by NIC burst-tonic neurons (cf. Section 4.1) or vertical MLBs (84, 6, 7, 8) in the same species. The peak frequency reached during bursts differs from one MN to the next. Some trochlear motoneurons reach frequencies that do not surpass 200 spikes/s while others reach frequencies as high as 550 spikes/s; the peak frequency of most of them is in the range of 350 - 450 spikes/s (105). The same is true of oculomotoneurons; their peak frequencies can be as low as 150 spikes/s for some cells and as high as 620 spikes/s for other (mean: 377, n = 35). Actually, the histogram of maximum burst rates during saccades seems to follow a bimodal distribution with one peak at about 250 spikes/s and a second peak at about 400 spikes/s (10). As, with MLBs, the number of spikes emitted by extraocular motoneurons during such bursts is proportional to the size of saccades in their pulling directions; slopes range between 0.25 - 2.5 spikes per deg (r: 0.8 - 0.9, n = 20; ref. 106). For off direction saccades, units usually pause with a mean latency of 6.7 ms (range 2.5 - 11 ms; ref. 105). As with NIC cells, some primate oculomotoneurons slow down instead of pausing, in particular after presaccadic discharge rates higher than 100 spikes/s (105).

5.4. Integrator projections to motoneurons

The notion that the vertical neural integrator projects to vertical extraocular motoneurons is supported by considerable evidence. Older efforts to demonstrate NIC projections to vertical extraocular motoneurons involved electrical lesion or stimulation of the NIC to visualize degenerating fibers into the oculomotor nucleus in the monkey (107) and the cat (108), or record PSPs intracellularly from vertical extraocular motoneurons in the cat (109). These results were not compelling because fibers originating elsewhere (e.g., from UMLBs and DMLBs) and deploying terminal fields in areas occupied by vertical extraocular motoneurons can be lesioned or activated inside the NIC. Recent results provide more confidence in the existence of connections between the NIC and contralateral oculomotor and trochlear nuclei (#5, figure 4). Their existence is supported by autoradiographic evidence in the cat (110) and the retrograde labelling of NIC neurons (mainly contralaterally) following bulk injections of HRP in the oculomotor nucleus of several mammalian species (rabbit, 111; cat, 112; monkey, 5, 73). Moreover, it is supported by the recovery of transynaptically labeled NIC neurons after the retrograde transport of tetanus toxin injected into the IO and IR muscles of the rabbit (113). Finally, injections of PHA-L and biocytin in the NIC disclose a strong contingent of fibers that cross to the opposite side with the posterior commissure (PC) to terminate in the contralateral oculomotor and trochlear nuclei (114) as well as the NIC.

To examine what are the signals that NIC cells send to the oculomotor complex, output fibers of the NIC were studied intraaxonally in alert monkeys and then injected with a tracer (either horseradish peroxidase or biocytin) and studied light-microscopically after proper histochemical processing of the brain (115). Figure 5 illustrates the discharge pattern of one such neuron. It displays a burst-tonic discharge pattern, with tonic intersaccadic activity, pauses for upward saccades, and bursts before downward saccades (latency: 5.74 ± 4.4 ms, mean ± S.D.). A plot of the same cell's mean firing rate (figure 6A, ordinate) versus mean intersaccadic vertical eye position (abscissa) demonstrates an excellent linear relationship between the two variables (r = 0.94). In contrast, the firing rate of this cell was weakly related with the mean horizontal position that the eyes occupied between saccades (right inset of figure 6A). To establish whether this neuron belongs to the "regularly" or the "irregularly" discharging class of NIC cells (54), the coefficient of variation of interspike intervals (CV = SD/mean) was computed for different intersaccadic intervals (115). As shown in the left inset of figure 6A, the SD of this neuron's interspike intervals increased in proportion to the mean interspike interval. Its C.V. was equal to 0.04 at primary position which indicates that it is a regularly discharging cell. Parameters of the same neurons bursts were also related to parameters of saccades. Figure 6B illustrates the excellent linear relationship (r = 0.88) between the number of spikes in the burst and the size of the downward component of saccades. No relationship was found between the number of spikes in the burst and the size of the horizontal component of saccades (inset of figure 6B). An excellent correlation (0.94) was also found between the duration of its bursts and the duration of saccades (figure 6C). Finally, an excellent correlation (r=0.97) was found between its firing rate and the vertical velocity of the eyes during smooth pursuit (figure 6D).

Figure 5. Discharge pattern of the NIC neuron illustrated in figure 7 (modified from ref. 115, with permission). Traces from top to bottom illustrate vertical (V) and horizontal (H) instantaneous eye position and instantaneous firing rate.

Figure 6. Quantitative analysis of the relationship between saccade metrics and the discharge pattern of the NIC neuron illustrated in figure 7 (modified from ref. 115, with permission). A) Plot of the mean intersaccadic firing rate (ordinate) versus mean intersaccadic vertical eye position (abscissa). Plots of firing rate (ordinate) versus mean intersaccadic horizontal eye position (abscissa) and S.D. (ordinate) versus mean (abscissa) of interspike intervals (ISI) are illustrated as insets (right and left, respectively). B) Plot of the number of spikes in the burst (ordinate) versus size of the downward component of saccades (abscissa). A plot of the number of spikes in the burst (ordinate) versus size of the horizontal component of saccades (abscissa) is illustrated as inset (solid circles). C) Plot of burst duration (ordinate) versus saccade duration (abscissa). D) Plot of firing rate (ordinate) versus smooth pursuit vertical eye velocity (abscissa).

A camera lucida reconstruction of the same cell is illustrated in figure 7. The axon emerged from a small soma in the NIC and run caudally to decussate with the posterior commissure. It then run ventrally around the edge of the periaqueductal gray (PAG) to enter the contralateral NIC where it ramified into preterminal branches before continuing towards the contralateral oculomotor nucleus. The fact that terminal fields in the contralateral NIC, as well as the oculomotor and trochlear nuclei virtually disappeared when the posterior commissure was lesioned before a biocytin injection in the NIC, indicates that the posterior commissure could be the conduit of most of the NIC output to these nuclei (114). The integrity of these fibers is necessary for normal velocity to position integration in the vertical plane, as demonstrated by the fact that inactivation of the posterior commissure disabled vertical gaze holding, and advanced the phase while reducing the gain of the vertical VOR in the dark (116).

Figure 7. Camera lucida reconstruction of the initial several millimeters of the axonal system of an NIC neuron (modified from ref. 115, with permission). Abbreviations: III, oculomotor nucleus; Aq, aqueduct of Sylvius; NIC, interstitial nucleus of Cajal; PAG, periaqueductal grey matter; pT, pretectum; RN, red nucleus.

The notion that the horizontal neural integrator projects to horizontal extraocular motoneurons (#6, figure 4) is also supported by considerable evidence. A projection of the NPH to the abducens nucleus has been demonstrated with the help of both retrograde (117, 118) and anterograde (97, 94) bulk tracer injection techniques in the cat as well as in the monkey (95). It is also supported by the demonstration that feline neurons of the rostral NPH, which reliably encode horizontal eye position and/or display type II responses to vestibular stimulation, have been backfired from the abducens nucleus in the cat; typical latencies of antidromic responses are 0.23 ± 0.04 ms (68) for the former and about 0.6 ms (119) for the latter. Field potentials recorded in the abducens nucleus and triggered by spikes of position-velocity cells of the ipsilateral (contralateral) NPH nucleus are consistent with a small distal excitatory (inhibitory) connection (120). When the spikes of these cells are used as events to generate peri-event time histograms for physiologically identified abducens motoneurons, a monosynaptic (latency = 0.76 ± 0.05 ms;range 0.7 -0.9 ms) facilitation of the latter is revealed (68). Similar evidence supports the existence of an inhibitory connection between eye position encoding neurons of the rostral NPH and contralateral abducens motoneurons (68). Moreover, an inhibitory projection of NPH cells to the contralateral abducens nucleus is supported by the fact that injection of tritiated glycine (a putative inhibitory neurotransmitter that is taken up by the terminals which release it) is transported to the cell bodies they originate from in the contralateral NPH (121). Finally, it is supported by descriptions of the somatodendritic and proximal axonal morphology of NPH neurons following their intrasomatic injection with horseradish peroxidase (122, 94). With this means, three morphologically distinct classes of cells have so far been shown to inhabit the NPH: 1) Multidendritic cells of the caudal part of the nucleus with medium or large size somata and many dendrites with rich branches; their thick axons do not deploy recurrent collaterals and project to the cerebellum. 2) Principal cells of the rostral part of the nucleus with medium size somata and relatively few sparsely branched dendrites; their axons deploy terminal fields in one side of the brain stem (including the NPH). 3) Small cells of the dorsolateral NPH which issue recurrent collaterals and bilateral projections. Of the three, it is the principal cells that have been shown to distribute terminal fields within the abducens nucleus of the ipsilateral (or the contralateral) side of the brain (94) and could thus be output elements of the horizontal neural integrator. A camera lucida reconstruction of one such neuron is illustrated in figure 8.

Figure 8. Camera lucida reconstruction of the initial several millimeters of the axonal system of a principal cell of the nucleus prepositus hypoglossi on two frontal sections through the brain stem with stereotaxic coordinates P 6.0 (A) and P 7.5 (B), respectively (modified from ref. 94, with permission). Stippled lines indicate borders of nuclei. Crosses surrounded by circles indicate points where the axon assumes a rostral trajectory. Abbreviations: VI, abducens nucleus; VIIn, facial nerve; DVN, descending vestibular nucleus; LVN, lateral vestibular nucleus; MVN, medial vestibular nucleus; RB, restiform body.

Consistent with the vertical and oblique on-directions of some NPH neurons, an excitatory, monosynaptic projection of the NPH to trochlear motoneurons has been shown with intracellular recording methods. This projection survives chronic parasagittal cuts of the brain stem designed to interrupt the ipsilateral inhibitory projection of the superior and the contralateral excitatory projection of medial vestibular nucleus to the trochlear nucleus (123). An NPH projection to areas of the oculomotor complex that house vertical motoneurons has also been demonstrated with bulk tracer injection techniques (92).

5.5. Motoneuronal projections of secondary vestibular neurons

The pattern of projections of secondary vestibular neurons onto motoneurons (#7 and #9 of figure 4, for vertical and horizontal neurons, respectively) depends on the input they receive from the peripheral end-organ and the nucleus that houses the cell bodies they originate from. Cells of the medial vestibular nucleus which respond to stimulation of the posterior canal project to the contralateral superior oblique and inferior rectus subdivisions of the oculomotor complex in both the rabbit and the cat (124). On the other hand, cells of the superior vestibular nucleus which respond to stimulation of the posterior canal project to the ipsilateral superior rectus and inferior oblique subdivisions of the oculomotor complex in the cat (124). In addition, cells which respond to stimulation of the anterior canal project to the contralateral superior rectus and inferior oblique (cells of the medial vestibular nucleus) or to the ipsilateral superior oblique and inferior rectus (cells of the superior vestibular nucleus) subdivisions of the oculomotor nucleus (125). The cells of the medial vestibular nucleus that participate in these projections are excitatory as indicated by the polarity of averages of field potentials recorded in the oculomotor complex and triggered from spikes of vestibular cells (126). In the horizontal system, neurons of the medial vestibular nucleus supply the ipsilateral (inhibitory cells) or the contralateral (excitatory cells) abducens and prepositus nuclei. This pattern of projections is consistent with the disynaptic EPSPs (IPSPs) that have been recorded from abducens motoneurons in response to stimulation of the contralateral (ipsilateral) vestibular nerve and the monosynaptic EPSPs (IPSPs) recorded from the same cells in response to the electrical stimulation of the contralateral (ipsilateral) medial vestibular nucleus (127). It is also consistent with averages of field potentials recorded in the abducens nucleus that were triggered from spikes of physiologically identified medial vestibular nucleus cells (antidromic latency equals 0.2 ms ± 0.05 for ipsilateral cells and 0.37 ms ± 0.1 for contralateral cells; ref. 68). In addition, Type I horizontal secondary vestibular neurons course with the ipsilateral ascending tract of Dieters to terminate in the medial rectus subdivision of the oculomotor nucleus (128).

With the exception of the pure vestibular cells, axons arising from all other functional classes of vertical secondary vestibular neurons (cf. Section 4.3) ascend through the MLF at least till the level of the trochlear nucleus (69). Accordingly, it is important to ask which of them establish connections onto vertical extraocular motoneurons (#7, figure 4). Unfortunately, only the pattern of terminations of PVPs has been established with the help of intraaxonal tracer injections first in the monkey (129) and later in the cat (130, 131). The pattern of axonal terminations of upward primate PVPs is illustrated in figure 9. Inhibitory cells of the superior vestibular nucleus have been shown to project to the ipsilateral inferior rectus and superior oblique subdivisions of the ipsilateral oculomotor complex. Conversely, excitatory upward primate PVPs of the medial and ventrolateral vestibular nucleus have been shown to project to the superior rectus and inferior oblique subdivisions of the contralateral oculomotor nucleus. Excitatory downward PVPs (figure 10) have been shown to project to the inferior rectus and superior oblique subdivisions of the contralateral oculomotor complex. When averages of field potentials recorded in the oculomotor nucleus are triggered from spikes of feline downward position-vestibular neurons (DPV) excitatory monosynaptic responses are evoked in the contralateral inferior rectus and superior oblique subdivisions and inhibitory responses are evoked in the ipsilateral superior rectus subdivision of the oculomotor complex (131). When DVP axons were injected with a tracer, terminals were observed in either the inferior rectus and superior oblique subdivisions of the oculomotor complex contralateral to the soma they originate from or in the ipsilateral superior rectus and inferior oblique subdivisions (131). The importance of these fibers is shown by the fact that pontine MLF lesions (which interrupts the ascending projections of vertical secondary vestibular neurons) cause vertical gaze nystagmus (24).

Figure 9. Schematic illustration of the axonal trajectories and projections (arrows) of one up PVP of the SVN (solid circle) and one up PVP of the MVN-VLVN (open circle; modified from ref. 129, with permission). Stippled lines are borders between nuclei. Abbreviations: IV, Trochlear nucleus; XII, hypoglossal nucleus; dPPRF, dorsal part of the paramedian pontine reticular formation; DR, caudal part of the dorsal raphe nucleus; IVN, inferior vestibular nucleus; NPH, nucleus prepositus hypoglossi; R, Rollers nucleus; RO, nucleus raphe obscurus; SVN, superior vestibular nucleus. Other abbreviations as in figures 7 and 8.

Figure 10. Schematic illustration of the axonal trajectory and projections of one down PVP of the MVN-VLVN (modified from ref. 129, with permission). Symbols and abbreviations as in figure 9.

As with the vertical system, of the several classes of functionally identified horizontal secondary vestibular nucleus neurons, it is the PVPs that have been shown to project to extraocular motoneurons with the help of intraaxonal tracer injections in the monkey (132). The pattern of axonal terminations of horizontal PVPs is illustrated in figure 11. In terms of vestibular responses, they are type I cells and they either project contralaterally to supply the abducens nucleus or they course with the ascending tract of Dieters to invade the ipsilateral subdivision of the oculomotor complex. McCrea and his colleagues were also successful in recovering BP units of the MVN projecting to the ipsilateral abducens and prepositus nuclei (132). A schematic illustration of the pattern of axonal projections of one such cell is illustrated in figure 12. Although their vestibular responsiveness is uncertain, these cells increase their discharge for ipsiversive abduction of the eyes; they are thus unlikely to correspond to the still elusive neuron that implements the inhibitory middle leg of the horizontal VOR. The fact that the primary vestibular projection to horizontal extraocular motoneurons is the excitatory one between the PVPs and the contralateral abducens motoneurons was demonstrated when averages of lateral rectus EMG were triggered from spikes of functionally identified secondary vestibular neurons (79). A small participation of BP, P and EHV cells in the same projection and a small inhibitory projection arising from ipsilateral PVP and EHV cells was also demonstrated in the same study.

Figure 11. Schematic illustration of the axonal trajectory and terminations of one ipsilateral (solid circle) and one contralateral (open circle) horizontal PVP of the MVN-VLVN (modified from ref. 132, with permission). Symbols and abbreviations as in figure 9.

Figure 12. Schematic illustration of the axonal trajectory and terminations of an ipsilateral projecting horizontal burst-tonic cell of the MVN (modified from ref. 132, with permission). Symbols and abbreviations as in figure 9.

5.6. Coennections between the vestibular nuclei, the NIC and the NPH

Besides supplying extraocular motoneurons with an eye position signal, secondary vestibular neurons supply the neural integrators with head velocity signals. The output of the neural integrators signal is then conveyed back to secondary vestibular cells and forward to extraocular motoneurons. Accordingly, the neural integrators of the saccadic system would contribute to integration in the vestibular system. This is a reasonable notion considering the responsiveness of NIC and NPH cells to vestibular stimuli and the symptoms of vestibular disease that follow their lesion (cf. Section 3).

The place of different classes of cells in the hierarchy of neural events that underlie the process of integration in the vestibular system can be estimated from of a simple calculation: it concerns the phase difference between head acceleration (the input to the vestibular system) or eye position (the output of the system) and neuronal firing rate as a function of the frequency of stimulation. Plots of phase (and gain) against frequency (Bode plots) are an important ingredient of the analysis of linear time invariant systems. Besides, they are conceptually related to the first order differential equations that describe the discharge of oculomotor related cells (e.g., Eqs. 1-4). Take for example the last one which describes the discharge of extraocular motoneurons. Letting eye position (E) be a sine wave (sin(omega.t)), where omega is the frequency of rotation in radians per second, t is time, and omega.t = fi (the phase difference between eye position and neuronal firing rate), then substituting into Eq. 4 and solving for fi when FR = F0, demonstrates that fi = tan-1(omega.r/k), where k and r have the usual meaning of neuronal position and velocity sensitivity. The phase difference (fi) between eye position and firing rate can be determined experimentally. This has been done for several frequencies and classes of oculomotor related cells (tables 1-3 and 5). For the sake of simplicity, I will concentrate on one frequency of rotation (roughly 0.2 Hz) which has been employed by several workers.

Starting with the input side of the system, the discharge of feline secondary vestibular neurons leads the sinusoidally modulated eye position by about 100 deg in horizontal cells (133, 68) and about 70 deg in vertical cells (134, 130). In contrast, the discharge of extraocular motoneurons is almost in phase with eye position (cf. table 5). It is this phase shift that is contributed by the neurons that comprise the neural velocity to position integrators. The phase difference between their discharge and eye position is summarized in tables 1 and 2. Estimates range between 10 - 60 deg of phase lead re eye position for NPH cells (cf. table 2) and about 40 deg for burst-tonic and irregular-tonic cells of the primate NIC (54). It must be noted that the vestibular modulation of the discharge of NPH BT and T cells and NIC BT cells must be related to the vertical eye movements they cause rather than the vestibular signals they receive, because it does not occur during VOR suppression (54, 64).

The fact that the NIC and the NPH can influence the response properties of vestibular nuclear cells and the VOR while NIC and NPH cells respond to vestibular stimuli would lead one to suspect the existence of reciprocal connections between the NIC the NPH and the vestibular nuclei. Indeed, NIC cells receive input from secondary vestibular nucleus neurons as indicated by the fact that they are di-synaptically excited from the contralateral vestibular nerve, in both the cat and the monkey (135, 136). There is some controversy concerning the signals carried by and the functional identity of the thus affected NIC cells; it is thought to be head velocity (pitch) cells according to some authors (137) and position cells according to others (138). However, this distinction is unlikely to be clear cut given the fact that there is no correlation between a neurons phase lead and its position sensitivity or CV (55).

Starting with the early reports of Muskens (139) and Lorente de N— (140), a projection from the vestibular nuclei to the NIC (#8, figure 4) has been consistently described over the past half century by many authors with the help of several techniques (reviewed in ref. 26). This projection primarily originates from the ipsilateral superior and the contralateral medial vestibular nucleus as demonstrated by the disclosure of retrogradely labelled cell bodies following biocytin injections in NIC (114). The projection that originates in the superior vestibular nucleus is inhibitory and the one that originates in the medial vestibular nucleus is excitatory as indicated by the fact that the electrical stimulation of these nuclei evokes monosynaptic IPSPs and EPSPs, respectively, in NIC neurons (141, 142, 136). Consistent with this, electrical stimulation of the labyrinth evokes disynaptic EPSPs in the contralateral NIC and disynaptic IPSPs in the ipsilateral NIC (136). Cells located in the contralateral medial and the ipsilateral superior vestibular nucleus which receive posterior canal information have been shown to project to the NIC in both the rabbit and the cat, on occasion bilaterally (124). Since as summarized in Section 4.3 secondary vestibular neurons are functionally diverse, it is important to ask which of them project to the NIC. So far it is vertical PVPs located in either side of the brain and thus corresponding to both excitatory and inhibitory cells, which have been shown to project to the NIC in both the cat (131) and the monkey (129), on occasion bilaterally.

Early anatomical descriptions of an NIC projection to the vestibular complex (#11, figure 4) were based on anterograde degeneration techniques in the cat (143, 108) and monkey (107), as well as the retrograde (74) and anterograde transport of tracers (144, 110). More recently, NIC projections to the vestibular nuclei (mainly the ipsilateral superior and medial vestibular nuclei) were demonstrated with injection of biocytin and PHAL in the NIC (114). The diameters of biocytin labelled NIC fibers that were traced to the vestibular nuclei (mean: 2.3 microns; SD: 1.1; N = 70) agree remarkably well with the conduction velocities of NIC cells that were antidromically activated from the MLF near the abducens nucleus (12 - 16 m/s; ref. 135). The projection of NIC neurons to the vestibular nuclei is excitatory (145). Yet, activation of NIC neurons inhibits type I neurons of the ipsilateral vestibular complex, presumably due to the excitatory influence they exert upon type II vestibular neurons (146). The importance of this projection is indicated by the fact that bilateral NIC lesions in the cat disrupt the responses of secondary vestibular neurons to head rotation. Thus, the depth of modulation (gain) of horizontal vestibular neurons decreases following lesions of the NIC but the amount of the phase by which their discharge lags head acceleration during sinusoidal vertical rotation remains unchanged (147). In contrast, the gain (re acceleration) of vertical secondary vestibular neurons decreases (mean±SD: 1.27±0.68 spikes/s/deg/s2; range: 0.22 - 2.9 spikes/s/deg/s2) by comparison to what is the case in normal cats (mean±SD:2.07±0.78 spikes/s/deg/s2; range: 0.88 - 3.3 spikes/s/deg/s2). Furthermore, the amount of the phase by which the discharge of vertical vestibular neurons lags head acceleration during sinusoidal vertical rotation of the head decreases by 30 deg following such lesions (26).

There is some information concerning the signals carried by (and the functional identity of) NIC cells that project to the vestibular nuclei. For example, NIC cells that encode down (but not up) eye position and are di-synaptically excited from the contralateral labyrinth have been backfired from the region of the ipsilateral medial vestibular nucleus (138). Additional information, largely rests on the physiological identification of NIC cells from the MLF at the level of the abducens nucleus. This applies to cells whose discharge is more deeply modulated by rotations in the plane of the vertical canals (134) as well as a few BT ones (3 of 25 tested; ref. 148). In contrast, interstitiospinal cells receive input preferentially from the horizontal canals (149).

NPH cells also receive input from secondary vestibular nucleus neurons (#10, figure 4) as indicated by the fact that they are di-synaptically excited from the contralateral vestibular nerve, and di-synaptically inhibited from the ipsilateral vestibular nerve in the cat (150). The vestibular projection to the primate NPH is provided by the type I (horizontal) PVPs as shown with the help of intraaxonal injection of tracer in functionally identified units (132). The vestibular responses of NPH cells are mainly type II (ipsiversive cells) but also type I (contraversive cells), in the cat (60) and mostly type I, in the monkey (64). The existence of a reciprocal projection of the NPH to vestibular nuclei (#12, figure 4) has also been demonstrated. These reciprocal connections are bilateral, concern several vestibular nuclei (medial, superior, descending and ventrolateral) and have been demonstrated with both retrograde (92, 95) and anterograde (151, 92, 95) techniques.