Physiological and Anatomical Evidence for an Inhibitory Trigemino-Oculomotor Pathway in the Cat

Abstract

During blink down-phase, the levator palpebrae superioris (levator) muscle is inactivated, allowing the orbicularis oculi muscle to act. For trigeminal reflex blinks, the excitatory connections from trigeminal sensory nuclei to the facial nucleus have been described, but the pathway whereby the levator is turned off have not. We examined this question by use of both physiological and anatomical approaches in the cat. Intracellular records from antidromically activated levator motoneurons revealed that periorbital electrical stimulation produced bilateral, long latency inhibitory postsynaptic potentials (IPSPs). Central electrical stimulation of the principal trigeminal nucleus produced shorter latency IPSPs. Intracellular staining revealed that these motoneurons reside in the caudal central subdivision and have 10 or more poorly branched dendrites, which extend bilaterally into the surrounding supraoculomotor area. Axons penetrated in this region could be activated from periorbital and central electrodes. Neurons labeled from tracer injections into the caudal oculomotor complex were distributed in a crescent-shaped band that lined the ventral and rostral aspects of the pontine trigeminal sensory nucleus. Double-label immunohisto-chemical procedures demonstrated that these cells were not tyrosine hydroxylase-positive cells in the Kölliker-Fuse area. Instead, supraorbital nerve afferents displayed a similar crescent-shaped distribution, suggesting they drive these trigemino-oculomotor neurons. Anterograde labeling of the trigemino-oculomotor projection indicates that it terminates bilaterally, in and above the caudal central subdivision. These results characterize a trigemino-oculomotor pathway that inhibits levator palpebrae motoneurons in response to blink-producing periorbital stimuli. The bilateral distributions of trigemino-oculomotor afferents, levator motoneurons, and their dendrites supply a morphological basis for conjugate lid movements.

During normal conscious activity, animals hold their eyelids open while scanning the world with their eyes. In mammals, the upper eyelid is held up through the tonic activity of the levator palpebrae superioris (levator) muscle, and the lid position is further adjusted with regard to vertical gaze by changes in levator activity levels (Bjork and Kugelberg, 1953; Evinger et al., 1984; Becker and Fuchs, 1988; Guitton et al., 1991; Porter et al., 1993). When the eyelid performs its cardinal functions of spreading the tear film across the cornea and protecting the delicate corneal surface, the orbicularis oculi muscle is activated to close the eye in a blink. In order for this ballistic action to occur, the tonic activity of the levator muscle is inhibited during the blink down-phase, as demonstrated from electromyography (EMG) recordings in rabbits and humans (Bjork and Kugelberg, 1953; Evinger and Manning, 1993). The motoneurons controlling the levator muscle are located within the caudal central subdivision (CC) of the oculomotor complex in cats and primates (Gacek, 1974; Akagi, 1978; Porter et al., 1989; Sun and May, 1993; Chen and May, 2002, 2007). Extracellular recordings from levator motoneurons in rhesus monkeys show that these cells are tonically active when the eyes are open and that this activity is sharply reduced during the down-phase of a blink (Fuchs et al., 1992). The importance of this pattern of behavior is emphasized by the deficits produced in the form of blepharospasm termed apraxia of lid opening. In this disorder, the tonic activity of the levator motoneurons is lost bilaterally at inappropriate times, producing temporary visual loss in the sufferer (Aramideh et al., 1995).

Blinks, with their concurrent activation of the orbicularis oculi muscle and inhibition of the levator muscle, can be elicited by a number of external stimuli (Tackmann et al., 1982; Manning and Evinger, 1986). Among the behaviorally important and better-studied forms of blink are those elicited by stimulation of the trigeminal nerves supplying the cornea and periorbital tissues (Kugelberg, 1952; Evinger et al., 1991; LeDoux et al., 1997). While blink movements are primarily accomplished by the orbicularis oculi muscle, which is supplied by facial motoneurons, the retractor bulbi muscles also play a role in many animals, including cats. These muscles are supplied by the accessory abducens nucleus and are positioned behind the eyeball to pull it back into the bony orbit as the eyelid closes (Baker et al., 1980; Delgado-Garcia et al., 1990). Eye movements are also present during blinks, and extraocular muscle activation has been proposed to withdraw the eye in species lacking retractor bulbi muscles (Evinger et al., 1984; Evinger and Manning, 1993). Thus, blinks may not only require inactivation of the levator motoneurons, they may also involve activation of the retractor bulbi and rectus muscles. The pathways and synaptic activity that underlies the shaping of neural activity in this diverse set of motoneurons during a blink are poorly understood.

Initial investigations of the sources of input to the oculomotor nucleus (III) that utilized retrograde neuronal tracers revealed a small number of retrogradely labeled neurons near the exiting trigeminal nerve, in addition to the projections from the pontine reticular formation, vestibular nuclei, and the abducens intranuclear neurons (Graybiel and Hartwieg, 1974; Graybiel, 1977; Steiger and Büttner-Ennever, 1979). It was suggested that the cells near the trigeminal nerve lay within the Kölliker-Fuse area of the pons. This region is associated with the parabrachial nuclei (Fulwiler and Saper, 1984), which contains noradrenergic neurons in the cat (Stevens et al., 1982). This association presents the interesting possibility that the cells observed by Graybiel might supply a noradrenergic input to oculomotor motoneurons. Such a projection could have implications for the changes in eye movement velocity and extraocular motoneuron firing rate that occur when the animal’s level of arousal varies. Alternatively, such a noradrenergic projection might terminate on cells in the centrally projecting Edinger-Westphal nucleus (EWcp), instead of the subjacent III, as the former has been shown to be involved in stress and in eating and drinking behaviors (May et al., 2008; Kozicz et al., 2011). If this does not represent a noradrenergic projection, then it may play a role in blinks. Anatomical studies in the cat and other species have suggested that these cells may be the source of a trigemino-oculomotor pathway (Ogasawara, 1985; May and Baker, 1987; Guerra-Seijas et al., 1993; May et al., 1996; Van Ham and Yeo, 1996b). One possible role for this pathway is to induce an inhibition of levator motoneurons at the same time other trigeminal pathways are activating motoneurons in the facial and accessory abducens nuclei. Other functions, such as activation of the superior rectus to produce Bell’s phenomenon, have been suggested (Ogasawara, 1985).

Full consideration of whether a trigemino-oculomotor pathway is present and whether it inhibits the levator motoneurons is deterred by the lack of data on a number of critical points. To remedy this situation, we recorded intracellularly from presumed levator motoneurons and trigemino-oculomotor axons within acutely prepared cats, while testing for trigeminal input. In addition, we intracellularly stained the motoneurons to reveal their morphological characteristics. We also used neuroanatomical methods to better characterize a pathway from the periorbital sensory receptors to the levator motoneurons. Finally, immunohistochemical techniques were utilized to determine whether these oculomotor afferent cells are, in fact, part of the noradrenergic population of the Kölliker-Fuse nucleus. A portion of the results from this study has been presented previously in abbreviated form (May and Baker, 1987; May et al., 2002).

MATERIALS AND METHODS

The electrophysiological experiments were carried out in 16 adult cats (Felis domesticus) of both sexes and the neuroanatomical experiments were carried out in 12 adult cats of both sexes. All experimental procedures conformed to institutional regulations, used approved protocols, and were in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals.

Neurophysiological procedures

The recording and injection procedures are similar to those described in detail in Vidal et al. (1988). Animals were initially anesthetized with sodium pentobarbital (35 mg/kg, intraperitoneally [ip]). An intravenous line was inserted to allow maintenance of the anesthesia level during surgery and recording sessions (supplemental doses = 10 mg/kg). Mucous secretion was inhibited by injection of atropine sulfate (0.2 mg/kg, subcutaneous [sc]) and central nervous system (CNS) edema was avoided by injecting dexamethasone sodium phosphate (2 mg/kg, intravenously [iv]). A tracheotomy was performed and the animal was artificially respired. This was necessary because the animal was paralyzed with gallamine triethiodide (5 mg/kg, iv) and received a bilateral pneumothorax in order to increase recording stability.

A pair of silver ball electrodes were placed retrobulbarly. These electrodes activated trigeminal primary afferents and also antidromically activated the nerves to the extraocular muscles. In some cases, Teflon-insulated silver wire electrodes were also placed either in the levator palpebrae superioris muscle or a cuff electrode was placed on the supraoculomotor branch of the ophthalmic division of the trigeminal nerve. To electrically stimulate the vestibular afferents, a silver ball electrode was placed in each middle ear cavity and a Teflon-insulated, fine silver wire electrode was placed in a hole drilled into a portion of the semicircular canal within the temporal bone. Central electrodes were employed to stimulate the cells of origin of the trigemino-oculomotor projection. These consisted of a twisted pair of Teflon-insulated silver wires placed in the principal trigeminal nucleus under stereotaxic guidance. Rectilinear, single shock stimuli of 0.1 ms duration, and 100–900 μA in intensity, as measured by voltage drop across a known resistance, were used both peripherally and centrally.

The head was tilted nose down 30°, the posterior cranium over the cerebellum was removed, and the anterior lobe of the cerebellum was aspirated to reveal the confluence of the cerebral aqueduct and the fourth ventricle. Glass micropipettes, held at an angle of 25–35° to the coronal plane, were inserted into the midbrain, on the midline at the base of the aqueduct, to access the CC. These recording electrodes were filled with either 3M KCl or 2M Kcitrate and had resistances in the range of 10–20 M Ω. The recorded potentials were lowpass-filtered, amplified, and displayed on an oscilloscope before being directly photographed. In some cases, records were digitized and displayed with an X–Y plotter. Low-gain DC records were displayed directly, while high-gain AC records were first high pass filtered (1 Hz). Electrical stimulation through the retrobulbar silver ball electrodes produced an antidromic field potential that could be used to locate the oculomotor nucleus. In the cases where the trigeminal nucleus was stimulated centrally, the medial longitudinal fasciculus was transected bilaterally between the fourth and sixth nucleus. This greatly lessened the possibility that synaptic potentials observed in levator motoneurons following central electrical stimulation were generated by current spread to the vestibular nuclei. In three cats, presumed trigemino-oculomotor axons were intercepted in the area immediately ventral and lateral to the oculomotor nucleus field. They were characterized by their responses to peripheral and central stimulation of the trigeminal sensory system. At the conclusion of the recording session, the animals were given an additional dose of sodium pentobarbital (50 mg/ml, iv), and were perfused with a 10% buffered formalin solution. The brains were then cryoprotected, frozen, and sectioned at 80 μm for histological confirmation of electrode placement and medial longitudinal fasciculus transection.

Intracellular injection and staining

In order to confirm that the recordings were indeed from levator motoneurons and to characterize their soma-todendritic morphology, examples were intracellularly injected. Glass micropipettes filled with 10% horseradish peroxidase (HRP, Sigma, St. Louis, MO) in 0.1M, pH 7.2 Tris buffer with 0.5M KCl that had resistances of 20–30 MΩ were used to record from and inject levator motoneurons. Positive current (5–10 nA) passed at a 50% duty cycle for 1–3 minutes was used to iontophorese the HRP into the cell. Only one or two cells were injected per animal.

These animals were perfused through the heart after a survival time under anesthesia of 2–6 hours. They received a buffered saline prewash, followed by a fixative containing 1% paraformaldehyde and 1.25% glutaraldehyde (Fisher, Pittsburgh, PA) in 0.1M, pH 7.2 phosphate buffer. The brainstem was cut into 100-μm-thick transverse sections. All sections through the midbrain were reacted using a modification of the diamino benzidene (DAB) method (Shaw and Baker, 1985) in which the sections were incubated in a 0.1M, pH 7.2 phosphate-buffered DAB (Sigma) solution which also contained nickel ammoniumsulfate and cobalt acetate (Fisher). Labeled neurons were observed at 400× and reconstructed using a microscope equipped with a drawing tube.

Neuroanatomical tracing procedures

To trace the projections from the lateral pons to the oculomotor nucleus, we utilized the anterograde and retrograde transport capabilities of wheatgerm agglutinin conjugated to horseradish peroxidase (WGA-HRP, Sigma) and HRP (Sigma). Animals were anesthetized for surgery with sodium pentobarbital (35 mg/kg, ip). Injections of WGA-HRP were made into III by utilizing either iontophoretic or pressure injections. For iontophoretic injections, a solution of 1.0% WGA-HRP in pH 7.2, 0.1M Tris buffer was utilized. The WGA-HRP was iontophoresed from glass micropipettes with tip diameters of 10–20 μm using 1–2 μA, positive, 0.5 second square wave current. For pressure injections, 0.04–0.1 μl of 2% WGA-HRP in a pH 7.2, 0.1M phosphate-buffered solution (PBS) was injected out of a 1.0 μl Hamilton syringe. To locate III, the occipital pole of cortex was removed by aspiration to allow visualization the midbrain. Surface features and stereotaxic coordinates were employed to direct the micropipette tip or syringe needle. Iontophoretic WGA-HRP injections centered in the lateral pons were also made. A hole was drilled in the tentorium after aspiration of occipital cortex, and a purely stereotaxic approach was used. In an additional experiment, the primary afferents traveling in the supraoculomotor nerve were labeled transcutaneously by injecting 25 μl of 1.0% WGA-HRP into the lid, brow, and forehead with a 10 μl Hamilton syringe. In one case, this superficial injection was made in conjunction with a central WGA-HRP injection into III. Following each injection, all surgical openings were closed and the animals survived for an additional 24–48 hours.

At the end of the survival period, animals were deeply anesthetized with sodium pentobarbital (50 mg/kg, ip) and perfused through the heart with a buffered saline rinse, followed by a fixative containing of 1.0% parafor-maldehyde (Fisher) and 1.25% glutaraldehyde (Fisher) in 0.1M, pH 7.2 PBS for normal transport experiments, and 2.0% paraformaldehyde and 0.2% glutaraldehyde for experiments in which immunocytochemical evaluation of the tissue was made. The brainstem was then frozen and sectioned at a thickness of 25 or 50 μm. For visualizing the WGA-HRP alone, tetramethylbenzidene (TMB, Sigma) methods were employed (Mesulam, 1978; May and Porter, 1998).

To allow simultaneous visualization of both the retrograde label and the presence of a marker for adrenergic neurons, an adaptation (see May et al., 1987) of the Sternberger et al. (1970) soluble peroxidase-antiperoxidase (PAP) method was employed to label neurons containing tyrosine hydroxylase (TH). The TMB chromagen from the initial reaction was stabilized with diaminobenzidine (DAB) as follows: sections were placed in 0.25% cobalt acetate (Fisher) and then reacted in 0.5% DAB with 0.001% H₂O₂ (Fisher) in PBS. This procedure results in a black granular reaction product that survives the immunocytochemical procedures. The immunocytochemistry was initiated by treating the sections with 3.0% normal goat serum (Sigma) in pH 7.2, 0.1M phosphate buffer (NG-PBS), before incubating overnight in rabbit antibody to TH (Baker et al., 1983) diluted 1:1,000 in 1.0% NG-PBS at 4°C. After rinsing in 1.0% NG-PBS, sections incubated in goat antirabbit IgG (Sigma) diluted 1:50 in 1.0% NG-PBS. They were then incubated in rabbit PAP solution (Sigma) diluted 1:100 in 1.0% NG-PBS. Finally, the sections were reacted to reveal the PAP complex using a solution of 0.5% DAB and 0.001% H₂O₂ in pH 7.2, 0.1M PBS. This produces a homogeneous brown reaction product in neurons that are TH-positive that can be differentiated from the black granular WGA-HRP reaction product if the cell is doubly labeled. The primary TH antibody was raised against purified trypsin-treated bovine protein (Joh et al., 1973; Baker et al., 1991). It recognizes a single band on western blot analysis (Joh et al., 1973) and has been fully characterized (Saino-Sato et al., 2007). Appropriate controls in which the primary or secondary were left out, or the primary was preabsorbed with the antigen, were performed.

The results presented photographically were obtained using either an Olympus BH-2 photomicroscope and conventional film, which was then digitized, or a Nikon Eclipse 80i photomicroscope with a Nikon DXM1200F digital camera. The contrast, brightness, and color balance of all digitized images were adjusted to match the image seen by eye.

RESULTS

Anterograde examination of oculomotor complex projection

Our first set of experiments was aimed at better determining the actual target of the cells observed near the exiting trigeminal nerve by Graybiel (Graybiel and Hartwieg, 1974; Graybiel, 1977). To accomplish this, the area of the lateral pons that contained their cell bodies was injected with WGA-HRP and the sections were reacted to reveal the anterogradely transported tracer within axons and their terminals. As shown in Figure 1G–I, the injection was largely confined to the trigeminal sensory nucleus at pontine levels. Numerous labeled axons crossed the midline of the pons at the level of the injection. Most of these presumably represent the ventral trigeminothalamic tract, for they joined with the ascending medial lemniscus and moved laterally as they ascended the brainstem (Fig. 1A–F). However, at the level of the trochlear nucleus (Fig. 1E) a small group of labeled axons disengaged from the main group of axons that continued on to terminate in the superior colliculus (Fig. 1A–C) and the ventral posteromedial nucleus of the thalamus (not shown). This small group of axons traveled dorsally (Fig. 1C–E) to terminate above the caudal half of the oculomotor nucleus (Fig. 1B–D). Terminals were absent over the rostral half of the oculomotor nucleus (Fig. 1A), and were not found in the EWcp. As shown in the darkfield photomicrographs (Fig. 1J,K), there was in fact a bilateral distribution of labeled terminals. The projection was predominantly contralateral, but ipsilateral terminals were also present. The terminations were densest in CC and supraoculomotor area (SOA) above III. There was also considerable terminal label along the midline between the two oculomotor nuclei. The labeled axons crossed through III, making it difficult to determine whether sparse terminations were present within III itself. These results indicated that the EWcp was not the target of the projection, and directed our physiological investigation toward the caudal end of the oculomotor complex, in general, and the levator motoneurons in the caudal central subdivision, in particular.

Figure 1.

Anterograde labeling of the trigemino-oculomotor pathway. The WGA-HRP injection into the principal trigeminal sensory nucleus (pV) (G–I) labeled axons (lines) that crossed the midline (G) and joined the medial lemniscus (F). A portion of these axons travel dorsally (D,E) to terminate bilaterally, with a contralateral predominance over the caudal aspect of the oculomotor nucleus (III) (stipple) (B–D). J,K: Darkfield photomicrographs demonstrating that the terminals are primarily located in the caudal central subdivision (CC) and the overlying supraoculomotor area (SOA). Note that in addition to the heavy contralateral terminal label, light terminal label is also present ipsilaterally. Abbreviations for this and following figures: Antidro: antidromic stimulation; BC: brachium conjunctivum, CC: caudal central subdivision, CN: cochlear nuclei, Cun: cuneiform nucleus, DR: dorsal raphe, DV: descending vestibular nucleus, EWcp: Edinger-Westphal centrally projecting nucleus, Extra: extracellular record, IC: inferior colliculus, III: oculomotor nucleus, Intra: intracellular record, IV: trochlear nucleus, KF: Kölliker-Fuse area, LC: locus coeruleus, LL: lateral lemniscus, LV: lateral vestibular nucleus, MG: medial geniculate, MLF: medial longitudinal fasciculus, MRF: mesencephalic reticular formation, MV: medial vestibular nucleus, NRTP: nucleus reticularis tegmenti pontis, PAG: periaqueductal gray, PB: parabrachial nuclei, PRF: pontine reticular formation, pV: principal trigeminal nucleus, R: red nucleus, SGI: intermediate gray layer, SGP: deep gray layer, SGS: superficial gray layer, SN: substantia nigra, SO: superior olive, SOA: supraoculomotor area, SV: superior vestibular nucleus, sVo: spinal trigeminal nucleus pars oralis, Vcn: central trigeminal electrode, Ves: vestibular, VI: abducens nucleus, VII: facial nucleus, Vm: motor trigeminal nucleus, Vop: ophthalmic branch of trigeminal nerve, WGA-HRP: wheat germ agglutinin conjugated horseradish peroxidase, 5n: trigeminal nerve, 6n: abducens nerve, 7n: facial nerve. Scale bars = 500 μm in J,K.

Intracellular recording and staining experiments

Records from two typical presumed levator motoneurons are shown in Figure 2. In the first cell (Fig. 2A–D), the DC trace in A shows an antidromic action potential that is triggered with a latency of less than 1.0 ms following electrical stimulation from the left orbit with silver ball electrodes. High-gain, AC traces recorded extracellularly just before impaling this neuron indicated it was located within the oculomotor antidromic field. Electrical stimulation of the retrobulbar electrodes on the other (right) side (Fig. 2B) produced long-duration hyperpolarizing potentials of greater than 1.5 mV. Comparison of the intra- and extracellular high-gain traces indicates a relatively long latency (3.3 ms) for the initiation of these potentials. When the antidromic potential was blocked with hyperpolarizing current, the arrival in the soma of the antidromic volley (m-spike) could be observed in the high- and low-gain traces (Fig. 2C). This procedure also unmasked a long latency (≈ 4 ms) hyperpolarizing potential, as shown in the high-gain trace. This potential had a long duration (over 10 ms). Electrical stimulation of the left and right labyrinthine electrodes (Fig. 2D) produced excitatory postsynaptic potential (EPSP), inhibitory postsynaptic potential (IPSP) sequences in the high-gain records that clearly differed from orbitally elicited potentials. In the same penetration, neurons like this one could be recorded that were antidromically activated from either the ipsilateral or the contralateral side. A small number of the recorded motoneurons showed evidence of sending a bifurcating axon that reached both orbits. In these, antidromic potentials were observed following stimulation of both the left and right orbits, and decreasing the time between the two stimuli led to a collision between the elicited action potentials.

Figure 2.

Records from two presumptive levator motoneurons showing evoked responses to single pulse electrical stimulation from retrobulbar silver ball electrodes (Eye) and electrodes placed in the vestibular labyrinth (Ves). The first cell (A–D), found within the oculomotor extracellular (Extra) field potential (Field), was antidromically driven from the retrobulbar electrodes in the left eye (A). B: The intracellular records (Intra) reveal a long latency IPSP evoked from the right retrobulbar electrodes. When only an m-spike was present as antidromic activity (C) following stimulation of the left retrobulbar electrodes, an underlying long latency IPSP was revealed. These potentials differed in latency and form from those evoked following electrical stimulation of the vestibular labyrinths (D). However, the matched pattern of left and right vestibular responses suggested that the recorded motoneuron was involved in vertical gaze. The second cell’s intracellular records also reveal an antidromic spike following left-sided electrical stimulation and a long latency IPSP following stimulation on the right (E). F: Antidromic activation and a long latency reverse IPSP seen during injection of HRP into this cell. This long latency IPSP could be reversed by injecting current into the recorded cell (G). The latency and form of the potentials evoked by vestibular stimulation differed from those evoked by the retrobulbar electrodes. H: The potentials evoked from the left side following injection of HRP into the cell. In addition to an antidromic m-spike, a reverse IPSP was present in an intracellular record taken just before leaving the cell. Scale in D applies to B; scale in H applies to F,G.

Records from a second example of a presumptive levator motoneuron are shown in Figure 2E–H. In E, electrical stimulation of the left oculomotor nerve produced an anti-dromic action potential in this neuron, while electrical stimulation of the retrobulbar electrodes on the right evoked a long latency hyperpolarization. This motoneuron was then intracellularly injected with HRP by passing current out of the tip of the micropipette. Confirmation that the appropriate neuron was stained is shown in F, where the antidromic potential is still present in the low-gain trace following stimulation of the left oculomotor nerve. Due to chloride entry during the injection, the neuron displays long latency, long duration, depolarizing potentials following electrical stimulation from the right orbit. These potentials could be reversed using depolarizing currents to reveal a latency of 4.8 ms. This suggests that these records represent chloride-dependent IPSPs. By comparison, the potentials evoked from the labyrinthine electrodes (Fig. 2G) have shorter latencies and a different wave form. In H, the high-gain records show the potentials observed just before and after the electrode leaves the cell. The intracellular record shows an m-spike from the antidromic invasion of the motoneuron and an underlying reverse IPSP. Comparison with the extracellular record shows a tight correlation in the latency of the m-spike and the antidromic field. The latency of the long duration reverse IPSP is 6.3 ms.

A serial reconstruction of the motoneuron whose records were presented in Figure 2E–H is shown in Figure 3A,B, along with a photomicrograph of another example. The latter shows the quality of the intracellular staining. The reconstructed cell was located within the CC, dorsal to the caudal pole of III (Fig. 3B). Its position within the CC confirms that it is a levator motoneuron. The location of the motoneuron on the left side of the brainstem indicates that it is one of the rarer cells that send their axon to the ipsilateral eye. As shown in A, the multipolar soma of this motoneuron had a long diameter of 37 μm and a short diameter of 17 μm. More than 10 primary dendrites extended from the soma in all directions. These long, gradually tapering dendrites exhibited very few branches as they extended nearly 1,000 μm from the soma. The dendrites freely traversed the midline to distribute throughout the CC, and they extended bilaterally into the surrounding SOA. Some dendrites extended along the midline between the oculomotor nuclei.

Figure 3.

Intracellularly identified and stained presumptive levator motoneurons. A: Serial reconstruction of the soma and dendritic field of the motoneuron characterized in Figure 2E–H. Cell has numerous, poorly branched dendrites radiating from it. B: It was located in the caudal central (CC) division of the oculomotor nucleus (III), and its dendrites extended across the midline and up into the overlying suprao-culomotor area (SOA), but not into the medial longitudinal fasciculus (MLF). A photomicrograph of another intracellularly stained cell with similar electrophysiological responses is provided in C to demonstrate the quality of the staining. Scale bar = 50 μm in C.

Two additional examples of contralaterally projecting, intracellularly labeled levator motoneurons injected in the same animal are illustrated in a set of six serial sections in Figure 4 to show the relationship between their dendritic fields and the cytoarchitectonic boundaries in the region. Both multipolar somata display numerous dendrites. The dendrites extend throughout both sides of the CC and throughout SOA over the caudal end of III. They also extend along the midline between the oculomotor nuclei. However, they do not extend into the dorsal raphe nucleus, and very little of the dendritic field of either of these two motoneurons is found within III proper.

Figure 4.

Rostral to caudal, serial sections (A–F) demonstrating the somata (arrows) and dendritic fields of two intracellularly stained levator motoneurons in relation to cytoarchitectonic boundaries. The dendritic fields extended bilaterally within the caudal central subdivision (CC) into the supraoculomotor area (SOA). Relatively few dendrites extended into the oculomotor nucleus (III) or the medial longitudinal fasciculus (MLF).

Occasionally, neurons were penetrated within the portion of the oculomotor field where the levator motoneurons were found that could not be antidromically activated. Although these neurons were difficult to hold, indicating a small somatic diameter, one example is shown in Figure 5. The high-gain, AC records (Fig. 5A) show that antidromic action potentials were not elicited following electrical stimulation by retrobulbar silver ball electrodes on either the left or right side, despite the fact that an antidromic field could be observed in both the extracellular and intracellular traces following stimulation of the left side. Stimulation of the right side produced a small amplitude, long latency (5.8 ms) hyperpolarization, while stimulation of the left side produced a more pronounced, long latency (3.2 ms) hyperpolarization. The synaptic potentials produced following stimulation of the orbits differed in their form compared to the shorter latency (1.8–2.0 ms) hyperpolarizing potentials produced by electrical stimulation of the labyrinths (Fig. 5B). Intra-cellular staining of this neuron with HRP revealed that it was indeed located within CC (Fig. 5C). However, it was considerably smaller than the injected motoneurons, as shown in the serial reconstruction (Fig. 5D). Its soma had a long axis of 20 μm and a short axis of 15 μm. Numerous thin, poorly branched dendrites extended from the soma up to 550 μm, and were largely confined to the CC and overlying SOA.

Figure 5.

A nonmotoneuron found in the caudal central subdivision (CC). A: Intracellular records indicate that this cell was not antidromically driven from retrobulbar electrodes placed in either the right or left eye. However, long latency hyperpolarizations were present. The IPSP observed from left-side stimulation was of greater amplitude and shorter latency. Electrical stimulation on this side also evoked an extracellular field potential (Field). B: Electrical stimulation of the labyrinths (Ves) revealed that this cell also received bilateral vestibular input. C: Intracellular staining showed that this nonmotoneuron was located within the CC, but its dendrites did not extend into the oculomotor nucleus (III). Serial reconstruction of the stained cell (D) reveals a relatively small diameter soma from which emerge several, sparsely branched dendrites.

In Figure 6 the effects of central stimulation are demonstrated. This levator motoneuron was antidromically activated (Fig. 6A) following electrical stimulation by retrobulbar silver ball electrodes on the right. The high-gain, AC traces in B show that electrical stimulation of the left orbit produced a long latency (3.0 ms), long duration hyperpolarization that could be reversed by injection of chloride ions. In comparison, central electrical stimulation through a bipolar electrode placed in the pontine trigeminal sensory nucleus also produces a hyperpolarization (Fig. 6C), which could be reversed by injection of chloride ions to reveal a latency of 1.6 ms. This latency is roughly similar to that observed in the reversal potentials of the vestibular inputs following electrical stimulation of the labyrinths (Fig. 6D). However, the short latency vestibular inputs were largely abolished following a transection of the medial longitudinal fasciculus (Fig. 6E), while central trigeminal stimulation still produced vigorous reversed IPSPs (Fig. 6F), even as the micropipette was withdrawn from the motoneuron. Comparison of the high-gain, intra- and extracellular traces suggests a latency of 2.0 ms or less.

Figure 6.

Central stimulation also produced IPSPs in presumptive levator motoneurons. A: Antidromic activation following electrical stimulation with retrobulbar electrodes in the right eye. B: Electrical stimulation from retrobulbar electrodes in the left eye revealed an evoked field followed by a long latency hyperpolarization that could be reversed by injection of chloride ions. C: Electrical stimulation of the trigeminal sensory nucleus through bipolar electrodes placed in the pons (Vcn) produced a long-lasting IPSP that could be reversed by injection of chloride ions. D: Comparison of these intracellularly (Intra) and extracellularly (Extra) recorded potentials with those in C reveals the differences in both form and latency from reversed IPSPs observed following left and right labyrinthine electrical stimulation (Ves). These were largely obliterated following sectioning of the medial longitudinal fasciculi (E), without affecting the potentials elicited from the trigeminal electrode (F). Scale in E applies to D.

Axons were encountered in the region ventrolateral to the oculomotor nucleus field that could be activated from electrodes electrically stimulating the trigeminal sensory system peripherally and centrally. Figure 7A shows an extracellular record from such an axon. Stimulation through a cuff electrode placed on the supraorbital nerve produced multiple action potentials, the earliest of which had a 2.0 ms latency. Central stimulation from an electrode placed in the principal trigeminal nucleus activated the axon with a latency of 0.5 ms. Note the presence of two spikes in response to single shock stimuli. A second example of these axonal recordings is shown in Figure 7B. In trace 1, a high-gain extracellular recording shows the presence of an axon that fires in response to electrical stimulation of both the peripheral and central electrodes. After penetrating the axon, orthodromic action potentials are demonstrated in the low-gain DC trace (4), as well as the high-gain AC trace (3). An average of three high-gain traces (2) shows the latency from stimulation of the ophthalmic branch of the trigeminal nerve is 3.5 ms, and the latency following central stimulation is 2.8 ms.

Figure 7.

Activity in two presumptive trigemino-oculomotor axons. A: Low (3) and high (1,2) gain extracellular (Extra) records showing potentials evoked following electrical stimulation of a cuff electrode on the supraorbital portion of the ophthalmic branch of the trigeminal nerve (Vop) and a bipolar electrode placed in the pontine portion of the trigeminal sensory nucleus (Vcn). B: Extracellular (1) and intracellular (Intra) (2–4) high (1–3) and low (4) gain records from another axon that could be ortho-dromically activated from electrical stimulation through both the peripheral and central electrodes. Note that multiple spikes were elicited from single pulse stimuli in both of these axons.

Retrograde neuroanatomical tracer experiments

In order to better define the population of cells in the lateral pons that project to the oculomotor complex, an iontophoretic injection of WGA-HRP was made into this region (Fig. 8A–C). It was centered caudally on the left side, and only slightly involved the medial longitudinal fasciculus and the overlying SOA. In addition to the expected labeling of axons in the medial longitudinal fasciculus, labeled axons could be observed extending ventrally in a paramedian position (Fig. 8C) and then decussating at the level of the trochlear nucleus (Fig. 8D). As expected, retrogradely labeled cells were present contralaterally in the abducens nucleus and bilaterally in the vestibular nuclei (Fig. 8J–L). In addition, retrogradely labeled lateral pontine neurons appeared at the level of the trigeminal motor nucleus (Fig. 8G) and extended as far caudally as the level of the abducens nucleus (Fig. 8K). These labeled neurons were scattered dorsal and ventral to the fascicles of the exiting trigeminal and facial nerves. Those at more rostral levels (Fig. 8F,G) appeared to be included within the Kölliker-Fuse area. The entire population was observed bilaterally, although the bulk of the injection was on the left side. Viewed in this plane, there was little pattern to the distribution of these neurons and relatively few cells were present in any individual section. Examples of these labeled neurons from another case are shown in the photomicrographs of Figure 9. These neurons can be seen dorsal and ventral to the exiting facial nerve, as well as interspersed among its fascicles. The vast majority are quite small, with long axes of around 10 μm (black arrowheads). There are a small number of cells of intermediate size, with long axes of around 20 μm (white arrowheads).

Figure 8.

Distribution of retrogradely labeled neurons supplying the caudal oculomotor nucleus (III). The injection of WGA-HRP includes III and the overlying supraoculomotor area on the left (A–C). Labeled axons extend ventrally (C,D) and then cross the midline (D). Labeled cells (dots) can be found bilaterally in the lateral pons (F–K) within the Kölliker-Fuse area (KF) and around the exiting fifth (5n) and seventh (7n) nerves. Sections ordered rostral to caudal.

Figure 9.

Neurons in the lateral pons labeled retrogradely from an oculomotor nucleus injection. Two sections at the level of the abducens nucleus (VI) are shown (C,D). Boxes indicate the regions shown at higher magnification (A for D, B for C). The labeled cells (arrowheads) are found above (in the principal trigeminal sensory nucleus (pV)), below, and between the fascicles of the exiting facial nerve (7n). Small neurons (black arrowheads) predominate, although a few medium-sized cells (black arrowheads) are present. Scale bars = 1.0 mm in C; 100 μm in A. Scale in A applies to B; in C applies to D.

To determine whether all or a portion of the labeled neurons supplying oculomotor complex might be noradrenergic, sections were reacted using a double-label procedure that revealed whether cells containing retrogradely transported WGA-HRP were also TH-positive. The distribution of the two labels is charted in Figure 10. Following a large oculomotor nucleus injection (Fig. 10A), retrogradely labeled neurons (red circles) were seen in the ventral Kölliker-Fuse nucleus (Fig. 10D) and around the exiting trigeminal and facial nerves (Fig. 10E–G), as described above. TH-positive neurons (blue triangles) were observed extending from the locus coeruleus, through the parabrachial nuclei along either side of the brachium conjunctivum (Fig. 10B–E), into the Kölliker-Fuse area (Fig. 10C,D). While a few doubly labeled neurons were noted in the locus coeruleus and parabrachial nucleus (Fig. 9B,E, green stars), none were observed in the Kölliker-Fuse area, even though cells labeled retrogradely and via the immunohistochemical procedure were in close proximity. This relationship is further demonstrated in photomicrographs showing retrogradely labeled neurons found ventrally in the Kölliker-Fuse area with black particulate label (Fig. 11C,D), and adjacent, more dorsal, cells with the brown homogeneous labeling produced by the immunohistochemical localization of TH (Fig. 11C). No doubly labeled cells like the parabrachial neuron shown in Figure 11B were observed within the Kölliker-Fuse area. These findings indicate that the Kölliker-Fuse area does not project to the oculomotor complex, in accordance with its known function in respiration (Lavezzi et al., 2004), and that the border of the Kölliker-Fuse area suggested by cytoarchitecture (Graybiel and Hartweig, 1974) may not precisely agree with the border based on TH-staining or connectivity.

Figure 10.

Comparison of the distribution of tyrosine hydroxylase (TH)-positive and retrogradely labeled neurons. Cells labeled by antibody to TH are indicated by filled blue triangles. These are primarily found in the parabrachial nuclei (PB), locus coeruleus (LC), and the Kölliker-Fuse area (KF) (B–E). Those neurons labeled retrogradely from a WGA-HRP injection into the oculomotor nucleus (III) shown in A are indicated by red circles. Those of interest are primarily found near the exiting trigeminal and facial nerves (5n, 7n) (E–G), although some are found in ventral KF (D). Double-labeled cells (green stars) were rare, were located in PB (B) and LC (E), and were not seen in KF.

Figure 11.

Appearance of singly and doubly labeled neurons. A: Tyrosine hydroxylase (TH)-positive cells were present in the locus coeruleus (LC), parabrachial nuclei (PB), and Kölliker-Fuse area (KF). Box indicates region of KF shown at higher magnification in C. The retrogradely labeled neurons (white arrowheads) lay ventrolateral to those tagged with antibody to TH (black arrowheads). The box indicates region shown at higher magnification in D, where the particles signifying the retrograde label are evident. B: Examples of singly retrogradely labeled, singly antibody labeled, and doubly labeled (black arrow) cells in the parabrachial nucleus. Scale bars = 1.0 mm in A; 250 μm in C; 50 μm in D. Scale in D applies to B.

To obtain a better understanding of the distribution of cells in the lateral pons that target the oculomotor complex, the brainstem was cut in the sagittal plane. Figure 12A shows an iontophoretic injection of WGA-HRP into the left caudal oculomotor complex with little spread of tracer beyond slight involvement of the right nucleus and the overlying supraoculomotor area. Retrogradely labeled neurons observed contralaterally in the lateral pons (Fig. 12C–F) were clearly arranged in a crescent-shaped, continuous distribution that lined the rostral and ventral edge of the trigeminal sensory nuclei within the pons. Thus, their scattered appearance in the frontal plane is due to the fact the trigeminal and facial nerves pass through this group, whose major axis is rostrocaudally oriented. This arrangement is also shown in a darkfield photomicrograph (Fig. 13A) from this case.

Figure 12.

Distribution of cells projecting to the oculomotor nucleus in the parasagittal plane. A: Injection of WGA-HRP into the caudal oculomotor nucleus (III). B–F: medial to lateral series of parasagittal sections showing the distribution of retrogradely labeled neurons (dots). Of note is the crescent-shaped band of labeled cells lining the rostral and ventral aspect of the principal trigeminal sensory nucleus (pV) in C–F.

Figure 13.

Darkfield photomicrographs showing the patterns of label observed in parasagittal sections through the lateral pons. Three different cases are shown. A: Retrogradely labeled cells from the injection shown in Figure 12 are indicated by arrows in a section through the exiting fifth and seventh nerves (5n, 7n). Most lie along the rostral and ventral border of the principal trigeminal nucleus (pV). B: Distribution of anterograde terminal label following a WGA-HRP injection into the brow is indicated by arrows in this parasagittal section. Bright terminals form a crescent-shaped band that runs along the rostral and ventral aspects of pV and extends into pars oralis of the spinal trigeminal nucleus (sVo). C,D: Retrogradely labeled neurons from the case shown in Figure 14. Arrows indicate cells labeled retrogradely from an oculomotor injection of WGA-HRP. Punctate terminal label from a brow injection of WGA-HRP is distributed in the vicinity of the retrogradely labeled cells. Scale bars = 1.0 mm in A; 100 μm in C,D. Scale in A applies to B.

The experiments described above suggest that the lateral pontine neurons projecting to the oculomotor complex are not a source of a noradrenergic input to extraocular motoneurons. Another possibility is that these neurons are part of the trigeminal blink reflex pathways that are activated by corneal and periorbital stimulation. The location of the labeled neurons along the ventral border of the trigeminal sensory nucleus is noteworthy in this regard. Studies in cats have indicated that the primary afferents from the cornea and supraorbital nerve terminate ventrally in the principal sensory nucleus and pars oralis of the spinal trigeminal nucleus (Marfurt, 1981; Panneton and Burton, 1981). In order to compare this terminal distribution to that of the neurons projecting to III, the area of the supraorbital nerve peripheral field was injected with WGA-HRP and the brainstem was cut in the sagittal plane. Figure 13B shows the terminal distribution in one section from this case. These terminals are located immediately dorsal to the exiting trigeminal and facial nerves, and the terminal field extends in a crescent shape along the rostral edge of the principal trigeminal nucleus. Thus, this terminal field has a location and shape similar to that of the distribution of neurons projecting to III (compare Fig. 13A with B).

We further tested for overlap in their distribution by injecting WGA-HRP into the brow and lid, as well as into III and CC of the same animal (Fig. 14D). The peripheral injection once again anterogradely labeled a crescent-shaped terminal field (stipple, Fig. 14A–C) in the ventral and rostral part of the trigeminal sensory nucleus. It also retrogradely labeled facial motoneurons (triangles, Fig. 14C). This terminal field overlapped with some members of the population of retrogradely labeled cells (dots) from the III injection that lined the trigeminal nucleus (Fig. 14A–C). Labeled puncta that were presumed to be anterogradely labeled terminals could be seen in the immediate vicinity of these retrogradely labeled cells (Fig. 13C,D). The number of labeled puncta was much more extensive than that which was observed with just III injections (compare Fig. 13A with C,D).

Figure 14.

Comparison of the distribution of trigeminal afferent terminations and cells projecting to the oculomotor complex following peripheral and central injections of WGA-HRP in the same case. A–C: Medial to lateral series of parasagittal sections through the trigeminal sensory nucleus. D: The central WGA-HRP injection site was located in the caudal end of the oculomotor nucleus (III) and included the caudal central subdivision (CC). It resulted in retrogradely labeled neurons (dots) arranged in a crescent along the rostral and ventral aspects of the principal trigeminal sensory nucleus (pV) (A–C). The injection of WGA-HRP into the brow and lid resulted in terminal label (stipple) in a crescent-shaped band lining the rostral and ventral aspects of the principal nucleus (pV) and extending into pars oralis of the spinal trigeminal nucleus (sVo). This injection also retrogradely labeled neurons (triangles) in the facial nucleus (VII). The distributions of trigeminal afferent terminals (stipple) and trigemino-oculomotor cells (dots) partially overlapped along the ventral and rostral edge of the principal trigeminal nucleus.

DISCUSSION

The electrophysiological and anatomical results from this study present a compelling case for the presence of a trigemino-oculomotor projection that derives from cells lining the rostral and ventral portion of pontine trigeminal sensory nucleus, and which produces IPSPs in levator motoneurons. This system is characterized by its bilaterality. Not only does the trigemino-oculomotor terminal field have an ipsilateral component, levator motoneurons can project contralaterally, ipsilaterally, or bilaterally, and their dendritic fields do not respect the midline. Thus, the anatomy of this system in the cat appears to favor conjugate lid closure. This circuit is schematically illustrated in Figure 15.

Figure 15.

Schematic illustration of the findings from this series of experiments. Ophthalmic division trigeminal afferents from the lids and brow terminate in a crescent-shaped field that lines the rostral and ventral aspect of the principal trigeminal sensory nucleus (gray field). This terminal field partially overlaps with the cell bodies (dots) of a population of neurons that project to the caudal central subdivision and overlying supraoculomotor area. The axons of the these trigemino-oculomotor cells terminate bilaterally, with a contralateral predominance, presumably contacting levator palpebrae motoneurons (soma, hexagon). The dashed line indicates the midline. The orbicularis oculi muscle is only shown in the lower lid for clarity.

The anatomical results from this study confirm the presence of a projection by a set of neurons in the lateral pons to III that were initially noted by Graybiel and Hartwieg (1974). They extend these previous findings by showing that these neurons do not represent a source of noradrenergic input from the Kölliker-Fuse area to the oculomotor complex. While this finding rests on negative evidence, i.e., a lack of doubly labeled Kölliker-Fuse neurons, it is backed by the fact that doubly labeled neurons were found elsewhere, and the fact that the populations of retrogradely labeled neurons and antibody-labeled neurons in the Kölliker-Fuse area were adjacent, but nonoverlapping. The present results instead indicate that this pathway represents part of a circuit linking the trigeminal sensory nucleus to the oculomotor complex. Specifically, their terminal distribution precisely matched the location of levator motoneuron dendrites. Furthermore, examination of the distribution of these neurons in the parasagittal plane reveals that they are not a sparse scattered population. On the contrary, they are a continuous band of cells that line the rostral and ventral aspect of the principal trigeminal nucleus. This distribution is very similar to that which we observed when the central terminations of the supraorbital branch of the trigeminal sensory nerve were labeled, suggesting that the trigemino-oculomotor neurons could be activated by trigeminal primary afferents from the regions that normally elicit blinks, and that they are responsible for the IPSPs observed in levator motoneurons with electrical stimulation.

Comparison to previous studies

Eye movements, like somatic movements, vary with respect to level of alertness. One possible role for the projections of these lateral pontine neurons to III might have been to provide this state-dependent modification, for the parabrachial complex has been implicated by some studies in control of state (Fulwiler and Saper, 1984). However, the only doubly labeled neurons lay outside the Kölliker-Fuse area. Moreover, their small number suggests that changes in oculomotor motoneuron activity due to alertness levels may be primarily regulated above the level of the motoneuron.

Based on retrograde studies, lateral pontine cells projecting to III have been previously reported in cats (Graybiel and Hartweig, 1974; Graybiel, 1977), as well as in other species (rabbit: Guerra-Seijas et al., 1993; macaque: Steiger and Büttner-Ennever, 1979), suggesting the pathway described here may be a general mammalian feature. However, only ipsilateral neurons were plotted in the macaque, and the neurons described in the rabbit appear to be concentrated dorsally in the intertrigeminal area lateral to the trigeminal motor nucleus. Nevertheless, in all cases neurons distributed around the exiting 5th and/or 7th nerve were described. The present cat results show that these neurons represent a continuous, rostrocaudally oriented band of cells, whose position is very similar to that of the terminal field of the supraorbital branch of the trigeminal nerve. The ventral location of the central terminations within the principal nucleus and pars oralis of the spinal trigeminal nucleus has been demonstrated in the cat for the cornea, and the supraorbital and frontal nerves (Marfurt, 1981; Panneton and Burton, 1981; Shigenaga et al., 1986). In the Shigenaga et al. (1986) study, the extension of the terminal field along the rostral pole of the principal nucleus was also indicated. More varied results have been observed in other species. In rats, Marfurt and Del Toro (1987) did not see a rostral termination zone from the cornea in the principal nucleus or pars oralis. However, a small terminal zone from the supraorbital nerve was observed ventrally in the principal nucleus by Pellegrini et al. (1995). In rabbits, a small projection is present from the lower eyelid and cornea, but not upper eyelid; while in macaques, only a very minor projection was observed from the cornea and upper eyelid (Marfurt and Echtenkamp, 1988; VanHam and Yeo, 1996a; May and Porter, 1998). In general, these anatomical studies correlate with physiological recording studies that indicate the periocular area is represented ventrally within the principal and spinal trigeminal nuclei (Kerr et al., 1968; Warren and May, 2002).

These studies, together with the present data, suggest that the terminals from the orbital part of the trigeminal representation terminate in a position where they would either overlie or lie immediately adjacent to the cells of origin for the trigemino-oculomotor projection (Fig. 15). This relationship makes primary afferent synaptic contact with trigemino-oculomotor neurons a distinct possibility. The latency difference we observed between peripheral and central stimulation (1.4 ms) could be produced by either a monosynaptic or disynaptic connection. In either case, these afferents would be in a position to activate the trigemino-oculomotor projection when the cornea or periorbital skin was stimulated, as would be expected when a blink is elicited by the trigeminal sensory system.

The trigemino-oculomotor pathway’s course and terminal distribution, as shown here, confirm and extend previous work in the cat (Ogasawara, 1985) and the rabbit (Guerra-Seijas et al., 1993; Van Ham and Yeo, 1996b). Specifically, Ogasawara (1985) injected tritiated amino acids into the rostral trigeminal sensory nucleus and showed a patch of terminals present in the supraoculo-motor area above the contralateral, caudal oculomotor nucleus. The present findings elucidate the axonal path of the projection, and show that there is also a smaller ipsilateral projection. The trigemino-oculomotor projection of the rabbit follows the same path (Guerra-Seijas et al., 1993), and also has an ipsilateral component (Van Ham and Yeo, 1996b). However, in the rabbit, a study using anterograde transport of biocytin (Guerra-Seijas et al., 1993) demonstrated a much larger portion of the terminals within III, while the one using anterograde transport of WGA-HRP (Van Ham and Yeo, 1996b) primarily demonstrated a terminal field along the border region between III and SOA. The latter pattern is similar to that demonstrated with WGA-HRP in the present experiments.

Functional significance

When a blink is evoked by the trigeminal sensory system, closure of the eyelid is produced by activation of the orbicularis oculi muscle, which is controlled by motoneuron activity in the facial nucleus (Schultz et al., 2010). In some cases, the eyeball may also be retracted back into the orbit. This is primarily accomplished by activation of the retractor bulbi muscles in those animals that posses them, and it allows the nictitating membrane to slide over the eye in those species that have this additional eye protection. The retractor bulbi muscles are supplied by motoneurons in the accessory abducens nucleus, and these motoneurons receive a strong monosynaptic input from the spinal trigeminal nucleus (Baker et al., 1980; Berthier and Moore, 1983; Holstege et al., 1986a,b; Van Ham and Yeo, 1996b). In mammals without retractor bulbi muscles, the eyeball is believed to be withdrawn into the orbit by co-contraction of the recti muscles (Evinger et al., 1984; Delgado-Garcia et al., 1990). However, it has been suggested that even in animals that have retractor bulbi muscles, all the extraocular muscles except the superior oblique are brought into play (Evinger and Manning, 1993). The result of the contraction of the retractor and recti muscles, in addition to pulling the eyeball into the orbit, is a variable upward movement of the eye termed Bell’s phenomenon.

Ogasawara (1985) suggested that the trigemino-oculomotor projection he observed in the cat specifically terminated on superior rectus motoneuron dendrites in the supraoculomotor area, and that this pathway was responsible for the upward movement of the eyes in Bell’s phenomenon. However, eye movements also accompany blinks evoked by other stimuli such as light flashes (Tackmann et al., 1982; Manning and Evinger, 1986), and the neurons located in or near the trigeminal nucleus, which are the source of the trigemino-oculomotor pathway, seem somewhat unlikely candidates to carry this signal. The caudal terminal distribution of the trigemino-oculomotor projection seen in the present study certainly matches the caudal distribution of superior rectus motoneurons within III (Gacek, 1974; Akagi, 1978), but cannot be said to exclude other groups. In rabbit, the termination of the trigemino-oculomotor projection has been suggested to overlap the distribution of superior rectus motoneurons in one study (Guerra-Seijas et al., 1993), but such a close match with superior rectus motoneuron distribution was not observed in a second study (Van Ham and Yeo, 1996b). The latter authors suggested that this difference may relate to the part of the trigemino-oculomotor pathway involved. While we did not specifically investigate the superior rectus in the present study, we did penetrate nonlevator motoneurons lateral to the caudal central subdivision, and did not observe EPSPs in these cells from our trigeminal stimulation sites. Nevertheless, the possibility exists that the rarer intermediate-sized neurons observed in the trigemino-oculomotor population might be the source of an excitatory drive to non-levator motoneurons.

Two muscles play a crucial role in producing a proper blink movement. Normally the upper eyelid is held up by the tonic action of the levator muscle, but this activity abruptly subsides as the orbicularis oculi muscle is activated (Evinger and Manning, 1993). This has been shown in monkeys to be caused by a loss of activity in levator motoneurons (Fuchs et al., 1992). Thus, a likely role for the trigemino-oculomotor pathway is to provide inhibition of levator motoneurons during the down-phase of a blink elicited by trigeminal stimulation. We have demonstrated that both peripheral and central trigeminal stimulation produce IPSPs in levator motoneurons. Furthermore, this interpretation receives support from the present data in which the caudal distribution of terminals matches the location of the levator motoneurons, i.e., terminals and levator motoneuron dendrites are found in the CC, SOA, and along the midline between the oculomotor nuclei (Fig. 15). A similar match has been reported in the rabbit, even though the levator motoneurons are located dorso-lateral to III in this species (Guerra-Seijas et al., 1993).

Unilateral stimulation of the cornea or brow in frontal-eyed species often produces a bilateral blink response in which the ipsilateral response occurs earlier and is stronger (Pellegrini et al., 1995). Our physiological recordings did not reveal this pattern with respect to the latency and strength of the IPSPs in levator motoneurons, but the small sample produced by the intracellular study might not reveal the organization across the population. On the other hand, the pattern of laterality of the projections demonstrated anatomically would seem to support this behavior. Specifically, the projection of the trigemino-oculomotor pathway is primarily crossed, as is the innervation of the levator muscle (Chen and May, 2002, 2007). Thus, the greater effect would be on the eye ipsilateral to the stimulation. In addition, the ipsilateral portion of the trigemino-oculomotor projection could inhibit the levator motoneurons innervating the contralateral eyelid, in agreement with the observed bilaterality of the reflex, or it could terminate specifically on those motoneurons projecting to the ipsilateral levator muscle. In the latter case, a circuit interconnecting the trigeminal nuclei would be needed to produce levator inhibition for conjugate blinks (Fig. 1I). On the other hand, if the trigemino-oculomotor projection excites the other extraocular motoneurons in III, its laterality is more curious. As a primarily crossed projection, it would mainly excite the muscles in the contralateral eye, with the exception of the superior rectus, which is supplied contralaterally.

That said, the morphology of the levator motoneurons appears to be particularly specialized for bilateral effects. Both ipsilaterally and contralaterally projecting cells were observed, along with rare cells in which bilateral antidromic activation was found. The latter finding agrees with anatomical evidence for bilaterally projecting levator motoneurons in monkeys (VanderWerf et al., 1997). In addition, the dendritic fields of the levator motoneurons extended throughout the CC into the SOA on both sides of the midline. Not only could this provide a substrate for conjugate blink reflexes, this bilaterality may be particularly important for levator action during vertical eye movements (Becker and Fuchs, 1988). Due to the fact that, unlike vertical extraocular muscles, the levator palpebrae superioris muscle has no secondary actions, the signals from both sides can be shared to produce conjugate action in the vertical plane (Porter et al., 1989). The dendritic fields of levator motoneurons appear to be specialized for this feature. On the other hand, these dendritic fields largely avoid III proper. This suggests that levator motoneurons do not simply share the vertical gaze signals provided to the superior rectus motoneurons. This finding agrees with evidence in cats and monkeys that a separate set of cells in or near the rostral interstitial nucleus of the medial longitudinal fasciculus provide levator motoneurons with the appropriate gaze signal (Horn et al., 2000; Chen and May, 2002).

The physiological and anatomical data provided here clearly indicate the presence of an inhibitory pathway suitable for inactivation of the antagonist of the orbicularis oculi muscle during blink down-phase (Fig. 15). These findings also strongly support the idea a trigemino-oculomotor projection is the central link in this pathway. However, the long latencies for inhibition leave the precise circuit details uncertain. First, it remains to be determined whether trigeminal afferents directly contact trigemino-oculomotor neurons, or whether other interneurons within the trigeminal nucleus proper are included in the circuit. Second, it is possible that trigemino-oculomotor axons are excitatory and terminate on inhibitory inter-neurons in the CC and/or SOA. However, since all the neurons we recorded, including the small nonmotoneurons observed in the CC, show IPSPs, they could not be the hypothetical inhibitory interneurons. Furthermore, we have observed direct synaptic projections in the macaque monkey (May et al., 2002). Thus, the more parsimonious assumption is that this projection is inhibitory, and that the numerous, very small trigemino-oculomotor cells have thin axons that largely account for the long latency and duration of the IPSPs.