Retrograde transneuronal tracing with rabies virus from the right orbicularis oculi muscle was used to identify neural networks underlying spontaneous, reflex, and learned blinks. The kinetics of viral transfer was studied at sequential 12 hr intervals between 3 and 5 d after inoculation. Rabies virus immunolabeling was combined with the immunohistochemical detection of choline acetyltransferase expression in brainstem motoneurons or Fluoro-Ruby injections in the rubrospinal tract. Virus uptake involved exclusively orbicularis oculi motoneurons in the dorsolateral division of the facial nucleus. At 3–3.5 d, transneuronal transfer involved premotor interneurons of trigeminal, auditory, and vestibular reflex pathways (in medullary and pontine reticular formation, trigeminal nuclei, periolivary and ventral cochlear nuclei, and medial vestibular nuclei), motor pathways (dorsolateral quadrant of contralateral red nucleus and pararubral area), deep cerebellar nuclei (lateral portion of interpositus nucleus and dorsolateral hump ipsilaterally), limbic relays (parabrachial and Kölliker–Fuse nuclei), and oculomotor structures involved in eye–eyelid coordination (oculomotor nucleus, supraoculomotor area, and interstitial nucleus of Cajal). At 4 d, higher order neurons were revealed in trigeminal, auditory, vestibular, and deep cerebellar nuclei (medial, interpositus, and lateral), oculomotor and visual-related structures (Darkschewitsch, nucleus of the posterior commissure, deep layers of superior colliculus, and pretectal area), lateral hypothalamus, and cerebral cortex (particularly in parietal areas). At 4.5 and 5 d the labeling of higher order neurons occurred in hypothalamus, cerebral cortex, and blink-related areas of cerebellar cortex. These results provide a comprehensive picture of the premotor networks mediating reflex, voluntary, and limbic-related eyelid responses and highlight potential sites of motor learning in eyelid classical conditioning.
- eyelid responses
- parietal cortex
- rabies virus
- red nucleus
- reflex blinks
- reticular formation
The kinematics and frequency domain properties of eyelid responses have been reported in a quantitative manner for spontaneous, reflex, passive, and learned movements in rabbits, cats, and humans (McCormick et al., 1982; Gormezano et al., 1983; Woody, 1986; Evinger et al., 1991; Welsh, 1992; Thompson and Krupa, 1994; Domingo et al., 1997; Gruart et al., 2000a). The activity of selected neural structures has been related convincingly to different aspects of reflex and learned eyelid responses. Thus, the properties of orbicularis oculi motoneurons involved in eyelid closing have been described previously (Trigo et al., 1999), and a relationship between neural firing responses and eyelid movements has been reported for several brain sites, e.g., red nucleus, cerebellar cortex and nuclei, motor cortex, and hippocampus (Berger et al., 1983; McCormick and Thompson, 1984; Berthier and Moore, 1986, 1990; Aou et al., 1992;Keifer, 1993; Gruart et al., 2000b; Múnera et al., 2001). A map of the sites involved in the generation of classically conditioned eyelid responses has been proposed (Kim and Thompson, 1997) on the basis of lesion or inactivation of given neural structures.
Nevertheless, despite some really valuable attempts (Courville, 1966;Takeuchi et al., 1979; Hinrichsen and Watson, 1983; Travers and Norgren, 1983; Fanardjian and Manvelyan, 1984, 1987a,b; Takada et al., 1984; Holstege et al., 1986a,b; Isokawa-Akesson and Komisaruk, 1987;Fort et al., 1989), available information regarding the organization of the eyelid premotor system is limited. Eyelid movements are specialized motor responses, related not only to corneal protection but also to emotional expression, visual perception, and eye movements and are susceptible to motor learning. Given the diversity of sensory sources able to activate eyelid muscles and the variety of behavioral displays in which they are involved, the underlying neuronal network should be rather complex. A comprehensive map of such networks could not be obtained with conventional tracers (Mesulam, 1982; Kuypers and Huisman, 1984; Köbbert et al., 2000) but now can be accomplished with retrograde transneuronal tracers that are able to reveal neuronal networks in their entirety. The most effective transneuronal tracers are viruses because of their ability to function as self-amplifying markers (Ugolini et al., 1989; Kuypers and Ugolini, 1990; Ugolini, 1995a,b). Rabies virus is the most valuable, particularly for studying motor networks, because after its injection into muscles or nerves it is taken up exclusively by motoneurons, without uptake by sensory or sympathetic neurons (Ugolini, 1995b; Tang et al., 1999; Kelly and Strick, 2000; Graf et al., 2002). Rabies virus propagates exclusively by retrograde transneuronal transfer. Transfer is time-dependent, enabling a sequential visualization of serially connected neurons across an unlimited number of synapses. Rabies-infected neurons remain viable and intact (Ugolini, 1995b; Tang et al., 1999; Graf et al., 2002).
We have exploited this powerful method to study the neural networks controlling the orbicularis oculi muscle. Our results provide a comprehensive map of such networks, showing the involvement of different sensory modalities (trigeminal, vestibular, auditory, and visual) in the genesis and control of eyelid responses and the participation of specific portions of sensorimotor cortex, red nucleus, reticular formation, and cortical and nuclear cerebellar areas.
Some results already have been published in abstract form (Ugolini et al., 1999).
MATERIALS AND METHODS
Viral tracer. The virus used in this study was the challenge virus standard (CVS), fixed strain 11 of rabies virus (Ugolini, 1995b). Concentrated virus [titer 1–1.5 × 1010 plaque-forming units (pfu)/ml] was prepared by pelleting the supernatant of baby hamster kidney-21 (BHK-21) cells infected for 72 hr through a cushion of 25% glycerol. The virus stock was kept frozen at −70°C until a few minutes before use.
Animals, virus injection, and perfusion procedures.Experiments were performed in 55 albino Wistar rats (Iffa Credo, Les Oncins, France), weighing 250–300 gm. Animal care and experimental procedures conformed with French Government, European Union Directive (86/609/EU), and National Institutes of Health guidelines. Surgery was performed aseptically under general anesthesia (Avertin 1.3–1.5 ml/100 gm, i.p., plus supplementary doses as required to maintain areflexia). The right orbicularis oculi muscle was exposed after a sagittal incision of the scalp. Using a glass micropipette connected to a pressure pump, we placed two injections of rabies virus (3–4.5 μl each) into the preseptal and pretarsal portions of the muscle, respectively, to map premotor innervation of both muscle components. The pipette was inserted in the muscle under microscopic guidance and kept in place for 5 min after each injection. Any eventual leakage of the inoculum was removed with cotton swabs. The wounds were sutured with sterile surgical silk. Rats were caged individually and examined daily. They were perfused for histological examination at sequential 12 hr intervals from 3 to 5 d after injection (3 d, n= 11 rats; 3.5 d, n = 16 rats; 4 d,n = 14 rats; 4.5 d, n = 4 rats; 5 d, n = 10 rats). None of the rats exhibited neurological symptoms or behavioral changes at these times, which correspond to the asymptomatic period of rabies (Ugolini, 1995b; Tang et al., 1999). At the chosen time points the rats were anesthetized deeply (Avertin, 2.5 ml/100 gm of body weight) and perfused transcardially with 500 ml of PBS, followed by 500 ml of 4% paraformaldehyde in 0.1 m phosphate buffer and 500 ml of 15% sucrose in 0.1 m phosphate buffer (perfusates, pH 7.4). The brains were dissected out and transferred to a 15% sucrose solution in which they were kept overnight at 4°C for cryoprotection before histological processing.
In five rats the rabies virus injections as above were combined in the same surgical session with ipsilateral spinal injections of the retrograde fluorescent tracer Fluoro-Ruby to identify rubrospinal neurons in the same experiment. For this purpose a hemi-laminectomy was performed to expose the right side of the C3–C5 spinal segments, and two injections (3 μl each) of Fluoro-Ruby 10% were made into the white and gray matter, centered on the dorsolateral funiculus in which the rubrospinal tract is located. These rats were killed for analysis 5 d later.
Histology. Brains were cut on a cryostat in coronal sections (50 μm) that were collected free-floating in four parallel series. In two alternative series the rabies immunolabeling was detected by using the peroxidase–antiperoxidase method (Ugolini, 1995b). One series was counterstained with cresyl violet. Sections from the remaining series were used for dual-color immunofluorescence detection of rabies virus and choline acetyltransferase, for double fluorescence detection of rabies virus and Fluoro-Ruby, or as a positive control in the various immunohistochemical reactions.
For immunoperoxidase staining the free-floating sections were incubated at room temperature in 0.3% H2O2 in PBS for 1 hr, followed by 0.2% swine normal serum in PBS for 1 hr, and then were incubated overnight at 4°C with rabbit polyclonal antibodies raised against rabies virus nucleocapsid (dilution, 40 μg/ml; Sanofi Pasteur, Paris, France). Subsequently, the sections were incubated at room temperature for 2 hr in swine anti-rabbit IgG (Dako, Trappes, France) diluted 1:200, followed by 2 hr in rabbit immunoperoxidase complex (Dako) diluted 1:200. Several washes with PBS were performed between steps. Peroxidase activity was revealed by incubation (2–10 min) in a metal-enhanced diaminobenzidine substrate kit (Pierce, Rockford, IL). Staining specificity was checked by incubating each series together with positive controls (i.e., sections from brains having shown rabies immunolabeling in previous reactions) and negative controls (sections from noninfected brains). Sections were mounted on gelatin-coated slides, air-dried, and coverslipped with Entellan (Merck, Whitehouse Station, NJ).
One series of sections from rats killed at 4 d was treated by using a dual-color immunofluorescence protocol for simultaneous detection of rabies virus and choline acetyltransferase, used here as a marker for motoneurons. Free-floating sections were incubated in 0.4% Triton X-100 for 30 min and in 3% donkey serum (Chemicon, Temecula, CA) for 1 hr, followed by incubation (20 hr) at 4°C in a mixture of mouse monoclonal antibodies recognizing the rabies P-protein (Laboratoire de Génétique des Virus, Gif-sur-Yvette, France) diluted 1:100 and goat anti-choline acetyltransferase (Chemicon AB144P) diluted 1:100. After being rinsed in PBS, the sections were incubated for 2 hr at room temperature in a mixture of fluorescein isothiocyanate (FITC)-conjugated donkey anti-mouse IgG (1:100; Jackson ImmunoResearch, West Grove, PA), and indocarbocyanine (Cy3)-conjugated donkey anti-goat IgG (1:100; Chemicon). Primary and secondary antibodies were diluted in PBS containing 3% donkey normal serum and 2% bovine serum albumin. Staining specificity was checked by omitting the primary or secondary antibody and by using positive and negative controls. After several washes in PBS the sections were mounted on slides and coverslipped with Vectashield (Vector Laboratories, Burlingame, CA).
In the experiments involving rabies immunolabeling in combination with retrograde labeling of rubrospinal neurons by means of the red fluorescent tracer Fluoro-Ruby, the sections were treated for immunofluorescent (FITC, green) visualization of the rabies antigen as described above.
Analysis. Fluorescent preparations were observed with a Leitz epifluorescence microscope. Selected images were captured by using a digital camera (Leitz DC-250, Wetzlar, Germany) and the IM-1000 Image Manager (Leica, Nussloch, Germany).
Rabies-immunolabeled neurons in the brain were plotted by using an analog x–y plotter connected by potentiometers to the microscope stage or the computer-assisted Neurolucida system (MicroBrightField, Colchester, VT). The location of labeled neurons was illustrated in various drawings. In all drawings the retrogradely labeled orbicularis oculi motoneurons are indicated by asterisks. Each dot represents one rabies transneuronally labeled neuron. The nomenclature is according to Paxinos and Watson (1986).
Kinetics of infection of orbicularis oculi motoneurons (first-order neurons)
After rabies virus injection into the right orbicularis oculi muscle, rabies-immunolabeled motoneurons were found exclusively within the dorsolateral subdivision of the ipsilateral facial nucleus (Fig.1 A), where orbicularis oculi motoneurons are known to be located (Martin and Lodge, 1977;Faulkner et al., 1997). Primary sensory neurons in Gasser's ganglion, which also innervate the orbicularis oculi muscle, were not labeled, confirming the lack of peripheral uptake of rabies virus by sensory neurons (Tang et al., 1999; Kelly and Strick, 2000; Graf et al., 2002). At none of the time points that were explored (3–5 d after inoculation) were labeled motoneurons found in other subdivisions of the facial nucleus (Fig. 1 A), showing that no spurious diffusion of the virus tracer occurred either from the orbicularis oculi muscle to neighboring facial muscles or within the facial nucleus itself.
Rabies-immunolabeled orbicularis oculi motoneurons showed normal size, morphology, and Nissl staining as well as normal levels of expression of the choline acetyltransferase marker. Glial cells were not infected. As illustrated in Figure 1 A, the number of positive motoneurons quantified at 3 d after inoculation was low (12 ± 4 SD; n = 8 animals) and did not show any significant increase at later time points (13 ± 3; counted fromn = 10 animals at 5 d). The low number of retrogradely labeled motoneurons is probably attributable to the fact that the tracer injections had involved only a few motor endplates because of their restricted, nonintermingled distribution in the orbicularis oculi muscle. Notably, the finding that labeled orbicularis oculi motoneurons did not become more numerous with time indicates that rabies virus uptake remained restricted to the injection sites, i.e., the virus tracer did not diffuse even within the muscle itself.
Retrograde transneuronal labeling of eyelid premotor networks
The kinetics of retrograde transneuronal transfer of rabies virus was studied at sequential 12 hr intervals from 3 to 5 d after injection. Transneuronal transfer was time-dependent. The results obtained at successive time points were highly reproducible, showing a progressive increase in the number of labeled neural sites and in the number of rabies-immunolabeled neurons. As observed in other rodent models (Ugolini, 1995b; Tang et al., 1999), 3 d after injection was sufficient for the onset of transneuronal labeling of second-order neurons, i.e., neurons projecting monosynaptically onto orbicularis oculi motoneurons. At this time point, however, transneuronal labeling occurred only in some cases and involved only a subset of the total population of second-order neurons, presumably those that are connected more heavily with orbicularis oculi motoneurons and that synapse proximally on the motoneurons soma-dendritic trees (Ugolini, 1995b;Graf et al., 2002). Thus, only a few labeled neurons were found in specific portions of the brainstem reticular formation (nuclei reticularis ventralis, dorsalis, magnocellularis, and cuneiformis) and auditory (periolivary) structures (see below) and at the ventral border of the spinal trigeminal nucleus in some cases.
At 3.5 d, transneuronally labeled neurons at these locations became more numerous, and additional sites became labeled (Fig.2). On the basis of the correspondence with the results of tracing studies (Courville, 1966; Takeuchi et al., 1979; Hinrichsen and Watson, 1983; Travers and Norgren, 1983; Takada et al., 1984; Holstege et al., 1986a,b; Isokawa-Akesson and Komisaruk, 1987; Fort et al., 1989), neuronal labeling at this time point still involved second-order neurons (Ugolini, 1995b). Notably, only a few labeled neurons were seen within the medial border of the ipsilateral spinal trigeminal nucleus, pars spinalis, interpolaris, and oralis (Fig. 2). Most of the labeled neurons were located ipsilaterally in specific portions of the caudal medullary lateral reticular formation (labeled already at 3 d), i.e., in the nucleus reticularis dorsalis and in the ventral part of the caudalmost portion of the nucleus reticularis parvocellularis as well as in the adjoining nucleus reticularis ventralis (Figs. 1 B,2 B–D). Contralaterally, labeling in the reticular formation at these levels was sparse and located mainly ventrally in the nucleus reticularis ventralis and parvocellularis (Fig.2 B–D). More rostral portions of the medullary lateral reticular formation showed very little labeling. In the medullary medial reticular formation the labeling involved primarily the nucleus reticularis magnocellularis bilaterally, mainly ipsilaterally (Fig. 2 E–G). A few labeled neurons also were found in the nucleus reticularis gigantocellularis, in the medial part of the nucleus reticularis paragigantocellularis dorsalis, and in the adjoining medial longitudinal fasciculus. More rostrally, some labeled neurons were seen in the nucleus reticularis pontis caudalis and oralis, particularly in the ventral reticular area neighboring the superior olivary complex. Notably, a dense accumulation of labeled neurons was found, mainly ipsilaterally, in auditory structures in which labeling already had been observed 12 hr previously, particularly in the caudal periolivary nuclei, in the nucleus of the trapezoid body and the ventral periolivary nucleus, and, to a lesser degree, in the rostral periolivary nucleus (Fig.2 H–J). Labeling also occurred bilaterally in the caudal part of the ventral cochlear nucleus (Figs.1 D, 2 F). A dense accumulation of labeled neurons was seen in the nucleus of Kölliker–Fuse and ventral parabrachial nuclei ipsilaterally (Fig. 2 J). Sparse labeling was found in the nucleus of the solitary tract in only two (of 16) cases. A few labeled neurons were seen in the medial vestibular nucleus at rostral levels (Fig. 2). Notably, dense labeling at this time point appeared in specific portions of the deep cerebellar nuclei, i.e., in the caudolateral part of the anterior interpositus nucleus and dorsolateral hump ipsilaterally (Fig. 2 G). In the mesencephalon a dense accumulation of labeled neurons was seen in the dorsolateral quadrant of the contralateral red nucleus and in the dorsally adjoining pararubral area (Ruigrok and Cella, 1995), which is part of the nucleus reticularis cuneiformis (Figs.1 F, 2 K,L), suggesting strong synaptic links with orbicularis oculi motoneurons. Sparse labeling occurred in the interstitial nucleus of Cajal (Fig.2 L). Some labeled neurons also were found consistently within the ipsilateral oculomotor nucleus and in the dorsally adjoining supraoculomotor area (Fig. 2 K). The neurons labeled within the oculomotor nucleus appeared to be smaller than motoneurons. They were identified clearly as oculomotor internuclear neurons and not motoneurons because they were not cholinergic, as shown by the results of dual-color immunofluorescence for simultaneous visualization of rabies and choline acetyltransferase antigens (see below).
A great increase in the distribution and number of labeled neurons occurred at 4 d after inoculation. This additional 12 hr interval is consistent with an additional synaptic step of transfer (Ugolini, 1995b; Tang et al., 1999; Graf et al., 2002). Correspondingly, transneuronal labeling obtained at this time point clearly involved higher order (presumably third-order) neurons connected polysynaptically to orbicularis oculi motoneurons.
The distribution of labeling in the medulla, pons, and mesencephalon at 4 d is illustrated in Figures 3-5. At medullary levels one of the characteristic features of this time point was the appearance of extensive labeling ipsilaterally in the principal and spinal trigeminal nuclei and in the dorsal horn of the first cervical segments (Figs. 1 C, 3 B, 4). Labeling was particularly dense and widely distributed in deep layers of the dorsal horn (Figs. 3 B,4 A) and in the adjoining caudalmost portion of the spinal trigeminal nucleus, pars caudalis (Fig. 4 B). At more rostral levels of the ipsilateral spinal trigeminal nucleus (pars caudalis, interpolaris, and oralis) and in the principal trigeminal nucleus, labeled neurons were seen only in the ventral part of these nuclei (Figs. 1 C,4 C–I). Only sparse labeling was found in the contralateral spinal trigeminal nuclei and dorsal horn (Fig. 4).
A considerable increase in the number and distribution of labeled neurons occurred also in the medullary and pontine reticular formation, where labeling became much more bilateral but still showed a clear dominance ipsilaterally in the portions of the nuclei reticularis dorsalis, ventralis, and parvocellularis that already had shown labeling at previous time points (Fig. 4). The increase in areal distribution of reticular neurons was particularly noticeable in the intermediate zone of the upper cervical segments, in the nucleus reticularis gigantocellularis, reticularis pontis caudalis and oralis, and reticularis cuneiformis (Fig. 4). Rabies immunolabeling provided a complete visualization of neuronal morphology, including distal dendrites (Fig. 3 A).
Another feature of the 4 d time point was the dense labeling bilaterally in the medial, descending, and superior vestibular nuclei, particularly in the medial vestibular nucleus (Figs.1 E, 4 F–I). Labeling of the parabrachial nuclei became more pronounced and bilateral, and some labeling occurred consistently in the nucleus of the solitary tract (Fig. 4 C,D).
Besides a moderate increase in the number of labeled neurons in the auditory structures labeled at previous time points, additional labeling appeared bilaterally in the dorsal cochlear nucleus (Fig.4 H), in external layers of the inferior colliculus, and in the area ventral to it corresponding to the dorsal nucleus of the lateral lemniscus (Figs. 3 C,5 A,B). Sparse labeling occurred also in more ventral portions of the nucleus of the lateral lemniscus.
In the contralateral red nucleus the labeled neurons increased in number but remained restricted to the dorsolateral quadrant of the red nucleus and the adjoining pararubral area (Figs. 3 D,5 C,D). The contralateral prerubral and retrorubral areas also were labeled densely (Fig. 5).
At 4 d, labeling also appeared in intermediate and deep layers of the superior colliculus, mainly contralaterally (Fig.5 C–E). Several mesencephalic oculomotor-related structures were labeled, such as the nucleus of Darkschewitsch, the interstitial nucleus of Cajal, and the nucleus of the posterior commissure (Fig.3 E), mainly ipsilaterally, the supraoculomotor area, and some internuclear interneurons located inside the oculomotor complex (Fig. 5 C–F). Besides the supraoculomotor area, labeling involved more dorsal portions of the central gray and the dorsal raphe nucleus (Fig. 5 B–D). Some labeling also was seen in the pretectal nuclei (Fig. 5 F). Because the pretectal area receives afferences from the retina (Holstege et al., 1986a,b), this pathway is potentially involved in flash-evoked blinks (Evinger et al., 1991; Gruart et al., 1995). Some labeling also was found bilaterally in the substantia nigra, lateral hypothalamus, and zona incerta (Fig. 5 E–G).
In the deep cerebellar nuclei (Figs. 4 H,6 A, 7), dense labeling occurred in the dorsolateral hump and in the lateral part of the anterior interpositus nucleus and posterior interpositus nucleus, except its caudalmost portion (Figs. 4 H, 6 A,7 B–E). Labeling was bilateral with a clear ipsilateral dominance (contralateral labeling involving mainly the dorsolateral hump) (Fig. 7 C). Bilateral labeling with a clear ipsilateral dominance appeared also in the medial (fastigial) nucleus and in the lateral (dentate) nucleus (Figs. 4 H,6 A, 7 B–E).
Another distinctive feature of the 4 d time point was the appearance of labeling in the cerebral cortex (Figs.8 C, 9), which is a clear reflection of polysynaptic links with orbicularis oculi motoneurons. Labeling involved exclusively pyramidal neurons in layer V (see example in Fig. 8 A). Most of the labeled neurons were located in parietal cortex areas 1 and 2 (Pa1 and Pa2; Paxinos and Watson, 1986) bilaterally, with a clear contralateral dominance (Fig. 9). A few labeled neurons also were located in frontal (F1, F2) and occipital (Oc2L) cortices contralaterally. In four (of 14) animals some labeled pyramidal cells also were found in the perirhinal and temporal (Te1) cortices (data not shown).
Data collected at 4.5 and 5 d were characterized by an increase in the number of labeled interneurons in trigeminal, vestibular, auditory, and oculomotor-related structures as well as in the reticular formation plus labeling at some additional sites, for example some neurons within the lateral vestibular nucleus. Even at these time points the specificity of transfer was illustrated by the lack of invasion of brainstem sites that clearly are unrelated to eyelid motoneurons, such as other brainstem motor nuclei (hypoglossus, abducens, ambiguous, etc.).
A noticeable increase in the number of labeled structures occurred at diencephalic levels. In the hypothalamus the labeling appeared in the anterior hypothalamus and the dorsomedial hypothalamic nuclei, besides the lateral hypothalamic area, already labeled at 4 d. The ventral posteromedial and the posterior thalamic nuclei showed some labeling bilaterally, presumably via their connections to the labeled cortical areas. Labeling in the cerebral cortex became more extensive. As illustrated in Figure 8, B and C, the number of labeled cells in the parietal cortex showed a great increase (>300%) compared with that obtained at 4 d (Fig.8 B), and the labeling of individual pyramidal neurons in layer V became Golgi-like (Fig. 8 C, compare withB). Both parietal areas Pa1 and Pa2 (Paxinos and Watson, 1986) were heavily labeled bilaterally, still with a contralateral dominance. Labeling also occurred consistently in some portions of temporal (Te1 and Te3) and Oc2L cortices contralaterally and in frontal (F1 and F2 areas) and perirhinal cortices bilaterally, whereas no labeling was found in entorhinal cortex or hippocampus.
In the cerebellum (Fig. 6 B,C) the labeling appeared in vermal and paravermal (c1–c3 areas) Purkinje cells (for references, see Gruart et al., 1997). The labeled Purkinje cells showed a clear zonal organization, being arranged in narrow rostrocaudal bands (Fig. 6 B,C).
Identification of oculomotor internuclear neurons projecting to orbicularis oculi motoneurons
Some rabies-immunolabeled neurons were found in the oculomotor nucleus, mainly ipsilaterally, in all cases from 3.5 d onward (monosynaptic time point). Their number and distribution did not increase substantially with time. To clarify whether they were interneurons or motoneurons, we treated sections through the oculomotor nuclei with a dual-color immunofluorescence protocol for simultaneous visualization of rabies and choline acetyltransferase. Sections through the facial nucleus, containing rabies-immunolabeled orbicularis oculi motoneurons, were reacted together as positive controls. Choline acetyltransferase is an excellent motoneuron marker, and we have shown previously that rabies-infected motoneurons are able to express choline acetyltransferase antigen at normal levels (Tang et al., 1999; Graf et al., 2002). The results showed unequivocally that rabies-immunolabeled neurons in the oculomotor nucleus are interneurons (and not motoneurons), because they are not cholinergic (Fig. 10). The possibility of false negative results for colocalization of rabies and choline acetyltransferase antigen in motoneurons clearly can be ruled out in view of the positive expression of the choline acetyltransferase antigen in rabies-infected orbicularis oculi motoneurons in the same experiments.
Specificity of rubral projection to orbicularis oculi motoneurons
Until now, it was unknown whether the rubral projections to the facial nucleus are mediated by separate rubral population or are axon collaterals of rubrospinal-projecting neurons (Mizuno and Nakamura, 1971). To clarify this issue, retrograde transneuronal labeling with rabies virus of the rubral neurons innervating orbicularis oculi motoneurons was combined in the same experiments with conventional (single-step) retrograde labeling of rubrospinal neurons by using Fluoro-Ruby. As expected, the injections of Fluoro-Ruby into the rubrospinal tract at upper cervical levels resulted in comprehensive labeling of rubrospinal neurons, as shown by their widespread distribution in all spinal-projecting portions of the red nucleus, including its lateral horn (Fig.11 A,C) (Huisman et al., 1981; Strominger et al., 1987; Ugolini, 1992; Tang et al., 1999). In contrast, the transneuronally labeled rubrofacial neurons occupied exclusively the dorsolateral quadrant of the red nucleus and the dorsally adjoining pararubral area (Fig. 11 B,D). Our results (Fig. 11) show that rubrofacial neurons innervating orbicularis oculi motoneurons monosynaptically and disynaptically are clearly a separate population from the rubrospinal-projecting ones. Despite the topographical overlap between the two populations in the dorsolateral quadrant, we found only a very small percentage (2%) of double-labeled neurons projecting with axon collaterals to both targets (Fig.11 C,D, arrows). This has negligible functional significance, if any, considering also that the projections from the red nucleus to functionally different targets generally are more collateralized in the rat than in other mammals (cat and monkey) (Huisman et al., 1981, 1982).
Neuronal networks involved in the generation and control of spontaneous, reflex, and learned eyelid responses were revealed here by transneuronal transfer of rabies virus from the orbicularis oculi muscle. Because of its highly specific propagation by retrograde transneuronal transfer across an unlimited number of synapses (Ugolini, 1995b; Tang et al., 1999; Kelly and Strick, 2000; Graf et al., 2002), the rabies virus tracer represents a formidable step forward compared with previous attempts to map such networks with conventional tracers (Takeuchi et al., 1979; Hinrichsen and Watson, 1983; Holstege et al., 1984, 1986a,b; Takada et al., 1984; Fort et al., 1989; Travers, 1995), which could help to locate putative sources of monosynaptic input to orbicularis oculi motoneurons, but not higher order relays. Virus uptake involved exclusively orbicularis oculi motoneurons in the dorsolateral facial nucleus division (Martin and Lodge, 1977; Faulkner et al., 1997). Infected motoneurons remained viable and did not become more numerous or more widely distributed with time, showing that rabies virus did not diffuse either within the muscle or within the facial nucleus. Transneuronal transfer was time-dependent. The observation that all known sources of monosynaptic input to orbicularis oculi motoneurons were labeled at 3.5 d confirms that the efficacy of transfer of rabies virus is not dependent on transmitter types (Ugolini, 1995b; Graf et al., 2002). However, second-order neurons that synapse proximally on the motoneurons soma-dendritic tree may be visualized earlier than neurons establishing more distal or weaker synaptic contacts (Ugolini, 1995b; Graf et al., 2002).
The results highlight the reflex pathways by which sensory inputs of trigeminal, auditory, vestibular, and visual origins can evoke eyelid responses and the participation of reticular, rubral, cerebellar, and cortical neurons to eyelid control. Potential sites of interactions between networks controlling eye and eyelid movements also are revealed here. Based on our results and as pointed out previously (Holstege et al., 1986a,b; Bracha and Bloedel, 1996), multiple pathways (with coinciding nodal points for different sensory modalities) are available in the eyelid premotor system as putative memory storage sites for classical conditioning of eyelid responses, a widely used model for associative motor learning (Gormezano et al., 1983; Woody, 1986;Thompson and Krupa, 1994).
Sensory modalities to orbicularis oculi motoneurons
Projections to orbicularis oculi motoneurons from the ipsilateral trigeminal nuclei (Erzurumlu and Killackey, 1979; Holstege et al., 1986a, 1988; Pellegrini et al., 1995; Van Ham and Yeo, 1996) mediate the R1 component of the blink reflex (Hiraoka and Shimamura, 1977). Such projections, visualized here at 3–3.5 d, were not substantial. Extensive neuronal labeling in the ipsilateral principal and spinal trigeminal nuclei and upper cervical dorsal horn appeared later (4 d), compatible with additional synaptic steps. The labeled sites receive trigeminal afferents of corneal and periocular origin (Panneton and Burton, 1981; Pellegrini et al., 1995). Polysynaptic trigeminal afferents to orbicularis oculi motoneurons mediate the R2 blink reflex component via reticular formation and dorsal horn synaptic relays (Hiraoka and Shimamura, 1977; Pellegrini et al., 1995).
Auditory pathways were already labeled heavily at monosynaptic time points, an unexpected result according to the paucity of tone-evoked reflex blinks in rabbits and cats (Gruart et al., 1995, 2000a). A likely explanation is that the rat presents a noticeable startle response. Our results indicate that the underlying connections are derived from the nucleus of the trapezoid body; the caudal, ventral, and rostral periolivary nuclei and neighboring portions of the nucleus reticularis pontis caudalis and oralis (López et al., 1999; Sinex et al., 2001); and the caudal part of the ventral cochlear nuclei bilaterally, mainly ipsilaterally. Higher order neurons are located in the dorsal cochlear nucleus, lateral lemniscus nuclei, and inferior colliculus.
Monosynaptic and disynaptic excitatory and inhibitory potentials have been evoked in facial motoneurons by stimulation of the medial and superior vestibular nuclei (Shaw and Baker, 1983). Our results reveal a few monosynaptic and numerous polysynaptic vestibular afferents (from the medial, superior, and descending vestibular nuclei) to orbicularis oculi motoneurons. Such pathways may be involved in eye–eyelid coordination, for example to prevent blinking during vestibulo-ocular reflex performance.
Visual information involved in light-evoked blinks reaches orbicularis oculi motoneurons via the olivary pretectal nucleus and related mesencephalic structures (Itoh et al., 1983), as confirmed here. Although both orbicularis oculi and accessory abducens motoneurons receive such visual input in the cat (Holstege et al., 1986a,b), only orbicularis oculi motoneurons are able to fire to flashlight presentation in alert behaving animals. Visual afferents to accessory abducens motoneurons are an example of “silent” pathways (Delgado-Garcı́a et al., 1988; Delgado-Garcı́a, 1998).
Reticular formation and limbic brainstem afferents
Our results show that specific areas of the medullary, pontine, and mesencephalic reticular formations project onto orbicularis oculi motoneurons. Particularly, ipsilateral projections have substantial monosynaptic (and polysynaptic) components. From our results, these reticular areas are connected more heavily with orbicularis oculi motoneurons than are the trigeminal nuclei. Reticular neurons in such areas that project to the orbicularis oculi division are activated by ipsilateral supraorbital nerve stimulation (Inagaki et al., 1989). Bilateral reticular projections to orbicularis oculi motoneurons explain the bilateral blinks (R2 component; Kugelberg, 1952) evoked by unilateral electrical stimulation of the supraorbital nerve (Pellegrini et al., 1995). They also may explain why eyelid classically conditioned responses present a (weaker) component contralateral to the side of the unconditioned stimulus presentation (Gruart et al., 1995).
The different reticular nuclei probably are involved in the generation and/or integration of commands of different origin (motor cortex, basal ganglia, limbic system). Based on in vitro effects of acetylcholine on facial motoneurons (Magariños-Ascone et al., 1999), it may be proposed, for instance, that cholinergic neurons in the dorsal medullary reticular formation, near the hypoglossal nucleus, are involved in the generation of orbicularis oculi motoneuron tonic firing, which is characteristic of R2 responses, eyelid-friendly displays, and classically conditioned responses (Holstege et al., 1986a,b; Fort et al., 1989; Travers, 1995; Trigo et al., 1999).
Some of the identified pathways, such as those derived from the parabrachial and Kölliker–Fuse nuclei, may be involved in the genesis of premotor signals related to the expression of internal emotional states, because limbic structures project to these nuclei through the central amygdala and hypothalamus. Such pathways are not affected by lesions of the pyramidal fibers, explaining the possibility of emotional expression in the absence of voluntary eyelid responses (Holstege et al., 1986a,b).
Eye muscles–eyelid premotor relationships
As shown here, structures belonging to eye movement networks (oculomotor internuclear neurons, supraoculomotor area, interstitial nucleus of Cajal, nucleus of Darkschewitsch, superior colliculus) also innervate monosynaptically or polysynaptically the orbicularis oculi motoneurons. This common network is likely to be involved in coordination of eyelid and eye movements during blink (Bour et al., 2000), intentional saccades, and fast phases of the vestibulo-ocular and optokinetic reflexes, mostly for eye movements in the vertical plane. No labeling occurred in the prepositus hypoglossi nucleus, which is involved in horizontal eye movements. Projections to facial motoneurons from these structures were reported in tracing studies (Takeuchi et al., 1979; Hinrichsen and Watson, 1983; Takada et al., 1984; Holstege et al., 1986a; Isokawa-Akesson and Komisaruk, 1987; Fort et al., 1989) and electrophysiological experiments (Fanardjian and Manvelyan, 1987b; Vidal et al., 1988; May et al., 1990).
Red nucleus and cerebellum
Our results indicate that eyelid-related rubrofacial neurons are clustered in the dorsolateral quadrant of the contralateral red nucleus and pararubral area. As shown here, eyelid-related rubrofacial pathways are clearly independent (i.e., not a collateral branch) of the rubrospinal tract. This was unknown previously (Courville, 1966; Mizuno and Nakamura, 1971; Yu et al., 1972).
Monosynaptic and polysynaptic projections to orbicularis oculi motoneurons also are derived from the lateral part of the anterior and posterior interpositus nuclei, mainly ipsilaterally. Notably, interpositus neurons at these locations project to the red nucleus and to the labeled oculomotor-related nuclei (interstitial nucleus of Cajal, nucleus of Darkschewitsch, oculomotor interneurons, superior colliculus) (Fanardjian and Manvelyan, 1984; Gonzalo-Ruiz and Leichnetz, 1987). This double-projecting system probably is involved in regulating and/or reinforcing eyelid responses by exciting the rubrofacial pathway and disfacilitating the (antagonist) eyelid levator palpebrae muscle (Gruart et al., 2000b). Moreover, the interpositus nuclei project to the medullary reticular formation and other labeled brainstem structures (Mehler, 1983; Rubertone et al., 1990; Voogd, 1995). Evidence of polysynaptic connections to orbicularis oculi motoneurons from the medial and lateral cerebellar nuclei also was obtained. Labeling of cerebellar nuclei is in keeping with the involvement of motor cortex, red nucleus, and reticular formation in eyelid motor control. At later time points consistent with additional steps of transfer, labeling appeared, with a zonal distribution, in vermal and paravermal (c1–c3 zones) Purkinje cells. Notably, the c1–c3zones have been related to trigeminally evoked eyelid blinks (Gruart et al., 1997).
Higher order forebrain structures
Cerebral cortical areas and hypothalamic and thalamic nuclei were labeled starting from 4 d, consistent with polysynaptic connections with orbicularis oculi motoneurons. Correspondingly, cortical afferents to facial motoneurons in the cat are not monosynaptic but reach them through the trigeminal nuclei and/or the nearby reticular formation (Fanardjian and Manvelyan, 1987a). Labeling was restricted to corticofugal pyramidal cells in layer V. Parietal cortex labeling was bilateral, with a contralateral dominance. Bilateral cortical projections to the orbicularis oculi division of the facial nucleus exist in other mammals, including humans (Morecraft et al., 2001). In a study of the early gene c-fos expression during classical conditioning of eyelid responses, wide areas of parietal cortex were labeled (Gruart et al., 2000c). However, labeling was restricted to nonpyramidal cells located mainly in layers 2 and 3. Apparently, layer V pyramidal cells, which as shown here represent the highest level of eyelid motor control, are not involved in this plastic response.
This work was supported by European Union Grant BIO4-CT98-0546. We are grateful to Dr. A. Gruart for her comments and suggestions and to R. Churchill for help in the editing of this manuscript. S.M. was a visiting doctoral fellow at G.U.'s laboratory (fellowship supported by the Spanish Ministerio de Educación y Ciencia program FP-96).
Correspondence should be addressed to Dr. Gabriella Ugolini, Laboratoire de Virologie Moléculaire et Structurale, Institut de Neurobiologie Alfred Fessard, Bâtiment 32, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette, France. E-mail:.