OUP user menu

Cortical innervation of the facial nucleus in the non-human primate
A new interpretation of the effects of stroke and related subtotal brain trauma on the muscles of facial expression

Robert J. Morecraft, Jennifer L. Louie, James L. Herrick, Kimberly S. Stilwell-Morecraft
DOI: http://dx.doi.org/10.1093/brain/124.1.176 176-208 First published online: 1 January 2001

Summary

The corticobulbar projection to musculotopically defined subsectors of the facial nucleus was studied from the face representation of the primary (M1), supplementary (M2), rostral cingulate (M3), caudal cingulate (M4) and ventral lateral pre- (LPMCv) motor cortices in the rhesus monkey. We also investigated the corticofacial projection from the face/arm transitional region of the dorsal lateral premotor cortex (LPMCd). The corticobulbar projection was defined by injecting anterograde tracers into the face representation of each motor cortex. In the same animals, the musculotopic organization of the facial nucleus was defined by injecting fluorescent retrograde tracers into individual muscles of the upper and lower face. The facial nucleus received input from all face representations. M1 and LPMCv gave rise to the heaviest projection with progressively diminished intensity occurring in the M2, M3, M4 and LPMCd projections, respectively. Injections in all cortical face representations labelled terminals in all nuclear subdivisions (dorsal, intermediate, medial and lateral). However, significant differences occurred in the proportion of labelled boutons found within each functionally characterized subdivision. M1, LPMCv, LPMCd and M4 projected primarily to the contralateral lateral subnucleus, which innervated the perioral musculature. M2 projected bilaterally to the medial subnucleus, which supplied the auricular musculature. M3 projected bilaterally to the dorsal and intermediate subnuclei, which innervated the frontalis and orbicularis oculi muscles, respectively. Our results indicate that the various cortical face representations may mediate different elements of facial expression. Corticofacial afferents from M1, M4, LPMCv and LPMCd innervate primarily the contralateral lower facial muscles. Bilateral innervation of the upper face is supplied by M2 and M3. The widespread origin of these projections indicates selective vulnerability of corticofacial control following subtotal brain injury. The finding that all face representations innervate all nuclear subdivisions, to some degree, suggests that each motor area may participate in motor recovery in the event that one or more of these motor areas are spared following subtotal brain injury. Finally, the fact that a component of the corticofacial projection innervating both upper and lower facial musculature arises from the limbic proisocortices (M3 and M4) and frontal isocortices (M1, M2, LPMCv and LPMCd) suggests a potential anatomical substrate that may contribute to the clinical dissociation of emotional and volitional facial movement.

  • brainstem
  • cranial nerves
  • limbic cortex
  • motor cortex
  • BDA = biotinylated dextran amine
  • DY = diamidino-yellow
  • FB = fast blue
  • FD = fluorescein dextran
  • FR = fluoro ruby
  • LPMCd = dorsal lateral premotor cortex (or area 6V)
  • LPMCv = ventral lateral premotor cortex (or area 6V)
  • LYD = lucifer yellow dextran
  • M1 = primary motor cortex
  • M2 = supplementary motor area (or medial area 6)
  • M3 = rostral cingulate motor cortex (or area 24c)
  • M4 = caudal cingulate motor cortex (or area 23c)
  • MCA = middle cerebral artery
  • PB = phosphate buffer
  • PHA-L = Phaseolus vulgaris leucoagglutinin

Introduction

The complex nature of facial expression has greatly influenced the notion that the cerebral cortex plays an important role in mediating facial movement (Duchenne, 1862; Darwin, 1872; Jackson, 1884; James, 1890; Denny-Brown, 1950; Damasio, 1994). Underscoring this relationship are the many alterations in facial expression that follow subtotal brain injury and occur in psychiatric illness and cranial–cervical dystonia. Despite the critical nature of these disorders, we know little about how the cerebral cortex and facial nucleus interact structurally, and how alterations in this interaction may disrupt our ability to appropriately control the muscles of facial expression. For example, over the past century little has changed in the way we interpret the most common post-stroke condition, in which occlusion of the middle cerebral artery and its branches results in paralysis of the contralateral lower face, but sparing of the upper face. The clinical consequences remain explained by the suggestion that the upper facial musculature receives bilateral innervation from the primary motor cortex (M1) and the lower face only contralateral innervation from M1 (Papez, 1927; Pearson, 1946; Reed and Sheppard, 1976; Brodal, 1981; Wilson-Pauwels et al., 1988; Adams et al., 1997; Kandel et al., 2000). The continued teaching of this tenet is quite surprising, considering the fact that neuronal tracing studies conducted in human and non-human primates do not fully support the classic interpretation (Kuypers, 1958a, b; Jenny and Saper, 1987). For example, it has been shown that parts of the facial nucleus innervating the lower facial musculature receive heavy contralateral input from M1. However, parts of the facial nucleus containing motor neurones innervating the upper facial musculature receive little input from M1. Contributing further to the enigma is evidence indicating that the supplementary motor area (M2 or medial area 6), which contains a face representation, may not innervate the facial nucleus (Jürgens, 1984).

Further associated with the mysteries underlying facial deficits following localized brain damage are numerous behavioural and clinical observations that have demonstrated a dissociation between voluntary and emotional facial movements (Darwin, 1872; Jackson, 1884; Wilson, 1924; Monrad-Krohn, 1924, 1939; Allen, 1931; Karnosh, 1945; Podolsky, 1945; Brown, 1967; Brodal, 1981; Duchenne, 1990; Hopf et al., 1992; Damasio, 1994, 1995; Topper et al., 1995; Husain, 1997; Ross and Mathiesen, 1998; Urban et al., 1998; Hopf et al., 2000). The most frequently reported example is `volitional facial paralysis'. This occurs in patients who present with impaired voluntary movement in their lower facial muscles contralateral to a cortical or subcortical lesion but overcome the paralysis when responding to an emotionally provocative or humorous remark. Volitional facial paresis has been associated with lesions involving M1 and the underlying subcortical white matter (Hopf et al., 1992). Similarly, but less commonly encountered, is a condition where emotionally affiliated facial movement is not only spared but excessive. In contrast to this is the inverse condition historically known as `amimia', or more commonly known as `emotional facial paralysis'. This disorder is characterized by an inability to smile on one side of the face; however, voluntary control over the same set of facial muscles is largely unaffected. Emotional facial paresis has been reported in patients with lesions involving the midline cortex, insula, thalamus, striatocapsular region and pons (Laplane et al., 1977; Hopf et al., 1992, 2000; Damasio, 1994; Urban et al., 1998). Collectively, these observations have led to the hypothesis that separate neural systems may mediate emotional and voluntary facial movements. Closely related is the recent suggestion that the anterior cingulate cortex may mediate emotionally associated movements in the upper face and the lateral frontal cortices volitional movements in the lower face (Damasio, 1994). However, specific neural systems that may form an integrated network between voluntary and emotional portions of the cortex and musculotopically defined parts of the facial nucleus have not been localized.

Several lines of research have revealed that there are at least five cortical face representations in the human and non-human primate brain, with each face representation being affiliated with five distinct cortical motor areas. (Fig. 1A and B) (Ferrier, 1886; Horsley and Schäfer, 1888; Penfield and Welch, 1951; Woolsey et al., 1952, 1979; Bates, 1953; Luschei et al., 1971; Muakkassa and Strick, 1979; McGuinness et al., 1980; Sirisko and Sessle, 1983; Luppino et al., 1991; Morecraft and Van Hoesen, 1992; Paus et al., 1993; Tanji, 1994; Godschalk et al., 1995; West and Larson, 1995; Morecraft et al., 1996; Picard and Strick, 1996; Tokuno et al., 1997; Wu et al., 2000; Morecraft and Yeterian, 2001). They include the face representation of the primary motor cortex (M1-areas F1 and F4), ventral lateral-premotor cortex (LPMCv-area 6V), supplementary motor cortex (M2-area 6m), rostral cingulate motor cortex (M3-area 24c) and caudal cingulate motor cortex (M4-area 23c). The fact that there is widespread face representation at the cortical level raises the possibility that each face representation may innervate the facial nucleus, i.e. much like all cortical arm representations, selectively innervate brachial levels of the spinal cord and leg representations the lumbosacral levels (Biber et al., 1978; Murray and Coulter, 1981; Hutchins et al., 1988; Galea and Darian-Smith, 1994; Luppino et al., 1994; Rouiller et al., 1994; Morecraft et al., 1997). It is possible to speculate that all cortical face representations may not only innervate the facial nucleus but also preferentially target subsectors controlling different groups of facial muscles, this premise being based upon the fact that the various subsectors of the primate facial nucleus are musculotopically organized (Fig. 1C) (Jenny and Saper, 1987; Satoda et al., 1987; Porter et al., 1989; Welt and Abbs, 1990; VanderWerf et al., 1998) and the fact that M1 preferentially innervates only the contralateral lateral subnucleus (Jenny and Saper, 1987). Finally, defining the organization of the corticofacial projection from M3 and M4, whose origins are embedded in the proisocortical portion of the `emotive' limbic lobe, may assist in interpreting the clinical observations that have long suggested a dissociation of voluntary and emotional facial movements.

Fig. 1

(A) Schematic diagrams of the medial (top) and lateral (bottom) surfaces of the cerebral cortex of the rhesus monkey depicting the basic organization of the frontal lobe and adjacent cingulate cortex. (B) Enlarged representation of A illustrating the basic somatotopical organization of the motor cortices. (C) Transverse section through the lower pons illustrating the location of the facial nucleus (top), the major nuclear subdivisions (bottom right) and the basic musculotopic organization (bottom left). A = arm; as = arcuate sulcus; cf = calcarine fissure; cgs = cingulate sulcus; cs = central sulcus; Ea = ear; F = face; FEF = frontal eye fields; Fr = frontalis; ios = inferior occipital sulcus; ips = intraparietal sulcus; L = leg; lf = lateral fissure; LL = lower lip; LPMCd = dorsal lateral premotor cortex; LPMCv = ventral lateral premotor cortex; ls = lunate sulcus; M1 = primary motor cortex; M2 = supplementary motor cortex; M3 = rostral cingulate motor cortex; M4 = caudal cingulate motor cortex; OO = orbicularis oculi; P = platysma; pre-SMA = pre-supplementary motor cortex; rs = rhinal sulcus; SEF = supplementary eye field; sts = superior temporal sulcus; poms = medial parieto-occipital sulcus; sts = superior temporal sulcus; UL = upper lip.

The occurrence of altered facial expressions that accompany subtotal brain trauma and present in psychiatric illness and cranial–cervical dystonia implies that a strong and highly organized structural linkage occurs between the cerebral cortex and the facial nucleus. The fact that selective muscle groups are adversely affected following localized brain damage indicates that the various cortical face representations may preferentially influence different groups of facial muscles. To advance our understanding of these important neurological issues, we studied the organization of the descending projection from the face representation of five major cortical motor areas to musculotopically defined subsectors of the facial nucleus. In addition, we investigated a potential corticofacial projection from the ventral part of the dorsal lateral premotor cortex (LPMCd), which primarily gives rise to arm movements following stimulation, but occasionally face movements. Our findings provide new insight to the long-standing question of cortical innervation of the upper and lower facial musculature. Together, these observations offer a new and alternative explanation for the clinical deficits that occur following localized cortical damage to the lateral and medial regions of the cerebral hemisphere.

Material and methods

The corticobulbar projection to the facial nucleus was studied in seven adult rhesus monkeys (Macaca mulatta). All protocols received Institutional Animal Care and Use Committee approval and were conducted in accordance with USDA and the Society for Neurosciences guidelines. Each monkey was immobilized with ketamine hydrochloride (10 mg/kg) then anaesthetized intravenously with pentobarbital (2.5–4 mg/kg/h). The head was shaved, scrubbed with betadine and the animal placed into a NIH head holder or Kopf stereotaxic head holder.

Surgery

The lateral surface of the cerebral cortex was exposed as described previously (Morecraft and Van Hoesen, 1992; Morecraft et al., 1992). Bridging cortical veins draining into the superior sagittal sinus were cauterized to expose the medial wall of the cerebral cortex. Intracortical microstimulation was used to localize the arm and face representations in M1, LPMCd, LPMCv and M2. Standard anatomical landmarks were used to localize the face representation in M3 and M4. Graded pressure injections (0.2–0.3 μl per penetration) of anterograde neuronal tracers were made into the various cortical face representations using a Hamilton microsyringe inserted 2–3 mm below the cortical surface under microscopic guidance. To minimize white matter involvement following M2 injections, the superior frontal lobule was gently retracted laterally and the tip of the needle was inserted directly into the intended grey matter lining the vertical wall without involving cortex on the dorsomedial convexity and its underlying white matter. For injections into the lower bank of the cingulate sulcus (e.g. M3 and M4) the arachnoid matter bridging over the cingulate sulcus was removed and the superior frontal lobule was retracted laterally exposing cortex in the lower bank. The needle tip was then inserted directly into the cingulate cortex, avoiding the adjacent frontal cortex (Fig. 2B and C). The location of each stimulation point and injection site was recorded on a JVC videocassette recorder (HR-S9400U, Dage-MTI, Inc., Michigan City, Ind., USA) directly connected to a three-chip colour camera (MTI DC-330, JVC, Aurora, Ill., USA) that was mounted on the photo/video port of the surgical microscope (Stortz M-703F, Olympus, Minneapolis, Minn., USA). These images were metrically calibrated and captured with respect to anatomical landmarks (e.g. sulci) and used for developing an accurate post-mortem reconstruction of the brain surface, the location of each physiological stimulation point and the penetration point of each tract tracer injection. Intramuscular injections of fluorescent retrograde neuronal tracers were made in upper and lower facial muscles in four of the seven monkeys studied by using a microsyringe inserted into a 20 g needle, which was then inserted into the muscle tissue. Each muscle was localized using the criteria of Huber (Huber, 1933).

Fig. 2

Plate of low-power bright-field photomicrographs of representative BDA injection sites in different motor fields, following tissue processing for immunohistochemical visualization. (A) Coronal section illustrating the ventral lateral premotor cortex (LPMCv) injection site in case SDM 22. (B) Coronal section depicting the rostral cingulate motor cortex (M3) injection site in case SDM 9. (C) Horizontal section showing the M3 injection site in case SDM 14. Scale bar in B = 5 mm and is applicable to all plates. CgS = cingulate sulcus; ILAS = inferior limb of the arcuate sulcus; PS = principle sulcus; SLAS = superior limb of the arcuate sulcus.

Intracortical microstimulation

A monopolar tungsten microelectrode (impedance 1.0–2.0 MΩ) was attached to a Grass S88 stimulator and a constant current stimulus isolation unit. The system was grounded and the electrode was lowered to a depth of 100 μm and advanced at 500-μm intervals. Cortex was mapped with a train duration of 50–100 ms and a pulse duration of 0.2 ms delivered at 330 Hz. Stimulation points were spaced 0.5–2.5 mm apart to minimize tissue damage. The lowest level of current needed to evoke movement was considered the movement threshold. During all experiments, applied current levels were continuously monitored by measuring the voltage with the use of a digital storage oscilloscope (LG Precision Co., Ltd, Ceritos, Calif., USA) and determining the system resistance. The current values ranged between 3 and 80 μA (Figs 3–6). All movements at threshold were discrete and confined to a specific body part (e.g. upper lip, tongue, wrist, finger).

Fig. 3

Summary diagrams of the experimental design in monkey SDM 14. (A) Corticobulbar terminals were studied in the facial nucleus following cortical injections (black irregular spheres) made into five face representations in the right cerebral cortex. The view of the medial wall is a mirror image, thus illustrating the right medial wall injection sites. Top pullout depicts the location of the LYD injection in M2 in relation to physiological mapping, the BDA injection in M3 and the FR injection in M4. Bottom pullout demonstrates the location of the PHA-L injection in M1 and the FD injection in the LPMCv in relation to physiological mapping. Each black dot represents a stimulation point labelled with corresponding threshold level (μA) and body part where the movement was observed. (B) In the same monkey, the musculotopic organization in the facial nucleus was studied following injections of retrograde dyes into selected facial muscles. In the left upper face, DY was injected into the frontalis (squares) and fast blue (FB) was injected into orbicularis oculi (x). In the right lower face, DY was injected into the upper lip (triangles) and FB was injected into the lower lip (inverted triangles). ac = anterior commissure; cc = corpus callosum; cf = calcarine fissure; D = digits; Ea = ear; ecs = ectocalcarine sulcus; El = elbow; ilas = inferior limb of the arcuate sulcus; ipcd = inferior precentral dimple; J = jaw; NR = no response; ps = principal sulcus; Sh = shoulder; spcd = superior precentral dimple; Th = thumb; To = tongue; Wr = wrist. For other conventions see Fig. 1.

Fig. 4

Summary diagrams illustrating the experimental design in monkey SDM 19. (A) Corticobulbar terminals were studied in the facial nucleus following cortical injections (black irregular spheres) made into two face representations in the right cerebral cortex. The view of the medial wall is a mirror image. The top pullout depicts the location of the BDA injection into M2. FR was also injected into the arm representation of M2. The bottom pullout demonstrates the location of the LYD injection in M1 and the FD injection in the arm representation of M1. Each black dot represents a stimulation point labelled with corresponding threshold level (μA) and body part where the movement was observed. (B) In the same monkey, the musculotopic organization in the facial nucleus was studied following injections of retrograde dyes into selected facial muscles. In the left upper face, DY was injected into the frontalis (squares) and FB was injected into orbicularis oculi (x). In the right lower face, DY was injected into the upper lip (triangles) and FB was injected into the lower lip (inverted triangles). For abbreviations see Figs 1 and 3.

Fig. 5

Diagrams illustrating the experimental design in monkey SDM 22. (A) Corticobulbar terminals were studied in the facial nucleus following cortical injections (black irregular spheres) made into five face representations in the right cerebral cortex. The view of the medial wall is a mirror image. The top pullout depicts the location of the FR injection into M2, the PHA-L injection into M3 and the LYD injection into M4. The bottom pullout demonstrates the location of the FD injection in M1 and BDA injection into LPMCv. Each black dot represents a stimulation point labelled with corresponding threshold level (μA) and body part where the movement was observed. (B) In the same monkey, the musculotopic organization in the facial nucleus was studied following injections of retrograde dyes into selected facial muscles. In the left upper face, DY was injected into the orbital auricularis (squares) and FB was injected into orbicularis oculi (x). In the right lower face, DY was injected into the upper lip (triangles) and FB was injected into the lower lip (inverted triangles). For abbreviations see Figs 1 and 3.

Fig. 6

Diagrams illustrating the experimental design in monkey SDM 23. (A) Corticobulbar terminals were studied in the facial nucleus following an injection of FD (irregular black spheres) made into arm/face area of LPMCd in the right cerebral cortex. The pullout demonstrates the location of the FD injection in relation to physiological mapping. Each black dot represents a stimulation point labelled with corresponding threshold level (μA) and body part where the movement was observed. For abbreviations see Figs 1 and 3.

Experimental design

Cortical and intramuscular injections

In all monkeys, anterograde tracers were injected into various cortical face representations to study the patterns of terminal labelling in the facial nucleus (Figs 2–6). Injections and representative volumes were made according to the relative size of each face representation and to minimize arm representation involvement when possible. The anterograde tracers used were 10% biotinylated dextran amine (BDA), 10% lucifer yellow dextran (LYD), 2.5% Phaseolus vulgaris leucoagglutinin (PHA-L), 10% fluorescein dextran (FD) and 10% fluoro ruby (FR). The FD injectate was composed of an equal mixture of 3000 and 10 000 MW volumes, as was the FR injectate. In monkeys SDM 14, 19, 20 and 22, specific facial muscles were injected with the retrograde tracers 4% fast blue (FB) and 4% diamidino-yellow (DY). This enabled us to examine the topographical relationship between corticofacial terminals and the general organization of facial lower motor neurones. The specific injection site parameters in each experiment are described below.

Monkey SDM 9.

FB was injected into the face representation of M1 (four penetrations; total volume 0.8 μl) (see Fig. 1 of Morecraft et al., 1996). In the contralateral hemisphere, BDA was injected into the face area of M3 (two penetrations; total volume 0.4 μl), positioned directly above the genu of the corpus callosum (Fig. 2B).

Monkey SDM 10.

FB was injected into the arm representation of M2 (one penetration; total volume 0.2 μl) and DY was injected into the face representation of M2 (one penetration; total volume 0.2 μl) (see Fig. 3 of Morecraft et al., 1996). On the contralateral hemisphere, BDA was injected into the rostral half of M4 (one penetration; total volume 0.4 μl), positioned at coronal levels involving the arcuate spur to levels extending 2 mm caudal to this landmark.

Monkey SDM 14.

All cortical injections were made into the right hemisphere (Figs 2C and 3A). PHA-L was injected into the face representation of M1 (four penetrations; total volume 0.8 μl) and FD was injected into the physiologically defined face representation of LPMCv (total volume 0.8 μl). On the medial wall of the hemisphere, LYD (total volume 0.6 μl) was injected into the face representation of M2, and BDA was injected into the rostral portion of M3 (lower bank of the cingulate sulcus) corresponding to coronal levels including the genu of the corpus callosum (two penetrations; total volume 0.6 μl) (Fig. 2C). Finally, FR (one penetration; total volume 0.4 μl) was injected into the rostral portion of M4 located at coronal levels including the arcuate spur. In the upper facial musculature located contralateral to the cortical injections, DY was injected into the frontalis/corrugator supercilii complex (three penetrations; 10 μl total volume) and FB was injected into the upper and lower orbital and preseptal portions of the orbicularis oculi (six penetrations; 10 μl total volume) (Fig. 3B). In the lower facial musculature located ipsilateral to the cortical injections, DY was injected into the upper orbicularis oris (upper lip) (three penetrations; 10 μl total volume) and FB was injected into lower orbicularis oris (lower lip) (three penetrations; 10 μl total volume) (Fig. 3B).

Monkey SDM 19.

All cortical injections were made in the right hemisphere (Fig. 4A). LYD was injected into the face representation of M1 (four penetrations; total volume 1.2 μl) and FD was injected into the arm representation of M1 (four penetrations; total volume 0.9 μl). FR was then injected into the arm representation of M2 (two penetrations; total volume 0.6 μl). However, during surgery a large bridging vessel required ligation over the rostral part of M2, yielding the underlying cortex (face area) unresponsive to stimulating current. Therefore, the face area was approximated as being localized immediately anterior to the physiologically defined arm representation and caudal to the pre-SMA (M2 pre-supplementary motor area) which resides in the superior frontal gyrus in a plane above the genu of the corpus callosum (Fig. 1A). BDA was injected into the face area of M2 (two penetrations; total volume 0.6 μl). In the upper facial musculature located contralateral to the cortical injections, DY was injected into the frontalis–corrugator supercilii complex (three penetrations; 10 μl total volume) and FB was injected into the upper and lower orbital and preseptal portions of the orbicularis oculi (six penetrations; 10 μl total volume) (Fig. 4B). In the lower facial musculature located ipsilateral to the cortical injections, DY was injected into the upper orbicularis oris (three penetrations; 10 μl total volume) and FB was injected into the lower orbicularis oris (three penetrations; 10 μl total volume) (Fig. 4B).

Monkey SDM 20.

All cortical injections were made in the right hemisphere. BDA was injected into the face representation of M1 (two penetrations; total volume 1.0 μl) and LYD was injected into the arm representation of M1 (two penetrations; total volume 1.0 μl). FR was injected into cortex lining the lower bank of the cingulate (M3) from a location established at coronal levels, including the genu of the corpus callosum, to a caudal location corresponding to a coronal plane located 2 mm rostral to the genu of the arcuate sulcus (10 penetrations; total volume 10 μl). FB was injected bilaterally into the orbital, preseptal and pretarsal portions of the orbicularis oculi (10 penetrations; total volume 25 μl).

Monkey SDM 22.

All cortical injections were made in the right hemisphere (Fig. 5A). On the lateral wall, FR was injected into the face representation of M1 (four penetrations; total volume 1.2 μl), and BDA was injected into the face representation of LPMCv (four penetrations; total volume 1.2 μl). FR was injected into the face area of M2 (four penetrations; total volume 0.4 μl). PHA-L was injected into the face area of M3 (two penetrations; total volume 0.5 μl) and LYD was injected into the face area of M4 (one penetration; total volume 0.4 μl). In the upper facial musculature contralateral to the cortical injections, FB was injected into the upper and lower orbital, preseptal and pretarsal portions of the orbicularis oculi (seven penetrations; 10 μl total volume) and DY was injected into the orbito-auricularis muscle (two penetrations; 10 μl total volume) (Fig. 5B). In the lower facial musculature located ipsilateral to the cortical injections, FB was injected into the upper orbicularis oris (three penetrations; 10 μl total volume) and DY was injected into the lower orbicularis oris (three penetrations; 10 μl total volume) (Fig. 5B).

Monkey SDM 23.

FD was injected into the ventral part of the arm area of LPMCd as defined using microstimulation (three penetrations; total volume 1.0 μl) (Fig. 6).

Tissue processing

Following a survival period of 21–30 days each monkey was deeply anaesthetized with Nembutal and perfused with 300–500 ml of 0.9% saline solution followed by 2 l of 4% paraformaldehyde in 0.1 M phosphate buffer (PB), then 1 l of 10% sucrose in 0.1 M PB. The perfusion was completed by an infusion of a 10% then 30% solution of sucrose in 0.1 M PB. The brain was removed, placed in 30% sucrose in 0.1 M PB and stored for 2–4 days at 4°C. Post-mortem photographs and electronic images were taken of the dorsal, ventral, lateral and medial surfaces of the cerebral cortex, brainstem and spinal cord. Each injected facial muscle in monkeys SDM 14, 19, 20 and 22 as well as the adjacent non-injected muscles were dissected and placed in 30% sucrose in 0.1 M PB, stored at 4°C.

The cortex was frozen then sectioned coronally (monkeys SDM 9, 10, 20 and 22) or horizontally (monkeys SDM 14, 19 and 23) on an American Optical sliding microtome (AO 860) at a thickness of 50 μm in cycles of 10. The brainstem was frozen with dry ice and cut transversely (perpendicular to its long axis) at a thickness of 50 μm in cycles of five. In monkey 5, brainstem serial sections were cut with a vibratome at a thickness of 200–300 μm. Finally, each muscle was frozen sectioned and examined for tracer localization (Vander Werf et al., 1996, 1998).

The first section of each cortical and brainstem series was mounted on subbed slides, dried overnight, coverslipped with DPX mounting medium (Aldrich Chemicals, Milwaukee, Wis., USA) and used for fluorescent analysis (Morecraft and Van Hoesen, 1992, 1993, 1998; Morecraft et al., 1993) (Fig. 7). After analysis, each fluorescent section was stained for Nissl substance using thionin. The second section in each series was processed for BDA using the avidin–biotin method (Fig. 8). Following successive rinses in PB, the sections were incubated in 0.05% DAB (diaminobenzidine tetrachloride) and 1% nickel ammonium sulphate for 10 min. Then, 0.009% hydrogen peroxide was added and the tissue incubated for another 10–15 min, rinsed in PB, serially dehydrated, mounted on subbed slides and coverslipped using Permount (Fisher Chemicals, Denver, Co., USA). The BDA injection site, axons and cells stained by nickel-enhanced DAB appeared blue-black upon visualization.

Fig. 7

Plate of photomicrographs showing examples of fluorescent dye-labelled facial motor neurones following intramuscular injections. (A) DY labelled neurone in the medial subnucleus following an injection in the orbito-auricularis in case SDM 22. (B) FB-labelled neurones in the intermediate subnucleus following an injection into the orbicularis oculi in SDM 14. (C) DY-labelled neurone in the lateral subnucleus following an injection into the upper lip in SDM 19. (D) FB-labelled neurone in the lateral subnucleus following an injection into the lower lip in SDM 22. Scale bar in D is applicable to all plates.

Fig. 8

Plate of photomicrographs illustrating terminal labelling in the facial nucleus following injections of anterograde tracers in the various face representations in the cerebral cortex. Unlabelled arrows identify examples of terminal boutons. (A) LYD fibres and terminals (blue reaction product) in the lateral subnuclei following an injection of LYD in M1 in case SDM 19 (×40). (B) BDA fibres and terminals (brown reaction product) in the lateral subnucleus following an injection in LPMCv in case SDM 22 (×40). (C) BDA fibres and boutons (brown reaction product) found in the medial nucleus following an injection of BDA in M2 in case SDM 19 (×40). (D) FR fibres and terminals (red reaction product) found in the medial subnucleus following an injection of FR in M2 in case SDM 22 (×40). (E) BDA fibres and terminals (brown reaction product) in the intermediate subnucleus following an injection of BDA in M3 in case SDM 14 (×60). (F) LYD fibres and terminals (blue reaction product) in the lateral subnucleus following an injection of LYD in M4 in case SDM 22 (×40). The brown reaction product in the same tissue section represents BDA-labelled fibres that occurred as a result of the BDA injection in LPMCv in the same cerebral hemisphere (see B).

Monkeys SDM 14, 19 and 22 received injections of multiple anterograde tracers (e.g. BDA, LYD, PHA-L, FR and FD). In these experiments, BDA was processed alone in a complete series of tissue sections as indicated above, mounted on glass slides and coverslipped. In additional serial sets of unprocessed tissue sections through the cortex, brainstem and spinal cord, double labelling immunohistochemistry was performed (Fig. 8F). In all of the double labelling experiments, BDA was reacted first according to the above protocol with the exception that BDA was stained brown using DAB in the absence of nickel ammonium sulphate (Fig. 8B and C). After reacting for BDA, the same tissue sections were then incubated in avidin–biotin blocking reagent (Vector SP-2001), rinsed in 10% normal goat serum for 2 h then incubated overnight, with the appropriate biotinylated antibody directed against a second neuronal tract tracer (e.g. biotinylated anti-LYD). The tissue was then incubated in a solution of avidin–biotinylated peroxidase complex, rinsed in buffer and stained blue using the Vector SG peroxidase substrate kit (Vector SK-4700). Thus, BDA was colourized brown and the second tracer (e.g. LYD) was stained blue in the same tissue sections (Fig. 8A and F). All sections were rinsed, dehydrated, mounted and coverslipped on glass slides. This process was conducted on another set of serial sections staining BDA brown with DAB, then another anterograde tracer blue (e.g. PHA-L), again using the appropriate antibody and blue reaction process to visualize the second tracer (e.g. biotinylated anti-PHA-L). FR and FD were also processed immunohistochemically in the same manner (e.g. BDA brown then FR blue). In these series, FR and FD labelling was verified by comparing the locations of immunochemically localized reaction products to fluorescent labelling in the non-reacted fluorescent series. The specific antibodies used for each anterograde tracer were biotinylated anti-Lucifer Yellow rabbit IgG fraction, biotinylated anti-PHA-L rabbit IgG fraction, biotinylated anti-FR rabbit IgG fraction and biotinylated anti-FD rabbit IgG fraction (all from Vector Laboratories, Burlingame, Calif., USA). The remaining cortical tissue sections were stained for myelin using the gold chloride method (Schmued, 1990) and cytochrome oxidase for evaluating histochemical boundaries of the frontal motor areas (Matelli et al., 1985).

Data analysis

Fluorescent material was studied using epi-fluorescent illumination and immunohistochemically processed material was examined under bright-field illumination. Data from all tissue sections were collected with the use of a conventional Hewlett Packard X-Y plotter (HP-7045B, Hewlett Packard, Palo Alto, Calif., USA) that was attached to the xy axes of a microscope stage (Olympus BX60, Minneapolis, Mass., USA). The outline of the tissue section, anatomical landmarks such as blood vessels and ventricles, the white and grey matter interfaces, periphery of the injection sites, and transported tracing substances were plotted in each tissue section. The criterion of Condé was used to assess the effectiveness of the fluorescent injection sites (Condé, 1987). Accordingly, zones 0–II were interpreted as corresponding to the effective uptake site. For immunohistochemically developed material, the effective injection site was defined by the heavily stained area that surrounds the point of tracer administration and extends to the edge of the deposit. The edge of the injection site corresponding to the region where the dense precipitate diminishes, neurones and labelled axons are distinguishable, and the intensity of fibre staining is similar to fibre staining intensity occurring in local, primary projection targets (Fig. 2). Cortical sections were spaced at intervals of 500 μm. Nissl-, myelin- and cytochrome-stained sections were used for plotting architectonic boundaries directly onto the chartings. Brainstem tissue sections were analysed at intervals of 250 μm. Fluorescent labelling (retrogradely labelled motor neurones with FB and DY and corticofacial terminals labelled with FD and FR) and immunohistochemically developed sections (BDA, LYD, PHA-L, FD and FR) were also plotted using the xy plotter. To study in greater detail facial nucleus labelling patterns in immunohistochemically processed sections, axon varicosities in the nucleus were charted a second time, using maximum range conditions of the xy recorder or with a drawing tube. White and grey matter borders were added to the brainstem chartings using Nissl-stained sections. In monkey SDM 20, vibratomed sections through the brainstem were charted for FR terminals. FR labelling in the facial nucleus was verified in this case using a Bio-Rad 1024 confocal microscope (Bio-Rad Laboratories, Hercules, Calif., USA).

The terminal projection from representative cases of each motor area (M1, LPMCv, LPMCd, M2, M3 and M4) was quantified within the facial nucleus using immunohistochemically processed material (Fig. 9). To accomplish this, the total number of labelled boutons was counted in every immunohistochemically reacted serial section through the ipsilateral and contralateral facial nucleus (e.g. at 250 μm intervals through a rostrocaudal distance of 1.7–2.2 mm). Terminal boutons were further categorized by ipsilateral and contralateral nuclear subdivision (medial, lateral, dorsal and intermediate). Terminal boutons, or axon varocosities, represent putative synaptic contacts between a projection axon and local neurones (Wouterlood and Groenewegen, 1985). The total number of boutons in each subdivision of each side was divided by the total number of boutons counted in the entire case. The ratio was converted to a percentage and expressed by subdivision in each side.

Fig. 9

Histograms showing the percentage of boutons found in each subnucleus of each side in representative cases involving each motor area. The data from M1 (SDM19-LYD), LPMCv (SDM22-BDA), LPMCd (SDM23-FD), M2 (SDM19-BDA), M3 (SDM14-BDA) and M4 (SDM22-LYD) is further illustrated in serial cross-sections through the facial nucleus in Figs 10, 11, 12, 13, 14 and 15, respectively.

The cortex was reconstructed using surface drawings developed from metrically calibrated photographs and video images taken of the brain prior to tissue sectioning (Figs 3–6). Nissl-, myelin- and cytochrome-oxidase-stained tissue sections were used to delineate architectonic areas of the frontal lobe and cingulate gyrus. A physiological map of the cortical surface illustrating the locations of the stimulation points, electrolytic lesions, sulci and vessels was constructed from images taken during surgery of each penetration point and surrounding landmark. The stimulation map was superimposed on the injection site map (Figs 3–6). The brainstem sections were reconstructed by stacking each serial section in a caudal to rostral sequence depicting terminal labelling in all cases and motor neuronal labelling in selected cases (Figs 10–15).

Fig. 10

A series of line drawings through the facial nucleus in case SDM 19 following an injection of LYD into the face representation of M1 and injections of retrograde tracers into representative upper and lower facial muscles. Illustrated on the tissue sections are the locations of anterograde and retrograde labelling in each subnucleus.

Fig. 11

A series of line drawings through the facial nucleus in case SDM 22 following an injection of BDA into the face representation of LPMCv and injections of retrograde tracers into representative upper and lower facial muscles. Illustrated on the tissue sections of the facial nucleus are the locations of anterograde and retrograde labelling in each subnucleus.

Fig. 12

A series of line drawings through the facial nucleus (top) and spinal cord (bottom) in case SDM 23 following an injection of FD into the arm representation of LPMCd. Illustrated on the representative tissue sections is anterograde labelling (black dots).

Fig. 13

A series of line drawings through the facial nucleus in case SDM 19 following an injection of BDA into the face representation of M2 and injections of retrograde tracers into representative upper and lower facial muscles. Illustrated on the tissue sections of the facial nucleus are the locations of anterograde and retrograde labelling in each subnucleus.

Fig. 14

A series of line drawings through the facial nucleus in case SDM 14 following an injection of BDA into the face representation of M3 and injections of retrograde tracers into representative upper and lower facial muscles. Illustrated on the tissue sections of the facial nucleus are the locations of anterograde and retrograde labelling in each subnucleus.

Fig. 15

A series of line drawings through the facial nucleus in case SDM 22 following an injection of LYD into the face representation of M4 and injections of retrograde tracers into representative upper and lower facial muscles. Illustrated on the tissue sections of the facial nucleus are the locations of anterograde and retrograde labelling in each subnucleus.

Cytoarchitectonic organization of the facial nucleus

The facial nucleus can be subdivided into several distinct cellular groups or subnuclei (Papez, 1927; Courville, 1966; Jenny and Saper, 1987; Satoda et al., 1987; Baker et al., 1994). In this report, the organization of the monkey facial nucleus was determined according to the criteria of Jenny and Saper (Jenny and Saper, 1987) using Nissl-stained sections to facilitate comparison of our work with previously published results on the corticofacial projection. Thus, the nucleus was subdivided into intermediate, dorsal, medial and lateral subsectors (Fig. 1C, right). Furthermore, patterns of motor neurone labelling following injection of retrograde tracers in the facial muscles and organization of corticofacial terminals following injections into cortex correlated well with the basic designation of intermediate, dorsal, medial and lateral subsectors (Fig. 1C, left). Cell sparse bands, packing, somal diameter and staining intensity were considered for parcellation.

Definition and characterization of cortical face areas

The face area of M1 is located rostral to the ventral extent of the central sulcus and on average has a low physiological threshold for excitability (Fig. 1A and B) (Weinrich and Wise, 1982; Huang et al., 1988; Huntley and Jones, 1991; Murray and Sessle, 1992; Preuss et al., 1996) (Fig. 1B). The caudal portion of the M1 face area corresponds to area 4 in Nissl-stained sections (Brodmann, 1909; Barbas and Pandya, 1987; Preuss and Goldman-Rakic, 1991) and area F1 in cytochrome oxidase preparations (Matelli et al., 1985, 1991). The rostral portion corresponds to the caudal portion of area 6V in Nissl-stained sections (Barbas and Pandya, 1987; Preuss and Goldman-Rakic, 1991) and area F4 in cytochrome-oxidase-stained sections (Matelli et al., 1985, 1991). Immediately anterior and adjacent to the face representation of M1 is the face area of the ventral lateral premotor cortex (LPMCv) (Fig. 1B). This cortex corresponds to the rostral portion of area 6V (Barbas and Pandya, 1987; Preuss and Goldman-Rakic, 1991) as well as area F5 (Matelli et al., 1985, 1991). Physiologically, the face representation of LPMCv has a higher threshold of excitability in unanaesthetized monkeys when compared with the caudal portion of M1 (Rizzolatti et al., 1988; Preuss et al., 1996). Electrophysiological evidence indicates the arm and face representations overlap in the dorsal portion of LPMCv (Gentilucci et al., 1988; Rizzolatti et al., 1988; Godschalk et al., 1995; Preuss et al., 1996). Within this transitional cortex, arm movements are heavily represented, but face movements are occasionally evoked. Similarly, the ventral part of LPMCd is primarily related to arm movements and less so to face movements (Godschalk et al., 1995; Preuss et al., 1996). Histologically, the ventral part of LPMCd corresponds to area 6D in Nissl preparations and area F2 in cytochrome oxidase preparations. This overlap in somatotopy was taken into consideration in our experimental design as we investigated the corticofacial projection from the ventral part of LPMCv as well as the corticofacial projection from the ventral portion of LPMCd. The face representation of M2 lies within the rostral portion of medial area 6 (Mitz and Wise, 1987; McGuire et al., 1991; Preuss and Goldman-Rakic, 1991; Godschalk et al., 1995; Morecraft et al., 1996). This same location corresponds to area F3 using cytochrome oxidase staining (Matelli et al., 1985, 1991; Luppino et al., 1993). The face representation of M2 is relatively smaller than the lateral face areas and has, on average, a higher threshold of movement excitation when compared with the average threshold values of cortex residing in the caudal portion of M1 (Mitz and Wise, 1987; Luppino et al., 1991; Huntley and Jones, 1991). The face area of M2 lies 1–3 mm rostral to a coronal plane established through the genu of the arcuate sulcus (Mitz and Wise, 1987; Matsuzaka and Tanji, 1996; Morecraft et al., 1996). Rostral to area F3 is area F6, which corresponds to the pre-SMA (Matelli et al., 1985; Luppino et al., 1991). The pre-SMA is related primarily to complex arm movements (Alexander and Crutcher, 1990; Matsuzaka and Tanji, 1996) and is much less excitable than adjacent area F3 (Luppino et al., 1991). The cingulate motor cortex lines the depths of the cingulate sulcus and is formed by the rostral and caudal cingulate motor cortices (Biber et al., 1978; Muakkassa and Strick, 1979; Morecraft and Van Hoesen, 1988, 1992; Hutchins et al., 1988; Dum and Strick, 1991; Devinsky et al., 1995; Nimchinsky et al., 1996). The rostral cingulate motor cortex (M3) corresponds cytoarchitecturally to area 24c (Vogt et al., 1987; Dum and Strick, 1991; Morecraft and Van Hoesen, 1992; Morecraft et al., 1996). The face area of M3 is located in the rostral portion of area 24c, in coronal levels that include the genu of the corpus callosum (Muakkassa and Strick, 1979; Morecraft and Van Hoesen, 1992; Morecraft et al., 1996; Tokuno et al., 1997). The genu and its anterior limit were visualized during surgical exposure to assist in localizing the M3 face area. The caudal cingulate motor area (M4) corresponds cytoarchitecturally to area 23c (Vogt et al., 1987; Hutchins et al., 1988; Dum and Strick, 1991; Morecraft and Van Hoesen, 1992; Morecraft et al., 1996). The face representation of M4 is located in the rostral part of area 23c (Morecraft et al., 1996; Tokuno et al., 1997). This general area gives rise to face movements using intracortical microstimulation in the unanaesthetized monkey (Luppino et al., 1991; Godchalk et al., 1995). Anatomically, this location corresponds to coronal levels that include the anterior commissure medially and spur of the arcuate sulcus laterally (Morecraft and Van Hoesen, 1992; Morecraft et al., 1996; Nimchinsky et al., 1996; Tokuno et al., 1997). If the spur is absent, the location of the M4 face area was approximated 1.0–2.0 mm caudal to the genu of the arcuate sulcus.

Results

Retrograde labelling of facial motor neurones

Labelled neurones were found in the facial nucleus in all monkeys receiving injections of FB and DY into the facial muscles (Figs 3–6). All injections of retrograde tracer placed into the frontalis–corrugator muscle complex labelled motor neurones in the dorsal subnucleus (Figs 10, 13 and 14, right). Occasionally, a few labelled cells appeared in the medial edge of the intermediate subnucleus. Labelled neurones occurred in the dorsal region of the medial subnucleus following injections in the orbito-auricularis muscle (Figs 11 and 15, right). Injections in the orbicularis oculi muscle consistently labelled neurones that were confined to the intermediate subnucleus (Figs 10, 11 and 13–15, right). Injections in upper orbicularis oris muscle labelled motor neurones in the dorsolateral sector of the lateral subnucleus (Figs 10, 11 and 13–15, left) and injections in the lower orbicularis oris muscle labelled neurones in the ventrolateral sector of the lateral subnucleus (Figs 10, 11 and 13–15, left).

Location of cortical injection sites

In all cases presented in this report, the injection sites were evaluated for somatotopic affiliation on the basis of multiple criteria. First, the location and peripheral extent of the injection sites made in M1, M2, LPMCv and LPMCd were reconstructed in direct relation to all electrophysiological stimulation points. Secondly, matching Nissl and cytochrome oxidase sections were used to evaluate histological boundaries in each motor area. The face areas of M1, LPMCv and LPMCd correlated to the ventral extent of each respective histological region. The face areas of M2, M3 and M4 correlated to the rostral extent of each respective histological region. Thirdly, corticocortical interconnections amongst the various injection sites in the same animal were evaluated for topographical reciprocity. For example, cortical arm representations are selectively interconnected, as are cortical face representations (Pandya and Vignolo, 1971; Muakkassa and Strick, 1979; Morecraft and Van Hoesen, 1988, 1992; Luppino et al., 1994; Nimchinsky et al., 1996; Tokuno et al., 1997). Fourthly, descending projections from each injection site to the facial nucleus and spinal cord were analysed. Injection site involvement of cortical face areas was supported by the presence of projections to the facial nucleus (Kuypers, 1958a, b; Jenny and Saper, 1987; Morecraft et al., 1996). Injection site involvement of cortical arm areas was supported by the presence of direct projections to the brachial spinal cord (Biber et al., 1978; Hutchins et al., 1988; Dum and Strick, 1991; Galea and Darian-Smith, 1994; Rouiller et al., 1994; Luppino et al., 1994; Morecraft et al., 1997).

Monkey SDM 9.

The BDA injection site in M3 was confined to the face representation as determined by FB tracer transport from M1 to the BDA injection site in M3 (Fig. 2B) [also see Morecraft et al. (1996), Fig. 1]. This was verified by the presence of BDA labelling in the facial nucleus [see Morecraft et al. (1996), Fig. 4]. No BDA labelling was found in the spinal cord, indicating no BDA involvement of the M3 arm representation.

Monkey SDM 10.

The BDA injection site involved the arm and face representations of M4 as determined by the corticocortical projection patterns and the fact that BDA labelling was found in both the facial nucleus and spinal cord (see Morecraft et al., 1996, Figs 3 and 4).

Monkey SDM 14.

The PHA-L injection in M1, FD injection in LPMCv and LYD injection in M2 were all confined to the physiologically defined face representation of each motor area (Fig. 3A). All injections produced terminal labelling in the facial nucleus (Fig. 14) with the exception of PHA-L, which only labelled descending fibres. A few terminals were found in the spinal cord from the LPMCv injection, indicating minimal involvement of the adjacent arm representation. At the cortical level these injection sites were found to be topographically interconnected with each other as well as with the BDA injection in the rostral part of M3 (face area) and FR injection site in the rostral part of M4 (face area). Anatomically, the M3 injection site corresponded to coronal levels including the genu of the corpus callosum. This injection produced terminal labelling in the facial nucleus verifying the involvement of the M3 face area (Fig. 8E). The anatomical location of the M4 injection site corresponded to coronal levels including the arcuate spur as well as the anterior commissure. Terminal labelling was found in the facial nucleus and very light labelling was found in the spinal cord that was largely confined to the region of the spinal accessory nucleus.

Monkey SDM 19.

On the lateral surface of the hemisphere (Fig. 4A, bottom), the LYD and FD injection sites were confined to the physiologically defined face and arm representations of M1, respectively. The LYD injection labelled terminals in the facial nucleus (Figs 8A and 10) and the FD injection labelled terminals in the spinal cord. On the medial wall of the hemisphere (Fig. 4A, top), the FR injection site was confined to the arm representation of M2, which was verified by physiological reconstruction, and its heavy projections to the arm area of M1 and the spinal cord. The BDA injection site was located in the rostral part of M2. This injection gave rise to heavy projections to the face representation of M1 as well as the facial nucleus, indicating primary involvement of the face representation of M2 (Fig. 8C). Light labelling was found in the arm representation of M1 as well as the spinal cord, indicating minimal involvement of the arm representation of M2.

Monkey SDM 20.

The BDA and LYD injection sites were restricted to the physiologically mapped face and arm representations of M1, respectively. Projections from these injection sites were analysed in the cortical region of M3 for the purpose of evaluating the somatotopic affiliation of the FR injection site placed in M3. The FR injection was large by design, involving most of M3 including its face and arm representations. This was supported by the finding that the M1 face injection labelled terminals in the rostral part of M3 (face area of M3) and the M1 arm injection labelled terminals immediately caudal to this field (arm area of M3). This was subsequently verified by the finding of FR labelling in both the facial nucleus and spinal cord.

Monkey SDM 22.

On the lateral surface of the hemisphere (Fig. 5A, bottom), the FD and BDA injection sites were found to be confined to the physiologically mapped face representations in M1 and LPMCv, respectively, and each gave rise to terminal labelling in the facial nucleus (Figs 8B and 11). On the medial surface of the hemisphere (Fig. 5A, top), the PHA-L injection was confined to the face representation of M3 as determined by its anatomical location and projections to the face areas of M1, LPMCv and M2. However, this injection did not label terminals at mesencephalic and pontine levels, only fibres. The LYD injection site in M4 involved the face area and to a minimal extent the arm area (Fig. 5A, top). Its somatotopically affiliated projections to the face areas of M1, LPMCv and M2 as well as its projection to the facial nucleus verified this deduction (Figs 8F and 15). Light labelling in the dorsomedial part of the spinal grey indicated minimal injectate involvement of the arm representation of M4.

Monkey SDM 23.

The FD injection site was confined to the physiologically defined arm representation of LPMCd (Fig. 6). Heavy labelling occurred in the spinal cord (Fig. 12, bottom) and very light labelling was found in the facial nucleus (Fig. 12, top). This indicated overlap between the arm and face representations of the lateral premotor cortices (e.g. intermingling of corticospinal and corticofacial neurones) with arm representation predominating in this location.

Topographical organization of anterograde projections to the facial nucleus

Thirteen anterograde injection sites involving the various face representations were used for analysis. All face representations were found to project to the facial nucleus but major differences were found in the topographical distribution of each projection. In all cases, labelling took the form of concentrated patches and occurred throughout most of the rostral and caudal extent of the facial nucleus (Figs 10–15). The results from individual injection sites described below are identified by monkey (e.g. SDM19) and tract tracing compound (e.g. LYD).

Corticofacial projections from M1

Two injection sites were evaluated (SDM19-LYD and SDM22-FD). Both gave rise to similar and extensive patterns of labelling in the facial nucleus. The overall projection was characterized by a contralateral predominance, where the primary target was the contralateral lateral subnucleus. Terminal boutons were counted in each nuclear subsector in injection SDM19-LYD (Figs 4, 8A, 9 and 10). In this case, 3065 boutons were found bilaterally. The most extensive distribution of labelled boutons occupied the contralateral lateral subnucleus (55.38%). This took the form of dense patches of label in both the dorsal and ventral portions of the lateral subnucleus, where motor neurones innervating the upper and lower perioral musculature, respectively, were located (Fig. 10). Comparatively lighter and patchy labelling occupied the ipsilateral lateral subnucleus (15.09%). Sparse terminal labelling was also found in the intermediate (ipsilateral, 5.62%; contralateral, 9.21%), medial (ipsilateral, 6.41%; contralateral, 4.29%) and dorsal (ipsilateral, 1.41%; contralateral, 2.59%) subnuclei.

Corticofacial projections from LPMCv (area 6V)

Two injection sites were studied (SDM14-FD and SDM22-BDA) (Figs 3, 5, 8B, 9 and 11). A total number of 3731 boutons were counted bilaterally in injection SDM22-BDA. Much like the M1 projection, the LPMCv projection was characterized by a contralateral predominance, with the primary target being the contralateral lateral subnucleus (42.97%). Surprisingly, a substantial number of labelled boutons occurred in the ipsilateral lateral subnucleus (29.64%), suggesting a significant influence of LPMCv projection on the ipsilateral lower facial musculature. In both lateral subnuclei, labelling was densely distributed in the dorsal and ventral regions where lower motor neurones innervating the upper and lower perioral musculature were localized (Fig. 11). Much lighter axonal labelling occurred in the remaining subnuclei with a slight predominance in the contralateral projection (ipsilateral intermediate, 2.09%; contralateral intermediate, 7.64%; ipsilateral dorsal, 1.39%; contralateral dorsal, 3.16%; ipsilateral medial, 4.37%; contralateral medial, 8.74%).

Corticofacial projections from LPMCd (area 6D)

The LPMCd injection site (SDM23-FD) gave rise to the weakest projection to the facial nucleus, totalling 264 boutons (Figs 6, 9 and 12). Labelling was diffuse with large label-free zones located between small clusters or patches of terminal fields (Fig. 12). The projection was primarily to the contralateral lateral subnucleus (28.02%) with a moderate projection ending in the ipsilateral lateral subnucleus (20.88%). A relatively moderate projection, similar in intensity to the ipsilateral projection to the lateral subnucleus, occurred in the contralateral medial subnucleus (15.33%). Few boutons were found in the ipsilateral medial subnucleus (5.55%). The dorsal and intermediate subnuclei also received a comparatively minor projection that was slightly more contralateral in its intensity (ipsilateral dorsal, 8.79%; contralateral dorsal, 10.99%; ipsilateral intermediate, 4.40%; contralateral intermediate, 6.04%).

Corticofacial projections from M2 (area 6m)

Three injection sites involving the face representation of M2 gave rise to similar patterns of terminal labelling in the facial nucleus (SDM14-LYD, SDM19-BDA and SDM22-FR) (Figs 3–5, 8C and D and 13). All the injection sites in M2 produced dense terminal labelling in the medial subnucleus, where lower motor neurones innervating the orbito-auricular muscle were localized in our study (Fig. 13), and the posterior auricular muscle in other studies (Jenny and Saper, 1987; Welt and Abbs, 1990). The injection in SDM19-BDA was the largest by design, producing 2738 boutons and was selected as the representative M2 injection site. In this case, axon varicosities were concentrated bilaterally in the medial subnucleus (ipsilateral, 30.86%; contralateral, 29.52%). A lighter, but also bilateral projection occurred in the lateral (ipsilateral, 7.41%; contralateral, 9.57%), dorsal (ipsilateral, 6.17%; contralateral, 5.51%) and intermediate (ipsilateral, 5.59%; contralateral, 5.37%) subnuclei.

Corticofacial projections from M3 (area 24c)

Three injections involving the face representation of M3 were studied (SDM9-BDA, SDM14-BDA and SDM20-FR) (Fig. 4). All cases gave rise to a general pattern of labelling that was dispersed throughout the dorsomedial region of the facial nucleus. Specifically, this projection was mainly located in the intermediate and dorsal subnuclei, where lower motor neurones innervating the upper facial muscles were located and the medial portion of the medial subnucleus where auricular muscles were represented (Figs 8E and 14). Overall, this projection was characterized as being predominantly bilateral, with slightly more ipsilateral labelling distributed within the intermediate subnucleus in some cases. There was also a moderate gradient in the overall density of terminal labelling, where the boutons were slightly more numerous in the intermediate subnucleus. Comparatively fewer labelled profiles were found in the dorsal subnuclei and fewer boutons were localized in the medial and lateral subnuclei. For example, a total of 759 boutons were counted in the corticofacial projection in injection SDM14-BDA. The majority of boutons were observed in the ipsilateral intermediate subnucleus (24.82%) with fewer boutons found in the contralateral intermediate subnucleus (21.20%). A moderate number of boutons were found in the dorsal subnucleus (ipsilateral, 11.19%; contralateral, 16.20%). Progressively fewer boutons occurred in the medial (ipsilateral, 9.48%; contralateral, 8.16%) and lateral (ipsilateral, 4.74%; contralateral, 4.21%) subnuclei.

Corticofacial projections from M4 (area 23c)

Three injection sites involving the face representation of M4 were analysed for corticofacial projections (SDM10-BDA, SDM14-FR and SDM22-LYD) (Figs 3 and 5). All injections in the face representation of M4 produced similar patterns of axonal label in the facial nucleus that were concentrated in the dorsal portion of the lateral subnucleus (Figs 8F and 15). This location contained labelled motor neurones following muscular injections in the upper portion of the orbicularis oris. A total of 638 boutons were found bilaterally in case SDM22-LYD. Of these, the largest number of labelled boutons was concentrated in the contralateral lateral subnucleus (43.86%), with significantly fewer labelled terminal profiles occurring in the ipsilateral lateral subnucleus (3.29%). The remaining projections were comparatively lighter and predominantly contralateral in the intermediate (ipsilateral, 2.19%; contralateral, 14.89%), dorsal (ipsilateral, 3.45%; contralateral, 12.70%) and medial (ipsilateral, 8.46%; contralateral, 11.16%) subnuclei.

Discussion

The contemporary viewpoint on cortical innervation of the facial nucleus has deep roots in neurological history dating back for more than a century. Along with the corticospinal projection, the corticofacial projection is often used as a classical example for teaching the basic principles of lesion localization as well as diagnosing the upper motor neurone syndrome. The premise of our existing knowledge of this projection is based upon the interpretation of upper motor neurone facial palsy following the most common cortical stroke. Specifically, this view is formed in the context of unilateral destruction of M1 following middle cerebral artery (MCA) occlusion and the suggestion that each M1 supplies upper facial motor neurones with a significant bilateral projection and lower facial motor neurones with only a contralateral projection. Presumably, the bilateral projection is responsible for upper face sparing and the contralateral projection underlying the deficit. Results from the present study do not fully support this interpretation but, importantly, provide an alternative explanation for the common clinical occurrence of upper face sparing following MCA infarction. This report also advances new information that may assist in interpreting facial deficits following midline brain injury. Furthermore, the present observations shed new light on multiple supranuclear projection systems that may participate in functional recovery of facial movement following localized cortical damage. Finally, our results are discussed and interpreted in the context of localization of emotional and volitional facial paresis.

Cortical innervation of the facial nucleus

The corticofacial projection from M1

Following injections of anterograde tracer into the face region of M1 we found labelling in all subdivisions of the facial nucleus that would be in agreement with previous observations in the monkey (Kuypers, 1958b; Jenny and Saper, 1987) (Figs 9 and 10). Also in agreement with these reports was our finding of a substantial projection from M1 to the contralateral lateral subnucleus, which innervates the lower facial musculature (Fig. 16). The heavy labelling found in the lateral subnucleus and comparative light labelling found in the remaining subnuclei correlate with the common motor responses evoked in the contralateral lower facial muscles and weak, but occasionally notable, activity in the upper facial muscles following intracortical microstimulation in monkey (McGuinness et al., 1980; Sirisko and Sessle, 1983; Huang et al., 1988; Murray and Sessle, 1992). Similarly, this would correspond to the movement patterns elicited using epicortical and magnetic transcranial stimulation of M1 in humans (Penfield and Welch, 1951; Woolsey et al., 1979; Cohen and Hallett, 1988; Cruccu et al., 1990; Cocito et al., 1993; Urban et al., 1997). The weak but consistent projection to the intermediate and dorsal subnuclei may underlie the mild weakness in upper facial movement occasionally reported in the literature (Mills, 1889a; Monrad-Krohn, 1924; Kojima et al., 1997). Our observations support the classical observation of contralateral lower facial paralysis following damage to M1. However, in agreement with Jenny and Saper (1987), our results do not suggest that sparing of the upper facial muscles is due to significant bilateral projections to upper facial lower motor neurones arising from M1 located in the undamaged hemisphere.

Fig. 16

Summary diagram illustrating cortical face areas giving rise to contralateral innervation of the lateral subnucleus. The lateral subnucleus, in turn, was found to innervate the lower facial musculature. These observations support the classic teaching that middle cerebral artery occlusion that involves M1 and or LPMCv gives rise to paralysis of the contralateral lower facial musculature.

The corticofacial projection from LPMCv and LPMCd

The corticofacial projection from LPMCv and LPMCd was found to be primarily contralateral and target the lateral subnucleus (Figs 9, 11, 12 and 16). This would correlate with the limited physiological observations indicating that mouth movements are primarily represented in the ventral lateral premotor region of macaque monkeys (Gentilucci et al., 1988). The finding of a relatively moderate ipsilateral projection from LPMCv to the lateral subnucleus suggests a potentially significant ipsilateral influence on lower facial motor neurones from the lateral surface of the hemisphere. These observations indicate that the overall extent of a cortical lesion affecting the lateral convexity may correlate with the degree of facial paralysis, on the one hand, and the potential for motor recovery on the other. For example, our results suggest that a large unilateral lesion damaging both M1 and the adjacent LPMCv are likely to give rise to severe deficits in the contralateral lower facial musculature with a poor prognostic outcome in terms of motor recovery. In contrast, a more limited lesion leaving one of these areas intact may yield milder deficits in the contralateral lower face and correlate with greater potential for motor recovery. Our findings also suggest that recovery in the contralateral lower facial muscles following unilateral destruction of M1 and or LPMCv may be facilitated through the intact contralateral projection from M4 to the lateral subnucleus. Furthermore, the intact ipsilateral projection to the lateral subnucleus from LPMCv in the undamaged hemisphere may play an important role in the recovery process that follows MCA infarction.

We found a greater number of labelled boutons in the facial nucleus following injections in LPMCv (SDM22-BDA) than in M1 (SDM19-LYD), which was not anticipated based upon what is known about the relative density of the corticospinal projection from M1 and LPMCv (Figs 10 and 11). Although it is possible that the LPMCv projection may be slightly heavier, this result may be attributable to several experimental factors. It is possible that the rate of uptake and transport of BDA and LYD are somewhat different. Indeed, caution must be exercised when comparing the intensity of two projection systems determined with two different neuronal tract tracers. However, we consistently found robust labelling in many primary subcortical targets such as the pontine nuclei and medullary reticular formation in all of our experiments using both of these dextran tracers as well as with FR and FD. Another potential factor may be related to the finding that the LPMCv injection site was slightly larger in the dorsal and ventral dimension than both M1 injection sites studied. Thus, the LPMCv injection site may have involved a greater number of corticofacial projection neurones. Importantly, the facial nucleus in case SDM22-BDA was larger along the longitudinal axis (e.g. superior–inferior dimension) than the facial nucleus in both M1 cases. This resulted in an additional tissue section to examine in case SDM22 (e.g. note seven facial nucleus levels in Fig. 10 and eight levels in Fig. 11). Also adding to the total number of labelled boutons in the LPMCv case was the comparatively significant ipsilateral projection to the lateral nucleus. Indeed, future experiments are needed to address the relative strength of each projection using alternative methodologies. Notwithstanding, our findings indicate that the corticofacial projection from LPMCv appears to be quite important anatomically and possibly functionally.

The corticofacial projection from M2

We found the M2 projection to the facial nucleus to be moderate in intensity and to target preferentially and bilaterally the medial subnucleus (Figs 9, 13 and 17). The medial subnucleus, in turn, was found to contain lower motor neurones supplying the auricular musculature. Possibly related to this finding are observations in monkeys demonstrating that neurones in M2 are responsive to auditory stimuli in addition to other forms of sensory cues (Kurata and Tanji, 1985). Our finding of a corticofacial projection from M2 is in disagreement with an earlier study (Jürgens, 1984) that did not localize a facial nucleus projection from M2. This discrepancy could be due to the possibility that the injection in Jürgens' study did not involve the face representation of M2 and the highly sensitive immunohistochemical detection method used in our study. The potential clinical implications of the preferential projection to auricular lower motor neurones are unclear for several reasons. First, human auricular musculature is not as developed and specialized as demonstrated in the rhesus monkey (Huber, 1933; Williams et al., 1989). Secondly, a clear or commonly noted facial deficit does not appear to accompany localized M2 lesions. For example, an absence of cranial nerve abnormalities in patients with localized M2 lesions has been reported previously (Meador et al., 1986; Chan and Ross, 1988). Thirdly, pathologically confirmed lesions affecting the superior frontal lobule (M2) that produce facial deficits also involve adjacent portions of the cingulate cortex. Under these circumstances patients have been characterized as having a blunted facial expression and weakness in the lower face, which is particularly notable when smiling (Laplane et al., 1977; Masdeu et al., 1978). The above, in conjunction with our observations, suggests a careful re-evaluation of the potential role of M2 in mediating facial expression.

Fig. 17

Summary diagram illustrating cortical face areas giving rise to bilateral innervation of the intermediate, dorsal and medial subnuclei. These subnuclei were all found to innervate upper facial musculature. Our findings suggest that sparing of the upper facial musculature following middle cerebral artery infarction is due to sparing of the projection from M2 and M3, which reside in the territory of the anterior cerebral artery. The present findings and the fact that the rostral cingulate motor cortex (M3) receives widespread limbic and prefrontal inputs (Morecraft and Van Hoesen, 1993, 1998; Morecraft et al., 1998, 2000), and is itself considered one of the higher-order motor areas in the cortex (Shima et al., 1991; Shima and Tanji, 1998), supports the view that M3 may be involved in mediating upper facial expression of higher-order emotions (Damasio, 1994). FN, facial nucleus.

The corticofacial projection from M3 and M4

Our findings support previous observations of a direct projection from M3 and M4 to the facial nucleus (Morecraft et al., 1996) and extend this finding by demonstrating the topographical organization of each projection. In the present study, we found the projections from M3 and M4 to be very different in terms of their primary subnuclear target as well as their mode of laterality. The corticofacial projection from M3 ended bilaterally, within the dorsal and intermediate subnuclei, which were found to innervate the orbicularis oculi and frontalis–corrugator muscular complex, respectively (Figs 9, 14 and 17). This complex consists of anatomically interdigitating, histochemically unique and functionally complementary circumorbital musculature (Freilinger et al., 1990; Goodmurphy and Ovalle, 1999). The contralateral portion of the projection found in our study would, in part, support Damasio's observation of weakness occurring in the contralateral orbicularis oculi in patients with unilateral anterior cingulate damage (Damasio, 1994). The fact that the projection in the monkey was bilateral raises the possibility that there may be species differences in the laterality of the projection to the upper face.

The difficulty in interpreting deficits in the upper facial musculature following anterior cingulate damage may, in part, be associated with the potential bilateral nature of this projection found in our study and the fact that facial paralysis in patients is commonly assessed on the basis of asymmetry. For example, in applying our non-human primate data to the human, it suggests that unilateral destruction would yield deficits similar in magnitude in the ipsilateral and contralateral orbicularis oculi and frontalis–corrugator musculature. Presumably, the bilateral M3 projection to the upper face from the intact hemisphere may mask significant facial deficits and contribute to rapid compensatory mechanisms underlying motor recovery.

We have emphasized the lack of experimental support affiliated with the classic interpretation of upper face sparing following superficial MCA infarctions. Since the origin of the M3 projection described in our study is located within the vascular territory of the anterior cerebral artery, we suggest that bilateral sparing of the upper face following superficial MCA occlusion is due, in part, to sparing of anterior cingulate projection to the dorsal and medial regions of the facial nucleus.

The topography of the corticofacial projection from M4 was very similar to the general distribution found in experiments conducted on M1, LPMCv and LPMCd as it preferentially targeted the contralateral subnucleus (Figs 9, 15 and 16). Furthermore, the projection from M4 was highly specific in that it ended within the dorsolateral sector of the lateral subnucleus, which consistently contained motor neurones innervating the upper, but not the lower lip. Theoretically, this observation indicates that damage to the human caudal cingulate motor region may impair elevation of the contralateral upper lip. This finding may underlie the clinical observations of Laplane and co-workers who carefully noted contralateral lower facial weakness only when smiling, following unilateral resection of M2 and the midportion of the adjoining cingulate cortex (Laplane, 1977).

Musculotopic organization of the facial nucleus

The topographic representation of the peripheral branches of the facial nerve in the macaque facial motor nucleus have been reported in several previous studies (Jenny and Saper, 1987; Satoda et al., 1987; Porter et al., 1989; Welt and Abbs, 1990; Baker et al., 1994; VanderWerf et al., 1998). The present study was not intended to re-examine the musculotopic organization of the primate facial nucleus but to include intramuscular injections in our design to strengthen our deductions regarding topographical association formed between each descending corticofacial projection and lower motor neurone organization. However, the general patterns of retrogradedly labelled motor neurones found in our study are consistent with most of the reported observations in the macaque monkey. For example, injections of fluorescent dyes into the orbicularis oculi consistently labelled motor neurones in the intermediate subnuclei (Figs 10, 11 and 13–15). This observation would be in agreement with several recent studies (Porter et al., 1989; Welt and Abbs, 1990; Baker et al., 1994; Huffman et al., 1996; VanderWerf et al., 1998). However, we did not find labelled motor neurones in the contralateral intermediate subnucleus following unilateral injections in the orbicularis oculi as reported by Porter and co-workers (1989). Injections located in the frontalis muscle labelled neurones primarily in the dorsal subnucleus, with few labelled neurones occurring in the adjacent portion of the intermediate subnucleus (Figs 10, 13 and 14). This finding would be in agreement with both Jenny and Saper (1987) and Welt and Abbs (1990). However, we were unable to find labelled neurones in the dorsal part of the lateral subnucleus as reported by Jenny and Saper (1987). We found the motor neuronal population devoted to upper lip innervation to be located in the dorsal portion of the lateral subnucleus and lower lip innervation to be located in the ventral portion of the lateral subnucleus (Figs 10, 11 and 13–15). This finding would parallel the topographical observations of perioral innervation demonstrated by Welt and Abbs (Welt and Abbs, 1990).

Face representation in the cerebral cortex

Multiple face areas have been localized in the human cerebral cortex that appear to have homologous counterparts in the monkey cortex (Mills, 1889b; Penfield and Welch, 1951; Woolsey et al., 1952, 1979; Muakkassa and Strick, 1979; Morecraft et al., 1996; Picard and Strick, 1996). For example, non-human primate face areas have been identified in M1, LPMCv, M2 and M4 using intracortical microstimulation (Gentilucci et al., 1988; Godschalk et al., 1995; Preuss et al., 1996). Patterns of corticocortical interconnections have also been supportive by demonstrating that the face areas of M1, LPMCv, M2, M3 and M4 are selectively interconnected at the cortical level (Muakkassa and Strick, 1979; Morecraft and Van Hoesen, 1992; Morecraft et al., 1996; Tokuno et al., 1997). Our results reinforce and extend the concept of cortical face representation by demonstrating that each major face area sends direct corticobulbar projections to the facial nucleus. Furthermore, our observations suggest that each cortical face representation may, in part, function independently from M1 by preferentially influencing different facial movements. However, it is important to recognize that integrated facial movements may be heavily mediated at the cortical level through reciprocal corticocortical projections forming interconnecting matrices amongst all the face areas studied.

The various patterns of descending projections to the facial nucleus and spinal cord from LPMCv and LPMCd suggest that some overlap occurs between face and arm representations in the lateral premotor region. For example, injections located in the ventral portion of LPMCv gave rise to a heavy projection to the facial nucleus but also a small projection to the spinal cord (Fig. 11). This observation indicates that corticofacial and corticospinal neurones are probably intermingled in the LPMCv, with corticofacial projections predominating. We found the inverse relationship to occur when investigating the descending projection from LPMCd. For example, an injection into LPMCd gave rise to a heavy corticospinal projection and a light corticofacial projection (Fig. 12). In contrast to both projection patterns of the lateral premotor cortex (LPMC), we failed to find corticospinal projections following tracer injection in the face area of M1, only corticofacial projections. Therefore, it appears that the anatomical and functional border between the face and arm representations in LPMC is less distinct than that found in M1. These observations parallel the intermingled somatotopy of lateral area 6 found using intracortical microstimulation (Gentilucci et al., 1988; Rizzolatti et al., 1988; Godschalk et al., 1995) and lack of intermingled somatotopy characterizing the ventral part of M1 using the same experimental approach (McGuinness et al., 1980; Sirisko and Sessle, 1983; Huang et al., 1988; Huntley and Jones, 1991). The integrated somatotopy in the lateral premotor region may, in part, form a basis for unit recordings suggesting this area plays an important role in mediating combined movements of the hand and mouth (Rizzolatti et al., 1988).

Clinical dissociation of emotional and voluntary facial movements

Formulated largely on the basis of clinical correlates is the suggestion that separate neural systems exist for mediating voluntary and emotional facial movements (Darwin, 1872; Jackson, 1884; Wilson, 1924; Monrad-Krohn, 1924, 1939; Allen, 1931; Karnosh, 1945; Podolsky, 1945; Brown, 1967; Brodal, 1981; Rinn, 1984; Duchenne, 1990; Weddell, 1990; Arroyo et al., 1993; Weller, 1993; Damasio, 1994, 1995; Topper et al., 1995; McConachie and King, 1997; Ross and Mathiesen, 1998; Urban et al., 1998; Hopf et al., 2000). Although our comments related to this issue must be tempered since our findings reflect only structural patterns, they may contribute key pieces of information to develop new and testable hypotheses designed to advance our understanding of this clinically derived tenet. Of particular significance was our finding of a bilateral projection from M3 to the upper facial musculature and a contralateral projection from M4 to the lower facial musculature. These diverse systems may represent important descending projections contributing to a complex neural network subserving emotional expression of facial movement. For example, the origin of the corticofacial projection from M3 and M4 is located in the dorsal edge of the cingulate cortex. The cingulate cortex, in turn, has long been viewed as a critical component of the `emotive' limbic lobe (Broca, 1878; Papez, 1927; MacLean, 1954; Baleydier and Maugiere, 1980; Damasio and Van Hoesen, 1983; Mesulam, 1988; Pardo et al., 1993; Devinsky et al., 1995; George et al., 1995; Ketter et al., 1996; Van Hoesen et al., 1996) and both cingulate motor cortices receive powerful and widespread prefrontal and limbic lobe inputs (Bates and Goldman-Rakic, 1993; Morecraft and Van Hoesen, 1993; Lu et al., 1994; Morecraft and Van Hoesen, 1998; Morecraft et al., 2000). Furthermore, the amygdala, which plays a critical role in regulating emotional processes (Kling and Brothers, 1992; Damasio, 1994, 1995; Adolphs et al., 1995; LeDoux, 1996; Rolls, 1999) projects not only to the cingulate motor field (Amaral and Price, 1984) but to its somatotopic representations in a highly specific manner (Morecraft et al., 1998). Specifically, we have found that the amygdalo-cingulate motor projection targets primarily the face area of M3 and to a much lesser extent the remaining somatotopic representations in M3 and M4 (Morecraft et al., 1998). Therefore, there is good ground to consider the possibility that blunted emotional expression in the upper facial musculature accompanying anterior cingulate damage (Critchley, 1930; Laplane et al., 1981; Damasio, 1994) may, in part, be a result of disconnecting the limbic lobe and amygdala projections to the rostral cingulate motor area which, in turn, projects bilaterally to the dorsomedial region of the facial nucleus. It is also possible that flattened facial expression in patients with anterior cingulate damage is a general consequence of disruption of the cingulate motor cortex, which forms a pivotal interface between the limbic lobe and many cortical and subcortical motor targets such as the isocortical motor fields, basal ganglia and facial nucleus as reported here (Van Hoesen et al., 1993; Morecraft and Van Hoesen, 1998). The above would, in part, lend structural support to the behavioural view suggesting that the anterior cingulate controls emotion-related movements in the face (Damasio, 1994). Finally, functional imaging data as well as observations from naturally occurring and surgically guided lesions affecting the anterior cingulate region have demonstrated that the anterior cingulate cortex encodes pain and perceived unpleasantness (Foltz and White, 1962; Hurt and Ballantine, 1974; Davis et al., 1997; Rainville et al., 1997; Lenz et al., 1998; Cohen et al., 1999). Our observations suggest that bilateral contraction of the upper face, a hallmark response directly associated with the perception of pain in man (Darwin, 1872) may in part be mediated by descending corticofacial projections emanating directly from the rostral cingulate motor cortex.

Our observations indicate that the corticofacial projection from M4 to the dorsolateral portion of the lateral subnucleus may be associated with the classical observation of preserved reflex smiling in the presence of contralateral paralysis in the lower facial musculature following damage to the lateral face areas (M1 and LPMCv). As shown in our study and in agreement with others, the dorsolateral region of the lateral subnucleus contains lower motor neurones supplying the upper perioral musculature. More speculative, but possibly associated, are the observations from patients with localized cingulate seizures suggesting that the cingulate cortex plays an important role in regulating the expression of laughter (Arroyo et al., 1993; McConachie and King, 1997).

Although our series of experiments identifies several new upper motor neurone pathways that may be involved in mediating volitional and emotional facial responses, it is important to recognize that structures interconnected with the origin of each corticobulbar projection, such as the thalamus, insula and temporal lobe, are likely to contribute as well to the appropriate execution of volitional and emotional facial movement. This supposition is based upon the fact that each upper motor neurone projection field must be activated or influenced by neural inputs from their `extended' projection system. In support of this would be the observation that emotional facial paralysis occurs following damage to the insula, thalamus and striatocapsular region (Hopf et al., 1992). Therefore, the projections described in our report must be considered in the context of potential large-scale neural systems mediating complex behaviours (Pandya and Kuypers, 1969; Jones and Powell, 1970; Mesulam and Greswind, 1978; Goldman-Rakic, 1988; Gloor, 1990; Mesulam, 1990, 1998; Morecraft et al., 1993).

Study limitations

It is important to note that our deductions regarding cortical innervation of the facial musculature are based upon the topographical association formed between anterogradely labelled boutons and retrogradely labelled groups of facial motor neurones. Thus, our study does not reveal whether monosynaptic contacts directly link these projection systems. However, our findings provide critical baseline data to initiate experiments designed to examine the microcircuitry of these important neural systems.

Our predictions regarding the potential clinical implications of our data have been developed on the theoretical basis of cortical damage that is restricted to the origin of each corticobulbar projection. However, it is important to recognize the possibility that subcortical lesions are likely to disrupt the various descending projection systems as they pass inferiorly to reach the facial nucleus. We are aware that this issue is particularly critical when considering the fact that each pathway courses through different regions of the corona radiata and internal capsule (unpublished observations), where each pathway is selectively vulnerable to different forms of localized insult. Due to the extensive scope of this data and its potential clinical correlates, this material will be described in a subsequent report.

Finally, although we have identified several new pathways that interconnect the cerebral cortex with specific parts of the facial nucleus controlling upper facial musculature, it is important to recognize that numerous subcortical pathways are likely to contribute as well to mediating activity in the upper facial musculature. These would include the retrorubral, medullary reticular, spinal trigeminal, parvocellular reticular, lateral tegmental and nucleus ambiguus projections to the dorsal and medial subregions of the facial nucleus (Holstege et al., 1977, 1986; Isokawa-Akesson and Komisaruk, 1987).

Conclusions

The facial nucleus received input from all face representations studied underscoring their recognition as cortical face representations and potential roles in regulating facial movement. M1 and LPMCv gave rise to the heaviest corticofacial projection. In comparison, M2 gave rise to a moderate projection, whereas the projection from M3 and M4 was lighter. Evidence for a weak corticofacial projection from LPMCd was found, but the corticospinal projection was more prominent from this region of cortex. Injections in all cortical face representations labelled terminals in all nuclear subdivisions (dorsal, intermediate, medial and lateral). However, significant differences occurred in the proportion of labelled boutons found within each musculotopically characterized subdivision. M1, LPMC and M4 projected primarily to the contralateral lateral subnucleus, which innervated the lower facial musculature (Fig. 16). M2 projected bilaterally to the medial subnucleus which supplied the orbito-auricularis as determined in our study (Fig. 17) and the posterior auricularis musculature as reported by others (Jenny and Saper, 1987; Satoda et al., 1987; Welt and Abbs, 1990). M3 projected bilaterally to the dorsal and intermediate subnuclei, which innervated the frontalis and orbicularis oculi muscles, respectively (Fig. 17).

Potential disturbances in the projection patterns found in the present study could take the form of highly predictable deficits in facial expression in the event that selective insult disrupts the origin or descending axons of each corticofacial projection. Our results suggest that unilateral damage to the lateral motor cortices (LPMC and M1) is likely to lead to deficits in the contralateral lower facial muscles as classically interpreted. Recovery may occur, in part, from the contralateral projection from M4 within the damaged hemisphere as well as the moderate ipsilateral projection from LPMCv within the undamaged hemisphere. Our findings also suggest that upper facial sparing following lateral cortical damage is due to sparing of the bilateral projection from M2 and M3, which reside medially and originate from the cortical territory supplied by the anterior cerebral artery. Disturbances affecting the reported circuitry in the form of excessive stimulation, which may occur from inappropriate sensory stimulation, or in the absence of an appropriate inhibitory influence might yield excessive muscle contractions in isolated groups of muscles, such as that which characterizes the various forms of cranial–cervical dystonia.

Acknowledgments

This work was supported by NIH grants NS 33003 and NS 36397, the South Dakota Health Research Foundation and a grant from the Benign Essential Blepharospasm Research Foundation (R.J.M.).

References

View Abstract