1 Anatomy

A complete list of structures involved in movement is very extensive and includes the motor cortex, the premotor cortices cortico-spinal tracts, cerebellum, many of the ‘basal ganglia’, various brainstem nuclei and their spinal connections, plus the lower motor neuron. In addition, sensory input is required at many levels in the nervous system to modulate motor activity appropriate to the changing environment. ‘A motor system without sensory control is a fiction, not even a useful fiction’ (Jung and Hassler 1960). An inclusive review of all structures involved in control of movement would cover much of the whole topic of neuro-anatomy. The present chapter is restricted to those that are affected in the diseases grouped under the heading of ‘movement disorders’. Thus it concentrates on the basal ganglia and their connections plus the cerebellum and part of the cerebral cortex.

Parts of the basal ganglia form a major portion of the ‘extrapyramidal system’ of clinical neurology. There are good reasons for abandoning this term as such a grouping is an anatomical abstraction. These structures, however, do have common functions and their groupings have considerable relevance, although this term does not imply that they are a single unit. The term is thus unlikely to disappear from clinical neurology in the foreseeable future and it has been used here to a limited extent.

The extrapyramidal system is impossible to define precisely. Taken literally it means parts of the nervous system not included in the pyramidal or cortico-spinal system. By convention, sensory pathways and neurons are excluded. Most authors also exclude the cerebellum with its afferent and efferent projections. Thus, the motor system is considered to consist of pyramidal, extrapyramidal, and cerebellar subsystems. This unsatisfactory view of the extrapyramidal system is based on exclusions. Structures involved include neostriatum, globus pallidus, subthalamic nucleus, substantia nigra, parts of the thalamus, and pathways linking these structures, including connections with cerebral cortex.

The anatomy of the nervous system can be considered on several different levels. First, there is the gross anatomy of major brain structures and their spatial interrelationships. Second, there is the microscopy and cellular ultrastructure of these areas. Third, there are the interconnecting pathways between these regions. The question of function arises out of such considerations and overlaps with neurophysiology, biochemistry, and pharmacology. Generally speaking, the degree of certainty of knowledge decreases as we pass from gross anatomy through ultrastructure to interconnections and function. It is tempting to attach importance to those structures and connections which are visually obvious. There are, however, pathways that may appear insignificant but which have great functional importance. An example is the nigro-striatal system which carries dopamine between the substantia nigra and the striatum. This fine fibred pathway is not apparent on ordinary light microscopy and its demonstration had to await the development of fluorescent histochemistry (Carlsson et al. 1962, Falck 1962). There are undoubtedly many important interconnections yet to be discovered which may radically alter present concepts of motor function.

When it comes to considering possible axonal interconnections between structures there are also varying degrees of certainty. The weakest evidence of a connection is a hypothetical relationship because of similar functions. The demonstration of biochemical alteration in one area following manipulation in another is more positive evidence, as is an electrophysiological change following stimulation at a distant site. Such methods, however, are liable to misinterpretation due to involvement of neighbouring structures or axons of passage. The possibility of interneurons mediating these effects has to be borne in mind. The most definite evidence of interconnection is direct visualization of the interconnecting pathways. Though traditional histological stains have revealed much detail they are relatively insensitive. Degeneration in tracts following natural or experimental brain lesions has enabled the study of many major myelinated pathways.

Cells and axons which contain certain neurotransmitters, such as catecholamines, have been visualized with special biochemical techniques. In addition, a range of specialized techniques have allowed the anterograde or retrograde labelling of axons, neurons, and dendrites. Thus horseradish peroxidase, cholera toxin, radioactive tracers, fluorescent dyes, and the like can be injected into very discrete brain regions to be transported retrogradely to neurons whose axons terminate in that area, or be transported anterogradely from soma at the injection site to the termination of their processes. Herpes simplex virus has also been used in this way. These techniques have taken some pathways from the realms of conjecture to certainty, and have demonstrated many connections that were previously unsuspected. The use of multiple labelling techniques, in which different axon termination sites are injected with different materials, allows multiple substances to be detected in a single neuron, and has greatly increased our knowledge of the collateral branching of axons. It has thus been shown that many neurons innervate widely different sites.

Techniques that involve transport of material along living neuronal processes are not suitable for human studies. Studies in man are restricted to traditional methods using autopsy material from normal brains, neurological diseases, and degeneration after spontaneous or therapeutic brain lesions. Much recent knowledge of pathways within the brain is derived from animal experiments, and interspecies variation makes extrapolation to man uncertain. There has, however, been shown to be reasonable conformity between different mammalian species and differences tend to be those of emphasis rather than direction. Research continues to reveal a baffling array of interconnections. In order to make this intelligible, emphasis has to be given to some at the expense of others. In this chapter on anatomy, emphasis has been determined by several factors. These include structures and connections which are large and prominent, pathways that are presumed to have major functional importance, those that are established with a reasonable degree of certainty, and those that have most relevance to man. With regard to this latter aspect, human material is used where available and primate studies take precedence over those of other mammals.

In this field, knowledge of structure greatly exceeds that of function. Even the purpose of such large masses as the claustrum remains quite uncertain, while the recently discovered myriad of interconnections between smaller structures presents a baffling puzzle. Under such circumstances the following account may overemphasize the importance of some connections, while not giving others their proper place. There remain undiscovered many others which are, no doubt, of considerable functional importance. With these reservations, the following account outlines the structural basis on which the ‘movement disorders’ occur.

The ‘basal ganglia’ refers to groups of cells or nuclei which lie at the base of the cerebral hemispheres. Unfortunately the term has no precise meaning and different authors have included different structures in this category. Some have allowed only telencephalic masses, while others have included diencephalic or even mesencephalic nuclei. Thus the thalamus, substantia nigra, and red nucleus have been variably included. ‘Basal ganglia’ refers to an anatomical juxtaposition, which is not always relevant in terms of physiology. As the function of these nuclei has become better understood, there has been a tendency to bend the definition to embrace structures functionally connected to the corpus striatum and exclude those less definitely associated. Thus, some have excluded the amygdala, which is more closely related to the limbic system. This review includes a number of structures that could not be considered as part of the basal ganglia even under a widened definition, but, because of their relevant connections, are functionally related. Conversely, structures, such as some of the thalamic nuclei, which are less directly related to movement disorders, are excluded from consideration. This dichotomy between structural relationships and function is reflected in this review, which groups nuclei according to their topographical relationships when

considering their gross anatomy, but groups them functionally when considering their connections.

The major parts of the basal ganglia are formed by the archistriatum, paleostriatum, and neostriatum (see Table 1.1 and Fig. 1.1).

The archistriatum is phylogenetically the oldest and consists of the amygdaloid nuclear complex. In lower animals it is mainly concerned with olfaction and other ‘visceral’ functions, but in higher species it forms part of the limbic system. Although the paleostriatum is more recent it is well developed in fish. In man it is represented by the globus pallidus. The neostriatum is only fully formed in mammals and is closely related to the extensive development of the neocortex. In birds and reptiles, in which there is apparently only rudimentary neocortical development, the thalamus may function as the main sensory centre in the brain and the striatum the main motor one. The motor activity in these animals is largely stereotyped, automatic, and instinctive. These striatal centres continue to subserve such activity in mammals.

In man the neostriatum consists of the claustrum, caudate nucleus, and putamen. Some authors exclude the claustrum as it is embryologically different and develops from an enfolding of neocortex rather than from the depths of the telencephlon. The neostriatum is sometimes simply referred to as the striatum. This and the paleostriatum, or globus pallidus, form a fairly contiguous mass which is sometimes called the corpus striatum. The term ‘striatum’, which forms part of many of these names, refers to the striped appearance of these structures in freshly cut specimens. This is due to bundles of pale myelinated fibres radiating between grey cellular masses. Because the putamen and the globus pallidus are in close juxtaposition, with their combined shapes resembling a bean, they are referred to as the lenticular nucleus. This grouping is unfortunate as, although the structure and function of the caudate and putamen are very similar, this does not apply to the putamen and globus pallidus.

The caudate nucleus and putamen have been regarded as virtually identical in structure, interconnections, and function. In spite of marked similarities, however, there are important differences (see later under ‘Structure and connections of the basal ganglia – neostriatum’). In lower mammals they are fused together in a single mass, but in man the development of neocortex has led to a profusion of ascending and descending fibres in the internal capsule, which separate the caudate from the putamen over most of its length.

The caudate nucleus forms an elongated ‘C’ shaped structure which opens rostro-inferiorally and lies immediately medial to the internal capsule (Figs 1.1–1.7). Its bulbous head is situated rostrally where it forms the lateral wall of the anterior horn of the lateral ventricle. As it arches dorsally and caudally it progressively narrows and this middle portion or body lies dorso-lateral to the thalamus (Figs 1.5, 1.6). The progressively narrowing tail swings round the postero-lateral margin of the thalamus as it passes laterally and rostrally towards the tip of the temporal lobe. Throughout its length the caudate nucleus is immediately adjacent to the lateral ventricle and it follows its course. A thin band of fibres, the stria terminalis, runs along its medial border (Figs 1.3–1.7). The tip of the head of the caudate nucleus is fused with the laterally placed putamen at the rostral end of the anterior limb of the internal capsule (Fig. 1.2). Interspersed along the length of the caudate are small bands of cells passing through the internal capsule connecting it with the putamen.

At the extreme anterior end of the caudate there is a small group of cells, the nucleus accumbens, which form part of its head. It has a peripheral shell and a central core. This nucleus is a relatively larger structure in lower mammals. It is closely related to the septal nuclei which lie medially (Fig. 1.2). The tip of the tail of the caudate nucleus lies immediately adjacent to the amygdaloid nuclear complex (Figs 1.1, 1.5, 1.6).

The putamen and the globus pallidus form the lenticular nucleus, which is a conical structure with its apex placed medially against the genu of the internal capsule and its base laterally against the external capsule (Figs 1.1–1.6). The globus pallidus forms the apex and the putamen is the base. The putamen thus extends rostrally, caudally, dorsally, and ventrally beyond the limits of the globus pallidus. It is the single largest part of the corpus striatum. Sections reveal bundles of fibres organized in a regular way radiating through it from the lateral surface towards the apex of the globus pallidus. It is encircled by the caudate nucleus, from which it is separated by the internal capsule. The medial surface of the putamen is separated from the lateral surface of the globus pallidus by a thin layer of white matter called the lateral or external medullary lamina (Figs 1.3–1.6).

The globus pallidus contains a greater proportion of myelinated fibres radiating through it and hence appears paler than the putamen in the fresh specimen. It is divided by a vertically placed sheet of white matter, the medial or internal medullary lamina, into lateral or external and medial or internal segments (Figs 1.4, 1.5). The medial segment lies at the extreme apex of the cone formed by the lenticular nucleus. The superior and medial aspect of the globus pallidus is in contact with the internal capsule, which separates it rostrally from the head of the caudate nucleus and anterior horn of the lateral ventricle, and caudally from the antero-lateral surface of the thalamus (Figs 1.1–1.6). In its mid and posterior portions the superior-medial surface of the globus pallidus is also separated from the thalamus by the subthalamic nucleus and the zona incerta. The caudal end of the subthalamic nucleus overlaps the rostral tip of the substantia nigra (Fig. 1.6). Anteriorly the inferior surface of the globus pallidus is adjacent to the substantia innominata and the more laterally placed fibres of the anterior commissure. The hypothalamus lies medial to this area (Figs 1.3–1.5). More caudally the inferior surface of the globus pallidus is closely related to the optic tract which winds around the lateral aspect of the cerebral peduncle to the lateral geniculate body (Figs 1.5–1.7). This part of the inferior surface of the globus pallidus is also adjacent to the amygdaloid nuclear complex and the medial temporal lobe containing the tail of the caudate nucleus, stria terminalis, hippocampus, fornix, and the temporal horn of the lateral ventricle. The caudate nucleus, stria terminalis, and fornix all curve caudally, superiorly, and then rostrally following the lateral ventricle to come again into close relationship with the globus pallidus, this time on its rostro-medial aspect at the tip of the anterior horn of the lateral ventricle (Figs 1.1–1.7).

The claustrum is a thin layer of grey matter, which lies lateral to the outer surface of the putamen, from which it is separated by a layer of white matter, the external capsule. It is thus lying beneath the surface of the insula in the sylvian fissure. The lateral surface of the claustrum is covered by the external capsule, which is another sheet of white matter.

The thalamus is the largest of the four divisions of the diencephalon, the others being the hypothalamus, subthalamic area, and epithalamus (i.e. the pineal gland and roof of the third ventricle including its attachment to the thalamus at the stria medullaris and the habenular (Figs 1.5–1.7). The thalamus is an ovoid body with its long axis running rostro-caudally. Its medial surface rostrally forms the upper part of the lateral wall of the third ventricle while caudally it is adjacent to the forward extension of the quadrigeminal cistern (i.e. the cerebro-spinal fluid (CSF)-filled space overlying the quadrigeminal bodies). The pineal gland lies in this space and is thus closely related to the caudal part of the medial thalamic surface. The lateral surface of the thalamus lies against the posterior limb of the internal capsule which separates it from the lenticular nucleus (Figs 1.5, 1.6). The ventral relationships of the thalamus are more complex. In its rostral part it is dorsal to the hypothalamus and subthalamic area. This latter zone contains the subthalamic nucleus and the zona incerta (Fig. 1.5). More caudally, the position of the subthalamus is taken by the red nucleus which separates the thalamus from the substantia nigra (Fig. 1.7). At this level the ventral aspect of the thalamus is adjacent to the medial lemniscus and medial and lateral geniculate bodies respectively, when passing from medial to lateral. The dorsal surface of the thalamus is partially encircled by the caudate nucleus. The head of the caudate nucleus lies rostro-medial to the thalamus and its body arches dorso-laterally over the dorsal thalamic surface to wind inferiorly around its caudal end. The medial part of the caudal end of the dorsal surface of the thalamus is in contact with the lateral ventricle (Fig. 1.6). The medial surfaces of the thalami may be partially fused, thus dividing the third ventricle. This interthalamic adhesion is called the massa intermedia.

The structure of the thalamus is complex. It contains a large number of nuclei of varying sizes, many of which overlap so that it is difficult to clearly visualize their relationships in sections restricted to any one plane. In addition the names of many of the nuclei do not accurately reflect their position in the adult human brain. Some of the thalamic nuclei are not directly related to movement disorders and these are not covered in detail.

Up to 60 different subnuclei have been distinguished, but in many cases anatomical subdivision has been arbitrary and based on minor differences in morphology, rather than on more valid criteria. There have been major differences between the anatomical and clinical literatures, between subhuman primates and man and between German and English speaking workers, which has added to confusion. This section is much simplified and follows a commonly used English speaking anatomical classification. In many of the chapters related to individual diseases, however, the clinical literature has been quoted, and terminology used by authors of these various papers has been retained. To aid interpretation and allow comparison, two main classification schemas are given in Table 1.2: that of Ohye (1990) in Table 1.2A and that of Hirai and Jones (1989) in Table 1.2B. It should be noted, however, that an exact match of terminologies between the two systems is not possible as in many places they do not exactly coincide.

A thin plate of myelinated fibres, the external medullary lamina, runs within the thalamus close to its ventral and lateral borders. This separates a thin laterally placed layer, the reticular nucleus (R), which is embryologically part of the subthalamus, from the main bulk of the thalamus, lying medially. Laterally the reticular nucleus is adjacent to the internal capsule and ventrally it is continuous with the zona incerta (Fig. 1.5). Another layer of white matter subdivides the main mass of the thalamus. It runs in an antero-posterior direction along the length of the thalamus and curves ventromedially from the dorsal surface. This is the internal medullary lamina and it contains a number of cell groups, which are collectively known as the intralaminar nuclei. The largest of these is the centromedian nucleus (CM) which lies almost in the centre of the thalamus.

The ventral nuclear group is divided into three main nuclei of about equal size which are respectively, from rostral to caudal, the ventral anterior (VA), the ventral lateral (VL), and the ventral posterior (VP) nuclei. A smaller ventral intermediate (VIM) nucleus is often delineated. The thalamic nuclei have been further subdivided into small groups of cells (Table 1.2), but although there may be cytoarchitectural differences the functional significance of some of these is doubtful (Asanuma et al. 1983[a,b]). The major divisions which have been distinguished in the ventral anterior nucleus are a small cell or parvocellular part (VApc) and a large cell or magnocellular part (VAmc). In the ventral lateral nucleus the major division is into oral (VLo), caudal (VLc), and medial parts (VLm). The ventral posterior nucleus is further subdivided into ventral postero-lateral (VPL) and ventral postero-medial nuclei (VPM). The former has oral (VPLo) and caudal (VPLc) parts and the latter has parvocellular (VPMpc) and ventral posterior inferior (VPI) divisions. Lying between the anterior parts of the dorso-medial and ventral anterior nuclei is the anterior nuclear group of the thalamus, which forms its rostral pole. This is subdivided into ventral (AV), medial (AM), and dorsal (AD) nuclei. A group of small nuclei lie adjacent to the third ventricle forming the midline nuclear group of the thalamus. These include cellular accumulations in the massa intermedia. The medial (MG) and lateral geniculate (LG) bodies are immediately infero-lateral to the pulvinar separating it from the brainstem.

The subthalamic area lies ventral to the thalamus, lateral to the hypothalamus, and dorso-medial to the globus pallidus, from which it is separated by the internal capsule (Figs 1.5, 1.6). It contains several structures, the most prominent of which are the subthalamic nucleus and the zona incerta. The subthalamic nucleus is shaped like a biconvex lens and lies parallel to the internal capsule at its medial border close to the apex of the globus pallidus. At its posterior end it overlaps slightly with the rostral margin of the substantia nigra which inserts itself between the subthalamic nucleus and the internal capsule–cerebral peduncle (Figs 1.6, 1.7). The subthalamic nucleus is separated from the thalamus by the zona incerta. This is a rather ill-defined area consisting of diffuse cell groups in a reticular network. It is continuous dorso-laterally with the reticular nucleus of the thalamus. It is closely related to Forel's fields (see later) and the pallidothalamic fibres, which entwine around its ventral and medial borders before arching back over its dorso-medial surface as the thalamic fasciculus. These fibres thus tend to separate the zona incerta from the subthalamic nucleus below and the thalamus above. The fields of Forel and the thalamic fasciculus are structures that are included in the subthalamic area.

The substantia innominata is an ill-defined area of grey matter beneath the rostral end of the lenticular nucleus, particularly the globus pallidus (Figs 1.2–1.5). It extends from the region of the septal nuclei and rostral hypothalamus medially, towards the amygdaloid complex laterally. Fibres of the anterior commissure lie dorsally and the optic tract lies ventrally. The substantia innominata contains a number of small groups of cells (Table 1.3).

The basal nucleus of Meynert is the most prominent group of cells lying in the substantia innominata just beneath the globus pallidus. Rostrally it is continuous with the nucleus of the horizontal limb of the diagonal band. Its cells become continuous superiorly with those in the medullary laminae of the globus pallidus. The basal nucleus of Meynert has been divided into two portions by some authors (i.e. the nucleus of the ansa peduncularis, which lies below the medial portion of the posterior limb of the internal capsule, and the nucleus of the ansa lenticularis, which extends laterally below the lenticular nucleus [see Mettler 1968]). The nucleus of the diagonal band lies between the anterior commissure and the optic chiasm. The diagonal band connects the amygdaloid complex with the area of the septal nuclei (Figs 1.2–1.5). The nucleus of the stria terminalis encircles the rostral end of the stria and the adjacent anterior commissure (Fig. 1.3). The substantia innominata contains a dense plexus of somatostatin – immunoreactive nerve terminals, which also ascend in the medial medullary lamina of the globus pallidus. The cells of origin may be in the amygdala (Mufson et al. 1988, Desjardins and Parent 1992). The medial medullary lamina (MML) of the globus pallidus comprises a thin layer of white matter intervening between the external segment of the globus pallidus and the internal segment of the globus pallidus and has cells with high levels of glycine receptors (Waldvogel et al. 2007). Some authors have viewed the substantia innominata as consisting of elements of the ventral striatopallidal system rostrally and an ‘extended amygdala’ caudally, with large cholinergic and non-cholinergic corticopedal neurons interwoven between them (Alheid and Heimer 1988).

The rostral end of the lenticular nucleus, caudate nucleus, and substantia innominata lie in close relationship to and have interconnections with the olfactory apparatus and the septal nuclei. Axons from the olfactory bulb pass posteriorly in the olfactory tract. Dorsal to the optic chiasm the olfactory tract splits into medial and lateral olfactory striae, which sweep medial and lateral to a small area of cortex perforated by many small blood vessels, the anterior perforated substance (Figs 1.2, 1.4). The medial olfactory stria initially lies on the ventral surface of the cortex but turns dorso-medially to form the anterior fibres of the anterior commissure. These cross and pass forwards into the contralateral olfactory tract and bulb. It is doubtful if any of these fibres pass into the septal nuclei. The lateral olfactory stria also initially lies on the ventral surface of the cortex, but as it curves laterally and posteriorly it penetrates the adjacent temporal lobe to reach the amygdaloid complex and the adjacent prepyriform cortex, which overlies the peri-amygdaloid area. Prominence of the anterior perforated substance is due to a small cellular mass, the olfactory tubercle, which lies immediately dorsal to it. This receives fibres from the olfactory striae and the amygdaloid nuclear complex and projects to the stria medullaris.

The septal nuclei lie medial to the substantia innominata where the rostral tip of the head of the caudate nucleus joins with the accumbens nucleus (Fig. 1.2). The septal nuclei are in the septum pellucidum rostral to the anterior commissure. Medial and lateral septal nuclei can be distinguished.

The amygdaloid nuclear complex lies in the dorsal part of the medial temporal lobe just anterior to the tail of the caudate nucleus and the tip of the temporal horn of the lateral ventricle (Figs 1.1, 1.4, 1.5). The claustrum is lateral and corpus striatum superior. The medial aspect of the amygdala is covered by a thin layer of cortex and superiorly it lies adjacent to the optic tract and substantia innominata (Figs 1.4, 1.5). There are two main nuclear masses, a larger basolateral and a smaller corticomedial nuclear group, the latter lying dorso-medial. Each is subdivided into several constituent nuclei.

The hypothalamus is largely related to endocrine, autonomic, and visceral functions. Although these may be disturbed in some movement disorders, this is usually not a major feature and the anatomy is only briefly considered here. The hypothalamus is formed by the lower parts of the lateral walls of the third ventricle and its floor. It lies below the thalamus and medial to the subthalamic area, internal capsule, globus pallidus, substantia innominata, and optic tract respectively, when passing from dorsal to ventral (Figs 1.3–1.6). The ventral aspect of the hypothalamus extends from the mammillary bodies posteriorly to the optic chiasm anteriorly. It contains a ventral bulge, the tuber cinereum, to which is attached the infundibulum or stalk leading to the pituitary gland. A number of separate nuclei are identifiable and these can be divided into several groups. The preoptic group lies above and anterior to the optic chiasm in the walls of the preoptic recess. The supraoptic group lies above the posterior and superior part of the optic chiasm where it indents the floor of the third ventricle (Fig. 1.3). Nuclear masses in the lateral walls of the third ventricle are divided into medial and lateral groups by two bands of fibres. The first band is the fornix, which passes downwards and backwards from the region of the anterior commissure to the mammillary body, and the second is the mamillothalamic tract, which runs upwards and forwards from the mammillary body to the anterior nuclear group of the thalamus (Figs 1.5, 1.6). A final group of hypothalamic nuclei is made up from the mammillary bodies and surrounding groups of cells.

The red nucleus consists of a large oval mass of cells with its long axis running rostro-caudally. It has a slight pinkish coloration in the freshly cut specimen. It extends over a considerable length, but the substantia nigra is longer and is closely related to the whole of its ventral aspect (Figs 1.1, 1.7, 1.8).

In its rostral part, the posterior thalamus lies dorsal and the red nucleus occupies a position similar to that of the more anteriorly placed subthalamic nucleus. Caudally the dorsal aspect of the red nucleus is related to the periaqueductal grey matter and the nucleus of the third cranial nerve. The fibres of this cranial nerve pass through the substance of the red nucleus on their way to exit in the interpeduncular cistern. The medial lemniscus and the geniculate bodies lie dorso-lateral to the red nucleus. Rostrally the red nucleus is related to the retroflex fasciculus, which is a band of axons passing from the habenular nucleus (see section on the ‘Thalamus’) ventrally and caudally to the interpeduncular nucleus (Fig. 1.8). Caudally fibres from the contralateral superior cerebellar peduncle in the brachium conjunctivum enter the red nucleus. Some of them pass through it and continue rostrally to the thalamus. The prerubral area lies just rostral to the red nucleus and contains a number of scattered cells and many fibres of passage. Efferent fibres from the globus pallidus forming Forel's field H are the major component (see later under ‘Efferent connections of the globus pallidus’).

The substantia nigra is an elongated nucleus, which has a dark appearance in the freshly cut specimen. Its long axis lies antero-posteriorly, parallel and dorso-medial to the descending corticospinal fibres in the cerebral peduncle. It extends throughout the length of the midbrain and its rostral tip insinuates itself just ventral to the caudal end of the subthalamic nucleus (Figs 1.1, 1.6, 1.7, 1.8). The red nucleus is dorso-medial while the medial lemniscus and geniculate bodies are dorsal. The dorsal portion of the substantia nigra is the pars compacta, which consists of numerous closely packed cells containing melanin pigment. Ventrally, separating the pars compacta from the cortico-spinal fibres in the cerebral peduncles, is a larger zone with fewer cells, the pars reticulata. A smaller pars lateralis lies at the dorso-lateral tip of the substantia nigra. There is a small group of scattered cells which lie medial to the main mass of the substantia nigra and just dorsal to the interpeduncular nucleus (Fig. 1.8). These cells of the ventral tegmental area of Tsai have similar functions to the substantia nigra pars compacta and may really be considered an extension of this. (See later under ‘Efferent connections of the substantia nigra pars compacta’)

The most prominent structures on the dorsal aspect of the midbrain are the paired swellings of the superior and inferior colliculi (Figs 1.1, 1.8). The superior colliculi are immediately caudal to the third ventricle and thalamus, being overlapped dorsally by the latter, although separated from it by the anterior extension of the CSF-filled quadrigeminal cistern. The pineal gland lies dorsal, between the two superior colliculi. The inferior colliculi are immediately caudal to the superior ones. Both colliculi consist of a complex collection of cells. The superior colliculi are related to eye movement and the inferior colliculi form a relay station for ascending auditory information passing to the medial geniculate bodies.

The pedunculo-pontine nucleus is a small mass of cells which is part of the midbrain reticular formation. It lies just ventral to the inferior colliculus and dorsal to the caudal end of the substantia nigra. The superior cerebellar peduncles emerge from the anteroventral aspect of the cerebellum and run rostrally over the dorsal surface of the pons (Fig. 1.1). They form part of the walls of the rostral fourth ventricle (Figs 1.9, 1.10).

At the junction of the pons and midbrain the fibres pass ventro-medially and decussate at the level of inferior colliculi, forming the brachium conjunctivum. Just prior to the decussation they pass around and through the pedunculo-pontine nucleus. From the level of the decussation rostrally, the fibres lie ventral to the aqueduct of Sylvius and central grey matter. They pass forward in this relative position to enter the posterior end of the red nucleus, many of them passing through it on the way to the thalamus. There are several other structures relevant to the output from the cerebellum as follows:

The cerebellum is attached to the brainstem by superior, middle, and inferior cerebellar peduncles (Fig. 1.1). The middle cerebellar peduncle is formed by fibres from the pontine nuclei (Fig. 1.9), which decussate before passing laterally into the cerebellum. These nuclei receive afferents from the cortico-pontine tracts. This peduncle lies lateral to the superior and inferior ones as its fibres run dorsally into the cerebellum. The inferior cerebellar peduncle contains axons from the inferior olivary complex and dorsal spinocerebellar fibres from the spinal cord. This peduncle is formed by a band of fibres on the dorso-lateral surface of the medulla and enlarges as it passes rostrally. As it turns dorsally into the cerebellum it comes to lie just infero-lateral to the superior peduncle and medial to the middle peduncle (Figs 1.9, 1.10). The cerebellum has several divisions but the main ones are the two laterally placed cerebellar hemispheres joined by the midline vermis. In cross section the cerebellum can be seen to consist of a superficial cerebellar cortex separated from central nuclei by white matter. There are four deep cerebellar nuclei on each side, and from lateral to medial these are the dentate, emboliform, globose, and fastigial nuclei (Fig. 1.10). The emboliform and globose nuclei are separate in man, but in other mammals they form a continuous mass known as the interposed nucleus. These nuclei act as the main stations for efferent fibre projection from the cerebellum.

The caudate and putamen have a highly cellular structure, which is broken up by numerous bundles of afferent and efferent fibres. These nuclei have a more abundant capillary network than the globus pallidus. Earlier studies suggested that there were two basic types of striatal neurons. The first were small rounded or oval achromatic cells. Although some of them were truly small cells of 6–10 µm in diameter, others were really medium-sized cells being 15–30 µm. Some have interpreted these smaller cells as being glial and their nature remains uncertain (Graybiel and Ragsdale 1983). The second type of neuron was large, at about 50–60 µm in diameter with a polymorphic multipolar soma, a large nucleus containing a dense central nucleolus, and a peripheral ring of chromatic material. There was only a little cytoplasm containing coarse clumps of Nissl material. These larger neurons tended to have satellite groups of glial cells around the soma and had fewer dendrites than the small cells. The large cells were thought to be Golgi type I neurons (i.e. axons projecting distally beyond the neostriatum) and the small cells Golgi type II neurons (i.e. axons arborizing nearby and not projecting to a distance) (see Mettler 1968 for a review). The normal human neostriatum has approximately 10 million small-to-medium and 670,000 large cells; i.e. a ratio of approximately 20:1 (Schroeder et al. 1975). While the morphology of the cell bodies and their relative numbers is accurate, it is now known that the concept of distribution of their axons is incorrect.

Electron microscopic and other techniques have revealed that the dendrites of some neurons are covered in small processes or ‘spines’. These spines are in synaptic contact with boutons of afferent fibres. Neurons have thus been classified as spiny or aspiny and medium or large sized (Fox et al. 1971, Fox et al. 1971/1972). This classification is probably too simple, as at least six and possibly seven different types of neurons have been identified in the striatum based on their morphological and neurochemical characteristics, as outlined below (DiFiglia et al. 1976, Graybiel and Ragsdale 1983, Kawaguchi 1995; Table 1.4.

Spiny I (SI) neurons (Pasik et al. 1979, DiFiglia et al. 1980) have numerous spines on distal, but not proximal, dendrites and considerable local axonal branching. The nuclei are not indented and lack inclusions. Spiny II (SII) (Pasik et al. 1979, DiFiglia et al. 1980) neurons have fewer spines, more profuse dendritic branching, and less local axonal arborization than SI neurons. The nuclei are indented. Aspiny I (AI) neurons have varicose curved dendrites which divide close to the soma and substantial local axonal branching. The nuclei are deeply indented. Aspiny III (AIII) neurons have only occasional spines, elongated thin dendrites, and beaded collaterals forming a local axonal plexus.

Spiny II (SII) neurons (Pasik et al. 1979, DiFiglia et al. 1980) have infrequent spines and elongated somas. They are probably a larger version of the medium-sized spiny II neurons. Aspiny II (AII) neurons (Pasik et al. 1979) have varicose smooth dendrites and ovoid somas and are the biggest neurons in the striatum. Striato-nigral type 2 neurons (Bolam et al. 1981[a]) have neither varicosities nor dendrites, but as yet their morphology and relationship to other cells is uncertain. These neurons seem infrequent.

There is a lot of work suggesting that fibres which project beyond the neostriatum belong to the medium-sized spiny neurons, which are thus Golgi type I cells (Somogyi and Smith 1979, Preston et al. 1980, Kitai 1981). It has also been postulated that large spiny cells are projection neurons (Pasik et al. 1979). Although the situation is not completely clarified, some large aspiny cells do not project beyond the neostriatum and thus fit into the category of Golgi type II cells (Kitai 1981, Woolf and Butcher 1981, Chung and Hassler 1982). The ‘Strionigral type 2’ neuron of Bolam et al. (1981[a]), however, seems at least one type of large aspiny cell which projects beyond the striatum. The concept of two types of cells which project their axons either locally or distally is too simple and it is now known that many of the neurons which send axons to the globus pallidus also have extensive collateral systems which ramify in the neostriatum adjacent to the parent cell and form contacts with other spiny neurons by way of symmetric synapses (Kitai 1981, Bolam et al. 2000). Intracellular labelling techniques have shown that the axons that do project distally may have several collaterals (Kitai 1981).

Yelnik et al. (1991) have argued for a return to a simpler system and have suggested that previous taxonomy has been based on identification of specific morphological features, rather than valid classification criteria. They have proposed a subdivision into only four neuronal categories in primates as follows: a) spiny neurons 96% (includes SI and some SII); b) leptodendritic neurons 2% (basically aspiny, although occasional spines have been described in young individuals – includes some SII and some aspiny neurons); c) spidery neurons 1% (includes some AII); and d) microneurons 1% (includes AI and AIII).

These authors distinguish only three types of cell bodies: small and rounded corresponding to either spiny neurons and microneurons, medium-sized and elongated corresponding to leptodendritic neurons, and large and rounded corresponding to spidery neurons.

The addition of histochemical and antibody labelling techniques has provided new information about the probable neurotransmitters used by certain morphological types of cells. Although several types of neurotransmitters are probably produced within the neostriatum, the two major ones appear to be acetylcholine (ACh) and gamma-aminobutyric acid (GABA).

Unfortunately the detection of acetylcholine is difficult and most studies have relied on identifying its degrading enzyme acetylcholinesterase (AChE). The presence of AChE on or in a neuron does not necessarily imply that ACh is the neurotransmitter produced by that cell. In non–cholinergic neurons AChE may act to degrade ACh from afferent synapses. Other possibilities include the catabolism of substance P or release into neighbouring blood vessels (Butcher and Woolf 1982). Cells showing AChE reactivity stain either lightly or intensely. While the status of cells which stain lightly is uncertain, most of those that stain intensely probably produce ACh as a neurotransmitter, although the dopaminergic neurons of the substantia nigra pars compacta may be an exception (Butcher and Woolf 1982). In the neostriatum there are two morphological types of cells which stain for AChE and these are the small to medium aspiny (AI) and the large aspiny (AII) neurons. Only 3–4% of striatal cells stain for AChE and not all of these show an intense reaction. Most studies show that AChE staining cells do not project beyond the neostriatum and appear to act as a local interneuron (Henderson 1981, Woolf and Butcher 1981, Butcher and Woolf 1982). Occasional discrepancies suggesting projections to substantia nigra (Kayia et al. 1979) and neocortex (Parent et al. 1981) are unresolved.

Studies using monoclonal antibody to choline acetyl transferase (CAT), an enzyme involved in the formation of ACh, have shown that cholinergic neurons are large aspiny cells (Bolam et al. 1984), and not the small to medium sized cells. The large rounded spidery neurons of Yelnik et al. (1991) are AChE positive. Although the vast majority of CAT immunoreactive fibres in the neostriatum are derived from these cells, a few appear to come from the basal forebrain (Mesulam et al. 1992) and possibly substantia nigra dopaminergic neurons, raphe nuclei, thalamus, and cerebral cortex (Bernard et al. 1995).

Medium sized spiny cells projecting from the neostriatum seem to use GABA as their neurotransmitter and this probably applies to strio-pallidal and strio-nigral connections (Wilson and Groves 1980, Kitai 1981). These neurons also have GABA receptors at synapses on their dendrites, spines, and perikarya (Fujiyama et al. 2000). Although some of the medium sized aspiny neurons take up [³H] GABA and are thus probably GABAergic, they seem to be interneurons and do not project beyond the neostriatum. Approximately 15% of rat neostriatal cells can be labelled in this way and may be interneurons (Bolam et al. 1983[a]). They express very strong immunoreactivity for GABA and glutamic acid decarboxylase (Katsumaru et al. 1988, Kita and Kitai 1988). Most neostriatal interneurons also contain GABA receptors, although there is considerable regional variation in the distribution of subunit types (Waldvogel et al. 1999). GABAergic striatal interneurons synapse with a variety of structures including probable cholinergic neurons (Bolam et al. 1985). There are several subtypes of GABAergic interneurons, the biggest of which is characterized by the presence of the calcium-binding parvalbumin. These receive direct input from cortical neurons by way of AMPA receptors (Bernard et al. 1997). After stimulation of the cortex these neurons show increased expression of Fos over a bigger striatal zone than is seen with spiny neurons (Parthasarathy and Graybiel 1997). The main efferent output of GABAergic interneurons is to the GABAergic spiny projection neurons and the terminals of the former form basket-like aborizations around the perikarya of the latter. Individually such interneurons may receive input from both motor and sensory cortical regions and it has been suggested that they may help integrate activity from different cortical areas. They may mediate lateral inhibition within the striatum by providing a feed-forward surrounding zone of inhibition, thus focusing and perhaps limiting the duration of cortical excitation. Stimulation of GABAergic interneurons results in large inhibitory post-synaptic potentials in spiny neurons (Koos and Tepper 1999). Thus, although they are present in relatively low numbers, they may be able to exert a major effect on striatal function (see Bolam et al. 2000).

Substance P is also present in the neostriatum and there is evidence that this acts as a neurotransmitter in strio-nigral and strio-pallidal pathways (Brownstein et al. 1977). Fibres to the globus pallidus go to the internal segment (Graybiel and Ragsdale 1983). Antibody labelling shows that there are two morphological types of substance P-containing cells in the neostriatum. One seems to correspond to the medium sized spiny neurons and probably gives rise to strio-nigral fibres as well as axon collaterals, which terminate within the neostriatum itself. It has been suggested that substance P in these axons might all be released by collaterals in the neostriatum onto cholinergic or somatostatinergic neurons and none may pass to the substantia nigra (Lee et al. 1997). Striatal substance P may act at synaptic and extrasynaptic sites (Li et al. 2000). A second type of medium sized cell has been suggested, but the classification and projections of this remain uncertain (Bolam et al. 1983[b]). Using immunohistochemistry, islands of dense staining for substance P stand out from a background of lesser reactivity (Beach and McGeer 1984). There is also evidence suggesting a dynorphin (an opioid peptide) pathway from the neostriatum to the substantia nigra and globus pallidus and substance K may also be involved (Vincent et al. 1982, Weber et al. 1982, Lee et al. 1986, Graybiel, Gerfen and Young 1988). Substance P and dynorphin appear to be co-localized in projection neurons which also contain GABA (Anderson and Reiner 1990, Karle et al. 1992) and express the D1 subtype of dopamine receptor (see Bolam et al. 2000).

Enkephalin-containing neostriatal neurons project to the external segment of globus pallidus (Del Fiacio et al. 1982, Graybiel and Ragsdale 1983). Enkephalin seems to be in different neurons from substance P and dynorphin, but these also have GABA within them (Reiner and Anderson 1990) and express the D2 subtype dopamine receptor (see Bolam et al. 2000 and ‘Nigro-striatal fibres’ below for distribution of dopamine receptors throughout the neostriatum). Neurotensin is found in medium sized spiny neurons and in some of them it co-localizes with enkephalin, with or without GABA. In the cat these cells project to the globus pallidus, pars compacta, and pars lateralis of the substantia nigra and retrorubral area, unless they contain GABA, in which case they end mainly in the globus pallidus and pars reticulata of the substantia nigra (Sugimoto and Mizano 1987).

Neurokinin B may co-localize with substance P or enkephalin and fibres may project to the globus pallidus in the rat (Burgunder and Young 1989).

The basic division of striatal output GABAergic neurons into those containing substance P and dynorphin and expressing D1 receptors on the one hand and those containing enkephalin and expressing D2 receptors on the other has been seen as reflecting division into the direct and indirect striatofugal pathways respectively (see later and Chapter 2). Some authors, including Parent et al. (2000), have challenged this concept as being simplistic and inaccurate. Individual efferent axons from the striatum have been observed to arborize in both pallidal segments and the substantia nigra in the primate (Parent 1995, Levesque and Parent 2005).

Somatostatin is localized in striatal interneurons, along with neuropeptide Y, and these medium sized aspiny cells have high levels of nicotinamide adenine dinucleotide phosphate-diphospherase (NADPH-diphospherase or nitric oxide synthetase) (Aronin et al. 1983, Graybiel and Ragsdale 1983, Graybiel 1986, Figueredo-Cardenas et al. 1996). These interneurons form a population that is quite distinct from those that stain positively for CAT (Selden et al. 1994), although some contain AChE (Bernard et al. 1995). Cholecystokinin (Emson et al. 1980) and angiotensin (Arregui et al. 1977) are also found in the neostriatum, but morphological details are sparse. The former, however, is in some intrinsic aspiny neurons (Adams and Fisher 1990), as is vasoactive intestinal polypeptide (Theirault et al. 1987). GABAergic interneurons may also express mRNA for nerve growth factor, acidic fibroblast growth factor, and glial cell line-derived neurotrophic factor, whereas only the first of these is seen in cholinergic cells (Bizon et al. 1999). Parvalbumin is a Ca-binding protein present in the basal ganglia. The leptodendritic neurons of Yelnik et al. (1991) are parvalbumin-immunoreactive and are GABAergic interneurons, which are distinct from the somatostatin-containing ones (Kita et al. 1990). Calretinin is another Ca-binding protein, which is also found in interneurons which do not express parvalbumin (Schlosser et al. 1999). Vitamin D-dependent calcium binding protein (calbindin D_28k) (Gerfen et al. 1985, DiFigia et al. 1989, Celio 1990) and transforming growth factor alpha (Fallon 1987) are located in spiny neurons and the former is also found in interneurons (Prensa et al. 1998) including those mentioned above, which stain positively for somatostatin and NADPH-diaphosphorase (Selden et al. 1994). There is also a small population of what appear to be striatal dopaminergic interneurons, which do not express immunoreactivity for parvalbumin, calbindin D_28k, and NADPH-diaphosphorase (Betarbet et al. 1997).

Glutamate and aspartate from cerebral cortex, dopamine from substantia nigra pars compacta, serotonin (5-hydroxytryptamine) from raphe nuclei, cholecystokinin from claustrum, neocortex, amygdala, thalamus, and ventral mesencephalon (Graybiel 1986, Adams and Fisher 1990), vasoactive intestinal polypeptide (Emson et al. 1979), substance P (Sugimoto et al. 1984), calcitonin gene-related peptide from the thalamus (Inagaki et al. 1990), and noradrenaline from the locus coeruleus are thought to be afferent transmitters to the neostriatum, but much of the morphological detail of synaptic contacts is uncertain (Pasik et al. 1979, Sugimoto et al. 1985, Adams and Fisher 1990). Enkephalin (Somogyi et al. 1982), vasoactive intestinal polypeptide (Emson et al. 1979), thyrotrophin releasing hormone (Spindel et al. 1981), and several other compounds may also be afferent transmitters from unknown sources.

Major afferent pathways terminate in the neostriatum in multiple, small clusters. These include fibres from cerebral cortex, thalamus, amygdala, substantia nigra, and raphe nuclei. Similarly, neostriatal neurons projecting to globus pallidus and substantia nigra have a patchy distribution (Graybiel and Ragsdale 1983). A number of neurotransmitters and related substances are also dispersed in mosaic or small ‘compartments’ rather than being spread evenly throughout the neostriatum. Zones of low acetylcholinesterase activity, called ‘striosomes’, form a three-dimensional network. They are separated by a background of high acetylcholinesterase activity, known as ‘matrix’.

Striosomes are also low in CAT immunoreactivity (Hirsch et al. 1989). These may be areas of high interneuron density (Graybiel et al. 1979) and are zones of termination of some dopaminergic fibres (Graybiel and Ragsdale 1983), although the matrix contains a greater number of tyrosine hydroxylase-immunoreactive terminals (Martin et al. 1991). During striatal development the striosomes are areas of high acetylcholinesterase and tyrosine hydroxylase activity compared with the matrix, but in the mature primate the converse applies (Graybiel 1984[b]). The dopaminergic input into the neostriatum in the primate is compartmentalized, with the retrorubral and ventral tegmental area of Tsai projecting to the matrix, while the striosomes receive input from the ventral tier and downwards projecting finger-like processes of the pars compacta (Langer and Graybiel 1989; see later under ‘Substantia nigra – structure’). Dopamine receptors are also influenced by these divisions, with D1 receptor density being greater in striosomes and D2 receptors being more common in matrix (Loopuijt et al. 1987). Some authors have claimed, however, that virtually all neostriatal neurons contain both of these types of receptors (Aizman et al. 2000), although others have suggested the opposite (Aubert et al. 2000). Muscarinic M1 receptor binding is also denser in the striosomes, although cholinergic cell bodies, neutrophil, and uptake sites are more abundant in the matrix (Nastuk and Greybiel 1988). GABA(A) receptors are more frequent in striosomes, but there is a complex difference in distribution of the various receptor subunits (Waldvogel et al. 1999). In contrast, GABA(B) receptors seem to be more homogeneously distributed between the striosome and matrix compartments (Waldvogel et al. 2004) where they are found not only on terminals that appear to be of cortical origin, but also at extrasynaptic sites (Smith et al. 2000).

The relationship of neuropeptide immunoreactivity to striosomes and matrix shows variability between different parts of the striatum (Martin et al. 1991, Holt et al. 1997), but some generalizations can be made. Striosomes show high levels of immunoreactivity to enkephalin, neurotensin, and dynorphin and are the site of benzodiazepine, kainite (Dure et al. 1992), and opiate-like receptor activity (Graybiel et al. 1980, Graybiel et al. 1981, Graybiel 1986, Faull and Villiger 1988). Mu-opiod receptors in striosomes are often co-localized with NMDA-type glutamate receptors on spiny neurons (Wang et al. 1999). All human striatal neurons, however, express mRNA for NMDA receptors, although there is considerable variation in distribution of receptor subtypes (Kuppenbender et al. 2000). Enkephalin-containing neurons are probably mainly located in the matrix, while axons and terminals seem to predominate in striosomes (Bolam et al. 1988). Substance P-containing neurons, however, are common in both compartments (Bolam et al. 1988, Martin et al. 1991), although the striosomes show the highest level of immunoreactivity (Bennett and Bolam 1994). The latter tends to be concentrated in neurons adjacent to striosomes and it has been suggested that these and other interneurons (see later) might transfer information from one compartment to another (Bolam et al. 1988, Augood et al. 1991). In primates these boundary zones form rings of neurochemically distinct tissue (Graybiel 1984[a]). Histochemical and immunohistochemical techniques in humans have shown that striosomes are composed chemically of two distinct domains: a core and a peripheral region (Faull 1989, Jakab 1996). The core is slightly higher in AChE activity than the periphery, but is lower in calbindin D_28k, while the periphery has high levels of CAT, calretinin, and NADPH-diaphorase activity. The entire striosomes stain weakly for parvalbumin and intensely for substance P. There is, however, considerable variation throughout the striatum, particularly in a rostro-caudal direction (Prensa et al. 1999). Calbindin immunoreactivity is also somewhat greater in the matrix, as is staining for NADPH-diaphosphorase (Martin et al. 1991, Sadikot et al. 1992) and parvalbumin (Waldvogel and Faull 1993, Bennett and Bolam 1994).

In the rat and the cat, striosomes project to the substantia nigra pars compacta, while the matrix sends axons to the pars reticulata and the globus pallidus (Graybiel et al. 1979, Gerfen 1985,  Jimenez-Castellanos and Graybiel 1987, Bolam et al. 1988, Jimenez-Castellanos and Graybiel 1989). The situation in primates is unspecified (Joel and Weiner 2000). These mosaic patterns in striatum seem to be a reflection of topographical compartmentalization elsewhere in the nervous system. Thus prefrontal and insular cortices plus amygdala may project to striosomes, while sensory and motor cortices and angular gyrus might radiate to the matrix (Jimenez-Castellanos and Graybiel 1989). It has been suggested that division into striosomes and matrix may represent two parallel input–output systems linked by striatal interneurons (Gerfen 1985). At a cellular level striatal projection neurons largely respect this compartmentalization. Thus, the majority of medium sized spiny neurons have dendritic arbors which remain entirely within their compartment, although about a quarter have at least one dendrite crossing the boundary. In comparison, interneurons, such as those immunoreactive for CAT, neuropeptide Y, or parvalbumin, which are near the border, have dendrites which extend into both compartments (Kitai et al. 1990, Kubota and Kawaguchi 1993, Bennett and Bolam 1994). These traversing dendrites may allow for ‘cross-talk’ between these largely separate systems (Walker et al. 1993). The matrix itself has been found to also have distinct input and output zones, which have been termed ‘matrisomes’ (Graybiel et al. 1991). The matrix is thus not homogenous but has a highly organized modular structure. Hence, the sensorimotor cortical afferents to the striatum innervate only distinct regions (Malach and Graybiel 1986, Flaherty and Graybiel 1991). Overall, the connections of the matrix suggest it has a mainly motor function.

The relationship of cholinergic neurons to nigro-striatal terminations has been a matter of debate. An hypothesis that cholinergic interneurons formed a link between the dopaminergic neostriatal system and neostriatal output pathways is unlikely. Dopamine terminals have been reported to synapse on GABAergic, enkephalinergic, and substance P-containing spiny neurons (Groves 1980, Pickel et al. 1981, Kubota et al. 1986[a], Kubota et al. 1987,) and not on aspiny neurons, which are probably the cholinergic interneurons. Nonetheless, dopamine receptor stimulation and blockade alters the activity of striatal cholinergic neurons, which appear to have reciprocal effects to those of dopaminergic neurons (Lehmann and Langer 1983) and cholinergic neurons contain dopamine (D2) receptors (Joyce and Marshall 1987). In addition, substance P-containing terminals have been reported to synapse on cholinergic neurons and may act via neurokinin 1 receptors (Bolam et al. 1986, Pickel et al. 2000). The situation remains to be clarified.

With the development of silver impregnation techniques it became possible to demonstrate a projection from most parts of the cerebral cortex to the neostriatum (Carman et al. 1963, Webster 1965, Kemp and Powell 1971[a]). Thus, afferents have been demonstrated from the prefrontal (Goldman and Nauta 1977), sensory-motor (Künzle 1975, Künzle 1977, Jones et al. 1977), auditory (Reale and Imig 1983), visual (Battaglini et al. 1982), and olfactory (Sorensen and Witter 1983) cortices. Most of these cells are in infragranular cortical layers (V and to a lesser extent VI), but some projecting to the caudate arise from supragranular levels (II and III) (Royce 1982). The dendritic trees of cortico-striate neurons reach towards the superficial cortical layers and in some cases extend to layer I (Hersch and White 1982). Although each cortical layer projects to both striosomes and matrix, the deep part of V and layer VI innervate mainly striosomes, whereas the converse applies with the superficial part of V and layers II and III in the rat (Gerfen 1989). In this animal three kinds of cortico-striatal axonal arborizations are observed. One preferentially, but not exclusively, innervates the matrix, a second strictly innervates the striosomes, and a third focuses terminals in groups in the matrisomes (Kincaid and Wilson 1996).

Parts of the cerebral cortex in one hemisphere may project to the neostriata bilaterally and individual striatal areas receive bilateral innervation from homotopic cortical areas (Royce 1982). In some species all major cortical areas radiate bilaterally, although the ipsilateral connections predominate (McGeorge and Faull 1989). In the rat, cat, and rabbit the sensory and motor cortices both project bilaterally to the neostriatum, but in the monkey the projection from the somatic sensory cortex (areas 3a, 3b, 1, and 2) is unilateral (Jones et al. 1977, Künzle 1977, Malach and Graybiel 1986, McGeorge and Faull 1987), whereas the fibres from the somatic motor cortex (area 4) radiate bilaterally (Künzle 1975). Bilateral fibres cross in the corpus callosum and enter the putamen via the external capsule. Some fibres may reach the caudate through the subcallosal fasciculus. Certain areas of cortex project chiefly to matrix while others are directed to striosomes (Malach and Graybiel 1986). Thus, in the rat, allocortex mainly innervates striosomes, whereas neocortex principally sends fibres to matrix (Gerfen 1989).

There are distinct differences between these projections to the caudate and putamen so that they can no longer be considered as identical functional units. In the monkey the motor and sensory cortices project to the putamen but only negligibly to the caudate, although prefrontal area 9 has an extensive efferent connection with the caudate (Goldman and Nauta 1977). The premotor cortex, area 6, may also project to the caudate (Royce 1982). Details vary between species, but there is a general topographical arrangement such that rostral to caudal and ventral to dorsal cortical areas tend to project to similar locations in the neostriatum. In the monkey the leg and tail representation of the somatic motor area is rostro-dorsal in the putamen, while the face and upper limbs are caudo-ventral (Künzle 1975, Fotuhi et al. 1989, Flaherty and Graybiel 1991). One feature of the organization of the putamen is that there is an overlapping layering of inputs from sensory and motor cortices (Kitai 1981). Thus, there is a highly organized topographical innervation of the striatum from the cortex. In general, neocortex mainly projects to the dorsal striatum and the allocortex; particularly the hippocampal formation goes to the ventral striatum. Frontal cortex projects especially to the head of the caudate; sensory, motor, and parietal cortices to the putamen; temporal cortex to medial and caudal putamen as well as caudate; and occipital cortex to the tail of the caudate. Any single cortical area, however, projects not only to the above regions but also much more widely and can in a patchy fashion reach almost any striatal area (Fotuhi et al. 1989, Flaherty and Graybiel 1991). On the whole the putamen seems to be chiefly connected to motor and sensory regions of the cortex, suggesting that its functions are mainly motor, whereas the caudate's principal innervation from the prefrontal cortex hints at a greater role in the planning, memory-based, and psychological aspects of extrapyramidal functions. A major exception to this is that the frontal eye fields, and medial or supplementary eye fields, predominantly innervate the caudate and hence this latter structure seems to subserve eye movements (Stanton et al. 1988, Huerta and Kass 1990, Parthasarathy et al. 1992). In addition, there is considerable convergence of inputs from different sources onto striatal cells. Thus the same cell may receive inputs from the cerebral cortex, thalamus, and midbrain (Buchwald et al. 1973, Kitai et al. 1976[a], Kocsis et al. 1976, Van der Maelen and Kitai 1980). It has been proposed, however, that individual cortical neurons make very few synaptic contacts with individual striatal neurons and that neighbouring striatal cells do not share common cortical inputs (Kincaid et al. 1998).

Whether the cortico-striatal fibres are collaterals of other descending projections or whether they are an independent system has been a matter of debate (Kitai et al. 1976[b], Jones et al. 1977, Oka and Jinnai 1978, Jinnai and Matsuda 1979). Combined intracellular labelling and electrophysiological techniques have demonstrated that at least some striatal afferents from somatic sensory-motor cortex are collaterals of fibres which continue to descend in the internal capsule, possibly passing to the brainstem or spinal cord (Donoghue and Kitai 1981). On the other hand, some cortico-striate neurons do not receive direct excitation from ventro-lateral thalamus, unlike cortico-pontine and cortico-spinal cells (Kitai 1983). The problem is unresolved. The cortico-striate fibres have asymmetric synaptic terminals on the dendritic spines of medium sized spiny neurons (Hassler et al. 1978, Hattori et al. 1979). Evidence suggests that glutamate and possibly aspartate are neurotransmitters at these synapses (Carter 1982, Graybiel and Ragsdale 1983). Activation of the cortico-striate neuron leads to glutamate release with activation of AMPA and NMDA receptors that are found almost exclusively within the synapses (Bernard et al. 1997, Bernard and Bolam 1998). There is considerable regional variation in NMDA receptor subtype (Kuppenbender et al. 2000, Ravenscroft and Brotchie 2000). The initial response of striatal cells following cerebral cortical excitation is similar to that produced by activation of striatal inputs from the intralaminar thalamus, substantia nigra pars compacta, and dorsal raphe nucleus, namely excitation followed by inhibition (Kitai 1981). An action potential is generated if there is sufficient convergent excitatory input onto an individual spiny neuron (Stern et al. 1997, 1998).

Afferent fibres from cerebral cortex, thalamus, and midbrain probably converge on the same spiny striatal neurons (Kitai 1983). The cortical projection seems the largest, the thalamic one smaller, and that from the midbrain smallest (Kemp and Powell 1971[b]). Nonetheless, the input from the midbrain, particularly the dopaminergic terminals from the substantia nigra pars compacta, seems especially important. These fibres form symmetric synaptic contacts with the necks of dendritic spines of the spiny projection neurons, the heads of which invariably seem to receive input from cortico-striatal fibres. They are thus well placed to modulate the excitatory cortical input. This synaptic relationship to the cortico-striatal terminals also applies to other inputs onto spiny neurons, including both cholinergic and GABAergic interneurons (see Bolam et al. 2000 for a review). Cortical input to striatum dominates in that excitation from this source is not blocked by inhibition generated by thalamic and mesencephalic stimulation, whereas inhibition from cortico-striatal fibres abolishes the excitatory influence of pathways from the thalamus and mesencephalon (Hull et al. 1973, Van der Maelen et al. 1978).

Another major projection to the neostriatum comes from the substantia nigra (see later under ‘Efferent connections of the substantia nigra’). The existence of this pathway had been suspected, but it was not until the development of fluorescent histochemistry that it was clearly demonstrated (Carlsson et al. 1962, Falck 1962). It arises chiefly from neurons in the substantia nigra pars compacta, although some cells in the pars lateralis, ventral tegmental area of Tsai, retrorubral region, γ group of Olszewski and Baxter (1954), and non-dopaminergic pars reticulata are probably also involved (Sotelo and Riche 1974, Kocsis and Van der Maelen 1977, Szabo 1977; Oertel et al. 1982). Its axons are fine, unmyelinated, and varicose and 90–95% are dopaminergic (Pritzel et al. 1983). The majority of these fibres pass to the ipsilateral striatum, but approximately 5%, most of which are also dopaminergic, pass to the contralateral neostriatum. Some of these contralateral fibres arise as collaterals from the ipsilateral projection. These contralateral fibres cross in the inferior thalamic peduncle and the massa intermedia (Pritzel et al. 1983). The ipsilateral projection is topographically arranged so that cells distributed lateral to medial in the pars compacta and ventral tegmental area of Tsai have terminations arranged dorsal to ventral in the putamen. Caudally arising cells project mainly to the putamen. The strio-nigral projection has a similar pattern of organization (Szabo 1962, Szabo 1967, Szabo 1970; see later under ‘Efferent connections of the neostriatum’). The nigro-striatal fibres to the caudate nucleus and putamen arise from two largely different cell populations, although there are a small number of neurons which send collaterals to both. The neurons are grouped in separate small clusters, which project to either the caudate nucleus or the putamen. These groups are intermingled like a mosaic (Parent et al. 1983[a]). The dorsal tier neurons and those in the retrorubral nucleus, ventral tegmental area, and γ area of the substantia nigra (Olszewski and Baxter 1954) probably project to the matrix in humans and other mammals, while the ventral tier and the pars reticulata innervate the striosomes (Gerfen et al. 1987, Jimenez-Costellanos and Graybiel 1987, Kiyama et al. 1987, Gibb 1992, Hanley and Bolam 1997, Smith and Kieval 2000). The former are more numerous, thinner, and have fewer varicosities. In addition, in primates the dorso-lateral or ‘sensorimotor’ striatum tends to receive fibres from neurons that are located almost exclusively in the ventral tier of the pars compacta, including from the projections that interdigitate with the pars reticulata. The ventral or ‘limbic’ striatum, however, is well innervated from the dorsal and, to a lesser extent, the ventral tier of the pars compacta, as well as from the ventral tegmental area (Lynd-Balta and Haber 1994[a&b]). The head and body of the caudate nucleus or ‘associative’ striatum receives inputs from a range of dopaminergic neurons, but especially from the denso-cellular part of the ventral compacta (Haber and Fudge 1997, Francois et al. 1999). A small number of nigro-striatal neurons have collaterals which pass to both segments of the globus pallidus and subthalamic nucleus (see later under ‘Afferent connections of the globus pallidus’ and ‘Afferent connections of the subthalamic nucleus’), where they arborize profusely. The striatal branching of these collaterals is very restricted, compared with the profuse arborization of the nigro-striatal axons, which do not have collaterals (Parent et al. 2000).

The nigro-striatal fibres pass rostrally along the dorso-medial border of the substantia nigra into Forel's field H just lateral to the hypothalamus (see under ‘Efferent connections of the globus pallidus’). Fibres to the posterior neostriatum leave the main bundle in Forel's field and turn laterally dorsal to the subthalamic nucleus to pass through the posterior internal capsule and globus pallidus. Fibres to the more rostral neostriatum continue forwards before penetrating the internal capsule in a similar way. Within the neostriatum they form symmetric synapses mainly on the necks of dendritic spines of projection neurons (Kemp 1968, Fox et al. 1971/1972, Freund et al. 1984, Smith and Kieval 2000) and release dopamine as their neurotransmitter. They may thus be in a position to modulate the effect of input from the cerebral cortex, which is largely on the heads of the spines. There is strong convergence between the nigro-striatal and cortico-striatal projection onto the same neostriatal neurons (Smith and Kieval 2000). Although evidence is conflicting, most studies suggest that their initial effect on striatal cells is excitatory, although this is followed by inhibition. The inhibitory phase may be due to collaterals exerting a secondary effect (Kitai 1981, 1983). Other studies have suggested that two populations of fibres might be responsible for these responses: a fast-conducting (2–10 m/s) non-dopaminergic projection from the pars reticulata and a slow-conducting (0.3–1 m/s) dopaminergic projection from the pars compacta (Deniau et al. 1976, Guyenet and Aghajanian 1978). This remains unresolved.

The highest concentration of dopamine receptors (D1 and D2) in the human brain is found in the caudate and putamen and the next highest is in the nucleus accumbens (De Keyser et al. 1988). As mentioned above, the D1 and D2 receptors are on medium sized spiny projection neurons and they are spread throughout both the dorsal and the ventral neostriatum. Their mRNAs tend to be segregated into the two main groups of striatal projection neurons, namely those containing substance P, which project to the substantia nigra and contain high levels of D1 receptors, and the enkephalin-immunoreactive striato-pallidal neurons, which contain D2 receptors (Gerfen et al. 1990). D3 receptors are restricted to the ventral limbic-related striatum, where they are partly co-localized with D1 and D2 receptors on striatal projection neurons (Lemoine and Bloch 1996). Cholinergic interneurons have high levels of D2 and D5 receptors (Lemoine and Bloch 1990, Bergson et al. 1995). While D5 and D4 receptors have also been reported on striato-fugal neurons, their targets are uncertain (Bergson et al. 1995, Mrzljak et al. 1996, Defagot et al. 1997). Dopamine receptors are widely distributed on neurons at both synaptic and non-synaptic sites and the latter may be more numerous, at least in some species (Caille et al. 1996, Descarries et al. 1996). This suggests that much of the dopamine derived from the nigro-striatal projections may act through ‘volume transmission’ and function as a neuromodulator rather than a conventional neurotransmitter.

Acetylcholinesterase is present in the nigro-striatal system and stimulation of substantia nigra produces an increase in nigral and a decrease in striatal acetylcholinesterase release ipsilaterally (Greenfield et al. 1980).

Cholecystokinin is found in some neurons of the ventral tegmental area and substantia nigra. It may co-exist in dopaminergic fibres projecting to the neostriatum, amygdala, accumbens, and bed nucleus of the stria terminalis (Graybiel and Ragsdale 1983) and it has been suggested that both may act as neurotransmitters but utilize different types of synapses (Hattori et al. 1991). Some of these axons may also contain neurotensin (Seroogy et al. 1988). The possibility of serotonergic and GABAergic fibres in the nigro-striatal projection has also been raised (Graybiel and Ragsdale 1983). In rats the latter fibres pass only to the matrix and are non-dopaminergic (Hanley and Bolam 1997). Some pars lateralis pars reticulata neurons project to both the neostriatum and the superior colliculus (see under ‘Efferent connections of the substantia nigra’) and in some species such pars lateralis neurons might also send fibres to the thalamus (Takada 1992). Nigro-striatal fibres from the pars compacta may send collaterals to the cingulate cortex in some species (Takada and Toshiaki 1986).

Some thalamic nuclei project to the neostriatum and their axons penetrate through the internal capsule. Most fibres arise from the intralaminar nuclei, but there are also some from the ventro-anterior, ventro-lateral, and dorso-medial nuclei (Royce 1983). The centro-median and parafascicular nuclei are the major source of afferents, but the central medial, central lateral, and paracentral nuclei also contribute. There is probably considerable species variation and although these connections have been described in animals, the extent to which they parallel the human situation is uncertain, although a thalamo-striatal connection has been demonstrated in man (McLardy 1948, Simma 1951).

Although the centro-median nucleus in the cat projects to both caudate nucleus and putamen, in the monkey it may project virtually entirely to the latter and shows a high degree of topographical organization (Kalil 1978, Nakano et al. 1990). Asymmetric synapses are formed with striatal interneurons which are immunoreactive for AChE, somatostatin, and parvalbumin (Sidibe and Smith 1999). In the monkey the ventro-anterior nucleus has its major connection with the caudate, although additionally sending fibres to the putamen. The ventro-lateral nucleus, however, projects almost exclusively to the putamen (Nakano et al. 1990). The fibres from these two ventral nuclei pass to the dorsal or ‘sensorimotor’ striatum where they synapse with neurons that also have direct input from motor cortical regions, including primary motor, premotor, supplementary motor, and cingulate motor areas. As the ventro-anterior and ventro-lateral nuclei also have specific connections with the same cortical areas it seems likely that the thalamo-striatal input modulates the cortico-striato input (Gimenez-Amaya et al. 2000, McFarland and Haber 2000, 2001). In the cat, thalamic nuclei which connect with frontal and limbic cortices project mainly to medial caudate and accumbens nuclei, while fibres from thalamic nuclei with major motor connections pass mainly into lateral caudate (Jayaraman 1985). The zones of termination tend to be focused and only a small proportion of thalamic cells projecting to the caudate send fibres to more than one region of the nucleus (De las Heras et al. 1999).

Work in the rat suggests that about three-quarters of the axons from the parafascicular nucleus synapse with dendrites and the majority of the remainder with spines. Only a very small number appear to act on parvalbumin-immunoreactive dendrites through asymmetric synapses (Rudkin and Sadikot 1999). These parafascicular-striatal fibres seem to play a critical role in D1-mediated stimulated acetylcholine release in the dorsal striatum via NMDA receptors (Consolo et al. 1996). The accumbens receives fibres from the paraventricular thalamic nucleus, which is possibly involved in circadian timing and sends efferents to limbic-associated areas of cerebral cortex and amygdaloid nucleus (Berendse and Groenewegen 1990, 1991, Moga et al. 1995). The neostriatal terminations group in small mosaics similar to those seen with the cortico-striatal projection. Those from the parafascicular nucleus, which has received afferents from the premotor cortex, end in the matrix and although the input from the centro-median nucleus, which has innervation from the motor cortex, is more complex, it has similarities (Sadikot et al. 1992, Flaherty and Graybiel 1994, Deschenes et al. 1996). The lateral dorsal nucleus also sends fibres to the matrix where they appear to make direct contact with the striato-fugal neurons projecting to the globus pallidus and substantia nigra (Funaki et al. 1998). The midline thalamic nuclei tend to project to the striosomes. Double labelling and electrophysiological techniques have shown that some of these thalamic neurons not only ramify extensively within the neostriatum, but also send collaterals to the motor cortex, medial segment of the globus pallidus, subthalamic nucleus, and region of the red nucleus (Jones and Leavitt 1974, Jinnai and Matsuda 1981, Moran 1982, Royce 1983, Deschenes et al. 1996). Some thalamo-striatal fibres also have collaterals which terminate in other thalamic nuclei (Nguyen Legros et al. 1982, Deschenes et al. 1996). The thalamus receives major motor input from the globus pallidus, as described below under ‘Efferent connections of the globus pallidus’. In addition, thalamo-striatal neurons receive input from a number of other structures, including the substantia nigra (De las Heras et al. 1998), superior colliculus (Ichinohe and Shoumura 1998), and the pedunculo-pontine nucleus (Erro et al. 1999). This suggests that thalamo-striato fibres have feed-back circuits from other parts of the basal ganglia. Firing of the appropriate thalamic neurons results in excitation or excitation followed by inhibitation in the caudate nucleus (Purpura and Malliani 1967, Kocsis et al. 1976, Van der Maelen and Kitai 1980). Stimulation of some thalamic areas causes a ‘recruiting response’ in the neostriatum similar to that produced by the collaterals to the central cortex (Jinnai and Matsuda 1981). The main neurotransmitter involved in the thalamo-striatal pathway is probably glutamate.

A direct connection from the subthalamic nucleus to the caudate and putamen has been shown in monkeys, with the larger part going from the dorso-medial part of the subthalamic nucleus to the putamen (Nakano et al. 1990). About 17% of subthalamic neurons contribute to this projection, the fibres of which do not seem to have collaterals (Sato et al. 2000[a]).

The pedunculo-pontine nucleus projects to the striatum bilaterally in primates, but the ipsilateral connection is the major one (Nakano et al. 1990). Most ascending fibres from the pedunculo-pontine nucleus, however, pass to structures related to the striatum, including the substantia nigra, globus pallidus, subthalamic nucleus, and most thalamic nuclei (Inglis and Winn 1995, Lee et al. 2000), rather than terminating in it directly.

The dorsal raphe nucleus projects to the neostriatum. Both serotonergic and non-serotonergic fibres are involved. Although most are ipsilateral, a few terminate contralaterally (Miller et al. 1975, Bobillier et al. 1976; Steinbusch et al. 1980). Serotonergic fibres are abundantly distributed through the neostriatum, but density is especially high caudally in the putamen adjacent to the globus pallidus (Okumura et al. 2000) and in the ventro-caudal striatum which shows the highest concentration of serotonin (Bobillier et al. 1975, Ternaux 1977). The striatal response to stimulation of this pathway may be predominantly excitatory and it has been shown to reduce sensory evoked responses in the caudate nucleus (Kitai 1981, Andersen and Dafny 1982, Kitai 1983). Raphe-striatal fibres may provide collaterals to the substantia nigra (van der Kooy and Hattori 1980). The median raphe nucleus, at least in the cat, does not contribute fibres to the striatum (Vertes et al. 1999).

The locus coeruleus has been said to provide noradrenaline to neostriatum and nucleus accumbens via the dorsal noradrenergic bundle. The latter recipient seems to be the most definite (Berridge et al. 1997). Cell groups caudal to this project to the same structures via the ventral noradrenergic bundle (Speciale et al. 1978, O’Donohue et al. 1979). There is a discrepancy, however, between moderate or high levels of alpha- and beta- noradrenergic receptors in the striatum and very low levels of noradrenaline (Graybiel and Ragsdale 1983).

It has been shown that there is an extensive input into the neostriatum from limbic structures in animals, although details of similar connections in humans are uncertain (Nauta 1982, Jayaraman 1984). The cingulate cortex, hippocampus, and amygdaloid complex have direct connections. Amygdaloid fibres reach the accumbens nucleus via the striae terminalis, but there is also a large direct projection from amygdala to caudate and putamen (Kelly et al. 1982). A small contralateral component to this projection crosses in the anterior commissure. The limbic input is to most of the neostriatum apart from the lateral antero-dorsal segment, which in the rat is the zone of termination of the cortico-striatal fibres from the sensory-motor cortex. Fibres from the raphe nuclei, prefrontal cortex, substantia nigra, and ventral tegmental area of Tsai also avoid this lateral antero-dorsal area. Similar segregation of striatal inputs into medial ‘limbic’ and lateral sensory-motor compartments have been described in the cat (Jayaraman 1984). In the primate the limbic system is largely connected with the ventral striatum, which includes accumbens, ventromedial caudate, and the adjacent ventromedial putamen (Poletti and Cresswell 1977, Van Hoesen et al. 1981, Russchen et al. 1985, Haber et al. 1990). The accumbens gets afferent input from the prefrontal cortex, hippocampus, amygdala, and ventral pallidum (Groenewegen 2001). The input from the amygdala into the nucleus accumbens produces excitation, which seems to be suppressed by the action of dopaminergic and non-dopaminergic input from the substantia nigra and ventral tegmental area (Yim and Mogenson 1982). Thus, as well as having these direct afferents, the limbic system influences the neostriatum indirectly via connections through the midline thalamus, substantia nigra, ventral tegmental area, and raphe nuclei (Nauta 1982, Jayaraman 1984, Groenewegen 2001).

A cholecystokinin-containing projection from claustrum to neostriatum has been proposed, but a cortical origin of the pathway has not been excluded (Krettek and Price 1978, Meyer et al. 1982).

Early descriptions of projection from the globus pallidus to the neostriatum (Nauta 1979, Staines et al. 1981) lead to the suggestion that these fibres may really be destined for cerebral cortex (Graybiel and Ragsdale 1983). Subsequent studies, however, confirmed its existence in several species (Beckstead 1983, Parent et al. 1983[a], Kita et al. 1991). Co-distribution of pallido-nigral and pallido-striatal neurons in the external segment of the globus pallidus resulted in the idea that the two projections might be co-localized (Staines and Fibiger 1984, Parent 1986). Subsequent work has shown that about one-quarter of neurons in the external globus pallidus provide collaterals which have been said to selectively innervate striatal interneurons (Bevan et al. 1998). The other collateral fibres pass not only to the substantia nigra, but also to the subthalamic nucleus and the internal globus pallidus (Bolam et al. 2000). The pallido-striatal fibres are topographically organized (Spooren et al. 1996), terminate in a patchy way, and are distributed mainly to the matrix, where the majority synapse on parvalbumin-positive GABA-containing interneurons. In addition, a smaller proportion contact neuropeptide Y and NADPH-diaphorase-containing interneurons (Staines and Hincke 1991, Rajakumar 1994). Synapses are primarily in the proximal regions of neurons, and the pallido-striatal pathway is thus in a powerful position to be able to influence the activity of these interneurons, which themselves seem capable of controlling the GABAergic spiny striatal projection neurons (Bolam et al. 2000). Some authors, however, have claimed there are also direct monosynaptic connections between pallido-striatal terminals and striatal neurons which send axons to the external segment of the pallidum (Williams and Faull 2001).

Electrophysiological studies have shown that neostriatal cells respond to various forms of sensory input, including somatosensory, vestibular, and auditory stimuli (Potegal et al. 1971, Matsurami and Cohen 1975, Andersen and Dafny 1982, Richards and Taylor 1982). The somatosensory input seems to be mainly, but not entirely, nociceptive and is somatotopically arranged, with the major projection being contralateral (Munez et al. 1976, Pazo and Medina 1982, Richards and Taylor 1982). The pathways by which these inputs reach the neostriatum are uncertain.

Electrophysiological evidence suggests that there may be direct connections between corresponding positions in the two caudate nuclei. Only the head and body seem to be involved and connections are via the corpus callosum (Medina and Pazo 1981, Pazo and Medina 1982).

The strio-pallidal system, along with the strio-nigral system, forms the major output of the neostriatum. Fibres arise from the medium sized spiny neurons (see earlier under ‘Neostriatum – structure’). The axons are thin and sparsely myelinated (Adinolfi and Pappas 1968, Fox et al. 1975). Those from the caudate nucleus pass ventrally through the internal capsule into the dorsal part of the globus pallidus, while those from the putamen radiate medially to the ventral globus pallidus. The fibres are arranged in an orderly manner, like the spokes of a wheel, and topographical relationship is thus maintained. The caudate nucleus and putamen both project to the external and internal pallidal segments. Fibres from the neostriatum have two medio-lateral zones of termination within the globus pallidus. The first lies in the lateral part immediately adjacent to the neostriatum and has a two-dimensional topographical organization. The second lies medially and extends throughout the rest of the globus pallidus. This has a three-dimensional topographical organization. Axons from striatal projection neurons are highly co-lateralized. They provide collaterals close to their somata in the neostriatum (Preston et al. 1980, Kitai 1981, Wilson and Phelan 1982; see earlier under ‘Neostriatum – structure’) and some of these seem to synapse with the dendrites of other spiny neurons (Oorschot et al. 2001).

Although virtually all striato-fugal fibres terminate in the external globus pallidus, at least in the rat, the majority also send fibres to the medial globus pallidus and/or the substantia nigra pars reticulata (Fox et al. 1975, Fox and Rafols 1976, Preston et al. 1980, Kawaguchi et al. 1990, Parent et al. 1995, Wu et al. 2000). Some of the pallidal neurons contacted by striato-pallidal projection in turn send fibres to the nigra (Smith and Bolam 1991; see also under ‘Substantia nigra’). Collaterals to different target areas from the same neostriatal cells have been claimed to be less important in higher mammals than in the rat (Staines and Fibiger 1984, Loopijt and Van der Kooy 1985, Beckstead and Cruz 1986, Feger and Crossman 1994). Overall, the putamen tends to project more strongly to the globus pallidus than the substantia nigra pars reticularis, whereas the converse applies to the caudate, including its eye movement zone (see Flaherty and Graybiel 1994). The part of the putamen which lies rostral to the anterior commissure may project only to the globus pallidus while more caudal parts may supply the globus pallidus and substantia nigra (Szabo 1967). A projection from the accumbens nucleus to the entopeduncular nucleus, which is the homologue of the internal segment of the globus pallidus in rodents and cats (Fox and Schmitt 1944), terminates in its antero-ventral part (Williams et al. 1977, Pacitti et al. 1982, Mogenson et al. 1983). In the rat, fibres from the nucleus accumbens also pass to subpallidal structures including the substantia innominata, lateral preoptic, and lateral hypothalamic areas (Mogenson et al. 1983). Although the existence of such connections in primates was initially questioned (Powell and Lenman 1976), subsequent studies suggest that the ‘limbic’ or ventral striatum, which includes the accumbens, projects to what might be called the ‘limbic’ or ventral pallidum. The latter area involves the rostral globus pallidus and the entire region ventral to the anterior commissure (Harber et al. 1990). This is topographically organized (Klitenick et al. 1992). To a large extent the subpallidal and ventral pallidal areas are synonymous. The accumbens nucleus also projects to the substantia nigra, ventral tegmental area, and caudal mesencephalic areas (Groenewegen 2001). These fibres may be collaterals of axons to the globus pallidus. These projections from the neostriatum and accumbens nucleus to the globus pallidus and substantia nigra use gamma-aminobutyric acid (GABA) as a neurotransmitter (Pycock and Horton 1976, Pycock et al. 1976, Kitai 1981). Most evidence suggests they have an inhibitory effect (Kitai 1981, Kitai 1983, Mogenson et al. 1983, Scarnati et al. 1983).

Using immunohistochemistry, fibres containing substance P, dynorphin, and enkephalin can be seen in globus pallidus, where they ensheath pallidal dendrites in a striking pattern. These are termed ‘woolly fibres’ and extend into the subpallidal region, the nucleus accumbens, and olfactory tubercle (Haber and Nauta 1983, Beach and McGeer 1984, Haber and Watson 1985). As mentioned above, these probably arise in the neostriatum and enkephalin terminals predominate in the external segment of the globus pallidus, while substance P and dynorphin terminals predominate in the internal segment. Only a small proportion of neostriatal enkephalin positive cells seem to project into the globus pallidus (Brann and Emson 1980). In primates the substance P-containing area of the ventral pallidum is restricted to a small area beneath the anterior commissure (Haber and Elde 1981).

Strio-pallidal fibres probably conduct slowly, in the order of 0.8–1.5 m/s (Kitai 1983). The rich collateral plexus adjacent to the soma may produce inhibition of striatal medium spiny neurons by way of GABA receptors (Park et al. 1980).

The second major output from the neostriatum is to the substantia nigra. The cells of origin are the small and medium sized spiny neurons. As mentioned above, many, if not all, axons projecting to the substantia nigra arise as collaterals from those passing to the globus pallidus (external segment or external and internal segments) and, as might be expected, rates of conduction in these two fibres are similar (Fox et al. 1975, Fox and Rafols 1976, Kitai 1981, 1983). Strio-nigral axons pass caudally through the posterior limb of the internal capsule and lie dorsal to the cerebral peduncle. They terminate mainly in the substantia nigra pars reticulata, in a precisely ordered topographical arrangement (Domesick 1977, Gerfen 1985; see previously under ‘Neostriatum – structure’). The axon terminals contain large pleomorphic vesicles and establish symmetrical synapses (Rinvik and Grovofa 1970, Fox and Rafols 1976, Oka and Jinnai 1978). There are also fibres that topographically form symmetric synapses with dopaminergic neurons and proximal dendrites of the pars compacta and these may be derived mainly from the ventral striatum, which has limbic connections (Bolam and Smith 1990, Flaherty and Graybiel 1994, Joel and Weiner 2000). The shell of the accumbens nucleus tends to project to the ventral tegmental area, while the core is more directed to the pars compacta (Klitenick et al. 1992). The projection from the dorso-lateral or sensorimotor-related striatum innervates a limited region of the ventro-lateral nigra and avoids the dorsal pars compacta in the primate (Lynd-Balta and Haber 1994[c]). For further details see later under ‘Afferent connections of the substantia nigra’.

As mentioned above, it seems that this system is GABAergic and, like the strio-pallidal pathway, its action is probably inhibitory (Fonnum et al. 1974, Ribak et al. 1976, Kitai 1983). In addition it is possible that some strio-nigral fibres use substance P as a transmitter (Brownstein et al. 1977, Hong et al. 1977, Kanazawa et al. 1977, Bolam et al. 1983[b], Bolam and Smith 1990) and others may even use the opioid peptide dynorphin (Vincent et al. 1982). There is also evidence to support the existence of an excitatory striato-nigral projection to the pars reticulata (Rodriguez et al. 2000). It is uncertain what proportion of neostriatal neurons project to the substantia nigra and estimates in the rat vary from 20 to 70% (Bolam et al. 1981[b], Henderson 1981).

The strio-nigral pathway may serve as part of a negative feed-back loop controlling the activity in the dopaminergic neostriatal system. Parts of it may be related to the other motor functions of the substantia nigra pars reticulata, including the control of saccadic eye movements (see later under ‘Efferent connections of the substantia nigra’).

A projection from the neostriatum to the cerebral cortex has been described and passes mainly to the auditory and adjacent ‘associative’ cortex. The cells of origin are large and probably cholinergic (Jayaraman 1980, Parent et al. 1981, Oleshko and Maisky 1993). The significance of this is uncertain.

In man the external and internal segments of the globus pallidus lie in juxtaposition, separated only by the medial medullary lamina. In rats and cats the internal segment is separated from the external by part of the internal capsule and is termed the entopeduncular nucleus (Fox and Schmitt 1944). This is important when considering information about the connections of the globus pallidus derived from these animals. The cellular structure of the globus pallidus is quite different from that of the neostriatum. The external and internal pallidal segments are, however, very similar histologically, although the biggest cells in the internal segment may be slightly larger (Mettler 1968). The neurons are large and fusiform, being 25–40 µm wide and approximately double this in length. Unlike the neostriatum, small cells are relatively infrequent. The dendritic trees are made of thick, smooth fibres, which may end in complex aborizations or simple thin, beaded processes. The complex endings branch in a discoidal zone, which lies parallel to the medullary laminae. Neostriatal axons perpendicularly traverse the complex endings and make synaptic contact with them. Both complex endings and thin processes contact dendrites and soma of the other large cells, which may allow communication between large parallel neurons (Francois et al. 1984, Yelnik et al. 1984; see later under ‘Substantia nigra’). The pallidal neurons tend to be orientated radially parallel to the neostriatal fibres (Fox et al. 1966, Mettler 1968). The dendritic trees of the pallidal neurons are very large, so that each is contacted by many striato-pallidal axons. This produces a major convergence of input (Percheron et al. 1984). The axons of these pallidal cells divide early into numerous collaterals and enter bundles of adjacent fibres of passage.

In addition there are less frequent small cells (approximately 12 µm in diameter), which have relatively short dendrites and short sparsely branching axons. These have a small number of synapses and are probably interneurons (DiFigila et al. 1982). The pallidal projection neurons are considered to be GABAergic in nature. Approximately two-thirds of the globus pallidus neurons in the rat are parvalbumin-immunoreactive and their somata tend to be larger than those which stain negatively. Both of these types of cells, however, seem to receive similar synaptic inputs from outside of the pallidum and they are connected through local axon collaterals (Kita 1994).

Reciprocal connections between internal (entopeduncular nucleus) and external segments have been identified (Smith and Bolam 1990, Fink-Jensen and Mikkelsen 1991, Hazrati et al. 1991, Kincaid et al. 1991). In the rat such input from the globus pallidus (external segment) to the entopeduncular nucleus (internal segment) synapses on the same neurons that receive terminals from the striatum (Bolam and Smith 1992, Bolam et al. 1993). Although this situation has not been explored in the same detail in the primate, it has been demonstrated that in monkeys, external pallidal neurons give rise to large GABA-containing terminals that form symmetric synapses with perikarya and proximal dendrites, often in a basket-like manner. They are thus in an excellent position to modulate the excitatory influence generated more distally in the dendritic tree of these same internal segment neurons by synapses from the subthalamic neurons (Smith et al. 1994, Smith et al. 1998; see later). In fact, all external pallidal projection neurons seem to give rise to collaterals which pass to the internal segment, subthalamic nucleus, and substantia nigra, as well as providing local axon collaterals within the external segment (Bolam et al. 2000). As mentioned above, the connections and functions of the ventral part of the globus pallidus are somewhat different from the dorsal part, with input particularly from ventral striatal areas, including the accumbens nucleus, and output being particularly directed to the medio-dorsal thalamus. This ‘limbic’-associated ventral pallidum is thus somewhat distinct from the dorsal pallidum, which mainly has sensorimotor and associative functions (Bolam et al. 2000, Groenewegen 2001). In monkeys the dorsal one-third of the internal segment of the globus pallidus, which receives input preferentially from the head and body of the caudate nucleus, has been regarded as the ‘associative striatum’, while the ventro-lateral two-thirds, which receive input chiefly from the putamen, have been termed the sensorimotor striatum (Shink et al. 1997).

Emphasis in the current discussion particularly relates to the dorsal connections of the globus pallidus, but relevant material related to the ventral pallidum is also mentioned. These systems are closely integrated and individual neurons in the globus pallidus interna, substantia nigra pars reticulata, and compacta and subthalamic nucleus receive convergent afferent input from both the dorsal and ventral parts of the external globus pallidus (Bolam et al. 2000).

The main afferent connection of the globus pallidus is from the neostriatum (as mentioned above and in the section on ‘Efferent connections of the neostriatum’).

After the neostriatum, this is the most important afferent connection of the globus pallidus. The axons pass ventro-laterally from the subthalamic nucleus penetrating the adjacent internal capsule and entering both external and internal pallidal segments via their dorso-medial surface (Nauta and Cole 1974, 1978, Carpenter et al. 1981, Parent 1990). These fibres may be collaterals of the subthalamo-nigral projection (Hammond and Yelnik 1983, Hammond et al. 1983[b]), but in the monkey almost half the subthalamic neurons project to both external and internal pallidal segments and provide axons to no other structure. Another 11% project only to the external pallidum (Sato et al. 2000[a]). There is a high degree of convergence of striatal and subthalamic input upon single pallidal cells (Hazrati and Parent 1992). In the rat some entopeduncular (internal segment) neurons which receive such dual innervation in turn project to the thalamus (Bevan et al. 1994[a]). In the monkey, fibres from subthalamic nucleus to internal pallidum form asymmetric synapses on dendrites distal to the input from the external segment (Smith et al. 1994). The subthalamo-pallidal pathway is probably excitatory and possibly glutaminergic (Nakanishi et al. 1987, Smith and Parent 1988).

A dopamine projection from the substantia nigra and ventral tegmental area to the globus pallidus has been demonstrated and terminals of this also end in the neostriatum. These fibres pass to both the internal and external pallidal segments as well as the subthalamic nucleus. The arborization in the neostriatum of such axons is relatively limited. This pathway is also substantially smaller than that which goes directly from the substantia nigra to the neostriatum without collaterals (Gerfen et al. 1982, Smith et al. 1989, Schneider and Dacko 1991, Parent et al. 2000). The nigro-pallidal fibres entering the globus pallidus by way of the ansa lenticularis and the lenticular fasciculus may be mainly distributed to the internal segment (Lavoie et al. 1989, Besson et al. 1990, Smith and Kieval 2000), although presumptive dopaminergic terminals are present in both external and internal pallidal segments in humans (Hedreen 1999). The ventral pallidum receives substantial input from dopaminergic neurons of the ventral tegmental area (Klitenick et al. 1992, Napier and Maslowski-Cobuzzi 1994).

Several other afferent projections to the globus pallidus have been described, but their significance in humans remains uncertain. These include fibres from the cerebral cortex (Mettler 1945, DeVito and Smith 1964, Naito and Kita 1994), thalamus (Mettler 1945,  Johnson 1961, Kincaid et al. 1991), central amygdaloid nucleus (Shinonaga et al. 1992), pedunculo-pontine nucleus (Charara and Parent 1994), dorsal raphe nucleus (Charara and Parent 1994, Flaherty and Graybiel 1994), and basal forebrain cholinergic neurons (Mesulam et al. 1984, Charara and Parent 1994).

The major efferent output of the globus pallidus is to the thalamus. This projection arises only from the internal pallidal segment and separate fibre bundles arise from its ventral and dorso-medial surfaces. The ventral bundle emerges from the lateral part of the internal segment and passes medially through the dorsal substantia innominata where it mingles with the cells of the basal nucleus of Meynert. This group of axons comprise the ansa lenticularis which forms a hook curving medially and dorsally around the posterior limb of the internal capsule to enter Forel's field H.

The fibres arising from the dorso-medial surface of the internal segment of the globus pallidus have their soma located closer towards its apex. They emerge medial and caudal to those forming the ansa lenticularis and group in a bundle known as the lenticular fasciculus. These fibres pass medially and most lie rostral to the subthalamic nucleus. A few have to arch around the dorso-lateral aspect of the rostral tip of the subthalamic nucleus. In this situation they penetrate the internal capsule and lie between the subthalamic nucleus and the zona incerta. These fibres of the lenticular fasciculus are known as Forel's field H2. They pass ventro-medially and curve around the medial edge of the zona incerta to join with the ansa lenticularis and form the thalamic fasciculus, which is known as Forel's field H1. The thalamic fasciculus contains other fibres including cerebello-thalamic ones. It lies ventral to the thalamus and its pallido-thalamic fibres move rostrally and dorsally to project to some thalamic nuclei. Fibres enter into the ventral anterior, parvocellular part (VApc), ventro-lateral (VL), centro-median (CM), and parafascicular (Pf) nuclei.

There has been debate over which subdivisions of the ventro-lateral nucleus are involved. VLo (pars oralis) and VLm (pars medialis) have been shown to receive projections by some workers (Nauta and Mehler 1966, Kuo and Carpenter 1973, Kim et al. 1976) but VLc and not VLo have been found to be terminal zones by others (DeVito and Anderson 1982). More recent studies confirm that VLo is a major recipient (Sakai et al. 2000). Some authors have also found the densocellular part of the ventral anterior nucleus (VAdc) to be a terminal zone (Ilinsky and Kultas-Ilinsky 1997, Kultas-Ilinsky et al. 1997). The adjacent lateral habenular nucleus also receives some fibres (Parent and De Bellefueille 1982). The pallido-thalamic projection is topographically arranged with rostro-caudal, dorso-ventral, and possibly medio-lateral correspondence between pallidal and thalamic areas (Kuo and Carpenter 1973, Kim et al. 1976). Some of these axons appear to have collateral fibres passing to both the ventro-anterior–ventro-lateral nuclear complex and the centro-median nucleus, or to extrathalamic sites such as the pedunculo-pontine nucleus (Harnois and Filion 1982, Parent et al. 1999). There is also a smaller projection to the contralateral ventro-anterior, ventro-lateral, and centro-median nuclei and pedunculo-pontine nuclei (Nakano et al. 1983, Hazrati and Parent 1991, Parent et al. 1999). The thalamic regions of termination of the pallidal fibres especially project to the supplementary motor area (Sakai et al. 1999, 2000). The areas of termination of the pallidal fibres are largely distinct from those of the cerebellum and substantia nigra (see later under ‘Afferent connections of the cerebellum’), although there is some overlap (Sakai and Patton 1993).

It has been suggested that the internal pallidal segment has two distinct sites of origin of efferent fibres. One, a central ‘motor’ zone, sends axons to the thalamus and pedunculo-pontine nucleus as mentioned above, and the other, a peripheral ‘limbic’ zone, sends axons to the lateral hypothalamus and habenular (Parent and De Bellefeuille 1982). This concept has been extended to distinguish the globus pallidus proper, which projects mainly via the ventral thalamic nuclei to the motor cortices, from the ventral pallidum. The latter overlaps with the substantia innominata and receives input from ‘limbic’ striatum and other forebrain limbic structures, while projecting to prefrontal cortex via the dorso-medial thalamic nucleus (Graybiel 1986). This thalamic zone of termination in primates, however, is relatively sparse and the ventral or ‘limbic’ pallidum sends its major efferents to the subthalamic nucleus, lateral hypothalamus, lateral habenular nucleus, and substantia nigra (see later) (Haber et al. 1993). This concept is largely derived from animal work and its relevance to the human brain is uncertain. The pallido-thalamic fibres have been postulated to be excitatory, inhibitory, or produce both these effects (Kitai 1983). It seems likely, however, that the pallido-thalamic pathway is inhibitory and that its neurotransmitter is GABA (Graybiel 1986, Parent 1986), although it has been claimed to be glutamatergic and excitatory (Finkelstein et al. 2001).

The external division of the globus pallidus gives rise to axons which sweep dorso-medially through the internal capsule and terminate in the subthalamic nucleus. These have medio-lateral and dorso-ventral topographical organization and lie somewhat caudal to the ansa lenticularis (Carpenter et al. 1968, Grofova 1969). The projection from the ventral pallidum, mentioned above, is also topographically organized (Haber et al. 1993). Together with the reciprocally projecting subthalamo-pallidal fibres they form a bundle called the subthalamic fasciculus. In primates all pallido-subthalamic fibres have collaterals and the majority go to the substantia nigra pars reticulata. Smaller numbers go to the reticula and globus pallidus interna or to the interna alone (Sato 2000[b]). The terminal large synaptic boutons have a basket-like distribution across the soma and proximal dendritic trees of the subthalamic neurons. Pallido-subthalamic fibres in this fasciculus may be GABAergic (Fonnum et al. 1978, Nauta and Cuenod 1982, Smith and Parent 1988). The pallido-subthalamic fibres have been claimed to mediate both inhibition and excitation (Frigyesi and Rabin 1971, Tsubokawa and Sutin 1972, Ohye et al. 1976), but it seems likely they are inhibitory. Excitation in sub-thalamic nucleus probably arises when the pallido-subthalamic projection is itself inhibited by the action of the striatal input to the external globus pallidus.

Fibres from the globus pallidus project to the substantia nigra. These were initially thought to come from the internal segment and go to the pars compacta but not the pars reticulata (Grofova 1975, 1979, Kim et al. 1976, Tulloch et al. 1978, Gerfen 1982). It is now known, however, that the pars reticulata receives substantial pallidal input, which seems to come mainly, but not exclusively, from the external segment (Smith and Bolam 1989, Smith and Bolam 1991, Hay-Schmidt and Mikkelsen 1992). Well over half of the projection neurons from the external pallidum go to the substantia nigra reticulata in the monkey, although, as mentioned above, they all have collaterals to the subthalamic nucleus alone or to this structure and the globus pallidus interna (Sato et al. 2000[b]). The terminal synaptic arrangement in the substantia nigra pars reticulata is similar to that of the external pallidal axons in the internal pallidum. The projection from the ventral pallidum in the monkey goes to both reticulata and compacta and does not seem to be topographically organized (Haber et al. 1993). The pallido-nigral pathway is probably GABAergic (Hattori et al. 1973[b]), inhibitory (Scarnati et al. 1983) and terminates on dopaminergic (Hattori et al. 1975) or nigro-collicular neurons (Smith and Bolam 1989, Smith and Bolam 1991).

This group of fibres arises from the internal pallidal segment and passes caudally, lying dorso-medial to the subthalamic nucleus and ventro-lateral to the red nucleus. It then moves dorsally to the pedunculo-pontine nucleus and parabrachial area (Nauta and Mehler 1966, Carpenter and Strominger 1967, DeVito and Anderson 1982, Parent and De Bellefeuille 1982, Shink et al. 1997, Parent et al. 1999). As mentioned above, pallido-pedunculo-pontine fibres are collaterals of pallido-thalamic ones. As they are thicker than the latter, however, the pedunculo-pontine nucleus may receive input from the internal pallidum before the thalamus (Parent et al. 2000). The ‘associative’, ‘sensorimotor’, and ‘limbic’ areas in the internal pallidum all project to the pedunculo-pontine nucleus, where they contact the non-cholinergic neurons of the pars dissipata, avoiding the compacta (Shink et al. 1997).

Some large neurons in the external segment of the globus pallidus (but not the entopeduncular nucleus) project to widespread areas of the cerebral cortex, but particularly to motor regions. They stain intensely for acetylcholinesterase and may be cholinergic. They are similar to cells in the substantia innominata and basal nucleus of Meynert (Jackson and Crossman 1981[a], Parent et al. 1981, Parent and De Bellefeuille 1982, Reinoso-Suarez 1982), which also provide a diffuse cholinergic projection to the cerebral cortex. Some authors regard these latter cholinergic subpallidal structures as a ventral extension of the globus pallidus (Russchen 1982[a,b]). Alternatively the putative cholinergic component of the globus pallidus may be thought of as a dorsal extension of the subpallidal zone.

The projection from globus pallidus to neostriatum is mentioned previously under ‘Afferent connections of the neostriatum’.

Other connections of the pallidum, such as the fibres from the ventral pallidum to the hypothalamus, medio-dorsal thalamic nucleus, and habenular nucleus, (Haber et al. 1993), have been mentioned above.

The cells of the subthalamic nucleus are small to medium in size being 20–40 µm in diameter. They are fairly closely packed and this is more marked medially, where they tend to be a litle smaller. This difference in size has been emphasized in the monkey, in which magnocellular and parvocellular cell groups have been distinguished (Fisher et al. 1991). They vary in shape being polygonal, spindle shaped, or round, with prominent Nissl substance, which extends into the bases of the fine dendritic trees. Most neurons have five to eight long sparsely spined dendrites which arborize mainly in one plane (Sato et al. 2000[a]). As mentioned above under ‘Globus pallidus’, the pallido-subthalamic pathway is probably GABAergic. In addition, acetylcholinesterase (Poirier et al. 1977), neurotensin (Kataoka et al. 1979), enkephalin (Sar et al. 1978, Wamsley et al. 1980), and dopamine (Meibach and Katzman 1979) have been found in the subthalamic nucleus. Different subpopulations of cells exists in dorsal and ventral regions based on calcium binding protein expression (Augood et al. 1999). It seems likely most subthalamic neurons utilize the excitatory neurotransmitter glutamate (Smith and Parent 1988).

This connection, which arises in the external pallidal segment, is described above under ‘Efferent connections of the globus pallidus’.

Various other projections to the subthalamic nucleus have been described, but conflicting results cast doubt on their existence.

Thus, fibres from the substantia nigra (pas compacta – see later under ‘Efferent connections of the substantia nigra’), dorsal raphe nucleus, pedunculo-pontine nucleus, and locus coeruleus were described in the monkey by Rinvik et al. (1979), but could not be confirmed by Carpenter et al. (1981). The very small numbers of fibres in these projections may help to explain this discrepancy (Hammond et al. 1983[a], Lavoie et al. 1989). The existence of a nigro-subthalamic pathway, however, has been confirmed (see earlier under ‘Afferent connections of the neostriatum’). The fibres in this pathway arise in the substantia nigra pars compacta and retrorubral and ventral tegmental areas (Hassani et al. 1997, Gauthier et al. 1999). They terminate throughout the subthalamic nucleus and probably provide dopaminergic innervation (Hedreen 1999, Francois et al. 2000), which in humans may be by way of dopamine D1 receptors (Augood et al. 2000). An afferent connection from the motor cortex has been described and this may monosynaptically excite Golgi type 1 cells, which have distally projecting axons (Kitai and Deniau 1981). It is dense, somatotopic, and accompanied by a weaker projection from surrounding cortex (see Flaherty and Graybiel 1994). This pathway seems capable of conveying information from cortex to the globus pallidus via the subthalamo-pallidal connection more rapidly than via the cortico-striato-pallidal route (Nambu et al. 2000). There is also evidence for input to the subthalamic nucleus from the supplementary and presupplementary motor areas (Nambu et al. 1997, Inase et al. 1999). Thalamo-subthalamic fibres may arise in the centro-median and parafascicular nuclei passing through the zona incerta to reach the ipsilateral rostral subthalamic nucleus. Their pathway has been demonstrated in the rat and cat, but their functional significance remains to be determined (Sugimoto et al. 1983).

The major output of the subthalamic nucleus is to the internal and external segments of the globus pallidus as mentioned previously under ‘Afferent connections of the globus pallidus’. In the monkey the projection to the external segment is from the magnocellular division, while the parvocellular division supplies the internal segment and ventral pallidum (Fisher et al. 1991).

Fibres pass caudally from the subthalamic nucleus to the substantia nigra and terminate in the pars compacta and pars reticularis (Kanazawa et al. 1976, Carpenter et al. 1981), but less than a quarter of subthalamic neurons provide such axons (Sato et al. 2000[a]). They arise in the parvocellular division in the monkey (Fisher et al. 1991) and are collaterals of subthalamo-pallidal fibres as mentioned above, which go either to the external segment or to the external and internal segments (Hammond and Yelnick 1983,

Hammond et al. 1983[b], Sato et al. 2000[a]). Their activity produces excitation in the substantia nigra (Hammond et al. 1983[b]). Some of the nigral cells with which they make contact may project to the superior colliculus (Hammond et al. 1983[b]).

A small number of neurons in the parvocellular division of the subthalamic nucleus appear to project to the substantia innominata and ventral pallidum. This zone also sends fibres caudally to the pedunculo-pontine nucleus and this connection is probably reciprocal (Rinvik et al. 1979, Jackson and Crossman 1980, Jackson and Crossman 1981[b], Hammond et al. 1983[a], Fisher et al. 1991). Fibres from the lateral part of the subthalamic nucleus to the ipsilateral cerebral cortex have been described in some species (Jackson and Crossman 1981[a], Miyata 1986) and other studies show that the subthalamic nucleus innervates the sensory motor and prefrontal cortices (Degos et al., 2008). There is also a subthalamo-striatal projection (see earlier under ‘Afferent connections of the neostriatum’) and some axons may be directed to the spinal cord (Takada et al. 1987).

Histologically the substantia nigra is divided into two major areas: the pars compacta and the pars reticularis (Fig. 1.8). The dorsal neurons of the compacta tend to be fusiform and arranged vertically, while the ventral neurons are stellate (Mettler 1968). Compacta neurons are sometimes referred to as type I and reticulata neurons as type II. Cells in the compacta are closely arranged whereas those in the reticulata are separated by a network of fibres and are more varied in shape.

There have been various attempts to subdivide and classify the pars compacta and adjacent dopaminergic cell groups. The exact arrangement varies from one species to another. Thus, while the boundaries between the pars compacta, ventral tegmental area, and retrorubral area are reasonably distinct in the rodent, they are difficult to define in monkeys and humans. In an early attempt to subdivide this region Olszewski and Baxter (1954) delineated α, ß, and γ groups of cells. The γ group lay most dorsal and medial and the cells of this were contiguous with the tegmental nuclei. The α group was most ventral and lateral, with the ß group lying in between. The immunostaining properties of these γ neurons are somewhat different from those of the α and ß regions and subsequent publications have tended to exclude the γ group from the compacta (Gibb 1992). In addition, the compacta has been reclassified in humans and other primates into a dorsal tier, which corresponds to the ß group and medial part of the α group of neurons, and a ventral tier, which is the lateral section of the α group (Gibb 1992). The neurons in the dorsal tier are rather loosely arranged and can be fairly readily distinguished from those of the more closely packed ventral tier (denso-cellular zone). Clusters of neurons in this latter region form finger-like extensions, which pass deeply into the pars reticulata (Haber and Groenewegen 1989).

The ventral tegmental area includes the midline cell groups, the paranigral nucleus, and the pigmented parabrachial nucleus. It is continuous with the retrorubral group of cells, the γ region of Olszewski and Baxter (1954), and other various defined cellular regions, such as the pars mixta of Francois et al. (1984) and the pars dorsalis of Poirier et al. (1983). Together these neurons make up a morphologically similar group of cells which lie dorsal to the majority of the pars compacta cells and share a number of immunostaining properties (see later).

The most prominent feature of the substantia nigra is the blackness of the pars compacta, which is due to melanin pigment granules within the neuronal cytoplasm. Although this melanin first appears in the foetus, it is not visible macroscopically until about 5 years of age. It reaches maximum intensity by about 15 years (Pilcz 1895). Its formation is due to the auto-oxidation of dopamine, which is being formed in these cells. As such it represents a by-product without known useful function. The dorsal tier cells have a high melanin content, while in those of the ventral tier it is low. It is also present in the cells of the pars lateralis and the ventral tegmental area of Tsai, which are dopaminergic. In addition to the presence of melanin, dopaminergic cells can be distinguished by large masses of Nissl substance and darkly staining basophilic cytoplasm. Electron microscopy shows this cytoplasm to be filled with free ribosomes associated with complexes of rough endoplasmic reticulum. These features are less marked in non-dopaminergic cells (Domesick et al. 1983). The neurons in the γ group of Olszewski and Baxter (1954) and in the retro-rubral nucleus are somewhat different from those in the pars compacta. Not only do many have a low content of melanin, but also they are like the other tegmental nuclei and unlike most compacta cells, in that they stain strongly for tyrosine hydroxylase and calbindin D_28k (Gibb 1992). In monkeys only the dorsal tier dopaminergic neurons stain for the latter (Francois et al. 1999) while the majority of pars compacta and pars lateralis cells are positive for calretinin (Isaacs and Jacobowitz 1994, Rogers 1994). Compacta neurons, particularly those in the ventral tier, contain high levels of RNA for dopamine transporter (Haber et al. 1995). The pars reticulata looks reddish-brown in the fresh specimen and is without melanin, although it contains a high concentration of iron. Some of its cells may be GABAergic (Oertel et al. 1982). It is traversed, particularly in its lateral part, by descending fibres which originate mainly in the neostriatum.

Two varieties of dendrites are distinguished. Type I arises from type I compacta neurons and sometimes course in bundles into the reticulata. The dendrites of the ventral tier neurons tend to be dorso-ventrally orientated, whereas those of the dorsal tier are aligned medio-laterally (see Lynd-Balta and Haber 1994[a,b]). They have relatively few synapses, but they lie in close contact with each other. The reticulata has abundant type II dendrites, which show many synapses without the membrane contact of the type I dendrites. While it seems certain that type I dendrites arise from type I neurons, the origin of type II dendrites is not entirely clear (Henderson and Greenfield 1984).

The synaptic arrangement of the substantia nigra remains uncertain. An earlier simplistic view was that the strio-nigral GABAergic pathway synapsed with pars reticulata cells, which acted as interneurons and projected to the pars compacta. This in turn had its major efferent projection back to the neostriatum. It is now known that this is erroneous and that the situation is much more complex. There are many more afferent and efferent connections involved and the pars reticulata has motor functions independent of the nigro-striatal pathway (Cools et al. 1983). The nigra contains interneurons, including GABAergic ones (Hebb and Robertson 2000). It is uncertain if some interneurons are involved in the connection between the strio-nigral fibres and dopaminergic pars compacta (Kitai et al. 1975, Bunney and Aghajanian 1976[a,b]), but many such contacts are monosynaptic (Bolam and Smith 1990). It has been shown that neurons in the pars reticulata receive symmetric synapses from striato-nigral fibres on their distal dendrites and from pallido-nigral fibres on their perikarya and proximal dendrites. A substantial number of such nigro-collicular neurons get convergent input from both of these sources. Pallido-nigral neurons receive direct synaptic input from the neostriatum (Smith and Bolam 1991). In addition to these imperfectly known details of the axo-dendritic connections, which are discussed below under the ‘Afferent connections of the substantia nigra’, the situation is further complicated by the possible existence of dendro-dendritic synapses between the type I dendrites of the pars compacta. The purpose of these contacts is uncertain (Groves and Linder 1983, Henderson and Greenfield 1984). Some nigral dendrites release dopamine and both D1 and D2 and possibly D3 and D4 receptors are involved in mediating the intranigral effects of dopamine. For example, dopamine is known to facilitate GABA and possibly glutamate release from striato-nigral fibres in the pars reticulata (see Smith and Kieval 2000), Conn et al. 2001). Many such dopamine receptors are at extrasynaptic sites (Caille et al. 1996). Whatever the exact cellular arrangement within the substantia nigra it appears to contain several other neurotransmitters or neuromodulators, including dynorphin substance P, substance K, enkephalin, neurotensin, somatostatin, cholecystokinin, serotonin, and possibly acetylcholinesterase and acetylcholine (see Graybiel 1986, Bolam and Smith 1990, Kalivas and Duffy 1995 and Smith and Kieval 2000 for further details). Using immunostaining techniques to demonstrate chemical profile, a complex array of both dopaminergic and GABAergic neurons can be demonstrated and up to 10 types of cells have been delineated in animals (Gonzalez-Hernandez and Rodriguez 2000).

These fibres have been described above under ‘Efferent connections of the neostriatum’. The GABA-containing fibres are known to terminate in the pars reticulata and seem to make monosynaptic contact with long dendrites from cells of the pars compacta, which extend down into the pars reticulata (Yoshida and Precht 1971, Hattori et al. 1975, Kitai et al. 1975, Smith et al. 1981). This input comes mainly from the ‘sensorimotor’ and ‘associative’ regions of the striatum (Joel and Weiner 2000). GABAergic axons, which originate chiefly in the ‘limbic striatum’, also form symmetric synapses with dopaminergic neurons within the pars compacta (Chang 1988, Bolam and Smith 1990, Joel and Weiner 2000). These include a projection from the nucleus accumbens (Swanson and Cowan 1975, Bunney and Aghajanian 1976[b], Nauta et al. 1978, Sugimoto and Mizuno 1987). The ‘limbic striatum’, however, also extends its zone of influence to cover most of the dopaminergic neurons which project to the sensorimotor and associative striatum (Joel and Weiner 2000). In addition similar fibres make monosynaptic contacts with reticulata neurons projecting to the thalamus and tectum (Deniau et al. 1978, Smith et al. 1981, Williams and Faull 1985). The cells of origin of nigro-tectal fibres to the superior colliculi may receive contacts from strio-nigral axons which have arisen from the neostriatal projection area of the visual cortex – see later (Rhoades et al. 1982, Williams and Faull 1985). Substance P, substance K, enkephalin, and neurotensin-containing fibres from striatum may terminate in both pars reticulata and compacta (Inagaki and Parent 1984, Graybiel 1986, Sugimoto and Mizuno 1987, Bolam and Smith 1990). There is evidence to suggest that projections to these two nigral zones arise from separate striatal compartments, at least in rats and cats (see earlier under ‘Neostriatum – structure’).

The precise arrangement in humans is uncertain, but both enkephalin and substance P-positive terminals form woolly fibres (Haber and Nauta 1983), which are restricted to the pars reticulata and the ventral tier of the pars compacta. While substance P is found throughout these areas, enkephalin is located only in their medial section. As the latter axons may arise in the ventro-medial striatum it has been suggested they represent input from the limbic system (Haber and Groenewegen 1989).

These fibres have been described previously under ‘Efferent connections of the globus pallidus’. It has been demonstrated that in some species fibres from the pallidum (external segment) and neostriatum converge on the same nigral neurons in a situation similar to that seen in the internal segment of the globus pallidus (Smith and Bolam 1991, Von Krosigk et al. 1992, Bolam et al. 1993).

There is a major projection from the subthalamic nucleus to the substantia nigra (Nauta and Cole 1978) which has been described previously under ‘Efferent connections of the subthalamic nucleus’.

The central nucleus of the amygdala seems to project ipsilaterally to both the pars reticulata and the pars compacta of the substantia nigra (Ungerstedt 1971, Bunney and Aghajanian 1976[a], Barasi and Pay 1980), but there may be significant species variation. Thus, terminations have been reported to be restricted to the compacta in the rat (Gonzales and Chesselet 1990) and the reticulata in the cat (Shinonaga et al. 1992). The terminations in the reticulata may be excitatory and those in the compacta inhibitory (Barasi and Pay 1980). Metenkephalin, dynorphin, and neurotensin are probably present in this pathway (Vankova et al. 1992).

There are a number of other structures which may also project to the substantia nigra. These include the bed nucleus of the stria terminalis (Tulloch et al. 1978, Grofova 1979), the hypothalamic area (Gerfen et al. 1982), the parafascicular nucleus of the thalamus (Gerfen et al. 1982), the habenular (Bunney and Aghajanian 1976[a]), and the nucleus accumbens (see under ‘Strio-nigral fibres’). Connections with the prefrontal, premotor, and motor areas have been described by a number of authors, but it seems likely that they are relatively minor (Bunney and Aghajanian 1976[a], Tulloch et al. 1978, Grofova 1979, Carter 1982, Gerfen et al. 1982, Kornhuber et al. 1984, Gariano and Groves 1988). Some of these synapse in other midbrain regions, such as the ventral tegmental area, rather than in the nigra proper (Carr and Sesack 2000). Axons from the serotonergic dorsal raphe nucleus (Bunney and Aghajanian 1976[a], Tulloch et al. 1978) and the pedunculo-pontine nucleus (Gerfen et al. 1982, Gould et al. 1989) have been found. The latter make symmetric synapses with dendritic shafts of dopaminergic neurons and seem to be glutamatergic (Charara et al. 1996).

The major output from the substantia nigra pars compacta is via the nigrostriatal pathway to the neostriatum as described previously under ‘Afferent connections of the neostriatum’. The dendrites of these dopaminergic cells make contact with fibres in the strio-nigral system as mentioned previously under ‘Afferent connections of the substantia nigra’.