The basal ganglia system is a major cerebral system that has only recently been recognized. In the past, part of it was presented as "motor" or "extrapyramidal", complementary to the corticospinal (pyramidal) system. Contrary to what was thought, the basal ganglia sytem has no direct output to the spinal cord. As already known by Charcot, the motor effects are exerted through the motor cortex and the pyramidal system. The unilateral effects are thus controlateral. The basal ganglia system participates in much more than motor functions. As it is highly evolutive, it appeared opportune to consider the primate system apart. Major changes occured in the last years which drastically modify the subject, which thus deserves to be detailed.

=History= The first anatomical identification of distinct subcortical structures (at the "base" of the brain) was published by Willis in 1664. For many years, the term corpus striatum was used to describe a large group of subcortical elements, some of which were later discovered to be functionally unrelated. In the same time, the putamen and the caudate nucleus were not linked together. The putamen was associated to the pallidum in what was called the "nucleus lenticularis". Pioneering work by Cécile and Oskar Vogt (1941) greatly simplified the description of the basal ganglia in proposing the term striatum to unite together the caudate nucleus, the putamen and the mass linking them ventrally, the fundus. The striatum gets its name from the striated appearance created by dense radiating bundles of striato-pallido-nigral axons, described by Kinnier Wilson (1912) as "pencils". The anatomical link of the striatum with its primary targets, the pallidum and the substantia nigra was discovered later. The similitude between the structure of the substantia nigra and that of the "glous pallidus" was noted very early by Mirto. However their regrouping as an ensemble (pallidonigral set) is still not commonly accepted. The striato-pallido-nigral ensemble represents the core of the basal ganglia. The striato-pallidonigral bundle passes through the pallidum, forms the so-called "comb bundle of Edinger" when it crosses the internal capsule, then finally reaches the substantia nigra. Additional structures later became associated with the basal ganglia such as the "body of Luys" (1865) or subthalamic nucleus, whose lesion was known to produce hemiballism. More recently, other areas such as the central complex (centre médian-parafascicular) of Luys and the pedunculopontine complex have been considered as parts of the system.

=Definition and content= The basal ganglia system, identified lately, met problems even in its naming. The term basal comes from the fact that most of its elements are located in the basal part of the brain (the basal nucleus of Meynert is however not a part of the system). The term ganglion has never been adequate. Terminologia anatomica (1998), the international authority for anatomical naming retained "nuclei basales", which is not used. The International Basal Ganglia Society (IBAGS) considers as basal ganglia, the striatum, the pallidum (with two nuclei), the substantia nigra (with two distincts parts) and the nucleus subthalamicus. To this is added today the central region (centre median-parafascicular) (Percheron et al. 1991, Parent and Parent, 2005) and for some, the pedunculopontine complex (Mena-Segovia et al. 2004). The basal ganglia system may be defined as the set of subcortical cerebral elements starting from the striatum, in interaction together and with parts of the thalamus and cortex. This ensemble must not be seen as a collection of elements but as a system.

=Corticostriatal connection= The whole system starts as a major output of the cortex, about the same size as the corticopontine system opening the cerebellar system. The cortico-striatal connection represents a significant portion of the whole cortical output. Almost every part of the cortex, except for the primary visual and auditory cortices, sends axons to the striatum. The origin of the connection is in the pyramidal neurons of the layer V of the cortex. Corticostriate contributors, of the motor cortex at least, may be collaterals of axons descending lower in the nervous system. However, in primates, the majority of corticostriate axons are monotarget, pure cortico-striate, thin and unbranched until they arrive in the striatum (Parent and Parent, 2006). The corticostriatal connection is glutamatergic and excitatory. This connection is not topologically as simple as was initially described by Kemp and Powell (1970); where the frontal lobe projected anteriorly and the occipitotemporal lobes posteriorly. Part of this distribution grossly remains but the distribution is much more complex. One small cortical place can send terminal arborisations to several and distal striatal places (Goldman and Nauta, 1977, Selemon and Goldman-Rakic, 1985). The cortico-stiatal connection is the substrate of cortical information separation or regrouping: axons from distinct areas can systematically end together or systematically end separately. There is also a spatial reorganisation: a "remapping"(Flaherty and Graybiel, 1991).

The corticostriate connection is the first in the chain of a strong reduction in numbers between emitter and receiver neurons (Percheron et al. 1987), i.e. a numerical convergence. The effect of this is that if each striatocortical neurons has its own message, this will be mixed or compressed, leading to a lesser definition of the input map.

=Striatum and basal ganglia core= The striatum is the first element of the basal ganglia system. The basal ganglia core includes the striatum and its direct targets (reached through the major striato-pallidonigral bundle) the two nuclei of the pallidum and the substantia nigra.

The striatum is a huge neuronal telencephalic mass located close to the lateral cerebral ventricles. It has a toric topology but is a single close space.

Structure

---

In primates, the striatum has four neuronal genera: spiny neurons (96%), leptodendritic (2%), spidery neurons (1%) and microneurons(1%) Yelnik et al. (1991). The dendritic arborisations of the spiny neurons are spherical unless close to a border. Their diameter depends on the animal species. Spines are of the same type than those of two other (telencephalic) acanthodendritic (acanthos means spine) genera, the pyramidal neurons of the cerebral cortex and the spiny neurons of the amygdala. Most of these spines synapse with cortical afferents. The spontaneous activity of striatal neurons of awake monkeys is surprinsigly "low or absent" (DeLong, 1980), "quiescent". Neurons are activated by cortical stimulations. In the sensorimotor striatum they respond to movements. Their axons have abundant and dense initial axonal collateral acting fastly on neighbour neurons (Czubayko and Plenz, 2002). The further part is long and myelinated. The spiny neurons are GABAergic, instauring the first part of the inhibitory 2-path so particular to the system . The leptodendritic neurons (or Deiter's), stain for parvalbumin and have all the morphological properties of the pallidal neurons. The spidery neurons are specific to primates. They have a big soma and short dendritic and axonal branches. They are the cholinergic neurons of the primate, with a morphology entirely different from that of non-primates. This must lead to a great care in making physiopathological references. They are the "tonically active neurons" or TANs (Kimura et al. 2003). The microneurons are local circuit neurons similar to those found in the thalamus for instance. They are GABAergic, and some may be dopaminergic (Cossette et al.2005). The long oblique split of the striatum by the internal capsule creates the classic division into putamen, caudate and fundus. In fact, the striatum is a continuous mass. The gross anatomical division does not correspond exactly with the presently accepted anatomofunctional subdivision of the striatum in primates. There are several levels of organization of the striatum. One level is brought by the differentiated topographic endings of the corticostriatal axons. Endings of axons from the central region of the cortex, primary somatosensory, motor, premotor (Künzle 1975 and several other papers), accessory motor and anterior parietal constitute a sensorimotor territory (or for short sensorimotor striatum) essentially putaminal, but which does not cover the total extent of the putamen. Conversely it includes intracapsular fringes and the inferolateral border of the caudate Percheron et al. (1984).It is grossly somatotopically arranged with three oblique bands, laterally the inferior limb, then the upper limb and medially the face. To this is opposed an associative territory, essentially caudate, above all orally and dorsally, which, also, does not cover the entire caudate volume. The separation between the two sensorimotor /associative territories may be in some places clearcut and observed using calbinding immunochemistry (the sensorimotor territory being negative)François et al. (19 ). The isolation of a third ventral striatal part often qualified as "limbic" is more difficult. Only one part is distinctive, the "nucleus accumbens" (in fact a pars and not a nucles) having same morphological features than elsewhere but particular immunostaining properties and, above all, a selectivety for the reception of axons from the subiculum. A "shell" and a "core" are said to be also present in primates. They are of small size (Brauer et al. 2000) relatively to the other parts. Another level of organisation is that of "compartments". Histochemistry has showed inhomogeneities with regards to the distribution of different molecules. The major compartment, the matrix, as its name indicates, is considered as the basic element. It contains contrasting islands or striosomes that contain opiate receptors (Pert, ) and stain for acetylcholinesterase. This opposition, obvious in the head of the caudate, is not clear everywhere. Striosomes have no links with amygdalar afferents in primates. Striosomes would rather represent the insular segregation of particular frontal axonal endings (posterior orbitofrontal/anterior insula and mediofrontal/anterior cingulate cortex) (Eblen and Graybiel,1995). Matricial neurons are those contained in the matrix. Striosomal neurons are those contained in the striosome. They have been opposed as sources of distinctive efferences, which will be shown below no to be really true.

Hodology to targets

---

There have been disputes concerning the origin of striatal axons to the different targets. For a long time spiny neurons were considered as local circuit neurons! They are long axoned neurons. Due to difficulties linked to the geometry of the system, the first data from tracing techniques led to believe that there were specialised striato-pallidal or striato-nigral neurons having each histochemical particularities. A recent study in macaque (Levesque and Parent et al. 2005), following another one in the rat, has drastically changed the situation. Spiny neurons do have generally several targets. This is not an archaic pattern since it is found in 90% of the cases in macaque monkeys versus 63,6% in the rat. Virtually all striatal axons have the lateral pallidum (the most voluminous) as their first target. 24/27 of the studied axons projected to the three consecutive targets, lateral pallidum, medial pallidum and nigra (lateralis and reticulata)(3-target). There are no striatal axons projecting to the single medial pallidum, single nigra or both. Between matricial and striosomal axons, the only difference in axonal hodology is that striosomal axons cross the whole lateral to medial extent of the nigra and emit (in macaques) 4 to 6 vertical collaterals, forming vertical columns entering deep inside the pars reticulata. The matricial neurons emit more sparsely branched axons. This general pattern of connectivity raises new problems. The main mediator of the striato-pallidonigral system is GABA; but with comediators. Since Haber and Elde (1981), it is known that the lateral pallidum stains for met-enkephalin, the medial for substance P and /or dynorphin and the nigra for both. This could mean that a single axon is able to concentrate different comediators in different subtrees depending on the target. This considerably modifies several decades old schemes and raises new questions.

= Pallido-nigral set and pacemaker= The pallidonigral set comprises the direct targets of the striatal axons: the two nuclei of the pallidum and the pars lateralis and pars reticulata of the "substantia" nigra (not a substance but a nucleus). One character of this ensemble is given by the very dense striato-pallidonigral bundle giving it its whitish aspect (pallidus means pale). By no ways has the pallidum the shape of a globe. Cecile and Oskar Vogt (1941) simplified the case by selecting the term pallidum, also offered by the Terminologia Anatomica (1998). They also proposed nigrum that may be replaced by nigra. The whole set is made up the same neuronal components. The majority is made up of very large neurons, strongly stained for parvalbumin, having very large dendritic arborisations (much larger in primates than in rodents) with straight and thick dendrites (Yelnik et al. 1987). Only the shape and direction of the dendritic arborizations differ between the pallidum and the nigra neurons. The pallidal dendritic arborisations are very large flat and discoidal (Yelnik et al. 1984). Their principal plane is parallel one to the others. They are all parallel to the lateral border of the pallidum and thus perpendicular to the axis of the afferences (Percheron et al. 1984). Since the pallidal discoidal discs are thin, they are crossed only on a short distance by striatal axons, but, since they are large they are crossed by many of them. Since they are loose, the chances of contact are not very high. Striatal arborisations however emit perpendicular branches participating to flat bands parallel to the lateral border, which increases the density in this direction. This is true for the striatal afferent but also for the subthalamic (see below). The synaptology of the set is uncommon and characteristic allowing to easily attribute one place to this ensemble (Fox). The dendrites of the pallidal or nigral axons are entirely covered by synapses, without any apposition of glia. More than 90% of synapses are from striatal origin (Di Figlia et al. 1982 ). One noticeable property of this ensemble is that not one of its elements receives cortical afferents. Initial collateral are present. However, in addition to the presence of various appendages at the distal extremity of the pallidal neurons (di Figlia et al. 1982, François et al. 1984) that could act as elements of local circuitry, there are weak or no functional interrelations between pallidal neurons (Bar-Gad et al 2003).

Pallidum laterale (external g. pallidus, GPe)

The lateral pallidum is flat, curved and very extended. The three-dimensional shape of arborisations is discoid and flat. The arborisations are parallel to one another and to the lateral border of the pallidum. They are perpendicular to the striatal afferences (Percheron et al. 1984). In addition to the striato-pallidal afference, the lateral pallidum receives a major connection from the subthalamic nucleus (see below). It also receives dopaminegic afferences from the nigra compacta. The lateral pallidum constitutes the most ancient part of the pallidum in mammals. Contrary to two other elements of the basal ganglia core, it is not a source of output to the thalamus as it sends its axons essentially to other basal ganglia elements (intrasystemic connections). To some extent, it may be seen as an inner basal ganglia regulator. Its mediator is GABA. Contrary the medial pallidum, the very fast spontaneous activity is discontinuous with long intervals and silence. Lateral pallidal neurons are often multitargets and may correspond to several hodotypes (neuronal varieties according to the topology of their targets). From Sato et al. (2000), in macaques,the lateral pallidal neurons sends axons in the direction of the striatum in only 15.8%. The other striatal neurons projects to three consecutive targets (medial pallidum, nigra reticulata and subthalamic nucleus) in 13,2% of the cases. The neurons projecting to the medial pallidum and subthalamic targets are 18,4%. Those projecting to the subthalamic nucleus and nigra reticulata 52,6%. The subthalamic nucleus is thus, in 84,2% of the cases, the target of lateral pallidal neurons. In return, the subthalamic nucleus, the priviledged target of the lateral pallidum sends the majority of its axons to it (see below).

Pallidum mediale (internal g. pallidus, GPi)

The medial pallidum, though absolutely similar to the lateral, is phylogenetically youngest. It appears only in primates. The entopeduncular nucleus of non-primate is not its equivalent. It does not have indeed a separate territory in the thalamus since its axons end together with nigral ones. In this respect the entopeduncular nucleus would rather be a lateral intracapsular part of the nigra. The medial pallidum is separated into two parts (medial and lateral) by the lamina intermedia. The medial pallidum (as well as the lateral and the nigra reticulata, see below) is a "fast-spiking pacemaker" with spontaneous discharges in awake monkeys at about 90 Hz (Mink and Thach, 1991), 70 to 80 for Fillion and Tremblay (1991). In opposition to that of the lateral pallidum the activity is continuous (DeLong, 1960). In addition to the massive striatopallidal connection, the medial pallidum receives a dopaminergic innervation from the nigra compacta. Contrary to the lateral pallidum, it is one major source of basal ganglia outputs. The first axonal component (10%) in macaque is in direction of the habenula. The main group (90%) sends long axons directed posteriorly that, through collaterals, furnish several successive targets: the lateral region of the thalamus (VO)(see thalamus), the pars media of the central complex (see below), the pedunculopontine complex (Percheron et al., 1996) and to the retrorubral area (Parent and Parent (2004). The highly evolutive development of medial pallidum carries along that of a major output, the pallido-thalamic bundle, (successively the ansa and fasciculus lenticularis, the cam system, the Forel's fields H2, H and H1) and, above all, the appearance of a distinct nucleus in the lateral region of the thalamus, the nucleus ventralis oralis, VO (see thalamus). The mediator is GABA forming the second segment of the inhibitory 2-path.

Nigra lateralis (SNl)

The substantia nigra is inhomogeneous and complex. The first separation must be made between the dopaminergic ensemble (including the pars compacta) and the GABAergic ensemble pursuing the pallidum in the other side of the capsule. Experience shows that it is very difficult to convince people to extract mentally the substantia nigra from the mesencephalon and to place it fully in the basal ganglia system. Hence there is a continuity of the bundle from the pallidum to the nigra. The two parts of the nigra that belong to the basal ganglia core are the pars lateralis and the pars reticulata receiving a dense projection from the striato-pallido-nigral bundle and having the same structure as that of pallidal neurons. The nigral neurons are also sparsely branched and long (Yelnik, et al. 1987). The difference between pallidal and nigral neurons is only in the three-dimensional extension of their dendritic arborizations (François et al. 1987). Nigral dendrites, as well as pallidal, tend to be perpendicular to the arriving stiatal axons. The particular synaptology is also the same. The pars lateralis is the most lateral part of the nigra. It is frequently not considered separately as the main difference from the pars reticulata is that it sends axons to the superior colliculus (François et al. 1984). This seems yet to be a sufficient reason.

Nigra reticulata(SNr)

The pars reticulata or diffusa, is the most often considered alone including the pars lateralis. It may be either only the most medial part of the pallido nigral ensemble, when a pars lateralis is retained.This must be checked in papers. Its name is simply an opposition to the dense pars compacta located above it. The border between the two is highly convoluted with deep fringes.Its neuronal genus is the same as that of the pallidum with the same thick and long dendritic trees.It receives its synapses from the striatum in the same way as the pallidum. Striatonigral axons from the striosomes may form columns vertically oriented entering deeply in the pars reticulata (Lesvesque and Parent, 2005). The ventral dendrites of the pars compacta from the reverse direction go also deeply in it. The nigra reticulata is another "fast-spiking pacemaker" (Surmeier et al. 2005). In addition to the massive striatopallidal connection, the nigra reticulata receives a dopamine innervation from the nigra compacta and glutamatergic axons from the pars parafascicularis of the central complex. It sends nigro-thalamic axons. There is no conspicuous nigro-thalamic bundle. Axons arrive medially to the pallidal afferences at the anterior and medial most part of the lateral region of the thalamus: the nucleus ventralis anterior (VA, differentiated from the VO receiving pallidal afferences).The mediator is GABA.

The pallidonigral pacemaker

One of the most important recent discovery is that the basal ganglia machinery is not simply set in motion from the outside. It has indeed several "autonomous pacemakers" defined as a set of "neurons capable of periodic spiking in the absence of synaptic input" (Surmeier et al. 2005) i.e. able of producing an own activity. The pallidonigral ensemble is one of them. Among autonomous pacemakers, it belongs to "fast-spiking pacemakers" "capable of discharges rates in excess of 200 Hz for sustained periods" (Surmeier et al. (2005).In the so evolutive basal gaglia system it is important to be precise and careful. The majority of what is known about the production and regulation of the pallido(nigral) pacemaker comes from works on slices in rodents. These having no true medial pallidum, the data deal with the external, which belong to a particular subsystem. From these data, the regularity and frequence of the pacemaker is linked to cyclic nucleotide-gated channels (HCN2 and HCN1, Chan et al. 2004) present on the dendrites of pallidal neurons, modulated by striatopallidal axons.

The fact that there is an autonomous activity deplaces the problem on what really influences this activity?

Striato-pallidonigral connection

---

The striato-pallidonigral connection is a very particular one. It engages a considerable number of fine striatal axons. Estimated numbers are 110 million in man, 40 in chimpanzees and 12 in macaques (see Percheron et al. 1984, 1987). The striato-pallido-nigral bundle is made up of very numerous thin, fewly myelinated axons from the striatal spiny neurons grouped into pencils "converging like the spokes of the wheel" (Papez, 1941). It gives its "pale" aspect to the receiving areas. It strongly stains for iron using Perls technique.

Convergence

After the huge reduction in numbers of neurons between the cortex and the striatum (see corticostriate connection), the striatopallido-nigral connection is a further reduction in the number of transmitter vs receiving neurons. Numbers given by Percheron et al. (1987, 1989) are 31 millions striatal spiny neurons in macaques. There are 166000 lateral pallidal neurons, 63000 medial pallidal, 18000 lateral nigral and 35000 in the pars reticulata, which makes 283000 target neurons. If the number of striatal neurons is divided by this number, as an average, each target neuron may receives information from 117 striatal neurons. (Numbers in man are 555000 lateral pallidal, 157000 medial pallidal, 167000 nigral -lat+ret- and about the same ratio). Another different approach starts from the mean surface of the pallidonigral target neurons and the number of synapses that they may receive. Each pallidonigral neurons may receive 70000 synapses. Each striatal neuron may contribute for 680 synapses. This leads again to an approximation of 100 striatal neurons for one target neuron. This represents a huge, unfrequent reduction in neuronal connections. The consecutive compression of maps cannot preserve finely distributed maps (as would be the case for instance in sensory systems). The fact that a strong possibility of convergence exists does not means that it is constantly used. Percheron and Filion's (1991) argued for a "dynamically focused convergence". A recent modeling study starting from entirely 3-d reconstructed pallidal neurons showed that their morphology alone was able to create a center-surround pattern of activity (Mouchet and Yelnik, 2004). The very particular geometry of the connection between striatal axons and pallidonigral dendrites in fact offers the possibility for a very large number of combinations such as local addition of simultaneous inputs to one tree or to several distant foci. The pallidum would offer an extended keyboard on which various cortical signals from close or remote areas can play.

Synaptology and combinatory

The synaptology of the striato- pallidonigral connection is so peculiar as to be recognized easily. Pallidonigral dendrites are entirely covered with synapses without any apposition of glia (Fox et al.,1974, Di Figlia et al. 1982). This gives in sections characteristic images of "pallissades" or of "rosettes". More than 90% of these synapses are of striatal origin. The few other synapses such as the dopaminergic or the cholinergic are interspersed among the GABAergic striatonigral synapses. The way striatal axons distribute their synapses is a disputed point. The fact that striatal axons are seen parallel to dendrites as "woolly fibers" has led to exagerate the distances on which dendrites and axons are parallel. Striatal axons may in fact simply cross the dendrite and give a single synapse. More frequently the striatal axon curves its course and follow the dendrite forming "parallel contacts" for a rather short distance. The average length of parallel contacts was found to be 55 micrometres with 3 to 10 boutons (synapses). In a last case the afferent axon bifurcate and give two or more branches, parallel to the dendrite increasing the number of synapses given by one striatal axon. The same axon may reach other parts of the same dendritic arborisation (forming "random cascades"). With this pattern, it is more than likely that 1 or even 5 striatal axons are not able to influence (to inhibit) the activity of one pallidal neuron. Certain numeral, spatio-temporal conditions are necessary for this.

=Nigra compacta (SNc) and nearby dopaminergic elements =

Sensu stricto, the pars compacta is a part of the core of basal ganglia core since it receives directly synapses from striatal axons through the striatopallidonigral bundle. The long ventral dendrites of the pars compacta indeed plunge deep in the pars reticulata where they receive synapses from the bundle. However, its constitution contrasts with the rest of the nigra. Aging leads to the blackening of its cell bodies, by deposit of melanin, visible by naked eye. This is the origin of the name of the ensemble ('substantia nigra' meaning black substance). The densely distributed neurons of the pars compacta have larger and thicker dendritic arborizations than those of the pars reticulata and lateralis. Contrarilly to the last, they are "low-spiking pacemakers" (Surmeier et al. 2005), spiking at low frequency (0,2 to 10 Hz)(below 8, Schultz). Groups of neurons located more dorsally and posteriorly in the tegmentum are of same type without forming true nuclei. The "cell groups A8 and A10" are spread inside the peduncule (François et al. 1999). They are not known to receives striatal afferences and are not in the topographical position to do so. The dopaminergic ensemble is thus also on this point inhomogeneous. This is another major difference with the pallidonigral ensemble. The fact that the efferent dopaminergic connection attracts the attention more than its input explains its intermediate position in our plan. The axons of the dopaminergic neurons, that are thin and varicose, leave the nigra dorsally. They turn round the medial border of the subthalmic nucleus, enter the H2 field above the subthalamic nucleus, then cross the internal capsule to reach the upper part of the medial pallidum where they enter the pallidal laminae, from which they enter the striatum (Percheron et al. 1989). They end intensively but inhomogeneously in the striatum, rather in the matrix in the anterior part and rather in the striosomes dorsalwards (Prensa et al.2000). These authors insits on the extrastriatal dopaminergic innervation of other elements of the basal ganglia system: pallidum and subthalamic nucleus . The role of the dopaminergic neurons has been the source of a considerable literature.It will be just remembered that due to its widespread distribution, it may regulate the system in many places.

= Regulators of the basal ganglia core=

Subthalamic nucleus, or corpus Lyuisi

---

As indicated by its name, the subthalamic nucleus is located below the thalamus; dorsally to the substantia nigra and medial to the internal capsule. The subthalamic nucleus is lenticular in form and of homogeneous aspect. It is made up of a particular neuronal species having rather long ellipsoid dendritic arborisations, devoid of spines, mimicking the shape of the whole nucleus (Yelnik and Percheron,1979).The subthalamic neurons are "fast-spiking pacemakers" (Surmeier et al. 2005) spiking at 80 to 90 Hz. There are also about 7,5% of GABA microneurons participating in the local circuitry (Levesque and Parent 2005). The subthalamic nucleus receives its main afference from the lateral pallidum. Another afference comes from the cerebral cortex (glutamatergic), particularly from the motor cortex, that is too neglected in models. A cortical excitation, via the subthalamic nucleus provokes an early short latency excitation leading to an inhibition in pallidal neurons (Nambu et al. 2000). Subthalamic axons leave the nucleus dorsally. Except for the connection to the striatum (17,3% in macaques), most of the principal neurons are multitargets and in direction to the other elements of the core of the basal ganglia (Sato et al. 2000). Some send axons to the substantia nigra medially and the medial and lateral nuclei of the pallidum lateraly (3-target 21,3%). Some are 2-target with the lateral pallidum and the substantia nigra (2.7%) or the lateral pallidum and the medial(48%). Less are single target for the lateral pallidum. If one adds all those reaching this target, the main afference of the subthalamic nucleus is,in 82,7% of the cases, the lateral pallidum (external segment of the globus pallidus. When striatopallidal and the pallido-subthalamic connections are inhibitory (GABA), the subthalamic nucleus utilises the excitatory neurotransmitter ^*. This gives it a particular (sometimes excessive) interest in models. Its lesion resulting in hemiballismus is known for long. Chronic stereotactic stimulation of the nucleus suppress most of the symptoms of the Parkinson' syndrome, particularly dyskinesia induced by dopatherapy.

subthalamo-lateropallidal pacemaker

As said before, the lateral pallidum has purely intrinsic basal ganglia targets. It is particularly stongly linked to the subthalamic nucleus by bothways connections. Contrary to the two output bases (medial pallidum and nigra reticulata), neither the lateral pallidum or the subthalmic nucleus sends axons to the thalamus. The subthalamic nucleus and lateral pallidum are both fast-firing pacemakers (Surmeier et al.2005). Together they constitute the "central pacemaker of the basal ganglia" (Plenz and Kitai,1999) with synchronous bursts. They are thus coupled regulators. The lateral pallidum receives a lot of striatal axons, the subthalamic nucleus not. The subthalamic nucleus receives cortical axons, the pallidum not. The pallido-subthalamic connection is inhibitory, the subthalamo-pallidal is excitatory.In this couple, this contrast has been insufficiently explored.

Central complex (C)

---

The central complex is the so-called centre-médian- parafascicular complex. Contrary to the current claim it does not topographically, histologically or functionally belong to the intralaminar group (Percheron et al. 1991). Located at the inferior part of the thalamus, it is almost everywhere surrounded by a capsule making it a closed region. In upper primates, starting from the cercopithecidae, it is made up not of two but of three parts with their own neuronal species (Fenelon et al. 1994). From there, two opposed interpretations were proposed concerning the belonging of the intermediate part: either to the centre médian (the Vogts, 1941) or to the parafascicular nucleus (Niimi et al. 1960). This is undecidable. It has thus been proposed to group the three elements together in the regio Centralis (since it is a classical nucleus) and to name them from medially to laterally: n. centralis pars parafascicularis, pars media and pars paralateralis. The whole is parvalbumin rich. The first two medial parts are acetylcholinesterase rich. They are the source of the major, centralo-striatal, part of the thalamo-striatal connection, with glutamate as the mediator. The pars parafascicularis receives afferences from the substantia nigra and the superior colliculus. It sends axons to the associative striatum. The pars parafascicularis sends also axons to the substantia nigra. The main afference of the pars media is the medial pallidum. This pars media sends axons to the matrix compartment of the sensorimotor striatum through an important bundle (François et al. 1991). The pars media is a part of the subcortical Nauta-Mehler's circuit (striatum-medial pallidum-pars media-striatum). There are thus strong interconnections of the complex with the basal ganglia. The pars paralateralis has essentially cortical relations particularly with the motor cortex. The structure of the complex being different from that of the close intralaminar formation and having different connections, it has been proposed two decades ago to remove the central complex from the intralaminar elements and to link it to the basal ganglia system, where it may be classified among the regulators of the core. Lesions of the complex have no known clinical effects. There are few physiological data in awake monkeys. For Matsumoto et al. (2001) the axons of the complex would supply striatal neurons with information about behaviorally significant sensory events. For Minamimoto and Kimura (2002) the region plays a role in attentional orienting to events occuring in the controlateral side.

Pedunculopontine complex

---

The pedunculopontine complex is not a primary part of the basal ganglia. It is a part of the reticulate formation (Mesulam et al. 1989) having strong interrelations with the basal ganglia system. As indicated by its name, it is located at the junction between the pons and the cerebral peduncle, lateral to the decussation of the brachium conjunctivum. The complex is not homogeneous. An important part is made up of cholinergic (Ch5)(excitatory) neurons, which is also the case for the laterodorsal tegmental nucleus (Ch6) (Mesulam et al. 1989). Other neurons are GABA. The tracing of axons from the pedunculopontine complex has shown that it ends intensively in the nigra reticulata first and to the compacta. Another strong innervation is observed in the subthalamic nucleus (Lavoie and Parent, 1994). Other targets are the pallidum (mainly medial) and the striatum. The complex receives abundant direct afferences from the medial pallidum (Percheron et al. 1998)(inhibitory).It sends axons to the pallidal territory of the lateral region VO. All this led Mena-Segovia et al. (2004) to propose that the complex be linked in a way or another to the basal ganglia system. A review on its role in the system and in diseasesis given by Pahapill and Lozano (2000). It plays an important role in awakeness and sleep. The complex must be left its double position and function. It is a part of the reticular formation. It is a regulator (regulating and being regulated) of the basal ganglia system.

= Outputs of the basal ganglia system=

Many connections of the basal ganglia are between elements of the basal ganglia. There are few output external targets. One is the superior colliculus, from the nigra lateralis. The two other major output subsystems are in the direction to the thalamus and from there to the cortex. Starting from cercopidae, the ending from the two sources of the basal ganglia are located without mixture in front of the cerebellar territory (VIm or VL)(see thalamus). From there, there is also a complete separation of medial pallidal elements from nigral. Pallidal and nigral terminal arborisations do not mix (Percheron et al. 1998). The development of the medial pallidum creates the appearance of a new distinctive pallidal nucleus, the nucleus ventralis oralis VO, lateral to the nigral VA (Percheron, 2003). This distinction is of major importance(see thalamus).

Nigra lateralis to superior colliculus

---

The nigra lateralis made up of the same cell type than the pars reticulata differs by its targets. The now well establihed connection in macaques (Jayaraman et al.1977, François et al.1984) is not given its true value. The superior colliculus indeed sends axons to the thalamus VImM, VA, Cpf with links with the oculomotor cortex.

Medial pallidum to thalamic VO and cortex

---

Axons from the pallidum to the thalamus form the ansa lenticularis and the fasciculus lenticularis, making in fact a single entity. The axons arrive at the medial face of the pallidum; from there, they cross the internal capsule where they form the comb system ("Kamm system" of Edinger, 1900). The axons arrives at the lateral border of the subthalamic nucleus. Passing above it they constitute the field H2 of Forel (1877). From there, they curve down towards the hypothalamus. At field H, they turn abruptly. This has been the cause of historical mistakes as it was thought that the bundle had to pursue its ventral course. In fact the bundle goes up in a dorsolateral direction (forming the H1 field) and reach in this manner the ventral border of the thalamus. Pallidal axons have their own thalamic territory in the lateral region of the thalamus; everywhere separated from the cerebellar and from the nigral territories. The VO nucleus remains everywhere lateral in macaques and humans. It stained for calbindin and acetylcholinesterase. The axons ascend in the nucleus where they emit branches that widespreadly distribute "bunches" of axonal branches (Arrrechi-Bouchhiouia et al.1996,1997). The distribution is such that if any somatotopical organisation exists, it may be only poor. The thalamocortical neurons of VO go preferentially to the supplementary motor cortex (SMA), to preSMA and to a lesser extent to the motor cortex. The pallidothalamic neurons also give branches to the pars media of the central complex (see above), which sends axons to the premotor and accessory motor cortex.

Nigra reticulata to thalamic VA and cortex

---

Nigral axons go up dorsally without forming a clear bundle. They reach the inferomedial border of the thalamus. The nigral territory (VA) is medial to the pallidal. It is crossed by the mammillothalamic bundle. In the monkey, the nucleus is usually divided into a magnocellular part, medial and close to the mammillothalamic bundle, and a mediocellular part. In the human brain, the majority of the nucleus is composed of the magnocellular component. In any case, in macaques, the afferences from the nigra do not care about these cytoarchitectonic subdivisions.In addition to the nigral afference, VA receives axons from the tectum (superior colliculus) and from the amygdala (basal complex), which makes a singular set of afferences. Thalamocortical axons from VA send their axons to an also particular set made up of the frontal, the cingular cortex and the oculomotor cortex (FEF and SEF), not the motor cortex. This set of output likely indicates a visuomotor role.

= Subsystems and models= Systemic representations very frequently use the "box-and-arrow model", in which boxes are elements and arrows connections.In many contributions, boxes are presummed to be homogeneous and clearly distinct one from any other and closed. The cortex in general is one box. The thalamus is reduced to a single VA/VL complex.Several models have been proposed at about the same time. Anatomical works have demonstrated that there is a strong compression of numbers of neurons meaning a numerical convergence (Yelnik et al.1984 and Percheron et al. 1984). This objective "funneling" has been attacked in the opposed model of Alexander et al.(1986), Alexander and Crutcher (1990), presented under several other forms. These authors priviledged distinct chains of anatomical connections that would escape funneling to preserve 5 to 6 "basal ganglia-thalamocortical circuits": motor, oculomotor, prefrontal (dorsolateral prefrontal and lateral orbitofrontal) and limbic (or anterior cingulate), which through the basal ganglia and the thalamus would return to the initial point of the cortex. This "parallelist" view does not fit with corticostriatal anatomy. As repeatedly proven, the corticostriate connection does not follow the Kemp and Powell topography (1970). In addition to compression there is an intrication of subsystems (e.g.the oculomotor component intricated with the frontal one). The thalamo-cortical connection do not follow simple rules and usually have several cortical targets. Another model, and probably the most famous (Albin et al.1989)(later admitted to have been too simplistic) selected two criteria: the inhibitory/excitatory character of connections and the mediator involved. Current models place the role of the subthalamic nucleus in a privileged position due to the fact that it is excitatory when the striatum, the pallidum and the nigra are inhibitory. To a "direct pathway" (cortex-striatum-medialpallidonigro-thalamo-cortical)(5-circuit) was opposed an "indirect pathway"(cortex-striatum-lateral pallidum-subthalamic nucleus-medialpallidonigral-thalamo-cortical) (6-circuit). Connections are links without defined numbers of axons and no topological varieties. This particularly complex system yet incites to more refined systemic analyses, with systems and subsystems (defined very simply as elements in interactions). The opposition between output subsystems and regulator subsystems does not correspond exactly with that between the core and the regulators. One part of the core, the lateral palliduum, is indeed a part of the core and a regulator.It does not send axons to the thalamus and from there to the cortex. All its efferent axons instead are returning inside the basal ganglia system. With its main target the subthalamic nucleus they constitutes together a particular subsystem, with two high frequency autonomous pacemakers, one inhibitory (GABA) and one excitatory (glutamate). The subthalamic nucleus sends axons to another regulator: the pedunculo-pontine complex (id). This and the central complex are elements of other basal ganglia regulating subsystems.

In addition to the nigrotectal connection there are two output subsystems starting from the striatum, forming 3-paths to the cortex. They do not send axons, then no regulatory messages, either to the striatum, lateral pallidum or subthalamic nucleus. The first output path from the striatum has its first relay in the medial pallidum (GABAegic, inhibitory). This sends axons (again GABAergic and inhibitory) to a particular place of the thalamus, the nucleus ventralis oralis VO. VO sends its axons to the accessory motor and the motor cortex (with glutamate as the mediator). The second output subsystem follows exactly the same pattern, but, this time, starts from the nigra reticulata (GABA). From there nigral axon go to the nucleus ventralis anterior VA. This VA sends axons to the frontal cortex and the oculomotor areas (glutamatergic).Striato-pallidonigral and pallidonigral-thalamic connection are inhibitory. The separation of the two subsystems is finally a differentiated cortical distribution. One of the characteristics of the basal ganglia system is thus the succession of two inhibitory connections. At first,there is not one "direct" but two output circuits in primates: one is cortex-striatum-medial pallidum- VO-SMa in one hand and the other is cortex-striatum-nigra reticulata-VA- anterior cortex in the other. Furthermore,there are many arguments against treating in the same manner the "indirect circuit" involving a regulator circuit and the output circuits. Among other ways of looking to the basal ganglia system, that indicated by morphology and physiology raises problems. The pallidonigral set (as defined above) is a high-frequency pacemaker (1) emitting inhibitory signals (2) receiving at low frequency but in possibly large numbers messages from the striatum that are also inhibitory. Adaquate striatal patterns thus might carve, by desinhibition, an appropriate pattern of signals (a message) to the thalamus and from there to the cortex.

Some axons from the lateral pallidum go to the striatum (Sato et al.2000). Above all, many of them go to other basal ganglia elements: the medial pallidum, the nigra reticulata and the subthalamic nucleus. The activity of the medial pallidum is thus influenced by afferences from the lateral pallidum and from the subthalamic nucleus (Smith, Y., Wichmann,T.,DeLong, M.R. 1994). The same holds true for the nigra reticulata (Smith, Y., Hazrati, L-N, Parent, A. 1990).

=Models of the basal ganglia system= At first,there is not one "direct" but two output circuits in primates: one is cortex-striatum-medial pallidum- VO-SMa in one hand and the other is cortex-striatum-nigra reticulata-VA- anterior cortex in the other. Furthermore,there are many arguments against treating in the same manner the "indirect circuit" involving a regulator circuit and the output circuits. Among other ways of looking to the basal ganglia system, that indicated by morphology and physiology raises problems. The pallidonigral set (as defined above) is a high-frequency pacemaker (1) emitting inhibitory signals (2) receiving at low frequency but in possibly large numbers messages from the striatum that are also inhibitory. Adaquate striatal patterns thus might carve, by desinhibition, an appropriate pattern of signals (a message) to the thalamus and from there to the cortex.

=References=