,letter to the editor Parallel processing in the basal ganglia: up to a point SIR:
In a recent special issue Alexander and Crutcher ~ argue that the architecture of the basal ganglia provides the 'substrate' for 'parallel processing' of information within separate neuronal 'loops'. Goldman-Rakic and Selemon2 endorse this view. Chevalier and Deniau 3 suggest contrarily that different aspects of striatal processing appear to be 'compiled in the same output pathway, perhaps even within single neurons'. Alexander and Crutcher have neglected our contribution to the demonstration of convergence in the core of the basal ganglia. 'Parallel processing' must not be opposed to 'funnelling', which is not 'a past view of basal ganglia organization '~. 'Funnelling' is a recent term and to understand the concept it is necessary to distinguish various kinds of information processing (Fig. 1), and two successive corticostriatal and striatopallido-nigral processing steps. The corticostriatal funnelling !s due mainly to the 'intrication 4 (Fig. 1) of islands of axonal endings from wide cortical regions within the striatum. Percheron and collaborators 5-7 made the cartographic demonstration that axonal endings from cortical neurons sharply delineate two distinctly distributed, intricated territories within the striatum: the sensorimotor and the associative territories. The segregation of axons originating in these two territories is maintained in the palliclum and the substantia nigra. Thus, the single consideration of the axonal trajectories might lead to the impression that there are separate sub-systems. Recently, however, Percheron and collaborators 5--9 demonstrated, using new morphological methods, that the striatopallidonigral funnelling is not simply due to a strong reduction in the number of neurons from the source to the target (cardinal convergence) (Fig. 1), but above all to a drastic contrast between the threedimensional geometry of the emitting axonal and receiving dendritic arborizations. The wide TINS, Vol. 14, No. 2, 1991
dendritic arborization of pallidal and nigral neurons, oriented at right angles to the almost unbranched incoming striatal axons, leads to the confluence of information from the sensorimotor and associative territories on many neurons, and to an extreme reception convergence (Fig. 1). We believe that the morphologically measured informational convergence must be taken as a static datum that possibly corresponds to the maximal aperture of a focusing system. In given conditions, the striatopallidonigral system seems to focus information in time and threedimensional space so as to give relatively localized responses. Dynamically focused convergence is the probable essential informational device of the striatopallido-nigral system.
SYSTEMIC DI SIRIBUTION
The projection from the sensorimotor cortex to the putamen is only roughly somatotopically arranged 9'1°, as shown by threedimensional cartography 5-7. However, Alexander and Crutcher 1 maintain that there is a 'pronounced degree of functional specificity and somatotopic coding among neurons at all stages of the motor circuit'. The occurrence of some somatotopy in the pallidum of monkeys has been confirmed recently 11. However, the monkeys were submitted to the same, probably too narrow, experimental protocol. The correlations between discharges of single pallidal neurons and movements of distinct parts of the body can hardly be used to rule out striatopallidal convergence. Electrophysiological results that were obtained by Filion and his
SYSTEMIC SEPARATION
source CONFLUENCE
INTRICATION
INPUT source(s)
Z
V
,~,,...,,..,..H..~,
NNNN
i
V OUTPUT ta rget(s) processing
addressing
PARALLEL PROCESSING
isolating
cardinal CONVERGENCE
mixing
emission CONVERGENCE
reception CONVERGENCE
INPUT I source
V
i
.
I
I
I
I
V OUTPUT processing
sharing
compressing
Fig. 1. The first row of diagrams explains various kinds of connections between sources and targets and their effect on the output. The second row shows various kinds of information processing due to the organization of neuronal elements at the interface between input and output. These elements might be segregated, they might show cardinal or numerical differences, or the geometric relationships between axonal and dendritic arborizations might differ. Applications to basal ganglia circuitry are made in the text. Abbreviations: ax arbor, axonal arborizations; dd arbor, dendritic arborization. 55
letter
to the editor colleagues ~2-~4 bring strong arguments in favor of such a convergence. A large number of pallidal neurons were shown to be influenced individually by passive manipulation of several joints of both upper and lower, and ipsiand contralateral limbs in monkeys. Admittedly, the latter monkeys had been treated with the neurotoxin 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP), and so much convergence was not characteristic of intact monkeys. But the dopaminergic denervation apparently resulted in the wide opening of the striatopallidal focusing system, unmasking the morphologically quantified convergence. Striatopallidal convergence at a neuronal level was also demonstrated after electrical stimulation of the striatum in the monkey 13. Again, it was exaggerated after MPTP treatment 14. The parallel processing metaphor, which implies the concomitant realization of the same task ~5, should be used more cautiously. The theory of parallel processing in the basal ganglia is questionable when it relies on adimensional 'box-and-arrow' diagrams. It is excessive when it discards any data in favor of convergence. We would like to challenge Alexander and Crutcher on the hypothesis that in monkeys left free to use their full motor repertoire the same pallidal or nigral neuron might respond to signals from different parts of the body, and might integrate 'motor', 'oculomotor', 'limbic' and two types of 'prefrontal' signals, according to context. Gdrard Percheron INSERM Laboratoire de Neuromorphologieinformationnelleet de NeurologieexpEfimentale du mouvement,Pav.ClaudeBernard,H6pitalde la Sa/pEtriEre,47 Boulevardde I'H6pital, 75651 Paris Cedex 13, France. Michel Filion Centre de Rechercheen Neurobiologie,H6pital de/'Enfant JEsus, 1401, 18EmeRue, QuEbec GIJ 1Z4, Canada.
References 1 Alexander, G. E. and Crutcher, M. D. (1990) Trends Neurosci. 13,266-271 2 Goldman-Rakic, P. S. and Selemon, L. D. (1990) Trends Neurosci. 13, 241-244 3 Chevalier, G. and Deniau, J. M. (1990) Trends Neurosci. 13,277-280
56
4 Goldman, P. S. and Nauta, W. J. H. (1977) J. Comp. Neurol. 171, 369-386 5 Percheron,G., Yelnik,J. and Francois, C. (1984)J. Comp. Neurol. 227, 214-227 6 Percheron,G., Yelnik,J. and Francois, C. (1984) in The Basal Ganglia (McKenzie, J. S., Kemm, R. E. and Wilcock, L. N., eds), pp. 87-105, Plenum Press 7 Percheron,G,, Franqois,C. and Yelnik, J. (1987) in The Basal Ganglia II (Carpenter, M. B. and Jarayaman,A., eds), pp. 205-226, Plenum Press 8 Yelnik,J., Percheron,G. and Fran(~ois, C. (1984) J. Comp. Neurol, 227, 200-213 9 Fran~:ois,C., Yelnik, J. and Percheron, G. (1987) J. Comp. Neurol. 265, 473-493 10 Ki~nzle,H. (1977) Exp. Brain Res. 30, 481-492 11 Nambu, A., Yoshida,S. and Jinnai, K. (1990) Brain Res. 519, 183-191 12 Filion, M., Tremblay, L. and B~dard, P. J. (1988) Brain Res. 444, 165-176 13 Tremblay, L. and Filion, M. (1989) Brain Res. 498, 1-16 14 Tremblay, L., Filion, M. and B~dard, P. J. (1989) Brain Res. 498, 17-33 15 Nelson, M. E., Furmanski, W. and Bower, J. M. (1989) in Methods in Neuronal Modeling (Koch, C. and Segev, I., eds), pp. 397-437, MIT Press
some of their observations. Moreover, they ignore or dismiss an enormous amount of anatomical and physiological data that are simply incompatible with their perspective. Their central argument is that 'the wide dendritic arborization of pallidal and nigral neurons, oriented at right angles to the almost unbranched incoming striatal axons, leads to the confluence of information from the sensorimotor and associative territories on many neurons, and to an extreme receptive convergence 'I. While the axons of striatal projection neurons do cut perpendicularly across the dendritic fields of pallidal neurons en route to their ultimate targets, the terminal arborizations of the striatopallidal axons are oriented largely in parallel with these same dendritic fields 3,4, and branches from a single striatal efferent fiber can follow a given pallidal dendrite and its branches over a considerable distance, making numerous synaptic contacts along the way s . Moreover, electron microscope studies have shown that the main, radially oriented axons ('radial fibers') of striatoReply pallidal neurons remain myelinSIR: ated as they course through the Percheron and Filion I have drawn pallidum, and that less than 10% attention to an important issue of the synaptic contacts within that was addressed only cursorily the pallidum are derived from in our recent reviewL; that is, myelinated axons 5. It seems to us, what is the nature of, and struc- therefore, that in the face of clear tural basis for, information pro- tract-tracing evidence to the concessing within the basal ganglia trary 4'6, it is implausible for output nuclei (globus pallidus and Percheron and Filion to infer that substantia nigra)? They contend individual pallidal neurons are that we have neglected their evi- likely to receive 'a confluence of dence concerning the degree of information from sensorimotor structural convergence and func- and associative territories', simply tional integration within the basal on the basis that the dendritic ganglia, and insist that their find- fields of such neurons might be ings provide strong evidence crossed by incoming radial fibers against the concept of parallel, that originate from different functionally segregated basal regions within the striatum. After ganglia-thalamocortical circuitry. all, there are examples throughThe data of Percheron and Filion out the brain of nuclei, e.g. in are certainly relevant to the issue various sensory systems, in which of integration versus segregation the incoming axons of topoof information within the basal graphically organized projections ganglia, and they do deserve cross through one functional more attention than was possible region before terminating in in our brief review. But their another, and such arrangements conclusions are based on rather are known not to nullify the questionable inferences concern- corresponding visual, auditory or ing the functional significance of somesthetic maps. TINS, VoL 14, No. 2, 1991
In a recent special issue Alexander and Crutcher ~ argue that the architecture of the basal ganglia provides the 'substrate' for 'parallel processing' of information within separate neuronal 'loops'. Goldman-Rakic and Selemon2 endorse this view. Chevalier and Deniau 3 suggest contrarily that different aspects of striatal processing appear to be 'compiled in the same output pathway, perhaps even within single neurons'. Alexander and Crutcher have neglected our contribution to the demonstration of convergence in the core of the basal ganglia. 'Parallel processing' must not be opposed to 'funnelling', which is not 'a past view of basal ganglia organization '~. 'Funnelling' is a recent term and to understand the concept it is necessary to distinguish various kinds of information processing (Fig. 1), and two successive corticostriatal and striatopallido-nigral processing steps. The corticostriatal funnelling !s due mainly to the 'intrication 4 (Fig. 1) of islands of axonal endings from wide cortical regions within the striatum. Percheron and collaborators 5-7 made the cartographic demonstration that axonal endings from cortical neurons sharply delineate two distinctly distributed, intricated territories within the striatum: the sensorimotor and the associative territories. The segregation of axons originating in these two territories is maintained in the palliclum and the substantia nigra. Thus, the single consideration of the axonal trajectories might lead to the impression that there are separate sub-systems. Recently, however, Percheron and collaborators 5--9 demonstrated, using new morphological methods, that the striatopallidonigral funnelling is not simply due to a strong reduction in the number of neurons from the source to the target (cardinal convergence) (Fig. 1), but above all to a drastic contrast between the threedimensional geometry of the emitting axonal and receiving dendritic arborizations. The wide TINS, Vol. 14, No. 2, 1991
dendritic arborization of pallidal and nigral neurons, oriented at right angles to the almost unbranched incoming striatal axons, leads to the confluence of information from the sensorimotor and associative territories on many neurons, and to an extreme reception convergence (Fig. 1). We believe that the morphologically measured informational convergence must be taken as a static datum that possibly corresponds to the maximal aperture of a focusing system. In given conditions, the striatopallidonigral system seems to focus information in time and threedimensional space so as to give relatively localized responses. Dynamically focused convergence is the probable essential informational device of the striatopallido-nigral system.
SYSTEMIC DI SIRIBUTION
The projection from the sensorimotor cortex to the putamen is only roughly somatotopically arranged 9'1°, as shown by threedimensional cartography 5-7. However, Alexander and Crutcher 1 maintain that there is a 'pronounced degree of functional specificity and somatotopic coding among neurons at all stages of the motor circuit'. The occurrence of some somatotopy in the pallidum of monkeys has been confirmed recently 11. However, the monkeys were submitted to the same, probably too narrow, experimental protocol. The correlations between discharges of single pallidal neurons and movements of distinct parts of the body can hardly be used to rule out striatopallidal convergence. Electrophysiological results that were obtained by Filion and his
SYSTEMIC SEPARATION
source CONFLUENCE
INTRICATION
INPUT source(s)
Z
V
,~,,...,,..,..H..~,
NNNN
i
V OUTPUT ta rget(s) processing
addressing
PARALLEL PROCESSING
isolating
cardinal CONVERGENCE
mixing
emission CONVERGENCE
reception CONVERGENCE
INPUT I source
V
i
.
I
I
I
I
V OUTPUT processing
sharing
compressing
Fig. 1. The first row of diagrams explains various kinds of connections between sources and targets and their effect on the output. The second row shows various kinds of information processing due to the organization of neuronal elements at the interface between input and output. These elements might be segregated, they might show cardinal or numerical differences, or the geometric relationships between axonal and dendritic arborizations might differ. Applications to basal ganglia circuitry are made in the text. Abbreviations: ax arbor, axonal arborizations; dd arbor, dendritic arborization. 55
letter
to the editor colleagues ~2-~4 bring strong arguments in favor of such a convergence. A large number of pallidal neurons were shown to be influenced individually by passive manipulation of several joints of both upper and lower, and ipsiand contralateral limbs in monkeys. Admittedly, the latter monkeys had been treated with the neurotoxin 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP), and so much convergence was not characteristic of intact monkeys. But the dopaminergic denervation apparently resulted in the wide opening of the striatopallidal focusing system, unmasking the morphologically quantified convergence. Striatopallidal convergence at a neuronal level was also demonstrated after electrical stimulation of the striatum in the monkey 13. Again, it was exaggerated after MPTP treatment 14. The parallel processing metaphor, which implies the concomitant realization of the same task ~5, should be used more cautiously. The theory of parallel processing in the basal ganglia is questionable when it relies on adimensional 'box-and-arrow' diagrams. It is excessive when it discards any data in favor of convergence. We would like to challenge Alexander and Crutcher on the hypothesis that in monkeys left free to use their full motor repertoire the same pallidal or nigral neuron might respond to signals from different parts of the body, and might integrate 'motor', 'oculomotor', 'limbic' and two types of 'prefrontal' signals, according to context. Gdrard Percheron INSERM Laboratoire de Neuromorphologieinformationnelleet de NeurologieexpEfimentale du mouvement,Pav.ClaudeBernard,H6pitalde la Sa/pEtriEre,47 Boulevardde I'H6pital, 75651 Paris Cedex 13, France. Michel Filion Centre de Rechercheen Neurobiologie,H6pital de/'Enfant JEsus, 1401, 18EmeRue, QuEbec GIJ 1Z4, Canada.
References 1 Alexander, G. E. and Crutcher, M. D. (1990) Trends Neurosci. 13,266-271 2 Goldman-Rakic, P. S. and Selemon, L. D. (1990) Trends Neurosci. 13, 241-244 3 Chevalier, G. and Deniau, J. M. (1990) Trends Neurosci. 13,277-280
56
4 Goldman, P. S. and Nauta, W. J. H. (1977) J. Comp. Neurol. 171, 369-386 5 Percheron,G., Yelnik,J. and Francois, C. (1984)J. Comp. Neurol. 227, 214-227 6 Percheron,G., Yelnik,J. and Francois, C. (1984) in The Basal Ganglia (McKenzie, J. S., Kemm, R. E. and Wilcock, L. N., eds), pp. 87-105, Plenum Press 7 Percheron,G,, Franqois,C. and Yelnik, J. (1987) in The Basal Ganglia II (Carpenter, M. B. and Jarayaman,A., eds), pp. 205-226, Plenum Press 8 Yelnik,J., Percheron,G. and Fran(~ois, C. (1984) J. Comp. Neurol, 227, 200-213 9 Fran~:ois,C., Yelnik, J. and Percheron, G. (1987) J. Comp. Neurol. 265, 473-493 10 Ki~nzle,H. (1977) Exp. Brain Res. 30, 481-492 11 Nambu, A., Yoshida,S. and Jinnai, K. (1990) Brain Res. 519, 183-191 12 Filion, M., Tremblay, L. and B~dard, P. J. (1988) Brain Res. 444, 165-176 13 Tremblay, L. and Filion, M. (1989) Brain Res. 498, 1-16 14 Tremblay, L., Filion, M. and B~dard, P. J. (1989) Brain Res. 498, 17-33 15 Nelson, M. E., Furmanski, W. and Bower, J. M. (1989) in Methods in Neuronal Modeling (Koch, C. and Segev, I., eds), pp. 397-437, MIT Press
some of their observations. Moreover, they ignore or dismiss an enormous amount of anatomical and physiological data that are simply incompatible with their perspective. Their central argument is that 'the wide dendritic arborization of pallidal and nigral neurons, oriented at right angles to the almost unbranched incoming striatal axons, leads to the confluence of information from the sensorimotor and associative territories on many neurons, and to an extreme receptive convergence 'I. While the axons of striatal projection neurons do cut perpendicularly across the dendritic fields of pallidal neurons en route to their ultimate targets, the terminal arborizations of the striatopallidal axons are oriented largely in parallel with these same dendritic fields 3,4, and branches from a single striatal efferent fiber can follow a given pallidal dendrite and its branches over a considerable distance, making numerous synaptic contacts along the way s . Moreover, electron microscope studies have shown that the main, radially oriented axons ('radial fibers') of striatoReply pallidal neurons remain myelinSIR: ated as they course through the Percheron and Filion I have drawn pallidum, and that less than 10% attention to an important issue of the synaptic contacts within that was addressed only cursorily the pallidum are derived from in our recent reviewL; that is, myelinated axons 5. It seems to us, what is the nature of, and struc- therefore, that in the face of clear tural basis for, information pro- tract-tracing evidence to the concessing within the basal ganglia trary 4'6, it is implausible for output nuclei (globus pallidus and Percheron and Filion to infer that substantia nigra)? They contend individual pallidal neurons are that we have neglected their evi- likely to receive 'a confluence of dence concerning the degree of information from sensorimotor structural convergence and func- and associative territories', simply tional integration within the basal on the basis that the dendritic ganglia, and insist that their find- fields of such neurons might be ings provide strong evidence crossed by incoming radial fibers against the concept of parallel, that originate from different functionally segregated basal regions within the striatum. After ganglia-thalamocortical circuitry. all, there are examples throughThe data of Percheron and Filion out the brain of nuclei, e.g. in are certainly relevant to the issue various sensory systems, in which of integration versus segregation the incoming axons of topoof information within the basal graphically organized projections ganglia, and they do deserve cross through one functional more attention than was possible region before terminating in in our brief review. But their another, and such arrangements conclusions are based on rather are known not to nullify the questionable inferences concern- corresponding visual, auditory or ing the functional significance of somesthetic maps. TINS, VoL 14, No. 2, 1991
