Anatomical structure alone cannot predict function

The central hypothesis of the target article that tidal waves of parallel fiber excitation precisely activate Purkinje cell spiking is hard to reconcile with recent neurophysiological and modeling data. The assumed pattern of mossy fiber input seems unrealistic, inhibition is likely to interfere with the proposed excitatory responses, and moreover, computer simulations show that the Purkinjecell is a poor coincidence detector. While providing an interesting framework for understanding cerebellar function, the theory proposed by Braitenberg and colleaguesis limited mostly to neuroanatomical speculation. In this comment we will show that recent physiological and modeling data cast doubt on three essential components of the theory: the pattern of mossy fiber input, the purely lateral inhibition and the sensitivity of Purkinje cells to waves of parallel fiber input. Unless major enhancements to the theory are made, and direct experimental evidence in vivo is found, we find it unlikely that the cerebellum is operating according to the principle of parallel fiber tidal waves. The authors propose that only those spatiotemporal patterns of mossy fiber input that evoke waves of parallel fiber activity are effective in stimulating Purkinje cells. The simplicity of this system may seem attractive, but if it is true most patterns of mossy fiber input would not change cerebellar output at all. Even if the fractured somatotopic maps of the granule cell layer (Shambes et al. 1978) were set up to generate such waves, one would be able to code at most a few input-output sequences for each beam. Moreover, it has been shown that these fractured somatotopic maps show very little plasticity (Gonzalez et al. 1993), which means that an animal would be born with a limited fixed repertoire of effective input sequences. This seems a rather inefficient way to control the complex adaptive dynamics of the motor system. A second component of the theory is the purely excitatory response along the hypothetical narrow parallel fiber beam. The authors specifically dismiss the presence of inhibitory responses of Purkinje cells within this beam. Instead, inhibition is suggested to suppress activation of Purkinje cells lateral to the beam. We believe that available experimental data contradictthis hypothesis. Anatomically, stellate cell axons connect to Purkinje cells inthe immediate vicinity of parallel fibers activating these cells (Sultan et al.1995), which does not indicate a spared region of pure excitation. Even earlyrecordings looking for beams of activated Purkinje cells with direct electricalparallel fiber stimulation in primates (Bloedel et al. 1972), foundpredominantly inhibitory responses along the center of the activated parallelfiber beam. A similar finding was made in the anesthetized rat, when parallelfibers were activated with sensory stimulation (Bower & Woolston 1983).The final component of the theory is the "decoder", i.e. the Purkinje cellwhich is supposed to act as a coincidence detector responding mostly towaves of parallel fiber activity. The authors’ own experimental data, however,show that the Purkinje cell seems to be a poor decoder of wave-like parallelfiber activity. Their figure 10B shows that for identical "ideal" inputs thelatency of the Purkinje cell response varied by 7 ms and had a 20% failurerate. In fact, input movement in the "wrong" direction sometimes causedresponses with similar latencies to 40% of the stimuli (Fig. 10C). Withresponses already so variable in the slice preparation where Purkinje cells aremostly silent, one wonders how these cells firing at rates of 40 Hz andhigher in vivo could reliably signal the presence of a wave during actualmotor behavior. These experimental results match our modeling studieswhich suggest that even in vivo Purkinje cells can be quite sensitive to fastchanges in the rate of parallel fiber input, but with a large jitter of theresponse (De Schutter 1994) and with little sensitivity to actual coincidenceof the input (unpublished results). In our view it is unlikely that the Purkinjecell is a coincidence detector, instead it performs a complex integration overtime of excitatory and inhibitory input (Jaeger et al. 1996). ReferencesBloedel, J. R., Gregory, R. S. & Martin, S. H. (1972) Action of interneuronsand axon collaterals in cerebellar cortex of a primate. Journal ofNeurophysiology 35: 847-863.Bower, J. M. & Woolston, D. C. (1983) Congruence of spatial organization oftactile projections to granule cell and Purkinje cell layers of cerebellarhemispheres of the albino rat: vertical organization of cerebellar cortex.Journal of Neurophysiology 49: 745-766.Clark, B. & Häusser, M. (1995) Synaptic inhibition of cerebellar Purkinje cellsby single interneurons. Abstracts Society for Neuroscience 21: 595.De Schutter, E. (1994) Modelling the cerebellar Purkinje cell: Experiments incomputo. In The self-organizing brain: from growth cones to functionalnetworks, ed. J. van Pelt, M. A. Corner, H. B. M. Uylings & F. H. Lopes daSilva, Progress in Brain Research 102: 427-441.De Schutter, E. & Bower, J. M. (1994) An active membrane model of thecerebellar Purkinje cell: II. Simulation of synaptic responses. Journal ofNeurophysiology 71: 401-419.Gonzalez, L., Shumway, C., Morissette, J. & Bower, J. M. (1993)Developmental plasticity in cerebellar tactile maps: fractured maps retain a fractured organization. The Journal of Comparative Neurology 332:487-498.Heck, D. (1994) Rat cerebellar cortex in vitro responds specifically to movingstimuli. Neuroscience Letters 157: 95-98.Jaeger, D., De Schutter, E. & Bower, J. M. (1996) The role of synaptic andvoltage-gated currents in the control of Purkinje cell spiking: a modelingstudy. The Journal of Neuroscience in print.Shambes, G. M., Gibson, J. M. & Welker, W. (1978) Fractured somatotopy ingranule cell tactile areas of rat cerebellar hemispheres revealed bymicromapping. Brain Behavior and Evolution 15: 94-140.Sultan, F., Ellisman, M. H. & Bower, J. M. (1995) Quantitative anatomicalaspects of the inhibitory interneurons of the rat cerebellar molecular layer: alight and electron microscopic Golgi study. Abstracts Society forNeuroscience 21: 1191.