Synaptic Integration in Cerebellar Granule Cells
- First Online:
- Cite this article as:
- Ekerot, C. & Jörntell, H. Cerebellum (2008) 7: 539. doi:10.1007/s12311-008-0064-6
- 191 Views
To understand the function of cerebellar granule cells, we need detailed knowledge about the information carried by their afferent mossy fibers and how this information is integrated by the granule cells. Recently, we made whole cell recordings from granule cells in the non-anesthetized, decerebrate cats. All recordings were made in the forelimb area of the C3 zone for which the afferent and efferent connections and functional organization have been investigated in detail. Major findings of the study were that the mossy fiber input to single granule cells was modality- and receptive field-specific and that simultaneous activity in two and usually more of the afferent mossy fibers were required to activate the granule cell spike. The high threshold for action potentials and the convergence of afferents with virtually identical information suggest that an important function of granule cells is to increase the signal-to-noise ratio of the mossy fiber–parallel fiber information. Thus a high-sensitivity, noisy mossy fiber input is transformed by the granule cell to a high-sensitivity, low-noise signal.
KeywordsSensoryGranule cellsMossy fibers
Cerebellar granule cells comprise about half of the number of neurons in our nervous system. In several respects, the properties of granule cells are unique. The size of the cell body is only in the order of 5–7 μm and a granule cell is contacted by only by three to five afferent mossy fibers. The unique properties of granule cells lead to the formulation of different hypotheses of their function. Braitenberg and Onesto  proposed, based on the length and slow conduction velocity of parallel fibers, that they function as delay lines and that granule cells played an important role for timing. Based on the large number of granule cells and the limited numbers of mossy fibers converging onto single granule cells, Marr  suggested that the function of the cerebellar cortical network was in pattern recognition. He suggested that the threshold for action potentials in granule cells was high and that the mossy fiber input to granule cells were randomly organized. Thus, an event causing activation of mossy fibers would activate a relatively limited number of granule cells. The pattern of activated granule cells would be specific for the event.
To disclose the function of granule cells, we have studied these cells in a well-characterized cerebellar system. From previous investigations, we have a detailed knowledge of the functional organization of the forelimb area of the C3 zone and its afferent and efferent connections . The zone is divided into about 40 narrow microzones, each characterized by the receptive field of its climbing fiber input. The forelimb area of the C3 zone receives mossy fiber afferents from four main sources, the exteroceptive and proprioceptive components of the cuneo-cerebellar tract (E-CCT and P-CCT), which carry information from peripheral receptors, the lateral reticular nucleus (LRN) monitoring activity in segmental interneuronal circuits, and finally from pontine nuclei mossy fibers carrying information from the cerebral cortex. The cutaneous mossy fiber input from the E-CCT has been studied in great detail. These mossy fibers have small cutaneous receptive fields, similar to those of climbing fibers to the C3 zone. Mossy fibers and climbing fibers with similar receptive fields terminate in overlapping narrow microzones .
The cutaneous parallel fiber input to Purkinje cells and inhibitory interneurons was recently studied in detail . The parallel fiber receptive fields of the individual neurons were similar to those of single mossy fibers. This finding in itself suggested that mossy fiber afferents with the same receptive field converge onto single granule cells. Also, as a reflection of the parallel fiber synaptic plasticity mechanisms, the relationship between climbing fiber and mossy fiber receptive fields differed between Purkinje cells and inhibitory interneurons. In Purkinje cells, the parallel fiber receptive field never overlapped that of the climbing fiber whereas in interneurons the parallel fiber receptive fields were almost identical to that of locally terminating climbing fibers.
Analysis of the evoked responses in a large number of granule cells indicated that all afferent mossy fibers to a granule cell were of the same modality and had similar peripheral receptive field. For granule cells with cutaneous input, it was also possible to show that the afferent mossy fibers were activated by the same submodality. The modality–specificity of the mossy input to single granule cells seen in the present study suggest that mossy fibers with different modalities have an orderly termination in the granule layer. Our data showed that this was the case: granule cells with cutaneous input were located in the superficial part of the granule layer, cells with muscle and joint input and cells with input from interneuronal circuits at intermediate depths. Granule cells in the deepest part of the granule layer were not driven by peripheral stimulation but had a spontaneous mossy fiber EPSP activity. The latter granule cells were probably receiving information from mossy fibers originating from the pontine nuclei which were disconnected from their driving input by the decerebration.
Analysis of the spontaneous and evoked mossy fiber activity indicated that at least two and usually several mossy fiber afferents had to be simultaneously active to reach the threshold value for action potential generation (see Fig. 1). In no case were granule cells found to be activated by a single mossy fiber. This was also directly tested by using electrical white matter stimulation to activate single mossy fiber inputs in frequencies up to 1,000 Hz . Recently, a paper was published claiming that a granule cell can be activated by a single afferent mossy fiber . However, in this paper, it is important to note that this issue was tested directly only in vitro, even though mossy fibers were activated in patterns that had previously been recorded in vivo. Since the time constants of mf-grc EPSPs in vitro are substantially longer and their amplitudes much larger than in vivo, the effectiveness of the synaptic summation is greatly exaggerated in this case.
What is the function of granule cells? One obvious function is to distribute the mossy fiber input to as many cells as possible. The total number of synapses formed by parallel fibers on Purkinje cell dendrites in humans reaches astronomical numbers, and is in the order of 5,000 billions . In addition to the synapses on Purkinje cells, parallel fibers make synapses with all other types of neurons in the cerebellar cortex. Recent studies have shown that only a few percent of parallel fiber synapses on Purkinje cells are electrically effective and that these synapses have been selected by climbing fiber controlled plasticity [5, 9, 10]. The large number of “silent” parallel fiber synapses on Purkinje cells secures a high learning capacity .
A second very important role of granule cells is to decrease the signal-to-noise ratio of the mossy fiber input. This is necessary since already the individual peripheral mechanoreceptor is subjected to thermodynamic noise and the spike generator of all neurons, including the precerebellar neurons, is subjected to stochastic noise. The high threshold for action potentials in granule cells and the convergence of mossy fibers with virtually identical information would filter out temporally uncorrelated events in the signals from mossy fibers. This makes it possible to use a high-sensitivity, noisy input, to mossy fibers without deteriorating the signal-to-noise ratio of the granule cell output. For instance, spontaneous, asynchronous mossy fiber activity will be filtered out and have little influence on granule cell output.
This work was supported by the Swedish Research Council (projects K2006-04X-08291-19-3 and K2005-04X-14780-03A) and EU (SENSOPAC, project no. 028056).