Information

Why cerebellar input fibers use 2 ways to send a siganl to DCN?

Why cerebellar input fibers use 2 ways to send a siganl to DCN?

Both groups of input fibers of cerebellum (mossy, climbing) start 2 pathways: 1) project directly to the deep nuclei 2) project to cerebellar cortex, which then (after some processing) sends projection back to deep nuclei through Purkinje cells.

What is the purpose of such an architecture?

And closely related question: is the input information to DCN (the first pathway) and to cerebellar cortex (the second pathway) the same?


Short answers:

1) The true purpose of this architecture is not known (yet).

2) Yes, it is the same information, since the signal comes from the same axon (for climbing fibers see e.g. De Zeeuw et al., 1997).

Comment on the first answer:

Of course, though, there are many theories about the function of the cerebellar circuits, including these collaterals of the mossy and the climbing fibers. I would refer to the articles on wikipedia and scholarpedia as entry points.

To give you an extremely simplified scenario where such an architecture could be useful:

Let's say the mossy fibers contain the motor plan, coming from the cerebral cortex and the climbing fibers contain the true state of the limbs coming from the peripheral sensors (formulated like this, these statements are certainly not true). Now the direct signals to the deep cerebellar nuclei (DCN) could be used to quickly compute a rough "movement error" (i.e. planned trajectory vs. real trajectory) that already initiates a quick, rough movement correction. The signals to the cerebellar cortex would be used to compute an exact "movement error" which is also integrated at the DCN "a bit later" (i.e. some synaptic transmissions later) to further refine the movement correction.

Again, this is certainly not what is going on in the cerebellum but might give you some intuition of why such a circuitry could exist.

References

  • De Zeeuw CI, Van Alphen AM, Hawkins RK, Ruigrok TJ. Climbing fibre collaterals contact neurons in the cerebellar nuclei that provide a GABAergic feedback to the inferior olive. Neuroscience. 1997 Oct; 80(4):981-6.

Discussion

The ability to learn properly timed outputs is central to cerebellar function. We have used a large-scale computer simulation of the cerebellum and eyelid conditioning experiments to demonstrate and characterize temporal subtraction, a network mechanism of temporal coding in cerebellar cortex. We showed that the cessation of one of two ongoing mossy fiber inputs to a cerebellar simulation produces a robust and specific temporal code in the population of granule cells. Because this granule cell code is unique to the interval after the cessation of the shorter mossy fiber input, i.e., the offset interval, it predicts precisely timed learning for mossy fiber inputs that are otherwise too long to normally support learning. We showed that cerebellar-dependent learning in rabbits supports these unusual predictions of the simulation. We then characterized how each input contributes to learning under these atypical conditions. Finally, we showed how feedforward inhibition present in the synaptic organization of the cerebellum (Eccles et al., 1967 Palkovits et al., 1971, 1972 Ito, 1984) may contribute to this enhanced temporal code and learning of precisely timed responses.

Temporal subtraction via feedforward inhibition adds to the list of connectivity-based mechanisms that can contribute to the generation of temporal codes, which includes recurrent inhibition, feedback inhibition, and feedforward excitation (Durstewitz et al., 2000 Ikegaya et al., 2004 Mauk and Buonomano, 2004 Buhusi and Meck, 2005 D'Angelo et al., 2009). Recurrent inhibition has been invoked as a means to generate oscillations in the activity of individual cells that could be used in timing (Buhusi and Meck, 2005 D'Angelo et al., 2009). Similarly, feedforward excitation has been hypothesized as a means to generate delay lines or synfire chains (Durstewitz et al., 2000 Ikegaya et al., 2004). In this case, a series of connected neurons generates temporal codes by activating their downstream targets in a highly regulated chain. Previous simulations of the cerebellum have also highlighted the ability of sparse (non-recurrent) feedback inhibition to generate time-varying activity (Medina et al., 2000). Although feedback inhibition and feedforward excitation certainly operate in this simulation of the cerebellum, the present results also illustrate how the feedforward inhibition that is present in many brain regions (Lawrence and McBain, 2003 Swadlow, 2003 Mauk and Buonomano, 2004 Tepper et al., 2004) can contribute to and enhance temporal coding for patterns that involve one input ending before the other.

Temporal subtraction requires only that one of two ongoing mossy fiber inputs terminate, and thus it may represent an often engaged computational principle in the cerebellum. Subtraction may occur during the presentation of peripheral stimuli, such as tones, in which certain mossy fibers respond phasically to the onset of the tone and others respond for the duration of the tone (Boyd and Aitkin, 1976 Aitkin and Boyd, 1978). The stimuli used in eye movement studies produce variety in the temporal pattern of mossy fiber input (Chubb et al., 1984 Lisberger and Pavelko, 1986 Ramachandran and Lisberger, 2006), which may also give opportunity for subtraction. Similarly, the subtraction pattern of inputs is reminiscent of the pattern of inputs that the cerebellum appears to be presented with during trace eyelid conditioning, in which the offset of the CS and onset of the US is separated by a stimulus-free trace interval. Specifically, our results are consistent with evidence that the cerebellum uses mossy fiber input driven by the forebrain that persists to the US and tone-driven input that is active only for the duration of the tone (e.g., a subtraction pattern of input) to generate granule cell activity during the trace interval (Kalmbach et al., 2009, 2010b). Furthermore, they suggest that this coding during trace eyelid conditioning depends on feedforward inhibition that is driven by the tone-driven mossy fiber input.

These data underscore the importance of a precise temporal code for cerebellar learning. LTD of the granule cell-to-Purkinje cell synapse is a form of synaptic plasticity implicated in cerebellar learning (Ito and Kano, 1982 Linden, 1994 Ito, 2001). The induction of LTD requires a specific temporal relationship between the activation of granule cell and climbing fiber synapses onto Purkinje cells granule cell synapses must be active within ∼300 ms of a climbing fiber input to undergo LTD (Chen and Thompson, 1995 Wang et al., 2000 Safo and Regehr, 2008). Conversely, granule cell synapses active in the absence of climbing fiber input undergo LTP (Lev-Ram et al., 2003). Because the activation of weak synapses would tend to increase cerebellar output, synapses that have undergone LTD would be important for the initiation of responses. Although we observed granule cell activity within the window for LTD induction during training with both inputs presented for 3.5 s, this activity was not unique to the LTD induction window. These granule cells were active in the time window for the induction of both LTD (because they were active within ∼300 ms of a climbing fiber input) and LTP (because they were active for seconds in the absence of climbing fiber input). As such, the strength of these synapses either did not change or even increased. This interpretation may explain why learning declines as the length of the CS used in conditioning increases (Schneiderman and Gormezano, 1964 Ohyama et al., 2003): longer CSs are incapable of generating granule cell activity that is specific to the time interval for LTD induction. Conversely, during presentation of mossy fiber input in the subtraction pattern to our cerebellar simulation, there were synapses active precisely in the window for the induction of LTD. This observation suggests that the cessation of one of two ongoing mossy fiber inputs enables learning by creating activity in a subset of granule cells that is unique to the period for LTD induction.

Our results also suggest how the unipolar brush cells (UBCs) found within cerebellar cortex may contribute to temporal coding. UBCs receive a single mossy fiber input and contact granule cells, Golgi cells, and other UBCs (Nunzi et al., 2001). In vitro evidence suggests that UBCs are capable of converting a phasic mossy fiber input into a tonic train of action potentials lasting hundreds of milliseconds (Diño et al., 2000 Nunzi et al., 2001). Thus, although these cells were not included in the cerebellar simulation, they may be well suited to transform phasic mossy fiber input into prolonged “mossy fiber inputs” with variable firing durations, which could in turn engage the subtraction mechanism and enhance temporal coding and learning. In this way, UBCs may act as a source of time-varying “intrinsic” mossy fiber input.

In summary, these data suggest how feedforward inhibition present in the synaptic organization of the cerebellar cortex contributes to the generation of precisely timed motor movements by contributing to the generation of a precise temporal code in a population of granule cells. Feedforward inhibition present in other brain regions may similarly function to enhance temporal coding.


Biography

Henrik Jörntell received his PhD degree in Neurophysiology from Lund University. He is currently employed as a senior lecturer at Lund University where he heads the lab ‘Neural Basis for Sensorimotor Control’ at the Department of Experimental Medical Science. His interests are the neurophysiological analysis of neuronal microcircuits involved in movement control, spanning cerebellar, spinal, brainstem and neocortical circuitry as well as models of how these structures interact during movement performance and motor learning.


Motion sickness

A common treatment of motion sickness is Dramamine, which helps to reduce the sensitivity of input from your vestibular system to the rest of your body. [Image: Mike Baird, https://goo.gl/zfeqqr, CC BY 2.0, goo.gl/BRvSA7]

Although a number of conditions can produce motion sickness, it is generally thought that it is evoked from a mismatch in sensory cues between vestibular, visual, and proprioceptive signals (Yates, Miller, & Lucot, 1998). For example, reading a book in a car on a winding road can produce motion sickness, whereby the accelerations experienced by the vestibular system do not match the visual input. However, if one looks out the window at the scenery going by during the same travel, no sickness occurs because the visual and vestibular cues are in alignment. Sea sickness, a form of motion sickness, appears to be a special case and arises from unusual vertical oscillatory and roll motion. Human studies have found that low frequency oscillations of 0.2 Hz and large amplitudes (such as found in large seas during a storm) are most likely to cause motion sickness, with higher frequencies offering little problems.

Summary

Here, we have seen that the vestibular system transduces and encodes signals about head motion and position with respect to gravity, information that is then used by the brain for many essential functions and behaviors. We actually understand a great deal regarding vestibular contributions to fundamental reflexes, such as compensatory eye movements and balance during motion. More recent progress has been made toward understanding how vestibular signals combine with other sensory cues, such as vision, in the thalamus and cortex to give rise to motion perception. However, there are many complex cognitive abilities that we know require vestibular information to function, such as spatial orientation and navigation behaviors, but these systems are only just beginning to be investigated. Future research regarding vestibular system function will likely be geared to seeking answers to questions regarding how the brain copes with vestibular signal loss. In fact, according to the National Institutes of Health, nearly 35% of Americans over the age of 40 (69 million people) have reported chronic vestibular-related problems. It is therefore of significant importance to human health to better understand how vestibular cues contribute to common brain functions and how better treatment options for vestibular dysfunction can be realized.


Introduction

The “Realistic” Modeling Approach

In contrast to the classical top-down modeling strategies guided by researcher’s intuitions about the structure-function relationship of brain circuits, much attention has recently been given to bottom-up strategies. In the construction of bottom-up models, the system is first reconstructed through a reverse engineering process integrating available biological features. Then, the models are carefully validated against a complex dataset not used to construct them, and finally their performance is analyzed as they were the real system. The biological precision of these models can be rather high so that they merit the name of realistic models. The advantage of realistic models is two-fold. First, there is limited selection of biological details that might be relevant to function (this issue will be important in the simplification process considered below). Secondly, with these models it is possible to monitor the impact of microscopic variables on the whole system. A drawback is that some details may be missing, although they can be introduced at a later stage providing proofs on their relevance to circuit functioning (model upgrading). Another potential drawback of realistic models is that they may lose insight into the function being modeled. However, this insight can be recovered at a later stage, since realistic models can incorporate sufficient details to generate microcircuit spatio-temporal dynamics and explain them on the basis of elementary neuronal and connectivity mechanisms (Brette et al., 2007).

Realistic modeling responds to the general intuition that complexity in biological systems should be exploited rather that rejected (Pellionisz and Szentágothai, 1974 Jaeger et al., 1997 De Schutter, 1999 Fernandez et al., 2007 Bower, 2015). For example, the essential computational aspects of a complex adaptive system may reside in its dynamics rather than just in the structure-function relationship (Arbib et al., 1997, 2008), and require therefore closed-loop testing and the extraction of rules from models running in a virtual environment (see below). Moreover, the multilevel organization of the brain often prevents from finding a simple relationship between elementary properties (e.g., neuronal firing) and higher functions (e.g., motor control or cognition). Network connectivity on different scales exploits local neuronal computations and eventually generates the algorithms subtending brain operations. An important new aspect of the realistic modeling approach is that it is now much more affordable than in the past, when it was less used due to the lack of sufficient biophysical data on one hand and of computational power and infrastructures on the other. Now that these all are becoming available, the realistic modeling approach represents a new exciting opportunity for understanding the inner nature of brain functioning. In a sense, realistic modeling is emerging as one of the most powerful tools in the hands of neuroscientists (Davison, 2012 Gerstner et al., 2012 Markram, 2013). The cerebellum has actually been the work bench for the development of ideas and tools fuelling realistic modeling over almost 40 years (for review see Bhalla et al., 1992 Baldi et al., 1998 Cornelis et al., 2012a D𠆚ngelo et al., 2013a Bower, 2015 Sudhakar et al., 2015).

Cerebellar Microcircuit Modeling: Foundations

In the second half of the 20th century David Marr, in a classical triad, developed theoretical models for the neocortex, the hippocampus and the cerebellum, setting landmarks for the development of theoretical and computational neuroscience (for review see, Ito, 2006 Honda et al., 2013). Since then, the models have advanced alternatively in either one or the other of these brain areas.

The striking anatomical organization of the cerebellar circuit has been the basis for initial models. In 1967, the future Nobel Laureate J.C. Eccles envisaged that the cerebellum could operate as a neuronal “timing” machine (Eccles, 1967). This prediction was soon followed by the theoretical models of Marr and Albus, who proposed the Motor Learning Theory (Marr, 1969 Albus, 1971) emphasizing the cerebellum as a “learning machine” (for a critical vision on this issue, see Llinás, 2011). These latter models integrated a statistical description of circuit connectivity with intuitions about the function the circuit has in behavior (Marr, 1969 Albus, 1971). These models have actually been only partially implemented and simulated as such (Tyrrell and Willshaw, 1992 see below) or transformed into mathematically tractable versions like the adaptive filter model (AFM Dean and Porrill, 2010, 2011 Porrill et al., 2013).

While Marr himself framed his own efforts to understand brain function by contrasting 𠇋ottom up” and “top down” approaches (he believed his approach was 𠇋ottom up”), in initial models the level of realism was limited (at that time, little was known on the ionic channels and receptors of the neuronal membrane, by the way). Since then, several models of the cerebellum and cerebellar subcircuits have been developed incorporating realistic details to a different extent (Maex and De Schutter, 1998 Medina et al., 2000 Solinas et al., 2010). In the most recent models, neurons and synapses incorporate Hodgkin-Huxley-style mechanisms and neurotransmission dynamics (Yamada et al., 1989 Tsodyks et al., 1998 D𠆚ngelo et al., 2013a). As far as microcircuit connectivity is concerned, this has been reconstructed by applying combinatorial rules similar to those that have inspired the original Marr’s model. Recently, an effort has allowed the reconstruction and simulation of the neocortical microcolumn (Markram et al., 2015) showing construction rules that may also be used for different brain areas. The approach used for the neocortical microcircuit is based on precise determination of cell densities, on cell morphologies and on a set of rules for synaptic connectivity based on proximity of the neuronal processes (density-morphology-proximity or DMP rule). One question is now whether the construction rules used for the neocortex can also be applied to the cerebellar network. Moreover, since ontogenetic factors play a critical role in network formation, taking a snapshot of the actual state of the mature cerebellar network may not be enough to implement its connectivity and investigate its function. Again, while developmental models have been devised for the cerebral cortex (Zubler et al., 2013 Roberts et al., 2014), their application to the cerebellum remains to be investigated. Therefore, advancement on the neocortical front may now inspire further development in cerebellar modeling.

The most recent realistic computational models of the cerebellum have been built using an extensive amount of information taken from the anatomical and physiological literature and incorporate neuronal and synaptic models capable of responding to arbitrary input patterns and of generating multiple response properties (Maex and De Schutter, 1998 Medina et al., 2000 Santamaria et al., 2002, 2007 Santamaria and Bower, 2005 Solinas et al., 2010 Kennedy et al., 2014). Each neuron model is carefully reconstructed through repeated validation steps at different levels: at present, accurate models of the GrCs, GoCs, UBCs, PCs, DCN neurons and IOs neurons are available (De Schutter and Bower, 1994a,b D𠆚ngelo et al., 2001 D𠆚ngelo et al., 2016 Nieus et al., 2006, 2014 Solinas et al., 2007a, b Vervaeke et al., 2010 Luthman et al., 2011 Steuber et al., 2011 De Gruijl et al., 2012 Subramaniyam et al., 2014 Masoli et al., 2015). Clearly, realistic models have the intrinsic capacity to resolve the still poorly understood issue of brain dynamics, an issue critical to understand how the cerebellum operates (for e.g., see Llinás, 2014).

That understanding cerebellar neuron dynamics can bring beyond a pure structure-function relationships was early recognized but the issue is not resolved yet. There are several correlated aspects that, in cascade from macroscopic to microscopic, need to be considered in detail (see below). Eventually, cerebellar functioning may exploit internal dynamics to regulate spike-timing and to store relevant network configurations through distributed plasticity (Ito, 2006 D𠆚ngelo and De Zeeuw, 2009 Gao et al., 2012). The testing of integrated hypotheses of this kind is exactly what a realistic computational model, once properly reconstructed and validated, should be able to promote.

A further crucial consideration is that the cerebellum has a similar microcircuit structure in all its parts, whose functions differentiate over a broad range of sensori-motor and cognitive control functions depending on the specific anatomical connections (Schmahmann and Sherman, 1998 Schmahmann, 2004 Ito, 2006 Schmahmann and Caplan, 2006 D𠆚ngelo and Casali, 2013 Koziol et al., 2014). It appears therefore that the intuition about the network role in learning and behavior of the original models of Marr-Albus-Ito can be implemented now by integrating realistic models into a closed-loop robotic environment. This allows the interaction of the microcircuit with ongoing actions and movements and the subsequent learning and extraction of rules from the analysis of neuronal and synaptic properties under closed-loop testing (Caligiore et al., 2013, 2016). In this article, we are reviewing an extended set of critical data that could impact on realistic modeling and are proposing a framework for cerebellar model development and testing. Since not all the aspects of cerebellar modeling have evolved at similar rate, more emphasis has been given to those that will help more in exemplifying prototypical cases.

Realistic Modeling Techniques: The Cerebellum as Workbench

Realistic modeling allows reconstruction of neuronal functions through the application of principles derived from membrane biophysics. The membrane and cytoplasmic mechanisms can be integrated in order to explain membrane potential generation and intracellular regulation processes (Koch, 1998 De Schutter, 2000 D𠆚ngelo et al., 2013a). Once validated, neuronal models can be used for reconstructing entire neuronal microcircuits. The basis of realistic neuronal modeling is the membrane equation, in which the first time derivative of potential is related to the conductances generated by ionic channels. These, in turn, are voltage- and time-dependent and are usually represented either through variants of the Hodgkin-Huxley formalism, through Markov chain reaction models, or using stochastic models (Hodgkin and Huxley, 1952 Connor and Stevens, 1971 Hepburn et al., 2012). All these mechanisms can be arranged into a system of ordinary differential equations, which are solved by numerical methods. The model can contain all the ion channel species that are thought to be relevant to explain the function of a given neuron, which can eventually generate all the known firing patterns observed in real cells. In general, this formalism is sufficient to explain the properties of a membrane patch or of a neuron with very simple geometry, so that one such model may collapse all properties into a single equivalent electrical compartment. In most cases, however, the properties of neurons cannot be explained so easily, and multiple compartments (representing soma, dendrites and axon) have to be included thus generating multicompartment models. This strategy requires an extension of the theory based on Rall’s equation for muticompartmental neuronal structures (Rall et al., 1992 Segev and Rall, 1998). Eventually, the ionic channels will be distributed over numerous different compartments communicating one with each other through the cytoplasmic resistance. Up to this point, the models can usually be satisfactorily constrained by biological data on neuronal morphology, ionic channel properties and compartmental distribution. However, the main issue that remains is to appropriately calibrate the maximum ionic conductances of the different ionic channels. To this aim, recent techniques have made use of genetic algorithms that can determine the best data set of multiple conductances through a mutation/selection process (Druckmann et al., 2007, 2008).

As well as membrane excitation, synaptic transmission mechanisms can also be modeled at a comparable level of detail. Differential equations can be used to describe the presynaptic vesicle cycle and the subsequent processes of neurotransmitter diffusion and postsynaptic receptor activation (Tsodyks et al., 1998). This last step consists of neurotransmitter binding to receptors, followed by the opening ion channels or modulation of intracellular cascades, and it is often accounted by stochastic receptor models. The synapses can also be endowed with mechanisms generating various forms of short- and long-term plasticity (Migliore et al., 1995). Appropriate synaptic modeling provides the basis for assembling neuronal circuits.

In all these cases, the cerebellum has provided a work bench that has remarkably contributed to write the history of realistic modeling. Examples are the development of integrated simulation platforms (Bhalla et al., 1992 Bower and Beeman, 2007), the definition of model optimization and evaluation strategies (Baldi et al., 1998 Vanier and Bower, 1999 Cornelis et al., 2012a,b Bower, 2015), the generation of complex neuron models as exemplified by the Purkinje cells (De Schutter and Bower, 1994a,b Bower, 2015 Masoli et al., 2015) and the GrCs (D𠆚ngelo et al., 2001 Nieus et al., 2006 Diwakar et al., 2009) and the generation of complex microcircuit models (Maex and De Schutter, 1998 Medina and Mauk, 2000 Solinas et al., 2010). Now, the cerebellar neurons, synapses and network pose new challenges for realistic modeling depending on recent discoveries on neuron and circuit biology and on the possibility of including large-scale realistic circuit models into closed loop robotic simulations.


Introduction

An intriguing subpopulation of neurons displays the ability to sustain spiking at high rates in excess of ∼200 Hz (Chow et al., 1999 Henderson et al., 2004 Bartos et al., 2007 Chesselet et al., 2007). To date, voltage-gated potassium channels (Kv) of the Kv3 subfamily are invariably present and necessary for maximal firing rates in these fast-spiking neurons (Lau et al., 2000 Matsukawa et al., 2003 McMahon et al., 2004 Song et al., 2005 Kasten et al., 2007 Espinosa et al., 2008). Kv3 channels are distinguished by rapid activation and deactivation, which confer upon them the ability to rapidly repolarize action potentials. By activating quickly, they maintain action potential brevity and, by deactivating quickly, allow for spiking at high rates. Four separate genes (Kcnc1-4) encoding subunits Kv3.1–Kv3.4 assemble into homotetrameric and heterotetrameric channels, exhibiting distinctive and overlapping expression patterns (Rudy et al., 1999 Rudy and McBain, 2001).

In the cerebellar cortex (Fig. 1), the projection neurons, Purkinje cells, express Kv3.3 throughout the cell and Kv3.4 largely in dendrites (Martina et al., 2003 McMahon et al., 2004 Chang et al., 2007). Granule cells express Kv3.1 and Kv3.3 (Weiser et al., 1995 Sekirnjak et al., 1997 Ozaita et al., 2002), whereas basket cells express Kv3.2 and Kv3.4 channel subunits in the pinceaux (Veh et al., 1995 Bobik et al., 2004). In the deep cerebellar nuclei (DCN) and analogous vestibular nuclei, mRNA for all subfamily members is present (Weiser et al., 1994), with Kv3.1 and Kv3.3 protein is expressed in large, presumably glutamatergic projection neurons (Weiser et al., 1995 McMahon et al., 2004). Kv3.2 protein is also present in these nuclei (Lau et al., 2000), and Kv3.4 has not been examined here.

Simplified cerebellar circuitry illustrating direction of information flow and Kv3 channel localization. Interneurons and inhibitory projection neurons from the DCN to the inferior olive have been omitted for simplicity. Cerebellar input arrives from climbing and mossy fibers to stimulate neurons in the deep nuclei and the cortex, in which climbing fibers stimulate Purkinje cells directly and granule cells excited by mossy fibers stimulate Purkinje cells by way of parallel fibers. Purkinje cells in turn inhibit deep nuclear neurons also, at a latency, leading to rebound excitation.

Mice expressing Kcnc1 but lacking Kcnc3 (+/+ −/−) exhibit ataxia and abnormal spiking in Purkinje cells which rely on Kv3.3 for brief action potentials (McMahon et al., 2004). Re-expression of Kv3.3 selectively in Purkinje cells restored spike parameters and rescued motor coordination in +/+ −/− mice as well as in mice additionally lacking one Kcnc1 allele (+/− −/−) (Hurlock et al., 2008). As Purkinje cell firing is potentially related to the control of fine motor timing or speed, determination of what aspect of motor function was affected in the ataxia of +/+ −/− mutants is of interest.

For electrophysiological alterations to be consequential for behavior, the intrinsic firing properties of DCN neurons must be reasonably intact for the rescue of fast repolarization in Purkinje cells to be efficacious in averting ataxia in +/+ −/− mutants (Fig. 1). Large, non-GABAergic DCN projection neurons express Kv3.1 and Kv3.3 channel subunits. Thus, we assessed Kcnc1/Kcnc3 double-knock-out (−/− −/−) mice for a behavioral and electrophysiological rescue. To extricate whether the ataxia stems from hypermetric speed/force, timing of limb movement, or relative timing across limbs, we also measured high-speed motor performance and gait pattern alterations. To define which electrophysiological changes could most plausibly account for the ataxia, we examined intrinsic spiking of DCN neurons in mice lacking Kcnc3 as well as Kcnc1 alleles.


Somatic Sensory Pathways

The somatosensory pathway is composed of three neurons located in the dorsal root ganglion, the spinal cord, and the thalamus.

Learning Objectives

Describe the somatosensory area in the human cortex

Key Takeaways

Key Points

  • A somatosensory pathway will typically have three neurons: primary, secondary, and tertiary.
  • The cell bodies of the three neurons in a typical somatosensory pathway are located in the dorsal root ganglion, the spinal cord, and the thalamus.
  • A major target of somatosensory pathways is the postcentral gyrus in the parietal lobe of the cerebral cortex.
  • A major somatosensory pathway is the dorsal column–medial lemniscal pathway.
  • The postcentral gyrus is the location of the primary somatosensory area that takes the form of a map called the sensory homunculus.

Key Terms

  • parietal lobe: A part of the brain positioned superior to the occipital lobe and posterior to the frontal lobe that integrates sensory information from different modalities and is particularly important for determining spatial sense and navigation.
  • reticular activating system: A set of connected nuclei in the brain responsible for regulating wakefulness and sleep-to-wake transitions.
  • postcentral gyrus: A prominent structure in the parietal lobe of the human brain that is the location of the primary somatosensory cortex, the main sensory receptive area for the sense of touch.
  • thalamus: Either of two large, ovoid structures of gray matter within the forebrain that relay sensory impulses to the cerebral cortex.

A somatosensory pathway will typically have three long neurons: primary, secondary, and tertiary. The first always has its cell body in the dorsal root ganglion of the spinal nerve.

Dorsal root ganglion: Sensory nerves of a dorsal root ganglion are depicted entering the spinal cord.

The second has its cell body either in the spinal cord or in the brainstem this neuron’s ascending axons will cross to the opposite side either in the spinal cord or in the brainstem. The axons of many of these neurons terminate in the thalamus, and others terminate in the reticular activating system or the cerebellum.

In the case of touch and certain types of pain, the third neuron has its cell body in the ventral posterior nucleus of the thalamus and ends in the postcentral gyrus of the parietal lobe.

In the periphery, the somatosensory system detects various stimuli by sensory receptors, such as by mechanoreceptors for tactile sensation and nociceptors for pain sensation. The sensory information (touch, pain, temperature, etc.,) is then conveyed to the central nervous system by afferent neurons, of which there are a number of different types with varying size, structure, and properties.

Generally, there is a correlation between the type of sensory modality detected and the type of afferent neuron involved. For example, slow, thin, unmyelinated neurons conduct pain whereas faster, thicker, myelinated neurons conduct casual touch.

Ascending Pathways

In the spinal cord, the somatosensory system includes ascending pathways from the body to the brain. One major target within the brain is the postcentral gyrus in the cerebral cortex. This is the target for neurons of the dorsal column–medial lemniscal pathway and the ventral spinothalamic pathway.

Note that many ascending somatosensory pathways include synapses in either the thalamus or the reticular formation before they reach the cortex. Other ascending pathways, particularly those involved with control of posture, are projected to the cerebellum, including the ventral and dorsal spinocerebellar tracts.

Another important target for afferent somatosensory neurons that enter the spinal cord are those neurons involved with local segmental reflexes.

Spinal nerve: The formation of the spinal nerve from the dorsal and ventral roots.

Parietal Love: Primary Somatosensory Area

The primary somatosensory area in the human cortex is located in the postcentral gyrus of the parietal lobe. This is the main sensory receptive area for the sense of touch.

Like other sensory areas, there is a map of sensory space called a homunculus at this location. Areas of this part of the human brain map to certain areas of the body, dependent on the amount or importance of somatosensory input from that area.

For example, there is a large area of cortex devoted to sensation in the hands, while the back has a much smaller area. Somatosensory information involved with proprioception and posture also target an entirely different part of the brain, the cerebellum.

Cortical Homunculus

This is a pictorial representation of the anatomical divisions of the primary motor cortex and the primary somatosensory cortex namely, the portion of the human brain directly responsible for the movement and exchange of sensory and motor information of the body.

Homunculus: Image representing the cortical sensory homunculus.

Thalamus

The thalamus is a midline symmetrical structure within the brain of vertebrates including humans it is situated between the cerebral cortex and midbrain, and surrounds the third ventricle.

Its function includes relaying sensory and motor signals to the cerebral cortex, along with the regulation of consciousness, sleep, and alertness.

Thalamic nuclei: The ventral posterolateral nucleus receives sensory information from the body.


Functional Module of the Cerebellum: Microcomplex

The cerebellar cortex is organized into seven longitudinal (A, B, C1, C2, C3, D1, and D2) zones. Each zone sends Purkinje cell axons to a certain cerebellar or vestibular nucleus. Thus, zone A projects to fastigial and vestibular nuclei, zone B to vestibular nuclei, zones C1 and C3 to the rostral part of the interpositus nucleus and zone C2 to the caudal part of it, and zones D1 and D2 to the medial and lateral parts of the lateral (dentate) nucleus.

Each longitudinal zone of the cerebellar cortex is composed of a number of microzones. Each microzone projects to a small group of vestibular or cerebellar nuclear neurons and receives climbing fibers from a small group of inferior olive neurons, which project collaterals to a small group of nuclear neurons projected by the microzone. A microcomplex is an interconnected set of a microzone, a small group of nuclear neurons, and a small group of inferior olive neurons. The microzones defined in the paravermis and flocculus may measure about 10 mm2. In rats, one microzone of this size contains about 10,000 Purkinje cells and 2,740,000 granule cells. The human cerebellum is about 50,000 mm2 wide, so that it may contain as many as 5,000 microzones as its functional unit.

The microcomplex would function as a module of the cerebellum in the following manner. First, input signals from a precerebellar nucleus (except those from the inferior olive) drive the nuclear neurons, which generate output signals of the microcomplex under inhibitory influences of Purkinje cells. Second, the same input signals pass via mossy fibers to a microzone, where they are relayed by granule cells and in turn excite Purkinje cells and other cortical neurons, eventually evoking simple spikes in Purkinje cells. Simple spike discharges of Purkinje cells driven by moss-fiber signals produce a unique functional state of a microzone arising from concerted activities of excitatory and inhibitory synapses. Third, climbing fibers convey error signals pertaining to the operation of the neural system that includes the micrcomplex, as generated by various neuronal mechanisms in diverse preolivary structures. Fourth, climbing-fiber error signals induce LTD in the conjointly activated parallel fiber-Purkinje cell synapses (learning rule) and thereby modify the operation of the micro-complex until the error signals are minimized. Climbing-fiber signals evoke complex spikes in Purkinje cells and thereby induce conducting impulses in Purkinje cell axons, which eventually evoke IPSPs in the nuclear neurons. However, the effects of the IPSPs on nuclear neurons are counteracted by the EPSPs evoked via collaterals of olivocerebellar fibers. The signal content of complex spikes has been analyzed in various motor behaviors and is related partly to consequence errors and partly to internally computed errors.

The postulated operation of the microcomplex, including inhibitory neurons in the cerebellar cortex, has been computer-simulated. However, since several forms of synaptic plasticity other than conjunctive LTD were observed in the cerebellum, further studies are needed to reproduce the performance of a micro-complex that incorporate these forms of synaptic plasticity.


Results

Experiment 1: CBI decreases after visuomotor adaptation

When the 30° CW visuomotor transformation was applied in Adapt1, large initial errors were observed (first epoch mean ± SE was −26.45 ± 2.24° Fig. 3A). When the visuomotor transformation was removed for a Catch epoch after 48 movements, subjects displayed partial learning of the rotation, as indicated by counter-CW errors (11.08 ± 0.97°). When the perturbation resumed at the beginning of Adapt2, error reduction continued and, by the end of Adapt2, subjects had compensated for 25.92 ± 1.03° of the original 30° perturbation.

Experiment 1 Results. A, End point error (blue line) with SEs (shaded region) during baseline (Base), adaptation (Adapt1 and Adapt2), and catch trials (C). Negative values indicate clockwise deviations caused by the visuomotor perturbation. B, Physiological measure of CBI for both the right FDI (blue) and TA (red). CBI was recorded before any movements (Base) and immediately after catch trials (Catch) and late adaptation (Adapt2). *CBI decreased significantly for both muscle effectors after early perturbation exposure.

The magnitude of CBI for both the FDI and TA muscles (hand and leg) was reduced after adaptation with the right hand, indicating disinhibition (Fig. 3B). An ANOVARM on CBI ratio values revealed a time effect (F(2,36) = 9.785 p < 0.01), but no effect on muscle group (F(1,18) = 2.74 p = 0.12) and, importantly, no interaction (F(2,36) = 0.16 p = 0.90). Post hoc tests revealed that CBI after Catch and after Adapt2 was significantly different from baseline CBI (p < 0.01 p = 0.01, respectively), suggesting that visuomotor adaptation with the hand results in a significant reduction in CBI for both FDI and TA muscles.

Experiment 2: Right hand learning transfers to right foot movements

As in Experiment 1, subjects in Experiment 2 were able to almost completely adjust their hand movements to the 30° CW visuomotor rotation (mean error of final epoch in Adapt2 was −2.32 ± 1.11° Fig. 4). When comparing behavioral blocks across right hand and right leg (Baseline, Catch1, Catch2, and Catch3), ANOVARM revealed a significant effect of time (F(2,36) = 24.94 p < 0.01), limb (F(1,36) = 9.44, p = 0.01) and a limb × time interaction effect (F(2,36) = 5.63 p < 0.01). Post hoc analysis revealed that movement error in Catch Foot1 and Catch Foot2 were each different from Base Foot (p = 0.05, 0.01), suggesting that the right hand's adaptation transferred to the foot. Critically, Post hoc analysis also revealed a difference in Catch Hand1 and Catch Hand2 being different from Base Hand (both p < 0.01). A comparison of foot aftereffects (Catch Foot 1–2) with hand aftereffects (Catch Hand 1 and the first epoch of Post 1) indicated that the amount of transfer at Foot Catch1 was 42.3%, and 42.2% at Foot Catch2. Post hoc tests also showed that Catch Foot2 and Catch Foot3 were different from each other (p < 0.01), suggesting that transfer of learning from hand to foot had degraded after washout of learning from the hand.

Experiment 2 Results. A, End point error and SEs for right hand (blue) and right leg (red) movements. Negative values indicate clockwise deviation. B, C, Mean end-point errors in degrees (±SEM) for right leg (red) and right hand (blue) for the baseline and three catch trial epochs. Post hoc analysis revealed significant changes in error for both effectors, indicating hand-to-foot transfer.

Experiments 3 and 4: CBI changes in a somatotopy-specific manner

Participants did not improve their RT throughout the experiment. For Experiment 3, the baseline average RT for the hand (176.1 ± 6.2 ms) was not significantly different from RT from the last 30 hand trials (174.0 ± 5.6 ms t(11) = 1.242, p = 0.24). Baseline foot RT (191.9 ± 8.7 ms) was also not different from the last 30 foot trials (190.3 ± 7.5 ms t(11) = 0.8, p = 0.44). Similarly, in Experiment 4, baseline foot RT (196.0 ± 5.7 ms) was not different from the last 30 foot trials (193.3 ± 5.4 ms t(7) = 0.971, p = 0.36) and baseline hand RT (181.2 ± 7.3 ms) was not different from the last 30 hand trials (179.9 ± 6.7 ms t(7) = 0.759, p = 0.47).

CBI recruitment curve (Experiment 3 only)

ANOVARM comparing the CBI ratio across CS intensity revealed a significant effect for CS intensity (F(1,11) = 7.926, p < 0.01). Critically, TS MEP amplitudes were not significantly different across different CS conditions (F(1,11) = 0.029, p = 0.87), suggesting that the CBI changes were due to changing CS intensities. As a result, the mean CS intensity set for CBImove was 72.9 ± 0.74% of the stimulator output, yielding a rest CBI ratio response of 0.82 ± 0.07.

Premovement CBImove

To assess whether CBI changes are somatotopy specific, we measured CBImove for the FDI (Experiment 3) and TA (Experiment 4) muscle representation when participants were asked to either move the right index finger or right foot. In Experiment 3, when participants made movements with their finger, ANOVARM revealed an effect of FDI CBImove for RT (F(4,44) = 8.192, p = 0.02 Fig. 5A). Conversely, when FDI CBImove was recorded in preparation of foot movements, ANOVARM failed to find the same effect (F(4,44) = 3.214, p = 0.22 Fig. 5B). These results indicate that, in the absence of learning, CBI changes of the FDI muscle occurs during the preparation of finger movements, but not the foot. In Experiment 4, when participants made movements with the right foot, ANOVARM on TA CBImove values revealed a time effect (F(4,28) = 4.920 p < 0.01 Fig 5D). Conversely, when TA CBImove was recorded in preparation of finger movements, ANOVARM did not reveal the same effect (F(4,28) = 0.241, p > 0.50 Fig. 5E). Together, the results from Experiment 3 and 4 suggest a somatotopic effect of CBI changes during movement preparation.

CBImove. A, B, The x-axis represents CBImove for the right FDI in preparation to moving the hand (A) and foot (B). FDI CBImove measured at five timings (T1–T5) with respect to individual mean response times separately for the hand (blue) and foot (red). *CBI was reduced significantly only in preparation of hand movements. C, CBImove was calculated as the percentage difference from FDI CBI obtained at rest. D, E, The x-axis represents CBImove for the right TA in preparation to moving the foot (D) and hand (E). F, Percentage difference from TA CBI obtained at rest. Positive values indicate disinhibition and negative values increased inhibition. *Only hand movements at 90% RT (T5) modulated CBImove. Data are shown as mean ± SEM.

Furthermore, to compare directly the changes in FDI CBImove between preparation of finger movements and foot movements, we subtracted the amount of CBI measured at rest for each session. Here, ANOVARM revealed an effect of FDI CBImove for the group (finger, foot F(1,22) = 6.214, p = 0.02) and RT (Go, T1, …, T5) × group interaction (F(4,88) = 4.713, p = 0.01 Fig. 5C). Specifically, CBImove recorded during movement preparation at 90% RT (T5) was significantly different from CBI assessed at cue representation (p = 0.03), indicating that specific FDI CBI changes occur just before movement onset. We performed the same analysis to compare the changes in TA CBImove. Similar to the results of Experiment 3, ANOVARM showed an effect of TA CBImove for group (foot, finger F(1,14) = 7.911, p = 0.02) and the RT (Go, T1, …, T5) × group interaction (F(4,56) = 3.210, p = 0.02 Fig. 5F).

Importantly, we determined that modulation of CBI were due to changes in conditioned (CS + TS) responses from the cerebellum and not due to changes in TS responses from M1. Although ANOVARM revealed a significant effect of TS MEP amplitude during finger movement preparation (Go, T1, …, T5 F(4,44) = 14.823, p = 0.001), we controlled for this confound by measuring FDI CBI at rest with a matched TS MEP amplitude observed at the 90% RT of CBImove (TS ∼2 mV Fig. 6). When we compared TS and CS+TS MEP amplitudes between rest and premovement measurements using matched TS responses, two-way ANOVARM showed a significant interaction of MEP amplitudes for state (rest, premovement) × condition (TS, CS+TS F(1,22) = 6.295, p = 0.043). Specifically, CS + TS MEP amplitude for premovement was different from rest (p = 0.02) despite having comparable TS MEP amplitude (p = 0.458). This indicates that the changes in FDI CBImove observed at 90% RT are due to changes in cerebellar excitability (CS + TS), not to higher M1 excitability.

Rest and premovement TS and CS + TS MEP amplitudes. For each participant, we assessed CBI at rest matching the TS MEP amplitudes obtained during CBImove at 90% of RT (TS). CS + TS MEP amplitude (CBI) was only present at rest (green), not when assessed in the context of movement (purple). This indicates that the reduction of CBImove is not due to increased excitability in M1.


Introduction

The “Realistic” Modeling Approach

In contrast to the classical top-down modeling strategies guided by researcher’s intuitions about the structure-function relationship of brain circuits, much attention has recently been given to bottom-up strategies. In the construction of bottom-up models, the system is first reconstructed through a reverse engineering process integrating available biological features. Then, the models are carefully validated against a complex dataset not used to construct them, and finally their performance is analyzed as they were the real system. The biological precision of these models can be rather high so that they merit the name of realistic models. The advantage of realistic models is two-fold. First, there is limited selection of biological details that might be relevant to function (this issue will be important in the simplification process considered below). Secondly, with these models it is possible to monitor the impact of microscopic variables on the whole system. A drawback is that some details may be missing, although they can be introduced at a later stage providing proofs on their relevance to circuit functioning (model upgrading). Another potential drawback of realistic models is that they may lose insight into the function being modeled. However, this insight can be recovered at a later stage, since realistic models can incorporate sufficient details to generate microcircuit spatio-temporal dynamics and explain them on the basis of elementary neuronal and connectivity mechanisms (Brette et al., 2007).

Realistic modeling responds to the general intuition that complexity in biological systems should be exploited rather that rejected (Pellionisz and Szentágothai, 1974 Jaeger et al., 1997 De Schutter, 1999 Fernandez et al., 2007 Bower, 2015). For example, the essential computational aspects of a complex adaptive system may reside in its dynamics rather than just in the structure-function relationship (Arbib et al., 1997, 2008), and require therefore closed-loop testing and the extraction of rules from models running in a virtual environment (see below). Moreover, the multilevel organization of the brain often prevents from finding a simple relationship between elementary properties (e.g., neuronal firing) and higher functions (e.g., motor control or cognition). Network connectivity on different scales exploits local neuronal computations and eventually generates the algorithms subtending brain operations. An important new aspect of the realistic modeling approach is that it is now much more affordable than in the past, when it was less used due to the lack of sufficient biophysical data on one hand and of computational power and infrastructures on the other. Now that these all are becoming available, the realistic modeling approach represents a new exciting opportunity for understanding the inner nature of brain functioning. In a sense, realistic modeling is emerging as one of the most powerful tools in the hands of neuroscientists (Davison, 2012 Gerstner et al., 2012 Markram, 2013). The cerebellum has actually been the work bench for the development of ideas and tools fuelling realistic modeling over almost 40 years (for review see Bhalla et al., 1992 Baldi et al., 1998 Cornelis et al., 2012a D𠆚ngelo et al., 2013a Bower, 2015 Sudhakar et al., 2015).

Cerebellar Microcircuit Modeling: Foundations

In the second half of the 20th century David Marr, in a classical triad, developed theoretical models for the neocortex, the hippocampus and the cerebellum, setting landmarks for the development of theoretical and computational neuroscience (for review see, Ito, 2006 Honda et al., 2013). Since then, the models have advanced alternatively in either one or the other of these brain areas.

The striking anatomical organization of the cerebellar circuit has been the basis for initial models. In 1967, the future Nobel Laureate J.C. Eccles envisaged that the cerebellum could operate as a neuronal “timing” machine (Eccles, 1967). This prediction was soon followed by the theoretical models of Marr and Albus, who proposed the Motor Learning Theory (Marr, 1969 Albus, 1971) emphasizing the cerebellum as a “learning machine” (for a critical vision on this issue, see Llinás, 2011). These latter models integrated a statistical description of circuit connectivity with intuitions about the function the circuit has in behavior (Marr, 1969 Albus, 1971). These models have actually been only partially implemented and simulated as such (Tyrrell and Willshaw, 1992 see below) or transformed into mathematically tractable versions like the adaptive filter model (AFM Dean and Porrill, 2010, 2011 Porrill et al., 2013).

While Marr himself framed his own efforts to understand brain function by contrasting 𠇋ottom up” and “top down” approaches (he believed his approach was 𠇋ottom up”), in initial models the level of realism was limited (at that time, little was known on the ionic channels and receptors of the neuronal membrane, by the way). Since then, several models of the cerebellum and cerebellar subcircuits have been developed incorporating realistic details to a different extent (Maex and De Schutter, 1998 Medina et al., 2000 Solinas et al., 2010). In the most recent models, neurons and synapses incorporate Hodgkin-Huxley-style mechanisms and neurotransmission dynamics (Yamada et al., 1989 Tsodyks et al., 1998 D𠆚ngelo et al., 2013a). As far as microcircuit connectivity is concerned, this has been reconstructed by applying combinatorial rules similar to those that have inspired the original Marr’s model. Recently, an effort has allowed the reconstruction and simulation of the neocortical microcolumn (Markram et al., 2015) showing construction rules that may also be used for different brain areas. The approach used for the neocortical microcircuit is based on precise determination of cell densities, on cell morphologies and on a set of rules for synaptic connectivity based on proximity of the neuronal processes (density-morphology-proximity or DMP rule). One question is now whether the construction rules used for the neocortex can also be applied to the cerebellar network. Moreover, since ontogenetic factors play a critical role in network formation, taking a snapshot of the actual state of the mature cerebellar network may not be enough to implement its connectivity and investigate its function. Again, while developmental models have been devised for the cerebral cortex (Zubler et al., 2013 Roberts et al., 2014), their application to the cerebellum remains to be investigated. Therefore, advancement on the neocortical front may now inspire further development in cerebellar modeling.

The most recent realistic computational models of the cerebellum have been built using an extensive amount of information taken from the anatomical and physiological literature and incorporate neuronal and synaptic models capable of responding to arbitrary input patterns and of generating multiple response properties (Maex and De Schutter, 1998 Medina et al., 2000 Santamaria et al., 2002, 2007 Santamaria and Bower, 2005 Solinas et al., 2010 Kennedy et al., 2014). Each neuron model is carefully reconstructed through repeated validation steps at different levels: at present, accurate models of the GrCs, GoCs, UBCs, PCs, DCN neurons and IOs neurons are available (De Schutter and Bower, 1994a,b D𠆚ngelo et al., 2001 D𠆚ngelo et al., 2016 Nieus et al., 2006, 2014 Solinas et al., 2007a, b Vervaeke et al., 2010 Luthman et al., 2011 Steuber et al., 2011 De Gruijl et al., 2012 Subramaniyam et al., 2014 Masoli et al., 2015). Clearly, realistic models have the intrinsic capacity to resolve the still poorly understood issue of brain dynamics, an issue critical to understand how the cerebellum operates (for e.g., see Llinás, 2014).

That understanding cerebellar neuron dynamics can bring beyond a pure structure-function relationships was early recognized but the issue is not resolved yet. There are several correlated aspects that, in cascade from macroscopic to microscopic, need to be considered in detail (see below). Eventually, cerebellar functioning may exploit internal dynamics to regulate spike-timing and to store relevant network configurations through distributed plasticity (Ito, 2006 D𠆚ngelo and De Zeeuw, 2009 Gao et al., 2012). The testing of integrated hypotheses of this kind is exactly what a realistic computational model, once properly reconstructed and validated, should be able to promote.

A further crucial consideration is that the cerebellum has a similar microcircuit structure in all its parts, whose functions differentiate over a broad range of sensori-motor and cognitive control functions depending on the specific anatomical connections (Schmahmann and Sherman, 1998 Schmahmann, 2004 Ito, 2006 Schmahmann and Caplan, 2006 D𠆚ngelo and Casali, 2013 Koziol et al., 2014). It appears therefore that the intuition about the network role in learning and behavior of the original models of Marr-Albus-Ito can be implemented now by integrating realistic models into a closed-loop robotic environment. This allows the interaction of the microcircuit with ongoing actions and movements and the subsequent learning and extraction of rules from the analysis of neuronal and synaptic properties under closed-loop testing (Caligiore et al., 2013, 2016). In this article, we are reviewing an extended set of critical data that could impact on realistic modeling and are proposing a framework for cerebellar model development and testing. Since not all the aspects of cerebellar modeling have evolved at similar rate, more emphasis has been given to those that will help more in exemplifying prototypical cases.

Realistic Modeling Techniques: The Cerebellum as Workbench

Realistic modeling allows reconstruction of neuronal functions through the application of principles derived from membrane biophysics. The membrane and cytoplasmic mechanisms can be integrated in order to explain membrane potential generation and intracellular regulation processes (Koch, 1998 De Schutter, 2000 D𠆚ngelo et al., 2013a). Once validated, neuronal models can be used for reconstructing entire neuronal microcircuits. The basis of realistic neuronal modeling is the membrane equation, in which the first time derivative of potential is related to the conductances generated by ionic channels. These, in turn, are voltage- and time-dependent and are usually represented either through variants of the Hodgkin-Huxley formalism, through Markov chain reaction models, or using stochastic models (Hodgkin and Huxley, 1952 Connor and Stevens, 1971 Hepburn et al., 2012). All these mechanisms can be arranged into a system of ordinary differential equations, which are solved by numerical methods. The model can contain all the ion channel species that are thought to be relevant to explain the function of a given neuron, which can eventually generate all the known firing patterns observed in real cells. In general, this formalism is sufficient to explain the properties of a membrane patch or of a neuron with very simple geometry, so that one such model may collapse all properties into a single equivalent electrical compartment. In most cases, however, the properties of neurons cannot be explained so easily, and multiple compartments (representing soma, dendrites and axon) have to be included thus generating multicompartment models. This strategy requires an extension of the theory based on Rall’s equation for muticompartmental neuronal structures (Rall et al., 1992 Segev and Rall, 1998). Eventually, the ionic channels will be distributed over numerous different compartments communicating one with each other through the cytoplasmic resistance. Up to this point, the models can usually be satisfactorily constrained by biological data on neuronal morphology, ionic channel properties and compartmental distribution. However, the main issue that remains is to appropriately calibrate the maximum ionic conductances of the different ionic channels. To this aim, recent techniques have made use of genetic algorithms that can determine the best data set of multiple conductances through a mutation/selection process (Druckmann et al., 2007, 2008).

As well as membrane excitation, synaptic transmission mechanisms can also be modeled at a comparable level of detail. Differential equations can be used to describe the presynaptic vesicle cycle and the subsequent processes of neurotransmitter diffusion and postsynaptic receptor activation (Tsodyks et al., 1998). This last step consists of neurotransmitter binding to receptors, followed by the opening ion channels or modulation of intracellular cascades, and it is often accounted by stochastic receptor models. The synapses can also be endowed with mechanisms generating various forms of short- and long-term plasticity (Migliore et al., 1995). Appropriate synaptic modeling provides the basis for assembling neuronal circuits.

In all these cases, the cerebellum has provided a work bench that has remarkably contributed to write the history of realistic modeling. Examples are the development of integrated simulation platforms (Bhalla et al., 1992 Bower and Beeman, 2007), the definition of model optimization and evaluation strategies (Baldi et al., 1998 Vanier and Bower, 1999 Cornelis et al., 2012a,b Bower, 2015), the generation of complex neuron models as exemplified by the Purkinje cells (De Schutter and Bower, 1994a,b Bower, 2015 Masoli et al., 2015) and the GrCs (D𠆚ngelo et al., 2001 Nieus et al., 2006 Diwakar et al., 2009) and the generation of complex microcircuit models (Maex and De Schutter, 1998 Medina and Mauk, 2000 Solinas et al., 2010). Now, the cerebellar neurons, synapses and network pose new challenges for realistic modeling depending on recent discoveries on neuron and circuit biology and on the possibility of including large-scale realistic circuit models into closed loop robotic simulations.


Results

Experiment 1: CBI decreases after visuomotor adaptation

When the 30° CW visuomotor transformation was applied in Adapt1, large initial errors were observed (first epoch mean ± SE was −26.45 ± 2.24° Fig. 3A). When the visuomotor transformation was removed for a Catch epoch after 48 movements, subjects displayed partial learning of the rotation, as indicated by counter-CW errors (11.08 ± 0.97°). When the perturbation resumed at the beginning of Adapt2, error reduction continued and, by the end of Adapt2, subjects had compensated for 25.92 ± 1.03° of the original 30° perturbation.

Experiment 1 Results. A, End point error (blue line) with SEs (shaded region) during baseline (Base), adaptation (Adapt1 and Adapt2), and catch trials (C). Negative values indicate clockwise deviations caused by the visuomotor perturbation. B, Physiological measure of CBI for both the right FDI (blue) and TA (red). CBI was recorded before any movements (Base) and immediately after catch trials (Catch) and late adaptation (Adapt2). *CBI decreased significantly for both muscle effectors after early perturbation exposure.

The magnitude of CBI for both the FDI and TA muscles (hand and leg) was reduced after adaptation with the right hand, indicating disinhibition (Fig. 3B). An ANOVARM on CBI ratio values revealed a time effect (F(2,36) = 9.785 p < 0.01), but no effect on muscle group (F(1,18) = 2.74 p = 0.12) and, importantly, no interaction (F(2,36) = 0.16 p = 0.90). Post hoc tests revealed that CBI after Catch and after Adapt2 was significantly different from baseline CBI (p < 0.01 p = 0.01, respectively), suggesting that visuomotor adaptation with the hand results in a significant reduction in CBI for both FDI and TA muscles.

Experiment 2: Right hand learning transfers to right foot movements

As in Experiment 1, subjects in Experiment 2 were able to almost completely adjust their hand movements to the 30° CW visuomotor rotation (mean error of final epoch in Adapt2 was −2.32 ± 1.11° Fig. 4). When comparing behavioral blocks across right hand and right leg (Baseline, Catch1, Catch2, and Catch3), ANOVARM revealed a significant effect of time (F(2,36) = 24.94 p < 0.01), limb (F(1,36) = 9.44, p = 0.01) and a limb × time interaction effect (F(2,36) = 5.63 p < 0.01). Post hoc analysis revealed that movement error in Catch Foot1 and Catch Foot2 were each different from Base Foot (p = 0.05, 0.01), suggesting that the right hand's adaptation transferred to the foot. Critically, Post hoc analysis also revealed a difference in Catch Hand1 and Catch Hand2 being different from Base Hand (both p < 0.01). A comparison of foot aftereffects (Catch Foot 1–2) with hand aftereffects (Catch Hand 1 and the first epoch of Post 1) indicated that the amount of transfer at Foot Catch1 was 42.3%, and 42.2% at Foot Catch2. Post hoc tests also showed that Catch Foot2 and Catch Foot3 were different from each other (p < 0.01), suggesting that transfer of learning from hand to foot had degraded after washout of learning from the hand.

Experiment 2 Results. A, End point error and SEs for right hand (blue) and right leg (red) movements. Negative values indicate clockwise deviation. B, C, Mean end-point errors in degrees (±SEM) for right leg (red) and right hand (blue) for the baseline and three catch trial epochs. Post hoc analysis revealed significant changes in error for both effectors, indicating hand-to-foot transfer.

Experiments 3 and 4: CBI changes in a somatotopy-specific manner

Participants did not improve their RT throughout the experiment. For Experiment 3, the baseline average RT for the hand (176.1 ± 6.2 ms) was not significantly different from RT from the last 30 hand trials (174.0 ± 5.6 ms t(11) = 1.242, p = 0.24). Baseline foot RT (191.9 ± 8.7 ms) was also not different from the last 30 foot trials (190.3 ± 7.5 ms t(11) = 0.8, p = 0.44). Similarly, in Experiment 4, baseline foot RT (196.0 ± 5.7 ms) was not different from the last 30 foot trials (193.3 ± 5.4 ms t(7) = 0.971, p = 0.36) and baseline hand RT (181.2 ± 7.3 ms) was not different from the last 30 hand trials (179.9 ± 6.7 ms t(7) = 0.759, p = 0.47).

CBI recruitment curve (Experiment 3 only)

ANOVARM comparing the CBI ratio across CS intensity revealed a significant effect for CS intensity (F(1,11) = 7.926, p < 0.01). Critically, TS MEP amplitudes were not significantly different across different CS conditions (F(1,11) = 0.029, p = 0.87), suggesting that the CBI changes were due to changing CS intensities. As a result, the mean CS intensity set for CBImove was 72.9 ± 0.74% of the stimulator output, yielding a rest CBI ratio response of 0.82 ± 0.07.

Premovement CBImove

To assess whether CBI changes are somatotopy specific, we measured CBImove for the FDI (Experiment 3) and TA (Experiment 4) muscle representation when participants were asked to either move the right index finger or right foot. In Experiment 3, when participants made movements with their finger, ANOVARM revealed an effect of FDI CBImove for RT (F(4,44) = 8.192, p = 0.02 Fig. 5A). Conversely, when FDI CBImove was recorded in preparation of foot movements, ANOVARM failed to find the same effect (F(4,44) = 3.214, p = 0.22 Fig. 5B). These results indicate that, in the absence of learning, CBI changes of the FDI muscle occurs during the preparation of finger movements, but not the foot. In Experiment 4, when participants made movements with the right foot, ANOVARM on TA CBImove values revealed a time effect (F(4,28) = 4.920 p < 0.01 Fig 5D). Conversely, when TA CBImove was recorded in preparation of finger movements, ANOVARM did not reveal the same effect (F(4,28) = 0.241, p > 0.50 Fig. 5E). Together, the results from Experiment 3 and 4 suggest a somatotopic effect of CBI changes during movement preparation.

CBImove. A, B, The x-axis represents CBImove for the right FDI in preparation to moving the hand (A) and foot (B). FDI CBImove measured at five timings (T1–T5) with respect to individual mean response times separately for the hand (blue) and foot (red). *CBI was reduced significantly only in preparation of hand movements. C, CBImove was calculated as the percentage difference from FDI CBI obtained at rest. D, E, The x-axis represents CBImove for the right TA in preparation to moving the foot (D) and hand (E). F, Percentage difference from TA CBI obtained at rest. Positive values indicate disinhibition and negative values increased inhibition. *Only hand movements at 90% RT (T5) modulated CBImove. Data are shown as mean ± SEM.

Furthermore, to compare directly the changes in FDI CBImove between preparation of finger movements and foot movements, we subtracted the amount of CBI measured at rest for each session. Here, ANOVARM revealed an effect of FDI CBImove for the group (finger, foot F(1,22) = 6.214, p = 0.02) and RT (Go, T1, …, T5) × group interaction (F(4,88) = 4.713, p = 0.01 Fig. 5C). Specifically, CBImove recorded during movement preparation at 90% RT (T5) was significantly different from CBI assessed at cue representation (p = 0.03), indicating that specific FDI CBI changes occur just before movement onset. We performed the same analysis to compare the changes in TA CBImove. Similar to the results of Experiment 3, ANOVARM showed an effect of TA CBImove for group (foot, finger F(1,14) = 7.911, p = 0.02) and the RT (Go, T1, …, T5) × group interaction (F(4,56) = 3.210, p = 0.02 Fig. 5F).

Importantly, we determined that modulation of CBI were due to changes in conditioned (CS + TS) responses from the cerebellum and not due to changes in TS responses from M1. Although ANOVARM revealed a significant effect of TS MEP amplitude during finger movement preparation (Go, T1, …, T5 F(4,44) = 14.823, p = 0.001), we controlled for this confound by measuring FDI CBI at rest with a matched TS MEP amplitude observed at the 90% RT of CBImove (TS ∼2 mV Fig. 6). When we compared TS and CS+TS MEP amplitudes between rest and premovement measurements using matched TS responses, two-way ANOVARM showed a significant interaction of MEP amplitudes for state (rest, premovement) × condition (TS, CS+TS F(1,22) = 6.295, p = 0.043). Specifically, CS + TS MEP amplitude for premovement was different from rest (p = 0.02) despite having comparable TS MEP amplitude (p = 0.458). This indicates that the changes in FDI CBImove observed at 90% RT are due to changes in cerebellar excitability (CS + TS), not to higher M1 excitability.

Rest and premovement TS and CS + TS MEP amplitudes. For each participant, we assessed CBI at rest matching the TS MEP amplitudes obtained during CBImove at 90% of RT (TS). CS + TS MEP amplitude (CBI) was only present at rest (green), not when assessed in the context of movement (purple). This indicates that the reduction of CBImove is not due to increased excitability in M1.


Discussion

The ability to learn properly timed outputs is central to cerebellar function. We have used a large-scale computer simulation of the cerebellum and eyelid conditioning experiments to demonstrate and characterize temporal subtraction, a network mechanism of temporal coding in cerebellar cortex. We showed that the cessation of one of two ongoing mossy fiber inputs to a cerebellar simulation produces a robust and specific temporal code in the population of granule cells. Because this granule cell code is unique to the interval after the cessation of the shorter mossy fiber input, i.e., the offset interval, it predicts precisely timed learning for mossy fiber inputs that are otherwise too long to normally support learning. We showed that cerebellar-dependent learning in rabbits supports these unusual predictions of the simulation. We then characterized how each input contributes to learning under these atypical conditions. Finally, we showed how feedforward inhibition present in the synaptic organization of the cerebellum (Eccles et al., 1967 Palkovits et al., 1971, 1972 Ito, 1984) may contribute to this enhanced temporal code and learning of precisely timed responses.

Temporal subtraction via feedforward inhibition adds to the list of connectivity-based mechanisms that can contribute to the generation of temporal codes, which includes recurrent inhibition, feedback inhibition, and feedforward excitation (Durstewitz et al., 2000 Ikegaya et al., 2004 Mauk and Buonomano, 2004 Buhusi and Meck, 2005 D'Angelo et al., 2009). Recurrent inhibition has been invoked as a means to generate oscillations in the activity of individual cells that could be used in timing (Buhusi and Meck, 2005 D'Angelo et al., 2009). Similarly, feedforward excitation has been hypothesized as a means to generate delay lines or synfire chains (Durstewitz et al., 2000 Ikegaya et al., 2004). In this case, a series of connected neurons generates temporal codes by activating their downstream targets in a highly regulated chain. Previous simulations of the cerebellum have also highlighted the ability of sparse (non-recurrent) feedback inhibition to generate time-varying activity (Medina et al., 2000). Although feedback inhibition and feedforward excitation certainly operate in this simulation of the cerebellum, the present results also illustrate how the feedforward inhibition that is present in many brain regions (Lawrence and McBain, 2003 Swadlow, 2003 Mauk and Buonomano, 2004 Tepper et al., 2004) can contribute to and enhance temporal coding for patterns that involve one input ending before the other.

Temporal subtraction requires only that one of two ongoing mossy fiber inputs terminate, and thus it may represent an often engaged computational principle in the cerebellum. Subtraction may occur during the presentation of peripheral stimuli, such as tones, in which certain mossy fibers respond phasically to the onset of the tone and others respond for the duration of the tone (Boyd and Aitkin, 1976 Aitkin and Boyd, 1978). The stimuli used in eye movement studies produce variety in the temporal pattern of mossy fiber input (Chubb et al., 1984 Lisberger and Pavelko, 1986 Ramachandran and Lisberger, 2006), which may also give opportunity for subtraction. Similarly, the subtraction pattern of inputs is reminiscent of the pattern of inputs that the cerebellum appears to be presented with during trace eyelid conditioning, in which the offset of the CS and onset of the US is separated by a stimulus-free trace interval. Specifically, our results are consistent with evidence that the cerebellum uses mossy fiber input driven by the forebrain that persists to the US and tone-driven input that is active only for the duration of the tone (e.g., a subtraction pattern of input) to generate granule cell activity during the trace interval (Kalmbach et al., 2009, 2010b). Furthermore, they suggest that this coding during trace eyelid conditioning depends on feedforward inhibition that is driven by the tone-driven mossy fiber input.

These data underscore the importance of a precise temporal code for cerebellar learning. LTD of the granule cell-to-Purkinje cell synapse is a form of synaptic plasticity implicated in cerebellar learning (Ito and Kano, 1982 Linden, 1994 Ito, 2001). The induction of LTD requires a specific temporal relationship between the activation of granule cell and climbing fiber synapses onto Purkinje cells granule cell synapses must be active within ∼300 ms of a climbing fiber input to undergo LTD (Chen and Thompson, 1995 Wang et al., 2000 Safo and Regehr, 2008). Conversely, granule cell synapses active in the absence of climbing fiber input undergo LTP (Lev-Ram et al., 2003). Because the activation of weak synapses would tend to increase cerebellar output, synapses that have undergone LTD would be important for the initiation of responses. Although we observed granule cell activity within the window for LTD induction during training with both inputs presented for 3.5 s, this activity was not unique to the LTD induction window. These granule cells were active in the time window for the induction of both LTD (because they were active within ∼300 ms of a climbing fiber input) and LTP (because they were active for seconds in the absence of climbing fiber input). As such, the strength of these synapses either did not change or even increased. This interpretation may explain why learning declines as the length of the CS used in conditioning increases (Schneiderman and Gormezano, 1964 Ohyama et al., 2003): longer CSs are incapable of generating granule cell activity that is specific to the time interval for LTD induction. Conversely, during presentation of mossy fiber input in the subtraction pattern to our cerebellar simulation, there were synapses active precisely in the window for the induction of LTD. This observation suggests that the cessation of one of two ongoing mossy fiber inputs enables learning by creating activity in a subset of granule cells that is unique to the period for LTD induction.

Our results also suggest how the unipolar brush cells (UBCs) found within cerebellar cortex may contribute to temporal coding. UBCs receive a single mossy fiber input and contact granule cells, Golgi cells, and other UBCs (Nunzi et al., 2001). In vitro evidence suggests that UBCs are capable of converting a phasic mossy fiber input into a tonic train of action potentials lasting hundreds of milliseconds (Diño et al., 2000 Nunzi et al., 2001). Thus, although these cells were not included in the cerebellar simulation, they may be well suited to transform phasic mossy fiber input into prolonged “mossy fiber inputs” with variable firing durations, which could in turn engage the subtraction mechanism and enhance temporal coding and learning. In this way, UBCs may act as a source of time-varying “intrinsic” mossy fiber input.

In summary, these data suggest how feedforward inhibition present in the synaptic organization of the cerebellar cortex contributes to the generation of precisely timed motor movements by contributing to the generation of a precise temporal code in a population of granule cells. Feedforward inhibition present in other brain regions may similarly function to enhance temporal coding.


Functional Module of the Cerebellum: Microcomplex

The cerebellar cortex is organized into seven longitudinal (A, B, C1, C2, C3, D1, and D2) zones. Each zone sends Purkinje cell axons to a certain cerebellar or vestibular nucleus. Thus, zone A projects to fastigial and vestibular nuclei, zone B to vestibular nuclei, zones C1 and C3 to the rostral part of the interpositus nucleus and zone C2 to the caudal part of it, and zones D1 and D2 to the medial and lateral parts of the lateral (dentate) nucleus.

Each longitudinal zone of the cerebellar cortex is composed of a number of microzones. Each microzone projects to a small group of vestibular or cerebellar nuclear neurons and receives climbing fibers from a small group of inferior olive neurons, which project collaterals to a small group of nuclear neurons projected by the microzone. A microcomplex is an interconnected set of a microzone, a small group of nuclear neurons, and a small group of inferior olive neurons. The microzones defined in the paravermis and flocculus may measure about 10 mm2. In rats, one microzone of this size contains about 10,000 Purkinje cells and 2,740,000 granule cells. The human cerebellum is about 50,000 mm2 wide, so that it may contain as many as 5,000 microzones as its functional unit.

The microcomplex would function as a module of the cerebellum in the following manner. First, input signals from a precerebellar nucleus (except those from the inferior olive) drive the nuclear neurons, which generate output signals of the microcomplex under inhibitory influences of Purkinje cells. Second, the same input signals pass via mossy fibers to a microzone, where they are relayed by granule cells and in turn excite Purkinje cells and other cortical neurons, eventually evoking simple spikes in Purkinje cells. Simple spike discharges of Purkinje cells driven by moss-fiber signals produce a unique functional state of a microzone arising from concerted activities of excitatory and inhibitory synapses. Third, climbing fibers convey error signals pertaining to the operation of the neural system that includes the micrcomplex, as generated by various neuronal mechanisms in diverse preolivary structures. Fourth, climbing-fiber error signals induce LTD in the conjointly activated parallel fiber-Purkinje cell synapses (learning rule) and thereby modify the operation of the micro-complex until the error signals are minimized. Climbing-fiber signals evoke complex spikes in Purkinje cells and thereby induce conducting impulses in Purkinje cell axons, which eventually evoke IPSPs in the nuclear neurons. However, the effects of the IPSPs on nuclear neurons are counteracted by the EPSPs evoked via collaterals of olivocerebellar fibers. The signal content of complex spikes has been analyzed in various motor behaviors and is related partly to consequence errors and partly to internally computed errors.

The postulated operation of the microcomplex, including inhibitory neurons in the cerebellar cortex, has been computer-simulated. However, since several forms of synaptic plasticity other than conjunctive LTD were observed in the cerebellum, further studies are needed to reproduce the performance of a micro-complex that incorporate these forms of synaptic plasticity.


Somatic Sensory Pathways

The somatosensory pathway is composed of three neurons located in the dorsal root ganglion, the spinal cord, and the thalamus.

Learning Objectives

Describe the somatosensory area in the human cortex

Key Takeaways

Key Points

  • A somatosensory pathway will typically have three neurons: primary, secondary, and tertiary.
  • The cell bodies of the three neurons in a typical somatosensory pathway are located in the dorsal root ganglion, the spinal cord, and the thalamus.
  • A major target of somatosensory pathways is the postcentral gyrus in the parietal lobe of the cerebral cortex.
  • A major somatosensory pathway is the dorsal column–medial lemniscal pathway.
  • The postcentral gyrus is the location of the primary somatosensory area that takes the form of a map called the sensory homunculus.

Key Terms

  • parietal lobe: A part of the brain positioned superior to the occipital lobe and posterior to the frontal lobe that integrates sensory information from different modalities and is particularly important for determining spatial sense and navigation.
  • reticular activating system: A set of connected nuclei in the brain responsible for regulating wakefulness and sleep-to-wake transitions.
  • postcentral gyrus: A prominent structure in the parietal lobe of the human brain that is the location of the primary somatosensory cortex, the main sensory receptive area for the sense of touch.
  • thalamus: Either of two large, ovoid structures of gray matter within the forebrain that relay sensory impulses to the cerebral cortex.

A somatosensory pathway will typically have three long neurons: primary, secondary, and tertiary. The first always has its cell body in the dorsal root ganglion of the spinal nerve.

Dorsal root ganglion: Sensory nerves of a dorsal root ganglion are depicted entering the spinal cord.

The second has its cell body either in the spinal cord or in the brainstem this neuron’s ascending axons will cross to the opposite side either in the spinal cord or in the brainstem. The axons of many of these neurons terminate in the thalamus, and others terminate in the reticular activating system or the cerebellum.

In the case of touch and certain types of pain, the third neuron has its cell body in the ventral posterior nucleus of the thalamus and ends in the postcentral gyrus of the parietal lobe.

In the periphery, the somatosensory system detects various stimuli by sensory receptors, such as by mechanoreceptors for tactile sensation and nociceptors for pain sensation. The sensory information (touch, pain, temperature, etc.,) is then conveyed to the central nervous system by afferent neurons, of which there are a number of different types with varying size, structure, and properties.

Generally, there is a correlation between the type of sensory modality detected and the type of afferent neuron involved. For example, slow, thin, unmyelinated neurons conduct pain whereas faster, thicker, myelinated neurons conduct casual touch.

Ascending Pathways

In the spinal cord, the somatosensory system includes ascending pathways from the body to the brain. One major target within the brain is the postcentral gyrus in the cerebral cortex. This is the target for neurons of the dorsal column–medial lemniscal pathway and the ventral spinothalamic pathway.

Note that many ascending somatosensory pathways include synapses in either the thalamus or the reticular formation before they reach the cortex. Other ascending pathways, particularly those involved with control of posture, are projected to the cerebellum, including the ventral and dorsal spinocerebellar tracts.

Another important target for afferent somatosensory neurons that enter the spinal cord are those neurons involved with local segmental reflexes.

Spinal nerve: The formation of the spinal nerve from the dorsal and ventral roots.

Parietal Love: Primary Somatosensory Area

The primary somatosensory area in the human cortex is located in the postcentral gyrus of the parietal lobe. This is the main sensory receptive area for the sense of touch.

Like other sensory areas, there is a map of sensory space called a homunculus at this location. Areas of this part of the human brain map to certain areas of the body, dependent on the amount or importance of somatosensory input from that area.

For example, there is a large area of cortex devoted to sensation in the hands, while the back has a much smaller area. Somatosensory information involved with proprioception and posture also target an entirely different part of the brain, the cerebellum.

Cortical Homunculus

This is a pictorial representation of the anatomical divisions of the primary motor cortex and the primary somatosensory cortex namely, the portion of the human brain directly responsible for the movement and exchange of sensory and motor information of the body.

Homunculus: Image representing the cortical sensory homunculus.

Thalamus

The thalamus is a midline symmetrical structure within the brain of vertebrates including humans it is situated between the cerebral cortex and midbrain, and surrounds the third ventricle.

Its function includes relaying sensory and motor signals to the cerebral cortex, along with the regulation of consciousness, sleep, and alertness.

Thalamic nuclei: The ventral posterolateral nucleus receives sensory information from the body.


Motion sickness

A common treatment of motion sickness is Dramamine, which helps to reduce the sensitivity of input from your vestibular system to the rest of your body. [Image: Mike Baird, https://goo.gl/zfeqqr, CC BY 2.0, goo.gl/BRvSA7]

Although a number of conditions can produce motion sickness, it is generally thought that it is evoked from a mismatch in sensory cues between vestibular, visual, and proprioceptive signals (Yates, Miller, & Lucot, 1998). For example, reading a book in a car on a winding road can produce motion sickness, whereby the accelerations experienced by the vestibular system do not match the visual input. However, if one looks out the window at the scenery going by during the same travel, no sickness occurs because the visual and vestibular cues are in alignment. Sea sickness, a form of motion sickness, appears to be a special case and arises from unusual vertical oscillatory and roll motion. Human studies have found that low frequency oscillations of 0.2 Hz and large amplitudes (such as found in large seas during a storm) are most likely to cause motion sickness, with higher frequencies offering little problems.

Summary

Here, we have seen that the vestibular system transduces and encodes signals about head motion and position with respect to gravity, information that is then used by the brain for many essential functions and behaviors. We actually understand a great deal regarding vestibular contributions to fundamental reflexes, such as compensatory eye movements and balance during motion. More recent progress has been made toward understanding how vestibular signals combine with other sensory cues, such as vision, in the thalamus and cortex to give rise to motion perception. However, there are many complex cognitive abilities that we know require vestibular information to function, such as spatial orientation and navigation behaviors, but these systems are only just beginning to be investigated. Future research regarding vestibular system function will likely be geared to seeking answers to questions regarding how the brain copes with vestibular signal loss. In fact, according to the National Institutes of Health, nearly 35% of Americans over the age of 40 (69 million people) have reported chronic vestibular-related problems. It is therefore of significant importance to human health to better understand how vestibular cues contribute to common brain functions and how better treatment options for vestibular dysfunction can be realized.


Biography

Henrik Jörntell received his PhD degree in Neurophysiology from Lund University. He is currently employed as a senior lecturer at Lund University where he heads the lab ‘Neural Basis for Sensorimotor Control’ at the Department of Experimental Medical Science. His interests are the neurophysiological analysis of neuronal microcircuits involved in movement control, spanning cerebellar, spinal, brainstem and neocortical circuitry as well as models of how these structures interact during movement performance and motor learning.


Introduction

An intriguing subpopulation of neurons displays the ability to sustain spiking at high rates in excess of ∼200 Hz (Chow et al., 1999 Henderson et al., 2004 Bartos et al., 2007 Chesselet et al., 2007). To date, voltage-gated potassium channels (Kv) of the Kv3 subfamily are invariably present and necessary for maximal firing rates in these fast-spiking neurons (Lau et al., 2000 Matsukawa et al., 2003 McMahon et al., 2004 Song et al., 2005 Kasten et al., 2007 Espinosa et al., 2008). Kv3 channels are distinguished by rapid activation and deactivation, which confer upon them the ability to rapidly repolarize action potentials. By activating quickly, they maintain action potential brevity and, by deactivating quickly, allow for spiking at high rates. Four separate genes (Kcnc1-4) encoding subunits Kv3.1–Kv3.4 assemble into homotetrameric and heterotetrameric channels, exhibiting distinctive and overlapping expression patterns (Rudy et al., 1999 Rudy and McBain, 2001).

In the cerebellar cortex (Fig. 1), the projection neurons, Purkinje cells, express Kv3.3 throughout the cell and Kv3.4 largely in dendrites (Martina et al., 2003 McMahon et al., 2004 Chang et al., 2007). Granule cells express Kv3.1 and Kv3.3 (Weiser et al., 1995 Sekirnjak et al., 1997 Ozaita et al., 2002), whereas basket cells express Kv3.2 and Kv3.4 channel subunits in the pinceaux (Veh et al., 1995 Bobik et al., 2004). In the deep cerebellar nuclei (DCN) and analogous vestibular nuclei, mRNA for all subfamily members is present (Weiser et al., 1994), with Kv3.1 and Kv3.3 protein is expressed in large, presumably glutamatergic projection neurons (Weiser et al., 1995 McMahon et al., 2004). Kv3.2 protein is also present in these nuclei (Lau et al., 2000), and Kv3.4 has not been examined here.

Simplified cerebellar circuitry illustrating direction of information flow and Kv3 channel localization. Interneurons and inhibitory projection neurons from the DCN to the inferior olive have been omitted for simplicity. Cerebellar input arrives from climbing and mossy fibers to stimulate neurons in the deep nuclei and the cortex, in which climbing fibers stimulate Purkinje cells directly and granule cells excited by mossy fibers stimulate Purkinje cells by way of parallel fibers. Purkinje cells in turn inhibit deep nuclear neurons also, at a latency, leading to rebound excitation.

Mice expressing Kcnc1 but lacking Kcnc3 (+/+ −/−) exhibit ataxia and abnormal spiking in Purkinje cells which rely on Kv3.3 for brief action potentials (McMahon et al., 2004). Re-expression of Kv3.3 selectively in Purkinje cells restored spike parameters and rescued motor coordination in +/+ −/− mice as well as in mice additionally lacking one Kcnc1 allele (+/− −/−) (Hurlock et al., 2008). As Purkinje cell firing is potentially related to the control of fine motor timing or speed, determination of what aspect of motor function was affected in the ataxia of +/+ −/− mutants is of interest.

For electrophysiological alterations to be consequential for behavior, the intrinsic firing properties of DCN neurons must be reasonably intact for the rescue of fast repolarization in Purkinje cells to be efficacious in averting ataxia in +/+ −/− mutants (Fig. 1). Large, non-GABAergic DCN projection neurons express Kv3.1 and Kv3.3 channel subunits. Thus, we assessed Kcnc1/Kcnc3 double-knock-out (−/− −/−) mice for a behavioral and electrophysiological rescue. To extricate whether the ataxia stems from hypermetric speed/force, timing of limb movement, or relative timing across limbs, we also measured high-speed motor performance and gait pattern alterations. To define which electrophysiological changes could most plausibly account for the ataxia, we examined intrinsic spiking of DCN neurons in mice lacking Kcnc3 as well as Kcnc1 alleles.