The Potential Role Of Corticothalamic Feedback Projections In Tactile Information Processing

The FF model of touch completely ignores the fact that, like all other mammalian sensory systems, the somatosensory system also contains massive corticocortical and corticofugal feedback projections. For instance, in primates, feedforward soma-tosensory pathways terminate in four distinct somatosensory areas located in the anterior parietal cortex (areas 3a, 3b, 1, and 2). These areas are connected through feedforward corticocortical projections. Projections from the anterior parietal cortex also reach motor cortical areas in the frontal lobe, the secondary somatosensory cortex, and the somatosensory and multi-modal cortical areas of the posterior parietal cortex.61 What is often ignored in any model of touch is that cortical neurons located in frontal motor areas and in the posterior somatosensory areas give rise to massive corticocortical feedback projections that target somatosensory fields in the anterior parietal cortex. Anterior parietal areas are also reciprocally connected through feedforward and feedback projections. In addition, corticofugal projections which originate in the infragranular layers of the primary and higher order somatosensory cortical areas project to all intermediary subcortical relays (i.e., spinal cord, brainstem, and thalamic nuclei) of the somatosensory system.61 Indeed, once the full domain of these feedback projections is considered, the somatosensory system can only be defined as a highly recurrent network, in which multiple feedback projections are intertwined with several parallel feedforward pathways.

The importance of these corticofugal projections to any neurophysiological model of touch can be illustrated by a brief description of the anatomical organization and the physiological effects mediated by these pathways. Recently, the anatomical organization of corticothalamic projections has been investigated in great detail in rodents. As in every other mammalian species,w7.110 feedback projections from several somatosensory (e.g., SI, SII, PV, etc.) cortical areas 749 converge on neurons located in primary and secondary thalamic nuclei (e.g., VPM, POM, and ZI) of the trigeminal system of rodents. Studies in mice51,53 and rats7,26 have shown that these corticothalamic projections terminate primarily in the distal dendrites of these thalamic neurons.100 In the case of the ventral posterior medial (VPM) nucleus, the primary thalamic relay of the trigeminal system, these corticothalamic projections are organized in a topographic manner7,27,51,130 (see Figure 12.1). In this arrangement, corticothalamic projections originating from layer VI neurons, which are located under a particular cortical barrel127 (e.g., barrel D1 in layer IV), terminate on thalamic neurons located across the thalamic barreloids 123 that define the representation of a whisker arc or column (e.g., barreloids A1, B1, C1, D1, and E1) in the VPM (see Figure 12.1). Corticothalamic projections from layer VI also reach the reticular nucleus (RT) of the thalamus,98,100 the main source of GABAergic inhibition in the rat VPM.99 These cortical-RT projections are also organized in a topographic arrangement, which seem to be orthogonal to that observed in the VPM nucleus.51,52 Thus, axons from layer VI neurons located under a given cortical barrel (e.g., C1) target neurons located across the representation of a whisker row (e.g., C1, C2, C3, and C4) in the RT nucleus.51,52 Neurons located in secondary somatosensory thalamic nuclei, such as the posterior medial nucleus (POM), also receive corticothalamic terminals,51,53 albeit these are primarily derived from pyramidal neurons located in layer V of the somatosensory cortex. The morphology of corticothalamic terminals also varies according to whether they terminate in the primary (e.g., VPM) or secondary thalamic relay (e.g., POM).52,53

Physiological studies have shown that corticothalamic projections are primarily excitatory and likely employ glutamate as their main neurotransmitter.119,120 The glutamate released from these corticothalamic terminals acts on NMDA, AMPA, and metabotropic receptors located in the distal dendrites of thalamic neurons.73,105,120 Activation of NMDA, AMPA, and metabotropic receptors by in vitro stimulation of corticothalamic axons produces long-lasting, slow-rising EPSPs in the thalamus.106,119 Based on some of these findings, corticothalamic-mediated activation of metabotropic receptors has been suggested to produce the modulation of neuronal firing in the VPM nucleus.106,119 For instance, it is conceivable that the slowly rising depolarization produced by activation of corticothalamic projections could allow thalamic neurons to reach firing threshold in the presence of subthreshold synaptic input. In addition, corticothalamic afferents could also contribute to the slow activation of a low-threshold calcium conductance that underlies the production of bursts of action potentials by thalamic neurons.110

Despite a wealth of anatomical, pharmacological, and in vitro physiological information, the role played by corticothalamic projections in tactile information processing has remained elusive. For instance, penicillin-induced epileptic discharge in the cat somatosensory cortex96 and cortical spreading depression in the rat cortex2 were found to induce a depression of sensory evoked responses in the thalamus. In another series of experiments, carried out in both anesthetized and awake preparations, Yuan et al.128-129 reported that lidocaine-induced inactivation of SI cortex resulted in reduced thalamic responses to electrocutaneous stimulation without any effect on the spontaneous activity, stimulus threshold, response latency, and receptive fields of the same thalamic neurons. Other studies, however, have reported a facil-itatory influence of SI cortex on evoked thalamic discharges3'4 using cortical spreading depression125 or electrical stimulation.5

It is likely that part of the confusion in the literature arises because corticotha-lamic pathways can mediate both a monosynaptic excitatory and a dysynaptic inhibitory (via RT nucleus) postsynaptic potential in the thalamus. Thus, depending on how the cortex is stimulated or blocked, a variety of facilitatory and inhibitory response effects could be induced in the thalamus. Thus, by microstimulation of small territories of the SI cortex, Shin and Chapin115 described a range of thalamic effects, in addition to an overall suppressive influence of thalamic sensory responses, that depended upon the topographic location of neurons in the ventral posterior nucleus of the thalamus.

In our hands, pharmacological block of SI cortical activity by focal infusion of the GABAa agonist muscimol, and consequent silencing of pools of cortical neurons that give rise to corticofugal projections to the thalamus and brainstem, produced a series of physiological effects in the rat VPM.66 First, we observed that blocking cortical activity altered both the short- and long-latency components of the tactile responses of VPM neurons. The end result of these modifications was the demonstration that corticofugal projections contribute to the definition of the complex spatiotemporal structure66 of the RFs of VPM neurons. These results were obtained by using traditional single-whisker stimuli. When more complex tactile stimuli were used in our experiments, we observed that the ability of VPM neurons to integrate complex tactile stimuli (e.g., multi-whisker deflections) in a non-linear way was also significantly reduced by a pharmacological block of cortical activity. Both supra-and sub-linear summation of multi-whisker stimuli 41 was reduced in these experi-ments.39 Overall, these findings not only support the hypothesis that corticothalamic projections may mediate both facilitatory and suppressing effects on thalamic neurons, but they also suggest that the action of these corticofugal projections may also depend on the type of tactile stimulus provided to the somatosensory system. As we will see below, there is indirect evidence that the physiological contribution of these descending pathways to tactile information processing may also depend on the behavioral state of the animal.

Further evidence for the functional relevance of corticofugal projections in the rat somatosensory system was obtained in studies carried out in our laboratory to evaluate the contribution of corticofugal projections to the ability of subcortical neurons to express unmasking of novel tactile responses following a peripheral deafferentation.66 This reorganization process, which we dubbed immediate or acute plasticity, is known to trigger a system-wide reorganization of the somatotopic maps located at cortical, thalamic, and brainstem levels.33 The most conspicuous effect of this immediate reorganization is the shifting of receptive fields of individual neurons away from the deafferented region due to the unmasking of neuronal tactile responses that were not present before the peripheral block. Interestingly, such unmasking tends to occur almost simultaneously in the brainstem, thalamus, and cortex.33 In a recent series of experiments, we observed that blocking of neuronal activity in the infragranular layers of the SI cortex, a procedure that silences the projecting neurons that give rise to corticobulbar and corticothalamic feedback projections, reduces by almost 50% the number of VPM thalamic neurons that exhibit unmasking of tactile responses following a partial and reversible peripheral deafferentation.66 Although plastic reorganization in the VPM nucleus is still observed after cortical inactivation, its spatial extent is reduced significantly. These findings have been confirmed and extended further by the recent demonstration that the immediate, but not the late phase of plastic reorganization in the ventral posterior lateral nucleus (the thalamic relay for somatosensory fibers from the rest of the body), are reduced or eliminated by removal of corticofugal projections.97

Hebbian-based rules of synaptic plasticity have been proposed to account for the occurrence of adult sensory plasticity in a feedforward model of touch.76 Despite this addition, a strict feedforward view is clearly at odds with the evidence implicating cortical feedback projections in the early stages of thalamic plasticity. Thus, the available empirical evidence seems to indicate that any attempt to apply a pure FF model of the somatosensory system to account for the phenomenon of adult plasticity will be short lived.

Further support for the functional relevance of descending corticofugal projections comes from the observation that these projections have been demonstrated to affect the physiological properties of several other subcortical relays of the soma-tosensory system. For instance, block of neuronal activity in the SI cortex has been reported to eliminate most of the tactile responses of neurons located in the POM nucleus of the thalamus.28 In addition, corticobulbar projections have also been shown to influence the physiological properties of neurons located in the brainstem nuclei that relay ascending somatosensory information to the thalamus.55 For instance, removal of corticofugal projections in rats increases the responsiveness of neurons in spinal trigeminal brainstem complex to whisker stimuli.55

Overall, the results reviewed above make a compelling case for the need to incorporate recurrent corticofugal projections as an integral part of a comprehensive and realistic model of touch. Indeed, the recurrent nature of the somatosensory system strengthens our hypothesis that the mammalian somatosensory system relies on highly distributed neuronal interactions, which emerge from the dynamic interplay of multiple ascending and descending pathways, to represent tactile informa tion.85-88'90'92-94 Although the concept of distributed processing is not new, and many investigators have proposed schemes based on population coding,25'30'31'37'49'81'109 this distributed encoding scheme has recently attracted the attention of neuroscientists because of the successful application of artificial neural networks in pattern recognition problems.6'44'46 In a distributed coding scheme, divergent neural connections ensure that specific units of information are not held in single or small groups of neurons, but instead are widely distributed, or "encoded," by large neural ensembles located at multiple cortical and subcortical levels of the system.49 Consequently, each neuron contributes in some way to processing of most of the information handled by the network. In line with this hypothesis, a series of studies in our lab, as well as other labs,42'64'71'77'85'89'92'101 has begun to re-examine traditional views of information encoding by the somatosensory system. Anatomical evidence in favor of a distributed model includes the fact that ascending feedforward (FF) somatosen-sory pathways that carry information from the periphery to the SI cortex exhibit different degrees of divergence,18'19'54'68-70'99'104'124 which contribute to the large multi-whisker RFs observed in the VPM and SI.42'86 Thus, the effects of even small but incremental changes at each processing level of the pathway (e.g., from brainstem to thalamus) would tend to multiply through successive relays and could be markedly amplified by the time they reached the cortex. In addition, wide-field sensory inputs, such as high-threshold mechanical and noxious stimuli, which are transmitted through paralemniscal pathways, could also converge on cortical neurons. These effects could be further amplified by corticocortical connections within the SI and between the SI and other cortical areas.12'32'91 In this context, the existence of massive divergent corticofugal feedback projections to all subcortical somatosensory relay nuclei provide almost unlimited opportunity for increasing the ultimate radius of influence from a single sensory event.80 In this model, single neurons would not function as single feature detectors to serve as the functional unit of the system. Instead, neurons would work as part of ensembles that are capable of representing and processing multiple tactile attributes of a given complex stimulus simultaneously.

Massive corticofugal projections, that reach somatosensory relay structures located in the thalamus, brainstem, and the spinal cord, could offer the anatomical substrate for massive neural ensembles dedicated for tactile information processing. Such structures could be formed by somatosensory, motor, limbic, and association cortical areas, and would influence the activity of neurons located in subcortical centers, even before mechanoreceptors in the skin are activated by a tactile stimulus. According to this view, corticofugal feedback projections could incorporate subcortical nuclei into the computational processes required for the emergence of tactile percepts. Although rarely discussed in the literature, reciprocal loops between cortical and thalamic nuclei could also mediate a different type of corticocortical communication, in which thalamic networks could be employed to transform convergent signals from one or more cortical areas and then disseminate the result of this transformation to vast cortical territories. Such an interactive view of the soma-tosensory system would predict that top-down influences would be capable of modulating the activity of subcortical neurons during different behavioral states. But is there evidence for the existence of such top-down influences in the somatosensory system? In the next section, we describe a well-known phenomenon that may provide the key for unraveling the primordial physiological role played by corticofugal feedback on tactile perception and, hence, serve as the basis for mounting a formidable challenge to the FF model of touch.

0 0

Post a comment