Cortico Striatallike Auditory Processing Module

FIGURE 16.2 A proposed cortico-striatal loop representing the auditory processing module. A cortico-striatal module comprising auditory regions of the avian telencepahlon is proposed to make up the auditory processing module. The caudal nucleus of the neostriatum sends a projection to the underlying striatal area, caudal paleostriatum. PC projects to dorsal pallidum, and dorsal pallidal neurons extend to the thalamus. The thalamic neurons project back to the pallium. The operational aspect of the auditory processing module is represented by two axonal collaterals to the midbrain dopaminergic neurons. The axonal collaterals we allude to originate from two sources: those that bifurcate from PC efferents to the dorsal pallidum, and those that diverge from the dorsal pallidum projection to the DLM. This anatomical feature provides an interface with the anterior forebrain pathway.

FIGURE 16.2 A proposed cortico-striatal loop representing the auditory processing module. A cortico-striatal module comprising auditory regions of the avian telencepahlon is proposed to make up the auditory processing module. The caudal nucleus of the neostriatum sends a projection to the underlying striatal area, caudal paleostriatum. PC projects to dorsal pallidum, and dorsal pallidal neurons extend to the thalamus. The thalamic neurons project back to the pallium. The operational aspect of the auditory processing module is represented by two axonal collaterals to the midbrain dopaminergic neurons. The axonal collaterals we allude to originate from two sources: those that bifurcate from PC efferents to the dorsal pallidum, and those that diverge from the dorsal pallidum projection to the DLM. This anatomical feature provides an interface with the anterior forebrain pathway.

16.5.1 The Cortico-Striatal Module as a Pattern Detector

The spiny striatal neuron is well suited to detecting patterned isocortical activity to the extent that the cortical firing pattern reflects a unique, behaviorally significant context (Houk, 1995). In both the mammalian and avian telencephalon, the spiny striatal cell's membrane potential oscillates between a low state, when it is hyper-polarized, and a high state, when it is in a slightly depolarized state. The downstate does not allow for action potentials to occur, whereas the upstate is particularly favorable to neural output (Wilson, 1993). In particular, the vast convergence onto single spiny striatal cells requires patterns of temporally coherent activation in order to drive these cells into an upstate. A specific detection of cortical input may emerge through coincidence detection mechanisms so that a cortical pattern comes to be favored by long-term potentiation (LTP)-mediated plasticity (e.g., Charpier and Deniau, 1997).

In light of the observed properties of these neural circuits, the striatal beat frequency (SBF) model was proposed to account for interval timing within the mammalian brain (see Matell and Meck, 2000; Matell et al., this volume). The SBF model builds on the conceptual framework embodied by contemporary views of cortico-striatal architecture and integrates it with an earlier neural network model of interval timing (Miall, 1989) by hypothesizing that event durations may be detected through the spiny striatal neuron's capacity to detect patterns of cortical input. Given the number of different cortical afferents that project to a single spiny neuron within the striatum, many different permutations of cortical afferents will fire simultaneously onto a spiny neuron throughout a timed interval. The idea of the SBF model is that persistent feedback is responsible for selectively weighting a pattern of simultaneous cortical inputs onto a spiny neuron that happens to be active at the time of occurrence of a biologically pre-potent event. Additionally, the SBF model assumes that ensembles of spiny striatal neurons are recruited to time an interval rather than individual neurons so that ensemble activity would progressively increase and reach a maximum average around the criterion time. Furthermore, members of the ensemble must have their established pattern of cortical inputs start firing together at the beginning of the interval so that the temporally effective pattern of coincident activity will be maximal at the criterion time. The MDNs could serve this function as their activation is gradually transferred from the end of the interval to the beginning with repeated training. This excitatory response to predictive stimulus onsets is assumed to synchronize the cortical inputs and innervate the ensembles of spiny neurons in order to initialize them for the timing of another interval (Matell and Meck, 2000; Matell et al., this volume).

16.5.2 The Auditory Processing Module

In this manner, the architecture of an auditory template in the auditory processing module permits the development of a temporal representation of the tutor song in order to guide vocal learning. The perception of a time marker (see Section 16.2.2) within each syllable engages a population of NCM neurons, and these are afferent to spiny cells in the PC. The consistent temporal relationship among the time markers that compose the song and the feedback provided by the completion of the song permits the ensembles of spiny neurons in the PC to be trained so that their mean firing rate is expected to gradually rise and peak at the time of the expected song completion. However, due to secondary conditioning (see Rescorla, 1980) and the song's consistent temporal organization, the various syllables would be expected to accrue associative strength so that they too could serve as feedback for temporally antecedent syllables. This brings up the potential for recruiting additional PC spiny neurons in order to coordinate their activity in relation to temporal criteria that fall between syllables, rather than only adjusting their firing rate in relation to song completion. Encoding the time of song completion is paralleled by the MDNs transferring an excitatory signal associated with this event to the onset of each syllable's time markers, as they are temporally predictive of feedback.

This system embodies the crux of simultaneous temporal processing as described by Meck and Church (1984). Several signals (syllables in our case), each one indicating the onset of a different temporal criteria, can be combined in such a way that they provide temporal information about the occurrence of a single biologically pre-potent stimulus. In this fashion, the songbird could come to behave as if it is independently timing each syllable without interference (see Pang and McAuley, this volume).

For the moment, a foundation has been established in order to explain the generation of a comparison signal due to real-time auditory feedback. One could suppose that the extent to which a perceived sequence of syllables matches the auditory template, in terms of both the spectral characteristics of each syllable and the temporal relationships among them, would influence the degree of overlap in the PC spiny neurons activated with relation to those originally recruited to encode the auditory template. Therefore, this overlap establishes the degree of the excitatory signal generated within the MDNs that is associated with predictive stimuli. Due to the reciprocal connections within the striatum via the MDNs (see Section 16.3), this excitatory signal is transmitted to the spiny cells within Area X of the AFP, modulating its activity (see Section 16.4.3).

16.5.3 Role of the Comparison Signal's Transmission to the AFP in Relation to the Posterior Motor Pathway

Recall that the songbird learns its song by producing prototypes and using the auditory template to shape the primitive song into the final song (Tchernichovski et al., 2001) (see Section 16.2.1). The activity in HVC that occurs with the onset of song production can precede the song by up to 2 sec (McCasland, 1987). The premotor drive for each prototype comes to transmit two motor signals: a primary signal toward RA via HVC[RA] and a secondary signal to Area X (see Figure 16.2). The secondary signal for each syllable prototype will activate a population of HVC[X] neurons. Keep in mind that with each vocalized syllable, there is ongoing real-time auditory feedback into the auditory processing module that contains a representation of the memorized tutor's song. The objective for the juvenile songbird is to modulate the prototype syllables so that vocal feedback engages these same PC ensembles that are tuned to the memorized song. In doing this continuously, a sequence of dopaminergic bursts is transmitted to the spiny striatal neurons in Area X in a consistent temporal pattern. This provides a substrate with which the Area X spiny neurons may be temporally entrained using the same SBF-like process that was used to construct the auditory representation. The only difference is that populations of HVC[X], rather than NCM neurons, are recruited due to the secondary signal. Individual HVC[X] neurons typically burst reliably a few times in the song motif. The bursts are not as robust or reproducible as the HVC[RA] neurons. Thus, at the level of individual HVC[X] neurons, there is no visible relationship between their bursts and song elements (M. Fee, personal communication, June, 2002). From our standpoint, however, individual HVC[X] neurons need not have a relationship with specific elements of the song because the SBF model stipulates that populations of regularly firing HVC[X] neurons are recruited as a unique pattern to project onto Area X spiny neurons.

Various ensembles of Area X spiny neurons would modulate their activity such that the net output from its target neurons, the pallidal-like cells in Area X, would come to reflect a pattern of activity corresponding to the learned durations in the auditory representation. There is a paucity of published experimental data characterizing spiny-neuronal activity in Area X of an awake, behaving songbird. Some preliminary single-unit recordings from Area X demonstrate time-locked activity for each syllable in the song, though other units had a modulatory shape in their activity profile marked by inhibition (Margoliash, 1997). The latest findings of single-unit activity in Area X found an increase in firing rate during a vocalized motif relative to no singing, while no particular relationship between the firing rate increase and elements in the song was found. On the other hand, the properties of the single cells were similar to the pallidal-like cell types found in Area X (Hessler and Doupe, 1999).

Interestingly, the increase in firing rate observed in these pallidal-like cells of Area X was accompanied by an increase in momentary pauses that interrupted activity; these pauses may be indicative of spiny-like striatal inputs, which are inhibitory (Hessler and Doupe, 1999). From the published data alone, the extensive convergence onto pallidal cells makes it difficult to draw any strong conclusions regarding the temporal pattern of the spiny striatal input. These pauses are important for current computational models involving cortico-striatal loops. The pallial location of LMAN and connectivity relative to the other telencephalic structures suggest a homology with the mammalian area of the brain equivalent to the frontal cortex (Bottjer and Johnson, 1997). This area includes the dorsolateral prefrontal cortex (dlPFC), an area most often used in computational models for cortico-striatal modules. In applying these models to the AFP, pauses in pallidal cell firing would ultimately lead to disinhibition of LMAN neurons that would permit them to go into a persistently active state. Faithfully matching the auditory template would result in a consistent temporal pattern of activity innervating LMAN, permitting a reorganization in its circuitry that would reflect this temporal-structured sequence. This process may be guided through presynaptic LTP mechanisms (Bottjer and Johnson, 1997).

In fact, during development, stabilized spatial patterns of activity appear to emerge in LMAN that reflect an accurate rendition of the bird's learned song (Boettiger and Doupe, 2001). Blocking N-methyl-D-aspartate (NMDA) receptors in LMAN prior to a juvenile bird's exposure to the tutor song impairs its ability to acquire song (Basham et al., 1996). The biophysical properties of the NMDA receptor are particularly amenable to LTP mechanisms and to the capability for sustaining activations (Wang, 1999). Moreover, depriving the young songbird of auditory information before the sensory phase prevents the development of the topographical organization from LMAN to RA even though all other connections to and from LMAN remain topographical (Iyengar and Bottjer, 2002). These results suggest that the proper connectivity between LMAN and RA requires the existence of an auditory representation, the necessity for vocal feedback, or both. The intrinsic circuitry, or perhaps the recurring pathway back to the same topographical point in Area X, might help stabilize the activation of LMAN while the interval is being timed. Similar types of reasoning have been used in computational models of the cortico-striatal modules.

How does this process facilitate encoding of the primary signal passing through the posterior motor pathway? The circuitous root of the AFP imparts a delay on the secondary signal to RA relative to the primary signal's arrival in RA. The delay on the secondary signal is sufficiently great, such that the activity from the secondary signal evoked by LMAN cells may coincide with the arrival of the primary signal for the subsequent syllable in the sequence. The significance of this is that the vocal learning system may be able to construct a song by associating the response to a syllable's secondary signal via the AFP with the pattern of activity evoked by the subsequent syllable's primary signal (Dave and Margoliash, 2000).

0 0

Post a comment