maps is that inhibitory circuits undergo compensatory modification of the enlarged receptive fields. To test this hypothesis, we used electrophysiological techniques to map out the contribution of inhibitory inputs to the receptive fields of SC neurons in partial tectum (PT) and 2-amino-5-phosphonovaleric acid (APV) cases. We found that the extent and the strength of inhibition were increased, maintaining a balance between excitatory and inhibitory subfields that compensated for the increased number of excitatory inputs in the APV group (Razak et al., 2002, 2003).

These results together show that an increase in the number of afferents can be compensated for by homeostatic, matching alterations in both excitatory and inhibitory circuitry, maintaining the ability of the system to identify visual objects. This has obvious utility from an evolutionary perspective. Any individual variation producing a change in the number of inputs could not only be incorporated but could also produce an adaptive advantage. Role of Sensory Deprivation in Specificity of Thalamocortical Pathways

Connections between sensory thalamus and sensory cortex are demonstrably quite specific from early in development (Crandall and Caviness, 1984; Miller et al., 1991), and thus are a good experimental system in which to study target specificity. If sensory inputs can specify cortical identity through thalamo-cortical axon (TCA) pathfinding, then alteration of TCAs could change cortical identity. Multiple, large tracer injections in primary auditory cortex (A1) of deafened ferrets revealed an anomalous thalamocortical pathway in the deaf ferret brains, connecting visual thalamus (lateral geniculate nucleus (LGN) and lateral posterior nucleus (LP)/pulvinar) with auditory cortex (Pallas et al., 2002). This result is particularly surprising because previous thinking holds that thalamocortical targeting is hard-wired by molecular guidance factors and not affected by sensory experience. This LGN-A1 projection is not a normal developmental exuberance that was stabilized, rather it is novel and a response to the deafferentation occurring several synapses away. This finding is relevant clinically because it could provide a substrate for the recovery of function seen in deaf individuals (Neville, 1990). If a remodeling of the thalamocortical pathway could occur evolu-tionarily, it could provide a means of generating a new cortical area (Kaas, 1993). Our observation thus provides us with a model system for brain evolution. If a change in thalamocortical afferent source can direct cortical circuitry for processing to its own ends, then peripheral change could induce a matching central change without the need for genomic modification. Cross-Modal Plasticity in the Retinogeniculocortical System Parcellation of mammalian neocortex

The evolution of the mammalian neocortex has involved a spectacular amount of parcellation and diversification (Felleman and Van Essen, 1991; see Rosa and Krubitzer, 1999, for review). Several explanations for this evolutionary change seem possible. Perhaps the least likely is simultaneous, matching mutations that affect both afferent inputs and their cortical targets. Alternatively, intrinsic genetic information, either from afferent inputs or from cortex itself, could drive parcellation and circuit arrangement (Rakic, 1988). Third, unique spatiotemporal patterns of afferent activity reaching each presumptive cortical area could instruct parcellation and circuit organization (Hebb, 1949; O'Leary, 1989; Kaas, 1995; Northcutt and Kaas, 1995; Kaas and Reine, 1999). Finally, cortical circuitry could be sufficiently similar in different regions that changes would be unnecessary to accommodate adaptive changes in afferent inputs. The cross-modal plasticity paradigm

Cross-modal plasticity provides a way of examining the mechanism behind cortical parcellation during development and evolution. The paradigm permits experimentally changing the modality and activity pattern within TCAs without changing their molecular identity. The hypothesis being tested with this approach is that the sensory information received by a region of cortex during development (or evolution) plays an essential role in organizing its functional identity. In contrast to specification according to intrinsic instructions, afferent instruction would allow animals to be responsive to a changing sensory environment or to alterations in peripheral sensory systems. This dependence on experience would have obvious adaptive significance. However, although some advances in understanding cortical parcellation have been made, it is not known which characteristics distinguishing different cortical areas are due to intrinsic, preprogrammed differences, and which are due to activity in the sensory inputs, and in the latter case, whether sensory activity is permissive or instructive (Levitt et al., 1997; Crair, 1999).

In cross-modal plasticity experiments, afferents of one sensory modality are induced to innervate thalamus of a different modality, which then carries the cross-modal information to the sensory cortex (Figure 4). These experiments were first performed in hamsters (Schneider, 1973; Frost, 1981) and then in ferrets (Sur et al., 1988), which are carnivores with sensory cortical physiology similar to that of the cat, except that they are born earlier in development. The manipulation is performed before sensory information has gained access to cortex (Luskin and Shatz, 1985; Jackson et al., 1989). In ferrets, external visual and auditory cues are not available until after postnatal day 30 (Moore and Hine, 1992; Akerman et al., 2002), in part because the ears and eyes open at this time. This protracted developmental period allows a more detailed examination of the process than is possible in rodents.





Figure 4 Procedure for inducing retinal projections into auditory thalamus (MGN). Midbrain lesions simultaneously deafferent MGN and eliminate retinal target space. As a result, retinal axons sprout into MGN. MGN in turn carries visual information into primary auditory cortex (A1). Cross-modal A1 can process and map visual inputs Experiments performed in the Frost lab demonstrated that retinal axons lacking target space could invade either somatosensory or auditory thalamus, inducing visual responses and some retinotopic organization in the cortical area affected (Frost, 1981; Metin and Frost, 1989). Extending the cross-modal paradigm to ferrets (Sur et al., 1988) showed that early visual activation of A1 causes auditory cortex to resemble visual cortex in both its topography and receptive field properties (Roe et al., 1990, 1992; see Sur et al., 1990, for review). Remarkably, cross-modal A1 in ferrets or hamsters can mediate rudimentary visual perception (Frost et al., 2000; von Melchner et al., 2000). Thus it appears that supplying A1 with early visual input transforms it in a functional sense into a visual cortical area. The remaining question is how this transformation occurs. Either visual and auditory cortex are so similar that they can process each other's inputs without modification, or the visual input has caused a modification in A1's circuitry that allows it to process visual information appropriately. Our goal has been to distinguish between these two possibilities. Anatomical basis for visual responses and topography in cross-modal A1 Using an anatomical approach, we showed that the early anomalous visual inputs do in fact cause modifications to A1's circuitry on multiple levels, including intracortical (horizontal) and corticocortical (cal-losal) connectivity patterns and local inhibitory circuitry. In normal auditory cortex of cats and ferrets, callosal connections are organized into the so-called binaural bands along the tonotopic axis, uniting A1 neurons in both brain hemispheres that are excited by sound presented to either ear (EE cells: Imig and Brugge, 1978; Figure 5, normal). Another class of A1 neurons is inhibited by sound in one ear (EI cells) and does not project callosally. The EI cells form interdigitating bands with the callosally projecting EE cells (Imig and Brugge, 1978). The pattern of callosal bands arises during development from a more diffuse early pattern (Feng and Brugge, 1983), suggesting that auditory experience is involved in its refinement. Perpendicular to the tonotopic axis, intracortical, horizontal connections in A1 unite neurons that have similar frequency tuning (Matsubara and Phillips, 1988; Wallace and Bajwa, 1991; Gao and Pallas, 1999). In this case, the early projections are somewhat specific in their targeting, although there is some refinement occurring at about the time of hearing onset that is lost with early deafening (Figure 5, deaf; Gao et al., 1999a;

Horizontal connections-isofrequency axis Normal Cross-modal Deaf

Callosal connections-binaural bands Normal Cross-modal Deaf

Callosal connections-binaural bands Normal Cross-modal Deaf

Figure 5 Injection of local or long-distance tracers reveals the connectivity of auditory cortex in normal, cross-modal, or deafened ferrets. Local horizontal projections within A1 interconnect neurons with similar sound frequency tuning. Callosal connections run along the binaural bands of neurons receiving excitatory inputs from both ears. Deafferentation of A1 results in loss of specificity in these connection systems, suggesting an activity-dependent basis. Visual input to A1 in the left hemisphere alters the connections with respect to both pattern and location. Adapted from Gao, W.-J. and Pallas, S. L. 1999. Cross-modal reorganization of horizontal connectivity in auditory cortex without altering thalamocortical projection. J. Neurosci. 19, 7940-7950; Pallas, S. L., Littmann, T., and Moore, D. R. 1999. Cross-modal reorganization of callosal connectivity in auditory cortex without altering thalamocortical projections. Proc. Natl. Acad. Sci. USA 96, 8751-8756.

Moerschel and Pallas, 2001). Interestingly, it seems that there is more error correction occurring in visual cortex during development than in auditory cortex (Ruthazer and Stryker, 1996; White et al., 2001), but in both cases neural activity is critical for normal development.

We reasoned that, if the cortical connectivity pattern in A1 of adults is dependent on auditory experience, then supplying visual inputs to A1 in cross-modal ferrets should result in a pattern more closely resembling that in visual cortex. Horizontal connections in primary visual cortex (V1) are organized as a periodic, symmetrical array of clusters that interconnect neurons with similar visual orientation tuning (Callaway and Katz, 1990; Malach et al., 1993). These clusters are refined under the influence of visual activity (Callaway and Katz, 1991; Ruthazer and Stryker, 1996). Callosal connections in visual cortex of adults are made between neurons at the border of visual cortical areas 17 and 18 (Lewis and Olavarria, 1995; Olavarria, 2001; Riederer et al., 2004; see The Role of Transient, Exuberant Axonal Structures in the Evolution of Cerebral Cortex for review).

The rewiring manipulation is performed on one side of the brain only; thus the normal, contralateral auditory cortex can communicate auditory information across the corpus callosum. This raises the interesting possibility that A1 may have both visual and auditory-responsive neurons. Our anatomical results were consistent with this possibility (Figure 5, cross-modal). Horizontal connections in cross-modal A1 were arranged in a radially symmetric array of clusters as in V1, and, unlike in normal A1, extended toward the medial part of cross-modal A1 (Gao and Pallas, 1999). Callosal connections, in contrast, were pushed laterally and were entirely absent from the medial part of cross-modal A1. The remaining callosally projecting neurons were organized in patches instead of the binaural bands seen in normal A1 (Pallas et al., 1999).

Because we found that the callosal and horizontal connections in cross-modal A1 were arranged in a mutually exclusive pattern, we have proposed that A1 is split in two by the anomalous visual and normal auditory inputs (Pallas, 2002). Laterally, callosal connections would interconnect sound-responsive neurons in both hemispheres. Visually responsive neurons would be preferentially located in medial A1 of cross-modal ferrets, where they would be interconnected according to their orientation tuning. Evidence for iso-orientation connectivity in cross-modal A1 has been provided using optical imaging methods (Sharma et al., 2000), supporting this idea. If it is the case that cross-modal A1 is subdivided by its bimodal inputs, then we can model the evolution of a new cortical area on a developmental timescale. This would provide an ideal system in which to study how evolutionary changes in sensory input could trigger development of a cortical area to process those inputs (Kaas, 1995). In either case, the results show that changing only the pattern of activity without changing the molecular identity of thalamic inputs can configure cortical circuits adaptively. Role of inhibition in cross-modal plasticity We have also tested the hypothesis that rearrangements of inhibitory circuitry occur in cross-modal A1. The rationale is that the pathway from the auditory thalamus (medial geniculate nucleus (MGN)) to the auditory cortex projects as a one-dimensional sheet along the isofrequency axis (Pallas et al., 1990), suggesting that in order to represent visual spatial topography in two dimensions, suppression of select subsets of these one-dimensional projections would be necessary. Using immunocyto-chemical methods to catalog subsets of GABAergic nonspiny, nonpyramidal interneurons, we found that there was an increase in number and a change in morphology of a calbindin-containing subset of inhibitory neurons in A1 as a result of early visual inputs (Gao etal, 1999b, 2000; see Pallas, 2001a, for review). The next question that must be addressed is whether changes in inhibitory circuitry are necessary or sufficient to reconfigure auditory cortex for a visual processing role. We predict that local blockade of intracortical inhibition will eliminate response properties that depend on the two-dimensional nature of visual stimuli, as opposed to the one-dimensional arrangement of sound frequency coding. Molecular specification of cortical areas through axon guidance There are several possible explanations of how the functional identity of each cortical area is established. Thalamic projections are well targeted during pathfinding (Crandall and Caviness, 1984; Miller et al., 1993), forming sharp boundaries between projections to adjacent cortical areas, and thus could instruct cortical identity. How these precise projections of thalamo-cortical afferents are targeted remains to be determined, though previous studies have pointed to the ventral telencephalon and the cortical subplate as containing important guidance cues (Ghosh et al., 1990; Catalano and Shatz, 1998; Lopez-Bendito and Molnar, 2003), and to matching adhesion factors (cadherins) in thalamus and cortex (Gil et al., 2002; Poskanzer et al., 2003). Information intrinsic to cortex, such as transcription factors and markers of positional identity, may be sufficient to set up guidance of TCAs to their appropriate final cortical targets or to establish other unique features. Conversely, cortex may be a tabula rasa that is instructed by the TCAs to develop in a particular fashion. More likely the answer lies in between. Our recent effort has been to explore the expression and action of genes identified as candidates for a role in the specification of cortical identity.

Regionally patterned transcription factors and adhesion molecules are thought to be responsible for areal patterning, via specification of both positional information and TCA targeting (Grove and Fukuchi-Shimogori, 2003). Patterned gene expression is known to be involved in the specification of axonal projection patterns in the hindbrain (Keynes and Krumlauf, 1994) and spinal cord (Stoeckli and Landmesser, 1998). This protomap idea (Rakic, 1988) has been tested by cortical gene knockout studies (Bishop et al., 2000; Mallamaci et al., 2000) and by examination of mutants in which thalamocortical projections do not form or form incorrectly (Miyashita-Lin et al., 1999; Nakagawa et al., 1999; Tuttle et al., 1999; Garel et al., 2002). The finding that many patterning genes are expressed whether or not the TCAs are present supports the idea that they instruct TCA pathfind-ing. None of the genes found so far are restricted to areal boundaries in their expression, however (Donoghue and Rakic, 1999). Thus, a combinatorial, threshold effect, similar to that for guidance of retinal axons by the repulsive molecules called ephrins (Hansen et al., 2004; Tadesse et al., 2004), has been suggested (Donoghue and Rakic, 1999).

Knockout of the transcription factor gene Pax6, which is expressed in a rostrolaterally increasing gradient, results in an apparent caudalization of the remaining cortical epithelium, as defined by rostral shifts in expression of the adhesion factor gene Cad8 and other caudal marker genes (Bishop et al., 2000; Mallamaci et al., 2000). Conversely, knockout of Emx2, another transcription factor gene that is expressed in an opposite gradient to Pax6, ante-riorizes positional identity in cortex, and is associated with caudal shifts in the normally anteriorly restricted expression of Cad6 and other anterior marker genes. In addition, expansion or contraction of an FGF8 expression domain induced by manipulating Emx2 levels results in opposite shifts in the location of the somatosensory barrel field (Fukuchi-Shimogori and Grove, 2001, 2003). If there were corresponding defects in TCA path-finding, then a role for the genes in areal specification via TCA guidance would be supported. In Emx2 mutants, a caudal shift of TCAs to soma-tosensory cortex occurs, and the projection of TCAs to visual cortex is reduced concomitantly (Bishop et al., 2000; Mallamaci et al., 2000), prompting a suggestion that the loss of Emx2 redirects TCAs through a change in positional identity of caudal cortex. However, this interpretation has been challenged by reports that TCAs become mistargeted in the ventral telencephalon, before they come into contact with the altered cortical regions (Caric et al., 1997; Garel et al., 2002; Molnar, 2000). Furthermore, another study has shown that similar Fgf8-induced shifts in patterned gene expression, and a resulting caudalization of the cortical epithelium, are not associated with any shift in TCA projection patterns in neonates (Garel et al., 2003). It is possible that, rather than respecifying cortical identity, loss of Pax6 or Emx2 transcription factors interferes with the generation of neurons in cortex. Indeed, the cortex is smaller in the mutants (Bishop et al., 2000), supporting this alternative interpretation. Another complication is that the knockout method eliminates gene function everywhere, throughout life, but these genes likely play multiple roles in development across space and time. In addition, the mutants die before the borders of cortical areas can be confirmed functionally, or have other profound defects that complicate interpretation of the data (Caric et al., 1997; Molnar, 2000; Pratt et al., 2000; Garel et al., 2002). Thus the question of the role of these genes in cortical specification requires further study, with complementary approaches.

Our approach affords the unique advantage of combining physiological and molecular investigations in an animal, the ferret, with protracted postnatal development and well-characterized cortical physiology, providing a model that has major advantages over mice. We combined quantitative reverse transcriptase polymer chain reaction (RT-PCR) with neuroanatomical tracing of both TCA and corticocortical projections in ferrets, comparing A1 and V1 in normal animals in topography and levels of gene expression in relation to the timing of axon ingrowth. We found that, in normal animals, Pax6 and Emx2 expression gradients were declining by postnatal day 10 (P10) to P14, prior to TCA ingrowth. Differential expression of Cad6 and Cad8, however, was maximal during the period of TCA targeting and synapse formation (P14 to P25) and overlapped with the later development of specific corticocortical connectivity (Xu et al., 2003; Figure 6). The expression patterns of Pax6 and Emx2 across development and between cortical areas are consistent with a role in orchestrating gradients of neurogenesis, providing early regional patterning in cortex, and/or triggering a signal cascade that continues in their absence, but not with direct specification of axon targeting to cortex. These genes likely trigger expression of downstream genes such as the cadherins that could be more directly involved with correct pathfinding by TCA axons. The expression patterns between different cortical areas, correlated with the timing of afferent ingrowth, provided the impetus for our further investigations into the roles of cadherins in cortical plasticity.

Individual functional areas within the cerebral cortex have characteristic connectivity patterns which exhibit experience-dependent development and plasticity. Another of our efforts has been to look for a relationship between connectivity patterns and gene expression patterns. The classical cadherins may instruct, or be instructed by, thalamocortical ingrowth (Bishop et al., 2000; Mallamaci et al., 2000; Gil et al., 2002). We tested the hypothesis that, if cadherins are instructive for TCA guidance, then in experimental models producing mistargeting of TCAs, cadherins should be misexpressed. Ferrets deafened prior to the onset of hearing exhibit mis-targeting of both retinal and thalamocortical projections (see above and Pallas and Moore, 1997; Pallas et al., 2002). We predicted that these deaf

Pax6 expression Emx2 expression

Emx2 Cortex

Postnatal age Postnatal age

Cad6 expression Cad8 expression

Postnatal age Postnatal age

Figure 6 Expression levels of four putative cortical patterning genes were measured with quantitative RT-PCR. Top: Early in development, prior to TCA innervation, Pax6 is high in auditory cortex and Emx2 is higher in visual cortex, potentially providing positional information about regional cortical identity. As the TCAs are growing into the cortical plate and corticocortical connections are being formed, Pax6 and Emx2 levels drop, but the cadherin pair Cad6 and Cad8 form similar opposing gradients. The cadherins are thought to facilitate patterning of connections throughout the brain. Adapted from Xu, M., Baro, D. J., and Pallas, S. L. 2003. A quantitative study of gene expression topography in visual and auditory cortex during thalamocortical development in postnatal ferrets. Soc. Neurosci. Abstr. 29, 673-677.

ferrets with ectopic projections between visual thalamus (LGN) and auditory cortex would exhibit alterations in the spatial pattern of cadherin expression. However, this was not the case. Results from quantitative RT-PCR of cadherin mRNA and Western blots of cadherin protein were similar in normal and deafened ferrets. These data could be interpreted as evidence against a role for cadherins in TCA targeting. The LGN to A1 projections are vastly outnumbered by normal MGN to A1 projections in the deafened ferrets, however, and thus any alteration of cadherin expression may be undetect-able with our method. Further study is warranted to uncover how TCA pathfinding strategies can respond to deafferentation of neighboring cortical areas, because it could provide an important basis for the recovery of function seen in humans with impaired sensory ability such as blindness or deafness (Neville,

1990; Sadato et al., 1996; Cohen et al., 1997; Finney et al., 2001; Bavelier and Neville, 2002).

In contrast to this negative result, examination of cadherin mRNA and protein expression in cross-modal animals with altered corticocortical connectivity patterns revealed intriguing changes in cross-modal compared to normal animals. The patterning of corticocortical connectivity reflects the modular organization of the information-processing circuitry of neocortex and is essential to sensory perception. Connectivity within and between cortical areas starts out somewhat diffusely organized and refines during development, at least partly under the influence of activity (Innocenti, 1981; Feng and Brugge, 1983; Callaway and Katz, 1990; Callaway and Katz, 1991; Schlaggar and O'Leary, 1991). There is also evidence for intrinsic specification of corticocortical connectivity patterns by restricted distribution of gene products, including cadherins (Korematsu and Redies, 1997; Suzuki et al., 1997; Bekirov et al., 2002; Huffman et al., 2004). We reasoned that, if adhesion factors such as the cadherins are involved in targeting of cortico-cortical projections, at the direction of upstream transcription factors such as Pax6 and Emx2, then the cross-modal ferrets with experience-dependent alterations in horizontal and callosal connectivity patterns described above might have cadherin expression patterns reflective of that change. Specifically, the cadherin expression pattern in cross-modal A1 should resemble that in normal V1, given that the connectivity patterns in A1 are organized as they are in V1. Consistent with this idea, we found that expression levels of cadherin 6 and 8 mRNA and protein are similar across A1 and V1 of cross-modal ferrets, and no longer exhibit the graded differential expression seen in normal animals (Xu and Pallas, 2005). Thus cadherins could no longer provide information that could be used to differentiate auditory from visual cortex during the development and refinement of connectivity patterns. Evolutionary Evidence of Cross-Modal Respecification of Cortical Identity

The ability to induce cross-modal projections in an experimental situation is useful only to the extent that it can model a natural event. In addition to the parallels between clinical reports of sensory substitution in the deaf and blind (Bavelier and Neville, 2002), and our findings in cross-modal ferrets, there are evolutionary experiments that also show parallels. Perhaps most well studied are the blind mole rats of the Middle Eastern desert, Spalax ehrenbergi. These fossorial creatures are born with a rudimentary eye that does not form images (Bronchti et al., 1991). Wollberg and colleagues have shown that visual cortex in these rodents is activated by auditory stimuli (Bronchti et al., 1989, 2002; Sadka and Wollberg, 2004). This evolutionary inverse of our cross-modal manipulation in ferrets provides the proof of principle that peripheral alterations that occur in evolution can be seamlessly incorporated by plasticity inherent in sensory cortex.

Was this article helpful?

0 0

Post a comment