Does activity play an instructive or permissive role in the development of the visual system

It is clear that activity plays an important role in the development and plasticity of eye-specific segregation in the LGN and in at least the plasticity of ocular dominance columns. A long-standing question in the field of visual system development is whether the role of activity is instructive (the development of connections is specifically guided by patterns of activity) or permissive (some threshold level of activity is necessary for connections to be guided by activity-independent cues) (for reviews, see Crair, 1999; Katz and Shatz, 1996; Sur et al., 1999). At its most basic level, this question was answered by the earliest experiments on ocular dominance plasticity. The rearrangements of connections caused by monocular deprivation cannot be explained by a permissive role for activity. If the deprived eye were losing cortical territory merely because it did not maintain a necessary level of activity, then binocular deprivation would result in the loss of inputs to the cortex from both eyes. Similarly, the plasticity of ocular dominance in response to experimentally induced strabismus rules out a purely permissive role for activity. In strabismic animals, the levels of activity in both eyes are normal; it is only the patterns of activity that have been changed by misaligning the two eyes.

It has been proposed that the instructive role of activity could work through "Hebbian" synapses, that is, synapses that are strengthened when pre- and postsynaptic cells fire together and (by extension) weakened when they fire asynchronously (Hebb, 1949). Such synapses between LGN afferents and cortical cells could explain the monocular deprivation effect. Retinotopy in the geniculocortical projection is well established during the critical period, so neighboring LGN afferents giving input to a single cortical cell would tend to be driven together by visual stimuli in the open eye and would therefore tend to activate the postsynaptic cell, strengthening the synapse. Firing in the deprived eye in the absence of vision, on the other hand, would tend to be random, so neighboring LGN afferents giving input to a single cell would be unlikely to fire together. Therefore, the cortical cell would be less likely to reach threshold, weakening the synapse between an active axon and a silent cortical cell. Hebbian synapses could also explain the effects of strabismus, in which neighboring LGN axons from each eye are driven by visual stimuli to fire in synchrony, but the visual stimulation of the two eyes is different, so (unlike in normal vision) there is no synchrony of neighboring axons from the two eyes. In this situation, each eye's LGN axons will strengthen synapses with any cell where they had an initial advantage and weaken other synapses, resulting in a loss of binocularity in the cortex. Correlated activity in the two eyes driven by vision is important not only for maintaining binocularity during normal development, but also for the physiological and behavioral recovery from monocular deprivation. In kittens subjected to monocular deprivation and then allowed to recover normally with both eyes open, substantial recovery of vision through the deprived eye is seen. If, however, strabismus is induced following the monocular deprivation, recovery of vision in the deprived eye is much less complete (Kind et al., 2002)

One approach to testing whether this sort of patterning of activity might be involved in the development of ocular dominance columns is to produce artificial manipulations of the patterns of activity during development. Such manipulations are obviously difficult to perform, but one such experiment has been successfully completed. In this experiment, kittens were raised with all natural retinal activity pharmacologically blocked, but with artificial activity produced by stimulating electrodes placed in the optic nerves or optic chiasm (Stryker, 1986; Stryker and Strickland, 1984). If stimulation alternated between the two nerves such that the two eyes' inputs were never firing at the same time, then sharper than normal ocular dominance columns developed, similar to those seen in strabismic animals. If, on the other hand, the same amount of stimulation was performed at the chiasm, so that both eyes' inputs were always firing at the same time, then no ocular dominance columns were seen and LGN axons from the two eyes remained overlapping (or desegregated) in visual cortex. These results support a role for Hebbian synapses. When both eyes are equally but asynchronously active (as is also the case in strabismus), both eyes will maintain equal input to the cortex by strengthening synapses to any cells where they had an initial advantage, but binocularity will be lost. When the two eyes are equally and synchronously activated, however, synapses from both eyes onto all cortical cells, are strengthened, resulting in no segregation of the inputs from the two eyes.

Other experiments have provided more direct evidence for the existence of Hebbian synapses involved in the plasticity of ocular dominance columns. If an animal is monoc-ularly deprived, and its cortical cells are pharmacologically silenced without affecting the geniculocortical afferents, then the open eye actually become less effective than the deprived eye at driving cortical cells (Reiter and Stryker, 1988), and the open eye loses territory in the cortex (Hata and Stryker, 1994). Thus, the effects of monocular deprivation do not depend on the levels of activity in the two eyes, but rather on correlations between pre- and postsynaptic activity. If cortical cells are responsive, they will be better driven by the open eye, so their activity will be correlated with that of the open eye, and the open eye will "win." If, on the other hand, the cortical cells are silenced, their activity will be not be correlated with the open eye and so the open eye will "lose." Such Hebbian synapses have been directly demonstrated in another series of experiments in which the activity of LGN afferents was controlled by visual stimulation, and the activity of postsynaptic cortical cells was artificially controlled by current injections (Fregnac et al., 1988). In these experiments the ocular dominance of a cortical cell was determined, and then visual stimulation through one eye was paired with artificial firing of the cell while visual stimulation through the other eye was paired with artificial silencing of the cell. After a period of such pairings, the cell increased its response to visual stimulation through the eye whose stimulation had been paired with postsynaptic firing and decreased its response to stimulation through the other eye. This result shows that when pre- and postsynaptic activity of the cell was correlated, the synapse was strengthened, while when activity was uncorrelated, the synapse was weakened.

The results of such experiments, in which manipulating the patterns of activity by changing the correlations either between two sets of inputs (in alternating or synchronous stimulation experiments) or between inputs and target cells (in experiments manipulating postsynaptic activity) causes altered patterns of connectivity, clearly demonstrate that activity plays an instructive role in the plasticity of ocular dominance columns. It has been more difficult to rule out a purely permissive role for activity in the development and plasticity of eye-specific segregation in the LGN. Since the development of eye-specific segregation in the LGN occurs prior to the development of functional photoreceptors, it has not been possible to determine the effects of manipulations of visual experience like monocular deprivation or strabismus. Instead, until recently, manipulations of activity in experiments studying LGN development involved reducing levels of activity in one or both eyes (Penn et al., 1998; Shatz and Stryker, 1988; Sretavan et al., 1988). Therefore, the results of these manipulations could merely reflect a permissive role for activity; retinal ganglion cell axons in manipulated animals could have changed their behavior due to lack of a threshold level of activity necessary for them to interact with some molecular cue that would normally guide their connections. The recent experiment pharmacologically increasing levels of activity in one eye rules out such a purely permissive role for activity in LGN development (Stellwagen and Shatz, 2002). It remains to be determined, however, whether patterns of activity are truly instructing the development of eye-specific segregation in the LGN; so far, no experiments have directly addressed this question. It is interesting to note that while activity does not play a purely permissive role in the development of eye-specific segregation, it may play such a role in the development of normal lamination of retinal inputs to the LGN. Activity is clearly involved in the development of layers of retinal input to the LGN; blocking activity blocks the formation of layers (Penn et al., 1998; Shatz and Stryker, 1988; Sretavan et al., 1988). However, a number of experiments show that although normal retinal activity leads to normal segregation of retinal afferents to the LGN, it may not in all cases lead to normal lamination of those afferents. Such a pattern is seen in animals with coat color mutations which affect the number of retinal axons crossing at the chiasm; in such animals, retinal activity is normal and retinal axons from the two eyes to the LGN segregate, but they segregate into patches instead of into layers (Cucchiaro and Guillery, 1971; Guillery, 1969). A similar situation is seen in "rewired" ferrets in which retinal input is surgically rerouted to terminate in auditory thalamus; in these animals retinal activity is normal, and inputs from the two eyes are again segregated into patches instead of layers (Angelucci et al., 1997). Finally, segregation of retinal axons into patches instead of layers is also seen in animals in which the segregation of retinal axons in the LGN has been temporarily blocked during the time when retinal axons normally segregate by either a genetic (Muir-Robinson et al., 2002) or a pharmacological (Huberman et al., 2002) manipulation of activity, and then retinal activity is allowed to return to normal. These results suggest that although lamination of retinal axons in the LGN is activity-dependent, patterns of activity are not driving the layout of the layers. Instead there may be molecular cues in the LGN which are involved in layer formation, and whose proper function requires activity during a specific time period of development. Additional evidence for molecular cues involved in layer formation comes from the study of mutant dogs lacking an optic chiasm. In normal animals, ipsilater-ally projecting fibers come from the temporal (lateral) portion of the retina, while contralateral projections come from the nasal (medial) portion of the retina. In the achias-matic dogs, all retinal fibers from each eye remain ipsilateral. However, the LGN in these animals is relatively normal, anatomically, with well-developed laminae where temporal retinal fibers project to what would normally be the ipsilateral layer and nasal retinal fibers project to what would normally be the contralateral layer (Williams et al., 1994). This result is consistent with the possibility that molecular affinities between retinal axons from the two sides of the retina and LGN cells from the two laminae guide the development of LGN layers in both normal and achiasmatic animals.

0 0

Post a comment