Factors influencing the regeneration of axons in the central nervous system

Damage to the central nervous system (CNS) causes damage to neurons. This damage can result in the complete death of neurons, or in them becoming disconnected from their inputs or target structures due to disruption of axons. The main reason why damage to the human CNS is so disastrous and disabling is that axons will not in general regenerate in the mammalian brain, and neurons once lost are not replaced. In order, therefore, to repair the CNS, techniques will have to be developed to replace dead neurons, and induce axon regrowth. Central to the technologies necessary for brain repair is the ability to induce and control the growth of axons, since in a damaged brain both surviving and newly implanted neurons must grow axons to make or remake appropriate synaptic connections. Worthwhile treatments, however, do not necessarily require the repair of all the damaged circuits in the CNS, it may be possible to substantially improve the function of patients with relatively few reconnected axons, if those axons are ones which mediate particularly important behaviours, such as respiration, bladder control, or hand and arm movements.

Interactions between meningeal cells and astrocytes in vivo and in vitro

At the interface between the meninges and the central nervous system there is a characteristic structure known as the glia limitans, consisting of many fine interdigitating astrocyte processes which contain both GFAP and vimentin, and a basal lamina. A similar structure is set up after brain injury where meningeal cells invade the lesion. We have experimentally put astrocytes and meningeal cells in contact with one another, both in vivo and in vitro, to see whether this results in the formation of a glia limitans. Cultured meningeal cells were injected into the hippocampus of adult rats, and from 1 to 12 weeks later brains were stained were stained for GFAP and vimentin. One week after injection there was a widespread astrocytic reaction stretching up to 2 mm from the injection, the cells being stained intensely for both GFAP and vimentin. Over the next 4-6 weeks this widespread reaction subsided, the only remaining vimentin stained astrocytes, apart from those at the normal glia limitans, being in contact with the injected meningeal cells, or with meningeal cells which had migrated into the injection needle track. In vitro a structure reminiscent of the glia limitans formed where patches of astrocytes abutted meningeal cells; the astrocytes formed a layer of fine interdigitating processes all running parallel to the interface between the two cell types, and there was heavy staining for laminin and fibronectin. We conclude that a glia limitans forms wherever astrocytes and meningeal cells come into contact.

The effects of embryonic retinal neurons on neural crest cell differentiation into Schwann cells

Amongst the many cell types that differentiate from migratory neural crest cells are the Schwann cells of the peripheral nervous system. While it has been demonstrated that Schwann cells will not fully differentiate unless in contact with neurons, the factors that cause neural crest cells to enter the differentiative pathway that leads to Schwann cells are unknown. In a previous paper (Development 105: 251, 1989), we have demonstrated that a proportion of morphologically undifferentiated neural crest cells express the Schwann cell markers 217c and NGF receptor, and later, as they acquire the bipolar morphology typical of Schwann cells in culture, express S-100 and laminin. In the present study, we have grown axons from embryonic retina on neural crest cultures to see whether this has an effect on the differentiation of neural crest cells into Schwann cells. After 4 to 6 days of co-culture, many more cells had acquired bipolar morphology and S-100 staining than in controls with no retinal explant, and most of these cells were within 200 microns of an axon, though not necessarily in contact with axons. However, the number of cells expressing the earliest Schwann cell markers 217c and NGF receptor was not affected by the presence of axons. We conclude that axons produce a factor, which is probably diffusible, and which makes immature Schwann cells differentiate. The factor does not, however, influence the entry of neural crest cells into the earliest stages of the Schwann cell differentiative pathway.

The effects of protease inhibitors on axon growth through astrocytes

We have shown in a previous paper (Devl Biol. 135, 449, 1989) that axons regenerating from postnatal neurons are unable to penetrate three-dimensional cultures of mature astrocytes, while axons from embryonic dorsal root ganglia (DRGs) and retina will grow through such cultures for considerable distances. We have now investigated the role of proteases in the penetration of three-dimensional astrocyte cultures by axons from embryonic DRGs. Embryonic DRGs were grown in association with three-dimensional astrocyte cultures, with astrocyte monolayers, and with-air dried collagen. The effects of inhibitors of the three families of proteases that have been shown to be involved in tumour cell invasion were investigated. The serine protease inhibitors EACA and Trasylol both reduced growth in three-dimensional astrocyte cultures to around 50% of control, but had little effect on growth on astrocyte monolayers or on collagen. TIMP, which inhibits collagenases, had no effect on growth on two- or three-dimensional cultures. Cbz-gly-phen-amide, an inhibitor of enteroproteases, reduced growth in all three types of culture.

A comparison of the initial retinal ganglion cell projection to the contralateral superior colliculus in albino and pigmented rats

Newborn albino and pigmented rats received localised Fast blue (FB) injections into the most caudal part of the contralateral superior colliculus (SC). A proportion of retinal ganglion cells (RGC's) from the temporal retina which in the adult projects exclusively to the rostral half of the colliculus are labelled by injections to the caudal colliculus in neonatal animals. The majority of these cells die during the period of naturally occurring cell death in the retina, which occurs during the first 10 postnatal days. The object of this experiment was to see whether albino rats, which have well documented abnormalities in axon pathfinding in their visual system, had a larger number of cells in temporal retina which initially project to caudal colliculus than pigmented animals. On postnatal day 2 (P2), the ratio of temporal to nasal RGCs projecting to the caudal SC is greater in albino than pigmented rats (6.2% vs 2.65%). After the wave of naturally occurring cell death, at P14, when many of the neonatal errors have been eliminated, the ratio of temporal to nasal RGC's is reduced to 1.71% for albinos versus 1.53% for pigmented rats.

The growth of axons in three-dimensional astrocyte cultures

The environment of the adult central nervous system (CNS) does not support axon regeneration. We have been unable to replicate this behaviour using monolayer cultures of glia, so we have developed a technique for three dimensional culture of glial cells. We have examined the growth of axons from embryonic and postnatal retina and dorsal root ganglia (DRG's) through purified three-dimensional astrocyte cultures. Neither postnatal DRG's nor adult retina were able to grow axons through astrocytes from cultures 3 weeks or more old, although some DRG axons grew in astrocyte cultures which were 10 days or less old. However axons from embryonic DRG's and retina grew axons profusely into even elderly astrocyte cultures. All the tissues grew axons into three-dimensional Schwann cell cultures. The behaviour of axons in three-dimensional glial cultures therefore reproduces the behaviour of axons in vivo.

Rapid induction of the major embryonic alpha-tubulin mRNA, T alpha 1, during nerve regeneration in adult rats

The mRNAs for 2 isotypes of alpha-tubulin, termed T alpha 1 and T26, are differentially regulated in the developing rat nervous system. T alpha 1 alpha-tubulin mRNA is expressed at high levels when neurons extend processes whereas T26 mRNA is expressed constitutively (Miller et al., 1987b). We have examined the expression of these 2 alpha-tubulin mRNAs in regenerating facial and sciatic motor neurons of the rat using Northern blot and in situ hybridization analyses. T alpha 1 alpha-tubulin mRNA is rapidly induced in axotomized motor neurons of the facial nerve: increased levels of mRNA are detectable 4 hr after a lesion is made 1.5 cm distal to the neuronal cell bodies. T alpha 1 mRNA levels are highest from 3-7 d postcrush and decline slowly to control levels following functional reinnervation of facial muscles. In contrast, T26 mRNA levels remain constant throughout the regeneration process. Total alpha-tubulin mRNA levels do not change until 1 d postaxotomy; otherwise the changes in expression are similar to T alpha 1 mRNA, although the relative increase is not as great. Enhanced T alpha 1 alpha-tubulin mRNA expression also occurs in motor neurons of crushed or tied sciatic nerve. Ligature or crush of the sciatic nerve leads to approximately the same peak in the expression of T alpha 1 mRNA at 7-15 d postaxotomy. Following the facial nerve transection, under conditions in which reinnervation is prevented, T alpha 1 alpha-tubulin mRNA levels remain elevated significantly longer than when the nerve is crushed. Taken together, the data indicate that T alpha 1 alpha-tubulin mRNA is rapidly induced following neuronal axotomy, remains elevated during the period of axonal regrowth, and is subsequently down-regulated at the approximate time of target contact. These results are reminiscent of changes in T alpha 1 mRNA that occur during neuronal development. This growth-associated pattern of T alpha 1 gene expression can be modified by inhibiting appropriate regeneration of the damaged nerve.

Expression of Schwann cell markers by mammalian neural crest cells in vitro

During embryonic development, neural crest cells differentiate into a wide variety of cell types including Schwann cells of the peripheral nervous system. In order to establish when neural crest cells first start to express a Schwann cell phenotype immunocytochemical techniques were used to examine rat premigratory neural crest cell cultures for the presence of Schwann cell markers. Cultures were fixed for immunocytochemistry after culture periods ranging from 1 to 24 days. Neural crest cells were identified by their morphology and any neural tube cells remaining in the cultures were identified by their epithelial morphology and immunocytochemically. As early as 1 to 2 days in culture, approximately one third of the neural crest cells stained with m217c, a monoclonal antibody that appears to recognize the same antigen as rat neural antigen-1 (RAN-1). A similar proportion of cells were immunoreactive in cultures stained with 192-IgG, a monoclonal antibody that recognizes the rat nerve growth factor receptor. The number of immunoreactive cells increased with time in culture. After 16 days in culture, nests of cells, many of which had a bipolar morphology, were present in the area previously occupied by neural crest cells. The cells in the nests were often associated with neurons and were immunoreactive for m217c, 192-IgG and antibody to S-100 protein and laminin, indicating that the cells were Schwann cells. At all culture periods examined, neural crest cells did not express glial fibrillary acidic protein. These results demonstrate that cultured premigratory neural crest cells express early Schwann cell markers and that some of these cells differentiate into Schwann cells. These observations suggest that some neural crest cells in vivo may be committed to forming Schwann cells and will do so provided that they then proceed to encounter the correct environmental cues during embryonic development.

Oligodendrocytes repel axons and cause axonal growth cone collapse

We have examined the interactions between axons regenerating from dorsal root ganglia (DRGs) derived from newborn rats and oligodendrocytes cultured by three different techniques. Cultures examined after 2 days have a profuse outgrowth of axons from the DRGs, forming a dense mat on the culture surface. However, the axons avoid growing on oligodendrocytes; axons are seen all around these cells, but do not grow over them. We have also performed time-lapse video studies of the interactions between axonal growth cones and oligodendrocytes. Axons grow normally until their growth cone comes into direct contact with an oligodendrocyte, following which the growth cone remains motile for 30–60 min, but without making any progress over the cell. The growth cone then suddenly collapses, and the axon retracts, leaving a thin strand in contact with the cell. After this a new growth cone is usually elaborated and the process repeated. Oligodendrocytes are therefore inhibitory to axonal growth, and this may partially explain the failure of axons to regenerate in the mammalian central nervous system.

Formation of topographic maps

The catalogue of data presented here form many systems demonstrates that multiple mechanisms are involved in the formation of topographic maps. We are not yet in a position to explain why a particular mechanism appears to dominate in some situations and not in others. Certain generalizations can be made, however. First, at least some form of chemospecificity can be invoked to help explain connectivity in all of the experiments we have cited. Often, the differential identities of a population of neurons can be reflected in an orderly pattern of axon outgrowth and in the actively maintained preservation of neighbor relations as the axons grow toward their targets; such orderly arrangements are not obligatory, but, where present, they facilitate the speedy establishment of orderly maps when the axons reach their target nuclei. Within a terminal zone, chemospecific cues may dominate and constrain a given axon to terminate in a specific location, but axon-axon interactions commonly supercede chemospecific matching. At least two forms of axon-axon interaction occur, one based on some sort of biochemical properties related to the axon's embryological identity and another based on the axons' electrical activity. Tasks for the future are to identify the cellular bases of each of these mechanisms and to understand the situations in which each is manifested.

Topographic targeting errors in the retinocollicular projection and their elimination by selective ganglion cell death

In adult rats, as in other rodents, the retinocollicular projection is topographically organized in a very precise manner. Experiments involving the use of the retrogradely transported fluorescent dye fast blue as either a short- or long-term marker in neonatal rats indicate that the precision of this retinotopic projection does not arise ab initio, but rather is brought about by the preferential elimination of those ganglion cells whose axons project to topographically inappropriate regions of the colliculus. Such topographic targeting errors have been identified along both the rostrocaudal and mediolateral axes of the colliculus, and their elimination occurs during the period of naturally occurring ganglion cell death, which is completed by about postnatal day 10. When impulse activity in the retinal ganglion cell axons is blocked by repeated intraocular injections of the sodium channel-blocking agent tetrodotoxin (TTX) throughout the postnatal period of ganglion cell death, the preferential loss of the incorrectly projecting ganglion cells does not occur in the activity-blocked eye, although, as reported elsewhere, the overall loss of ganglion cells is comparable to that seen in normal animals. This supports the notion that the mechanism for selecting against incorrectly projecting ganglion cells is based on impulse activity among the competing ganglion cell axons. However, under activity-block conditions, the aberrantly projecting axons appear to retract from the caudal margin of the colliculus. The death of retinal ganglion cells during development thus seems to serve 2 purposes: It provides for the quantitative matching of the ganglion cell population to the needs of its central projection fields, and, at the same time, it serves to selectively eliminate those cells whose axons project to inappropriate targets or to inappropriate regions within the correct target fields.

The effect of intraocular tetrodotoxin on the postnatal reduction in the numbers of optic nerve axons in the rat

We have examined the effects of blocking retinal ganglion cell activity with the sodium channel blocker tetrodotoxin (TTX) on the postnatal reduction in the number of optic nerve axons (and, by interference, the degree of ganglion cell loss in the retina). TTX was injected every other day into the left eyes of a series of albino rats beginning on the day of birth, continuing through the 3rd, 7th, 12th or 14th days when the animals were killed and the optic nerves from both eyes were prepared for electron microscopy. The numbers of axons in the TTX treated and untreated optic nerves from the opposite side were determined from electron micrographs, and compared to the number seen in normal rats at the same ages. Both the magnitude and the time course of the reduction in the number of axons in the TTX-treated and untreated nerves were found to be similar to those seen in normal animals. However, there was a slight reduction in the loss of optic axons in the untreated nerves on the side opposite the TTX injections; this attenuation in axon loss could be mimicked by large systemic injections of TTX, and is probably attributable to a general systemic effect following repeated intraocular injections. These findings indicate that blocking ganglion cell activity with intraocular injections of TTX has little effect on the normal rate of axon loss from the optic nerve and on the numbers of ganglion cells that die during the first two weeks of postnatal life.

Muscle basal lamina: a new graft material for peripheral nerve repair

The suitability of muscle basal lamina as a graft material for the repair of peripheral nerves was investigated. Grafts were prepared by evacuating the myoplasm from muscles excised from rats and rabbits. This produced a material consisting mainly of basal lamina and connective tissue, with the basal lamina arranged as parallel tubes. Rat- and rabbit-derived graft material in 0.5-cm lengths was sutured into rat sciatic nerves, and 4-cm lengths of rabbit-derived graft material were interposed into rabbit sciatic nerves. For controls, 0.5-cm nerve autografts were grafted into rats and 4-cm autografts into rabbits. After 2 to 3 months, the success of the grafts was assessed functionally, electrophysiologically, and anatomically. By all these criteria the basal lamina grafts were as successful as nerve autografts; essentially the same number of axons of the same size grew through both graft types, animals recovered their limb function equally well, and the nerve conduction velocities and relative refractory periods were the same in both groups of animals. In rats, following both basal lamina and nerve autografts, the number of axons distal to the grafts was approximately the same as that proximal to them, but axon diameter and speed of conduction were significantly less than normal. The authors conclude that muscle basal lamina grafts are as effective as nerve autografts for repairing severed rat or rabbit peripheral nerves, and suggest that grafts prepared in this way may prove to be useful for nerve repair in humans.

The migration of neural crest cells and the growth of motor axons through the rostral half of the chick somite

We have studied the pathway of migration of neural crest cells through the somites of the developing chick embryo, using the monoclonal antibodies NC-1 and HNK-1 to stain them. Crest cells, as they migrate ventrally from the dorsal aspect of the neural tube, pass through the lateral part of the sclerotome, but only through that part of the sclerotome which lies in the rostral half of each somite. This migration pathway is almost identical to the path which presumptive motor axons take when they grow out from the neural tube shortly after the onset of neural crest migration. In order to see whether the ventral root axons are guided along this pathway by neural crest cells, we surgically excised the neural crest from a series of embryos, and examined the pattern of axon outgrowth approximately 24 h later. In somites which contained no neural crest cells, ventral root axons were still found only in the rostral half of the somite, although axonal growth was slightly delayed. These axons were surrounded by sheath cells, which had presumably migrated out of the neural tube, to a point about 50 micron proximal to the growth cones. With appropriate antibodies we found that the extracellular matrix components fibronectin and laminin are evenly distributed between the rostral and caudal halves of the somite. Neither of these molecules therefore plays a critical role in determining the specific pathway of neural crest cells or motor axons through the rostral half of the somite.

Factors guiding regenerating retinotectal fibres in the frog Xenopus laevis

I have examined the pathways of retinotectal fibres regenerating back to the contralateral tectum, and also to innervated and ‘virgin’ ipsilateral tecta in postmetamorphic Xenopus. The fibres were visualized by HRP labelling of either the whole optic nerve, or a selected quadrant of the retina.

Most fibres grow into either the ipsilateral or contralateral optic tract, although a small proportion go down the outside of the contralateral optic nerve. In the tracts, many fibres grow superficially, close beneath the pia, but a variable proportion runs more deeply. Axonal growth is not, therefore, restricted absolutely to the subpial region in the postmetamorphic Xenopus brain.

Fibres growing onto the contralateral, or a ‘virgin’ tectum mostly grow straight onto the rostral margin of the tectal lobe, without growing around its margin in the form of a medial or lateral brachium. Most of these fibres grow through the deeper part of the tectal layer which normally contains optic neuropil, but a proportion of them grow immediately deep to the pia. Fibres regenerating to an innervated ipsilateral tectum mostly enter either the medial or lateral brachium of the optic tract, and only leave this close to their site of termination. In the brachia the fibres run superficially under the pia, but when they leave the brachia they mostly run through the deeper retinorecipient layers. These observations provide further evidence that ingrowing optic fibres have their pathways influenced by the axons which have preceded them.

The migration of neural crest cells and the growth of motor axons through the rostral half of the chick somite

We have studied the pathway of migration of neural crest cells through the somites of the developing chick embryo, using the monoclonal antibodies NC-1 and HNK-1 to stain them. Crest cells, as they migrate ventrally from the dorsal aspect of the neural tube, pass through the lateral part of the sclerotome, but only through that part of the sclerotome which lies in the rostral half of each somite. This migration pathway is almost identical to the path which presumptive motor axons take when they grow out from the neural tube shortly after the onset of neural crest migration. In order to see whether the ventral root axons are guided along this pathway by neural crest cells, we surgically excised the neural crest from a series of embryos, and examined the pattern of axon outgrowth approximately 24 h later. In somites which contained no neural crest cells, ventral root axons were still found only in the rostral half of the somite, although axonal growth was slightly delayed. These axons were surrounded by sheath cells, which had presumably migrated out of the neural tube, to a point about 50 μm proximal to the growth cones. With appropriate antibodies we found that the extracellular matrix components fibronectin and laminin are evenly distributed between the rostral and caudal halves of the somite. Neither of these molecules therefore plays a critical role in determining the specific pathway of neural crest cells or motor axons through the rostral half of the somite.

Factors guiding regenerating retinotectal fibres in the frog Xenopus laevis

I have examined the pathways of retinotectal fibres regenerating back to the contralateral tectum, and also to innervated and 'virgin' ipsilateral tecta in postmetamorphic Xenopus. The fibres were visualized by HRP labelling of either the whole optic nerve, or a selected quadrant of the retina. Most fibres grow into either the ipsilateral or contralateral optic tract, although a small proportion go down the outside of the contralateral optic nerve. In the tracts, many fibres grow superficially, close beneath the pia, but a variable proportion runs more deeply. Axonal growth is not, therefore, restricted absolutely to the subpial region in the postmetamorphic Xenopus brain. Fibres growing onto the contralateral, or a 'virgin' tectum mostly grow straight onto the rostral margin of the tectal lobe, without growing around its margin in the form of a medial or lateral brachium. Most of these fibres grow through the deeper part of the tectal layer which normally contains optic neuropil, but a proportion of them grow immediately deep to the pia. Fibres regenerating to an innervated ipsilateral tectum mostly enter either the medial or lateral brachium of the optic tract, and only leave this close to their site of termination. In the brachia the fibres run superficially under the pia, but when they leave the brachia they mostly run through the deeper retinorecipient layers. These observations provide further evidence that ingrowing optic fibres have their pathways influenced by the axons which have preceded them.

On the formation of eye dominance stripes and patches in the doubly-innervated optic tectum of the chick

By surgically dividing the region of the presumptive optic chiasm in chick embryos on the third day of incubation (around stage 15), we have been able to induce substantial numbers of optic nerve fibers to grow aberrantly into the ipsilateral optic tract. As a result, many of the visual centers that are normally innervated only by fibers from the contralateral retina received fibers from both eyes. The proportion of fibers going to each tectal lobe varied from case to case, but in about one-third of the animals the tectal lobes received approximately equal numbers of fibers from each eye. In animals that survived until embryonic days 17-19 (which is beyond the period of retinal ganglion cell death) labeling of the two eyes with WGA-HRP and [3H]proline respectively, revealed a pattern of sharply defined eye dominance stripes or patches in the stratum griseum et fibrosum superficiale (SGFS) of the optic tectum, and in the ventral lateral geniculate nucleus. Less clearly segregated eye dominance zones were seen in the ectomammillary nucleus and the nucleus externus. The size and distribution of the stripes varied depending on the number of fibers projecting from each eye to a given tectal lobe; the minimum size was about 75 micron, while the maximum was large enough to occupy almost the entire tectal lobe. In animals in which the tectal input from the two eyes was roughly equal, the stripes varied in width between 75 micron and about one-third of the surface of the tectal lobe. The orientation of the stripes was consistently orthogonal to the direction of fiber ingrowth from the optic tract. From the earliest stages of optic fiber ingrowth, the fibers from the two eyes are completely intermixed in the stratum opticum (SO). However, on embryonic day 12, shortly after they have begun to penetrate into the SGFS, they are already segregated into stripes, although the stripe borders are very fuzzy. This suggests that the fibers from the two eyes may overlap at this stage. The phase of stripe formation coincides with that of naturally occurring retinal ganglion cell death, and we suggest that the two processes are interlinked.

The use of backscattered electrons to examine selectively stained nerve fibers in the scanning electron microscope

Selective staining of neuronal tissues using standard light microscopic techniques has been combined with backscattered electron scanning electron microscopy. This technique allows neurons to be readily distinguished from their surrounding tissues and examined at high resolution. The technique overcomes some of the problems involved in scanning electron microscopy of nervous tissue in situ.

Fibre order in the normal Xenopus optic tract, near the chiasma

In juvenile Xenopus retinotopic fibre order in the optic tract near the chiasma was investigated by labelling small groups of optic fibres from peripheral retina with HRP. This selective fibre labelling with HRP was combined with autoradiography following administration of tritiated thymidine to the eye, so that the HRP-labelled fibres could be located within the borders of the optic tract. Fibres arising from the periphery of all four retinal quadrants were superficially located in the optic tract near the chiasma, with dorsal retinal fibres showing the greatest tendency to travel deep in the diencephalon. Retinal lesions closer to the optic nerve head labelled fibres which ran deeper in the optic tract. Near the chiasma, fibres from ventral retina tended to group rostrally while fibres from dorsal retina tended to group caudally. However, no obvious localization of fibres arising in temporal or nasal retina was seen in the lower optic tract.

Fibre order in the normal Xenopus optic tract, near the chiasma

In juvenile Xenopus retinotopic fibre order in the optic tract near the chiasma was investigated by labelling small groups of optic fibres from peripheral retina with HRP. This selective fibre labelling with HRP was combined with autoradiography following administration of tritiated thymidine to the eye, so that the HRP-labelled fibres could be located within the borders of the optic tract.

Fibres arising from the periphery of all four retinal quadrants were superficially located in the optic tract near the chiasma, with dorsal retinal fibres showing the greatest tendency to travel deep in the diencephalon. Retinal lesions closer to the optic nerve head labelled fibres which ran deeper in the optic tract. Near the chiasma, fibres from ventral retina tended to group rostrally while fibres from dorsal retina tended to group caudally. However, no obvious localization of fibres arising in temporal or nasal retina was seen in the lower optic tract.

Regressive events in neurogenesis

The development of most regions of the vertebrate nervous system includes a distinct phase of neuronal degeneration during which a substantial proportion of the neurons initially generated die. This degeneration primarily adjusts the magnitude of each neuronal population to the size or functional needs of its projection field, but in the process it seems also to eliminate many neurons whose axons have grown to either the wrong target or an inappropriate region within the target area. In addition, many connections that are initially formed are later eliminated without the death of the parent cell. In most cases such process elimination results in the removal of terminal axonal branches and hence serves as a mechanism to "fine-tune" neuronal wiring. However, there are now also several examples of the large-scale elimination of early-formed pathways as a result of the selective degeneration of long axon collaterals. Thus, far from being relatively minor aspects of neural development, these regressive phenomena are now recognized as playing a major role in determining the form of the mature nervous system.