Two stages in the development of a mammalian retinocollicular projection

Overexpression of Ten-m3 in the retina alters ipsilateral retinocollicular projections in the wallaby (Macropus eugenii)

Retinal projections to the superior colliculus are organised into retinotopic maps. Binocular vision requires that inputs from the two eyes map in register with each other. Studies in mice lacking Ten-m3, a homophilic transmembrane protein, indicate that it plays a key role in this process by influencing ipsilateral projections. The postnatal, ex utero development of the wallaby allows the targeted manipulation of molecules of interest during development. The distribution of mRNA for Ten-m3 in the retina and superior colliculus of the wallaby, and the effects of its spatiotemporally restricted retinal overexpression was investigated, in particular on the mapping of ipsilateral projections. Quantitative polymerase chain reaction found that Ten-m3 mRNA is expressed at relatively higher levels in the retina and colliculus early in development. Further, it is higher in ventral than dorsal retina, and increased in the retinotopically corresponding medial compared to lateral superior colliculus. In situ hybridisation demonstrated an increasing dorsoventral gradient in retinal ganglion cells was matched to an increasing lateromedial gradient in the superior colliculus. Overexpression of Ten-m3 by in vivo retinal electroporation produced an increase in ipsilateral projections to the binocular rostromedial colliculus, fitting with the proposal that Ten-m3 mediates mapping by attractant homophilic interactions. Retrograde labelling of the projection from this region suggested that overexpression produces a shift in the axons of existing ipsilaterally projecting ganglion cells rather than a rerouting of the axons of contralaterally projecting cells. Retinal manipulation of Ten-m3 levels produces changes in ipsilateral mapping, supporting a role for it in binocular mapping.

Regenerating optic axons restore topography after incomplete optic nerve injury

Following complete optic nerve injury in a lizard, Ctenophorus ornatus, retinal ganglion cell (RGC) axons regenerate but fail to restore retinotectal topography unless animals are trained on a visual task (Beazley et al. [ 1997] J Comp Neurol 370:105-120, [2003] J Neurotrauma 20:1263-1270). Here we show that incomplete injury, which leaves some RGC axons intact, restores normal topography. Strict RGC axon topography allowed us to preserve RGC axons on one side of the nerve (projecting to medial tectum) while lesioning those on the other side (projecting to lateral tectum). Topography and response properties for both RGC axon populations were assessed electrophysiologically. The majority of intact RGC axons retained appropriate topography in medial tectum and had normal, consistently brisk, reliable responses. Regenerate RGC axons fell into two classes: those that projected topographically to lateral tectum with responses that tended to habituate and those that lacked topography, responded weakly, and habituated rapidly. Axon tracing by localized retinal application of carbocyanine dyes supported the electrophysiological data. RGC soma counts were normal in both intact and axotomized RGC populations, contrasting with the 30% RGC loss after complete injury. Unlike incomplete optic nerve injury in mammals, where RGC axon regeneration fails and secondary cell death removes many intact RGC somata, lizards experience a "win-win" situation: intact RGC axons favorably influence the functional outcome for regenerating ones and RGCs do not succumb to either primary or secondary cell death.

Retinocollicular synaptogenesis and synaptic transmission during formation of the visual map in the superior colliculus of the wallaby (Macropus eugenii)

Spontaneous retinal activity has been implicated in the development of the topographic map in the superior colliculus (SC) but a direct demonstration that it reaches the colliculus is lacking. Here we investigate when the retinocollicular projection is capable of transmitting information from the retina in a marsupial mammal, the wallaby (Macropus eugenii). The projection develops postnatally, allowing in vivo analysis throughout development. Quantification of retinocollicular synaptogenesis has been combined with electrophysiology of the development and characteristics of retinocollicular transmission, including in vivo and in vitro recording in the same animals. Prior to postnatal day (P) 12-14 in vitro recording detected only presynaptic activity in retinal axons in the colliculus, in response to stimulation of the optic nerve. Postsynaptic responses, comprising both N-methyl-d-aspartate (NMDA) and non-NMDA responses, were first detected in vitro at P12-14 and retinal synapses were identified. In contrast, postsynaptic responses to optic nerve stimulation could not be detected in vivo until P39, around the time that retinal axons begin arborizing. Around this age density and numbers of total synapses began increasing in the retinorecipient layers of the colliculus. By P55-64, the numbers of retinal synapses had increased significantly and density and numbers of retinal and total synapses continued to increase up to P94-99. During this time the map is undergoing refinement and degenerating axons and synapses were present. The discrepancy between in vitro and in vivo onset of functional connections raises the question of when retinal activity reaches collicular cells in the intact, unanaesthetized animal and this will require investigation.

Brain-derived neurotrophic factor is expressed in a gradient in the superior colliculus during development of the retinocollicular projection

Abstract Theoretical models of topographic map formation have postulated a gradient of attractant in addition to a gradient of repulsion in the target. In species where many axons grow past their correct positions initially, it has also been argued that a parallel gradient of attractant or branching signal is required to ensure collateral formation at the correct position (O'Leary et al., 1999). Brain-derived neurotrophic factor (BDNF) is a known attractant and promotes branching of retinal axons. We have examined its distribution in the superior colliculus and that of its receptor, trkB, in the retina, using immunohistochemistry and in situ hybridization, respectively, during the development of the topographic retinocollicular projection in the wallaby, a marsupial mammal. The number of glial endfeet expressing BDNF at the surface of the colliculus was found to be in a high caudal-to-low rostral gradient during the time when the retinocollicular projection was developing. When the projection was mature the rostrocaudal gradient had disappeared and the number of detectable endfeet expressing BDNF was very low. Messenger RNA for TrkB was expressed in the retinal ganglion cell layer throughout the time when the retinocollicular projection was developing, with no difference in expression across the nasotemporal axis of the retina. The low rostral to high caudal distribution of BDNF in glial endfeet supports the idea that it is providing a parallel gradient of attractant or branching signal in the colliculus.

Analysis of EphB receptors and their ligands in the developing retinocollicular system of the wallaby reveals dynamic patterns of expression in the retina

The expression of EphB1 and B2 receptors and ephrins-B1, -B2 and -B3 in the retina and superior colliculus of the wallaby (Macropus eugenii) was examined during the development of the retinocollicular projection, using reverse transcription-polymerase chain reaction and immunohistochemistry. There was an early transient differential expression of EphB2 that was higher in ventral retina and restricted to the outer neuroblast layer, whereas a high ventral to low dorsal gradient of ephrin-B2 expression occurred there throughout the study period. However, there was no dorsoventral gradient of receptors or ligands in retinal ganglion cells or a mediolateral gradient of ephrins in the colliculus. These findings suggest a limited role for these molecules in topographic mapping across the mediolateral colliculus in the wallaby. Early in retinal development there is a complementary pattern of expression of ephrin-B1 and -B2 in the outer neuroblast layer that overlaps with expression of EphB2. Ganglion and amacrine cells also express EphB2. As development proceeds subpopulations of putative horizontal and bipolar cells, also expressing EphB2, come to reside in the inner nuclear layer and ephrin-B1 is expressed throughout the outer nuclear layer. At the same time cells expressing ephrin-B2, and subpopulations of horizontal and bipolar cells come to reside in the inner nuclear layer and there is a corresponding decrease in ephrin-B2 expression in the outer nuclear layer. This pattern of coexpression of receptors and ligands suggests a role for them in cell migration and maintenance of laminar boundaries.

Macrophages contribute to the maintenance of stable regenerating neurites following peripheral nerve injury

Normal adult uninjured nerve is unable to support axonal regeneration. We have studied the mechanisms underlying the regeneration of peripheral nerve by culturing adult mouse dorsal root ganglia (DRG) explants on unfixed, longitudinal cryosections of either the uninjured sciatic nerve or the distal segment of the transected sciatic nerve. We found that, initially, DRG grew vigorously on cryosections of both uninjured and postinjury sciatic nerves. However, the neurites began to degenerate shortly after contact with the uninjured nerve, whereas those growing on postinjury nerve substrate remained healthy for up to 9 days in culture. This ability to support stable outgrowth peaked at 8 days, gradually decreased by 10 days, and disappeared by 20 days after injury. Macrophages appeared in the distal segment by 4 days postinjury and had infiltrated its entire length by 8 days. Uninjured nerve cryosections could be rendered supportive of stable outgrowth by preincubation with macrophage-conditioned medium or by brief trypsinization. The activity of the macrophage-conditioned medium was augmented upon activation of macrophages. Together these findings suggest that the environment of the sciatic nerve undergoes a transformation during Wallerian degeneration such that it becomes transiently supportive of the stable outgrowth of neurites. This transformation may be mediated by a proteolytic activity, generated by activated macrophages, that removes a putative "degeneration signal" protein normally present in the adult nerve and thus contributes to the maintenance of stable regenerating neurites.

Developmental expression and subcellular localization of wallaby gonadotropin-releasing hormone receptor and its splice variants

The developmental expression of gonadotropin-releasing hormone receptor (GnRH-R) and its splice variants was examined in the gonads of tammar wallaby pouch young in order to elucidate the functional role of GnRH-R in the developing testis and ovary. Wallaby GnRH-R, like eutherian GnRH-Rs, contains three exons and two introns. In the present study, the transcripts of two splice variants (GnRH-R Delta 1 and GnRH-R Delta 2) were cloned from the pituitary. GnRH-R Delta 1 contained a 291 bp deletion from nucleotide positions 232 to 522 within exon 1. This transcript appears to be distinctive in the wallaby and has not been reported in other species. GnRH-R Delta 2 contained a 220 bp deletion from nucleotide positions 523 to 742, corresponding to exon 2. We examined the subcellular localization of the wild type GnRH-R and its splice variants with confocal microscopy, showing that both the wild type receptor and the splice variants were membrane-associated molecules. The different pattern of expression of the wild type receptor and the variants transcripts found in adult and neonatal tissues suggests a specific developmental regulation of the GnRH-R Delta 2 transcript. In addition, the developmental expression of the GnRH-R and GnRH-R Delta 1 transcripts showed a possible association with key physiological events during gonadal development in the wallaby pouch young, suggesting that GnRH-R may be involved in the regulation of early development in the testis and ovary.

Differentiation and morphological integration of neural progenitor cells transplanted into the developing mammalian eye

Transplantation of neural stem/progenitor cells has been proposed as a novel approach for the replacement and repair of damaged CNS tissues. We have evaluated the influence of the host cellular microenvironment upon the survival, differentiation, and integration of neural progenitor cells transplanted into the CNS. Using this approach, we have investigated the fate of neural progenitor cells in vivo following transplantation into the developing mammalian eye. Murine brain progenitor cells (mBPCs) isolated from neonatal mice expressing the green fluorescent protein (GFP) transgene were transplanted into the eyes of Brazilian opossums (Monodelphis domestica). Monodelphis pups are born in an extremely immature, fetal-like state. The eyes of neonatal pups provide a fetal-like environment in which to study cellular interactions between host tissues and transplanted neural progenitor cells. mBPCs were transplanted by intraocular injection in hosts ranging in age from 5 days postnatal to adult. The transplanted cells were easily identified because of their GFP fluorescence. Extensive survival, differentiation, and morphological integration of mBPCs within the host tissue was observed. We found that the younger retinas provided a more supportive environment for the morphological integration of the transplanted mBPCs. Cells with morphologies characteristic of specific retinal cell types were observed. Moreover, some transplanted mBPCs were labeled with antibodies characteristic of specific neural/retinal phenotypes. These results suggest that the host environment strongly influences progenitor cell differentiation and that transplantation of neural progenitor cells may be a useful approach aimed at treating degeneration and pathology of the CNS.

Differential localization of SNARE complex proteins SNAP-25, syntaxin, and VAMP during development of the mammalian retina

SNARE complex proteins have critical functions during regulated vesicular release of neurotransmitter. In addition, they play critical roles during neurite outgrowth and synaptogenesis. Although it is clear that the function of any one SNARE complex protein during release of neurotransmitter is dependent on its association with other members of the complex, it is less certain whether their function during development and differentiation is dependent on interaction with one another. Previously, we have observed transient high levels of SNARE complex protein SNAP-25 in developing cholinergic amacrine cells (West Greenlee et al. [1998] J Comp Neurol 394:374-385). In addition, we detected, high levels of SNAP-25 in developing and mature photoreceptors. To better understand the functional significance of these high levels of SNAP-25 expression, we used immunocytochemistry to examine the developmental expression of the three members of the SNARE complex, SNAP-25, Syntaxin, and vesicle associated membrane protein (VAMP/also Synaptobrevin). Our results demonstrate that the high levels of SNAP-25 in cholinergic amacrine cells and photoreceptors are not accompanied by the same relatively high levels of other SNARE complex proteins. These results suggest that high levels of SNAP-25 in specific cell types may function independently of association with Syntaxin and VAMP. In this analysis, we characterized the changing patterns of immunoreactivity for the three SNARE complex proteins during the development and differentiation of the mammalian retina. We have compared the pattern of expression of the core SNARE complex proteins in the Brazilian opossum, Monodelphis domestica, and in the rat and found common patterns of expression between these diverse mammalian species. We observed temporal differences in the onset of immunoreactivity between these three proteins, and differences in their localization within synaptic layers in the developing and mature mammalian retina. This study is the first to characterize the changing expression patterns of the three SNARE complex proteins in the developing central nervous system. The differential distribution of SNAP-25, Syntaxin, and VAMP may indicate additional roles for these proteins during vesicle trafficking events, which are independent of their association with one another.

Differentiation of the chick retinotectal topographic map by remodeling in specificity and refinement in accuracy

To understand the development of the retinotopic map, differentiation of the topographic map was quantitatively examined in the chick. Labeling the retinal ganglion cell (RGC) axons anterogradely with the local injections of DiI revealed the relative anteroposterior positions of their growth cones (GCs) on the tecta as a function of the nasotemporal positions of the injected sites in the retinae, which allowed a graphic representation of the map. The topographic map was depicted by combination of two parameters: specificity which indicates strictness of the topographic relationship between locations of the RGC bodies and their GCs on the tectum, and accuracy which indicates an extent of the GC displacement on the tectum. A crude projection with low specificity emerged at embryonic day 11 (E11). The initial crude projection was remodeled into the inaccurate map with high specificity by E13; thereafter, it was refined to the accurate map with higher specificity by E15. The results suggest that the elements of the guidance mechanism operate stage by stage through the formation of the crude projection, the remodeling in specificity, and the refinement in accuracy to establish the final topographic map.

Marsupial retinocollicular system shows differential expression of messenger RNA encoding EphA receptors and their ligands during development

The protracted development of the wallaby (Macropus eugenii) has allowed study of messenger RNAs encoding Eph receptors EphA3 and EphA7 and ligands ephrin-A2 and -A5 in the retina and superior colliculus at intervals throughout the development of the retinocollicular projection: from birth, before retinal innervation, to postnatal day 95, when the projection is mature. Reverse transcription-polymerase chain reaction showed messenger RNAs for both receptors and ligands were expressed at all ages. EphA7 was expressed more highly in the rostral superior colliculus. Ephrin-A2 and -A5 were expressed more highly in the caudal colliculus. EphA3 was expressed in a complementary manner, more highly in temporal than in nasal retina. There are higher levels of expression of the ligands when the projection is only coarsely topographically organised. This suggests a role for them and their receptor EphA3 in this stage, by repulsive interactions which restrict temporal axons to rostral superior colliculus. This is the first account in a marsupial mammal of the appearance of this molecular family, substantiating its ubiquitous role in topographically organised neuronal connections. Nevertheless, expression is not the same as in the mouse, suggesting differences in the details of topographic coding between species.

Differential withdrawal of retinal axons induced by a secreted factor

To understand the development of the topographic map in the chick retinotectal projection, we studied the long-term interactions between retinal axons and tectal cell processes using a novel coculture system, the ryomen chamber. Both nasal and temporal retinal axons initially grew equally well on a substrate consisting of posterior tectal cell processes; however, subsequently most temporal axons withdrew from this surface, whereas most nasal axons did not. Experiments using conditioned media indicate that posterior tectal cells induced withdrawal of the temporal axons by secreting a soluble factor. This withdrawal seems to be distinct from the immediate repulsive effect of ephrin-A2 (ELF-1) and ephrin-A5 (RAGS) seen in the stripe assay because (1) the withdrawal-inducing factor was diffusible, whereas ephrin-A2 and -A5 are membrane-bound, and (2) the withdrawal-inducing factor appeared later in development than ephrin-A2 and -A5. Furthermore, sensitivity to the withdrawal-inducing factor decreased continuously from the temporal to nasal retina. These results suggest that target cell-induced axonal withdrawal may be involved during a late stage of the development of the retinotectal map.

Onset of optic nerve conduction and synaptic potentials in superior colliculus of fetal rats studied in vitro

This article describes the onset of electrical excitability and synaptic transmission in the retinocollicular pathway of the fetal and early postnatal rat, utilizing a novel in vitro preparation. Although the optic nerve is visible in embryonic day (E) 14 brain, its stimulation produced no response in the superior colliculus (SC) until E16 when a low voltage simple negative wave was evoked. At E17 these potentials were blocked rapidly, completely, and reversibly when choline was substituted for sodium or with the addition of cobalt ions. In the course of establishing the block with either of the above agents the latency of response increased, indicating an action on axonal transmission. By E20 the collicular evoked potential showed a short followed by a longer latency wave. The latter was blocked by the glutamate antagonist kynurenic acid, with latency unaffected. Further examination of potentials with the addition of glutamatergic receptor subtype blockers aminophosphonopentanoic acid (APV) and 6-cyano-7-nitroquinoxaline-2,3-dione/6,7-dinitroquinoxaline- 2,3-dione (CNQX/DNQX) showed a clear abolition of the elicited potentials by E20 and older. Thus, fetal rat optic nerve fibers are capable of conduction in response to electrical stimulation as soon as they reach the SC at E16. Both sodium and calcium are involved. GABA-mediated modulation of axonal conduction is evident by E18. Glutaminergic synaptic transmission is established by E20. The timetable of fetal onset of capability to conduct and support synaptic transmission in the retinocollicular pathway is earlier than had previously been reported in vivo in the rat in which the superior colliculus neurones are said not to be driven by the optic nerve until 6 days post natal. This has relevance to the possible role of impulse activity in development of the pathway.

Conduction and synaptic transmission in the optic nerve and the superior colliculus during development of the retinocollicular projection in the wallaby (Macropus eugenii)

When do the developing connections between mammalian retinal ganglion cells and the superior colliculus become functional? Evoked potentials elicited by optic nerve stimulation in the pouch young of the wallaby were used to answer the question. Up to 42 days after birth, the evoked potentials in the colliculus appeared to be generated by axon conduction. Synaptic activity was first recorded from the rostral colliculus at 45 days, and was found to be progressively more caudal, spreading to cover the colliculus, by 65 days. From the earliest indication of synaptic activity until eye opening at 140 days, current source density (CSD) analysis consistently showed the same basic pattern: an initial deep sink from synaptic activity of fast (Y type) fibres, and a more superficial longer-latency sink from slower (W type) fibres. All features became more clearly delineated with age. The indirect retinocorticocollicular connection appeared between 134 days and 146 days. The ability of optic nerve fibres to sustain action potentials precedes their formation of functional synapses with collicular neurons, which happens abruptly at three months before eye opening. CSD analysis showed that the relationship between the conduction velocity of optic nerve fibres and their depth of termination is evident from the first signs of synapse formation.

Spatiotemporal pattern of ontogenetic expression of calbindin-28/kD in the retinorecipient layers of rat superior colliculus

Using an antibody against calbindin-28kD, we have studied the spatial pattern of expression of this protein in the superior colliculi (SC) of four strains of mature laboratory rats. In all four strains, calbindin-expressing cells (CECs) formed horizontally oriented tiers in the retinorecipient and intermediate gray layers but were diffusely distributed throughout the deep layers. Ontogenetically, calbindin-28kD was expressed for the first time in the retinorecipient layers at postconceptional day 20 (PCD 20), by cells located in the rostrolateral region where the first born retinal ganglion cells (RGCs) are represented. Although on the day of birth (PCD 22/23), the CECs were distributed more widely, they were still absent in the most medial part of the SC, that is, the region where the latest born RGCs are represented. The spatial distribution of CECs became adultlike only by PCD 29, that is, at the end of the period of the naturally occurring death of the RGCs. Monocular eye enucleations on PCD 23 prevented the expression of calbindin in the medial fifth of the retinorecipient layers of the contralateral SC, while the unilateral removal of the visual cortices had no discernable effect on the numbers and distribution of the CECs in either SC. Thus, the spatiotemporal pattern of ontogenetic expression of calbindin-28kD in the retinorecipient layers of SC reflects the spatiotemporal pattern of generation of the RGCs, and the retinal input appears to induce neuronal expression of calbindin-28kD in these layers.

The initial stages of development of the retinocollicular projection in the wallaby (Macropus eugenii): distribution of ganglion cells in the retina and their axons in the superior colliculus

The time course of ingrowth of retinal projections to the superior colliculus in the marsupial mammal, the wallaby (Macropus eugenii), was determined by anterograde labelling of axons from the eye with horseradish peroxidase, from birth to 46 days, when axons cover the colliculus contralaterally and ipsilaterally. The position of retinal ganglion cells giving rise to these projections over this period was determined in fixed tissue by retrograde labelling from the colliculus with a carbocyanine dye. Axons first reach the rostrolateral contralateral colliculus 4 days after birth and extend caudally and medially, reaching the caudal pole at 18 days and the far caudomedial pole at 46 days. The first contralaterally projecting cells are in the central dorsal and temporal retina, followed by cells in the nasal and finally the ventral retina. They are distributed closer to the periphery with increasing age. The first sign of a visual streak appears by 18 days. Axons reach the ipsilateral colliculus a day later than contralateral axons and come from a similar region of the retina. The sparser ipsilateral projection reaches the caudal and medial collicular margins by 46 days but by 16-18 days, ganglion cells giving rise to this transient projection are already concentrated in the temporoventral retina. The orderly recruitment of ganglion cells from retinotopically appropriate regions of the retina as axons advance across the contralateral colliculus suggests that the projection is topographically ordered from the beginning. The ipsilateral projection is less ordered as cells are located in the temporoventral crescent at a time when their axons are still transiently covering the colliculus prior to becoming restricted to the rostral colliculus. Features of mature retinal topography such as the visual streak and the location of ipsilaterally projecting cells begin to be established very early in development, before the period of ganglion cell loss and long before eye opening at 140 days.

The effect of dark-rearing, strobe-rearing and acute visual cortex removal on the visual responses in the superficial superior colliculus of the guinea-pig

Extracellular multi-unit responses to visual stimuli were recorded in the cells of the superficial layers of the superior colliculus (SC) in four groups of adult guinea-pigs: a control group, a strobe-reared group, a dark-reared group and a group with the ipsilateral visual cortex removed acutely. Single unit visual responses were also recorded in a control and a dark-reared group. When guinea-pigs were either strobe or dark-reared from birth, the number of directionally selective responses in the superficial SC decreased significantly. Acute removal of the visual cortex had no affect on the number of directionally selective cells recorded in the SC. The correlation between azimuthal visual receptive field and rostrocaudal position of the recording electrode in the SC was not significantly different from the control group following strobe, dark-rearing or acute visual cortex removal. These data imply that, during early development, visual information is necessary for directional selectivity of the visual responses in the superficial SC. However, the map of visual azimuthal space is essentially unperturbed by visual restriction (in the form of dark or strobe-rearing) or acute visual cortex removal.

Absence of topography in precociously innervated tecta

The retinotectal map in Xenopus forms very early: retinal axons are topographically ordered along the dorsoventral axis of the tectum by stage 39, as they first arrive. To test whether topographic cues are present even earlier, we forced retinal axons to innervate the tectum prematurely by transplanting stage 28 eye primordia into stage 20 hosts, then assayed dorsoventral topography using focal injections of lipophilic dye into dorsal and ventral retina at donor stages 39–40. Unoperated and isochronic control projections showed normal dorsoventral ordering both in the optic tract and in the tectum. In contrast, projections from heterochronically transplanted eyes were ordered in the tract, but spread out upon entering the tectum and did not show significant dorsoventral ordering. Individual axons entering the tectum precociously often made abnormally abrupt and topographically incorrect turns. Thus, the topographical cues normally expressed in the tectum at stage 39 appear to be absent a few hours earlier. However, this lack of cues is only temporary, since heterochronic transplants allowed to survive to donor stages 45–46 showed normal topography. The absence of tectal topography at a stage when retinal axons can navigate to the young tectum strongly suggests that the molecules that provide tectal topographical signals are distinct from those used for pathfinding in the diencephalon and target recognition at the tectum.

Wiring up the visual system

1. In this review we describe some of our recent studies on the developing marsupial visual pathway. The description focuses on retinal ganglion cells, considering the formation of their dendritic trees, the outgrowth of axons and the formation of connections within the brain. 2. Both dendritic trees and outgrowing axons undergo a period of exuberance, followed by one of refinement. The dendritic tree transiently develops a more complex branching pattern than is found in adults. Short side branches, referred to as spines, are a feature of immature dendrites and, to a lesser extent, of axons. These structures are mostly lost as development proceeds. However, they are retained on the dendritic trees of small-field ganglion cells and, for a proportion of axons, on that part within the nerve fibre layer of the retina. Although most axons navigate fairly direct routes towards their targets, a minority follow inappropriate courses, such as doubling back towards the eye or entering the opposite optic nerve at the chiasm. As such errant axons are not seen in the adult, we assume that their parent cell bodies die during development. 3. Throughout development, optic axons are arranged in an approximate retinotopic order along the length of the visual pathway; as a result, axons approach the visual centres aligned to form, at least, a crude retinotopic map. Axons from dorsal and ventral retina exchange locations along the optic nerve and in this way correct for the inversion of the image brought about by the lens.(ABSTRACT TRUNCATED AT 250 WORDS)