Self-organizing Neural Maps: The Retinotectal Map and Mechanisms of Neural Development: From Retina to Tectum

By John T. Schmidt

Self-organizing Neural Maps: From Retina to Tectum describes the underlying processes that determine how retinal fibers self-organize into an orderly visual map. The formation of neural maps is a fundamental organizing concept in neurodevelopment that can shed light on developmental mechanisms and the functions of genes elsewhere. The book presents a summary of research in the retinotectal field with an ultimate goal of synthesizing how underlying mechanisms in neural development harmoniously come together to create life. A broad spectrum of neuroscientists and biomedical scientists with differing backgrounds and varied expertise will find this book useful.

• Describes the mechanisms relating to the developmental wiring of the retinotectal system
• Brings together the state-of-the-art research in axon guidance and neuronal activity mechanisms in map formation
• Focuses on topographical maps and inclusion of multiple animal models, from fish to mammals
• Explores the molecular guidance and activity dependent cue components involved in neurodevelopment

• Language: English
• Publisher: Academic Press
• Release date: Oct 15, 2019
• ISBN: 9780128185803

John T. Schmidt

Dr. Schmidt is Professor Emeritus at the University of Albany (SUNY) in the Department of Biological Sciences. He has worked in the retinotectal area since the early 1970s and published more than 60 articles and chapters. For the last four decades, he was Professor of Biological Sciences at SUNY-Albany, serving for 25 years as the Director of the Center for Neuroscience Research. In the 1990s, he together with Professor Jonathan Wolpaw organized an international conference on the wider subject and edited a volume of the proceedings for the Annals of the New York Academy of Sciences, entitled Activity-Driven CNS Changes in Learning and Development (Volume 627). This volume was the NY Academy’s bestselling issue of all time.

Book preview

manuscript.

Chapter 1

Overview and basics of the retinotectal system

Abstract

An important research area in developmental neurobiology concerns the mechanisms behind the formation of topographic maps, which constitute the lion’s share of the central nervous system. The mechanisms must build in both considerable flexibility as well as stability and reproducibility. The retinotectal projection, the basis of visual localization, has long been the leading model for such studies. Its advantages include ease of manipulation of physiological stimuli and ease of surgical exposure of both retina and tectum. The early theory of its formation is Sperry’s chemoaffinity, which envisaged selective matching of homophilic receptors on retinal fibers and tectal cells, which in one version were arranged as corresponding gradients across the two sheets of neurons. This chapter introduces the development of ideas in this area, centering on the early finding of seemingly contradictory results, and continuing with the uncovering of gradients, exploration of plasticity, and computer modeling of results.

Keywords

Neural development; Topographic projections; Retinotopic projections; Map formation; Chemoaffinity; Visual localization; Visual plasticity

Even before you understand them, your brain is drawn to maps. There’s just something hypnotic about maps.

Ken Jennings, Author and Jeopardy champion

I am told there are people who do not care for maps, and I find it hard to believe.

Robert Louis Stevenson, Treasure Island

Two important characteristics of maps should be noticed. A map is not the territory it represents, but, if correct, it has a similar structure to the territory, which accounts for its usefulness.

Alfred Korzybski, Semantic scholar

1.1 Overview of the development of circuits

Highly ordered (specific) connections within the central nervous system (CNS) are essential to behavioral function of the organism. The most common feature of CNS organization is the spatial or topographic map—maps of the body surfaces, maps of body musculature, maps of auditory space, and maps of visual space around the organism. Maps are one of the fundamental ways in which the brain analyzes events in the outside world, and they help to organize how we respond appropriately to each event. In fact, in many parts of the CNS, maps from one modality are overlaid on maps of another. It is widely held that organizing sensory projections topographically—that is, with neighboring places in the external world projecting to neighboring parts of their CNS representation—facilitates both the developmental wiring and the analysis of the incoming sensory information. This is because the interconnections necessary to compare information from neighboring parts of the map are shortest and can be organized with simple rules (Chklovskii and Koulakov, 2004). Because of the prevalence of topographic maps, the developmental mechanisms by which such maps are established are of great interest. The retinotectal projection is assumed to be representative, similar in most respects to other topographically ordered projections, and studied much more than any other.

What makes the brain’s development so fascinating lies in its ability to self-assemble neural circuits that can then function in mediating adaptive behaviors. The computer’s circuits are also highly ordered, but the ordering is the work of outside agents—the computer engineers and programmers, which are missing in the brain’s development. Initially the brain must generate the correct number of neurons in each of its nuclei, and then connect these structures properly. To accomplish the orderly connection, the neurons themselves must grow out long cellular processes (called axons) behind their growing tips (called growth cones) as they recognize and grow along specific paths to their target nuclei, and then form appropriate synaptic connections on specific neurons within the target. These are complex processes, but the overall outlines are now well understood.

The primary mechanisms for axonal pathway navigation and subsequent synapse formation involve triggering specific gene expression in exact spatiotemporal patterns so that neurons can interact first with pathway cells to guide their growth along the pathways and then with their target neurons to form the circuits that will connect them together. These steps usually involve a cascade of cellular and genetic interactions. There is an initial expression of genes for secreted and membrane factors on the cells and their corresponding receptors on the surface of the axons that, when activated, not only steer axonal growth cones, but subsequently regulate expression of other gene products for the next step—either via the activation of transcription factors to express (or suppress) a further set of genes or more simply by triggering the translation of preexisting mRNAs already present within the growth cone itself. These gene products in turn generate another round of receptors and/or signals, thus bootstrapping the continuing pathway navigation. Once target contact has been made, a different set of genes supporting synaptic development and the function of neural circuits is expressed and swings into action. This bootstrapping process of interactions is what is generally known as the genetic program for development.

After the circuits are established, they must be able to function sufficiently such that the organism is fully capable of performing all the behaviors necessary for the survival of the individual and, ultimately of course, to reproduce for the survival of the species. The question arises whether the synapses and circuits formed are solely the product of this programmed pattern of gene expression, or whether a process of synaptic adjustment is necessary to produce smoothly functioning neural circuits. Indeed a few simple circuits appear to be immediately fully functional at birth, such as the suckling reflex in newborn mammals, swimming in tadpoles and larval fish, or the escape reflexes triggered by shadows seen in either fish or birds. However, the idea of later synaptic adjustments is consistent with our understanding that practice makes perfect in so many everyday tasks—catching a baseball, hitting a pitch, shooting a basketball, driving a car, etc.

We know from the vast literature on learning that synaptic modification is involved in storing learned tasks. While our basic reflexes seem to be genetically determined, they can still be modified (or new ones produced) via processes such as associative and operant conditioning of learned reflexes. An example of associative conditioning was demonstrated more than a century ago in Pavlov’s conditioning experiments, but examples such as learned flinching, eye blinks, etc., are frequent in everyday life. Somewhat more complex are things learned by operant conditioning. When driving a car, we apply the brakes when we see a stop sign or red traffic light, and do so very rapidly without thinking. Those who have not learned to drive do not, of course, have these reflexes. In these cases, initially neutral stimuli eventually come to reliably elicit new responses, simply by repeatedly pairing them with unconditioned stimuli that reliably trigger the response. This implies that at some point within the nervous system, the signal from the associated conditional stimulus has come to make effective synapses with the response pathway, synapses that either did not exist before or more likely were so weak that they were ineffective and therefore not detected. Since the repeated experience of an associated condition is carried by neural activity in the nervous system, we can progress from saying that experience modifies circuitry to saying that patterns of neural activity drive the synaptic changes in the circuitry.

Developmental studies of how neural activity drives changes in connections have produced an extensive literature, concentrating both on the sensitive periods during early life when activity modifies synapses as well as the exact cellular and molecular mechanisms behind these synaptic changes. One view is that most, if not all, parts of the nervous system are initially capable of being modified by the activity that flows through them, but that most areas largely lose this plasticity upon maturation, so that in the adult brain plasticity is confined mainly to those brain areas that mediate learning and memory. This may be only a matter of degree, because even in maturity the pain of injuries inevitably changes our reflexes so that we may lose coordination or continue to walk with a limp long after the pain is gone. Animal studies have shown that even the most basic reflexes like the stretch reflex in muscles can be modified in adults (Wolpaw et al., 1991).

Inevitably much research has examined the exact demarcation between the genetically specified aspects of development and those aspects that are dependent upon experience and/or neural activity: where do genetically determined molecular mechanisms end? And also what is the widest extent that experience can alter our synaptic connections? This is sometimes referred to as the nature/nurture problem or the genetically determined vs experience-dependent development. I hesitate to use the terms genes vs environment, since the molecular machinery in brain that both carries the neural signals of our experiences and modifies the synapses is in itself coded for by our genes, reminding us that there is a complex interaction between the two. In fact, as described in Chapter 9, the genetic program can even generate spontaneous, fixed patterns of activity, thereby driving activity-dependent (Hebbian) mechanisms in a genetically determined manner before visual experience even begins at eye opening. Such interactions can be studied across all parts of the nervous system, but one area that is particularly advantageous for such studies is the formation of topographic maps of visual space that are set up between retina and brain. As we examine the molecular mechanisms involved in forming the visual maps, a further question arises as to the degree that these mechanisms are also used in establishing other sensory and motor maps.

1.2 Topographic maps in the CNS

An oft repeated example of patterned brain circuits is the topographic map, because much of the brain contains various kinds of maps. In such maps, neighboring neurons on surface A are always connected to neighboring neurons on brain surface B. If the map is further relayed within the brain, neighboring neurons in surface B are always connected to neighboring neurons in surface C, etc. There are maps of sensory receptors on the skin surface, maps of the muscles across the body, maps of auditory space, and finally visual maps. Maps are one of the fundamental ways that the brain analyzes events in the outside world, and they help to organize how we coordinate responses to each event.

In particular, humans, like most vertebrates, are very visual animals, and how the brain develops the ability to identify and track visual objects is a subject of intense interest to Developmental Neurobiologists. Visual systems convert the optical image of the organism’s environment that the lens projects onto the photoreceptors in the retina into a pattern of neuronal activity that encodes both the object’s characteristics and local features as well as its location. This conveys answers to two basic questions: what is it? and where is it? This information is encoded by retinal ganglion cells (RGCs), which report on small retinal areas (called receptive fields) around their cell bodies, and then project their axons topographically to form a map onto a corresponding sheet of neurons in the brain. This organization is accomplished by having neighboring ganglion cells’ axons terminate next to each other in brain, one advantage of which is that analyzing objects larger than one receptive field is done with minimum length of axons for interconnecting between neighboring points. The organism can then use this pattern of neuronal activity to guide its subsequent actions, including identifying food objects for approach or identifying predators for avoidance. Organisms that make the best decisions are more likely to survive and leave offspring than those that make worse ones. Thus, evolution shapes both the structural design as well as the level of plasticity within the nervous system.

1.3 Stability, reproducibility, and flexibility

The building of a functional nervous system juxtaposes two seemingly contradictory issues. First, there must be a mechanism that reproducibly produces a consistent organization of the visual (or other) system that can then function appropriately—with the same function from individual to individual, and also from point to point across the entire visual field within each animal. Second, however, if there is any variability at all in the structure of the retina in the eye or in the target lobe or tectum within the brain, then mechanisms must accommodate these differences. These might be due to differential mutation of various genes controlling development or just the variability in implementing the developing structures produced by the activation of these developmental genes. The adaptive mechanisms must also be flexible enough to accommodate spatial differences in the development of the head—the separation between the eyes, for example—which would affect the ability to generate effective binocular vision from the two eyes’ separate streams of information. From this we can intuit that a completely fixed visual system would not function very well, and that flexibility must be built into the system—something we call visual plasticity.

Visual plasticity is the study of the sum of all those mechanisms that are dependent on the ongoing flow of experiential information after eye opening that provide the capability to make alterations in the basic genetically—and developmentally—determined plan. There is a vast literature on these experience-driven plasticity mechanisms, coming initially from the mammalian retinogeniculocortical system, in the higher level analysis of visual input. Many plasticity mechanisms occur only during initial sensitive periods during development, but others seem to be present throughout life. They are certainly much more pronounced in the visual predators, cats, monkeys, and primates, including humans, which have large cerebral cortices. However, these complex mammals and their visual cortices are often difficult to study at the cellular and the molecular level. Thus, much of our current understanding about how fundamental visual maps are initially formed, as well as how images on the retina are processed by centers in the brain, comes from studies of a relatively small number of simpler model systems—fish, frog, chick, and mice—that we consider here.

This is the story of the retinotectal projection, long studied as a model for both the development of maps in general as well as the mechanisms of experience-dependent plasticity that also shape their final forms. Of course, the stable genetic programs were emphasized most prominently in early studies, and this book traces both the elucidation of the molecular gradients responsible for specification as well as the emergence of the seemingly conflicting evidence for plasticity in what was initially considered a genetically fixed, inflexible mechanism. Indeed, the Eph and ephrin molecular gradients discovered over 25 years ago were often considered to be absolute markers of the position that each retinal axon must innervate in tectum. But further molecular experiments showed that axons are not absolutely bound to innervate their corresponding molecular sites. From this we observe that the ephrin gradients indicate relative rather than absolute positions of innervation, and that other mechanisms such as activity-driven plasticity and competition contribute to the final spatial order and the unfolding of the visual map.

The particular focus of this book is the interaction between the genetically determined and activity-driven mechanisms and how our understanding has unfolded based on various, selected experiments. Much of what has been learned here is applicable to more complex areas of the visual system and the brain in general. But first we need to introduce the retinotectal projection and its advantages for such studies. We begin with a brief presentation of the salient points of the mature retinotectal system, followed by the reasons why it has attracted so much attention, and close with a summary of Sperry’s chemoaffinity theory.

1.4 The mature retinotectal system

The retinotectal projection has three parts: the retina, the pathway (optic nerve and tract), and the tectum (Fig. 1.1). The retina (Fig. 1.2) is the thin, roughly hemispherical shell of tissue at the back of the eye receiving the visual image formed by the lens. The photoreceptors—rods and cones—are the most numerous elements in retina. Their light-evoked signals are sent to second-order neurons (bipolar cells) and laterally conducting elements (horizontal and amacrine cells) that in turn converge—directly and indirectly—on the ganglion cells, the only retinal neurons that project axons into the brain to relay visual information (Fig. 1.2).

Fig. 1.1 A dorsolateral view of the goldfish brain and eyes.

Dorsal is up, anterior is to the right. The optic nerve leaves the back of the right eye, crosses at the chiasm in the ventral diencephalon and ascends as the optic tract to enter the opposite left tectum. The left eye (not shown) innervates the right tectum (Te, arrows). Here a larval goldfish (just after hatching) is draped across the midline between the two tecta. This illustrates the point that this visual system must function as the fish’s CNS grows tremendously in size to that of the adult, a topic covered in Chapter 5. Abbreviations: C, cerebellum; F, forebrain. (Reproduced from Schmidt JT, Buzzard M: Activity driven sharpening of the retinotectal projection in goldfish: development under stroboscopic illumination prevents sharpening, J Neurobiol 24: 384–399, 1993, John Wiley and Sons.)

Fig. 1.2 The structure of the vertebrate retina in the eye.

Light comes from the lens (from bottom upward in diagram). At the top is the black pigmented epithelium at the back of the eye. Below is a diagram of the retinal layers showing the ganglion cells at the inner surface, the only neurons sending axons out of the optic nerve to connect with the brain. They receive information via the bipolar cells from the photoreceptors (rods and cones) above via their cell dendrites (vertical pathway), and from lateral elements (amacrine and horizontal cells) carrying visual input from flanking areas on either side. (Reproduced from Gartner LP, Hyatt JL: Color textbook of histology, ed 3, 2007, with permission from Elsevier.)

The photoreceptors that are functionally connected to a ganglion cell make up that cell's receptive field, the region of the retina where light influences the activity of that ganglion cell, as measured by changes in firing rate of action potentials (APs). The receptive field is generally circular or elliptical, with a diameter of tens to hundreds of micrometers, depending on the species, the location within the retina (smaller in the central area), and the particular kind of ganglion cell. Because the photoreceptors are only a few micrometers in diameter and are generally packed very tightly together, the receptive field often includes hundreds to thousands of them (except in the central area of fine grained vision). The receptive field is a physiological concept, but it is fairly well understood in anatomical terms. Its central part is defined by the dendritic arbor of the ganglion cell, that is, the photoreceptors contributing to the central part of the receptive field are those that project onto the cell's dendritic arbor (Yang and Masland, 1992). The receptive fields, and the dendritic arbors, of neighboring ganglion cells overlap considerably, so they share some of the same photoreceptors, and therefore differ from electronic pixels in a computer image, which are totally separate, nonoverlapping, areas. Thus, the signal from a single photoreceptor projects to more than one ganglion cell, and the signals from a population of photoreceptors converge onto a single ganglion cell. The result is that individual optic nerve fibers report on light activity in small regions of the visual field with some overlap, since several fibers receive information from each point in the field. More details about the vertebrate retina can be found in the books by Rodieck (1973) and Dowling (1968).

The receptive field of a ganglion cell is defined by its location, spatial extent, and its trigger features, those properties of the light stimulus that activate the cell. For example, in the frog, Rana pipiens, there are at least five classes of RGCs. One gives a sustained response to the onset of light, while another responds to small dark moving objects, and has been called a bug detector, implying its naturalistic function in the animal (Barlow, 1953; Lettvin et al. 1959). A third responds to decreased light and is called a dimmer. In most animals unlike the frog, extraction of such features is done at higher levels in brain. Clearly, activity in different types of ganglion cells may signal very different visual stimuli. However, the location of the receptive field depends only on the cell's retinal location, and this has been the property of greatest interest in retinotectal studies.

The axons of the ganglion cells run along the inner retinal surface (the one facing the vitreous) toward the optic nerve head near the center of the retina (Fig. 1.2). There they gather in the optic nerve which runs postero-medially through the orbit to the optic foramen, through which it enters the cranium, and meets the optic nerve from the other eye at the optic chiasm. In most of the animals considered in this book, the retinotectal projection is almost entirely crossed, that is, the great majority of axons in each optic nerve that are destined for the tectum cross the midline to enter exclusively into the contralateral optic tract to innervate the contralateral tectal lobe. A few small direct projections to the ipsilateral tectal lobe have recently been described in adult amphibians, fish, and birds, but they will be ignored here. The ipsilateral projection of mammals, in contrast, is too large to be ignored, and a good part of the interest in mammalian retinotectal development has focused on it.

The optic tract meets the rostral pole of the tectum and the retinal axons deploy their terminals in the tectal lobe as well as some smaller visual nuclei in front of it (e.g., see Sharma, 1972). The tectal lobe, like the retina, is a roughly hemispherical shell, convex both dorsally and laterally, and is usually smaller than the retina. It is a laminated cortical structure with most of its neuronal cell bodies in deeper layers. The optic fibers, which provide the main input, enter relatively superficially in all species and terminate in several well-defined strata (Crossland et al., 1975; Sharma, 1972). Terminal arbors typically extend over an area ranging from tens to hundreds of micrometers wide. Usually, they are flattened and restricted to one narrow sublamina within the tectum. Thus, at the cellular level, the retinotectal projection is from a small area of the retina to a comparably small area of the tectum. The retinotectal map, therefore, is not point-to-point, but area-to-area.

Both the retina and the tectum are also connected to other structures in the brain, but in our treatment, most of these will be ignored except where they figure into mechanisms of map formation. One prominent structure is the nucleus isthmi (n. isthmi; called the parabigeminal nucleus in mammals), which is an important structure in the indirect ipsilateral retinotectal projection in frogs (Fig. 1.3). The pathway is from retina to contralateral tectal lobe; from this tectal lobe to n. isthmi on the same side; from n. isthmi through the postoptic commissure to the tectal lobe on the opposite side (ipsilateral to the retina in which the signal originated; Glasser and Ingle, 1978; Grobstein et al., 1978; Gruberg and Udin, 1978; Keating and Gaze, 1970). There is additionally a reciprocal projection from n. isthmi to the tectal lobe on the same side. Thus, both tectal lobes receive a direct projection from the contralateral eye (and its reflection from n. isthmi) as well as an indirect one from the ipsilateral eye.

Fig. 1.3 Schematic of ipsilateral retinal input converging in tectum with contralateral input via a pathway through nucleus isthmi (NI).

Ganglion cells (circles) in the two retinas receive the image of a common point in the binocular visual field. The left eye projects directly to the right tectal lobe (triangle = synaptic ending). The right eye similarly projects to the left tectum, where a neuron (circle) relays the signal to the left NI. Axons from neurons in the left NI cross the midline in the postoptic commissure en route to the right tectum, converging with left eye input. The left NI also projects back to the left tectum. Binocular input converges on the left tectum by a symmetrical route. (Reproduced from Titmus MJ, et al.: Effects of choline and other nicotinic agonists on the tectum of juvenile and adult Xenopus frogs: a patch clamp study, Neuroscience 91:754, 1999, with permission from Elsevier.)

Neighboring RGCs project their terminals to neighboring and overlapping regions on the tectal hemisphere. The location of a terminal on the tectal surface is predictable from the retinal address of its ganglion cell body. In goldfish, for instance, temporal, nasal, dorsal, and ventral retinal quadrants project to anterior, posterior, lateral, and medial tectal quadrants (Schwassman and Kruger, 1965). This orderly arrangement, in which the neighbor relations on the retinal surface are reproduced on the tectal surface, is described as being a retinotopic projection. In frogs, both retinotectal projections, direct and indirect, are retinotopic and in register with one another (Gaze and Jacobson, 1962). Apparently, the two projections converge on single cells in the tectum, as binocular tectal units have been described (Fite, 1969). In mammals, binocular units are present in visual cortex, which is not well developed in the other vertebrate classes. But mechanisms for forming binocular visual projections in tectum can be compared to what is known of the formation of binocular cortical areas in mammals.

Retinotopic organization is the feature of the retinotectal projection that has attracted the most attention. The retinotectal system in the study of neuronal connections is used to search for the mechanisms of formation of such retinotopically organized projections.

1.5 Advantages of studying the retinotectal system

The retinotectal projection is assumed to be representative, similar in most respects to other ordered projections, but has been studied much more than any other. Why is this so? What makes the retinotectal projection so amenable to study?

The answer lies mostly in convenience. The visual system, of which it is a part, is surely the most heavily studied region of the nervous system, as a scan of recent abstracts of the Society for Neuroscience confirms. It has certain attractive features. Physiological studies are facilitated because the function of the visual system is known and because stimulus variables (location, size, and intensity) can be controlled with relative ease. Anatomical studies are facilitated by the physical isolation of the eye from the rest of the CNS, which makes it readily accessible to surgery, to drug treatment by intraocular injection, or to the introduction of axon-tracing molecules. Finally, in fish and amphibians, the tectum is large and nearly as accessible as the retina, as it occupies a superficial position in the brain, and is readily exposed and visualized after a few minutes of simple surgery.

Both of these features—the order of the projection and the convenience of the preparation—are powerful reasons for studying it, but the interest of developmental neurobiologists was initially stimulated by another attribute, the ability of adult fish and amphibians to regenerate their optic nerve fibers. When the set of connections is interrupted by severing the optic nerve, the axonal endings are separated from their cell bodies and degenerate, so the animal is blinded. But after a few weeks, fish and amphibians regain vision, and can localize objects and discriminate hues. Such tasks almost certainly require the reestablishment of the same classes of connections that existed before the lesion. For several decades, regeneration was studied as a model for development, because effective techniques for manipulating and assessing the much smaller embryos (Fig. 1.1) had not yet been developed. Now most studies focus on how connections develop, and only a few on how regeneration of connections may differ.

1.6 Sperry's chemoaffinity theory

Roger Sperry's chemoaffinity theory has been the dominant hypothesis in the study of the retinotectal development for several decades. As stated in his seminal paper, Chemoaffinity in the orderly growth of nerve fiber patterns and connections, (Sperry, 1963): The patterning of synaptic connections in the nerve centers, including those refined details of network organization heretofore ascribed mainly to functional molding in various forms, must be handled instead by the growth mechanism directly, independently of function, and with very strict selectivity governing synaptic formation from the beginning. The establishment and maintenance of synaptic associations were conceived to be regulated by highly specific cytochemical affinities that arise systematically among the different types of neurons involved via self-differentiation, induction through terminal contacts, and embryonic gradient effects… It seemed a necessary conclusion… that the cells and fibers of the brain and cord must carry some kind of individual identification tags, presumably cytochemical in nature, by which they are distinguished from one another almost, in many regions, to the level of a single neuron… (pp 703–704).

The key element is that growth cones were thought to recognize and begin to form functional contacts on cells on the basis of their chemical markers. This contrasted with the alternative view, according to which initial contacts were made rather randomly, and only the ones that were adaptive for the animal were validated, and retained for adult life. This feature, that only useful connections should be kept, was the one that Sperry attacked experimentally over nearly three decades.

1.7 Organizational overview of the material

As described in Chapter 2, he was correct with respect to the primary level of organization; a molecular chemoaffinity mechanism appears to be responsible for the initial order of the retinotectal map, regardless of its value to the organism as demonstrated after surgical rotation of the eye. The molecular basis for this organization is introduced in Chapter 3, but it turned out to be a gradient of chemorepulsion rather than chemoaffinity. In addition, the early projection is very roughly ordered, then further modified and refined by several other processes. Experiments outlining the effects that organize maps contrary to rigid chemoaffinity predictions are introduced first at the end of the next chapter and then covered in detail in Chapters 4 and 5. Among these additional processes are, first, the competition between fibers for synaptic space in the tectum, and second, a type of adaptive process for validating effective synapses—later referred to as activity-driven (Hebbian) synaptic stabilization. The latter is covered in detail in Chapters 9 and 10, after we look more extensively at the details of the developing projection, from developmental, anatomical, and molecular perspectives (Chapters 6–8).

References

Barlow H.B. Summation and inhibition in the frog’s retina. J Physiol Lond. 1953;119:69–88.

Chklovskii D.B., Koulakov A.A. Maps in the brain: what can we learn from them?. Annu Rev Neurosci. 2004;27:369–392.

Crossland W.J., Cowan W.M., Rogers L.A. Studies on the development of the chick optic tectum IV: an autoradiographic study of the development of retinotectal connections. Brain Res. 1975;91:1–23.

Dowling J.E. Synaptic organization of the frog retina: an electron microscopic analysis comparing the retinas of frogs and primates. Proc R Soc Lond Ser B. 1968;170:205–228.

Fite K.V. Single unit analysis of binocular neurons in the frog optic tectum. Exp Neurol. 1969;24:475–486.

Gaze R.M., Jacobson M. The projection of the binocular visual field on the optic tecta of the frog. Q J Exp Physiol. 1962;47:273–280.

Glasser S., Ingle D. The nucleus isthmus as a relay station in the ipsilateral visual projection to the frog’s optic tectum. Brain Res. 1978;159:214–218.

Grobstein P., Comer C., Hollyday M., Archer S.J. A crossed isthmotectal projection in Rana pipiens and its involvement in the ipsilateral visuotectal projection. Brain Res. 1978;156:117–123.

Gruberg E.R., Udin S. Topographic projections between the nucleus isthmi and the tectum of the frog Rana pipiens. J Comp Neurol. 1978;179:487–500.

Keating M.J., Gaze R.M. The ipsilateral retinotectal pathway in the frog. Q J Exp Physiol Cogn Med Sci. 1970;55:284–292.

Lettvin J.Y., Maturana H.R., McCulloch W.S., Pitts W.H. What the frog’s eye tells the frog’s brain. Proc I R E. 1959;47:1940–1951.

Rodieck R.W. In: Francisco S., Freeman W.H., eds. The vertebrate retina: principles of structure and function. 1973.

Schwassman H.O., Kruger L. Organization of the visual projection upon the optic tectum of some freshwater fish. J Comp Neurol. 1965;124:113–126.

Sharma S.C. The retinal projection in the goldfish: an experimental study. Brain Res. 1972;39:213–223.

Sperry R.W. Chemoaffinity in the orderly growth of nerve fiber patterns and connections. Proc Natl Acad Sci U S A. 1963;50:703–710.

Wolpaw J.R., Lee C.L., Carp J.S. Operantly conditioned plasticity in Spinal Cord. Ann N Y Acad Sci. 1991;621:338–348.

Yang G., Masland R.H. Direct visualization of the dendritic and receptive fields of directionally selective retinal ganglion cells. Science. 1992;258:1949–1952.

Further reading

Dowling J.E., Boycott B.B. Organization of the primate retina: electron microscopy. Proc R Soc Lond B. 1966;166:80.

Schmidt J.T., Buzzard M. Activity driven sharpening of the retinotectal projection in goldfish: development under stroboscopic illumination prevents sharpening. J Neurobiol. 1993;24:384–399.

Udin S.B., Fisher M.D. Development of the nucleus Isthmi in Xenopus laevis. I Cell genesis and the formation of connections with the tectum. J Comp Neurol. 1985;232:25–35.

Chapter 2

Early work supports chemoaffinity with one contradictory result

Abstract

The early work in the retinotectal field supported Sperry’s chemoaffinity, but one result with compound eyes directly contradicted it. Sperry found that regeneration of the optic nerve in fish and frog reestablished the original connections even when maladaptive. Following eye rotation, this resulted in reversed visual responses. Eye transplants from left to right side showed a reversal in one axis only. Behavioral assays after tectal lesions showed the existence of a visual map, while neuronal recordings verified the map and its reestablishment after nerve regeneration. Compound eyes, formed by replacing the nasal, temporal, or ventral half of one embryonic eye with the same half of the opposite eye, resulted in mirror symmetric maps that covered not just the appropriate half of tectum but the entire tectum. This expansion into the inappropriate half contradicted rigid chemoaffinity. Further chapters identify the chemoaffinity gradients, the capacity for plasticity, and how these two are reconciled.

Keywords

Visual maps; Sperry’s chemoaffinity; Optic nerve regeneration; Compound eyes; Visual plasticity; Visual behaviors

How wonderful that we have met with a paradox. Now we have some hope of making progress.

Niels Bohr, Atomic physicist, Nobel prize

The chemoaffinity hypothesis evolved and held center stage from the 1940s through the 1960s (Gaze, 1970, 1982; Hunt and Cowan, 1990), largely due to Sperry’s work on the retinotectal projection, although similar ideas had been proposed earlier by Langley, Ramon y Cajal, and Sherrington. This chapter summarizes the work to 1963, which was a pivotal year in two respects. First, the most widely cited statement of the theory appeared in that year, in the paper quoted from Chapter 1 (Sperry, 1963). Second, a very provocative paper was published by Gaze et al. (1963) in which chemoaffinity was shown to fail rather badly in predicting the outcome of an interesting experiment. But first, the early work.

2.1 Grafted eyes regenerate optic nerves to restore vision

The restoration of vision after grafting an eye or cutting an optic nerve was first convincingly demonstrated by R. Matthey, in 1927, followed later by both L.S. Stone and Roger Sperry. Working with adult Triturus cristatus, a urodele amphibian (newts and salamanders), Matthey either grafted a new eye into an orbit from which the original had been removed surgically, or he removed a portion of the optic nerve of an intact eye. After a long recovery period, the opposite normal eye was removed so that vision could be unequivocally tested through the experimental eye. The animals demonstrated restored vision by attacking objects such as a worm that was enclosed in glass, which removed the possibility of using other sensory inputs. Histological examination also revealed axons that extended from eye to brain. Although only a small percentage of animals successfully regenerated, the work had great credibility because of his rigorous testing.

The reconnection of the eye and brain then became the primary issue, but eye grafting is a poor way to study it both due to the low success rate (many grafted eyes did not adequately revascularize and so did not regenerate) and also due to the long delay before optic nerve fibers begin to grow. Therefore, attention switched to the regeneration of the optic nerve following a cut or crush. Nerve crush, which leaves the sheath intact, generally does not damage the retinal blood supply so that the retina remains intact and the optic axons regenerate quite reliably with only a short delay. Do they reestablish the same connections in the brain that existed before the lesion?

2.2 Optic nerve regeneration restores original connections even when maladaptive

Roger Sperry had been studying regeneration in the peripheral nervous system and found that motor performance was never as good after regeneration as before. He concluded that regenerated neuromuscular contacts in mammals formed randomly in the sense that individual axons in a severed motor nerve did not necessarily reinnervate the same muscle as before (Sperry, 1945a). Although the issue had been settled in the peripheral nervous system, Sperry acknowledged that the central nervous system might be different—more adaptable and plastic, and perhaps influenced by experience. The visual system seemed a good place to study this, since humans, fitted with spectacles that reversed the visual field, adapted quite well (Stratton, 1897; Ewert, 1930 (both cited in Sperry, 1943a)).

Sperry reversed the visual fields of newts (Triturus viridescens) by rotating their eyes 180 degrees about the visual axis after having cut the extraocular muscles (Sperry, 1943a,b). The retina and optic nerve remained functional, as their blood supply remained intact, and the eyes reattached to the orbit in their new orientation, as indicated by the different coloring of the dorsal and ventral iris. The function of these experimental newts and controls (with the eyes sham-rotated) was tested using two visual behaviors: prey location (same as Matthey and Stone used) and the optomotor response. For the latter, the animal was placed inside a rotating drum with vertical black/white stripes on its interior surface. Newts have very little eye movements so they typically turn their heads to track in the direction of horizontal drum rotation, a reaction that minimizes the movement of the image across the retina and thereby facilitates the extraction of visual information. This is a convenient indicator of vision, as the animal is unrestrained, as it involves the whole body, and no learning is required.

Animals with both eyes rotated, or with one rotated and the other removed, showed the most abnormal behavior, evident in three settings. Presented with prey, they systematically oriented in a direction 180 degrees off target, indicating that they saw objects in front of them as if they were behind them. In the rotating drum, their heads rotated against, not with, the movement of the stripes. Again, this is the expected result if they saw the world reversed, and this behavior is clearly maladaptive, as it speeded up the movement of the retinal image of the stripes rather than slowing it down. Finally, even in stationary environments, they circled continuously, so called circus movements that are readily interpretable in terms of the same optomotor reflex involved in drum-following. When an animal's head makes a slight leftward rotation in a stationary environment, the visual scene moves to the right, but because of the eye rotation, the animal sees leftward movement which evokes a reflexive leftward following response that, rather than decreasing the movement of the retinal image, actually increases it. This positive feedback continues to evoke the inappropriate response, and the animal turns continuously in circles. The normal vestibuloocular reflex is apparently too weak to overcome the reversed optokinetic response. The same logic ought to apply to the self-induced movement of the retinal image in the vertical direction, but circus movements are usually restricted to the horizontal plane—apparently the optomotor reflex in response to vertical movement is controlled effectively by the vestibular and tactile input.

The animals never altered these responses. The power of these results derives both from the persistence of these behaviors, and from their maladaptiveness (turning away from prey rather than approaching it, rotating the head to increase the movement of the retinal image, and turning in circles continually). In Sperry's words: Their perverse intractability to correction by experience suggests that like the spinal motor patterns…, they are organized in the beginning by the growth process itself rather than through trial and error adjustment… (Sperry, 1943b).

Next, Sperry combined eye rotation with optic nerve section and found two very important results (Sperry, 1943b). First, after the optic nerve regenerated, the animals still behaved as if the world were reversed, strong evidence that maladaptive connections were reestablished and maintained (Fig. 2.1). Second, histological examination of the silver-stained nerve and tract revealed that the site of the lesion contained a tangle of fibers, in contrast to the orderly laminar fascicles of the normal nerve. The fact that visual function was still restored despite the disorder of fiber regeneration argued strongly against mechanical pathway guidance, the idea that fibers had to travel through precisely specified pathways in order to terminate at their correct sites. These experiments were repeated in anurans with the same results (Sperry,