Molecular and Cellular Mechanisms of Brain Development

David A. Morilak, Matthew H. Porteus, and Roland D. Ciaranello

D. A. Morilak, M. H. Porteus, and R. D. Ciaranello: Nancy Pritzker Laboratory of Developmental and Molecular Neurobiology, Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA 94305.

Researchers and clinicians concerned with severe psychiatric disorders are focusing increasing attention on brain development. There are several reasons for this, including an increasing recognition that exogenous or endogenous disruptions in brain development may play an important etiologic role in psychiatric disorders, particularly such severe disturbances as infantile autism and schizophrenia. There has also been a veritable explosion in our knowledge of developmental neurobiology, that area of neuroscience that focuses on factors regulating the development of neurons, neural circuitry, and the complex regional organization of systems in the brain. Finally, the application of molecular genetics to neurologic disease has enabled us for the first time to understand the genetic bases of certain diseases without requiring foreknowledge of the underlying biochemical abnormalities.

It is difficult, if not impossible, to overstate the importance of this last point. We know virtually nothing about the biologic substrates of the severe neurodevelopmental disorders, despite years of intensive study. The same statement can be made about psychiatric diseases in adults, where an even greater, more prolonged research endeavor has failed to disclose even a single consistent biologic deficit which can guide the diagnosis or treatment of any psychiatric disorder. Until recently, this lack of knowledge was a major obstacle to progress in psychiatric research. Although this remains a formidable impediment today, the ability to identify defective genes without a priori knowledge of their protein products or underlying pathophysiology has contributed recently to breathtaking advances in medicine, and there is every reason to believe that such advances also hold great hope for breakthroughs in psychiatric diseases, and in particular for the developmental disorders.

This chapter attempts to guide the reader through the basic principles of neuronal and brain development. The relatively new discipline of developmental neurobiology draws, as one might expect, from many other disciplines, including neuroanatomy, neurophysiology, neuropharmacology, endocrinology, and genetics, as well as molecular biology, and derives its clinical sustenance from pediatrics, neurology, and psychiatry. This chapter is organized so as to provide the reader with a basic understanding of the genetic, metabolic, and cellular processes underlying brain development before considering how disruptions in these processes might lead to disturbances in brain maturation and ultimately to disturbances in behavior and social functioning (see Basic Concepts and Techniques of Molecular Genetics, Proto-Oncogenes: Beyond Second Messengers, The Development of Brain and Behavior, and Neuroendocrine Interactions).

This section summarizes what is known about mammalian brain development at a molecular level in a four-part way. The first gives a broad overview of the basic embryology of the developing brain. This is intended as a review and to establish a common ground from which further discussion will emanate. The second outlines certain cell biological events and concepts that are important in brain development. The third part addresses the possible molecular bases underlying some of these developmental processes. Finally, in the fourth part, we speculate about the future focus of molecular brain research as it relates to issues in developmental neurobiology. In this section, we emphasize the emergence of new and powerful techniques and approaches that are just starting to contribute to our general understanding of the molecular basis for development (for related issues and discussion, see Basic Concepts and Techniques of Molecular Genetics, Proto-Oncogenes: Beyond Second Messengers, The Development of Brain and Behavior, and The Neurobiology of Treatment-Resistant Mood Disorders). One point that will become obvious from this overview is that there is not yet a comprehensive picture of mammalian brain development; many hypotheses have been derived from experiments in other developmental systems and remain to be confirmed or repudiated in the mammalian brain.

Neurogenesis begins with the formation of the neural plate, a thickening of ectodermal cells on the dorsal aspect of the developing embryo. Ridges form at the lateral edges of the plate, curling up to meet at the dorsal midline to form the neural tube. The internal cavity created by the tube is called the ventricle. Initially, the neural plate appears uniform along its entire axis. However, even as closure of the neural tube is occurring, specialized regions of the nervous system begin to emerge through differential cell division and migration. Major subdivisions include the mylencephalon and metencephalon (giving rise to the medulla, pons, and cerebellum), the mesencephalon, and the prosencephalon, which matures into the diencephalon and telencephalon. Through this process, the subdivision of the developing brain lays the foundation for regional specialization in the mature brain.

The development of the laminar organization of the neocortex (see Fig. 1) presents a salient example of many of the issues discussed in this chapter. Excellent and detailed reviews of this process can be found in refs. 1, 34, 44 and 45. First, the ventricular zone forms. This is a layer of ectodermal pseudostratified neuroepithelial cells lining the ventricle of the neural tube. Ventricular zone cells are small round mitotically active cells that eventually give rise to all of the cell types, including both neurons and glia, in the mature brain. Thus, the ventricular zone is also known as the germinal zone. After ventricular zone cells stop dividing, becoming terminally postmitotic, they migrate to their target destination, which is remote from the ventricular zone, where they differentiate and mature.

As proliferation of precursor cells in the germinal layer of the primordial cortex progresses, a second layer of mitotic cells emerges above the ventricular zone, called the subependymal or subventricular zone. A third layer then forms above it, called the intermediate zone. Cells in this region have undergone their final mitosis and have begun to migrate upward to what will become the cerebral cortex. Throughout development, the intermediate zone represents a region of transition and cell migration, as well as a staging area, where cells begin to extend processes and attain specific orientation. This process determines the direction of their migration. The next layer to form in the developing cortex is the primordial plexiform layer, formed by the first neurons to arrive in the cortex. As more neurons are born (i.e., undergo final mitosis) and migrate, the plexiform layer is split by the new arrivals into two layers, the marginal zone and the subplate. The marginal zone becomes layer I in the adult cortex, a superficial, cell-sparse zone underlying the pial surface. The cortical subplate lies under the newly arriving cortical neurons. Cells in the subplate are the targets of early afferent fibers from subcortical structures. This innervation is a transient phenomenon and may serve an organizational or coordinating function in the formation of the developing cortex and in establishing appropriate mature contacts with subcortical afferents. Finally, the neurons of the cortical gray matter are born and migrate up through the intermediate zone and subplate to reside in the cortical plate.

There is a radial inside–out gradient of neurogenesis, such that the earliest born neurons occupy the deepest layers of the cortex. Successive generations of cortical neurons migrate up through the already established layers to reside in the next most superficial layer until all layers VI–II have been established. The timing of the passage of migrating neurons through the subplate seems to coincide precisely with the gradients of innervation of the cortex by the major subcortical afferent systems, including specific thalamic inputs, noradrenergic innervation from the locus coeruleus, serotonergic innervation from the raphe nuclei, dopaminergic innervation from the ventral tegmental region, and cholinergic innervation from the basal forebrain.

The proliferation, organization, and specialization of the brain at a regional level subsumes a variety of cell biological processes unique to development, including division, migration, differentiation, and the establishment of appropriate synaptic connectivity. These, and related developmental cellular processes, are addressed in the next two sections.

In this section, we describe research relating to five important developmental questions: i) What is the cell lineage of neurons in the brain? ii) How do cells migrate to their appropriate destinations? iii) How do neurons make the correct axonal connections? iv) How do cells and regions differentiate? v) How plastic is the developmental process? These issues are based on the decisions that an undifferentiated ventricular zone cell must make to become a functionally mature neuron. Moreover, these issues take on added significance for understanding human psychopathology, as presumed abnormalities in these developmental processes of migration, differentiation, cell death and regression, and formation of appropriate circuit connections are hypothesized to play a role in the etiology of certain neuropsychiatric disorders, especially schizophrenia (see The Development of Brain and Behavior).

To summarize what is known about cell lineage of neurons in the brain, we focus on three types of experiments: birthdating studies, transplantation studies, and retroviral tracing experiments. Birthdating studies can reveal if the final destination of a cell is related to the time at which it undergoes final mitosis. These studies are done by giving a pulse of tritiated thymidine to animals in utero. When cells replicate, they then incorporate tritiated thymidine into their DNA and become radiolabeled. If a cell continues to divide from that point, the label will dilute among its daughter cells and become undetectable by autoradiography. Thus, the most intensely labeled cells are those that will have divided only once (final mitosis) after the thymidine pulse. After a period of maturation, usually when the animal has become an adult, the brain is sectioned, and the location of the radiolabeled cells are determined. By varying the time during development at which the thymidine pulse is delivered, the distribution in the mature brain of cells that underwent their final mitoses together can be determined. Conversely, the different times at which neurons in a particular location became terminally postmitotic can be established.

It was through such birthdating studies that the timing and developmental gradients of cortical neogenesis described in the preceding section were established. Similar studies have been conducted in other regions of the brain, including those with a nuclear, rather than laminar, pattern of organization. For example, cells in the neostriatum are organized into clumps, or patches, surrounded by more loosely arranged matrix cells; birthdating studies have shown that patch cells become postmitotic at a distinctly earlier time than matrix cells (54). Thus, both the cerebral cortex and the striatum are examples in which there is a temporal determinant to the final cell location.

However, even though birthdating studies have shown that the layers of the cortex develop in a sequential manner, such studies do not show whether the cells migrate to appropriate layers in response to specific signals in the surrounding environment or if their destination is predetermined. Transplantation studies have revealed some complex answers to this question. For example, thymidine-labeled ventricular zone cells that normally would have migrated to layer 6 were transplanted into a ventricular zone that was making layer 2/3 cells (32). If the cells underwent final mitosis while in the layer 6 environment, they still migrated to layer 6 when placed in the new environment. However, if they incorporated labeled thymidine while temporarily placed in culture, they responded to the new environment and migrated to layer 2/3. Thus there is a complicated relationship between the time at which the cell becomes postmitotic, the environment in which it undergoes its final DNA synthesis, and the layer to which it will eventually migrate.

Retroviral lineage studies are designed to determine whether a given ventricular zone cell gives rise to different cell types or to cells of a single cell type. These studies are carried out by infecting ventricular zone cells with a small number of disabled retroviruses bearing a genetic marker. The retrovirus inserts itself into the host cell genome during DNA replication and gives the cell a novel genetic marker, but is unable to infect other cells. Thus, once in the genome, that marker is passed on uniquely to all of the infected cell's progeny, and all the cells that show the marker are clonally related to the single original cell that was infected. These studies can address whether cells within a clonal population are restricted in their final destination and whether cells within a clonal population are restricted in their mature phenotype.

Retroviral lineage studies in the cerebral cortex have shown that early in brain development any given ventricular zone cell can give rise to every cell type in the brain. However, at later stages of development, the multipotency of ventricular zone cells becomes restricted. An early split in lineage occurs between the astrocytes of the grey matter and other cells of the brain (41). Some studies suggest that certain germinal cells become restricted to only generating neurons (29). Germinal cells appear able to give rise to cells of all cortical layers within a vertical column. On the other hand, some ventricular zone cells can generate offspring cells separated widely in the horizontal plane (41, 56). These studies highlight the fact that we are only beginning to understand the process that determines the lineage of cells in the brain.

Once a cell becomes postmitotic, how does it get from the ventricular zone to its final destination? In cortex, as neurons begin to migrate out of the germinal zone, the beginnings of process formation occur. An apical dendrite extends vertically toward the pial surface, and this defines the path along which the cell will migrate. These processes are in close association with radial glial cells, which themselves extend processes that are anchored at both the pial and ventricular surfaces. These glia form a scaffold through the developing cortex and act as guide wires along which radial migration occurs into and through the cortical plate (17, 44, 45). With the electron microscope, putatively migrating cells can be seen in close apposition to radial glial fibers and appear to be creeping up the fiber. Real-time video microscopy shows that cerebellar neurons in vitro move along Bergmann glial cells and resemble the migrating brain cells that are seen in the electron microscope (17). It has been suggested that migration along these radial glia forms the developmental basis for the vertical, columnar organization of functional cortical modules in the mature brain.

Genetic mouse mutants that show defects of neuronal migration highlight the importance of the interaction between migrating neurons and radial glia. In the weaver mutant, migrating cerebellar granule cells are unable to attach to the glia and do not migrate correctly, ultimately leading to the death of the granule cells (4, 17). In the reeler mutant, the cerebral cortex develops outside–in rather than inside–out, but there is no cell death (4, 17). In this mutation, it seems that the migrating cells are unable to detach from the radial glia cell and thus form a roadblock for the migration of subsequent waves of neurons.

As the primitive neurons traverse the intermediate zone, there is often a pause in their migration, during which their orientation and their direction of movement may adjust, to be resumed again in a vertical direction just below their final cortical destinations. During this sojourn, axons begin to emerge, and the temporary shifts in orientation may in part determine the direction in which axons are initially extended.

The mechanisms are vaguely understood for establishing a neural pathway in the brain, sometimes over a very long and circuitous route. Axons grow toward a successive series of intermediate targets, like signposts, along the route to their final destinations. Along the way, a striking degree of organizational specificity is attained by the establishment of appropriate axonal connections over long distances and, equally important, by the omission and preclusion of a myriad of potentially inappropriate connections.

This process represents a complex interaction between cell-surface proteins, extracellular matrix proteins, and diffusible attractant and repellent factors secreted by target regions in the brain (reviewed in refs. 6 and 51). Axons emanating from any general region adhere to each other, forming a pathway or bundle (fasciculation) from the interaction of surface proteins expressed on the axons. Similarly, they adhere to and migrate along the extracellular matrix and along cell-to-matrix boundaries by virtue of interactions between surface proteins and proteins on the surface of other cells or in the matrix itself.

The most studied example of axonal pathfinding is the developing visual system. In the mammalian visual system, retinal ganglion cells form synapses with neurons in the lateral geniculate nucleus (LGN), which then project to layer 4 of the visual cortex. A striking aspect of this pathway is the organization and maintenance of a precise retinotopic map in both the LGN and visual cortex. One strategy by which such organization is believed to develop is by the initial projections of pioneer neurons from target regions, which then serve as a template upon which the mature pathway is established. The concept of pioneer neurons was first proposed in the developing grasshopper (2), and it has been proposed that cells in the cortical subplate act as pioneer neurons in the cerebral cortex (31). Subplate cells are present only in development and are among the first cells to become postmitotic in the developing cortex. As determined in dye tracing studies, the cortifugal subplate neurons make connections with the LGN even before layer 4 cells are born. Moreover, these studies suggested that LGN axons dock on the subplate cells of the visual cortex. The LGN axons then wait in the subplate until layer 4 cells, the mature targets of LGN axons, are born and migrate into place before the axons themselves grow into layer 4 to form synapses (13). If the subplate cells beneath the developing visual cortex were removed, the LGN axons grew past the visual cortex and never found a place to dock. Although such studies point out the importance of a transient cellular scaffold in guiding axons as they approach their final targets, they do not address how the LGN axons got to the subplate or how the pioneer subplate axons got to the LGN in the first place.

At the tips of developing axons are specialized, enlarged structures called growth cones. These exhibit a highly active, ameboidlike array of filamentous processes that extend and retract continuously. The direction of outgrowth of these processes determines the direction of axon extension and hence the direction of axon pathway formation. Whereas axon fasciculation and adherence is mediated by surface proteins, growth cone motility and hence the direction of pathway formation is influenced by diffusible attractant or repellent substances secreted by their intermediate or final destinations. A chemoaffinity hypothesis, proposed by Sperry in the 1940s, suggests that axons follow such chemical gradients to find their correct destination (43). Specific secretion of diffusible factors establishes gradients to which the growth cones respond by either extending and growing toward (an attractant gradient) or retracting and turning from (a repellent gradient). Coculture experiments show that axons preferentially grow toward appropriate neural target tissue and not toward other tissues (19). Even in the early stages of axon formation, the floor plate of the developing neural tube releases a substance that attracts axons to the midline (6). When the floor plate is prevented from forming by removing the notochord, axons fail to migrate to the midline.

In development of cortical connections, there is an overabundance of synapse formation and more neurons than in the adult. Thus, an important part of cortical development and connectivity is the retraction and elimination of excess cells and exuberant synaptic contacts (reviewed in ref. 9). As many as 25% to 50% of the maximal number of cortical neurons seen during development in a given layer ultimately die and are eliminated in the early postnatal period (10, 37, 58). These regressive processes serve to refine the cortical circuitry and again are not well understood. Cell death may result from competition for limited supplies of neurotrophic growth factors (see below) or from a failure to establish a functional synaptic connection with a target region, leading to a reduced availability of such factors for uptake and transport.

Cell differentiation is the process by which a cell changes from an immature, pluripotent phenotype to a more mature, specialized one. This process encompasses cell lineage, cell migration, axonal outgrowth, and plasticity, as well as the establishment of such specific neuronal characteristics as neurotransmitter content and interneuronal patterns of synaptic connectivity and responsivity. These processes, occurring at the cellular level, also reflect a higher level of organization resulting in the specification of particular regions of the brain as cells within those regions mature. Regional specification is thus the process by which a part of the brain takes on, through characteristic patterns of cellular phenotype, organization, and connectivity, a functional identity that distinguishes it from other regions. The question of regional specification is one of the most actively studied and debated areas in developmental neurobiology. There are two fundamental views of regional specification: (i) that it is an intrinsic, genetically controlled process or (ii) that it is an extrinsic, environmentally controlled process.

Cell transplantation has been used to explore this question. Cells are transplanted from one environment to another, and then they are examined as to whether the transplanted cells have adopted the phenotype of the donor environment or of the acceptor environment. For example, part of the developing occipital cortex has been transplanted into the developing somatosensory cortex. The transplanted tissue adopted the organizational and connectivity characteristics of the acceptor environment. Occipital donor tissue, which normally does not project through the pyramidal tract, established a pyramidal projection when transplanted into the rostral neocortex, an area that normally maintains such a projection (36). In another experiment, transplanted occipital cortex cells formed "barrels," a functional architectural feature unique to the somatosensory cortex (46). Thus, based on these results, local environment does play an important role in specifying morphologic and functional phenotype. However, the retroviral tracing and transplantation studies described earlier highlight the importance of timing in interpreting these results. If the transplantations had been performed later in development, after some critical period had passed, the donor tissue may not have taken on host tissue characteristics, and the interpretation could have been that environment had no influence. A more serious concern is the issue of selective attrition; it is possible that only the tissue that was capable of assuming host characteristics could survive the transplantation. Thus, the experiment may have selected for tissue that could adapt, rather than revealing the differentiation potential of the donor tissue. Nevertheless, it seems certain that the local environment does play an important role in differentiation, although exactly what the local environment contributes is not known.

Another approach has been to alter the input to a particular region of the cerebral cortex. One way of altering the input to the visual cortex is to study an anophthalmic animal. Rakic (45) showed that the visual cortex of an anophthalmic mouse maintained its correct topography and connections. This suggests that the visual cortex does not need environmental cues to develop correctly. A second way to alter the input to the visual cortex was to enucleate an eye of an embryonic animal. This caused cell death in the lateral geniculate nucleus and a corresponding decrease in the number of thalamic inputs to the visual cortex. The visual cortex of such an animal appeared biochemically and cytologically normal (45). There was a decrease in the size of the visual cortex that was attributable to a decrease in the number of columns. But each column seemed functionally normal, despite the lack of normal input. A third type of experiment was to reroute retinal ganglionic afferents to the medial geniculate nucleus, the thalamic auditory relay center (45). After such an alteration, what would normally have been the auditory cortex became capable of processing visual information. Thus, the functional specificity of a cortical region was altered by changing its functional input.

The final developmental phenomenon we wish to discuss is the plasticity of the developing brain. The immature brain shows an amazing ability to adapt to changes in its environment and to fine tune its connections during development. Again, the visual system provides the best example to illustrate plasticity. The mature visual cortex consists of a series of alternating columns that respond preferentially to one eye or the other, called ocular dominance columns. When a radioactive tracer is injected into one eye or the other, it is retrogradely transported back to the visual cortex, revealing a series of alternating stripes corresponding to the ocular dominance columns (21). This organization is created by the axonal arborization of LGN neurons, which are driven by one eye or the other, with the projections of a given LGN neuron limited to a single column. During development, however, the axon terminals of LGN neurons that are driven by opposite eyes overlap in the cortex. It is only by subsequent pruning that the ocular dominance columns are formed. If activity from one eye is blocked, either pharmacologically or physically, during a particular time called the critical period, the normal ocular dominance columns do not form (47). Instead, the stripes corresponding to the blocked eye become narrower and the stripes corresponding to the unblocked eye get wider. The unblocked eye seems to take over the columns normally controlled by the blocked eye. It has been proposed that the organization into ocular dominance columns is, therefore, the result of some sort of activity-dependent competition among neurons. In the next section, we discuss a possible molecular mechanism for this competition. Moreover, this activity-dependent competition has a temporal profile (47). If activity is blocked after a critical period, there is no effect, and if the block is removed before the critical period, there is also no effect. The block has to span a specific period of time during development to affect the organization of ocular dominance columns.

The techniques of molecular biology allow scientists to study individual genes and gene products in a unique and powerful way. Only recently, however, have those techniques been applied to the study of mammalian brain development. Thus, our knowledge of possible molecular mechanisms underlying the developmental phenomena described in the previous two sections is just beginning to accumulate.

An important decision for ventricular zone cells is when to divide and when to stop dividing. It appears that, as a cell progresses through the sequential steps of the cell cycle, it must pass certain checkpoints, without which further division and proliferation will not occur. A long-standing issue in developmental biology has been what determines the initiation of cell division, the progression to a new cycle of division, and the cessation of division. Recent molecular studies have indicated the importance of specific gene products or factors that are involved in regulating certain aspects of the cell cycle and the sequential passage of a dividing cell through its various stages. For example, studies in yeast have shown that the product of the cdc2 gene, called p34cdc2, must be phosphorylated for the cell to pass from the G1 into the S phase (27). The mammalian homolog of the cdc2 gene also appears to serve such a checkpoint function (27). Phosphorylated p34cdc2 is itself an active protein kinase. A number of genes have been isolated that regulate the state of phosphorylation of p34cdc2 in both yeast and vertebrates (27). However, although it is likely that the division of neuronal precursors in the brain are also regulated by a similar mechanism, there is little data on what gene products are present in ventricular zone cells that may regulate the cell cycle. Moreover, there is presently no molecular data on what triggers a neuron to become terminally postmitotic.

As discussed previously, cells are thought to migrate to their final destination along radial glia cells. Understanding neuronal migration will require knowledge about the molecular relationship between the radial glial cell and the migrating neuron. Cloning of the genes involved in mouse mutants, for instance reeler and weaver mutants, that show migration defects might be a first step in understanding that relationship. One molecule that is missing or defective in weaver mouse granule cells is astrotactin. Antibodies against astrotactin, but not against other cell-adhesion molecules such as L1 or NCAM, were able to block the adherence of granule cells to radial glia in culture (17). Thus, astrotactin is possibly involved in neuron–glia interactions.

A number of substances that play important roles in axonal outgrowth have been identified. In Drosophila, certain genetic mutants show axonal path-finding defects, either choosing the wrong tract to follow or simply not growing (16). The genes encoding the cell-adhesion proteins neuroglian and the fasciclins-1, -2, and -3 were shown to underly some of the path-finding mutations (3, 16, 50, 60). Biochemical analyses showed that these proteins engage in homophilic adhesion; that is, the surface protein binds to itself on the surface of another cell. Axons may find the correct tract by matching the homophilic adhesion molecule on its surface with the tract that expresses the same homophilic adhesion molecule.

In the vertebrate spinal cord, certain axons must cross the midline. The floor plate is thought to release a substance that causes the axon to grow toward the midline. These axons express the cell adhesion molecule Tag-1 on their surface while they are on the same side of the spinal cord as the parent cell body. After crossing the midline, however, the axons no longer express Tag-1. Instead, they express another cell-adhesion molecule called L1. Both Tag-1 and L1 promote neurite outgrowth (6), suggesting that these molecules act as site-specific guides for axonal pathway formation.

Another class of substances that are important in the establishment and maintenance of appropriate patterns of cell growth and connectivity in the brain are the so-called target-derived neurotrophic growth factors, the most extensively studied of which is nerve growth factor (NGF) (see reviews in refs. 8, 52, and 55 and also the chapter by Patterson, this volume). Related members of the NGF family are brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3), NT-4, and NT-5. These substances are secreted by the innervated targets of nerve fibers, taken up into the terminal via a receptor-mediated process and transported retrogradely to the cell body, where they serve multiple functions to promote the survival, growth, and differentiation of neurons. The synthesis and secretion of neurotrophic substances is influenced by many factors, but the effect of afferent activity is of particular interest. In cells that secrete NGF, expression is increased by excitation of the cells and is decreased by neural inhibition. Excitatory inputs to a target cell will thus activate NGF secretion and promote the survival of other afferents to the same cell. Likewise, the activity of a cell that is dependent on NGF will itself influence the production of NGF in its target, and thereby promote its own survival. This may explain in part the observation that neuronal activity promotes survival and the need for a cell to establish functional synaptic connections to survive. The various NGF-like substances, although related, show a great deal of specificity in that they are differentially distributed, influence different populations of neurons, and interact with different receptors.

In terms of molecular mechanisms, the maturational transitions that take place during differentiation at either the cell or regional level can be attributed in part to both cell-autonomous (i.e., intracellular) and nonautonomous (i.e., intercellular) factors. Gene transcription factors are an example of cell-autonomous agents, whereas cell-to-cell signaling molecules and morphogens are examples of nonautonomous factors.

In the development of the Drosophila nervous system, there is a cluster of genes, called proneural genes, that when mutated cause an alteration in the number of sensory neurons in the periphery (14, 22). This set of genes codes for a class of transcriptional regulatory proteins that contain a common domain called the "basic-helix-loop-helix" (bHLH) domain. Studies of these and other bHLH proteins have shown that they are intimately involved in specifying particular cell fates (57). For example, when MyoD (a mammalian muscle determination gene) is induced in myoblasts, the myoblasts mature into multinucleate muscle cells. Two mammalian homologs of the Drosophila genes have been cloned and have been found to be expressed in many parts of the nervous system during development (23, 28). These are likely to control some aspect of cell differentiation in the mammalian nervous system during development.

Also in Drosophila, there is a class of genes that play a role in body segmentation and in specification of body parts. These genes share a common 180 nucleotide motif called a homeobox that encodes a 60 amino acid protein "homeodomain" (12). Homeodomain-containing proteins mediate their developmental effects by regulating the transcription of many target genes, and are expressed with an orderly regional or segmental distribution. There have been over 30 mammalian homologs of homeobox genes cloned, and there is evidence that these mammalian homeogenes play similar developmental roles as their counterparts in flies, controlling the identity of particular segments (25, 24). A subset of mammalian homeobox genes show a distribution of expression that parallels specific organizational patterns in the nervous system (20, 25), such as layer-specific expression in the cerebral cortex (18), or differential dorsal-ventral expression in the developing forebrain and diencephalon (40, 42). These genes may thus participate in the regional differentiation of the developing brain.

Cells can also differentiate based on nonautonomous factors, such as secreted proteins. During early neural development, the notochord, a mesodermal derivative, induces the overlying neural plate, an ectodermal derivative, to form nervous tissue. Transplantation studies have shown that the notochord induces the floor plate to release a diffusable factor, that has not yet been identified, which causes the overlying spinal cord to form cells of the appropriate types, such as motoneurons (38, 59). Other factors are important later in development. For instance, a homozygous null mutation in the Wnt-1 gene, which codes for a secreted protein that is normally expressed in the hindbrain, led to abnormal hindbrain development in mice (33, 53 ). It is likely, therefore, that this and other related Wnt gene products are important in cell determination via their intercellular signaling properties in the developing brain.

Certain substances can induce differential patterns of morphologic maturation depending on a spatial concentration gradient. Retinoic acid (RA) or its metabolites are thought to act as such morphogens in vertebrate development. The receptor for RA belongs to the steroid receptor superfamily of genes (15), which alter the transcription of target genes after binding their appropriate ligand. In cell culture, different concentrations of RA can have different effects on the expression of certain homeobox genes (48). During development of the chick wing bud, RA can cause a mirror duplication of wing structures in a concentration-dependent manner (49) and causes anteroposterior transformations in the frog nervous system and in the mouse vertebral column (7, 26). The concept of a graded morphogen is an important one in mammalian developmental biology, and even though RA is a likely candidate, it is not a proven one, and there are few, if any, others.

Plasticity of neurons and their connections have been studied in many systems. The molecular studies of desensitization and sensitization in the invertebrate sea slug Aplysia are likely to have important ramifications in our understanding of brain development. However, we focus on the role of correlated activity as an important determinant during brain development. Tetrodotoxin (TTX), a toxin from puffer fish, blocks the sodium channel in axons and thereby blocks action potentials. When TTX is injected into the pathway of the developing visual system, the normal segregation of the LGN into eye-specific layers and the visual cortex into ocular dominance columns is blocked (47). At a cellular level, TTX prevents the normal pruning of the extensive axonal arborizations that are present during development.

The N-methyl-D-aspartate (NMDA) receptor is a glutamate neurotransmitter receptor. When activated, it allows Na+, K+, and Ca2+ to flow through its channel. It is the only known glutamate-gated channel that allows Ca2+ to pass, eliciting a host of calcium-dependent effects, some of which may be involved in the molecular aspects of plasticity. However, activation of the NMDA receptor is dependent both on binding of glutamate and on the prior depolarization of the cell. Thus, Ca2+ only enters the cell if the receptor is activated in a correlated manner with another excitatory input. This property of the NMDA channel suggests that it may be involved in strengthening inputs that are temporally coactivated. If NMDA receptor antagonists are used to treat developing or regenerating visual system pathways, the normally precise retinotopic maps are disrupted (47). This disruption may be the result of the cell being unable to correlate normally synchronous inputs. The recent cloning of the NMDA receptor will help to unravel the role it plays in the normally developing brain (35) (see Exitatory Amino Acid Neurotransmission and Neuronal Growth and Differentiation Factors and Synaptic Placticity).

In this section, we have sketched the characteristics of various neurotrophic factors, transcription factors, cell signaling molecules, morphogens, cell adhesion molecules, and neurotransmitter receptors to show how brain development is beginning to be understood at a molecular level. In the next section, we speculate about the future directions of molecular research in brain development (see Neuroendocrine Interactions, for a discussion of gender-dependent brain development).

In the immediate future, new approaches will lead to a better understanding of the function of developmentally important genes that have been and will continue to be cloned and identified. There are now techniques for studying and manipulating mammalian genes transgenically. Transgenic mouse technology gives the ability to put an altered copy of a gene back into the genome, and, in the last few years, gene targeting by homologous recombination has allowed the generation of mice that have null mutations in specific genes (30). In the previous section, we discussed how null mutations in two Hox genes and one Wnt gene have helped underscore their importance in development. In the future, many mouse mutants will be created by gene-targeting approaches, and mammalian brain development can be understood more directly at a genetic level.

We believe that a major focus of future molecular research in brain development will be on the higher order relationship between the genome and the developing brain. The classic genetic code translates nucleotide sequence into amino acid sequence. However, this is but one level of genetic understanding. Higher order processes are involved in translating the genome, with all of its associated structure, into an organism. Understanding of these higher order processes requires, but is by no means limited to, characterizing regulatory elements and interactions of specific genes, characterizing the signals for alternative splicing of a gene, characterizing how chromatin is folded and bent to make certain regions transcriptionally active or unactive, and characterizing the role that DNA modifications, such as methylation, have in gene expression.

Researchers estimate there are approximately 100,000 different genes in higher mammals, perhaps 30% to 60% of which are uniquely expressed in the brain. At present only 5% have been cloned. The nature of cloning suggests that this 5% is skewed toward the most abundantly expressed genes. Thus, there are 95,000 genes that still remain to be cloned; these genes are likely to control the subtle aspects of brain development. The use of subtractive hybridization (39), which isolates genes based on their preferential expression in one tissue over another, and enhancer/promoter traps (11), which identify genes by trapping the regulatory elements of that gene using a marker transgene, are promising methods to isolate genes involved in brain development.

At the outset of this chapter we expressed our conviction that advances in developmental and molecular neurobiology could provide important new insights into the mechanisms underlying developmental psychopathology. We believe the following issues are pertinent to this discussion: (i) the relative importance of environmental and genetic factors in brain development; (ii) the need for fresh conceptualizations of developmental disorders, and (iii) the importance of biologic markers to researchers in this field.

First, we have learned that the normal development of the nervous system unfolds as a series of timed genetic events, the coordinated expression of which depends on appropriate environmental stimuli. We also know that the degree to which this genetic programming can be influenced by environmental events varies greatly across cell types, brain regions, and time periods. Deprived of light, the visual cortex will not develop properly, even though the genetic information for normal development is not affected. In rat pups deprived of maternal tactile stimulation, important metabolic enzymes are not activated, even though their structural genes are perfectly intact. Indeed such genetic–environmental interactions should be considered the norm rather than the exception.

Second, we believe a new approach is required to understand how perturbations of normal brain development can lead to severe childhood psychopathology, such as infantile autism. Autism is a devastating clinical disorder characterized by extreme social dysfunction, stereotypic behaviors, failure to develop normal language, and, in most cases, mild to moderate mental retardation. It begins in infancy; both its temporal course and outcome suggest it is a disorder of CNS development. But with few exceptions, no structural neurologic or neuropathologic deficits have been found in the brains of autistic children, and those that have been found are inconsistent and inadequate to explain the clinical deficits.

What new concepts emerging from basic research in developmental neurobiology can increase our understanding of disorders such as autism? One is the notion of the transient expression of important developmental factors or functions, either at the gene or cellular level, at critical time points in ontogeny, examples of which we have described in the sections above. Certain genes are actively expressed only during certain times in development, and we currently know the most about the class of transiently expressed transcription factors that themselves regulate the expression of other genes and are critical to normal nervous system development. Similarly, certain neurons in the developing subplate exist only transiently, helping developing cortical neurons establish appropriate connections, and then disappear. Both examples suggest a potentially important point regarding developmental disorders: despite their clinical severity, their underlying neuropathology must be very subtle, because we do not observe gross abnormalities with existing research methodologies (5). Aberrations in the developmental processes of migration and differentiation, establishing appropriate synaptic connectivity, axonal pruning, dendritic arborization, neurite regression, and programmed cell death all could give rise to considerable neurologic impairment (see chapter by Weinberger, this volume), but may not be obvious on autopsy or microscopic examination. Defects in developmentally transient processes could be even more subtle and less likely to be detected with current methodologies. However, specific hypotheses derived from basic models in developmental neurobiology could be applied to the understanding of clinical developmental disorders. For instance, in the case of transiently expressed factors, there are several candidate genes, derived from research in flies and mice, that could be tested for mutations in children with developmental disorders. Even though such a gene may no longer be expressed, mutations can still be identified using the standard techniques of molecular biology.

This brings us to the third point, the need for reliable and consistent biological markers that can be used to unambiguously classify developmental disorders. Clinically, rather than being single, defined syndromes, these disorders typically represent clusters of heterogenous and overlapping entities classified by sets of diagnostic criteria. Attempts to differentiate among them along such symptomatologic criteria, however refined and sophisticated, do not substitute for objective and conclusive biologic markers. Decades of biologic psychiatric research have failed to identify a single marker that can be used to diagnose, plan treatment, or predict the outcome of any psychiatric disorder. The lack of reliable, objective markers greatly hinders our ability to frame testable hypotheses that could lead to important advances. As new findings emerge from basic cellular, genetic, and developmental biologic research, it will be essential to incorporate these findings into a better clinical understanding of the neurodevelopmental bases of psychiatric disorders.

This work was supported by a program-project grant from the National Institute of Mental Health (MH 39437), by a Research Scientist Award (RDC) from the National Institute of Mental Health (MH 00219), by the endowment fund of the Nancy Pritzker Laboratory, and by the Spunk Fund, the Meyer Fund, the John Merck Fund, the Rebecca and Solomon Baker Fund, and the Edward and Marjorie Gray Endowment Fund (DAM).

published 2000