The Development of Brain and Behavior

Thomas R. Insel

In the course of a few months of gestation, the human central nervous system (CNS) changes from a microscopic band of embryonic neuroblasts to a 350 g mass with more than 109 interconnected highly differentiated neurons in the cortex alone. By age 5, the human brain reaches 90% of its adult weight (1.3 Kg) and the density of synapses in the cortex has already peaked and is beginning to decline. How this extraordinary growth with its intricate connectivity results in sensorimotor, cognitive, affective, and behavioral development is one of the great mysteries of biology. Developmental neurobiologists have begun to discover some of the mechanisms by which neurons migrate to their correct position, send efferents to the appropriate targets, and receive afferents from distant and adjacent sources. As various neurotransmitters and growth factors as well as specific pharmacologic treatments affect these processes, the insights of developmental neurobiology have recently become important for neuropharmacology. In addition, abnormalities of neuronal migration or differentiation might underlie various behavioral disorders, offering the hope that developmental approaches might prove important for understanding psychopathology and possibly offer new strategies for treatment.

This chapter is a brief summary of some of the major findings in recent studies of both neural and behavioral development. This is not a comprehensive review nor is it a careful critique of new findings. The intent here is to provide an organizational context with which the reader can approach more focused parts of this volume or from which the reader can search in other books on development to find in-depth coverage. Accordingly, three major perspectives are provided. First, some of the basic principles and recent directions in neuroanatomic development are summarized. Any reader interested enough to read through this section should know that two superb texts, one by Jacobson (1) and a second by Purves and Lichtman (2), will prove enormously helpful guides to developmental neurobiology, even for the non-anatomically minded neuropharmacologist. Next, aspects of neurochemical development are described. There are fewer general references here, although several focused reviews on the development of monoaminergic or neuropeptide systems are now available (3, 4). Finally, some recent new directions in behavioral development are noted (see ref. 5 for further reading).

After summarizing the anatomic, chemical, and behavioral approaches, this chapter addresses the epigenetic questions of how drugs (or other environmental influences) affect neural development, and conversely, how the stage of development alters drug response. These areas of research are of critical importance, not only for the practical therapeutic questions of how drugs work in children, but more fundamentally for determining how early experience can have longterm consequences for neural organization and behavior. The study of neural and behavioral development is one of the growth industries of modern neuroscience. Hopefully, this review will provide some impetus to search further into this exciting literature (see Molecular and Cellular Mechanisms of Brain Development, The Neurobiology of Treatment-Resistant Mood Disorders, Basic Biological Overview of Eating Disorders, Pharmacological Treatment of Obesity, and Early-Onset Mood Disorder).

The course of neural development can be traced through four overlapping processes: cell birth (neurogenesis), migration, formation of connectivity (including elaboration of processes, synapse formation, cell death, and axonal regression), and myelination (Fig. 1). In mammals, there are important species differences in the duration of each of these processes as well as their timing with respect to birth. In all mammalian species, however, the basic pattern of neural development involves exuberant production of neural elements followed by widespread elimination of neurons or regression of some of their axon collaterals.

It has long been known that the nervous system originates from embryonic ectoderm, but identifying the transition from ectoderm to neural precursor cells has never been easy. The identification of several neuron-specific proteins, such as the neural cell adhesion molecules (N-CAMs), has provided a clearer picture of when and where the CNS originates. This has been most elegantly demonstrated in frog embryos, where N-CAM can be detected as early as the 128-cell stage, 2.5–3 hours after the beginning of gastrulation (6). This initial differentiation of ectoderm into the neural plate occurs largely under the inductive influence of the underlying mesoderm which is believed to release some diffusible factor (possibly a steroid or peptide hormone or growth factor). Soon after the emergence of neuroblasts, the CNS begins to develop at a phenomenal rate with an orderly progression of cell birth and differentiation.

The time when specific neurons are born can now be detected with great precision using 3H-thymidine autoradiography (reviewed in ref. 1). These studies, in which labeled thymidine is permanently incorporated into the DNA of dividing cells, have suggested several basic principles of neurogenesis in rodents which are relevant to all mammals: (a) there is an orderly pattern of development with littermates injected at the same time showing nearly identical patterns of 3H-thymidine labeling; (b) in laminar structures (e.g., cerebral cortex) there is a characteristic "inside-out" pattern of origin—that is, neurons that are born later migrate past those born earlier to attain the most peripheral location; and (c) in a given region, large neurons are produced before small neurons, motor nuclei are born before sensory nuclei (at the same level of the neuraxis), and cells in phylogenetically older regions are born before cells in regions that are more recently evolved (e.g., ventral thalamus before dorsal thalamus).

It should also be noted that although neurogenesis is predominantly a prenatal event, in some regions neurons are generated postnatally in the mammalian brain. Granule cells continue to be produced in the olfactory bulb and dentate gyrus through postnatal development (7) and even into adulthood in some species (8).

The mechanisms by which neurons born in the ventricular zone reach their final position in either cortical or sub-cortical structures have been the focus of intense study recently. Early stages of migration to the cortex may occur rapidly in a relatively unguided fashion, but as the distance increases and new cells have to migrate through layers of older cells, the journey becomes more complex. Rakic has described the importance of radial glia for the normal guidance of cells born in the ventricular zone to reach their appropriate destination in the cortical mantle (9). Radial glia are found only during a distinct period of development and appear to differentiate later into fibrillary astrocytes. This mechanism for cell migration may account for the columnar organization of the neocortex which has been recognized with both functional and anatomic mapping. In addition, considerable attention has recently focused on the cortical subplate (10, 11), a transient band of cells that appears to (a) serve as a staging ground for arriving afferents, (b) provide a chemically-rich matrix through which cortical cells must migrate, and (c) offer an "axonal scaffold" for early cortical efferents. The cells in this cortical subplate disappear early in postnatal life, but may be essential for the appropriate organization of cortical architecture during the early phases of development.

It now appears that some forms of migration even occur in the adult brain. The cell adhesion molecules (especially the neuron specific N-CAM) are important for neural migration as well as neural induction. During development, N-CAM changes from a highly sialylated form to several isoforms with less sialic acid. In a recent study, Theodosis et al. demonstrated the persistence of the highly sialylated form of N-CAM in the adult brain, specifically in parts of the hypothalamus known to undergo a remarkable form of reorganization in adulthood (12). During lactation, magnocellular neurons in this region of the hypothalamus reversibly change their orientation due to retraction of the intervening glia and migration of neural processes to permit synchronous depolarization. This embryonic isoform of N-CAM is found specifically in the processes of magnocellular neurons in the adult hypothalamus.

The discovery of ectopic cells and migratory arrests in post-mortem tissue from patients with dyslexia, autism, schizophrenia, and fetal alcohol syndrome have suggested the hypothesis that abnormal migration may be of pathophysiologic importance (reviewed in ref. 13). In fact, migration shows considerable individual variation and the functional significance of even severely disordered migration remains uncertain. Studies of the reeler mutant mouse have been particularly instructive in this regard. This mouse shows extensive disorganization of laminar structures due to widespread deficits in neuronal migration, yet connectivity of hippocampal and neocortical neurons is generally intact (14). Pathology within the late-developing cerebellum is probably responsible for the tremor, hypotonia, and ataxia characteristic of the reeler mutant (see Molecular and Cellular Mechanisms of Brain Development, The Neurobiology of Treatment-Resistant Mood Disorders, Pharmacological Treatment of Obesity, and Phencyclidine).

Most research in developmental neurobiology over the past three decades has been focused on the mechanisms of how axons find their targets and how the mature pattern of synapses forms. Although much has been learned about axonal growth, heralded by studies of the axonal growth cone and its unique protein markers such as GAP-43, the mechanisms by which axons find their targets remain mysterious. Axonal outgrowth from cortical neurons is a surprisingly early event, beginning even before a neuron has finished migrating and often reaching the appropriate target even when migration is obstructed (15).

It is clear that the brain produces many axons that ultimately are eliminated, but the reason for this ostensibly inefficient strategy is still a puzzle. Three examples of this phenomena are worth noting. Initially, there is a diffuse connection between cortical fields via the corpus callosum. In the monkey, these axon collaterals are eliminated in late fetal life (about one month before birth) to result in the adult pattern in which each cortical neuron sends a single axon to a specific contralateral region (16). Within a hemisphere there are multiple cortical–cortical connections that are eliminated about the same time (17). Perhaps most remarkable, layer V neurons in the visual cortex transiently send axons via the pyramidal tract to the spinal cord (18). In each of these cases, the neurons survive, but axons to some targets are retracted while others are preserved. It should be noted that all of these studies require the injection of retrograde tracers into a presumed target site several days prior to examining the animal for cortical cells of origin. As a result, we still know very little about the development of circuitry in the human brain. The development of the carbocyanine dye, DiI, that can be used as a retrograde tracer in post-mortem tissue might allow for similar studies in the developing human brain (see e.g., ref. 19), but this technique is slow and may be of limited usefulness for long pathways or tract-tracing in the more mature brain (20).

Axonal regression has been attributed to trophic factors at the target, trophic factors in the region of the cell body, or activity dependent changes which favor some projections over others. The existence of soluble growth factors, such as NGF, has been known for many years. Recently, several new protein growth factors, such as BDNF, CNTF, and IGF-1 have been identified (reviewed in ref. 21). Perhaps of greatest interest is the identification and cloning of a novel factor, glial cell-line derived neurotrophic factor, GDNF, which is not only a highly potent growth factor, but one that thus far appears specific for dopamine cells in dissociated cell cultures from rat midbrain (22). One might predict that the discovery of related specific growth factors (and their receptors) for other neuronal phenotypes, such as serotonin, GABA, and glutamate, will have an enormous impact on our understanding of neural development as well as offering a new class of potential therapeutic agents for degenerative diseases of the nervous system (see Neuronal Growth and Differentiation Factors and Synaptic Placticity).

The process of over-production followed by elimination occurs for neurons as well as their axons. In some regions of the rat brain, as many as 50% of neurons die before birth (23). Studies of cell lineage have been done most extensively in the transparent nematode Caenorhabditis elegans, which consists of 1090 postmitotic somatic cells of which 959 survive into adulthood (24). The developmental death of 131 cells (including 20% of all presumptive neural cells) is surprisingly invariant, suggesting that early death is destined or programmed for specific cells. It is now clear that "programmed" cell death is characteristic of vertebrates as well as invertebrates and occurs in many organs including the brain. Of course, the problem with studying this phenomenon is recognizing which cells will die before the terminal event.

Almost two decades ago, Kerr et al. described specific morphologic changes that precede death, such as cell shrinkage, condensation of chromatin, and changes in membrane morphology (25). They called these highly predictable changes apoptosis (Greek for "a flower losing its petals"). More recently, various investigators have identified those genes that are expressed with apoptosis to examine the molecular basis of programmed cell death during development. It is now clear from studies in C. elegans that there is a family of genes associated with cell death (ced-3 and ced-4) and another associated with survival (ced-9). Recently, the oncogene bcl-2 has been hypothesized to function in mammalian neurons as a homologue to ced-9 of the nematode. Although there are still questions about the extent to which cell death within the mammalian CNS is "programmed," one can imagine how normal development might be altered by changes in the expression of genes for either cell death or survival. Conversely, the possibility that oncogenic transformation is initiated by the failure of genes normally associated with apoptosis may prove of great therapeutic significance.

In addition to increased numbers of neurons and axons, synaptic density is higher during development than adulthood. Synaptogenesis begins very early in development (as soon as axons reach postsynaptic cells) and continues well into postnatal life. In the human visual cortex, the peak of synaptic density is at age 8 to 11 months with a gradual decrease to adult levels by puberty (26). Changes in synaptic density, while ostensibly impressive, may be neither ubiquitous nor informative. Neurons which receive only a single afferent axon (such as cells within the submandibular ganglion) show a gradual ontogenetic increase in the number of synapses although the number of axons innervating each cell decreases (27). In cortical neurons, where hundreds of axons innervate a given neuron with its extensive dendritic arbor, changes in synaptic density involve several different processes, including elimination of some axonal input, increased numbers of synaptic boutons on surviving axons, and developmental elaboration of the dendritic arbor (28). Synaptic density is thus a consequence of many different changes in connectivity, and interpretation of changes in density depends on which specific aspect of afferent input is altered.

With the current emphasis on regressive events in brain development, it is easy to forget that the brain is undergoing tremendous growth during postnatal life. The human brain weighs 350 g at birth and nearly 1.2 Kg at 5 years of age (29). This growth can be attributed partly to increased neuropil from the elaboration of dendrites, but also to increased myelin formation by oligodendrocytes. In the rat, myelin formation begins after the period of cellular proliferation and migration and extends into adulthood. Jacobson estimates that myelin content increases 1500% between 15 days and 6 months after birth in the rat brain (1). In the human brain, myelination continues at least through the first decade of life, with some phylogenetically recent structures only appearing fully myelinated in the second decade (30). Although the time course of myelination suggests that this phase of neural development would be most vulnerable to postnatal environmental influences, little is known about either the influence of epigenetic factors on myelin formation or the functional consequences of subtle changes in either retarded or reduced myelination.

Concurrent with the dramatic morphologic changes of neural development, neurotransmitters and their receptors fluctuate, in some cases markedly, during ontogeny. Indeed, for certain neurochemical systems, the changes in regional expression that occur normally in development dwarf any changes that are observed with experimental manipulations in adulthood. Phenotypic plasticity, that is the capacity of a cell to shift from one neurochemical phenotype to another, has been known in sympathetic neurons (which shift from noradrenergic to cholinergic) and adrenomedullary cells (which shift from noradrenergic to adrenergic), but now appears to occur within the CNS as well (31). As a first approach to this broad area, the development of neurotransmitters, their receptors, and their effectors in the brain will be briefly summarized.

Monoaminergic neurons are born early in the course of CNS development (Fig. 1). Serotonin, norepinephrine, and dopamine are detectable in nuclear groups by E13 in the rat brain (prior to the peak of cortical neurogenesis) (32). Monoaminergic cells from the brainstem send off processes quickly and are among the first afferents to colonize the cortex, migrating to the wall of the telencephalon (33). Norepinephrine has long been implicated in cortical development not only because of this early invasion of the cortex [E18 in the rat (34) and E70 in the monkey (33)] but because of several different forms of experimental evidence linking noradrenergic input to synaptic plasticity in the visual system (reviewed in ref. 1). Serotonergic fibers also invade the cortex early and have been implicated as trophic for the developing cortical architecture (reviewed in ref. 32). Serotonin in the early postnatal neocortex is present at roughly twice the adult concentration (35). Autoradiographic labeling of serotonin uptake sites has demonstrated the transient appearance of serotonin terminals specifically in the barrel fields of the developing somatosensory cortex (36). While the function of this highly specific distribution has not yet been demonstrated, the timing and the distribution of these transient fibers suggest a role in the organization of the arriving thalamocortical axons.

The ontogeny of amino acid neurotransmitters such as GABA and glutamate appears somewhat different from the picture with monoamines. Glutamate has been detected mostly postnatally in the rat brain, reaching adult concentrations by postnatal day 15 (37). Remarkably, careful immunocytochemical studies for glutamate in the embryonic brain have not been published. In the rat visual cortex, GABA immunocytochemically-identified cells appear to be born from E14 to E20, concurrent with the appearance of the pyramidal cells that they innervate (33). GABA immunoreactivity can be detected even earlier in less differentiated neural elements. Indeed, GABA may be the earliest neurotransmitter to emerge in the CNS, although its role as a developmental signal may turn out to be quite different from its traditional role as a fast, inhibitory neurotransmitter.

Acetylcholine, as detected by ChAT immunoreactivity, appears relatively late in development. In the cortex, most ChAT-positive cells are not observed until the second or third post-natal week (38, 39). However, a transient expression of ChAT-positive cells has been noted in the cortex as early as E17, disappearing by P1. Acetylcholine, like norepinephrine has been implicated in visual cortical plasticity (40).

Neuropeptides during development are also expressed in a unique distribution in several brain regions, although some, like oxytocin, are not fully processed until the first postnatal week. In the cortex of the embryonic macaque, for instance, proenkephalin, tachykinin, and somatostatin appear transiently (41). It is not clear if the cells expressing these peptides change phenotype or if these cells are eliminated during cortical development (see Structure and Function of Colonergic Pathways in the Cerebral Cortex, Limbic System, Basal Ganglia, and Thalamus of the Human Brain, Development of Mesencephalic Dopamine Neurons in the Nonhuman Primate: Relationship to Survival and Growth Following Neural Transplantation, Central Norepinephrine Neurons and Behavior, Anatomy, Cell Biology, and Maturation of the Serotonergic System: Neurotrophic Implications for the Actions of the Psychotropic Drugs, General Overview of Neuropeptides).

Given the concurrent reorganization in both axonal and dendritic branching, it is not surprising that the distribution of receptors in development might appear quite different from the patterns observed in adulthood. Even in developing muscle where cholinergic receptors can be detected before the arrival of cholinergic innervation, the distribution of a-bungarotoxin binding is initially widespread, migrating in the membrane to the point of innervation (42). Although monoaminergic receptors and amino acid receptors show some local differences, receptor binding in these systems generally develops monotonically with an overshoot (relative to adult levels of binding) in the second or third postnatal week of the rat. For 5-HT2 and 5-HT1c receptors, the developmental changes in binding have been associated with concurrent changes in mRNA levels, suggesting that binding increases represent increased receptor synthesis (43). In the case of peptide receptors, particularly in the cortex, the degree of overshoot is far more dramatic, amounting to severalfold increases relative to adult concentrations (Fig. 2 and Fig. 3).

The function of these transient receptors remains unclear. Indeed, these receptors may not be functional in the sense of being linked to an intracellular effector, but nevertheless may serve some other role as cell surface markers. In any case, three general rules should be appreciated: (a) receptor binding may develop independently of innervation; (b) the pattern of receptor binding is remarkably diverse with specific brain regions often showing intense binding transiently in the first through third postnatal weeks (in the rat); and (c) the expression of binding precedes receptor-effector coupling (45).

In addition to the unique patterns of receptor distribution, one also finds patterns of coupling to second-messengers during development which are absent or less obvious in the adult brain. For instance, a 5-HT receptor (5-HT4) in the superior colliculus coupled to adenylate cyclase was first described in the neonate (46). Similarly the high levels of phosphoinositol generated in the first three postnatal weeks in rat hippocampal slices exposed to glutamate provided the first recognition of a metabotropic receptor distinct from the channel-linked site originally thought to be the only form of glutamate receptor in the adult brain (47).

The developmental patterns of second messengers, such as various G proteins, is of note independent of their coupling to receptors. Fig. 4 shows the developmental pattern of stimulation of cyclase in the rat brain. Clearly, the capacity for neurotransmitters to function will depend on the maturation of their respective intracellular effectors (see also Cholinergic Transduction, Signal Transduction Pathways for Catecholamine Receptors, Serotonin Receptors: Signal Transduction Pathways, and Intracellular Messenger Pathways as Mediators of Neural Plasticity).

Although the most dramatic discoveries in development have been in the realm of neuroscience, research into how behavior becomes organized in development has also undergone some major conceptual shifts in the past decade. Three new areas of research should be noted by neuropharmacologists.

Viktor Hamburger, who is arguably the father of modern developmental neurobiology, became in his later years, a student of behavioral development. In classic studies carried out in the 1960s, Hamburger showed that motor development in the embryonic chick precedes sensory stimulation (49). Recently, Smotherman and Robinson have begun to follow Hamburger's lead by studying fetal behavior in mammals (50). Studies in the exteriorized rat fetus have demonstrated a surprising capability for rapid associative learning, at least some of which is mediated by opioid peptides.

A major shift in developmental psychology has been led by the work of Kagan and colleagues (51). These investigators have demonstrated the relatively consistent expression of specific clusters of personality traits (i.e., temperaments) across development. One cluster, labeled as behavioral inhibition, is operationally defined as a tendency to be shy, timid, and constrained in novel situations provided in a laboratory setting. Children who appear inhibited at 21 months and who remain inhibited at age 7.5 years show increased rates of anxiety disorders and have parents with increased rates of anxiety disorders. The focus on temperament, as opposed to manifest symptoms, provides an opportunity to study vulnerability to various mental disorders. Moreover, the search for specific genes that predispose to mental illness may ultimately find that the phenotype is not a specific psychiatric disorder, but a more general temperamental predisposition to develop a form of psychopathology.

The field of attachment theory, which has been at the center of ego psychology for the past three decades, has undergone an important conceptual shift from a focus on a single process that affects the infant to a focus on the interaction of caregiver and infant. Hofer has described a number of "hidden regulators" which maintain this dyad at a physiological as well as behavioral level (52). The concept is that the caregiver provides several different forms of comfort—nutritional, tactile, thermal, and olfactory—each of which regulates a specific aspect of the infant's physiology (Table 1). As a result, withdrawal of one aspect of care (as opposed to the more global concept of "loss") could lead to dysregulation of a specific physiologic response (53). For instance, Levine and colleagues have described how normal maternal contact regulates the infant rat's hypothalamic-pituitary-adrenal axis. With prolonged (i.e., 24 hours) separation from maternal contact (in the presence of warmth, nutrition, and maternal odors), plasma corticosterone increases several-fold (54).

For the neuropharmacologist (as for the psychobiologist), the two major questions provoked by this generation of studies in development are: (1) How do environmental events (hormones, drugs, stressors, infection) affect neural development? and (2) How does the stage of neural development affect the response to environmental events? In fact, neither question admits to a satisfactory answer, but there are useful data for consideration in both cases.

Many of the processes of neuroanatomic and neurochemical development described above are modified by activity in the developing CNS. The formation of ocular dominance columns in the cat visual system (55), somatosensory representations of both the thalamus and the cortex (56), and the alignment of auditory and visual maps in the bird's optic tectum (57) are perhaps the most intensively studied forms of experience-dependent plasticity in development. The period during which experience can alter the neural representation defines the sensitive period. In the case of the owl's visual-auditory maps, Knudsen and Knudsen have defined the sensitive period by elegant experiments in which either visual or auditory input is altered (58). The critical period, which is often used interchangeably with the sensitive period, is more strictly defined as the developmental interval during which normal maps can be restored following a period of deprivation.

Neurophysiologic and anatomic studies in kittens by Hubel et al. (59), Singer (55), and Spinelli et al. (60) have demonstrated that monocular deprivation leads to lasting changes in ocular dominance columns in the cortex and to smaller neurons in the lateral geniculate nucleus. Remarkably, experience confers a selective advantage for a particular aspect of vision. For instance, in the normal kitten visual cortex, virtually all neurons respond to visual orientation, with a roughly equal number showing preference for horizontal, vertical, and oblique patterns. When a kitten is reared in an environment that restricts vision to a single orientation, the majority of cells adopt preference for this orientation, apparently through the repeated excitation of the circuits subserving this pattern, with the consequent elimination of connections for the other visual orientations (60, 61). Within this specific sensitive period, certain connections within the mammalian CNS develop in an experience-dependent fashion with lasting morphologic and behavioral consequences.

Similar conclusions might be drawn from a literature on the influence of the complexity of the environment on neural development. Studies in primates as well as rodents have described a variety of morphologic effects resulting from rearing animals in a complex rather than simple environment. These effects include cortical thickening, alterations in cortical dendritic branching, increases in the number of spines per dendrite, and changes in the morphology of synaptic contacts (reviewed in ref. 62).

How experience sculpts the developing neural circuitry is one of the most intriguing questions in developmental neurobiology. D. O. Hebb, in his classic volume, The Organization of Behavior, formulated the concept that neural activity in some way reinforces specific circuits, so that when cell A excites cell B some change takes place in both cells to increase A's efficiency at firing B (63). In this way, synapses that are active become more efficient and synapses that are inactive might be ultimately eliminated. This basic idea has been elaborated in a number of ways to suggest that competition for some trophic factor, bursts of asynchronous activity, or changes in the stability of membrane bound receptors mediate the effects of neural activity on synaptic survival (see ref. 2) for a more detailed discussion). These mechanisms may not be mutually exclusive—all of them recognize the importance of competition, of the timing of activity, and of the remarkable magnitude of the effects in an organ where as many as 50% of the cells may be eliminated during a brief period of development. Norepinephrine, acetylcholine, NMDA, and GABA have all been implicated in this process (see above).

Just as stimulation of the visual system at a sensitive period of development appears essential for subsequent vision, exposure to gonadal steroids at particular phases of development is essential for the subsequent expression of normal adult sexual behavior. This concept was first introduced by experiments 40 years ago demonstrating that testosterone administration in developing guinea pigs could alter sex behavior in adulthood (64). Similar results have been reported in various species from frogs to primates, leading to the well-accepted notion that steroids, which functionally resemble growth factors, may have effects in development that are quite distinct from their effects in adulthood (65). These long-term effects of gonadal steroid exposure in development have been termed organizational effects as exposure appears to organize the system for later responsiveness. For instance, androgens administered to female rat pups in the first week of postnatal life confer an altered sensitivity to subsequent physiologic levels of gonadal steroids such that masculinization (enhancement of mounting behavior) and defeminization (reduced capacity for lordosis) are manifested after puberty (66). The first postnatal week is normally associated with a transient increase in hypothalamic receptors for gonadal steroids, possibly providing for this sensitive period (67). An analogous "feminization" of adult sexual behavior is evident in males that are not exposed to adequate levels of testosterone prenatally (66). Prenatal stress is also associated with subsequent feminized sexual behavior in male offspring, an effect which may be mediated by stress-induced decreases in testosterone concentrations in fetal plasma (68, 69).

Although there is little question that gonadal steroids in development can have longterm consequences for adult behavior, the mechanisms for these organizational effects are not entirely elucidated. Some of these effects may be related to trophic actions, as estrogen induces neurite outgrowth in hypothalamic cells in vitro (70). In addition, androgens appear to reduce cell loss from certain targets during the normal period of neuronal elimination (71). Gonadal steroids appear essential for the full development of the sexually dimorphic area in the preoptic region of the hypothalamus—an area which appears markedly more cellular in the male brain (reviewed in ref. 72). Although the functional role of this sexually dimorphic area remains unclear, analogous dimorphisms are also found in the human hypothalamus (73, 74) and may also be sensitive to neonatal sex steroid levels in man. The reader should note that in spite of a considerable literature on the functional consequences of increases or decreases in gonadal steroids in development, at present there is little evidence that gonadal steroids have cellular or molecular effects in the infant distinct from their effects in the adult (75).

Recent evidence suggests that gonadal steroids are not the only compounds to have longterm effects on the organization of neural systems: various neurotransmitters and neuromodulators appear to have similar properties during development. In addition, the interaction of neurotransmitters with their receptors may have different consequences in development. In adulthood, administration of an agonist for a neurotransmitter receptor generally leads to a compensatory downregulation and administration of an antagonist may lead to upregulation of receptor number. These changes are usually rapid (reflecting either local changes such as membrane internalization of the receptor or genetic changes in receptor transcription) and reversible, suggesting that the cell has some homeostatic control over receptor sensitivity. Several studies have demonstrated that during ontogeny, when receptor number (i.e., genetic control), receptor regulation (i.e., the presumed homeostat), and second messenger or channel coupling are incompletely developed, exposure to agonists or antagonists may have effects which are paradoxical and enduring. For instance, offspring of mothers treated with haloperidol show lasting decreases in dopamine receptors (76). Curiously, exposure to haloperidol in the early postnatal period results in increased not decreased numbers of dopamine receptors (76). As another example, morphine given to rat pups from day 1 until day 7 confers a lasting increase in mu opiate receptors as well as behavioral analgesia (77).

Similar longterm consequences of neonatal drug administration have been reported for several compounds including neuroleptics, antidepressants, substance P, vasopressin, benzodiazepines, and corticotropin releasing hormone (reviewed in ref. 44). These actions, which have been variously described as organizational (78) or chemical imprinting (44) effects, all suggest that exposure to an agonist during the neonatal period induces enduring functional increases in receptor responsiveness. As noted above, receptors for many of these systems show a profound "overshoot" during development, but it is not clear that this is causally related to these "imprinting" effects.

While the data are still lacking, reasonable hypotheses for the chemical imprinting effects include increases in survival of post-synaptic cells with a given receptor (79), increases in processes on those cells that do survive (80, 81), and increases in coupling to a second messenger or ion channel (48). Whatever the mechanism, it appears that classical neurotransmitters might function as neurotrophic factors in development with organizational effects similar to what has been previously reported with gonadal steroids.

Although infants are uniquely sensitive to certain environmental insults, they are either relatively spared or show enhanced recovery from others. Kennard was perhaps the first to demonstrate behavioral sparing following lesions to the motor cortex in neonatal monkeys (82). Certain excitotoxins, such as quinolinate, which cause marked neurodegeneration in the adult rat hippocampus and striatum, have little effect when injected during the first two post-natal weeks (83). Neonatal administration of the catecholamine neurotoxin, 6-hydroxydopamine, causes an enduring depletion of norepinephrine and dopamine associated with locomotor hyperactivity, but several behavioral consequences observed in lesioned adults are not seen when the toxin is given to developing rats (84). And rat pups with extensive depletion of serotonin following the selective neurotoxin methyl,diethylmethamphetamine (MDMA) show virtually no change in thermoregulation, weight gain, or locomotor activity (85). In each case, the targeted system is still developing and one might assume, as with structural lesions, that plasticity persists through this period of development. Plasticity however is a description, not a mechanism. How these systems adapt remains an open question (86). How the adult organism might be induced to adapt similarly to neurochemical insults is clearly an exciting future focus for research.

The developing organism responds in a unique fashion not only to neurotoxins, but also to various psychopharmacologic agents. The a-2 adrenergic agonist clonidine provides one of the most remarkable examples of a developmentally determined drug response. In the rat, clonidine increases locomotor activity from Day-1 to Day-14 postnatal and dramatically decreases locomotor activity thereafter (87). Both the stimulating and the suppressant effects are dose-dependent and both are blocked by yohimbine (author's unpublished data). This apparent reversal of clonidine's locomotor effects may be attributed to the late development of a-2 receptors in motor pathways (approximately Day-21 postnatal) (88). The early stimulatory effects of clonidine have not been fully explained. Destruction of presynaptic a-2 receptors potentiates both the stimulatory and the inhibitory effects on locomotion (89). The transient expression of a-1 receptors in the globus pallidus of the rat during the first two postnatal weeks matches the temporal course of the stimulatory effect of clonidine, but this correlation may be only coincidental (90).

Similar developmental patterns have been described with a number of drugs, emphasizing the importance of age on the "normal" pharmacologic response. In addition to the ontogeny of receptors and their effectors, metabolic enzymes and the blood-brain barrier are maturing postnatally, ensuring very different pharmacokinetic profiles during development (91). While this point is generally appreciated by neuropharmacologists, there has been less recognition of these developmentally idiosyncratic responses as "experiments of nature" providing an opportunity for studying the mechanisms of drug response. For example, the study of how adrenergic receptor expression and coupling in discrete brain regions changes between days 14 and 21 should provide important insights into those pathways that mediate the locomotor suppressant effects of clonidine.

As the research base grows, the fields of developmental neurobiology and developmental psychobiology seem to grow further apart. We have begun to address phenotypic plasticity and regressive events with more powerful cellular and molecular techniques, but unfortunately, the functional importance of these events has become less of a focus of inquiry. Similarly, the problems of how behavioral patterns are organized through development has been remarkably uninformed by the rapid advances in developmental neurobiology.

Neuropharmacology may provide one of the bridges to link these two disciplines. For instance, studying the neurochemical events that underlie experience-dependent changes and determining the range of functions of neurotransmitters in development are important goals for neuropharmacology that should advance our understanding of both the neurobiological and behavioral aspects of development. Moreover, the immature organism provides a unique opportunity for investigating the mechanisms of drug action. By a careful choice of ages, the investigator can study receptor gene expression, receptor-effector coupling, and cellular consequences of activation in tissues in which various aspects of cellular machinery are developmentally modulated. Finally, the investigator interested in the pathophysiology of major mental illnesses may find abnormalities in one or more of the aspects of neural development which ultimately result in the emergence of pathological behavior—an observation that will bridge the behavioral and anatomic aspects of development.

published 2000