| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Biology, University of Washington, Seattle, Washington
ABSTRACT I. INTRODUCTION II. COREGULATION OF ION CHANNEL DEVELOPMENT, SPONTANEOUS ACTIVITY, AND ACTIVITY-DEPENDENT DEVELOPMENT A. Oocyte Maturation and the Block to Polyspermy 1. Nature and developmental function of spontaneous activity 2. Relationship to channel development B. Early Postfertilization Changes and Selective Channel Elimination C. Retina and Refinement of Visual Connections 1. Retina: nature of spontaneous activity 2. Retina: refinement of retinal ganglion cell connections by activity 3. Retina: outgrowth of retinal ganglion cell axons 4. Retina: dendritic patterning of RGCs 5. Retina: relationship to channel development D. Hippocampus and Excitatory GABA Responses 1. Nature of spontaneous activity 2. Developmental roles of spontaneous activity 3. Relationship to channel development E. Cerebral Cortex and Coordinated Na+ and Resting Channel Development 1. Cortex: nature of spontaneous activity 2. Cortex: developmental effects of spontaneous activity: migration 3. Cortex: developmental effects of spontaneous activity: later differentiation of neurons and circuits 4. Cortex: activity and neural migration: relationship to channel development 5. Cortex: perinatal spontaneous synchronized activity: relationship to channel development F. Cerebellar Neurons, Ca2+ Currents, and Neuronal Migration G. Hindbrain and Synchronized Activation of Motor Neurons and Cranial Nerves 1. Nature of spontaneous activity 2. Developmental roles of spontaneous activity 3. Relationship to channel development H. Spinal Cord and Emerging Motor Patterns 1. Nature of spontaneous activity 2. Developmental roles of spontaneous activity 3. Relationship to channel development I. Cochlear Hair Cells and the Loss of Excitability After Activity 1. Nature of spontaneous activity 2. Developmental roles of activity 3. Relationship to channel development J. Dorsal Root Ganglion Cells, Myelination, and Cell Adhesion Molecules 1. Nature of activity 2. Developmental roles 3. Relationship to channel development K. Amphibian Spinal Neurons, Transmitter Phenotype, and Low-Frequency Spontaneous Activity 1. Nature of activity 2. Developmental roles 3. Relationship to channel development L. Ascidian Muscle, Inward Rectifier, and Activity-Dependent Ion Channel Development 1. Nature of spontaneous activity 2. Developmental roles of spontaneous activity 3. Relationship to channel development M. Insect Neurons and the Refinement of Dendritic Trees During Metamorphosis N. Mammalian Muscle and Activity-Dependent Fusion of Myoblasts O. Amphibian Muscle and Multiple Windows of Activity-Dependent Development P. Cajal-Retzius Cells, Rohon-Beard Neurons, and Activity-Dependent Cell Death Q. Summary III. HOW SPONTANEOUS ACTIVITY CARRIES OUT ITS DEVELOPMENTAL FUNCTIONS A. Role of [Ca2+]i Transients 1. Amplitude 2. Frequency 3. Physical pathway of Ca2+ entry 4. Spatial distribution of Ca2+ entry B. Developmental Regulation of Intracellular Ca2+ Stores and Buffering C. Release of Developmentally Active Neurotransmitters D. Neurotrophins as Major Activity-Dependent Pathway E. Relationship Between Synaptic Plasticity in the Adult and Developing Nervous Systems IV. SOME PRINCIPLES OF HOW ION CHANNELS DEVELOP TO REGULATE SPONTANEOUS ACTIVITY A. Immature Voltage-Gated Channels With Properties Different From Their Mature Counterparts B. Immature Ligand-Gated Channels With Properties Different From Their Mature Counterparts C. Different Immature Channel Function Due to Different Ion Gradients Early in Development 1. Optimization 2. Coordination 3. Self-limiting nature of activity 4. Incompatibility of immature and mature properties D. Nonlinear Developmental Profiles of Channels That Create Early Periods With Unique Firing Properties 1. Transient channel expression: disappearing channels 2. Relative timing differences of the development of different channels 3. Late appearance of mature, low resting resistance E. Changes in the Spatial Distribution of Channels During Development F. Changes in the Coupling of Channels to Intracellular Events During Development G. Differences in Intracellular Trafficking of Channel Subtypes V. ACTIVITY-DEPENDENT ION CHANNEL DEVELOPMENT AS PART OF THE ESSENTIAL TRANSITION BETWEEN IMMATURE AND MATURE PHYSIOLOGICAL PROPERTIES A. Voltage- and Ca2+-Gated Channels B. Ligand-Gated Channels C. Summary VI. CLINICAL IMPLICATIONS OF ACTIVITY-DEPENDENT NERVOUS SYSTEM DEVELOPMENT VII. SUMMARY REFERENCES
 |
ABSTRACT |
|---|
 Top NextReferences |
|---|
|
I. INTRODUCTION |
|---|
|
TopPreviousNextReferences |
|---|
For this reason, the configuration of ion channels and receptors expressed at early stages of development are optimized to mediate spontaneous activity and unique to the early stages when this kind of activity occurs. This optimization creates the appropriate electrical waveform of spontaneous activity, synchronizes it among cells, and mediates the Ca2+ influx that transduces activity into developmental programs. In addition to optimization of immature ion channels and their incompatibility with many mature functions, two other general principles will arise repeatedly in this review. The first is coordination of the development of multiple channel types, both ligand- and voltage-gated, which must cooperate to create periods of spontaneous activity. The second is the self-limiting nature of spontaneous activity. The transition between this immature period of spontaneous activity and the mature, information processing functions of the cells is critical. It is managed in part by making the expression of mature ion channels and receptors, whose expression tends of terminate spontaneous activity, dependent on the spontaneous activity created by their immature predecessors.
The existence of distinct electrical properties early in development that are geared toward creating spontaneous activity has important clinical implications, especially in the field of pediatric seizure disorders.
In this review, we analyze how the patterns of ion channel development give rise to spontaneous activity and how that activity carries out its developmental functions. In section II we use several cell types to illustrate the wide variety of developmental processes regulated by spontaneous activity, and how the waveform of activity in different cells is regulated by the expression patterns of ion channels early in development. In section III, we discuss the mechanisms by which spontaneous activity controls developmental processes. In section IV, we propose some general principles that govern how the ion channels expressed at early stages differ from mature channels and how they regulate spontaneous activity. In section V, we discuss the critical role played by spontaneous activity in regulating the maturation of ion channels so that cells successfully make the transition between embryonic and mature signaling properties. Finally, in section VI, we discuss clinical ramifications of the idea that immature neurons have electrical properties that favor spontaneous activity.
|
II. COREGULATION OF ION CHANNEL DEVELOPMENT, SPONTANEOUS ACTIVITY, AND ACTIVITY-DEPENDENT DEVELOPMENT |
|---|
|
TopPreviousNextReferences |
|---|
A. Oocyte Maturation and the Block to Polyspermy
We often think of complex patterns of ion channel development as characteristic of the terminal differentiation of neurons and muscle cells. But they actually begin even before fertilization. The ways in which ion channel properties are modulated during development of oocytes can be understood in the same context that governs similar later events: by knowing the developmental function of electrical signals at these stages and asking how particular ion channels ensure that the properties of those signals are consistent with that function.
1. Nature and developmental function of spontaneous activity
In many organisms (including echinoderms, amphibians, nemertean worms, but not mammals), fertilization is accompanied by a large, long-lasting depolarization known as the fertilization potential (see Refs. 264, 265 for review). Fertilization potentials last from several minutes to more than 1 h, depending on the species. The fertilization potential is mediated by sodium, nonspecific cation, or chloride channels that are gated by the rise in [Ca2+]i that occurs at fertilization, with contributions at early times by voltage-gated Ca2+ or Na+ channels and possibly by channels donated to the oocyte membrane by the sperm. The long duration of the fertilization potential is caused by the long duration of the [Ca2+]i transients at fertilization and in many egg cells by the virtual absence of delayed K+ currents to repolarize the membrane. In addition, the resting resistance of most oocytes is very high, creating a long time constant. It is an interesting, but as yet unexplained, observation that oocytes (such as those of mammals) that do not depolarize at fertilization still have voltage-gated channels and are excitable to direct stimulation (477). This suggests that electrical activity has functions in oocytes that we do not yet understand.
While the fertilization potential is not strictly spontaneous because it is triggered by sperm binding, it emphasizes that electrical activity plays a role in development from the earliest stages. A variety of experiments have shown that fertilization potential mediates the fast block to polyspermy in many organisms, acting to prevent supernumerary sperm from fusing with the oocyte at short times before physical mechanisms of polyspermy block have been established (264, 292, 266, 267; see Ref. 265 for reviews). Direct current injection experiments have shown that the depolarization alone is sufficient to block sperm entry (264).
2. Relationship to channel development
The populations of ion channels in the egg cell membrane at the time of fertilization result from a complex earlier process of development that has been studied in detail in starfish. At the end of oogenesis in starfish, the fully grown, immature oocyte awaits a hormonal signal that will trigger its maturation and ovulation in preparation for fertilization. In starfish this hormone is 1-methyladenine (1-MA; Ref. 275). Maturation involves breakdown of the nuclear membrane, the reinitiation of meiosis, and other events that prepare the egg to be fertilized, events that are similar to those triggered in mammalian oocytes by progesterone.
Voltage-clamp of oocytes of the starfish Leptaserias shows that the fully grown, immature oocyte has only two depolarization-activated currents: an inward Ca2+ current and a transient outward (A-type) K+ current. Action potentials can be elicited by depolarization, but because of the large ratio of A-current to Ca2+ current, they do not overshoot 0 mV. During maturation, which can be triggered in vitro and takes only 30–45 min, the A-current decreases by 
50%, while the Ca2+ current remains unchanged. This increases the amplitude of the action potential so that it now overshoots 0 mV (424). The decrease in A-current is caused by loss of plasma membrane during maturation: cell capacitance decreases by precisely the same amount and with the same time course as the A-current, and electron micrographs show an almost complete elimination of the microvilli that characterize the surface of the oocyte before maturation (422; see Ref. 530). Quantitative measurements of membrane surface area from the micrographs show a close correspondence with the change in surface area measured electrically by capacitance. The selective loss of the A-current during maturation is necessary for the mature egg to be fertilized successfully. Without the resulting increase in action potential amplitude, the fast electrical block to polyspermy is much less efficient, fertilization is polyspermic, and abnormal development ensues (414, 415). Thus a prefertilization process of ion channel development that relates to the first activity-dependent developmental event in the life of the organism is required for embryonic life to begin.
The lack of decrease in Ca2+ current during this massive membrane loss implies that Ca2+ channels are protected from endocytosis, possibly by cytoskeletal anchoring. During oogenesis in this species, the A-current increases gradually, accurately tracking membrane area during a 2-yr growth interval just as it tracks the 30-min period of membrane loss during maturation. The Ca2+ current, on the other hand, appears abruptly at the end of the growth phase, dissociated from membrane addition just as it is from membrane loss during maturation (421). The appearance of the Ca2+ current coincides with the migration of the nucleus to the animal pole at the end of oogenesis. This raises the possibility that Ca2+ channels are inserted into the membrane only at the animal pole and are then protected from endocytosis by mechanisms that anchor the nucleus in that position (529). Protection from endocytosis by cytoskeletal anchoring or by accessory subunits influences ion channel development during terminal neuronal differentiation as well (6).
Selective loss of some currents and preservation or increases in the amplitudes of others have also been observed during maturation of amphibian oocytes (25). This kind of selective modulation also extends to exogenous channels expressed in oocytes (80, 536).
The patterns of ion channel development in this relatively simple system encapsulate many of the principles that are seen in more complex central nervous system structures. Individual currents may increase and decrease during development, changing at distinct stages and with specific relationships to other cellular events, such as changes in membrane surface area or cell cycle progression and arrest. The timing and specificity of such changes dictate changes in action potential threshold and waveform in ways that are critical for the developmental roles of electrical signaling at specific stages.
B. Early Postfertilization Changes and Selective Channel Elimination
Complex changes in ion channel expression also occur between the time of fertilization and the beginning of nervous system formation. Most of the information about ion channel development at these early cleavage stages comes from work in ascidian embryos. Ascidians are chordates whose embryos have long been a classic preparation in which to study development, partly because many cell lineages are committed very early in development without substantial cell interactions (520; a notable exception is the nervous lineage, which does require induction by the notochord). Some species have the additional advantage of an intensely pigmented muscle lineage, so muscle lineage cells can be recognized even at early cleavage stages and in dissociated preparations.
Before fertilization, eggs of the ascidian Boltenia villosa express Na+, Ca2+, and inwardly rectifying K+ currents. After fertilization these three currents are eliminated, each at a specific stage. The Na+ current is lost within 2 h, at the time of first cleavage, but the Ca2+ current is retained (51, 240, 120). The Ca2+ current is eliminated later, at gastrulation, and the inward rectifier even later, at neurulation. The inward rectifier is particularly interesting. In the 12 h before it is eliminated, the inward rectifier is maintained in all cells of the embryo at a constant density. Because the total surface area of all cells in the embryo at gastrulation is 10 times that of the egg, there must have been a 10-fold upregulation of the inward rectifier during this period, either by addition of new channels or unmasking of preexisting ones. Then, after this intense period of upregulation, the inward rectifier is eliminated in all cells in only a few hours (51, 208). This kind of active maintenance of channel density followed by abrupt disappearance suggests roles both for the presence of the channel early and its absence later. In this case, it is likely that maintenance of the inward rectifier, which is the only resting conductance of these cells, combined with the loss of the Na+ current, prevents generation of action potentials during the cleavage phase of early development. As discussed below, the disappearance of the inward rectifier is the trigger for spontaneous activity in these cells. Superimposed on all of these changes is the cyclical appearance and disappearance of a hyperpolarization-activated Cl– channel with each cell division (52, 598).
A very interesting analogous result to the above selective elimination of Na+ current (INa) after fertilization, but retention of Ca2+ current (ICa), is seen in another ascidian, Ciona intestinalis. The egg cell of this species does not have a low-threshold INa, but rather a second, low-threshold ICa similar to the T-type Ca2+ current. It is this low-threshold ICa that is lost after fertilization in Ciona, while the high-threshold ICa is retained (11). This suggests that the functional significance in both species is the elimination of the low-threshold inward current that could lead to aberrant spiking during cleavage, independent of the identity of the channel that carries it. Such aberrant spiking might be triggered by the activation of mechanosensitive ion channels, which are a prominent feature of ascidian oocytes (423).
Some of the most extensive work on ion channel development in early embryos has been done by Takahashi and colleagues, working on cleavage-arrested embryos of the ascidian Halocynthia roretzi. The egg cell of this species expresses Na+, Ca2+, delayed K+, and inwardly rectifying K+ currents (460). If embryos at various early stages of development are cleavage-arrested with cytochalasin, the cells differentiate into a variety of mature cell types without dividing. This differentiation is accompanied by specific changes in the patterns of ion channel expression (245, 456). So, for example, if cleavage is arrested at or before the four-cell stage, cells develop into an epidermal type, characterized by expression of Ca2+, inwardly rectifying K+, and Ca2+-activated K+ currents. Later cleavage arrest yields cells that differentiate into neural or muscle types, which express different patterns of ion currents. The cleavage-arrested one-cell embryo is large enough that the time course of ionic current expression can be followed in real time as it differentiates from egg to epidermal cell (244). The egg Ca2+ current disappears before gastrulation, and then a mature epidermal form reappears later. The mature Ca2+ current shows Ca2+-dependent inactivation, whereas the egg form shows voltage-dependent inactivation (243). Oddly enough, in the closely related species H. aurantium, the egg-type Ca2+ current reappears and then it subsequently changes into the mature form (243). The Na+ current disappears entirely from the cleavage-arrested egg as it develops, since the epidermal cell type into which it develops does not express a functional Na+ current. The disappearance of the Na+ current is very gradual and follows a complex time course, with a transient peak of density as the larval tail elongates (243). In a further elegant series of experiments, this group developed an in vitro two-cell neural induction system, in which individual cleavage-arrested neural- and notochord-lineage cells are placed into physical contact and subsequently separated to study the exact developmental timing of induction, as well as the timing of competence in each cell (454, 455, 457). With the use of this system, it was shown that neural induction triggered the expression of a neural-type Na+ channel whose biophysical properties were distinct from the Na+ channel expressed in the egg (461, 462).
Early postfertilization channel development has also been studied in mammalian oocytes. These cells express functional Ca2+ currents and generate action potentials (459, 477, 630), acquiring this property during the growth phase of oogenesis (162). After fertilization, this Ca2+ current shows a transient increase in density near the second meiotic division (627) and then decreases in amplitude and disappears by the eight-cell stage (413).
Some of these changes in ion channel expression in cleavage-stage embryos may be related to the fact that electrical activity and Ca2+ influx appear to play a role in neural induction. Induction by the lectin ConA is accompanied by long-lasting increase in [Ca2+]i and is inhibited by L-type Ca2+ channel blockers. L-type Ca2+ channel agonists can trigger neural induction of ectoderm (426). L-type Ca2+ channels are expressed in the embryo in dorsal ectoderm at the appropriate stages (154), and induction by mesoderm in vivo activates these channels and causes [Ca2+]i transients in ectoderm (321, 322). In vivo imaging of Xenopus embryos reveals spontaneous [Ca2+]i transients in the presumptive forebrain region of the dorsal ectoderm, although not in the presumptive spinal cord. L-type Ca2+ channel blockers produce defects in anterior nervous system structures (322).
Later, specific patterns of ion channel expression appear in presumptive neural tissue. Outward K+ currents and T-type Ca2+ currents are expressed in rat floor plate epithelium (188). As discussed below, proliferative cells of the cortical ventricular zone also express delayed outward K+ currents (480).
The above studies show that complex patterns of ion channel development are triggered by fertilization. These proceed throughout the early cleavage stages of embryogenesis, and usually involve specifically timed elimination of voltage-gated currents, as well as cyclical activation of channels by the cell cycle. Thus the electrophysiological properties seen at the start of terminal differentiation in many excitable cells reflect a complex history of ion channel development that begins even before fertilization. Aside from polyspermy block, the roles of electrical signaling during oogenesis, maturation, and early cleavage stages are as yet poorly understood.
C. Retina and Refinement of Visual Connections
1. Retina: nature of spontaneous activity
It has been known for quite some time that the mammalian retina generates spontaneous activity early in development, before it is capable of responding to light, and that this activity involves retinal ganglion cells (RGCs) and cholinergic synaptic transmission (381). With the understanding that the spontaneous activity was intimately involved in the patterning of retinogeniculate connections (535), more attention began to be paid to its mechanisms. Using extracellular unit recording, Galli and Maffei (192) and Maffei and Galli-Resta (362; both in rat retina) demonstrated that neighboring RGCs showed temporally correlated action potentials during this activity, and proposed that such correlations might act in a Hebbian manner to strengthen their connections to downstream targets. We now know that this activity takes the form of spontaneous waves of action potentials and [Ca2+]i transients (169, 616). Such spontaneous activity before the onset of patterned vision occurs in a wide variety of vertebrates, including mouse (17), rat (192), rabbit (637), ferret (389), cat (389), chick (620), turtle (533), and salamander (76) (see Refs. 167, 615, for reviews).
The basic properties of retinal waves have been established using a combination of multielectrode array recording, whole cell recordings from single cells, and [Ca2+]i imaging. Activity in the form of bursts of action potentials lasting 2–4 s, occurring at intervals of 1–2 min, sweeps across large regions of the retina at speeds of 80–140 µm/s (389). RGCs participate in this activity, generating brief bursts of action potentials riding on a large depolarizing wave (618). These waves spread across "domains" in the retina, initiated apparently at random in different regions (169, 170). Some kind of refractory period prevents the rapid reoccurrence of waves in single regions. The idea of some form of postevent refractoriness as a determinant of the interval between waves is likely to hold in other spontaneously active developing structures, like the spinal cord (576). In retina it is supported by the finding of postburst depression of RGC-RGC synapses, whose time course of recovery is similar to the interval between waves (234), and by computational models (85, 86). In chick retina the propagation of these waves seems somewhat more widespread, with activity often moving outward to the edges of the retina (90, 535, 620). Although early experiments indicated that this activity was sensitive to block by tetrodotoxin (TTX), thus implying that the [Ca2+]i transients required Na+-dependent action potential activity (389), it is now understood that this is not strictly true. TTX reduces the amplitude of the [Ca2+]i transients during activity (563), indicating that although RGC action potentials are not essential for wave propagation, they are necessary for the full amplitude of Ca2+ entry during waves and probably also to permit detection of the waves with extracellular recording methods.
Although spontaneous activity persists for a long time during development before the establishment of patterned vision, the mechanisms generating that activity and propagating it across the retina change dramatically. At early stages, experiments in ferret and rabbit retina indicate that ACh is the primary transmitter involved in wave propagation, consistent with the fact that the retinal circuitry at those stages is mainly dependent on the cholinergic starburst amacrine cells (168, 619, 637). At later times in this early period, GABA becomes involved as an excitatory transmitter in the waves (563, 619), reflecting the presence of GABAergic amacrine cells. At these stages (and perhaps later), the spread of activity appears to involve endogenous adenosine acting via A2 receptors to increase intracellular levels of cAMP (563).
At later stages, as glutamatergic synapses between bipolar cells and RGCs develop, glutamate becomes essential for wave generation and ACh becomes less important (619, 637). GABA also becomes inhibitory (178), reflecting its transition from excitatory to inhibitory action due to changing intracellular chloride concentration (see sect. IVC). The appearance of glutamatergic bipolar cell participation in the activity also correlates with the appearance of differences in the participation of ON and OFF RGCs in waves (323), reflecting a developing role for the waves in segregation of ON and OFF terminals within the lateral geniculate nucleus (see below). This developmental change in the transmitters and circuitry underlying spontaneous activity is also seen in spinal cord and reflects a remarkable stability in spontaneous activity even as the cells and circuits mediating it undergo developmental changes. The timing of this change in retina is plastic. When the early cholinergic waves are eliminated in mouse knockouts of the 
2-subunit of the nicotinic ACh receptor, glutamate-dependent waves appear several days earlier than normal (17). This may reflect a form of compensation that ensures stability of spontaneous activity during certain critical stages of development, even in the face of disruptions to the mechanisms that create it.
Gap junctional communication also seems to be involved in retinal waves. In chick, blockers of gap junctions suppress spontaneous waves (620). In salamander retina, they disrupt the short-time correlations between firing of neighboring cells (76), and in mouse they increase the interval between waves and decrease the number of cells participating in a wave (544).
As is true in developing cortex (see sect. IIE), the retina can be induced to generate similar waves under conditions that increase neuronal excitability (L-type Ca2+ channel agonists). Also as in cortex, these waves are not generated by the same mechanisms as normal spontaneous activity. In retina, the induced waves persist in the presence of antagonists of DL-
-amino-3-hydroxy-5-methylisoxazole-propionic acid (AMPA), N-methyl-D-aspartate (NMDA), glycine, and GABA receptors, which block normal waves at various stages of development. Like normal waves, however, induced waves are suppressed by gap junction blockers and by agents that disrupt the action of adenosine (544). The overlap in properties of these two forms of activity implies that retinal circuitry has several mechanisms of propagating waves of activity, involving classical chemical synaptic transmission pathways, electrical communication via gap junctions, and spread of activity via adenosine action on the cAMP second messenger system. The degree to which each participates may depend on developmental stage as well as the physiological states of the participating cells.
These results, like those in cortex, emphasize the difference between the kinds of activity that neuronal circuits are capable of generating and the activity that actually occurs. Inducing activity by artificial means may yield valuable information about the underlying functional circuitry and potential mechanisms of activity that a given structure may draw upon in creating spontaneous activity. Studying the actual spontaneous activity itself reveals how the structure makes use of that circuitry and those mechanisms.
An interesting recent finding points to the possibility that retinal waves may also propagate into the retinal ventricular zone (VZ), providing some kind of feedback to the zone from which retinal cells arise (574). Waves in the VZ showed close spatial and temporal relation to the retinal waves, and pharmacological studies indicated that the VZ waves involve muscarinic ACh receptors, and likely require the retinal waves, but not vice versa. Mature retinal glia can also generate [Ca2+]i waves, which can modulate retinal ganglion cell light responses (439, 440).
2. Retina: refinement of retinal ganglion cell connections by activity
The parameters of retinal waves, such as propagation speed and emerging differences in the participation of various types of RGCs, are critical for how activity encodes RGC identity to target structures. The establishment of correct patterns of connections between RGCs and their primary targets in the brain is one of the best known examples of how intrinsic molecular tags and electrical activity cooperate during brain development. To understand the roles of electrical activity in RGC projections, it is important to understand the anatomical and functional differences among the various species of animals in which the work has been done. Three points are particularly important: 1) the location to which RGC axons project. In cold-blooded vertebrates (amphibians, fish), RGCs project to the optic tectum, which is the main visual processing center in these species. In rodents and birds, RGCs project both to the tectum (superior colliculus) and to the lateral geniculate nucleus in the thalamus (LGN). The relative projections to those two structures differ among species (200, 337). In the higher mammals (cats, ferrets, primates), the visual cortex has evolved as the main processing center, and RGCs project primarily to the LGN as the synaptic relay center that sends visual information to the cortex. The superior colliculus in the higher mammals and birds serves important functions in the control of eye movement, but not as a processing center for visual perception.
2) The second important point is the developmental timing and initial accuracy of RGC projections to their primary targets in relation to eye opening and the appearance of patterned visual input (see Refs. 543, 615 for review). In cold-blooded vertebrates, the retinotopic map of initial RGC projections is fairly accurate, although there is substantial remodeling and refinement of that map, partly to compensate for retinal and tectal growth (494). This period of refinement occurs after the establishment of patterned vision, and thus the activity dependence of refinement in these animals could reflect activity driven by visual input, although the relative roles of visually driven and spontaneous activity are not entirely clear (see Ref. 296). In rodents, RGC projections to the superior colliculus are less accurate initially, and subsequent refinement takes place before the retina has become functional. Birds are similar in this regard. Thus the activity dependence of refinement reflects spontaneous retinal activity. In higher mammals, RGC projections to the LGN are also initially fairly inaccurate, and refinement occurs before the establishment of functional vision.
3) The degree of binocular input to the RGC projection site is important. Although adult cold-blooded vertebrates have binocular pathways in the tectum, there is not the eye-specific layering that is seen in the LGN of higher mammals. However, eye-specific tectal fields can be created by surgical manipulations that create direct binocular innervation of the tectum (tectal ablation, third eye implants) (see, e.g., Refs. 495, 499, 512, 513 for reviews). Thus both retinotopic projections and eye-specific termination fields can, and have been, studied in these animals. In rodents, although there is binocular projection to the LGN, it is small compared with that in higher mammals, and so true eye-specific layers are not present, although eye-specific fields do exist and are studied (200; see, e.g., Ref. 255). In birds, retinotectal projections are almost exclusively contralateral, a pattern that arises by elimination, in an activity-dependent manner, of ipsilateral projections that are present earlier (159, 615, 624). In higher mammals with substantial binocular vision, true binocular projections to the LGN exist in the form of eye-specific layers, created in part by activity-dependent pruning of initial connections (see below; Ref. 615).
Axons of RGCs projecting to the tectum in lower mammals, chicks, and cold-blooded vertebrates undergo a period of refinement into restricted terminal zones in a retinotopic pattern. In fish in which this pattern is being reestablished during regeneration of cut optic nerves, blockade of retinal activity by TTX injection prevents this topographic map refinement, although axon outgrowth and initial projections to the tectum are normal (397). The same activity dependence has also been shown during initial development of this map in zebrafish, by abolishing RGC activity using the macho mutant, which reduces RGC Na+ currents and blocks their activity, or TTX (199). Similar experiments have been done in developing chick, in experiments using TTX or the Na+ channel opener grayanotoxin to disrupt retinofugal activity (295). To test the hypothesis that patterned visual input provides activity that is correlated between neighboring RGC axons as a means of map refinement, fish were raised in stroboscopic light to synchronize activity across wide regions of RGCs. Stroboscopic light, but not diurnal light or darkness, phenocopied the TTX effects on map refinement (118, 525). Thus artificially synchronizing all RGC inputs does disrupt map formation, but on the other hand, the retinotopic map forms normally in darkness. These results imply a more complex scheme in which spontaneous activity even during stages when visual input is functional, is necessary for map refinement. Recent experiments in regenerating fish optic nerve add another layer of complexity: the retinas do not generate spontaneous waves of activity during regeneration, and in fact, overall firing rates of RGCs were depressed as their tectal projections refined during regeneration. Furthermore, blocking retinal activity during this period did not affect activity in tectal neurons, making a Hebbian scheme of refinement more difficult to envisage (296).
In amphibians there is direct evidence that coactivity of adjacent RGCs during spontaneous retinal waves serves to mediate long-term changes in synaptic efficacy where the two RGCs converge onto a single tectal neuron (634). These experiments were done in frog at early stages when RGC axon terminals are still widespread in the tectum, and synapses have a combination of NMDA and AMPA receptors. Repetitive stimulation of a single input to a tectal neuron causes homosynaptic potentiation, which requires action potentials in the postsynaptic neuron, but which does not affect other inputs. Pairing of two inputs potentiates both as long as the postsynaptic cell spiked. This included potentiation of a previously subthreshold input as long as it fired within 20 ms before the suprathreshold input. Simultaneous stimulation of two subthreshold inputs could potentiate both as long as they summed to trigger postsynaptic activity.
In mice, the retinotopic map in the superior colliculus (tectum) is refined to a much greater degree during early development, before development of visual input (see above). Nonetheless, activity is still required from RGCs for map refinement, as shown by using mice lacking the 2-subunit of the ACh receptor, which is essential for spontaneous waves of activity at these early stages (388). This activity-dependent refinement requires functional NMDA receptors (543).
When innervation of one tectum by both eyes is induced, eye-specific layers are formed (see above), implying that differential activity between the two eyes can drive segregation of their projecting axons, since presumably the molecular targeting cues are the same for axons from the two eyes. Blocking RGC activity with TTX prevents eye-specific segregation (67) and can even reverse segregation that has already occurred (495). These results imply that activity is not simply permissive for formation of retinotectal maps predetermined by other factors, but that activity can instruct the formation of a map that does not occur under normal circumstances. Block of tectal NMDA receptors can also desegregate the RGC terminals in the presence of continued afferent activity, and exogenous NMDA can sharpen the borders of eye-specific regions (113). Activation of NMDA receptors appears to act by a combination of elimination of inappropriate axonal branches and stabilization of appropriate ones (488, 512). Interestingly, nitric oxide does not seem to be involved downstream of these NMDA effects (499).
In higher mammals, such as cats, ferrets, and primates, there is extensive binocular innervation of the LGN by RGCs, and eye-specific layers form in each LGN before visual input is possible (see Refs. 513, 615 for reviews). These layers form by pruning of an initial innervation pattern in which axon terminals originating from different eyes are intermixed (560). Blockade of spontaneous activity originating in the retina at the stages during which this layering occurs (see above) prevents eye-specific layers from forming and leaves the earlier, wider branching patterns of axons in the LGN in place (535, 561). Although the original experiments were done with intracranial infusions of TTX, later intraocular TTX injections confirmed that the activity in question does indeed arise from the retina (476). The long-term effects of TTX treatment on layering are not completely clear. Layers appear to form almost normally at long times during chronic TTX treatment (119), but activity block after layering is complete can cause desegregation (97). The mechanisms by which activity acts involves competition between RGC terminals from different eyes: increasing activity in one eye expands its territory in the LGN, but increasing activity in both eyes leaves layering mostly intact (476, 562). Even so, NMDA receptor activity does not appear to be involved in this competition-based formation of layers (548), although it does seem to be involved in formation of LGN sublaminae containing different RGC cells types (see below). After the period of eye-specific layer formation, retinal spontaneous activity continues, although its mechanism changes from cholinergic to glutamatergic (see above). Selective elimination of the early, cholinergic activity in mice deficient in the 2-subunit of the ACh receptor disrupts layering, demonstrating that the early retinal activity is required (432, 508). Interestingly, in these experiments, eye-specific patches of RGC terminals were still formed in the LGN (432).
The patterns of spontaneous activity recorded from the retina at these stages seem well-suited to encode both spatial location of RGCs within one eye and ipsilateral versus contralateral identity. The movement of waves of activity across one retina results in contiguous RGCs showing correlated activity (362), and the short duration of the waves compared with their frequency of occurrence would result in activity from the two eyes occurring at different times. Confirming this is the finding that disrupting correlations in activity between neighboring RGCs while leaving the overall frequency of activity intact did not disrupt eye-specific layering (253). Reinforcing the idea of the role of correlated activity strengthening retino-geniculate synapses in the Hebbian-like mechanism is the finding that retinal activity is passed onto geniculate neurons, and that stimulated RGC firing at frequencies near those occurring during spontaneous waves can induce long-term potentiation (LTP)-like synaptic strengthening (425). However, findings that blockade of RGC activity by TTX does not completely eliminate eye-specific layer formation at early stages, and that even in the chronic presence of TTX, delayed layer refinement does occur (although not to normal levels of sharpness)(119), indicate that neural activity may not be the sole player in regulating retinogeniculate mapping. It is also possible that redundant mechanisms exist that can at least partially compensate for loss of activity (see Ref. 513). Recent evidence suggests the involvement of immune system molecules in activity-dependent LGN layering (255).
Activity also is involved in the finer-grain segregation of mammalian retinogeniculate connections. ON- and OFF-center RGCs innervate distinct sublaminae in the LGN (567), and this segregation requires activity in the retina before visual input is functional (124, 221). Unlike eye-specific layer formation (548), formation of ON and OFF sublaminae does require activity of NMDA receptors (221), and subsequent activation of the neuronal nitric oxide (NO) synthase-NO-cGMP pathway (123, 125, 320). Although these findings reinforce the idea of LTP-like synaptic strengthening by correlated pre- and postsynaptic activity, ON/OFF segregation does not involve the activation of "silent" synapses by induction of AMPA receptors in LGN neurons (247). Segregation of ON and OFF RGC connections in the geniculate appears to rely on their different patterns of activity, which in turn appears to be due to a divergence of their intrinsic ion channel properties early in retinal development (436).
Activity-dependent refinement and pruning of axons in the visual system is not restricted to RGC connections. Spontaneous activity regulates branching and organization of LGN axon projections in the visual cortex. Block of activity by TTX prevents correct branching and causes some axons to project to the subplate of areas outside of their normal visual cortex target area (89, 238).
In many of these experiments, as well as those described below, one must be aware that blocking spontaneous activity with TTX, for example, does not necessarily eliminate all periodic activity in the system. Ca2+-dependent action potentials, spontaneous transmitter release, and possibly other forms of activity may still occur and be responsible for some developmental phenomena. In systems such as the retinotectal and retinogeniculate pathways, the assumption is that activity in the form of Na+-dependent action potentials must propagate along axons to carry out its developmental functions. In other preparations, however, local effects of activity may be more resistant to TTX.
3. Retina: outgrowth of retinal ganglion cell axons
It has been known for some time that direct electrical stimulation can reversibly arrest axon outgrowth and cause growth cone filopodial retraction in both invertebrate and vertebrate neurons (115, 176). As in other cases of activity-dependent events, stimulation in a burst, or phasic, pattern is more effective than tonic stimulation (176). But early in development, activity may stimulate initial axon outgrowth. Recent experiments indicate that activity interacts with various trophic factors in intricate ways to regulate axon outgrowth and pathfinding. In retinal ganglion cells, peptide trophic factors stimulate axon outgrowth, but only at slow rates in the absence of activity. Electrical stimulation at physiological frequencies greatly speeds outgrowth stimulated by these factors (201). Interestingly, the pattern of stimulation that proved most effective was brief bursts of action potentials delivered at 1-min intervals, closely approximating the pattern of spontaneous synchronous activity seen normally in developing retina and many other areas of the mammalian central nervous system (see sect. IIIA2; Ref. 201).
4. Retina: dendritic patterning of RGCs
The spontaneous activity that sweeps across the developing retina appears to have a function in the elaboration of dendritic trees in the retina, as well as on the patterns of RGC axon elaboration in the LGN. Retinal spontaneous activity helps to segregate the dendrites of ON and OFF RGCs within the retina, much as activity segregates their axon terminals into different LGN sublaminae, although the degree to which activity instructs this is a matter of debate. Bodnarenko and Chalupa (54) and Bisti et al. (43) report a strong requirement for metabotropic glumatergic transmission in this process, whereas Bansal et al. (17) find more subtle effects. Blocking activity in RGCs with TTX also eliminates their ability to extend dendrites into RGC-free areas created by injury (142).
5. Retina: relationship to channel development
Ion channel development in various retinal cell types is likely timed to regulate spontaneous activity, although many details remain unclear. In cat and mouse retinal ganglion cells, a negative shift in the voltage dependence of activation and a positive shift in the inactivation curve of INa combine with increased INa density to help bring about the early appearance of repetitive firing ability (510, 546). Later expression of Ca2+-activated K+ currents and speeding of recovery from inactivation of INa may contribute to changing the firing patterns of retinal ganglion cells from bursting during spontaneous retinal waves to more sustained firing needed for encoding visual information (509, 604, 605).
Divergence of ion channel properties in different retinal cell types is also likely to regulate how spontaneous activity occurs, although not always in obvious ways. Late-emerging differences in the intrinsic properties of ON and OFF RGCs allow them to participate differentially in spontaneous retinal waves of activity, a difference which probably instructs their differential projections in the LGN (436). More perplexing are changes in excitability of starburst amacrine cells. In the rabbit, these cells express large Na+ currents and action potentials just before eye opening, and then lose INa and excitability over the next several weeks (636). Because spiking ability coincides with the period of spontaneous retinal activity, it was presumed that the transient expression of INa allows starburst amacrine cells to participate in this activity. Direct recordings showed, however, that despite their ability to spike, these cells do not generate action potentials during retinal waves (635; see Ref. 563). This may imply a different function for Na+ currents, perhaps in the development of intrinsic properties of these cells.
Optimization of channel properties is also likely to occur downstream in the visual system to tune the responses of LGN neurons to spontaneous activity in RGCs. During these stages, LGN neurons express NMDA receptors containing the NR2B subunit, which gives glutamate-induced synaptic currents a much longer time course than in the adult. This would clearly favor temporal summation, and in fact, such summation is observed in neonatal rat LGN in response to retinal spontaneous activity (349; see sect. IVB).
D. Hippocampus and Excitatory GABA Responses
1. Nature of spontaneous activity
During the first postnatal week, rat hippocampal neurons generate spontaneous and highly synchronous bursts of activity known as early network oscillations (ENOs) or giant depolarizing potentials (GDPs) (35, 194). These take the form of large synaptically driven depolarizations and bursts of action potentials, with associated bursts of [Ca2+]i transients, occurring at an overall frequency of 0.4–2/min (194). GDPs occur in the entire population of CA1 and CA3 pyramidal cells, in interneurons, and in hilar, septal, and dentate gyrus neurons (194, 284, 324, 326, 391, 566). GDPs are primarily GABAergic events, but also have substantial NMDA components and, at least at later stages, AMPA components as well (35, 55, 194, 326). Similar events have been recorded in vivo in both anesthetized and freely moving neonatal rats (325). The disappearance of spontaneous GDPs at the end of the second postnatal week correlates closely with the change in intracellular Cl– concentration that converts GABA action from excitatory to inhibitory (283). [Note that there is some disagreement as to the timing of GDP disappearance and GABA switchover. Khazipov et al. (283) postulate that earlier estimates are biased by the use of intracellular recording methods, which might disrupt intracellular Cl– concentration and/or introduce leak resistance effects into the measurements.]
It is not completely clear whether the synchronous GDPs that occur throughout the hippocampus are driven by a specific pacemaker region. Strata et al. (566) presented evidence that hilar neurons serve this function, at least for GDPs in the CA3 region. In hippocampal slices, surgical isolation of the hilus from CA3 abolished CA3 GDPs but left those in hilar neurons intact. Paired recordings showed that hilar GDPs preceded CA3 GDPs with a consistent 5- to 10-ms latency. Dye injections showed gap junctional coupling among hilar neurons and block of gap junction channels with octanol suppressed GDPs. Finally, voltage clamp of hilar neurons showed the presence of a putative pacemaker current (Ih), and Cs+ block of Ih slowed or blocked GDPs. Other experiments point to different pacemaker regions. Leinekugel et al. (324) used an intact two hippocampi plus septum preparation to ask whether septal neurons might act as pacemakers for spontaneous GDPs. They found that spontaneous GDPs propagate temporally to the hippocampi from the septum and that when partially isolated, the septum maintains a higher GDP frequency than the hippocampi. When the hippocampi were completely isolated from the septum, however, the hippocampi retained the ability to generate GDPs while the septum did not. They propose a pacemaker role from the septum, but an additional requirement of activity generated in the hippocampi to perhaps raise the level of excitability in the septum so that it can serve its pacemaking function. These two results are not necessarily incompatible. Strata et al. (566) worked in transverse slices, with CA3-hilus connections intact but with no septal-hippocampal connections. It remains possible that the hilar neurons generate a pacemaker signal that propagates to both the CA3 region and to the septum, which in turn propagates a pacemaker-like signal back into the hippocampus. Work by Menendez de la Prida and Sanchez-Andres (392, 393) and Menendez de la Prida et al. (391) makes an equally convincing case that GDPs are an emergent network property that can be generated by almost any subset of the hippocampal circuit. This hypothesis is based on data showing that isolated "islands" of CA1, CA3, and dentate gyrus can each generate spontaneous GDPs. Further data from paired recordings show that GDPs are triggered when the overall network activity rises to a level that can generate a threshold frequency of excitatory postsynaptic potentials (EPSPs) within participating neurons. Thus the origin of spontaneous GDPs in the neonatal hippocampus remains somewhat of a mystery, with evidence pointing to more than one potential pacemaker region and to a network property that can operate without a single discrete pacemaker.
2. Developmental roles of spontaneous activity
Much of this work focuses on the apparent paradox of "silent synapses" early in development (see Ref. 226). In the hippocampus, the large majority of synapses at P0 are pure NMDA, or silent, synapses, so named because in the absence of other receptor types, glutamate cannot activate them due to the voltage-dependent Mg2+ block of the NMDA receptor. It is known that repeated pairing of presynaptic activity with postsynaptic depolarization can "AMPA-fy," or induce, functional synapses by inducing functional AMPA receptors. The problem is that, early in development, there may not be a sufficiently high density of AMPA receptors on the postsynaptic cell for repeated presynaptic activity to depolarize the postsynaptic cell enough to trigger induction. It has been proposed (226) that spontaneous, GABAergic GDPs might provide the coincident pre- and postsynaptic depolarization to activate Hebbian mechanisms of synapse strengthening. Indeed, it has been recently shown that pairing of mossy fiber stimulation with spontaneous GDPs can induce a form of LTP, including induction of previously silent synapses (279).
At least one of the activity-dependent developmental programs in hippocampal neurons may serve to make activity self-limiting. Spontaneous activity in hippocampus depends on excitatory GABAA actions (see sect. IVC). Application of GABAA blockers blocks activity and delays the appearance of KCC2 chloride pump mRNA, whose expression lowers intracellular Cl– concentration and converts GABAA action to inhibition. Enhancing activity by KCl depolarization accelerates the switchover and the KCC2 mRNA expression (193).
3. Relationship to channel development
Spontaneous activity in hippocampus relies on the excitatory action of GABA, which is unique to developing neurons (reviewed in sect. IVC). It has been known for some time that GABAA actions are excitatory early in development, due to the high intracellular Cl– concentration at these stages (reviewed in Refs. 34, 469, 470). Some of the early reports of this phenomenon were of experiments done in hippocampus (430, 431). That spontaneous activity in hippocampus depends on excitatory GABA transmission is apparent from both the effects of GABAA blockers on activity (see above) and from the fact that the developmental disappearance of spontaneous GDPs parallels closely the switchover to inhibitory GABAA action (283).
The developmental profile of the hyperpolarization-activated cation current Ih may also influence hippocampal spontaneous activity. Ih is well known as a pacemaker current in a variety of cells (473, 504), and in particular, hilar neurons in the hippocampus appear to rely on Ih to drive spontaneous GDPs in other hippocampal regions (566). Ih density peaks in the early postnatal hippocampus (596) and thus may play a role in spontaneous activity. Another potential pacemaker current, the T-type Ca2+ current, is also present at higher density in neonatal hippocampal neurons than later (94).
E. Cerebral Cortex and Coordinated Na+ and Resting Channel Development
1. Cortex: nature of spontaneous activity
There are a large number of reports of spontaneous activity in rat and mouse cortical neurons, particularly in the first postnatal week. Most of these involve activity that is synchronous in small contiguous clusters of neurons, or involves more subtle correlations of activity among slightly more scattered cells. In many of these instances, activity is not actually spontaneous, but is elicited by altered ionic conditions, ion channel blockers, or transmitter agonists. These experiments show that the perinatal neocortex has the ability to generate activity with a variety of complex forms of spatial and temporal synchronicity.
A) NEURONAL DOMAINS. In rat cortex during the first postnatal week, clusters of 5–50 cells in a columnar orientation generate synchronous [Ca2+]i transients (632). Different clusters can be seen to generate this activity apparently randomly, with an interval of 4 min between events in different clusters. Although columnar in orientation, the clusters do not correspond to obvious functional units such as barrels. This activity does not appear to be caused by synchronous electrical activity, but rather by an inositol 1,4,5-trisphosphate (IP3)-mediated [Ca2+]i release that spreads from a trigger cell through the cluster, which seems to be defined by gap junctional coupling (277, 631). A similar form of activity is seen in mouse cortex at these stages, and knockout experiments suggest the involvement of the 
2 subunit of the NMDA receptor in regulating the spread of the [Ca2+]i transient (453), although how is unclear. It is possible that these coupled units and the [Ca2+]i signaling within them might represent a precursor to a modular architecture in the mature cortex.
B) CORRELATED ACTIVITY PATTERNS AMONG LAYER I NEURONS. A very interesting observation in the context of early cortical development is the presence of correlated patterns of [Ca2+]i transients among neurons of layer I, including both Cajal-Retzius and non-Cajal-Retzius cells (4, 531). This activity is not synchronous across large numbers of cells, but correlations among groups of cells can be detected by comparing correlation coefficients in pairs of cells to those expected from random activity at the same mean frequency (531). This pattern of activity in layer I was only rarely observed spontaneously, but could be evoked by high levels of extracellular K+ (50 mM). The [Ca2+]i transients were quite long in duration (>100 s), and were not blocked by TTX, although TTX did block the correlations among the transients (531). Correlations among cells involved chemical synaptic transmission, including glutamate, GABA, and ACh receptors. Correlations were not blocked by blockers of gap junction channels, indicating that direct electrical communication is not involved, thus distinguishing this activity from the neuronal domains discussed above. It is not clear whether these [Ca2+]i transients are caused by Ca2+ entry during electrical activity, or represent release from internal stores. Their long duration suggests the latter, but their sensitivity to block by Ni2+ suggests that Ca2+ entry may at least be the initial trigger for internal Ca2+ release. This hypothesis is also more compatible with the action of TTX in blocking correlations among cells within the network. It seems possible that Ca2+-dependent activity in cell bodies may trigger the long [Ca2+]i transients, but axonal Na+-dependent action potentials may be required to propagate the activity to other cells in the correlated network. The function of this type of activity in layer I is not clear, although communication between layer I cells and the apical dendrites of developing pyramidal neurons is likely to be involved (531). Activity that is synchronous between layer I neurons (both Cajal-Retzius and non-Cajal-Retzius) can be evoked under low-Mg2+ conditions, which activates silent NMDA receptors (549). The question remains, though, as to whether this activity is intense enough under normal (not high [K+]) conditions to carry out the proposed functions. The possibility of a deep cortical trigger for layer I activity is raised by the work of Dammerman et al. (131), who reported that electrical stimulation of GABAergic axons passing through layer I could excite cortical pyramidal neurons in neonatal rat cortex. These fibers arise in the zona incerta (ZI) of the thalamus and could represent a subcortical pathway capable of driving activity across large regions of the developing cortex. This hypothesis is strengthened by the direct recording of spontaneous activity in ZI neurons (131).
C) INDUCED LARGE-SCALE WAVES OF ACTIVITY. Cortical slices can generate waves of activity (measured either as electrical activity or [Ca2+]i transients) when treated with cholinergic agonists (287) or TEA (475). Although induced under artificial conditions, these waves reveal the capabilities of the neonatal cortex to initiate and propagate large-scale waves, synchronous among many neurons, over large physical distances. These waves require the activity of voltage-gated Na+ channels and seem to be propagated via glutamatergic synapses more than GABAergic ones. In the case of TEA, gap junctions seem also to be involved in wave propagation (475). The triggering of these waves by cholinergic agonists suggests that these agonists may substitute for cholingergic inputs from subcortical structures that are disrupted in brain slices (287). Such inputs may serve to raise the overall level of excitability in the neonatal cortex to a point where glutamatergic synaptic interactions can synchronize large neuronal populations. Cholinergic agonists can trigger such waves only during the first postnatal week, indicating that the intrinsic neuronal properties and synaptic circuitry of the neonatal cortex are specifically optimized for such functions.
D) SPONTANEOUS WIDESPREAD SYNCHRONOUS ACTIVITY. Two reports indicate that both rat and mouse neonatal cortical neurons can generate spontaneous [Ca2+]i transients that show widespread synchrony among a very large percentage of neurons in the cortex (121, 195). These [Ca2+]i transients result from electrical activity, as judged both by TTX sensitivity and by extracellular field potential recordings. Although GABA is excitatory to cortical neurons at these stages (E18-P5), activity is not blocked by GABAA antagonists, at least in rat. Activity is blocked by antagonists to NMDA and non-NMDA glutamate receptors. Unlike in the retina, the pharmacological profile of activity in cortex does not change as development progresses (195). The [Ca2+]i transients occur at low frequencies of 1/min to 1/12 min and propagate across the cortex at 1.5–2.5 mm/s. The activity emerges in the cortex just before birth, peaks at P0, and ceases by about P5 (121, 195). In each case, transients were studied in somewhat elevated (4.5–5 mM) [K+]. The posterior-anterior propagation of this activity, starting in the entorhinal cortex (195), might suggest that hippocampal activity acts as a pacemaker for cortical activity. Cortical activity, however, does not consistently propagate in this direction and occurs at lower frequencies than that in hippocampus (195). Analysis of slices in which the participation of neurons in synchronous activity was <50% showed that participating neurons were spatially clustered, implying a network in which mechanisms of synchronicity are widely distributed and are weaker between distant neurons than contiguous ones (121).
A recent report documents synchronous [Ca2+]i transients induced by blockers of glutamate transport in P0-P5 rat cortex (139). This raises the question of how to interpret these experiments where activity is induced by slightly elevated external [K+] or glutamate transport blockers. It is likely that both stimuli compensate for loss of activity during preparation and maintenance of slices, due to loss of extracellular glutamate by diffusion or loss of subcortical excitatory inputs. But it is also possible that cortex in vivo is not spontaneously active, or at least less so than in vitro. In that case, activity induced by these stimuli indicate that in vivo cortex is only marginally subthreshold for generating this kind of widespread spontaneous activity. The former possibility is strengthened by the finding that spontaneous synchronized activity recorded in hippocampal slices also occurs in vivo (325).
E) SPONTANEOUS ACTIVITY IN CULTURED CORTICAL NEURONS. A strikingly similar form of widespread spontaneous activity appears in cortical neurons from E15-E16 rat brain that have been in culture for 1–3 wk (467, 602), somewhat later chronologically than widespread activity in acute slices of neonatal cortex. The spontaneous [Ca2+]i transients occur at similar frequencies and had similar durations to those found in intact neonatal cortex. It is unclear whether this activity arises from the same mechanisms as that in cortical slices. In culture, this activity seems to be driven by a pacemaker population of neurons in the subplate and is as a result blocked by GABAA antagonists. GABAA blockers do not block similar activity in rat neocortical slices, indicating that a GABAergic subplate pacemaker is not necessary in that preparation.
