Abstract: Seizures are one of the most common yet frightening neurological conditions encountered in humans. While seizures can occur at any age, affecting at least 2% of the population at one point or another, they are far more common in children than adults. The highest incidence of seizures occurs during the first year of life. While some infants with neonatal seizures fare well, in many cases the seizures are a sign of an ominous neurological disorder. There are many unanswered questions regarding neonatal seizures. What are the features of the immature brain that make it so prone to seizures? What are the consequences of seizures during early brain development? Because of the ethical and logistical difficulties in performing studies in ill newborns, much of our insight into the pathophysiology and consequences of neonatal seizures have necessarily come from animal studies. Why is the immature brain prone to seizures? Results from animal studies are consistent with the clinical observation that the immature brain is far more prone to seizures than the adult brain. The increased excitability in the developing brain appears to be secondary to a developmental imbalance between maturation of excitatory and inhibitory circuits (7Gaiarsa J.-L. Tseeb V. Ben-Ari Y. J. Neurobiol. 1994; 73: 246-255Google Scholar, 2Ben-Ari Y. Khazipov R. Leinekugel X. Caillard O. Gaiarsa J.-L. Trends Neurosci. 1997; 20: 523-529Abstract Full Text Full Text PDF PubMed Scopus (669) Google Scholar, 12Leinekugel X. Medina I. Khalilov R. Ben-Ari Y. Khazipov R. Neuron. 1997; 18: 243-255Abstract Full Text Full Text PDF PubMed Scopus (330) Google Scholar, 10Jensen F.E. Wang C. Stafstrom C.E. Liu Z. Geary C. Stevens M.C. J. Neurophysiol. 1998; 73: 73-81Google Scholar). The main ionotropic receptors (GABAA, NMDA, and AMPA) display a sequential developmental pattern of participation in neuronal excitation in the neonatal hippocampus. GABA, the main inhibitory transmitter in the adult, provides the main excitatory drive to hippocampal neurons at early stages of postnatal development, because of a Cl− gradient that leads to depolarization of young neurons rather than the hyperpolarization observed in adults. Consequently, in the immature brain, GABAA responses are associated with the Cl− efflux, depolarization, and activation of voltage-dependent Na+ and Ca2+ channels (12Leinekugel X. Medina I. Khalilov R. Ben-Ari Y. Khazipov R. Neuron. 1997; 18: 243-255Abstract Full Text Full Text PDF PubMed Scopus (330) Google Scholar). The depolarization produced by GABA is sufficient to remove the voltage-dependent Mg2+ block from NMDA channels, thereby inducing large Ca2+ influx into immature neurons. The other major postsynaptic inhibitory system, the postsynaptic GABAB, adenosine, and 5-hydroxytryptamine/G protein–coupled K+ channels, also have a delayed maturation, suggesting that the neonatal circuit operates without transmitter-gated inhibition. In contrast, presynaptic inhibition, mediated by adenosine, GABAB, or other metabotropic receptors, is fully operational at birth; this observation suggests that the major form of inhibition in the neonatal circuit is the control of transmitter release. Other factors contribute to the increased excitability in the developing brain. For instance, a large input resistance in developing neurons leads to the generation of major changes in membrane potentials in immature dendrites. The prevalence of gap junctions during early development could potentiate the generation of synchronized activity, amplifying small imbalances in neuronal activity into the large-scale synchronization characteristic of seizures. In addition, a shorter period of postictal refractoriness in young animals leads to a quick progression through early stages of kindling and to a rapid generalization of seizures. Thus, while the precocious development of excitatory networks before inhibitory systems may play an essential role in plasticity and learning, this imbalance also appears to place the infant at high risk for seizures. How do seizures affect the developing brain? While the immature brain is more susceptible to seizures than the mature brain, there is now a considerable body of evidence suggesting that the consequences of seizures in the immature brain are considerably different from those occurring in adults. In the mature animal, status epilepticus causes neuronal loss in hippocampal fields CA1 and CA3 and the dentate hilus. This leads to aberrant growth (sprouting) of granule cell axons in the supragranular zone of the fascia dentata and stratum infrapyramidale of CA3, sprouting of CA1 axons, and long-term deficits in learning, memory, and behavior. Surprisingly, a substantial number of studies have demonstrated that a single prolonged seizure in an immature animal results in less cell loss and sprouting and fewer deficits in learning, memory, and behavior than seizures of similar severity in adults. Indeed, even following a brain injury, such as a hypoxic–ischemic insult, prolonged seizures do not appear to incur any additional cell loss (5Cataltepe O. Vannucci R.C. Heitjan D.F. Towfighi J. Pediatr. Res. 1995; 38: 251-257Crossref PubMed Scopus (42) Google Scholar). There are a number of reasons why seizures produce less cell loss in the immature brain than in the mature brain. While prolonged seizures are associated with elevations of glutamate, there are clear developmental differences in glutamate-induced damage during seizures. When equal concentrations of glutamate are injected in the CA1 subfield of the hippocampus in young and mature rats, there is far more damage in the mature brain (14Liu Z. Stafstrom C.E. Sarkisian M.R. Tandon P. Yang Y. Hori A. Holmes G.L. Dev. Brain Res. 1997; 97: 178-184Crossref Scopus (78) Google Scholar). Postsynaptic action of glutamate-induced Ca2+ entry through the NMDA receptor exhibits age-dependent changes, as may each step in the intracellular cascade following Ca2+ entry. For example4Bickler P.E. Gallego S.M. Hansen B.M. J. Cereb. Blood Flow Metab. 1993; 13: 811-819Crossref PubMed Scopus (72) Google Scholar noted an increase in the elevations of intracellular Ca2+ of 240% from P1–P2 to P28 following application of equal concentrations of glutamate in rat cerebral cortex slices. There are also indications that intracellular Ca2+ buffering capabilities may be greater in young than in older animals. In addition, the smaller density of synapses and larger extracellular space in the developing brain may reduce the concentration of potentially excitotoxic glutamate. Furthermore, the fact that GABA provides a significant component of the excitatory driving force will reduce the pathological consequences of seizures because of the smaller driving force and the shunting effects of GABA. However, it is becoming clear that seizures do not necessarily need to cause cell death to result in adverse outcomes. While prolonged seizures may cause less cell loss in young animals, there are nonetheless clear indications that they may have other potentially deleterious effects. Following kindling, immature rats, like adult rats, have a permanent reduction in seizure threshold. Recurrent seizures during the neonatal period also result in clear deficits in learning and memory when the animals are studied as adults, despite the lack of clear cell loss (17Neill J. Liu Z. Sarkisian M. Tandon P. Yang Y. Stafstrom C.E. Holmes G.L. Dev. Brain Res. 1996; 95: 283-292Crossref PubMed Scopus (60) Google Scholar). This may be due to the fact that seizures may perturb a wide range of phenomena that are activity dependent, including cell division and migration, sequential expression of receptors, and formation and probably stabilization of synapses. Paroxysmal discharges may facilitate an early expression of glutamatergic receptors and alter the formation of functional entities as “cells that fire together wire together” (8Goodman C.S. Shatz C.J. Cell. 1993; 72: 77-98Abstract Full Text PDF PubMed Scopus (994) Google Scholar). We have found that recurrent flurothyl-induced seizures during the first days of life are associated with sprouting of mossy fibers in the CA3 subfield and supragranular region and a reduction in seizure threshold (9Holmes G.L. Gaiarsa J.-L. Chevassus-Au-Louis N. Yang Y. Ben-Ari Y. Ann. Neurol. 1998; 44: 845-857Crossref PubMed Scopus (342) Google Scholar). Figure 1 shows an example of Timm staining, which detects zinc in the axons and terminals of the dentate granule cells in a control rat (Figure 1A) and a rat subjected to multiple, brief seizures (Figure 1B). Excessive Timm granules can be seen in the pyramidal cell layer in the rat with recurrent seizures. These rats, when studied as adults, had a lower seizure threshold and impaired learning on a task of spatial memory. Since this CA3 and supragranular sprouting is not a result of cell loss, what type of mechanism, then, might account for it? One important variable to take into account is neurogenesis. Increases in granule cell neurogenesis following seizures occurs in both the immature and mature brain. Granule cell development occurs during an extended period that begins during gestation and continues well into adulthood. Increased granule cell neurogenesis has been found following both prolonged and brief seizures in the adult rat (3Bengzon J. Kokaia Z. Elmir E. Nanobashvili A. Kokaia M. Lindvall O. Proc. Natl. Acad. Sci. USA. 1997; 94: 10432-10437Crossref PubMed Scopus (672) Google Scholar, 18Parent J.M. Yu T.W. Leibowitz R.T. Geschwind D.H. Sloviter R.S. Lowenstein D.H. J. Neurosci. 1997; 17: 3727-3738Crossref PubMed Google Scholar). Following a prolonged seizure induced by pilocarpine18Parent J.M. Yu T.W. Leibowitz R.T. Geschwind D.H. Sloviter R.S. Lowenstein D.H. J. Neurosci. 1997; 17: 3727-3738Crossref PubMed Google Scholar found enhanced neurogenesis, when compared to the controls, for several weeks. Neurogenesis may be stimulated through direct synaptic activation of precursor cells by the seizure or may occur as a result of cell loss. A multitude of biochemical changes occurring during a seizure could alter the rate of neurogenesis. For example, several studies using immunohistochemistry, in situ hybridization, or biochemical assays have demonstrated that neurotrophins are increased in specific brain regions following seizures. 11Kornblum H.I. Sankar R. Shin D.H. Wasterlain C.G. Gall C.M. Mol. Brain Res. 1997; 44: 219-228Crossref PubMed Scopus (49) Google Scholar found that seizures induced by lithium-pilocarpine or kainic acid resulted in dramatic elevations of brain-derived neurotrophic factor mRNA in immature rats. It is further known that neurotrophins are important in the regulation of neurogenesis and may have played a role in the noted increase in dentate granular cells (19Tao Y. Black I.B. DiCicco-Bloom E. J. Comp. Neurol. 1996; 376: 653-663Crossref PubMed Scopus (94) Google Scholar). However, the story may be more complex, since there are a variety of other activity-dependent processes that could have effects on brain development. For example, LoTurco and colleagues (1995) found that during early stages of cortical neurogenesis, both GABA and glutamate depolarize cells in the ventricular zone and inhibit DNA synthesis. While seizure-induced neurogenesis occurs in both the immature and mature brain, it is possible that the consequences of neurogenesis may be quite different. In the mature brain, neurogenesis occurs in the context of cell loss in the pyramidal cell layer and dentate granule cell layer. Thus, depending on the degree of cell death and of granule cell neurogenesis, there may be a paucity of target cells for the newly formed cells to synapse upon. In the immature brain, however, neurogenesis occurs without cell loss, so the ratio of granule cells to pyramidal cells is markedly increased. In addition, whether neurogenesis ensues in the presence or absence of cell loss may result in fundamental differences of connectivity. As demonstrated in Figure 2A, dentate granule cells send axons (mossy fibers) to the CA3 pyramidal cells and interneurons. Following a prolonged seizure in the adult animal, there is genesis of new neurons in the dentate granule cell layer and loss of existing neurons in the dentate granule cell layer, hilus, CA3, and CA1. Because of the loss of target cells, there is resultant aberrant sprouting of the mossy fibers on remaining cells in the granular cell layer, hilus, and CA3 (Figure 2B). While not shown in the diagram, sprouting of CA1 axons may also occur. In the immature brain, where dentate granule cells are increased without any cell loss (Figure 2C), there is a resultant increase in the ratio of granule cell axons to pyramidal cells; this imbalance between the number of granule cells and target cells results in an increase in axons to the CA3 subfield, and to a lesser degree in the supragranular region. In neonatal rats, seizures occur before all of the mossy fibers have connected with the CA3 principal neurons; sprouting may therefore be an example of formation of entirely abnormal circuits as opposed to sprouting of already established pathways. Whether the alterations in cognitive function and seizure susceptibility in adults versus neonates depend upon these differences in cell loss and ensuing anatomical changes is a subject of current research. It is not yet entirely clear why enhanced neurogenesis and sprouting would be detrimental to the animal. Since glutamate is the neurotransmitter of the mossy fibers, it is tempting to suggest that increased numbers of glutamatergic synapses could conceivably increase excitability and lower seizure threshold. However, recently 1Áscady L. Kamondi A. Sik A. Freund T. Buzsaki G. J. Neurosci. 1998; 18: 3386-3403PubMed Google Scholar demonstrated that terminals from dentate granule cells are more likely to innervate GABA inhibitory interneurons than excitatory pyramidal cells, and that this may therefore be a mechanism by which the brain may try to regulate increased excitatory activity. Regardless of whether the morphological changes seen with recurrent seizures are directly responsible for the reduced seizure threshold, there is some evidence that the cognitive and behavioral effects of increased numbers of neurons and increased connectivity are detrimental to the animal. For example, the degree of CA3 mossy fiber projections has been inversely correlated with learning. Lipp and colleagues (1988) compared the number of trials for rats to learn to avoid a 10 s electrical shock by moving from one compartment to another following a conditioning stimuli (two-way avoidance learning) with magnitude of the stratum pyramidale projections of mossy fibers. Learning was tightly related to extent of mossy fiber projections to the intra- and infrapyramidal layers of CA3, with animals having more CA3 mossy fiber terminals doing less well than animals with fewer terminals. These authors also found an inverse relationship between the extent of infrapyramidal mossy fiber projections and two-way avoidance learning in rats treated with L-thyroxine. Furthermore, Wimer and colleagues (1983) noted a negative correlation of granule cell density in the dentate gyrus with two-way avoidance conditioning in the mouse. However, the relationship between the size of the hippocampal mossy fiber projections and learning and memory may be task dependent (6Crusio W.E. Schwegler H. Brust I. Eur. J. Neurosci. 1993; 5: 1413-1420Crossref PubMed Scopus (72) Google Scholar). For example, we have found that rats with recurrent seizures during the first weeks of life demonstrated impairment in the water maze and open field test (9Holmes G.L. Gaiarsa J.-L. Chevassus-Au-Louis N. Yang Y. Ben-Ari Y. Ann. Neurol. 1998; 44: 845-857Crossref PubMed Scopus (342) Google Scholar) and in an auditory localization task but not in a quality discrimination task (17Neill J. Liu Z. Sarkisian M. Tandon P. Yang Y. Stafstrom C.E. Holmes G.L. Dev. Brain Res. 1996; 95: 283-292Crossref PubMed Scopus (60) Google Scholar). Future advances in understanding network properties in learning and development will undoubtedly clarify the role that increased neurogenesis and sprouting has in the cognitive impairments resulting from early seizures. While the immature brain does appear to have some biological advantages over the mature brain with regard to seizure-induced cell loss, it is now clear that recurrent seizures do result in some maladaptive changes. Unfortunately, neonates and young children, for reasons described above, are at high risk for recurrent seizures. Currently, the most commonly used drug to treat neonatal seizure is phenobarbital, which enhances postsynaptic GABA inhibition. However, numerous animal and human studies have suggested that phenobarbital can be detrimental to the developing brain, leading to impaired brain growth and cognitive dysfunction (16Mikati M.A. Holmes G.L. Chronopoulos A. Hyde P. Thurber S. Gatt A. Liu Z. Werner S. Stafstrom C.E. Ann. Neurol. 1994; 36: 425-433Crossref PubMed Scopus (114) Google Scholar). Since GABA has also been shown to be excitatory in rodents during the first few days of life, phenobarbital may not be an optimal drug for neonatal seizures and, in fact, appears to have limited efficacy in treating neonates. Phenytoin, a drug that causes use-dependent inhibition of Na+ channels necessary for activation of action potentials, is sometimes used, but the poor absorption in neonates limits its usefulness as an oral agent. Other drugs used in adults that work on Na+ channels, such as carbamazepine and lamotrigine, have not been studied in neonates because of the logistical difficulties in studying drugs in very ill neonates as well as the ethical concerns about drug testing in infants. Based on the sequential developmental pattern of pre- and postsynaptic neurotransmitter receptor development described previously, drugs that are likely to be beneficial in neonatal seizures would be those that have their effect on the NMDA receptor, presynaptic GABAB receptors, or voltage-gated channels. Unfortunately, drugs that block the NMDA receptor, such as topiramate, have been tested sparingly in neonates because of the concern that treatment with excitatory amino acid antagonists may interfere with learning, memory, and brain plasticity. No antiepileptic drugs are currently available that are specific agonists of the GABAB receptor. Thus, while there has been tremendous progress made in the pharmacological treatment of seizures in older children and adults, the current methods to treat neonatal seizures have not changed in the last 50 years. While a better understanding of the mechanisms responsible for seizure-induced damage is critical, an even greater challenge is to develop drugs that inhibit neonatal seizures without otherwise interfering with brain development or physiology.‡To whom correspondence should be addressed (e-mail: [email protected]).