Potential consequences of altered neurogenesis on learning and memory in the epileptic brain


Address correspondence to Professor J. Martin Wojtowicz, Department of Physiology, University of Toronto, 1 King's College Circle, Toronto, ON M5S 1A8, Canada. E-mail: martin.wojtowicz@utoronto.ca


Studies of epilepsy and memory are tied by their common dependence on the hippocampal formation and the adjacent brain structures in the temporal lobe. With the discovery of adult neurogenesis and the consequent revisions of our understanding of how the hippocampus works, the role of neurogenesis in epilepsy needs to be addressed. In this article, we outline two theories describing how neurogenesis contributes to the hippocampus-dependent learning. We speculate that any drastic changes in neurogenesis will negatively impact the hippocampal memory processing.

Historically, the fields of epilepsy and memory research have been intertwined in many fascinating ways. Patient H.M., who suffered memory deficits caused by an ill-advised surgical procedure to cure epilepsy, is one of the best-studied medical cases in modern neurology (Ingram, 1994). On the other side of the spectrum is the case of Daniel Tammet, an autistic savant, whose prodigious memory abilities are presumably caused by a developmental anomaly. One notable event in his early childhood was the occurrence of seizures at age 4 or 5 (Tammet, 2007).

In most cases of childhood epilepsy involving temporal lobes, the consequences tend to be negative in terms of a disrupted structural integrity of the dentate gyrus and subsequent impairment of performance on cognitive tests. However, enhanced neurogenesis, as indicated by the expression of PSA-NCAM, a marker of immature neurons (Mathern et al., 1996, 1999), suggests a possibility of circuit rewiring that in some cases could produce unexpected improvements in learning performance such as those seen in savant subjects. Somewhere in the middle of the spectrum is a still mysterious amnesia-inducing effect of electroconvulsive shock (Squire, 1992; Squire et al., 2001) and its use in therapy for psychiatric disorders, for example depression, a disease with a substantially impaired memory component (Becker & Wojtowicz, 2007). Alterations in the adult neurogenesis could be involved in all these cases, but in different ways. In individuals suffering from chronic epilepsy, such as patient H.M., the neurosurgery consisting of a radical excision of large portions of the temporal lobes would eliminate neurogenesis to the detriment of learning and memory. It is well established that, in general, epileptic patients suffer from learning and memory problems and there is some (Mathern et al., 1999; Blumcke et al., 2001), albeit controversial (Fahrner et al., 2007), evidence that neurogenesis is enhanced in such patients. Such uncontrolled neurogenesis is likely to be disruptive to a normal brain, as suggested by animal experiments reviewed below. In cases of patients experiencing seizures at a very early age when neurogenesis is proceeding at a high rate, the rewiring of the hippocampal circuit could lead to permanent changes in the hippocampal circuitry with unpredictable consequences.

A possible, but still not understood, link between epilepsy and adult neurogenesis has to be seen in the context of a tremendous promise of neurogenesis as a mechanism to cure various neurological and psychiatric diseases. The possibility of improving memory and other cognitive defects in the aging brain (Abrous & Wojtowicz, 2008), stimulation of regeneration after injury (Wojtowicz, 2008), treatment of depression (Becker & Wojtowicz, 2007), and improvements in cognitive decline in diabetic dementia (Zhang et al., 2008) have all been mentioned in connection with adult neurogenesis (Fig. 1). Thus, adult-onset neurogenesis could be a double-edged sword with either beneficial or damaging effects depending on the circumstances.

Figure 1.

Multiple roles of adult neurogenesis in dentate gyrus. Reductions in neurogenesis in diseases such as depression, diabetes, or dementia are likely to be deleterious. Enhancement of neurogenesis to compensate for cell loss after injury or disease may be a potential therapeutic strategy. The enhanced neurogenesis after epileptic seizures and its possible consequences on memory is the subject of this article.

Seizures and Neurogenesis in Adult Brain

This article is based on the premise that adult-born new neurons are an integral and essential part of the hippocampal circuit. Thus, modern views on how the hippocampus works within the larger framework of brain circuitry must include the afferent and efferent connectivity and the intrinsic properties of newborn neurons.

Even before the considerable details of the neurogenic process had been described, the sensitivity of neurogenesis to seizures was already known. Pioneering studies of Parent and Lowenstein in late 1990s showed that seizures in an animal model stimulate adult neurogenesis. The meaning of this stimulation and its relationship to clinical symptoms of epilepsy is only now being revealed. Since these early studies of using a pilocarpine model (Parent et al., 1997), several other models have been tested and proven to stimulate neurogenesis: amygdala kindling, kainic acid administration, and electroconvulsive shock, among others (Parent et al., 1998; Scott et al., 1998, 2000; Madsen et al., 2000; Jessberger et al., 2005). This article will attempt to explain both expected and observed consequences of enhanced neurogenesis on formation of memory, the primary function of the hippocampus. First, we will present two distinct theories that include neurogenesis in the learning process, and then, we will argue that the expected effects of enhanced neurogenesis after seizures are likely to be negative.

Two Theories of Adult Neurogenesis

There are two contrasting opinions regarding possible functions of the young granule neurons that can be presented as opposing theories: the neuronal replacement theory (I) and the neuronal recruitment theory (II) (see Fig. 3 below).

Figure 3.

Two hypothesis of neurogenesis. (I) The cell replacement hypothesis proposes that mature cell death triggers cell birth so that the neurons can be replaced by new viable units. (II) The cell recruitment hypothesis proposes an activity-dependent selection of new neurons that subsequently take part in learning. The selection takes place during the critical period in the second week after cell birth. The young neurons surviving the selection process are available for “use” in the formation of memory trace at ages of 4 weeks and older. GCL, granule cell layer; SGZ, subgranular zone.

The replacement theory assigns a maintenance role to new neurons. This theory is based on the observation that the new neurons constitute only a small fraction (<0.5%) of the total mature population, and that ultimately, the new neurons become mature and undistinguishable from the preexisting neurons that were born during the embryonic and early postnatal periods (van Praag et al., 2002; Kempermann et al., 2003; Laplange et al., 2006; Wiskott et al., 2006). Furthermore, the supply of neurons is strongly age-dependent and appears to decline exponentially with a time constant of approximately 3–4 months in laboratory rodents (Table 1; Fig. 2A). Consequently, the rate of neuronal production is minimal in the aging animals. According to the replacement theory, the mature neurons have a limited lifespan (Dayer et al., 2003) and are replaced only as the need arises. The rate of replacement would be relatively high in young animals, but slows to a trickle past middle age. Overall, this pattern of neurogenesis may still be significant, since over the lifespan of the animal from 1 month to 2 years of age, the number of generated neurons is estimated to be in hundreds of thousands. For example, in laboratory rats, the total number of granule neurons is estimated to be around 1,000,000, and this number does not change much during adulthood (Boss et al., 1985). Our theoretical estimate of cell production takes into account the declining cell proliferation, similar rate of differentiation, and slightly reduced rate of maturation, as reported by McDonald & Wojtowicz, 2005 (Fig. 2B). The calculations assume that the age-related decline is exponential. This assumption is supported by the doublecortin (DCX) data illustrated in Fig. 2A. The graph in Fig. 2B shows that, in spite of the decay in cell production, the number of mature neurons accumulated over time would be substantial (Fig. 2B). The existing techniques for detection of cell death are not sufficiently accurate to demonstrate directly how many neurons die, but given a steady supply of neurons and no, or minimal, change in the total cell population, there must be a large number of neurons dying every day. Thus, there must be a balance between cell death and cell birth. Consistent with this hypothesis are the reports that reduced cell production occurs in parallel with reduced cell death as the animal ages (Heine et al., 2004). It is still uncertain what the trigger for cell replacement might be. One possibility is that the death of a mature neuron sends a signal to the proliferative zone to begin a replacement process (Fig. 3).

Table 1.  Number of DCX+ cells (neuroblasts and young neurons) per dentate gyrus (unilateral)
Age (days)Number of cellsStrainReference
  1. These estimates are based on several studies in Wojtowicz's lab. Same data are plotted in Fig. 2A. A single exponential curve provides satisfactory fit to the data, with an extrapolated initial value of 60,000 and a rate constant of 0.011. The time constant of decay is estimated at 91 days.

 3937,528S-DMcDonald & Wojtowicz (2005)
 6632,970S-DMcDonald & Wojtowicz (2005)
12020,000S-DWang et al. (2005)
13012,000S-DZhang et al. (2008)
13511,800L-EWinocur et al. (2006)
150 7,300L-E( J.M. Wojtowicz, personal communication)
15011,000S-DZhang et al. (2007b)
270 6,300S-D( J.M. Wojtowicz, personal communication)
310 4,800S-D( J.M. Wojtowicz, personal communication)
335 1,800S-D( J.M. Wojtowicz, personal communication)
360 3,000S-DMcDonald & Wojtowicz (2005)
390 1,300S-D( J.M. Wojtowicz, personal communication)
Figure 2.

Analysis of neurogenesis using young (doublecortin, DCX) and mature (calbindin, CaBP) neuronal markers. (A) Neurogenesis declines exponentially with age. Summary of data from several studies with references to the individual points are given in Table 1. The data represent the number of DCX+ cells (neuroblasts and young neurons) per dentate gyrus (unilateral). A single exponential curve provides satisfactory fit to the data, with an extrapolated initial value of 60,000 and a rate constant of 0.011. The time constant of decay is estimated at 91 days. This simple, exponential decline facilitates the analysis shown in Fig. 2B.
(B) Calculated numbers of DCX+ cells and CaBP+ cells on the basis of quantitative data from McDonald and Wojtowicz (2005). The predicted cumulative curves incorporate the age decline of proliferation and maturation, but assume the same differentiation rate in young and old animals. The calculations predict that despite a steep decline in cell proliferation, the production of new neurons, over the course of an animal's life, will result in a large number of mature neurons. This suggests that during the course of adult life, a large fraction of the granule cell population could be replaced by new neurons. The formula gives cumulative number of immature, DCX+ (declining curve) and mature, CaBP+ (rising curve) neurons. M, function describing reduced maturation with age (a); P, function describing reduced proliferation with age. The time constant of DCX curve is 139 days and of CaBP curve is 213 days. The delay is explained by the time required for cells to mature and accumulate.

According to this theory, any changes in neurogenesis induced by seizures will have relatively minor short-term consequences, since the phenomenon of cell replacement is relatively unimportant for the day-to-day functioning of the hippocampus. In the long term, however, the consequences could be significant and probably damaging, since the normal resupply of neurons would be disrupted by a sudden and transient production of new cells, causing a temporary oversupply followed by a dearth of new neurons. Moreover, the sudden burst of neurogenesis could overwhelm the tissue support that is presumably necessary to supply nutrients and neurotrophins necessary for cell growth and result in abnormal cells and/or abnormal synaptic connections. Later in life, seizures will have less impact because neuronal production is minimal past the middle age and the dentate gyrus appears to function essentially without neurogenesis.

Another strength of this theory is its basic conformance with the existing thinking about the hippocampal function as a whole. According to the current concepts of the hippocampal learning, the cortical input flows into the hippocampus either via the dentate gyrus or directly into the CA3 or CA1 fields (Witter, 1993). The role of the dentate gyrus appears to be to disperse the information from the entorhinal pathway into individual channels, perhaps mediated by individual or small groups of dentate granule neurons. This may be visualized as a “strainer” where a substance enters in bulk but exits in individual streams. This pattern separation is adequately performed by the mature granule neurons (Leutgeb et al., 2007) and is thought to facilitate unambiguous coding of similar environmental cues. The sparse output of the dentate gyrus is carried by the mossy fibers that exert a strong effect onto the pyramidal neurons in CA3 by virtue of their phenomenally plastic mossy fiber synapses. The association among the incoming bits of information is thought to occur via excitatory, recurrent circuits in CA3. This results in memory engrams that are next sent to CA1 for further processing. The two direct pathways from the entorhinal cortex to either CA3 or CA1 are not well understood, but they may serve preferential role in spatial learning, an important function of the hippocampus (Morris, 2006). CA3 neurons, in particular, are able to perform the pattern separation in a manner similar to the dentate gyrus (Leutgeb et al., 2007). This view of the hippocampal function has been developed without prior knowledge of ongoing neurogenesis, so it does not require addition of new neurons, except perhaps as a means to renew the dying granule neurons.

The second (II) theory is far more radical and, if true, will have more profound consequences on epilepsy-induced changes in the hippocampus. This theory assigns new, previously unknown role to the newly born neurons that changes our thinking about the function of the dentate gyrus. According to the recruitment theory, the new neurons no longer just separate the entorhinal inputs, but instead, actively participate in input integration due to their superior plasticity (Becker & Wojtowicz, 2007). In order to understand how the new neurons function, one must realize that the neuronal development proceeds through a series of choreographed steps, beginning with the embryonic and early postnatal development when neuronal stem cells migrate from the subependymal zone toward the subgranular zone of the dentate gyrus. These cells form a germinal proliferative layer that is strategically located along the border of the hilus and the granule cell layer (GCL). The principle cells that give rise to new neurons are characterized by certain expression factors, such as Sox1 (a marker of neural stem cells), and produce the astrocyte-specific protein glial fibrillary acidic protein (GFAP) (Encinas et al., 2006; Hattiangady & Shetty, 2007). These cells are thought to divide slowly and give rise mostly to neuronal progenitors that divide and differentiate further to become neuroblasts. These proliferating neuroblasts express the specific protein DCX, which has become a standard marker of immature neurons. During the second week of the neuronal life, the DCX+ neuroblasts pass through a critical period when they are particularly sensitive to environmental stimuli. Some of these stimuli, such as an enriched environment, promote neuronal survival and stimulate permanent exit from the proliferative cycle (Tashiro et al., 2007), while others, such as stress, reduce the neuronal survival (Thomas et al., 2007). Under laboratory conditions where rats are kept in simple cages, there is a net loss of about 60% of the neuroblasts during the second week (McDonald & Wojtowicz, 2005). This cell death is a part of the neurogenic process, but its magnitude can be regulated by environmental and physiological factors.

The original report on the existence of a critical period in adult neurogenesis suggested that new neurons were responsive to very selective stimuli such as hippocampal-dependent learning (Gould et al., 1999). During this period, the neurons “learn to survive” or die (Shors et al., 2002). This phenomenon has been further characterized by Abrous and colleagues who showed that the survival is accentuated by “pruning” of cells that are younger than 1 week (Dupret et al., 2007). What might be the purpose of such selective learning-induced cell survival? Hypothetically, neurons that are made to survive by a learning experience will be connected by afferents and will be responsive to specific sensory stimuli involved in the behavior that shaped it (Fig. 3). There is some suggestive evidence that such a process takes place during adult neurogenesis, i.e., that neurons that are made to survive the critical period by enriched environment preferentially respond, at an older age of 4–6 weeks, to the same environment (Tashiro et al., 2007). In agreement with this concept, the neurons that are 4–6 weeks old are characterized by heightened plasticity (Wang et al., 2000; Piatti et al., 2006; Ge et al., 2007) and respond to learning stimuli by expressing immediate early genes (Kee et al., 2007). The exact age at which the neurons become functional is still debatable (Ge et al., 2007), and neurons of different ages may be optimally suited for different functions (Winocur et al., 2006; Abrous & Wojtowicz, 2008).

Some behavioral experiments suggest that new neuronal pathways formed during learning are indeed essential for learning. For example, Shors and colleagues showed that neuronal survival is more pronounced in individuals that learn the task well (Dalla et al., 2007). Abrous and coworkers showed that among the aging rats, the group that was cognitively impaired had fewer surviving neurons in comparison to the unimpaired rats (Drapeau et al., 2007). Both of these studies suggest a correlation between the ability to stimulate cell survival and the ability to learn. However, it remains to be shown whether a specific learning experience that stimulates the neuronal survival at 1–2 weeks is the same as the one that “utilizes” the surviving neurons at 4–6 weeks. It is also possible that the survival-promoting effect of learning on neurogenesis is an epiphenomenon of marginal or no importance. Until a specific blocker of this action can be found, one cannot be sure that the survival of a few newborn neurons is a necessary mechanism for subsequent learning to take place.

Other experiments cast doubt on the validity of this theory. For example, experiments that used lesions to drastically reduce neurogenesis with either antimitotic agents or with irradiation failed to show an impairment in spatial learning (Leuner et al., 2006). Although these results can be explained by a compensatory mechanism, such as the use of an alternative afferent pathway from the entorhinal cortex to CA1, they clearly suggest that the mechanism of stimulated neuronal survival is not necessary for spatial learning.

Experiments that used alternative hippocampal tasks (other than spatial water maze task) showed that performance is strongly dependent on neurogenesis. The trace eye-blink conditioning, delayed nonmatching to sample, and contextual fear conditioning have all been shown to be severely affected when neurogenesis was reduced (Saxe et al., 2006; Winocur et al., 2006). Paradoxical enhancement of learning in the neurogenesis-depleted animals has also been reported (Saxe et al., 2007). Thus, most, but not all, hippocampal learning tasks appear to be dependent on ongoing neurogenesis in agreement with theory II. Also consistent with theory II is the presence of a substantial number of immature neurons that are available for recruitment. This would be necessary since cell production is a relatively slow process taking weeks, while the recruitment needs to draw rapidly on the existing pool of the plastic neurons. The presence of such a large pool has been well documented (Cameron & McKay, 2001). Furthermore, an in vitro electrophysiological assay has been devised that can distinguish the pharmacological and physiological characteristics of the young neuronal population from the mature population (Saxe et al., 2006; Snyder et al., 2001). This assay supports the idea of two distinct neuronal populations within the dentate gyrus. One includes the mature, strongly inhibited neurons that are suitable for pattern separation, and the other includes a highly flexible, heterogenous, and uninhibited young neurons suitable for pattern integration (Fig. 3).

Assuming that the intricate mechanisms predicted by theory II do exist, it is tempting to predict that abrupt, uncontrolled changes in the rate of neurogenesis following seizures would be disruptive and produce cognitive impairments. This disruption could be exacerbated if the cells produced after seizures are themselves intrinsically abnormal and unfit to properly integrate into the hippocampal circuit. Thus, new cells formed as a consequence of seizures may not respond properly during the critical period and may erroneously produce faulty pathways. Several steps could go wrong. The new synapses formed on the new abnormal neurons could have a wrong balance of excitation and inhibition. The synapses on basal dendrites extending toward the hilus are thought to be excitatory and would make the new cells hyperexcitable. The cells could migrate toward the hilus instead of toward the GCL. This again would make them hyperexcitable and wrongly connected with the perforant pathway (Shapiro et al., 2007). The cells could make excessive recurrent collaterals in the inner molecular layer and connect with hilar neurons (sprouting), thus upsetting a balance of excitation and inhibition of the preexisting granule neurons. All these anomalies would disrupt normal development and could upset the proposed mechanisms in hypothesis II.

Structural Changes in Hippocampus Related to Enhanced Neurogenesis

The reported structural changes related to neurogenesis are the massive increase in the rate of proliferation (Parent et al., 1997), accelerated growth (Overstreet-Wadiche et al., 2006b), abnormal cell migration toward hilus or GCL (Jessberger et al., 2005; Jakubs et al., 2006; Walter et al., 2007), and growth of basal dendrites (Shapiro et al., 2007). Cognitive consequences of these changes are difficult to determine since they occur in parallel with the numerous changes in the mature granule cell population. However, considering that the newly formed neurons are generally more excitable and subjected to reduced gamma amino butyric acid (GABA)ergic inhibition (Wang et al., 2000; Ge et al., 2005), one may reasonably hypothesize that the new cells contribute to seizures. It is unfortunate that numerous studies characterizing altered synaptic properties in dentate gyrus after seizures failed to recognize that neurogenesis plays a major part in the process (Pathak et al., 2007; Zhang et al., 2007a). Future studies should include information on GABA-mediated tonic and phasic inhibition, taking the differences between the mature and immature neurons into account. Two recent studies examined the new neurons specifically and compared their properties to mature neurons and to neurons generated in a normal, nonepileptic brain. Overstreet-Wadiche et al. (2006b) found that pilocarpine-induced seizures produced neurons that were hyperexcitable due to enhanced drive from the perforant path and from the recurrent collaterals. In addition, the generated cells were growing faster and dispersed toward the hilus (see review by Zhao and Overstreet-Wadiche, 2008). Such a dispersal was earlier shown to potentially contribute to enhanced recurrent activity (McCloskey et al., 2006). Another study demonstrated that new neurons generated after electrically induced seizures tend to be less excitable due to reduced excitatory drive (Jakubs et al., 2006). Thus, different epileptic models could produce varied effects due to excessive neurogenesis. In all cases, the disturbances in the connectivity and in the relative balance of excitation and inhibition are likely to be deleterious to the normal signal processing in the hippocampal circuit (Jessberger et al., 2007a, 2007b).

As mentioned above, disturbances due to excessive neurogenesis are expected to have severe consequences if the new neuronal population is actively and preferentially involved in learning and memory (theory II). On the other hand, if theory I is correct, the effects of disturbed neurogenesis may be easier to offset by the vast majority of the mature granule neurons, provided that some of them remain normal. Theory I predicts that the effects of neurogenesis should be less noticeable in mature animals, where the rate of neurogenesis is minimal in comparison to the young ones. Theory II is less clear, since the additional neurogenesis in the context of relatively low rate of neurogenesis encountered in aging animals, even if abnormal, could be somewhat beneficial in memory processing, particularly if offset by slower neuronal development due to aging (Overstreet-Wadiche et al., 2006a).


Two distinct, but not mutually exclusive, theories of neurogenesis are presented. Both theories try to include neurogenesis as an integral and vital process within the hippocampus. The balance of evidence seems to indicate that seizure-enhanced neurogenesis is damaging to the animals regardless of which theory is proven correct. This may partly explain why early childhood and chronic epilepsy is usually associated with cognitive problems in human subjects. It may also serve as a warning that more neurogenesis is not always better, and that, any attempts to stimulate neuronal production in the adult brain for therapeutic purposes should take the proper integration of the new neurons within the existing circuitry into account.


This work was supported by grants from CIHR and NSERC, Canada.

My particular thanks to Ms. Yao Fang Tan for the preparation of figures for this article.

Conflict of interest: I confirm that I have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The author has declared no conflicts of interest.