Abstract
The common co-occurrence of autism and epilepsy suggests that certain neurobiological mechanisms are shared between these disorders. In particular, the profusion of novel genetic mutations being discovered in autism and epilepsy points to abnormalities in synapse formation and function that alter the balance between neuronal excitation and inhibition. Animal models can be informative in sorting out the medical and behavioral complexities in autism and epilepsy and the relationship between them. As mechanistic information accrues, it is anticipated that mutation- and pathway-specific targeted treatments can be developed.
The relationship between autism and epilepsy is well established yet complex. The increased prevalence of autism spectrum disorders (ASD) in epilepsy, and epilepsy in ASD, has resulted in each disorder being considered a “comorbidity” of the other. Up to 30% of nonsyndromic autistic individuals have seizures, and an even higher percentage has epileptiform discharges on EEG (1, 2). The overlapping prevalence suggests that epilepsy and ASD share at least some common biological mechanisms (3, 4). However, complexity arises in that each disorder has multiple heterogeneous etiologies, and some of these etiologies overlap. The purpose of this brief review is to discuss current thinking about the relationship between epilepsy and autism and provide a perspective on how animal models might help to decipher their relationship. Do seizures cause autism? Does autism lead to seizures? Both? Neither? Sometimes? We hypothesize that the imbalance between neuronal excitation and inhibition that underlies both epilepsy and autism allows for testable hypotheses and a search for rational therapies (4).
In the accompanying review, Roberto Tuchman (this issue) concludes from clinical and epidemiological studies that autism is not a primary cause of epilepsy and that autism is not an epileptic encephalopathy. Furthermore, epilepsy does not commonly lead to autism unless intellectual disability (ID) is present (5). ID is also the major risk factor for epilepsy in autistic patients. Eight percent of children with autism develop epilepsy if there is no ID present, but 20% of such children develop epilepsy if they are also intellectually disabled (6, 7). Further, ASD is often coupled with ID (range 30–60%) (8, 9). Age is an additional critical risk factor for the development of epilepsy in children with ASD. It is important to identify ASD early in the care of children with epilepsy because interventions are available to enhance language function, social skills, and behavior (10; Tuchman, this issue). The additional cost of educating learning-disabled children can be 2 to 5 times that of their peers (11). Health care costs for children with ASD/ID are 18 times that of their peers (12). In the United States, more than 3.5 million individuals have ASD (1 in 88 males [13]) with a total annual cost of $260 billion/year (14). Substantial savings could come from even marginal improvements in ASD care (15).
ASD is not a singular entity but rather a clinical syndrome with many etiologies (16, 17). ASD etiologies are either syndromic or nonsyndromic (formerly called “idiopathic”). The syndromic forms (perhaps 10% of cases) might be better conceptualized as “neurodevelopmental disorders with autistic features as part of the phenotype” and include disorders such as tuberous sclerosis complex (TSC), fragile X syndrome, and Rett syndrome. It is likely that nonsyndromic ASD (representing the majority of cases) is polygenic, possibly with contributions from identified or unidentified environmental factors (18, 19); many nonsyndromic cases probably involve subtle genetic abnormalities, such as copy number variations (20). It has been hypothesized that aberrant protein translation leads to brain overgrowth underlying abnormal connectivity and altered synaptic function associated with some of these syndromes (19, 21). An imaging study of children with nonsyndromic ASD has documented early-onset brain overgrowth (22). Longitudinal clinical studies will allow more direct testing of the causes and effects of seizures in ASD (23).
Nearly all syndromic and many nonsyndromic forms of human ASD include epilepsy, suggesting one of several possibilities (not mutually exclusive). A genetic mutation may cause disruption of neuronal development (migration, channel function, synapse function, etc.); as a result, seizures, ASD, or both develop. Alternatively, seizures (or EEG abnormalities) might enhance the emergence or progression of ASD, or ASD could engender epilepsy and its consequences (4, 17, 24).
Features of epilepsy in individuals with autism are diverse—there is no consistent seizure type or cortical localization (25). For example, in TSC, the hamartoma (tuber) or adjacent cortex is highly epileptogenic, and early-life seizures are very common. The number and localization of tubers is related to eventual intellectual level and degree of autistic features (26, 27). However, both ASD and epilepsy may also result from tuber-independent cellular and molecular mechanisms. Furthermore, it is not clear whether seizures cause ASD or whether this is simply an association related to the underlying disease pathophysiology of TSC. In nonsyndromic ASD, the association between early seizures and ASD is even less certain, but in both cases, these questions are amenable to study in animal models.
The “core features” of ASD are considered to be impairments in language/communication, social interactions, and motor behaviors with restricted interests and activities. These three symptom categories have been used to characterize autism since the original description by Kanner (28), yet they lack biological markers and validation. The most recent version of the Diagnostic and Statistical Manual of Mental Disorders combines social and communication deficits into a single diagnostic criterion and retains restricted interests and behaviors as a separate category (29). Since the severity of each symptom domain can vary within an individual child, autism is considered a “spectrum” disorder. Difficulties with language and communication, social interactions, and restricted activities and behaviors are often not equally prominent in a given autistic individual, and each of these core features may follow a different time course—further evidence that each symptom has a different biological basis and underlying neurocircuitry (30).
Perspective on Animal Models of Autism
For the neuroscientist interested in studying ASD in the laboratory, the above clinical considerations are central and should drive hypotheses and model generation. What, exactly, would a rodent with autism look like? It is often hard enough to diagnose a child with autism. In laboratory animals, underlying cognition must be inferred from observable behaviors. Extreme caution must be exercised to avoid attributing any behavioral deviance in rodents to autism. Animal models cannot be expected to recapitulate all of the human aspects of ASD, yet some animal models do manifest social/communication and repetitive behaviors that can be quantified. For example, well-validated tests of socialization in rodents are the social chamber (SC) and the social partition (SP) tests. The SC tests an animal's preference for a novel object versus an unfamiliar rodent. If an animal shows no preference for the unfamiliar rodent, a socialization deficit is diagnosed. The SP tests an animal's preference for a familiar versus an unfamiliar mouse—again, a social deficit is concluded if the test animal fails to prefer the unfamiliar mouse. Repetitive behavior can be assessed with the marble burying task. Normally, rodents tend to bury objects in their bedding material, but excessive or decreased burying behavior reflects abnormal perseverative or restricted motor activity. As a surrogate for communication abnormalities, ultrasonic vocalizations and responsiveness to vocalizations have been studied in rodent genetic models of autism (31–33); therefore, several rodent behavioral tests can be used to assess autistic-like behaviors, but these are inherently indirect and inferential. Behavioral tests are also available to differentiate the core autism features from anxiety (elevated plus maze), intellectual dysfunction such as learning and memory deficits (water maze, radial arm maze), and other aspects of cognition that are not part of the autism spectrum but commonly occur as ASD and epilepsy comorbidities (34, 35).
Nonrodent models of epilepsy and ASD should be exploited to expand molecular understanding of these disorders. For example, zebrafish allow detailed developmental and molecular investigations of mutations relevant to human disease, though behavioral manifestations are even less direct (36). Nonhuman primates would more closely mimic human behavior but are severely limited by cost and ethical disadvantages. If laboratory investigations focus on discrete dimensions of behavior and their neurobiological pathways, useful information about ASD (and epilepsy) mechanisms can be gained from diverse animal models.
Numerous animal models, especially in mice, have been created that harbor genetic mutations with reported autism-like behaviors, and many include seizures or seizure susceptibility in the phenotype (though actual epilepsy is not commonly reported, see below). It is not our goal to review exhaustively the ever-expanding list of such mutations (17, 37–41). Rather, we emphasize common aspects that could help to understand how autism and epilepsy emerge. With any mutation, it is assumed that a genetic defect was present throughout most of early development. Therefore, both seizures and ASD can be independent consequences of the mutation, or seizures can worsen the ASD phenotype. Models that involve human-related mutations, seizures and ASD-like features are potentially valuable in the search for targeted therapies. For example, animal models of fragile X syndrome, like FMR1 knock-out mice, resemble the human disorder with autistic-like features, heightened seizure propensity, and learning impairments (2). Furthermore, the molecular and synaptic basis of fragile X syndrome has been elucidated to the point that rational therapy trials are being conducted (43); this model is ripe for exploration of seizure mechanisms as well (24). However, epilepsy per se (spontaneous, unprovoked seizures) has not been documented in fragile X knock-out mice or in many other genetic models of neurodevelopmental disorders. Therefore, each genetic model will have specific features related to the function of the mutated gene in neuronal function and neural circuits, and importantly, animal models of human ASD/ID do not always phenocopy epilepsy (e.g., Rett and CDKL5).
A related question is whether seizures during early development lead to later autism-like behaviors. As mentioned above, in humans, epilepsy is not thought to cause ASD, yet there is abundant data that early-life seizures are associated with adverse consequences, such as impairments in learning, memory, and cognition (44–46). The odds-ratio of ASD is threefold higher in preterm infants with early-life seizures (47). Others have suggested that early-life seizures may simply be a marker of the severity of brain injury (48–50), a conundrum that reflects the lack of proven and effective treatments of neonatal seizures to isolate the effects of seizures themselves. To address this issue, seizures induced in animals early in life have been shown to impair social interaction, communication, and repetitive motor behaviors, that is, some of the behaviors intrinsic to ASD (51, 52). Specifically, following multiple early-life seizures induced with the convulsant flurothyl, mice have deficits in socialization (SC and SP tests) (51, 52). Even a single postnatal seizure can result in abnormal, autistic-like behaviors in adulthood including reduced marble burying and social interactions (53). Treatment of early-life flurothyl seizures with bumetanide in a rat model may prevent the development of autistic-like behaviors (54). Rat pups subjected to neonatal hypoxia develop autism-like deficits (on SC test) as adults, an effect that is reversible with the mTOR inhibitor rapamycin (55). Behavioral tests discussed above can be readily applied to autism models as well. Mice with a deletion of PTEN (a gene involved in the regulation of the mTOR pathway) had deficits in SP, SC, and marble burying (56). Mice with BTBR mutations express autistic features (on SC test) but are not seizure-prone. Yet, the ketogenic diet improves the social deficits in BTBR mice, suggesting a dissociation between the anti-seizure and anti-autism effects of the diet (57).
Excitation/Inhibition (E/I) Imbalance Underlies Epilepsy and ASD
The idea that disruption of the delicate balance between neuronal excitation and inhibition (E/I imbalance) leads to the generation of seizures and epilepsy is well recognized. While this concept is admittedly oversimplified, it stands as a useful construct to consider both pathophysiological reasons for seizures to occur as well as targets for therapeutic intervention. Traditionally, increased excitation has been attributed to dysfunction of ion channels that mediate depolarization (e.g., sodium or calcium channelopathy) or enhanced excitatory neurotransmission (e.g., more glutamatergic synapses, greater glutamate release, glutamate receptor subunit changes). Similarly, decreases in inhibitory function (potassium channelopathy) or GABAergic pathophysiology (decreased synthesis, enhanced degradation, impaired receptor function) can facilitate seizures. Resulting shifts in the E/I balance are dependent on whether excitatory circuits, inhibitory circuits, or both are affected (58). In addition, all of these factors have developmental profiles that can contribute to enhanced seizure susceptibility at different ages (45, 46, 59, 60).
Likewise, in ASD, an E/I imbalance has been proposed to result in altered neuronal function that could lead to the core autism signs of impaired communication, aberrant socialization, and restricted activity/stereotypies (4, 37, 38). In particular, dysfunction of synapse formation, development, maintenance or function that disrupts the E/I balance has been implicated in ASD as well as in other neurodevelopmental disorders such as schizophrenia and intellectual disability (27, 40, 41).
Despite the appeal of the simple E/I imbalance hypothesis, newer data demands expansion of this concept (61). Many genetic mutations recently found to cause ASD, epilepsy, or both, involve aspects of neuronal function beyond ion channels and synaptic physiology. It is now appreciated that structural proteins that anchor synaptic machinery, regulate synaptic vesicle release, control subcellular signaling pathways, and govern migration of neurons and organization of network connections have important roles in neuronal excitability. Mutations in proteins involved in each of these functions are associated with ASD and epilepsy (27, 62).
The co-occurrence of E/I imbalance in epilepsy and ASD suggests that treating one disorder—for example, by addressing the cause of the imbalance—might benefit the other as well. There is some evidence that anticonvulsant treatment improves autistic symptoms in humans (63) and animal models (54, 55). Animal models provide a rich source of diverse pathophysiologies and genetically modifiable substrates to evaluate the effects of potential treatments, acknowledging that face and construct validities are inherently imprecise. While the best model of human autism will always be a human, rodent models that fulfill face, construct, and predictive validities will move translational research forward. The difficulty of attributing “autism” to a nonhuman animal is obvious (face validity), though an autistic (or at least “autistic-like”) phenotype can be inferred by quantifying behaviors (the outward manifestation of neuronal circuit activity). The mechanistic underpinnings of systems, circuits, and molecules can never be identical between and human and rodent ASD (construct validity). A nearly infinite number of genetic and environmental combinations could result in ASD. Thus, more generalizable models should complement single-gene models that may only represent a small percentage of human ASD.
Highlight Points
Epilepsy and autism frequently occur together, suggesting that there are shared neurobiological mechanisms underlying these two disorders.
Animal models are useful in exploring the relationship between autism and epilepsy, permitting genetic, molecular, physiological, and behavioral analyses that can be compared with human studies.
Novel genetic mutations in both autism and epilepsy are illustrating the wide diversity of etiologies and mechanisms that can lead to these disorders.
Discovering common mechanisms in autism and epilepsy may lead to targeted therapies applicable to both disorders. This task requires close cooperation between researchers in epilepsy and autism.
Creating an animal model of any disease process is inherently difficult, especially for disorders involving highly complex behaviors with multiple etiologies. Genetic mutants provide much information, but nongenetic models can also be informative. For example, prenatal valproate induces clinical and behavioral changes consistent with an autistic-like syndrome, manifesting decreased social interactions, stereotyped behaviors, and altered pup vocalizations, as well as heightened anxiety (64, 65)—but not epilepsy. The prenatal VPA model has construct validity with a significantly increased risk of autism in offspring after maternal in utero VPA exposure (66). This model can be used to test drug therapies as well as environmental enrichment to test predictive validities. Indeed, once predictive validity is established in a model of any disorder, it is time to move on to the next disorder.
Future Considerations
The recent explosion of genetic mutations found to underlie both ASD and epilepsy raises a number of questions of relevance to experimental neuroscience. First, with so many different mutations affecting such varied aspects of neuronal physiology, there is no single road to ASD or epilepsy. The task is to discover general principles among the specific individual mutations. To date, few studies have specifically investigated the link between autism and epilepsy in animal models. Second, the concept of E/I imbalance, though useful, needs widening to include mechanisms not traditionally thought to result in excitability disorders. In turn, these new concepts will open new therapeutic possibilities and targets. Third, we have much more to learn about region-specific cell types, circuits, and networks involved in the development of ASD and epilepsy. For all of these questions, animal models will serve as a focal point for investigations, acknowledging appropriate caveats about species-related differences. Our experimental repertoire needs to expand beyond rodents to exploit the full spectrum of molecular and behavioral deficits in ASD and epilepsy. Animals allow for comparison of specific synaptic cellular, molecular defects with social behaviors, and epilepsy propensity, but we need ask very critical questions of the various models in relation to human disease. Addressing the above questions requires a better working relationship between researchers with expertise in epilepsy and those who specialize in autism (67).