Utilizing Animal Models of Infantile Spasms

Abstract

Infantile spasms are a devastating epileptic encephalopathy characterized by early life spasms and later seizures. Clinical outcomes of infantile spasms are poor and therapeutic options are limited with significant adverse effects. Therefore, new strategies to treat infantile spasms are of the utmost importance. Animals models of infantile spasms are a critical component of developing new therapies. Here, we review current chronic animal models of infantile spasms and consider future advances that may help improve patient care, as well as our scientific understanding of this debilitating disease.

Infantile spasms (IS), also known as West Syndrome, is a devastating epileptic encephalopathy presenting with early life flexion extension spasms evolving into severe chronic epilepsy later in life. IS is also a component of many other epilepsy etiologies, including tuberous sclerosis, fragile X syndrome, and focal cortical dysplasia and has been considered a “final common outcome” in response to a variety of acquired and genetic brain insults at this stage of brain development. A number of unique clinical features define IS: First, and most obvious, is the onset of disease-defining spasms in the first year of life. Second, the presence of hypsarrhythmia, a high-amplitude chaotic EEG signal, is also a common clinical feature of IS. Third, and most challenging, spasms evolve into chronic seizures later in life. Finally, severe neurodevelopmental disabilities are a common comorbidity, adding to the difficulty for patients and their families. Many in the field believe that IS may be a form of epilepsy in which we can make rapid clinical advances. In support of this, a portion of children treated early with existing therapies can do extremely well and avoid the progression from spasm to later epilepsy. These cases are extremely exciting, but that excitement is tempered by many other patients who do not respond to therapy or who have only modest improvements in their outcomes. One strategy to developing new therapies is to utilize preclinical animal models. In this review, I will highlight recent advances in chronic IS animal models and outline how they might be used to enhance our understanding of disease mechanism and to provide preclinical platforms for drug testing.

Using a Human Clinical Framework to Establish and Evaluate Animal Models of IS

To robustly model a disease in rodents, its etiology must be understood. In IS, however, there are multiple insults, genetic and otherwise, that can cause disease. Genetic mutations in ARX, CDKL5, FOXG1, GRIN1, GRIN2A, MAGI2, MEF2C, SLC25A22, SPTAN1, TSC1/2, LIS1, DCX, NR2F1, and STXBP1 can lead to IS (1–12), with ARX and STXBP1 being the most common (13). Acquired forms of IS also exist and can arise from hypoxic-ischemic insults, infection, brain malformation, and, rarely, metabolic/vascular abnormalities (14). Animal models of IS have been developed in which the human genetic mutation leads to mouse phenotypes that approximate human disease. Similarly, a number of acquired models in which various insults are delivered to the developing brain can be strikingly similar to the human condition. Most perplexing, however, are genetic models that replicate human mutations but do not recapitulate spasms or seizures.

Similar to mimicking etiology, another goal of animal models is to recapitulate disease pharmacosensitivity. The most utilized current treatments for IS are adrenocorticotropic hormone (ACTH) and vigabatrin, as well as high-dose steroids (15). ACTH is a polypeptide released from the pituitary, and it plays an important role in regulating the activation of the hypothalamus/pituitary/adrenal HPA axis. It is unknown how ACTH acts to control IS, although effects on the HPA axis via corticotropin-releasing hormone may be involved (16). Vigabatrin is a GABA transaminase (enzyme that degrades GABA) inhibitor. Again, the mechanism of action in IS is unknown, although it stands to reason that inhibiting GABA breakdown potentiates GABAergic inhibition. Both of these frontline treatments are associated with significant adverse effects, and a substantial portion of IS patients do not respond to existing treatments. Clearly, new therapies are needed. Preclinical scientists currently have animal models at their disposal that respond to standard of care treatment, as well as models that are insensitive to treatments used in patients. This is advantageous because it is important to develop novel treatments that work in patients who do not respond to currently used treatments. Utilizing treatment-resistant animal models offers the opportunity to assay novel potential therapies in a preclinical environment.

Defining the Essential Parameters of a Model of IS

Now that we have established a clinically relevant framework in which to consider animal models, the next task is to define the essential properties of a model of IS. A number of recent reviews and an NINDS workshop have done an excellent job addressing this question (15, 17–19). To summarize, ideal animal models of IS should demonstrate neonatal behavioral spasms, ictal EEG complexes, interictal hypsarrhythmia, and seizures in adults. Additionally, specific seizure properties (such as onset during wakefulness or on arousal), response to standard of care treatments, and behavioral/cognitive dysfunction would also replicate key phenotypes of human IS. While these qualifications are useful, they need not be universally applied to validate a new model of IS. For example, hypsarrhythmia does not occur in all IS cases and, as mentioned earlier, animal models that do not respond to established treatments are needed to identify novel therapies for treatment-resistant IS. Furthermore, etiologically realistic models (i.e., rodents harboring known disease-causing mutation) that do not manifest with spasms or seizures can provide important insight into molecular and cellular aspects of disease progression. With that in mind, we can now explore existing animal models of IS as well as IS-relevant models.

Currently Available Rodent Models of Infantile Spasms

This review will focus on chronic models of IS, although acutely induced models exist, lead to long-term cognitive deficits, and are unquestionably useful to the field (15, 17, 20, 21). When focusing on chronic models, there are two general classes of models: genetic and acquired. Genetic models involve disruption of a specific gene, while acquired models expose genetically normal animals to various chemical/pharmacological agents to induce IS-like phenotypes. For a more thorough review of models, please see the excellent book chapter from Galanopoulou and Moshe (17), from which this short review draws heavily.

Genetic Models of Chronic Infantile Spasms

Arx Knock-In (Arx^(GCG)10+7) Mouse Model

The Aristaless-related homeobox (Arx) protein is a homeobox transcription factor important in the maturation and migration of developing GABAergic interneurons (22, 23). Mutations in Arx can lead to IS/epilepsy as well autism, mental retardation, and dystonia (24–27). A commonly seen mutation in IS is an expansion of a polyalanine tract in the Arx gene. To replicate that human mutation, mice were generated with 7 additional alanine codons in the polyalanine tract (28). Arx ^(GCG)10+7 mice exhibit behavioral spasms when examined from post-natal day (P) 7–11 and have high-voltage slow wave EEG transients followed by EEG attenuation. Brief myoclonic jerks occurred at the onset of the EEG attenuation. More mature Arx ^(GCG)10+7 animals (3.5–10 weeks) had low-voltage fast activity onset seizures associated with slow head bobbing and clonic movements. Occasional tonic–clonic seizures have also been noted, and cognitive impairments are seen in Arx ^(GCG)10+7 mice. In line with Arx‘s known role in interneuron maturation, Arx ^(GCG)10+7 mice have a reduced abundance of calbindin-positive interneurons in the cortex and granule cell layer of the dentate gyrus and neuropeptide Y-expressing and choline acetyltransferase-expressing interneurons in the striatum (28). To date, hypsarrhythmia has not been documented in Arx ^(GCG)10+7 mice (or any other genetic IS model), and there are no published reports of their response to ACTH or vigabatrin. Interestingly, Arx ^(GCG)10+7 show reduced spasms and seizures in response to neonatal treatment with 17β-estradiol (29), with an important caveat to be discussed later.

Arx Conditional Knock-Out (cKO) Mouse Model

In this model, Arx was genetically removed from DLX5/6-expressing cells (30). This eliminates Arx expression in ganglionic eminence derived neurons, included cortical GABAergic interneurons. Young (P14–17) pups exhibit behavioral seizures, abnormal cortical EEG, including slowed background activity of a low voltage and limited high-amplitude spikes. All Arx cKO mice show spontaneous seizures typified by either behavioral arrest or spasm-like behavior in adults. In addition, Arx cKO show a variety of altered cortical and hippocampal rhythms with specific disruptions in males versus females. As in Arx ^(GCG)10+7 mice, Arx cKO mice show decreased labeling of calbindin-positive interneuron in the cortex. Interestingly, genetic excision of Arx from cortical projection neurons using Emx1-Cre does not result in seizures or spasms but does cause a variety of behavioral abnormalities (31).

APC cKO Mouse Model

Adenomatous polyposis coli (APC) is a negative regulator of β-catenin/Wnt signaling. Interestingly, multiple human mutations that cause IS involve proteins that modulate, or are modulated by, β-catenin signaling (see Pirone et al. for a summary [32]). In line with a hyper-excitation phenotype, genetically removing APC from excitatory forebrain neurons using CamKII-cre results in increased dendritic spines on hippocampal CA1 neurons, increased excitatory neurotransmission, and autistic-like behavior in adults (31). When neonatal APC cKO mice were examined, robust behavioral spasms and abnormal high-amplitude discharges on the neonatal EEG were observed, both peaking in frequency at P9 (35). In the neonatal cortex, increased excitatory input onto layer V pyramidal cells was reported, similar to what is seen in the adult hippocampus. Approximately 80% of adult APC cKO mice showed electro-graphic/behavioral seizures characterized by head-bobbing, tail stiffening, forelimb clonus and freezing. Seizures were not frequent and tended to cluster. To date, there have been no published reports of how APC cKO mice respond to IS standard of care treatments.

TSC Mouse Model

TSC is a common cause of IS. We will not go into detail regarding TSC models (reviewed here [34]), but a recent report indicates spasms occur in a mouse models of TSC (Tsc1 ^+/− [35]). Importantly, most of the abnormal EEG recordings in neonates (P9–18) suggest tonic–clonic seizures rather than spasms, but more work could elucidate the specific spasm versus seizure categorization.

Acquired Models of Chronic Infantile Spasms

TTX Rat Model

Unilateral infusion of the sodium channel blocker tetrodotoxin (TTX) into the dorsal hippocampus from P10–40 in rats generates a robust model of IS (36). Approximately one-third of TTX-infused animals exhibit behavioral spasms, which start ≈1 week after infusion begins and increase in frequency over time. Spasms often, but not always, occur with an abnormal EEG signature characterized by a high-amplitude slow wave followed by electrodecrement with superimposed fast activity. Focal seizures also occur in the TTX model and are characterized by tonic head posturing and forelimb clonus. Importantly, this is the only rodent model that generates a hypsarrhythmia EEG signal, making it especially interesting to IS-relevant circuit dysfunction. Hypsarrhythmia generated in the TTX model includes high-frequency oscillations that occur most commonly contralateral to the site of TTX infusion (37, 38). The TTX model has been reported to respond to vigabatrin treatment (39), but no studies of ACTH treatment have been published.

Multiple-Hit Rat Model

In the multiple hit model, doxorubicin and lipopolysaccharide are infused unilaterally into the lateral ventricle of Sprague-Dawley rat pups at P3 (40). Two days later, pups receive i.p. injections of p-chlorophenyalanine. This protocol induces structural lesions in the brain and spasms begin at P4. Spasm frequency peaks around P4–7 and are absent by P13. Epidural EEG recorded in pups show both electrodecrement and spike- and sharp-wave discharges during behavioral spasms. Similar to most IS models, behavioral spasms do not always occur simultaneously with an EEG correlate. In adult multiple-hit animals, clonic and tonic–clonic seizures are seen in approximately two-thirds of animals. Multiple-hit animals show cognitive (Barnes Maze [38]) and sociability (social chamber test) impairments. Spasms are transiently suppressed by vigabatrin treatment, suggesting that multiple-hit animals are generally refractory to standard of care treatment. One of the most exciting uses of the multiple-hit model is its application for robust drug screening approaches (41–43).

Chronic Early Stress Rat Model

Early life stress is induced by mixing SD rat pups from different litters in cages with restricted bedding and nesting materials. Animals subjected to this treatment showed behavioral spasms (≈50%) accompanied by EEG spikes. Later in life, seizures were detected in ≈10% of animals stressed early in life (44).

Future Models of Infantile Spasms on the Horizon?

Multiple research groups are focused on understanding how genes linked to IS lead to disease. This has resulted in the generation of mouse models that have disruptions in IS risk genes but that have not been validated as models of IS. Two of the most interesting of these mouse models of human IS risk genes are SPTAN1 and NR2F1. Genetic knock-out of SPTAN1 in CNS cells (Nestin-cre) leads to disrupted cortical lamination and seizures in young animals (>P20). The loss of SPTAN1 appears to specifically disrupt the structure and function of the axon initial segment and leads to complete animal mortality within the first month of life (45). Hopefully, future work from this group will examine whether behavioral spasms occur and will further explore neonatal EEG activity in Nestin-Cre SPTAN1 cKO mice. Also of great interest is NR2F1, a gene important in the maturation of inhibitory GABAergic interneurons. Interestingly, conditional loss of NR2F1 (COUP-TFI) in developing GABAergic interneurons (Dlx5/6-Cre) leads to a seizure-resistant phenotype. This may occur due to a compensatory increase in parval-bumin-positive neurons in response to the loss of other cortical GABAergic cell types (VIP- and calretinin-positive) in this model (46). A number of other IS risk genes lead to behavioral abnormalities without clear evidence of spasms or seizures. For example, STXBP1 haploinsufficient mice have impaired spatial learning and memory, synaptic abnormalities, but no reports of spasms or seizures (47). Various genetic manipulations have also been made in CDKL5. These tend to disrupt synaptic function, neuronal morphology, and cognitive and social behaviors (48, 49). To date, no manipulation of CDKL5 has shown seizures or spasms, although altered responses to convulsant treatments (50, 51) have been identified. Similar studies exist for FOXG1 (52), MAGI2/S-SCAMa (53), and MEF2C (54), demonstrating that genetic disruption leads to behavioral abnormalities without documentation of spasms or seizures.

What's Next for Preclinical Models of IS?

With all the available tools, what is the next advance the field can hope to make? Two areas are especially interesting: mechanistic understanding of IS pathology and therapeutic advancement. Our current understanding of the IS mechanism appears to focus on the role of inhibitory interneurons in IS pathology. Arx and NR2F1 are both specific regulators of interneuron maturation, while specific knockout of Arx in excitatory neurons fails to cause seizures or spasms. Additionally, genes for NMDA receptor subunits are also implicated in IS. It will be interesting to determine if disrupted NMDA-dependent activation of interneurons is involved in IS. Also important is determining what brain regions are involved in spasm generation. While most preclinical studies focus on the cortex or hippocampus, growing evidence implicates the HPA axis, striatum, and other subcortical structures. Rapidly advancing technology for in vivo recording and cell-type specific activation may shed light on this critical question. Furthermore, with more animal models becoming available, we may be able to see a “common pathway” of disease eitiology begin to emerge. Our own work using the APC cKO model suggests that multiple IS-linked genes control, or are controlled by, Wnt/β-catenin signaling. Why this is an exciting idea which we are currently pursuing, it is likely not the whole story. Many types of IS are likely unique and may necessitate a “precision medicine” approach rather than a “common pathway” treatment. With regards to therapeutic development, more IS preclinical models are useful—but with limitations. An important example is the treatment of Arx ^(GCG)10+7 with 17β-estradiol. For this treatment to be efficacious, it has to occur within the first month of life. This means that early detection and treatment of IS is essential for improving outcome. This finding also highlights another important use of preclinical models, that of providing information to motivate clinical trials. IS has two frontline therapies (ACTH and vigabatrin), and preclinical studies suggest early treatment is essential. Therefore, compelling preclinical data can help motivate both neurologists and parents to participate in early-treatment clinical studies when frontline therapies fail. This approach is essential to identifying drugs and treatment regimens that have the greatest chance of improving patient outcomes. In closing, I would like to highlight the role that the CURE Infantile Spasms Initiative made in the evidence presented in this review and for promoting a team-based approach to improving the lives of patients with infantile spasms.