Translational animal models of autism and neurodevelopmental disorders.

Autism is a neurodevelopmental disorder whose diagnosis is based on three behavioral criteria: unusual reciprocal social interactions, deficits in communication, and stereotyped repetitive behaviors with restricted interests. A large number of de novo single gene mutations and chromosomal deletions are associated with autism spectrum disorders. Based on the strong genetic evidence, mice with targeted mutations in homologous genes have been generated as translational research tools. Mouse models of autism have revealed behavioral and biological outcomes of mutations in risk genes. The field is now poised to employ the most robust phenotypes in the most replicable mouse models for preclinical screening of novel therapeutics.

Genetic causes of autism spectrum disorder

Autism spectrum disorders were originally diagnosed by Kanner and Asperger in the 1930s.[1,2] However, the diagnostic criteria were not codified until the 1994 Diagnostic and Statistical Manual of Mental Disorders (DSM).[ Astonishingly high heritability of autism spectrum disorders, reaching 90% concordance for monozygotic twins, as compared with less than 10% concordance for dizygotic twins and siblings, along with a 4:1 male:female ratio of prevalence, quickly led to an major international search for genes causing autism. By assembling large numbers of simplex and familial cases, several research consortia have discovered single gene mutations, rare and common polymorphisms, and epigenetic modifications associated with autism.[4,5] Copy number variants, including duplications of a sequence of genes within defined chromosomal loci, were reported to be relatively common in autism.[6-9] Clearly, autism is not a single-gene disorder.

To parse the role of each of these many genetic abnormalities in the etiology and symptomology of autism spectrum disorders, and in other neurodevelopmental disorders in which autism is concomitantly diagnosed, homologous genetic mutations have been generated in experimental animals. Because the targeted gene mutation technology was perfected in the mouse, mice are currently used throughout biomedical research as the primary model organism for generating transgenic and knockout mouse models of human genetic disorders. Table I and the descriptions in this review illustrate a small portion of the wealth of available mouse models of autism. In addition, recent advances in generating knockout rats[10] are leading to the development of mutant rat models of neurodevelopmental disorders.

Because concordance for autism spectrum disorder is not 100% between identical twins, whose genomes are presumably identical, environmental and epigenetic causes of autism spectrum disorders are also under investigation. Hypotheses about prenatal exposure to toxicological and immunological insults, and neuroanatomical lesions, have been modeled in mice, rats, and monkeys.[11-18] The challenge now is to understand the consequences of each of these genetic and environmental perturbations, and their interactions. Animal models employing behavioral assays relevant to the specific symptoms offer excellent translational research tools to identify the biological mechanisms underlying the core features of autism spectrum disorder.

How do we model the behavioral symptoms of autism in mice?

As defined in the DSM-IV,[3] the diagnosis of autism requires the presence of at least six symptoms, including a minimum of two measures of qualitative impairment in social interaction, one symptom of qualitative impairment in communication, and one symptom of restricted and repetitive behaviour.[19,20] The DSM-5 is expected to redefine Autism Spectrum Disorder into two symptom domains: (i) Social interaction and social communication deficits; (ii) Restricted, repetitive patterns of behavior, interests, or activities (http://www.dsm5.org/ProposedRevisions/Pages/proposedrevision.aspx?rid=94, January 2011). Associated symptoms that appear in subsets of individuals with autism include seizures, anxiety, intellectual impairment, hyperactivity, hyper-responsiveness and hyporesponsiveness to sensory stimuli, sleep disruption, and gastrointestinal distress.[20-26]

Given that the defining criteria for autism are behavioral, investigations employing mouse models require considerable insight into which specific behaviors in the mouse repertoire are sufficiently relevant to each category of the diagnostic symptoms of autism. Inclusion of behavioral assays relevant to associated symptoms further enhance the heuristic value of animal models of autism spectrum disorders. We and other behavioral neuroscientists have generated a comprehensive set of assays for social interaction, social communication, and repetitive behaviors in mice, to test hypotheses about the causes of autism.[27-45] Social approach, reciprocal social interactions, olfactory communication, ultrasonic vocalizations, motor stereotypies such as circling and vertical jumping, repetitive behaviors such as self-grooming and digging, and perserveration in spatial tasks, are now in routine use for phenotyping mouse and rat models of autism and other neurodevelopmental disorders. Procedures for assaying behaviors relevant to associated symptoms of autism, including neurodevelopmental milestones, cognitive abilities, anxiety-like tendencies, seizures, motor dysfunctions, hyperactivity, responsiveness to sensory stimuli, and altered sleep patterns in mice have been adapted from the available behavioral neuroscience literature.[46]

In each case, we begin with the human endophenotype. In what ways are mouse behaviors similar to a defining feature of autism? Luckily, Mus musculus is a social species. Laboratory mice display a social repertoire that includes approach to olfactory pheromones emitted by other mice, approach to familiar and new conspecifics, reciprocal social interactions, ultrasonic vocalizations, communal nesting, sexual and parenting behaviors, territorial scent marking, and aggressive behaviors.[47-49] Standardized methods for scoring adult social approaches, reciprocal social interactions, nesting, sexual interactions, parental behaviors, and aggressive encounters are available in the behavioral neuroscience literature.[28,50-61]

First diagnostic category

We employ social assays that have been refined from standard tests in the behavioral neuroscience literature.[48,49] These choices are designed to maximize relevance to the types of social deficits specific to autism, including playing alone with inanimate toys rather than engaging in social interactions, and inappropriate responses to social cues. To quantify tendencies to engage in reciprocal social interactions, each subject mouse is paired with a novel partner mouse, inside a testing arena that permits free interactions over a test session of 10 to 30 minutes' duration. Digital videocameras record the session for later scoring of multiple parameters of social interactions. Ratings are performed by investigators who are blind to the genotype or treatment of the subject mice. Parameters routinely scored include sniffing, following, physical contact, and allogrooming.[58,62] Automated videotracking systems can accurately score some of the simpler elements of social interaclions.[34,35,44,63,64]

Tendency to spend time with a novel mouse versus a novel nonsocial object is evaluated in a social approach apparatus (. This 3-chambered assay, which was developed by our team to provide a simple measure of general sociability,[30] is widely used as an initial, highthroughput test for social deficits in mouse models of autism.[27,32,34,38,41,64-70]

Second diagnostic category

Social communication in rodents is mostly through the emission and detection of olfactory pheromones, and perhaps to a lesser extent, the emission and detection of ultrasonic vocalizations.[40,44,55,57,59,60,71-74] Olfactory communication is assayed by time spent sniffing olfactory stimuli from novel mice, and identification of novel versus familiar mice through olfactory cues.[36,75,76]

Vocal communication is assayed by recording ultrasonic vocalizations emitted during social interactions[77] (. Number of calls and their properties are subsequently scored by investigators. Software is available for quantifying some of the simpler parameters. Different patterns of ultrasonic vocalizations are emitted by separated pups, adult males interacting with estrus females or urine from estrus females, adult females interacting with each other, and adult male residents in response to an intruder.[72,78]

The anatomical and neurobiological substrates of olfactory and ultrasonic communication in mice do not precisely map onto the biological substrates of language and visual social communication in humans. In addition, considerably more work is needed to fully understand the communication value of ultrasonic vocalizations in mice.

At present, the existing tasks offer a reasonable start for discovering mechanistic similarities between species. Face validity of these mouse social interaction and communication tasks to the tendencies of people with autism to engage in less social approach and interaction, and to respond appropriately to complex social cues, remains inferential. We cannot know what a mouse is thinking, feeling, or intending, but only the quantifiable external expressions of those internai states.

Third diagnostic category

Mice with various genetic mutations exhibit spontaneous motor stereotypies such as circling and vertical jumping, and spontaneous repetitive behaviors such as long bouts of self-grooming and excessive digging in the litter (. Assays generally focus on the number of bouts of the behavior, or the cumulative time engaged in the behavior, during a defined test session of 10 minutes or longer.[40,58,65,79-81] Restricted interests, insistence on sameness, and special interests are more challenging diagnostic features of autism to model in mice. Perseveration of spatial habits, such as difficulty in learning a new location of a re inforcer in a T-maze or water maze after the initial learning of a first location, has been employed with some success in mouse models of autism.[82,83]

Associated symptoms

Established, standardized tests are available in the voluminous behavioral neuroscience literature for most of the associated symptoms of autism.[46,84-86] Neurodevelopmental milestones are scored from postnatal day 2 to postnatal day 18 on physical attributes such as body length and eye opening, and behavioral attributes such as responses to handling and the righting reflex. Spontaneous seizures are scored with a rating scale or EEG electroencephalography. Learning and memory tests for mice include Morris water maze spatial navigation tasks, contextual and cued fearconditioned freezing after exposure to an aversive footshock, and operant nose-poke and touch-screen reinforcement schedules.

Anxiety-related tests for rodents are primarily approach-avoidance conflicts. Mice are nocturnal. They prefer to be in dimly lit, enclosed environments. The current gold standard for anxiety-like tests for mice is the elevated plus-maze, which consists of two open and two enclosed arms, raised 1 meter from the floor, thus offering the choice between enclosed spaces and a high dropoff ledge.[85] The corroborating light-dark test consists of a two-compartment apparatus in which one chamber is dark and enclosed while the other chamber is open and brightly lit.[87,88] Mice spend more time in the closed arms of the elevated plus-maze and more time in the dark compartment of the lighte↔dark apparatus. Excessive anxiety-like traits are interpreted when the preference is unusually high for the closed arms and for the dark compartment. Anxiolytic-like treatment responses are interpreted when mice venture out more frequently into the open arms of the elevated plus-maze and the brightly lit chamber of the light↔dark box.

Responses to sensory stimuli include acoustic startle, olfactory habituation and dishabituation to a series of non-social and social odors, and the hot plate and tail flick thermal tests. Hyperactivity is scored from automated parameters of locomotion in a novel open field. Unusual sleep patterns are scored by observations of the home cage during the daylight sleeping hours and during the nighttime active hours, and/or by EEG recordings. Optimal animal models should incorporate: (i) face validity, ie, close analogies to the defining features of the human syndrome; (ii) construct validity, ie, the biological dysfunction that causes the human disease, such as a gene mutation or anatomical abnormality; and (iii) predictive validity, ie, responsiveness to treatments that prevent or reverse symptoms in the human disease. The best animal models of autism and related developmental disorders will maximize face, construct, and predictive validities. At present this combination represents a very small subset of the model systems in use, particularly for neurodevelopmental disorders in which no effective therapeutics exist to test predictive validity of the animal model. The selected examples below are designed to illustrate the progress and promise of the mouse modeling approach in autism basic research and therapeutic development. Our goal is to systematically analyze the wealth of emerging animal models of neurodevelopmental disorders, understand the strengths and weaknesses of each, gain basic knowledge about phenotypic outcomes, and employ the best model systems for treatment discovery.

How do we discover therapeutics using mouse models?

The unmet medical need for effective treatments for neurodevelopmental disorders is striking. The number of reported cases of autism lias risen rapidly over the past décade.[89,90] This rapid rise is largely a fonction of better diagnostic instruments and public awareness, allhough possible environmental causes and gene x environment interactions are under investigation.[91-93]

Personal and financial costs are high, to the affected individuals, their families, schools, and health care providers. At present the only effective interventions are intensive behavioral therapies.[26,94] The only pharmacological treatments approved by the US Food and Drug Administration are risperidone and aripiprazole: Risperdol™ and Abilify.™ Their approved use is solely for the associated “irritability,” which includes aggression, self-injury, and tantrums.[95]

A major revelation from the genetic association studies is that the most frequent mutations in autism are in genes that mediate the formation and maturation of synapses, particularly the postsynaptic densities, dendritic spines, and signaling mechanisms downstream from receptors mediating excitatory neurotransmission.[96-100] Pharmacological agents that alter synaptic functions are already available to some extent, and next-generation compounds are under development.[101-104] To evaluate the ability of novel drug treatments to reverse and/or prevent the symptoms of diseases, biomedical researchers often begin by testing exploratory compounds in appropriate animal models. Robust behavioral pnenotypes with face validity to autism, In mouse models with construct validity to autism spectrum disorders, hold great promise as predinical tools for discovering effective treatments for components of autism spectrum disorders.

Because rodents are similar to humans in many aspects of biochemistry, physiology, anatomy, and genetics, mice and rats are routinely employed in biomedical research as translational systems. Compounds that reverse behavioral and biological phenotypes in mouse models of autism offer leads which may be worth pursuing in human clinical trials. However, species differences exist in drug metabolism, alternate biochemical pathways, genetic variants, and toxicology. As in any field of biomedical research employing model systems, 100% predictive validity of efficacy and practicality in humans cannot be expected.

Keeping these caveats in mind, we design rigorous methods to evaluate proposed therapeutic interventions for autism spectrum disorders for their ability to reverse and/or prevent the major phenotypes in mouse models.[29,43] Behavioral pharmacologists test acute and chronic drug treatments, across a dose range, at various time points after administration, and assay for the most robust autism-relevant behaviors, in the strongest mouse models. Examples of some of the successes are described in the next section.

Mouse models with high translational value

Cell surface adhesion glycoproteins

Cell surface adhesion glycoproteins are a primary mechanism through which connections of presynaptic axons and postsynaptic dendrites are elaborated in neuronal synapses.[97,105] Mutations in cell surface protein genes have been reported with comparatively high frequency in neurodevelopmental disorders. Individuals with autism have been identified with mutations in NEUREXIN1, NEUROLIGIN3, NEUROLIGIN4, SHANK2, SHANKS, and CNTNAP2. For each of these rare mutations, a small number of individuals with the mutations who meet the diagnostic criteria for autism spectrum disorder has been identified.[106-109] Mice with homologous mutations in these genes are available from several excellent molecular genetics laboratories and from The Jackson Laboratory repository.

Shank3 knockout mice

Shank3 knockout mice present a particularly fascinating example of the importance of the location of the mutation within the gene. The Shank3 gene includes an ankyrin repeat domain, a PDZ domain, and a Homer binding domain.[110-112] Five distinct lines of Shank3 knockout mice with mutations at these various sites were generated and phenotyped in the past 2 years.[71,81,113,114] Two lines of Shank3 knockouts containing the mutation at the ankyrin domain displayed impairments in excitatory neurotransmission and long-term potentiation, but were predominantly normal on standard measures of sociability, with only small genotype differences detected in ultrasonic vocalizations and repetitive behavior.[71,81] Inserting the mutation at the Homer binding site resulted in mice with more social interactions, primarily in the form of aggression, along with mostly normal dendritic spines, reduced long-term potentiation, and enhanced long-term depression.[113] When the mutation was in the PDZ domain, Shank3 knockouts displayed much more severe phenotypes, including high spontaneous self-grooming resulting in skin lesions, impaired sociability, reduced corticostriatal excitatory transmission, longer dendritic spines, and lower density of dendritic spines, as compared with wild-type controls.[81] These divergent outcomes of mutations at differing sites within the same gene provide a unique opportunity to understand the binding partners and their downstream signaling actions that determine the severity of symptoms in humans. For example, deficits in mGluRS signaling have been reported after Shank3 knockdown in neuronal cultures.[115] Augmentation of mGluRS activity could be beneficial in cases of autism with SHANK3 mutations, and in individuals with Phelan-McDermid syndrome, an intellectual disability syndrome in which the SHANK3 mutation is central to the 22q13 chromosomal deletion.[108]

Contactin associated protein 2 (Cntnap2)

Contactin associated protein 2 (Cntnap2), a member of the neurexin superfamily, plays a role in neuron-glia interactions and neuronal migration during early brain maturation. Mutations in CNTNAP2 are associated with autism in a small number of individuals, particularly with language disabilities.[107,116] Cntnap2 knockout mice were generated to understand the actions of this protein on brain development and autism-relevant behaviors.[40] Seizures were detected in 9 out of 10 null mutants. Social behaviors were impaired on the 3-chambered task, during reciprocal interactions, and in home cage nesting. Repetitive self-grooming was elevated. Resistance to change was seen in the Morris water maze, in which the initial learning was normal but the Cntnap2 knockouts failed the reversal test when the escape platform location was changed. Less spontaneous alternation in a T-maze was seen in the null mutants, concomitant with moderate hyperactivity Reduced number of GABAergic interneurons and impaired migration of cortical projection neurons in this line of Cntnap2 mice underlie their seizures and some of their behavioral abnormalities. The Geschwind team proceeded to test risperidone, the antipsychotic approved by the US Food and Drug Administration for the treatment of irritability in autism. At 0.2 mg/kg IP daily for 7 days, a dose and regimen which did not affect locomotion in the wildtype controls, risperidone reduced the hyperactivity and repetitive selfgrooming in Cntnapl null mutant mice.[40] Social behaviors were unaffected by the treatment with risperidone, which is an atypical antipsychotic.

Single gene mutations, chromosomal deletions, and duplications cause a variety of neurodevelopmental disorders, including Fragile X, Rett, Angelman, PraderWilli, Smith-Lemli-Opitz, Timothy, Williams, and PhelanMcDermid syndromes, and tuberous sclerosis.[97,108] A surprisingly large number of these de novo mutations code for signaling proteins that mediate the biochemical events downstream to postsynaptic neurotransmitter receptors. Interactome network analyses revealed convergences in genes that mediate transcriptional and splicing mechanisms that may be dysregulated in autism spectrum disorders.[117] Mutant mouse models of many of these syndromes have been generated.[43,44,114,118-122] While clinically distinct disorders caused by known single gene mutations suggest straightforward targets, as compared with complex disorders such as cases of autism in which the genetic substrates are unknown, increasing knowledge about the actions of downstream signaling proteins could identify pharmacological interventions which target key mechanistic sites in convergent biochemical cascades. Mice with homologous mutations are being employed as translational tools to evaluate convergent downstream target mechanisms, and to screen compounds that yield useful interventions at those sites.

Tuberous sclerosis

Tuberous sclerosis, caused by a mutation in the Tsc1 or Tsc2 gene, is characterized by benign tubers in the cerebral cortex, seizures, a high incidence of intellectual impairment, and frequent comorbidity with autism.[123,124] Tscl and Tsc2, which dimerize, are downstream targets of the PI3K/Akt postsynaptic signaling pathway elements that bind to mTORC2, and regulate mTORCl at a further downstream site.[125] mTOR, the mammalian target of rapamycin, is a serine/threonine kinase which regulates many facets of brain development and cytoskeletal organization.[126] Deletion of Tsc1 in knockout mice, hippocampal slices, or cortical cultures resulted in enlarged brains, large dysmorphic astrocytes, decreased myelination, reductions in γ-aminobutyric acid (GABA)-ergic interneurons in the cerebral cortex, and loss of mGluR-dependent long-term depression.[126,127] Mice with mutations in Tsc2 display neuronal hypertrophy, reduced long-term potentiation in hippocampal slices, impaired hippocampally mediated fear conditioning, and impaired water maze learning.[128] Treatment with the mTOR inhibitor rapamycin for 5 days reversed the fear conditioning deficit and improved water maze learning, along with reducing brain weight and increasing survival.[128]

This early demonstration of a pharmacological rescue of phenotypes in a mouse model of a neurodevelopmental disorder sparked optimism for treating disorders caused by perturbations in signal transduction.[129] In a separate mutant line, 4 weeks of treatment with rapamycin reduced the macroencephaly and increased the low social interaction in mice with a mutation in Pten, an upstream regulator of mTOR that is implicated in cancers, seizures, and autism.[38] Rapalogs, analogs of mTOR, are in clinical trials for cancers.[130] Rapalogs and compounds targeting PI3K and Akt[131] present possibilities for therapeutic interventions in neurodevelopmental disorders with underlying mechanisms in the mTOR signaling pathway.

Fragile X syndrome

Fragile X syndrome is the most frequent genetic cause of intellectual disabilities. Constriction at the end of the X chromosome, termed a fragile site, is associated with a dramatic expansion of CGG triplet repeats, which transcriptionally silence the FMR1 gene.[132,133] Fragile X mental retardation protein (FMRP) is highly expressed in the brain, where it negatively regulates the synthesis of a large number of downstream proteins.[134]'[1] Mice with a mutation in Fmr1 display impairments in long-term potentiation, unusual social behaviors, and some unusual cognitive and anxiety-related behaviors.[135-139]

One functional consequence of the FMR1 mutation is upregulation of mGluRS receptors.[140] Bear and colleagues discovered that crossing mGluRS knockout mice with Fmrl knockout mice rescued the impaired longterm depression, elevated the dendritic spine densities in the hippocampus, and attenuated seizures.[141] Negative allosteric modulators of the mGluR5 receptor were therefore postulated as potential treatments for Fragile X Syndrome. Clinical trials are in progress to test this hypothesis.[142] Approximately 30% of individuals with Fragile X syndrome meet the diagnostic criteria for autism.[143] Considering this high comorbidity, we reasoned that a treatment effective in Fragile X Syndrome might act through a pathway convergent with other risk genes for autism. [144,145] We discovered that the prototypic mGluR5 antagonist 6-methyl-2-(phenylethynyl) pyridine (MPEP), and GRN-529. a more selective negative allosteric modulators of the mGluRS receptor, reduced the high levels of repetitive self-grooming in BTBR mice,[79,146] an inbred strain that displays robust social deficits, low vocalizations in social settings, and high repetitive self-grooming and digging.[42,58,60,65,78,79,147,148] In addition, GRN-529 reduced the high levels of stereotyped jumping that characterize another inbred strain, C58/I[80,146,149] Further, MPEP reduced marble burying in Fmr1 knockout mice, reduced stereotypies in Swiss-Webster mice, and reduced repetitive self-grooming and marble burying in mice pretreated prenatally with valproic acid.[14,150,151] These reports lend credence to the notion that interventions acting through mGluR5 receptors could confer specific benefits for treating repetitive behaviors, a major component of the third diagnostic symptom of autism.

Conclusions

Promising early findings of therapeutic rescues in mouse models have energized the rational search for pharmacological treatments of autism spectrum disorder. While the optimal developmental period for pharmacological intervention remains to be determined, adults with autism will likely be recruited for the first clinical trials,[152] since the risks of adverse drug reactions are predicted to be greater in children. Challenges will include discovering the critical window during development and/or adulthood at which interventions are useful, dosages, and treatment regimens which minimize toxicity. We have taken the first step in a long journey.

Table I

Autism-relevant behavoral phenotypes in selected mouse models with targeted mutations in associated with autism.[155-169]

	Gene	Protein	Autism-relevant behavoral phenotypes
	Integrin β3	Integrin β3	Lack of preference for social novelty
			Increased repetitive self-grooming
	Nign1	Neuroligin 1	Increased repetitive self-grooming
	Nign2	Neuroligin 2	Increased repetitive self-grooming
Synaptic	Nign3	Neuroligin 3	Reduced pup ultrasonic vocalizations
cell-adhesion			Sensory abnormalities
proteins	Nign4	Neuroligin 4	Reduced sociability and vocalizations
	Neurexin-1α	Neurexin-1α	Increased repetitive self-grooming
	Cntnap2	Contactin-associated protein2	Seizures, reduced social behaviors
	Shank3	Shank3	Mild reduction in social interactions and adult vocalizations
.
	En2	Engrailed-2	Reduced social behaviors - Impaired learning and memory
Signaling and	Met	Tyrosine kinase/ hepatocyte	Cognitive deficit
developmental		growth factor receptor
proteins	Foxp2	Forkhead box protein 2	Reduced pup ultrasonic vocalizations
	Pten	Phosphatase and tensin homolog	Reduced reciprocal social interactions - Reduced sociability
.
	Avrp1	Vasopressin receptor	Impaired social recognition - Reduced reciprocal social interaction
			Reduced ultrasonic vocalizations
	Cadps2	Calcium-dependent secretion	Reduced reciprocal social interactions
Neurotransmitters		activator 2
and receptors	Gabrb3	GABA A receptor beta3 subunit	Low sociability - Lack of preference for social novelty
			Repetitive stereotyped circling behavior
	Oxtr	Oxytocin receptor	Impaired social gecognition - Reduced pup ultrasonic vocalizations
	Slc6a4	Solute carrier family 6, member 4	Low sociability
		(Serotonin transporter)	Lack of preference for social novelty