The utility of rodent models of autism spectrum disorders

Abstract

Purpose of Review

This review discusses the ways rodent models of autism spectrum disorders (ASDs) have been used to gain critical information convergent molecular pathways, the mechanisms underlying altered microcircuit structure and function, and as a screen for potential cutting edge-treatments.

Recent Findings

There is convergent evidence that impaired developmental pruning of connections may be a common finding among several mouse models of ASDs. Recent studies have uncovered impaired autophagy by pathological mTOR activation as a potential contributor to microcircuit dysfunction and behavior. ASD related disinhibition and exaggerated synaptic plasticity in multiple distinct circuits in cortex and reward circuits in striatum also contribute to social dysfunction and repetitive behaviors. New exciting molecular therapeutic techniques have reversed cognitive deficits in models of ASD, indicating that mouse models could be used for preclinical translational studies of new treatments.

Summary

Rodent models of ASDs coupled to new emerging technologies for genome editing, cell-specific functional and structural imaging, and neuronal activity manipulation will yield critical insights into ASD pathogenesis and fuel the emergence of new treatments.

I. Introduction – Autism Spectrum Disorders (ASDs)

Autism spectrum disorders (ASDs) are developmental disorders characterized by deficits in social interactions and language, as well as repetitive and restrictive behaviors. Autism is clinically heterogeneous, and often accompanied by one or more comorbidities, including intellectual disability, hyperactivity, sensory processing abnormalities, motor deficits, and seizures. The etiology of ASDs is complex and has been attributed to genetic factors as well as poorly understood non-genetic causes (reviewed in [1]). Large-scale next generation sequencing studies from large cohorts of ASD patients and controls have found highly validated de-novo and inherited genetic changes implicating dozens of susceptibility genes contributing to the disorders [2–5]. With increasing size of sequencing cohorts, this list of validated causative gene mutations is expected to increase to include several hundred genes [2]. However, identification of causative genetic changes is only the first essential step for understanding how each genetic change alters downstream molecular cascades, perturbs brain development and function and ultimately leads to autism-related behaviors. Rodent models incorporating this growing list of genetic changes are one of the prime methods used to dissect the effects of gene mutations on neuronal anatomy, connectivity, physiology and behavior. In this article, we review many of the more recent advances made in the field using mouse models of ASDs and discuss ways these models could be used to study ASD pathogenesis and find new treatments (Table 1).

Table 1.

Table demonstrating each ASD model, and associated ASD-related behaviors, functional disruptions, and response to treatments.

Model	ASD-related behaviors	Functional disruptions	Treatment	Reference
15q11-13dup	Decreased sociability Increased vocalizations in pups Reduced vocalizations in adults Behavioral inflexibility Increased grooming Hyperactivity Anxiety	Spine pruning deficits in Purkinje neurons Enhanced LTD	None	Piochon et al. 2014
16p11.2del	Reduced sociability Repetitive behaviors Hyperactivity	Alterations in dopaminergic pathways in MSNs Decreased striatal and nucleus accumbens volume	None Decreased sensitivity to risperidone	Portmann et al. 2014
BTBR	Reduced sociability Repetitive grooming Increased vocalizations in pups Altered vocalizations in adults	Decreased inhibition Alterations in multisensory integration	Oxytocin Diazepam	Bales et al. 2014 Gogolla et al. 2014
FMR1	Altered social behavior Increased vocalizations Stereotypies Repetitive behaviors Anxiety Hyperactivity Cognitive deficits	Spine pruning deficits Delayed inhibitory maturation and inhibitory function Altered developmental synchrony Delayed inhibitory maturation Enhanced LTD Increased dendritic and cortical excitability (BK-Ca channels)	mGluR antagonists BK-Ca channel openers Prenatal bumetanide	Bear et al. 2014 Patel et al. 2014 Patel et al. 2013 Gonclaves et al. 2013 Tyzio et al. 2014
Mecp2	Altered sociability Decreased vocalizations Increased grooming	Excessive spine pruning Altered inhibition	Levodopa Dopa-decarboxylase inhibitors IGF1	Jiang et al. 2014
Nlgn3	Altered sociability Repetitive behaviors Inflexibility	Alterations in dopaminergic pathways (Nac) Impaired inhibition onto MSNs Decreased perisomatic inhibition	None	Rothwell et al. 2014
PTEN	Decreased sociability Increased grooming	Decreased LTD Decreased intrinsic excitability (SK channels)	None	Takeuchi et al. 2013 Garcia-Junco-Clemente et al. 2013
TSC1	Decreased sociability Inflexibility Increased grooming Increased pup calls	Cerebellum Decreased inhibition Decreased LTD	Rapamycin/mTOR	Bateup et al. 2014
TSC2	Decreased sociability Increased grooming	Deficits in developmental spine pruning and autophagy	Rapamycin/mTOR	Tang et al. 2014
Ube3a	Decreased sociability Decreased vocalizations Repetitive behaviors	Decreased synaptic pruning Decreased excitatory neurotransmission	ASOs	Smith et al. 2011 Meng et al. 2014 Piochon et al 2014

II. Why use mouse models

Mouse model systems have long been used to provide mechanistic insights onto the etiology of disease and have been instrumental in the discovery of pharmacologic therapies for human illness. Because of their close evolutionary relationship, mice and humans share great preservation of genes, biological processes, brain circuitries, and to some extent, behaviors. Although ASDs are uniquely human disorders, many of the core deficits of the disorders can be paralleled in mice through close behavioral investigation [6,7]. This evolutionary conservation, together with our ability to employ experimental manipulations through genetic engineering and other cutting-edge technologies [8,9], may not only help us probe the underlying mechanism of the disorder, but may also lead to the development of targeted and effective therapeutic approaches that can later be translated to humans.

Rodent models of ASDs have been useful in a number of mutually reinforcing ways for ASD research. First, these models present a relevant and tractable intact biological system for understanding the complex interaction of a specific mutated gene product with other proteins, helping to define convergent molecular pathways that that can later be targeted for treatment. Second, these models can be used to define the anatomical and physiological changes in precisely defined microcircuits that may contribute to ASD. Examining these changes across multiple models carrying gene mutations in distinct molecular pathways would help identify fundamental changes in neural circuitry that are the cumulative result of the genetic mutations and the ultimate cause of social behavioral dysfunction. Accordingly, these models would help identify biomarkers that may be translated to use in humans, predicting outcome and gauging response to treatment. Third, these models would be useful for screening therapeutic effects of behavioral and pharmacological treatments. A battery of behavioral tests has been developed that can provide careful assessment of behavioral deficits in rodent models [7,10]. Behavioral measures are indispensable for gauging of effect and drug treatments and therapeutics. Fourth, the advent of models that allow temporally-specific genetic deletion and rescue of ASD-related genetic changes, would allow us to define the critical developmental windows where interventions would be effective [11–14]. Many of these genetic changes may not only alter brain development but also affect the function of fully-developed brain circuits [15]. Thus these conditional genetic perturbation techniques hold the potential for effective treatment of adults with certain forms of ASD. Finally, these models would be useful in the future for assessing the interaction of specific environmental insults and with autism susceptibility genes.

III. Search for convergence: molecular cascades (the mTOR pathway)

To better understand the core pathological mechanisms leading to ASD, it is critical to comprehend the biological function of a given gene and its protein product, and to find convergent pathways among the many genes. The best example of a molecular pathway that was identified and dissected using mouse models involves the PI3K-mammalian target of rapamycin (mTOR) signaling cascade where mutations Tuberous Sclerosis Complex 1 and 2 (TSC2, TSC2), and Neurofibromatosis 1 (NF1) and PTEN all cause syndromes which include ASD-related behaviors. Multiple studies in these mouse models have shown the beneficial effects of the mTOR inhibitor rapamycin in ameliorating anatomical, physiological and behavioral deficits [16–21]. Parallel studies in mice and humans have also shown that other seemingly unrelated ASD related proteins also be incorporated into the canonical mTOR pathway. One example is the loss of FMR1 that causes Fragile X syndrome; the lack of FMR1 also upregulates mTOR activation in dendritic spines of neurons, leading to elevated cap-dependent translation and impaired plasticity in Fragile X mice [22]. These findings suggest that effects on mTOR dependent translation could act as a final common pathway for a number of ASD-related molecular cascades. Moreover, pharmacological down-regulation of mTOR using rapamycin could be beneficial not only for the traditional mTOR related disorders but also for a wide range of other ASD syndromes. Mouse models can help us determine which syndromes would respond to particular treatments to therapies targeting the mTOR system and help us tailor clinical trials to patient populations that are most likely to respond and benefit from certain therapies.

IV: Search for Convergence: macro and microcircuits

Mouse models have also been extremely useful for helping us discover autism-related changes in cortical circuit connectivity during brain development (Table 1). Tang, Sulzer and colleagues made an important advance in our understanding of cortical connectivity by showing that the developmental pruning of cortical dendritic spines is defective in TSC2 knockout mice [23]. In addition they were able to further dissect the cause of this pruning impairment by mating their animals to mice with impaired autophagy and showing that the blockade of autophagy mediates the specific spine pruning phenotype as well as the social impairment in these animals. Interestingly, rapamycin could no longer rescue the pruning deficits in the TSC2 knockout mice when they were bred with mice with impaired autophagy. This discovery of an entirely novel molecular mechanism downstream of mTOR, which links changes in cortical connectivity with impaired social behavior, was made possible by the creative use of mouse genetic models [23]. Similar to findings in the TSC2 model, electrophysiological studies in the FMR1 model of Fragile X syndrome show a developmental deficit in the pruning of connections between L5A cortical neurons, suggesting that deficits in developmental pruning may extend across multiple models of ASD [24]. Similar deficits in pruning induced by loss of chemokine receptor Cx3cr1 in microglia, also induce deficits in social interactions and increase in repetitive behaviors, providing convergent evidence that pruning deficits could directly contribute to abnormal autism-related behaviors [25]. Excessive dendritic spine pruning in cortical neurons later in life, as seen in the MECP2 duplication syndrome mouse, also strongly correlates with onset of behavioral abnormalities, indicating that spine density needs to be precisely controlled for proper motor, cognitive and social function [26]. Impaired developmental pruning of connections in autism models is not limited to cortical pyramidal neurons, but has also been observed at the climbing fiber-Purkinje cell synapse in the 15q11-13 duplication model, suggesting that impaired developmental pruning of connections may extend to other circuits across the brain, contributing to a global deficit [27].

In addition to identifying new molecular mechanisms contributing to ASD-related changes in connectivity, mouse models of ASD have also been highly effective for localizing the precise circuit abnormalities causing specific ASD related behaviors. A prime example of this was a recent study by Rothwell, Sudhof and colleagues who through cell type specific alterations of the autism related gene NLGN3 showed that deletion of the gene in dopamine (D1) receptor positive nucleus accumbens (NAc) neurons (Drd1+ neurons) (but not in D2 receptor positive neurons [Drd2+ neurons] or cerebellar neurons) was sufficient to induce a motor phenotype [28]. Furthermore, they showed that this deletion impaired inhibition onto Drd1+-medium spiny neurons (MSNs) of the NAc [28]. This suggests that ASD-related genetic changes can result in highly specific changes onto specific microcircuit elements within specific subcortical structures. Another study in the mouse model of 16p11.2 deletion syndrome, used single-cell transcriptomics to discovered an increase in number of striatal Drd2+ neurons, and decrease in the number of Drd1+ neurons in the deep cortical layers. In this model, excitatory input onto striatal MSN in the NAc show decreased NMDA/AMPA ratios and decreased probabilities of release as assayed by paired pulse ratios. Behaviorally, the mice show hyperactivity, circling, deficits in movement control and a lack of habituation [29]. Together, both studies highlight the importance of striatal reward circuits for understanding repetitive behaviors and motor control in autism

These findings are somewhat at odds with findings in the TSC1 model that found that cerebellar dysfunction was key for all ASD related behaviors in this model. Specifically, deletion of TSC1 in cerebellar Purkinje cell resulted in abnormal social interactions, repetitive behaviors and abnormal vocalizations [17]. These findings supported earlier histological work implicating the cerebellum in autism [30–32]. They also highlight the importance of unbiased screens in searching for circuits inducing ASD-related behaviors. Further studies in multiple models would be needed to understand the complex interactions of the multiple cortical, cerebellar and subcortical regions in mediating abnormal motor and non-motor behaviors [17]. A mouse model for 15q11-13 duplication syndrome, which in humans causes autism, intellectual disability and seizures also shows profound changes in cerebellar physiology, as well as both motor and social behaviors, providing convergent evidence of cerebellar pathology in autism [27].

Loss of inhibition seems to be a convergent theme among many ASD models. There is a specific loss of inhibition in the hippocampus of TSC1 mice [33], and a massive loss of parvalbumin neuron specific inhibition in the hippocampus of NLGN3 R451C mice [34]. In contrast inhibition from cholecystokinin-positive (CCK+) basket cells was strongly increased through loss of tonic endocannabanoid signaling, suggesting highly precise cell-type specific changes to perisomatic inhibition in the NLGN3 R451C model [34]. In the FMR1 knockout mouse, there is a delayed developmental switch of excitatory to inhibitory chloride reversal potential [35], and a reduction of excitation onto fast-spiking cortical interneurons, which ultimately reduces inhibitory output [36]. These alterations correlate with delayed developmental desynchronization of network activity in the Fragile X mouse model, suggesting that alterations in synaptic connectivity directly impact network synchrony and potentially activity-dependent circuit development [37]. In the BTBR model of autism, an inbred mouse strain with severe social deficits and repetitive behaviors, loss of inhibition leads to abnormal multisensory integration in the insular cortex [38]. In both the BTBR and Fragile X models, there is diminished oxytocin-dependent decrease of intracellular chloride at birth, leading to excitatory GABAergic responses. Treatment of pregnant females with bumetanide, a blocker of the Na-Cl-K co-transporter NKCC1, prevented these pathological changes and improved autism-related behaviors by normalizing intracellular chloride concentrations and therefore the driving force for GABAergic transmission [39]. This study highlights the long-lasting effects of altered inhibition in early development. In support of these findings, bumetanide administration to a small group of children with autism resulted in some improvements in autism related behaviors; these findings will need to be replicated in larger studies [40].

These changes in inhibition will likely not only impact basal synaptic transmission but also alter synaptic plasticity. For example, TSC1 [41] and PTEN models [42] show a loss of metabotropic receptor dependent LTD in the in CA1 and dentate gyrus regions of the hippocampus, respectively, while the FMR1 model of Fragile X and the Ube3a knockout model of Angelman Syndrome both show a enhancements of mGluR-dependent LTD [43–45]. In the FMR1 model, mGluR dependent LTD pathologically persists in the absence of protein synthesis [46]. This suggests that either diminished or excessive plasticity could result in abnormal circuit function. The Fragile X studies led to the mGluR theory of Fragile X syndrome that posited that many of the physiological and behavioral deficits in Fragile X arise from increased mGluR signaling [43]. This led to several studies which showed that mGluR5 inhibition rescue cognitive deficits, auditory hypersensitivity, aberrant dendritic spine density, overactive ERK and mTOR signaling [47,48] or social behavioral deficits [49] in FMR1 knockout mice. Unfortunately these findings have not translated to successful treatment of individuals with Fragile X in clinical trials. This may require more careful selection of patients and outcome measures [50].

Changes in connectivity are functionally translated into alterations in neuronal output by engaging intrinsic ion channels. In the PTEN mouse model of autism, adult single copy deletion leads to diminished intrinsic excitability of L2/3 visual cortical neurons through increased expression of calcium-activated small conductance (SK-type) potassium channels This decreased intrinsic excitability in turn leads to decreased visual cortical response magnitudes without altering selectivity [51]. Conversely, in the FMR1 model of Fragile X, there is decreased expression of dendritic BK-type calcium activated potassium channels which increases dendritic excitability and heightens sensitivity to incoming somatosensory inputs [52]. Therefore autism-related changes in intrinsic excitability can dramatically alter sensory responses and may respond to pharmacological interventions [52,53].

New discoveries using optogenetic and DREADD (Designer Receptors Exclusively Activated by Designer Drugs)-based interventions in wild-type mice have greatly improved our understanding of specific circuit elements driving social and non-social behaviors. Increasing excitatory activity in the medial amygdala, for example, inhibits social behaviors including aggression and mating behaviors, while increasing inhibitory neuron activation in the same structure promotes social behaviors, and inhibits repetitive self-grooming behavior [54]. Similar but not identical effects were observed when activating or inhibiting amygdala projections within the ventral hippocampus, highlighting the importance of this projection for regulating social behavior [55]. Future studies in mouse models of autism can focus on dissecting microcircuit changes in the medial amygdala or determine whether pharmacological or cell-specific treatments in the medial amygdala can be used to treat social behavioral dysfunction in these models.

V: The utility of rodent models: testing new generation treatments

Mouse models can also enable us to test a new generation of rationally designed treatments for ASDs in validated models. A remarkable example is the use of antisense oligonucleotides (ASOs) against the nuclear-localized long non-coding RNA, UBE3A antisense transcript (UBE3A-ATS) which silences the paternal copy of UBE3a in a model of Angelman’s syndrome. This treatment caused sustained unsilencing of paternal UBE3a, both in-vitro and in-vivo, and improved cognitive deficits in the mouse model of the disorder [56]. While many details remain to be worked out on timing and mode of delivery of the ASOs, this approach shows the essential role of mouse models for serving as an intermediate pre-clinical step for the development of novel human therapeutics.

Conclusions

The utility of rodent models for autism will increase non-linearly with improvements in emerging transgenic, genome editing, cellular functional imaging, and activity modulation techniques [8,57,58]. These techniques will allow us to genetically perturb highly specific neuronal subpopulations and follow activity patterns in these in cells and their connected neighbors before, during, and after potential treatments, over months, as rodents interact with other animals [59]. Mouse models will complement the results obtained in other in-vitro models such as induced pluripotent stem cell derived cultured neurons [60]. Most importantly understanding of basic neuron circuit function at the most fundamental level will likely yield the most impact for understanding ASDs in the long run.

Key Points.

1. Rodent models of ASDs are useful for unraveling the molecular pathways associated with ASD-related gene mutations.

2. Rodent models of ASDs can help us find convergent microcircuit changes associated with ASDs.

3. Rodent models of ASDs may be useful for preclinical behavioral studies before testing drugs in humans.