Modeling Dopamine Dysfunction in Autism Spectrum Disorder: From Invertebrates to Vertebrates

Abstract

Autism Spectrum Disorder (ASD) is a highly heterogeneous neurodevelopmental disorder characterized by deficits in social communication and by patterns of restricted interests and/or repetitive behaviors. The Simons Foundation Autism Research Initiative’s Human Gene and CNV Modules now list over 1,000 genes implicated in ASD and over 2,000 copy number variant loci reported in individuals with ASD. Given this ever-growing list of genetic changes associated with ASD, it has become evident that there is likely not a single genetic cause of this disorder nor a single neurobiological basis of this disorder. Instead, it is likely that many different neurobiological perturbations (which may represent subtypes of ASD) can result in the set of behavioral symptoms that we called ASD. One such of possible subtype of ASD may be associated with dopamine dysfunction. Precise regulation of synaptic dopamine (DA) is required for reward processing and behavioral learning, behaviors which are disrupted in ASD. Here we review evidence for DA dysfunction in ASD and in animal models of ASD. Further, we propose that these studies provide a scaffold for scientists and clinicians to consider subcategorizing the ASD diagnosis based on the genetic changes, neurobiological difference, and behavioral features identified in individuals with ASD.

1. Introduction

Autism Spectrum Disorder (ASD) is a highly heritable neurodevelopmental disorder estimated to affect 1 in 59 children (Baio, 2012). This disorder is characterized by deficits in social communication and by patterns of restricted interests and/or repetitive behaviors that persist throughout life (“Diagnostic and Statistical Manual of Mental Disorders ∣ DSM Library,” n.d.). Since the initial description of autism by Leo Kanner in 1943 (Kanner, 1943), mounting evidence has suggested there is no single cause of ASD. Rather, the diagnosis of ASD is structured around a core set of behavioral symptoms that serve to unify individuals with a heterogeneous collection of genetic and mechanistic differences.

Recurrence within families and twin studies have implicated a strong genetic component to ASD (Bailey et al., 1995; Bolton et al., 1994; Constantino et al., 2010; Sumi et al., 2006). However, this disorder is not monogenic – there is not one gene responsible for all causes of autism. A growing number of genetic changes, including de novo single nucleotide variants (Iossifov et al., 2012; Neale et al., 2012; Sanders et al., 2012) as well as both de novo and transmitted copy number variants (CNVs) (Levy et al., 2011; Pinto et al., 2014, 2010; Sanders et al., 2015) have been identified in individuals with autism. Concordance rates and linkage studies suggest a multigenic inheritance pattern, although a subset of ASDs may have a monogenic etiology (Risch et al., 1999).

Within this context, a key question is how these various genetic risk factors ultimately result in the constellation of behavioral symptoms that are the hallmark of autism. Many have suggested that these seemingly disparate genetic risk factors ultimately converge on downstream mechanisms. However, the high clinical and genetic heterogeneity of ASD instead suggests that rather than considering ASD as a disorder with a singular pathway abnormality, it may be more appropriate to define subgroups of the ASD population based on genetic risk factors, biomarkers, or specific clinical presentation (e.g. area(s) of greatest impairment(s), existing comorbidities, etc.). These subgroups could be a result of mechanistic commonalities, and thus could be used to delineate the populations that might benefit from differential, targeted therapies.

In this review, we will consider evidence for one such possible subgroup: individuals in whom dopaminergic dysfunction may underlie part or all of their autism symptomatology. We will consider evidence for primary dopaminergic system dysfunction (i.e. dysfunction of either dopaminergic brain structures or brain structures receiving dopaminergic input) in both individuals with ASD and in various animal models that demonstrate many of the phenotypic characteristics of ASD, from Drosophila melanogaster to the mouse. We will discuss the growing evidence suggesting DA dysfunction as a contributing factor to ASD, the endophenotypes of ASD that may be characteristic of this group, and possible therapeutic strategies for such individuals. We will first discuss the anatomy and physiology of the dopaminergic system to briefly review the requisite knowledge for our discussion of ASD and the possible role of dopamine (DA) in ASD.

2. Autism Spectrum Disorder

2.1. Core symptomatology

The fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-V) defines ASD as a neurodevelopmental disorder characterized by two key diagnostic criteria. The first is core deficits in social communication and social interaction. This includes persistent impairment in social reciprocity, nonverbal communication, and in developing, maintaining, and understanding relationships. Examples of such social deficits include (but are not limited to): avoiding eye contact, flat or inappropriate facial expressions, failure to understand personal space boundaries, deficits in reciprocal to-and-fro communication, and difficulty using and understanding gestures, tone of voice, or body language. The second diagnostic criterion is restricted, repetitive patterns of behavior, interests, or activities. This can include such behaviors as repetitive motor movements, inflexible adherence to routines, highly restricted interests that are abnormal in intensity, and hyper- or hyporeactivity to sensory input. The addition of hypo- or hyperreactivity to sensory input to the diagnostic criteria for ASD in the DSM-V is a tribute to the growing recognition of and evidence for sensory processing and perceptual differences in ASD (Baum et al., 2015; Cascio et al., 2016; Lane et al., 2010; Robertson and Baron-Cohen, 2017; Stevenson et al., 2014).

The behavioral manifestations of ASD are highly variable and depend on the severity of the condition, the patient’s developmental level, and the patient’s chronological age (although ASD is a life-long disorder). The current diagnostic rubric for ASD encompasses a number of disorders under previous diagnostic guidelines including: infantile autism, childhood autism, Kanner’s autism, high-functioning autism, atypical autism, pervasive developmental disorder not otherwise specified, childhood disintegrative disorder, and Asperger’s disorder. These prior classifications highlight the true heterogeneity of the ASD population and the need to consider ASD by subtype in our scientific endeavors.

The National Institutes of Mental Health Research Domain Criteria (RDoC) initiative suggests such a way to classify individuals with ASD and other neuropsychiatric disorders. The DSM relies on assemblages of clinical symptoms to determine diagnosis and, in doing so, neglects to incorporate objective biological measurements into diagnosis. This limits the ability of scientists and clinicians to effectively subcategorize patients with ASD and other neuropsychiatric disorders. RDoC proposes to classify mental disorders based on not only the clinically observable behavior (the basis of the DSM-5), but also on laboratory-based neurobiological measures, such as genetic makeup and psychophysical and cognitive evaluations. This framework, by incorporating biology, behavior, and context, is more directed toward identifying the mechanistic underpinnings of the various manifestations of autism, and thus could hold greater promise for personalized or precision treatment approaches. The RDoC framework is particularly important in conditions as heterogeneous as ASD, and raises an immediate set of questions around the disorder and which are centered on what measures or co-morbidities (i.e. co-occurring conditions) could be used to subcategorize ASD.

2.2. Physiologic and psychiatric comorbidities of ASD

To explore possible subcategories of the ASD diagnosis, it is important to assess the conditions and maladaptive features that can co-occur with ASD. In addition to the aforementioned core symptomatology that defines autism, there exist a number of psychiatric conditions and physiologic features and disorders that frequently co-occur with autism. The specific presentation and combination of these comorbidities varies greatly across individuals with autism. However, the patterns in which these features co-occur may ultimately inform our understanding of ASD and permit the identification of diagnostic subcategories within the ASD diagnosis. While a complete review of all disorders and behavioral features frequently co-occurring with ASD is beyond the scope of this review, we will consider some of the most common disorders and features and how they may provide the scaffolding for partitioning the ASD population based on shared neurobiological and behavioral features. Of great utility in this endeavor is a study performed in 2014 that used an electronic health record time-series analysis to identify three comorbidity clusters in ASD (Doshi-Velez et al., 2014): one characterized by psychiatric disorders, one characterized by multisystem disorders (including gastrointestinal disorders), and one characterized by seizures (Doshi-Velez et al., 2014). These subgroups encompass the most common comorbidities of ASD and provide a useful context for our discussion.

Focusing first on psychiatric comorbidities, in one study of 112 children with autism, 70.8% had at least one current psychiatric disorder in addition to autism and 41% had two or more (Simonoff et al., 2008). The most common comorbid psychiatric disorders in these children were attention deficit hyperactivity disorder (ADHD; 28.2%, 95% confidence interval [CI] 13.2-45.1), oppositional defiant disorder (ODD; 28.1%, 95% CI 13.9-42.2), social anxiety disorder (29.2%, 95% CI 13.2-45.1), and generalized anxiety disorder (13.4%, 95% CI 0-27.4) (Simonoff et al., 2008). A larger and more recent study of 658 children with ASD seeking treatment for disruptive behavior demonstrated high rates of ADHD (81.2%), ODD (45.5%), and any anxiety disorder (45.5%) in those individuals (Lecavalier et al., 2019). In this particular study, 66.1% of the sample had two or more concomitant psychiatric disorders (Lecavalier et al., 2019).

The particularly high rate of ADHD in children with ASD is of great interest as this could point both to a shared neurobiological etiology for at least some individuals with these disorders as well as a potential diagnostic subcategory within ASD (e.g. ASD with ADHD). ADHD is characterized by inattention, hyperactivity, and impulsivity. ADHD alone can impair social functioning (Harpin et al., 2016; Ros and Graziano, 2018) and can lead to low levels of reciprocal friendships (Boer and Pijl, 2016), as is also seen in ASD. Individuals with ADHD may, for example, talk excessively and interrupt or intrude into the conversations and activities of others, which are behaviors that can also be associated with ASD. It is not surprising therefore that, when ADHD co-occurs with ASD, this disorder may exacerbate the social dysfunction pathognomonic of ASD.

While the precise etiology of ADHD is elusive, disruptions in the neurochemical environment created by the neuromodulators DA and NE are thought to play a central role in this condition (Arnsten and Pliszka, 2011; Sharma and Couture, 2014). Many studies directly link DA dysregulation and ADHD. PET imaging in individuals with ADHD shows that these individuals have reduced DAT and D2/D3 receptor availability in the midbrain and accumbens (Volkow et al., 2009). A recent systematic meta-analysis of DA receptor genes and ADHD demonstrated that there is strong evidence for the association of specific variants of the D4 receptor and ADHD (Wu et al., 2012). Taken together with the high rates of ADHD in ASD, it is possible that dysfunction in either of these neuromodulatory systems could contribute to the etiology of ASD, especially in those individuals with concurrent ASD and ADHD.

Functional gastrointestinal disorders (fGID) are reported in 30-70% of individuals with ASD (Penzol et al., 2019). In one study of 845 individuals with ASD, at least one fGID was present in 30.5% of these individuals (Penzol et al., 2019). The Rome Foundation defines fGIDs as disorders of gut-brain interaction (DGBI). This group of disorders is classified by gastrointestinal (GI) symptoms related to any combination of motility disturbances, visceral hypersensitivity, altered mucosal and immune function, gut microbiota, and/or central nervous system processing (Schmulson and Drossman, 2017). This combination of symptoms causes an illness experience in the patient that is not due to clearly identifiable anatomic disorder (Penzol et al., 2019). It has been shown that children with ASD are more likely to experience the symptoms of abdominal pain, constipation, and diarrhea than those without ASD (McElhanon et al., 2014). Some have suggested that these GI symptoms may exacerbate the behavioral symptoms exhibited by children with ASD by presenting a source of emotional distress that then manifests as problematic behaviors (Thulasi et al., 2019). Indeed, maladaptive behaviors correlate with GI issues in individuals with ASD (Chaidez et al., 2014). In children with ASD, behavior scores for irritability, social withdrawal, stereotypy, and hyperactivity are significantly higher in children with frequent abdominal pain, gaseousness, diarrhea, and constipation (Chaidez et al., 2014).

There is also evidence suggesting that metabolites produced by microbes in the gut (including neurotransmitters such as DA and 5-HT (Clarke et al., 2014)) can influence brain function and behavior (Li and Zhou, 2016). For example, germ-free (GF) mice (who lack bacterial colonization of the intestine) have several behavioral and neurobiological differences when compared with colonized mice. GF mice exhibit significant increases in hippocampal 5-HT compared with colonized animals (Clarke et al., 2013). Such mice also demonstrate increased metabolism of NE, DA, and 5-HT in the striatum compared with mice with normal gut microbiota and reduced expression of proteins required for synaptogenesis (specifically synaptophysin and PSD-95) in the striatum, suggesting that gut microbiota can affect normal brain development (Heijtz et al., 2011). Moreover, these neurobiological differences are accompanied by behavioral changes including increased motor activity (Heijtz et al., 2011), increased anxiety-like behavior (Heijtz et al., 2011), and abnormal social behaviors (Desbonnet et al., 2014), suggesting a behaviorally-relevant impact of such neurobiological changes.

Many studies have reported differences in the composition of the gut microbiota in patients with ASD (Dinan and Cryan, 2015; Strati et al., 2017; Tomova et al., 2015) and it is possible that these differences in gut flora may contribute to both behavioral and GI symptoms (Li and Zhou, 2016; Vuong and Hsiao, 2017). Fecal bacterial analyses have shown increased abundance of Clostridium (Parracho et al., 2005), Lactobacillus, and Desulfovibrio species in children with ASD compared with typically developing controls, with the amount of Desulfovibrio species correlating with ASD severity (Tomova et al., 2015). However, while several studies have shown differences in the composition of the intestinal microbiome between ASD and non-ASD subjects, there is variability between studies regarding the specific bacterial species which are altered (Finegold et al., 2010; Parracho et al., 2005; Tomova et al., 2015), which raises questions regarding the directionality of the effect of observed changes in gut flora in ASD.

As communication between the brain and the gut is bidirectional, neurobiological differences in ASD may drive dysfunction of the enteric nervous system and/or drive changes in microbiota composition (Collins et al., 2012; Sun et al., 2013) that then contribute to the GI symptoms experienced by individuals with ASD. The enteric nervous system (ENS), which consists of a system of neurons that spans the length of the digestive system, serves to regulate digestive function (Spiller and Grundy, 2004). As many neurotransmitters and signaling pathways are common between the ENS and the CNS (including DA, 5-HT, and acetylcholine), pathophysiologic changes affecting CNS function can similarly impact ENS function (Rao and Gershon, 2016). ENS deficits are reported to co-occur with a number of CNS disorders that involve disturbances in the DA system. One such disorder is Parkinson’s Disease (PD), which as described above, is defined by degeneration of nigrostriatal dopaminergic neurons resulting in a movement disorder characterized by rigidity, hypokinesia, resting tremor, and postural instability. Dopaminergic neurons are also found in the ENS and are important for proper gut motility (Li et al., 2006). Studies in mice (Anderson et al., 2007; Wang et al., 2012) and humans (Singaram et al., 1995) have demonstrated that dopaminergic neurons in the ENS are similarly susceptible to degeneration in PD. Indeed, PD is often accompanied by disturbances in GI motility including dysphagia, impaired gastric emptying, and constipation(Jost, 2010), which are thought to be due to ENS DA neuron degeneration and dysfunction.

Epilepsy has been recognized as a common comorbidity of ASD for decades (Tuchman et al., 2010). Epilepsy is defined as the occurrence of two or more unprovoked seizures, which are the physical manifestation of sudden, abnormal, excessive, hypersynchronous neuronal firing (Adams et al., 1997; Bromfield et al., 2006). Studies estimate this neurologic disorder occurs in 8.6%-22% of individuals with ASD (Bolton et al., 2011; Surén et al., 2012; Thomas et al., 2017), a prevalence considerably higher than that reported in all children (0.32%-4.4%) (Camfield and Camfield, 2015). Importantly, children with ASD and epilepsy have higher rates of hyperactivity and are at an increased risk for more severe autism symptoms and maladaptive behaviors due to the increased likelihood of these children having low IQ (Viscidi et al., 2014). Children with ASD and epilepsy (but without intellectual disability) have higher rates of irritability and hyperactivity symptoms (Viscidi et al., 2014). Given the high rates of epilepsy in ASD, there likely exists a shared biological etiology for these two conditions when they co-occur. Indeed, several genetic syndromes have been identified in which ASD and epilepsy co-occur, including conditions caused by mutations in single genes (or monogenic disorders) and by genomic copy number variation (Lee et al., 2015). Evidence points toward shared neurobiological features between these two conditions, including alterations in cortical architecture and changes in the excitatory-inhibitory neurotransmission balance (Frye et al., 2016) (which we will discuss separately). Both of these neurobiological changes, alone or in combination, could lead to instability of neural networks, resulting in altered neuronal excitability, and ultimately seizures and the behaviors associated with ASD (Casanova et al., 2002a, 2003; Rubenstein and Merzenich, 2003).

These clinical comorbidities demonstrate two important concepts. First, that considering co-occurring conditions in individuals with ASD is potentially an effective means to subcategorize the ASD diagnosis. Second, that a shared genetic etiology between ASD and these comorbidities may point us toward an understanding of the neurobiological mechanisms within these subcategories. Therefore, perhaps as important to the identification of subtypes of ASD as considering common comorbidities is to consider the genetic susceptibilities and variants that may give rise to dysfunction in specific neurobiological circuits. Doing so will not only improve identification of individuals with ASD by determining genetic risk factors, but also will improve clinical efficacy in identifying possible comorbidities of these genetic changes which may complicate or exacerbate the clinical presentation of ASD in these individuals. We will briefly review the genetic and environmental risk factors associated with ASD and how these may point us toward a better understanding of the various forms of ASD that may exist and their unique neurobiologies.

2.3. Genetic and environmental risk factors

The earliest evidence for the contribution of genetic factors to autism came from twin and family studies conducted in the 1970s (Folstein and Rutter, 1977). Since these earliest studies, twin studies have shown pairwise concordance rates of 10-31% for dizygotic (DZ) and 60-92% for monozygotic (MZ) twins (Bailey et al., 1995; Folstein and Rutter, 1977; Hallmayer et al., 2011; Rosenberg et al., 2009). Sibling recurrence rates have been estimated between 7% to 25.9% and increase to 13.5% to 32.2% if 2 or more siblings are diagnosed with ASD (Grønborg et al., 2013; Ozonoff et al., 2011). A study of children born in Sweden found that risk of autism is increased 10-fold if a full sibling has a diagnosis (Sandin et al., 2014). These studies collectively demonstrate the high heritability of ASD and emphasize the importance of dissecting the genetic etiology of ASD.

Syndromic causes of ASD, such as fragile X syndrome (FXS) and Rett syndrome, were among the first identified genetic causes of ASD (Amir et al., 1999; Brown et al., 1982; “Identification of a gene (FMR-1) containing a CGG repeat coincident with a breakpoint cluster region exhibiting length variation in fragile X syndrome,” 1991). Syndromic forms of ASD are typically caused by chromosomal abnormalities or single gene mutations and differ from non-syndromic ASD in their association with co-occurring phenotypes or dysmorphic features (Sztainberg and Zoghbi, 2016). While syndromic forms of autism are now considered separate clinical entities from non-syndromic ASD, their genetic bases can provide important insights that may help better elucidate the genetic underpinnings of non-syndromic ASD.

In 2007, de novo copy number variants (CNVs) were first associated with autism (Sebat et al., 2007) and, in 2012, whole-exome sequencing revealed the contribution of de novo single-nucleotide variants (SNVs) to ASD risk (Sanders et al., 2012). CNVs, segments of DNA 1kb or larger present in the genome at a variable copy number in comparison to a reference genome(Redon et al., 2006), include deletions, duplications, triplications, insertions, and translocations (Stankiewicz and Lupski, 2010). Well-known CNVs include large-scale chromosomal abnormalities, such as trisomy 21 in patients with Down syndrome, but smaller-scale microdeletions and microduplications have also been linked with various diseases (Campbell et al., 2019). In comparison, SNVs are single base differences in the genome. CNVs and SNVs are considered de novo when these genetic changes are present for the first time in a family member as a result of spontaneous genetic variation during the formation of germ cells (sperm or egg) or after the union of sperm and egg cells (Leyfer et al., 2006; Roach et al., 2010). An SNV is rare if it occurs with a minor allele frequency of less than 1% and a CNV is considered rare if it occurs in less than 1% of the population (Bomba et al., 2017; Grozeva et al., 2010).

Both rare SNVs (especially de novo mutations (Iossifov et al., 2012; Neale et al., 2012; Sanders et al., 2015)) and common SNVs are now known to contribute to risk for ASD (Gaugler et al., 2014). Indeed, whole exome sequencing has shown that de novo insertions and deletions that disrupt the proteins that are encoded by the affected genes are significantly more common in affected individuals than in their unaffected siblings (Iossifov et al., 2012). Chromosomal microarray studies have shown that de novo CNVs are found in 8-10% of sporadic ASD cases (Levy et al., 2011; Sebat et al., 2007), however de novo SNVs in protein-coding regions of the DNA are more common than large de novo CNVs (Krumm et al., 2014). While common and de novo variants in both copy number and single nucleotides underlie a portion of the genetic etiology of ASD (Gaugler et al., 2014; Sanders et al., 2012), the detection of SNVs in individuals with ASD in particular permits us to consider the contribution of discrete, single-gene changes on specific neurobiological pathways. This ultimately will lead to improved understanding of the contribution of dysfunction in these specific pathways to the behaviors associated with ASD and its common comorbidities. It is therefore of utmost importance to continue identifying and studying these SNVs to further define subgroups of ASD based on the observed neurobiological and behavioral consequences of these mutations.

To date, over one thousand genes have been associated with risk for ASD. The Simons Foundation Autism Research Initiative’s Human Gene and CNV Modules list 1,089 genes implicated in ASD and 2,291 CNV loci reported in individuals with ASD, respectively (“Human Gene Module,” n.d.). The roles of these genes in brain development and function are as varied in scope as they are in number. Some are involved in brain development, including the production, growth, and organization of neurons, while others are involved in the formation and maintenance of synapses and ongoing cell-cell communication (Chahrour et al., 2016). A more thorough understanding of the function of these genes and how changes in these genes impact their function is required before we can begin to understand how they may inform the creation of subtypes of ASD based upon common neurobiological differences.

To state that genetic risk is the only, or even the most important, factor contributing to the development of autism would be a misstatement. There is overwhelming evidence suggesting that genetic risk alone may not be sufficient to drive the development of ASD. This is apparent even in the seminal twin and sibling studies of ASD. These studies found that while the concordance rates of ASD in MZ twins is high, it is incomplete (Bailey et al., 1995; Hallmayer et al., 2011; Rosenberg et al., 2009). Furthermore, genetic risk factors for ASD can be found in individuals who do not meet the diagnostic criteria for ASD (Robinson et al., 2016). This has led to great work dedicated toward identifying both environmental and sex-based risk factors.

There is a well-documented male-biased prevalence of ASD (Loomes et al., 2017; Xu et al., 2019), with a 3:1 male-to-female ratio. This has led to suggestion of a “female protective effect” (Robinson et al., 2013; Sanders et al., 2015; Werling and Geschwind, 2013), in which a yet incompletely understood female biological mechanism is protective against the development of ASD. This is evidenced by work demonstrating that genetic load of functional de novo mutations (i.e. loss-of-function and predicted deleterious mutations) is higher in females with ASD as compared to males with ASD suggesting a higher minimum genetic burden required to develop ASD in females. Additionally, recent in vivo studies demonstrate that increased genetic risk burden in males with ASD is associated with higher functional connectivity between the substantia nigra and somatosensory regions of the brain, while genetic load had no correlation with functional connectivity in females with ASD. These findings suggest that genetic risk differentially impacts brain connectivity in males and females with ASD (Lawrence et al., 2021). However, there are data to suggest that non-biological factors and methodological biases contribute toward this apparent male-bias prevalence. Indeed, data suggest that females are less likely than males to meet the diagnostic criteria for ASD with equivalent levels of autism traits (Dworzynski et al., 2012) and the diagnosis of ASD is often delayed in females compared to males (Begeer et al., 2013; Giarelli et al., 2010). This may be due to subtle differences in symptom manifestation, differences in referral practices, and gender bias in assessment tools (Kreiser and White, 2014).

In addition to sex differences that may contribute to ASD risk, there is evidence to suggest that environmental factors may play a role in ASD risk. It has been suggested that environmental events internal or external to an individual, especially during critical developmental periods, lead to altered gene expression via epigenetic mechanisms or differential activation of cellular signaling pathways, which in turn lead to permanent changes in brain development and ultimately behavior (Tordjman et al., 2014). There exists evidence from human studies to support such a hypothesis in neuropsychiatric disorders. Specifically, parental adversity in the early post-natal period is correlated with lifetime prevalence of neuropsychiatric disorders (Green et al., 2010). Additionally, the prevalence of perinatal complications including fetal distress during delivery, caesarean section delivery, and poor condition at birth is higher in children diagnosed with ASD than in nonaffected siblings (Glasson et al., 2004). Additionally, maternal infection in the first trimester and maternal bacterial infection in the second trimester of pregnancy are associated with diagnosis of ASD in the offspring (Atladóttir et al., 2010). While these studies do not provide definitive proof of causality, they are suggestive of a role for environmental factors in the development of neurodevelopmental and neuropsychiatric disorders, including ASD.

There is direct evidence demonstrating a link between various environmental factors and exposures and risk for ASD. For example, in utero exposure to the anticonvulsant valproic acid (VPA) is associated with a significantly increased risk of autism (with an absolute risk of 4.42% in children exposed to VPA in utero compared to 1.53% in children not exposed to VPA in utero) (Christensen et al., 2013). A systematic review of 111 publications examining estimated environmental toxicant exposures and toxicant biomarkers in ASD found that toxins and toxic exposures, such as pesticides, phthalates, polychlorinated biphenyls, solvents, toxic waste sites, air pollutants, and pesticides may promote neurotoxic mechanisms leading to the development of ASD (Rossignol et al., 2014). However, this review also found that the majority of these studies have significant limitations including small sample size, inadequate matching of controls and cases, retrospective design, and recall and publication biases that limit their validity (Rossignol et al., 2014). This does not exclude the possibility that toxins and toxic exposures may play a role in the pathogenesis of autism but does underline the necessity of future studies designed to prospectively determine the role of environmental exposures in ASD risk.

The possible role of environment and sex-based differences in ASD risk only serves to further highlight the importance of modeling detected de novo mutations and rare inherited variants in the laboratory to determine their contribution to the observed clinical presentation of ASD. These models allow us to examine the molecular and neurophysiological contributors to ASD and its comorbidities in isolation from, or in controlled combination with, potentially interacting influences. Indeed, a number of animal models of the identified de novo and rare inherited genetic variants associated with ASD have been created. These models allow the discovery of important insights regarding the differential impact of these mutations on neurobiological pathways and how these neurobiological changes may promote the behaviors associated with autism. Before discussing these models, it is important to review some of the prevailing perspectives on the neurobiological changes that have been hypothesized to result in or contribute to ASD.

2.4. Neurobiological theories of autism

Given the immense variability in the clinical presentation of children with ASD and the numerous genetic risk factors associated with ASD described above, a single cause of ASD is unlikely to exist. Many theories attempt to unify the myriad genetic and environmental risk factors now associated with ASD by suggesting shared changes in synaptic function or structure, cortical architecture, and/or network function as a result of these risk factors. Of these many theories, among the most popular are the excitatory-inhibitory balance (E/I balance) disruption theory (Gogolla et al., 2009; Nelson and Valakh, 2015; Rubenstein and Merzenich, 2003), the altered network connectivity theory (Anderson et al., 2011; Hahamy et al., 2015a; Schipul et al., 2011; Supekar et al., 2013), and the altered predictive coding theory (Pellicano and Burr, 2012; Van Boxtel and Lu, 2013). In the following, we will consider each of these theories individually and also speak to the commonalities they share.

The E/I balance disruption theory proposes that ASD is due to alterations in the ratio of excitatory to inhibitory neurotransmission (Casanova et al., 2003; Rubenstein and Merzenich, 2003). Much data exists to support alterations in both excitatory and inhibitory neurotransmission in the brains of individuals with ASD. A large portion of this research has focused on alterations in neuronal activity and network communication in response to the primary inhibitory neurotransmitter in the brain, GABA, and the primary excitatory neurotransmitter in the brain, glutamate. In a meta-analysis of multiple ASD mouse models, Gogolla and colleagues noted reductions in the number of parvalbumin-positive, GABAergic interneurons (a specific inhibitory interneuron subpopulation) in the cerebral cortex (Gogolla et al., 2009). In another review, Cellot et al. describe alterations in GABAergic signaling in eight of the most commonly used animal models of ASD (Cellot and Cherubini, 2014). Findings in human cortical tissue also support alterations in the E/I balance. Disruptions in cortical minicolumn architecture have been reported in the brains of individuals with ASD (Casanova et al., 2002a, 2002b). Since the formation and function of these cortical minicolumns is dependent on lateral inhibition by GABAergic interneurons during development (Favorov and Kelly, 1994a, 1994b), the observed changes may be due to altered inhibitory neurotransmission. One of the common comorbidities associated with ASD also supports changes in the E/I balance. As discussed above, there are high rates of seizures in individuals with ASD (Bolton et al., 2011; Surén et al., 2012; Thomas et al., 2017). Seizures are the result of excessive, hypersynchronous neuronal activity, consequently reflecting an imbalance between excitatory and inhibitory neurotransmission. Therefore, in individuals with ASD and epilepsy, a shift in the E/I balance could govern the observed clinical phenotype. We must note there are many pathologic processes that can disturb the E/I balance beyond primary disturbances in GABAergic interneuron populations, including disruption of extracellular ion homeostasis, changes in cellular metabolism, altered receptor function, or altered neurotransmitter uptake (Bromfield et al., 2006). One example of such an indirect effect can be seen in the striatum. As described previously, a large population of GABAergic neurons, called medium spiny neurons (MSNs), reside in the striatum and receive input from the thalamus, various association areas of cortex, sensory cortex, as well as from dopaminergic neurons projecting from the SNc (Yager et al., 2015). MSNs are typically active immediately prior to the initiation of a movement and may encode the decision to make a movement toward a specific target (Murray et al., 2011). By perturbing dopaminergic input to these MSNs, one could disrupt the E/I balance in the striatum. Given the important role of the striatum in action selection and reward-mediated behaviors (Murray et al., 2011; Schultz, 2013), both of which are disrupted in ASD, change in striatal function represents a plausible etiology for ASD-associated behaviors (Fuccillo, 2016). We will return to the possible role of dopamine and dopaminergic signaling in ASD in a later section.

The altered network connectivity theory holds that distortions in neuronal communication within the cerebral cortex are the primary drivers of the behaviors observed in ASD (Belmonte et al., 2004). Many studies supporting this theory rely on functional magnetic resonance imaging (fMRI), an imaging technique that measures changes in blood oxygenation associated with neuronal activity (Huettel et al., n.d.). The interregional correlations between fluctuations of the fMRI signal can be used to determine the operational interactions (i.e. functional connectivity) of various brain regions, whether or not an apparent physical connection between these regions exists (Rogers et al., 2007). Somewhat paradoxically, results of several fMRI studies reveal both decreases (Belmonte et al., 2004; Just et al., 2004; Schipul et al., 2011) and increases (Keown et al., 2013; Supekar et al., 2013) in functional connectivity within and between a number of brain regions in individuals with ASD as compared to typically developed individuals. Hahamy and colleagues posit that these seemingly contradictory results could be the product of distinct distortions of the typical functional connectivity pattern in individuals with ASD (Hahamy et al., 2015a). Their study of a large database of fMRI resting state scans demonstrated that individuals with ASD have higher inter-subject variability in their connectivity patterns, both within and between hemispheres, as compared to controls. Moreover, the authors found that greater distortions of the typical interhemispheric connectivity maps correlated with more severe ASD symptoms(Hahamy et al., 2015a). Importantly, these patterns of resting state functional connectivity are thought to reflect both genetic and environmental factors, including the manner in which an individual with ASD interacts with the environment(Hahamy et al., 2015b).

The predictive coding hypothesis rests on the assumption that our brain creates internal models of the external sensory environment that serve as the basis for our predictions about, and consequently perceptions of, the world. These models (or “priors”) are generated based on incoming sensory information (so-called “bottom-up” information) and are generally updated when incoming sensory information does not match the predictions of these models (i.e. when there is prediction error). If there is a mismatch in the degree of importance (i.e. weighting) the brain places on either the internal model or on the incoming sensory information, this would lead to a fundamental shift in one’s perception of the sensory environment. According to this theory, individuals with ASD both improperly form and fail to appropriately update their internal models of the sensory environment. Some have posited this as the basis for the observed hyper- and hypo-responsivity to sensory stimuli observed in individuals with ASD (Karvelis et al., 2018). As the fidelity of the incoming sensory information and the sensory representations they inform is essential to the formation of a coherent percept of the world, any change could cascade into altered processing of social cues and communication in individuals with ASD (Baum et al., 2015; Stevenson et al., 2016). Incoming (“bottom-up”) sensory information is also subject to the influences of these internal models (so called “top-down” predictions). These “top-down” predictions are built on the previously received sensory information. Therefore, dysfunction at either level would likely drive dysfunction in the other (Van Boxtel and Lu, 2013; Van de Cruys et al., 2014). Pellicano and Burr suggest that individuals with ASD have hypo-priors, or weakened internal models of the external environment, leading to an over-weighting of incoming sensory information (Pellicano and Burr, 2012). These authors further suggest that an increased reliance on incoming sensory information (i.e. reduced ability to predict the sensory environment) could support many of the behavioral features associated with ASD (including sensory hypersensitivities, insistence on sameness, and repetitive behaviors) (Van Boxtel and Lu, 2013).

Much overlap exists among these hypotheses and they may have some degree of mechanistic commonality. For example, since proper E/I balance is essential to synaptic function in the brain, which in turn forms the basis of network communication, the observed changes in functional connectivity could be the result of E/I imbalances. The converse is also true: connectivity changes, by altering the patterns of ongoing neuronal activity, could drive E/I dyshomeostasis. Similarly, E/I balance disruptions are not entirely separable from the neurobiological mechanisms underlying the proposed changes in predictive coding. E/I balance disruptions could distort the representation of primary sensory information in the brain via changes in lateral inhibition and local communication or could alter communication between brain regions (i.e. network connectivity) and, ultimately, drive a failure of the brain to construct and update accurate internal models of the world.

These theories, when taken in combination with the large number of genetic variants associated with ASD, the pleiotropic nature of these variants, and the many comorbidities associated with ASD, point toward the distinct possibility that many disparate neurobiological mechanisms ultimately give rise to the complex and variable behavioral symptomatology associated with ASD. This heterogeneity provides a framework from which to study the genetic variants associated with ASD, with the ultimate goal of understanding their impact on brain function and, in turn, behavior. Such studies will undoubtedly provide greater clinical insight as to the various molecular subtypes of ASD that may exist and, in doing so, allow for more precise outcome predictions and the development of more targeted interventions.

Among the many neurobiological mechanisms that could contribute to the aforementioned neurobiological changes (i.e. E/I imbalance, altered network communication, and sensory prediction error or altered predictive coding) is dysfunction of the dopaminergic system. DA is known to regulate both the release of excitatory and inhibitory neurotransmitters as well as neuronal responses to these neurotransmitters (Bamford et al., 2004; Goldman-Rakic et al., 1989; Seamans et al., 2001; Tseng and O’Donnell, 2004). Inhibitory neurotransmitters such as GABA can similarly modulate dopaminergic function (Stoof et al., 1979; Xi and Stein, 1998). Therefore, a change in dopaminergic function could result in an E/I imbalance or an E/I imbalance could result in dopaminergic system dysfunction. Additionally, as DA is a regulator of ongoing neuronal activity, it is highly likely that DA dysfunction could lead to abnormal patterns of neuronal activation. These abnormal patterns of activation could in turn cascade into altered neuronal connectivity and therefore altered connectivity patterns of the cerebral cortex. In regards to the predictive coding theory of ASD, there is evidence to suggest a fundamental role for DA in the appropriate weighting of prior expectations and new sensory information (Cassidy et al., 2018). Cassidy et al. found that people who weight their priors more strongly (i.e. overestimate the precision of their predictions) during perceptual inference have higher levels of striatal DA. Specifically, this study found that individuals with more striatal DA are more likely to report target auditory stimuli embedded in a stream. Taken together, these lines of reasoning suggest several possible mechanisms by which DA dysfunction could cascade into the behavioral symptoms associated with ASD.

3. The dopaminergic system & ASD

To understand the possible role of DA in neuropsychiatric disorders such as ASD, we must first consider the basics of the dopaminergic system. Dopamine (DA, 3-hydroxytyramine) is one of several monoaminergic neurotransmitters produced in the brain (the other monoaminergic neurotransmitters being serotonin [5-HT], epinephrine, norepinephrine [NE], and histamine). DA is synthesized from the amino acid tyrosine in the neurons whose cell bodies reside in the substantia nigra (SN) and the ventral tegmental area (VTA; See Figure 1). A small amount of DA is also produced by the hypothalamus.

Figure 1. Sites of DA synthesis and dopaminergic projection patterns in the brain.

DA is primarily synthesized in the substantia nigra (SN), the ventral tegmental area (VTA), and the hypothalamus. Projections from the SN to the striatum form the nigrostriatal pathway (here in purple). Projections from the VTA to the cortex form the mesocortical pathway (here in blue). The VTA also projects to limbic structures in the brain, forming the mesolimbic pathway (pink). The hypothalamus projects to the pituitary, forming the tuberoinfundibular pathway (green), where it functions to regulate prolactin release.

The dopaminergic neurons of the SN project to the dorsal striatum, a contiguous group of subcortical nuclei that collectively form the input structure of the basal ganglia (Purves et al., 2001), forming the nigrostriatal pathway. This pathway is thought to mediate motor activity. A large population of neurons that produce GABA (γ-aminobutyric acid) are found in the striatum. These cells, called medium spiny neurons (MSNs), account for 95% of the striatal neuron population. These neurons are the primary target of dopaminergic neurons whose cell bodies reside in the SN. MSNs also receive input from the thalamus, various association areas of cortex, and sensory cortex (Yager et al., 2015). MSNs are typically active immediately prior to the initiation of a movement and may encode the decision to make a movement toward a specific target (Murray et al., 2011).

Dopaminergic neurons of the VTA project to the ventral striatum (which includes the nucleus accumbens, NAc), the bed nucleus of the stria terminalis (BNST), and the amygdala to form the mesolimbic pathway and to the prefrontal cortex to form the mesocortical pathway (see Figure 1). The mesolimbic pathway is thought to mediate the interpretation of potential positive and negative reinforcers and to assess the value of these reinforcers (Adinoff, 2004). DA in the NAc is a sensorimotor integrator that modulates the response output of an organism to motivational cues from the environment, thus allowing an organism to overcome response costs for a reward (Salamone et al., 1997). Much like the dorsal striatum, the NAc is composed mainly of MSNs. The amygdala is an integrative site for emotion and memory and is involved in memory consolidation for emotionally charged events (Adinoff, 2004). The BNST, which is considered a part of the extended amygdala, is involved in the behavioral response to fearful events and stress(Adinoff, 2004). Mesocortical DA neuron input to the PFC is neuromodulatory in nature and is thought be involved in working memory, planning, and attention (Seamans and Yang, 2004). DA in the cortex modulates ongoing inhibitory and excitatory neurotransmission to modulate how strongly a certain cortical representation is maintained. This pathway is thought to be dysfunctional in the brains of individuals with schizophrenia (Knable and Weinberger, 1997).

The temporal and spatial aspects of the DA signal are regulated, in part, by the dopamine transporter (DAT, see Figure 2). The DAT is a Na⁺/Cl⁻-dependent membrane transporter that acts to rapidly clear released DA from the synapse. Both cocaine and amphetamine (AMPH) inhibit DA uptake by the DAT and increase extracellular DA, though through distinct mechanisms. AMPH is a substrate of the DAT that competitively inhibits DA uptake into the presynaptic terminal and into synaptic vesicles via the VMAT isoform VMAT2. AMPH also promotes an inward-facing DAT conformation that favors DA efflux (Guptaroy et al., 2009). Cocaine elevates synaptic levels of DA by binding to the DAT in a way that prevents DA uptake (Carroll et al., 1992).

Figure 2. Illustration of a dopaminergic synapse.

DA (blue spheres) is packaged into vesicles in the presynaptic terminal via VMAT. Released DA stimulates post-synaptic D1-like and D2-like G-protein coupled receptors. Released DA is cleared from the synaptic cleft via the DAT or degraded via membrane-bound COMT. DA cleared from the synapse by the DAT is either repackaged into vesicles pre-synaptically or degraded by MAO in the presynaptic terminal.

3.1. The role of DA in behavior

DA plays an essential role in both learning and in motivated behavior and can be conceptualized as providing an estimate for how worthwhile it is for an organism to devote limited resources to a particular motor or cognitive task (Berke, 2018). DA plays an important role in motivated behavior, action selection, and reward processing (Hamid et al., 2016; Salamone et al., 2012). DA signaling has been shown to drive behavioral activation and to promote effortful behavior. Rats in whom DA has been depleted or antagonized in the NAc redirect their behavior away from tasks that have high response requirements in favor of lower cost options (Salamone et al., 2012). Other groups have shown that reward rate is reflected by increased concentration of DA in the NAc and that DA levels in this brain region covary with motivational vigor (Hamid et al., 2016). That is, phasic changes in the concentration of DA affect an animal’s willingness to do work for a reward (Hamid et al., 2016).

In the same manner that DA can encode whether a physical movement warrants the allocation of energy, DA can also encode whether various cognitive processes (including attention and working memory) are similarly worthwhile(Westbrook and Braver, 2016). Early work aimed at understanding DA’s role in behavior demonstrated that DA depletion in the prefrontal cortex (PFC) of rhesus monkeys causes deficits in spatial working memory function (Brozoski et al., 1979). A more recent study in humans demonstrates increased PFC DA during a verbal working memory task when compared to a less demanding attention task (Aalto et al., 2005), supporting functional involvement of DA in working memory. DA signaling has also been linked with allocation of attention. When DA is depleted in the brains of rats, these rats display impairments in the integration of sensory information with motor performance in the home cage or neutral testing environments. However, when the same animals are placed in a threatening or activating situation (such as in a cage with cats) where the salience of environmental stimuli is heightened, these rats demonstrate an immediate (although transient) restoration of sensorimotor performance (Marshall et al., 1976). This study demonstrates that, without appropriate DA signaling, an animal will not dedicate attentional resources to external cues that normally generate orienting movements.

Some of the earliest and most striking illustrations of the impact of DA dysfunction came from the study of Parkinson’s Disease (PD). In PD, progressive degeneration of the dopaminergic neurons of the SN pars compacta (SNc) results in significant motor impairment characterized by rigidity, hypokinesia, resting tremor, and postural instability. These motor impairments may be the most well-recognized symptom of PD, but the disease is also accompanied by a number of non-motor symptoms (“Non-motor symptoms of Parkinson’s disease,” 2006). Neuropsychiatric symptoms commonly associated with PD include depression (Aarsland et al., 1999), anxiety (Menza et al., 1993), repetitive behaviors (Kurlan, 2004), and deficits in attention (Bronnick et al., 2007). These symptoms and disorders result from neurobiological factors associated with the underlying neurodegeneration (i.e. central loss of monoamines) (Marsh, 2013) and are not due to psychosocial factors or disability (Ehmann et al., 1990). Impulsive behaviors (such as pathological gambling, shopping, eating, and hobbying) and repetitive behaviors (Friedman, 1994) can result from or can be worsened by DA replacement therapy (McKeon et al., 2007), the primary treatment for PD. That is, artificially increasing levels of DA in the brain results in impulsive and repetitive behaviors.

Perhaps most interesting and directly related to the symptomatology of ASD is a behavior called punding that is observed in Parkinson’s patients on high levels of DA replacement therapy and in chronic AMPH users (Evans et al., 2004). Punding refers to a constellation of stereotyped behaviors including intense fascination with repetitive manipulation of mechanical objects, handling and examining of common objects, excessive grooming, and engagement in extended monologues. Punding includes such behaviors as, for example, taking apart and putting together watches, flashlights, or radios, sorting and arranging common objects, repetitions of single words and phrases, etc. (Evans et al., 2004; O’Sullivan et al., 2007). In both Parkinson’s patients on DA replacement therapy and chronic AMPH users (Schiørring, 1981), there is an excess of available DA, which is thought to result in the observed obsessive and stereotyped behavior. Attempts by caretakers to interrupt punding leads to irritability in patients with Parkinson’s (Evans et al., 2004), suggesting a persistent, irresistible, compulsive need for the patient to engage in these behaviors, despite potentially adverse personal, familial, or occupational consequences. Punding, and the insistence on performing the behaviors which punding defines, are reminiscent of the repetitive behaviors, restricted interests, and strict adherence to routines observed in individuals with ASD (Kanner, 1944). Work in animals also supports a role for DA in compulsive and repetitive behaviors. For example, D1 receptor agonism has been shown to drive increased repetitive grooming behavior in mice, as has D2 receptor knockout (Lee et al., 2018).

In addition to the role that DA has been shown to play in motor behavior, there is a great deal of evidence linking DA with a second key clinical domain in autism - social function. Work in rats has demonstrated that DA is phasically released in response to pro-social cues and that this release is associated with social-seeking behavior, indicating an important role for the dopaminergic system in driving social behavior (Willuhn et al., 2014). Studies in prairie voles found that administering a DA receptor antagonist prior to mating abolished mating-induced partner preference formation, indicating that DA within the NAc can directly influence social choice (Aragona et al., 2003; Aragona and Wang, 2009). These studies also found that D1-like receptors were upregulated in the NAc of prairie voles who have pair bonded compared to those who have not. The authors further demonstrate that this D1-like receptor upregulation was important in maintaining pair bonding (Aragona and Wang, 2009), again highlighting the role of the dopaminergic system in social choice. Such findings have been replicated in a number of vertebrates, including non-mammalian species. For example, when zebrafish are given a D1-receptor antagonist, a significant reduction in social preference is observed (without change in motor function or vision) (Scerbina et al., 2012). In humans, activation of the VTA and the caudate nucleus (two DA-rich areas) occurred when subjects were shown images of their romantic partner (Fisher et al., 2005). Given that social dysfunction is pathognomonic of ASD and that DA is intricately involved in social regulation, it seems highly probable that dysfunction in the dopaminergic system could underlie at least some forms of ASD, either through altered social reward processing or altered social choice.

3.2. Evidence for DA dysfunction in ASD

The role disruption to the dopaminergic system plays in ASD has been of interest since at least the 1970s (Damasio and Maurer, 1978). Many studies between 1975 and 1990 focused on attempting to identify biomarkers of ASD by using indices of DA function derived from plasma, platelets, urine, and cerebrospinal fluid. These indices included measures of the levels of DA, of its downstream metabolite HVA (the end product of DA metabolism), and of the enzymes responsible for the metabolism of DA (i.e. dopamine-β-hydroxylase [DBH] and monoamine oxidase [MAO]). These studies, summarized in Table 1, failed to show replicable differences between individuals with ASD and controls in the concentration of DA, the concentration of products of DA metabolism, or in the activity of the enzymes involved in DA metabolism (Boullin et al., 1975; Cohen et al., 1977; Garnier et al., 1986; Gillberg and Svennerholm, 1987; Lake et al., 1977; Launay et al., 1987; Minderaa et al., 1989). While disappointing in their inability to identify a consistent biomarker of ASD, these studies do not, however, exclude the possibility of central dopaminergic dysfunction playing an important role in ASD. Indeed, as is often noted in these studies, there exists much inter-individual variability in these measures, which could support the idea of multiple subtypes of ASD, each with a unique neurophysiological disturbance at its core. Hence, DA dysfunction may represent one of many different subtypes of ASD.

Table 1.

DA-Related Biochemical Studies in Individuals with ASD

Year	Authors	n	Measures	Sample Type	Findings	Limitations & Considerations
1987	Gillberg and Svennerholm	ASD = 25 Other psychotic disorders = 12 Control = 20	HVA	CSF	Mean CSF HVA elevated in ASD group (p < 0.001) Mean CSF HVA elevated in other psychotic disorders group (p < 0.01)	ASD group included 4 with FXS
1987	Launay et al.	ASD = 22 Control = 22	DA, NE, Epinephrine MAO in plasma only DOPAC and MHPG in urine only	Plasma	Mean NE elevated in ASD (p < 0.02) Mean epinephrine elevated in ASD (p < 0.01) No difference in mean DA	Wide age range of subjects
				Platelets	Mean DA lower in ASD group (p < 0.05) Mean epinephrine lower in ASD group (p < 0.05) No difference in mean MAO activity or norepinephrine
				Urine	ns
1975	Boullin et al.	ASD = 9 Control = 9	MAO activity	Platelet	ns
1989	Minderaa et al.	ASD = 36-40 Control = 20-28	PRL, HVA, DA	Plasma	No difference in mean HVA or PRL
				Urine	No difference in mean HVA or DA
1977	Cohen et al.	ASD = 10 Nonautistic psychotic = 10 Central processing disturbance = 7 Aphasic = 7 Control = 9	HVA, 5-HIAA, Probenecid	CSF	Lower 5-HIAA in ASD group (p = 0.04)	Control group consisted of children with other medical conditions including vertebral disk disease, nonspecific headache, and conversion disorder Probenecid administered to artificially block egress of metabolites of dopamine and serotonin
1986	Garnier et al.	ASD = 19 Control = 15	DBH, HVA	Plasma	Higher DBH in more impaired ASD group (s < 0.01)	No difference was noted when ASD group as a whole was compared to control group
				Urine	Higher HVA in ASD group (s < 0.02)
1977	Lake et al.	ASD = 11 Control = 12	NE, DBH activity, MAO activity	Plasma	Higher NE in ASD group (p < 0.001) Lower DBH activity in ASD group (p < 0.05)

Contrary to these often-conflicting results in peripheral measures of dopaminergic system function in ASD, brain imaging studies have revealed more consistent results (see Table 2). One of the earliest imaging studies of DA-system dysfunction in ASD, a positron emission tomographic (PET) scanning study in 1997, found a 39% reduction in the accumulation of fluorine-18-labelled fluorodopa (FDOPA, a radiolabeled analog of DOPA used to visualize dopaminergic nerve terminals) in the anterior medial prefrontal cortex in individuals with autism compared to the control group (Ernst et al., 1997). Since then, MRI studies have repeatedly demonstrated abnormalities in dopaminergic structures and their connectivity (Abbott et al., 2018; Delmonte et al., 2013; Di Martino et al., 2011; Langen et al., 2014, 2012, 2011, 2009, 2007; McAlonan et al., 2005; Padmanabhan et al., 2013; Schuetze et al., 2016; Sears et al., 1999). Multiple groups have found evidence for altered growth trajectory of the caudate nucleus in individuals with ASD (one of the major targets of the dopaminergic system) (Akkermans et al., 2018; Langen et al., 2014, 2009, 2007; Schuetze et al., 2016; Sears et al., 1999). Sears and colleagues demonstrated significant enlargement in caudate volume in patients with ASD in two independent sample groups (Sears et al., 1999). This study was limited by its inclusion of subjects exposed to neuroleptics. A follow up study in medication-naïve individuals with ASD (again in two independent sample groups) confirmed the finding of caudate nucleus enlargement in individuals with ASD (Langen et al., 2007). McAlonan and colleagues found significant localized grey matter reduction in fronto-striatal networks, suggesting abnormal connectivity of dopaminergic structures with cortical structures in subjects with ASD (McAlonan et al., 2005).

Table 2.

DA-Related Imaging Studies in ASD

Year	Authors	n	Technique	Region	Findings	Limitations & Considerations
1997	Ernst et al.	ASD = 14 Control = 10	PET - FDOPA	Anterior medial cortex	39% reduction in FDOPA in ASD group (p = 0.016)
1999	Sears et al.	Sample 1: ASD = 35 Control = 36 Sample 2: ASD = 13 Control = 25	MRI (volumetric)	Caudate nucleus	Increased volume of the caudate nuclei in ASD group in both samples (Sample 1: p = 0.01; Sample 2: p = 0.003)	Wide age range of subjects; no information on neuroleptic use in subjects
2004	McAlonan et al.	ASD = 17 Control = 17	MRI (connectivity defined as interregional grey matter volume correlation)	Fronto-striatal networks	Significant localized grey matter reduction in ASD group (p < 0.01)
2007	Langen et al.	Sample 1: ASD = 21 Control = 21 Sample 2: ASD = 21 Control = 21	MRI (volumetric)	Caudate nucleus	Significant enlargement in the ASD group (Sample 1: p < 0.05; Sample 2: p < 0.001)
2009	Langen et al.	ASD = 99 Control = 89	MRI (volumetric & gray matter density)	Caudate nucleus	Increased volume with age in ASD subjects (p < 0.042) & decreased volume with age in control subjects (p = 0.002)
2011	Di Martino et al.	ASD = 20 Control = 20	MRI (functional connectivity)	Cortico-striatal networks	Prominent ectopic striatal functional connectivity in ASD group (increased functional connectivity between striatum and associative and limbic cortex) (p < 0.05)
2012	Langen et al.	ASD = 21 Control = 22	DTI	Fronto-striatal networks	Lower fractional anisotropy of white matter tracts connecting putamen to frontal cortex (p < 0.019) in the ASD group
2013	Padmanabhan et al.	ASD = 42 Control = 48	fMRI (resting-state)	Cortico-striatal networks	Increased connectivity of striatal regions with parietal cortex and decreased connectivity with prefrontal cortex in the ASD group
2013	Delmonte et al.	ASD = 28 Control = 27	MRI (functional & structural)	Fronto-striatal networks	Increased functional connectivity between striatum and multiple regions of the frontal cortex in the ASD group Increased functional connectivity between ACC and caudate was associated with deactivation to social rewards in the caudate (p = 0.006) Greater connectivity between the MFG and caudate was associated with higher restricted interests and repetitive behaviors (p = 0.008)
2013	Di Martino et al.	ASD = 56 ADHD = 45 Control = 50	fMRI	Whole brain	ASD + ADHD group shared ADHD-specific abnormality of increased local connectivity in the right striatum (p = 0.017)
2014	Langen et al.	ASD = 49 Control = 37	MRI	Striatum	Increased growth rate of striatal structures in ASD group specific to the caudate nucleus (p = 0.005), which correlated with repetitive behavior (p = 0.009)
2016	Schuetze et al.	ASD = 373 Control = 384	MRI (surface-based)	Basal ganglia	Greater surface area in bilateral dorsal medial globus pallidus in individuals with more severe restricted, repetitive symptoms on ADOS (q < 0.01) Steeper increase in concavity of the caudal putamen and pallidum with age in the ASD group
2017	Abbott et al.	ASD = 50 Control = 52	MRI (intrinsic functional connectivity)	Cortico-striatal networks	Cortico-striatal overconnectivity of limbic and frontoparietal seeds in ASD group	ASD group not controlled for medications or comorbidities
2018	Akkermans et al.	ASD = 24 OCD = 25 Control = 29	fMRI (resting-state)	Fronto-striatal networks	Significant positive association of repetitive behavior with functional connectivity between the left cuneate nucleus and the right premotor cortex (p < 0.05)

Subsequent studies focused on dissecting changes in functional brain networks repeatedly demonstrated increased connectivity of striatal regions with cortex in individuals with ASD relative to control groups (Abbott et al., 2018; Delmonte et al., 2013; Di Martino et al., 2011; Padmanabhan et al., 2013). In several studies, authors uncovered a significant positive association between functional connectivity in fronto-striatal networks and severity of repetitive behaviors in individuals with ASD (Akkermans et al., 2018; Delmonte et al., 2013; Langen et al., 2014), linking altered dopaminergic system function with the behaviors associated with ASD. While these studies provide compelling evidence that differences in dopaminergic structures and their connectivity patterns exist in individuals with ASD, they do not provide definitive causal evidence. These changes could occur as a result of dysfunction in other neurotransmitter signaling pathways or structural protein differences that cascade into the observed differences in DA-related structure and connectivity.

There are many reasons why studies aimed at determining possible peripheral biomarkers of dopaminergic abnormalities in individuals with ASD might produce such variable findings, while the results of the many imaging studies related to the DA system in ASD are more consistent. First, as discussed above, most studies related to peripheral biomarkers of DA dysfunction in ASD focus on measurements of the concentration of DA and its metabolites (e.g. HVA) in peripheral body fluids. This is problematic as (1) DA cannot cross the blood-brain barrier and (2) there is no evidence to suggest that plasma or cerebral spinal fluid (CSF) HVA concentration measurements are sensitive enough to estimate DA activity in the brain (Elsworth et al., 1987). In one study, the concentration of HVA was measured in four brain areas (dorsal frontal cortex, orbital frontal cortex, caudate, and putamen), in CSF, and in plasma. Correlations were tested between CSF and plasma and between these fluids and brain regions. Only the concentration of HVA in dorsal frontal cortex and in CSF were significantly correlated, limiting the value of plasma HVA measurements for determining DA activity in the brain (Elsworth et al., 1987). Moreover, DA is also produced and released by various peripheral tissues including the pancreas (Mezey et al., 1996), the adrenal medulla (Rosol et al., 2001), the kidney (Wahbe et al., 1982), and peripheral leukocytes (Kokkinou et al., 2009), complicating our interpretations. Finally, primary DA dysfunction is not required for observable changes in dopaminergic brain structures to occur. Genetic variants that impact synaptic scaffolding could lead to abnormal patterning of cortex and subcortical brain structures (Betancur et al., 2009, p.), such as the striatum and basal ganglia, without impacting the gross availability of DA and production of its metabolites. Indeed, recent PET/CT studies show no difference in striatal DA synthesis capacity and no association between DA synthesis capacity and feelings of social defeat in adults with ASD (Schalbroeck et al., 2021), while D2R expression is increased in the MSNs of caudate and putamen based on post-mortem analyses of basal ganglia in ASD (Brandenburg et al., 2020).

3.3. Interventions and therapeutics for ASD targeting the DA system

Treatment for ASD is largely focused on developmental and behavioral therapies designed to improve social function, improve communication, promote academic functioning, and decrease maladaptive behaviors (Myers et al., 2007). Currently, there are no psychopharmacological agents that treat the core symptoms of ASD. Psychopharmacotherapy is initiated only after potential contributors to problematic behaviors (e.g. psychosocial stressors, co-occurring psychiatric disorders, medical issues, and difficulties using functional communication (Myers et al., 2007)) are either addressed with behavioral and educational interventions or ruled out (McGuire et al., 2016). Psychotropic medications are then be used to treat specific symptoms such as hyperactivity, inattention, impulsivity, aggression, and anxiety.

Two of these symptoms, inattention and hyperactivity, are behaviors which have been linked with dysregulation or dysfunction of the DA system. These behaviors may result as a response to the sensory changes associated with ASD (e.g. hyperarousal resulting from changes in sensory processing (Myers et al., 2007)) or may be due to co-occuring psychiatric disorders, such as ADHD. Given the known role of the DA system in ADHD and the emerging evidence for DA dysfunction in ASD, individuals with ASD and co-occurring ADHD maybe represent a distinct subgroup that may have DA dysfunction or dysregulation as a core component contributing to their clinical presentation.

Strong evidence suggests that drugs targeting the DA system may reduce hyperactivity and improve inattention in some children with ASD (Huffman et al., 2011). In one randomized, controlled, crossover trial of methylphenidate, which acts via blockade of the DA and NE transporters (Volkow et al., 2005), 49% of enrolled subjects with ASD showed an improvement in parent- and teacher-rated hyperactivity, while 18% of enrolled subjects discontinued the medication due to adverse effects (most commonly irritability) (Research Units on Pediatric Psychopharmacology Autism Network, 2005). In this study, methylphenidate did not improve Aberrant Behavior Checklist (ABC) subscale ratings for social withdrawal, stereotypy, or inappropriate speech (Research Units on Pediatric Psychopharmacology Autism Network, 2005). This study highlights both the possible responsivity of ASD-related behavioral symptoms to modulation of DA systems as well as the inherent inter-individual variability that exists in ASD. In the future, it will be important to define, either genetically or through a comprehensive behavioral profile, individuals who might benefit most from therapies targeting the DA system and those who might be classified as non-responders to these targeted therapeutics.

The atypical antipsychotic drug risperidone (which functions via D₂ and serotonin 5-HT_2A receptor antagonism (Horacek et al., 2006)) may also hold therapeutic value in the treatment of hyperactivity and inattention in some children with autism (Huffman et al., 2011). In a randomized, double-blind, placebo-controlled experiment, treatment with risperidone resulted in significant reduction in ABC subscale ratings for hyperactivity, lethargy/social withdrawal, and irritability. There is also strong evidence for the benefit of risperidone in treating other features of autism (including restricted, repetitive, and stereotyped behaviors, interests, and activities) (Huffman et al., 2011). At this time, evidence suggests limited efficacy of risperidone therapy for improving social communication in individuals with ASD (Canitano and Scandurra, 2008). Still, the positive value of risperidone for the treatment of the non-social symptoms of ASD highlights the possibility that dopaminergic dysfunction may underlie at a least a portion of the behavioral changes observed in a subset of individuals with ASD. Identifying these individuals through genetic testing, cognitive testing, and psychophysical evaluation is essential for the practice of personalized, precision medicine to improve outcomes and quality of life for affected individuals.

4. Modeling DA dysfunction in ASD

While human studies certainly suggest the involvement of the dopaminergic system in ASD in at least a subset of individuals, the question remains how the various genetic risk factors associated with ASD might lead to dysfunction of dopaminergic systems and, ultimately, the associated clinical phenotype. Animal models provide a particularly useful system for exploring the impact of de novo mutations and rare inherited variants on the neurobiological mechanisms that may contribute to the constellation of symptoms associated with ASD. We will consider here evidence from invertebrate and vertebrate models demonstrating how changes in DA system function, either due to directed mutation to the DA system or via genetic variants associated with ASD, provide important insights into the links between DA signaling and phenotypic characteristics of ASD. The impact of the variants discussed on the dopaminergic signaling pathway is graphically summarized in Figure 3.

Figure 3. Reported sites of action of ASD-associated variants on DA signaling pathway.

Red circles indicate downregulation, reduction, or inhibition of action. Green circles indicate upregulation or potentiation of action.

Before we consider the various model systems that have been used to further our understanding of the role of DA in behavior and the biological basis of ASD, it is prudent to discuss the concept of validity of animal models. At the highest level, an animal model is an animal in which the biology, behavior, and pathological processes associated with a human condition can be studied. The validity of an animal model therefore relates to the level of fidelity with which a model represents a given human condition.

The criteria for model validity have been conceptualized in a number of ways, but most often include face validity, construct validity, and predictive validity (Belzung and Lemoine, 2011; Rand, 2008; Tordjman et al., 2007). Face validity refers to the reliability with which the model to reproduces the features of a given condition (e.g. the behavioral symptoms of a neuropsychiatric disorder). Construct validity refers to how well a model measures what it is intended to measure as it relates to etiology and pathogenesis of a disorder (Rand, 2008; Tordjman et al., 2007). Finally, predictive validity refers to the fidelity with which an animal model predicts the relation of a triggering factor on an observable effect of the disease and the fidelity with which a model predicts the observable effects of a therapeutic agent on the disease (Rand, 2008).

These concepts are critical when considering a constructive approach to applying animal models to a given research question, especially in complex neurobiological disorders such as ASD. For example, while behavioral features of ASD may be present in fly and fish models (i.e. the models may have face validity), these models may be most useful not for understanding ASD-associated behaviors themselves, but rather may be most appropriately used to dissect in detail the neurobiological or neurodevelopmental mechanisms of ASD in a way that is not possible in another model system. As an extension of this concept, while using flies or fish may be appropriate as a screening tool for construct validity, the inherent limitations in face validity of these animals may also limit their predictive value, for example, in testing novel therapeutic agents. Here, rodent models may become particularly useful given their greater ethological similarity to humans. As these animal models are used to validate new therapeutic targets and drug candidates for human use, the imperative to use models with the highest fidelity for a given research purpose or question is clear. We suggest that many of the models we will discuss below may be best or most accurately used not alone, but in combination to better understand the biological mechanisms and behavioral consequences of DA dysregulation in ASD.

4.1. Drosophila melanogaster (Fruit Fly)

The common fruit fly, Drosophila melanogaster, has long been used as a model system in neuroscience (Bellen et al., 2010). The utility of the fly model is reinforced by the numerous structural, functional, and genetic parallels between the brains of vertebrates and Drosophila. At the genetic level, 77% of human disease genes have homologs in Drosophila (Reiter et al., 2001) that can be easily manipulated to study the association between genotype and phenotype. Structurally, in the same way the vertebrate brain is organized into a tripartite pattern (forebrain, midbrain, and hindbrain), the Drosophila brain also follows a tripartite organization(Reichert, 2005). At the functional level, Drosophila brains contain the same key neurotransmitters (i.e. glutamate, GABA, DA, homologs of epinephrine and NE, and 5-HT) as the vertebrate brain and the requisite synthetic enzymes, transporters, and receptors for these neurotransmitter systems (O’Kane, 2011). Given these numerous parallels, it is perhaps not surprising that there now exist many studies of ASD-associated genes and their associated neurobiology in Drosophila (Tian et al., 2017). Several of these studies, summarized in Table 3, reveal clear interplay between the genes associated with ASD, the DA system, and the behaviors associated with both ASD and DA.

Table 3.

Animal models of ASD, their known behavior changes, and evidence for dopaminergic dysfunction

Year	Authors	Species	Mutation or Manipulation	Social Behavior	Motor & Repetitive Behavior	Other Behavior	Evidence for DA or Neuronal Dysfunction
Danio rerio (Zebrafish)
2013	Mahabir et al.	Zebrafish	WT (2 strains: AB & TU)	Maturation of shoaling behavior (index of social behavior) differs between the two strains and correlates with DA & DOPAC levels			DA & DOPAC levels increase differentially with age in the two strains & correlates with behavioral differences between the strains
2012	Scerbina et al.	Zebrafish	WT (AB Strain) treated with D1 receptor antagonist	Significant reduction of social preference
2008	Thirumalai and Cline	Zebrafish	WT				Endogenous DA release suppresses swim circuits in developing zebrafish
2017	Kacprzak et al.	Zebrafish	DAT-KO			KO: anxiety-like phenotype in fish WT + COC: anxiety-like phenotype D1R inhibition:rescue of anxiety-like phenotype (not seen with D2/3R inhibition)	WT + COC: reduced DAT mRNA abundance
2015	Kozol et al.	Zebrafish	syngap1 knockdown		Disruptions in motor behaviors (unproductive swim attempts)	Seizure-like behaviors	Delayed mid- and hindbrain development (responsible for locomotor behaviors)
			shank3 knockdown
2018	Liu et al.	Zebrafish	shank3b loss-of-function mutant	Reduced social interaction (increased inter-fish distance during shoaling) and reduced social preference	Reduced locomotor activity	Repetitive behaviors (circling and figure “8” swimming patterns)
2011	Souza et al.	Zebrafish	DA		90% decrease in larval movement (p = 0.007)
			Quinpirole (D2R agonist)		94% decrease in larval movement (p = 0.011)
			SCH-23390 (D1R antagonist)		94% reduction in larval movement (p = 0.011)
			SKF-38392 (D1R agonist)		120% increase in larval movement (p = 0.005)
			MPTP		60% decreased in larval motor activity (p = 0.002)		25% decrease in TH expression in brains of 5 dpf larvae
2015	Jay et al.	Zebrafish	Targeted ablation of supraspinal DAergic neurons		Decreased locomotion
2012	Lange et al.	Zebrafish	lphn3.1 (risk factor for ADHD)		Hyperactivity and swimming bursts (impulsivity) rescued by methylphenidate and atomoxetine		Reduction and misplacement of DA neurons in the ventral diencephalon
2015	Zimmerman et al.	Zebrafish	VPA exposure	Social interaction deficit (loss of social preference)	Hyperactivity	Anxiety
2017	Baronio et al.	Zebrafish	VPA exposure	Loss of social preference			Reduced th1 Reduced dbh Reduced number of TH1-immunoreactive cells
Drosophila melanogaster (Fruit fly)
1997	Yellman et al.	Decapitated Drosophila	D1-like agonists		Stimulate hind-leg grooming
			D2-like agonist (quinpirole)		Stimulate hind-leg grooming and locomotion
2006	Chang et al.	Drosophila	VMAT-A overexpression	Prolonged courtship behavior & decrease in successful mating	Stereotypic grooming behavior Increased locomotion
2011	Riemensperger et al.	Drosophila	TH knockout in neurons		Reduced activity & locomotor deficits that increase with age4/11/20 1:35:00 PM	Hypophagia
2017	Fernandez et al.	Drosophila	VMAT loss-of-function mutants	Increase in social spacing (similar to non-social flies)
			Tyrosine hydroxylase knockout	Females exhibit increased social spacing	Reduced locomotion		1
			RNAi against inhibitor of tyrosine hydroxylase (catsup)	Males exhibit increased social spacing
2011	Tauber et al.	Drosophila	dfmr1 mutant	Reduced courtship (social interaction)	Impaired climbing and impaired flight as adults	Excessive grooming	VMAT blockade suppresses excessive grooming Dfmr1 mutant flies exhibit elevatedlevels of VMAT mRNA Reserpine (depletes monoamines) suppresses excessive grooming
2005	Zhang et al.	Drosophila	dfmr1 null mutant				Significant increase in dopamine in brain Elevation of the dense core vesicles that package monoamine neurotransmitters for secretion
2015	Cartier et al.	Drosophila	Syntaxin 1 mutant (STX1 R26Q)		No effect of AMPH on locomotion		Reduced AMPH-induced DA efflux Enhanced DAT-mediated uptake of DA
			DAT mutant (hDAT R51W)		Significantly reduced effect of amphetamine on locomotion		Reduced AMPH-induced DA efflux
2013	Hamilton et al.	Drosophila	hDAT T356M		Significantly increased locomotion		Reduced DAT-mediated DA uptake Anomalous dopamine efflux
2009	Lu et al.	Drosophila	dUBE3A overexpression in sensory neurons				Decreased dendritic branching
2011	Ferdousy et al.	Drosophila	dUBE3A overexpression		Hyperactivity		Elevated THB (rate-limiting cofactor in monoamine synthesis) and significantly increased DA
			Dube3a knockout		Hypoactivity		Loss of Dube3a results in decreased tetrahydrobiopterin (rate-limiting cofactor in monoamine synthesis) and significant reduction of DA pools
2013	Hahn et al.	Drosophila	dnl2 (orthologue of neuroligin) mutant	Impaired social interaction Alters acoustic communication signals Less female-directed courtship
Mus musculus (Mouse)
2014	Mergy et al.	Mouse	DAT A559V		Reduction in novelty-induced rearing Increased darting speed Blunted behavioral response to AMPH		Anomalous DA efflux (significant elevation of basal extracellular DA) Loss of AMPH-induced DA efflux
2011	Panayotis et al.	Mouse	Mecp2-null		Reduced grip strength Poorer performance on rotarod		Morphological and function alterations in substantia nigra pars compacta (decreased Th protein levels; fewer Th-positive neurons; reduced Th phosphorylation; reduced [DA] in caudate-putamen)
2015	Su et al.	Mouse	Mecp2-null				MeCP2 mantains local dopamine content in a non-cell autonomous manner in the rostral striatum
2015	Kao et al.	Mouse	Mecp2-null		Hypoactivity Impaired motor coordination Impaired motor skill learning		Significant reduction in striatal dopamine content Down-regulation of TH Up-regulation of dopamine D2 receptors
2001	Guy et al.	Mouse	Mecp2-null		Stiff, uncoordinated gait Reduced spontaneous movement Hindlimb clasping Irregular breathing
2002	Mineur et al.	Mouse	Fmr1 knockout			Diminished learning on radial maze Increased locomotor activity
2008	Wang et al.	Mouse	Fmr1 knockout			Increased locomotor activity	Impaired D1 receptor signaling D1 receptor hyperphosphorylation D1 receptor agonist partially rescuses hyperactivity
2013	Rogers et al.	Mouse	Fmr1 knockout				Attentuated cerebellar-evoked medial prefrontal cortex dopamine release Inactivation of the VTA decreased dopamine release by 50% in WT and 20-30% in knockouts Inactivation of the ventrolateral thalamus decreased dopamine release by 15% in WT and 40% in knockout animals [altered cerebellar modulation of mPFC dopamine release related to reorganization of neuronal pathways mediating this release]
2011	Peça et al.	Mouse	Shank3B null	Reduced interaction with social partners		Repetitive grooming Reduced rearing Anxiety-like behavior (reduced time in open arms of elevated zeromaze)	Altered morphology of striatal medium spiny neurons (increased dendritic length and surface area) Volumetric enlargement of the caudate
2011	Wang et al.	Mouse	Shank3e4-9	Abnormal social behaviors (reduced social interaction) and communication patterns (abnormal USVs)	Repetitive behaviors (increased head pokes) Poorer performance on rotarod	Learning and memory deficits
2016	Bariselli et al.	Mouse	Shank3 knockdown in VTA neurons	Loss of social preference			Altered glutamatergic transmission in DA neurons of the VTA Bursting rate of DA neurons is significantly lower
2016	Uchigashima et al.	Mouse	Neuroligin-2 (NL2) knockdown			Reduced density of dopamine synapses on MSN dendrites and increased GABAergic synapses
2008	Jamain et al.	Mouse	Nlgn4 knockout	Reduced social interaction Reduced ultrasonic vocalizations		Impaired olfaction	Reduced total brain volume Highest levels of NL-4 is in the olfactory bulb, striatum, cortex, and hippocampus
2014	Rothwell et al.	Mouse	Nlgn3 knockout and R451C mutation		Poorer performance on rotarod Increased activity in open field test (hyperactivity) Enhanced acquisition of repetitive motor routines		Selective reduction of synaptic inhibition on NAc D1-MSNs
2012	Reith et al.	Mouse	Tsc2 knockout in Purkinje cells	Loss of social preference	Repetitive marble burying Poorer performance on rotarod at 5 months of age		Loss of Purkinje cells
2012	Tsai et al.	Mouse	Tsc1 knockout in Purkinje cells	Loss of social preference Increased ultrasonic vocalizations	Increased grooming		Purkinje cell abnormalities including: Increased area of the soma, increased dendritic spine density, and abnormal axonal collaterals Reduced excitability of purkinje cells
2019	DiCarlo et al.	Mouse	DAT T356M	Loss of social preference Loss of social dominance	Repetitive rearing Increased spontaneous locomotor activity		Reduced [DA] in the striatum Reduced pTH and pERK Slowed clearance of DA from the extracellular space
2005	Spencer et al.	Mouse	Fmr1 knockout	Loss of social dominance (tube test) and increased sniffing behavior on social tests; initial social anxiety followed by enhanced social investigation		Increased anxiety-like responses to reflected images of mice
2008	Wang et al.	Mouse	Fmr1 knockout		Increased locomotor activity		Impaired D1 receptor signaling D1 receptor hyperphosphorylation D1 receptor agonist partially rescuses hyperactivity
2013	Rogers et al.	Mouse	Fmr1 knockout				Attenuated cerebellar-evoked medial prefrontalcortex dopamine release Inactivation of the VTA decreased dopamine release by 50% in WT and 20-30% in knockouts Inactivation of the ventrolateral thalamus decreased dopamine release by 15% in WT and 40% in knockout animals
2020	Chao et al.	Mouse	Fmr1-KO	Intranasal DA alleviated impairment in social novelty			Reduced TH expression in the substantia nigra, VTA, and dorsal striatum Abnormal morphology of TH-positive axons in the striatum Decreased expression of striatal DAT
2008	McFaralane et al.	Mouse	BTBR	Reduced social approach Low reciprocal social interactions Impaired juvenile play	Repetitive self-grooming
2010	Silverman	Mouse	BTBR		Repetitive self-grooming		mGluR5 antagonist (MPEP) blocks repetitive self-grooming
2014	Squillace	Mouse	BTBR				Impaired D2R function resulting in hypoactivation of the reward system upon GBR 12909 (DAT inhibitor) challenge
2020	Chao et al.	Mouse	BTBR				Reduced TH expression in the substantia nigra, VTA, and dorsal striatum Decreased expression of striatal DAT Intranasal DA rescued deficits in non-selective attention, object-based attention, and social approach
Rattis norvegicus (Rat)
2018	Modi et al.	Rat	Shank2 knockout	Deficits in juvenile play behaviors Decreased social-seeking behavior (olfactory exporation) in adult rats	Hyperlocomotion	Repetitive behaviors	Upregulation of mGluR1 expression, enhanced LTD in MSNs, and increaseddendritic branching in the striatum D1/D5R antagonism reduces repetitive behaviors
2018	Berg et al.	Rat	Shank3 knockout	Reduced social interactions Reduced social approach
2019	Song et al.	Rat	Shank3 knockout	Impaired social memory		Anxiety-like behavior	Striatal scaffolding protein deficiencies (reduced Homer, PSD-95) and reduced GluR1 expression
2005	Schneider & Przewlocki	Rat	Prenatal VPA-exposure	Reduced social exploration Reduced juvenile social play		Stereotypic-like hyperactivity
2019	Schiavi	Rat	Prenatal VPA-exposure	Reduced pay responsiveness Impaired sociability			Increased D2R Hyperexcitability of MSNs in the NAc

At the broadest level, genetic and physiologic work in Drosophila has sought to determine the role of DA and the constituent components of the DA system in behavior. This body of evidence shows that perturbations to the DA system can result in a number of abnormal social and motor behaviors. For example, loss-of-function mutations to the VMAT (the transporter responsible for packaging monoamines into vesicles for synaptic release) have been shown to drive increased social spacing in flies (Fernandez Robert W. et al., 2017), while overexpression of VMAT drives abnormal courtship behaviors, decreases successful mating, drives stereotypic grooming, and increases locomotion (Chang et al., 2006). Alterations to the synthetic pathway of DA can result in similarly abnormal social and motor behaviors in Drosophila. Loss of tyrosine hydroxylase (TH), the rate limiting enzyme in the synthesis of DA, results in abnormal social spacing in female flies and reduced locomotion, while disinhibition of TH results in increased social spacing in male flies (Fernandez Robert W. et al., 2017). Knockout of TH in neurons specifically results in reduced activity and locomotor deficits (Riemensperger et al., 2011). Stimulation of D1 and D2 receptors in flies has been shown to drive repetitive grooming behavior (Yellman et al., 1997). These studies provide direct links between dysfunction of the DA system and behaviors associated with ASD (i.e. social deficits and repetitive behaviors) as well as behaviors associated with other neuropsychiatric disorders, including ADHD.

A great deal of work has linked genetic mutations associated with syndromic forms of ASD with DA dysfunction in Drosophila. Fragile X Syndrome (FXS) is the most common inherited cause of intellectual disability and the most frequent monogenic cause of ASD. To be clear, FXS is considered a separate clinical entity from ASD, however one in three individuals with FXS also have ASD. As such, studying this syndrome provides a valuable window into the neurobiology of ASD. This syndrome results from a CGG repeat expansion in the FMR1 gene (Fu et al., 1991), leading to transcriptional silencing of FMR1 and resultant reduction or absence of fragile X mental retardation protein (FMRP) (Sutcliffe et al., 1992; Verheij et al., 1993). At least two groups have demonstrated that mutation to the Drosophila homolog of FMR1, dfmr1, has significant impacts on the DA system in flies (Tauber et al., 2011; Zhang et al., 2005) and drives behavioral abnormalities typically associated with ASD including social deficits and repetitive behaviors (Tauber et al., 2011). Zhang and colleagues demonstrated that dfmr1 null mutants have significant increases in the synthesis of DA and 5-HT in the brain and in the number of dense core vesicles that package these neurotransmitters for secretion (Zhang et al., 2005). Subsequent work by Tauber et al. confirmed and extended these findings by showing that dfmr1 mutant flies exhibit elevated levels of VMAT mRNA, which is necessary for the packaging of DA into presynaptic vesicles (Tauber et al., 2011). This group also identified behavioral abnormalities consistent with FXS and ASD in dfmr1 mutant flies including reduced courtship (a series of stereotypical behaviors carried out by male flies that is used to assay social behavior that permits the exchange of somatosensory, chemosensory, auditory, and visual information between male and female flies (Mackay et al., 2005)), impaired climbing, impaired flight as adults, and repetitive grooming behavior (Tauber et al., 2011). Importantly, depletion of monoamines with reserpine (a drug that reduces uptake and intracellular stores of monoamines via irreversible blockade of VMAT) in these flies suppressed the observed repetitive grooming behavior, directly linking DA dysregulation due to dfmr1 mutation with the observed behavioral changes (Tauber et al., 2011). Beyond ASD-related behavioral symptoms, dfmr1 mutants also demonstrate FXS-related behavioral changes including deficits in commonly used learning/memory assays. Specifically, dfmr1 mutants show deficits in assays of olfactory association (i.e. learning to associate an odor with an aversive or appetitive stimulus) (Busto et al., 2010; Gatto and Broadie, 2011). While these findings are not specific to DA or to ASD, the presence of learning/memory deficits in dfmr1 mutants supports the validity of these mutants for modeling FXS and studying the neurobiological mechanisms underlying this condition. Together, these studies in Drosophila provide suggestive evidence that FXS, a syndromic form of ASD, may include a component of DA dysfunction.

The UBE3A gene encodes a protein that acts both as a ubiquitin ligase and as a co-activator of transcription (Nawaz et al., 1999; Vatsa and Jana, 2018). While maternally inherited deletion of this gene is known to cause Angelman Syndrome (Matsuura et al., 1997), duplication or triplication of chromosomal regions encompassing UBE3A causes a highly penetrant form of ASD (Yi et al., 2015). In Drosophila, overexpression in sensory neurons of dUBE3A (the Drosophila homologue of UBE3A) results in decreased dendritic branching, suggesting that the appropriate level of dUBE3A is essential for the proper neuronal development (Lu et al., 2009). While this finding is not specific to dopaminergic neurons, overexpression of dUBE3A has also been shown to elevate the rate-limiting cofactor in monoamine synthesis (tetrahydrobiopterin [THB]) and to significantly increase DA levels in Drosophila (Ferdousy et al., 2011). This increase in DA drives hyperactivity in flies (Ferdousy et al., 2011), a behavior commonly observed in both ADHD. Conversely, loss of dUBE3A has been shown to decrease THB and significantly reduce DA pools, resulting in hypoactivity in Drosophila (Ferdousy et al., 2011). These findings provide additional evidence of a possible role for an ASD-associated genetic mutation in DA homeostasis and behaviors associated with ASD.

Members of the neuroligin family (NLGN) have been implicated as candidate genes for ASD (Jamain et al., 2003; Ylisaukko-oja et al., 2005). Neuroligins are post-synaptic cell surface proteins that bind with neurexins to guide the formation of functional synapses (Ichtchenko et al., 1996; Nam and Chen, 2005; Scheiffele et al., 2000; Song et al., 1999) and are essential for normal synaptic function (Südhof, 2008). Five neuroligin genes have been identified in humans, with at least three of these also being identified in rodents (Bolliger et al., 2001; Ichtchenko et al., 1996). Neuroligins-1, −2, and −3 are principally expressed in the CNS, where neuroligin-1 is enriched post-synaptically at exclusively at excitatory synapses (Song et al., 1999, p. 1) and neuroligin-2 is expressed post-synaptically at inhibitory synapses (Varoqueaux et al., 2004). Neuroligin-4 is expressed primarily in the cerebral cortex and is localized to excitatory synapses (Marro et al., 2019). Early work connected NLGN3 and NLGN4 with ASD (Jamain et al., 2003; Laumonnier et al., 2004), but more recent work also suggests a role for NLGN2 (Parente et al., 2017). Thus, Drosophila deficient in neuroligin-2 exhibit a number of abnormal behaviors including reduced social interaction (as measured by increased inter-individual spacing when placed in groups and reduced motivation to engage in courtship behaviors) and impaired transitions between behaviors, including transitions between walking and turning and from courtship singing to subsequent behavior (Hahn et al., 2013).

Increasing empirical evidence associates DAT variants with ASD and with the behavioral features associated with ASD (Bowton et al., 2014; Campbell et al., 2019; Gadow et al., 2008; Hamilton et al., 2013). A number of these ASD-associated DAT variants have been modeled in Drosophila. Results of these studies show that these variants disrupt DA signaling and drive behavioral abnormalities consistent with ASD (Campbell et al., 2019; Hamilton et al., 2013). The human DAT (hDAT) T356M mutation (a threonine to methionine substitution at position 356) results in an “open to the outside” conformation of the transporter (thereby driving anomalous DA efflux through the transporter) and reduces DAT-mediated DA uptake in cells. Drosophila expressing the T356M form of the hDAT display increased locomotion (Hamilton et al., 2013). Similarly, the hDAT ΔN336 mutation promotes a DAT intracellular gate conformation that is “half-open and inward facing”, also resulting in reduced DA uptake. Drosophila expressing the hDAT ΔN336 exhibit impaired social behavior during escape, hyperactivity, and prolonged freezing and reduced fleeing in response to predatory cues (Campbell et al., 2019). These two mutations, identified in individuals with ASD, directly link altered DA function with ASD-associated behaviors in Drosophila.

Collectively, these studies in Drosophila demonstrate that ASD-associated genetic changes can lead to behavioral abnormalities that recapitulate many aspects of the ASD phenotype, and that these mutations can drive abnormalities in the dopaminergic system. Drosophila reproduce rapidly, reach adulthood in 10-15 days, and can be easily genetically manipulated (Venken and Bellen, 2005), making them ideal candidates for first-pass identification of behavioral and physiological changes that may result from genetic variants. The advent of new techniques and technologies for studying complex behavior (i.e. social behaviors and repetitive behaviors) in Drosophila further serves to increase the attractiveness of this organism for use in screening studies of new variants associated with ASD as they are discovered.

4.2. Danio rerio (Zebrafish)

Danio rerio (zebrafish) are attractive as a model organism to study ASD for a number of reasons. These vertebrate animals are inexpensive, highly prolific, develop quickly, have high genetic homology with humans, are easy to manipulate genetically, and display robust behavioral phenotypes (including social and reward-related behaviors) (Howe et al., 2013; Kalueff et al., 2013). Zebrafish are also appealing from a neuroscientific perspective as zebrafish possess all the neurotransmitters, receptors, transporters, and enzymes required for glutamate, GABA, acetylcholine, and biogenic amine (DA, NE, 5-HT, and histamine)-mediated neurotransmission (Kalueff et al., 2014; Panula et al., 2010; Stewart et al., 2015, 2014). Although zebrafish lack the midbrain dopaminergic system seen in mammals (e.g. the substantia nigra and ventral tegmental area) (Du et al., 2016), they do have a significant telencephalic population of DA neurons that may be the zebrafish homolog of the substantia nigra (Panula et al., 2010). Despite this difference in the organization of the dopaminergic system, zebrafish do respond behaviorally to pharmacological manipulations of the DA system in a manner consistent with mammalian models (Anichtchik et al., 2004). For these reasons, zebrafish have much potential for high-throughput genetic screening (similar to the proposed screening studies put forth above) in the study of DA dysfunction in ASD and other neuropsychiatric disorders.

Behaviors that are frequently altered in ASD (social behavior and motor behaviors) have been associated with DA in zebrafish. Mahabir and colleagues demonstrated that, in two different strains of zebrafish, maturation of shoaling (the aggregation of swimming fish for social purposes) correlates with increases in the level of DA and DOPAC in the brain with age (Mahabir et al., 2013). In another study, zebrafish exposed to the D₁ receptor antagonist SCH23390 showed significant reductions in social preference, again providing evidence for the role of the dopaminergic system in regulating social behaviors (Scerbina et al., 2012). Similarly, motor behaviors have been linked to the zebrafish dopaminergic system. Targeted ablation of dopaminergic neurons in zebrafish results in decreased locomotor activity (Jay et al., 2015; Souza et al., 2011). Exposing developing zebrafish to exogenous DA during key periods during development suppresses episodes of swimming, while ablating dopaminergic neurons during the same period in development increases episodes of swimming (Thirumalai and Cline, 2008). Exposing zebrafish larvae to either quinpirole (a D₂ receptor agonist) or SCH23390 (a D₁ receptor antagonist) reduces larval movement, while exposing larvae to SKF-38392 (a D₁ receptor agonist) increase larval movement (Souza et al., 2011). These results demonstrate a key role for DA in the development of social behaviors and motor function in zebrafish. With this information in hand, the question then becomes whether mutations associated with ASD can drive ASD-like behaviors in zebrafish and to what degree these changes in behavior stem from dopaminergic dysfunction. Multiple transgenic zebrafish models have been created as a tool to study the role of ASD-associated mutations on behavior and brain morphology. We will consider some of the more important of these models here, with an emphasis on the evidence for dopaminergic dysfunction resulting from these mutations.

Mutation to SHANK3 has been identified as a causative factor of ASD (Durand et al., 2007, p. 3). SHANK3 is a postsynaptic protein whose major functions include modulating the scaffolding of glutamatergic postsynaptic densities and promoting normal development of MSN morphology in the striatum (Peça et al., 2011). As described earlier, MSNs are the major target of dopaminergic projections arising from the substantia nigra and the ventral tegmental area (Yager et al., 2015). In zebrafish, knockdown of shank3 (the zebrafish orthologue of SHANK3) results in disruptions in motor behaviors (i.e. unproductive swim attempts), seizure-like behavior (a common comorbidity of ASD, as described above), and delayed mid- and hindbrain development (regions of the zebrafish brain responsible for motor control) (Kozol et al., 2015). Zebrafish with shank3b loss-of-function mutations show impaired shoaling behavior (as measured by increased inter-fish distance resulting in larger and looser schools) and reduced frequency and duration of social contacts with conspecifics (Liu et al., 2018). These animals also exhibit repetitive swimming patterns, such as repetitive or stereotyped figure “8” swimming, circling, cornering, and walling (swimming repetitively up and down one side of the test chamber) (Liu et al., 2018). Given the known role of dopaminergic neurons in the development of both motor and social behaviors in zebrafish (Jay et al., 2015; Mahabir et al., 2013; Souza et al., 2011; Thirumalai and Cline, 2008), it is likely that DA dysfunction plays a role in the phenotype observed in these shank3 mutants.

As described earlier, exposure to the anticonvulsant VPA in utero is a known risk-factor for ASD (Christensen et al., 2013). Zebrafish exposed to VPA during development exhibit loss of social preference, hyperactivity, and anxiety-like behavior (Zimmermann et al., 2015). Importantly, Baronio and colleagues demonstrated that exposing zebrafish embryos to VPA results in reduced mRNA expression of the tyrosine-hydroxylase encoding gene th1 (the rate limiting enzyme in the production of dopamine) and reduced mRNA expression of the dopamine-β-hydroxylase encoding gene dbh (the enzyme that converts DA into NE). The authors also report reductions in the number of TH1-immunoreactive cells. However, this does not mean VPA acts only to alter expression of mRNAs associated with DA synthesis. VPA inhibits histone deacetylase (which is associated with epigenetic regulation of gene expression). Studies suggest VPA alters the expression of over 1300 genes, including subunits of the GABA receptor (Fukuchi et al., 2009). While VPA does not solely target the DA system, there is reasonable evidence that this known risk factor for ASD (in utero VPA exposure) can drive ASD-associated behaviors in zebrafish that are accompanied by alterations to the dopaminergic system in these animals.

Given the high comorbidity of ASD and ADHD, it is likely that the two disorders share neurobiological substrates. As such, considering dysfunction of the DA system in the context of ADHD is also useful to our understanding of how DA dysfunction may play a role in ASD. Dopaminergic dysfunction has been specifically studied in zebrafish models of ADHD-risk alleles. One such ADHD-susceptibility gene is LPHN3, which encodes latrophilin 3. Latrophilins function as both adhesion molecules (essential for the formation of synapses) and as G-protein-coupled receptors (whose activation sets in motion an intracellular signaling cascade with multiple targets that act in concert to regulate neuronal activity) (Moreno-Salinas et al., 2019). At the behavioral level, loss of function of the LPHN3 ortholog (lphn3.1) in zebrafish results in hyperactivity and increased bursts of swimming, suggesting motor impulsivity (Lange et al., 2012). This is accompanied by impaired DA system development, including misplacement of DA neurons in the ventral diencephalon (Lange et al., 2012).

Much work remains to establish a definitive link between dopaminergic dysfunction and alterations in behaviors associated with ASD-associated mutations in zebrafish. However, the current evidence suggests that ASD-associated mutations do drive dysfunction in social and motor behaviors in zebrafish. Given that these behaviors are behaviors known to be under the influence of the dopaminergic system, future work should aim to determine whether changes in dopamine signaling are sufficient to drive changes in these behaviors consistent with the pattern observed in ASD (i.e. social deficits and restricted and repetitive interactions or activities).

4.3. Mus musculus (Mouse)

Perhaps the strongest evidence supporting a role for DA dysfunction in ASD comes from work performed in mouse genetic models. These models shed light on the impact of genetic mutations associated with ASD on underlying neurobiological pathways and the complex behaviors supported by these pathways. As in Drosophila melanogaster and Danio rerio, work in mice has historically focused on dissecting the impact of mutations associated with the most common, syndromic forms of ASD. However, a growing number of studies now model rare variants associated with ASD and their role in the pathophysiology of ASD, providing new insight into possible mechanistic underpinnings.

Among the earliest mouse genetic models of ASD-associated conditions were the Mecp2-null mouse and the Mecp2^308/y mouse. These models replicate the protein truncating mutation causative of Rett Syndrome, a syndrome which (while considered a separate clinical entity from ASD by the DSM-5) shares many clinical characteristics with ASD (including withdrawal from social engagement and repetitive behaviors). Both the Mecp2-null mouse and the Mecp2^308/y mouse have a number of behavioral changes similar to the behaviors observed in Rett Syndrome (of which many are applicable for ASD). These behavioral changes include loss of social preference (i.e. reduced exploration of novel social targets) (Berger-Sweeney, 2011) and a number of motor deficits. These motor deficits include impaired motor coordination as measured by a reduced latency to fall on the accelerating rotarod task and impaired motor skill learning evidenced by limited improvement on the accelerating rotarod task over consecutive days of training (Guy et al., 2001; Kao et al., 2015; Panayotis et al., 2011).

In addition to the behavioral abnormalities observed in Mecp2-null mice are a number of abnormalities in dopaminergic synapses and dopaminergic brain structures (Kao et al., 2015; Panayotis et al., 2011; Su et al., 2015). Such abnormalities include marked reductions in the population size of tyrosine hydroxylase-expressing neurons in the substantia nigra pars compacta (Panayotis et al., 2011) and reduced DA content in the striatum (Panayotis et al., 2011; Su et al., 2015, 2015). Importantly, Su et al. demonstrated that selective deletion of MeCP2 in the striatum was sufficient to disrupt DA content and locomotor activity, directly linking disruption of a gene known to be causative of ASD-associated behaviors and DA dysfunction.

As discussed previously, mutation to SHANK3 (a postsynaptic scaffolding protein involved in the normal development of medium spiny neurons in the striatum) has been identified as a causative factor of ASD (Durand et al., 2007, p. 3). Shank3B-null mice and mice with disruptions in major isoforms of Shank3 (i.e. Shank3^e4-9) exhibit behaviors associated with ASD, including reduced social interaction and repetitive behaviors (i.e. repetitive grooming) (Shahbazian et al., 2002; Wang et al., 2011). Shank3B-null mice also display alterations in the morphology of striatal MSNs and volumetric enlargement of the caudate (Shahbazian et al., 2002). Knock-down of Shank3 specifically in ventral tegmental neurons results in a similar behavioral phenotype (abnormal social behaviors and repetitive behaviors) as well as altered neuronal activity patterns in DA neurons (Bariselli et al., 2016), suggesting that reduction of Shank3 in dopaminergic neurons is sufficient to drive the ASD-associated behavioral changes exhibited by these mice.

As discussed previously, a number of neuroligin family genes (NLGN) have been implicated as candidate genes for ASD (Jamain et al., 2003; Ylisaukko-oja et al., 2005). The NLGN genes encode post-synaptic cell surface proteins critical for guiding the formation of synapses (Ichtchenko et al., 1996; Nam and Chen, 2005; Scheiffele et al., 2000; Song et al., 1999). In mice, knock-down of neuroligin-2 significantly impacts the formation of dopaminergic synapses in the striatum, resulting in a reduction in the density of dopaminergic synapses on medium spiny neurons (MSNs) and an increase in the number of GABAergic synapses on MSN dendrites, indicating a shift in the excitatory-inhibitory ratio (Uchigashima et al., 2016). Overexpression of neuroligin-2 similarly results in the reduction of the excitatory-inhibitory ratio and drives a number of behavioral abnormalities (including stereotyped jumping behavior, anxiety, and impaired social interactions) (Hines et al., 2008). While the authors of this study attribute these behavioral changes to potentiation of inhibitory responses in the frontal cortex (Hines et al., 2008), it is likely that the observed excitatory-inhibitory imbalance extends to and results in the dysfunction of other regions of the brain, including striatal structures, which are both a major target of and source of input to the cortex. This concept is supported by work demonstrating that knockout of and mutation in neuroligin-3, another member of the neuroligin family, results in selective reduction of synaptic inhibition on D1-dopamine receptor-expressing MSNs in the ventral striatum that is accompanied by enhanced formation of repetitive motor routines (Rothwell et al., 2014). Loss of neuroligin-4, which is most highly expressed in the olfactory bulb, striatum, cortex, and hippocampus, results in social deficits in mice (reduced social preference and reduced ultrasonic vocalizations) as well as a reduction in total brain volume. These mice also exhibit impaired olfaction, a function that is highly dependent on dopaminergic signaling (Pignatelli et al., 2005; Tillerson et al., 2006).

The DAT T356M mutation, as previously discussed, was identified in an individual with ASD and results in dysfunction of the DAT (i.e. constitutive reverse transport of DA) (Hamilton et al., 2013). Mice homozygous for this mutation (DAT T356M^+/+ mice) displayed significant impairments in the uptake of released DA and reduced total tissue content of DA (DiCarlo et al., 2019). DAT T356M^+/+ mice also exhibited a number of behavioral changes similar to the behavioral characteristics of ASD. These behavioral changes included reduced social preference (i.e. equal time spent exploring an inanimate target as spent with a social target), repetitive rearing behavior, and profound increases in spontaneous locomotor activity (DiCarlo et al., 2019). Antagonism of the DAT reduced the observed hyperlocomotion in DAT T356M^+/+ animals, suggesting that DAT-mediated leak of DA may underlie hyperactivity in these animals (DiCarlo et al., 2019). This research (summarized in Table 3) provides new, direct evidence for a role of DA dysfunction (specifically anomalous DA efflux) in the behavioral changes typically associated with ASD and ADHD.

Fmr1 KO mice demonstrate ASD-associated behaviors, including impairments in social behaviors, anxiety-like behavior, and hyperlocomotion (Chao et al., 2020; Spencer et al., 2005; Wang et al., 2008). There is increasing evidence that transcriptional silencing of Fmr1 also results in profound changes in DA signaling systems. In a study by Wang et al., Fmr1 KO mice demonstrate impaired D1R-mediated signaling, which is partially rescued by D1 receptor agonists. This work was expanded upon by Rogers et al. who demonstrated that Fmr1 KO mice have reduced cerebellar-evoked mPFC DA release. This pathway is defined by projections from the cerebellum to DA neurons in the VTA which then project to mPFC. Inactivation of the VTA decreased DA release less in Fmr1 KO mice than in WT mice, indicating a relative weakening of this pathway in Fmr1 KO mice (Rogers et al., 2013). More recently Chao et al. demonstrated that Fmr1 KO mice express less TH in the substantia nigra, VTA, and dorsal striatum that WT mice and express less striatal DAT. This is accompanied by abnormal morphology of TH-positive axons. Importantly, in this study, Fmr1 KO mice treated with intranasal DA administration had enhanced social novelty seeking compared to untreated Fmr1 KO mice (Chao et al., 2020), suggesting both that these behaviors are indeed DA-mediated and that DA-targeted therapy may be efficacious for treating the symptoms associated with ASD.

The BTBR mouse model is an inbred mouse strain that is considered a model of idiopathic ASD that exhibits social and communication deficits (Babineau et al., 2013; McFarlane et al., 2008; Scattoni et al., 2011; Wöhr et al., 2011) and repetitive behaviors (McFarlane et al., 2008; Silverman et al., 2010). The BTBR mouse has been found to have hundreds of noncoding SNP differences in 124 candidate genes (McFarlane et al., 2008). As such, the BTBR mouse might best represent a polygenic form of ASD in which many small-effect SNPs contribute to alter neurobiology in a way that results in the clinical syndrome we call ASD. Significant research has been dedicated to understanding the behavior and neurobiology of this model (Babineau et al., 2013; McFarlane et al., 2008; Meyza and Blanchard, 2017; Servadio et al., 2015; Silverman et al., 2010). Among these works, several studies have demonstrated a connection between dopaminergic dysfunction and the abnormal behaviors exhibited by BTBR mice. Silverman et al. demonstrated that antagonism of mGluR5, a receptor expressed abundantly in the mouse and rat striatum (Shigemoto et al., 1993; Testa et al., 1995) and co-expressed in D1 receptor-expressing neurons important for incentive learning (Novak et al., 2010), blocks repetitive self-grooming in BTBR mice. fMRI studies demonstrate that DAT blockade fails to activate DA-dependent forebrain circuits, which are responsible in part for reward processing, in the setting of impaired D2R function in BTBR mice (Squillace et al., 2014). Most recently, Chao et al. demonstrated that BTBR mice have reduced TH expression in the DA-neuron rich regions of the brain (including substantia nigra, VTA, and dorsal striatum) as well as reduced striatal DAT expression. In these mice, intranasal administration of DA increased social exploration, improved deficits in object-based attention, and improved deficits in non-selective attention. Taken together, these studies implicate relative hypoactivity of the DA system and compromised striatal connectivity in the behavioral deficits observed in the BTBR mouse and provide a potential therapeutic target for the symptoms of ASD.

Given the proposed contribution and evidence for environmental influences on the development of ASD in humans (Dawson, 2008), many studies have been performed in various mouse models of ASD to assess the malleability of the behavioral and neurobiological changes observed in these animals. These studies have demonstrated varying levels of success. Environmental enrichment has been shown to improve ASD-associated and other behavioral deficits in a number of mouse models of ASD, including Mecp2-null mice, BTBR mice, neuroligin-3 variant mice, and Fmr1-KO mice. Mecp2-null mice raised in enriched environments demonstrated improved motor coordination and learning in male mice and learning and anxiety-like behaviors in female mice (Lonetti et al., 2010). VPA-exposed mice and BTBR mice raised in enriched environments both demonstrated improved social affiliation and reduce anxiety-like behaviors (Queen et al., 2020; Yamaguchi et al., 2017). Fmr1-KO mice raised in an enriched environment had complete rescue of the adult behavioral phenotype including social deficits, hyperactivity, and cognitive deficits (Oddi et al., 2015).

However, not all studies of environmental enrichment have demonstrated benefit. Environmental enrichment failed to improve behavioral deficits (i.e. self-grooming and exploratory behaviors) in Shank3 variant mice and increased anxiety-like behavior and decreased motor performance in these mice (Hulbert et al., 2018). Similarly, while neuroligin-3 R451C mice respond to environmental enrichment with increased affiliative and social interaction, both WT and NL3^R451C mice exhibited increases in aggression behaviors with environmental enrichment (Burrows et al., 2020).

There is evidence to suggest that environmental enrichment impacts dopaminergic structures in the brain. Mice reared in enriched environments, for example, have alterations in mRNA expression for proteins involved in cell proliferation, cell survival, and signal transduction specifically in the striatum (Thiriet et al., 2008) and reduced DAT surface expression (Kim et al., 2016). Additionally, environmental enrichment reduces the activating and rewarding effects of stimulants such as cocaine, whose primary mechanism of action is at the dopamine transporter (Solinas et al., 2009). These behavioral changes are accompanied by altered gene expression in the striatum and reduced cocaine-induced dopamine levels in the striatum (Solinas et al., 2009; Thiriet et al., 2008). This work provides intriguing evidence for the role of environment in modulating the dopamine system, which may in turn contribute to variable expressivity of the genetic variants associated with ASD.

4.4. Rattus norvegicus

A large body of work supports the use of rats as model systems in the study of neuropsychiatric disorders and, more specifically, in the study of ASD. Rats are particularly useful model organisms in the study of ASD given their large social repertoire (Schweinfurth, 2020; Servadio et al., 2015). Moreover, the role of DA in social behaviors in rats has been studied extensively and provides a strong foundation upon which to consider how disruptions to this system may impact DA-dependent social behaviors. DA receptor antagonism has been demonstrated to inhibit social play in rats (Beatty et al., 1984; Niesink and Van Ree, 1989; Trezza and Vanderschuren, 2009), while DAT antagonism and DA receptor agonists (when administered systemically) have inconsistent effects on social play (Achterberg et al., 2014; Beatty et al., 1984; Niesink and Van Ree, 1989). This finding is intriguing as it suggests that broad increases or decreases in DA tone in the brain may have varying effects on social behaviors and that region-specific DA function may be required for normal social play. To this effect, Trezza et al. demonstrated that DAT inhibition specifically in the NAc increases social play in a D1 and D2-receptor dependent manner (Manduca et al., 2016). These studies suggest an important role for DA in normal social behavior and the region-specific importance of functional DA neurotransmission in these behaviors.

To further study the neurobiological mechanisms that contribute to the behavioral abnormalities observed in ASD, several rat models of ASD have been developed. These models include both genetic and environmental models. Environmental models are useful in understanding idiopathic ASD, where a single causative gene does not underlie the disorder. We will discuss in particular one such environmental model, the VPA-exposure model, as well as two genetic models, the Shank2 and Shank 3 model, in which alterations in DA signaling have been linked with ASD-associated behaviors.

Valproic acid (VPA) is a known human teratogen that has been linked with neural tube defects (Nau et al., 1991), developmental delay (Shallcross et al., 2011), cognitive impairment (Meador et al., 2009), and ASD (Bromley et al., 2008). Mice and rats exposed to VPA prenatally demonstrate repetitive behaviors, deficits in social interactions, and reduced isolation-induced vocalization as pups (Nicolini and Fahnestock, 2018; Schneider and Przewłocki, 2005; Servadio et al., 2015). In vivo microdialysis in the PFC of VPA-exposed rats revealed elevated basal concentration of DA and increasingly elevated DA in response to swim stress (Nakasato et al., 2008). The authors of this study hypothesize that such hyperactive mesocortical DA signaling may play a role in the abnormal behaviors exhibited by individuals with ASD during periods of stress (Nakasato et al., 2008). Similarly, Schiavi et al. found that prenatal exposure to VPA causes striatal MSN hyperexcitability and alterations in DA receptor expression in adolescent and adult rats. Specifically, VPA-exposed rats have increased D2R expression in the NAc in adolescence and adulthood and increased D1R expression in the NAc and hippocampus in adulthood (Schiavi et al., 2019). However, behavioral testing in these rats revealed that VPA-exposed rats perform similarly to WT rats in tasks of socially-induced place preference and sucrose preference despite these dopaminergic changes. Therefore it is most likely that the aforementioned alterations to DA signaling in VPA-exposed rats leads to ASD-like behaviors not due to the inability to experience pleasurable aspects of social interaction, but perhaps due to abnormal processing of and stress response to unfamiliar or changing social and physical circumstances (Schiavi et al., 2019).

As discussed in previous sections of this review, variants in the genes coding for SHANK proteins (which are critical for synaptic development and function (Monteiro and Feng, 2017)) have been associated with ASD (Berkel et al., 2010; Cheng et al., 2018; Leblond et al., 2012; Monteiro and Feng, 2017). Recent work by Modi et al. demonstrated that Shank2 loss-of-function rats demonstrate ASD-associated behaviors and striatal dysfunction (Modi et al., 2018). Specifically, Shank2 KO rats demonstrate deficits in juvenile play behaviors, reduced olfactory exploration (social recognition) in adult rats, repetitive circling behavior, and hyperlocomotion. In this study, the MSNs of the striatum (a major dopaminergic target) were found to have a number of molecular alterations including upregulated mGluR1 protein expression in male rats, larger soma size, and increased proximal dendritic branching. Importantly, Shank2 KO rats treated with a selective D1/D5R antagonist (SCH-39166) demonstrated reduced repetitive circling, linking altered DA signaling with some of the abnormal behaviors observed in these rats (Modi et al., 2018). Similarly, several studies demonstrate that rats deficient in Shank3 have reduced reciprocal social interactions, reduced social approach (Berg et al., 2018), and anxiety-like behavior (i.e. increased thigmotaxis) (Song et al., 2019). Scaffolding proteins (i.e. Homer and PSD-95) and glutamate receptor subunits (i.e. GluR1) that interact with SHANK3 are deficient specifically in the striatum of Shank3-deficient rats (Song et al., 2019), indicating that abnormal striatal function may contribute to these behavioral deficits.

5. Discussion

This review highlights a number of invertebrate and vertebrate models with behavioral changes that recapitulate those seen in humans with ASD (i.e. social behaviors, repetitive behaviors, anxiety, and hyperactivity) that may also have underlying dopaminergic dysfunction or changes in structures known to be involved in the DA system. These models illustrate several important concepts. First, that modeling of complex behaviors known to be altered in ASD (i.e. social behaviors, repetitive behaviors, anxiety, and hyperactivity) is possible in invertebrates such as Drosophila and zebra fish and in vertebrates (i.e. mice). Second, that the behavioral changes observed as a result of a given ASD-associated mutation are often conserved across species. Finally, that DA dysfunction may play a role in these behavioral changes. This is particularly important as the scientific and clinical communities continue to uncover a growing number of genetic variants associated with ASD. As these variants are detected the question becomes: how do these variants (alone or in concert) act to alter brain function and contribute to the ASD phenotype?

As we look forward to the future of autism research, we must set our sights on an improved mechanistic understanding of ASD so we may improve diagnostics and interventions to optimize individual outcomes. Not all individuals with ASD are equally affected and not all individuals exhibit the same constellation of symptoms. We must endeavor to better understand the factors contributing to the presentation of ASD, including genetic and environmental influences, if we hope to effect meaningful change upon the status quo. The animal models reviewed here, summarized in Table 3, represent the beginning of such efforts. Through these models, we can begin to gain a picture of the neurobiological commonalities, and differences, that may exist among the various genetic risk factors we now associate with ASD. The possibility exists that doing so will ultimately reveal a unifying thread between these biological changes, but more likely we will uncover the intricacies of each that will lead to better understanding of and appreciation for the unique and varied clinical presentations of ASD.

We propose a standardized, high-throughput screening process for genetic variants identified in individuals with ASD in the interest of furthering these goals. Drosophila and zebrafish are inexpensive, reproduce rapidly, reach adulthood quickly, and can be easily genetically manipulated, making these model species ideal candidates for first-pass identification of behavioral and physiological changes that may result from genetic variants. The advent of new techniques and technologies for studying complex behavior (i.e. social behaviors and repetitive behaviors) in Drosophila and zebrafish further serves to increase the attractiveness of these organisms for use in screening studies of new variants associated with ASD as they are discovered. Such screening studies could involve targeted genetic manipulation followed by a standardized, pre-specified battery of behavioral and neurobiological assays designed to characterize the impact of the particular genetic manipulation. Behavioral assays could include social testing (e.g. social spacing and courtship assays in Drosophila, which are paralleled by indices of shoaling and social preference in zebrafish), observation for repetitive grooming or locomotor behaviors using high speed cameras, hyperactivity testing (e.g. observation of locomotion), and fear/anxiety testing (e.g. measurement of fleeing in response to predatory cues and measures of anxiety-like behaviors such as thigmotaxis). Targeted drug screens on mutant Drosophila and zebrafish can be used both to guide studies related to the mechanistic underpinnings of the variants as well as determine which, if any, of the behavioral changes observed are responsive to potential therapeutic agents. For example, drug screens with various dopamine-receptor agonists might inform whether hypo- or hyper-sensitivities to these drugs exist, which may in turn highlight the potential neurobiological pathways involved.

Based on these behavioral findings and the hypothesized function of the gene in question, one could then rely on a tailored set of neurobiological assays to assess neuronal function. For example, if a novel variant in the gene encoding the DAT were discovered and animals expressing this variant displayed reduced sensitivity to cocaine, this would be suggestive of for altered DAT function. To determine the nature of this dysfunction, one could then use amperometric techniques and DA-uptake assays to determine the impact of the variant on transporter function (i.e. DA release and uptake profiles of the variant transporter). Similar strategies could be applied for the other neurotransmitter systems. These first-pass experiments would then serve as the rationale for pursuing genetic manipulations, further behavioral testing, and further neurobiological testing in mammalian (i.e. rodent) models. While this proposal is not necessarily novel (Dourlen et al., 2018), there are also few laboratories with the technical ability and requisite knowledge to perform all of the potential studies outlined. This highlights the critical importance of the establishment of collaborative networks designed specifically to facilitate this goal.

Here we have delineated how a collection of such experiments, from Drosophila to the rat, has built support for the concept of a DA-dependent form of ASD, which may represent a unique subtype of ASD. We have reviewed literature suggesting DA dysfunction, either resulting from or contributing to perturbances in network connectivity, excitatory and inhibitory signaling, and allocation of attention, energy, and working memory may indeed form a plausible explanation for the behavioral and physiological symptoms observed in ASD and in the conditions that co-occur frequently with ASD (including epilepsy, GI disorders, and ADHD). We have reviewed evidence suggesting that DA dysfunction may be a direct result of variants associated with ASD or may be impacted secondarily by variants in neurobiological systems that interact with the DA system. We have found evidence across multiple institutions, laboratories, and model organisms for DA disturbances as a result of many genetic variants associated with ASD. Many of the specific behavioral and neurobiological changes reviewed replicated between the different species considered, again supporting the use of more rapidly generated animal models as a screening method before proceeding to rodent models.

We must also consider that disruption of the DA system and dysregulation of dopaminergic signaling has been invoked in many other neuropsychiatric disorders, including schizophrenia, ADHD, OCD, and Tourette Syndrome. The question then becomes: how can dysregulation of the same neurotransmitter system lead to such a variety of clinical presentations? We are in the infancy of understanding the answer to such a question. We are beginning to undercover the complex mechanisms by which these systems are regulated and by which they lead to phenotypic variability, but cannot claim to fully understand these mechanisms. This is complicated by our current diagnostic criteria, which rely solely on interpretation of behavioral symptoms. Many neuropsychiatric disorders share behavioral symptoms that can be misinterpreted. Therefore, what may be diagnosed as ASD by one provider may be relabeled as ADHD by another. We believe our proposal outlined here provides one possible path forward to disentangling these confounding factors. Moreover, given the discussed impact of environmental factors on the variable phenotypic expression of ASD-related behaviors, it is certainly possible that DA dysregulation, under certain mitigating environmental circumstances, may lead to a clinical syndrome consistent with ASD, while under other conditions may be asymptomatic or lead to another phenotypic expression of DA dysregulation, such as ADHD. We propose that DA may be more highly dysregulated in some individuals with ASD and that identification of these individuals (via genetic testing or other yet to be identified clinical biomarker or constellation of co-occurring symptoms) is essential to more precisely tailor a therapeutic regimen for these patients. This will require investment of resources to systematically study the genetic variants and behavioral symptomatology associated with these variants, which will guide both our future research efforts and therapeutic interventions.

While DA system dysfunction may not be the primary driver of disease as it relates to some variants, the reviewed experiments have provided important insight into the how these variants may cascade into DA disturbances, which in turn may manifest as the behavior associated with ASD or exacerbate these behaviors. For some individuals in whom DA dysregulation is the primary insult, we have reviewed evidence suggesting that therapeutics targeting the DA system may alleviate at least some of the observed behavioral changes. These findings not only lend credence to the idea of the existence of a DA-dependent subtype of ASD, but also provides a path forward for identifying and understanding the mechanistic underpinnings of other forms of ASD, beyond the strict behavioral guidelines outlined in the DSM-V.

Table 4.

DA-related genetic variants associated with ASD

Gene	Encoded Protein	Finding	Year Reported	Authors
DRD3	D3 subtype of DA receptor	Genetic association found between the DRD3 gene and autism in the Dutch and UK populations	2009	Krom et al.
DRD1	D1 substype of DA receptor	DRD1 haplotype was associated with risk for autism spectrum disorders in male-only affected sib-pair families	2008	Hettinger et al.
SLC6A3	Dopamine transporter	SLC6A3 gene dentified in a simplex ASD case as part of a whole-exome sequencing study in 175 ASD parent-child trios	2012	Neale et al.
DRD2	D2 subtype of DA receptor	Increased frequency of the rs1800498 TT genotype in affected males	2012	Hettinger et al.
DDC	DOPA decarboxylase	Significant association between the DDC gene and ASD in a case-control genetic association study consisting of 326 unrelated autistic patients and 350 gender-matched controls from Spain	2012	Toma et al.
SLC29A4	Membrane protein that catalyzes reuptake of monoamines	Two novel missense variants in the SLC29A4 gene (c.412G>A/p.Ala138Thr and c.978T>G/p.Asp326Glu) identified in six individuals out of a cohort of 248 ASD cases	2014	Adamsen et al.

Highlights.

• This review discusses dopamine and dysregulation of the dopamine system in Autism Spectrum Disorder.

• Mutations in dopamine system genes have been associated with autism.

• Animal models of autism demonstrate a number of dopaminergic system abnormalities.

• Changes in dopamine signaling may represent a contributing factor to autism in some individuals.