The serotonin system in autism spectrum disorder: from biomarker to animal models

Abstract

Elevated whole blood serotonin, or hyperserotonemia, was the first biomarker identified in autism spectrum disorder (ASD) and is present in more than 25% of affected children. The serotonin system is a logical candidate for involvement in ASD due to its pleiotropic role across multiple brain systems both dynamically and across development. Tantalizing clues connect this peripheral biomarker with changes in brain and behavior in ASD, but the contribution of the serotonin system to ASD pathophysiology remains incompletely understood. Studies of whole blood serotonin levels in ASD and in a large founder population indicate greater heritability than for the disorder itself and suggest an association with recurrence risk. Emerging data from both neuroimaging and postmortem samples also indicate changes in the brain serotonin system in ASD. Genetic linkage and association studies of both whole blood serotonin levels and of ASD risk point to the chromosomal region containing the serotonin transporter (SERT) gene in males but not in females. In ASD families with evidence of linkage to this region, multiple rare SERT amino acid variants lead to a convergent increase in serotonin uptake in cell models. A knock-in mouse model of one of these variants, SERT Gly56Ala, recapitulates the hyperserotonemia biomarker and shows increased brain serotonin clearance, increased serotonin receptor sensitivity, and altered social, communication, and repetitive behaviors. Data from other rodent models also suggest an important role for the serotonin system in social behavior, in cognitive flexibility, and in sensory development. Recent work indicates that reciprocal interactions between serotonin and other systems, such as oxytocin, may be particularly important for social behavior. Collectively, these data point to the serotonin system as a prime candidate for treatment development in a subgroup of children defined by a robust, heritable biomarker.

Findings from multiple domains of research have implicated the serotonin system in autism spectrum disorder (ASD). Initial findings more than fifty years ago identified elevated whole blood serotonin levels, termed hyperserotonemia, in a subset of children with autism. Since then, work on the serotonin system in ASD has extended intermittently as new tools have become available, from pharmacology to genetics to neuroimaging. Emerging findings, driven largely by advances in mouse models, point to the importance of serotonin for social function, repetitive behavior, and sensory development. With the increasing recognition of heterogeneity in ASD, biological markers will play an important role in identifying subsets of children who are more likely to share common risk or benefit from the same treatments. The combination of a robust biomarker with a rich understanding of serotonin neurobiology suggests that the serotonin system is particularly ripe for treatment development in a subset of children with ASD.

Brief overview of the serotonin system

For millions of years, serotonin (5-hydroxytryptamine, 5-HT) has existed as a signaling molecule across phylogeny (Hay-Schmidt, 2000). It is produced from the essential amino acid, tryptophan, via a two-step synthetic pathway. In the first step, tryptophan is converted to 5-hydroxytryptophan (5-HTP) by the rate-limiting enzyme in 5-HT synthesis, tryptophan hydroxylase (Figure 1). There are two isoforms of tryptophan hydroxylase, TPH1 and TPH2, which are primarily responsible for 5-HT synthesis in the periphery and central nervous system (CNS), respectively (Lovenberg et al., 1967, Walther et al., 2003). In the final step, the intermediate product, 5-HTP, is converted to 5-HT by aromatic acid decarboxylase (AADC). Degradation of 5-HT primarily occurs by the mitochondrial bound protein monoamine oxidase A (MAOA), leading to the production of the metabolite, 5-hydroxyindoleacetic acid (5-HIAA). Importantly, serotonin also serves as an intermediate substrate for melatonin synthesis.

Figure 1. Overview of the Serotonin System.

This cartoon depicts the serotonin synthesis, release, reuptake, and degradation, as well as the canonical roles of a subset of serotonin receptors. TPH = tryptophan hydroxylase. AADC = aromatic acid decarboxylase. VMAT = vesicular monoamine transporter. SERT = serotonin transporter. MAO = monoamine oxidase. 5HT1-7 = serotonin receptor subgroups (oval shape) and individual receptors (goblet shape). Figure created using Ingenuity Pathway Analysis (Qiagen, Valencia, CA).

Though more recognized for its role as a neurotransmitter, the vast majority of 5-HT produced in the body is located in the periphery. Specifically, enterochromaffin cells that line the lumen of the gastrointestinal tract are the primary source of peripheral 5-HT, which is then taken up by platelets as they pass through enteric circulation (Anderson et al., 1987b). Interestingly, peripheral production of 5-HT expands during pregnancy, with the pancreas contributing to maternal 5-HT levels and the placenta producing 5-HT that contributes to fetal forebrain 5-HT levels (Kim et al., 2010, Bonnin et al., 2011), suggesting an additional developmental role for peripheral 5-HT.

Although gut-derived 5-HT influences numerous aspects of peripheral physiology, it is unable to cross the mature blood brain barrier and interact with neural tissue (Hardebo and Owman, 1980). Consequently, 5-HT found in the brain is produced by TPH2-expressing serotonergic neurons in the midbrain and hindbrain. Serotonergic neurons are organized into nine discrete clusters (B1–B9), collectively known as the raphe nuclei (Dahlstroem and Fuxe, 1964). While the more caudal raphe nuclei (B1–B5) project to the peripheral nervous system (PNS), the rostral groups (B6–B9), the dorsal and median raphe nuclei, primarily send their projections to forebrain structures (Conrad et al., 1974, O’Hearn and Molliver, 1984). Despite their relatively small number, serotonergic neurons innervate a broad collection of brain regions (Jacobs and Azmitia, 1992), allowing 5-HT to influence neural circuitry underlying a myriad of behaviors.

Reflective of its seemingly ubiquitous presence in the brain and periphery, 5-HT exerts its effects through 14 genetically distinct receptor subtypes (Hannon and Hoyer, 2008, Millan et al., 2008). Moreover, mRNA editing and alternative splicing adds another layer of complexity to an already diverse signaling system (Burns et al., 1997, Kishore and Stamm, 2006). While a thorough description of 5-HT receptor function in the brain and periphery is beyond the scope of this review, it should be clearly noted that the unique temporal and spatial patterns of 5-HT receptor expression implicate 5-HT as a key developmental molecule (Lidow and Rakic, 1992, Bonnin et al., 2006).

Arguably the most crucial aspect of 5-HT homeostasis is the clearance of extracellular 5-HT. In addition to terminating 5-HT receptor signaling, reuptake of 5-HT back into the cell helps recycle the monoamine for future use (Blakely and Edwards, 2012). The serotonin transporter (SERT, 5-HTT) is an integral plasma membrane protein that actively transports 5-HT in a Na⁺/Cl⁻-coupled dependent and antidepressant-sensitive manner (Blakely et al., 1991). In addition to being found on presynaptic terminals at serotonergic synapses in the brain, SERT is expressed at numerous sites in the periphery, including in the platelet (Qian et al., 1995). Since the discovery of SERT, there has been considerable interest in understanding its role in psychiatric disorders. Initially, this interest was driven by the efficacy of selective serotonin reuptake inhibitors (SSRIs), which bind to SERT and block 5-HT uptake, in conditions such as depression, anxiety, and obsessive-compulsive disorder (Vaswani et al., 2003). Genetic studies have now implicated SERT gene variation in moderating risk of depression and other neuropsychiatric symptoms following environmental stressors, particularly during childhood (Caspi et al., 2002, Caspi et al., 2003). As highlighted in other reviews, genetic and pharmacological studies have pointed to the critical role of SERT and the serotonin system during neurodevelopment (Suri et al., 2014, Kepser and Homberg, 2015).

Serotonin and neurodevelopment

Serotonin has now been linked to developmental processes such as cell proliferation, migration and differentiation, but the most definitive evidence of a developmental role for 5-HT first emerged from studies of the rodent somatosensory system. Barrel field architecture within rodent primary somatosensory cortex is the cortical representation of peripheral snout whiskers (Welker, 1971). Both 5-HT immunostaining and ³[H]-Citalopram binding identified a transient serotonergic innervation of the barrel cortex and other primary sensory areas that peaked during the first postnatal week of life and disappeared by postnatal day 21 (P21) in rodents (Fujimiya et al., 1986, D’Amato et al., 1987). The 5-HT in barrel cortex is actually contained in glutamatergic thalamocortical axons that lack synthetic capacity but take up extracellular 5-HT via SERT (Lebrand et al., 1996). It remains unclear, however, how 5-HT captured by TCAs plays a functional role during neurodevelopment.

Several groups have manipulated 5-HT levels in the developing brain to elucidate 5-HT-mediated mechanisms of sensory map formation. Decreased or ablated 5-HT during development causes, at most, subtle changes in barrel cortex formation, including delayed sensory map maturation (Bennett-Clarke et al., 1994, Bennett-Clarke et al., 1995, Persico et al., 2000, van Kleef et al., 2012). In contrast, manipulations that increase 5-HT levels during development show a 5-HT_1B-dependent disruption of barrel field architecture (Cases et al., 1996, Salichon et al., 2001). Further, thalamocortical 5-HT_1B signaling also modulates responsiveness to axon guidance cues (Bonnin et al., 2007). Beyond structural disruption, indirect measures of somatosensory function in various serotonin transporter mouse models indicate altered processing of whisker stimulation (Esaki et al., 2005, Dawson et al., 2009). In parallel, other groups have found that serotonin also impacts visual and auditory cortex development and function (Salichon et al., 2001, Simpson et al., 2011).

Emerging evidence also points to a broader role for 5-HT extending beyond sensory brain regions. Ansorge and colleagues demonstrated that exposure to a serotonin reuptake inhibitor early in postnatal life results in increased anxiety-like behavior in adult mice (Ansorge et al., 2004), with evidence that this effect is mediated by the medial prefrontal cortex (Rebello et al., 2014). Genetic ablation of SERT also alters other aspects of brain development, including overall brain growth (Page et al., 2009) and interneuron migration into the cortex (Riccio et al., 2009) (Figure 2). The importance of SERT specifically and 5-HT generally in neurodevelopment converges with accumulating evidence pointing to abnormalities in the peripheral and central serotonin system in autism spectrum disorder.

Figure 2. Venn Diagram of the Serotonin System and Autism Spectrum Disorder.

Key roles or research findings within the serotonin system are shown in the purple circle on the left. Research or clinical findings in autism spectrum disorder are depicted in the teal circle on the right. The overlapping blue region demonstrates findings directly related to serotonin in ASD. Some findings in ASD, such as constipation, overlap with known roles of the serotonin system, such as gut motility – even though they have not been directly connected – and these are shown in black type in parallel positions within the corresponding circles. Other findings, such as somatic overgrowth or regulation of vasoconstriction, so no such overlap, and these are shown in gray type.

Hyperserotonemia: the first biomarker in autism spectrum disorder

The identification of hyperserotonemia as a biomarker in autism spectrum disorder (ASD) preceded the broad acceptance of 5-HT as a neurotransmitter (Folk and Long, 1988). In 1961, Schain and Freedman first reported that six children in a cohort of 23 individuals with a diagnosis of “infantile autism” had dramatically increased blood 5-HT levels (Schain and Freedman, 1961). The 5-HT contained in whole blood is now understood to be almost completely contained in platelets (Anderson et al., 1987a). In contrast with platelet stores of 5-HT, only 1% of blood 5-HT is contained in plasma, and it is unclear if plasma 5-HT levels are changed in autism (Cook et al., 1988a, Spivak et al., 2004).

The description of hyperserotonemia occurred in an era when autism was rarely diagnosed and was attributed by many to poor parenting. Only sixteen years earlier, Leo Kanner had published his initial case series of children with abnormal social behavior and repetitive patterns of behavior (Kanner, 1943). The diagnostic criteria and terms for autism have evolved substantially over the years, with the current definition of autism as a spectrum disorder (American Psychiatric Association, 2013) reflecting the substantial heterogeneity and unclear boundaries of these symptoms. Even while the diagnosis has climbed from very rare to more than 1% of school-aged children (Developmental Disabilities Monitoring Network, 2014), rates of hyperserotonemia, typically defined as whole blood 5-HT levels above the 95^th percentile in the control population, have remained stable in ASD. A meta-analysis indicated that this biomarker is present in more than 25% of the ASD population (Gabriele et al., 2014). Of note, probands with ASD from multiplex families (more than one affected child) have been found to have higher whole blood 5-HT levels than probands from simplex families (Piven et al., 1991). Interestingly, elevated whole blood serotonin has been reported in children with obsessive-compulsive disorder (OCD), which shares some symptoms with ASD such as repetitive behaviors, who have a family history of OCD, but not in those without a family history (Hanna et al., 1991).

Initial debate focused on whether hyperserotonemia was unique to autism. While there were a few initial reports of elevated serotonin levels in intellectual disability, these occurred when ASD was rarely diagnosed in the severely affected or non-verbal population (Partington et al., 1973, Hanley et al., 1977a). Subsequent work has found a weak inverse correlation between whole blood 5-HT and measures of intelligence but has not identified significantly increased rates of hyperserotonemia in intellectual disability (Cook et al., 1988b, Cook et al., 1990, Mulder et al., 2004, Weiss et al., 2004b, Gabriele et al., 2014). One study in an ethnically more homogeneous population suggested a bimodal distribution of whole blood 5-HT levels in ASD, with one group similar to the normal distribution of controls and the second mode higher than the others (Mulder et al., 2004). Despite its strong and specific association with ASD, no prospective study has yet assessed whether hyperserotonemia may predict ASD risk in infants, including baby siblings of children with ASD.

Several groups have attempted to find clinical correlates of hyperserotonemia in ASD but with mixed results for symptoms such as stereotypic and self-injurious behavior and with some inconsistency across studies (McBride et al., 1998, Mulder et al., 2004, Kolevzon et al., 2010, Sacco et al., 2010). The relationship between hyperserotonemia and other ASD symptoms or frequently co-occurring disorders has been underexplored, including systems that are known to relate to 5-HT function, such as sensory function and gastrointestinal symptoms.

Potential mechanisms of hyperserotonemia

The potential mechanisms underlying platelet 5-HT levels have been explored both in ASD families and in the general population. Essentially, elevated 5-HT levels in platelets could be due to one of four mechanisms: increased 5-HT production by enterochromaffin cells in the intestine, increased uptake of 5-HT into the platelet, decreased metabolic breakdown of 5-HT, or altered platelet release (Cook et al., 1993a). Most studies have focused on the platelet, rather than on the source of 5-HT in the gut. We recently identified a small positive correlation between higher whole blood 5-HT levels and more gastrointestinal symptoms in children with ASD (Marler et al., In Press), although it isn’t possible to assess the nature of this relationship. Decreased blood tryptophan levels have been described in a few but not all studies in ASD, which could be consistent with altered 5-HT synthesis but could also reflect dietary or other metabolic changes (Anderson et al., 1987c, D’Eufemia et al., 1995, Croonenberghs et al., 2000, Boccuto et al., 2013). To date, no studies have directly evaluated the relationship between 5-HT levels in the platelet and gut 5-HT production or content.

Most platelet studies in ASD have focused on the serotonin transporter. Studies with various SERT ligands have yielded little evidence of changes in specific binding, either in children with ASD or in family members (Anderson et al., 1984, Cook et al., 1993b, Anderson et al., 2002). Two studies have found a positive correlation between the maximal velocity (V_max) of platelet 5-HT uptake and whole blood 5-HT levels in either participants with ASD or their family members (Cook et al., 1993b, Anderson et al., 2002). Neither study found a significant difference in 5-HT uptake between hyperserotonemic and non-hyperserotonemic participants; although they had less power for this analysis than for the correlation.

A few studies have also evaluated other elements of the platelet 5-HT system, including serotonin receptor binding. In separate studies, decreased platelet binding of the 5-HT receptor agonist LSD (³H- or ¹²⁵I-labeled) has been identified in participants with ASD and first-degree relatives of hyperserotonemic probands (McBride et al., 1989, Cook et al., 1993b); although one study using a different ligand found no difference (Perry et al., 1991). The primary metabolite of serotonin, 5-HIAA, is also increased in the urine in a few ASD studies, with little evidence that there is altered metabolism of 5-HT (Hanley et al., 1977b, Minderaa et al., 1994, Mulder et al., 2010). Interestingly, one study reported that N-acetylserotonin (NAS), which is a precursor in melatonin synthesis, is also increased in ASD, but neither NAS nor melatonin show a correlation with 5-HT levels despite a robust negative correlation between NAS and melatonin levels (Pagan et al., 2014).

In addition to platelet physiological and biochemical studies, genetic approaches have also been applied to understand the control of whole blood 5-HT levels. Evidence for heritability of whole blood 5-HT levels initially emerged from correlations between whole blood serotonin levels in ASD probands and their parents and siblings (Kuperman et al., 1985, Abramson et al., 1989, Cook et al., 1990, Leventhal et al., 1990, Leboyer et al., 1999). A large study of whole blood 5-HT levels in a Hutterite founder population confirmed that whole blood serotonin levels are strongly heritable in an unaffected population, with both additive and dominance components that sum up to a broad heritability of 0.99 (Abney et al., 2001), substantially higher than estimates of heritability for ASD itself. In a small twin study, the maximum rate of 5-HT uptake was also found to be highly heritable (Meltzer and Arora, 1988).

Based upon its high heritability, linkage and association techniques were applied to map 5-HT levels as a quantitative trait in the Hutterite population, identifying significant association at a functional polymorphism in the integrin β3 subunit gene (ITGB3) as well as suggestive association at the Vitamin D receptor gene (Weiss et al., 2004a). Recently, vitamin D was shown to regulate 5-HT synthesis (Patrick and Ames, 2014), in further support of the relationship between vitamin D and 5-HT. Follow-up analyses on the quantitative trait loci revealed that ITGB3, as well as the serotonin transporter gene (SLC6A4), were primarily associated with whole blood 5-HT levels in males (Weiss et al., 2005). Carneiro and colleagues subsequently demonstrated physical and functional interaction between the corresponding proteins in controlling platelet 5-HT uptake as well as aggregation (Carneiro et al., 2008). Remarkably, three studies have reported gene-gene interaction between ITGB3 and SLC6A4 in association with autism (Weiss et al., 2006, Coutinho et al., 2007, Mei et al., 2007), supporting the idea that whole blood 5-HT may serve as an intermediate phenotype for mapping autism susceptibility genes. Candidate gene association studies in ASD have also identified an association between SLC6A4 polymorphisms and whole blood 5-HT levels or SERT function (Coutinho et al., 2004, Coutinho et al., 2007, Cross et al., 2008).

Complementing analyses in the human population, the Blakely laboratory reported that whole blood 5-HT levels are also heritable in recombinant inbred BxD mouse strains that comprise multiple congenic lines derived from an initial cross of C57BL/6J with DBA/2J, with a heritability estimate of 0.60 (Ye et al., 2014). Genetic mapping identified two loci with suggestive evidence of linkage with whole blood 5-HT levels in the BxD population, again with evidence of sexual dimorphism in control of this quantitative trait (Ye et al., 2014). Denser mapping of sequence variation in the Hutterite population and more statistical power in the BxD analyses would be necessary to more accurately assess the consistency of findings between human and mouse studies.

Of note, in relation to the consistency of hyperserotonemia as a biomarker, the serotonin system has been largely unrepresented in the list of high confidence ASD genes identified by de novo, gene-disrupting mutations in two or more children with ASD (De Rubeis et al., 2014, Iossifov et al., 2014). This could be taken as evidence that the serotonin system is less important than previously thought; however, the high heritability of whole blood 5-HT could suggest that individuals with hyperserotonemia are less likely to have a de novo event and more likely to fit the common, inherited variant model still favored by epidemiological data (Gaugler et al., 2014). It is also possible that hyperserotonemia is less well represented in the group of patients with intellectual disability and/or epilepsy that shows an over-representation of de novo, gene-disrupting variants (Robinson et al., 2014, Sanders et al., 2015). These ideas could be tested by assessing the contribution of de novo mutations in individuals with or without hyperserotonemia.

The central serotonin system in autism spectrum disorder

Accumulating findings indicate that the brain 5-HT system is also altered in ASD. Based upon the peripheral findings of increased platelet 5-HT, one might hypothesize that increased 5-HT uptake or storage in the presynaptic neuron would lead to decreased synaptic 5-HT. Obviously, it is not possible to get a direct assessment of synaptic 5-HT in humans. One surrogate measure is to assess brain 5-HT synthesis, which appears to follow an altered developmental pattern in autism (Chugani et al., 1999, Shoaf et al., 2000). Other studies have considered serotonin receptor or transporter binding. Paralleling platelet binding studies, two neuroimaging studies have found decreased 5-HT₂ receptor binding: a SPECT study in adults with Asperger’s syndrome (Murphy et al., 2006) and a PET study in parents of children with autism (Goldberg et al., 2009). A postmortem study found decreases in both 5-HT_2A and 5-HT_1A binding in ASD (Oblak et al., 2013). Consistent findings showing decreased 5-HT₂ receptor binding in platelet, neuroimaging, and post-mortem studies support the idea that peripheral alterations in the serotonin system may be an important marker of central abnormalities in autism.

Findings have been less consistent for the serotonin transporter. Two reports have found decreased binding to SERT in ASD: a SPECT study in children with autism (Makkonen et al., 2008) and a PET study in young adults with autism (Nakamura et al., 2010). Another report found no changes in a PET study of SERT binding in adults with Asperger’s disorder (Girgis et al., 2011). One postmortem study found decreased SERT binding in deep layers of the fusiform gyrus but no difference in superficial layers or in the posterior cingulate cortex (Oblak et al., 2013). In contrast, the Azmitia laboratory described an increase in axons showing SERT immunoreactivity in postmortem tissue from individuals with autism spectrum disorder, spanning two to twenty-nine years old (Azmitia et al., 2011). While it is tempting to view this as contradicting the neuroimaging data, the antibody recognizes a different protein domain than the binding ligand, and the Azmitia analyses were focused on immunoreactive axons rather than total SERT binding. In sum, it remains uncertain whether the serotonin transporter is altered in ASD at a group level, let alone whether any alteration is due to changes in the number or projections of serotonergic neurons or to changes in SERT expression specifically.

Genes in the serotonin system have also been studied in relation to structural or functional MRI findings in ASD. Besides hyperserotonemia, one of the most consistent biomarker findings in ASD is an altered developmental trajectory of brain growth, with increased brain size during childhood that normalizes over time (Courchesne et al., 2001). In the first ASD neuroimaging study to examine the functional SERT gene (SLC6A4) promoter polymorphism (5-HTTLPR), Wassink and colleagues found that the low SERT expressing short allele was significantly associated with increased cerebral cortex grey matter volume in children with autism (Wassink et al., 2007).

In contrast, a subsequent study examining 5-HTTLPR effects on cortical grey matter volume in adults with Asperger’s disorder reported no significant genotype associations (Raznahan et al., 2009). In addition to basic structural MRI measures, emerging literature indicates that SLC6A4 variants may impact the connectivity and function of neural circuits in ASD. The Monk laboratory reported that affected children and adolescents with high-expressing 5-HTTLPR genotypes had stronger connectivity of the default network, a broadly distributed, interconnected group of brain regions that are active during wakeful rest, with the opposite finding in typically developing children (Wiggins et al., 2012). They have also explored 5-HTTLPR genotype effects on amygdala responsiveness in ASD, based upon previous reports in healthy adults (Hariri et al., 2002). Using fMRI to examine potential SERT genotype differences in amygdala habituation to face stimuli in children and adolescents with ASD, they found that affected individuals with low SERT expressing genotypes had decreased amygdala habituation upon repeated exposure to faces (Wiggins et al., 2014).

Finally, limited pharmacology studies also point to the central 5-HT system in autism. Tryptophan depletion, which is expected to cause decreased synaptic 5-HT, leads to worsened repetitive behaviors and irritability in autism (McDougle et al., 1993, McDougle et al., 1996). Neuroimaging studies show differential effects of tryptophan depletion on brain region activity by fMRI in the ASD population in contrast to typical controls, particularly in response to presentation of emotional faces (Daly et al., 2012, Daly et al., 2014). Adult studies suggested that serotonin reuptake inhibitors relieve symptoms of irritability and rigid-compulsive behavior in autism (Gordon et al., 1993, Hollander et al., 2005, Hollander et al., 2012); however studies in children have been less supportive, perhaps because of greater adverse events in children or because of methodological difficulties (King et al., 2009, King et al., 2013). More consistent data supports the use of risperidone and aripiprazole (McCracken et al., 2002, McDougle et al., 2005), atypical antipsychotic medications with antagonism at multiple monoamine receptors, including the serotonin receptor 5-HT_2A.

Some studies have also examined prenatal exposure to serotonin reuptake inhibitors (SRIs) in relation to ASD risk. It is very difficult to disentangle exposure from the indications for SRI use, such as depression, anxiety disorders, or OCD, each of which is more common in family members of children with ASD (Harrington et al., 2013). An initial small study in a California population reported association of prenatal exposure to SRIs with ASD risk; although they also described more severe psychiatric history in women taking SRIs (Croen et al., 2011). Subsequently, a large study in the Swedish registry supported this association (Rai et al., 2013), whereas a large study in the Danish population found no significant association (Hviid et al., 2013). Another smaller study found no significant effect of SRI exposure overall but with some secondary analyses suggesting risks specific to male fetuses or exposure during the first trimester (Harrington et al., 2014). Overall, it is difficult to assess whether these mixed data indicate a contribution of SRI exposure to ASD risk or whether this represents an association driven by a separate, common factor, such as maternal psychiatric history (Levitt, 2011).

Across all of these studies of the central serotonin system in relation to ASD, from neuroimaging to pharmacology, investigators have not evaluated whether there is a relationship with peripheral 5-HT measures. Since hyperserotonemia appears to define a subgroup of children, it would be very helpful to know if this subgroup drives 5-HT receptor binding differences or altered response to tryptophan depletion observed in ASD. Accounting for the peripheral biomarker could also potentially clarify ambiguous or contradictory results.

Genetic linkage and association studies of SLC6A4 in ASD

Genetic findings in ASD have largely centered on the serotonin transporter gene SLC6A4, initially based upon its position as the primary candidate gene for the hyperserotonemia biomarker. Unfortunately, very few of these studies have included measurement of whole blood 5-HT levels as a heritable biomarker in ASD. Family-based association studies involving common SLC6A4 variation were inconsistent, with the largest studies finding association with the 5-HTTLPR short allele (Devlin et al., 2005). A recent analysis indicated that mixed effects could explain some of this inconsistency, with risk associated with the long allele in mothers and the short allele in probands (Kistner-Griffin et al., 2011). The first linkage study in ASD found the strongest single-point evidence of linkage for an intron 2 polymorphism in SLC6A4 (International Molecular Genetic Study of Autism Consortium, 1998), and two subsequent linkage studies have identified significant evidence of linkage at the chromosome 17q11 region harboring SLC6A4 (Yonan et al., 2003). Two groups subsequently found that linkage signals on 17q strengthened when considering only families with affected males but no affected females (Stone et al., 2004, Sutcliffe et al., 2005).

Common polymorphisms in SLC6A4 did not appear to explain this linkage signal (McCauley et al., 2004), leading Sutcliffe and colleagues to focus on rare variants. In families with evidence of linkage near SLC6A4 (Sutcliffe et al., 2005), they identified multiple rare variants of SERT, including Gly56Ala and the novel SERT variants, Ile425Leu, Phe465Leu, and Leu550Val, located in highly conserved transporter transmembrane domains. Interestingly, the Ile425Leu variant displayed a unique segregation pattern indicative of the male-biased linkage signal, with the maternally transmitted Leu425 allele present in two affected sons as well as three unaffected daughters. Furthermore, another rare SERT mutation located at the same protein residue, Ile425Val, was previously observed in two unrelated families with a history of OCD and other psychiatric comorbidities, including ASD (Ozaki et al., 2003). The most common rare variant, Gly56Ala, was enriched in families with linkage to 17q11 but not in families without evidence of linkage, a finding corroborated by a subsequent screen of SLC6A4 in a case-control association study (Sakurai et al., 2008).

Interestingly, affected carriers of the Gly56Ala, Ile425Leu, Phe465Leu, and Leu550Val variants exhibited increased rigid-compulsive behaviors as measured by the Autism Diagnostic Interview (ADI-R). The Gly56Ala variant had a unique association with ADI-R ratings of sensory aversion in affected individuals (Sutcliffe et al., 2005). In support of these findings, we found association between high-expressing SLC6A4 genotypes and tactile hypersensitivity, the most common form of sensory aversion, in children with ASD (Schauder et al., 2015). Aberrant sensory behavior could be related to the reported role of SERT in the establishment of sensory-related circuits during neurodevelopment (Persico et al., 2001, Salichon et al., 2001, Esaki et al., 2005).

In addition to genetic and behavioral analyses, the Blakely laboratory examined the impact of the Ala56 variant on SERT function in lymphocytes taken from genotyped study participants. Strikingly, lymphocytes expressing the Ala56 variant displayed significantly increased 5-HT transport at basal conditions. 5-HT transport in Ala56 lymphocytes was also insensitive to regulation by p38 mitogen-activated protein kinase (MAPK) and protein kinase G (PKG) activators, which normally enhance SERT function (Blakely et al., 2005). Moreover, these findings were subsequently confirmed in heterologous cell systems devoid of potential genetic modifiers of SERT (Prasad et al., 2009).

The SERT Ala56 mouse as a model of ASD risk and hyperserotonemia

As the most common of the SERT amino acid variants, the Ala56 allele was the natural initial target for a genetic knock-in mouse model (Veenstra-VanderWeele et al., 2012). As long hypothesized, altered SERT function in these animals led to a significant increase in whole blood 5-HT levels in comparison to wildtype littermate controls. Paralleling findings in cell models, these animals also showed increased 5-HT clearance in the brain, as well as increased basal, p38-MAPK-dependent phosphorylation of SERT. Both 5-HT_1A and 5-HT_2A receptors showed hypersensitivity to agonist drugs, suggesting a homeostatic response to chronic decreases in synaptic 5-HT levels. Consistent with increased sensitivity of 5-HT_1A receptors, electrophysiology studies in SERT Ala56 midbrain slides revealed greater inhibition of neuronal firing with bath application of 5-HT, in comparison to wildtype littermate controls. Surprisingly, 5-HT-sensitive neurons in SERT Ala56 animals showed diminished baseline firing in this preparation, suggesting that heightened receptor sensitivity may overcompensate for diminished synaptic 5-HT, at least in this preparation (Veenstra-VanderWeele et al., 2012).

SERT Ala56 mice also exhibited changes in social and repetitive behavior. The common three-chamber test of sociability was complicated by the inbred strain background used (129S6/S4), with some animals showing minimal exploration of the apparatus, regardless of genotype. When mice with fewer than four chamber entries were excluded from the analysis, wildtype littermate controls showed significant preference for the social chamber, whereas the mutant animals did not. Two other tests supported a change in social behavior in these animals. First, the tube test for dominance revealed that SERT Ala56 animals were much more likely to back away when confronted with an animal of the opposite genotype. Second, SERT Ala56 pups separated from their dam at P7 vocalized much less than their wildtype littermate controls. Further, SERT Ala56 mice showed a repetitive cage lid climbing/hanging behavior in their home cage, with some animals showing over a thousand separate bouts of this behavior in twenty-four hours. Finally, the mutant animals showed altered prepulse inhibition, with a significant genotype by prepulse intensity interaction effect, reflecting increased startle response at lower prepulse intensities (Veenstra-VanderWeele et al., 2012).

The challenge of assessing social and cognitive behaviors on the original 129S6/S4 inbred strain background motivated a backcross of the SERT Ala56 allele onto a pure C57BL/6J (B6) inbred strain background. Since the B6 SERT sequence differs from 129S SERT at two amino acids (Glu39Gly and Arg152Lys) (Carneiro et al., 2009), the native 129S SERT was also backcrossed in parallel with the mutant SERT backcross, to greater than 99% congenic status (Kerr et al., 2013). Unfortunately, no statistically significant differences in synaptosome 5-HT uptake, whole blood 5-HT, or 5-HT_1A or 5-HT_1B receptor sensitivity were observed in B6 SERT Ala56 mice, indicating a background strain dependence of the mutant serotonin transporter phenotype. Not surprisingly, most of the behavioral phenotypes were also non-significant, with the exception of the tube test for dominance, which showed consistent effects. Ultrasonic vocalizations at P7 were actually increased in the mutant animals on the B6 background, which is difficult to interpret. Overall, these findings suggest that the Ala56 variant has less impact on SERT function on the B6 background, perhaps owing to altered regulatory partners in B6 animals, which express two other SERT variants that impact function (Carneiro et al., 2009). The observed differences across strains may enable mapping of modifier loci that could be relevant in humans.

There are a number of logical next steps in the SERT Ala56 animals. One key question is whether the brain and behavioral phenotypes in these animals are developmental or dynamic. As has been found for the SERT null mouse, it is quite possible that enhanced SERT function during development leads to altered sensory and medial prefrontal cortex development (Rebello et al., 2014, Chen et al., 2015). As noted above, the somatosensory system would be of particular interest, as would multisensory integration, which is abnormal in some children with ASD (Stevenson et al., 2014). The developmental impact could extend to maternal effects of altered SERT function, as suggested by association studies (Kistner-Griffin et al., 2011) and as observed in other models of altered 5-HT synthesis or receptor function (Cote et al., 2007, Gleason et al., 2010). The original linkage and association findings implicated the SERT gene region specifically in males (Sutcliffe et al., 2005, Weiss et al., 2005), and published studies thus far have not examined which observed phenotypes extend to female animals. Potential rescue experiments could target SERT regulation in presynaptic neurons or 5-HT receptors in post-synaptic neurons but could also target the consequences of developmental alterations due to increased SERT activity (Rebello et al., 2014, Chen et al., 2015). Finally, in addition to opportunities to probe the neural mechanisms of autism-relevant behavior, these animals also offer the potential to understand the mechanisms underlying peripheral changes in the 5-HT system in ASD, including platelets but also other organ functions impacted by 5-HT including gut motility and bone density (Yadav et al., 2008, Gorrindo et al., 2012, Gershon, 2013, Neumeyer et al., 2013, Margolis et al., 2014).

Other rodent models of altered 5-HT function relevant to ASD

At this point, knockout of most of the known genes in the 5-HT system has been reported. Hundreds of papers have now been published describing the various phenotypes of the SERT knockout mouse (Bengel et al., 1998), including increased anxiety-like behavior, decreased aggression, and disruption of brain architecture in various regions, including the sensory cortex, paralleling the MAOA knockout described above, and the amygdala (Murphy and Lesch, 2008). Despite seeing decreased platelet 5-HT levels, the opposite of what is observed in the hyperserotonemia of autism, these mice also show decreased sociability in the three-chamber test, potentially reflecting the confound of anxiety-like behavior (Moy et al., 2008). Interestingly, when bred with mice haploinsufficient for Pten, a putative autism susceptibility gene, SERT heterozygous knockout mice show an exacerbation of macrocephaly and decreased sociability (Page et al., 2009). These findings may be consistent with the idea that SERT activity may be constrained within a certain range for normal function and behavior, with elevated or diminished SERT function leading to vulnerability to autism-like behavior.

A number of other ASD-relevant phenotypes have been reported in mice lacking critical proteins in the 5-HT system. For example, mice lacking tryptophan hydroxylase 2, responsible for central 5-HT synthesis, show decreased ultrasonic vocalizations, decreased sniffing of social odors, apparent deficits in social memory, and cognitive inflexibility (Kane et al., 2012, Del’Guidice et al., 2014, Mosienko et al., 2015). Mice lacking MAOA, responsible for breaking down 5-HT, show decreased social approach, decreased ultrasonic vocalizations, and impaired reversal learning (Bortolato et al., 2013). Mice lacking 5-HT_1B show increased aggression, with some suggestion that 5-HT_1A has opposite effects (Saudou et al., 1994, Olivier et al., 1995). Interestingly, mice lacking 5-HT_1A also show elevated platelet 5-HT levels that emerge in the second week of life (Janusonis et al., 2006). Female mice lacking the 5-HT_3A receptor show decreased reciprocal social interaction, and both male and female animals show decreased social transmission of food preference (Smit-Rigter et al., 2010).

Beyond the enzymes, transporters, and receptors that are typically considered as part of the 5-HT system, a number of transcription factors and regulatory partners have emerged. As noted above, the integrin β3 gene was associated with whole blood 5-HT levels in the human population. Mice lacking integrin β3, which impacts SERT function both in platelets and in the brain (Carneiro et al., 2008, Whyte et al., 2014), appear to show a deficit in social memory, and exhibit repetitive grooming behavior (Carter et al., 2011). Using translational profiling of 5-HT neurons, Dougherty and colleagues identified CELF6 as a differentially expressed gene that was disrupted in one proband with ASD (Dougherty et al., 2013). In addition to decreased whole brain 5-HT levels, mice lacking Celf6 show decreased ultrasonic vocalizations and a perseverative-like phenotype on a hole board assay (Dougherty et al., 2013).

Mouse models of ASD risk with abnormal 5-HT

Curiously, despite consistent reports of hyperserotonemia in autism, as well as data suggesting that whole blood serotonin levels are strongly heritable, few labs have assessed this biomarker in genetic mouse models of ASD. Mouse models of 15q11-q13 duplication and Smith-Lemli-Opitz syndromes show alterations in the central 5-HT system (Waage-Baudet et al., 2003, Nakatani et al., 2009, Farook et al., 2012, Korade et al., 2013). The BTBR inbred mouse strain, which shows many autism-relevant behavioral phenotypes, shows decreased baseline SERT binding throughout the brain and increased 5-HT_1A activity in the hippocampus (Gould et al., 2011, Gould et al., 2014). Further, short-term tryptophan supplementation as well as acute treatment with the serotonin reuptake inhibitor fluoxetine or the 5-HT_1A partial agonist buspirone result in increased sociability in BTBR mice (Chadman, 2011, Gould et al., 2011, Gould et al., 2014, Zhang et al., 2015).

Interestingly, abnormalities in the 5-HT system have been reported in mice lacking Gtf2ird1, a key gene in the region deleted in Williams-Beuren syndrome (WBS), which includes intellectual disability and a hypersociable personality in many affected children. In addition to decreased anxiety-like behavior and some evidence of increased social interaction, animals lacking Gtf2ird1 show increased levels of 5-HIAA in amygdala and frontal cortex and enhanced sensitivity to 5-HT_1A stimulation in the prefrontal cortex (Young et al., 2008, Proulx et al., 2010). Two reports have described hyperserotonemia in a few patients with WBS and co-occurring ASD, suggesting that the 5-HT system could contribute to or modulate the WBS phenotype (Reiss et al., 1985, Tordjman et al., 2013).

A few reports indicate that the 5-HT system is perturbed in environmental models of ASD. The most robust environmental risk factor to be modeled in rodents is prenatal valproic acid exposure, which has been assessed primarily in rats. Reports of altered brain and gut 5-HT system function in these animals have been intriguing, with most studies reporting decreased 5-HT levels or innervation in various brain regions (Miyazaki et al., 2005, Tsujino et al., 2007, Winter et al., 2008, Dufour-Rainfray et al., 2010, de Theije et al., 2014).

Maternal immune activation or viral exposure has also become a popular rodent model. Of note, while exposure to infection in utero clearly contributes to risk of schizophrenia, it remains unclear when and how – and even whether – maternal immune activation contributes directly to ASD risk (Atladottir et al., 2012, Brown, 2012, Langridge et al., 2013, Zerbo et al., 2013a, Zerbo et al., 2013b). In a mouse model of maternal immune activation, Hsiao and colleagues described elevated serum 5-HT (Hsiao et al., 2013), although it is somewhat difficult to interpret serum 5-HT levels in relation to ASD, where platelet-free plasma levels appear unchanged (Anderson et al., 1987a, Anderson, 2007, Anderson et al., 2012). Other papers reported that maternal immune activation or virus exposure resulted in abnormal development of 5-HT neurons or decreased brain 5-HT levels (Miller et al., 2013b, Ohkawara et al., 2015). More epidemiological research is needed to better understand the right way to interpret these models, but substantial evidence suggests bidirectional interplay between serotonin and the immune system, including evidence that inflammation triggers increased SERT function (Baganz and Blakely, 2013).

Interplay between the serotonin and oxytocin systems

The serotonin system intersects with many other signaling systems, both peripherally and in the brain, with oxytocin (OT) as one of the most obviously relevant to the social deficits observed in ASD. Biomarker and genetic data for OT and its receptor have been promising but inconsistent in ASD and in relation to various measures of human social behavior (Modahl et al., 1998, Jacob et al., 2007, Hammock et al., 2012, Miller et al., 2013a, Parker et al., 2014, Skuse et al., 2014). Intranasal OT has gained momentum as a potential treatment for ASD but with similarly promising but inconsistent results thus far (Gordon et al., 2013, Aoki et al., 2014, Preti et al., 2014, Aoki et al., 2015, Auyeung et al., 2015, Guastella et al., 2015). Evidence points to multiple intersections of the 5-HT and OT systems, and interactions between these systems influence behaviors such as sociability, aggression, and anxiety that are relevant to ASD (Dolen, 2015). Interactions between genes or polymorphisms in the 5-HT and OT systems have been associated with ASD or related behaviors (Montag et al., 2011, Thanseem et al., 2012, Nyffeler et al., 2014), while animal models have been particularly useful for elucidating the mechanisms of 5-HT and OT interactions in specific brain regions and how they impact behaviors.

Examining the effects of “entactogens,” drugs that elicit social behavior, in animal models can reveal particular neural mechanisms crucial for social behavior. In particular, studies of MDMA (3,4-methylenedioxymethamphetamine), which causes reverse transport of 5-HT via SERT, have implicated interactions between the 5-HT and OT systems. Male rats given MDMA demonstrated increased plasma OT levels, which were blocked by systemic pre-administration of a 5-HT_1A receptor antagonist (Thompson et al., 2007). In a related experiment, MDMA also increased time the rats spent in social interaction, and intracerebroventricular pre-administration of an OT receptor antagonist blocked the effect (Thompson et al., 2007). Similarly, male mice given MDMA showed increased sociability and locomotor activity, behaviors which were blocked by a 5-HT_1A or OT receptor antagonist (Kuteykin-Teplyakov and Maldonado, 2014). Furthermore, MDMA-induced changes in c-Fos expression were reduced in hypothalamic OT-positive cells following pre-administration of a 5-HT_1A receptor antagonist in male rats (Hunt et al., 2011). Collectively, these findings indicate that the 5-HT system impacts social behavior via OT release.

Other work has investigated the reciprocal impact of the OT system on 5-HT release. Using a transgenic reporter, Yoshida and colleagues (2009) localized OT receptor gene expression in 5-HT cells in the raphe. Based upon this observation, they demonstrated that administration of OT directly into the median raphe led to increased 5-HT release. Further, central administration of OT led to increased time spent in the center of the open field test, an indication of anxiolysis, which was prevented by a 5-HT_2A/2C receptor antagonist (Yoshida et al., 2009). In humans, intranasal OT results in increased 5-HT_1A receptor availability as measured by PET in the dorsal raphe and a number of target regions, suggesting decreased, rather than increased, 5-HT release (Mottolese et al., 2014). Further work will be necessary to better understand the impact of OT on 5-HT release across species, especially given the limitations of indirect measurement in human studies.

Dölen and colleagues (2013) demonstrated that input from both the 5-HT and OT systems to the nucleus accumbens are essential for establishing social conditioned place preference (Panksepp and Lahvis, 2007). They found that presynaptic OT receptors on dorsal raphe projections within the accumbens induce release of 5-HT and, further, that the activation of the post-synaptic 5-HT_1B receptors is necessary to establish social preference (Dölen et al., 2013). Another study used conditional knockout mice to demonstrate that male mice lacking the OT receptor on 5-HT neurons showed lower frequency of aggression than wild type mice; whereas female mice did not exhibit genotype effects on behavior (Pagani et al., 2015). This sex difference is important, but few other studies in mice have included female animals in their analyses.

Predictably, the interactions between 5-HT and OT also influence development of these systems and behaviors. When OT was administered to prairie vole pups on postnatal day 1, it led to greater density of 5-HT axons by weaning, in a brain region-specific manner (Eaton et al., 2012). Conversely, when a non-selective 5-HT agonist was given to rats or voles prenatally, beginning from gestational day 12, fewer OT cells were evident in the paraventricular nucleus of the hypothalamus in adulthood, and time spent in typical affiliative social behaviors was lower than in saline-treated animals (McNamara et al., 2008, Martin et al., 2012). A similar treatment given to male and female rats pre- and postnatally revealed that the effects on the number of OT or 5-HT receptor positive cells were age- and sex-dependent, and effects on juvenile play behavior were only seen in females, not males (Madden and Zup, 2014). One area for further research includes development of multisensory integration, with evidence that both 5-HT and OT play critical roles in cross-modal plasticity within the sensory cortex (Jitsuki et al., 2011, Zheng et al., 2014).

Interestingly, one study reported a correlation between whole blood 5-HT and plasma OT levels in children and adolescents with ASD (Hammock et al., 2012). Serotonin levels correlated negatively with age and with OT levels. A parallel experiment in the same report demonstrated that 5-HT levels were higher in juvenile mice lacking the OT receptor, compared with their wildtype littermates (Hammock et al., 2012). This suggests that the OT system also affects the peripheral 5-HT system. Importantly, it does not rule out a reciprocal role for 5-HT affecting the OT system, as has been shown in the brain. This work suggests that, in addition to further elucidating effects of changes in specific neural systems, animal models can be useful for understanding potential biomarkers for ASD (Hammock et al., 2012).

Next steps to understand the contribution of the serotonin system to ASD

Despite more than fifty years of research, the hyperserotonemia biomarker has yet to fully reveal its relationship with ASD risk and pathophysiology. The pleiotropic effects of 5-HT, both peripherally and in the brain, present a challenge to developing a simple model of its contribution. One central question is whether its contribution is developmental, dynamic, or both. For instance, the impact of altered 5-HT uptake or breakdown on sensory development is clearly developmental and may not relate to targeted treatments in adulthood. In contrast, short-term tryptophan depletion or SRI administration in adult rats and humans alters reversal learning and other cognitive measures relevant to the repetitive or compulsive behavior observed in ASD (Clarke et al., 2004, Cools et al., 2008, Worbe et al., 2015).

In addition to the temporal dynamics of its influence as both a morphogen and a neurotransmitter, we need a better understanding of the regional specificity of serotonin’s effects on ASD-relevant behavior. A variety of techniques, from the cre-mediated excision (Chen et al., 2015) to optogenetic, pharmacogenetic, or viral silencing (Dölen et al., 2013), allow either temporary or permanent manipulation of the 5-HT system with varying levels of temporal and spatial control. These approaches can also better examine the impact of the peripheral 5-HT system on brain development and behavior. For instance, the placenta is the key source of 5-HT for early forebrain development, using tryptophan from the maternal circulation as a substrate for 5-HT synthesis (Bonnin et al., 2011). By walking back and forth between rodent models, human epidemiology, and cognitive neuroscience using short-term manipulation of the 5-HT system, we can further dissect the developmental and spatial impacts of the serotonin system on the social, communication, repetitive, and sensory behaviors underlying ASD.

More complete understanding of the temporal and spatial influences of altered 5-HT function on ASD-relevant behavior may yield new opportunities for rescue experiments, but our existing knowledge already points to potential targets. On the presynaptic side, the influence of p38-MAPK and PKG signaling in the SERT Ala56 animals suggests that it may be possible to tune down SERT function without blocking it entirely (Veenstra-VanderWeele et al., 2012). Drugs targeting serotonin receptors, particularly 5-HT_1A and 5-HT_2A, have shown promise for increasing social interaction or decreasing cognitive rigidity (File et al., 1996, Edwards et al., 2006, Boulougouris and Robbins, 2010, Gould et al., 2011, Amodeo et al., 2014).

Beyond considering what targets are logical within the serotonin system, it is critical that we start to measure biomarkers that define ASD subgroups when conducting studies of novel treatments. For instance, the intersection between the serotonin and oxytocin systems suggests that studies of intranasal oxytocin should likely measure whole blood 5-HT levels in participants. Given the results of Dölen and colleagues (2013), it is possible that participants with hyperserotonemia could have decreased synaptic 5-HT and therefore not benefit from oxytocin administration. Measuring serotonin levels adds little burden to treatment studies that are collecting blood to monitor drug metabolism and safety. Remarkably, whole blood 5-HT levels have not even been reported for trials of serotonin reuptake inhibitors or for studies of tryptophan depletion, each of which was motivated in large part by the hyperserotonemia biomarker. By subgrouping participants using a robust, heritable biomarker, future studies can identify individuals who may be more likely to benefit from a treatment, or more likely to experience adverse events. Potentially, narrowing in on a more specific population could rescue clinical trials that would otherwise be viewed as negative.

Highlights.

• Elevated whole blood serotonin is a well-replicated biomarker found in 25% of children with autism spectrum disorder.

• In addition to a well-defined role in adults, serotonin modulates neurodevelopment, including sensory development.

• The serotonin transporter gene is associated with whole blood serotonin levels and with autism risk but only in males.

• A serotonin transporter knock-in mouse shows elevated blood serotonin levels and altered social and repetitive behavior.

• The serotonin system is a prime candidate for treatment development in children with this heritable biomarker.