Background
Serotonin (5-hydroxytryptamine (5-HT)) is a prominent monoamine neurotransmitter in the central and peripheral nervous systems, influencing mood, aggression, anxiety, impulsivity, and other behaviors. Serotonergic signaling has been implicated in multiple neuropsychiatric phenotypes, including major depression, obsessive-compulsive disorder (OCD), anxiety disorders, and autism spectrum disorder (ASD), among others (reviewed in ref. [
1]). The presynaptic, antidepressant-sensitive 5-HT transporter (SERT; gene symbol:
SLC6A4) is a critical regulator of 5-HT signaling by modulating synaptic 5-HT levels via presynaptic, Na
+/Cl
- dependent re-uptake. Given the significance of SERT in regulating 5-HT function and its targeting by widely used medications,
SLC6A4 has been an attractive target for genetic studies in neuropsychiatric disorders. A significant focus of
SLC6A4 genetic studies is a common insertion/deletion polymorphism (5-HTTLPR) in the promoter region reported to impact
SLC6A4 gene expression [
2]. Although some evidence supports association of 5-HTTLPR with psychiatric phenotypes including ASD, results overall are mixed [
3‐
5], potentially influenced by the inherent difficulty of diagnosing behaviorally defined disorders and heterogeneity within and across cohorts examined. Additionally, single nucleotide variants (SNPs) within 5-HTTLPR and the promoter region are inherent confounds to many earlier studies [
6‐
8].
The phenomenon of hyperserotonemia, or elevated whole blood or platelet 5-HT seen in approximately 35% of ASD cases, is the oldest ASD biomarker and is a highly heritable trait [
9]. The presence of SERT on the platelet surface and its role in acquiring 5-HT from the blood provides a plausible biological mechanism for SERT involvement in hyperserotonemia [
10‐
12]. Genetic association related to
SLC6A4 gene expression [
13], as well as an interaction of
SLC6A4 with the gene encoding integrin β3 (
ITGB3), which physically interacts with SERT, supports this idea [
14‐
16].
Following observations of significant genetic linkage at 17q11.2 (harboring
SLC6A4) in multiplex ASD families [
17‐
19], we screened exons of
SLC6A4 specifically in families contributing to the observed linkage and found multiple, novel coding variants (Ile425Leu, Phe465Leu, and Leu550Val) and an elevated frequency of a previously documented coding variant (Gly56Ala) to a degree that profoundly deviated from expectations under Hardy-Weinberg equilibrium [
17]. Further support for a role of these variants in ASD comes from studies reporting an Ile425Val variant that segregated in pedigrees harboring multiple psychiatric phenotypes, with Asperger syndrome (an ASD), OCD, and other anxiety disorders being the most prominent [
20‐
22]. Functional characterization of these SERT variants revealed that each elevated 5-HT transport function, as well as altered protein kinase G (PKG) and p38 mitogen activated protein kinase (MAPK) regulation [
23,
24]. Our characterization of one of these variants (Gly56Ala) in knock-in transgenic mice revealed elevated 5-HT clearance and p38 MAPK-dependent transporter hyperphosphorylation
in vivo accompanied by deficits in the three classical behavioral domains associated with ASD [
25]. Collectively, these results suggest that altered 5-HT signaling, and SERT activity and/or regulation represents an important biological endpoint for the functional impact of genetic variation at other genes contributing to SERT regulation and ASD risk.
Modulation of synaptic 5-HT is a dynamic and tightly controlled process, subject to influence through multiple signaling pathways and interacting proteins that act on SERT (reviewed in [
26]). Enhanced SERT activity can be achieved via PKG and p38 MAPK signaling pathways acting through trafficking-dependent and trafficking-independent (that is, functional modulation) mechanisms. A trigger for both of these uptake-enhancing pathways, and the focus of this paper, is activation of the A3 adenosine receptor (A3AR; gene symbol:
ADORA3), a G-protein-coupled receptor (GPCR) that is expressed by 5-HT synthesizing neurons at synaptic terminals [
27,
28]. A3ARs physically interact and influence SERT [
29] through a Gq-linked stimulation of guanyl cyclase (GC)-mediated cGMP synthesis. cGMP activation of PKG elevates SERT surface expression and, in parallel, a p38 MAPK-dependent elevation of surface resident SERT proteins [
26,
28,
30]. Importantly, A3AR agonist stimulation of SERT is lost in A3AR knockout mice [
30], providing evidence for the specificity of the current tools used to study receptor/transporter coupling.
Given a role of A3ARs in SERT regulation, we targeted ADORA3 as a candidate locus to determine whether rare or common variants at this locus are correlated with ASD risk and/or with altered A3AR-mediated SERT function or regulation. To test for common allele effects on ASD risk, single nucleotide polymorphisms (SNPs) that index common haplotypes at ADORA3 were assessed using family-based association methods. The alternative rare variant hypothesis was tested by Sanger sequencing of ADORA3 exons in a sample of ASD probands and ethnically matched controls and followed by a replication analysis using data from whole exome sequencing of independent ASD cases and controls. Nonsynonymous variants identified as being increased in cases from these studies were evaluated functionally through heterologous expression of wildtype and variant A3ARs to test for changes in basal and agonist-activated modulation of SERT activity.
Discussion
Based upon our knowledge that SERT and 5-HT have a longstanding connection to ASD, that A3AR plays a key role in SERT regulation via PKG and p38 MAPK signaling pathways, and that SERT is an essential regulator of 5-HT signaling, we screened
ADORA3 for SERT-altering and ASD-associated alleles. Our experiments were premised on a specific hypothesis: functional risk variants at
ADORA3 would lead to a signaling-mediated elevation of SERT-dependent 5-HT uptake activity, phenocopying
in vitro the elevated function seen with rare, ASD-associated SERT coding variants [
17,
20,
23]. Regarding association of ASD with SERT variants, we note that sequencing of
SLC6A4 by another group testing unrelated ASD probands from singleton or multiplex families (and controls) but without obligate allele sharing (linkage) at 17q11.2 did not show a similar pattern of enriched coding variants, and thus association with ASD risk [
60]. We believe therefore that screening of multiplex probands with allele sharing in this region was an important factor in our initial discovery of novel, functional variants, given the extreme genetic heterogeneity underlying ASD. Moreover, results from genotyping of ASD and related cohorts by Delorme and colleagues [
20,
21] further support ASD association initially suggested by Ozaki et al. [
20,
21] and reinforced by our own studies.
Consistent with results from large-scale GWA scans for common allele susceptibility effects [
35,
43], we found no evidence to support a main effect on ASD association attributable to common variants at
ADORA3. The current study is limited in power to detect alleles of small effect size (for example, OR <1.3), so we cannot exclude the possibility that common
ADORA3 variants confer very small effect sizes or interact with alleles at other genes to confer risk.
Mechanistically, we anticipated that a more likely scenario was for
ADORA3 to harbor coding variants, likely rare, that might impact A3 function directly and SERT function indirectly. Numerous, recent studies focused on CNV or sequencing [
44,
50,
58,
61‐
64] have documented that rare variation affecting a large number of genes is collectively a major source of genetic liability in autism. Our results are consistent with
ADORA3 being one such gene that contributes to ASD liability in rare cases. Here, we present the identification of novel coding variants in
ADORA3 in a Sanger sequencing-based screen of cases and non-clinical comparison samples ethnically matched to the case sample. This screen showed a statistically significant increase in coding variants in cases
vs. controls (
P=0.013). Subsequent availability and analysis of exome sequence data from cases and clinically-screened controls [
50] showed a greater number of ‘functional’ variants at
ADORA3 in cases compared with controls (15/524
vs. 4/644), but a difference that was not significant (
P=0.07). Nevertheless, combining Sanger discovery and newly available exome sequence data strengthens evidence for association, even after Bonferroni correction for multiple comparisons (Fisher’s exact 1-tailed
P=0.0025; OR = 4.72, CI: 1.56-14.30).
Encouraged by the initial discovery of Leu90Val and Val171Ile, functional studies focused on these two rare non-synonymous variants identified from the ASD discovery sample. To test our central SERT regulatory network hypothesis, analysis of Leu90Val and Val171Ile examined the effect of receptor stimulation to induce cGMP synthesis via G-protein coupling and on downstream SERT-dependent 5-HT uptake. IB-MECA stimulation of the Val90-encoded A3AR showed enhanced cGMP synthesis compared with wildtype A3AR under basal conditions, and enhanced cGMP levels extended over the full time course ending with a return to the elevated baseline. These results mirrored our findings that the IB-MECA induced increase in SERT-dependent 5-HT uptake activity across the experimental time course including a delayed return to baseline in co-transfections with the Leu90Val mutant A3 receptor compared with wildtype A3AR. Thus, overall 5-HT uptake is significantly increased in cells expressing the mutant A3 receptor. Increases in both cGMP and 5-HT uptake activity parallels our recent report of enhanced Leu90Val-A3AR/SERT complex formation [
29], which implies (1) enhanced basal receptor-G protein coupling, (2) reduced receptor desensitization, and/or possible differences in the binding kinetics of IB-MECA to Val90-A3AR as a consequence of its location proximal to the A3AR ligand binding site upon agonist stimulation. Taken together, our results are consistent with our prior expectation, based on increased 5-HT uptake caused by rare autism-associated coding mutations in SERT [
17,
20,
23]. In particular, the Gly56Ala SERT variant, exhibited increased catalytic activity as would be achieved through stimulation by PKG and p38 MAPK signaling, which represent the primary pathways for A3AR-mediated upregulation of SERT.
In contrast to Leu90Val, the Val171Ile variant rendered A3AR insensitive to the selective IB-MECA agonist to induce increased cGMP synthesis and a downstream increase in SERT-dependent 5-HT uptake activity. The molecular mechanism underlying this effect is not yet clear, however, we postulate that the proximity of the Val171 residue to the ligand binding pocket may prevent or hinder the ability of the adenosine analog IB-MECA to bind A3AR, resulting in a more rapid dissociation and/or less efficient (or absence of) receptor-G protein coupling. Additional experiments will be required to fully elucidate the molecular mechanisms of these two coding variants on receptor function. Although the functional impact on SERT of the Leu90Val and Val171Ile variants are in opposite directions, it is possible that both elevated and diminished capacity for regulation of SERT through A3AR pathways can impact 5-HT clearance in a manner that disrupts 5-HT’s ability to coordinate brain development [
65,
66] and/or adult 5-HT signaling [
67,
68].
We initially conceptualized dysregulation of a SERT regulatory network based not only on the molecular impact ASD-associated SERT mutations, rather within a broader context that implicates disruptions in 5-HT signaling in autism. Hyperserotonemia in 35% of ASD cases and efficacy of selective 5-HT re-uptake inhibitors (for example, fluoxetine, citalopram) and atypical antipsychotics (for example, risperidone) in ameliorating irritability and other anxiety-related problems in ASD are just two of many themes that implicate 5-HT dysregulation in ASD. We previously demonstrated a proof-of-principal for genetic variation in SERT-binding and regulatory proteins being associated with ASD. Here we refer to the common Leu33Pro variant in
ITGB3, which has been statistically associated with both elevated 5-HT levels in blood [
9] and ASD risk [
69], and which causes allele-dependent effects on SERT activity and regulation [
70]. We recently observed hyperserotonemia in a knock-in mouse model of the Gly56Ala SERT variant, along with p38 MAPK-dependent hyperphosphorylation of SERT, increased SERT-dependent 5-HT clearance and 5-HT receptor hypersensitivity
in vivo, as well as social behavior impairments, repetitive behaviors, and deficits in communication [
25]. Nonetheless, we recognize that statistical association of rare
ADORA3 variants (especially Leu90Val) requires further validation in larger samples. This is the case for rare variants in any specific locus implicated in ASD, and has led many investigators to emphasize the network as a better substrate to elucidate the underlying mechanisms. We recognize that the magnitude of ASD risk conferred by these variants is unknown. Indeed, the single male case harboring an Ala69Ser variant (of unknown functional effect) at
ADORA3 also possessed a
de novo duplication of an X-linked interval including
RPL10, a gene for which inherited and
de novo point mutations and gene-disrupting and/or CNV deletion has been associated with ASD and ID [
71]. Duplication effects here are unknown, but this variant is likely to confer risk. Nevertheless, we believe these studies will add to the growing body of data implicating specific gene/protein networks in contributing to ASD liability. Taken together our studies support the case for 5-HT and more specifically SERT regulatory pathways as one gene/protein network in which perturbations contribute to the underlying pathophysiology of ASD. Further studies within this network may provide new leads to ASD therapeutics.
There are a few limitations or caveats of the studies we present. First, while the discovery sequence sample was ethnically matched, subjects were not matched based on genome-wide genotype data. It is possible therefore, that subtle population stratification effects could lead to inflation of the observed increase in numbers of rare, ‘functional’ variants in cases
vs. controls. Given that the case and control samples from the AASC were pair-matched based on genotype data, and the greater number of functional variants in those cases (15/524
vs. 4/644) was not statistically significant in a Fisher’s Exact test, we recognize that both discovery (
P=0.0143) and combined (
P=0.0025) evidence for gene-based association of rare, ‘functional’ variants should be interpreted with caution. Second, while multiple comparisons were conducted in functional experiments, we are reassured that the increases in cGMP production and 5-HT uptake parallel one another, supporting our conclusions regarding the ability of Leu90Val A3AR to augment SERT-dependent 5-HT uptake over time. Finally, we note that the functional experiments were conducted in CHO cells, and may not reflect the function during or after development
in vivo. While this is certainly possible, our study of multiple variants in SERT and other proteins that influence SERT regulation (for example, ITGB3) present a consistent picture of results from
in vitro transfection-based experiments ultimately relating very well to effects on SERT function in mouse models harboring these variants [
70,
72,
73].
Acknowledgements
This work was supported by grants NS049261, MH061009 and MH084962 to JSS and MH094527, MH078028 and MH096972 to RDB, the NIH NCRR grant UL1 RR024975 for the Vanderbilt CTSA, and P30 HD15052 to the Vanderbilt Kennedy Center for Research on Human Development. This work would not have been possible without the DNA Resources Core, invaluable assistance by Melissa Potter for Data Coordination and statistical genetics support from Lan Jiang, both in the Computational Genomics Core at Vanderbilt. In particular, we recognize important technical contributions to this work by Emily L. Crawford and Anna G. McGrew.
The authors gratefully acknowledge investigators in the NIH ARRA Autism Sequencing Consortium: Eric Boerwinkle (University of Texas Health Science Center), Joseph D. Buxbaum (Mount Sinai School of Medicine), Edwin H. Cook, Jr. (University of Illinois at Chicago), Mark J. Daly (Communicating PI; Massachusetts General Hospital, Harvard Medical School and Broad Institute of Harvard and MIT), Bernie Devlin (University of Pittsburgh Medical School), Richard Gibbs (Baylor College of Medicine), Elaine T. Lim (Broad Institute of Harvard and MIT), Benjamin M. Neale (Massachusetts General Hospital, Harvard Medical School and Broad Institute of Harvard and MIT), Kathryn Roeder (Carnegie Mellon University), Aniko Sabo (Baylor College of Medicine), Gerard D. Schellenberg (University of Pennsylvania), Christine Stevens (Massachusetts General Hospital, Harvard Medical School, Broad Institute of Harvard and MIT) and James S. Sutcliffe (Vanderbilt University).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
NGC, RDB, and JSS designed the study. NGC performed genetic studies including sequencing, genotyping, and exome replication analysis. NGC, CBZ, KML, and WAH performed
in vitro functional assays. BLY and EKG preformed ancestry determination. CGT compared the A3AR with A2aAR and generated Figure
2. ARRA Sequencing Consortium performed exome sequencing and variant calls. NGC, RDB, and JSS prepared the manuscript. All authors read and approved the final manuscript.