The serotonin system is implicated in autism
Elevated whole blood serotonin (5-hydroxytryptamine, 5-HT) was first reported in autism by the lab of Daniel X. Freedman in 1961 [1
], well before 5-HT was fully accepted as a neurotransmitter [2
]. Serotonin was discovered as a serum factor leading to vasoconstriction in 1948 by Maurice Rapport and colleagues [3
]. This work converged with that of Vittorio Erspaner, who identified in 1952 that 5-HT was the agent secreted by enterochromaffin cells previously designated as ‘enteramine’ [4
]. Subsequent work demonstrated that 5-HT synthesized in enterochromaffin cells is released into the enteric circulation and taken up by platelets, which contain more than 99% of the 5-HT in peripheral blood [5
]. Due to the vastly greater amount of 5-HT contained in platelets, it is challenging to measure free 5-HT in plasma, and it is unclear if plasma 5-HT levels are changed in autism [6
]. Thus, hyperserotonemia appears to arise from elevated platelet stores of 5-HT.
Nearly fifty years have passed since the initial description of whole blood 5-HT as a biomarker in autism, described sixteen years earlier by Leo Kanner [8
]. Subsequent studies have repeatedly demonstrated that more than one-quarter of autistic children are found to have elevated whole blood 5-HT levels [9
], and, indeed, hyperserotonemia is arguably the most highly reproduced biomarker linked to a neurobehavioral disorder. A recent study in a more homogeneous Dutch population suggests that whole blood 5-HT may actually show a bimodal distribution in autism [10
]. Despite the long-standing observation of hyperserotonemia in autism, our understanding of its origin and relationship to the etiology of autism remains incomplete.
Functional studies of the platelet serotonin system in autism have evaluated both the serotonin transporter and serotonin receptors. The 5-HT transporter (5HTT, SERT) is essential for the uptake of the biogenic amine into the platelet. Some but not all studies indicate alterations in platelet 5-HT uptake as well as SERT binding levels in autism [11
]. Some studies have also shown decreased platelet serotonin receptor 5-HT2
], which could possibly influence platelet 5-HT levels by modulation of platelet activation and release. Of note, recent work from our laboratory indicates that the 5-HT2A
receptor can modulate SERT associations with the platelet membrane skeleton [23
]. Inconsistency in transporter and receptor studies in autism likely reflects the complex regulatory systems that control platelet 5-HT utilization as well as platelet activation. Certainly, an important consideration is the regulation of gut production of 5-HT and its regulation by intrinsic signaling and transport mechanisms [24
The robustness of hyperserotonemia as an autism biomarker has helped focus research on the genetic control of platelet 5-HT levels, with the belief that common regulatory mechanisms in the platelet and the central nervous system could reveal genes important to autism susceptibility. Indeed, one small study found higher whole blood 5-HT levels in children with ASD who had an affected sibling than in children with ASD but no affected siblings [29
]. In the Hutterites, a large founder population not ascertained for brain disorders, whole blood 5-HT levels were found to be under nearly complete genetic control, with a heritability of 0.99 [30
]. In a small twin study, the maximum rate of 5-HT uptake was also found to be highly heritable [31
]. The Cook lab has used linkage and association to map quantitative trait loci for whole blood 5-HT levels in the Hutterites [32
]. The initial study found significant association at a functional polymorphism in the integrin β3 subunit gene (ITGB3
) as well as suggestive association at the Vitamin D receptor gene [32
]. Follow-up analyses revealed that ITGB3
, as well as the 5-HT transporter gene (SLC6A4
), were primarily associated with whole blood 5-HT levels in males [33
]. Carneiro and colleagues subsequently demonstrated physical and functional interactions between the ITGB3 and SERT proteins that regulates both platelet 5-HT uptake and aggregation [26
]. Consistent with these findings, three studies have now reported a gene-gene interaction between ITGB3
in association with autism [34
], supporting the idea that whole blood 5-HT represents a genetically-determined, intermediate phenotype in autism that can be mined for elucidation of disease susceptibility genes.
Recent data support the idea that the 5-HT system is also altered in the brain in autism. If the peripheral findings translate into the brain, increased 5-HT uptake or storage in the presynaptic terminal would be expected to diminish extracellular 5-HT availability. Unfortunately, a direct assessment of synaptic 5-HT in patients with autism is not possible with present technologies. A surrogate measure is the assessment of brain 5-HT synthesis, and indeed this measure demonstrates an altered developmental pattern in autism [37
]. Other studies have evaluated 5-HT receptor binding. Two neuroimaging studies have found decreased 5-HT2
receptor binding, one a SPECT study in adults with Asperger’s syndrome [38
] and the second a PET study in parents of children with autism [39
]. Blatt’s group has also presented data demonstrating decreased 5-HT2A
binding in post-mortem samples from autism subjects [40
]. The overall consistent findings of decreased 5-HT2
receptor binding in platelet, neuroimaging, and post-mortem studies supports the hypothesis that peripheral alterations in 5-HT systems contribute to the behavioral abnormalities associated with autism.
Finally, limited pharmacology studies underscore a role for the 5-HT system in autism. Dietary depletion of the 5-HT precursor tryptophan, which is expected to cause decreased synaptic 5-HT, leads to worsened repetitive behaviors and irritability in autism [41
]. Consistent with these findings, several studies have reported that 5-HT reuptake inhibitors can relieve symptoms of irritability and rigid-compulsive behavior in autism [42
]; more recent data from clinical trials are less supportive, but the wide spectrum of autism traits complicates selection of appropriate target symptoms in these trials. More consistent data supports the use of risperidone [44
], an atypical antipsychotic with antagonism at multiple monoamine receptors, including the 5-HT2A
receptor. It is striking that this agent has clinical utility in autism despite findings of reductions in 5-HT2A
receptor levels. This discrepancy may be an important clue that a primary effect impacting 5-HT2A
receptor sensitivity in these subjects may exist which may then trigger compensatory receptor down-regulation.
Rare variants offer one way to approach complexity at the SERT gene
Multiple lines of evidence favor involvement of the 5-HT transporter gene (SLC6A4
) in autism; although the cumulative data are not straightforward or uniform in their support, which is perhaps not surprising in a complex behavioral disorder with a polygenic origin. Most of this evidence has emerged from candidate gene studies oriented around the platelet 5-HT system. Importantly, linkage studies have also implicated the chromosome 17q11-q21 region containing SLC6A4
]. Follow-up linkage studies [48
] have detected more significant evidence for linkage in this region when considering families with two or more affected male children and no affected females. These findings suggest that SERT might harbor sex-specific risk alleles for autism and possibly contribute to, or be responsive to, the gender bias of autism. Such an idea is in line with the findings that variation in SLC6A4
is primarily associated with whole blood 5-HT levels in males [33
]. It also matches accumulating evidence in the 5-HT transporter knockout mouse that altered expression of SLC6A4
has sexually dimorphic effects on brain 5-HT systems and resulting behavior (reviewed in [51
With the accumulated evidence of alterations in the platelet 5-HT system in autism, it is not surprising that some of the earliest candidate gene association studies in autism focused on the SERT gene. Unfortunately, studies of SLC6A4
in autism have been complicated by the presence of multiple common functional variants in gene regulatory regions. The best-studied of these variants, the promoter length polymorphic repeat (5-HTTLPR) [52
], was initially reported as associated with autism in haplotype analyses by the Cook lab [53
]. Subsequent studies have yielded inconsistent findings. Indeed, although most studies have reported association with the short allele, which leads to less gene transcription, others have reported association with the long allele (reviewed in a recent meta-analysis [54
]). Genotyping errors could possibly account for some of this inconsistency given deviation from Hardy-Weinberg equilibrium in some samples and discrepancies in genotypes of the same samples across different laboratories [55
Other studies have explored the possible involvement of SLC6A4
in autism by expanding the range of polymorphisms studied. Indeed, multiple, additional functional polymorphisms have been discovered within the SLC6A4
gene regulatory regions [56
], including a single nucleotide polymorphism (SNP) that reduces the functional impact of the 5-HTTLPR long allele (La, Lg) [57
]. To date, these polymorphisms have not been systematically studied in large autism samples. Other SLC6A4
polymorphisms and haplotypes that have been reported to associate with autism have not been fully characterized for either their own functional significance or linkage disequilibrium with other possible functional variants [59
]. Indeed, one haplotype in linkage disequilibrium with the 5-HTTLPR short allele, which leads to decreased SERT expression, has actually been associated with increased platelet 5-HT uptake [61
]. These findings signify a still incomplete understanding of SERT genetic regulation. A comprehensive analysis focusing on the impact of multiple alleles on predicted SERT expression or function should greatly enhance our understanding of the SERT-serotonin-autism connection.
Another fruitful approach to understanding the possible involvement of SLC6A4
in autism susceptibility is to focus on either specific symptom clusters or on intermediate phenotypes. Four studies have reported association at various SLC6A4
polymorphisms when performing sub-analyses of repetitive behavior in autism [49
]. One of these studies also reported association of the 5-HTTLPR short allele with poor non-verbal communication [63
]. Two studies have reported association of various SLC6A4
polymorphisms with either whole blood 5-HT [64
] or with 5-HT uptake in autism [61
]. Finally, one study reported association of the 5-HTTLPR short allele with increased cortical gray matter in two- to four-year-old children with autism [65
]. These collected studies of subphenotypes or intermediate phenotypes add strength to the hypothesis that SLC6A4
represents a susceptibility gene for subgroups of children with autism.
One intriguing possibility to emerge from the data on common SLC6A4
variants is that optimal SERT activity is likely to be highly constrained, with both high and low SERT activity impacting autism susceptibility, though likely with different patterns of symptoms and intermediate phenotypes. Such an idea parallels ideas emerging from recent findings at the MECP2
gene implicated in Rett syndrome [66
]. Additionally, this idea fits with the observation that genetic variation can lead to increased SERT function in association with repetitive behavior in autism [32
], as well as with OCD [56
]; whereas genetic variation that supports decreased SERT function is associated with impaired social behavior [63
] and increased cortical size in young children with autism [65
], as well as in a mouse model of PTEN
-associated autism [68
Association of human SERT gene variants with autism traits
Despite intriguing findings of both common and rare functional SLC6A4
variants, the collective evidence does not come close to explaining the strong linkage findings in male sibling pairs in this chromosomal region [48
]. Indeed, the collected evidence suggests that no single, common functional polymorphism at this locus underlies autism susceptibility, at least when considered individually. Recently, the Blakely and Sutcliffe labs explored the alternative hypothesis that multiple, rare functional variants account for some of the findings of linkage in this region. In a comprehensive analysis of SERT noncoding and coding exons from 120 families that contributed most to linkage at 17q11.2, our groups found 5 coding substitutions and 15 variants in 5´ noncoding and intronic regions transmitted in families [49
]. In the aggregate, these variants showed significant association to autism and tracked specifically with rigid-compulsive traits.
In the initial study, one of the autism-associated variants, Gly56Ala, showed constitutively elevated SERT activity that was not further up-regulated through the PKG and p38 MAPK pathways in transformed lymphoblasts from autism probands [49
]. This polymorphism, at an amino acid that is conserved across all mammalian SERTs, was the only SERT variant frequent enough to attempt independent statistical analyses, which revealed that the Ala56 allele: 1) was overtransmitted 3:1 to subjects with autism (P
0.025), 2) displayed a gender-bias (males) when considering affected subjects within linked pedigrees (P
0.017), and 3) was associated with Autism Diagnostic Interview component factors corresponding to rigid-compulsive traits (P
0.0085) and sensory aversion (P
0.0005). The SERT amino-terminus where Gly56 resides is an important regulatory area [69
]. These considerations make the Gly56Ala allele an important allele to model for system level effects in an animal model (see below).
Two of the SERT coding variants found in autism subjects in our studies, Gly56Ala and Lys605Asn, were both shown to exhibit a gain of function despite normal surface expression [72
]. Our autism studies uncovered three additional SERT coding variants: Ile425Leu, Phe465Leu and Leu550Val. All three variants occur within transmembrane domains at amino acids that are highly conserved [49
]. Rigid-compulsive behaviors were associated with these three novel SERT variants when analyzed as a group (P
]. When all linked families with individuals carrying a coding variant were considered together, the significance in elevated severity of rigid-compulsive behaviors increased (P
Even these rare variants are unlikely to present a simple story given the presence in the SERT gene of multiple common functional variants that can affect gene expression. It is entirely possible that the functional effects of these rare variants, particularly the more subtle increased activity and dysregulation of Gly56Ala, could be partially offset in some individuals by common variants leading to decreased SERT expression or variation in other genes within 5-HT signaling networks. Indeed, one subsequent study did not find association of rare SERT variants with autism or rigid-compulsive behavior in families without evidence for linkage to this region [73
]. Another study did not find association of the Ala56 allele when considered alone in OCD [74
]; although it is quite possible that these seemingly related disorders arise from distinct origins. Regardless, the discovery of functional amino acid variants in SERT offers a critical opportunity to translate altered SERT activity into an animal model for comprehensive assessment of genotype/phenotype assessments beyond what is feasible in humans.
In vitro analysis of autism-associated SERT coding variants
Of the five coding variants that we identified in probands from multiplex autism families, two of these, Gly56Ala and Lys605Asn, were actually first characterized, along with eight other rare SERT variants, independent of their potential roles in autism [72
]. In comparison to the reference hSERT cDNA (often termed “wild-type”) originally identified by Ramamoorthy and colleagues [76
], both Ala56 and Asn605 demonstrated elevated 5-HT transport activity in transfected HeLa cells [72
]. This gain-of-function transport property was shared (and even exceeded) by several other SERT coding variants including Ile425Val, a variant reported by Ozaki and colleagues in association with two pedigrees bearing a complex neuropsychiatric phenotype including OCD, autism spectrum disorder, social phobia, and alcohol abuse [77
]. Since these studies, additional cases of Val425 in OCD have been reported [78
]. Clearly, the identification of two mutations at the same site (Val425 vs Leu425) in disorders with trait overlap increases the probability that both are functional and relevant to disease status.
Recently, we described the functional properties of Leu425, Leu465 and Val550, expressed in HeLa cells in parallel with Ala56 and wildtype human SERT [79
]. These studies (Fig. ) revealed for the first time that all autism-associated SERT coding variants confer a gain-of-function phenotype
. Kinetic studies provided the first evidence, however, that the elevated SERT activities of these variants arise from distinct mechanisms. Thus, Ala56 displays a reduction in 5-HT KM
, consistent with a shift to a high-affinity state for 5-HT, whereas the other three variants display elevated VMAX
and no change in KM
(Table ). VMAX
increases in transport kinetic studies are typically associated with changes in the number of transporter molecules expressed at the cell surface, though other explanations are possible. Interestingly, Kilic and colleagues reported that Ile425Val exhibits a KM
shift with no change in VMAX
. As we have discussed elsewhere [71
], the Kilic studies utilized an overexpression system (COS cells), whereas we expressed SERT constructs at levels comparable to that seen in natively expressing cell lines (e.g. RBL-2H3, RN46A). Possibly, the high expression levels of the COS overexpression system preclude detection of Vmax changes whereas a trafficking impact in our HeLa model might somehow mask KM
effects, though they are detectible for Gly56Ala and Lys605Asn.
Fig. 1 Location and 5-HT transport activity of autism-associated SERT coding variants. A) Autism-associated variants are overlayed on a 12 TM model of a single SERT subunit, with NH2 and COOH termini oriented inside the cell. Variants in extramembrane domains (more ...)
Rare serotonin transporter coding variants identified in autism and autism spectrum (Jeremy)
In part due to the discrepancies with kinetic analyses noted above, we performed surface binding and biotinylation studies on all autism-associated SERT variants [79
]. These studies reveal that all variants exhibit comparable total [125
I]RTI-55 binding, compatible with evidence of equivalent protein expression as detected by Western blots (Fig. ). In contrast, differences in [125
I]RTI-55 surface expression were evident between Gly56Ala and the other SERT coding variants, findings verified in surface protein biotinylation studies. Whereas surface Gly56Ala levels were equivalent to, or lower than that found for wildtype SERT, surface expression of Leu425, Leu465 and Val550 were significantly elevated. These data indicate that the increase in uptake activity for Ala56 in transiently-transfected HeLa cells comes about from catalytic activation, whereas the increases in uptake for Leu425, Leu465 and Val550 arise from enhanced surface expression. Whether the latter effects arise from elevated insertion of transporters, reduced endocytosis or enhanced recycling after endocytosis is not known but presently under evaluation.
Fig. 2 Analysis of protein expression of autism-associated SERT coding variants. Impact of SERT coding variants on total and cell surface [125I]RTI-55 binding. HeLa cells transiently transfected with hSERT or one of the SERT coding variants were subjected to (more ...)
A striking feature of the Ala56 variant (and Asn605) observed in our initial characterization of deposited SERT variants [72
] is its lack of responsiveness to two protein kinases previously shown to influence SERT surface expression [80
] (protein kinase G, PKG) and catalytic activation [80
] (p38 mitogen activated protein kinase, p38 MAPK). These observations compelled us to evaluate all autism-associated SERT coding variants for their ability to support acute regulation by these kinases. As before, Ala56 was insensitive to activators of either PKG or p38 MAPK (Fig. ). In contrast, stimulation of these kinases produced significant increases in activity for Leu425, Leu465 and Val550, though slightly lower than that observed for wildtype SERT. Regardless, these observations support the presence of two distinct classes of functional perturbations for these variants. Similar distinctions were also evident with protein kinase C regulation, where enhanced sensitivity is evident for Ala56 but not for Leu425, Leu465 and Val550. The Ser/Thr protein phosphatase 2A (PP2A) has been implicated in SERT regulation and forms a reversible physical complex with the transporter [81
]. Although PKC, PKG and p38 MAPK activators distinguish among SERT variants, PP2A antagonist sensitivity appears to be lost for both Ala56 and Leu425, two variants that represent the two distinct classes of regulatory defects observed in SERT variants. These findings indicate that altered PP2A interactions or SERT dephosphorylation may be a more fundamental alteration that extends across all variants, with secondary effects linked to the physical site of mutations. Additional studies are needed to localize the site(s) of interaction of PP2A with SERT, as well as the sites of SERT phosphorylation relevant to PP2A action, further clarifying the mechanism of loss of PP2A sensitivity.
Fig. 3 Impact of 8BrcGMP and p38 MAPK on SERT activity of autism-associated hSERT coding variants. A) Activity modulation. HeLa cells transfected with hSERT or autism-associated hSERT coding variants were examined for 5-HT transport activities following pretreatments (more ...)
The studies reviewed above for altered SERT regulation in autism-associated coding variants derive from transiently-transfected cells. In an attempt to generate a stable model for SERT variants, where expression is driven from the same genomic locus, we produced stable cell lines with the wildtype SERT, Ala56 and Leu425 using the CHO-Flp-In model. Surprisingly, 5-HT transport assays and Western blots of CHO cell extracts revealed a significant reduction in SERT activity and expression for both SERT Ala56Ala and Leu425 compared to hSERT (Fig. ). Our current hypothesis for this effect is that it is a result of growth suppression and cell death that arises with elevated 5-HT accumulation in cells that do not have a vesicular storage compartment for the amine. Although this finding may appear to be a symptom of heterologous expression, we recognize that SERT is expressed in contexts without such vesicular compartments, raising the possibility that alterations in cell growth and viability arise as a consequence of heightened SERT expression and contribute in unexpected ways to the complex phenotypes observed in autism. Despite these basal effects on expression, we were able to use this model to investigate changes in catalytic rates. In the CHO-Flp-In model, when functional activity is normalized by surface expression, a significant catalytic activation is evident for both Gly56Ala and Ile425Leu. Together, these findings suggest that heterologous expression systems provide important, but likely only partial windows on the impact of autism-associated SERT variants.
Fig. 4 Altered activity of SERT variants is differently exhibited in CHO-Flip-In™ stable cells. A) 5-HT transport activity of autism-associated SERT coding variants in CHO-Flip-In™ stable cells. All variants were assayed for 5-HT transport activity. (more ...)
Ideally, one would like to evaluate the impact of SERT coding variants on the brain of autism subjects. While we hope to pursue this need with transgenic mice, it seems unlikely that such analyses in humans will be available in the near future. It is well known, however, that SERT is expressed in transformed lymphoblasts [52
], the very same models used to archive DNA from subjects recruited for genetic studies. We took advantage of this opportunity to examine basal transport activity and regulation in cells derived from autism subjects. Indeed, Ala56 [79
], as well as the other autism-associated SERT variants, display elevated 5-HT transport activity (Fig. ). As a control in the latter studies, L-glutamate transport was found to be equivalent to that in wildtype SERT, consistent with alterations derived from a specific influence on SERT expression and/or regulation. Interestingly, although transformed lymphoblasts do not express a vesicular storage mechanism for 5-HT (to our knowledge), SERT expression is low in these cells relative to platelet and raphe neurons, such that the growth perturbations noted above for stable CHO-Flp-In cells do not appear to reduce our ability to detect 5-HT transport enhancement. Studies with multiple lymphoblast lines expressing the Ala56 variant [49
] also reveal that activators of PKG and p38 MAPK display the loss of kinase sensitivity seen with transfected HeLa cells. Together, these findings reinforce the properties of autism-associated SERT variants as gain-of-function mutants that also display complex regulatory deficits.
In vivo manipulations of the mouse SERT gene
A number of methods are available to manipulate genes in the mouse, including gene knockout, regional or inducible gene knockout, and transgenic overexpression, including inducible and reversible systems. The first knockout of the mouse SERT gene was performed by deletion of exon 2 in the labs of Dennis Murphy and Klaus-Peter Lesch [82
]. Over two hundred papers have now been published describing this mouse’s phenotype, which includes increased anxiety-like behavior, increased stress reactivity, decreased aggression, obesity, and disruption of brain architecture in the sensory cortex and amygdala (reviewed in [51
]). 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 Crawley test [83
]. Interestingly, when bred with mice haploinsufficient for Pten
, a putative autism susceptibility gene, SERT +/− knockout mice show an exacerbation of the macrocephaly and decreased sociability characteristic of the Pten
+/− mice [68
]. These findings may be consistent with the idea that SERT activity may be constrained within a certain range, with elevated or diminished SERT function leading to vulnerability to autism-like behavior.
Another genetic manipulation was performed in a mouse line expressing a human SERT transgene generated in the lab of Trevor Sharp [84
]. These mice show a 2- to 3-fold increase in SERT binding sites and a resulting decrease in extracellular 5-HT. Their platelet 5-HT levels have not yet been assessed. Providing a contrast to the SERT knockout mice, these hSERT transgenic overexpressing mice show a decrease in anxiety-like behavior that reverses with acute treatment with a 5-HT reuptake inhibitor. Furthermore, they show increased sensitivity to 5-HT2
receptor stimulation [85
], again in contrast to the SERT KO mice, which show decreased sensitivity [86
]. This mouse could potentially serve as a surrogate model of the increased SERT function hypothesized in autism and OCD, but the use of a human SERT locus introduced using a yeast artificial chromosome construct may lead to alterations in gene regulation and expression patterns, as well as altered protein regulation and function, when introduced into the mouse.
Based upon our belief that manipulation of a native gene was likely to yield a meaningful brain and behavioral read-out of increased SERT function, we chose to introduce a knock-in of the Ala56 allele into the native mouse SERT gene. We chose the Ala56 allele for three reasons: 1) it shows association with rigid-compulsive behavior and sensory aversion in autism; 2) it is the most common amino acid variant of human SERT; 3) it offers an opportunity to probe the importance of altered SERT regulation in vivo.