Genetic background modulates phenotypes of serotonin transporter Ala56 knock-in mice

The serotonin (5-hydroxytryptamine, 5-HT) system has long been implicated in Autism Spectrum Disorder (ASD) [1]. Elevated whole blood 5-HT, termed hyperserotonemia, is found in about 30% of ASD patients [2, 3]. Blood 5-HT levels are more heritable than ASD itself [4‐7], but the link between hyperserotonemia and ASD pathophysiology has remained elusive. In the blood, 5-HT is found almost exclusively in platelets that cannot produce 5-HT, but take it up via the antidepressant-sensitive 5-HT transporter (SERT). Peripheral serotonin is synthesized and released by enterochromaffin cells that release 5-HT into the enteric circulation. Consistent with a primary role in platelet 5-HT uptake, polymorphisms in the SERT gene (SLC6A4) have been associated with human whole blood 5-HT levels; however, this has been observed only in males and not females [8], with a similar pattern observed for the SERT-interacting protein integrin β3 [8‐10]. Several linkage studies have implicated the chromosome 17q region containing SLC6A4 in ASD, also with stronger evidence in males [11‐15].

In contrast to linkage studies, multiple groups have found, at most, modest association between common SLC6A4 variants and ASD [16‐18]. Allelic heterogeneity may account for difficulty in assessing association in this region. Indeed, multiple common functional variants have been identified in the SLC6A4 gene (reviewed in [19]). We identified multiple, rare SERT coding variants in ASD families with evidence of linkage to 17q. We also found that the most common of these variants, SERT Ala56, was significantly associated with rigid-compulsive behavior and sensory aversion. Lending credence to a functional contribution of the SERT variants identified, each of the variants was found to confer increased 5-HT transport activity in cell models relative to wildtype SERT [14, 20, 21]. In our study, the Ala56 variant showed an approximately 2:1 transmission bias to affected versus unaffected children [14]. Importantly, however, some variant carriers were unaffected by ASD, and a case–control study has reported no association of ASD with rare SERT amino acid variants in the absence of linkage [22], indicating that other genetic or non-genetic risk factors likely impact the resulting phenotype.

To assess the functional relevance of the SERT Ala56 variant in vivo, we produced knock-in mice, expressing the variant on a 129S6/S4 inbred background. In these mice, we observed multiple biochemical and behavioral phenotypes that recapitulate some features of ASD, including hyperserotonemia, altered social function in the tube test and three-chamber sociability test, decreased ultrasonic vocalizations, and repetitive climbing/hanging behavior in the home cage [23]. We also observed hypersensitivity to in vivo activation of 5-HT_1A and 5-HT_2A receptors, reasonably attributable to diminished synaptic 5-HT availability and functional upregulation of 5-HT receptors. Unfortunately, this inbred strain background shows a low level of activity in novel environments across multiple behavioral tests [23], which limits the assessment of anxiety-like and social behaviors that are dependent upon spontaneous locomotion.

As 5-HT is a modulator of multiple CNS pathways subserving complex behaviors, we suspected that SERT coding variation could be influenced by genetic variation in other genes, which might account for the variable phenotype observed in humans with the SERT Ala56 allele. To explore this issue, we moved the SERT Ala56 allele to a C57BL/6 (B6) background for two purposes: 1) to examine autism-relevant behavior in a more active and social strain than 129, and 2) to study strain dependence of SERT function, the 5-HT system, and resulting behavior. Previous work identified a functional and behavioral impact of variation at two amino acids that differ between 129 and B6 inbred strains [24]. We therefore also backcrossed the 129 SERT onto the B6 background to allow a comparison with the mutant SERT on this background. Our findings reveal a striking influence of genetic background on the impact of the SERT Ala56 variant, suggesting that further dissection of strain differences could illuminate other genes modulating risk of ASD.

As our previous work identified two sites of functional, coding variation in SERT between 129 and B6 inbred strains (Glu39Gly; Arg152Lys) [24], it was necessary to backcross 129-derived animals expressing wildtype Gly56 SERT in parallel with engineered 129 animals expressing SERT Ala56 onto the B6 background to permit a proper analysis of the unique effects of the Ala56 variant. This effort yielded animals with >99% C57BL/6 strain identity. A female C57BL/6 SERT ER-Gly56 mouse (SERT Glu39/Gly56/Arg152, hereafter described as Gly56) was bred with a male C57BL/6 SERT ER-Ala56 mouse (SERT Glu39/Ala56/Arg152, hereafter described as Ala56) to establish animals heterozygous for the SERT Ala56 variant, permitting the generation of homozygous Gly56/Gly56, heterozygous Gly56/Ala56, and homozygous Ala56/Ala56 littermates used in the experiments reported.

Whereas we found the 129-expressed Ala56 variant to induce no changes in locomotor activity in a novel environment, as assessed in the open field test [23], we detected a small but significant decrease in activity in homozygous male Ala56 animals on the B6 background, both for total ambulatory time (t = 3.45, P = 0.003) and ambulatory distance (t = 3.11, P = 0.006) (Figure 1A, [see Figure S1 in Additional file 1]). In contrast, females exhibited no genotype effects in open field activity [See Figure S2 in Additional file 2; See Figure S3 in Additional file 3].

Results for anxiety-like behavior conflicted across three tests. The ratio of time spent in the center of the open field, compared to total time, over the first five minutes of the test was not different in males (t = 0.97, P = 0.35) or females (t = 0.79, P = 0.44) (Figure 1B, [See Figure S2 in Additional file 2]). Similarly, no difference was seen in the time spent in the open arms of the Elevated Zero Maze over 5 minutes for males (t = 0.79, P = 0.44) or females (t = 1.52, P = 0.15) (Figure 1C, [See Figure S2 in Additional file 2]). In contrast, a genotype effect was observed with SERT Ala56 males spending less time in the light compartment of the light–dark test than homozygous Gly56 littermates (t = 2.51, P = 0.02) (Figure 1D). Again, females displayed no difference in this test (t = 1.55, P = 0.15) [See Figure S2 in Additional file 2].

Results on social testing were also mixed. In the three-chamber sociability test, all three genotypes showed a significant preference for the social stimulus over the non-social stimulus, both when chamber time was recorded (overall stimulus F = 96.6, P <0.0001 for each genotype) and when time <1 cm from the stimulus mouse was recorded (overall stimulus F = 135.1, P <0.0001 for each genotype) (Figure 2A,B). On the preference for social novelty portion of the three-chamber test, the Gly56 animal showed no significant preference for the novel social stimulus [See Figure S1 in Additional file 1], which makes the comparison data uninterpretable. On the tube test, homozygous Ala56 animals showed a significant propensity to back out of the tube when encountering a homozygous Gly56 animal (McNemar’s exact test P <0.0001; n = 126 encounters), as well as when encountering a Ala56/Gly56 animal (P <0.0001; n = 119 encounters) (Figure 2C,D). No difference was detected when heterozygous Ala56/Gly56 encountered Gly56/Gly56 animals (P = 1.0, n = 93 encounters) (Figure 2E). On the resident-intruder test, we observed no difference in latency to attack between genotypes (t = 0.41, P = 0.69) (Figure 2F). Similarly, of the homozygous Ala56 animals, 13/18 showed aggression toward the intruder during the 10-minute test period, as compared to 10/18 homozygous Gly56 animals (Fisher’s exact test P = 0.49).

In our prior study on the 129 background, we found decreased ultrasonic vocalizations (USVs) in homozygous Ala56 pups when separated from the dam at postnatal day 7. In contrast, on the B6 background, homozygous Ala56 pups displayed increased USVs in comparison to wildtype littermate controls (Mann–Whitney U = 368.5, P = 0.017) (Figure 2G). Secondary Kruskal-Wallis comparison of all three genotypes showed a trend for an overall genotype effect (K-W 5.59, P = 0.06).

No changes in repetitive or sensory behavior were induced by the Ala56 variant on the B6 background. Behavior in the home cage did not show any differences by genotype (Genotype x Behavior Interaction F = 0.82, P = 0.60) (Figure 3A). There was also no difference between the genotypes when hanging behavior, which was increased on the 129 background, was considered alone (t = 0.35, P = 0.73) (Figure 3B). Importantly, this behavior was less prevalent on the B6 background in general in comparison to the 129 background [23]. Further, no differences were observed in spontaneous alternations on the Y maze (t = 1.20, P = 0.25) (Figure 3C) nor in tactile sensitivity assessed using reaction to von Frey filaments (t = 0.07, P = 0.95) (Figure 3D).

Since we observed evidence of a genetic background effect on behavioral phenotypes, we examined whether Ala56-induced serotonergic phenotypes were preserved. On the B6 background, we observed no genotype influences on adult male whole blood 5-HT levels (Figure 4A). To decrease variability and maximize sensitivity to detect a differences, we moved to a fluorometric detection method at post-natal day 7 [35] and here observed a trend for higher whole blood 5-HT levels in males, but only when we performed a one-tailed test (t = 1.64, P = 0.058, n = 12 per genotype) (Figure 4B). No difference in whole blood 5-HT levels was observed with females (one-tailed t = 0.27, P = 0.40, n = 12 per genotype) [See Figure S4 in Additional file 4]. Other studies of serotonergic phenotypes were performed in males only.

Ex vivo and in vivo measures of brain 5-HT phenotypes also suggested an attenuated impact of the Ala56 variant on a B6 background. Midbrain synaptosome 5-HT uptake showed no significant difference between genotypes at either 20 nM (t = 0.11, P = 0.93) or 100 nM of 5-HT (t = 2.93, P = 0.10) (Figure 4C,D). Since ex vivo synaptosome studies may not reproduce changes in 5-HT systems in vivo, we also assessed the impact of intraperitoneal (i.p.) injections of 5-HT receptor agonists [23]. In contrast to our prior findings, injection of the 5-HT_2A/2C agonist 2,5-dimethoxy-4-iodoamphetamine (DOI) induced a similar number of head twitches in Ala56 and Gly56 animals (t = 0.79, P = 0.44) (Figure 4E). Similarly, no significant difference was observed between the genotypes in the maximum hypothermia response to injection of the 5-HT_1A/7 agonist 8-hydroxy-2-(di-n-propylamino)tetralin (8-OH-DPAT) (t = 0.16, P = 0.87) (Figure 4F).

Strain comparisons have been profitably used to identify quantitative trait loci (QTLs) and specific genes that modify traits, as for example with the recombinant inbred (RI) mouse reference panel derived from C57BL/6 and DBA2/J parental strains [24, 36]. In this regard, we recently identified functional coding variation in SERT that distinguishes these two lines (SERT variation is the same in 129 strains as in DBA2/J) and were able to identify multiple anatomical, biochemical and behavioral traits linked to this variation [24]. Strain-dependent genetic variation provides a significant opportunity to identify genetic background effects on the impact of risk alleles, including the SERT Ala56 variant.

Biochemical and behavioral assays displayed a striking divergence of phenotypes between SERT Ala56 knock-in mice on the 129 versus on the B6 genetic background. Whereas a few tests showed quantitative differences in 5-HT related biochemical measures on the B6 background in the same direction as seen for the SERT Ala56 allele on the 129 background, none of these assays achieved statistical significance. We suspect that the complexity of 5-HT homeostasis and signaling [37, 38], allows variation in multiple different proteins to moderate the impact of the SERT Ala56 allele. Importantly, we previously showed that the native 129 SERT (Glu39) has higher 5-HT uptake than the Gly39 variant found in the B6 SERT [24]. It is conceivable that the effect of moving the Glu39 SERT onto the B6 background is large enough that additional effects of the Ala56 variant are not detectable. Further studies directly comparing Glu39 and Gly39 SERTs on a B6 background could shed light on this possibility.

Our findings of genetic background influences on behavioral measures are consistent with strain-dependent behavioral phenotypes observed in mouse models of other ASD-associated genes, most prominently including the Fragile X syndrome model Fmr1 null mouse, which shows variable phenotypes across multiple inbred strains and hybrid crosses [39]. The NLGN3 R451C knock-in mouse also shows differences in behavior between mixed 129S/B6 and pure B6 inbred strains [40‐42]. Further, the SERT null mouse also shows variable behavioral phenotypes across inbred strain backgrounds [43].

The one behavioral phenotype that was consistent across both genetic backgrounds was the tendency for SERT Ala56 mice to back away from mice of other genotypes in the tube test. Not only did the homozygous mutants back out when confronted with wildtype animals, they also backed away when confronted with heterozygous animals. No difference was seen when heterozygous animals and wildtype animals were paired, suggesting that the Ala56 allele only has a significant effect on this phenotype in the homozygous state. From published data, the tube test appears to be a very sensitive test of altered social function in genetic mouse models related to ASD, with significant findings in the Fragile X syndrome model [44], the Rett syndrome model [45], the Arx((GCG)10+7) infantile spasms model [46], the Potocki-Lupski model Dp(11)/17/+ mice [47], and the Dhcr7 heterozygous model of Smith-Lemli-Opitz Syndrome [48]. On the other hand, the tube test may also reveal significant differences in mice with altered motor function or impulsivity, or under conditions of social stress [49], and was originally developed as a test of social hierarchy/dominance, making it a sensitive but not specific test for ASD-relevant social behavior. Social interaction in the tube test does not generate aggression and usually includes only whisker contact. Dominance in this task sometimes [50‐52] but not always [53] predicts territorial aggression, and we did not observe any genotype differences in aggression on the resident-intruder test. We isolated resident mice for 7 days, and it is possible that a longer period of social isolation may have generated more aggression and a potential genotype difference. It seems more likely, however, that the tube test phenotype reflects some genotype difference in social judgment that is not simply a difference in territoriality or aggression.

Ultrasonic vocalizations on postnatal day 7 also showed a significant difference between genotypes on the B6 background, but this difference was in the opposite direction of that observed on the 129 background. Importantly, B6 animals are well-known to have age-related hearing loss beginning around 2 to 3 months of age [54], particularly at higher frequencies, so it is possible that the maternal response to ultrasonic vocalizations would be altered on this genetic background. While maternal response could theoretically interact with pup genotype effects on social communication, it remains difficult to reconcile these opposite findings in the SERT Ala56 mice. Of note, some children with ASD are described as having been passive infants, with less crying, whereas others are described as more dysregulated, with more distress and crying [55, 56].

Other subtle but significant differences were also found in the SERT Ala56 mice on the B6 background. Mixed results in anxiety-like behavior contrast with decreased anxiety-like behavior in the human SERT over-expressing mouse (hSERT OE) [57, 58]. The attenuation of serotonergic phenotypes in the B6 SERT Ala56 mice may account for the difference in findings from reports in the hSERT OE mouse. Alternatively, these discrepant findings could be due to the difference in mechanism between increased (hSERT OE) and dysregulated activity (SERT Ala56 knock-in) [20, 21, 23].

The male SERT Ala56 mice also show a subtle but significant decrease in activity in the open field, which is actually convergent with findings in the hSERT OE mice, but also in the SERT null mice [43, 58, 59]. They do not show changes in locomotion in the home cage or in the Y maze, however, suggesting that this could also be a chance finding. A number of potential explanations could explain convergence in activity measures with increased or ablated SERT activity. Both SERT Ala56 and SERT null mice show a decrease in 5-HT neuron firing rate [23, 60, 61]. Further, Jennings and colleagues showed a loss of frequency sensitivity for evoked 5-HT release in both hSERT OE and SERT null mice, suggesting diminished information processing as a convergent phenotype for opposite disruptions of 5-HT homeostasis [62]. Multiple examples of convergent phenotypes with increased or decreased gene dosage or function have been described in ASD human genetic studies, including MECP2 deletion and duplication [63‐65], chromosome 16p11 deletion and duplication [66], and SHANK3 deletion and duplication [67].