Individual Differences in Autistic Traits are Associated with Serotonin Transporter Gene Polymorphism Through Medial Prefrontal Function: A Study Using NIRS

概要

Individual differences in autistic traits are associated with serotonin transporter gene
polymorphism through medial prefrontal function: a study using NIRS
Akiko Kawamotoa, Aiko Kajiumea, Hiroshi Yoshidab, Tamotsu Toshimac, Masao Kobayashia
a

Department of Pediatrics, Graduate School of Biomedical & Health Sciences, Hiroshima

University, 1-2-3 Kasumi, Minami-ku, Hiroshima 734-8551, Japan
b

Faculty of Contemporary Culture, Hijiyama University, 4-1-1 Ushitashinmachi, Higashi-ku,

Hiroshima 732-8509, Japan
c

Department of Psychology, Graduate School of Education, Hiroshima University, 1-7-1

Kagamiyama Higashi-Hiroshima 739-8521, Japan

INTRODUCTION
Autism spectrum disorder (ASD) is a neurodevelopmental disorder whose core symptoms
include impairments or peculiarities in social interaction and communication (Volkmar and
Pauls, 2003). Patients with ASD usually have abnormalities in brain function or structure (Ha
et al., 2015). The diagnosis of ASD is based solely on behavioral characteristics (Volkmar and
Pauls, 2003), with no objective criteria such as results on a blood test. Symptoms and their
severity differ considerably among patients generically diagnosed as “ASD”.
Autistic traits are also seen in the general population, even in people for whom a
diagnosis of ASD is not warranted. Indeed, autistic traits are distributed along a continuum,
independent of the diagnosis (Constantino and Todd, 2003; Posserud, 2006) and there is no
clear border between those who have autistic traits and who do not. ASD is located at the
extreme end of the distribution. Further, evidence suggests that some genetic factors that affect

1

autistic traits are shared by ASD and subclinical populations (Constantino and Pauls, 2003;
Mosconi et al., 2010; Robinson et al., 2011).
Genetic factors are thought to play an important role in the etiology of ASD (Bailey et
al., 1995; Vorstman et al., 2006), but recently its association with certain early environmental
factors has also become evident (Hallmayer et al., 2011). Because the behavioral characteristics
of ASD result from multi-factorial heritability and interactions with environmental factors,
individual differences in symptoms and their severity are considerable. This makes
understanding the pathogenesis of ASD complicated.
Recently, gene polymorphism has drawn interest as a likely contributor to the differences
in individual symptoms and severity found in ASD. In particular, serotonin transporter-linked
polymorphic region (5-HTTLPR) has been investigated by several groups for association with
ASD (Christine and Susan, 2008; Cook et al., 1997; Devlin et al., 2005; Kistner-Griffin et al.,
2011; Klauck et al., 1997; Yirmiya et al., 2001). 5-HTTLPR comprises two variants, a low
expressing short (S)-allele and a high expressing long (L)-allele, which result in altered
transcriptional activity and function of the serotonin transporter (5-HTT) (Heils et al., 1996).
5-HTT is a key regulator of serotonergic neurotransmission (Canli et Lesch, 2007) and is the
presumed site of action for selective serotonin reuptake inhibitors (SSRI). The serotonergic
system is of interest in the etiology and pathogenesis of autism for several reasons: (1) ASD is
associated with elevated whole blood serotonin levels (Cook end Leventhal, 1996; Dubravka
et al., 2007; Lam et al., 2006); (2) core ASD symptoms can sometimes be improved by
treatment with SSRIs (Dove et al, 2012; Gordon et al., 1993; Hollander et al., 2005; Hollander
et al., 2012; Kolevzon et al., 2006; McDougle et al., 1996); (3) the brain serotonergic system
is disrupted in patients with ASD (Chandana et al., 2005) and in animal models of autism
2

(Tamada et al., 2010). Serotonin is therefore a strong candidate molecule for influencing ASD
and 5-HTTLPR is a potential candidate gene polymorphism linked to ASD. However, current
findings are inconsistent, with different alleles or no alleles reportedly being associated with
risk of ASD (Klauck, 2006).
In the current study, we assessed the effect of 5-HTTLPR on the wide range of autistic
traits seen in the general population. By interposing an objective index of brain function
between autistic traits and 5-HTTLPR, we were able to examine the possibility that 5-HTTLPR
has an indirect effect on traits through mediation by brain function. We focused on the medial
prefrontal cortex (mPFC) because previous studies have reported that it is associated with both
ASD and the serotonergic system. The mPFC plays a role in social cognition (Grossmann,
2013), and altered mPFC activity in ASD has been reported to be associated with deficits in
social communication and interaction ability (Ohnishi et al., 2000), including poor processing
of facial affect (Harms, 2010; Watanabe et al., 2012). A study using near-infrared spectroscopy
(NIRS) has reported that the degree to which healthy people exhibit autistic traits is correlated
with how active the prefrontal cortex (PFC) is when people view images of facial expressions
(Hosokawa et al., 2014).
Studies also indicate that mPFC function can be affected by serotonin or 5-HTTLPR.
Indeed, serotonin is a major modulator of the PFC (Puig and Gulledge, 2011; Victoria) and
plays an essential role in PFC function (Challis and Berton, 2015). Different 5-HTTLPR
genotypes result in functional and anatomical differences in brain regions, including the mPFC
(Jasinska et al., 2012; Rao et al., 2007), which might be related to why S-allele carriers exhibit
higher sensitivity to social and emotional cues (Friedel et al., 2009; Heinz et al., 2005).

3

Additionally, altered 5-HTT binding capacity has been observed in the medial frontal area of
individuals with ASD (Makkonen et al., 2008).
In light of these reported links between 5-HTTLPR, mPFC function, and ASD or autistic
traits, we considered that the degree of mPFC activity during a social-emotional cognition task
could be a relevant factor linking autistic traits and 5-HTTLPR. We hypothesize that 5HTTLPR indirectly affects autistic traits via mediation by mPFC function. To date, no studies
have examined the relationships between 5-HTTLPR, mPFC activation, and autistic traits in
the same participants.
In this study, we evaluated mPFC function using a facial affect-labeling task that was
based on the task in Nishikawa et al. (2015). Individuals with ASD have difficulty recognizing
facial affect (Kasari et al., 1993; Lozier et al., 2014; Sigman et al., 1992), and children with
autistic-like social communication difficulties have been reported to have similar difficulties
(Kothari et al., 2013). Although patients with ASD show poor performance on facial affectlabeling tasks (Bölte and Poustka, 2018), training with this type of task can effectively improve
daily life skill in recognizing facial affect (Wakamatsu, 2014). Training by facial affect labeling
has also been effective in patients with stroke as it increases prefrontal activation in response
to facial recognition (Shibasaki and Yoshida, 2016). We predicted that those who have greater
autistic traits will show lower prefrontal activation during the facial affect-labeling task.
Moreover, according to previous findings, the serotonergic system is involved in processing
facial affect; scores for labeling facial affect increased in healthy participants who received
SSRIs (Harmer et al., 2003) and the 5-HTTLPR genotype affects the ability to recognize facial
emotion (Antypa et al., 2003).

4

This study comprises two experiments. Experiment 1 tested whether mPFC activation
differs between those who have ASD and those who do not. In Experiment 2, we used the same
task to examine the relationship between 5-HTTLPR, mPFC activation, and autistic traits in
sub-clinical volunteers. Clarifying the neurogenetic basis of autistic traits in the sub-clinical
population may help us do the same for ASD. Thus, we determined whether individual levels
of autistic traits are indirectly affected by 5-HTTLPR through mediation by mPFC function.

EXPERIMENTAL PROCEDURES
Experimental design
Experiment 1: The aim was to examine whether mPFC activation induced by the facial affectlabeling task differs between those who have ASD and those who do not. We measured mPFC
activity levels using NIRS and assessed levels of autistic traits using the Autism-Spectrum
Quotient (AQ). These measures were compared between the ASD group and a typical
development (TD) group.
Experiment 2: The aim was to determine whether individual autistic trait levels were associated
with 5-HTTLPR in those without ASD, and if so, whether the relationship was mediated by
neural activity involved in social functions. In addition to the measures used in Experiment 1,
participants were genotyped for 5-HTTLPR. The correlation and mediation analyses were
conducted using Structural Equation Models (SEM) among AQ scores, changes in mPFC
activation, and 5-HTTLPR genotype.

Participants
5

All participants were right-handed. Written informed consent was obtained for each
experiment; from all parents before Experiment 1 and from all participants before Experiment
2. The study was performed in accordance with the declaration of Helsinki and approved by
the Ethics Committee of Clinical Study, Hiroshima University Hospital.
Experiment 1: Twenty children participated (ASD: n = 9, 8 boys and 1 girl, aged 7–14 years;
age- and sex-matched TD: n = 11, 7 boys and 4 girls, aged: 5–12 years). All children with ASD
were recruited from the Department of Pediatrics, Hiroshima University and had been
diagnosed according to the criteria of the Diagnostic and Statistical Manual of Mental
Disorders, fifth edition (DSM-V). The intelligence quotients (IQs) of the children with ASD
were measured using the Wechsler Intelligence Scale for Children, third or fourth edition, and
all had a full-scale IQ above 70. TD children were recruited from the local community, and
none were suspected of any developmental abnormalities. None of the TD children had siblings
with ASD.
Experiment 2: One hundred eighty neurotypical Japanese adults participated (121 males, 59
females; mean age: 23.9 ± 2.8 years). None suffered from any psychiatric disease or were
suspected of any developmental abnormalities. In order to investigate a wide range of
individual differences, no restriction was made with respect to the AQ total score.

Autistic traits assessment
The Japanese version of the AQ was used to assess the level of autistic traits. In Experiment 1,
the parents were asked to fill out the children’s version of the AQ (Wakabayashi, 2007) and in
Experiment 2 the participants completed the adult version (Wakabayashi et al., 2004) by
themselves. The questionnaires consisted of 50 items, made up of five 10-item sub-domains
6

(social skills, attention switching, attention to detail, communication, and imagination). Each
item is worth one point, making the maximum AQ total score equal to 50 (and that for each
sub-domain equal to 10). Higher scores indicate higher levels of autistic traits.

Facial affect-labeling task
The facial affect-labelling task was based off the task in Nishikawa et al. (2015) and was
designed to evaluate mPFC activation related to social/emotional cognitive processing (Fig. 1).
The task was performed using a personal computer with a touch panel and consisted of pretask, task, and post-task periods. In the task period, participants were requested to view a
woman’s face displayed on the monitor and use the touch screen to select a verbal label that
best described how the woman was feeling. Five labels (rectangles) were displayed at the
bottom of the screen, each with one of the following words: “happy”, “sad”, “angry”,
“surprised”, and “I am not sure” (written in Japanese). Each of four affective faces (happy, sad,
angry, and surprised) was presented twice in random order for a total of 8 trials. Each trial
ended as soon as one of the labels was touched.
In the pre- and post-task periods, only neutral faces were displayed. The participants
were instructed to look at the neutral face and select the label “Neutral” (written in Japanese).
The other four labels were blank. The location of the “Neutral” label was randomly chosen for
each trial. Each trial ended as soon as any label was touched.
The pre-task, task, and post-task periods included 6, 8, and 8 trials, respectively. Each
period began with an instruction displayed for 3 seconds that informed the participants about
the period. Trials were separated by a 1 second interval during which a blank black screen was
presented. The affective and neutral face stimuli were the “averaged faces” used by Maki et al.
7

(2013). Images were made by synthesizing standardized photos of four Japanese women from
database DB99 (Advanced Telecommunications Research Institute International, Inc. Nara,
Japan) to exclude non-emotional confounding factors and individual features of expressing
emotions.

NIRS measurement
A two-channel NIRS machine (NIRO-200NX; Hamamatsu Photonics, Hamamatsu City, Japan)
was used to measure changes in oxygenated-hemoglobin (oxy-Hb) and deoxygenatedhemoglobin (deoxy-Hb) concentration in the bilateral mPFC. Two probes were symmetrically
placed on both sides of the forehead surface, at Fp1 and Fp2, according to the international 1020 system used in electroencephalography. This probe placement was chosen in reference to
previous studies to detect the regional cortical oxygenation in mPFC (Okamoto et al., 2004).
The NIRS machine used three different wavelengths of near-infrared light (735, 810, and 850
nm) and tracked the variation in oxy-Hb and deoxy-Hb concentrations that were calculated (via
the Beer-Lambert law) from the changes in light absorption. The distance between the emission
and detection probes was 3.0 cm. The NIRS machine measures changes in hemoglobin
concentration approximately 2–3 cm beneath the surface of the skull, which only incudes the
surface of the cortex. The time resolution of the NIRS signal was 1.0 second. Changes in oxyHb (ΔOxy-Hb) were used for the analysis because it is known to reflect cortical activity better
than changes in deoxy-Hb (Hoshi et al., 2001). To compare mPFC activation among
participants, we normalized the raw ΔOxy-Hb by converting it to a z-score at each NIRS
channel: the mean ΔOxy-Hb from baseline (10 seconds just before the task period) through the
task period (5–15 seconds) was divided by the standard deviation of the baseline ΔOxy-Hb.
8

Genotyping
Genotyping of 5-HTTLPR was performed in Experiment 2. Buccal cells were sampled using
mouth swabs (GE Healthcare Japan), and genomic DNA was extracted using a QIAmp DNA
Investigator Kit (Qiagen Inc, Tokyo, Japan). The SLC6A4 gene promoter region was amplified
by

polymerase

chain

reaction

GGCGTTGCCGCTCTGAATGC-3′)

(PCR).
and

The

reverse

forward
primer

primer

(5′-

sequences

(5′-

GAGGGACTGAGCTGGACAACCAC-3′) were the same as those previously described
(Tomoda et al., 2013). The 50 µl reaction mixture contained 50 ng genomic DNA, 1.25 U of
Tks Gflex DNA Polymerase (Takara Bio, Japan), 25 µl of 2 × Gflex PCR buffer (Takara Bio,
Japan), and 15 pmol of each primer. The protocol for PCR was as follows: initial denaturation
(94°C for 1 min) followed by 35 amplification cycles (denaturation at 98°C for 10 s, annealing
at 60°C for 15 s, and extension at 68°C for 1 min). The PCR products were separated using
electrophoresis in a 3% agarose gel stained with ethidium bromide and visualized by ultraviolet light. The observed 484 bp band indicated the short (S) allele and the 528 bp band
indicate the long (L) allele. Each participant was identified as one of 3 groups: homozygote for
the S-allele (S/S), heterozygote for the S- and L-alleles (S/L), and homozygote for the L-allele
(L/L).

Statistical analysis
Experiment 1: Analyses were performed using IBM SPSS version 23 (IBM Corp., Armonk,
NY, USA). Student’s t-test were conducted for age, AQ score, and mean ΔOxy-Hb z-score in
each channel between the ASD and TD groups, and Fisher’s exact test was conducted for
9

gender. Correlation analyses were conducted between AQ scores and the mean ΔOxy-Hb zscore for each channel of each diagnostic group.
Experiment 2: A one-way ANOVA was performed for age, and Pearson’s chi-square test (2sided) was conducted for gender between the 5-HTTLPR genotypes to verify the inter-group
variability. Correlation analyses were conducted among AQ scores, mean ΔOxy-Hb z-score in
each channel, and the number of 5-HTTLPR L-alleles. Mediation analysis with Structural
Equation Models (SEM) was performed using AMOS version 23 (an add-on to the SPSS
statistical software) to test the 5-HTTLPR-mPFC-autistic traits pathway. We assessed the
potential effect of mediation according to the “causal steps approach” described by Baron and
Kenny (1986). In the first analysis, a direct path from independent variable X (5-HTTLPR) to
dependent variable Y (AQ score) was drawn to analyze the direct relationship. In the second
analysis, an indirect path through variable M (oxy-Hb) was added to the first path. If the direct
effect from X to Y disappears or is strongly reduced in the second analysis, variable M can be
interpreted as a significant mediator. Model fit was assessed by chi-square test (χ2), the General
Fit Index (GFI), the Comparative Fit Index (CFI), and the Root Mean Square Error of
Approximation (RMSEA). Acceptable fitting models require nonsignificant chi-square test,
GFI > 0.90, CFI > 0.90, and RMSEA < 0.08.

RESULTS
Experiment-1
Basic participant characteristics

10

Age and gender did not differ significantly between the TD and ASD groups (Table 1). There
were no gender differences in age, ΔOxy-Hb, or AQ scores in either diagnostic group.
Participant age was not correlated with the mean ΔOxy-Hb z-score or the AQ scores in either
diagnostic group.
variables

Females (n, %)

ASD

TD

(n=9)

(n=11)

Group differences
t-value

p-value

1 (11.1)

4 (36.3)

11.5 (1.8)

9.9 (3.0)

-1.4

0.18

27.8 (5.4)

13.5 (5.6)

-5.8

<0.01

Social skills

5.0 (2.4)

2.7 (1.6)

-2.6

0.02

Attention switching

6.2 (1.1)

3.5 (1.8)

-4.1

<0.01

Attention to detail

4.1 (1.7)

2.7 (1.2)

-2.1

0.046

Communication

5.9 (2.0)

2.2 (1.8)

-4.3

<0.01

Imagination

6.6 (2.1)

2.4 (1.6)

-5.2

<0.01

Age

0.32

AQ scores
Total

Table 1. Number of participants in each group and their mean age and AQ scores. Higher AQ
scores indicate higher autistic traits. Age and AQ scores are presented as means (standard
deviations).

AQ score
The ASD group had higher AQ total scores (p < 0.01; Fig. 2A) and higher sub-domain scores
(social skills: p = 0.02; attention switching: p < 0.01; attention to detail: p = 0.04;
communication: p < 0.01; imagination: p < 0.01) than TD group (Table 1).
11

NIRS data
The oxy-Hb (z-score) in the right and left mPFC during the task period (from 5 to 15 s) was
significantly higher than baseline in the TD group (right mPFC: p = 0.01; left mPFC: p < 0.01)
but not in the ASD group (right mPFC: p = 0.1; left mPFC: p = 0.74). The mean ΔOxy-Hb zscores in bilateral mPFC were significantly lower in the ASD group than in the TD group (right
mPFC: p < 0.01; left mPFC: p = 0.01; Fig. 2B). The association between mPFC activation and
AQ scores was examined using Pearson’s correlation coefficient for each diagnostic group and
for all participants. Significant correlations were observed between several parameters (Table
2).
Mean z-score for the change in oxy-Hb
Parameters

ASD group

TD group

all participants

rmPFC

lmPFC

rmPFC

lmPFC

-0.687*

-0.657*

-0.658

-0.682* -0.819** -0.762**

-0.668*

-0.625*

-0.642

-0.601

-0.727** -0.698**

-0.462

-0.337

-0.308

-0.332

-0.689** -0.584**

0.446

0.379

-0.128

-0.605

Communication -0.693*

-0.282

-0.191

-0.193

-0.730** -0.535*

Imagination

-0.039

-0.226

-0.377

-0.538**

AQ scores Total
Social skills
Attention
switching
Attention
to detail

-0.254

rmPFC

-0.171

lmPFC

-0.312

-0.356

12

Table 2. Pearson's correlation analyses between the mean ΔOxy-Hb z-scores in the right and
left mPFC and AQ scores for each diagnostic group and for all participants. The values
represent the Pearson’s r correlation coefficients. * p < 0.05, **p < 0.01.

Experiment-2
Basic participant characteristics
We found no gender differences in age, ΔOxy-Hb, or AQ scores. We found no significant
differences in gender or age among the three genotypes (Table 3). Participant age did not
correlate with the mean ΔOxy-Hb z-scores or AQ scores. The frequency distributions of the 5HTTLPR genotypes were as follows: S/S: 103, S/L: 67, L/L: 10. Observed genotype
frequencies did not deviate significantly from Hardy-Weinberg equilibrium (χ2 = 0.04, p =
0.76).
variables

S/S genotype

S/L genotype

L/L genotype

Group differences

(n=103)

(n=67)

(n=10)

p-value

Females (n, %)
Age

35 (34)

19 (28.4)

5 (50)

0.37

23.8 (2.75)

24.0 (2.68)

24.5 (3.10)

0.65

Table 3. Number of participants in each group and their mean age. Ages are presented as means
(standard deviations).

5-HTTLPR genotype effect on autistic traits and mPFC activation
Analysis using Spearman’s rank correlation coefficients were used to examine whether the
number of L-alleles was correlated with AQ scores or the mean ΔOxy-Hb z-score in the mPFC
(Table 4). Analysis revealed significant correlations with scores on the AQ social skills
13

subdomain and with mean ΔOxy-Hb z-scores in right mPFC (ρ = 0.203 and ρ = 0.364
respectively, both p < 0.01).
Parameters
AQ scores

Total
Social skills

Mean z-score for the
change in oxy-Hb

Number of L-allele
0.125
0.203**

Attention switching

0.004

Attention to detail

0.035

Communication

0.027

Imagination

0.031

right mPFC

-0.364**

left mPFC

-0.121

Table 4. Spearman's rank correlation analysis between the number of 5-HTTLPR L-alleles
and other factors (AQ scores and mean ΔOxy-Hb z-scores for right and left mPFC). The
values represent the Spearman’s rho correlation coefficients. *p < 0.05, **p < 0.01.

Relationships between mPFC activation and AQ scores
Pearson’s correlation coefficient was used to examine the association between mPFC
activation and AQ scores. We found significant negative correlations between mean ΔOxy-Hb
z-score in the right mPFC and AQ total score (r = −0.271, p < 0.01), AQ social skills score (r
= −0.395, p < 0.01; Fig. 3), AQ attention-switching score (r = −0.228, p < 0.01) and AQ
communication score (r = −0.276, p < 0.01) (Table 5).
14

Parameters

AQ scores

Mean z-score for the change in oxy-Hb
right mPFC

left mPFC

Total

-0.271**

-0.191*

Social skills

-0.395**

-0.184*

Attention switching

-0.228**

-0.150*

0.142

0.038

-0.276**

-0.245**

-0.069

-0.057

Attention to detail
Communication
Imagination

Table 5. Pearson's correlation analysis between the mean ΔOxy-Hb z-scores in right and left
mPFC and AQ scores. The values represent the Pearson’s r correlation coefficients. *p < 0.05,
**p < 0.01.

SEM evaluation
Based on the correlation analyses, we hypothesized that the right mPFC activation mediated
the effect of 5-HTTLPR genotype on autistic traits related to social skills. We designed path
models to test this hypothesis. We used the number of 5-HTTLPR L-alleles as the independent
variable, AQ social skills score as the dependent variable, and the mean ΔOxy-Hb z-score in
right mPFC as the mediator in the pathway. The first analysis examined the model that includes
only a direct effect from 5-HTTLPR to AQ social skills score. This model was completely
saturated and provided a perfect fit (χ2 = 0, df = 0, p > 0.01, CFI = 1.00, RMSEA = 0.00). The
direct path was significant (β = 0.22, p < 0.01). Next, to assess the extent to which right mPFC
activation reduced the magnitude of the direct effect, we analyzed a mediation model
comprising the direct and indirect paths (Fig. 4A). This model was also completely saturated
15

and provided a perfect fit (χ2 = 0, df = 0, p > 0.01, CFI = 1.00, RMSEA = 0.00). The paths
from 5-HTTLPR to right mPFC (β = −0.33) and from mPFC to AQ social skills (β = −0.36)
were both significant (p < 0.01). However, the direct path from 5-HTTLPR to AQ social skills
was not significant (β = 0.097, p = 0.18). The reduction was from β = 0.22 (p < 0.01) to β =
0.097 (ns). Thus, controlling for the mediating effect of right mPFC eliminated the direct effect.
The model that only included the indirect path also had an acceptable fit (χ2 = 1.79, df = 1, p
= 0.18, GFI = 0.99, CFI = 0.98, RMSEA = 0.066, see Fig. 4B). All pathways were significant
in this model (p < 0.01).

DISCUSSION
In Experiment 1 we examined whether mPFC activation induced by the facial affect-labeling
task differs between those who have ASD and those who do not. The NIRS results showed that
the task normally activates bilateral mPFC—the region responsible for recognition of facial
emotion (Heberiein et al., 2008)—and that activity levels were significantly lower in the ASD
than in the TD group. Simultaneously, we confirmed that individuals who have higher AQ
scores (i.e., the ASD group), show lower mPFC activity when engaged in the facial affect
labeling task. The reduced mPFC activation that we observed in the ASD group can be
considered to be associated with impaired ability to recognize facial emotion. Previous findings
suggest that the role of the mPFC when viewing an affective face is to inhibit amygdala
responses (Hariri et al., 2000; Lieberman et al., 2007), which then allows the emotion to be
judged correctly (Heberiein et al., 2008). Indeed, patients with lesions confined to ventral
mPFC have been reported to show impaired recognition of facial emotion (Heberlein et al.,
16

2008). Furthermore, when participants evaluate the emotion of an affective face and label it by
putting the emotions into words, the induced activation in the amygdala is further suppressed
by the mPFC (Hariri et al., 2000; Lieberman et al., 2007). The level of right PFC activity during
facial affect labeling has been reported to be negatively correlated with activity levels in the
amygdala, perhaps reflecting cognitive control of emotional responses through appraisal and
evaluation of emotional stimuli (Hariri et al., 2000). This inhibitory function of the mPFC helps
to alleviate emotional distress and plays an important role in emotion regulation (Etkin et al.,
2012; Lieberman et al., 2007; Motzkin et al, 2015; Quirk and Beer, 2006). A previous study
has reported that poorer ability to recognize facial emotion is correlated with deficits in social
communication skills in children with ASD. This implies that the reduced mPFC activation
that we saw in the ASD group might contribute to the dysfunction in how cognitive emotions
are evaluated and underlie the social impairments of ASD. Therefore, we next examined the
link between mPFC activation, the autistic traits, and 5-HTTLPR.
In Experiment 2 we investigated whether autistic traits were associated with 5-HTTLPR
in sub-clinical participants, and if so, was the association mediated by neural activity related to
social functions (i.e., the reduced mPFC activity observed in Experiment 1). AQ scores
revealed that people with large numbers of 5-HTTLPR L-alleles have stronger autistic social
skill traits. This result is supported by a previous study demonstrating that individuals with
high 5-HTT-expressing LA/LA genotypes had lower social skills than those with low 5-HTT
expressing S genotypes or who were LG-allele carriers (Gadow et al., 2013). However, a
previous study has reported that S-allele carriers with ASD show more severe social and
communication deficits (Tordjman et al., 2001). We discuss this in more detail below.
More interestingly, we found that individuals carrying larger numbers of L-alleles
17

showed lower right mPFC activation in response to facial affect labeling. This tendency
suggests weaker engagement of emotion-evaluation processes during facial affect labeling in
these individuals. Our results are supported by previous studies showing that individuals with
high 5-HTT-expressing LA/LA genotypes are less able to accurately recognize facial
expressions of emotion (Boll and Gamer, 2014), and that the intensity threshold for recognizing
negative facial expressions is higher for individuals with L/L genotypes than it is for S-allele
carriers (Antypa et al., 2011).
More related to our hypothesis, correlation analysis showed that the right mPFC activity
was negatively correlated with the AQ social skills score, which is correlated with the number
of L-alleles. This result is consistent with the result observed in Experiment 1 in which lower
mPFC activity was observed in the ASD group, which also had higher total AQ and AQ social
skills subdomain scores. Together with Experiment 1, we have thus demonstrated the
possibility that right mPFC activity engaged in facial affect recognition is the neural correlate
of some autistic traits. Furthermore, our SEM analysis confirmed that right mPFC activation
plays a mediator role in connecting 5-HTTLPR and autistic traits related to social skills. The
mediating effect was not seen in the left mPFC. These results suggest a possible neural
mechanism underlying the effect that 5-HTTLPR has on individual differences in social skills
seen in ASD, and the possibility that the cognitive process involved in evaluating emotional
stimuli can mediate the genotype-dependent tendency seen in these autistic traits. Thus, our
findings offer evidence for genetic and neural correlates of autistic traits related to social skills.
Caution should be exerted when attempting to generalize the implication of our gene,
neuroimaging, and phenotype results. Although our results showed that the S-allele was
associated with lower autistic traits related to social skills, as mentioned above, another study
18

has reported that the S-allele is highly associated with deficits in sociability and communication
(Tordjman S, et al., 2001). Furthermore, many previous studies have reported that the S-allele
is linked to negative outcomes, such as increased anxiety-traits and risk of psychiatric
disturbances in response to stressful life events (Caspi et al., 2003; Juhasz et al., 2015; Lesch
et al., 1996). A possible explanation for this apparent discrepancy is that S-allele carriers tend
to be more sensitive to environmental stimuli (Kiser et al., 2012). Homberg and Lesch have
mentioned in their review that hypervigilance of S-allele carriers may be the common
denominator in social cognitive superiority and anxiety-related traits, and that environmental
conditions determine whether a response will turn out to be positive (cognitive, in conformity
with the social group) or negative (emotional) (Homberg and Lesch, 2011). Under favorable
environments such as supportive and enriching social conditions, S-allele carriers show
adaptive outcomes, improved cognition, and social conformity. In contrast, in adverse
environments such as distressing and oppressive social conditions, S-allele carriers tend to be
emotional, leading to maladaptive responses and increased risk for mood disorders (Homberg
and Lesch, 2011). Kochanska et al. (2011) have reported that while the social competence of
L/L genotype carriers is not affected by early rearing environment, S-allele carriers reared by
unresponsive mothers had significantly lower social competence compared to those reared by
responsive mothers or to those with L/L genotypes. These findings indicate that the social
competence of S-allele carriers is sensitive to individual experiences and environmental
conditions. Taken together, we can say that higher neural activity in response to emotional
stimuli and higher facial affect-recognition ability (both related to the S-allele) can lead to
maladaptive social behavior, depending on experiences and environmental conditions.
Therefore, our observation does not necessarily imply that S-allele carriers always show
19

lower autistic traits related to social skills. Further studies are needed to understand whether
the discrepancy between our results (a positive aspect of the S-allele) and those from previous
studies (negative aspects of the S-allele) can be explained by the higher sensitivity of S-allele
carriers. Furthermore, since the pathogenesis of ASD includes early negative environmental
factors, the higher social skills observed in the sub-clinical S-allele carriers of this study is not
necessarily identically expressed in patients with ASD. The discrepancy in results among
studies that have examined the relationship between the S-allele and social skills may depend
on individual participant characteristics, such as the ASD diagnosis or other environmental
factors. Future research should separately examine ASD and healthy participants, considering
the effects of environmental factors. Having said that, it is interesting to note that individuals
who scored 9 or 10 on AQ social skills in our study were all L-allele carriers (Fig. 3).
This study has four methodological limitations, which are described below.
1. This study only included Japanese participants. Previous studies have shown
differences in neural activation patterns and transcultural social behaviors among
cultural groups (Kobayashi et al., 2006; Koelkebeck et al., 2011). Furthermore, ethnic
difference in the influence of 5-HTTLPR on brain networks has also been reported
(Kong et al., 2014). Replication studies in other ethnicities are needed to confirm our
findings.
2. We did not consider gender difference because we could not obtain enough female
participants with ASD in Experiment 1 (only one female). Previous studies have
reported significant differences in ASD prevalence between males and females (Lai
et al., 2014; Fombonne, 2003; Werling et al., 2013) and significant gender differences
in clinical presentation, including social and communicative domains (Werling et al.,
20

2019; Kirkovski et al., 2013). This limitation means that the current findings should
be interpreted cautiously.
3. All participants were right-handed. Previous research has reported that left-handers
and mixed-handers are more common among people with ASD than among the
general population (Rysstad et al., 2018). The current study reports the role of the right
mPFC only in right-handed participants. Further research is needed to determine
whether our results can be replicated in left-handers or mixed handers.
4. Despite the significant correlation of genotype and AQ social skills scores in
Experiment 2, we must be cautious about inferring a gene-trait interaction because the
sample size is small (Only 10 participants had L/L genotypes). Future studies are
expected to confirm our findings using larger sample sizes.
In conclusion, the present study provided evidence of a link between 5-HTTLPR and the
degree of autistic traits related to social skills. The neuroimaging findings indicate that higher
autistic traits predict lower mPFC activation when trying to discern facial emotions. More
interestingly, our analysis revealed a mediator role for the right mPFC in linking 5-HTTLPR
and autistic traits related to social skills, suggesting a potential neural mechanism that controls
how 5-HTTLPR effects individual autistic traits. Further studies of patients with ASD will
provide additional biological evidence for the clinical features of ASD, and provide targets for
potential pharmacotherapy and medical treatment.

This study is published on Neuroscience 458 (2021) 43–53

21

Figure Captions
Figure 1. Experimental protocol for the facial affect-labeling task. The actual labels were in
Japanese.

Figure 2. (A) Distribution of AQ total score for TD and ASD. (B) Distribution of mean z-score
of Δoxy-Hb in the left and right mPFC for TD and ASD. Solid horizontal lines represent the
mean value. The mean values for ASD were both significantly different than those for TD (p <
0.05).

Figure 3. A scatter plot showing the relationship between AQ social skills score and mean
ΔOxy-Hb z-score in the right mPFC of each participant. A negative correlation is visible,
shown by the straight line that was obtained by a fit using all the data.

Figure 4. The mediating role of the right mPFC in the association between 5-HTTLPR and
autistic traits related to social skills. (A) Mediation model illustrating how right mPFC
activation mediates the direct effect between 5-HTTLPR and autistic traits related to social
skills. The number in parenthesis under the path from 5-HTTLPR to the social skills is the
coefficient for this direct path when the indirect path is not included. (B) Mediation model
including only the indirect effect through right mPFC, eliminating the direct effect from 5HTTLPR to autistic traits related to social skills. *p < 0.05, **p < 0.01, ns; not significant.

22

Figure 1

Figure 2

23

Figure 3

Figure 4

24