Fetal Alcohol Exposure Alters BOLD Activation Patterns in Brain Regions Mediating the Interpretation of Facial Affect

Abstract

Background:

Although deficits in the interpretation of affective facial expressions have been described clinically and in behavioral studies of fetal alcohol spectrum disorders (FASD), effects of prenatal alcohol exposure on the neural networks that mediate affective appraisal have not previously been examined.

Methods:

We administered a nonverbal event-related fMRI affective appraisal paradigm to 64 children (mean age=12.5 yr; 18 with fetal alcohol syndrome (FAS) or partial FAS (PFAS), 18 non-syndromal heavily exposed (HE), and 28 controls). Happy, sad, angry, fearful, and neutral faces and pixelated control images were presented sequentially in a randomized order. The child indicated whether the currently displayed face showed the same or different affect as the previous one.

Results:

Data from whole brain analyses showed that all groups activated the appropriate face processing neural networks. Region of interest analyses indicated that, compared to HE and control children, the FAS/PFAS group exhibited greater blood oxygenation level dependent (BOLD) signal changes when processing neutral faces than pixelated images in two regions that form part of the visual-sensory social brain network, which plays an important role in the initial processing of facial affect. By contrast, BOLD signal when processing angry faces was smaller for the FAS/PFAS group in a region involved in the processing of facial identity and facial expressions and in a region involved in the recognition and selection of behavioral responses to aggressive behavior.

Conclusions:

These findings of greater BOLD signal in the FAS/PFAS group in response to neutral faces suggest less efficient neural processing of more difficult to interpret emotions, and the weaker BOLD response to angry faces suggests altered processing of angry stimuli. Although behavioral performance did not differ in this relatively simple affective appraisal task, these data suggest that in children with FAS and PFAS the appraisal of neutral affect and anger is likely to be more effortful in more challenging and dynamic social contexts.

The adverse effects of prenatal alcohol exposure (PAE) on the developing brain have been recognised as a major global public health problem (May et al., 2018). PAE can result in physical, behavioral and cognitive deficits collectively referred to as fetal alcohol spectrum disorders (FASD). Fetal alcohol syndrome (FAS), the most severe of these disorders, is characterized by distinctive craniofacial dysmorphology, microcephaly, and pre- and/or postnatal growth retardation (Hoyme et al., 2005). Partial FAS (PFAS) is diagnosed in individuals with a confirmed history of PAE who have the characteristic pattern of facial anomalies and growth deficiency, microcephaly or neurobehavioral impairment. The effects of PAE on brain development are diffuse, and the neurobehavioral profile indicates widespread impairment, including lower IQ (Jacobson et al., 2004; Mattson et al., 2011) and poorer learning and memory (Pei et al., 2008; Lewis et al., 2015; Du Plooy et al, 2016) and executive function (see Mattson et al., 2019 review).

Numerous studies have examined the effects of PAE on the blood oxygenation level dependent (BOLD) response using task-based functional magnetic resonance imaging (fMRI). Differences in regional activations (both increases and decreases) have been observed in children with PAE compared to controls on a range of cognitive tasks. For example, during a number processing task, Woods et al. (2015) found that higher levels of PAE were associated with increased activation in the angular gyrus but decreased activation in the right intraparietal sulcus (Woods et al., 2015). Similarly, during an inhibition task O’Brien et al. (2013) found that alcohol-exposed children showed increased frontal and parietal activations compared to controls despite similar behavioral performance but reduced activation in the pre- and post-central gyri on cued trials, possibly explaining their impaired performance on these trials. Greater BOLD response was also observed in critical regions in children with PAE compared to controls on a response inhibition task, particularly as task difficulty increased (Ware et al., 2015). Similarly, Fryer and colleagues (2007) found increased BOLD responses in the prefrontal cortex in children with FASD during a response inhibition task but reported decreased activation in the right caudate nucleus. Differential recruitment of critical brain regions was also observed during a working memory task, in which children with FAS or PFAS recruited different regions than nonsyndromal heavily exposed (HE) children, who recruited different regions than controls (Diwadkar et al., 2013).

In addition to cognitive impairment, many children with FASD exhibit poor social judgment (Greenbaum et al., 2009; Mattson et al., 2000) and deficits in social skills (Roebuck et al., 1999; Schonfeld et al., 2006) that become more pronounced with age and persist into adulthood (Carmichael Olson et al., 1997; Dodge et al., 2014; Thomas et al., 1998). Secondary effects of poor social judgment include high rates of delinquent and criminal behaviour, which are believed to be attributable, in part, to poor social competence (Roebuck et al., 1999). Little is known about the underlying mechanisms that mediate these socioemotional deficits, which are often attributed to cognitive problems, including impairment in executive function (Streissguth et al., 1996; McGee et al., 2008). Other studies have reported poorer social skills in children with PAE, even after statistical adjustment for overall cognitive functioning (Thomas et al., 1998; Whaley et al., 2001) and executive function (Lindinger et al., 2016). One study, which examined Theory of Mind, the ability to infer mental states (i.e., beliefs, intents and desires) of others, found that 4- to 8-year-old alcohol-exposed children’s performance on “false-belief” tasks (which assess ability to infer that another person’s belief about the world may differ from one’s own) was related to difficulties with inhibitory control (Rasmussen et al., 2009; see Kully-Martens et al. 2012 review).

The ability to correctly interpret and respond to social cues, particularly facial affect, plays a major role in social and emotional functioning (Adolphs, 2010). To date, three studies have examined recognition of facial affect in FASD. Greenbaum et al. (2009) found that 6- to 13-year-old children with FASD had more difficulty matching an emotion label to photos displaying facial affect than controls. In the same study, emotion processing was found to be a predictor of social skills, assessed by parents and teachers. Kerns et al. (2016) found that 8- to 14-year-olds with FASD had difficulty recognizing facial emotions in photos showing adult but not children’s faces. In our study, school-aged children with PAE performed more poorly on “Reading the Mind in the Eyes” (Baron-Cohen et al., 2001), an advanced Theory of Mind task, in which the child was asked to infer mental states of others by only looking at the eye region of faces (Lindinger et al., 2016). These findings raise the question whether this deficit may be mediated, in part, by impairment in the ability to identify facial affective cues.

A recent systematic review (Alcala-Lopez et al., 2018) examined task-based social brain activations in 26 meta-analyses of 3972 neuroimaging studies and derived four functional social-affective processing networks. The hierarchical clusters consisted of 36 seeds, divided by type of input and their level of BOLD activation. The high- and intermediate-level networks involve higher-order social-cognitive processes, such as the interpretation of mental states and the cohesion of environmental social context and behavioral responses.

In the current study, we examined affective processing at a very basic processing level using a task that did not require any emotion labelling and included only basic emotions (e.g., happy, sad, angry), which have very distinctive facial expressions (Ekman, 1999). The child was asked to judge whether the affect in two faces displayed sequentially is the same or different (Fig. 1). The processes activated during this type of affective appraisal fall within the two lower-level social brain clusters identified by Alcala-Lopez et al. (2018), the visual sensory and limbic networks. The visual sensory network is involved in the initial pre-processing of sensory input, distinguishing facial features from other stimuli and ultimately recognizing a face as a face. This visual information is then relayed along the occipital and temporal neocortex, triggering the extraction of perceptual information pertaining to faces (Adolphs, 2009; Kanwisher and Yovel, 2006). Next, the configuration of the facial expression is analysed in relation to knowledge regarding the specific emotion signalled by a given expression (emotion recognition), using input from the limbic areas (Fusar-Poli et al., 2009).

Figure 1.

Schematic depiction of the affective appraisal task. The correct response to each stimulus is labelled here for exposition purposes only. Response to each stimulus (face or pixelated image) was assessed behaviorally from the participant’s pressing one of two buttons (“same” or “different”) on a response box during an fMRI scan. Stimulus presentation was 2 s; interstimulus-interval was randomly jittered between 2–4 s.

Different emotions are processed by distinct, at least partly dissociable, neural networks (Gallagher et al, 2003; McCabe et al., 2001). For example, the processing of neutral facial expressions has been shown to activate the inferior occipital gyrus, the lateral fusiform gyrus, and the superior temporal sulcus (Haxby et al., 2000). Similarly, increased activation in the orbitofrontal and anterior cingulate gyrus has been demonstrated for angry facial expressions but not for sad facial expressions (Blair et al., 1999). No previous neuroimaging studies have examined alterations in the BOLD response signal during facial affect processing in individuals with FASD.

In this study, we used fMRI to examine differences in BOLD signal activation during processing of five basic facial affects (happy, sad, fearful, angry, and neutral) in three groups of children: FAS or PFAS, heavily-exposed (HE) non-syndromal children, and community- and age-matched controls. Our aims were (1) to determine whether each group activated the anatomical regions associated with facial processing; (2) to compare the BOLD response in the regions comprising the visual sensory and limbic social brain networks in these groups during the presentation of faces displaying positive, negative, and neutral affect; and (3) to determine whether BOLD responses in these groups differed during their processing of three negative emotions (sad, fearful, angry) compared with a positive emotion (happy).

MATERIALS AND METHODS

Participants

Participants were right-handed children (aged 9–14 yr), who were participating in a larger prospective longitudinal cohort study of FASD (Jacobson et al., 2008). The children were born to women from a socioeconomically disadvantaged Cape Coloured (mixed ancestry) community in Cape Town, South Africa. Prevalence of heavy drinking during pregnancy and FAS in certain high-risk communities in the Western Cape province has been shown to be among the highest in the world (May et al., 2013). Mothers were recruited between 1999–2002 during their initial visit to an antenatal clinic selected based on its high incidence of heavy alcohol consumption (Jacobson et al., 2008).

Volume of each type of alcoholic beverage consumed was recorded using a timeline follow-back (TLFB) interview (Jacobson et al., 2002), adapted to reflect drinking patterns in the local community (Jacobson et al., 2008). Number of cigarettes smoked/day and illicit drug use were also reported. Mothers were invited to participate in the cohort if they drank at least 1 oz absolute alcohol (AA)/day (1 oz of AA ≈ 2 standard drinks) or engaged in binge drinking (at least 5 standard drinks/occasion). All women who reported drinking during pregnancy were advised to stop or reduce their alcohol intake and were offered a referral for help to do so. Controls were women from the same community who abstained from alcohol use or drank no more than minimally. Women < 18 years of age and those with chronic illnesses, including diabetes, hypertension, epilepsy, or cardiac problems, were excluded. Infant exclusionary criteria were major chromosomal anomalies, neural tube defects, multiple births, and seizures. Exclusionary criteria for the fMRI study included any metal implants, claustrophobia, or pregnancy.

Procedure

Following screening, the TLFB interview was repeated at mid-pregnancy and again at 1-month postpartum to reflect drinking during the third trimester. Alcohol consumption was converted to oz AA and averaged across the three interviews. Three summary measures were derived: average oz AA/day, average oz AA/occasion, and frequency of drinking (days/week). Mothers also provided information regarding sociodemographic background including their educational attainment, marital status, age at delivery, and socioeconomic status (SES; Hollingshead, 2011).

Mothers and children were transported to our University of Cape Town (UCT) Child Development Research Laboratory, where the children were administered a battery of cognitive tests. Handedness was assessed at 5 years on the Edinburgh Handedness Inventory, and left-handed and mixed left-handed participants were not included in the functional imaging study. At 10 years, IQ was assessed on the Wechsler Intelligence Scales for Children, Fourth edition (WISC-IV IQ) and visual acuity, on the functional acuity contrast test (FACT; Ginsburg, 1987).

In September 2005, we organized a clinic in which each child was independently examined for alcohol-related anomalies and growth by two FAS expert dysmorphologists (H.E. Hoyme (HEH), M.D., and L.K. Robinson (LKR), M.D.), using a standard diagnostic protocol (Hoyme et al., 2005); both dysmorphologists were blind regarding maternal alcohol history (see Jacobson et al., 2008). Case conferences including the dysmorphologists, SWJ, JLJ, and CDM were held to reach consensus regarding diagnosis of FAS and PFAS. Those who did not meet criteria for FAS or PFAS were categorized as either heavily exposed (HE) nonsyndromal or non-/minimally-exposed controls, depending on maternal alcohol consumption during pregnancy. The children were subsequently re-examined by HEH and LKR in 2009 and by a team of FAS dysmorphologists led by HEH in 2013 and 2016, at which time the FASD diagnoses were confirmed.

Our research nurse and driver transported each child and his/her primary caregiver to the Cape Universities Brain Imaging Centre (CUBIC) for scanning. There the child was familiarised with the procedures in a mock scanner, which helped minimise subsequent movement and reduce anxiety in the scanner (Meintjes et al., 2010). All the cognitive and neuroimaging sessions were conducted in Afrikaans or English, depending on the primary language of instruction used in the child’s school.

All research assistants were blind to the participant’s FASD diagnosis and history of PAE, except in the most severe cases in which FAS status was apparent. Approval for human research was obtained from the Wayne State University and UCT Faculty of Health Sciences (FHS) human research ethics committees. Informed consent was obtained from the mothers at recruitment and at each assessment visit; assent, from the children. Children received a small age-appropriate gift, and the primary caregiver received a photo of her child and compensation consistent with guidelines from the UCT FHS Ethics Committee.

Neuroimaging assessment

Data acquisition:

MRI.

Each child was scanned on a 3T Allegra MR scanner (Siemens, Erlangen Germany) using a single channel head coil. At the beginning of the scanning protocol, a magnetization-prepared rapid gradient echo (MPRAGE) structural (high-resolution T1-weighted anatomical) image was acquired in a sagittal orientation [TR = 2530 ms, TE = 1.53 ms, TI = 1100 ms, 128 slices, flip angle 7 degrees, voxel size = 1.3 × 1.0 × 1.3 mm³, scan time = 8:07 minutes]. During the fMRI protocol, functional T2*-weighted images sensitive to the BOLD contrast were acquired using a gradient echo, echo planar imaging (EPI) sequence [TR = 2000 ms, TE = 30 ms, 34 axial slices with interleaved acquisition, 3 mm thick, gap 0.9 mm, field of view 200 × 200 mm² (in-plane resolution 3.125 × 3.125 mm²), 90° flip angle].

Data acquisition:

Behavioral.

Each child completed an event-related affective appraisal task during the fMRI acquisition: 96 images (80 faces and 16 pixelated control faces) were displayed continuously in a pseudo-random order, using a randomized jittered event-related design (Fig. 1). An advantage of this 1-back paradigm is that it does not depend on naming the affect and thus minimizes the impact of deficits involved in verbal labelling. To ensure that our findings were not attributable to the working memory component of the task, we conducted a pilot study which found that PAE did not affect performance on a 1-back working memory task (Lindinger et al., 2014).

The facial images were comprised of normatively rated monochromatic photographs of Caucasian faces portraying basic emotions: positive (happy), negative (sad, angry, fearful) and neutral (Ekman et al., 1979). Although face stimuli that matched the ethnicity of our cohort would have been preferable, no suitable stimuli were available. Because all participants were recruited from the same socioeconomically disadvantaged Cape Coloured community and attended the same schools, it is unlikely that the use of Caucasian faces confounded the effects seen in this study.

The child was not required to label the affect displayed. Instead, s/he was asked to decide whether the image expressed the “same” or “different” emotion from that seen on the previous display by pressing a button on a Lumitouch MRI-compatible response box (Photon Control Inc., Burnaby, Canada) with either their index (“same”) or middle (“different”) finger. For each emotion in the task and for pixelated control stimuli, 25% of the trials required a “same” response. Numbers of male (n=7) and female (n=9) faces were the same for each affect.

The task was programmed using E-Prime software (Psychology Software Tools, Inc., Pittsburgh, USA). All stimuli were presented for 2 s and randomly jittered with 2- to 4-s intervals (in 0.5-s increments) between stimuli. Each image was presented once to avoid repetition effects on the BOLD response, which might enhance the processing of the repeated stimuli (Gilaie-Dotan et al., 2008). All data were acquired within a single fMRI session lasting 8 min 12 s.

The main behavioral outcome used to assess performance on the affective appraisal task was d-prime, which measures the number of correct button presses (hit rate), after adjustment for false alarm presses. Number of hits and false alarms were examined separately, and response bias (i.e., a tendency to respond “same” or “different”) was calculated. d-prime was also used to identify any participants whose proportion of correct responses was at or below chance (d-prime < 0.15).

Data processing and analyses

fMRI processing.

The neuroimaging data were analysed using SPM 8 and 12 (Statistical Parametric Mapping, Wellcome Department of Imaging and Neuroscience, London, UK). The first four images acquired during each session were not used in the analyses in order to allow the signal to reach steady state.

Preprocessing.

Each participant’s structural images were manually re-oriented and rotated into the AC-PC plane. Functional MR images were slice time corrected using an interleaved slice order and slice 17 as the reference, realigned and re-sliced to correct for head movement, and co-registered to the subject’s own high-resolution anatomical images. Participants with movement >3 mm in any of the three movement parameter axes (x, y, or z) were excluded from the analyses.

Due to morphological changes that occur during brain development, each subject’s anatomical images were co-registered to an adapted NIHPD paediatric template created using the SPM template-o-matic toolbox and matching age and sex of each participant in the current sample to the templates in the NIHPD database.

The images were smoothed spatially by a Gaussian filter of 5mm full-width half maximum to improve inter-subject comparability and to reduce noise. The default masking threshold in SPM was lowered from 0.8 to 0.7.

First-level analysis.

During first-level analysis, activation maps were generated for each child for each condition using general linear model analysis (Amaro et al., 2006). In the first model, each condition (happy, sad, angry, fearful, neutral, and pixelated) was set up as an individual regressor, and all the image conditions associated with negative emotions (sad, angry, and fearful) were combined to constitute a separate single regressor for negative emotions. Motor responses were not modelled separately as all stimuli required a button press.

Second-level analysis.

Second-level analysis was aimed at identifying clusters of significant activation within groups and differences in the BOLD response signal between the three groups: FAS/PFAS, HE and controls. Whole-brain analysis was conducted initially within each group for all faces contrasted to pixelated faces to determine whether all three groups activated the regions found in the literature to be associated with face perception. Pixelated faces were used for control comparisons rather than neutral faces, which have been found to be more difficult to interpret (Etcoff et al., 1992; Leppänen et al., 2004). In these analyses, we report clusters that survived at an uncorrected p < 0.0005 and cluster-level family-wise error (FWE) p < 0.05. Cortical regions were labelled using the WFU PickAtlas toolbox add-on for SPM and then reviewed and revised by an expert neuroanatomist (CMRW).

Differences in affective processing were examined between diagnostic groups in 14 regions-of-interest (ROIs) identified by Alcala-Lopez et al. (2018) as comprising the visual sensory and limbic social/affective networks (Table 1). The visual sensory network consists of three bilateral regions: the fusiform gyrus, posterior superior temporal sulcus and middle temporal V5 area. The limbic network includes the bilateral amygdala, hippocampus and nucleus accumbens, and the rostral anterior cingulate cortex and ventro-medial prefrontal cortex. Using the MarsBaR toolbox (Brett et al., 2002) ROI masks were created for each region by building spheres with a 6mm radius around each ROI’s MNI coordinates. The MarsBaR toolbox was then used to extract mean task-related BOLD parameter estimates in each of the 14 spheres, for each subject and for each condition. First, we examined whether groups showed any differences in the processing of positive, negative, and neutral faces compared to pixelated images. Second, given that “happy” has been reported as the most accurately and rapidly recognized facial emotion (Kirita et al., 1995; Leppänen et al., 2004), the three negative emotions (i.e., sad, angry, fearful) were then each separately contrasted against the positive condition and between-group differences examined.

Table 1.

Regions of interest within the visual sensory and limbic networks from the Social Brain Atlas of Social and Affective Processing

Network/anatomical region	MNI coordinates (x,y,z)
Visual sensory
Left fusiform gyrus	−42,−62,−16
Right fusiform gyrus	43,−57,−19
Left middle temporal V5 area	−50,−66,5
Right middle temporal V5 area	50,−66,6
Left posterior superior temporal sulcus	−56,−39,2
Right posterior superior temporal sulcus	54,−39,0
Limbic
Left amygdala	−21,−4,−18
Right amygdala	23,−3,−18
Left hippocampus	−24,−18,−17
Right hippocampus	25,−19-,15
Left nucleus accumbens	−13,11,−8
Right nucleus accumbens	11,10,−7
Rostral anterior cingulate cortex	−3,41,4
Ventro-medial prefrontal cortex	2,45,−15

Source: Alcala-Lopez et al., 2018

Statistical analysis

Statistical analyses of the behavioral data were conducted using the IBM Statistical Package for the Social Sciences (SPSS) version 22.0. The relation of diagnostic group (FAS/PFAS, HE, control) to each of the background characteristics and the behavioral outcome measures was examined using one-way analysis of variance (ANOVA) for continuous variables and chi-square for dichotomous variables.

Whole-brain analyses using SPM were conducted within each group separately by comparing the BOLD signal response for faces to pixelated control images using one-way ANOVA. For each of the six conditions (e.g., neutral faces > pixelated images), between-group differences in BOLD response were then examined in the 14 ROIs by subjecting the extracted parameter estimates to one-way ANOVAs using SPSS. To correct for testing of 14 regions, we used a Bonferroni corrected threshold for significance of 0.0036 (i.e. p = 0.05/14). Any significant between-group differences were examined for potential outliers by multiplying the first and third interquartile range cut-offs for each group’s parameter estimate values by 1.5 and subtracting and adding this value from/to the first and third quartile cut-offs, respectively. Any parameter estimates below or above these lower and upper limits were considered outliers, and the analyses were re-run excluding them.

Five control variables were considered as potential confounders in between-group contrasts: child sex and age at scan and maternal SES, education, and smoking during pregnancy. Propensity score matching was used to adjust for these confounders (Stürmer et al., 2006). The five control variables were regressed on diagnostic group (FAS/PFAS, HE, controls) using multinomial logistic regression to generate two propensity scores: one predicting the contrast FAS/PFAS vs. controls; the other, HE vs. controls. The ROI analyses in which significant FASD group effects were seen were then re-run to adjust for the two propensity scores in analyses of covariance. The analyses of covariance were also run on these contrasts adjusting for WISC-IV Full Scale IQ.

RESULTS

Sample characteristics

fMRI data were acquired from 76 children; 5 were excluded whose behavioral task accuracy was at chance (d-prime < 0.15); 7 due to excessive motion (movement parameters > 3 mm).

The demographic and background characteristics of the final sample of 64 right-handed children (mean age at scan: 12.2 yr ± 1.1) are summarised in Table 2. Mothers of children in the FAS/PFAS and HE groups drank very heavily (M = 9 drinks/occasion); whereas all of the control mothers abstained. Maternal smoking and marijuana use during pregnancy did not differ across the groups, and none of the mothers reported using cocaine. Although all of the IQ scores for this cohort were low, the FAS/PFAS group IQ scores were lower than those of the HE and control groups. The vision of all but five of the children (2 FAS/PFAS, 2 HE, 1 control) was in the normal range; the other five, in the near normal range.

Table 2.

Sample characteristics

	FAS/PFAS (n=18)	HE (n=18)	Controls (n=28)	F or χ2	p
Maternal characteristics
Age at delivery (years)	29.3 (6.7)	25.8 (4.8)	25.6 (5.4)	2.75	.072
Education (years)	8.9 (2.0)	9.3 (2.7)	10.3 (1.9)	2.63	.080
Socioeconomic statusa	16.2 (6.4)	22.1 (11.1)	27.1 (7.8)	8.90	<.001
Alcohol use during pregnancyb
n (%) who used	18 (100.0)	15 (83.3)	0.0 (0.0)
AA/day (oz) (users only)	1.0 (0.7)	1.0 (1.0)	0.0 (0.0)	18.68	<.001
AA/occasion (oz) (users only)	4.1 (1.5)	4.0 (3.6)	0.0 (0.0)	30.07	<.001
Frequency (days/week) (users only)	1.6 (0.8)	1.3 (1.0)	0.0 (0.0)	39.70	<.001
Smoking during pregnancy
n (%) who smoked	16 (88.9)	15 (83.3)	13 (46.4)
Cigarettes/day(smokers only)	7.9 (5.7)	6.9 (3.6)	5.7 (5.3)	0.70	.502
Marijuana during pregnancy
n (%) who smoked	1 (0.06)	5 (27.8)	1 (0.04)
Days/month (users only)	3.1 (0.0)	3.4 (2.2)	1.7 (0.0)	0.25	.794
Child characteristics
Age at scan (years)	13.1 (1.2)	11.6 (0.7)	11.9 (1.0)	13.50	<.001
Sex (% male)	55.6	50.0	39.3	1.26	.532
WISC-IV IQc	64.6 (9.2)	78.1 (17.8)	79.3 (13.5)	6.88	.002

Note. Values are mean (SD) or %.

^aBased on Hollingshead Four Factor Index of Social Status (2011).

^bEstimated for mothers of children with FAS who denied drinking by taking the mean of the other mothers of children with FAS.

^cWechsler Intelligence Scale for Children, Fourth edition.

Behavioral findings

No differences were found between the FAS/PFAS, HE and control groups on d-prime, hit or false alarm rate, indicating that all of the children were able to perform the task (Table 3). Similarly, there were no group differences in response bias.

Table 3.

Behavioral data of affective appraisal task outcome during the scan

	FAS/PFAS (n=18)		HE (n=18)		Controls (n=28)
	M	SD	M	SD	M	SD	F	p	η2
d-prime	1.03	0.55	1.23	0.52	1.25	0.55	1.00	0.374	0.032
Hit rate	0.46	0.22	0.52	0.24	0.56	0.17	1.45	0.242	0.045
False alarm rate	0.15	0.13	0.14	0.08	0.16	0.10	0.27	0.764	0.009
Response bias	0.65	0.50	0.57	0.51	0.44	0.37	1.26	0.290	0.040

M = mean; SD= standard deviation

Neuroimaging findings

Whole-brain within-group analyses.

All three groups showed a greater BOLD response signal when processing faces compared to non-face pixelated images in two crucial face processing regions, the fusiform gyrus and the left precentral gyrus (Table 4; Fig. 2). The control group also showed a greater BOLD response signal in the left medial frontal gyrus and the right cuneus, and the HE group showed a greater response in the right cingulate gyrus and the left posterior cuneus.

Table 4.

Regions within each diagnostic group showing greater activation for faces compared to pixelated control images; uncorrected p < 0.0005 and cluster-level FWE p < 0.05

Region	MNI coordinates (x,y,z)	Number of voxels	Peak t
Control group
Frontal lobe
L precentral gyrus	−34,−14,62	584	6.75
extending into the L postcentral sulcus	−40,−30,52		4.19
L medial frontal gyrus	−6,2,54	148	4.40
extending into the L mid cingulate sulcus	−8,10,46		4.24
Occipital
R cuneus	14,−98,14	959	7.61
extending into the R lingual gyrus and bilaterally into the fusiform gyrus	2,−86,−4		7.05
and the L cuneus	−14,−100,8		6.57
FAS/PFAS group
Frontal lobe
L precentral gyrus	−38,−20,58	670	6.21
Occipital lobe
R middle occipital gyrus	12,−98,12	813	6.25
extending into the R lingual gyrus and fusiform gyrus	12,−90,−6		6.09
HE group
Frontal lobe
L precentral gyrus	−40,−22,60	397	5.56
extending into the L postcentral gyrus	−44,−32,46		4.62
R cingulate gyrus	6,16,46	597	5.59
extending into the L cingulate gyrus	−2,20,40		4.68
and the L medial frontal gyrus	−4,16,48		4.61
Occipital lobe
L posterior cuneus	−10,−98,6	393	6.44
extending into the L lingual gyrus and the L fusiform gyrus	−18,−80,−12		4.74

Note. Submaxima within clusters are shown under maxima in italics.

Figure 2.

Within group analyses of the FAS/PFAS group showing activation in the right middle occipital gyrus (MNI of peak: 12,−98,12) extending into right lingual gyrus and fusiform gyrus (MNI of peak: 12,−90,−6) when contrasting all face stimuli vs pixelated control stimuli. SPM activation maps overlaid on a normalised T1 image of a single FAS/PFAS participant. Thresholded at an uncorrected p<0.0005 and cluster-level family-wise error (FWE) p<0.05. (a) coronal view; (b) sagittal view. L = left; R = right; A = anterior; P = posterior. The blue lines on the sagittal and coronal images in (a) and (b), respectively, indicate the positions of the slices shown.

A-priori ROI between-group comparisons.

Differences in the BOLD signal between neutral faces and pixelated images were greater in the FAS/PFAS group than the HE and control groups in the left fusiform gyrus and the right posterior superior temporal sulcus (Table 5; Fig. 3). No significant between-group differences in BOLD response were observed in any clusters of the limbic network.

Table 5.

Between-group comparisons of first-level contrast estimates in regions of interest showing significant FASD group differences

Network/Region	FAS/PFAS	HE	CTL	F	p	Post-hoc comparisons	Fa	p
neutral faces > pixelated images
Visual sensory
Left fusiform gyrus	0.54 (0.69)	−0.14 (0.56)	−0.24 (0.91)	8.93	<.001	FAS/PFAS>HE & CTL	6.04	.004
Right posterior superior temporal sulcus	0.36 (0.63)	−0.26 (0.56)	−0.35 (0.55)	8.65	.001	FAS/PFAS>HE & CTL	11.26	<.001
angry faces > happy faces
Visual sensory
Left posterior superior temporal sulcus	−0.41 (0.45)	−0.31 (0.46)	−0.15 (0.63)	6.58	.003	CTL>FAS/PFAS & HE	6.19	.004
Limbic
Right nucleus accumbens	−0.34 (0.53)	−0.11 (0.43)	0.24 (0.63)	6.11	.004	CTL>FAS/PFAS	7.53	.001

Note. Values are mean (SD) contrast estimates from analyses of variance for each group after exclusion of outliers and adjustment for potential confounders and after adjustment for IQ. All p-values are significant at < 0.05 after Bonferroni correction for multiple comparisons across 14 ROIs.

^aAfter adjustment for IQ.

FAS = fetal alcohol syndrome. PFAS = partial fetal alcohol syndrome. HE = non-syndromal heavily exposed.

Figure 3.

Regions of interest (ROIs) showing significant differences between diagnostic groups for the difference in BOLD signal between processing neutral faces and pixelated images. The box-and-whisker plots on the left show mean parameter estimates (as a proxy for BOLD signal difference) for neutral faces vs pixelated images by diagnostic group for the ROIs in the left fusiform gyrus (top row) and right posterior superior temporal sulcus (bottom row), respectively. The images on the right show the ROI locations. The blue line on the right-hand image indicates the position of the left-hand slice shown. Slice positions are given in MNI coordinates. FAS/PFAS=fetal alcohol syndrome/partial fetal alcohol syndrome; HE= non-syndromal heavily exposed; CTL= controls.

No significant between-group differences in BOLD response were seen for sad faces contrasted with happy in any of the ROIs. However, while control children showed similar BOLD signal for angry and happy faces in the left posterior superior temporal sulcus, the BOLD signal for angry faces was lower than for happy faces in the FAS/PFAS and HE groups (Table 5; Fig. 4). In the right nucleus accumbens, the controls also showed greater BOLD signal for angry than happy faces, while children with FAS/PFAS again showed lower BOLD signal. All of the contrasts shown in Table 5 continued to be significant after adjustment for IQ.

Figure 4.

Regions of interest (ROI) showing significant differences between diagnostic groups for the difference in BOLD signal between processing of angry and happy faces. The box-and-whisker plots on the left show mean parameter estimates (as a proxy for BOLD signal difference) for the contrast between angry and happy faces by diagnostic group for the ROI in the left posterior superior temporal sulcus (top row) and right nucleus accumbens (bottom row), respectively. The images on the right show the ROI locations. The blue line on the right-hand image indicates the position of the left-hand slice shown. Slice positions are given in MNI coordinates. FAS/PFAS=fetal alcohol syndrome/partial fetal alcohol syndrome; HE= non-syndromal heavily exposed; CTL= controls.

DISCUSSION

Although problems in social cognition have been reported by clinicians, teachers and parents, and behavioral studies have documented impairment in affective function in FASD, the neural mechanisms that mediate these deficits are poorly understood. To our knowledge, this is the first fMRI study to examine the degree to which neural networks mediating affective appraisal are altered by PAE. Happy, sad, angry, fearful, and neutral faces were presented sequentially during an fMRI sequence. Performance accuracy did not differ between the FAS/PFAS, HE and control groups, confirming that group differences in the BOLD response signal within ROIs in the affective appraisal network were not confounded by group differences in the ability to discriminate between these emotions. The children in all three groups showed a greater BOLD response in the regional networks for facial processing, including the fusiform gyrus (Vuilleumier et al., 2001), when viewing faces compared to pixelated images. These findings indicate that the ability to distinguish between the physical features of faces and objects and the ability to identify basic emotions from face stimuli are not affected by PAE.

We conducted a priori ROI analyses to examine potential valence-dependent between-group differences in the BOLD response signal. Based on a comprehensive social brain network model (Alcala-Lopez et al., 2018), we selected regions within two networks identified to be active during affective appraisal. We found that the FAS/PFAS group showed a greater BOLD response for neutral faces compared to pixelated images than the HE and control groups in two regions within the visual sensory network. One of these regions, the fusiform gyrus, is involved with the processing of static facial features and, therefore, the encoding of face identity (i.e., recognizing a face as a face; Haxby et al., 2000). The other region, the posterior superior temporal sulcus, is primarily involved with processing the changeable features of faces and contributes to the processing of both facial identity and the encoding of facial expressions (Fox et al., 2009). Given that neutral facial expressions are considered to be cognitively more difficult to interpret than positive and negative emotional expressions (Etcoff et al., 1992; Leppänen et al., 2004), our data suggest that individuals with FAS and PFAS have altered activation patterns in social-affective regions during the processing of more difficult to interpret emotions.

No significant between-group differences in BOLD response were found when sad and fearful faces were contrasted with happy faces. Whereas control children showed similar BOLD responses in ROIs in the left posterior superior temporal sulcus when processing angry compared to happy faces, the BOLD signal was smaller for FAS/PFAS and HE children when processing angry faces. Although the right posterior superior temporal sulcus shows greater activation during facial emotion recognition than the left posterior superior temporal sulcus (Specht and Wigglesworth, 2018), optimal facial emotion recognition requires processing of stimuli in both the right and the left posterior superior temporal sulci (Sliwinska and Pitcher, 2018). In addition, BOLD responses were slightly larger for angry than happy faces in control children but smaller in children with FAS and PFAS in the right nucleus accumbens, a region involved in integrating processes related to emotional salience and flexibility required for selection of an effective behavioral response (Grace, 2000). The nucleus accumbens has also been shown to be critical in the expression of anger and aggression (Ferrari et al. 2003) and in the recognition of aggressive behavioral signals (Calder et al. 2004). In a social context, anger may be displayed to curb the behavior of others who have violated social norms (Averill, 1982). Our findings provide empirical evidence that may help explain behavioral difficulties reported in children with FAS, who appear to have difficulty learning from their own inappropriate behaviors due to a failure to respond to angry cues from adults or peers who admonish their behavior.

The processes activated during the Reading the Mind in the Eyes task involve regions that are part of the higher- and intermediate-level social brain networks, which are connected and receive input from regions in the two lower level clusters examined in this study. For example, the posterior superior temporal sulcus and the fusiform gyrus (where the FAS/PFAS group showed greater BOLD response for neutral than pixelated and smaller response for angry than happy faces, respectively) have significant functional connectivity with the anterior insula (intermediate level network). In addition, the posterior superior temporal sulcus itself has also been shown to be activated during the Reading the Mind in the Eyes task and identified as contributing to the ability of perspective taking through the recognition of facial expressions (Frith and Frith, 2006). Similarly, the nucleus accumbens has connectivity to the ventro-medial prefrontal cortex, which is critical in the development of Theory of Mind (Frith et al., 2006). The ventro-medial prefrontal cortex, in turn, has strong connectivity with most hierarchically higher social brain regions, including the temporo-parietal junction, which is crucial for the higher-order social cognitive processing required for performance on the Reading the Mind in the Eyes task (Schurz et al., 2014). Our neuroimaging findings support the hypothesis that a deficit in the ability to identify facial affective cues contributes to the PAE-related deficit seen in the Reading the Mind in the Eyes performance behaviorally (Lindinger et al., 2016).

The current study extends previous findings showing that PAE-related alterations may manifest themselves in differences in the BOLD signal response in task-specific cortical regions in FASD. For example, an fMRI study on number processing found that children with FAS or PFAS recruited a more extensive range of cortical regions to perform the same number processing task than control children (Meintjes et al., 2010). More extensive activations in task-specific cortical regions by children with FAS or PFAS have also been reported in two fMRI studies of working memory despite a lack of differences in behavioral performance (O’Hare et al., 2009; Diwadkar et al., 2013).

Our findings are consistent with clinical reports that children with FASD may have difficulty “reading social cues” (Carmichael Olson et al., 1997; McGee et al., 2009; Streissguth et al., 1996). The finding that the FAS/PFAS group showed a greater BOLD response when processing neutral faces, which are known to be cognitively more difficult to interpret (Etcoff et al., 1992; Leppänen et al., 2004), compared to pixelated faces, should also hold for other more difficult to interpret emotions. Our findings also suggest greater difficulty in the appraisal of anger. These findings suggest that behavioral problems, such as social conflicts in school or trouble with the law, which have been characterised as “secondary” to the well documented cognitive deficits in FASD (e.g., Carmichael Olson et al., 1997; Streissguth et al., 1996) may be attributable, at least in part, to a primary deficit in affective appraisal.

Limitations and Future Directions

One challenge in research on affective function in FASD is that, because PAE is associated with poorer intellectual competence, behavioral performance on an emotion recognition task may be attributable to general cognitive impairment. In our previous research, we found that the effect of PAE on performance on the Reading the Mind in the Eyes task persisted after adjustment for IQ (Lindinger et al., 2016). Similarly, in this study, which focused on basic emotions, the FASD group differences all remained significant after adjustment for IQ, supporting the interpretation that the effects seen here are not dependent on cognitive proficiency. Subtle issues with visual acuity might affect neural activation in the process of discriminating emotion in faces. However, the vision of all the children was in the normal or near normal range and, given that the stimulus facial photos were displayed at a distance of less than 1 foot in front of the participant during fMRI, the children should have all been able to see them clearly.

This study would benefit from an evaluation of the functional capacity of the participants’ social skills using, for example, a teacher and/or parent questionnaire. Such data would provide external validity by confirming an association between the fMRI findings and the challenges these children experience in their daily lives. Another limitation is that findings from this community may not generalize to other populations. However, given that PAE effects on cognition in this cohort are similar to those seen in U.S. samples (e.g., Lewis et al., 2015; Dodge et al., 2009), it seems likely that these social cognitive findings can also be generalized to other populations. Replication in other cultures is, nonetheless, warranted.

This study grew out of our previous work of children with FASD, which showed an impairment in the ability to attribute mental states in others (Lindinger et al., 2016). The current study adds to our understanding of the social cognitive deficits in PAE-affected children by demonstrating alterations in brain activation patterns in response to neutral and angry faces. Previous research has shown that early intervention programs can mitigate PAE-related impairments at the behavioural and neuronal level (Petrenko and Alto, 2017; Nash et al., 2018). Given the accumulating evidence of PAE-related deficits in social cognition, intervention programs that have been shown to improve aspects of social cognition in FASD, such as the Alert Program for Self-Regulation (Nash et al., 2015), would seem warranted, as might remedial paradigms utilised in other clinical populations with social-cognitive deficits, such as autism spectrum disorders (see Tseng et al., 2020 review).

Conclusions

In the current study, all three groups of children accurately appraised all five emotions on a relatively simple affective appraisal task. However, compared with controls, the FAS/PFAS group showed greater BOLD response signal in two important affective processing regions when processing neutral faces vs. pixelated images and smaller BOLD signal response in two affective brain regions when processing angry compared to happy faces. Although the FAS/PFAS group’s affective appraisal is accurate in response to inanimate affective stimuli, these data suggest that these children will likely find it more difficult to appraise neutral affect and anger accurately in more challenging and dynamic social contexts. This failure to appraise affect correctly may, in turn, result in inappropriate responses to another’s behavior, based on an erroneous interpretation of his/her intentions (Lemerise et al., 2000).

FASD has not previously been included in the Diagnostic and Statistical Manual of Mental Disorders (DSM). However, the Fifth edition of the DSM (American Psychiatric Association, 2013) includes a new proposed diagnosis, Neurodevelopmental Disorder Associated with Prenatal Alcohol Exposure, that has been designated as “in need of further study.” One criterion for this diagnosis is impairment in social communication and interaction, including difficulty reading social cues and understanding social consequences. The current study provides the first brain-based evidence of this deficit in FASD. In terms of remediation, evidence of a deficit in distinguishing and interpreting specific emotions supports the need to develop interventions that go beyond training in global cognitive skills to focus specifically on affective appraisal and interpretation of facial emotions.