Prenatal alcohol exposure affects brain function during place learning in a virtual environment differently in boys and girls

Abstract Introduction Although performance deficits in place learning have been reported in fetal alcohol spectrum disorders (FASD), neural correlates of these deficits have not been investigated. This functional magnetic resonance imaging (fMRI) study of 57 children (41 alcohol‐exposed; 16 controls; mean age = 9.4 years; 29 boys) examined effects of prenatal alcohol exposure (PAE) on place learning in a virtual environment, the computer‐generated (CG) arena. Methods Functional magnetic resonance imaging data were acquired while children passively viewed a recording of an experimenter completing the task. Visible‐target blocks involved navigation to a visible platform. During invisible‐target blocks, the platform appeared only when the experimenter moved over it. After the scan, all children performed a post‐test during which they had to navigate to the location of the invisible platform. Results Although there were no group differences in post‐test performance for sex or FASD diagnosis, PAE in boys was associated with poorer performance and reduced activation in the parahippocampal gyrus (PHG), precuneus, posterior cingulate, frontal and temporal lobes, caudate, insula, claustrum, lentiform nucleus, and thalamus. By contrast, PAE was not associated with performance or activation in any regions in girls. Discussion and conclusion Girls and boys are known to use different navigation strategies. Boys rely more on an allocentric navigational strategy and girls more on landmarks. Poorer recruitment of the PHG, a region known to mediate allocentric navigation, in more heavily exposed boys may explain the observed dose‐dependent place learning deficit. The absence of PAE effects in girls suggests that landmark‐based navigational strategies may be less affected by alcohol exposure.


| INTRODUC TI ON
Prenatal alcohol exposure (PAE) is associated with impairment in brain structure and function that may lead to cognitive, behavioral, and neurological deficits of variable severity (Archibald, et al., 2001;Meintjes, et al., 2014;Riley & McGee, 2005;Sowell, et al., 2001). Fetal alcohol syndrome (FAS), the most severe of the fetal alcohol spectrum disorders (FASD), is characterized by small head circumference, pre-and/or postnatal growth retardation, and characteristic facial features, including short palpebral fissures, thin vermillion, and flat philtrum (Hoyme, et al., 2005).
Two of the three facial features are also seen in partial FAS (PFAS), together with either small head circumference, retarded growth, or neurobehavioral deficits. Heavily exposed (HE) individuals who lack the distinctive pattern of FAS dysmorphology are diagnosed with alcohol-related neurodevelopmental disorder (ARND) if they exhibit cognitive and/or behavioral impairment (Hoyme, et al., 2005;Stratton, Howe, & Battaglia, 1996).
In those that found sex differences, females activated the right inferior parietal lobule, right superior parietal lobule, left superior and right medial frontal gyri, and the right prefrontal cortex more than males (Grön, et al., 2000;Wunderlich, 2001), while males showed greater activation of the right and left parahippocampal gyri, left hippocampus, and left posterior cingulate (Grön, et al., 2000;Wunderlich, 2001). These sex differences in performance and patterns of brain activation suggest that males and females may use different navigation strategies.
Of particular interest for this paper is that in a T-maze spatial task, both male and female rats from the PAE group showed reference memory deficits, while working memory was impaired only in the PAE male rats (Zimmerberg, Sukel, & Stekler, 1991). Consistent with this finding, Hamilton et al. (2003) reported alcohol-related deficits in place learning in boys, although that study did not include any girls. These findings, combined with the known sex differences in strategy and brain activation during navigation, suggest that navigation in males and females may be differentially affected by PAE.
To our knowledge, this is the first study to examine effects of PAE on neural correlates of place learning. This study includes children with FAS and PFAS as well as nonsyndromal HE children. We also examine the association between navigation performance and brain activation during navigation with continuous measures of PAE and investigate whether there are sex differences in the effects of PAE on performance and brain activation during navigation. We hypothesized that alcohol-exposed children would perform more poorly than controls and would also show altered patterns of brain activation. Further, we predicted that boys would perform better than girls on this task and that there would be sex differences in alcohol-related alterations in brain activation during navigation.

| Participants
Participants were 57 right-handed 8-to 10-year-old children from the Cape Coloured (mixed ancestry) community in Cape Town, South Africa. Forty-one of these children had been heavily exposed to alcohol prenatally (Jacobson, et al., 2008). The Cape Coloured community is composed primarily of descendants of white European settlers, Malaysian slaves, Khoi-San aboriginals, and black African ancestors. The incidence of FASD in this population is exceptionally high due to poor socioeconomic circumstances and historical practices of compensating farm laborers with wine, both of which have contributed to a tradition of heavy recreational weekend binge drinking (May, et al., 2007(May, et al., , 2013.

| Procedure
The children's mothers were recruited between 1999 and 2002 at their first visit to the antenatal clinic. Each woman was interviewed, using a timeline follow-back approach adapted to reflect how pregnant women in this community drink (Jacobson, Chiodo, Sokol, & Jacobson, 2002;Jacobson, et al., 2008;Sokol, et al., 1985), regarding her alcohol consumption during pregnancy. At recruitment, the mother was interviewed regarding the incidence and amount of her drinking on a day-by-day basis during a typical 2-week period at the time of conception. She was also asked whether her drinking had changed since conception; if so, when the change had occurred, and how much she had drunk on a dayby-day basis during the preceding 2-week period. This procedure was repeated in midpregnancy and again at 1 month postpartum to provide information about drinking during the latter part of pregnancy. Volume was recorded for each type of beverage consumed each day, converted to ounces of absolute alcohol (AA) using multipliers proposed by Bowman, Stein, and Newton (1975), and averaged to provide three continuous measures of alcohol consumption around time of conception and during pregnancy: average ounces of AA consumed/day, AA/occasion, and frequency of drinking (days/week).
Two groups of women were recruited: (a) heavy drinkers, who consumed at least 14 standard drinks per week (1.0 oz AA ⁄day) on average or who engaged in binge drinking (5 or more drinks/occasion) and (b) controls whose mothers abstained or drank no more than minimally during pregnancy. The number of cigarettes smoked/ day was also recorded, as was the use of illicit drugs (days/week).
Mothers were also interviewed regarding their age at delivery, education (years completed), and marital status.
In September 2005, we organized a clinic in which each child was independently examined for growth and FAS dysmorphology by two expert FAS dysmorphologists (H.E. Hoyme, MD, and L.K. Robinson, MD) using the Hoyme et al. (2005) protocol (see Jacobson, et al., 2008). A subset of children who could not attend the clinic was examined by another FAS dysmorphologist (N. PFAS. The 14 alcohol-exposed children who did not meet criteria for either FAS or PFAS were designated nonsyndromal HE.
Written informed consent was obtained from each mother and written assent from each child. Approval for the research procedures was obtained from the Wayne State University and UCT Faculty of Health Sciences Human Research Ethics Committees.

| Neuropsychological assessment
IQ data were collected from the children on the Wechsler Intelligence Scale for Children-IV (WISC-IV) at 10 years (Diwadkar, et al., 2013;Jacobson, et al., 2011). At the 5-year follow-up of these children, we administered the Junior South African Individual Scales (JSAIS; Madge, Berg, & Robinson, 1981), which are available in Afrikaans and in English and have been normed for South African children. IQ scores from the JSAIS were strongly correlated with the WISC-IV scores for the children in this sample, r = 0.74, p < 0.001, supporting the validity of the WISC for use with this population.

| Functional MRI experimental tasks
We used a virtual navigation environment known as the computergenerated (CG) arena (Jacobs, Laurance, & Thomas, 1997;Jacobs, Thomas, Laurance, & Nadel, 1998;Thomas, Hsu, Laurance, Nadel, & Jacobs, 2001), which presents tasks similar in form and appearance to those presented by the virtual water maze task used in the Hamilton et al. (2003) study. All participants practiced the task using a desktop-based version of the CG arena before scanning and also listened to a recording of the scanner noises while lying in a mock scanner. During the scan, the CG arena task was presented using a data projector positioned in a room behind the scanner. Images were projected through a waveguide in line with the bore of the magnet onto a rear projection screen mounted behind the scanner, which the children viewed using the standard mirror system that mounts to the head coil. Auditory tones were presented using the standard Siemens headphones. The children were able to talk to the examiner using an intercom that is built into the scanner. Children could stop the scan at any time by squeezing a ball held in the left hand.
Children were scanned while passively viewing a recording of another person navigating around the CG arena room to the platform, which was sometimes visible (visible-target condition) and sometimes hidden (invisible-target condition). The task comprised four repetitions of the visible condition (duration 30 s each), followed by four repetitions of the invisible condition (duration 30 s each), with 21-s rest blocks between each active block ( Figure 1). An initial 19-s rest block preceded the first visible block, and a final 20-s rest block followed the last invisible block (total task duration 426 s). During  (Figure 2d). The location of the platform remained the same in each repetition of the invisible condition. Hence, the aim was for the subject to learn the location of the invisible platform using distal cues within the room, such as pictures on the walls.

| Behavioral performance
After completing the scan, the children performed a post-scan test in a small room adjacent to the scanner. The purpose of this test was to confirm that they had learnt the location of the platform during the invisible-target condition. The test, which comprised two 120-s trials, involved using a joystick to navigate within the CG arena to the location of the invisible platform. Time to navigate to the location of the platform ("latency") and length of path to that location were used to assess task performance. Because the post-scan assessment indicated that all participants had been attending to the task, none of the children were excluded based on performance.

| fMRI analysis
All fMRI analyses were performed using Brain Voyager QX (Brain Innovation, Maastricht, The Netherlands). Four dummy images were acquired that were excluded from all analyses. Images were motioncorrected relative to the first volume with trilinear/sinc interpolation. Images were corrected for different slice acquisition times and linear trends and temporally smoothed with a high-pass filter of two cycles/point.
Because the present data were acquired near the end of the imaging session, the data contained substantial motion artifacts.
For each subject, data from the largest continuous section with no movement greater than 3 mm displacement or 3.0° rotation were analyzed. Children were only included in the analysis if they had usable data from at least one block for each condition.
Each child's functional data were coregistered to his/her highresolution anatomic MRI, rotated into the AC-PC plane, and normalized to Talairach space using a linear transform calculated on the anatomic images. The 3.125 × 3.125 × 3 mm 3 fMRI voxels were interpolated during Talairach normalization to 3 × 3 × 3 mm 3 . Whole brain group analyses were performed with a random effect analysis of variance (ANOVA) using a generalized linear model (GLM) with predictors based on the visible-target and invisible-target experimental blocks convolved by the standard hemodynamic function. The six motion correction parameters were z-transformed F I G U R E 1 Timing diagram of the CG arena task V 30s 21s 21s 21s 21s 21s 21s 21s and added as predictors of no interest. Beta maps were created for each subject for each condition of interest (i.e., visible target and invisible target). To examine whether the difference in activation between the invisible-target and visible-target conditions differed by sex, the beta maps were analyzed at the second level using a repeated-measures ANOVA, with one within-subjects factor (i.e., invisible target vs. visible target) and one between-subjects factor (i.e., sex). The voxelwise threshold was set to p < 0.01, with cluster-size thresholding to control for multiple comparisons using the Monte Carlo simulation tool implemented in Brain Voyager (Forman, et al., 1995); minimum cluster size = 324 mm 3 . The analysis was repeated controlling for PAE.
Because previous studies have found that males and females activate different regions during navigation (Grön, et al., 2000;Sneider, et al., 2011;Wunderlich, 2001), we identified regions separately in boys and girls where differences in activation between invisible-and visible-target conditions were associated with extent of PAE. Beta maps were created for each subject for the invisible-versus visibletarget contrast and analyzed separately for boys and girls at the second level in a GLM in relation to AA/day. The voxelwise threshold was set to p < 0.05, with cluster-size thresholding (minimum cluster size =1,026 mm 3 ). In each identified cluster, mean % signal change was extracted for each subject for each condition to examine in each region whether the association of degree of PAE with the difference in activation between invisible-target and visible-target conditions survives after controlling for potential confounders.
To identify regions crucial for successful place learning, we repeated the second-level analysis, again separately for boys and girls, with path length and time to target, respectively, as continuous outcome measures. The voxelwise threshold was set to p < 0.05, with cluster-size thresholding (minimum cluster size = 1,107 and 1,134 mm 3 for path length and latency, respectively). Mean % signal change values were extracted in the clusters for each subject to examine whether associations survive after controlling for alcohol exposure. All analyses were repeated omitting those children whose mothers used illicit drugs during pregnancy.

| Behavioral performance
All 57 children in the sample were included in the behavioral analyses. We only report data from the first trial of the post-scan navigation test because performance on that trial best reflects whether the child learnt the location of the platform during the scan. Performance on the second post-scan trial, in contrast, is contaminated by any learning that might take place during the first trial. All children had practiced the task extensively before scanning and hence were comfortable with the procedure.
Among all children, as well as within boys and girls separately, neither path length nor latency differed between diagnostic groups (Table 2). Boys and girls performed similarly, both for the sample as a whole and within each diagnostic group (all ps > 0.20, Table 2). Path length to target and latency were also not related (at p < 0.20) to any of the other control variables (Table 3), except for lead exposure.
The relation of PAE to behavioral performance is shown in Table 4, both before and after controlling for lead exposure. There was an overall effect of PAE on latency, largely due to the effect within the boys. In boys, increased alcohol exposure was associated with both longer path length to target and longer latency to target.
The relation between alcohol exposure and latency to target remained significant after controlling for lead exposure. In girls, there were no significant relations between performance and alcohol exposure levels.

| Neuroimaging assessments
Boys showed greater increases than girls in activation during the invisible condition compared to the visible condition in the right middle frontal gyrus and inferior parietal lobule, and left precuneus, superior parietal lobule, lingual gyrus, middle occipital gyrus, and middle temporal gyrus (Table 5). Sex differences in activation remained significant after controlling for PAE. There were no regions where girls showed greater differences than boys in activation between the two conditions. Whereas boys showed increased activation in the invisible-target compared to visible-target condition, girls showed reduced activation during the invisible-target condition in diffuse regions across all lobes.
Because we observed differences in brain activation in boys and girls and because several of our previous imaging studies have shown that continuous alcohol measures are often more sensitive than diagnostic categorical measures (du Plessis, et al., 2014;Meintjes, et al., 2014;Woods, et al., 2015), we examined whether there were regions where differences in activation between the invisible-and visible-target conditions were associated with levels of PAE within each sex separately. In girls, there were no regions where differences in activity were related to degree of alcohol exposure. In contrast, as shown in  (Figures 3 and 4). Additional analyses showed that the findings persisted after removing the controls (all rs < −0.60, all ps < 0.01).
Analyses examining relations between control variables and differences in mean % signal change between invisible-and visible-target conditions in each of the regions where associations of activation with level of alcohol exposure were found in boys, detected only one significant correlation: Child's age at scan was related to differences in activation in the cluster centered on the right middle frontal gyrus (r = 0.27, p = 0.08; for all other correlations, p > 0.20). After controlling for child's age in this region, the association with degree of alcohol exposure remained significant (Table 6).
Across FASD diagnostic groups, greater activation increases in boys in the bilateral precuneus and a large left frontal region during the invisible-target condition compared to the visible-target condition were associated with improved place learning as reflected both by shorter path lengths (Table 7a) and by a shorter latency (Table 7b) to the target. These correlations remained significant after controlling for alcohol exposure (all ps < 0.042). In girls, there were no regions where activation increases were associated with better performance.
All findings were essentially unchanged when the analyses were rerun omitting the four children whose mothers used illicit drugs during pregnancy.

| D ISCUSS I ON
In this study, we detected no sex differences in behavioral performance. However, activation increases during the invisible-target condition compared to the visible-target condition were greater in TA B L E 2 Comparison of post-test performance by sex and diagnostic group Although most studies report better allocentric navigational performance in males, including prepubertal children (Newhouse, et al., 2007), some using computer-simulated versions of the MWM (Astur, et al., 1998(Astur, et al., , 2004Burkitt, et al., 2007;Daugherty, et al., 2014;Driscoll, et al., 2005;Mueller, et al., 2008;Nowak, et al., 2014;Sneider, et al., 2015;van Gerven, et al., 2012;Woolley, et al., 2010) have found no sex differences in performance (Jacobs, et al., 1997(Jacobs, et al., , 1998Sneider, et al., 2011). It is possible that inclusion of a probe trial, where the platform is removed unbeknownst to the participants, may have increased sensitivity in detecting sex differences in these studies as a male advantage has been demonstrated previously in probe trials even in the absence of performance differences in invisible-target trials (Nowak, et al., 2014;Sneider, et al., 2015).
Although effects of PAE on place learning have been reported in studies involving both males and females (Mattson, et al., 2010), at least three other studies have reported effects in males only-one, an animal study, in which only male rats showed impaired working memory in a T-maze (Zimmerberg, et al., 1991); another in which only male rats were examined and which showed poorer and perseverative performance on the MWM by those moderately exposed to alcohol (Hamilton, et al., 2014); and the third, a human study that also did not include girls in which boys with FAS took longer paths to find a hidden platform (Hamilton, et al., 2003).
PAE has been shown to have a greater impact on spatial location memory than on object memory. For example, Uecker and Nadal (1996; showed impairment in memory for object location (but not in the recall of the objects themselves) in children with FASD, a finding replicated by Kaemingk and Tanner Halverson (2000). Poorer performance has also been demonstrated on the visual learning task from the wide range assessment of memory and learning (Sheslow & Adams, 1990), in which subjects are asked to remember the location of designs (Kaemingk, et al., 2003;Mattson & Roebuck, 2002). The findings that PAE appears to affect location memory specifically and that, in general, males rely more on spatial strategies for navigation may explain why their performance is impaired by PAE, whereas performance of females, who tend to rely more on landmarks (object memory), appears largely unaffected.
It is possible that place learning in boys is more affected by PAE because they rely to a greater extent on brain regions more vulnerable to alcohol exposure. Notably, two of the regions affected by PAE in boys, the left precuneus and the right middle frontal gyrus (see Table 6), are among the regions showing greater activation increases during the invisible-target condition compared to the visibletarget condition in boys than in girls (Table 5). Increased activation during invisible conditions in a similar region in the precuneus was also associated with better place learning performance in boys only (Table 7), suggesting that in boys, the use of the precuneus is an efficient strategy, whereas it is not in girls. To explore this idea further, we compared brain activation in unexposed boys and girls in those regions showing alcohol-related impairment in boys. All the regions TA B L E 7 Regions where in boys smaller activation increases during the invisible condition compared to the visible condition were associated with (a) longer path lengths (p < 0.05, cluster-size corrected, all clusters > 1,107 voxels) and (b) increasing latency (p < 0.05, cluster-size corrected, all clusters > 1,134). Coordinates are Talairach coordinates of the peak voxel showing alcohol-related alterations in activation in boys were activated more by unexposed boys than by unexposed girls. The regions preferentially activated by control girls may be less affected by PAE.
A limitation of this study was the relatively small sample sizes, especially those of the HE and control groups. Sample sizes were smaller than planned because about 28% of the children (n = 16) were excluded from the functional analysis due to excessive motion. It is possible that a larger sample might have revealed a relation between PAE and activation in girls. Another limitation is that we scanned the children only while they were passively learning the virtual environment and not while actively navigating through it. We followed this procedure because it is difficult for children to navigate with a joystick and remain still in the scanner at the same time-even during the passive task we had problems with excessive motion.
Furthermore, our post-scan navigation task did not include a probe trial to test learning. Performance on that trial may have been more sensitive to sex differences in performance and could have been used to examine differences in navigational strategies.
Future studies could include an egocentric version of the arena, as well as the allocentric version used here. It would be interesting to see whether PAE affects the performance and neural correlates of boys and/or girls in the egocentric version.

| CON CLUS IONS
Girls and boys are known to use different navigational strategies.
Our findings suggest that the use of these different strategies requires the involvement of different brain regions and that whereas the regions used by boys are susceptible to PAE damage, those used by girls appear to be relatively spared. September, Emma Makin, Renee Sun, and Neil Dodge. We also thank the parents and children for their long-term participation in and contribution to the study.

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests.