Proceedings of the 2019 Annual Meeting of the Fetal Alcohol Spectrum Disorders Study Group

Abstract

The 2019 Fetal Alcohol Spectrum Disorders Study Group (FASDSG) meeting was entitled “Computational Approaches to Studying Behavioral Control and Individual Change.” The theme was reflected in the presentations of two keynote speakers: A. David Redish, Ph.D. who spoke about computational psychiatry and vulnerabilities in decision-making processes, and Kevin Grimm, Ph.D. who spoke about contemporary machine learning approaches to studying individual change. The conference presented updates from three government agencies, and short presentations by junior and senior investigators showcasing late-breaking FASD research. The conference was capped by H. Eugene Hoyme, M.D., FACMG, FAAP, the recipient of the 2019 Henry Rosett award for career-long contributions to the field.

The 2019 annual meeting of the Fetal Alcohol Spectrum Disorders Study Group (FASDSG) was held on June 21, in Minneapolis, Minnesota as a satellite meeting of the Research Society on Alcoholism. The theme of the 2019 meeting presentations was “Computational Approaches to Studying Behavioral Control and Individual Change”. The meeting was supported by FASDSG members and by a grant from the National Institute on Alcohol Abuse and Alcoholism (NIAAA) and National Institute on Child Health and Human Development (NICHD). Over 140 investigators attended, including individuals from the United States (from 23 states and the District of Columbia), Canada, South Africa, and England. The program included 2 keynote presentations, a roundtable discussion, two trainee award talks, the Rosett Award presentation, and 10 FASt Data talks (2-slide, 5-min presentations of original, “late-breaking” data). Most FASt Data presentations were given by graduate students and post docs, and 8 were supported by competitive travel awards that were ranked by the FASDSG officers. Two trainees were selected to receive the Timothy A. Cudd Award and the Kenneth R. Warren Award. The meeting included a career development roundtable, and all trainees had the opportunity to interact with senior researchers and clinicians at a networking lunch. Representatives from the NIAAA, the Interagency Coordinating Committee on Fetal Alcohol Spectrum Disorders (ICCFASD), and the Centers for Disease Control and Prevention (CDC) gave updates on FASD-related programs. A highlight of the meeting was the presentation of the Rosett Award to Dr. H. Eugene Hoyme, to recognize lifetime contributions, achievement and service in FASD research.

Keynote Presentations

A. David Redish, Ph.D., Professor of Neuroscience at the University of Minnesota, delivered the first keynote presentation entitled “The Neurophysiology of Decision Making and How Those Decisions Go Wrong”. Dr. Redish opened his talk with a discussion of the need for understanding how decision processes operate computationally, as computational dysfunction can manifest in observed behavioral and cognitive dysfunction. Dr. Redish first established that decision making is comprised of multiple processes and neural systems, and drew upon examples from maze studies in rodents demonstrating the distinction between dorsal-striatum dependent motor response learning and hippocampal-dependent place learning.

The initial segment of his talk was organized around the idea that how information is represented changes the computational operations needed to work with the information. Different decision processes are evident through distinct behaviors at choice points in mazes. For example, animals using a place strategy will tend to hesitate and sample alternative paths at choice points in the maze, whereas animals using a response strategy do not. Tolman (1939) characterized such behavior as vicarious trial and error (VTE) which he suggested allows the animal to consider future possibilities. Dr. Redish discussed data demonstrating that neural representations of places in the hippocampus can support imagination of a location other than where an animal is presently located (Jackson & Redish, 2003; Johnson, Fenton, Kentros, & Redish, 2009). During VTE, decoding of hippocampal place cell populations reveals coherent sweeps of activity ahead of the animal’s current position (Gupta, van der Meer, Touretzky, & Redish, 2012; Johnson & Redish, 2007). Dr. Redish suggested that VTE reflects indecision which is accompanied by planning and deliberation about subsequent behaviors. This requires creation of an imagined scenario that is then passed on for evaluation to other systems. He presented data showing that place field sequences are regulated by the theta rhythm such that place fields proceed from where the animal is to where the animal is going (Amemiya & Redish, 2018; Lisman & Redish, 2009; Schmidt, Duin, & Redish, 2019). Temporal delays in navigating to a goal (Schmidt et al., 2019) and increased physical length to a goal (Wikenheiser & Redish, 2015) cause the sweep associated with where the animal is going to lengthen, suggesting that these “look-ahead” sweeps include the goal. Collectively, the data from studies on spatial representations illustrate that non-human animals, like humans, imagine and evaluate simulated worlds during planning and deliberation about future behavior. Dr. Redish introduced a model in which the prefrontal cortex is involved an requesting an imagined scenario, the hippocampus generates the requested scenario, and the nucleus accumbens evaluates the scenario.

Dr. Redish went on to discuss decision processes in response-based systems such as the dorsal-striatum. In this system decision-making involves recognizing the current situation and releasing a well-learned action sequence when appropriate. Based on this framework, Dr. Redish derived three predictions: 1) cells that represent situational parameters should be observed, 2) no sweeps to the goal should be observed because the response-based strategy is procedural rather than goal-oriented, and 3) the beginning of each action sequence should be observed. Evidence for the first prediction comes from the observation that striatal neurons increase spatial information during performance of a spatial task (a multiple T-maze), but not during a non-spatial task (Regier, Amemiya, & Redish, 2015; N. Schmitzer-Torbert & Redish, 2004; N. C. Schmitzer-Torbert & Redish, 2008). With respect to the second prediction, cells in the dorsal striatum do not show increased “look-ahead” sweeps during VTE at a choice point, whereas hippocampal place cells do (Regier et al., 2015; N. Schmitzer-Torbert & Redish, 2004; N. C. Schmitzer-Torbert & Redish, 2008). With respect to the third prediction, Dr. Redish described the work of Smith and Graybiel (2013), who reported increased activity of dorsolateral striatal neurons when animals did not deliberate at a choice point, and decreased activity when they did, consistent with the involvement of hippocampus but not dorsal striatum in deliberation events. Switches in decision systems can also occur, in which an animal switches from deliberation during VTE to the automatic performance of an action.

Dr. Redish provided a definition of decision-making as the selection of an action that could be broadly applied to behavior in various species. The decision-making process is informed by past experience, current sensory cues, and goals. Different systems will represent the information differently which will influence the decision process (Redish, 2013). He then discussed additional decision systems, beyond the deliberation and procedural (response) system, including reflexive systems and instinctual/emotional systems. In reflexive systems the past experience and goals can be considered genetic and the sensory cues are the eliciting stimuli in the environment. Instinctual or emotional response systems involve learning to release species-appropriate, prewired actions based on experience with predictive stimuli; these critically depend on the amygdala. Dr. Redish presented an example from Choi and Kim (2010) in which an artificial (robotic) predator elicits fear and escape responses in rats. These responses are abolished in animals with amygdala damage. Noting Dr. Aaron Beck’s characterization of anxiety as imagination of a negative future, Dr. Redish described recent experiments from his laboratory demonstrating that anxiolytic drugs increase the likelihood that animals will retrieve food when an artificial predator is present, and increase the time spent deliberating (Walters, Jubran, Sheehan, Erickson, & Redish, 2019).

Dr. Redish subsequently provided a framework for thinking about neural representations and how such systems might be involved in neuropsychological disorders. A central idea is that memory can be represented as points in neural space and form basins of attraction to which the network will naturally move when appropriate inputs (e.g., stimuli) are present. Categorization of situations involves defining the space to be searched in planning systems and the important parameters that habit systems should respond to (Redish, Jensen, Johnson, & Kurth-Nelson, 2007). When basins of attraction are stable the appropriate state can be readily reinstated, however, pharmacological manipulations, such as dopamine agonism or antagonism, can modify the depth of basins such that it is easier or more difficult for activity to move to a basin of attraction. With multiple basins representing the world, processes that flatten out basins of attraction in the neural space make it more likely that noise could move the basin into a novel, unpredicted location. This could, for example, result in the manifestation of unusual beliefs or delusions and result in a new memory that will be retrieved in the future.

Dr. Redish ended his presentation by emphasizing the importance of understanding how a system works computationally, which can help identify how things go wrong. Dr. Redish suggested that we can rethink psychiatric disorders as “vulnerabilities in the decision process” (Redish & Gordon, 2016) which ultimately lead to the field of computational psychiatry. He applied the reliability engineering concept of “failure modes”, which is a vulnerability inherent in the machinery, to the neural systems involved in decision making. Different disorders will have unique vulnerability profiles, and individual differences in vulnerabilities can occur within a single disorder. Multiple factors can combine to lead to vulnerabilities within a system, and if the “failure modes” are known we can more effectively guide treatment.

The second keynote presentation was given by Kevin Grimm, Ph.D., Professor of Psychology at Arizona State University. Dr. Grimm’s presentation was entitled “Machine Learning Approaches to Understanding Differences in Change Trajectories” and addressed contemporary quantitative methods for the analysis of individual change over time and its determinants. Considering ongoing and potential future studies on individual change in FASD, Dr. Grimm sought to provide a broad overview of typically employed approaches to the analysis of change and its determinants, and provide a novel approach based on regression-based machine learning approaches. This work was illustrated using longitudinal data tracking children’s trajectories of mathematics skills through elementary and middle school. Dr. Grimm’s presentation included mathematical specifications for all approaches which the reader can access in Dr. Grimm’s slides (see appendix A). Dr. Grimm began his presentation by orienting the audience to the long-standing use of growth models (Wishart, 1938) as the principal method for the analysis of change in dependent variables (DVs). Growth models decompose data acquired in repeated measures studies to a smooth trajectory and error scores, in which the mathematical functions (e.g., linear, quadratic) are assumed to be the same for all individuals and the particular parameters of the function vary between individuals. Growth models are commonly specified in one of two frameworks, structural equation modeling (SEM) or mixed-effects modeling (MLM), which have become less distinct over time (see, Ghisletta & Lindenberger, 2004; Grimm, Ram, & Estabrook, 2017).

Dr. Grimm covered common approaches to studying determinants of individual differences in change within the context of growth models. The approaches reviewed were 1) a regression-based approach, where time-invariant variables are included as predictors of the latent variables, 2) a multiple groups approach, where time-invariant variables are used to create groups with growth models separately estimated within each group, and 3) a growth mixture modeling approach, where researchers examine whether there is evidence for latent classes with different change trajectories. The regression-based approach has the benefits of handling multiple variables simultaneously and the general familiarity of users in terms of the interpretation of regression models and their parameters. Limitations of the regression-based approach are the assumption of linear relationships, lack of consideration for interactions among variables, potential for inclusion of unimportant variables that complicate interpretation, emphasis on differences in means, and issues around scaling of variables. The benefits of the multiple group approach include the straight-forward interpretation of model parameters and the ability to examine mean, shape, variance, and covariance in the trajectories differences across groups. Understanding the change process with this approach is, however, more complicated. The main limitation of the multiple group approach is that only one grouping variable is typically used which limits the scope of evaluation to main effects and can create confusion regarding effects of correlated variables. The benefits of growth mixture modeling are that the approach allows for the study of mean, shape, variance, covariance differences in the change process, and the inclusion of predictors of class membership. Limitations of growth mixture modeling include model selection, difficulty in justifying specific parameter restrictions, and covariate inclusion.

Dr. Grimm then illustrated two regression-based machine learning approaches to the study of differences in change: Recursive partitioning and Lasso regression. Recursive partitioning (Breiman, Friedman, Stone, & Olshen, 1984; Morgan & Sonquist, 1963) with a continuous DV and many independent variables involves repeatedly partitioning the data into nodes based on a collection of explanatory variables until predetermined stopping criteria have been met. The resulting prediction model can be read as a sequence of conditional (if-then) statements. This exploratory approach, due to repeatedly splitting the data into nodes, inherently searches for nonlinear and interactive associations. Recursive partitioning was combined with SEM by Brandmeier, von Oertzen, McCardle, and Lindenberger (2013). In this approach an SEM provided by the researcher is fit in each node and the split that minimizes the −2 log likelihood (−2LL) is retained. This process is repeated for each node until stopping criteria are met (e.g., non-significance of the −2LL change). Lasso regression (Tibshirani, 1996) introduces a penalty on the size of the parameter of a regression model. The penalty added to the fit function is $\lambda {\sum }_{p=1}^q\left|{\beta }_p\right|$, where λ is the penalty and ${\sum }_{p=1}^q\left|{\beta }_p\right|$ is the sum of absolute values of regression coefficients. Greater penalties result in certain parameters being reduced to 0, effectively eliminating them effects from the model. This results in variable selection based on survival of effects. In lasso regression, a series of penalties are considered ranging from no penalty, where all effects are retained, to a large penalty, where all effects are eliminated from the model. The optimal penalty size is often determined through cross-validation or using a fit index that accounts for model complexity (e.g., Bayesian Information Criterion). Lasso regression is being implemented more often because it has been shown to outperform other variable selection algorithms, such as forward selection and stepwise regression (Hastie, Tibshirani, & Tibshirani, 2017). Lasso regression was combined with SEM by Jacobucci, Grimm, and McArdle (2016), where specific path coefficients can be penalized as the penalty function is added to the maximum likelihood fit function.

Dr. Grimm illustrated examples of recursive partitioning and lasso regression to study determinants of change with longitudinal data from the Early Childhood Longitudinal Study – Kindergarten (ECLS-K) Cohort of 1998 (National Center for Education Statistics, 2001). Data on children, parents, teachers, and schools were collected on more than 21,000 children from kindergarten through 8^th grade. The sample was reduced to 250 cases with data from the first four assessments. The DV of interest was mathematics scores and potential covariates included demographic and school-readiness indicators. All model approaches were based on a linear growth model for mathematics growth. For the recursive partitioning approach, the data were partitioned based on the child’s fine motor skills and then the data from those children with higher fine motor skills were partitioned based on maternal education. Children who had higher levels of fine motor skills and mothers with more education had a growth process associated with the highest levels of mathematics performance at kindergarten entrance, followed by children with higher levels of fine motor skills and mothers with less education, and then children with lower levels of fine motor skills. Thus, only fine motor skills and maternal education were seen as key predictors when examining the determinants of mathematical growth. In contrast, using lasso regression the retained effects on the intercept included maternal education, paternal education, poverty status, impulsivity, and fine motor skills. No effects were retained on the slope. Both approaches reduced complexity of the model compared to traditional approaches and identified some common important variables for explaining individual change in math scores such as fine motor skills and maternal education. The discrepancy in the variables selected (i.e., poverty status, paternal education, impulsivity) through the different approaches can be attributable to the variable selection criterion and the nature in which the variable impacted the model (recursive partitioning inherently searched for interactive effects, which were not examined in lasso regression). Subsequent research on the nature of these associations should entail the estimation of refined models to examine interactive effects in standard SEM.

Dr. Grimm ended his presentation with a discussion of some current challenges and recent developments in the field of machine learning for the analysis of individual change. A key challenge is determining what is an important change in model fit, as all approaches depend upon comparisons of model fit but use different procedures for defining criteria for significance.

Timothy A. Cudd Award

Warton, F.L., (University of Cape Town), Meintjes, E.M., Warton, C.M.R., Molteno, C.D., Carter, R.C., Undinger, N.M., Wintermark, P., van der Kouwe, A., Jacobson, J.L., and Jacobson S.W. Maternal choline supplementation in pregnancy mitigates adverse effects on regional brain volumes in infants heavily exposed to alcohol in utero.

Dr. Fleur Warton received the 2019 Timothy A. Cudd Award in recognition of her work describing effects of choline supplementation on brain structure in prenatal alcohol exposed (PAE) humans. Early post-mortem human studies identified that the corpus callosum (CC) was often malformed or even missing after PAE. Magnetic resonance imaging (MRI) allows examination of live tissues and can also allow identification of more subtle effects. MRI studies have identified many brain regions as being smaller after PAE. Dr. Warton described outcomes from a randomized, double-blind maternal choline supplementation trial in a community in Cape Town, South Africa. A total of 69 heavy drinking pregnant women were recruited before 23 weeks of gestation: 35 placebo, 34 choline. They were asked to consume a non-alcoholic drink twice each day from enrolment to delivery. Each drink contained 1 g choline or placebo. Treatment arms did not differ on sample characteristics (such as maternal age, education, socioeconomic status, or amount of alcohol used) but the choline arm contained significantly more girls and also included more mothers that used marijuana. On average, infants were small at birth (26–28 %ile for weight and head circumference) but by 6 months infants of the choline-treated mothers had reached the 44%ile. At 6 months children from the choline-treated mothers were twice as likely to meet criterion for eyeblink conditioning and at 12 months these children performed better on the Fagan Test of Infant Intelligence than children from mothers given the placebo (Jacobson et al., 2018). Structural MRI scans were acquired within one month of birth in unsedated infants using a 3T MRI and a multiecho FLASH protocol. Usable scans from 50 alcohol-exposed infants (27 choline-treated; 23 placebo-treated controls) were analyzed. Subcortical regions of interest (caudate, putamen, hippocampus, thalamus, cerebellar hemispheres) and CC were manually segmented. Volumes were normalized for total intracranial volume and compared between groups using analysis of covariance. Maternal socioeconomic status, smoking, marijuana and methamphetamine use, and infant gestational age at scan were included as potential confounders for those regional volumes with which they were related at p < 0.10. After statistical adjustment for potential confounders, larger volumes were found in the choline-treated group in four brain regions: left (p = 0.013) and right (p = 0.042) thalamus, CC (p = 0.027), and right putamen (p = 0.079). The caudate showed nominal differences that did not survive correction for age, sex, and marijuana exposure: left (nominal p = 0.018) and right (nominal p = 0.018). These data are the first to provide evidence that choline supplementation initiated in mid-pregnancy can protect against and/or repair structural brain deficits associated with PAE. The intervention was feasible and effective even in a highly disadvantaged community. Overall, the study is an important breakthrough in treatments of effects of maternal drinking during pregnancy.

Kenneth Warren Merit Award

Holman, P.J., (University of British Columbia) Ellis, L., Chao, A., Fung, C., Grewal, R., Haghighat, S., Morgan, E., Yu, W., Raineki, C., and Weinberg J. Effects of prenatal alcohol exposure and early life adversity on the neurobiology of social behavior during adolescent development.

Mr. Parker Holman received the Kenneth Warren Merit Award in recognition of his body of work on the effect of prenatal alcohol exposure (PAE) on social behaviors and the modifying effects of early life adversity (ELA). PAE affects many social behaviors including interaction with early life caregivers (including the mother) and this alone, or in combination with other early insults, may alter the trajectory of social behavior development. Impaired social behavior in individuals with a Fetal Alcohol Spectrum Disorder (FASD) is known to have widespread implications for function in other domains and may contribute to difficulties within the school environment, social rejection, trouble with the law, and later mental health problems. Early-life adversity can also modulate social behavior development, and individuals with PAE are more likely to experience ELA. Effects of alcohol on social behaviors (play and social recognition) were assessed using a liquid diet model of PAE to expose fetal rats in utero. Play behavior in triads of adolescent rats was aberrant; in triads with 2 controls and one PAE animal, the control animals preferentially played with each other. Social recognition was also disrupted, particularly in males, and this was associated with increased binding at the oxytocin receptor. A second set of studies examined the role of ELA, using reduced bedding material in the cage of dams with litters aged postnatal day (PD)8–12. In this environment the dam is unable to build an adequate nest and spends more time moving pups and interacting roughly with the litter. A social recognition test was done on 2 days either in early adolescence (PD30) or later adolescence (PD45). On PD30 ELA females did not show social recognition on test day 1 but did on test day 2, whereas on PD45 all females displayed social recognition, suggesting a delay in the development of this behavior. By contrast, in males on PD30, PAE animals did not show social recognition on test day 1 regardless of housing condition, but did demonstrate social recognition on test day 2. At PD45, all males exposed to ELA had a deficit in social recognition on test day 1, but control animals demonstrated social recognition on test day 2. However, PAE-ELA animals improved on day 2, but continued to show a behavioral deficit. Brains were assayed for c-fos mRNA expression, a proxy for neuronal activity, after test day 2. PAE males had hyperactivity in amygdala, while ELA males had hypoactivity in the PFC. PAE males and all ELA males displayed low activity in the lateral septum. These regions are the same as regions that had changes in binding at the oxytocin receptor. Hypothalamic expression of oxytocin was also blunted in 45-day old PAE and ELA males. Thus, the data demonstrate both unique and interactive effects of PAE and ELA. Outcomes were sexually dimorphic with greater effects in males than females. Interestingly, the delays in social recognition memory often improved with experience, and the observed behavior changes were associated with altered patterns of neural activity and reductions in hypothalamic oxytocin expression. Taken together, the results highlight that PAE and ELA directly impact multiple aspects of social behavior and its underlying neurobiology.

FASt Data Presentations

Keum, D. (University of Maryland School of Medicine), Krahe, T.E., and Medina, A.E. Developmental alcohol exposure causes reduced excitation probability of excitatory neurons in visual cortex layer 2/3.

Dr. Dongil Keum described preliminary findings on the effect of third trimester-equivalent alcohol exposure on intrinsic membrane properties of the pyramidal neurons in layer M/III of mouse primary visual cortex. Mice from both sexes received 5 g/kg of alcohol (25% injected intraperitoneally) or saline (control) on postnatal days (P) 5, 7 and 9. Another group of mice from the same litters remained undisturbed (naïve). Animals were euthanized between P25-P38 and slices prepared for whole cell patch clamp electrophysiology. Recordings were obtained from excitatory neurons in layer 2/3 of primary visual cortex. Current clamp results revealed a 30% reduction in the number of action potentials (APs) in the alcohol and control groups compared with naïve animals. Cell membrane properties (cell capacitance, input resistance, and resting membrane potential) and AP properties (threshold, half duration, rising and decay time) were not significantly changed. He suggested that the effect on APs may be due to alterations in potassium channels. These findings may be relevant to children with FASDs who can display amblyopia-like visual impairments even in the absence of optical and oculomotor abnormalities.

Wetherill, L. (Indiana University School of Medicine), Boes J., Coles, C., Jones, K.L., Kable, J.A., Mattson, S.N., Nudelman, K., Parnell, S.E., Schwantes-An, T.H., Sowell, E.R., Suttie, M., Wozniak, J.R., Riley, E.P., Foroud, T., and CIFASD. Huntingtin (HTT) coding variants are associated with increased risk for spatial working memory errors, lower brain volume, and facial dysmorphology in prenatally exposed individuals.

Dr. Leah Wetherill described results from a whole exome sequencing study from people enrolled in the Collaborative Initiative on Fetal Alcohol Spectrum Disorders (CIFASD). The study tested if gene variants contributed to susceptibility of being diagnosed with FAS compared to people who were exposed but had no physical features of alcohol exposure. The study focused on the Huntingtin (HTT) gene because it is associated with brain development and autophagy in rodents. All carriers of any rare coding variant in HTT (n=35) were compared with non-carriers (n=119) on brain volume, spatial working memory, and degree of facial dysmorphism from computer-based analysis of 3D images in African Americans (AA; sample size of carriers in Caucasians was too small for analysis at this time). Carriers had significantly smaller cerebellum volume and committed more errors on a spatial working memory task than non-carriers; caudate volume was not different between the two groups. There was more severe facial dysmorphology in AA carriers (n=21) versus non-carriers (n=51). Together these results show a role for genetic variation in the etiology of FASDs and, in particular, suggest that genetic variation in HTT plays an important role in the etiology of deleterious effects associated with FASD.

Olguin, S. (University of New Mexico) and Brigman, J. Prenatal alcohol exposure impairs cognitive control on a touch-screen five-choice continuous performance task.

Ms. Sarah Olguin described a novel touch-screen task that tests both attention and cognitive control at the same time: a 5-Choice Continuous Performance Task (5C-CPT). Pregnant mice were exposed to alcohol using the New Mexico Alcohol Research Center (NMARC) model of consuming alcohol or saccharin in drinking water for four hours in the dark cycle throughout gestation, which achieves blood alcohol concentrations ~ 90mg/dL. Eight-week-old prenatal alcohol- (PAE) or saccharin- (control) exposed offspring were trained to touch a white square in 1 of 5 available windows (“target” trials) as the duration of stimulus availability was systematically reduced. Once mice responded accurately at a stimulus duration of 2 sec., “non-target” trials were added in which all 5 squares were illuminated. Mice were only rewarded on non-target trials for withholding a response. PAE mice acquired the target trial task at similar rates to control with no effect of sex or treatment at any stimulus duration. When tested on a version with non-target trials, PAE mice had significantly more false alarms, i.e., responded during a non-target trial. To examine the effects of PAE on neurophysiological activity during attention and cognitive control, a separate cohort of male and female PAE and controls were fitted with dura-resting skull screws targeting medial prefrontal, parietal, and motor cortices. EEG-like oscillatory activity was recorded during a 5C-CPT test. Results showed that in control animals theta (4–10 Hz) oscillatory signaling in the posterior parietal cortex was decreased while beta (13–25 Hz) power was increased during non-target trials. Preliminary findings in PAE animals (3–4 per group), showed that theta oscillatory signaling in the medial prefrontal cortex did not show the expected decrease in the non-target trials. These results show that alterations in cortical activity when cognitive control is required may underlie attentional deficits associated with PAE.

Hasken J. (University of North Carolina at Chapel Hill) and May P.A. FASD diagnosis and the quantity and frequency of alcohol use as measured by biomarkers and self-report in antenatal clinics in South Africa.

Ms. Julie Hasken reported findings from a study comparing biomarkers and self-reported alcohol consumption with FASD diagnosis in a South African cohort. 112 mothers in a prenatal clinic in the Western Cape Province completed an Alcohol Use Disorders Identification Test (AUDIT) and provided bloodspots and/or fingernail clippings for quantification of phosphatidylethanol (PEth) and ethyl glucuronide (EtG), respectively. A value of ≤8ug was considered indication of alcohol consumption for both PEth and EtG; children from mothers with levels below the cut-offs were considered “Unexposed”. Children were followed for physical growth, dysmorphology, and development, and the revised IOM diagnostic guidelines were used to assign diagnoses. Children were classified as FASD/Deferred diagnosis, Not FASD & Exposed, or Not FASD & Unexposed. Among mothers with children with FASD/Deferred diagnosis, 91.7% were identified as alcohol consuming by PEth and/or EtG while 53.9% of mothers with children without FASD were identified as alcohol using by PEth and/or EtG. Among women who had children with FASD, the mean value for PEth was 226.3 (SD=223.3, range: 0–825) and the mean EtG value was 96.2 (SD=118.0, range: 0–331). These women reported 6.7 (SD=4.3) drinks per drinking day (DDD) in the 3 months prior to pregnancy. Among women who had drank but had children without FASD, the mean value for PEth was 49.6 (SD=61.9, range: 0–596) and they reported 4.3 (SD=3.7) DDD in the 3 months prior to pregnancy. After adjusting for maternal age, BMI, gravidity, household income, and week of specimen collection, the odds of having a child with FASD/Deferred diagnosis was 15.534 (95% CI: 4.638 – 52.029) times as likely when the mother was positive on both biomarkers and was 10.090 (95% CI: 2.562 – 39.742) times as likely if the mother scored >4 on the AUDIT. The odds for having a child with FASD/deferred diagnosis was 37.368 (95% CI: 4.462–312.883) as likely when the mother reported alcohol consuming 3 months prior to pregnancy regardless of biomarker status. Together these results suggest that South African women in these communities report alcohol use honestly and accurately, and that both PEth & EtG are needed to most accurately identify children at risk for FASD.

Breit K. (San Diego State University) and Thomas J.D. Prenatal alcohol and THC exposure via vapor inhalation alters motor development and coordination.

Dr. Kristen Breit described her work using a novel model in which pregnant Sprague-Dawley rats were exposed to vaporized ethanol (EtOH; 68 mL/hr for 3 hr, air), delta-9-tetrahydrocannabinol (THC; 100 mg/mL, vehicle) via e-cigarette, or the combination throughout most of the pregnancy. An estimated 3–10% of pregnant women report using cannabis and approximately half of those women also consume alcohol. Blood tests on the dams at several time points throughout the pregnancy showed that concurrent THC and EtOH exposure resulted in significantly higher blood alcohol concentrations than EtOH alone (~100 mg/dL cf. ~60 mg/dL), particularly following repeated exposure days. Food and water intake were unaffected, and gestational length, litter size, and pup body weights were also unchanged. Offspring were tested for early motor development (sensorimotor maturation) using a hanging wire test each day between postnatal day (PD) 12 and PD 20 and for later motor coordination using a parallel bar test from PD 30–32. THC alone or in combination with EtOH advanced sensorimotor maturation as animals in these groups were able to hang on the wire for 30 sec and/or raise a hindlimb onto the wire at younger ages than animals not exposed to THC. However, the same animals showed significant impairment in motor coordination later during adolescence compared with the animals not exposed to THC. EtOH alone did not alter performance on either task. These data suggest that prenatal cannabis disrupts behavioral development, which has important implications for individuals exposed to THC prenatally.

Gursky, Z.H. (University of Delaware) and Klintsova, A. Thalamic nucleus reuniens is highly sensitive to alcohol-induced neuron loss in third-trimester equivalent period of rat.

Mr. Zachary Gursky described data showing that the midline thalamic nucleus reuniens (Re) is sensitive to alcohol exposure. Re is essential in coordinating activity between the prefrontal cortex (PFC) and hippocampus (HPC), and this circuit likely plays an important role in cognition. Male and female rat pups were exposed to binge alcohol (5.25 g/kg) on postnatal day (PD) 7 and cell death and cell numbers were assessed stereologically on PD7, PD11, or PD72. A 30-fold increase in cell death was seen 12 hours after alcohol exposure, and on PD11 and PD72 there was a ~30% cell loss in Re. A follow-up study of NeuN-labeled neurons and Hoechst33342-stained cell nuclei in female rats demonstrated that 5.25 g/kg/day alcohol on PD4–9 produced a significant loss of neurons, but not non-neuronal cells, in Re that persisted into adulthood but did not affect cell number in an adjacent thalamic region, the rhomboid nucleus (Rh). A third study compared two doses of alcohol: “high-dose” 5.25 g/kg/day (BAC ~400 mg/dL) or “moderate-dose” 3.00 g/kg/day (BAC ~150 mg/dL). Preliminary data again showed a ~30% neuron loss in Re, at both doses, with no effect on the number of Iba1-positive microglial cells. Taken together, the data suggest that Re is vulnerable to developmental alcohol exposure and this may play a role in the previously identified alcohol-induced effects in the PFC-HPC circuit.

Lindinger, N.M., (University of Cape Town), Jacobson, J.L., Molteno, C.D., Malcolm-Smith, S., Meintjes, E.M., and Jacobson, S.W. Lasting impairment or developmental delay: Examining higher order social-cognitive functioning from school age through adolescence in children with FASD.

Dr. Nadine Lindinger showed test-retest data on a “Reading the Mind in the Eyes” (RME) task done with 62 children from the Cape Town Longitudinal Cohort; children were categorized as Fetal Alcohol Syndrome (FAS), partial FAS (PFAS), non-syndromal heavily exposed (HE), or control. They had been tested on RME at 11 years of age and were re-rested 6 years later during adolescence. There was very strong test-retest reliability on overall RME accuracy (r = 0.61, p<.001). Specific aims of the study were to determine whether alcohol-related impairment on the RME is indicative of developmental delay or is a persistent deficit and whether alcohol-related RME deficits differed by valence; positive (7 items), negative (12 items), neutral (9 items). RME performance was examined in relation to FASD diagnosis using ANOVAs. For RME accuracy, the younger children with FAS (n=8) and PFAS (n=15) performed more poorly than HE (n=19) and control children (n=20; all ps<.001), even after control for confounders. At the older age, individuals with PFAS appear to catch up in their ability to interpret the RME task and performed similarly to HE and control groups (ps>.05), whereas the FAS group continued to perform more poorly compared to the other groups (all ps<.05). No deficits were seen in relation to FASD group for the positively-rated items at 17 years of age (ps>.10). For negative and neutral items, younger children with FAS and PFAS performed more poorly than HE and control children (ps<.05), but by adolescence only the FAS group continued to perform more poorly than the HE and control groups (ps<.005). Accuracy in evaluating negative and neutral items improved with age for all participants, and adolescents with PFAS caught up, suggesting developmental delay rather than a persistent deficit. By contrast, mentalizing impairment persisted from 11 to 17 years in individuals with FAS. The RME effects persisted after adjustment for IQ and executive function at both ages, despite alcohol-related group differences in these domains, suggesting that alcohol deficits in this social cognitive domain are relatively independent of other cognitive and executive impairments.

Kwan, S.T.C., (University of North Carolina at Chapel Hill) Plesha, C., Huebner, S.M., Wolfenden, E.E., Kling, P.J. and Smith, S.M. Maternal iron status influences placental inflammatory profile in a rat model of prenatal alcohol exposure.

Dr. Sze Ting (Cecilia) Kwan’s talk discussed modulation of prenatal alcohol exposure (PAE) effects by nutritional status of the mother. The nutritional focus was on iron status as iron deficiency is common among pregnant women and PAE can also alter maternal iron status. Dr Kwan’s work examined the placenta as one role is to supply adequate nutrition to the fetus, thus, effects on this organ may affect fetal growth. The work addresses the hypothesis that maternal ID and PAE will impair placental nutrient supply efficiency and the corollary that improving maternal iron status will attenuate outcomes. Pregnant Long-Evans rats were fed an iron deficient (2–6 ppm), iron-sufficient (100 ppm) or iron fortified (500 ppm) diet from gestational day (GD) 13.5–20.5. Animals were gavaged with alcohol (5g/kg) or maltodextrin (MD) from GD13.5–19.5. On GD20.5, placentas and fetuses were weighed, and placental cytokine expression was quantified using RT-qPCR. A mixed linear model was used to analyze growth outcomes and 2-way ANOVA or Kruskal-Wallis tests were used to analyze cytokine expression. In iron deficient animals PAE (vs. MD; P < 0.05) significantly reduced fetal weight and placental efficiency (ratio of fetal body weight to placental weight), and increased the ratio of the pro-inflammatory cytokine IL-6 to the anti-inflammatory cytokine IL-10. In iron fortified PAE animals placental efficiency and cytokine levels were normalized, but fetal body weight remained significantly lower than control animals. Thus, the effects of PAE on body weight may be independent of maternal iron status, but a dietary manipulation can mitigate the effect of PAE on placental nutrient supply and inflammatory status.

Inkelis, S.M., (San Diego State University), May, A.C., Thomas J.D., and Tapert, S.F. Maternal alcohol use during pregnancy predicts child alcohol sipping in preadolescence.

Ms. Sarah Inkelis examined the relationship between prenatal alcohol exposure (PAE) and childhood alcohol sipping behavior. Previous studies have shown that PAE is associated with higher risk of an alcohol use disorder (AUD) at 21 years of age and that ~45% of adults with a Fetal Alcohol Spectrum Disorder (FASD) have had problems with alcohol in their lifetime. Data were obtained from the Adolescent Brain Cognitive Development (ABCD) Data Repository via the NIMH Data Archive. At the time, the database contained information from 4,521 children aged 9–10 years old as part of the baseline visit. Children were classified as PAE (n=71) if their parent noted drinking ≥1 drink per week on average during pregnancy. Using propensity scores, non-exposed control children (CON; n=71) were matched to PAE children on gender, age, race, parent education, family history of alcohol problems, and ease of access to alcohol. Children were classified as Sipping (n=47) vs. No Sipping (n=95) based on endorsement of having ever sipped/tried alcohol. Logistic regression was used to examine the independent effects of age, gender, race, parent education, family history, alcohol accessibility, and PAE on sipping status. Spearman correlation and median analysis were performed within the Sipping group to examine the relationship between the number of times they had tried alcohol (i.e., sipping occurrences) and demographic variables. There were significant effects of PAE (X2(1)=3.987, p=0.046) and trend-level effects of alcohol accessibility (X2(1)=3.159, p=0.076) on sipping status. Children with PAE had significantly greater odds of having sipped/tried alcohol than CON (Odds Ratio (OR)=2.154). Easier access to alcohol was also related to greater odds of sipping at the trend level (OR=1.317). Within the Sipping group, median number of sipping occurrences did not differ based on demographic variables (ps>.515). Sipping occurrence was similar across children with and without PAE (Median=2), and did not differ based on alcohol accessibility. Thus, children with PAE were around twice as likely to have tried a sip of alcohol than children whose parents did not report alcohol use during pregnancy. Additionally, ease of access to alcohol may relate to increased likelihood of trying/sipping alcohol. Further evaluation of both developmental and environmental factors that contribute to early life alcohol use patterns may inform risk factors for, and resilience against, harmful drinking later in life.

Heroux, N.A., (University of Delaware), Horgan, C.J., Rosen, J.B., and Stanton, M.E. Cholinergic rescue of neurobehavioral insult after third-trimester equivalent alcohol exposure in the rat.

Mr. Nicholas Heroux presented new data based on earlier work showing that alcohol exposure (AE) on postnatal day [PD] 4–9 in rats disrupts long-term context memory formation, resulting in abolished post-shock and retention test freezing in a prefrontal and hippocampal-dependent variant of contextual fear conditioning called the Context Preexposure Facilitation Effect (CPFE; Heroux et al., 2019). CPFE is a three day test: day 1 the animal explores and encodes the environment, day 2 the animal receives a footshock, and day 3 the animal is re-exposed to the environment. Freezing time is measured for day 2 (post-shock freezing) and day 3 (retention freezing). The behavioral impairment of AE animals is accompanied by disrupted medial prefrontal, but not dorsal hippocampal expression of the immediate early genes (lEGs) c-Fos, Arc, Egr-1, and Npas4 during context exposure. The current experiment sought to determine if systemic administration of the acetylcholinesterase inhibitor physostigmine (PHY) prior to context learning would rescue prefrontal IEG expression and freezing in the CPFE. From PD4–9, Long-Evans rats received oral intubation of ethanol (EtOH; 5.25 g/kg/day, split into two daily doses) or underwent sham intubation (SI). Animals received a systemic injection of saline (SAL) or PHY (0.01 mg/kg) just prior to context learning on day 1 in the CPFE from PD31–33. A subset of rats were sacrificed 30 min after context learning to assay changes in IEG expression in the medial prefrontal cortex (mPFC), dorsal hippocampus (dHPC), and ventral hippocampus (vHPC) via qRTPCR. Administration of PHY rescued both post-shock (day 2) and retention (day 3) test freezing in the CPFE in ETOH animals without altering performance in SI animals. Replicating previous findings, ETOH-SAL animals had significantly reduced mPFC but not dHPC expression of c-Fos, Arc, Egr-1, and Npas4. PHY specifically rescued c-fos expression in mPFC of AE animals and non-specifically increased Arc and Npas4 in mPFC and c-fos in hippocampus of AE and SI animals. These findings implicate impaired cholinergic and prefrontal function in cognitive deficits arising from 3^rd-trimester equivalent alcohol exposure and suggest a potential therapeutic acute treatment.

Rosett Award

The 2019 Rosett Award recipient was Dr. H. Eugene Hoyme, M.D., FACMG, FAAP of The Sanford School of Medicine at The University of South Dakota and The University of Arizona College of Medicine, who was honored for his career-long contributions to the field. His presentation, entitled “Fetal Alcohol Spectrum Disorders: Early and Accurate Diagnosis of the Continuum of Teratogenic Effects of Alcohol’, focused on the evolution of diagnostic guidelines for FASD, components of the diagnostic dysmorphology approach to FASD, physical traits in FASD and their correlates, and new diagnostic tools.

An Institute of Medicine (1996) report described guidelines for four diagnostic categories, including Fetal Alcohol Syndrome (FAS), partial FAS (PFAS), alcohol-related birth defects (ARBD), and alcohol-related neurodevelopmental disorder (ARND). Dr. Hoyme noted, however, that a universally accepted diagnostic system for the full continuum of FASD does not exist. Three systems have been used: The University of Washington 4-Digit Diagnostic Code (https://depts.washington.edu/fasdpn/htmls/4-digit-code.htm), the Canadian Diagnostic Guideline for FASD (Cook et al., 2016), and the Hoyme et al. (2005) guidelines which sought to operationalize the four diagnostic categories described by the IOM. The Hoyme et al. (2005) guidelines established specific growth and dysmorphology criteria for diagnoses in the original four IOM FASD “umbrella” categories. These criteria have been used most broadly in international prevalence studies of FASD. By 2015, the authors of Hoyme et al. (2005) had cumulative experience with the evaluation of >10,000 children for potential FASD in the US and overseas. Further, investigators in the broader field of FASD research had delineated more specific, data-driven cognitive and neurobehavioral phenotypes in children with FASD. Thus, it was recognized that increased diagnostic specificity could be possible with additional revisions to the guidelines. New NIAAA-sponsored studies in the United States were also underway through the Collaboration on FASD Prevalence (CoFASP) to establish the prevalence of FASD among school-age children using active case ascertainment methods. The studies were an impetus to apply new, comprehensive expert-driven criteria to the US prevalence studies.

The CoFASP guidelines, published by Hoyme et al. (2016), were prepared over a 12-month period through a series of conference calls and face-to-face meetings at the offices of NIAAA, and were based upon a thorough review of the literature and the collective experience of the CoFASP investigators. The collaborators included dysmorphologists, clinical geneticists, psychologists, neuropsychologists, educational diagnosticians, and epidemiologists. The original four IOM diagnostic categories were retained. Diagnostic criteria for FAS were updated to allow consideration of neurophysiological abnormalities. The CoFASP guidelines also detail cognitive and neurobehavioral guidelines for diagnoses of FAS, PFAS, and ARND, and require specific cognitive and/or neurobehavioral disabilities for FAS and PFAS diagnoses. The criteria for PFAS have also been simplified to further differentiate PFAS from FAS. Specific definition of significant prenatal alcohol exposure is also provided. Beyond the CoFASP guidelines, collaborative work by a larger number of experts toward a common set of diagnostic criteria that can be universally applied is ongoing.

Dr. Hoyme emphasized that diagnosis of FASD should be multidisciplinary and should be assigned after gathering data from multiple sources including dysmorphology examination, behavioral assessment, psychological and neuropsychological testing, and maternal interview. He provided a detailed overview of dysmorphology measurements and the dysmorphology score, which is a method of quantifying the degree of growth deficiency and the number of major and minor anomalies present in an individual. The dysmorphology score is weighted with those features most commonly observed in children with FAS being assigned the highest score (Autti-Ramo et al., 2006; Hoyme et al., 2016). Dr. Hoyme cautioned that FASD is a diagnosis of exclusion and that phenocopies of FASD should be considered, as frequencies with which children are born with other genetic and dysmorphic syndromes are not different in alcohol-exposed and non-exposed individuals. Differential diagnoses should always include conditions arising from other teratogens. Additionally, because head circumference, growth, and many cognitive and behavioral characteristics have moderate to high degrees of heritability, information about the biological parents should be considered in the final diagnostic decision when available. Dr. Hoyme then described the implementation of a multidisciplinary diagnostic conference in which diagnoses are assigned in confidential, structured, case conferences. Findings from each domain are discussed individually in a roundtable. Summary results from three independently collected data sets (growth and dysmorphology, neurobehavior, and maternal characteristics) are presented by the research team members who supervise assessment in each domain. Two-dimensional, digital photographs of the child’s face (frontal and 45 deg. views) are also displayed. Following discussion final diagnoses are made by consensus based on the updated CoFASP guidelines (Hoyme et al., 2016). In cases of disagreement the final diagnosis is assigned by the examining dysmorphologist.

Several debated aspects of the Hoyme et al. 2005 diagnostic criteria were noted, including whether 2 or 3 “cardinal” facial features be required for a diagnosis of FAS, the specific percentile cutoffs for height and weight in FAS and PFAS, whether the diagnosis of PFAS could proceed without direct confirmation of maternal ethanol consumption, and whether the terms PFAS, ARBD, and ARND should be used to describe cases on the milder end of the spectrum. The approach taken by Dr. Hoyme and colleagues has been to design criteria that are the most inclusive, taking into account all of the domains evaluated in assessing children with prenatal alcohol exposure. Dr. Hoyme also noted that the use of 3 facial features in FAS diagnosis is highly specific, whereas using only 2 features increases sensitivity, but does not provide the same level of specificity. Importantly, taken alone any of the facial features are not adequate predictors of FAS.

Dr. Hoyme ended his presentation with a discussion of ongoing and potential future research projects. He described an ongoing study in a cohort of infants identified prenatally as exposed to alcohol in which data on dysmorphology, growth, and neurodevelopmental features will be used to determined which measures are most predictive of FASD diagnosis at age 5. A separate study is evaluating the natural history of the clinical phenotype of FASD into adulthood. These studies will inform data-driven diagnostic criteria for FASD in infants, younger children, adolescents, and adults. Diagnostic criteria could potentially include the “dysmorphology score” as a discriminating feature. Dr. Hoyme dedicated the presentation to his mentor, Dr. Kenneth Lyons Jones, M.D. and thanked many colleagues and collaborators from his years of work on FASD in the United States and internationally.

Highlights.

The 2019 Fetal Alcohol Spectrum Disorders Study Group met in Minneapolis, Minnesota and was attended by more than 140 participants.

Keynote Presentations were given by Dr. A. David Redish of the University of Minnesota and Dr. Kevin Grimm of Arizona State University.

The Timothy A. Cudd Award was given to Fleur Wharton of the University of Cape Town

The Kenneth R. Warren Merit Award was given to Parker J. Holman of the University of British Columbia.

The 2019 Rosett Award was given to Dr. H. Eugene Hoyme of The University of South Dakota, Stanford University, and The University of Arizona.