Prenatal origins of several mental illnesses have been established from epidemiological studies of famines and pandemics, and observational studies of behavioral deficit newborns who later develop mental illness (Brown & Derkits, Reference Brown and Derkits2010; Erlenmeyer-Kimling & Cornblatt, Reference Erlenmeyer-Kimling and Cornblatt1987; Mednick, Machon, Huttunen, & Bonett, Reference Mednick, Machon, Huttunen and Bonett1988; Susser & Lin, Reference Susser and Lin1992; Walker, Savoie, & Davis, Reference Walker, Savoie and Davis1994). Many genes associated with both brain development and mental illnesses are more highly expressed before birth than afterward (Birnbaum, Jaffe, Hyde, Kleinman, & Weinberger, Reference Birnbaum, Jaffe, Hyde, Kleinman and Weinberger2014). The implication is that fetal brain development is a critical period of risk for later mental illness. Prenatal folic acid supplementation is a paradigmatic example of a treatment to prevent fetal abnormalities that are difficult to treat postnatally (Wald, Sneddon, Densem, Frost, & Stone, Reference Wald, Sneddon, Densem, Frost and Stone1992). Vitamin D and choline are being similarly considered as possible prenatal interventions for mental illness (McGrath et al., Reference McGrath, Eyles, Pedersen, Anderson, Ko, Burne and Mortensen2010; Zeisel, Reference Zeisel2006a).

Choline is concentrated in the amniotic fluid from the maternal plasma by active transport in the placenta (Zeisel, Epstein, & Wurtman, Reference Zeisel, Epstein and Wurtman1980). Some choline is synthesized by the mother, but the majority comes from dietary sources (Zeisel & da Costa, Reference Zeisel and da Costa2009). In addition to its roles as a major methyl donor along with B₁₂, betaine, and folate in DNA methylation and as the synthetic precursor to the phospholipids necessary for membrane stabilization, it is involved in neurotransmission (Albright, Tsai, Friedrich, Mar, & Zeisel, Reference Albright, Tsai, Friedrich, Mar and Zeisel1999; Ballard, Sun, & Ko, Reference Ballard, Sun and Ko2012; Gossell-Williams, Fletcher, McFarlane-Anderson, Jacob, & Zeisel, Reference Gossell-Williams, Fletcher, McFarlane-Anderson, Jacob and Zeisel2005; Zeisel, Reference Zeisel2000; Zeisel & da Costa, Reference Zeisel and da Costa2009), both as a direct agonist at cholinergic receptors and as the precursor to acetylcholine (Zeisel, Reference Zeisel2006b; Zeisel et al., Reference Zeisel, Epstein and Wurtman1980). Choline's agonist role requires the highest amniotic concentrations (Frazier et al., Reference Frazier, Rollins, Breese, Leonard, Freedman and Dunwiddie1998). One-carbon metabolism and membrane synthesis in the fetus can proceed at lower concentrations, and if the available choline supply is limited, these pathways consume most of the available choline. Therefore, maternal plasma choline levels are hypothesized to be a critical factor in the activation of cholinergic receptors during fetal brain development.

Cerebral interneurons are migrating into the cerebral plate and differentiating during the 16th week of gestation, when maternal choline levels were assessed in this study (Vasistha et al., Reference Vasistha, Pardo-Navarro, Gasthaus, Weijers, Müller, García-González and Khodosevich2019). Optogenetic inactivation of these interneurons in animal models is associated with decreased attention and social behavior (Yizhar et al., Reference Yizhar, Fenno, Prigge, Schneider, Davidson, Ogshea and Deisseroth2011). Activation of α7-nicotinic acetylcholine receptors by choline in the amniotic fluid facilitates the transition of the embryonic KCCN chloride pump to the mature KCC2 chloride pump, which increases the chloride gradient across the neuronal cell membrane and allows GABA to become an inhibitory neurotransmitter (Liu, Neff, & Berg, Reference Liu, Neff and Berg2006). Expression of α7-nicotinic receptors is higher before birth than in newborns or adults, consistent with their major role in the maturation of GABAergic transmission (Court et al., Reference Court, Lloyd, Johnson, Griffiths, Birdsall, Piggott and Perry1997). However, cholinergic innervation has not yet reached the cerebrum from the midbrain (Descarries, Aznavour, & Hamel, Reference Descarries, Aznavour and Hamel2005). Choline in the amniotic fluid is needed in early development to activate α7-nicotinic acetylcholine receptors to ensure maturation of chloride pumps and GABAergic neurotransmission (Frazier et al., Reference Frazier, Rollins, Breese, Leonard, Freedman and Dunwiddie1998). This maturation process is not fully complete in patients with schizophrenia (Hyde et al., Reference Hyde, Lipska, Ali, Mathew, Law, Metitiri and Kleinman2011).

We conducted a double-blind, randomized trial of prenatal choline supplementation beginning by 15 weeks of gestation that compared 7300 mg phosphatidylcholine, equivalent to 900 mg choline, to placebo to test the hypothesis that choline supplementation would enhance the development of cerebral inhibitory function (Ross et al., Reference Ross, Hunter, McCarthy, Beuler, Hutchison, Wagner and Freedman2013). Supplementation continued after birth with blinded treatment of the newborn until 52 gestational weeks of age. The development of cerebral inhibition was assessed in 1-month-old newborns by the inhibition of the P50-evoked potential to a repeated auditory stimulus. P50 auditory-evoked potentials have a source in the hippocampus (Goff, Williamson, VanGilder, Allison, & Fisher, Reference Goff, Williamson, VanGilder, Allison and Fisher1980). In animal models of the P50-evoked potential, the activity of hippocampal GABAergic interneurons is responsible for the inhibition to repeated stimuli (Miller & Freedman, Reference Miller and Freedman1995). Choline supplementation significantly enhanced the development of P50 inhibition, as assessed in the newborns. A subsequent report demonstrated that the decreased development of P50 inhibition was related to later childhood behavior problems on the Child Behavior Checklist 1½–5 (CBCL1½–5) (Ross et al., Reference Ross, Hunter, Hoffman, McCarthy, Chambers, Law and Freedman2016). Mean rating percentiles on the Attention Problems scale were significantly lower for those who received supplement (59.4, s.e. = 2.5) compared to those who received placebo (65.9, s.e. = 2.3); there was no effect of gender. Ratings on the Withdrawn Syndrome scale showed effects of supplement and gender. Overall, those who received supplement had lower mean rating percentiles than those who received placebo, and mean rating percentiles for females were lower compared to males. For males who received supplement, mean rating percentiles were significantly lower (60.2, s.e. = 3.2) compared to ratings for those who received placebo (69.4, s.e. = 3.4); this difference was not significant for females.

Other studies of the effects of prenatal choline on outcomes in early childhood include three double-blind, placebo-controlled trials and two observational studies. One trial reported decreased reactive saccadic latency from 4 to 13 months of age in children of healthy mothers who received third trimester choline supplementation, a finding which the authors interpreted as increased processing speed (Caudill, Strupp, Muscalu, Nevins, & Canfield, Reference Caudill, Strupp, Muscalu, Nevins and Canfield2018). Another trial of mid-pregnancy choline supplementation in women who had heavy alcohol use found increased novelty preference scores at 12 months of age on the Fagan Test of Infant Intelligence, which the authors interpreted as better visual recognition memory (Jacobson et al., Reference Jacobson, Carter, Molteno, Stanton, Herbert, Lindinger and Jacobson2018). A trial of phosphatidylcholine supplementation in healthy women beginning at 18 weeks of gestation found no effect on the Visuospatial Memory Delayed Response Task, the Mac-Arthur-Bates Short Form Vocabulary Checklist: Level I, or the Mullen Scales of Early Learning at 10–12 months of age placebo; scores were above the normal range for offspring of both placebo and supplemented mothers (Cheatham et al., Reference Cheatham, Goldman, Fischer, da Costa, Reznick and Zeisel2012). Higher maternal choline concentrations in non-randomized observational studies were associated with increased cognition scores at 18 months of age on the Bayley Scales of Infant Development (Wu, Dyer, King, Richardson, & Innis, Reference Wu, Dyer, King, Richardson and Innis2012). The longest duration study, based on the assessment of choline in the maternal diet in the first and second trimesters, found increased performance on the Wide Range Assessment of Memory and Learning, Design and Picture Memory subtests, in the 7-year-old offspring (Boeke et al., Reference Boeke, Gillman, Hughes, Rifas-Shiman, Villamor and Oken2013).

The current study was undertaken to expand the results of the initial trial of choline supplementation: (1) mothers who had infections or used substances including nicotine, alcohol, and cannabis, were included, whereas they were excluded from the initial trial, (2) to obtain a more diverse population, the women were not randomized or required to take supplements, because the increased compliance requirements exclude many women. All women received dietary instruction as in the previous trial. Maternal plasma choline concentrations obtained with the phosphatidylcholine supplement in the randomized trial were 15.21 μM (s.d. 8.14), compared to 7.85 μM (s.d. 2.54) for placebo (p = 0.03). Choline concentrations ⩾ 7.07 μM, 1 s.d. below the mean level obtained with phosphatidylcholine supplements in the previous trial, were used in the current analysis as a comparator to the supplement. The children have now reached 4 years of age. This paper analyzes their CBCL½–5 Attention Problems and Withdrawn Syndrome scales to assess the hypothesis that higher choline levels in this second cohort would be associated with decreased scores on the Attention Problems and Withdrawn Syndrome subscales, particularly in male offspring, as found in the initial trial of choline supplementation (Ross et al., Reference Ross, Hunter, Hoffman, McCarthy, Chambers, Law and Freedman2016). We also hypothesized, based on the previous trial, that the effects on CBCL½–5 problems would be related to the children's development of P50 inhibition, assessed when they were newborns.

Previous publications from this second cohort have shown interaction of higher choline concentrations with the adverse effects of infection and cannabis on the child through 52 weeks of gestational age (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Alessandro, Christians and Hoffman2019; Hoffman et al., Reference Hoffman, Hunter, D'Alessandro, Noonan, Wyrwa and Freedman2020). In these studies, higher maternal choline concentrations were associated with improved newborn P50 cerebral auditory-evoked response inhibition as well as improved outcomes on the Orienting/Regulation Index of the Infant Behavior Questionnaire-R (IBQ-R) (Gartstein & Rothbart, Reference Gartstein and Rothbart2003) at 52 weeks of gestational age. Reduced scores on the Orienting/Regulation Index of the IBQ-R are related to poorer performance on measures of reading readiness at 4 years of age and increased distractibility at 9 years of age (Gartstein, Putnam, & Kliewer, Reference Gartstein, Putnam and Kliewer2016). The pathogenic mechanism for most infections appears to be the effects of the mother's inflammatory response on the placenta and fetus. In animal models of maternal immune activation effects, dietary choline supplementation counteracts the increase in interleukin-6 in the fetal brain and the increased anxiety behaviors in the offspring (Wu et al., Reference Wu, Adams, Stevens, Chow, Freedman and Patterson2015). Cannabis effects on the developing brain may be related to the co-location of CB1 receptors on the same interneurons as α7-nicotinic receptors (Morales, Hein, & Vogel, Reference Morales, Hein and Vogel2008). Blockade of CB1 receptors by cannabis interferes with endogenous cannabinoid signaling that promotes neurite outgrowth. The effects of α7-nicotinic receptor activation on GABAergic function appear to offset the adverse effects of cannabis. We hypothesized that the interactions of maternal choline levels with either infection or cannabis use would also be associated with decreased CBCL½–5 Attention Problems through 4 years of age.

Results

Of the 183 women who consented to the study, 162 brought their newborns for their initial postnatal evaluation at 1 month of age, and 83 continued in the study as their offspring reached early childhood. The principal reason for attrition at each stage was the family moving away from the Denver metropolitan area. The proportion of women with choline concentrations ⩾7.07 μM at the newborn evaluation was 31%; it was 34% at the 40-month evaluation and 25% at the 48-month evaluation. Only one African American woman completed the previous randomized trial. In the current study, 22% who completed were African American or Native American women (Table 1).

Table 1. Difference between mothers with higher and lower plasma choline concentrations at 16 weeks of gestation

^a Student's t test or Fisher's exact test.

^b Compared to Caucasian women.

Mothers with choline concentrations ⩾7.07 μM were 3 years older on average compared to mothers with lower choline concentrations. There were no other maternal variables or neonatal birth outcomes related to choline concentrations (Table 1). Gestational betaine and dimethylglycine levels did not differ based on the choline level. Maternal body mass index (BMI), prenatal vitamin and folic acid use, which exceeded 90% in both groups, and maternal PEMT genotype also did not differ (Table 1).

Mean maternal choline concentration at 16 weeks of gestation in the group with choline concentration ⩾7.07 μM, 8.47 μM (s.d. 1.56) was significantly lower than the level obtained with phosphatidylcholine supplementation in the randomized trial, 15.21 μM (s.d. 8.14), t _df50 = 4.37, p < 0.001. Mean choline levels for the entire group of mothers in the current study rose at 28 weeks of gestation, consistent with other reports of increasing choline concentrations after mid-pregnancy (Orczyk-Pawilowicz et al., Reference Orczyk-Pawilowicz, Jawien, Deja, Hirnle, Zabek and Mlynarz2016; Wu et al., Reference Wu, Dyer, King, Richardson and Innis2012). Maternal plasma choline concentrations at 28 weeks were not associated with outcomes in childhood.

Several maternal covariates were associated with CBCL1½–5 problems scales. They clustered into three groups of intra-associated factors, headed by maternal age, gestational age at birth, and maternal lifetime depression disorder (Table 2).

Table 2. Co-variates associated with CBCL1½–5 Attention problems

*p < 0.05, **p < 0.01, ***p < 0.001.

Offspring of mothers with 16-week gestational choline levels ⩾7.07 μM had fewer problems with attention (both sexes) and social withdrawal (males) behaviors on the CBCL1½–5 compared to children of mothers with lower plasma choline concentrations (choline level, Wald χ²_df1 = 3.837, p = 0.050; Table 3, online Supplementary Table e1). Children whose mothers had higher plasma choline concentrations had lower Attention Problem T-scores of 52.8, s.e. 1.0 (percentile 58.5, s.e. 2.4) compared to children whose mothers had lower choline concentrations, Attention Problem T-score, percentile 55.2, s.e. 0.2 (percentile 65.3, s.e. 1.8); p = 0.050. For male children, Withdrawn T-scores were also lower if mothers had higher choline concentrations, T-score 52.0, s.e. 1.2 (percentile 59.2, s.e. 3.2), v. T-score 57.6, s.e. 1.2 (percentile 68.3, s.e. 2.5); p = 0.007, for children whose mothers had lower choline concentrations (choline level × child sex, Wald χ²_df1 = 5.362, p = 0.021; Fig. 1; online Supplementary Table e2). Higher choline concentrations were also associated with fewer Sleep Problems, T-score 52.1, s.e. 1.1 (percentile 58.3, s.e. 2.2), compared to T-score 55.0, s.e. 0.8 (percentile 62.2, s.e. 1.6); p = 0.037, for children of mothers with lower choline concentrations (choline level, Wald χ²_df1 = 4.354, p = 0.037, online Supplementary Table e3).

Fig. 1. Mean percentiles for scores on the CBCL1½–5 Withdrawn Syndrome Scale shown separately for males and females by maternal choline concentrations. Scores were significantly lower for male children of mothers with higher choline concentrations (p = 0.007).

Table 3. Relation of maternal choline plasma concentration at 16 weeks of gestation to childhood behavior problems

As was found in the clinical trial of phosphatidylcholine supplementation, maternal choline concentrations at 16 weeks of gestation were associated with increased inhibition of the P50 cerebral auditory-evoked potential, and poorer fetal development of this cerebral inhibition was associated with increased Attention Problems in childhood (Ross et al., Reference Ross, Hunter, Hoffman, McCarthy, Chambers, Law and Freedman2016). The correlation of choline concentration with increased newborn P50 cerebral-evoked potential amplitudes to the second of paired auditory stimuli, indicating poorly developed cerebral inhibition, was β = −0.122, p = 0.048 (online Supplementary Table e4). The correlation of increased newborn P50 cerebral-evoked potential amplitudes to the second of paired auditory stimuli, indicating poorly developed cerebral inhibition, to higher CBCL1½–5 Attention Problem T-scores, was β = 0.381, p = 0.003 (online Supplementary Table e5).

Both prenatal cannabis and common maternal viral and bacterial infections were associated with increased CBCL1½–5 syndrome scale scores, including Attention Problems (Table 4). Higher maternal choline concentrations were associated with decreased Attention Problems scores in children of mothers who used cannabis or had infections (choline level, Wald χ²_df1 = 5.367, p = 0.021, Table 5, online Supplementary Table e6). For children of mothers who used cannabis in gestation, Attention Problems scores were significantly lower if the mother also had higher choline concentrations, T-score 53.1, s.e. 3.4 (percentile 60.3, s.e. 8.4), than if the mother had lower choline concentrations, T-score 61.1, s.e. 1.6 (percentile 76.4, 4.0); p = 0.034. For children of mothers who had gestational infections, Attention Problems scores were also lower if the mother also had higher choline concentrations, T-score 53.5, s.e. 2.0 (percentile 60.9, s.e. 5.0) than if the mother had lower choline concentrations, T-score 58.0, s.e. 1.1 (percentile 69.9, s.d. 2.8; p = 0.050) (Fig. 2).

Fig. 2. Mean percentiles for scores on the CBCL1½–5 Attention Problems Scale shown separately for prenatal cannabis exposure, maternal prenatal infection, and for all participants by maternal choline concentrations. Scores are lower for children whose mothers had higher choline concentrations (all participants, p = 0.050). This relationship was also true for children with prenatal cannabis exposure (p = 0.034) as well as children whose mothers who experienced infection during gestation (P = 0.050).

Table 4. Effects of maternal cannabis and infection at 16 weeks of gestation on Child Behavioral Checklist/1½–5 problems

*Students t test: p < 0.05, **p < 0.01, *p < 0.001.

Table 5. Relation of choline plasma concentration at 16 weeks of gestation and gestational cannabis use or maternal infection with childhood attention problems

CBCL1½–5 includes scales that group problems into DSM-5 categories. Ratings ⩾92nd percentile on the Attention Deficit Hyperactivity Disorder (ADHD) scale occurred in 8 of the 83 children between ages 40 and 48 months. Ratings at this level are generally associated with children who present for clinical evaluations on the referral of parents, schools, and physicians, according to the CBCL1½–5 manual (Achenbach & Rescorla, Reference Achenbach and Rescorla2000). Seven of these eight children were from mothers who had lower choline concentrations.