Introduction
Retrospective epidemiological studies established maternal infection during the early second trimester as a risk factor for schizophrenia (Brown & Derkits, Reference Brown and Derkits2010; Clarke, Tanskanen, Huttunen, Whittaker, & Cannon, Reference Clarke, Tanskanen, Huttunen, Whittaker and Cannon2009). Animal models identify maternal immune activation (MIA) with cytokine induction as a pathophysiological mechanism (Wu et al., Reference Wu, Adams, Stevens, Chow, Freedman and Patterson2015). Maternal inflammation, often in response to respiratory and genitourinary infections in the early second trimester, decreases the development of inhibitory sensory gating and increases impulsivity in the offspring (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Allesandro, Christians and Hoffman2019; Ghassabian et al., Reference Ghassabian, Albert, Hornig, Yeung, Cherkerzian, Goldstein and Gilman2018; Melbye, Hvidsten, Holm, Nordbø, & Brox, Reference Melbye, Hvidsten, Holm, Nordbø and Brox2004). Pathogenic effects of maternal cytokines on both the placenta and the brain have been proposed (Brown & Meyer, Reference Brown and Meyer2018). In addition to infection, maternal inflammation has been associated with obesity, maternal depression and anxiety, and environmental pollution (Madan et al., Reference Madan, Davis, Craig, Collins, Allan, Quinn and Dammann2009; Osborne & Monk, Reference Osborne and Monk2013; van den Hooven et al., Reference van den Hooven, de Kluizenaar, Pierik, Hofman, van Ratingen, Zandveld and Jaddoe2012). Elevation in plasma C-reactive protein (CRP) at 16 weeks gestation and interleukin-6 (IL-6) later in pregnancy in women whose offspring develop schizophrenia adds molecular evidence for the pathogenic role of inflammation (Canetta et al., Reference Canetta, Sourander, Surcel, Hinkka-Yli-Salomaki, Leiviska, Kellendonk and Brown2014; Goldstein et al., Reference Goldstein, Cherkerzian, Seidman, Donatelli, Remington, Tsuang and Buka2014). These epidemiological findings are now well-accepted, but limited information exists about which specific aspects of human fetal brain development are altered by maternal inflammation to produce mental illness in later life (Graham et al., Reference Graham, Rasmussen, Rudolph, Heim, Gilmore, Styner and Buss2018; Spann, Monk, Scheinost, & Peterson, Reference Spann, Monk, Scheinost and Peterson2018). The first pathophysiological signs of schizophrenia, apparent as early as birth, have been proposed as indicators of aberrant fetal brain development (Erlenmeyer-Kimling & Cornblatt, Reference Erlenmeyer-Kimling and Cornblatt1987; Walker, Savoie, & Davis, Reference Walker, Savoie and Davis1994). RDoC provides a framework that can be applied prospectively to study the earliest development origins of schizophrenia by examining pathophysiological components of the illness that appear before diagnostic symptoms are apparent (Ross & Freedman, Reference Ross and Freedman2015).
This study assessed newborn cerebral inhibitory physiology as a primary outcome of fetal brain development using an auditory sensory gating measure, a biomarker of schizophrenia in adults that first appears in newborns (Kisley, Polk, Ross, Levisohn, and Freedman, Reference Kisley, Polk, Ross, Levisohn and Freedman2003). Development of inhibitory neurons was characterized by sensory gating of the P50 cerebral evoked response in a paired-auditory stimulus paradigm (S1, S2). P50 amplitude is normally decremented in response to S2, because of inhibitory mechanisms activated by the response to S1. This inhibition is impaired in persons with schizophrenia (Adler et al., Reference Adler, Pachtman, Franks, Pecevich, Waldo and Freedman1982). P50 sensory gating is a physiological measure in the RDoC Cognition Domain, Perception Construct, Auditory Sub-Construct. P50 sensory gating has been used in adults with schizophrenia to detect familial or genetic risk (Freedman et al., Reference Freedman, Coon, Myles-Worsley, Orr-Urtreger, Olincy, Davis and Byerley1997; Hall, Taylor, Salisbury, & Levy, Reference Hall, Taylor, Salisbury and Levy2011; Quednow et al., Reference Quednow, Brinkmeyer, Mobascher, Nothnagel, Musso, Gründer and Winterer2012). Abnormal P50 gating also occurs in autism spectrum disorder (Olincy, Blakeley-Smith, Johnson, Kem, & Freedman, Reference Olincy, Blakeley-Smith, Johnson, Kem and Freedman2016; Orekhova et al., Reference Orekhova, Stroganova, Prokofyev, Nygren, Gillberg and Elam2008). P50 inhibition can be recorded as soon as 1 month after birth (Kisley et al., Reference Kisley, Polk, Ross, Levisohn and Freedman2003). Diminished newborn P50 inhibition predicts early childhood problems in self-regulation, attention, and social function (Hutchison et al., Reference Hutchison, Hunter, Wagner, Calvin, Zerbe and Ross2007; Ross et al., Reference Ross, Hunter, Hoffman, McCarthy, Chambers, Law and Freedman2016). These early childhood behaviors are recognized as early developmental symptoms in children who later develop psychosis (Pine & Fox, Reference Pine and Fox2015; Rossi, Pollice, Daneluzzo, Marinangeli, & Stratt, Reference Rossi, Pollice, Daneluzzo, Marinangeli and Stratt2000; Rutter, Kim-Cohen, & Maughan, Reference Rutter, Kim-Cohen and Maughan2006). Newborn P50 inhibitory deficits are increased if either parent has schizophrenia, a finding replicated by another group (Hunter, Kisley, McCarthy, Freedman, & Ross, Reference Hunter, Kisley, McCarthy, Freedman and Ross2011; Smith, Crawford, Thomas, & Reid, Reference Smith, Crawford, Thomas and Reid2018).
Increased maternal choline promotes the development of newborn P50 inhibitory gating, and therefore, we also assessed the possible interaction of the maternal inflammatory response with choline in this study (Ross et al., Reference Ross, Hunter, McCarthy, Beuler, Hutchison, Wagner and Freedman2013).
Participants and methods
Maternal assessment and recruitment
Women were enrolled from a public safety-net prenatal clinic at 14–16 weeks gestation from July 2013 until July 2016. Gestational age was timed from the last menstrual period and by ultrasound. Exclusions were fetal anomaly and major maternal medical morbidity. The Colorado Multiple Institution Review Board approved the study; all participants gave informed consent. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. Women were asked to participate in a prospective study of stress in pregnancy on their child's development.
Psychiatric diagnoses were made using the Structured Clinical Interview for DSM-IV Axis I Disorders with DSM-5 criteria. Self-ratings on Center for Epidemiological Studies of Depression (CESD), State-Trait Anxiety Inventory-State Version (STAI-S), and the Perceived Stress Scale (PSS) were acquired. Maternal sociodemographics and health, including infections, substance use, BMI, and prenatal vitamin use, were assessed.
The medical record for all prenatal care was reviewed for infection. A mother's self-report of infection was considered significant if it was entered as a problem in the medical record. Treatment was provided for all reported genitourinary infections. Most respiratory infections were viral and were therefore not treated. In addition, mothers had an in-person review of systems for symptoms of infection at 16 weeks gestation by research personnel. The correlation between symptoms rated by the mother as moderate to severe on a custom-made scale from little or mild (1–3) to moderate (4–5) to severe (6–9) in the interview and problems in the medical record is r s = 0.96, p < 0.001 (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Allesandro, Christians and Hoffman2019). Mothers reported the use of alcohol, cannabis, and nicotine at 16 weeks gestation in the same interview and at 6-week intervals through the first 6 weeks postpartum (Hoffman et al., Reference Hoffman, Hunter, D'Alessandro, Noonan, Wyrwa and Freedman2019a, Reference Hoffman, Olincy, D'Alessandro, Reisz, Hansen, Hunter and Ross2019b).
Labor, delivery, and neonatal parameters were recorded from the medical record. Investigators were blinded to maternal choline levels during assessments.
Choline measurements
Maternal plasma was assayed for choline and its metabolite betaine at 16 weeks gestation by the Colorado Translational Research Center Metabolomics Laboratory using mass spectroscopy (online Supplementary material). Blood samples were obtained at least 2 h after breakfast. Plasma was quickly separated by refrigerated centrifugation to prevent platelet phosphatidylcholine release.
Mothers received information on diets higher in choline, but dietary intake was not estimated because of the low relationship of self-reported intake to maternal choline levels in pregnant women, r = 0.2 (Wu, Dyer, King, Richardson, & Innis, Reference Wu, Dyer, King, Richardson and Innis2012). The placental choline transporter CLT1 produces amniotic fluid levels approximately twice maternal plasma levels (Baumgartner et al., Reference Baumgartner, Trinder, Galimanis, Post, Phang, Ross and Winn2015; Ilcol, Uncu, & Ulus, Reference Ilcol, Uncu and Ulus2002). Uptake is proportional to maternal plasma concentration, which suggests that higher peak levels may be important determinants of amniotic fluid levels (Iwao et al., Reference Iwao, Yara, Hara, Kawai, Yamanaka, Nishihara and Inazu2016). Maternal levels obtained in non-fasting conditions, as in the present study, can be elevated, but only after high-choline meals that exceed the recommended daily intake (Abratte et al., Reference Abratte, Wang, Li, Axume, Moriarty and Caudill2009; Holm, Ueland, Kvalheim, & Lien, Reference Holm, Ueland, Kvalheim and Lien2003; Zeisel, Growden, Wurtman, Magil, & Logue, Reference Zeisel, Growden, Wurtman, Magil and Logue1980). Only choline activates α7-nicotinic receptors (Alkondon, Pereira, Cortes, Maelicke, & Albuquerque, Reference Alkondon, Pereira, Cortes, Maelicke and Albuquerque1997). We found no effects of its metabolite betaine.
Maternal CRP and cytokines
Plasma CRP at 16 weeks gestation was assayed by the Beckman–Coulter high sensitivity assay at the Colorado Translational Research Center, Colorado Children's Hospital. Tumor necrosis factor-α (TNFα), IL-6 and IL-8 were assayed by R&D Systems high sensitivity assays at the CTRC Core Laboratory, Children's Hospital of Colorado. IL-6 values above the detectable limit (1.9 pg/ml) were found for nine of 91 samples from women without infection and 17 of 61 women with infection (Fisher's exact test p = 0.007). TNF levels above detectable limits were observed in 81 of 91 women without infection and 60 of 61 women with infection (1.6 pg/ml). Levels below detectable limits were set to zero.
Neonatal physiological recording of cerebral inhibition
Newborns were studied 1 month (44 weeks after the last menstrual period) after birth adjusted for gestational age. Vertex electroencephalogram, electro-oculogram, submental electromyogram, and respiration were continuously recorded while infants napped (Hunter et al., Reference Hunter, Kisley, McCarthy, Freedman and Ross2011; Kisley et al., Reference Kisley, Polk, Ross, Levisohn and Freedman2003). Recording of the cerebral auditory evoked potential P50, a positive EEG wave 50 ms post stimulus, occurred in the second active sleep episode, the precursor of REM sleep, identified by low-voltage desynchronized vertex activity with the absence of K-complexes, change in respiration, and large eye movements with submental atonia (Anders, Emde, & Parmelee, Reference Anders, Emde and Parmelee1971). The second active sleep episode was reached approximately 45 min after sleep onset. In adults, P50 inhibition in REM and waking are equivalent (Griffith & Freedman, Reference Griffith and Freedman1995).
Two identical auditory stimuli are delivered 500 ms apart to elicit P50S1 and P50S2. P50 inhibition is often assessed as amplitude ratios P50S2/P50S1 or (P50S1–P50S2)/P50S1 (Adler et al., Reference Adler, Pachtman, Franks, Pecevich, Waldo and Freedman1982). However, the skew inherent in ratios limits their power for correlation with risk factors. P50S2 amplitude, covaried for P50S1, which is normally distributed, has therefore also been used (Smith, Boutros, & Schwarzkopf, Reference Smith, Boutros and Schwarzkopf1994). Two large NIMH genetic collaborations that included P50 sensory gating in their phenotypes [eight-site Collaboration on the Genetics of Schizophrenia (COGS) and five-site Bipolar-Schizophrenia Network on Intermediate Phenotypes (B-SNIP)] found effect sizes d’ = 0.28 and d’ = 0.22, respectively, for P50S2 amplitude in schizophrenia, compared to community controls (Clementz et al., Reference Clementz, Sweeney, Hamm, Ivleva, Ethridge, Pearlson and Tamminga2016; Olincy et al., Reference Olincy, Braff, Adler, Cadenhead, Calkins, Dobie and Freedman2010). A recent study using magnetoencephalography also reported that the sensory gating defect in schizophrenia is robust to the measurement method (Schubring, Tzvetan, Miller, & Rockstroh, Reference Schubring, Tzvetan, Miller and Rockstroh2018). Regardless of the method used, higher P50S2 amplitudes indicate decreased inhibition. The assumption is that P50S1 variance is small, compared to P50S2 variance. In 151 newborns, effect sizes for P50S1 differences between newborns whose mothers had no known risk v. women with depression or schizophrenia ranged from 0 to 0.16. Effect sizes for the decrease in P50S2 amplitude were 0.21–0.50 (Hunter et al., Reference Hunter, Kisley, McCarthy, Freedman and Ross2011). Intraclass correlation between two newborn recordings 1 week apart is r ICC = 0.84. Other technical aspects of recordings have been published (Hunter, Corral, Ponicsan, & Ross, Reference Hunter, Corral, Ponicsan and Ross2008; Reference Hunter, Gillow and Ross2015).
Childhood behavioral assessments
Parents completed the Infant Behavior Questionnaire-Revised Short Form (IBQ-R) when the infant was 3 months of age (Gartstein & Rothbart, Reference Gartstein and Rothbart2003; Putnam, Helbig, Gartstein, Rothbart, & Leerkes, Reference Putnam, Helbig, Gartstein, Rothbart and Leerkes2014). The Parental Distress Subscale of the Parenting Stress Index was also completed as a possible covariate for parental bias (Abidin & FL, Reference Abidin2012). The 91-item IBQ-R Short Form, commonly used to study behavior in children at this age, rates 14 aspects of child behavior, which the IBQ-R developers clustered into three indices by factor analysis: Surgency summarizes the child's level of activity and positive affect; Negativity summarizes fearfulness and anxiety; and Regulation summarizes duration of attention, responsiveness to parents, and enjoyment of quiet play. Two components of Regulation are also in Surgency, smiling and soothability, and the two indices are highly correlated in the present sample (r = 0.51, p < 0.001). Covariation with Surgency isolates elements of Regulation that are more specific to the early development of attention and less attributable to the child's general psychomotor activation. A similar covariance between Regulation and Surgency (0.80) has been documented by another group, who have also proposed revisions to the factor structure (Bosquet-Enlow, White, Hails, Cabrera, & Wright, Reference Bosquet-Enlow, White, Hails, Cabrera and Wright2016). Lower IBQ-R Regulation at 1 year of age is associated with decreased reading readiness at age 4 years and decreased conscientiousness, organization, and increased distractibility at age 9 years of age (Putnam et al., Reference Putnam, Helbig, Gartstein, Rothbart and Leerkes2014; Slobodskaya & Kozlova, Reference Slobodskaya and Kozlova2016).
Statistical analyses
Neonatal P50S2 amplitude and childhood IBQ-R Regulation were the two principal outcomes, based on previous work that found P50 inhibition was a biomarker of both infection and choline's effects and that regulatory behaviors were the most affected outcome (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Allesandro, Christians and Hoffman2019; Ross et al., Reference Ross, Hunter, Hoffman, McCarthy, Chambers, Law and Freedman2016). Kolmogorov–Smirnoff tests for each outcome did not find a significant deviation from normal distributions. General Linear Models and multiple regression analyzed child sex as a fixed effect and CRP and other cytokine levels and choline levels as a continuous effect. For P50S2 amplitude, maternal tobacco smoking, cannabis, and alcohol use were examined as covariates because of previously established effects on P50 inhibition (Hoffman et al., Reference Hoffman, Hunter, D'Alessandro, Noonan, Wyrwa and Freedman2019a, Reference Hoffman, Olincy, D'Alessandro, Reisz, Hansen, Hunter and Ross2019b; Hunter et al., Reference Hunter, Kisley, McCarthy, Freedman and Ross2011). For IBQ-R analyses, maternal education was a covariate because of its correlation with both CRP levels (r = −0.169, p = 0.038) and IBQ-R Regulation (r = 0.196, p = 0.021). In addition, maternal 3 months postpartum CESD, STAI-S, PSS, and PSI were examined as covariates, because of possible bias in ratings from maternal mood and stress. In a previous study, the effect size of choline on P50 inhibition was Cohen's d’ = 0.7 (Ross et al., Reference Ross, Hunter, McCarthy, Beuler, Hutchison, Wagner and Freedman2013). We expected 20% of the women would have adequate choline levels and 30% attrition (Hoffman et al., Reference Hoffman, Hunter, D'Alessandro, Noonan, Wyrwa and Freedman2019a; Reference Hoffman, Olincy, D'Alessandro, Reisz, Hansen, Hunter and Ross2019b; Wu et al., Reference Wu, Dyer, King, Richardson and Innis2012). Therefore, we enrolled 200 women to have power 1–β > 0.95, α = 0.05, 1-tail to observe an overall choline effect.
Results
Mothers' inflammatory response during early gestation
Of 316 mothers screened, 201 were enrolled by 16 weeks gestation, as dated from the last menstrual period and verified by ultrasound measurements. Of these, 162 brought their newborns to the 1-month postnatal assessment, where the auditory evoked potentials were obtained as previously reported (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Allesandro, Christians and Hoffman2019). CRP levels were assessed in plasma samples obtained at 16 weeks gestation from 150 of the women (online Supplementary Fig. S1). When the infants reached 3 months of age, 127 women brought them in for a behavioral assessment. Attrition during the study after enrollment generally reflected mothers who moved from the pre-birth residence to their mothers' homes to care for their infants and were then lost to follow-up. The significant differences between male and female babies were only in greater male birth weights and head circumferences (online Supplementary Table S1).
CRP levels correlated with maternal CESD self-ratings of depression, pre-pregnancy BMI, severity of gestational infection, and maternal education (Table 1). There was no correlation with infant sex. More educated women were older (r = 0.324, p < 0.001) and had lower self-ratings of depression (r = −0.190, p = 0.015) and lower BMI (r = −0.283, p < 0.001), which has been found in other populations (Bui & Miller, Reference Bui and Miller2018). CESD ratings correlated with both infection (r = 0.264, p = 0.001) and BMI (r = 0.210, p = 0.007). Multiple regression analysis found significant effects for CRP of infection severity [β = 0.202 (95% CI 0.047–0.356) p = 0.011] and BMI [β = 0.329 (95% CI 0.176–0.482) p < 0.001], but not depression or maternal education (online Supplementary Table S2). If the ratings were dichotomized to identify mothers likely to come to clinical attention as depressed (CESD > 15), obese (BMI > 30), or as having a significant infection (severity>5), then only infection was a significant factor for CRP elevation [β = 0.237 (95% CI 0.076–0.398) p = 0.004].
Table 1. Maternal factors associated with CRP elevation at 16 weeks gestation
CRP levels were not associated with either complications of labor and delivery, such as preterm birth, gestational diabetes, or chorioamnionitis, or neonatal outcomes, such as lower APGAR scores or small size for gestational age infants (online Supplementary Table S3). CRP levels were correlated with other maternal cytokines, specifically IL-6, IL-8, and TNF-α (Table 1).
Newborn expression of cerebral inhibition
A paired-stimulus auditory paradigm (S1, S2) delivered during active sleep was used to assess the newborn's development of cerebral inhibition, measured as the decrease in amplitude of P50S2, relative to P50S1 (Fig. 1). There were no differences in P50S1 or in P50S2 amplitude between male and female newborns (Table 1).
Fig. 1. Top: Examples of P50S1 and P50S2 auditory evoked responses recorded 1 month after birth in a female newborn. P50 amplitude is measured from the most positive peak relative to the preceding negativity. The mother's 16-week gestation CRP was 15.3 mg/ml and her choline was 5.87 µM. P50S1 was 1.50 µV and P50S2 was 0.27 µV. Bottom: Effects of CRP on P50S2 for male and female infants.
Maternal CRP levels had a sex-dependent effect on P50S2 amplitude (Wald χ2df1 = 6.522, p = 0.011; online Supplementary Table S4); P50S1 was unaffected. CRP levels correlated with increased P50S2 amplitude in males [β = 0.205 (95% CI 0.023–0.387) p = 0.03], but decreased P50S2 amplitude in females [β = −0.214 (95% CI −0.390 to −0.038) p = 0.016; Fig. 1]. Tobacco smoking, marijuana, and alcohol use during early pregnancy did not affect the significance of the effects of CRP on P50S2 amplitude (online Supplementary Table S4).
Infant behavior at 3 months of age
Maternal ratings of their infants' behavior showed no differences between male and female infants on the three principal IBQ-R indices – Surgency, Regulation, and Negativity (online Supplementary Table S1). Multivariate analysis of effects on CRP on the three IBQ-R indices found that maternal CRP increased 3-month IBQ-R Surgency regardless of the sex of the infant [F df1 = 4.645, p = 0.033; β = 0.207 (95% CI 0.036–0.378); online Supplementary Table S5]. The two most affected components of this scale were Vocalizations (Wald χ2df1 = 6.299, β = 0.195, p = 0.012); and High Intensity Pleasure (Wald χ2df1 = 7.478, β = 0.202, p = 0.006).
The effects of CRP on IBQ-R Regulation were sex-dependent (CRP × sex: F df1 = 4.291, p = 0.041; Fig. 3; online Supplementary Table S5). There were no significant effects of CRP on IBQ-R Negativity. In males, CRP decreased Regulation [β = −0.209 (95% CI −0.429 to −0.001), p = 0.049], but there was no significant effect in females (β = 0.062, p = 0.6). The two most affected IBQ-R components were Soothability (CRP × sex: Wald χ2df1 = 4.842, p = 0.042; in males β = −0.388, p = 0.005) and Cuddliness (CRP × sex: Wald χ2df1 = 6.043, p = 0.014; in males β = −0.245, p = 0.044).
Multiple regression found that newborn P50S2 amplitude was related to 3-month-old infant IBQ-R Regulation, also in a sex-dependent manner (P50S2 × sex: Wald χ2df1 = 7.052, p = 0.008; Fig. 2, online Supplementary Table S6).
Fig. 2. Top left: Effects of maternal CRP on infant IBQ-R Regulation. Top right: Relation of newborn P50S2 amplitude to infant IBQ-R Regulation. Bottom: Effects of maternal CRP on infant IBQ-R Surgency.
IL-6 levels correlated with infant IBQ-R Negativity across both sexes (r = 0.229, p = 0.009).
Interaction of maternal choline levels and prenatal vitamins with the maternal inflammatory response
Maternal choline levels at 16 weeks gestation did not directly affect CRP levels, but maternal choline levels decreased the effect of CRP levels on P50S2 amplitude. Multiple regression found effects of maternal choline for all newborns [β = −0.121 (95% CI −0.227 to −0.01); Wald χ2df1 = 2.017, p = 0.029] and sex-specific effects also (Wald χ2df1 = 7.888, p = 0.019; Fig. 3, online Supplementary Table S7). In males, choline significantly diminished P50S2 amplitude [β = −0.216 (95% CI −0.297 to −0.035) P = 0.019]. In females, the effect of choline was small and not significant (β = 0.010, p = 0.9). To demonstrate the three-way interaction of choline with CRP and child sex, choline levels were dichotomized based on the mean value in the sample (6.4 µM). For males whose mothers had choline levels >6.4 µM, the effect of CRP on P50S2 amplitude was considerably diminished, until CRP levels rose above 15 mg/ml (Fig. 3).
Fig. 3. Effects of maternal choline levels and maternal CRP on male newborn P50S2 (left) and IBQ-R Regulation (right).
Prenatal vitamins with folic acid decreased P50S2 amplitude in both sexes, but the effects did not reach significance (in males, β = −0.168, p = 0.07; in females, β = −0.148, p = 0.09).
Multiple regression found a significant interaction of child sex and maternal CRP and choline levels for IBQ-R Regulation (Wald χ2df1 = 7.663, p = 0.022; online Supplementary Table S8). For males whose mothers had choline levels below 6.4 µM, the effect of CRP on Regulation was β = 0.352 (95% CI 0.044–0.660), p = 0.028. If the mother had choline levels above 6.4 µM, the effect of CRP levels became non-significant (β = 0.093, p = 0.6; Fig. 3). The effect of CRP on Regulation was non-significant for females regardless of maternal choline level. Prenatal vitamins had no significant effect on IBQ-R Regulation for either sex.
Comment
Increased maternal CRP levels in early gestation were associated with decreased development of P50 cerebral inhibition in males but not females. The decrease in P50 sensory inhibitory gating was further reflected in decreased infant Regulation behaviors rated on the IBQ-R at 3 months of age. CRP levels were increased by maternal infections, obesity, and depression, with infection and BMI being predominant (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Allesandro, Christians and Hoffman2019). Unlike inhibitory gating and Regulation, increases in both infant Surgency, also related to elevated CRP, and infant Negativity, related to elevated IL-6, were observed in both sexes. The sex-specific effects on inhibition in male fetuses suggest that there may be specific effects of maternal inflammation on the development of inhibitory interneurons, particularly since their development in this study appears to be facilitated in females by the inflammatory response. Other aspects of early behavior, Surgency and Negativity, are similarly affected in both sexes, however, suggesting that there may be both sex-dependent and sex-independent mechanisms affecting fetal brain development. Effects on both the placenta's integrity and neuronal development have been investigated in the maternal immune response (Brown & Meyer, Reference Brown and Meyer2018).
In animal models, cerebral interneurons are sensitive to maternal inflammation and hypoxia (Lacaille et al., Reference Lacaille, Vacher, Bakalar, O'Reilly, Salzbank and Penn2019). Parvalbumin neurons of the hippocampus are specifically affected, as well as somatostatin-positive neurons (Canetta et al., Reference Canetta, Bolkan, Padilla-Coreano, Song, Sahn, Harrison and Kellendonk2016; Vasistha et al., Reference Vasistha, Pardo-Navarro, Gasthaus, Weijers, Müller, García-González and Khodosevich2019). A pathophysiological mechanism involving the astroglia response has been proposed (Sobue et al., Reference Sobue, Ito, Nagai, Shan, Hada, Nakajima and Yamada2018). Male mice have a greater reduction in hippocampal volume and more chronic macrophage infiltration than females after MIA (Dada et al., Reference Dada, Rosenzweig, Al Shammary, Firdaus, Al Rebh, Borbiev and Burd2014). Male mice are also more sensitive to other insults, such as prenatal radiation, with greater loss of hippocampal pyramidal and interneurons (Ganapathi & Manda, Reference Ganapathi and Manda2017). We did not find a report of an animal model of inflammation that enhanced female brain development. Moreover, female mice are more sensitive to the apoptotic effects of prenatal dexamethasone (Zuloaga et al., Reference Zuloaga, Carbone, Quihuis, Hiroi, Chong and Handa2012), and effects on NMDA receptors in the hippocampus after prenatal glucocorticoids are more marked in females than males (Owen, Setiawan, Li, McCabe, & Matthews, Reference Owen, Setiawan, Li, McCabe and Matthews2004). A second hit has also been proposed as a pathophysiological mechanism. Animals exposed to MIA during gestation are more likely to show deficits after pubertal stressors. This combination reduced the expression of parvalbumin inhibitory interneurons, but the analyses were performed only in male mice (Giovanoli et al., Reference Giovanoli, Engler, Engler, Richetto, Voget, Willi and Meyer2013). Differences in gene expression in placental trophoblasts have also been proposed as a mechanism of female resilience. The X-linked gene OGT is expressed at greater levels in females; it is responsible for protein glycosolation, including glycosolation of H3K27me3, which regulates repressive epigenetic modifications (Nugent, O'Donnell, Epperson, & Bale, Reference Nugent, O'Donnell, Epperson and Bale2018).
In humans, the retrospective epidemiological data are not as clear. Maternal CRP levels at 16 weeks gestation are associated with increased risk of schizophrenia and autism spectrum disorder in a Finnish cohort (Brown et al., Reference Brown, Sourander, Hinkka-Yli-Salomäki, McKeague, Sundvall and Surcel2014; Canetta et al., Reference Canetta, Sourander, Surcel, Hinkka-Yli-Salomaki, Leiviska, Kellendonk and Brown2014), but decreased autism spectrum disorder in a California cohort (Zerbo et al., Reference Zerbo, Traglia, Yoshida, Heuer, Ashwood, Delorenze and Croen2016). In a Dutch cohort, higher maternal CRP levels at 13 weeks gestation were related to autism traits in the general population (Koks et al., Reference Koks, Ghassabian, Greaves-Lord, Hofman, Jaddoe, Verhulst and Tiemeier2016). There were no sex-specific effects in any of these cohorts, despite the preponderance of schizophrenia and autism disorders in males. In a New England cohort, however, mothers of males with schizophrenia had higher IL-6 levels in the early third trimester (Goldstein et al., Reference Goldstein, Cherkerzian, Seidman, Donatelli, Remington, Tsuang and Buka2014). The double prenatal-adolescent hit is also apparently pathogenic in humans. In a Danish cohort, females had increased risk for schizophrenia after prenatal exposure to infection compared to males, but additional pubertal trauma resulted in males having a markedly increased risk (Debost et al., Reference Debost, Larsen, Munk-Olsen, Mortensen, Meyer and Petersen2017). Like animal models, human females are also more sensitive to the effects of corticosteroids released naturally when mothers become depressed (Kim et al., Reference Kim, Davis, Sandman, Sporns, O'Donnell, Buss and Hetrick2017). Thus, there may be a second pathway by which female fetuses are affected because the mother's infection is often accompanied by depression (Freedman et al., Reference Freedman, Hunter, Law, Wagner, D'Allesandro, Christians and Hoffman2019).
CRP does not generally cross from the maternal circulation to directly affect the fetus (Malek et al., Reference Malek, Bersinger, Di Santo, Mueller, Sager, Schneider and Raio2006). A possible mechanism of the effects of MIA is cytokine-mediated macrophage attack of the placenta, which the mother's immune system may treat as a foreign body. Placental cytokines are increased in the placenta in animal models of stress during pregnancy, specifically in males (Bronson & Bale, Reference Bronson and Bale2014). In humans, elevated CRP in early gestation is associated with chronic placental villitis (Ernst et al., Reference Ernst, Grobman, Wolfe, Huang, McDade, Holl and Borders2013). Maternal CRP is deposited in the human placenta during pregnancy, where increased levels are associated with chorioamnionitis, pre-eclampsia, and preterm delivery (Kim et al., Reference Kim, Yoon, Jeon, Lee, Hong, Lee and Kim2015). Genes expressed in the placenta are associated with schizophrenia in the subset of patients who had prenatal complications, including significant infection. The expression of these genes in the placenta is specifically upregulated in males compared to females (Ursini et al., Reference Ursini, Punzi, Chen, Marenco, Robinson, Porcelli and Weinberger2018). MIA of umbilical vein macrophages is also greater in males than in females (Kim-Fine et al., Reference Kim-Fine, Regnault, Lee, Gimbel, Greenspoon, Fairbairn and de Vrijer2012).
This maternal immune attack on the placenta, although not directly transmitted to the fetus, triggers a reaction in the fetus itself that directly affects fetal brain development, including interneuron migration specifically (Oskvig, Elkahloun, Johnson, Phillips, & Herkenham, Reference Oskvig, Elkahloun, Johnson, Phillips and Herkenham2012). These interneurons are responsible for inhibition of the cerebral auditory evoked response (Miller & Freedman, Reference Miller and Freedman1995). The timing of measurement of maternal CRP levels, early in pregnancy v. closer to term, is critical for assessing pathogenic v. normal effects. Cerebral interneurons are in a critical stage of development at 16 weeks gestation, when we found effects in this study and when effects are also seen on the later risk for schizophrenia (Bayatti et al., Reference Bayatti, Moss, Sun, Ambrose, Ward, Lindsay and Clowry2008; Canetta et al., Reference Canetta, Sourander, Surcel, Hinkka-Yli-Salomaki, Leiviska, Kellendonk and Brown2014; Vasistha et al., Reference Vasistha, Pardo-Navarro, Gasthaus, Weijers, Müller, García-González and Khodosevich2019; Zecevic, Hu, & Jakovcevski, Reference Zecevic, Hu and Jakovcevski2011). Later in pregnancy, inflammation and CRP are involved in the normal parturition process. The placenta at term synthesizes CRP (Malek et al., Reference Malek, Bersinger, Di Santo, Mueller, Sager, Schneider and Raio2006). Increased inflammatory cytokines in the last trimester are associated with enhanced brain development, but later childhood problems with impulsivity (Graham et al., Reference Graham, Rasmussen, Rudolph, Heim, Gilmore, Styner and Buss2018; Spann et al., Reference Spann, Monk, Scheinost and Peterson2018). Maternal IL-6 levels at 16 weeks gestation, which we found associated with infant Negativity, were not similarly associated when levels across the entire pregnancy were considered (Rudolph et al., Reference Rudolph, Graham, Feczko, Miranda-Dominguez, Rasmussen, Nardos and Fair2018). These studies did not report sex-specific effects.
The effects of choline on the development of cerebral inhibition are mediated by the activation of α7-nicotinic receptors (Alkondon et al., Reference Alkondon, Pereira, Cortes, Maelicke and Albuquerque1997; Ross et al., Reference Ross, Hunter, McCarthy, Beuler, Hutchison, Wagner and Freedman2013). α7-nicotinic receptors on hippocampal neurons are required for the induction of the chloride ion gradient necessary for neuronal inhibition (Liu, Neff, & Berg, Reference Liu, Neff and Berg2006). From Fig. 3, the effects of choline on inhibition appear to be competitive with the effects of maternal inflammation in males. In an animal model, deletion of CHRNA7, the gene associated with α7-nicotinic receptors, completely blocks the effects of choline on the development of inhibition (Stevens, Choo, Stitzel, Marks, & Adams, Reference Stevens, Choo, Stitzel, Marks and Adams2014). Direct comparison of CHRNA7 deletion with prenatal MIA found similar effects of both insults on the offsprings' development (Giovanoli, Werge, Mortensen, Didriksen, & Meyer, Reference Giovanoli, Werge, Mortensen, Didriksen and Meyer2019), providing further evidence for the possible competing effects of α7-nicotinic receptor activation and maternal inflammatory effects on brain development. The α7-nicotinic receptor activation mediates the anti-inflammatory action of the vagus nerve (Wang et al., Reference Wang, Yu, Ochani, Amella, Tanovic, Susarla and Tracey2003). In pregnant wild-type C57BL/6N mice subjected to MIA, choline supplementation did not change IL-6 levels in the placenta, but did lower IL-6 levels in the fetal brain, suggesting no general anti-inflammatory effect but rather a specific effect in the fetus itself. Effects were seen in both males and females (Wu et al., Reference Wu, Adams, Stevens, Chow, Freedman and Patterson2015). Effects of choline on maternal cytokine levels were observed in a rat gestational model with nearly fivefold dietary choline supplementation (Zhang et al., Reference Zhang, Han, Bao, Yang, Shi, Garfield and Liu2018), but there were no effects on mothers' cytokine levels observed in our study with normal human diets. A study of twofold choline supplementation on the expression of placental angiogenic factors and pro-inflammatory factors in a mouse model with a hemizygous Dlx3± gene deletion model of placental insufficiency found that the effects of supplementation in early gestation were greater in male placentas and in later gestation, greater in females (King et al., Reference King, Kwan, Yan, Jiang, Fomin, Levine and Caudill2019). Thus, like MIA, the sites of choline's effects appear to be complex, with different but interacting effects in mother, placenta, and fetal brain.
Effects of choline in the human infants in this study were most marked on the development of cerebral inhibition and behavioral Regulation where sex-dependent effects of inflammation were also seen. Sex-specific effects of choline have not been generally observed in other human studies of choline's effects (Jacobson et al., Reference Jacobson, Carter, Molteno, Stanton, Herbert, Lindinger and Jacobson2018; Ross et al., Reference Ross, Hunter, McCarthy, Beuler, Hutchison, Wagner and Freedman2013). Nonetheless, for the more vulnerable male fetus in particular, higher choline levels would seem to be a useful intervention to prevent the initial steps in the pathophysiology of later mental illness from occurring before birth in the mother who experiences inflammation. Three randomized clinical trials of choline-containing supplements have reported positive results on child cognition and behavior (Caudill, Strupp, Muscalu, Nevins, & Canfield, Reference Caudill, Strupp, Muscalu, Nevins and Canfield2018; Jacobson et al., Reference Jacobson, Carter, Molteno, Stanton, Herbert, Lindinger and Jacobson2018; Ross et al., Reference Ross, Hunter, McCarthy, Beuler, Hutchison, Wagner and Freedman2013, Reference Ross, Hunter, Hoffman, McCarthy, Chambers, Law and Freedman2016). The US Food and Drug Administration recommends at least 550 mg choline daily from diet, and the American Medical Association recommends choline supplements as part of prenatal vitamins for all women (American Medical Association, 2017; Food & Drug Administration, 2016).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291719003313.
Acknowledgements
Randal G. Ross (M.D.), who died during the course of this study, was responsible for its initial conceptualization and design. Viridiana Galicia-Rodriguez participated in the research effort.
Financial support
The study was supported by the Institute for Children's Mental Disorders; The Anschutz Foundation; The National Institutes of Health NIH/NCATS grant number UL1 TR001082 (all authors); NICHD grant number K12HD001271-11 (Dr Hoffman).
Conflict of interest
The authors report no conflict of interest.
