Maternal Prenatal Stress Is Associated With Altered Uncinate Fasciculus Microstructure in Premature Neonates

Background Maternal prenatal stress exposure (PNSE) increases risk for adverse psychiatric and behavioral outcomes in offspring. The biological basis for this elevated risk is poorly understood but may involve alterations to the neurodevelopmental trajectory of white matter tracts within the limbic system, particularly the uncinate fasciculus. Additionally, preterm birth is associated with both impaired white matter development and adverse developmental outcomes. In this study we hypothesized that higher maternal PNSE was associated with altered uncinate fasciculus microstructure in offspring. Methods In this study, 251 preterm infants (132 male, 119 female) (median gestational age = 30.29 weeks [range, 23.57–32.86 weeks]) underwent brain magnetic resonance imaging including diffusion-weighted imaging around term-equivalent age (median = 42.43 weeks [range, 37.86–45.71 weeks]). Measures of white matter microstructure were calculated for the uncinate fasciculus and the inferior longitudinal fasciculus, a control tract that we hypothesized was not associated with maternal PNSE. Multiple regressions were used to investigate the relationship among maternal trait anxiety scores, stressful life events, and white matter microstructure indices in the neonatal brain. Results Adjusting for gestational age at birth, postmenstrual age at scan, maternal age, socioeconomic status, sex, and number of days on parenteral nutrition, higher stressful life events scores were associated with higher axial diffusivity (β = .177, q = .007), radial diffusivity (β = .133, q = .026), and mean diffusivity (β = .149, q = .012) in the left uncinate fasciculus, and higher axial diffusivity (β = .142, q = .026) in the right uncinate fasciculus. Conclusions These findings suggest that PNSE is associated with altered development of specific frontolimbic pathways in preterm neonates as early as term-equivalent age.

. Relationship between white matter microstructure and predictors in regression models 9

Segmentation of uncinate fasciculus in 4 sub-regions
The uncinate fasciculus (UF) tract skeleton was manually segmented into 4 sub-regions and vertices on the skeleton surface were labelled as anterior frontal region, insula region 1, insula region II, posterior temporal region ( Fig S2). For each subject, AD, RD, MD and FA values were averaged across all vertices within each sub-region (Table S7).  (Table S8).
Of note, previous studies have reported white matter abnormality in the anterior section of the UF in generalized anxiety disorder (Phan et al., 2009) and sensitivity to early-life stress (Ho et al., 2017).
*values that survived correction for multiple comparisons ** for clarity, only those values that were significant at p<.05 before correction for multiple comparisons are presented.

SENSITIVITY ANALYSES
Please note that throughout this paper, R 2 refers to the whole model.
a. Imputed data for STAI-TR (n=32) Further, we repeated our analysis using multiple imputation with 5 imputations and compared this with our results obtained from imputation using the mean. For all cases, the median was 36.00 and the range was 20.00-68.00.    There were no significant relationships between uncinate fasciculus microstructure and trait anxiety.
Then, we excluded all the multiple births (n=52) and repeated the main analysis on singleton cases only and R-UF AD (R 2 =.39, β=.140, q=.037). There were no significant relationships between uncinate fasciculus microstructure and trait anxiety.

g. Pregnancy complications -Emergency C-section
We repeated our analysis introducing emergency C-section (yes/no) as a predictor in the model.
There were no significant relationships between uncinate fasciculus microstructure and trait anxiety.

j. Larger sample (n=277)
We repeated our analysis with a larger sample, which included all participants who had completed the stressful life events questionnaire, regardless of the data availability for the STAI-TR questionnaire.
k. GA/PMA range As our sample includes both a large range for gestational age at birth (23.57-32.86) as well as a large range for postmenstrual age at scan (37.86-45.71), we accounted for this in additional analyses.
First, we repeated our analysis including all gestational ages, but restricting the analysis to those scanned at PMA 38.00-43.00. The remaining sample was n=160 and the pattern of results remained largely the same. More specifically, after correcting for multiple comparisons, stressful life events was associated with L UF AD (R 2 =.31, β=.199, q=.025) and R-UF AD (R 2 =.31, β=.210, q=.014). The relationship between stressful life events and L UF MD approached significance (R 2 =.36, β=.172, q=.052, p value before correction=.009). The relationship between stressful life events and L UF RD did not survive correction for multiple comparisons (R 2 =.38, β=.157, q=.085, p value before correction=.016). There were no significant relationships between uncinate fasciculus microstructure and trait anxiety.

Motivation for selection of uncinate fasciculus and inferior longitudinal fasciculus as tracts of interest
The uncinate fasciculus (UF) was chosen based on previous literature relating abnormalities in its microstructure to a wide range of neurodevelopmental and psychiatric disorders, as well as to exposure to stressful events, as detailed below: Prenatal stress exposure has been linked to abnormal neurodevelopment of a number of brain regions including the limbic system and prefrontal cortex, in both animal (Uno et al. 1994;Salm et al. 2004;Kraszpulski et al. 2006;Tamura et al. 2011) and human studies (e.g. Qiu et al., 2013, Buss et al., 2010. The UF connects areas of the limbic system with the frontal lobes, and it has been suggested that limbic regions are particularly vulnerable to prolonged stress during early development (Olson et al., 2015), possibly due to excitotoxicity from glucocorticoids (Conrad et al., 2008). Abnormalities in functional connectivity between limbic and frontal areas have also been widely reported in anxiety disorders (Etkin et al., 2009, Etkin et al., 2010. Thus, in order to investigate whether maternal prenatal stress is associated with altered structural connectivity between the limbic system and frontal lobes, we focused specifically on the microstructural properties of the uncinate fasciculus. Abnormalities in UF microstructure are commonly reported in anxiety disorders and early-life stress (Phan et al., 2009, Hettema et al., 2012, Tromp et al., 2012, Costanzo et al., 2016, Koch et al., 2017. In a study of young adults, Hanson et al. (2015) reported reduced structural integrity of the UF in those who experienced childhood maltreatment, while in a study of early adolescence, higher sensitivity to early life stress was associated with altered microstructure of the uncinate fasciculus, which predicted higher levels of anxiety symptomatology (Ho et al., 2017).
The inferior longitudinal fasciculus (ILF) connects the occipital cortex to the temporal lobe (Ashtari et al., 2011) and alterations in ILF integrity have mostly been reported in relation to impaired visual processing and object naming (Ortibus et al., 2012, Shinouara et al., 2010, Shinouara et al., 2007, but also schizophrenia (Ashtari et al., 2007) and autism spectrum conditions (Koldewyn et al., 2014). To our knowledge, no research studies have implicated the ILF in the development of anxiety disorders in children. Our decision to include the ILF as a control tract was based on previous literature which included the ILF as a control tract in studies focusing on the UF (Sarkar et al., 2014). We chose to focus on 2 tracts to avoid multiple comparisons issues. A control tract was used to ensure that the findings did not stem from an overall global white matter impairment, but were specific to the uncinate fasciculus.
We repeated our analyses with other control tracts: corpus callosum, left and right superior longitudinal fasciculus, and left and right inferior fronto-occipital fasciculus. There was no significant relationship between FA, MD, RD and AD in these tracts and either stressful life events or maternal trait anxiety (no p values <.05 even before correction for multiple comparisons), with the exception of STAI-TR and L-IFO (Beta=.106, p=.024). However, this did not survive correction for multiple comparisons (Table   S12).
The UF and the ILF are late-developing association fibres (Lebel et al., 2012). The UF and ILF have similar developmental trajectories in the fetal and early postnatal period, with myelination commencing in the 3rd postnatal month (Yakolev & Lecours, 1967;Hasegawa, 1992;Gilles 1983, Dubois et al., 2014. Thus, the differences observed in this study are unlikely to occur as a result of differences in myelination. The neural mechanisms underlying the association between PNSE to UF development are unclear and we are not able to assess mechanisms in this observational study.
The neural mechanisms underlying the association between PNSE to UF development are unclear.
Maternal anxiety is associated with physiological changes including alterations in uterine blood flow (Fisk & Glover, 1999), cortisol production (Risbrough & Stein, 2006, Staufenbiel et al., 2013, Abelson et al., 2007 and fetal heart rate (Monk et al., 2000). Indeed, a recent study assessing structural and functional connectivity in infants exposed to maternal depression suggested that alterations in fetal heart rate may influence the development of the amygdala-prefrontal circuit (Posner et al, 2016). However, we do not have fetal physiological data in these infants or data regarding maternal cortisol levels, and we are not able to disentangle genetic versus environmental influences of uncinate fasciculus development.

Additional information on anxiety and stressful life events
On the STAI-TR a cut-off of 40 provides a high degree of accuracy in identifying women with a diagnosed anxiety disorder (Sinesi et al., 2019). In our sample, on the STAI-TR, n=97 women scored 40 or over (n=64 scored 40-49, n=24 scored 50-59, n=9 scored 60-69) and n=154 scored under 40 (n=56 scored 20-29, n=98 sored 30-39). In our sample, maternal trait anxiety scores are low (median 36) and well below clinical cut-offs (Dennis et al., 2013, Grant et al., 2008. Previous studies reporting associations between maternal antenatal anxiety and infant brain development have focused on state, rather than trait anxiety (Dean et al., 2018) or a combined score of state and trait (Rifkin-Graboi et al., 2015) and included participants with higher levels of anxiety (e.g. threshold ~43, Rifkin-Graboi et al., 2015). To our knowledge, only one study (Qiu et al., 2013) examined the relationship between maternal trait anxiety and infant brain development, and reported no significant association between trait anxiety and hippocampal volume at birth.
Trait anxiety, as measured using the STAI-TR, is seen as a relatively stable personality trait (Spielberger, 1972), which implies a generalized and enduring predisposition to respond to situations in an anxious manner. Dennis et al (2013) examined the stability of STAI scores in the perinatal period and reported that overall STAI scores at 1 week postpartum were significantly correlated to 4-week and 8-week scores. Further, high scores on STAI-ST have been reported to occur at similar rates between the third trimester (18.2%) and childbirth (18.6%); scores for STAI-TR were not reported in this study (Figuereido et al., 2011). Lastly, Grant et al (2008) reported considerable stability in self-reported anxiety between pregnancy and 7 months postpartum, arguing that trait anxiety persists from pregnancy into the postnatal period, rather than representing a transient state.
Several studies have investigated the reliability of maternal recall for pregnancy and birth related events (Sou et al., 2006, Tomeo et al., 1999, Quigley et al. 2007). For example, in the study by Tomeo et al.
(1999), recall of pregnancy-related events was highly accurate when comparing self-report measures collected approximately 30 years after birth to data collected during pregnancy. Further, the agreement between hospital records and maternal reporting of mode of delivery retrospectively assessed at 9 months postpartum was 94% (Quigley et al. 2007). To our knowledge, no studies have examined the reliability of maternal recall specifically for adverse life events. However, studies reporting on the validity of retrospective reports of adverse events have suggested that false positive reports are rare (Hardt et al., 2004). The myelogenetic cycles of regional maturation of the brain.
Regional development of the Brain in early life, Mikowski.