Childhood maltreatment experiences are associated with altered diffusion in occipito‐temporal white matter pathways

Abstract Background Childhood maltreatment may contribute to brain alterations in posttraumatic stress disorder (PTSD). We previously found that PTSD was associated with white matter compromise, or lower fractional anisotropy (FA), in the left inferior longitudinal fasciculus (ILF). In this study, including non‐PTSD controls, we examined whether ILF FA was associated with maltreatment exposures, including those that meet DSM‐IV criterion A (physical abuse, sexual abuse) and those that typically do not (emotional abuse, emotional neglect, physical neglect). We hypothesized that lower FA would be associated with PTSD diagnosis and with both categories of maltreatment. Methods Ninety‐three participants (51 female), ages 20–50, were enrolled, including 32 with lifetime DSM‐IV PTSD, 27 trauma‐exposed non‐PTSD controls, and 34 healthy controls. Participants completed structured interviews, the Childhood Trauma Questionnaire (CTQ), and diffusion‐weighted imaging (36 directions). Probabilistic tractography (using FreeSurfer's TRACULA) was used to assess diffusion metrics in the ILF. Results Contrary to our hypothesis, there was no significant effect of diagnostic group on FA. In contrast, higher CTQ scores were significantly associated with lower FA in the ILF bilaterally. This association of maltreatment with lower FA remained statistically significant after controlling for diagnostic group, and it was significant for both criterion‐A‐type and noncriterion‐A‐type maltreatment categories. Conclusions This work contributes to a growing body of literature indicating that different forms of childhood maltreatment are associated with altered white matter microstructure in the ILF, an association pathway involved in integrating visual information from occipital regions with emotion processing functions of the anterior temporal lobe.


| INTRODUC TI ON
Childhood maltreatment is associated with a broad range of negative developmental outcomes, including socioemotional, behavioral, and attachment problems (Cicchetti & Toth, 2005). Maltreated children have higher risk of adverse psychiatric outcomes (Kessler et al., 2010), including posttraumatic stress disorder (PTSD; Widom, 1999). This association likely reflects complex causal relationships.
Collectively, this literature suggests that childhood maltreatment moderates the neural expression of PTSD (Teicher & Samson, 2013).
Previously, we examined 10 white matter (WM) tracts in participants with PTSD versus trauma-exposed controls, including those connecting canonical fear circuitry regions (e.g., the uncinate fasciculus, UF), and found lower fractional anisotropy (FA) and higher radial diffusivity (RD) in the left inferior longitudinal fasciculus (ILF; Olson et al., 2017). FA and RD, along with mean diffusivity (MD) and axial diffusivity (AD) are diffusion metrics that assess the extent of directionality of water diffusion and reflect brain structural organization. The ILF is a visual limbic pathway that projects between the occipital and anterior temporal lobes, facilitating rapid transfer of visual information to and from temporal regions involved in memory, learning, and emotional processing (Catani, Jones, Donato, & Ffytche, 2003). Choi, Jeong, Polcari, Rohan, and Teicher (2012) also reported lower FA in the left ILF of individuals who witnessed parental domestic violence (WDV) early in life. Our finding was the first report of abnormalities in the ILF in association with PTSD. Because the trauma-exposed non-PTSD participants in our sample endorsed minimal exposure to childhood maltreatment, we could not statistically separate effects of childhood maltreatment from effects of PTSD diagnostic status.
One of the overarching goals of the current study was to address this limitation in a sample exposed to diverse forms of maltreatment.
One distinction between research on PTSD and research on childhood maltreatment relates to the types of index events under study. PTSD requires exposure to criterion A traumatic event(s) that involve "actual or threatened death or serious injury, or a threat to the physical integrity of self or others," including "violent personal assault (sexual assault, physical attack)." Physical abuse and sexual abuse typically meet criterion A, while other forms of maltreatment, including emotional abuse, emotional neglect, and physical neglect, typically do not. A growing body of literature suggests that exposure to forms of childhood maltreatment that do not meet criterion A is at least as detrimental as exposure to forms of maltreatment that do meet criterion A. For example, Khan et al. (2015) found that adolescent exposure to verbal abuse, emotional neglect, and nonverbal emotional abuse was more strongly associated with history of major depressive disorder than physical or sexual abuse. Additionally, brain imaging studies have shown that noncriterion-A-type maltreatment experiences have significant neurobiological consequences. Specifically, neglect and emotional abuse are associated with amygdala and hippocampus hyperactivity in response to negative emotion (Bogdan, Williamson, & Hariri, 2012;Lee et al., 2015;Maheu et al., 2010), and with amygdalahippocampus hyperconnectivity in response to psychosocial stress (Fan et al., 2015). Emotional neglect may be uniquely detrimental to social development; in a sample of maltreated adolescents, while controlling for other forms of maltreatment, only emotional neglect independently predicted antisocial features (Ometto et al., 2016).
In the childhood maltreatment literature, many prior reports have focused on reduced FA in the UF, a tract frequently selected for a priori analysis because of its role in facilitating prefrontal regulation of amygdala activity (e.g., Eluvathingal et al., 2006;Hanson, Knodt, Brigidi, & Hariri, 2015;Ho et al., 2017;Souza-Queiroz et al., 2016).
While the UF is often selected as an a priori target, altered diffusion in other tracts also has been identified in the context of diverse types of early adverse experiences, ranging from verbal abuse to early institutionalization, in samples with and without psychopathology. For example, Huang, Gundapuneedi, and Rao (2012) found lower FA in the superior longitudinal fasciculus (SLF), cingulum, inferior fronto-occipital fasciculus, and splenium of the corpus callosum in adolescents who had experienced physical abuse, sexual abuse, or witnessed domestic violence. In young adults without psychopathology but with prior exposure to chronic parental verbal abuse, Choi, Jeong, Rohan, Polcari, and Teicher (2009) found lower FA in the arcuate fasciculus, cingulum, and fornix. In a sample exposed to early institutional care, Hanson et al. (2013) found widespread alterations in FA that were particularly extensive in posterior cortical regions. Finally, abnormal diffusion in posterior portions of the corpus callosum has been reported in children with maltreatment-related PTSD (De Bellis et al., 2015;Jackowski et al., 2008). In addition to these cross-sectional findings, longitudinal studies have examined baseline diffusion as a predictor of the development of psychopathology in maltreated samples, with increased internalizing symptoms developing over time in the context of lower baseline FA in the UF (Hanson et al., 2015), SLF and cingulum (Huang et al., 2012), and corpus callosum and external capsule (Bick, Fox, Zeanah, & Nelson, 2017). Thus, the PTSD and childhood maltreatment tractography literatures both implicate fronto-temporal white matter tracts, though significant findings extend beyond those fibers to include more posterior cortical regions.
To summarize, the goal of the present study was to extend our previous finding of lower ILF FA in PTSD in a new sample that included comparison subjects with reported exposure to a broad range of childhood maltreatment severity, in order to dissociate possible effects of PTSD versus childhood maltreatment on white matter microstructure.
We hypothesized that lower ILF FA would be associated with both PTSD diagnosis and exposure to childhood maltreatment. Additionally, because of the demonstrated significance of emotional neglect, emotional abuse, and physical neglect on neurobiological development, we hypothesized that noncriterion-A-type maltreatment exposures would have comparable associations with ILF FA as criterion-A-type maltreatment exposures. As a control tract, we examined the corticospinal tract (CST), a pyramidal tract involved in motor control. We were unable to locate any prior reports of altered FA in the CST in individuals with PTSD or histories of childhood maltreatment.

| Participants
Ninety-three subjects who responded to study advertisements in the Boston metropolitan area participated in this study, including 32 participants with DSM-IV lifetime PTSD, 27 trauma-exposed non-PTSD controls (TENC), and 34 healthy controls (HC). Prior to participating, all subjects received a full explanation of study procedures and provided written informed consent. All subjects were paid for their time ($25/hr, up to $200). Study procedures were authorized by Partners HealthCare's Human Research Committee. For detailed inclusion/exclusion criteria, see Appendix S1.

| Clinical interviews and measures
Participants were evaluated with the Structural Clinical Interview for DSM-IV Axis I Disorders (SCID-I/P) (First, Spitzer, Gibbon, & Williams, 1996). PTSD symptom severity was assessed using the Clinician-Administered PTSD Scale, Current and Lifetime Versions (CAPS-DX) (Blake et al., 1995). The CAPS was administered to all participants who endorsed at least one DSM-IV criterion A event on the Life Event Checklist (Gray, Litz, Hsu, & Lombardo, 2004). Clinical interviews were conducted by doctoral-level clinical psychologists.
Lifetime trauma load was assessed using the Life Events Checklist (LEC: Gray et al., 2004); the total score was the sum of experienced and witnessed events).
Childhood Trauma Questionnaire (Bernstein et al., 2003) subscales were summed to generate a total maltreatment exposure variable. The physical and sexual abuse subscales were summed in order to create a combined criterion-A-type maltreatment exposure score; and the emotional neglect, physical neglect, and emotional abuse subscales were summed to create a noncriterion-A-type maltreatment exposure score.

| MR image acquisition
MRI data were acquired via a 3.0 Tesla Siemens Tim Trio Scanner (Siemens) with the use of a 32-channel phased-array design RF head coil operating at 123 MHz. Anatomical whole-brain images were obtained, including high-contrast T1-weighted MPRAGE images.
Whole-brain diffusion tensor images (DTI) were obtained using a 36 direction-weighting scheme. (For parameters, see Appendix S1.)

| Tractography processing
Automatic probabilistic tractography was performed using TRActs

| Childhood maltreatment exposure
Demographics and clinical characteristics of the sample are presented in Table 1. When maltreatment was coded as present or absent based on criteria for at least moderate exposure, 49 participants reported no maltreatment exposure, while 43 participants endorsed at least one category of moderate or severe maltreatment exposure (one participant did not provide complete data). Exposure to maltreatment types was largely overlapping (Figure 1): participants exposed to one maltreatment type tended to be exposed to  TA B L E 1 Demographic and clinical characteristics of posttraumatic stress disorder (PTSD), trauma-exposed non-PTSD control (TENC), and healthy control (HC) participants (mean ± SD or N (%))  Table 2). There were no significant associations between childhood maltreatment and the other diffusion metrics (MD, RD, AD).

| ILF FA: Relationships with childhood maltreatment types
The association between childhood maltreatment severity and ILF

| ILF FA: Associations of childhood maltreatment along the trajectory of the ILF pathway
To examine relationships with diffusivity along the trajectory of the ILF pathway, individual results were projected into MNI space along the tract, and p-values at each point along the normalized tract were computed for the association between total childhood maltreatment exposure and FA (at 53 points along the left ILF, and 52 points along the right ILF; Figure 4). In the left hemisphere, there were extensive regions along the trajectory of the ILF where FA was significantly associated with childhood maltreatment. These ran in parallel with much of the hippocampus, ranging from regions near the anterior hippocampus to regions near F I G U R E 1 Maltreatment exposures in this sample. Forty-three of 93 participants endorsed at least moderate maltreatment exposure. As illustrated here, maltreatment exposure categories were largely overlapping. Numbers reflect number of participants reporting exposure to the indicated maltreatment types. Numbers with carets reflect individuals exposed only to noncriterion-A-type maltreatment types, while numbers with asterisks reflect individuals exposed to only criterion-A-type maltreatment types. All others reported exposure to both maltreatment categories

| CST FA: Control tract associations with PTSD diagnosis and total childhood maltreatment severity
Highlighting the specificity of the ILF findings, there was no signifi-

| D ISCUSS I ON
Greater childhood maltreatment was associated with lower FA in the ILF bilaterally, an effect that persisted after accounting for diagnostic group. In contrast, there were no significant differences in ILF FA between diagnostic groups (PTSD vs. TENC vs. healthy control). Lower ILF FA emerged in association with both criterion-A-type childhood maltreatment (physical and sexual abuse) and noncriterion-A-type maltreatment (emotional abuse and emotional and physical neglect). There were no significant relationships with other diffusion metrics (MD, RD, AD), and there were no significant associations of childhood maltreatment or PTSD diagnostic status with WM integrity of the corticospinal tract, which was selected as a control tract.
In our previous report (Olson et al., 2017), we found lower ILF FA in adults with PTSD but were unable to determine whether this reflected an association with PTSD diagnosis versus childhood maltreatment because these two variables were statistically overlapping. The present analyses (a) confirm involvement of lower WM integrity in the ILF and (b) clarify that this effect tracks with childhood maltreatment and not PTSD diagnosis. Our findings align with previous research from Choi et al. (2012), who described lower FA in the ILF of young adults who had witnessed domestic violence as children compared with young adults with no maltreatment history.
Together, these results confirm that structural integrity of the ILF may be compromised in individuals who experience childhood maltreatment, with effects evident for multiple types of maltreatment.
The present study found significant effects on FA but not other diffusion metrics; this pattern can be generated in the context of modest changes in RD and AD that combine to produce significant TA B L E 2 Partial correlations between scores on Childhood Trauma Questionnaire (CTQ) subscales and right and left ILF FA, controlling for sex and age. Lower portion of table: in combined three-group sample (n = 92-93). Upper portion of table: in trauma-exposed sample (n = 58-59)  (Latini et al., 2017). The ILF directly contacts other association tracts, including the uncinate and arcuate fasciculi (Herbet, Zemmoura, & Duffau, 2018). Like all WM tracts, the developmental trajectory of the ILF is protracted, with peak FA reached after age 25; however, other tracts that have more often been the focus of trauma research show a much later peak FA age (e.g., mid-40s for the cingulum, over age 35 for the uncinate) (Lebel et al., 2012). It is possible that the earlier maturational process in the ILF makes it particularly vulnerable to insult in childhood or adolescence, since actively maturing regions may be more susceptible to disruption (Rice & Barone, 2000). From a functional perspective, the ILF is involved in a variety of functions including object recognition, facial recognition, and reading (Herbet et al., 2018), as well as semantic memory (Hodgetts et al., 2017). Hanson et al. (2013) found that reduced ILF FA in children exposed to emotional neglect via institutional rearing was correlated with deficits in visual paired associates learning. Importantly, ILF disruption also is associated with deficits in processing facial affect (Genova et al., 2015;Rigon, Voss, Turkstra, Mutlu, & Duff, 2018 Bellis et al., 2015). Additionally, because we had few participants who had experienced solely criterion-A-type maltreatment or solely noncriterion-A-type maltreatment, we were unable to directly compare effect sizes across groups. Our cell sizes are also too small for investigation of sex-specific effects, which may be present and important. Finally, our maltreatment measure did not include information about the timing and duration of exposure during childhood, so we are unable to further investigate those effects. Future work will employ measures (e.g., MACE: Maltreatment and Chronology of Exposure; Teicher & Parigger, 2015) that allow for assessment of these variables.

| CON CLUS ION
To summarize, the present study contributes to a growing body of literature indicating that childhood maltreatment is associated with altered WM microstructure in the ILF, a long association pathway involved in integrating visual information from the posterior cortices with anterior temporal lobe functions including emotion processing.
From a clinical perspective, the present work supports the proposition that neglect and emotional abuse can have comparable effects on neurodevelopmental maturation as physical and sexual abuse, highlighting the importance of early detection and intervention with individuals exposed to multiple forms of childhood maltreatment.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare in relation to this work.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request. LeŌ ILF R ight ILF