Functional Imaging Correlates of Childhood Trauma: A Qualitative Review of Past Research and Emerging Trends

Abstract

Childhood trauma exposure is common and is associated with poor clinical outcomes in adolescence along with mental health and sociodemographic challenges in adulthood. While many strategies exist to investigate the biological imprint of childhood trauma exposure, functional neuroimaging is a robust and growing technology for non-invasive studies of brain activity following traumatic experience and the relationship of childhood trauma exposure with psychopathology, cognition, and behavior. In this review, we discuss three popular approaches for discerning functional neural correlates to early life trauma, including regional activation, bivariate functional connectivity, and network-based connectivity. The breadth of research in each method is discussed, followed by important limitations and considerations for future research. We close by considering emerging trends in functional neuroimaging research of childhood trauma, including the use of complex decision-making tasks to mimic clinically-relevant behaviors, data-driven techniques such as multivariate pattern analysis and whole-brain network topology, and the applications of real-time neurofeedback for treatment of trauma-related mental disorders. We aim for this review to provide a framework for understanding the relationship between atypical neural functioning and adverse outcomes following childhood trauma exposure, with a focus on improving consistency in research methods and translatability of research findings for clinical settings.

1. Introduction

The field of research aimed at empirically examining the widespread effects of early life stressors maintains a vague boundary between the various terms used to categorize these stressors. Examples of early life stressors include parental divorce, parental mental illness or imprisonment, poverty and deprivation, death or illness in the family, bullying, peer rejection, exposure to community violence, and physical or emotional abuse (Finkelhor et al., 2015a). Often, these experiences are grouped together and encompassed by one term, such as adverse childhood experiences (ACE), childhood maltreatment, and childhood trauma. However, these experiences are vastly different and distinctions between the types of experiences and which term they fall under is of great importance. For the purpose of this review, we will begin by broadly discussing the prevalence of any and all early life stressors or adverse experiences. The remainder of the review will primarily focus on examining the existing research on individuals exposed to childhood trauma specifically. Types of childhood trauma are taken from the Childhood Trauma Questionnaire (CTQ; Bernstein and Fink, 1998) and the Clinician Administered PTSD Scale for DSM-5 (CAPS; Weathers et al., 2018), which include the following experiences: accidents, natural disasters, traumatic death of a loved one, major injuries or illness, physical abuse, sexual abuse, emotional abuse, emotional neglect, or physical neglect.

1.1. Burden and Prevalence of Early Life Adversity and Childhood Trauma Exposure

Exposure to adverse experiences in childhood is a prevalent and costly public health risk factor. Globally, roughly 39% of adults have experienced childhood adversity (Kessler et al., 2010). A study of adults from 25 U.S. states revealed that three-fifths of the sample experienced at least one ACE (Merrick, 2019). Further, in a national general population survey of children’s exposure to violence, roughly 50% of the sample had more than one direct experience of violence, crime, or abuse, roughly 10% of the sample experienced six or more, and roughly 1% experienced 10 or more (Finkelhor et al., 2015b). Within children and adolescents seeking treatment for trauma-related symptoms, 29% reported mild-to-moderate emotional abuse and 27% reported severe emotional abuse (Hoeboer et al., 2020), suggesting that emotional abuse is common in clinical samples as well. In a survey of adolescents, one-third of the sample experienced a potentially traumatic event as an adolescent (McLaughlin et al., 2013). Among the traumatic events experienced by children and adolescents, assault and abuse are particularly prevalent. In their sample of children, Finkelhor et al. (2015b) found that nearly 38% of children had experienced assault by a sibling or peer, nearly 7% experienced a sexual offense or assault, and 15.2% of the sample experienced childhood maltreatment within the last year. These prevalence rates are reflected in the financial costs associated with child maltreatment. The estimated per-victim lifetime cost for the victim of nonfatal and fatal child maltreatment was demonstrated to be $830,928 and $16,615,186, respectively and the economic cost of childhood maltreatment in the United States was estimated at up to two trillion dollars (Peterson et al., 2018). Experience of childhood trauma is common and expensive, and can have profound implications for physical and mental health throughout the life course. Adverse experiences during childhood increase the risk for broad forms of psychopathology, in particular mood and anxiety disorders, and this persists through the lifespan. Exposure to violence in childhood is associated with clinical symptoms of depression, social anxiety disorder, posttraumatic stress disorder (PTSD), attention-deficit hyperactivity disorder, and conduct problems (Briggs- Gowan et al., 2010). Compared to non-exposed children, clinical disorders are more likely among children exposed to violence (roughly 82% versus 40%) and non-interpersonal traumatic events (roughly 68% versus 41%). Interpersonal violence (IPV) specifically, is strongly related to PTSD (McLaughlin et al., 2013). The conditional probability of PTSD is roughly 40%, 37%, 31%, 29%, and 25% in youth who experienced rape, kidnapping, sexual assault, physical assault by a romantic partner, and physical abuse by a caregiver, respectively. Furthermore, emotional abuse and emotional neglect were demonstrated to be related to PTSD symptom severity in a clinical sample of children undergoing treatment for trauma-related symptoms (Hoeboer et al., 2020).

Childhood experiences of adversity are also related to poor clinical and sociodemographic outcomes in adulthood. In a global sample of adults, twelve types of childhood adversity were associated with an elevated risk for psychological disorders, and the rate of risk was greater with more experiences of adversity (Kessler et al., 2010). Furthermore, childhood adversities were related to about 45% of childhood-onset disorders and about 29% of later-onset disorders (Green et al., 2010). Within a U.S. sample of non-institutionalized adults, experience of an ACE was associated with health concerns and socioeconomic challenges as an adult (Merrick, 2019). More specifically, those who had experienced several traumas as a child had higher rates of “depression, heavy drinking, lower education attainment, lack of health insurance, and unemployment” (Merrick, 2019). Experiences of childhood trauma not only influence the individual but also their offspring. A survey of childhood trauma and psychiatric disorders revealed a high prevalence of every psychiatric disorder assessed in children who had a parent with a history of trauma, compared to children who did not have a parent with a history of trauma (Arditti and Strat, 2020). Furthermore, a parent’s trauma history was more influential on their child than the presence of a disorder. Thus, adverse experiences and childhood trauma exposure, more specifically, are associated with disruption in typical development, which underlies the relationship between trauma and resulting psychopathology in adulthood.

1.2. Neuroimaging Investigations of Childhood Trauma

Towards the goal of better understanding how early life trauma confers such broad risk for psychopathology, neuroimaging provides unique opportunities for the non-invasive study of brain structure and function. Functional Magnetic Resonance Imaging (fMRI) can be utilized to examine neurocircuitry differences between healthy individuals and individuals with a specific set of characteristics relating to a psychological construct of interest. Functional activation studies provide a specific piece of information regarding the complex biological profile of trauma as it relates to psychopathology. This profile not only includes functional neural activity, but also involves dysfunction in other physiological systems that is linked to chronic inflammation (Tursich et al., 2014), chronic pain (Burke et al., 2017), and altered neuroendocrine response (De Bellis and Zisk, 2014; Juruena et al., 2020). The effects of childhood trauma on individual survivors are wide-ranging and functional brain activity will only ever explain a small fraction of the biological consequences of exposure; however, through improvements in data acquisition and analysis strategies, neuroimaging has made significant contributions to our current understanding of human cognition and relationships between psychopathology and brain structure and function.

Research on disrupted cognitive and emotional processes in individuals who have experienced childhood trauma has primarily focused on altered emotional processing (McLaughlin and Lambert, 2017; Teicher et al., 2016). Specifically, evidence has shown that this population has disruption in processes such as fear acquisition, extinction, and generalization (Machlin et al., 2019; McLaughlin et al., 2016; Thome et al., 2018) and emotional regulation (Dvir et al., 2014; Pechtel and Pizzagalli, 2011). This altered emotional learning is accompanied by abnormal processing in fear- and emotional control circuits in the brain (Etkin and Wager, 2007; Marusak et al., 2015). Though fear- and emotion-regulation circuits are commonly implicated in studies of childhood trauma exposure, a smaller body of research has identified potential deficits in learning processes outside of the context of fear and emotions in this population (Pechtel and Pizzagalli, 2011; Yasik et al., 2007). In children, childhood maltreatment is often accompanied by delays in cognitive development, reduced memory performance, lower academic performance (in math and reading), lower performance in literacy and verbal comprehension, lower IQ scores, lower cognitive performance, lower scores on perceptual and non-verbal reasoning (Yingying et al., 2019). Despite the novel implications of this line of research, to our knowledge, the existing literature lacks comprehensive investigations connecting brain function to many cognitive impairments related to childhood trauma.

1.3. Purpose and Outline for Review

The purpose of this review is to discuss functional neurocircuitry alterations associated with childhood trauma exposure and psychopathology following exposure. We discuss three main approaches for discerning neurocircuitry correlates of early life trauma: regional activation, bivariate functional connectivity, and network-based connectivity approaches. The review begins with an examination of regional activation studies as they relate to emotion regulation and threat processing in this population, and the limitations of regional activation approaches. We then review the current literature investigating associations between functional connectivity (FC) during rest or during task engagement and trauma exposure and resulting psychopathology. A more contemporary understanding of brain organization posits that brain regions do not act independently; rather, spatially distributed brain regions function together to comprise a network. Therefore, we then discuss emerging trends in neuroimaging research focusing on network investigations of the brain, data-driven approaches to fMRI analyses, and fMRI tasks that probe complex decision-making processes through active, rather than passive, engagement. These trends have the potential to enrich our understanding of the effect of childhood trauma exposure on development of psychopathology and improve treatment outcomes.

2. Regional Neurocircuitry Studies in Childhood Trauma

fMRI studies of regional activation are designed to investigate the response of a specific brain area to a presented stimulus using a non-invasive imaging technology. With this approach, researchers can infer differential brain activity based on a contrast of the blood-oxygen-level-dependent (BOLD) response of a region, where a stronger BOLD response in one condition compared to another implies heightened activity and a weaker response implies reduced activity. These studies were motivated by the desire to understand alterations of regional specialization in mental disorders and to locate key areas of deficient processing. Theoretically, if one could locate the specific region or regions of altered activity in psychiatric disorders, one could then tailor treatments for symptom mitigation that are based on the unique biology of that region.

2.1. Emotion Regulation and Threat Processing

Individuals who experience childhood trauma or maltreatment often present with clinical symptoms related to heightened negative affect (e.g., anxiety, depression, irritability), threat reactivity, and difficulty regulating emotions (Merikangas et al., 2010; Mii et al., 2020). Consequently, much of the functional neuroimaging research in samples with childhood trauma exposure has utilized cognitive tasks that probe emotion- and threat-processing neural circuitry, particularly the amygdala, medial prefrontal cortex, and dorsal anterior cingulate cortex. During common forms of these tasks, participants lie in the fMRI scanner and passively view stimuli, generally emotional faces or threatening images, that are meant to evoke an emotional response. Regional or whole-brain activation across different stages of the task can then be compared between groups or in relation to clinical variables of interest. This approach is popular for studying the functional neural correlates of childhood trauma exposure and has resulted in a body of evidence supporting alterations in functional engagement of emotion and fear circuitry in trauma-exposed samples.

2.1.1. Enhanced Activation of the Amygdala.

Due to its involvement in the processing of fear- and emotion-related stimuli (Phelps, 2004), the amygdala is a primary target of investigation for many studies of childhood trauma exposure. In fact, the amygdala is frequently cited as showing elevated reactivity to emotional or threatening stimuli in a wide range of study samples with childhood trauma exposure. A recent meta-analysis of 32 fMRI studies revealed that early adversity was related to greater bilateral amygdala activation to sad faces (Saarinen et al., 2021). Elevated amygdala reactivity to emotional faces was also observed in youth with childhood abuse exposure (Ganzel et al., 2013; Garrett et al., 2012; McLaughlin et al., 2015, 2014), previously institutionalized youth with early deprivation (Tottenham et al., 2011), youth with sexual abuse exposure (van den Bulk et al., 2016), and youth with emotional neglect (Maheu et al., 2010). Amygdala hyperreactivity to threat may also be predictive of PTSD symptom improvement following treatment in adolescent girls with trauma exposure (Cisler et al., 2015) and altered amygdala activation to emotional stimuli may be related to specific sensitive periods of childhood abuse exposure (Zhu et al., 2019). Amygdala activation was also elevated in response to emotional faces in adult subjects with childhood abuse or neglect exposure (Dannlowski et al., 2013; van Harmelen et al., 2013), suggesting that heightened engagement of the amygdala during emotion processing may be a long-lasting imprint of childhood trauma exposure.

Though the amygdala is frequently cited in emotion processing studies of samples with childhood trauma, altered activity of the amygdala is not ubiquitous in this literature. Differential activation to emotional faces is less commonly identified in the amygdala in whole-brain analyses compared to small volume-corrected analyses where the amygdala is chosen as an a priori ROI. In a sample of youth with violence exposure, whole-brain voxelwise analyses revealed enhanced activation of the lateral occipital cortex and reduced activation of the dorsal anterior cingulate (dACC) to fearful faces in violence exposed youth compared to controls, but did not find evidence for altered amygdala activity (Weissman et al., 2019). Similar tasks in youth with posttraumatic stress symptoms and maltreatment exposure also failed to reveal group differences in amygdala activation to emotional faces at the whole-brain level (Garrett et al., 2019, 2012; van Harmelen et al., 2013). Finally, despite using an amygdala ROI, a study of adults with childhood maltreatment failed to find strong evidence for altered activity of the amygdala during an emotional valence conflict task (Fonzo et al., 2016). These investigations suggest that, while heightened activity of the amygdala during emotion-processing is perhaps the most commonly implicated neural correlate of childhood trauma exposure, this effect lacks robustness when considering whole-brain alterations in activity.

2.1.2. Altered Activation of the Medial Prefrontal Cortex and Dorsal Anterior Cingulate.

Similarly to the amygdala, the medial prefrontal cortex (mPFC) and dACC are implicated in disordered fear learning and emotion regulation (Feng et al., 2015; Milad et al., 2007; Phelps, 2004); therefore, these regions are also commonly studied in the context of childhood trauma exposure. Though the mPFC and dACC are often considered a singular brain structure, the mPFC comprises the anterior ventral portion of the PFC whereas the dACC extends dorsally though the posterior section of the PFC. Unlike the amygdala, trauma-related activation patterns of the mPFC and dACC are less clear and may depend highly on the chosen fMRI task or stimuli. With emotional faces and emotional Stroop tasks, childhood trauma was related to decreased activity in the rostral mPFC (Blair et al., 2019) and dACC (Weissman et al., 2019) in youth. mPFC activity was also decreased in adolescent girls with assault exposure compared to non-assaulted controls during a social trust task when reward was taken from the participant, indicating a reduced encoding of trust violations in this sample (Lenow et al., 2014). Conversely, increased dACC and mPFC activity has also been observed in samples with childhood trauma exposure. When presented with neutral expressions, youth with PTSD (Garrett et al., 2019) and previously-institutionalized youth (Tottenham et al., 2011) demonstrated increased dACC and mPFC, respectively, activity compared to controls, suggesting a heightened attention to neutral expressions or failure to discriminate emotional faces in these youth. dACC activity was also elevated in abuse-exposed youth during cognitive reappraisal (McLaughlin et al., 2015), during extinction recall (Marusak et al., 2020). These studies indicate that childhood trauma exposure is associated with heightened activation of the dACC during tasks engaging effortful cognitive processes, which may suggest impaired executive functioning / efficiency and/or the ability to disengage from emotional stimuli.

2.1.3. Other Regions of Altered Activity.

While the amygdala and dACC are common targets of investigation in fMRI studies of emotion regulation and threat processing in groups with childhood trauma, they are not the only regions with altered activity. While viewing emotional faces, youth with posttraumatic stress symptoms (PTSS) showed reduced activation in the dorsolateral prefrontal cortex (dlPFC) to neutral and happy faces (Garrett et al., 2012), indicating reduced encoding of facial emotions in a critical cognitive control region. During an affective Stroop task, youth with abuse, but not neglect, exposure demonstrated decreased differential responding on incongruent compared to congruent and view trials within the postcentral gyrus and inferior parietal lobule, precentral gyrus, middle cingulate cortex, middle temporal gyrus, superior temporal gyrus, and rostromedial prefrontal cortex (Blair et al., 2019). Taken together, these studies suggest that childhood trauma exposure and early life stress (ELS) are associated with reduced activity in regions of the default mode and frontoparietal networks, both of which are essential to cognitive and emotional control (Dosenbach et al., 2007; Spreng et al., 2010; Sripada et al., 2014).

Several regions other than the amygdala have also been demonstrated to be hyperactive during emotion- and threat-processing tasks in subjects with childhood trauma exposure, especially the anterior insula. The anterior insula is a critical structure that plays a role in interoception, bodily and emotional awareness, and assessment of risk and uncertainty (Craig, 2009), functions which are often altered in individuals with trauma exposure. Youth with trauma exposure showed enhanced anterior insula activation to negative images (McLaughlin et al., 2015), fearful faces (Wymbs et al., 2020), happy and neutral faces (Garrett et al., 2012), and during fear extinction recall (Marusak et al., 2020), suggesting that youth with early exposure to adverse events may have hyperactive salience detection to emotional or threatening cues. Maltreated youth also displayed greater activation than non-maltreated controls in the inferior frontal gyrus and middle frontal gyrus when presented with neutral faces (Tottenham et al., 2011), greater activation to fearful faces in the anterior cingulate/frontal pole, nucleus accumbens, and orbital frontal cortex (Wymbs et al., 2020), and greater activation of the putamen and thalamus to negative images and in the superior frontal gyrus and frontal pole during cognitive reappraisal (McLaughlin et al., 2015). In sum, these studies provide evidence for a relationship between childhood trauma exposure and enhanced cortical and subcortical functional activity on a variety of fMRI tasks which may be contributing to heightened negative affect, threat reactivity, and difficulty regulating emotions observed among individuals exposed to early life trauma.

2.2. Functional Activity Differences by Abuse Subtype

Childhood abuse, maltreatment, neglect, or trauma exposure can come in many forms. Though most studies utilize total scores from general trauma measures such as the CTQ (Bernstein and Fink, 1998) or the Youth Life Stress Questionnaire (Rudolph and Flynn, 2007) to assess relationships between childhood trauma exposure and brain activity, some studies attempt to parse apart the differential effects of abuse and neglect subtypes on brain function.

2.2.1. Assaultive Violence Exposure.

Childhood assaultive violence includes exposure to physical assault, physical abuse by a caregiver, sexual abuse, and witnessing domestic or community violence. Critically, assaultive violence exposure also confers higher risk for the development of PTSD and other psychological burdens relative to non-assaultive trauma exposures (Kessler et al., 2017; Kilpatrick et al., 2003). While these forms of trauma exposure carry different experiences, some trends in brain activity emerge when subjects with childhood assaultive violence are studied. Heightened functional activation of the amygdala during emotional faces viewing was observed in adolescents with PTSD related to childhood sexual abuse (van den Bulk et al., 2016) and predicted PTSD symptom reduction following treatment in a sample of adolescent girls with assaultive violence exposure (Cisler et al., 2015). Additionally, adolescents with sexual abuse exposure demonstrated increased activation in regions of the default mode network, including the middle and superior frontal gyri, posterior cingulate cortex (PCC), and superior and inferior temporal gyri, to looming social threat cues relative to neutral or receding threat cues (Blair et al., 2020), suggesting that sexual abuse may be related to increased activation of neural systems engaged during self-directed mentation during threat presentation. In youth, physical abuse and sexual abuse exposure were positively correlated with differential responsiveness to negative stimuli compared to positive stimuli and passive viewing compared to incongruent trials, respectively, on an affective Stroop task (Blair et al., 2019). Adolescents with physical abuse exposure also showed enhanced activation of dorsomedial frontal cortex regions during failed inhibition (Lim et al., 2015) and higher activation of the anterior insula and dACC during fear extinction recall (Marusak et al., 2020), which might indicate that childhood exposure to assaultive violence is related to increased salience-related engagement during tasks involving extinction recall or error detection.

Conversely, adolescents with assaultive violence exposure have shown reduced activation in brain regions implicated in attention, salience detection, and cognitive control, including the perigenual anterior cingulate cortex (pgACC), insula, mPFC, and striatum during social trust violations (Lenow et al., 2014) and reduced activity in frontoparietal clusters during behavioral inhibition (Cará et al., 2019). These findings provide evidence for decreased neural encoding of errors that may relate to difficulties in learning or social interactions. Finally, in a study of biotypes in adolescent girls derived from functional activation patterns during an emotion processing task, the biotype representing subjects with high levels of violence exposure and high internalizing symptoms showed reduced activity of default mode network regions during emotional faces processing compared to a low-violence biotype (Sellnow et al., 2020), suggesting that deactivation of default mode regions while viewing emotional faces may be particularly characteristic of childhood assaultive violence exposure.

2.2.2. Emotional Abuse or Neglect.

Childhood emotional abuse or neglect is often linked to enhanced activation of emotion- and cognitive-processing brain regions. Unsurprisingly, the amygdala is commonly cited as being hyperactive in groups with childhood emotional abuse or neglect during processing of angry faces (Maheu et al., 2010; van Harmelen et al., 2013), sad faces (Dannlowski et al., 2013; van Harmelen et al., 2013), and fearful faces (Maheu et al., 2010; Tottenham et al., 2011; van Harmelen et al., 2013). Greater activity to fearful faces in the ventral medial prefrontal cortex (vmPFC) and greater activity to neutral faces in the ACC, inferior and middle frontal gyri were also observed in a group of previously institutionalized maltreated youth (Tottenham et al., 2011). Conversely, previously institutionalized youth also showed reduced activation to fear faces in the sgACC on an emotional face inhibition task (Tottenham et al., 2011) and reduced activation of cognitive control regions including the lingual gyrus, anterior cingulate cortex (ACC), and bilateral occipital cortices during correct inhibition (Bruce et al., 2013). Finally, neglect exposure in youth negatively correlated with differential responsiveness to congruent vs. passive viewing trials in the cuneus on an affective Stroop task, suggesting decreased discrimination between affective signals in a region related to the default mode network (Blair et al., 2019). These studies indicate that emotional abuse and neglect may be related to reduced encoding of emotional and feedback signals necessary for learning.

2.2.3. Adults with Childhood Abuse or Neglect Exposure.

Though fewer studies investigating regional functional activation changes exist in adults with childhood abuse, neglect, or ELS compared to youth, similar regions of altered activity emerge. Heightened functional activation of the amygdala to sad face viewing was observed in university adults with childhood emotional abuse or neglect, with neglect-only individuals showing greater activity to angry faces in the hippocampus, subgenual anterior cingulate cortex (sgACC), somatosensory cortex, fusiform gyrus, and superior temporal sulcus compared to those with abuse only (Puetz et al., 2019). Young adult twins with early life adversity showed increased activity compared to their non-exposed co-twin in a varietal of cognitive processing and limbic regions throughout an emotional Stroop task, including the dACC, ventral lateral PFC, pregenual ACC, basal ganglia, hippocampus and parahippocampus, superior frontal gyrus, superior parietal cortex, and cerebellum (Godinez et al., 2016). Adults with childhood trauma exposure also demonstrated reduced functional activation to emotional stimuli in various regions. dACC activity to negative stimuli was reduced with treatment in adult females with PTSD (Thomaes et al., 2012) and frontoparietal activation throughout an emotional Stroop task was reduced in twins with ELS compared to non-exposed twins (Godinez et al., 2016). Further, activation of several clusters throughout the cortex, including the anterior insula and parietal cortex, was reduced in healthy young adult men with mild childhood trauma exposure compared to non-exposed controls and negatively predicted trait anxiety, depressive symptoms, and executive function ability (Mirman et al., 2020).

2.3. Summary: Regional Neurocircuitry Studies in Childhood Trauma.

Investigations of differential functional activity related to childhood trauma exposure typically use emotional or threatening task stimuli to evoke activity of specific brain regions in emotion- and threat-processing circuits. Tasks in pediatric samples with trauma exposure suggest that the amygdala is commonly hyperactive to emotional faces when the amygdala is used as a priori search region; however, this effect is less robust when considering findings from whole-brain studies. Though the amygdala is a focus of many studies in childhood trauma exposure, regional activation differs as a function of trauma exposure throughout the brain. Other common regions of altered activation include the mPFC, dACC, anterior insula, and frontoparietal cognitive control regions. Patterns of differential activation may vary depending on type of childhood maltreatment exposure and studies in adults with childhood maltreatment indicate similar alterations in functional activation to children, suggesting that childhood trauma exposure may have common neurocircuitry correlates that persist throughout the lifespan.

2.4. Limitations of Regional Neurocircuitry Approaches

While regional neurocircuitry investigations such as the studies described in this section have provided invaluable insights into the relationships between childhood trauma exposure and brain function, several limitations exist to this approach. First, regional approaches to functional neurocircuitry are designed such that single regions are studied in isolation from the rest of the brain, therefore precluding any conclusions about brain activity that may be influencing the region of interest (ROI; (van den Heuvel and Hulshoff Pol, 2010). Second, a major limitation is the use of a priori ROIs in functional activation studies, rather than a more data-driven approach (Cole et al., 2010; Tomasi and Volkow, 2010). In studies of childhood trauma exposure, the amygdala and mPFC/dACC are commonly chosen ROIs because these regions are implicated in emotion- and fear-regulation, which the literature suggests is altered in this population. Consequently, many ROI-based studies utilize these regions, resulting in a limited understanding of relationships between childhood trauma and brain activity outside of this small circuit. An ROI-based approach may also be used to enhance statistical power in small samples by reducing stringent thresholds for multiple comparisons; however, these ROIs are often large, resulting in ambiguity of activity throughout the region (Poldrack, 2007). Whole-brain, exploratory approaches reduce experimenter bias and therefore are more appropriate when statistical power allows.

Another limitation with some of the work in regional neurocircuitry related to childhood trauma exposure is the use of passive fMRI tasks. As described above, many of the tasks utilized in samples with childhood trauma involve the use of emotional or threatening stimuli that a participant passively views in a scanner without active interaction with the stimuli. While this is a common methodological approach, passive viewing of stimuli on a screen lacks ecological validity because humans actively choose how to interact with the world. As such, studies that utilize active decision-making or learning may provide more information about cognitive and emotional deficits related to childhood trauma exposure than studies where participants passively view stimuli in the scanner. fMRI tasks that involve active decision-making, behavioral inhibition, cognitive reappraisal, or associative learning attempt to create situations that mimic how an individual with trauma exposure may interact with the world. Further utilization of these types of tasks may assist with identifying targets for therapeutic treatment to correct emotional- and cognitive-regulation deficits in this population.

3. Functional Connectivity and Resting State Investigations of Childhood Trauma Exposure

Task-based, voxelwise fMRI investigations of differences in regional activation as a function of childhood trauma exposure are important for determining which brain regions merit more attention as potential drivers of trauma-related psychopathology. Consequently, the brain regions commonly observed to have altered activity during emotion- and threat-processing tasks (i.e. the amygdala and mPFC/dACC) are frequently utilized as seeds for functional connectivity analyses in trauma exposed populations. Functional connectivity (FC) represents a technique for examining spatially disparate, temporally correlated activity between an a priori-chosen seed region and another ROI or the whole brain. FC studies have a distinct advantage over regional activation approaches because researchers can use FC to examine neural circuits both in task and during rest. Activity of these neural circuits can provide clues into larger patterns of disorder-related neural processing towards which new and existing treatments can be targeted. Resting-state fMRI (rs-fMRI) is an essential tool because the organization of functional networks at rest is hypothesized to support cognitive function (van den Heuvel and Hulshoff Pol, 2010), is correlated with functional organization during cognitive tasks (Bullmore and Sporns, 2012, 2009) and is related to structural connectivity networks (Bullmore and Sporns, 2009; Chen et al., 2008). For clinical populations, rs-MRI is advantageous because participants need only to lay still and fix their eyes on the screen, making the scanning session less burdensome for participants. rs-fMRI, along with investigations of FC during the same tasks used in the regional activation studies, has been valuable for understanding alterations in neural circuitry in relation to childhood trauma exposure.

3.1. Functional Connectivity in Resting State

3.2.1. Youth with Childhood Trauma Exposure.

Similarly to the regional activation studies discussed in the previous section, FC of the fronto-limbic circuit is commonly implicated as altered in youth with childhood trauma. Specifically, resting-state functional connectivity (rsFC) between the amygdala and mPFC was demonstrated to be reduced in youth with general trauma exposure (Heyn et al., 2018) as well as sexual abuse exposure (Aghajani et al., 2016) compared to non-traumatized youth, and between the amygdala and ventral ACC in adolescent girls with trauma exposure compared to non-traumatized girls (Zielinski et al., 2018). Limbic-ACC rsFC may also be related to amount of childhood abuse and neglect exposure, as one group observed a negative association between CTQ scores and FC from the amygdala and hippocampus to the sgACC that was related to internalizing symptoms in youth (Herringa et al., 2013). Conversely, evidence also exists for increased rsFC in childhood trauma-exposed youth compared to controls, with reduced negative amygdala-ACC rsFC in urban youth with violence exposure (Marusak et al., 2015; Thomason et al., 2015). Though the direction of altered FC in the fronto-limbic circuit (i.e. increased or decreased rsFC) related to childhood trauma exposure is unclear, the existing evidence from the rsFC literature lends some support to the regional activation studies that implicate the limbic and prefrontal systems as important players that may be mediating hyperarousal and failures in fear memory extinction in youth with childhood trauma exposure.

While the fronto-limbic circuit is dominant in many rsFC studies in trauma-exposed populations, alterations in rsFC expand beyond this circuit. rsFC between the rostral and dorsal ACC and the precuneus was found to be reduced in adolescent girls with interpersonal trauma exposure compared to non-exposed peers (Zielinski et al., 2018). Additional evidence suggests that altered rsFC throughout regions of the default mode network, including the mPFC, PCC, lingual gyrus, and hippocampus is associated with trauma- and anxiety-related symptoms (Nooner et al., 2013; Sheynin et al., 2020; Viard et al., 2019) and varies by affective style (Dark et al., 2020). Taken together, these studies suggest rsFC changes in many different neural circuits that may impact the ability of youth with trauma exposure to regulate their emotions, leading to emotional regulation difficulties and psychiatric symptoms into adulthood.

3.2.2. Adults with Childhood Trauma Exposure.

In contrast to the many studies suggesting the dominant fronto-limbic circuit in youth with trauma exposure, studies utilizing adult subjects with childhood trauma exposure suggest more widespread rsFC deficits. rsFC from the amygdala to multiple regions of the prefrontal cortex was reduced in adults with childhood emotional maltreatment (van der Werff et al., 2013) and was negatively associated with childhood emotional abuse exposure (Fan et al., 2014). In addition, rsFC from the insula to the mPFC, middle temporal gyrus, and middle frontal gyrus was decreased in adults with childhood trauma compared to non-traumatized healthy adults (Lu et al., 2017). rsFC throughout the default mode network was also reduced in adults with PTSD stemming from childhood abuse compared to controls (Bluhm et al., 2008; Philip et al., 2013). Additionally, rsFC from the hippocampus and amygdala to the mPFC and lateral PFC was negatively associated with childhood maltreatment in male combat veterans (Birn et al., 2014). In adults with bipolar disorder, childhood maltreatment was associated with reduced vmPFC-amygdala rsFC (Souza-Queiroz et al., 2016) and physical and emotional neglect were also negatively correlated with rsFC from the caudate to the fronto-parietal cortex and from the caudate to middle frontal gyrus, respectively (Hsieh et al., 2020).

Though much of the rsFC literature in adults with childhood trauma exposure indicates reduced rsFC in neural circuits involved in emotional regulation and internalizing symptoms, a few studies suggest that childhood trauma exposure may be associated with increased rsFC in these circuits. In the same sample of male combat veterans with childhood trauma exposure where reduced limbic-prefrontal rsFC was observed, the authors also note positive associations between the amygdala and hippocampus to the cerebellum and PCC/precuneus (Birn et al., 2014) and increased rsFC as a function of childhood emotional maltreatment was demonstrated in the ACC-precuneus circuit in the same sample where reduced amygdala-PFC rsFC was associated with emotional maltreatment (van der Werff et al., 2013). In a group of women with interpersonal violence exposure, assaulted participants demonstrated increased rsFC from the rostral ACC to the superior frontal gyrus and from the amygdala to the ventral ACC in comparison to non-assaulted women (Zielinski et al., 2018). Finally, recent evidence suggests that childhood maltreatment may be associated with increased rsFC within a theory-of-mind circuit, where severity of abuse was positively related to rsFC from the middle temporal gyrus to the orbital frontal and middle frontal gyri (Boccadoro et al., 2019). In sum, rsFC appears to be altered in multiple circuits throughout the cortex and subcortex in adults with childhood trauma exposure, representing a complex interplay of increased and decreased rsFC that may be associated with a wide range of psychiatric symptoms.

3.2. Functional Connectivity during Cognitive Tasks

In keeping with the regional studies of childhood trauma exposure, the tasks utilized with FC approaches have typically involved the presentation of emotional or fearful stimuli and isolate the fronto-limbic circuit. During the presentation of fearful faces, previously institutionalized youth demonstrate stronger amygdala-mPFC coupling compared to non-institutionalized youth, which was positively related to separation anxiety symptoms (Gee et al., 2013). Similarly, childhood adversity was positively related to FC of the right amygdala to the mPFC and of the bilateral hippocampus to the dorsal PFC during the presentation of negative-valanced facial emotions in healthy 18-year-olds (Herringa et al., 2016). For adults with childhood trauma exposure, an emotional faces matching task revealed a positive correlation between the amygdala and several limbic and prefrontal regions, suggesting increased fronto-limbic connectivity (Jedd et al., 2015). Conversely, youth with PTSD demonstrated reduced amygdala-mPFC coupling during threat-image viewing (Wolf and Herringa, 2016) and reduced dACC-dorsal medial prefrontal cortex (dmPFC), amygdala-dmPFC, and amygdala dlPFC connectivity to angry faces compared to control youth (Keding and Herringa, 2015). Finally, a study utilizing a cognitive reappraisal task before and after trauma-focused cognitive-behavioral therapy among adolescent girls with PTSD found that less amygdala-insula coupling during the reappraisal portion of the task was related to better PTSD symptom reduction during treatment (Cisler et al., 2016). This study also found that greater pre-treatment amygdala-dACC coupling during fear blocks of an implicit threat processing task predicted less symptom change during treatment in adolescent girls with PTSD (Cisler et al., 2015). These studies suggest that individuals with childhood trauma and resulting psychopathology may present with different neurocircuitry deficits related to threat- and emotion-processing compared to traumatized individuals who are resilient to psychopathology. The relationships between FC and PTSD symptom change during treatment suggest that these circuits are mechanisms of the underlying psychopathology and are not merely epiphenomenal relationships.

3.3. Network Patterns in FC Alterations

Though the fronto-limbic circuit is highly represented in functional connectivity studies of subjects with childhood trauma exposure, many studies reveal interesting patterns of altered FC that take the shape of functional networks. That is, the specific brain regions where FC is related to childhood trauma are not uniformly distributed across the brain; rather, the regions where FC is related to childhood trauma tend to fall within specific well-defined networks (see (Ross and Cisler, 2020) for examples of rsFC studies exhibiting network patterns in PTSD). Consistent with more general models of cognition and psychopathology (Bressler and Menon, 2010; Menon, 2013, 2011), the networks in which these specific circuits tend to fall include the default mode and salience networks (reviewed in more detail below). Unsurprisingly, multiple studies report childhood trauma-related associations with changes in FC between and within the salience network. In a sample of urban youth, amygdala-insula rsFC was more positive in trauma exposed youth compared to controls and amygdala-dACC rsFC was more negative (Thomason et al., 2015). Additionally, compared to adolescents without trauma exposure, trauma exposed adolescents without PTSS demonstrated greater amygdala-dACC rsFC (Sheynin et al., 2020) and PTSD symptom improvement was associated with greater amygdala-insula FC during negative image viewing in adolescent girls with PTSD (Cisler et al., 2016). Though FC in the salience network appears to be altered in subjects with childhood trauma exposure, whether trauma is related to increased or decreased FC in this network is yet uncertain.

When examining FC within the default mode network, a clearer picture of childhood trauma-related changes in FC emerges. Evidence for reduced rsFC of the default mode network in subjects with childhood trauma has been noted in healthy adults (Lu et al., 2017; Philip et al., 2013), adult women with PTSD (Bluhm et al., 2008), and adolescents with severe PTSD, where PCC-hippocampus rsFC negatively correlated with PTSD symptom severity and anxiety (Viard et al., 2019). Though two studies report increased rsFC within and between default mode regions in women with a history of child abuse (Boccadoro et al., 2019) and adolescents with PTSD (Patriat et al., 2016), much of the evidence to date suggests that childhood trauma and resulting psychopathology may be associated with reduced coupling of default mode network regions.

3.4. Summary: Functional Connectivity and Resting State Investigations of Childhood Trauma Exposure

The introduction of functional connectivity to the childhood trauma literature opened the door to exploring changes in bimodal circuits that are related to early life trauma exposure and resulting psychopathology. This literature highlights the role of enhanced and/or reduced FC in the fronto-limbic circuit in resting-state, complementing the regional activation studies that note trauma-related variations in activity of the amygdala, mPFC, and dACC. Similarly, the use of cognitive tasks in FC studies further emphasizes the role of altered fronto-limbic connectivity in trauma-exposed subjects that is predictive of psychiatric symptom change with treatment. While studies in youth show distinct fronto-limbic dysfunction in trauma-exposed subjects during resting-state that may be related to emotion regulation difficulties, studies in adults with childhood trauma exposure demonstrate widespread FC changes in connectivity of cortical and subcortical circuits that correlate with psychiatric symptoms. Finally, patterns of reduced FC in circuits that reflect the salience network and default mode network suggest that childhood trauma exposure is related to larger changes in FC outside of the classic fronto-limbic circuit which may contribute to increased symptom severity and internalizing psychopathology.

3.5. Limitations of Functional Connectivity Investigations of Childhood Trauma Exposure

Functional connectivity approaches overcome some of the limitations of regional activation approaches, namely the ability to use resting-state data and the extension of investigation outside of a single, isolated region. However, this method is still limited by the need to choose a priori seed regions from which to examine connectivity. Consequently, the literature in the field of FC in childhood trauma exposure is similarly centered around fronto-limbic circuitry at the expense of investigations into other potentially important FC changes that may relate to cognitive control or executive function. Whole-brain FC target approaches are preferable to a priori target selection when statistical power allows and have revealed interesting patterns of FC that otherwise would be neglected. Though FC methods still operate with the assumption that bivariate relationships between regions are acting in isolation from the rest of the brain, these methods are an important intermediary step towards more data-driven investigations of altered FC patterns related to childhood trauma exposure.

4. Emerging Trends and Future Directions for Functional Neuroimaging in Childhood Trauma Exposure

4.1. Network Dynamics

As described in section 3.3, functional connectivity changes from seed-based analysis methods focused on FC between two regions suggest patterns of altered functional topology that mimic important functional networks. In addition, recent evidence suggests that connectivity dynamics within the fronto-limbic circuit fluctuate with the dynamics of the functional networks containing the amygdala and mPFC in adolescent girls with PTSD (Cisler, 2017), signifying that the changes in FC of the fronto-limbic circuit may actually be driven by larger-scale network connectivity changes in youth with trauma exposure. Critically, an emerging literature utilizes methods of network neuroscience to specifically investigate network-level changes in activity related to childhood trauma and resulting psychopathology. This literature is generally data-driven and conceptualizes functional activity within the brain as a large, coordinated graph rather than distinct circuits acting in isolation; therefore, network methods surpass some of the conceptual limitations of regional and FC approaches. Additionally, conceptualizing the brain as a network allows for testing of hypotheses regarding hierarchical activity of the brain, from connectivity of individual ROIs to connectivity dynamics of specific modules and whole-brain organization. The brain, like other biological systems, does not act in distinct parts; therefore, a network-based approach likely better captures the functional reality of brain activity. A popular framework for network investigations in clinical samples is the “Triple Network Model” of Psychopathology (Menon, 2011). Within this framework, deficits of within- and between network function of the default mode, central executive, and salience networks are thought to be involved in multiple mental health disorders and contribute to psychiatric symptoms. These three networks are also commonly implicated in relation to childhood trauma exposure, as described in section 3.3.

4.1.1. Default Mode Network.

The default mode network guides self-referential mental activity, regulation of internal emotional state, and recollection of previous experiences (Menon, 2011; Raichle, 2015). In a sample of adolescent girls with PTSD, assault exposure was positively associated with within-network connectivity of the default mode network (Cisler et al., 2013), which may be related to internalizing or re-experiencing symptoms in youth with trauma exposure. Youth with PTSD also demonstrated reduced between network connectivity of the default mode network and a task-positive network (Patriat et al., 2016), suggesting that childhood trauma may impair efficient switching from an internal state to meet external cognitive demands.

4.1.2. Central Executive Network.

Though the central executive network is essential for goal-directed behavior, executive functioning, cognitive control, error encoding, working memory, and attention (Bressler and Menon, 2010; Dosenbach et al., 2007; Seeley et al., 2007), few studies have isolated the role of this network in childhood maltreatment. The studies that have examined this network in adolescents with childhood trauma exposure provide support for a reduced capacity of the central executive network to encode latent-state beliefs during a social learning task (Letkiewicz et al., 2020) as well as reduced within-network connectivity during fearful face presentation as a function of increasing assault exposure and PTSD severity (Cisler et al., 2013). These studies suggest that central executive network activity may be reduced in young people with childhood assault exposure, which may possibly mediate difficulties in emotion and attention regulation. More studies are needed to better define the role of this network in patients with childhood maltreatment.

4.1.3. Salience Network.

The salience network, which is anchored in the insular cortex and dACC, is important for the detection and mapping of internally and externally relevant events and serves as a dynamic switch between central executive and default mode network function (Bressler and Menon, 2010; Menon and Uddin, 2010). Within the salience network, adolescent girls with assault exposure demonstrated increased activity in response to fearful faces compared to non-assaulted girls (Cisler et al., 2019, 2013), suggesting heightened encoding of fear-related stimuli in young people with childhood assaultive violence exposure. Conversely, childhood trauma was related to weaker rsFC of major salience network nodes in adolescent subjects which related to increased executive dysfunction (Silveira et al., 2020). Adolescent girls with severe maltreatment also demonstrated reduced salience network encoding of negative prediction errors compared to girls with less or no maltreatment in a social learning task (Cisler et al., 2019). Thus, the role of the salience network in differentiating the effects of childhood maltreatment is likely dependent on context and may impact attention to important details in cognitively-demanding tasks.

4.1.4. Other Networks.

Though the salience, default mode, and central executive networks are commonly studied in the context of psychopathology, some evidence suggests that other functional networks might also be altered as a function of childhood trauma exposure. In a sample of adult patients with major depressive disorder, childhood emotional abuse and neglect was associated with increased resting state connectivity between the dorsal attention and somatomotor networks, whereas physical abuse and neglect was associated with increased connectivity of the cingulo-opercular and visual network and dorsal attention and ventral attention networks (Yu et al., 2019). While sexual abuse was not related to between-network connectivity, within-network connectivity of the dorsal attention network was positively related to childhood sexual abuse exposure in this sample (Yu et al., 2019). In adults with childhood maltreatment, within-network strength of an inhibitory control network consisting of the bilateral insula, frontal cortex, dorsal anterior and middle cingulate cortex, and bilateral inferior parietal cortex was inversely associated with reaction times on a stop-signal task in females with high childhood maltreatment, suggesting a sex- and maltreatment-dependent association with network connectivity and inhibitory control (Elton et al., 2014). More work is needed to determine which functional networks show consistently altered activity in relation to childhood trauma exposure.

4.1.5. Summary: Network Dynamics.

Network-based investigations of brain function in childhood trauma are a promising new avenue for research into complex systems of disordered function in this population. Early research in childhood trauma suggests that the functional networks involved in the Triple Network Model of psychopathology (Menon, 2011), including the default mode, salience, and central executive networks, show altered patterns of within- and between-network connectivity that may facilitate deficits in cognitive control, attention, and emotion regulation.

4.2. Local and Global Graph Dynamics

Though the literature regarding functional network dynamics in the context of childhood trauma exposure is sparse at this point, the studies summarized in the preceding section reflect a growing trend towards using functional neuroimaging to examine larger patterns of altered FC in clinical samples. The previous section focused on alterations in dynamics within and between specific networks within a larger graph of brain activity; however, the dynamics of regions within the graph as well as the graph as a whole may also be altered as a function of childhood trauma exposure. Examining local and global graph dynamics as they relate to childhood trauma exposure provides a hierarchical, “bottom-up” (e.g., a brain region’s influence on the graph of activity) or “top-down” (e.g., the graph of activity’s influence on a single brain region) approach to understanding brain function and patterns of activity that underlie psychopathology. In terms of regional graph-connectivity alterations, pediatric earthquake survivors with PTSD demonstrated increased nodal centrality of default mode regions including the middle frontal gyrus and hippocampus as well as decreased nodal centrality of the ACC, paracentral lobule, and thalamus compared to survivors without PTSD (Xu et al., 2018). Nodal centrality reflects the overall connectivity of the node across the network, suggesting that trauma-related psychopathology in children may alter the importance of certain regions’ connectivity profile within a graph, potentially re-organizing functional networks in a less efficient manner. In a sample of adult women, those who were resilient to major depression following ELS demonstrated reduced node degree and clustering of the ventral lateral prefrontal cortex and reduced clustering and efficiency of the dACC compared to susceptible and control (non-ELS) women (Cisler et al., 2013), implying that resilience to major depression following ELS may rely on reduced integration of these regions into an emotion-processing network.

An emerging literature suggests that childhood trauma exposure may also be related to large-scale changes in connectivity and organization of the brain’s functional graph. Graph modularity (i.e., the degree to which a larger graph is comprised of sub-modules / networks) at rest was positively associated with childhood emotional abuse in a sample of adolescent girls with PTSD, suggesting that emotional abuse is related to increased segregation of the functional graph as a whole (Cisler, 2017). This may lead to inefficient communication between functional networks that results in failures to disengage from threatening stimuli or difficulties in regulating emotional states. Similarly, adolescent females with assault exposure demonstrate increased global modularity while viewing emotional faces that is positively related to early life trauma severity (Cisler et al., 2018). Interestingly, whole-brain modularity was inversely related to fronto-limbic FC across the task (Cisler et al., 2018), implying that graph segregation may be the driver of childhood trauma-related alterations in this circuit. Finally, pediatric earthquake survivors with PTSD demonstrated reduced global clustering and reduced path length compared to non-PTSD survivors (Xu et al., 2018), providing additional evidence for hyper-modular functional brain organization that may be driving trauma-related psychopathology in children with a history of abuse and neglect.

4.2.1. Summary: Local and Global Graph Dynamics.

Conceptualizing the brain as a dynamic graph of connectivity links allows for investigations of the relationships between childhood trauma exposure and changes in whole-brain functional organization, which may explain the complex symptom profiles of individuals with trauma-related psychopathology. Early research in this field suggests that childhood trauma exposure is related to re-organization of large-scale brain networks, which may result in altered efficiency of cognitive processing, attention biases to threat, and lack of resilience to psychopathology following trauma exposure.

4.3. Learning and Decision-Making in fMRI

Much of the functional MRI literature in childhood trauma exposure to date has utilized either resting-state or stimulus-viewing tasks in which brain activity is measured while participants passively react to the stimuli on a screen. While this literature is necessary foundational work for understanding basic fear- and emotion-related circuitry, it provides little insight into the neural correlates that support active cognition and how those mechanisms may be impacted by childhood trauma exposure. An emerging trend in fMRI investigations in this population is a shift away from passive tasks and towards active, learning- and goal-oriented tasks in the scanner. Translation of results from these tasks follows a more direct path from lab to clinic compared to passive tasks, as learning and decision-making tasks may capture behaviors that are relevant to an individual’s interactions with the world. Utilizing a task designed to assess valence processing within the context of goal-relevance, adults with PTSD and severe childhood maltreatment demonstrated increased activation to goal-irrelevant (“distractor”) fear words in the dlPFC compared to adults with PTSD and mild or no childhood maltreatment (Fonzo et al., 2016). Within the childhood maltreatment PTSD group, less dlPFC activation to the goal-relevant stimulus was associated with increased re-experiencing symptoms, suggesting that failure to encode goal-relevant valence information may contribute to intrusive symptoms in adults with childhood trauma exposure (Fonzo et al., 2016). Childhood adversity was also associated with slower reactions to fearful faces on an emotional Go-NoGo task, with adversity showing a positive relationship with activation of the amygdala and a frontal cortex cluster containing the OFC and ACC to fearful faces (Wymbs et al., 2020). Relative to healthy comparisons, adolescents with childhood abuse exposure demonstrated increased activation of a large cluster containing regions of the prefrontal cortex and supplementary motor area during failed inhibition on a stop-signal task, which may reflect enhanced sensitivity to errors in neutral learning environments (Lim et al., 2015).

Conversely, growing evidence suggests that individuals with childhood trauma exposure may show reduced encoding of outcomes in social, emotional, and neutral tasks. Important insights into the neural correlates of learning deficits associated with childhood trauma have come from a series of studies investigating various behavioral differences in adolescent girls with assaultive violence exposure compared to non-exposed control girls. The first study utilized a social trust task and regressed outcomes of the task against whole-brain activation between groups, finding that girls with trauma exposure demonstrated reduced encoding of social trust violations within the pgACC, insula, mPFC, and striatum that scaled negatively with assault exposure severity (Lenow et al., 2014). Similarly, a second study using an independent sample provided evidence for reduced salience network encoding of negative prediction errors during reinforcement learning, despite overall higher network activation to fearful faces in the group with high maltreatment exposure (Cisler et al., 2019). A secondary analysis of this dataset examined the encoding of latent-state beliefs (i.e., a form of model-based learning where an individual forms abstract representations about the causal structure underlying a learning environment) throughout the task and noted that adolescent girls with severe sexual abuse history demonstrated less model-based learning than girls with none-to-moderate sexual abuse and showed reduced encoding of latent-state beliefs within the frontoparietal network (Letkiewicz et al., 2020). These three studies in an adolescent female subset provide compelling evidence for the characterization of a neural model of childhood trauma exposure to include failures in activation of important networks to support learning in multiple contexts unrelated to fear.

Adding to this literature, adolescents with ELS exposure demonstrated lesser accuracy and slower learning on the punishment trials of an instrumental learning task compared to non-ELS controls, which was accompanied by reduced activation of attention-related regions including the middle frontal gyrus during the reversal learning stages of the task (Harms et al., 2018), suggesting an impairment in learning flexibility. This impairment may follow ELS-exposed individuals into adulthood, as young adults with ELS made poorer betting decisions and poor risk adjustment on a monetary incentive delay task compared to control adults, which was accompanied by reduced activation of default mode network regions during the anticipation of reward phase (Birn et al., 2017).

A common approach for investigating behavioral differences related to childhood trauma exposure in fMRI are Go/No-Go tasks that evaluate neural correlates of inhibitory control. Though this task has multiple modifications, it generally involves the subject in the scanner responding to stimuli that are associated with a “Go” instruction (e.g. pressing a button) or a “No-Go” instruction (e.g. refraining from pressing the button). Accuracy and brain activity on the No-Go trials can be measured in relation to clinical variables as an indicator of differences in inhibitory control. An early study utilizing this task in youth with trauma exposure revealed that youth with posttraumatic stress symptoms demonstrated increased activation in the middle frontal cortex, cuneus, and inferior temporal gyrus during No-Go trials compared to controls, and increasing activation of the insula was related to higher symptom severity, avoidance, and hyperarousal symptoms (Carrion et al., 2008). Similarly, youth with ELS switched more slowly than controls on change trials and showed increased activation of cognitive control regions during response inhibition (Mueller et al., 2010), indicating a breakdown in the efficiency of inhibitory systems. Adolescents with childhood physical abuse exposure also demonstrated increased activation compared to psychiatric controls of the dorsomedial frontal cortex (Lim et al., 2015) and maltreated foster children showed increased activation to failed inhibition in the parietal lobule (Bruce et al., 2013), which may reflect increased sensitivity to errors in maltreated youth. Conversely, cognitive and inhibitory control regions may be less active as a function of childhood trauma exposure. For example, a study of young adults with ELS suggested reduced encoding of No-Go errors in the dlPFC (Harms et al., 2017) and young adolescents with violence exposure demonstrated reduced activation in various frontoparietal clusters compared to non-exposed controls (Cará et al., 2019). Likewise, maltreated children in foster care showed reduced activation of cognitive control regions during correct No-Go trials (Bruce et al., 2013) and recruited fewer cognitive control regions for successful inhibition compared to non-maltreated controls (Jankowski et al., 2017). These studies suggest individuals with childhood trauma exposure likely encode feedback using different neural correlates than non-traumatized individuals, which may lead to failures in inhibition, challenges with attention to salient details, and deficits in cognitive control.

4.3.1. Summary: Learning and Decision-Making in fMRI.

Whereas passive stimuli-viewing tasks can provide information about differential brain activity related to childhood trauma exposure, emerging research of learning and decision-making in fMRI is able to link differential brain function to clinically relevant behaviors, such as behavioral inhibition, social trust, and goal-oriented learning. This literature suggests a general breakdown in efficient inhibition that is facilitated through heightened neural attention to goal-irrelevant threatening stimuli, as well as reduced encoding of both errors and correct responses in cognitive- and reward-processing regions. Changes in functional activity, either the reduced or increased encoding of task-relevant stimuli, were related to decreased task accuracy, slower reaction times, and poor adjustment of learning strategies. By combining neuroimaging and active decision-making, researchers can begin to understand not only the neural correlates of childhood trauma exposure, but also potential mechanisms of behavioral deficits that underlie trauma-related psychopathology.

4.4. Future Directions for Functional Neuroimaging in Childhood Trauma

Functional neuroimaging methods are constantly evolving to provide researchers with novel, data-driven, means to overcome some of the limitations of regional and FC approaches. A popular method for ameliorating biases in neuroimaging tasks involves the use of multivariate or multivoxel pattern analysis (MVPA; Haynes, 2015; Poldrack, 2011). Using MVPA, researchers can use machine learning techniques to decode the mental state of a subject in the fMRI scanner during the presentation of stimuli in order to predict which patterns of brain activation support a cognitive function (Poldrack, 2011). As such, MVPA approaches allow researchers to make data-driven predictions about cognitive states based on patterns and quantity of information encoded in brain activity (Wang et al., 2020) in contrast to the traditional approaches which infer brain function based on cognitive states. Importantly, MVPA decoding can provide insight on individual differences in brain activity and connectivity without reliance on group-level inference (Wang et al., 2020). Though the utility of MVPA in fMRI was first demonstrated in 2001 (Haxby et al., 2001), to our knowledge no studies have used MVPA to decode cognitive or emotional regulation function in a sample with childhood trauma exposure. Using MVPA to determine patterns of functional activity during learning and decision-making tasks with emotional or fear stimuli could be beneficial to understanding individual differences in functional activity that may be modified for each unique patient to improve clinical and behavioral outcomes.

Functional imaging tasks that utilize neurofeedback through MVPA to regulate participants’ brain states in real-time provide promising opportunities for investigation of modifiable neural correlates of childhood trauma exposure. Neurofeedback training with electroencephalogram (EEG) for assistance with trauma-related disorders was introduced in the 1990s as a non-invasive tool for training patients to regulate their emotional state via real-time feedback on their brain state (Peniston and Kulkosky, 1991). During a neurofeedback session, patients watch a computer screen while being instructed to manipulate the image on the screen (i.e. moving a line towards a target) using their brain states only. Patients receive feedback in real-time on the correctness of their brain state by the outcome on the screen (i.e. the line gets closer or further away from the target), and work to adjust their brain state to reach the goal. Many versions of this task exist, including versions with emotional stimuli, and recent work has applied the techniques from EEG to fMRI. Across clinical groups, the amygdala, anterior insula, and ACC are commonly targeted sites for neurofeedback training of emotional regulation with encouraging results (Linhartová et al., 2019). As these regions are also frequently studied in childhood trauma exposure, neurofeedback approaches may provide new avenues for brain-based interventions. While no studies to our knowledge have utilized fMRI neurofeedback training in subjects specifically with childhood trauma exposure, a recent study using EEG neurofeedback in youth with developmental trauma demonstrated reduced trauma-related symptoms in the neurofeedback treatment group compared to the treatment-as-usual group, suggesting that neurofeedback may be an effective tool for clinical improvement in children (Rogel et al., 2020). fMRI neurofeedback studies in adults with PTSD also provide evidence for differences in response to neurofeedback training (Weaver et al., 2020) and improvements in the recruitment of neural networks over the course of training in subjects with PTSD (Nicholson et al., 2018). While this literature is relatively new, fMRI neurofeedback and MVPA decoding are exciting directions for improving our understanding of regulatory neural deficits in individuals with childhood trauma exposure. Like all fMRI-based applications for clinical methods, however, fMRI neurofeedback and MVPA decoding are limited in their abilities to translate to a clinical setting due to the need for specialized, expensive equipment and clinicians trained in analyzing and interpreting fMRI data. Applications with EEG similar to those described by Rogel and colleagues, may be more practical than fMRI for clinical usage due to EEG’s portability and substantially lower cost.

4.4.1. Summary: Future Directions for Functional Neuroimaging in Childhood Trauma.

Functional neuroimaging is an ever-evolving field. New methods such as multivariate pattern analysis and real-time neurofeedback provide data-driven strategies for reducing bias in the neuroimaging literature and strive towards clinical applications of neuroimaging technologies. While much more work is needed before these technologies can translate from the lab to the clinic, early studies using MVPA and real-time neurofeedback suggest that these approaches are likely to aid the field’s understanding of clinically-relevant neural deficits in individuals with childhood trauma exposure.

5. Conclusions

This review provides a framework for understanding the relationship between atypical neural functioning and adverse outcomes following childhood trauma exposure. As described in section 2, neuroimaging studies of regional activation have demonstrated the prominent role of fronto-limbic circuits in emotional regulation and threat processing deficits in trauma exposed individuals with PTSD. FC studies that use similar emotion- or fear-processing tasks in this population also emphasize aberrant functioning of the fronto-limbic circuit, which translates to difficulty regulating emotions and may underlie treatment outcomes, as described in section 3. Despite the breadth of research on this topic, findings regarding whether these processing deficits are related to hyper- or hypo-activation or connectivity are relatively mixed. One explanation for this lack of consistency may be that regional activation and FC approaches define ROIs a priori, rather than disseminating the results through a data-driven approach. Therefore, the literature remains centered around the fronto-limbic circuit, neglecting the study of regions concerned with general cognitive functioning. In addition, the design of many fMRI tasks relies on a reverse inference assumption and are limited by passive engagement with stimuli that fail to reflect real-world contexts. Thus, aberrant activity within the fronto-limbic circuit is likely only a minor character in the story of fMRI correlates of childhood trauma exposure. Additionally, there have been numerous different types of cognitive tasks employed during neuroimaging, and the exact type of cognitive process engaged would be expected to moderate the neurocircuitry mechanism observed to vary as a function of early life adversity (Cisler et al., 2019).

Though limited by many of the same drawbacks as regional activation approaches, FC research provides a framework for understanding more widespread activation patterns by extending beyond a specific predetermined region. Evidence from rsFC studies specifically suggest that childhood trauma is associated with a complex interplay of impaired circuits, leading the way for network investigations. Examining network dynamics provides an avenue for investigating widespread neural deficits in trauma-exposed individuals, and reflects the importance of studying cognitive processing networks such as the default mode, central executive, and salience networks. Dysfunction of these key networks in trauma exposed individuals, along with other networks such as the dorsal attention, somatosensory, and visual networks, highlight that trauma exposure is not only associated with deficits in key emotion processing regions (i.e., the fronto-limbic circuit) but also with widespread functional alterations to brain regions involved in complex cognitive processing.

Broadly, the field concerned with identifying neurocircuitry correlates of childhood trauma is limited by inconsistencies in study samples, thus contributing to heterogeneity in research findings. As seen in the subsections of sections 2 and 3, findings from functional neuroimaging studies within the field vary depending on type of trauma exposure, age of the sample, and presence of psychopathology in trauma-exposed or control groups. Heterogeneity in findings obscures the interpretations that can be made and complicates the applications of the research for clinical use. In order to address these concerns, future research should exercise full transparency when reporting the characteristics of a sample. Special consideration should be given when reporting recruitment strategies, type or types of adverse experience under investigation, mental health diagnoses, age of the sample and age of adverse life experiences, and other specific inclusion and exclusion criteria. Due to the ambiguity in the term “adverse childhood experiences,” researchers should take care to characterize samples along multiple domains of experience and mental health conditions to improve translatability of neuroimaging literature to clinical applications.

Emerging trends in neuroimaging research offer promising contributions to studying the neurocircuitry of psychopathology following trauma exposure. Goal-oriented decision-making tasks that emulate real-world situations which individuals with trauma exposure navigate daily are providing valuable insights into new treatment models that may improve clinical outcomes for patients. Data-driven approaches such as MVPA and practical interventions such as neurofeedback are promising avenues for studying widespread neural deficits and psychopathology following trauma exposure that merit more investment for future research. Functional neuroimaging research has evolved substantially over the last thirty years, from simple regional activation studies to machine learning-based investigations of the entire brain. Understanding of the brain’s functional activity deficits in response to childhood trauma exposure has evolved along with advancements in neuroimaging, from narrow conceptualizations of heightened amygdala activation to fearful stimuli to knowledge of widespread neural deficits that relate to adverse clinical, cognitive, and behavioral outcomes. Ultimately, these new methodological considerations can enrich the implications drawn from research in order to identify treatment methods that address emotional- and cognitive-regulation deficits associated with psychopathology following childhood trauma exposure.

Box 1: Definitions for Acronyms Used in Review.

Acronym	Meaning	Definition	First Appearance in Paper
ACC	Anterior Cingulate Cortex	Large prefrontal cortex brain region composed of BA 24, 32, and 33	Section 2.2.2
ACE	Adverse Childhood Experience(s)	Index of early life stressors including parental divorce, parental mental illness or imprisonment, poverty and deprivation, death or illness in the family, bullying, peer rejection, exposure to violence, and abuse	Section 1
CAPS	Clinician Administered PTSD Scale for DSM-V	Structured interview to assess presence and severity of PTSD symptoms	Section 1
CTQ	Childhood Trauma Questionnaire	Self-report assessment of childhood maltreatment experiences with subscales for physical abuse, sexual abuse, emotional abuse, emotional neglect, and physical neglect	Section 1
dACC	Dorsal Anterior Cingulate Cortex	Dorsal and posterior division of the prefrontal cortex (BA 32)	Section 2.1.1
dlPFC	Dorsal Lateral Prefrontal Cortex	Lateral portion of the prefrontal cortex corresponding to BA 46	Section 2.1.3
dmPFC	Dorsal Medial Prefrontal Cortex	Large portion of the medial prefrontal cortex corresponding to BA 8 and 9	Section 3.2
EEG	Electroencephalogram	Portable device for measuring brain electrical activity with electrodes on the scalp	Section 4.4
ELS	Early Life Stress	Adverse events, including accidents, illnesses, deaths, neglect, and abuse, experienced before age 18	Section 2.1.3
FC	Functional Connectivity	Measure of temporal coherence between brain regions in a circuit	Section 1.3
fMRI	Functional Magnetic Resonance Imaging	Non-invasive technique that measures brain activity by detecting changes in blood flow	Section 1.2
IPV	Interpersonal Violence	Exposure to physical assault, physical abuse by a caregiver, sexual assault, domestic violence, or other assaultive violence	Section 1.1
mPFC	Medial Prefrontal Cortex	Anterior ventral division of the prefrontal cortex	Section 2.1.2
PCC	Posterior Cingulate Cortex	Posterior portion of the cingulate gyrus corresponding to BA 23 and 31	Section 2.2.1
PFC	Prefrontal Cortex	Portion of the cerebral cortex encompassing the frontal lobe	Section 2.1.2
pgACC	Perigenual Anterior Cingulate Cortex	Section of the anterior cingulate cortex located near the genu of the corpus callosum	Section 2.2.1
PTSD	Posttraumatic Stress Disorder	Mental health condition that develops following the experience of a trauma event(s). Characterized by changes in mood and cognitions, hyperarousal, avoidance of trauma reminders, and intrusive re-experiencing of the event	Section 1.1
PTSS	Posttraumatic Stress Symptoms	Symptoms that develop following the experience of a traumatic event, including changes in mood and cognitions, hyperarousal, avoidance of trauma reminders, and intrusive re-experiencing of the event	Section 2.1.3
ROI	Region of Interest	Brain region or regions chosen a priori for investigation based on theory or hypotheses	Section 1.3
rs-fMRI	Resting-State Functional Magnetic Resonance Imaging	Measurement of brain activity in MRI in the absence of a task or stimulus	Section 3
rsFC	Resting-State Functional Connectivity	Measurement of temporal coherence between brain regions in the absence of a task or stimulus	Section 3.1
sgACC	Subgenual Anterior Cingulate Cortex	Anterior portion of the anterior cingulate cortex, located below the genu of the corpus callosum	Section 2.2.3
vmPFC	Ventral Medial Prefrontal Cortex	Broad brain region in the lower central region of the prefrontal cortex, including the orbitofrontal cortex (BA 11)	Section 2.2.2

Highlights.

• Childhood trauma is prevalent and related to childhood and adult psychopathology

• Research on the neural correlates of trauma and psychopathology has primarily found functional abnormalities in limbic regions

• Past research has been limited by theory-driven approaches and passive tasks

• Future research should employ data-driven approaches and complex decision-making tasks

• Improvements will support the translatability of research findings to clinical settings