Amygdala connectivity is associated with withdrawn/depressed behavior in a large sample of children from the Adolescent Brain Cognitive Development (ABCD) Study ®

Many psychopathologies tied to internalizing symptomatology emerge during adolescence, therefore identifying neural markers of internalizing behavior in childhood may allow for early intervention. We utilized data from the Adolescent Brain and Cognitive Development (ABCD) Study ® to evaluate associations between cortico-amygdalar functional connectivity, polygenic risk for depression (PRS D ), traumatic events experienced, inter-nalizing behavior


The importance of examining neural correlates of internalizing behavior in childhood
Internalizing behavior refers to negative problematic behavior, thoughts, and emotions directed at the self, including depression, anxiety, withdrawal, and somatic complaints (Achenbach, 1978).In childhood, this behavior is associated with increased distress and functional impairment (Liu et al., 2011) and may indicate an increased risk for future mental health problems (Goodwin et al., 2004).With the prevalence of internalizing disorders dramatically increasing during adolescence late childhood represents a critically important window in which to examine potential brain-based markers of mood and anxiety problems (Herpertz-Dahlmann et al., 2013).Indeed, examining the neural correlates of internalizing behavior in childhood may help to elucidate mechanisms through which this behavior relates to future mental illness.
Evidence-based assessment of mental illness has revealed aspects of dimensionality with regard to many forms of psychopathology, including internalizing behavior (Hudziak et al., 2007;Kotov et al., 2017;Krueger et al., 2018;Watson et al., 2022).Most psychopathology symptoms occur across a range of conditions and are not confined to specific categorically defined mental disorders (Cai et al., 2021).Clinical and normative levels of internalizing symptoms are likely underpinned by overlapping neural substrates (Albaugh et al., 2017a(Albaugh et al., , 2017b)), and therefore, it may be advantageous to look at variation in a population as it relates to dimensional measures of internalizing symptoms rather than a categorical measure, such as what is employed by the DSM (American Psychiatric Association, 2013).

Amygdala connectivity as a predictor of internalizing behavior
The amygdalae have been shown to contribute to the processing of emotion through their rich structural and functional connections with other subcortical and cortical structures (Roy et al., 2009;Young et al., 1994;Ghashghaei et al., 2007;Ghashghaei and Barbas, 2002;Carmichael and Price, 1995;Ray and Zald, 2012).Prior associations between amygdala connectivity and internalizing behavior have been frequently reported in studies utilizing relatively small sample sizes (Padgaonkar et al., 2020;Feurer et al., 2021;Rogers et al., 2017;Kim et al., 2011;Gee et al., 2013;Qin et al., 2014).In particular, a substantial body of literature has implicated functional connectivity between the cerebral cortex and amygdalae in the etiology of internalizing problems in youth (Rogers et al., 2017;Burghy et al., 2012, Gee et al., 2013;Qin et al., 2014).However, small sample sizes have low statistical power (Button et al., 2013), often yield inflated effect sizes that are not reproducible (Marek et al., 2022), and are limited in their ability to fully represent the relevant variability of individual differences in behavior across populations.Furthermore, typical neuroimaging samples tend to be biased towards educated Western groups of people (Henrich et al., 2010).With a sample size of nearly 12,000 children collected across the continental United States, the Adolescent Brain and Cognitive Development (ABCD) Study can enable the generation of more reproducible and generalizable results in a population-based sample (Volkow et al., 2018).

Potential factors may impact the relationship between amygdala connectivity and internalizing behavior
Traumatic events, sex, age, and genetic risk for depression may impact the relationship between amygdala connectivity and internalizing behavior.Maltreatment has been linked to increased amygdala reactivity (Herringa et al., 2016), which may explain the observed associations between maltreatment in childhood and internalizing symptomatology (Cicchetti & Toth 1995, 2005;Kim & Cicchetti 2010).Sex-specific patterns of amygdala functional connectivity (Alarcón et al., 2015;Padgaonkar et al., 2020) may explain the significant sex differences observed in internalizing symptomatology and diagnoses later in life (Angold et al., 2002;Crick and Zahn-Waxler, 2003;Hankin et al., 1998;Lewinsohn et al., 1995).Changes in amygdala functional connectivity and associated white matter pathways (Gee et al., 2013;He et al., 2016;Albaugh et al., 2017;Albaugh et al., 2019) may contribute to observed associations between age-related changes in amygdala connectivity and anxiety (Clewett et al., 2014).Polygenic risk for depression (PRS D ), which has been linked to depressive symptomatology (Howard et al., 2022), amygdala connectivity patterns (Holmes et al., 2012), and prefrontal cortical measures (Cattarinussi et al., 2022;Schmitt et al., 2021), may also impact associations between depressive symptomatology and amygdala measures (Acosta et al., 2020).In infancy, PRS D has been shown to moderate an association between prenatal maternal depressive symptoms and infant amygdala volumes (Acosta et al., 2020).Thus, individual differences in amygdala circuitry may arise from genetic variability that contributes to risk for internalizing behavior.

Present study
In the current study, we utilize a large population-based sample of children to investigate associations between amygdala functional connectivity and dimensional assessments of internalizing psychopathology.Given the inherent heterogeneity within the internalizing construct, we wanted to evaluate if particular subscales had distinct neural supports.Additionally, we examine the extent to which these associations are impacted by genetic and environmental risk factors.Specifically, traumatic events experienced and polygenic risk for depression were examined as potential moderators.We focused our study on amygdala connectivity, given the numerous functional neuroimaging investigations showing that cortico-amygdalar connectivity is associated with internalizing behaviors (Rogers et al., 2017;Burghy et al., 2012, Gee et al., 2013;Qin et al., 2014;Padgaonkar et al., 2020;Feurer et al., 2020;Rogers et al., 2017;Kim et al., 2011;Gee et al., 2013;Qin et al., 2014).What remains to be determined is if these observed associations persist when examined in a large sample of children and when dimensional assessments of internalizing behaviors and additional risk factors are considered.

Participants
Participants included in these analyses were a part of the ongoing ABCD study, a population-based study tracking 9-10-year-old children recruited from 21 sites across the United States.The study was approved by the Institutional Review Board at the University of California, San Diego.Written informed consent and assent were obtained from parents and children, respectively.Data from the baseline and 1-year follow-up time points were included in the present study.Baseline data were examined, except for the child-reported Brief Problem Monitor scale and the parent-reported Short Social Responsiveness Scale, which were obtained at the 1-year follow-up.Data was accessed from the National Institutes of Mental Health Data Archives.Sample size varied between analyses based on the number of participants with data for each measure.Participant numbers and demographic and psychometric data are provided in Table S1 and Table S2 in Supplementary Materials.

Behavioral, genetic, and environmental factors 2.2.1. Internalizing problems
Parent-reported measures of internalizing behavior were obtained using the Child Behavior Checklist (CBCL; age 6 to 18 form) (Achenbach & Rescorla, 2001).Specifically, parents rated the extent to which behaviors were characteristic of their child over the past 6 months.Items were combined to create internalizing subscale scores for the following scales: withdrawn/depressed, anxious/depressed, and somatic complaints.These subscales were also combined to create a broader internalizing behavior composite score.Additionally, self-reported measures of internalizing behavior were obtained at the 6-month follow-up using the Brief Problem Monitor for Youth (BPM-Y) (Achenbach, 2009).

Traumatic events experienced
Traumatic experiences were assessed as part of the parent-report post-traumatic stress disorder module of the Kiddie Schedule for Affective Disorders and Schizophrenia (KSADS-5) (Barch et al., 2018).Data were obtained from a structured diagnostic interview with parents via computer in reference to their child (Kaufman et al., 1997).The post-traumatic stress disorder module contains 17 items (for example, witnessed death or mass destruction in a war zone) rated from 0 to 1 (0no, 1 -yes) to assess the number of traumatic events the children experienced.

Genotyping and polygenic risk score for depression
DNA was extracted from saliva and blood samples and genotyped using Affymetrix Axiom Smokescreen Array platform (see PMID: 37093311 for further detail).Genotypes were imputed using 1000 Genomes Project phase 3 reference panel for Europeans.SNPs that did not meet quality control criteria (Minor Allele Frequency < 0.01; Genotype Call Rate < 95%; Hardy-Weinberg Equilibrium < 1×10-6) were excluded.Genetic variants imputed with lower accuracy (R2 < 0.6), insertion/deletions, and palindromic SNPs were excluded, resulting in 5,183,147 SNPs.Related participants were excluded (grm-cutoff 0.05).The polygenic risk score (PRS) was calculated based on the largest genome-wide association study (GWAS) on depression (Howard et al., 2019), which meta-analyzed 33 cohorts of the Psychiatric Genomics Consortium and UK Biobank as described in Howard et al. (2018), including 500,199 participants of European ancestry (170,756 cases and 329,443 controls).This PRS was calculated using PLINK 2.0 (http s://www.cog-genomics.org/plink/2.0/).Index variants were identified by clumping using an r2 threshold of 0.1 with a 1000 kb window using the 1000 Genomes (EUR) as reference.Ancestry was determined by principal component loadings compared with population of reference from 1000 genomes.

MRI and fMRI data acquisition and processing
In the present study, we used resting-state functional magnetic resonance imaging (rsfMRI) data from the baseline time point obtained through NDA's Collection 3165 -DCAN Labs ABCD-BIDS (htt ps://collection3165.readthedocs.io/en/stable/#documentation-guide)using the NDA Collection 3165 ABCD-BIDS Downloader (https://github.com/DCAN-Labs/nda-abcd-s3-downloader).Image acquisition details can be found in Hagler et al., 2019.Data were minimally processed using the abcd-hcp-pipeline25 (https://github.com/DCAN-Labs/abcd-hcp-pipeline), a modified version of the HCP pipeline (Glasser et al., 2013).Current research indicates that vertex-level surface-based approaches to examine fMRI cortical connectivity data, rather than voxel-based approaches, allow for better alignment of cortical structures across individuals resulting in improved signal-to-noise (Glasser et al., 2013;Van Essen 2004;Van Essen et al., 1998;Sereno et al., 1995).The modified version of the HCP pipeline combined 2D surface meshes for cortical surfaces with 3D voxels for subcortical volumes to improve alignment and reduce noise across subjects (Glasser et al., 2016;2021).Details regarding data processing can be found in the documentation provided by the DCAN lab (https://collection3165.readthedocs.io/en/stable/pipeline/).After processing, data were curated based on head motion so that all frames with a framewise displacement (FD) threshold of greater than 0.2 mm were removed.Data included in our analyses met quality control standards set by the Data Analysis, Informatics & Resource Center (DAIRC) for preprocessed data and motion censoring (Power et al., 2012).Only resting-state scans from participants with at least 8 minutes of usable data were included in analyses.

Resting-state amygdala functional connectivity
We extracted right and left voxel-based amygdala regions of interest (ROIs) from the Gordon Parcellation.This parcellation was derived from the rsfMRI data of 120 healthy individuals (Gordon et al., 2016) and consists of 333 cortical surface-based parcels associated with 13 resting-state networks in addition to 19 subcortical voxel-based ROIs.In separate analyses, right and left amygdala ROIs were used as seed regions of interest to correlate with all other regions from the Gordon parcellation.This resulted in a vector of correlation values between each amygdala ROI and all other ROIs.The right and left amygdala were evaluated separately.

Statistical analyses
Linear mixed-effects models were used to identify patterns of amygdala connectivity to Gordon ROIs (which includes cortical and subcortical regions) associated with overall internalizing score and internalizing subscales somatic complaints, anxious/depressed, and withdrawn/depressed syndromes in separate models.Analyses included scanner ID, sex, age, parental education, handedness, and puberty status as fixed effects with family nested in scanner type as random effects.To correct for multiple comparisons, a false discovery rate (FDR) correction was applied (FDR correction= p<.05) to the cortical surface.In separate analyses, PRS D , age, sex, and traumatic events were tested as potential moderators through the creation of interaction terms.Only subjects without missing data were included in each analysis.Subject numbers for each analysis varied accordingly (see Supplementary Materials).

Amygdala connectivity is associated with withdrawn/depressed behavior
We first examined associations between right and left amygdala connectivity, overall internalizing behavior, and internalizing subscales.Results indicated left amygdala connectivity was associated with withdrawn/depressed symptomatology.Specifically, stronger connectivity to dorsal attention network regions (right midtemporal, β = 0.004, p = .04bilateral precentral gyrus: right, β = 0.005, p = .01;left, β = 0.004, p = .02,right inferior precentral gyrus β = 0.004, p = .02)and to frontoparietal network region (inferior parietal cortex; β = 0.004, p = .03)was associated with greater withdrawn/depressed symptoms.However, stronger connectivity to an auditory network region (posterior insula; β = -0.005,p = .02)was associated with less withdrawn/depressed behaviors.No associations were found between amygdala connectivity and internalizing behavior overall or anxious/depressed and somatic complaints subscales.Additionally, amygdala connectivity was not associated with the child-reported measure of internalizing.Thus, only aspects of left amygdala connectivity were associated with withdrawn/ depressed behavior (see Fig. 1).Connectivity of the right amygdala was not associated with any symptoms.

Potential moderators do not impact the relationship between amygdala connectivity and withdrawn/depressed behavior
Next, in a series of post-hoc exploratory analyses, we examined several additional factors that might impact the observed associations between amygdala connectivity and withdrawn/depressed behavior.In separate models, we examined PRS D , age, sex, and traumatic events as potential moderators of the association between cortico-amygdala connectivity and CBCL withdrawn/depressed score.Analyses included scanner ID, sex, age, parental education, handedness, and puberty status as fixed effects with family nested in scanner type as random effects (FDR correction = p < .05).To test the degree to which these variables moderate the relationship between amygdala connectivity and withdrawn/depressed behavior, the following model was run for each of these variables: Amygdala connectivity = intercept + ԁ 1 + β 1 (Age) + β 2 (Scanner ID) + β 3 (Sex) + β 4 (Education) + β 5 (Handedness) + β 6 (Puberty Status) + β 7 (Main Effect) + β 8 (moderating variable х CBCL withdrawn/ depressed) + e, where ԁ 1 represents the random effects of family nested in scanner type and e represents error.Given that the PRS D was derived from a GWAS of participants with only European ancestry, we tested PRS D as a moderator using the full sample of participants and, in a separate analysis, restricted our sample to participants with European-only ancestry.4 multidimensional scale (MDS) components were included in the model to account for population stratification.PRS D (in all participants and European-only participants), age, sex, and traumatic events did not moderate associations between amygdala connectivity and withdrawn/ depressed behavior.

Amygdala connections associated with withdrawn/depressed behavior are associated with social impairment
In a series of post hoc exploratory analyses, we tested the hypothesis that amygdala connectivity linked to withdrawn/depressed behavior may be tied to problems relating to social behavior.Specifically, we focused on measures of social anxiety, social impairment, and social problems obtained in the ABCD study (see Supplementary Materials).Results indicated significant associations between withdrawn/ depressed-related cortico-amygdalar connectivity and social difficulties.Specifically, stronger coupling to dorsal attention (precentral gyrus, β = 0.005, p = .05)and frontoparietal (inferior parietal cortex, β = 0.006, p = .03)network regions was associated with greater social anxiety.Stronger coupling to dorsal attention network regions (right midtemporal, β = 0.001, p = .05and precentral gyrus, β = 0.001, p = .05)was associated with greater social difficulties.Stronger coupling to dorsal attention (right midtemporal, β = 0.002, p = .05and precentral gyrus, β = 0.003, p = .01)network regions and weaker coupling to auditory (insula, β = -0.003,p = .01)network regions were associated with greater social problems.

Summary of results
The current study revealed significant associations between corticoamygdalar resting-state functional connectivity and CBCL withdrawn/ depressed problems in a large, demographically diverse sample of children.In particular, withdrawn/depressed behavior was associated with functional connectivity between the left amygdala and several cortical regions belonging to the dorsal attention, frontoparietal, and auditory networks.Post hoc analyses indicated that some of the corticoamygdalar connectivity patterns associated with withdrawn/depressed behavior were also linked to aspects of social behavior.Specifically, amygdala connectivity to dorsal attention network regions, a frontoparietal network region, and an auditory network region were associated with greater social anxiety, social impairment, and social problems.Although speculative, these results suggest that withdrawn/depressed-related amygdala connectivity patterns observed in the present study may stem from underlying impairments in social behavior.Future work is needed to directly test this possibility.

Amygdala connectivity largely unassociated with overall internalizing symptoms and other subscales despite prior findings
Many previous studies have shown strong associations between amygdala connectivity and both clinical and normative levels of internalizing behavior (Padgaonkar et al., 2020;Baur et al., 2013;Clewett et al., 2014;Kim and Whalen, 2009;Qin et al., 2014).However, the current study did not find such associations.There are a few potential reasons for this discrepancy.First and foremost, prior studies may have seen false significant relationships due to low sample sizes (Marek et al., 2022).Smaller sample sizes may have led to over-fitting and unreliable parameter estimates.Many published neurodevelopmental associations may represent severely inflated effect sizes (Button et al., 2013;Ioannidis, 2008).Further, studies with smaller sample sizes may have lacked the power to detect associations between amygdala connectivity and internalizing subtypes.Importantly, our null findings are in line with a few recent studies which have found no significant associations between internalizing problems and imaging measures (Belov et al., 2024;Nakua et al., 2022a;Winter et al., 2022).
Interestingly, some of our results are in line with prior literature in adult populations.Specifically, alterations in the dorsal attention network and auditory cortex have been seen in patients with major depressive disorder (Yan et al. 2019, Zwanzger et al., 2012).In particular, individuals with MDD have shown decreased network connectivity between the dorsal attention network and sensory motor network (Yan et al., 2019), and stronger auditory cortex activity in response to sine tones (Zwanzger et al., 2012).Further, increased amygdala connectivity to the dorsal attention network has been previously associated with depressive symptoms (Han et al., 2015).This suggests potential continuity in the neural circuitry underpinning withdrawn/depressed behavior from childhood to adulthood.Moreover, these connectivity patterns in childhood may indicate increased risk for future psychopathology.The degree to which cortico-amygdala connectivity relates to internalizing problems may be age-dependent, so it may be possible that it takes time for these brain-behavior associations to manifest as children age and these patterns stabilize (Gee et al., 2013).

Small associations in large studies are expected
Although associations found in our ABCD study were small, these associations are well powered (Dick et al., 2021).Small effect sizes are commonly observed between psychological variables in large data sets (Owens et al., 2021).Evidence for this has been shown in a meta-analysis of psychology and psychiatry literature, which found that most associations between clinically relevant variables are very small (r = 0.15 -0.3), and most clinically important effects are even smaller (Meyer et al., 2001).Further, in an analysis of significant associations between multimodal imaging and health-related outcomes, even the most significant associations explained only 1% of the variance in the outcomes (Miller et al., 2016).Given the inherent heterogeneity in any psychological construct, we might expect small effects when relating brain measures and behavior (Feczko et al., 2019b).

Limitations/ future directions
The current study characterized associations between internalizing behavior and cortico-amygdala resting-state functional connectivity in a large population-based sample of pre-adolescents.Some limitations should be noted.First and foremost, the GWAS, from which the polygenic risk score for depression was derived, was conducted in individuals of European-only descent, therefore our results are not generalizable to the larger population.There is consistent evidence that polygenic risk scores derived from predominantly European samples predict individual risk with higher accuracy in Europeans than in non-Europeans (Hoggart et al., 2023a;Li & Keating, 2014;Martin et al. 2017;Peterson et al., 2019;Ruan et al., 2022;Weissbrod et al., 2022;Wojcik et al., 2019).Though the lack of polygenic risk score effects was observed in both the full sample and European only sample, potentially suggesting that a poor fit to the genetics of non-Europeans may not be the only source of null results.Next, the current study is a cross-sectional study, which does not examine changes over time.Because we are only looking at a snapshot of development, it is important to acknowledge that it may take time for associations between amygdala connectivity and internalizing behavior to manifest.To address this possibility, in our future work, we aim to examine associations between trajectories of internalizing behavior over time in relation to change in amygdala connectivity over time as longitudinal data from the ABCD dataset becomes available.Further, there is inherent heterogeneity in any psychological construct, therefore by just focusing on internalizing behavior, we may be missing profiles of psychopathology as the manifestation of symptoms may arise from different mechanisms (Feczko et al., 2019).As additional data becomes available, we are considering additional constructs and factors that may impact associations between internalizing and amygdala connectivity, such as sleep quality (Klumpp et al., 2018), in our ongoing research efforts.While the current study builds upon prior work by examining associations between cortico-amygdalar connectivity and internalizing behavior in childhood, future work should explore potential patterns of internetwork connectivity that may be involved with internalizing behaviors.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Development (ABCD) Study (https://abcdst udy.org), held in the NIMH Data Archive (NDA).This is a multisite, longitudinal study designed to recruit more than 10,000 children age 9-10 and follow them over 10 years into early adulthood.The ABCD Study® is supported by the National Institutes of Health and additional federal partners under award numbers U01DA041048, U01DA050989, U01DA051016, U01DA041022, U01DA051018, U01DA051037, U01DA050987, U01DA041174, U01DA041106, U01DA041117, U01DA041028, U01DA041134, U01DA050988, U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038, U01DA041148, U01DA041093, U01DA041089, U24DA041123, U24DA041147.A full list of supporters is available at https://abcdstudy.org/federal-partners.html.A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/ consortium_members/.ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in the analysis or writing of this report.This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators.The ABCD data repository grows and changes over time.RBC was also supported by R25MH081482.Dr. Albaugh was supported by K08 MH121654 and a Young Investigator Grant from the NARSAD/Brain Behavior Research Foundation.