Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Independent and interactive effects of depression genetic risk and household socioeconomic status on emotional behavior and brain development
(USC Thesis Other)
Independent and interactive effects of depression genetic risk and household socioeconomic status on emotional behavior and brain development
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INDEPENDENT AND INTERACTIVE EFFECTS OF DEPRESSION GENETIC RISK AND
HOUSEHOLD SOCIOECONOMIC STATUS ON EMOTIONAL BEHAVIOR AND BRAIN
DEVELOPMENT
By
Claire Edwards Campbell
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(NEUROSCIENCE)
August 2024
Copyright 2024 Claire Edwards Campbell
ii
ACKNOWLEDGMENTS
This work would not be possible without all the Adolescent Brain Cognitive DevelopmentSM (ABCD)
Study (https://abcdstudy.org) participants and their families; a huge thank you to all who participated,
organized, and collected the data. This is a multisite, longitudinal study designed to recruit more than
10,000 children aged 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.
In addition to the ABCD Study, this work was also supported by funding from the National Institutes
of Health: NIEHS T32ES013678 (Gauderman and McConnell, trainee: Campbell) and NIMH
F31MH131347 (Campbell). Also, the Advanced Statistical Methods in Neuroimaging and Genetics
(NINDS R25NS117281) was crucial in developing the polygenic risk score calculations, with special
recognition to Dr. Andrey Shabalin. I also want to acknowledge three workshops that at the beginning
of my predoctoral career helped me understand the ABCD Study along with longitudinal modeling and
reproducible data analyses: ABCD Workshop on Brain Development and Mental Health (NIMH
R25MH120869), ABCD ReproNim Course - Reproducible Analyses of ABCD Data (NIDA
R25DA051675), and Modeling Developmental Change in ABCD Study (NIMH R25MH125545). At
my home university - the University of Southern California - I also want to acknowledge the Center for
Advanced Research Computing (CARC) for guiding me through utilizing the high performance
computing cluster. I would also like to thank the Divisions of Biostatistics and Environmental Health
within my home Department of Population and Public Health Sciences and my Neuroscience Graduate
Program for the seminars, talks, collaborations, support, and social events.
I have also had a lot of scientists along the way that helped guide and answer my questions, including
prior members of my committee and grant cosponsors, Dr. Farshid Sepehrband, Dr. Ryan Cabeen, Dr.
Sandrah P. Eckel, Dr. Rob McConnell, Dr. April Thames, Dr. Hosung Kim, Dr. Daniel Hackman, Dr.
Paul Thompson, Dr. Tiffany Ho, and Dr. J Michael Tyszka. In addition to these scientists, I would like
to say a huge thank you to my current committee who have worked with me throughout the years: Dr.
Lauren Salminen, Dr. Neda Jahanshad, and Dr. W. James Gauderman.
iii
I would like to acknowledge all the past and current members of my home lab, the Herting
Neuroimaging Lab. You all have been there with me through these past 5 years; you are all so open and
supportive, willing to collaborate, and I will miss you all so very much. To the postdocs for their
perspective, the older graduate students for commiserating, the new graduate students for their
enthusiasm, the staff for their expertise and support, and the undergraduates and volunteers for their
excitement, thank you all.
Next, I want to thank all my friends who were there for me during the tough times and celebrated with
me during the successes; you are all so supportive and truly wonderful people.
I want to thank my family who have been my biggest cheerleaders; thank you for all the brunches, phone
calls, vacations, dinners, puppy snuggles, and TV vegging sessions.
To my love, thank you for your continued support to be whoever I want to be and pursue my dreams.
You make me feel at peace, even when I have a grant due.
Now, to the real MVP, my mentor, Dr. Megan M. Herting. You are literally the best. I hit the jackpot
having you as a mentor. I worked for you for 2 years and then was your grad student for 5 more; after 7
years, I would whole-heartedly pick you as my mentor all over again. You have always been an
advocate, defender, and champion of mine. You taught me how to be an independent thinker,
conduct rigorous research, and opened doors for me, all with encouragement and compassion. I am so
grateful to be your first graduate student to defend. It is truly the end of an era; I grew so much not
just as a scientist but as a human. I will always look back at my time under your mentorship with so
much love and gratitude. Keep being your wonderful self.
iv
TABLE OF CONTENTS
ACKNOWLEDGMENTS...............................................................................................................................ii
LIST OF TABLES............................................................................................................................................vi
LIST OF FIGURES ........................................................................................................................................vii
ABSTRACT .....................................................................................................................................................viii
INTRODUCTION ...........................................................................................................................................1
CHAPTER 1: BEHAVIOR ASSOCIATIONS............................................................................................5
A. Introduction...............................................................................................................................................5
B. Methods......................................................................................................................................................7
B.1 Current Study Exclusionary Criteria................................................................................................7
B.1.1 General Exclusionary Criteria for the Complete Sample .....................................................7
B.1.2 Additional Chapter 1 Specific Exclusionary Criteria for Depression Prodromal Markers
Analyses..................................................................................................................................................7
B.2 Data Collection and Preparation......................................................................................................8
B.2.1 Behavioral Data...........................................................................................................................8
B.2.2 Depression Polygenic Risk Score (D-PRS) ............................................................................9
B.2.3 Incomes-to-Needs Ratio (INR)..............................................................................................10
B.2.4 Covariates and Confounders ..................................................................................................10
B.3 Data Analysis.....................................................................................................................................11
B.3.1 Depressed/Withdrawn Syndrome Scale Modeling Approach ..........................................12
B.3.2 Positive Affect Modeling Approach......................................................................................13
C. Results.......................................................................................................................................................14
C.1 Withdrawn/Depressed Symptom Scale........................................................................................14
C.1.1 Depression Polygenic Risk Score Effects.............................................................................18
C.1.2 Age Effects and Moderating Effect of Income-to-Needs Ratio.......................................19
C.2 Positive Affect ..................................................................................................................................20
C.2.1 Depression Polygenic Risk Score Effects.............................................................................24
C.2.2 Income-to-Needs Ratio Effects.............................................................................................24
D. Chapter 1 Discussion.............................................................................................................................25
E. Chapter 1 Summary................................................................................................................................29
CHAPTER 2: BRAIN ASSOCIATIONS....................................................................................................30
A. Introduction.............................................................................................................................................30
B. Methods....................................................................................................................................................34
B.1 Current Study Exclusionary Criteria..............................................................................................34
v
B.1.1 General Exclusionary Criteria for the Complete Sample ...................................................34
B.1.2 Additional Chapter 2 Specific Exclusionary Criteria for Neuroimaging Analyses.........34
B.2 Data Collection and Preparation....................................................................................................35
B.2.1 Neuroimaging Data ..................................................................................................................35
B.2.1.1 Resting-state Functional MRI (rs-fMRI) Connectivity....................................................35
B.2.1.2 Structural MRI (sMRI)..........................................................................................................35
B.2.3 Depression Polygenic Risk Score (D-PRS) ..........................................................................36
B.2.4 Income-to-Needs Ratio (INR) ...............................................................................................36
B.2.5 Covariates and Confounders ..................................................................................................36
B.3 Data Analysis.....................................................................................................................................37
B.3.1 Neuroimaging Modeling Approach.......................................................................................38
B.3.1.1 D-PRS and INR modeling in European-like and not European-like Group 1...........38
B.3.1.2. INR modeling in not European-like Group 2.................................................................40
C. Results.......................................................................................................................................................41
C.1 Resting-State Functional MRI........................................................................................................41
C.1.1 Income-to-Needs Ratios and Depression Polygenic Risk Scores Interactive Effects...43
C.2 Structural MRI Results....................................................................................................................45
C.2.1 Depression Polygenic Risk Scores Effects...........................................................................47
C.2.2 Income-to-Needs Ratio Effects.............................................................................................48
D. Chapter 2 Discussion.............................................................................................................................50
E. Chapter 2 Summary................................................................................................................................53
DISCUSSION...................................................................................................................................................54
CONCLUSIONS .............................................................................................................................................57
BIBLIOGRAPHY............................................................................................................................................58
APPENDIX......................................................................................................................................................78
vi
LIST OF TABLES
Table 1: Withdrawn/Depressed Symptoms Analyses Demographics for Baseline Data Collection ....................15
Table 2: Withdrawn/Depressed Symptoms Analyses Demographics for 1-Year Follow-Up Data Collection....16
Table 3: Withdrawn/Depressed Symptoms Analyses Demographics for 2-Year Follow-Up Data Collection....17
Table 4: Positive Affect Analyses Demographics for 6-Month Follow-Up Data Collection................................21
Table 5: Positive Affect Analyses Demographics for 1-Year Follow-Up Data Collection ....................................22
Table 6: Positive Affect Analyses Demographics for 18-Month Follow-Up Data Collection..............................23
Table 7: Cortical region names and their corresponding functional networks......................................................36
Table 8: Resting-State Functional MRI Analyses Demographics for Baseline Data Collection...........................42
Table 9: Resting-State Functional MRI Analyses Demographics for 2-Year Follow-Up Data Collection ..........43
Table 10: Structural MRI Analyses Demographics for Baseline Data Collection.................................................46
Table 11: Structural MRI Analyses Demographics for 2-Year Follow-Up Data Collection................................ 47
vii
LIST OF FIGURES
Figure 1: Collection timeline of data for emotional behavior analyses...................................................................11
Figure 2: Effect of D-PRS on Withdrawn/Depressed Symptoms..........................................................................19
Figure 3: Effect of Age and INR on Withdrawn/Depressed Symptoms................................................................20
Figure 4: Effect of D-PRS on Positive Affect ..........................................................................................................24
Figure 5: Effect of INR on Positive Affect ..............................................................................................................25
Figure 6: Location of brain regions associated with each of the large-scale brain networks ..................................31
Figure 7: Collection timeline of data for neuroimaging analyses............................................................................37
Figure 8: Trending interaction between INR and D-PRS for intra-network connectivity...................................44
Figure 9: Effect of D-PRS on Superior Frontal Gyrus in European-like Sample...................................................48
Figure 10: Effect of Age and INR on numerous sMRI metrics in the not European-like Group 2 Sample........49
Figure 11: Effect of INR on numerous sMRI metrics in the not European-like Group 2 Sample...................... 50
viii
ABSTRACT
Depression is one of the major contributors to the global burden of disease, with the World Health
Organization (WHO) ranking it as the number one non-fatal contributor. Most cases of depression
appear by an individual’s third decade of life, which is classified as early onset depression. The long-term
effects of early onset depression extend well into adulthood, usually leading to a high rate of recurrence
and significant health concerns. Research has shown that early intervention prior to disease onset leads
to the best outcomes. Therefore, detecting early markers of depression risk would help mitigate the
disease. Previous investigations have looked at the effect of environmental exposures or genetic
influences separately, with studies beginning to examine the interactive effects of genes and the
environment on risk for depression. Though, few studies have been done examining how gene-byenvironment interactions may map onto prodromal brain and behavioral biomarkers of risk for early
onset depression, which could greatly assist in early detection and treatment. Specifically, select brain
structure and functional networks as well as distinct emotional behaviors – such as, positive affect and
withdrawal symptoms – have been consistently associated with early onset depression. Ultimately, it
suggests that these may be important biomarkers in studying how gene-by-environment may contribute
to risk for depression that emerges prior to disease onset. Thus, this work will examine whether the wellknown environmental socioeconomic predictor of family income-to-needs may have independent
and/or interactive effects with an individual’s polygenic risk score for depression on the development of
emotional brain structure and function from pre- to early adolescence. To accomplish this goal, an
existing longitudinal dataset will be leveraged that examines approximately 8,000 subjects 9-10 year-ofage at baseline to 11-12 year-of-age at the 2-year follow-up from across the United States as part of the
larger Adolescent Brain Cognitive DevelopmentSM Study (ABCD Study®). Using up to three time
points for emotional behavior outcome data and up to two time points of data for the brain imaging,
we examine how genes, family income-to-needs, or their interactions affect changes in brain size and
function. Historically, genetic analyses have been conducted in European-like samples based on genetic
ancestry. Therefore, the first section of this research will test the independent and interactive effect of
an individual’s income-to-needs and polygenic risk for depression on prodromal emotional behaviors,
functional brain connectivity, and brain structure of key emotional regions previously associated with
depression in youth with European-like genetic ancestry. Next, to expand the field and work towards
more inclusive and generalizable findings, the same analyses will be conducted in youths who are not
European-like based on their genetic ancestry. Overall, we find that depression polygenic risk scores are
associated with brain and behavior within the European-like youths, but less so in youths who are not
European-like. We then see a potential moderating effect of family income-to-needs ratio on the
depression polygenic risk score within brain network connectivity, with further main effects of incometo-needs in youths from lower socioeconomic statuses. Ultimately, the findings from this project hold
the potential to identify potential brain-behavior biomarkers that may be important to consider in
establishing risk for early onset depression, in hopes of improving early detection and treatment.
1
INTRODUCTION
Half of all mental health disorders arise by age 141 with most cases going untreated. Major Depressive
Disorder (MDD) is one of the most burdensome disorders2
, inhibiting numerous individuals in living
fulfilling lives. Depression symptomatology and prevalence steadily increases from childhood through
adolescence3–5
, during a period of brain maturation that is highly dynamic6
. Moreover, most cases of
depression appear by an individual’s third decade of life, which is classified as early onset depression (age
of onset <30 years)7
. During this time, emotional and cognitive control brain regions are under
continued development, streamlining into efficient circuits8
. This leaves the brain vulnerable to
environmental factors9 that could be moderated by an individual’s genetic predisposition to
depression10. Therefore, identifying potential brain-behavior biomarkers associated with increased risk
of depression prior to onset could assist in depression mitigation and treatment.
When evaluating potential biomarkers of depression onset it is helpful to understand the varied factors
at play. The World Health Organization Commission of Social Determinants of Health found that
mental health is shaped partly by an individual’s environment, including socioeconomic factors11.
Systematic reviews of epidemiologic studies have shown pervasive negative associations between
common mental health disorders and socioeconomic status (SES) during childhood11. SES has been
linked to prodromal markers of depression - the period when initial symptoms emerge prior to disorder
onset -such as, increased depressive symptoms and decreased positive affect (i.e., positive emotions, such
as joy) during childhood12. Specifically, SES measured through household income has been shown to
positively associate with depressive symptomatology in children as measured through a child interview13;
for positive affect, a previous review examining self-esteem - which is highly associated with positive
affect14 - showed increased levels with higher SES measured through numerous methods, including
income in children of all ages15. Moreover, during childhood, the brain is undergoing extensive
maturation during this time16,17, making it vulnerable to outside factors9
. Given that the pediatric brain
is undergoing this dynamic change, the environmental influence of SES may affect depression risk by
altering neurodevelopmental trajectories. Support for this formulation includes studies where
household SES consistently influences emotional brain development. For instance, reductions in the
intra-network connectivity of specific networks has been linked to depression mainly in adults, with
findings showing negative or null effects in youths
18–21. In particular, in a cross-sectional analysis of 9-11
year-olds in the ABCD Study, lower household income was related to decreases in intra-network
connectivity of brain networks associated with depressionin childhood22; this work is further supported
by other smaller studies23,24. Beyond functional connectivity, household SES has also been linked both
cross-sectionally and longitudinally to structural brain alterations in children and adolescents. In these
studies, lower household SES has been associated with decreases in brain surface area25,26, as well as
decreases in amygdala27 and hippocampal volumes28, and cortical and subcortical volumes29. Taken
together, these initial findings point to SES as a key environmental factor to focus on when investigating
brain biomarkers related to emotional neurodevelopment and risk for early onset depression. However,
the majority of this previous work has utilized household income24,26,27 to represent household SES,
2
whereas fewer studies have utilized the preferred method30,31 of calculating a household income-to-needs
ratio (INR)28,32. Unlike total household income, INR takes into account the household size in order to
better represent the range of a family’s economic well-being and can examine a continuum of
experience31. While three initial cross-sectional studies have found lower INR to relate to brain
structure25 and resting-state33,34 at ages 9-10 years using large diverse samples, such as the nationwide
Adolescent Brain Cognitive Development (ABCD) Study dataset of over 11,800 children across the
U.S., questions remain as to how household INR effects longitudinal patterns of emotional
development and brain maturation in key emotional circuitry and if this effect varies alongside genetic
risk factors for depression.
While environmental factors, including SES, may impact neurodevelopment and risk for depression,
these effects may ultimately interact alongside an individual’s genetic make-up. Earlier studies have
examined monogenic (i.e., single gene) influences on emotional brain development35. Importantly,
depression is a complex, heterogeneous disease, with patients having varying symptoms, risk factors, and
treatments that are not consistent across individuals36. Therefore, given the heterogeneity of depression,
it is likely affected by many genes, with more recent work suggesting the importance of examining
polygenic (i.e., numerous genes) risk and its potential role in neurodevelopment37. Specifically, using
large genome-wide association studies (GWAS), depression polygenic risk scores (D-PRS) can be
calculated based on case-control studies including clinical diagnosis of patients in order to build weights
for each gene that shows an association with MDD diagnosis38. These scores can then be applied to a
new set of subjects to assess their risk for MDD based on their genetic code38. There have only been a
few longitudinal studies that have investigated the impact of D-PRS on depressed/withdrawn symptom
severity in youth39,40, with no studies - to our knowledge - examining the effect of D-PRS on positive
affect during childhood. Moreover, an individual’s neurodevelopment is, in part, regulated by their
genetics41. Previous work has found that structural brain alterations associated with D-PRS colocalize
to the prefrontal cortex of the brain in an adult depression case-control study42 and an adult non-clinical
sample42,43, but no work to our knowledge has investigated D-PRS effects on pediatric brain
development of the cortex and subcortical regions, let alone longitudinally. Moreover, there has been
minimal work examining D-PRS on functional brain activity, with only one study reporting a reduction
in brain activation to neutral faces among non-clinical adults with higher D-PRS44. To our knowledge,
no study has examined how D-PRS is associated with resting-state functional connectivity of the
salience, frontoparietal, and default mode networks, all of which undergo drastic development during
the transition to adolescence45. While initial studies suggest D-PRS may affect emotional neural circuitry
and ultimately risk for MDD, additional longitudinal pediatric research is needed to further investigate
if D-PRS influences emotional brain development and transdiagnostic prodromal behaviors linked to
depression. Moreover, while using D-PRS to examine genetic risk for depression offers a novel
opportunity, genetic studies are currently extremely biased to individuals of European-like descent.
Almost 80% of the genome-wide association studies (GWAS) used to generate polygenic risk scores are
calculated in individuals of European-like ancestry46. Overall, genetic diversity between humans is very
low given the relatively recent advent of the human species47. Though, with time, there has been a
3
gradual divergence, known as genetic drift; as populations mixed and moved around the world, a small
gradient in alleles emerged over time48. This has led to varying linkage disequilibriums (LDs)49 and
frequencies of causal loci50,51. Therefore, when generating polygenic risk scores in populations that are
European-like, these small variations most likely will lead to inaccurate predictions in diverse
populations. Specifically, previous research indicates that polygenic risk scores (PRS) generated from
samples of a specific ancestry often have reduced accuracy in diverse populations, and this accuracy can
vary depending on the trait for which the risk score is calculated52. Given all of this, more research is
needed that examines whether previously identified D-PRS are generalizable to individuals of diverse
ancestries utilizing diverse sample populations.
As stated above, the literature strongly suggests household SES and an individual’s D-PRS may each
impact the development of emotional neurocircuitry. However, little work has been done examining
how these two factors may interact with one another to impact longitudinal changes in depression
symptomatology, positive affect, and brain development. One study has examined both D-PRS and
household environment on the risk of early onset depression in a study sample from the United
Kingdom: they examined caregiver cruelty and bullying as their environmental variable instead of a
household socioeconomic status and did not examine an interaction between environment and
genetics39; in short, this longitudinal study of children from ages 10-24 years found that both depression
polygenic risk score (D-PRS) and environmental variables had positive associations with early-onset
depression39. Similarly, only a few studies have examined environmental interactions - as measured by
stress53, income54, or childhood trauma55 -with genes on neurodevelopment; for all of these studies, only
individual genes were examined, with modifying effects seen in functional connectivity and structural
associations in the brain53–55. Thus, to our knowledge, no study has examined how household SES may
interact with D-PRS to impact longitudinal changes in emotional neurodevelopment during the
transition to early adolescence. Specifically, a previous study has investigated how childhood household
SES and genetics interact to affect risk of early onset depression56, displaying the greatest depressive risk
in individuals with both low SES and high D-PRS; though this study did not examine if structural and
functional brain maturation were also affected.
Therefore, the current research examined - in a well-powered, diverse pediatric sample - the independent
and interactive effects of household SES and an individual’s D-PRS on longitudinal changes during the
transition to early adolescence in prodromal behavior markers of depression and emotional brain
networks using resting-state functional connectivity and structural MRI. To do so, the large,
longitudinal, and demographically diverse sample of participants from the Adolescent Brain Cognitive
Development℠ Study (ABCD Study®) was leveraged. This is important as many of the studies
mentioned offer insight cross-sectionally, but little to no work has been done examining the interplay
between D-PRS and household SES on longitudinal brain and behavior development. The ABCD
Study provides the required scale, scope, and individual variability necessary to address these important
gene-by-environment questions. Furthermore, given that the previous D-PRS studies were conducted
in European-like individuals we chose to first examine the D-PRS effect in the ABCD Study in
4
individuals who are genetically European-like; then to test the portability of these findings in more
diverse, non-European-like sample populations, we examined D-PRS in not European-like individuals.
Overall, this work is highly innovative and critically important, as examining an interactive effect of SES
and D-PRS in diverse populations is likely to offer a clearer image of an individual’s susceptibility.
Ultimately, studying these key emotional brain-behavior biomarkers for risk of MDD may help improve
detection, prevention, and intervention programs for early onset depression.
5
CHAPTER 1: BEHAVIOR ASSOCIATIONS
A. Introduction
Depression during adolescence is becoming an ever increasing concern. Particularly in the United States,
depression diagnosis has been heightened, with current rates around 16% of the adolescent population57.
Specifically, depression that onsets during adolescence is classified as early onset, whereas late onset
depression is defined as occurring after age 307
. Research has found that improved detection prior to the
onset of depression can help mitigate long term negative consequences that arise after depression onset58.
Conducting research to help pinpoint potential behavioral markers of risk for depression in youth can
help aid in this goal59. One of these methods is to identify prodromal markers of depression - the period
when initial symptoms emerge prior to disorder onset60 - while in tandem examining how these
prodromal markers of depression may be influenced by genetic and environmental factors61.
Previous studies of youth affected by depression provide important insight into the prodromal
behaviors that should be targeted in identifying early risk factors of depression prior to clinical
diagnosis12,62. Two transdiagnostic prodromal markers of Major Depressive Disorder (MDD) are
decreased positive affect (i.e., less joy)12 and increased depressed/withdrawn symptom severity62. In the
U.S., MDD is diagnosed clinically when an individual endorses, for at least 2 weeks, 5+ symptoms that
represent a significant change in functioning according to the Diagnostic and Statistical Manual of
Mental Disorders63. This can make it difficult, especially for early onset depression, to identify
individuals at risk before diagnostic criteria are met. By examining key prodromal markers of depression,
it places a focus on the current symptoms and does not solely rely on the presence of a depression
diagnosis, adopting a dimensional approach64. This is especially important while studying children, since
most diagnoses are not occurring during adolescence, but key interventions could be helpful during this
critical growth period to help minimize lasting disease effects58.
While there is decent evidence examining depression behavioral prodromes, there is less understanding
about how these behaviors can be impacted by genetic and environmental factors, such as family-level
socioeconomic status (SES) factors, during adolescence. As previously mentioned, lower household
income has shown to increase the risk for depression in youths13,15. Specifically, in the United States, a
cross-sectional study of 457 11-13 year-olds examining the effects of income in a relatively diverse cohort
found that lower income was associated with higher depressive symptoms13 through extracting
symptoms prevalence utilizing a diagnostic interview - the Computer-based Diagnostic Interview
Schedule for Children (C-DISC)65. In a review examining socioeconomic status effects on self-esteem -
which highly correlates with positive affect14 - a small but significant positive effect was found;
differences were seen across different populations of individuals, with steady increases in self-esteem
with increased SES seen in childhood15. Additionally, in a cross-sectional study on 8,568 9-year-old
children conducted in Ireland, household socioeconomic dynamics was positively associated with
children’s happiness66. Though, no studies - to our knowledge - examine the effects of SES specifically
on positive affect in youths.
6
In addition to these environmental influences, genetic factors can also play a crucial role in the risk of
depression onset. Much research has shown that early onset depression is more influenced by an
individual’s genetic make-up as compared to late onset depression67. Therefore, assessing genetic risk can
be a very helpful tool in unraveling strategies for personalized care pathways for youth at risk for early
onset depression. As previously stated, given the heterogeneity of depression, it is best to utilize a
polygenic (i.e., numerous genes) risk score68. Moreover, as most polygenic risk scores (PRS) are generated
in large adult samples, carefully utilizing one that is validated in an adolescent sample is ideal, especially
given the differences in depression with age at onset7
. As previously mentioned, a few longitudinal
studies investigated the effect of D-PRS on depression in children. In these studies, the depression
polygenic risk score (D-PRS) generated by Wray et al. (2018)69 was utilized. Specifically, in Europeanlike youths, this previous work has seen positive associations of this D-PRS on early-onset-depression39,
depression diagnosis40, and depressive symptoms -as measured through the Child Behavior Checklist40.
Therefore, future research testing whether this D-PRS can associate with other prodromal markers of
depression would be useful for early disease detection. Moreover, this previous research was conducted
in individuals of European-like ancestry. Given the bias towards individuals with European ancestry,
continued research is needed on more diverse populations to test the generalizability of these findings.
In addition to examining the main effects of SES and polygenic risk for depression in adolescents, it is
important to consider a potential interactive effect of these risk factors. In one novel case-control study
of 35,680 individuals, they examined the interaction between polygenic risk scores (PRS) and
socioeconomic status (SES) -as measured by maternal education; they found that both measures can be
helpful in conjunction to examine risk for early onset depression56. Though, this study was conducted
in Denmark with little generalizability to the diverse populations in the United States. Therefore,
examining a polygenic-by-socioeconomic interaction in a diverse population can assist in the
generalizability of the findings.
The aim of this first chapter of research was designed to fill these gaps in the literature by examining the
independent and interactive effects of household socioeconomic status, as measured by income-toneeds ratios, and an individual’s depression polygenic risk score (i.e., D-PRS) on longitudinal changes
during the transition to early adolescence in prodromal behavior markers of depression using data from
the nationwide, diverse ABCD Study. Specifically, the current research aimed to investigate the
independent and joint effects of household INR and D-PRS on longitudinal changes in caregiverreported depressive symptoms on the Child Behavior Checklist (CBCL)70 and the youth-reported
National Institute of Health (NIH) Toolbox Emotion Battery – Positive Affect Scale71 collected across
three time points. This current study also utilized a specific polygenic risk score69 that was validated in a
European-like youth cohort of depressed versus not depressed controls and when examining depressive
symptoms40. Moreover, as previously alluded to, given the D-PRS that has only been validated in
European-like adolescents, this research was first assessed in youths with European-like ancestry (<80%
genetic European-like ancestry). Then, we further examined whether the findings found in European-
7
like individuals were transferable to individuals who are not European-like based on genetic ancestry
(<80% genetic European-like ancestry) (see Chapter 1, Section B.1.2 & B.2.2, for details). We also
examine whether there is a potential moderating effect of age; there is usually significant behavior
development at the youths current age with increases in generating coping strategies for one’s own
emotions72. In terms of main effects, I hypothesized that lower income-to-needs ratios (INR) and
increased depression polygenic risk (D-PRS) would both lead to increased depressed symptoms and
decreased positive affect, and these effects will increase with age as emotional development progresses.
If an interaction is observed, I further hypothesized that increased D-PRS will lead to larger negative
effects of INR, increasing depressed symptoms and decreasing positive affect; though, the opposite
effect will occur with low D-PRS and high INR. Again, for this interaction, this effect would increase
with age. Additionally, I hypothesized that D-PRS values would be less predictive in the not Europeanlike youths, but SES effects would be consistent across samples.
B. Methods
This research utilized genetic, behavioral and demographic data from the ABCD Study - a longitudinal
single cohort study73. The study collected harmonized data from 21 U.S. sites, allowing for a socially,
economically, and racially diverse cohort of 11,873 children (48% female, baseline race/ethnicity
characteristics: 50% non-Hispanic white, 22% Hispanic, 15% non-Hispanic Black, 2% Asian, and 11%
mixed/other)73. The ABCD Study is intended to represent a population-based, non-clinical sample
(exclusion criteria included severe sensory, intellectual, medical, and neurological disorders; sample
recruitment, inclusion criteria, and protocol have been published in detail74,75).
B.1 Current Study Exclusionary Criteria
B.1.1 General Exclusionary Criteria for the Complete Sample
From the larger ABCD cohort, additional exclusionary criteria were implemented to improve scientific
rigor. Initially, subjects who were assessed within an initial site 22 were removed, since it was not
followed-up with during the ABCD Study. Next, given the mental health focus of this work, data
collected after March 1st, 2020 were removed to avoid confounding effects of stress on mental health
outcomes arising from the COVID-19 pandemic76–78. Next subjects who had poor genetic data (see
section B.2.2) and did not have socioeconomic values as represented by income-to-needs (see section
B.2.3) were excluded. Lastly, subjects whose caregiver incorrectly entered their sex-at-birth were also
removed. See Supplemental Figure 1 in the Appendix for a full flowchart by visit.
B.1.2 Additional Chapter 1 Specific Exclusionary Criteria for Depression Prodromal Markers Analyses
In addition to the above General Exclusionary Criteria, participants were removed from the current set
of analyses if they did not have genetic data, and at least one measure of sociodemographic and
behavioral data; this is possible because we chose a modeling approach that can handle missing
datapoints (see Chapter 1, Section B.3). Next, since polygenic risk scores were generated and there is
family relatedness among some subjects (i.e., twins, siblings), one subject per family was randomly
8
chosen to satisfy model assumptions regarding independent observations. Finally, given that almost all
genome-wide association studies (GWAS), including in this research, utilizes a discovery sample of
subjects who are of European-like ancestry79, our sample was separated based on ancestry-like markers
to test the usefulness of the depression polygenic risk score. Therefore, first the analyses was conducted
in subjects with ≥80% genetic European-like ancestry80 based on results from the 1000 Genome
reference panel phase 3 release81 with follow-up analyses conducted in individuals who are not
European-like (<80% genetic European-like ancestry) to test the generalizability of findings.
Furthermore, the not European-like subjects were further broken down into two groups given
collinearity issues between the depression polygenic risk score and the genetic PCs used to correct for
population stratification; this led to a not European-like Group 1 and Group 2 which was split based on
the first population stratification principal component (more details in section B.2.2 Depression
Polygenic Risk Score (D-PRS)).
Each of these three groups were then separated into two samples to analyze depressed/withdrawn
symptom scale and positive affect separately. For the withdrawn/depressed symptoms analyses, these
groups included: European-like (N=5,264), not European-like Group 1 (N=1,710), and not Europeanlike Group 2 (N=1,468); for positive affect analyses, these groups included: European-like (N=5,163),
not European-like Group 1 (N=1,685), and not European-like Group 2 (N=1,434). Detailed tree of
exclusionary criteria and subsequent groups can be found in Supplemental Figure 2 in the Appendix.
B.2 Data Collection and Preparation
B.2.1 Behavioral Data
Two behavioral questionnaires were utilized for the proposed research study: the Child Behavior
Checklist (CBCL)70 and the National Institute of Health (NIH) Toolbox Emotion Battery – Positive
Affect71. The CBCL is a 113-item questionnaire completed by the parent or primary caregiver at each
annual visit about the child’s behavior over the last six months. Each question (e.g., “Show little interest
in things around him/her”) is rated on a 3-point scale: not true (0 points), somewhat or sometimes true
(1 point), very often or always true (2 points). Syndrome scales are then calculated on a continuous scale.
The current study will focus on the withdrawn/depressed subscale. This subscale is calculated based on
the answers to 8 questions, meaning the total raw score can range from 0-16 points, with higher scores
indicating increased withdrawn/depressed symptom severity. The NIH Toolbox Emotion Battery –
Positive Affect questionnaire is completed by the child participant at both the 1 year annual in person
visit and at the 0.5 and 1.5 mid-year phone visit, where they are asked to answer nine questions about
how they felt in the past week (e.g., “I felt at ease”, “I felt delighted”) on a 3-point scale: not true (1
point), somewhat true (2 points), very true (3 points). This means the total raw score can range from 9-
27 points, with higher summary scores relating to higher levels of positive affect.
9
B.2.2 Depression Polygenic Risk Score (D-PRS)
At the baseline visit, saliva samples were collected to assess each individual’s genetic data by the ABCD
Study82. DNA is isolated from saliva and genotyped using the Affymetric NIDA Smokescreen™
Genotyping array that has over 600,000 single nucleotide polymorphisms (SNPs)83. DNA quality
control was then completed by Rutgers RUCDR on all calling signals and variant call rates. The
genotype dataset was then imputed using the TOPMed imputation server84. These data are then made
available in PLINK format by the consortium for further follow-up analysis. For the current study, these
SNP data were used to create depression polygenic risk score (D-PRS) values for each subject who has
available genetic data. Specifically, the D-PRS summary statistics are a single continuous value for each
individual based on the weight provided by the 2018 Psychiatric Genomics Consortium GWAS
summary statistics from Wray et al. (2018)69, with larger scores reflecting greater polygenic risk. This DPRS approach was chosen as it has been shown to replicate in a pediatric cohort40. This is especially
important as D-PRS are generated in an adult independent sample, yet the proposed research focuses
on early onset depression. Therefore, careful consideration was taken given the possibility that late onset
and early onset depression may have different genetic predispositions7,85. Therefore, using PRS weights
that replicate in a pediatric sample ensures the relevance of these genes in the risk for early onset
depression.
Specifically for calculating the D-PRS values, the discovery sample69 had 9.6 million SNPs with 135,458
subjects with depression and 344,901 controls. This sample was made up of seven cohorts from
numerous countries (Australia, Denmark, Germany, Iceland, Italy, Sweden, Switzerland, The
Netherlands, The United Kingdom, and The United States). For details on how the PRS weights were
generated please refer to the original research article69. To calculate the D-PRS with similar parameters
to the pediatric cohort replication study40, using the ABCD Study post-imputation files, bfctools86 was
used to filter the files with a call rate <98% and minor allele frequency (MAF) < 0.05 were excluded; a
Hardy-Weinberg Equilibrium (HWE) p < 0.0001 was not excluded given that this is done for
homogenous sample, and the ABCD Study is a non-homogenous group of subjects. Utilizing plink287
(www.cog-genomics.org/plink/2.0/)and R, samples with extreme heterozygosity - which could indicate
poor quality sample (high heterozygosity) or inbreeding (low heterozygosity)88 - were removed89. Next,
to generate the D-PRS values, we utilized the software program PRSice-290 with SNPs having an r2 >
0.25 with the index SNP being removed and SNPs within 200k of the index SNP considered for
clumping.
With this data, principal components (PCs) of the sample's ancestry were generated91. These PCs were
used to account for population stratification that is inherent within genome-wide association studies
(GWAS). To generate the PCs, plink2 was used to generate the first 50 PCs with a pruning window size
of 200 variants, a 50 variant step size, and filtering out any SNPs with a linkage disequilibrium (LD) r2
> 0.25. This was conducted within the European-like sample and not European-like sample separately
with the greatest number of subjects available for each group. Given the collinearity present between
the D-PRS and the PCs within the not European-like group, the first PC was used to split the sample
10
into two groups. This led to the creation of three samples that were used for analyses: European-like
sample, not European-like Group 1 sample, and not European-like Group 2 sample. Generated D-PRS
values were mean scaled to each of the groups given the differences between D-PRS values due to
population stratification (i.e., European-like, not European-like Group 1, and not European-like Group
2). Utilizing PCs generated within each group, a scree plot demonstrated that 10 PCs were sufficient to
capture the necessary variance inherent in each sample’s ancestry. Therefore, for each model, 10 PCs
were added as covariates to account for population stratification.
B.2.3 Incomes-to-Needs Ratio (INR)
Family size and household income are assessed every year starting at the baseline visit within the ABCD
Study. For family size, the caregiver reported the number of individuals living at the household as an
integer value; the few individuals who responded with 0 or 1 for household size were converted to NA
given the invalid nature of these answers (i.e., one caregiver plus child would indicate a minimum
household of 2). They also reported the household’s combined total income based on several binned
categories: 1 = Less than $5,000; 2 = $5,000 through $11,999; 3 = $12,000 through $15,999; 4 = $16,000
through $24,999; 5 = $25,000 through $34,999; 6 = $35,000 through $49,999; 7 = $50,000 through
$74,999; 8 = $75,000 through $99,999; 9 = $100,000 through $199,999; 10 = $200,000 and greater.
Utilizing this data, an INR was created as previously published using the ABCD Study25, which includes
converting binned categories into a continuous variable with mean adjustment and dividing by the
federal poverty level as reported by the U.S. Department of Health and Human Services92 for the
household size of a corresponding year. Moreover, moderate to good reliability was noted over the three
timepoints in the European-like sample (ICC=0.74), the not European-like Group 1 (ICC=0.79), and
not European-like Group 2 (ICC=0.81)93. Therefore, the mean of the household INR averaged over the
three time points was utilized. Overall, higher scores reflect greater SES advantage, with scores less than
1 indicating household living below poverty92.
B.2.4 Covariates and Confounders
For this research, the initial 3 years of data available were utilized, which covers the ages of 9 to 12 years,
with up to 3 waves of sociodemographic information (~1 year apart) and behavioral data (0.5 to 1 year
apart), and genetic data per participant. Given similarity in values over the three waves of data collection,
the sociodemographic variable of interest, incomes-to-needs ratio (INR), was averaged over three waves
of data collected (see section B.2.3 Incomes-to-Needs Ratio (INR)) (Figure 1). For the analyses, time
varying covariates were adjusted for in each model included: study-site and age, along with the timeinvariant covariates of sex-at-birth, caregiver-identified race/ethnicity, highest parents education, and
the first 10 genetic PCs. Previous work has shown that sex-at-birth can relate to differences in depressive
symptoms94. Moreover, in this work, race is defined as a socially constructed concept that has been used
to classify individuals into groups that have been politically driven to establish social hierarchies, while
ethnicity is another social construct that is used to categorize diverse populations which can inform
personal and group identities95. This combined variable of race and ethnicity is utilized in this work to
include individuals who personally identify as Hispanic or Latinx when asked about their race95. For
11
highest parents education96 and race/ethnicity97, these are two factors that can also highly confound the
effect of household income on depressive symptoms. When asked to the caregiver, the race/ethnicity
variable in the ABCD Study is defined as: non-Hispanic white, non-Hispanic Black, non-Hispanic
Asian (Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, other Asian), Hispanic, other
(American Indian, Native American, Alaska Native, Native Hawaiian, Guamanian, Samoan, other
Pacific Islander, or other race)95. When correcting for caregiver-identified race/ethnicity for each group
- European-like, not European-like Group 1, and not European-like Group 2 - the factors varied slightly;
this is because some of the caregiver-identified race/ethnicity categories had a small number of
individuals and needed to be collapsed into the ‘other’ category to prevent model convergence issues.
Therefore, for the European-like group, Black and Asian individuals were grouped into ‘other’; for not
European-like Group 1, white, Black, and Asian were grouped into ‘other’; for European-like Group 2,
white and Hispanic were grouped into ‘other’ with no individuals caregiver-identified as Asian. In short,
when representing race/ethnicity in each model, the European-like sample had 3 categories of
race/ethnicity: white, Hispanic, and other; the not European-like Group 1 sample had 2 categories of
race/ethnicity: Hispanic and other; the not European-like Group 2 sample had 2 categories of
race/ethnicity: Black and other.
Figure 1: Collection timeline of data for emotional behavior analyses
Abbreviations: CBCL = Child Behavior Checklist; NIH-TB = NIH Toolbox; PRS = Polygenic Risk Score.
B.3 Data Analysis
Initial descriptive and exploratory analysis were conducted to check for potential errors and outliers,
and assess variable distributions and correlations. Throughout analysis, models were evaluated to ensure
modeling assumptions were met and/or the results were not influenced by outliers. Prior to modeling,
D-PRS values were mean-scaled (centered around the mean with 1-standard deviation increments) for
12
ease of model interpretation in each group separately - European-like, not European-like Group 1, and
not European-like Group 2 - given differences in ancestry between groups. For INR, derived ratio values
were utilized without transformation to maintain the interpretability and generalizability across sample
groups. A multilevel modeling approach was implemented to take advantage of the large, longitudinal
dataset. The multilevel modeling approach taken was well equipped to handle missing datapoints;
therefore, all available data were utilized. For each model a two-level approach was used which allowed
for us to estimate the intercept (i.e., initial value at ages 9-10) and slope (i.e., change with age), if
necessary, for how an individual’s behavior was associated with D-PRS, INR, and/or their interaction.
B.3.1 Depressed/Withdrawn Syndrome Scale Modeling Approach
The depression symptom score was right skewed, therefore a standard linear model was not appropriate
given its propensity for artificial inflation of coefficient significance98–100. Consequently, this measure
was treated as a count metric and a negative binomial model was implemented, which incorporates an
extra parameter that accounts for overdispersion101. Below the negative binomial models are detailed,
which calculates the outcome at baseline and its rate of change with age between subsequent study visits.
Let �!" represent the outcome for of withdrawn/depressed symptom scale for participant � and visit �.
Let �!" denote age and �!" a set of time-varying adjustment variables (study site and age). The quantities
of interest in the Level 1 model are the subject-specific intercepts (�!, representing mean of the outcome
at age �∗) and the slopes (�!, representing the longitudinal change in � per year of age). Level 2 consists
of separate regression models for the subject-specific intercepts (Level 2a) and slopes (Level 2b). The
regressors of primary interest in these models are �!, the D-PRS for each individual and, �!, represents
the average INR across three study visits, and their interaction (�!�!). These models also included
adjustments for time-invariant covariates, �! (i.e., sex-at-birth, maximum highest-level of parents
education across all time points, caregiver-identified race/ethnicity, and the first 10 genetic PCs) and a
random effect of subject nested within site:
Level 1 (visit): ���(�!") = �! + �!(�!" − �∗) + �$�!"
Level 2a (intercepts): �! = � + �$�! + �%�! + �&�!�! + �%�! + �!
Level 2b (slopes): �! = � + �$�! + �%�! + �&�!�! + �&�! + �!
In the models, the Level 2 expressions for �! and �! are substituted into the Level 1 model and the
resulting single mixed-effect linear model is used to estimate the parameters of interest. The baseline
levels (�’s) and rate of change for the participants (�’s) were assessed in terms of the interactive influence
of D-PRS and INR on the depressed/withdrawn symptom scale. The primary parameters of interests
for this mainly focuses on the slope coefficients, �$, �% and �&, for how the individual changes over time
for the independent effect of D-PRS, INR, and their interaction, respectively. The secondary
parameters are the intercepts of these same terms – �$, �%, and �& – examining an individual’s starting
point. Utilizing �∗, it is possible to set this value to 10 and 12 (mean age rounded to the nearest half year
13
for baseline and 2-year follow-up) to examine the baseline effect of our independent and interactive
effect of where individuals begin and how that effect changes over time, respectively.
Given the complexity of this model, higher order terms were dropped in a stepwise manner in favor of
a parsimonious model if they were not significant (p>0.05). This allows us to see whether there are any
main effects present. First, if the three-way interaction of INR-by-PRS-by-Age was insignificant it was
removed; next, if not significant, 2-way age interactions were removed: INR-by-Age and/or PRS-byAge. Finally, the INR-by-PRS interaction was removed if not significant to test the main effects of INR
and PRS.
Following this approach, for all groups, the following interaction terms were dropped given their lack
of significance: PRS-by-Age, PRS-by-INR, and INR-by-PRS-by-Age. For INR-by-Age, it was
significant in both the not European-like groups; therefore, this interaction term was kept in the not
European-like groups, but dropped from the European-like model. Therefore, the final model used to
examine withdrawn/depressed symptoms in the European-like sample was:
Withdrawn/Depressed Symptoms ~ INR + D-PRS + Covariates
Here, as previously mentioned, the covariates are age, sex-at-birth, highest caregivers education,
caregiver-identified race/ethnicity, ABCD Study site, first 10 genetics PCs with subjects nested by site.
The final model used to examine withdrawn/depressed symptoms in both the not European-like
samples - Group 1 and Group 2 - was:
Withdrawn/Depressed Symptoms ~ Age + INR + INR*Age + D-PRS + Covariates
The covariates for these models are also sex-at-birth, highest caregivers education, caregiver-identified
race/ethnicity, ABCD Study site, and first 10 genetics PCs with subjects nested within site.
B.3.2 Positive Affect Modeling Approach
For the models examining positive affect, this outcome was not heavily skewed, therefore we employed
a linear mixed effects model. All the variables from above (see Section B.3.1) are represented in the same
manner in the following model, but the outcome is positive affect:
Level 1 (visit): �!" = �! + �!(�!" − �∗) + �$�!" + �!"
Level 2a (intercepts): �! = � + �$�! + �%�! + �&�!�! + �%�! + �!
Level 2b (slopes): �! = � + �$�! + �%�! + �&�!�! + �&�! + �!
14
Again, higher order terms were removed if they did not reach significance (see Section B.3.1). For all the
positive affect models, for all groups, no significant interactions with age (INR-by-Age, PRS-by-Age,
and INR-by-PRS-by-Age) and no interaction between INR and PRS were found, so again these terms
were dropped. Therefore, the final model used to examine positive affect in all the samples assessed -
European-like, not European-like Group 1, and not European-like Group 2 - was:
Positive Affect ~ INR + D-PRS + Covariates
For these models, as previously mentioned, the covariates are also age, sex-at-birth, highest caregivers
education, caregiver-identified race/ethnicity, ABCD Study site, and the first 10 genetics PCs with
subjects nested within site.
C. Results
For the groups used to assess the prodromal markers of depression, the European-like sample was heavily
skewed to include adolescents whose caregiver-identified them as non-Hispanic white (~88%); for the
not European-like Group 1, adolescents were mostly caregiver-identified as Hispanic (~70%), while
adolescents included in the not European-like Group 2 were mostly caregiver-identified as non-Hispanic
Black (~85%) (Tables 1-6).
C.1 Withdrawn/Depressed Symptom Scale
For distribution of variables in each ancestry group by collection date for the withdrawn/depressed
analyses see Tables 1-3 below; for graphical distributions of age, INR, D-PRS, and
withdrawn/depressed symptoms in each sample, see Supplemental Figure 3 within the Appendix. For
INR, it was found to be well distributed and consistent over time, with 16%, 21%, and 34% in poverty
(i.e, <1 INR) for samples at baseline in the European-like, not European-like Group 1, and not
European-like Group 2, respectively. After mean-scaling the D-PRS variable, there was no mean
difference between the groups, but the range varied, with the European-like sample having the greatest
spectrum. For the withdrawn/depressed symptom outcome, there was no significant difference
between the groups at baseline; there was a significant difference at 1- and 2-year follow-up visits, but
this mean difference only varied by ~0.3 points (Tables 1-3). Overall, no interactions between
depression polygenic risk score and income-to-needs ratios for either of the depression prodrome
markers were seen.
15
Table 1: Withdrawn/Depressed Symptoms Analyses Demographics for Baseline Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation; N-Miss = number of missing data.
16
Table 2: Withdrawn/Depressed Symptoms Analyses Demographics for 1-Year Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation; N-Miss = number of missing data.
17
Table 3: Withdrawn/Depressed Symptoms Analyses Demographics for 2-Year Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation; N-Miss = number of missing data.
18
C.1.1 Depression Polygenic Risk Score Effects
For the European-like sample, a main effect of D-PRS (Incidence Risk Ratio (IRR)=1.184, p=1.1x10-
8
) was found (Figure 2A; Supplemental Table 1). A non-linear relationship was seen between D-PRS
and withdrawn/depressed symptoms with higher D-PRS values leading to greater symptom counts. On
average, an individual with a mean D-PRS in our European-like sample, has a score of 0.74 out of a
possible 16 points, meaning they have little-to-no withdrawn/depressed symptoms. If an individual has
a lower D-PRS (2 standard deviations (SD) below the mean), this led to a 0.21 (29%) decrease in their
symptom score to 0.53, meaning less depressed/withdrawn symptoms. A higher D-PRS (2 SD above
the mean), would lead to a 0.30 (40%) increase as compared to mean D-PRS, to a score of 1.03. Though,
if an individual had a D-PRS towards the upper bounds of the range (6 SD above the mean), they would
have a 1.29 (175%) increase compared to the mean, having an estimated score of 2.03. To put these
results into perspective, if a 10-year-old female has a mean D-PRS value - based on our model -the CBCL
score would range from 0-1 points (50-52 t-score); with a score of 0-1, these symptoms are not clinically
relevant102. If a 10-year-old female has a D-PRS 6 standard deviations above the mean, the CBCL score
would be ~2 points (56 t-score), which would indicate that the symptoms are still not clinically relevant.
Next, for the not European-like Group 1 sample, there is a small main effect of D-PRS (IRR=1.223,
p=0.028) (Figure 2A; Supplemental Table 1). While this same effect was seen in the European-like
sample, this effect is deprecated in European-like Group 1. For the mean D-PRS value, an adolescent in
the not European-like Group 1, would have an estimated withdrawn/depressed symptom score of 0.71.
As the D-PRS decreased (2 SD below the mean), the predicted score decreased by 0.23 (33%) to 0.47,
while an increase (2 SD above the mean) led to an increase of 0.35 points (49%) to an estimated score of
1.06, which is not clinically relevant. Given the smaller range of D-PRS observed in this group, the larger
effects seen at high D-PRS in the European-like sample were not observed (Figure 2B). For the not
European-like Group 2, no significant effects of D-PRS were found for the withdrawn/depressed
symptom score (Figure 2A; Supplemental Table 1). Given that no significant D-PRS by age
interaction terms were seen for any of the groups, it suggests this genetic effect is consistent over time
between the ages of 9-12 years-old (i.e., from baseline to follow-up sessions).
19
Figure 2: Effect of D-PRS on Withdrawn/Depressed Symptoms
A) Effect of the depression polygenic risk score (D-PRS) on the withdrawn/depressed symptoms in each group; B)
Distribution of D-PRS by analysis group. Note, CBCL withdrawn/depression symptom scale can range from 0–16 points
(i.e., 8 questions). Abbreviations: pD-PRS = p-value for effect of D-PRS; n.s. = not significant.
C.1.2 Age Effects and Moderating Effect of Income-to-Needs Ratio
For the European-like sample, there was a significant age effect, but no effect of INR on
withdrawn/depressed symptoms (Figure 3A). In contrast, a significant interaction between INR and
age was found for the European-like Group 1 sample (IRR=1.02, p=0.008) and the European-like
Group 2 sample (IRR=1.021, p=0.041) (Figure 3A). In the not European-like Group 1, the age effect
was modified by INR, with larger age effects seen with increased INR. This effect was also seen in the
not European-like Group 2, but the effects were miniscule. As previously mentioned, it is important to
note that the INR was significantly different between ancestry groups. For the European-like sample,
the mean INR was 4.6, while it was 3.4 and 2.4 for the not European-like Group 1 and Group 2 samples,
respectively (Figure 3B). Since there is on average a relatively higher INR mean in the European-like
sample, this high INR could be obfuscating a potential INR-by-age interaction seen at lower INR values
that is demonstrated in the not European-like samples.
20
Figure 3: Effect of Age and INR on Withdrawn/Depressed Symptoms
A) Shows the effect of age in the European-like sample, plus the effect of age modified by income-to-needs ratio (INR) in
the not European-like Group 1 and Group 2 samples on the CBCL withdrawn/depressed symptoms, displaying age effects
at the 25th, 50th, and 75th percentiles within each sample, with all other variables held constant; B) violin plots showing the
distribution of income-to-needs (INR) across each group with width relative to the sample size. Note, CBCL
withdrawn/depression symptom scale can range from 0–16 points. Abbreviations: pAge = p-value for effect of age; pINR-by-Age
= p-value for effect of INR-by-age interaction.
C.2 Positive Affect
For distribution of variables in each ancestry group by collection date for the positive affect analyses see
Tables 4-6 below; for distributions of age, INR, D-PRS, and positive affect in each sample, see
Supplemental Figure 4 within the Appendix. While the positive affect has slightly fewer subjects than
the withdrawn/depressed sample, the subjects within the positive affect sample highly overlap with the
withdrawn/depressed sample (98-99%) across all three samples. Again, INR was well distributed and
consistent over time, with 16%, 21%, and 32% of subjects in poverty (i.e, <1 INR) at the 6-month followup session in the European-like, not European-like Group 1, and not European-like Group 2,
respectively. Similarly to the withdrawn/depressed analyses, after mean-scaling D-PRS, there was no
mean difference between the groups, but the range was larger in the European-like sample. For the
positive affect outcome, there was no significant difference at the 1-year follow-up in person visit, but a
significant difference between the groups at the 6-month and 18-month follow-up phone call visits.
Though the mean difference in the 6-month and 18-month phone visit total score between the samples
was only ~0.5 points out of a range of 9-27 points (Tables 4 & 6).
21
Table 4: Positive Affect Analyses Demographics for 6-Month Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR, income-to-needs; D-PRS, depression polygenic risk score; HS, high-school; SD
= standard deviation; N-Miss = number of missing data.
22
Table 5: Positive Affect Analyses Demographics for 1-Year Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR, income-to-needs; D-PRS, depression polygenic risk score; HS, high-school; SD
= standard deviation; N-Miss = number of missing data.
23
Table 6: Positive Affect Analyses Demographics for 18-Month Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR, income-to-needs; D-PRS, depression polygenic risk score; HS, high-school; SD
= standard deviation; N-Miss = number of missing data.
24
C.2.1 Depression Polygenic Risk Score Effects
In the European-like sample, there was a main effect of D-PRS (�=-0.181, p=0.002,), with higher DPRS leading to lower positive affect scores (Figure 4; Supplemental Table 2). Specifically, a 1 standard
deviation increase in a D-PRS decreased self-report of positive affect score by 0.181; albeit, this is
minimal given the range of the potential points (9-27 points). In the not European Group 1 (�=-0.121,
p=0.516) and Group 2 (�=-0.182, p=0.245) samples, a similar trend was seen, albeit was not significant,
potentially due to the minimized variance in D-PRS values generated for youths in these groups. Given
that no significant D-PRS-by-age interaction terms were seen for any of the groups, it suggests this
genetic effect is consistent over time between the ages of 9-12 years-old (i.e., from 6-month follow-up to
18-month follow-up sessions).
Figure 4: Effect of D-PRS on Positive Affect
A) Effect of the depression polygenic risk score (D-PRS) on the positive affect questionnaire in each group; B) Distribution
of D-PRS by analysis group. Note, Positive Affect scale can range from 9–27 points. Abbreviations: pD-PRS = p-value for
effect of D-PRS; n.s. = not significant.
C.2.2 Income-to-Needs Ratio Effects
An INR effect was only observed in adolescents within the not European-like Group 1 (�=0.1,
p=0.001,), with higher INR leading to higher positive affect. For a 1-unit increase in INR there was a
0.1 point increase in self-reported positive affect (Figure 5; Supplemental Table 2), which is again
25
minimal given the range of the potential points (9-27 points). INR was not significant in the Europeanlike (�=0.017, p=0.238) and not European-like Group 2 (�=0, p=0.997) samples (Figure 5;
Supplemental Table 2).
Figure 5: Effect of INR on Positive Affect
A) Effect of the income-to-needs ratio (INR) on the positive affect questionnaire in each group; B) Distribution of INR by
analysis group. Note, Positive Affect scale can range from 9–27 points. Abbreviations: pINR = p-value for effect of INR; n.s.
= not significant.
D. Chapter 1 Discussion
After investigating the effect of depression genetic risk and household socioeconomic status on
depression/withdrawn symptoms and positive affect behaviors during early adolescent development,
the findings indicate that against our hypothesis, there were no interactive effects between genetic risk
and income-to-needs ratios. We did, however, find both similarities and differences in the main effects
of polygenic risk for depression as well as age and incomes-to-needs among adolescents in the Europeanlike sample (≥80% European genetic ancestry; caregiver identified largely as non-Hispanic white (88%)),
not European-like Group 1 sample (<80% European genetic ancestry; caregiver identified largely as
Hispanic (70%) or other (23%)), and not European-like Group 2 sample (<80% European genetic
ancestry; caregiver identified largely as non-Hispanic Black (83%)). Main effects of polygenic risk for
depression on CBCL withdrawn/depressed scores were observed in both adolescents in the European-
26
like and not European-like Group 1 samples. This mimicked findings found in European-like
individuals, showing a positive relationship between D-PRS and depressive symptoms40. In addition,
an age effect was observed for withdrawn/depressed symptoms in the European-like group, whereas age
effects were modified by INR in both of the not European like groups, albeit with miniscule effects in
the latter. For the second depressive prodromal questionnaire, only adolescents in the European-like
group showed the expected association between higher genetic risk and decreases in self-reported
positive affect, while a positive main effect of income-to-needs ratios (INR) was observed for adolescents
from the non European-like 1 group with higher positive affect associating with increased INR. Overall,
the results indicate that this depression polygenic risk score in European-like individuals can associate
with increased depressed/withdrawn symptoms and decreased positive affect prior to adolescence.
Though, the size of this effect does not translate well across different ancestry groups. Moreover, we
find that the effects of INR are only seen in samples with on average lower INR values.
Since the sample used to generate the depression polygenic risk score (D-PRS) was European-like69, as
well as the analyses to validate in adolescents40, the focus of the first portion of this research was to
examine income-to-needs ratios and polygenic risk for depression in a European-like sample. What our
work found was that in the European-like sample, there is a non-linear effect of D-PRS of increased
withdrawn/depressed symptomatology regardless of participant’s age. This is inline with previous work
showing how other non-psychiatric PRS predictions tend to be age-independent earlier in life103;
specifically, in another study examining prostate cancer, the PRS values were less predictive later in
life104. Yet, it is important to note that while higher D-PRS significantly related to higher
withdrawn/depressed symptoms, it may be unlikely that individuals with very high D-PRS show
clinically relevant withdrawn/depressed symptoms as noted in our results. This is unsurprising since in
the European-like sample, at baseline, only 7% of subjects have a clinically relevant withdrawn/depressed
symptoms score (≥ 65)
102. Though, these findings are helpful, in that since the D-PRS does relate to
subclinical withdrawn/depressed symptoms it may be a good metric when examining depressive risk
prior to clinical diagnosis. Furthermore, these effects mimic previous findings examining the Child
Behavior Checklist depressive symptoms in European-like youths40. Moreover, in this same group of
adolescents, there is a further effect of D-PRS on positive affect, regardless of age, with a small but
significant negative relationship, with an approximately 2-point change between a participant with the
lowest D-PRS to the highest D-PRS; for the NIH-Toolbox Positive Affect Survey, there were a total of
9-items, with each item answered on a 3-point scale (1 = Not True; 2 = Somewhat True; 3 = Very True),
leading to a possible score that ranges from 9-27 points. While the change in positive affect across the
whole range of D-PRS is low (~2 points), we still demonstrate that D-PRS may be representing nuances
in depression onset with European-like individuals, which could later assist with depression diagnosis.
Though, this effect did not significantly translate to individuals who were not European-like,
demonstrating the necessity of incorporating SNPs from diverse populations in the D-PRS generation.
With this evidence, it appears that this D-PRS, which was initially generated in adults with depression69
and then later replicated in youths40, is generalizable to early adolescents with numerous prodromal
27
markers of depression. The rates of depression in the ABCD Study at these timepoints is incredibly low
(Supplemental Table 7). This is expected given prior research showing the progression and onset of
depression, with the peak of early onset depression diagnosis occurring between the ages of 20 and 30
years105. In fact, in the European-like sample used here to assess the effects of polygenic risk and incometo-needs ratios on depressive prodromes, only 0.13% participants would meet a current major depressive
disorder (MDD) diagnosis at ages 9-10 years (i.e., baseline) and 0.31% at ages 11-12 years (i.e., 2-year
follow-up) based on caregiver-report on the computerized Kiddie Schedule for Affective Disorders and
Schizophrenia for School-Aged Children (K-SADS). If a broader scope was utilized to assess present or
past MDD or Unspecified Depression, only 5.93% and 7.66% of European-like participants would meet
KSAD criteria for diagnosis at ages 9-10 years and 11-12 years, respectively (Supplemental Table 7).
Though studying earlier depression onset may help relieve the most burden of disease since an earlier
age-of-onset is linked to increased disease burden (i.e., symptom severity, suicide attempts, etc.)106.
Therefore, focusing on a dimensional characterization of depressive symptoms - such as
withdrawn/depressed symptoms and positive affect - allows for a more nuanced assessment of
depression related behaviors; this can aid in identifying early risk factors that may be relevant to the
progression of depression rather than traditional diagnosis.
While it is valuable to see that this D-PRS successfully associates with two depressive prodromes, this DPRS score seems to not fully generalize to adolescents that are not of European-like ancestry. In terms
of the effect of the D-PRS in the not European-like samples, there was a stunted association between DPRS and withdrawn/depressed symptoms in Group 1 and no effect in Group 2, with no significant
findings in either group when examining positive affect. Overall, when testing the transferability of
European generated PRS, they can perform poorly in diverse populations52, which is what is seen in
these findings. Linkage disequilibriums (LDs) can vary across populations which can lead to differences
in the selection of the best single nucleotide polymorphisms, along with alterations in allele frequencies
and the magnitude of the effect size107. For our work, the only significant effect of D-PRS in the
individuals who were not European-like, was in Group 1 on the withdrawn/depressed symptoms scale.
When examining the demographic make-up of Group 1, a majority of the participants are caregiveridentified as Hispanic. Previous work examining non-psychiatric PRS (i.e., height, white blood cell
counts, waist-to-hip ratio, etc.) has seen that European generated polygenic risk scores applied to
individuals who identify as Hispanic show an effect but there is an overall decreased effect size107.
Furthermore, our D-PRS did not generalize to adolescents in the not European-like Group 2 sample,
which had many adolescents whose caregivers identified them as non-Hispanic Black. It is important to
note, that both non-European like groups were based on showing less than 80% similarity to European
ancestry, and further categorized based on the first genetic ancestry PC, grouping them individuals by
shared genetic ancestry. As such, the demographics of each of these groups is based on race/ethnicity,
which is a socially constructed concept historically used to classify individuals into hierarchical groups.
Thus, self-identification as non-Hispanic Black does not equate to African ancestry; albeit many
individuals who self-identify as Black do tend to have higher rates of African ancestry108. Previous work
has also shown that European generated PRS values perform very poorly in individuals of African
28
ancestry because of the same issues discussed above due to the LD variability52. To improve D-PRS
values and their effectiveness in relating to depressed prodrome behaviors, new weights will need to be
created that utilize SNPs identified in ancestry-matched subjects to increase the D-PRS effectiveness in
diverse populations109.
When it comes to socioeconomic status as assessed by income-to-needs ratios (INR), there were only
significant effects seen in not European-like groups. For the withdrawn/depressed symptoms, there was
a significant age effect in the European-like sample, but this age effect was modified by INR in the
individuals within both the not European-like samples. At higher INR values in the not European-like
samples, there is a positive slope with age relating to increasing withdrawn/depressed symptoms as
youths get older. Thus, the age effect at higher INR values in not European-like samples shows a similar
trajectory to the age effect in European-like individuals, who - on average - had higher INR values. At
lower INR, there was no change in symptoms with age in adolescents within the not European-like
groups, however, the average withdrawn/depressed symptoms were higher as compared to those
reporting higher INRs. It is important to note, however, that the distributions and mean INR values
are not equal across these three samples of adolescents, with youths within the European-like group
having higher INR values on average than either of the other two not European-like groups. This is
reflective of systematic inequalities that have been pervasive throughout history in the United States,
leading to huge wealth gaps between races and ethnicities110. A lack of an interaction between age and
INR in the European-like group may stem from the differences in INR distributions. Specifically, the
interactive effect between ageand INR in withdrawn/depressed symptoms may be prevalent only when
factoring in lower socioeconomic statuses, which are less common in the European-like sample of the
ABCD Study.
Unfortunately, our findings suggest that when studying gene and environmental effects in Europeanlike groups, research may be missing potential important effects in minoritized individuals - in this case,
individuals who have lower SES - perpetuating biases. Furthermore, when examining the effect of INR
on positive affect, only a main effect is seen in the not European-like Group 1 sample, with higher
positive affect associating with increased INRs. Previously no work has assessed the relationship
between INR and positive affect during preadolescence, though, the trend is in the expected direction
based on previous work displaying positive associations between increased SES and positive health
outcomes111–113. This is potentially only seen in not European-like Group 1 since the INR values are the
most evenly distributed across the entire range of the variable, while not European-like Group 2 is
heavily skewed towards lower INR and the European-like sample is skewed towards higher INR. Age
by INR effects seen only in groups of adolescents with larger numbers of individuals reporting lower
INR could indicate that higher depressed symptoms at lower SES may be due to multiple factors, like
increased levels of psychosocial stress. Lower SES families may have less resources available to them,
which could in turn increase stress levels, increasing depressed symptoms; one potential mechanism of
action is changes in family environment (i.e., family structure, parental support, etc.) that could increase
depressive symptoms through SES changes13. This could cause individuals from low SES families to have
29
heightened levels of depressed symptoms at younger ages that persist over time as compared to youths
from high SES families which may display the more expected pattern of lower levels of symptoms that
increase over time with age. Overall, building upon the known increases in depressive symptoms with
age during this developmental period114, our work suggests that adolescents from low and high SES
backgrounds may eventually be on par with one another. Though, this increase in depressive symptoms
at earlier ages for individuals from lower SES families could lead to downstream effects - such as
increased comorbidities, suicidal ideations, and increased depression severity106 - that have lasting
consequences.
E. Chapter 1 Summary
Chapter 1 demonstrates the validity of using a depression polygenic risk score (D-PRS) in predicting
prodromal markers of depression in a European-like sample, with this slightly translating to some
individuals who were not European-like, while not being predictive at all for others. We also see that
socioeconomic status - as measured by income-to-needs ratios (INR) - has an interactive effect with age
in the individuals who were not European-like; this is likely only seen in these groups because of systemic
racism that has led to pervasive inequities that make income a moderating factor on age that is not
present in the age associations within samples with higher average INR. Therefore, we do see D-PRS
and INR both influencing depression prodromes, though they do not interact with one another in this
research. In Chapter 2 (below), this research further examines whether brain structure and function is
also affected by these genetic and environmental factors as potential biomarkers related to depression115.
30
CHAPTER 2: BRAIN ASSOCIATIONS
A. Introduction
The brain shows unique patterns of functional and structural maturation across childhood and
adolescence, with important implications for emotional processing and regulation116. During the preteenage years, the brain begins a period of rapid maturation, with key resting-state functional (rs-fMRI)
brain networks refining their connections117,118. These cortical networks include the salience (SN)119–121,
default mode (DMN)119,121, and frontoparietal network (FPN)121 (Figure 6). The SN responds to salient
(i.e., important) stimuli121, while the DMN – a network mainly activated when the brain is at rest –
focuses on cognitive activity related to the self121. The FPN is essential for high-level cognitive
functioning and behavioral regulation122. The SN is comprised of the anterior cingulate and ventral
anterior insular cortices123, while the DMN is made up of the medial prefrontal cortex, posterior
cingulate cortex, and inferior parietal lobes124; the FPN hubs are located within the lateral prefrontal and
posterior parietal cortices125 (Figure 6). With age, intra-network (i.e., within network) connectivity of
the SN126, DMN127, and FPN128,129 increases, whereas inter-network (i.e., between network) connectivity
begins to decrease, becoming less integrated, further specializing their function130,131. These increases in
functional intra-network and decreases in inter-network connectivity are paralleled by structural
changes beginning in the pre-teen years. Cortical thickness132–134 and surface area133,135 decrease with age
during childhood and adolescence. These decreases are thought to relate to the refinement and
streamlining of connections through synaptic pruning136. Beyond these cortical regions, the amygdala
and hippocampus are two subcortical brain regions that heavily connect with the SN137,138, DMN139,
and FPN140,141, which are critical for emotional development and function142, and show an increase in
connectivity with age from childhood to early adulthood143,144. Moreover, gray matter volumetric
increases are seen in the development of the amygdala and hippocampus, that continue from childhood
to early adolescence17,145. These age-related changes are important as it is the development of these
cortical networks and subcortical regions that are thought to be important for the integration of one’s
emotional responsivity to stimuli and higher order cognitive control working together to regulate both
arousal of emotional stimuli146,147 and negative affective signaling148. Hence, individual differences in the
maturation of these various cortical and subcortical regions may influence emotional
neurodevelopment and subsequent risk for emotional problems. Therefore, studying factors that may
influence longitudinal patterns of functional and structural brain network development may be helpful
in identifying early biomarkers of altered neurodevelopment that may ultimately contribute to risk for
various mental health conditions, such as early onset depression.
31
Figure 6: Location of brain regions associated with each of the large-scale brain networks
Representation of the large-scale brain networks in this research and their corresponding cortical regions; A) frontoparietal
network (FPN), B) default mode network (DMN), C) salience network (SN).
Given that intervention is best done at the earliest sign149, biomarkers of depression risk are invaluable
tools in mitigating depression onset. As such, a number of studies have aimed to identify key brain
biomarkers that may relate to depressive symptoms in otherwise healthy individuals, and/or are present
in those individuals affected by depression (i.e., diagnosed). In a study of adolescents120, and another
with young adults150, it was found that depression symptomatology had an inverse relationship with the
intra-network connectivity of the SN, with increased SN connectivity associated with lower positive
affect. A similar pattern is seen for the intra-network connectivity of the DMN in a cross-sectional
32
cohort of 57 adolescents, with individuals suffering from depression having elevated connectivity as
compared to controls without depression151. Alternatively, in another cross-sectional study, FPN intraconnectivity was found to be decreased in a sample of young/middle-aged individuals with increased
depressive symptomology152. Moreover, while investigating inter-network connectivity, depression has
also been associated with a reduction between networks, though this was in a small cross-sectional casecontrol study of depression in adults153. Beyond functional connectivity, previous research has shown
that depression is also associated with reductions in related structural MRI metrics, including cortical
thickness in a cross-sectional adolescent study154 and surface area in a longitudinal study in adults155.
Furthermore, the amygdala and hippocampus are two subcortical brain regions consistently implicated
as structurally and functionally altered in children and young adults diagnosed with depression156–164.
Specifically, decreases in functional connectivity between the SN and FPN with the amygdala and
hippocampus have been seen in depressed young adults159–161. In addition, depression is also associated
with increased connectivity between the DMN and the amygdala in adolescents162, and increased
connectivity between both the amygdala and hippocampus in adults163. For structural associations, the
hippocampus156,158 and amygdala164 displayed attenuated growth in children affected with early onset
depression. In terms of symptomatology, the hippocampus was shown to associate in children and
adolescents, with smaller volumes relating to increased withdrawal symptoms specifically estimated by
the Child Behavior Checklist165. Ultimately, only two studies examined longitudinal MRI scans157,158,
with most studies - except for one with 179 subjects165 - examining less than 100 subjects. Additionally,
in all the above studies, the cohorts either did not discuss the racial and ethnic make-up of the groups or
were mostly white. Therefore, future work studying large longitudinal changes in these features of brain
function, structure, and behavior that are linked to depression within diverse youth cohorts may hold
great promise for generalizable, timely detection and treatment of early onset depression.
When it comes to potential influential environmental predictors that affect brain structure and function
related to depression, lower household SES again emerges. Lower SES has been related to decreases in
intra-network connectivity of the SN24 in childhood. For the relationship between socioeconomic status
and intra-network connectivity of the DMN, two cross-sectional studies in the ABCD Study found
differing results23,33; the first study utilizing family income as their SES variable found decreases in DMN
intra-network connectivity with lower family income22, though the other study utilizing income-toneeds ratio found no relationship between any of our networks of interest33. This second study, though,
only examined the baseline time point and did not correct for race/ethnicity in their model33. Moreover,
functional connectivity between the FPN and the amygdala – key in emotion regulation166 – is found
to be reduced in children from lower household SES backgrounds32. Importantly, this reduction in
connectivity between the FPN and amygdala is also a known aforementioned hallmark biomarker of
depression160. Beyond functional connectivity, household SES has also been linked both crosssectionally and longitudinally to structural brain alterations; lower household SES has been associated
with decreases in brain surface area25,26, as well as decreases in amygdala27 and hippocampal volumes28,
and overall age associations across cortical and subcortical volumes in children and adolescents29.
33
Specifically in a large, diverse cohort - the ABCD Study - lower household SES, as measured by incometo-needs ratios, was also cross-sectionally associated with decreases in brain surface area25.
In addition to the effects of household socioeconomic status, there is also previous work to indicate that
numerous genes (i.e., polygenic risk) can also impact brain structure and function associated with
depression. Some research has been done examining the effect of depression polygenic risk scores in
adults with differences seen in the right orbitofrontal cortex in individuals who were not diagnosed with
depression43. Another study found that higher polygenic risk for depression was related to reductions in
brain activity to social stimuli44, smaller amygdala167, and less cortical gyrification167 in adults. Changes
in cortical thickness within the prefrontal regions have also been associated with depression polygenic
risk scores as evidenced by Cattarinussi et al. review paper (2022)168. However, all these studies were
conducted in adults. There has been very minimal research examining the effect of D-PRS on childhood
brain development, especially as it relates in a diverse cohort. One cross-sectional study in the ABCD
Study found a null effect of D-PRS on hippocampus volumes169. In addition to this study, there was
also a cross-sectional study conducted in China that found a D-PRS effect in the middle prefrontal
cortex that interacted with urbanicity170, demonstrating the need for research to consider examining DPRS-by-environment interactions as it pertains to brain development and risk for depression during
childhood and adolescence.
Ultimately, there is much to be gained by examining how environmental and polygenic risk can
independently and interactively affect the developing brain in a youth longitudinal sample. Gaining a
deeper understanding of the influence of these factors at a younger age may help to identify potential
biomarkers that could be used for early prevention and treatment. Such research may also lead to
increased specificity of treatment and more impactful interventions. Thus, the aim of this second
chapter of research was designed to fill these gaps in the literature by examining the independent and
interactive effects of household SES -as measured by income-to-needs ratios (INR) -and an individual’s
depression polygenic risk score (D-PRS) on longitudinal changes in functional brain connectivity and
gray matter morphology, again leveraging data from the nationwide ABCD Study. Specifically, the
current research aimed to investigate the independent and joint effects of household INR and D-PRS
on longitudinal changes in resting-state functional MRI (rs-fMRI) connectivity of the frontoparietal
network (FPN), default mode network (DMN), and salience network (SN), as well as their
corresponding cortical regions structural MRI metrics (i.e., cortical thickness and surface area) along
with hippocampal and amygdala volumes and connectivity collected at two time points, when the
participants were 9-10 and 11-12 years-old (i.e., baseline and 2-year follow-up). Given the D-PRS results
presented in Chapter 1, we chose to investigate the interactions of INR and D-PRS in only the groups
of adolescents in which D-PRS was predictive of depression behavior based on the Child Behavior
Checklist (CBCL). Therefore, since the D-PRS score in Chapter 1 did not associate with the
withdrawn/depressed symptoms in the not European-like Group 2 sample, it is likely that the D-PRS
may not be reflecting a predictive D-PRS value in this sample. Consequently, the association between
D-PRS and brain function and structure was not assessed in the not European-like Group 2 sample.
34
Accordingly, D-PRS results are reported only for children with European-like ancestry and the not
European-like Group 1. However, for completeness, we also examined and reported the age and INR
effects for the not European-like Group 2. For this particularwork, I hypothesized that high D-PRS and
a low INR will have a multiplicative effect on brain connectivity resulting in altered intra- and interconnectivity of the FPN, SN, and DMN and amygdala and hippocampus-based connectivity with age;
for the structural MRI analyses, I further hypothesized that high D-PRS and low INR will have a
multiplicative effect on structural development with age, as seen by greater reductions in cortical
thickness and surface area, and subcortical volumes that are normally seen to decrease during this period
of development133.
B. Methods
The longitudinal ABCD Study was again utilized73, including the genetic and demographic data
described in Chapter 1, as well as up to two timepoints of neuroimaging data. For more information
about the ABCD Study reference Chapter 1, section B.
B.1 Current Study Exclusionary Criteria
B.1.1 General Exclusionary Criteria for the Complete Sample
For the neuroimaging analyses, two waves of data were utilized, baseline and year-2 follow-up; this again
included children who were 9-12 years of age. The exclusionary criteria for this complete sample can be
found in Chapter 1, section B.1.1, along with Supplemental Figure 1.
B.1.2 Additional Chapter 2 Specific Exclusionary Criteria for Neuroimaging Analyses
Further exclusionary criteria, in addition to the above General Exclusionary Criteria, were also applied
to the resting-state functional magnetic resonance imaging (rs-fMRI) and the structural magnetic
resonance imaging (sMRI) data (Supplemental Figures 5 & 6). First, participants with no handedness
score and poor rs-fMRI and sMRI data were excluded; for both time points, this required no abnormal
findings or normal anatomical variant of no clinical significance on the MRI image (mrif_score=2 or 3),
plus a passing score (i.e., no motion artifacts) for the T1-weighted image based on ABCD Study rating
values (imgincl_t1w_include=1)171. For the rs-fMRI sample, a passing T2-weighted image
(imgincl_t2w_include=1) was also needed given the inclusion of the T2-weighted image in the rs-fMRI
functional calculations171. Next, participants were removed from the current set of analyses if they did
not have genetic data, and at least one measure of sociodemographic and either one measure of rs-fMRI
or sMRI, depending on the sample; this is possible because we chose a modeling approach that can
handle missing data (see Chapter 2, Section B.3). Again, since we are utilizing polygenic risk scores,
family relatedness is a factor, so we chose one subject per family to satisfy modeling assumptions of
having independent observations. Moreover, following the same methods in Chapter 1 (section B.1.2),
we again separated the sample into three samples based on ancestry-like markers: European-like sample,
not European-like Group 1, and not European-like Group 2. For more details on how these were
created, reference Chapter 1, Section B.1.2. For a detailed tree of exclusionary criteria and the
35
subsequent groups utilized for the rs-fMRI and sMRI analyses, see Supplemental Figure 5 & 6 in the
Appendix, respectively.
B.2 Data Collection and Preparation
B.2.1 Neuroimaging Data
An anatomical T1-weighted and functional T2*-weighted MRI scans were collected on 3-Tesla scanners
(Siemens Prisma, General Electric (GE) 750, and Philips) from each participant at baseline and 2-year
follow-up (Figure 7). Importantly, multi-site harmonization is implemented by the ABCD Study in
terms of image collection protocol, preprocessing, and quality control assessment171.
B.2.1.1 Resting-state Functional MRI (rs-fMRI) Connectivity
Functional connectivity was measured using resting-state functional magnetic resonance imaging (rsfMRI) data. These data were collected while the participant passively viewed a cross hair image on a
screen for up to 20 minutes, split up into four 5-minute sessions172. Each rs-fMRI scan was acquired at
2.4mm isotropic (TR=800ms) with a T2*-weighted image with a multiband EPI with slice acceleration;
for each scan a fieldmap was also collected for B0 distortion correction172. Functional correlations for
each region of interest (ROI) were then calculated based on the Fisher transformed z-statistic averaged
within and between the networks173. For the current proposal, a priori networks of interest include the
frontoparietal (FPN), default mode (DMN), and salience (SN) networks as defined by the Gordon
parcellations atlas174 (Table 7). Participants have a functional connectivity (i.e., correlational) value for
the baseline and two-year follow-up for both intra- and inter-connectivity of these 3 cortical networks
as well as connectivity between each cortical network (i.e., FPN, DMN, and SN) and our two a priori
subcortical ROIs bilaterally (i.e., the hippocampus, and amygdala), totaling 18 outcomes.
B.2.1.2 Structural MRI (sMRI)
Structural MRI (sMRI) data was measured using the T1-weighted image collected at 1-mm isotropic172
via an inverted RF-spoiled gradient echo scan sequence utilizing prospective motion correction175.
Cortical and subcortical segmentation were completed using the semi-automated FreeSurfer v. 7.1.1
brain segmentation protocol176–185. Again, all images underwent manual quality check post acquisition
and post segmentation with FreeSurfer by the ABCD Study. For the cortical regions, they are segmented
into regions based on the Destrieux atlas182,186; for this research we focused on cortical thickness and
surface area of the structural regions that correspond to the functional networks previously mentioned
(i.e., FPN, DMN, and SN), equaling 22 regions in two hemispheres with both a surface area and
thickness measurement; totaling 88 measurements (Table 7). The volume of the subcortical regions the
amygdala and hippocampus were also examined in the left and right hemispheres, totaling 4 volumetric
measurements.
36
Table 7: Cortical region names and their corresponding functional networks
This table represents the cortical regions that correspond to the frontoparietal network (FPN), default mode network
(DMN), and the salience network (SN); the corresponding Destrieux atlas short name is also listed. Abbreviations: G =
Gyrus; S = Sulcus.
B.2.3 Depression Polygenic Risk Score (D-PRS)
Identical methods were utilized to calculate depression polygenic risk score (D-PRS) as in Chapter 1;
reference Chapter 1, section B.2.2 for details.
B.2.4 Income-to-Needs Ratio (INR)
For the calculation of the income-to-needs ratio (INR) the exact procedure was followed as in Chapter
1; reference Chapter 1, section B.2.3 for details. Since the three timepoints across samples again showed
moderate to good reliability in both the rs-fMRI and sMRI European-like samples (ICC 0.74), not
European-like Group 1 samples (ICC 0.79), and not European-like Group 2 samples (ICC 0.81), we
took the average of the INR values reported across the three timepoints (baseline, 1-year follow-up, and
2-year follow-up). Again, higher scores reflected greater SES advantage.
B.2.5 Covariates and Confounders
The time varying covariates that were considered included age and MRI scanner serial number, along
with the time-invariant covariates of sex-at-birth, highest parent education, caregiver-identified
race/ethnicity, handedness, and first 10 genetic PCs. For a definition of the race/ethnicity variable,
please refer back to Chapter 1, section B.2.4. Handedness is added since it can have effects on the size of
brain regions187 given the bilateral nature of the brain. For the neuroimaging analyses, ABCD Study site
37
is replaced with MRI scanner serial number; these variables are highly correlated and the MRI scanner
serial number offers a more refined grouping for the MRI analyses. For the resting-state MRI analyses,
MRI motion estimates were also added to account for differences in rs-fMRI data that could be linked
to motion during scanning188. For the structural MRI analyses, intracranial volume, as estimated by
Freesurfer, were also included as a covariate for the structural MRI analyses of the subcortical regions of
the amygdala and hippocampus to account for overall brain size189.
Figure 7: Collection timeline of data for neuroimaging analyses
Abbreviations: PRS = Polygenic Risk Score; MRI = Magnetic Resonance Imaging.
B.3 Data Analysis
Similarly to Chapter 1, initial descriptive and exploratory analysis were conducted to check for potential
errors and outliers, and assess variable distributions and correlations. Models were evaluated to ensure
assumptions were met and our results were not influenced by outliers. Prior to modeling, D-PRS values
were again mean-scaled (centered around the mean with 1-standard deviation increments) for ease of
model interpretation in each group separately - European-like and not European-like Group 1 - given
differences in ancestry between groups. For INR, derived ratio values were utilized without
transformation to maintain the interpretability and generalizability across sample groups. Linear
multilevel mixed effect models were then implemented to examine the depression polygenic risk score
(D-PRS) by income-to-needs interaction (INR) as a function of age on: 1) resting-state functional MRI
(rs-fMRI) data of aforementioned relevant networks and subcortical regions and 2) related structural
MRI (sMRI) regions. Our multilevel modeling approach takes advantage of our large, longitudinal
dataset and is well equipped to handle missing data, therefore, all outcome data was utilized. For each
model, we again utilized a two-level approach to estimate the intercept (i.e., initial value at ages 9-10)
38
and slope (i.e., change with age), if necessary, for how an individual’s brain outcome is associated with
D-PRS, INR, and their potential interaction.
B.3.1 Neuroimaging Modeling Approach
All rs-fMRI and sMRI outcomes were normally distributed; therefore, we employed the linear mixed
effects model to calculate the baseline effect and its potential rate of change with age; for more details
refer to Chapter 1, section B.3.2. Again, given D-PRS was not found to be valid in predicting depression
behavior in the not European-like Group 2, we modeled D-PRS and INR in only the European-like and
not European-like Group 1, and investigated INR effects only in the not European-like Group 2. For
the modeling of the cortical regions, preliminary analyses examining each hemisphere separately found
similar trajectories across hemispheres. Therefore, both hemisphere measurements per cortical region
were utilized but they were nested within each subject to account for the similarity within an individual's
brain to utilize all available data. We describe each of these approaches below in greater detail.
B.3.1.1 D-PRS and INR modeling in European-like and not European-like Group 1
Longitudinal linear mixed effects models were implemented identical to Chapter 1, section B.3.2.
Briefly, �!" denotes age and �!" the time-varying adjustment variables. In the Level 1 model, the
quantities of interest are the subject-specific intercepts �!, representing mean of the outcome at age �∗)
and the slopes (�!, representing the longitudinal change in � per year of age). Level 2 consists of separate
regression models for the subject-specific intercepts (Level 2a) and slopes (Level 2b). The regressors of
primary interest in these models were �!, the D-PRS for each individual and, �!, which represents the
average INR across three study visits, and their interaction (�!�!). These models also included
adjustments for time-invariant covariates as described previously, �!:
Level 1 (visit): �!" = �! + �!(�!" − �∗) + �$�!" + �!"
Level 2a (intercepts): �! = � + �$�! + �%�! + �&�!�! + �%�! + �!
Level 2b (slopes): �! = � + �$�! + �%�! + �&�!�! + �&�! + �!
Level 2 �! and �! expressions are substituted into the Level 1 model resulting in a single mixed-effect
linear model to estimate the parameters of interest. The primary parameters of interests are the slope
coefficients, �$, �% and �&, for how the individual changes over time for the independent effect of DPRS, INR, and their interaction, respectively. The secondary parameters of interest are the intercepts
of these same terms – �$, �%, and �& – examining an individual’s starting point. If an age interaction is
detected, �∗ can be set to 10 and 12 to examine the baseline effect at age 10 and how they change over
time to age 12. All of the models were corrected for multiple comparisons with a False Discovery Rate
of �<0.05. Again, higher order terms were removed in a stepwise manner in favor of a parsimonious
model if they were not significant (pFDR>0.05). First, if the three-way interaction of INR-by-PRS-byAge was insignificant, it was removed. Next, the interaction of INR-by-Age and/or PRS-by-Age were
39
assessed for significance and removed if necessary. Finally, the INR-by-PRS interaction was removed if
it was not significant in order to test the main effects of INR and PRS.
Following this approach for rs-fMRI outcomes, there were no significant interactions between any of
the predictors (INR and D-PRS) with age in either sample, therefore, the INR-by-D-PRS-by-Age, INRby-Age, and D-PRS-by-Age interaction terms were removed from all models. Therefore, for this output,
the models in the European-like and not European-like Group 1 samples examined the main effect of
INR and D-PRS on rs-fMRI outcomes (Supplemental Table 3):
rs-fMRI outcome ~ INR + D-PRS + Covariates
Though, in the European-like sample, there was a trending effect of INR-by-D-PRS on the intranetwork connectivity of the DMN and FPN networks (pFDR<0.10). Therefore, we also report this
trending interactive effect of INR and D-PRS on the intra-network connectivity of the DMN and FPN
(Supplemental Table 4). Therefore, the models presented in this table represent the following model:
rs-fMRI outcome ~ INR + D-PRS + INR*D-PRS + Covariates
For both rs-fMRI models above, as previously mentioned, the covariates were age, sex-at-birth, highest
caregivers education, caregiver-identified race/ethnicity, handedness score, MRI motion, MRI device
serial number, and the first 10 genetics PCs with random effects of subject nested by site.
Following this approach for sMRI, within the European-like and not European-like Group 1 samples,
there were no age interactions (pFDR<0.05), therefore the INR-by-D-PRS-by-Age, INR-by-Age, and DPRS-by-Age interactions were removed. Next, there were no significant (pFDR<0.05) interactions
between INR and D-PRS and therefore, this interaction term was removed. Subsequently, the final
models in the European-like and not European-like Group 1 samples examined the main effect of INR
and D-PRS on the sMRI outcomes:
sMRI outcome ~ INR + D-PRS + Covariates
For the sMRI model above, as previously mentioned, the covariates were age, sex-at-birth, highest
caregivers education, caregiver-identified race/ethnicity, handedness score, MRI device serial number,
and the first 10 genetics PCs with random effects of hemisphere nested within subject, which is then
nested within site; for examining the subcortical regions - amygdala and hippocampus - intracranial
volume was also added.
40
B.3.1.2. INR modeling in not European-like Group 2
Identical longitudinal linear mixed effects models were implemented as outlined above, but the regressor
of primary interest in these models only included �!, representing the average INR across three study
visits (i.e., did not include D-PRS, �! , or D-PRS-by-INR interaction term, �!�!). Similar to above,
higher order terms were removed in a stepwise manner in favor of a parsimonious model if they were
not significant (pFDR>0.05). In this case, INR-by-Age was first assessed for significance and removed if
appropriate, and the main effect of INR was examined.
For all the rs-fMRI outcomes, INR and age did not show a significant interaction, therefore this
interactive term was dropped. Following this approach, for the rs-fMRI models in the not Europeanlike Group 2, the final model was:
rs-fMRI outcome ~ INR + Covariates
The covariates for this rs-fMRI model were age, sex-at-birth, highest caregivers education, caregiveridentified race/ethnicity, handedness score, MRI motion, and MRI device serial number with random
effects of subject nested by site.
For most of the sMRI outcomes, the INR-by-Age interaction was not significant and therefore dropped
from the model. Following this approach, for the sMRI models in the not European-like Group 2, most
final models examined INR:
sMRI outcome ~ INR + Covariates
The covariates for this sMRI model were age, sex-at-birth, highest caregivers education, caregiveridentified race/ethnicity, handedness score, and MRI device serial number with random effects of
hemisphere nested within subject, which is then nested within site; for examining the subcortical regions
- amygdala and hippocampus - intracranial volume was also added. For three cortical regions, when
examining their thickness, there were significant (pFDR<0.05) Age-by-INR interactions; this included
the transverse frontopolar gyri and sulci, the middle frontal gyrus, and the superior frontal gyrus.
Therefore, these models were:
sMRI outcome ~ Age + INR + Age*INR + Covariates
The covariates for this sMRI model were sex-at-birth, highest caregivers education, caregiver-identified
race/ethnicity, handedness score, and MRI device serial number with random effects of hemisphere
nested within subject, which is then nested within site.
41
C. Results
For most the European-like sample identified by genetic ancestry (≥80% European-like ancestry), most
youth’s caregivers identified them as non-Hispanic white (~90%); for the not European-like Groups
(<80% European-like ancestry) caregiver’s identified the adolescents in Group 1 as Hispanic (~71%) and
other (~23%), and in Group 2 as non-Hispanic Black (~84%). Only a trending interaction between DPRS and INR was observed for the intra-network connectivity of large-scale brain networks; further
results were seen though for the main effects of INR and D-PRS. Below, all of these effects are described
in detail.
C.1 Resting-State Functional MRI
For distributions of variables in each ancestry by collection date for the resting-state functional MRI (rsfMRI) analyses, see Tables 8 & 9 below. Graphical distributions of age, INR, and D-PRS for the rsfMRI sample are presented in Supplemental Figure 7. For INR, it was found to be well distributed
and consistent over time, with 16%, 20%, and 33% in poverty (i.e., <1 INR) for samples at baseline in
the European-like, not European-like Group 1, and not European-like Group 2, respectively. After
mean-scaling the D-PRS variable, there was no mean difference between the groups, but the range
varied, with the European-like sample having the greatest spectrum.
42
Table 8: Resting-State Functional MRI Analyses Demographics for Baseline Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation; N-Miss = number of missing data.
43
Table 9: Resting-State Functional MRI Analyses Demographics for 2-Year Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation.
C.1.1 Income-to-Needs Ratios and Depression Polygenic Risk Scores Interactive Effects
As alluded to in the methods, in the European-like sample there were trending interactions of INR and
D-PRS after FDR correction (pFDR<0.01) for intra-network connectivity of the DMN (puncorrected=0.006)
44
and FPN (puncorrected=0.005), regardless of age (Figure 8, Supplemental Table 4). This interaction
suggested a positive effect of D-PRS on intra-network connectivity of the DMN and FPN in adolescents
from families with a higher INR, but a small negative effect in the DMN, or no effect in FPN, in
adolescents who come from families with a lower INR. Given the trending nature of these interactions,
we also removed this interaction and examined the main effects. However, no main effects of INR or
D-PRS on any of the resting-state functional connectivity outcomes passed FDR correction
(Supplemental Table 3). In the not European-like Group 1 sample of adolescents, no significant
interactions or main effects of INR and D-PRS were observed after FDR correction (Supplemental
Table 3).
Figure 8: Trending interaction between INR and D-PRS for intra-network connectivity
Trending interaction between INR and D-PRS of the intra-network connectivity of the A) Default mode network (DMN)
and B) frontoparietal network (FPN) in adolescents in the European-like group; the INR categories are the 25th, 50th, and
75th percentile values of the distribution of INR in the European-like sample; this corresponds to an INR raw value of 2.5,
5.0, and 6.8, respectively. Age centered at 10 years; other covariates held constant (caregiver identified race/ethnicity, MRI
motion, MRI scanner serial number, sex-at-birth, highest parents education, handedness, first 10 genetic PCs).
45
Abbreviations: INR = income-to-needs ratio; D-PRS = depression polygenic risk score; pD-PRS-by-INR = p-value for the
interactive effect of D-PRS-by-INR.
C.2 Structural MRI Results
For distributions of variables in each ancestry by collection date for the structural MRI (sMRI) analyses,
see Tables 10 & 11 below. Graphical distributions of age, INR, and D-PRS for the sMRI sample are
presented in Supplemental Figure 8. For INR, it was found to be well distributed and consistent over
time, with 16%, 21%, and 34% in poverty (i.e, <1 INR) for samples at baseline in the European-like, not
European-like Group 1, and not European-like Group 2, respectively. After mean-scaling the D-PRS
variable, there was no mean difference between the groups, but the range varied, with the European-like
sample, again, having the greatest spectrum. Overall, there were no significant interactions between DPRS and INR; only main effects of D-PRS and INR were seen and described in more detail below.
Moreover, there were no significant effects of D-PRS (pFDR<0.05) on the volumes of the amygdala and
hippocampus (Supplemental Table 6).
46
Table 10: Structural MRI Analyses Demographics for Baseline Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation; N-Miss = number of missing data.
47
Table 11: Structural MRI Analyses Demographics for 2-Year Follow-Up Data Collection
For continuous variables, the p-value represents an ANOVA t-test with equal variances, while for categorical variables, it is
a chi-squared p-value. Abbreviations: INR = income-to-needs; D-PRS = depression polygenic risk score; HS = high-school;
SD = standard deviation; N-Miss = number of missing data.
C.2.1 Depression Polygenic Risk Scores Effects
For the European-like sample, there was a significant effect of D-PRS on the thickness of the superior
frontal gyrus (�=0.008, puncorrected=1.6x10-4
; Supplemental Table 5) - which is typically associated with
48
both the FPN and DMN (Table 7). Specifically, higher D-PRS values were associated with greater
thickness regardless of age of participants (Figure 9).
Figure 9: Effect of D-PRS on Superior Frontal Gyrus in European-like Sample
A) Effect of depression polygenic risk score (D-PRS) on the thickness of the superior frontal gyrus in the European-like
sample with B) corresponding visual of the brain region location. Age centered at 10 years; other covariates held constant
(caregiver identified race/ethnicity, income-to-needs ratio, MRI scanner serial number, sex-at-birth, highest parents
education, handedness, and the first 10 genetic PCs). Abbreviations: mm = millimeters; pD-PRS = p-value for effect of D-PRS;
sMRI = structural magnetic resonance imaging.
C.2.2 Income-to-Needs Ratio Effects
In the European-like sample, there was a trending effect of INR (�=3.381, puncorrected=0.004) in the
middle temporal gyrus - a brain region associated with the DMN - with higher INR associated with
greater surface area (Supplemental Figure 9; Supplemental Table 5). For the not European-like
Group 1 sample, no significant or trending effects were found for the main effects of D-PRS and INR.
In the INR models for the not European-like Group 2 sample, there was a significant interaction
49
between INR and age in a few regions (Figure 10) plus a few main effects of INR that passed FDR
correction and were trending (Figure 11 & Supplemental Figure 10, respectively). For INR-by-Age,
there was an effect with thickness changes in the fronto-marginal gyrus and sulcus (puncorrected=0.006),
middle frontal gyrus (puncorrected=2.2x10-4
), and superior frontal gyrus (puncorrected=0.002) (Figure 10),
which are brain regions associated with both the FPN and DMN (Table 7). For these interactions, as
INR decreases, the negative effect of age is amplified, meaning individuals with lower INR were found
to have greater thickness at 9 years-of-age, but at age 13 were found to have decreased thicknesses as
compared to youths with higher INR. For the main effect of INR, there were significant effects in the
surface area of the superior frontal sulcus (�=11.865, p=uncorrected0.001) and inferior part of the precentral
sulcus (�=7.461, puncorrected=0.001) (Figure 11, Supplemental Table 5), with higher INR relating to
increases in surface area within these affiliated DMN and FPN brain regions (Table 7), respectively.
Additional positive trends were seen in the surface area of the fronto-marginal gyrus and sulcus
(�=3.781, puncorrected=0.016) and middle frontal gyrus (�=12.709, puncorrected=0.011), which are affiliated
with the DMN and FPN, as well as the precuneus (�=7.196, puncorrected=0.013) (Supplemental Figure
10, Supplemental Table 5), which is associated with the DMN (Table 7). Overall, there were no
significant effects of INR (pFDR<0.05) on the volumes of the amygdala and hippocampus
(Supplemental Table 6).
Figure 10: Effect of Age and INR on numerous sMRI metrics in the not European-like Group 2 Sample
A-C) Interactive effect of income-to-needs ratio (INR) and age on the thickness of the above structural MRI metrics in the
not European-like Group 2 sample with D-F) corresponding visual of the brain region location below each graph. The INR
50
categories are the 25th, 50th, and 75th percentile values of the distribution of INR in the not European-like Group 2 sample;
this corresponds to an INR raw value of 0.7, 1.7, and 3.3, respectively. Covariates held constant (caregiver identified
race/ethnicity, MRI scanner serial number, sex-at-birth, highest parents education, handedness). Abbreviations: pINR-by-Age =
uncorrected p-value for interaction effect of INR and age; sMRI = structural magnetic resonance imaging.
Figure 11: Effect of INR on numerous sMRI metrics in the not European-like Group 2 Sample
A-B) Effect of income-to-needs ratio (INR) on the surface area of the above structural MRI metrics in the not Europeanlike Group 2 sample with C-D) corresponding visual of the brain region location below each graph. Age centered at 10 years;
covariates held constant (caregiver identified race/ethnicity, MRI scanner serial number, sex-at-birth, highest parents
education, handedness). Abbreviations: pINR = p-value of INR effect; sMRI = structural magnetic resonance imaging.
D. Chapter 2 Discussion
The current analyses investigated the potential independent or interactive effects of household
socioeconomic status and depression polygenic risk scores on longitudinal structural and functional
MRI data from youths within the Adolescent Brain and Cognitive Development (ABCD) Study.
Specifically, we examined the D-PRS effects in adolescents within both the European-like and not
European Group 1 given D-PRS scores were found to predict depression behavior in Chapter 1. In
terms of the effect of depression polygenic risk score (D-PRS) on the brains of these two samples, the
results showed that D-PRS effects are only evident in children from the European-like sample. In terms
of income-to-needs ratios (INR), there were trending effects in the European-like sample, with more
widespread main and age-interacting effects seen in the not European-like Group 2 sample. In contrast,
no significant INR effects were noted in the not European-like Group 1. Overall, there were trending
51
interactions between D-PRS and INR in the European-like groups when examining within-network
brain connectivity.
For the D-PRS in the European-like sample, there was a trending positive association after multiple
comparisons correction between default mode and frontoparietal intra-network connectivity and DPRS when the individual has a higher INR value. Though, as an individual’s INR decreases, this effect
begins to shift directions, becoming less prominent and eventually no longer positive. Increased intranetwork connectivity of the default mode network (DMN) and decreased intra-network connectivity
of the frontoparietal network (FPN) is associated with major depressive disorders in adulthood121,190,191.
In childhood, a longitudinal study examining depression severity over time with the Children’s
Depression Inventory192 in 124 6-12 year-olds found no association with the intra-network connectivity
of the DMN and FPN193. While this study corrected for income-to-needs ratios, they did not examine
its effect. In a relatively diverse cross-sectional case-control study examining 63 13-17 year-old
adolescents, increased intra-network connectivity within the DMN was associated with major
depressive disorder diagnosis151; another study in 282 adolescents found no relationship within the
FPN21. Therefore, in our research, our D-PRS findings show that, while previous work may show
conflicting findings, more work examining the interactive effects of genetic risk and household SES may
help elucidate these discrepancies. Our work extends research to suggest that utilizing intra-network
connectivity as a biomarker of depression risk will need to take SES into account when investigating
these patterns. It is important to note that many previous neuroimaging studies, especially in children,
have been conducted on individuals whose household SES is either not assessed or who report higher
SES, thus making results difficult to generalize to the broader population194. Subsequently, these
previously determined biomarkers of depression in the existing literature may be biased to individuals
of higher household SES, and may not transfer to minoritized populations as a brain biomarker of risk
for developing depression. Specifically, there are a few reasons why the D-PRS-by-INR effect on restingstate functional connectivity may not have been observed in adolescents from the not European-like
Group 1; it could be due to the minimized predictability of the European generated D-PRS.
Additionally, even in the European-like group these effects are small, suggesting the sample size of the
not European-like Group 1 sample (N=1,373) could also be under-powered to detect such an effect.
Also, given the mean differences in INR across samples, with not European-like Group 1 having much
lower INR averages as compared to the European-like sample, which is most likely due to systemic
racism95,110, it is feasible the absence of a D-PRS effect could be due to a lower household INR average
in this sample. In other words, since the average INR in the not European-like Group 1 is lower than in
the European-like sample, and the D-PRS associations with intra-network connectivity of the DMN
and FPN in the European-like group were null at lower values, a INR interaction may not be observed.
Additional research is needed with greater efforts to recruit and study similar sample sizes with a larger
INR range in not European-like children to more fully understand the current findings. Moreover,
more research should be undertaken to enhance the generalizability of D-PRS by incorporating single
nucleotide polymorphism (SNPs) from diverse ancestries and admixed populations.
52
In terms of depression genetic risk effects on structural brain metrics, there was a significant effect
observed only in the European-like sample within the superior frontal gyrus. Here, increased thickness
of the superior frontal gyrus was associated with increased D-PRS; the superior frontal gyrus is
associated with both the default mode and frontoparietal networks (Table 7). To our knowledge, this
is the first study utilizing the current D-PRS to examine polygenic effects on structural brain differences
in youth. However, in studies utilizing the same D-PRS risk score in European-like adults undiagnosed
with depression (N= 580, mean age=32.34 years), no cortical thickness changes were seen43; this,
however, may be due to the smaller number of participants, or it could indicate that the increase in
superior frontal gyrus thickness may be a biomarker of early onset depression, which may not translate
to undiagnosed adults. Nonetheless, with other research, the superior frontal gyrus has been implicated
in depression; previous studies in adults have reported increases in thickness in the superior frontal gyrus
in patients with major depressive disorder (MDD) versus participants undiagnosed with MDD195.
Though, when examining adolescents with MDD, a previous study found no change in thickness when
compared to individuals undiagnosed with MDD196. Additional research is needed to determine if a DPRS effect during childhood, but not in adults, may indicate that increases in superior frontal gyrus
thickness are more relevant to early onset depression as compared to late onset depression. In youths,
previous cross-sectional work specifically in the ABCD Study has also reported associations between
overall volume and surface area decreases in the superior frontal lobe with increased depressive
symptoms at the baseline visit (9-10 years-old)197. However, no associations were detected between
cortical thickness and depressive symptoms in this same study197. It is important to note that there are
methodological differences between this previous study and the current research. First, the prior paper
utilized a coarser atlas - the Desikan-Killiany198 - that generated larger parcellations as compared to the
one used in the current research - the Destrieux Atlas186 - which could potentially lead to this
discrepancy. Moreover, their measure of depressive symptoms was specific to MDD diagnosis, as it was
“derived from individual items of DSM-V MDD symptomology and diagnosis of suicidality”197.
When examining income-to-needs ratios (INR) effects in the European-like sample, the middle
temporal gyrus showed a trending relationship of increased surface area with increased INR. This region
is affiliated with the default mode network (DMN), which is congruent in part with the INR-by-DPRS interaction seen in intra-network DMN connectivity. Therefore, this could be an important region
of interest, but given its trending effects, follow-up work examining INR -either more dynamically (i.e.,
more time points) or through different measures - is needed to further understand the relevance of the
middle temporal gyrus.
For the not European-like Group 2 sample, given the lack of association between D-PRS and
withdrawn/depressed symptoms from Chapter 1, exactly what the D-PRS association would represent
is unknown; therefore, a D-PRS effect was not examined. Though, when the INR effect was assessed,
significant interactions with age and INR were found. Three cortical regions showed an age interaction
with INR: the fronto-marginal gyrus and sulcus, middle frontal gyrus, and superior frontal gyrus. All
these regions are located within the frontal lobe, and overall, they decrease with age, but the magnitude
53
of the age effect was larger in those from a family reporting lower INR. In other words, we found that
as INR decreases, the negative effect of age is amplified. These same regions also displayed a positive
trend between INR and surface area. There was also a main effect of INR that was significant in the
superior frontal sulcus and inferior part of the precentral sulcus in the not European-like Group 2, with,
again, increases in surface area relating to increases in INR. In a previous cross-sectional study, when
examining household income instead of INR, a main effect of household income was noted for total
cortical thickness, with greater thickness observed in those from families with higher income199; though
this work was conducted in participants 4-20 year-olds with a majority classified as European-like
through genetic ancestry.
E. Chapter 2 Summary
Chapter 2 demonstrates that the depression polygenic risk score (D-PRS) only associates with brain
function and structure in European-like individuals. We further find that the D-PRS associations in the
intra-network functional connectivity of the default mode and frontoparietal are modified by
household socioeconomic status, as measured by income-to-needs ratios (INR). Effects seen in the intranetwork connectivity match previous biomarkers associated with early-onset depression at high INR,
but are null at low INR, suggesting more diversity is needed in sample selection when investigating these
mechanisms. For the structural measures, again, D-PRS associations were only seen in European-like
individuals, with increases in superior frontal gyrus thickness with increasing genetic risk. For the not
European-like Group 1 sample there were no effects seen in brain function or structure. In contrast, for
the not European-like Group 2 sample, we find that there are numerous main effects of INR along with
INR interactions with age. This may indicate that INR effects mainly impact the brain at lower
household socioeconomic levels since on average the not European-like Group 2 had the lowest mean
INR. Again, these differences in INR across groups are highly likely due to systemic racism that is
pervasive across space and time in the United States, creating socioeconomic inequalities that ultimately
impact health and well-being95,110.
54
DISCUSSION
Overall in Chapter 1, for the depression polygenic risk score (D-PRS), we find that this metric can
significantly associate with numerous prodromal markers of depression. The greatest effect was seen in
European-like individuals, which is expected since the D-PRS was generated in a European-like cohort.
The generalizability of this D-PRS into not European-like samples was minimal, with small effects seen
in only one of the two samples examined. For socioeconomic status - as measured by income-to-needs
ratios (INR) - we do not find that it modifies D-PRS in the European-like sample, but interacts with age
in both the not European-like groups. As we followed-up this work to examine brain structure and
function related to depression in Chapter 2, we again start to see a discrepancy between the three groups
analyzed: D-PRS effects in the European-like sample and INR effects in the not European-like
individuals. Specifically for function we find a trending effect of D-PRS that is modified by INR in the
intra-network connectivity of the default mode (DMN) and frontoparietal networks (FPN) in
European-like individuals. At high INR levels, there is a positive association between D-PRS and intranetwork connectivity of both large networks but this effect is diminished when INR decreases. For the
structural regions associated with our large scale networks - the DMN, FPN, and salience network -
effects of D-PRS were only seen in the European-like sample, with positive associations in the superior
frontal gyrus thickness. For INR, many significant main effects were seen along with INR modifications
by age, with them all located in the prefrontal cortex in not European-like Group 2 individuals. Primarily
what we find is that D-PRS associates with the European-like individualsand INR has more widespread
effects in the not European-like groups. This indicates a lack of generalizability of the D-PRS to not
European-like people and that ingrained inequities have perpetuated INR differences between groups,
which could differentially impact risk for depression onset.
Specifically, in the European-like sample, higher D-PRS values - which correlate with increased
depression risk as shown in Chapter 1 - match prior research displaying an intra-network biomarker of
increased connectivity within the DMN within adolescents151 . Though, interestingly, this effect was
modified by INR, with lower INR values leading to little to no relationship between D-PRS and intranetwork connectivity. This may contradict the previous research associating intra-network connectivity
with depression since most neuroimaging studies are conducted on individuals who indicate they are
from higher socioeconomic backgrounds194. In one particular study, increased function connectivity
within the default mode network was seen in adolescent-onset major depressive disorder (MDD)200. This
study sample was recruited from Beijing Anding Hospital, but no information on their socioeconomic
status was discussed. In a longitudinal study from the United States examining subclinical depressive
symptoms and DMN connectivity in youths201, decreased DMN connectivity over time was associated
with more depressive symptoms, but again, no socioeconomic variables were included in this analysis.
The mixed findings of DMN intra-network connectivity in adolescents as compared to adults showing
increased DMN functional connectivity190 could be due to the modifying effects of socioeconomic
factors in evaluating potential biomarkers. Therefore, future work should at minimum report
socioeconomic status or examine its potential interaction. In terms of the FPN, hypoconnectivity in
adults with depression is observed190 with no relationship seen in the current research in youths193. Our
55
research could be indicating that there is a relationship in youths that is specific to prodromal markers
of depression which is modified by socioeconomic status. Future work seeing if this effect is replicable
and whether it predicts depression onset would help indicate its usefulness as a biomarker for early-onset
depression.
When examining structural relationships between brain and depression, previous studies in adults have
seen increases also in the superior frontal gyrus in patients with major depressive disorder (MDD) versus
subjects undiagnosed with MDD195. Though, when examining adolescents with MDD, a previous study
found no change in thickness when compared to individuals undiagnosed with MDD196. When
examining associations between D-PRS and cortical thickness in a non-clinical adult cohort, there were
no associations43. It is plausible that this may indicate that increases in superior frontal gyrus are
associated with early onset depression, and thus would only be seen in adults diagnosed before 30 yearsof-age or children with higher genetic risk. In the not European-like Group 1 sample, the small
relationship between D-PRS and withdrawn/depressed symptoms in Chapter 1, did not translate to
significant effects seen in brain structure and function in Chapter 2. Since the D-PRS showed small
associations between the D-PRS and withdrawn/depressed symptoms in Chapter 1, it is unlikely that
an association between D-PRS and brain would be detected in the not European-like Group 1 sample.
Moreover, no associations between INR and brain were seen in the not European-like Group 1, which
is again probably related to the even distribution of INR within this sample given systemic inequities
within our societies95,110. To see results related to INR, a sample that is skewed high or low - like our
European-like and not European-like Group 2 samples, respectively - would have the necessary power
to detect this interaction; a sample that is more evenly distributed (i.e., not European-like Group 1
sample) would find null results given the low number of subjects at the extreme values.
Given the minimized generalizability of the depression polygenic risk score, more work must be
undertaken to ensure the transferability of our polygenic risk scores to not European-like individuals
and admixed populations, especially given the heterogeneous make-up of the United State’s population.
Therefore, when policies are ultimately set into place either at the governmental level or within the
health department, they can be applicable not only to the historical majority but minoritized groups,
increasing equity. Furthermore, when examining biomarkers of depression - like intra-network
connectivity of the DMN and FPN - socioeconomic variables need to be incorporated, along with
recruiting individuals from varying socioeconomic statuses. We also see greater effects of income-toneeds ratios in the structural brain regions associated with depression in not European-like individuals.
Overall, these differences in income-to-needs ratios across groups is heavily influenced by structural
racism that is pervasive across the United States and across history95,110. While no interactive effects were
seen in our specific sample with these predictors of interest, the burgeoning field of gene-byenvironmental effects needs to pay closer attention to the population that is analyzed as to not
perpetuate these previous biases. Overall, adolescent rates of depression have been increasing over time,
with current rates shown closer to 16% in the United States57. As depression rates increase, more work
evaluating how to mitigate onset and progression becomes exceedingly more pertinent. Sample
56
recruitment from a variety of genetic and socioeconomic backgrounds is necessary to obtain a full
picture of functional and structural brain biomarkers related to depression to help pinpoint potential
avenues of interventions to reduce depression risk.
Overall, this genetic and environmental research had many strengths. We utilized longitudinal data from
individuals with diverse genetic ancestry. This allowed us to better test whether there were potential
changes over time especially during this huge period of developmental development, while also testing
whether there was transferability of our findings across genetic ancestry. Furthermore, we were able to
validate that this particular depression genetic risk score was valid for investigating nuanced prodromal
markers of depression in European-like adolescents. Additionally, we were able to utilize all the data
points available by employing models that could adapt to missing data, thus increasing our power. While
this research offers many strengths, there were limitations that were present as well. An important
limitation is the selection and retention bias present in our ABCD Study sample; for the
withdrawn/depressed symptoms and MRI analyses, there was a decreased overlap between subjects
analyzed in 2-year follow-up due to the exclusion of visits after the beginning of the COVID-19
pandemic (Supplemental Table 8 & 9). This decrease in overlap also varied by ancestry group, with
European-like individuals having the greatest overlap between study visits; the not European-like Group
1 sample had even lower overlap, with the not European-like Group 2 sample having the lowest overlap
between visits (Supplemental Table 8 & 9). Therefore, more work examining how to utilize data
collected during the COVID-19 pandemic to assess brain and behavior changes without heavily
confounding the results is necessary. Moreover, subjects within the ABCD Study have caregiverreported higher socioeconomic status than the general population within the United States202. Since we
are seeing income-to-needs ratio effects at lower values, we could be missing potential associations at
these further decreased levels. Furthermore, when examining INR, while we did see that INR values had
medium to high intraclass correlations and thus similarity over timepoints, previous work has seen that
fluctuations in socioeconomic status can be influential203,204; therefore, future work examining how the
change in INR could affect our outcomes would be beneficial. Moreover, this work did not examine
the potential mitigating or confounding effects of comorbidities associated with depressive
symptoms205, so the specificity of this effect is unknown.
There are numerous avenues for future work. Specifically in the ABCD Study, with further waves of
data, more timepoints for socioeconomic status (SES) will be available, and thus fluctuations in SES can
be examined. Also, given the influence of puberty on depression onset and development206,207, the
examination of the effect of sex and puberty as the children age will be crucial in understanding potential
mechanisms. Moreover, with more waves of data, the rates of major depressive diagnosis will increase,
allowing researchers to see whether our findings can be predictive of disease onset. In terms of genetics,
utilizing burgeoning work incorporating single nucleotide polymorphisms (SNPs) from diverse
populations to enhance the predictive power of polygenic risk scores in diverse populations could be
implemented in the ABCD Study; this is especially necessary given the diverse genetic backgrounds that
are present in this population. Furthermore, more care into the development of statistical models that
57
can adapt to the high-collinearity between genetic ancestry PCs and PRS to assess all subjects in one
model would be ideal.
CONCLUSIONS
The goal was to investigate depressive polygenic risk and socioeconomic status along with their potential
interaction on depressive prodromal markers and brain structure and function in diverse developing
youths. We were able to leverage the large, longitudinal cohort of the Adolescent Brain Cognitive
Development (ABCD) Study73 to tackle this research question. We utilized a depression polygenic risk
score (D-PRS) validated in adolescents40,69 along with income-to-needs ratios (INR) to best reflect
household socioeconomic status. Overall we found that the D-PRS values were useful in predicting
prodromal markers of depression, but were most useful in European-like individuals as compared to not
European-like youths. Furthermore, when examining these effects in brain structure and function, we
found a trending D-PRS-by-INR effect only in intra-network connectivity of the default mode network
(DMN) and frontoparietal network (FPN) within the European-like sample, with no effects in the not
European-like samples. Though, INR showed significant effects in the thickness and surface area of the
prefrontal cortex within the not European-like Group 2 sample - which had on average a lower INR
most likely due to pervasive systemic racism95,110. This could indicate that effects of INR are influential
at low socioeconomic statuses that may not be seen in other research given that most neuroimaging
studies skew towards higher socioeconomic statuses194, which even includes the ABCD Study202.
Overall, differences in behavioral and brain related to depression risk are associated with INR, and
future studies need to take into account the potential modifying effects of socioeconomic status. More
importantly, better D-PRS values that take into account single nucleotide polymorphisms from diverse
ancestries need to be implemented to increase the predictability of these scores to improve
generalizability and equitable research.
58
BIBLIOGRAPHY
1. Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJ, eds. Global Burden of Disease and
Risk Factors. The International Bank for Reconstruction and Development / The World Bank;
2006. Accessed July 8, 2024. http://www.ncbi.nlm.nih.gov/books/NBK11812/
2. Fleisher WP, Katz LY. Early onset major depressive disorder. Paediatr Child Health.
2001;6(7):444-448.
3. Kessler RC, Avenevoli S, Ries Merikangas K. Mood disorders in children and adolescents: an
epidemiologic perspective. Biol Psychiatry. 2001;49(12):1002-1014. doi:10.1016/s0006-
3223(01)01129-5
4. 2020 Methodological Summary and Definitions | CBHSQ Data. Accessed June 3, 2024.
https://www.samhsa.gov/data/report/2020-methodological-summary-and-definitions
5. Greimel E, Adams L, Zsigo C, et al. Psychopathological symptoms as precursors of depressive
symptoms in adolescence: a prospective analysis of the GINIplus and LISA birth cohort studies.
Soc Psychiatry Psychiatr Epidemiol. 2022;57(8):1627-1639. doi:10.1007/s00127-022-02267-1
6. Del Piero LB, Saxbe DE, Margolin G. Basic emotion processing and the adolescent brain: Task
demands, analytic approaches, and trajectories of changes. Dev Cogn Neurosci. 2016;19:174-189.
doi:10.1016/j.dcn.2016.03.005
7. Bukh JD, Bock C, Vinberg M, Gether U, Kessing LV. Differences Between Early and Late Onset
Adult Depression. Clin Pract Epidemiol Ment Health. 2011;7:140-147.
doi:10.2174/1745017901107010140
8. Marek S, Hwang K, Foran W, Hallquist MN, Luna B. The Contribution of Network
Organization and Integration to the Development of Cognitive Control. PLoS Biol.
2015;13(12):e1002328. doi:10.1371/journal.pbio.1002328
9. Sisk LM, Gee DG. Stress and adolescence: vulnerability and opportunity during a sensitive
window of development. Current Opinion in Psychology. 2022;44:286-292.
doi:10.1016/j.copsyc.2021.10.005
10. Bagana E. Prodromal Symptoms and Initial Episode of Depression-Genetic and Emotional
Correlations. Procedia - Social and Behavioral Sciences. 2013;78:451-455.
doi:10.1016/j.sbspro.2013.04.329
11. Social Determinants of Mental Health. World Health Organization and Calouste Gulbenkian
Foundation; 2014.
59
12. Fava GA, Tossani E. Prodromal stage of major depression. Early Intervention in Psychiatry.
2007;1(1):9-18. doi:10.1111/j.1751-7893.2007.00005.x
13. Tracy M, Zimmerman FJ, Galea S, McCauley E, Stoep AV. What explains the relation between
family poverty and childhood depressive symptoms? J Psychiatr Res. 2008;42(14):1163-1175.
doi:10.1016/j.jpsychires.2008.01.011
14. Stanculescu E. The self-conscious emotion of pride as mediator between selfesteem and positive
affect. Procedia - Social and Behavioral Sciences. 2012;33:263-267.
doi:10.1016/j.sbspro.2012.01.124
15. Twenge JM, Campbell WK. Self-Esteem and Socioeconomic Status: A Meta-Analytic Review.
Pers Soc Psychol Rev. 2002;6(1):59-71. doi:10.1207/S15327957PSPR0601_3
16. Giedd JN, Blumenthal J, Jeffries NO, et al. Brain development during childhood and adolescence:
a longitudinal MRI study. Nat Neurosci. 1999;2(10):861-863. doi:10.1038/13158
17. Herting MM, Johnson C, Mills KL, et al. Development of subcortical volumes across adolescence
in males and females: A multisample study of longitudinal changes. NeuroImage. 2018;172:194.
doi:10.1016/j.neuroimage.2018.01.020
18. Cullen KR, Gee DG, Klimes-Dougan B, et al. A preliminary study of functional connectivity in
comorbid adolescent depression. Neurosci Lett. 2009;460(3):227-231.
doi:10.1016/j.neulet.2009.05.022
19. Rzepa E, McCabe C. Anhedonia and depression severity dissociated by dmPFC resting-state
functional connectivity in adolescents. J Psychopharmacol. 2018;32(10):1067-1074.
doi:10.1177/0269881118799935
20. Zhang S, Chen JM, Kuang L, et al. Association between abnormal default mode network activity
and suicidality in depressed adolescents. BMC Psychiatry. 2016;16(1):337. doi:10.1186/s12888-
016-1047-7
21. Wen Y, Li H, Huang Y, et al. Dynamic network characteristics of adolescents with major
depressive disorder: Attention network mediates the association between anhedonia and
attentional deficit. Hum Brain Mapp. 2023;44(17):5749-5769. doi:10.1002/hbm.26474
22. Tomasi D, Volkow ND. Effects of family income on brain functional connectivity in US
children: associations with cognition. Mol Psychiatry. 2023;28(10):4195-4202.
doi:10.1038/s41380-023-02222-9
60
23. Rebello K, Moura LM, Pinaya WHL, Rohde LA, Sato JR. Default Mode Network Maturation
and Environmental Adversities During Childhood. Chronic Stress. 2018;2:2470547018808295.
doi:10.1177/2470547018808295
24. Gellci K, Marusak HA, Peters C, Elrahal F, Iadipaolo AS, Rabinak CA. Community and
Household-Level Socioeconomic Disadvantage and Functional Organization of the Salience and
Emotion Network in Children and Adolescents. NeuroImage. 2019;184:729.
doi:10.1016/j.neuroimage.2018.09.077
25. Gonzalez MR, Palmer CE, Uban KA, Jernigan TL, Thompson WK, Sowell ER. Positive
Economic, Psychosocial, and Physiological Ecologies Predict Brain Structure and Cognitive
Performance in 9–10-Year-Old Children. Frontiers in Human Neuroscience. 2020;14.
doi:10.3389/fnhum.2020.578822
26. Noble KG, Houston SM, Brito NH, et al. Family income, parental education and brain structure
in children and adolescents. Nat Neurosci. 2015;18(5):773-778. doi:10.1038/nn.3983
27. Merz EC, Tottenham N, Noble KG. Socioeconomic Status, Amygdala Volume, and Internalizing
Symptoms in Children and Adolescents. Journal of clinical child and adolescent psychology : the
official journal for the Society of Clinical Child and Adolescent Psychology, American Psychological
Association, Division 53. 2018;47(2):312. doi:10.1080/15374416.2017.1326122
28. Luby J, Belden A, Botteron K, et al. The Effects of Poverty on Childhood Brain Development:
The Mediating Effect of Caregiving and Stressful Life Events. JAMA pediatrics.
2013;167(12):1135. doi:10.1001/jamapediatrics.2013.3139
29. King LS, Dennis EL, Humphreys KL, Thompson PM, Gotlib IH. Cross-sectional and
longitudinal associations of family income-to-needs ratio with cortical and subcortical brain
volume in adolescent boys and girls. Developmental Cognitive Neuroscience. 2020;44.
doi:10.1016/j.dcn.2020.100796
30. McLoyd VC. Socioeconomic disadvantage and child development. American Psychologist.
1998;53(2):185-204. doi:10.1037/0003-066X.53.2.185
31. Dearing E, Mccartney K, Taylor BA. Change in Family Income-to-Needs Matters More for
Children with Less. Child Development. Published online 2001:1779-1793.
32. Kim P, Evans GW, Angstadt M, et al. Effects of childhood poverty and chronic stress on emotion
regulatory brain function in adulthood. Proc Natl Acad Sci U S A. 2013;110(46):18442-18447.
doi:10.1073/pnas.1308240110
61
33. Rakesh D, Zalesky A, Whittle S. Similar but distinct – Effects of different socioeconomic
indicators on resting state functional connectivity: Findings from the Adolescent Brain Cognitive
Development (ABCD) Study®. Developmental Cognitive Neuroscience. 2021;51:101005.
doi:10.1016/j.dcn.2021.101005
34. Ip KI, Sisk LM, Horien C, et al. Associations among Household and Neighborhood
Socioeconomic Disadvantages, Resting-state Frontoamygdala Connectivity, and Internalizing
Symptoms in Youth. Journal of Cognitive Neuroscience. Published online January 28, 2022:1-32.
doi:10.1162/jocn_a_01826
35. Bevilacqua L, Goldman D. Genetics of emotion. Trends in cognitive sciences. 2011;15(9):401.
doi:10.1016/j.tics.2011.07.009
36. Lynall ME, McIntosh AM. The Heterogeneity of Depression. AJP. 2023;180(10):703-704.
doi:10.1176/appi.ajp.20230574
37. Fullerton JM, Nurnberger JI. Polygenic risk scores in psychiatry: Will they be useful for clinicians?
F1000Res. 2019;8:F1000 Faculty Rev-1293. doi:10.12688/f1000research.18491.1
38. Lewis CM, Vassos E. Polygenic risk scores: from research tools to clinical instruments. Genome
Medicine. 2020;12(1):44. doi:10.1186/s13073-020-00742-5
39. Kwong ASF, López-López JA, Hammerton G, et al. Genetic and Environmental Risk Factors
Associated With Trajectories of Depression Symptoms From Adolescence to Young Adulthood.
JAMA Network Open. 2019;2(6):e196587. doi:10.1001/jamanetworkopen.2019.6587
40. Halldorsdottir T, Piechaczek C, Soares de Matos AP, et al. Polygenic Risk: Predicting Depression
Outcomes in Clinical and Epidemiological Cohorts of Youths. Am J Psychiatry. 2019;176(8):615-
625. doi:10.1176/appi.ajp.2019.18091014
41. Douet V, Chang L, Cloak C, Ernst T. Genetic Influences on Brain Developmental Trajectories on
Neuroimaging Studies: From Infancy to Young Adulthood. Brain imaging and behavior.
2014;8(2):234. doi:10.1007/s11682-013-9260-1
42. Howard DM, Adams MJ, Clarke TK, et al. Genome-wide meta-analysis of depression identifies
102 independent variants and highlights the importance of the prefrontal brain regions. Nature
neuroscience. 2019;22(3):343. doi:10.1038/s41593-018-0326-7
43. Schmitt S, Meller T, Stein F, et al. Effects of polygenic risk for major mental disorders and crossdisorder on cortical complexity. Psychol Med. Published online April 8, 2021:1-12.
62
doi:10.1017/S0033291721001082
44. Mareckova K, Hawco C, Dos Santos FC, et al. Novel polygenic risk score as a translational tool
linking depression-related changes in the corticolimbic transcriptome with neural face processing
and anhedonic symptoms. Transl Psychiatry. 2020;10(1):1-10. doi:10.1038/s41398-020-01093-w
45. Supekar K, Musen M, Menon V. Development of Large-Scale Functional Brain Networks in
Children. PLoS Biol. 2009;7(7):e1000157. doi:10.1371/journal.pbio.1000157
46. Morales J, Welter D, Bowler EH, et al. A standardized framework for representation of ancestry
data in genomics studies, with application to the NHGRI-EBI GWAS Catalog. Genome Biology.
2018;19(1):21. doi:10.1186/s13059-018-1396-2
47. Health (US) NI of, Study BSC. Understanding Human Genetic Variation. In: NIH Curriculum
Supplement Series [Internet]. National Institutes of Health (US); 2007. Accessed July 7, 2024.
https://www.ncbi.nlm.nih.gov/books/NBK20363/
48. Star B, Spencer HG. Effects of Genetic Drift and Gene Flow on the Selective Maintenance of
Genetic Variation. Genetics. 2013;194(1):235-244. doi:10.1534/genetics.113.149781
49. Charles BA, Shriner D, Rotimi CN. Accounting for linkage disequilibrium in association analysis
of diverse populations. Genet Epidemiol. 2014;38(3):265-273. doi:10.1002/gepi.21788
50. Wang Y, Guo J, Ni G, Yang J, Visscher PM, Yengo L. Theoretical and empirical quantification of
the accuracy of polygenic scores in ancestry divergent populations. Nat Commun.
2020;11(1):3865. doi:10.1038/s41467-020-17719-y
51. Wang Y, Tsuo K, Kanai M, Neale BM, Martin AR. Challenges and Opportunities for Developing
More Generalizable Polygenic Risk Scores. Annu Rev Biomed Data Sci. 2022;5:293-320.
doi:10.1146/annurev-biodatasci-111721-074830
52. Staerk C, Klinkhammer H, Wistuba T, Maj C, Mayr A. Generalizability of polygenic prediction
models: how is the R2 defined on test data? BMC Medical Genomics. 2024;17(1):132.
doi:10.1186/s12920-024-01905-8
53. van der Meer D, Hartman CA, Pruim RHR, et al. The interaction between 5-HTTLPR and
stress exposure influences connectivity of the executive control and default mode brain networks.
Brain Imaging Behav. 2017;11(5):1486-1496. doi:10.1007/s11682-016-9633-3
54. Parker N, Wong APY, Leonard G, et al. Income inequality, gene expression, and brain maturation
during adolescence. Sci Rep. 2017;7(1):7397. doi:10.1038/s41598-017-07735-2
63
55. Goltermann J, Redlich R, Grotegerd D, et al. Childhood maltreatment and cognitive functioning:
the role of depression, parental education, and polygenic predisposition.
Neuropsychopharmacology. 2021;46(5):891-899. doi:10.1038/s41386-020-00794-6
56. Agerbo E, Trabjerg BB, Børglum AD, et al. Risk of Early-Onset Depression Associated With
Polygenic Liability, Parental Psychiatric History, and Socioeconomic Status. JAMA Psychiatry.
2021;78(4):387-397. doi:10.1001/jamapsychiatry.2020.4172
57. Wilson S, Dumornay NM. Rising Rates of Adolescent Depression in the United States:
Challenges and Opportunities in the 2020s. J Adolesc Health. 2022;70(3):354-355.
doi:10.1016/j.jadohealth.2021.12.003
58. Gladstone TRG, Beardslee WR. The prevention of depression in children and adolescents: a
review. Can J Psychiatry. 2009;54(4):212-221. doi:10.1177/070674370905400402
59. Kovacs M, Lopez-Duran N. Prodromal symptoms and atypical affectivity as predictors of major
depression in juveniles: implications for prevention. Journal of Child Psychology and Psychiatry.
2010;51(4):472-496. doi:10.1111/j.1469-7610.2010.02230.x
60. Fava GA, Tossani E. Prodromal stage of major depression. Early Intervention in Psychiatry.
2007;1(1):9-18. doi:10.1111/j.1751-7893.2007.00005.x
61. Jacobs RH, Orr JL, Gowins JR, Forbes EE, Langenecker SA. Biomarkers of intergenerational risk
for depression: A review of mechanisms in longitudinal high-risk (LHR) studies. Journal of
Affective Disorders. 2015;175:494-506. doi:10.1016/j.jad.2015.01.038
62. Rueter MA, Scaramella L, Wallace LE, Conger RD. First Onset of Depressive or Anxiety
Disorders Predicted by the Longitudinal Course of Internalizing Symptoms and ParentAdolescent Disagreements. Archives of General Psychiatry. 1999;56(8):726-732.
doi:10.1001/archpsyc.56.8.726
63. Diagnostic and Statistical Manual of Mental Disorders : DSM-5. American Psychiatric
Association; 2013.
64. Cuthbert BN. The role of RDoC in future classification of mental disorders. Dialogues Clin
Neurosci. 2020;22(1):81-85. doi:10.31887/DCNS.2020.22.1/bcuthbert
65. Jensen P, Roper M, Fisher P, et al. Test-retest reliability of the Diagnostic Interview Schedule for
Children (DISC 2.1). Parent, child, and combined algorithms. Arch Gen Psychiatry.
1995;52(1):61-71. doi:10.1001/archpsyc.1995.03950130061007
64
66. McAuley C, Layte R. Exploring the Relative Influence of Family Stressors and Socio-Economic
Context on Children’s Happiness and Well-Being. Child Ind Res. 2012;5(3):523-545.
doi:10.1007/s12187-012-9153-7
67. Sullivan PF, Neale MC, Kendler KS. Genetic Epidemiology of Major Depression: Review and
Meta-Analysis. AJP. 2000;157(10):1552-1562. doi:10.1176/appi.ajp.157.10.1552
68. Dudbridge F. Power and predictive accuracy of polygenic risk scores. PLoS Genet.
2013;9(3):e1003348. doi:10.1371/journal.pgen.1003348
69. Wray NR, Ripke S, Mattheisen M, et al. Genome-wide association analyses identify 44 risk
variants and refine the genetic architecture of major depression. Nat Genet. 2018;50(5):668-681.
doi:10.1038/s41588-018-0090-3
70. Achenbach TM. The Achenbach System of Empirically Based Assessment (ASEBA): Development,
Findings, Theory, and Applications. University of Vermont, Research Center for Children, Youth,
& Families; 2009.
71. Salsman JM, Butt Z, Pilkonis PA, et al. Emotion assessment using the NIH Toolbox. Neurology.
2013;80(11 Suppl 3):S76-86. doi:10.1212/WNL.0b013e3182872e11
72. Saarni C. Emotional Development in Childhood. Encyclopedia on Early Childhood Development.
Published online January 1, 2011.
73. Garavan H, Bartsch H, Conway K, et al. Recruiting the ABCD sample: Design considerations and
procedures. Dev Cogn Neurosci. 2018;32:16-22. doi:10.1016/j.dcn.2018.04.004
74. Jernigan TL, Brown SA, ABCD Consortium Coordinators. Introduction. Dev Cogn Neurosci.
2018;32:1-3. doi:10.1016/j.dcn.2018.02.002
75. Volkow ND, Koob GF, Croyle RT, et al. The conception of the ABCD study: From substance
use to a broad NIH collaboration. Developmental Cognitive Neuroscience. 2018;32:4-7.
doi:10.1016/j.dcn.2017.10.002
76. Hamatani S, Hiraoka D, Makita K, Tomoda A, Mizuno Y. Longitudinal impact of COVID-19
pandemic on mental health of children in the ABCD study cohort. Sci Rep. 2022;12(1):19601.
doi:10.1038/s41598-022-22694-z
77. Kiss O, Alzueta E, Yuksel D, et al. The Pandemic’s Toll on Young Adolescents: Prevention and
Intervention Targets to Preserve Their Mental Health. Journal of Adolescent Health.
65
2022;70(3):387-395. doi:10.1016/j.jadohealth.2021.11.023
78. Yip SW, Jordan A, Kohler RJ, Holmes A, Bzdok D. Multivariate, Transgenerational Associations
of the COVID-19 Pandemic Across Minoritized and Marginalized Communities. JAMA
Psychiatry. 2022;79(4):350-358. doi:10.1001/jamapsychiatry.2021.4331
79. National Academies of Sciences, Engineering, and Medicine. Using Population Descriptors in
Genetics and Genomics Research: A New Framework for an Evolving Field. National Academies
Press (US); 2023. Accessed January 30, 2024. http://www.ncbi.nlm.nih.gov/books/NBK589855/
80. Cavazos TB, Witte JS. Inclusion of variants discovered from diverse populations improves
polygenic risk score transferability. HGG Adv. 2020;2(1):100017.
doi:10.1016/j.xhgg.2020.100017
81. Auton A, Abecasis GR, Altshuler DM, et al. A global reference for human genetic variation.
Nature. 2015;526(7571):68-74. doi:10.1038/nature15393
82. Uban KA, Horton MK, Jacobus J, et al. Biospecimens and the ABCD study: Rationale, methods
of collection, measurement and early data. Dev Cogn Neurosci. 2018;32:97-106.
doi:10.1016/j.dcn.2018.03.005
83. Baurley JW, Edlund CK, Pardamean CI, Conti DV, Bergen AW. Smokescreen: a targeted
genotyping array for addiction research. BMC Genomics. 2016;17(1):145. doi:10.1186/s12864-
016-2495-7
84. Das S, Forer L, Schönherr S, et al. Next-generation genotype imputation service and methods. Nat
Genet. 2016;48(10):1284-1287. doi:10.1038/ng.3656
85. Power RA, Tansey KE, Buttenschøn HN, et al. Genome-wide Association for Major Depression
Through Age at Onset Stratification: Major Depressive Disorder Working Group of the
Psychiatric Genomics Consortium. Biol Psychiatry. 2017;81(4):325-335.
doi:10.1016/j.biopsych.2016.05.010
86. Danecek P, Bonfield JK, Liddle J, et al. Twelve years of SAMtools and BCFtools. Gigascience.
2021;10(2):giab008. doi:10.1093/gigascience/giab008
87. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK:
rising to the challenge of larger and richer datasets. GigaScience. 2015;4(1):s13742-015-0047-0048.
doi:10.1186/s13742-015-0047-8
88. Marees AT, de Kluiver H, Stringer S, et al. A tutorial on conducting genome‐wide association
66
studies: Quality control and statistical analysis. Int J Methods Psychiatr Res. 2018;27(2):e1608.
doi:10.1002/mpr.1608
89. Choi SW, Mak TSH, O’Reilly PF. Tutorial: a guide to performing polygenic risk score analyses.
Nat Protoc. 2020;15(9):2759-2772. doi:10.1038/s41596-020-0353-1
90. Choi SW, O’Reilly PF. PRSice-2: Polygenic Risk Score software for biobank-scale data.
Gigascience. 2019;8(7):giz082. doi:10.1093/gigascience/giz082
91. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components
analysis corrects for stratification in genome-wide association studies. Nat Genet. 2006;38(8):904-
909. doi:10.1038/ng1847
92. ASPE. Prior HHS Poverty Guidelines and Federal Register References. Accessed July 27, 2022.
https://aspe.hhs.gov/topics/poverty-economic-mobility/poverty-guidelines/prior-hhs-povertyguidelines-federal-register-references
93. Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for
Reliability Research. J Chiropr Med. 2016;15(2):155-163. doi:10.1016/j.jcm.2016.02.012
94. CONLEY CS, RUDOLPH KD. The emerging sex difference in adolescent depression:
Interacting contributions of puberty and peer stress. Dev Psychopathol. 2009;21(2):593-620.
doi:10.1017/S0954579409000327
95. Cardenas-Iniguez C, Gonzalez MR. Recommendations for the responsible use and
communication of race and ethnicity in neuroimaging research. Nat Neurosci. 2024;27(4):615-
628. doi:10.1038/s41593-024-01608-4
96. Xiang Y, Cao R, Li X. Parental education level and adolescent depression: A multi-country metaanalysis. Journal of Affective Disorders. 2024;347:645-655. doi:10.1016/j.jad.2023.11.081
97. Zarei K, Kahle L, Buckman DW, et al. Parent-child nativity, race, ethnicity, and common mental
health conditions among U.S. children and adolescents. J Pediatr. 2023;263:113618.
doi:10.1016/j.jpeds.2023.113618
98. LAND KC, McCALL PL, NAGIN DS. A Comparison of Poisson, Negative Binomial, and
Semiparametric Mixed Poisson Regression Models: With Empirical Applications to Criminal
Careers Data. Sociological Methods & Research. 1996;24(4):387-442.
doi:10.1177/0049124196024004001
99. Stroup WW. Generalized Linear Mixed Models: Modern Concepts, Methods and Applications.
67
CRC Press; 2016.
100. Swartout KM, Thompson MP, Koss MP, Su N. What is the Best Way to Analyze Less Frequent
Forms of Violence? The Case of Sexual Aggression. Psychol Violence. 2015;5(3):305-313.
doi:10.1037/a0038316
101. Xu T, Zhu G, Han S. Study of depression influencing factors with zero-inflated regression models
in a large-scale population survey. BMJ Open. 2017;7(11):e016471. doi:10.1136/bmjopen-2017-
016471
102. Guerrera S, Menghini D, Napoli E, Di Vara S, Valeri G, Vicari S. Assessment of
Psychopathological Comorbidities in Children and Adolescents With Autism Spectrum Disorder
Using the Child Behavior Checklist. Front Psychiatry. 2019;10. doi:10.3389/fpsyt.2019.00535
103. Jiang X, Holmes C, McVean G. The impact of age on genetic risk for common diseases. PLoS
Genet. 2021;17(8):e1009723. doi:10.1371/journal.pgen.1009723
104. Schaid DJ, Sinnwell JP, Batzler A, McDonnell SK. Polygenic risk for prostate cancer: Decreasing
relative risk with age but little impact on absolute risk. The American Journal of Human Genetics.
2022;109(5):900-908. doi:10.1016/j.ajhg.2022.03.008
105. Park LT, Zarate CA. Depression in the Primary Care Setting. N Engl J Med. 2019;380(6):559-
568. doi:10.1056/NEJMcp1712493
106. Zisook S, Lesser I, Stewart JW, et al. Effect of Age at Onset on the Course of Major Depressive
Disorder. AJP. 2007;164(10):1539-1546. doi:10.1176/appi.ajp.2007.06101757
107. Grinde KE, Qi Q, Thornton TA, et al. Generalizing Polygenic Risk Scores from Europeans to
Hispanics/Latinos. Genet Epidemiol. 2019;43(1):50-62. doi:10.1002/gepi.22166
108. Bryc K, Durand EY, Macpherson JM, Reich D, Mountain JL. The Genetic Ancestry of African
Americans, Latinos, and European Americans across the United States. Am J Hum Genet.
2015;96(1):37-53. doi:10.1016/j.ajhg.2014.11.010
109. Majara L, Kalungi A, Koen N, et al. Low and differential polygenic score generalizability among
African populations due largely to genetic diversity. HGG Adv. 2023;4(2):100184.
doi:10.1016/j.xhgg.2023.100184
110. Himmelstein KEW, Lawrence JA, Jahn JL, et al. Association Between Racial Wealth Inequities
and Racial Disparities in Longevity Among US Adults and Role of Reparations Payments, 1992
to 2018. JAMA Netw Open. 2022;5(11):e2240519. doi:10.1001/jamanetworkopen.2022.40519
68
111. Poulain T, Vogel M, Sobek C, Hilbert A, Körner A, Kiess W. Associations Between SocioEconomic Status and Child Health: Findings of a Large German Cohort Study. Int J Environ Res
Public Health. 2019;16(5):677. doi:10.3390/ijerph16050677
112. Chen Q, Kong Y, Gao W, Mo L. Effects of Socioeconomic Status, Parent–Child Relationship,
and Learning Motivation on Reading Ability. Front Psychol. 2018;9.
doi:10.3389/fpsyg.2018.01297
113. Wang J, Geng L. Effects of Socioeconomic Status on Physical and Psychological Health: Lifestyle
as a Mediator. Int J Environ Res Public Health. 2019;16(2):281. doi:10.3390/ijerph16020281
114. Twenge JM, Nolen-Hoeksema S. Age, gender, race, socioeconomic status, and birth cohort
difference on the children’s depression inventory: A meta-analysis. Journal of Abnormal
Psychology. 2002;111(4):578-588. doi:10.1037/0021-843X.111.4.578
115. Shapero BG, Chai XJ, Vangel M, et al. Neural markers of depression risk predict the onset of
depression. Psychiatry Research: Neuroimaging. 2019;285:31-39.
doi:10.1016/j.pscychresns.2019.01.006
116.Mills KL, Tamnes CK. Longitudinal Structural and Functional Brain Development in Childhood
and Adolescence. In: Cohen Kadosh K, ed. Oxford Handbook of Developmental Cognitive
Neuroscience. Oxford University Press; 2024:0. doi:10.1093/oxfordhb/9780198827474.013.4
117. Lund MJ, Alnæs D, Rokicki J, et al. Functional connectivity directionality between large-scale
resting-state networks across typical and non-typical trajectories in children and adolescence. PLoS
One. 2022;17(12):e0276221. doi:10.1371/journal.pone.0276221
118. Váša F, Romero-Garcia R, Kitzbichler MG, et al. Conservative and disruptive modes of
adolescent change in human brain functional connectivity. Proceedings of the National Academy
of Sciences. 2020;117(6):3248-3253. doi:10.1073/pnas.1906144117
119. Ho TC, Walker JC, Teresi GI, et al. Default mode and salience network alterations in suicidal and
non-suicidal self-injurious thoughts and behaviors in adolescents with depression. Transl
Psychiatry. 2021;11(1):1-14. doi:10.1038/s41398-020-01103-x
120. Rzepa E, McCabe C. Decreased anticipated pleasure correlates with increased salience network
resting state functional connectivity in adolescents with depressive symptomatology. Journal of
Psychiatric Research. 2016;82:40-47. doi:10.1016/j.jpsychires.2016.07.013
121. Menon V. Large-scale brain networks and psychopathology: a unifying triple network model.
69
Trends Cogn Sci. 2011;15(10):483-506. doi:10.1016/j.tics.2011.08.003
122. Marek S, Dosenbach NUF. The frontoparietal network: function, electrophysiology, and
importance of individual precision mapping. Dialogues Clin Neurosci. 2018;20(2):133-140.
doi:10.31887/DCNS.2018.20.2/smarek
123. Seeley WW. The Salience Network: A Neural System for Perceiving and Responding to
Homeostatic Demands. J Neurosci. 2019;39(50):9878-9882. doi:10.1523/JNEUROSCI.1138-
17.2019
124. Ma Z, Zhang N. Chapter 22 - Brain-wide connectivity architecture: developmental aspects. In:
Factors Affecting Neurodevelopment. Academic Press; 2021:247-257.
125. Zanto TP, Gazzaley A. Fronto-parietal network: flexible hub of cognitive control. Trends Cogn
Sci. 2013;17(12):10.1016/j.tics.2013.10.001. doi:10.1016/j.tics.2013.10.001
126. Solé-Padullés C, Castro-Fornieles J, de la Serna E, et al. Intrinsic connectivity networks from
childhood to late adolescence: Effects of age and sex. Developmental Cognitive Neuroscience.
2016;17:35-44. doi:10.1016/j.dcn.2015.11.004
127. Fan F, Liao X, Lei T, et al. Development of the default-mode network during childhood and
adolescence: A longitudinal resting-state fMRI study. NeuroImage. 2021;226:117581.
doi:10.1016/j.neuroimage.2020.117581
128. Wendelken C, Ferrer E, Whitaker KJ, Bunge SA. Fronto-Parietal Network Reconfiguration
Supports the Development of Reasoning Ability. Cereb Cortex. 2016;26(5):2178-2190.
doi:10.1093/cercor/bhv050
129. Wendelken C, Ferrer E, Ghetti S, Bailey SK, Cutting L, Bunge SA. Frontoparietal Structural
Connectivity in Childhood Predicts Development of Functional Connectivity and Reasoning
Ability: A Large-Scale Longitudinal Investigation. J Neurosci. 2017;37(35):8549-8558.
doi:10.1523/JNEUROSCI.3726-16.2017
130. Stevens MC, Pearlson GD, Calhoun VD. Changes in the interaction of resting-state neural
networks from adolescence to adulthood. Hum Brain Mapp. 2009;30(8):2356-2366.
doi:10.1002/hbm.20673
131. Marek S, Hwang K, Foran W, Hallquist MN, Luna B. The Contribution of Network
Organization and Integration to the Development of Cognitive Control. PLOS Biology.
2015;13(12):e1002328. doi:10.1371/journal.pbio.1002328
70
132. Brown TT, Kuperman JM, Chung Y, et al. Neuroanatomical Assessment of Biological Maturity.
Current Biology. 2012;22(18):1693-1698. doi:10.1016/j.cub.2012.07.002
133. Tamnes CK, Herting MM, Goddings AL, et al. Development of the Cerebral Cortex across
Adolescence: A Multisample Study of Inter-Related Longitudinal Changes in Cortical Volume,
Surface Area, and Thickness. J Neurosci. 2017;37(12):3402-3412.
doi:10.1523/JNEUROSCI.3302-16.2017
134. Fjell AM, Grydeland H, Wang Y, et al. The genetic organization of longitudinal subcortical
volumetric change is stable throughout the lifespan. Fornito A, Makin TR, eds. eLife.
2021;10:e66466. doi:10.7554/eLife.66466
135. Mensen VT, Wierenga LM, van Dijk S, et al. Development of cortical thickness and surface area
in autism spectrum disorder. NeuroImage: Clinical. 2017;13:215-222.
doi:10.1016/j.nicl.2016.12.003
136. Selemon LD. A role for synaptic plasticity in the adolescent development of executive function.
Transl Psychiatry. 2013;3(3):e238-e238. doi:10.1038/tp.2013.7
137. Santos A, Mier D, Kirsch P, Meyer-Lindenberg A. Evidence for a general face salience signal in
human amygdala. NeuroImage. 2011;54(4):3111-3116. doi:10.1016/j.neuroimage.2010.11.024
138. Geng H, Li X, Chen J, Li X, Gu R. Decreased Intra- and Inter-Salience Network Functional
Connectivity is Related to Trait Anxiety in Adolescents. Frontiers in Behavioral Neuroscience.
2016;9:350. doi:10.3389/fnbeh.2015.00350
139. Alves PN, Foulon C, Karolis V, et al. An improved neuroanatomical model of the default-mode
network reconciles previous neuroimaging and neuropathological findings. Commun Biol.
2019;2(1):1-14. doi:10.1038/s42003-019-0611-3
140. Geng H, Wu F, Kong L, et al. Disrupted Structural and Functional Connectivity in PrefrontalHippocampus Circuitry in First-Episode Medication-Naïve Adolescent Depression. PLOS ONE.
2016;11(2):e0148345. doi:10.1371/journal.pone.0148345
141. Shou H, Yang Z, Satterthwaite TD, et al. Cognitive behavioral therapy increases amygdala
connectivity with the cognitive control network in both MDD and PTSD. Neuroimage Clin.
2017;14:464-470. doi:10.1016/j.nicl.2017.01.030
142. Tottenham N. The Brain’s Emotional Development. Cerebrum: the Dana Forum on Brain
Science. 2017;2017. Accessed November 28, 2021.
https://www.ncbi.nlm.nih.gov/labs/pmc/articles/PMC6132040/
71
143. Wu M, Kujawa A, Lu LH, et al. Age‐related changes in amygdala–frontal connectivity during
emotional face processing from childhood into young adulthood. Human Brain Mapping.
2016;37(5):1684. doi:10.1002/hbm.23129
144. Riley JD, Chen EE, Winsell J, et al. Network specialization during adolescence: Hippocampal
effective connectivity in boys and girls. NeuroImage. 2018;175:402.
doi:10.1016/j.neuroimage.2018.04.013
145. Wierenga L, Langen M, Ambrosino S, van Dijk S, Oranje B, Durston S. Typical development of
basal ganglia, hippocampus, amygdala and cerebellum from age 7 to 24. NeuroImage. 2014;96:67-
72. doi:10.1016/j.neuroimage.2014.03.072
146.Casey BJ, Cannonier T, Conley MI, et al. The Adolescent Brain Cognitive Development (ABCD)
study: Imaging acquisition across 21 sites. Developmental Cognitive Neuroscience. 2018;32:43-54.
doi:10.1016/j.dcn.2018.03.001
147. Hirota T, Milavić G, McNicholas F, Frodl T, Skokauskas N. Chapter 10 - Depression in Children
and Adolescents. In: Frodl T, ed. Systems Neuroscience in Depression. Academic Press; 2016:309-
324. doi:10.1016/B978-0-12-802456-0.00010-8
148. Manning KJ, Steffens DC. Chapter 11 - Systems Neuroscience in Late-Life Depression. In: Frodl
T, ed. Systems Neuroscience in Depression. Academic Press; 2016:325-340. doi:10.1016/B978-0-12-
802456-0.00011-X
149. Beames JR, Kikas K, Werner-Seidler A. Prevention and early intervention of depression in young
people: an integrated narrative review of affective awareness and Ecological Momentary
Assessment. BMC Psychology. 2021;9(1):113. doi:10.1186/s40359-021-00614-6
150. Qi D, Lam CLM, Wong JJ, Chang DHF, Lee TMC. Positive affect is inversely related to the
salience and emotion network’s connectivity. Brain Imaging and Behavior. Published online
October 8, 2020. doi:10.1007/s11682-020-00397-1
151. Ho TC, Connolly CG, Blom EH, et al. Emotion-dependent functional connectivity of the
default mode network in adolescent depression. Biological psychiatry. 2015;78(9):635.
doi:10.1016/j.biopsych.2014.09.002
152. Hwang JW, Egorova N, Yang XQ, et al. Subthreshold depression is associated with impaired
resting-state functional connectivity of the cognitive control network. Transl Psychiatry.
2015;5(11):e683-e683. doi:10.1038/tp.2015.174
72
153. de Kwaasteniet BP, Rive MM, Ruhé HG, et al. Decreased Resting-State Connectivity between
Neurocognitive Networks in Treatment Resistant Depression. Frontiers in Psychiatry. 2015;6:28.
doi:10.3389/fpsyt.2015.00028
154. Bos MGN, Peters S, van de Kamp FC, Crone EA, Tamnes CK. Emerging depression in
adolescence coincides with accelerated frontal cortical thinning. Journal of Child Psychology and
Psychiatry. 2018;59(9):994-1002. doi:10.1111/jcpp.12895
155. Opel N, Redlich R, Dohm K, et al. Mediation of the influence of childhood maltreatment on
depression relapse by cortical structure: a 2-year longitudinal observational study. The Lancet
Psychiatry. 2019;6(4):318-326. doi:10.1016/S2215-0366(19)30044-6
156. MacMaster FP, Kusumakar V. Hippocampal volume in early onset depression. BMC Med.
2004;2:2. doi:10.1186/1741-7015-2-2
157. Luking KR, Repovs G, Belden AC, et al. Functional Connectivity of the Amygdala in Early
Childhood Onset Depression. Journal of the American Academy of Child and Adolescent
Psychiatry. 2011;50(10):1027. doi:10.1016/j.jaac.2011.07.019
158. Whittle S, Lichter R, Dennison M, et al. Structural Brain Development and Depression Onset
During Adolescence: A Prospective Longitudinal Study. AJP. 2014;171(5):564-571.
doi:10.1176/appi.ajp.2013.13070920
159. Jacobs RH, Barba A, Gowins JR, et al. Decoupling of the amygdala to other salience network
regions in adolescent-onset recurrent major depressive disorder. Psychol Med. 2016;46(5):1055-
1067. doi:10.1017/S0033291715002615
160. Tahmasian M, Knight D, Manoliu A, et al. Aberrant Intrinsic Connectivity of Hippocampus and
Amygdala Overlap in the Fronto-Insular and Dorsomedial-Prefrontal Cortex in Major Depressive
Disorder. Frontiers in Human Neuroscience. 2013;7:639. doi:10.3389/fnhum.2013.00639
161. Guo H, Cheng C, Cao X, Xiang J, Chen J, Zhang K. Resting-state functional connectivity
abnormalities in first-onset unmedicated depression. Neural Regeneration Research.
2014;9(2):153. doi:10.4103/1673-5374.125344
162. Cullen KR, Westlund MK, Klimes-Dougan B, et al. Abnormal amygdala resting-state functional
connectivity in adolescent depression. JAMA Psychiatry. 2014;71(10):1138-1147.
doi:10.1001/jamapsychiatry.2014.1087
163. Liu Y, Chen Y, Liang X, et al. Altered Resting-State Functional Connectivity of Multiple
Networks and Disrupted Correlation With Executive Function in Major Depressive Disorder.
73
Frontiers in Neurology. 2020;11:272. doi:10.3389/fneur.2020.00272
164. Rosso IM, Cintron CM, Steingard RJ, Renshaw PF, Young AD, Yurgelun-Todd DA. Amygdala
and hippocampus volumes in pediatric major depression. Biol Psychiatry. 2005;57(1):21-26.
doi:10.1016/j.biopsych.2004.10.027
165. Koolschijn PCMP, van IJzendoorn MH, Bakermans-Kranenburg MJ, Crone EA. Hippocampal
volume and internalizing behavior problems in adolescence. European Neuropsychopharmacology.
2013;23(7):622-628. doi:10.1016/j.euroneuro.2012.07.001
166.Šimić G, Tkalčić M, Vukić V, et al. Understanding Emotions: Origins and Roles of the Amygdala.
Biomolecules. 2021;11(6):823. doi:10.3390/biom11060823
167.Miles AE, Dos Santos FC, Byrne EM, et al. Transcriptome-based polygenic score links depressionrelated corticolimbic gene expression changes to sex-specific brain morphology and depression
risk. Neuropsychopharmacol. 2021;46(13):2304-2311. doi:10.1038/s41386-021-01189-x
168. Cattarinussi G, Delvecchio G, Sambataro F, Brambilla P. The effect of polygenic risk scores for
major depressive disorder, bipolar disorder and schizophrenia on morphological brain measures: A
systematic review of the evidence. Journal of Affective Disorders. 2022;310:213-222.
doi:10.1016/j.jad.2022.05.007
169. Pine JG, Paul SE, Johnson E, Bogdan R, Kandala S, Barch DM. Polygenic Risk for Schizophrenia,
Major Depression, and Post-traumatic Stress Disorder and Hippocampal Subregion Volumes in
Middle Childhood. Behav Genet. 2023;53(3):279-291. doi:10.1007/s10519-023-10134-1
170. Zhang X, Yan H, Yu H, et al. Childhood urbanicity interacts with polygenic risk for depression to
affect stress-related medial prefrontal function. Transl Psychiatry. 2021;11:522.
doi:10.1038/s41398-021-01650-x
171. Hagler DJ, Hatton SeanN, Cornejo MD, et al. Image processing and analysis methods for the
Adolescent Brain Cognitive Development Study. NeuroImage. 2019;202:116091.
doi:10.1016/j.neuroimage.2019.116091
172.Casey BJ, Cannonier T, Conley MI, et al. The Adolescent Brain Cognitive Development (ABCD)
study: Imaging acquisition across 21 sites. Dev Cogn Neurosci. 2018;32:43-54.
doi:10.1016/j.dcn.2018.03.001
173. Van Dijk KRA, Hedden T, Venkataraman A, Evans KC, Lazar SW, Buckner RL. Intrinsic
functional connectivity as a tool for human connectomics: theory, properties, and optimization. J
Neurophysiol. 2010;103(1):297-321. doi:10.1152/jn.00783.2009
74
174. Gordon EM, Laumann TO, Adeyemo B, Huckins JF, Kelley WM, Petersen SE. Generation and
Evaluation of a Cortical Area Parcellation from Resting-State Correlations. Cerebral Cortex.
2016;26(1):288-303. doi:10.1093/cercor/bhu239
175. Hagler DJ, Hatton SeanN, Cornejo MD, et al. Image processing and analysis methods for the
Adolescent Brain Cognitive Development Study. NeuroImage. 2019;202:116091.
doi:10.1016/j.neuroimage.2019.116091
176. Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface
reconstruction. Neuroimage. 1999;9(2):179-194. doi:10.1006/nimg.1998.0395
177. Dale AM, Sereno MI. Improved Localizadon of Cortical Activity by Combining EEG and MEG
with MRI Cortical Surface Reconstruction: A Linear Approach. J Cogn Neurosci. 1993;5(2):162-
176. doi:10.1162/jocn.1993.5.2.162
178. Fischl B, Dale AM. Measuring the thickness of the human cerebral cortex from magnetic
resonance images. Proc Natl Acad Sci U S A. 2000;97(20):11050-11055.
doi:10.1073/pnas.200033797
179. Fischl B, Sereno MI, Tootell RB, Dale AM. High-resolution intersubject averaging and a
coordinate system for the cortical surface. Hum Brain Mapp. 1999;8(4):272-284.
doi:10.1002/(sici)1097-0193(1999)8:4<272::aid-hbm10>3.0.co;2-4
180. Fischl B, Liu A, Dale AM. Automated manifold surgery: constructing geometrically accurate and
topologically correct models of the human cerebral cortex. IEEE Trans Med Imaging.
2001;20(1):70-80. doi:10.1109/42.906426
181. Fischl B, Salat DH, Busa E, et al. Whole brain segmentation: automated labeling of
neuroanatomical structures in the human brain. Neuron. 2002;33(3):341-355. doi:10.1016/s0896-
6273(02)00569-x
182. Fischl B, van der Kouwe A, Destrieux C, et al. Automatically parcellating the human cerebral
cortex. Cereb Cortex. 2004;14(1):11-22. doi:10.1093/cercor/bhg087
183. Ségonne F, Pacheco J, Fischl B. Geometrically accurate topology-correction of cortical surfaces
using nonseparating loops. IEEE Trans Med Imaging. 2007;26(4):518-529.
doi:10.1109/TMI.2006.887364
184. Ségonne F, Dale AM, Busa E, et al. A hybrid approach to the skull stripping problem in MRI.
Neuroimage. 2004;22(3):1060-1075. doi:10.1016/j.neuroimage.2004.03.032
75
185. Fischl B, Sereno MI, Dale AM. Cortical surface-based analysis. II: Inflation, flattening, and a
surface-based coordinate system. Neuroimage. 1999;9(2):195-207. doi:10.1006/nimg.1998.0396
186. DESTRIEUX C, FISCHL B, DALE A, HALGREN E. Automatic parcellation of human
cortical gyri and sulci using standard anatomical nomenclature. Neuroimage. 2010;53(1):1-15.
doi:10.1016/j.neuroimage.2010.06.010
187. Jang H, Lee JY, Lee KI, Park KM. Are there differences in brain morphology according to
handedness? Brain Behav. 2017;7(7):e00730. doi:10.1002/brb3.730
188. Maknojia S, Churchill NW, Schweizer TA, Graham SJ. Resting State fMRI: Going Through the
Motions. Front Neurosci. 2019;13:825. doi:10.3389/fnins.2019.00825
189. Kijonka M, Borys D, Psiuk-Maksymowicz K, et al. Whole Brain and Cranial Size Adjustments in
Volumetric Brain Analyses of Sex- and Age-Related Trends. Front Neurosci. 2020;14:278.
doi:10.3389/fnins.2020.00278
190. Kaiser RH, Andrews-Hanna JR, Wager TD, Pizzagalli DA. Large-scale network dysfunction in
Major Depressive Disorder: Meta-analysis of resting-state functional connectivity. JAMA
Psychiatry. 2015;72(6):603-611. doi:10.1001/jamapsychiatry.2015.0071
191. Li B, Friston K, Mody M, Wang H, Lu H, Hu D. A brain network model for depression: From
symptom understanding to disease intervention. CNS Neurosci Ther. 2018;24(11):1004-1019.
doi:10.1111/cns.12998
192. Kovacs M. The Children’s Depression, Inventory (CDI). Psychopharmacology Bulletin.
1985;21(4):995-998.
193. Qu Y, Rappaport BI, Luby JL, Barch DM. No associations in preregistered study of youth
depression and functional connectivity of fronto-parietal and default mode networks.
Neuroimage: Reports. 2021;1(3):100036. doi:10.1016/j.ynirp.2021.100036
194. LeWinn KZ, Sheridan MA, Keyes KM, Hamilton A, McLaughlin KA. Sample composition alters
associations between age and brain structure. Nat Commun. 2017;8(1):874. doi:10.1038/s41467-
017-00908-7
195.Grieve SM, Korgaonkar MS, Koslow SH, Gordon E, Williams LM. Widespread reductions in gray
matter volume in depression. Neuroimage Clin. 2013;3:332-339. doi:10.1016/j.nicl.2013.08.016
196. Schmaal L, Hibar DP, Sämann PG, et al. Cortical abnormalities in adults and adolescents with
76
major depression based on brain scans from 20 cohorts worldwide in the ENIGMA Major
Depressive Disorder Working Group. Mol Psychiatry. 2017;22(6):900-909.
doi:10.1038/mp.2016.60
197. Shen X, MacSweeney N, Chan SWY, et al. Brain structural associations with depression in a large
early adolescent sample (the ABCD study®). eClinicalMedicine. 2021;42.
doi:10.1016/j.eclinm.2021.101204
198. Desikan RS, Ségonne F, Fischl B, et al. An automated labeling system for subdividing the human
cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage. 2006;31(3):968-
980. doi:10.1016/j.neuroimage.2006.01.021
199. Piccolo LR, Merz EC, He X, Sowell ER, Noble KG. Age-Related Differences in Cortical
Thickness Vary by Socioeconomic Status. PLoS One. 2016;11(9):e0162511.
doi:10.1371/journal.pone.0162511
200. Hou X, Liu R, Zhou Y, et al. Shared and unique alterations of large-scale network connectivity in
drug-free adolescent-onset and adult-onset major depressive disorder. Transl Psychiatry.
2024;14(1):1-8. doi:10.1038/s41398-024-02974-0
201. Son JJ, Schantell M, Picci G, et al. Altered longitudinal trajectory of default mode network
connectivity in healthy youth with subclinical depressive and posttraumatic stress symptoms.
Developmental Cognitive Neuroscience. 2023;60:101216. doi:10.1016/j.dcn.2023.101216
202. Heeringa SG, Berglund PA. A Guide for Population-based Analysis of the Adolescent Brain
Cognitive Development (ABCD) Study Baseline Data. Published online February 10, 2020.
doi:10.1101/2020.02.10.942011
203. Lorant V, Croux C, Weich S, Deliège D, Mackenbach J, Ansseau M. Depression and socioeconomic risk factors: 7-year longitudinal population study. The British Journal of Psychiatry.
2007;190(4):293-298. doi:10.1192/bjp.bp.105.020040
204. Thomson RM, Igelström E, Purba AK, et al. How do income changes impact on mental health
and wellbeing for working-age adults? A systematic review and meta-analysis. The Lancet Public
Health. 2022;7(6):e515-e528. doi:10.1016/S2468-2667(22)00058-5
205.Angold A, Costello EJ. Depressive comorbidity in children and adolescents: empirical, theoretical,
and methodological issues. Am J Psychiatry. 1993;150(12):1779-1791.
doi:10.1176/ajp.150.12.1779
206. McGuire TC, McCormick KC, Koch MK, Mendle J. Pubertal Maturation and Trajectories of
77
Depression During Early Adolescence. Front Psychol. 2019;10:1362.
doi:10.3389/fpsyg.2019.01362
207. Hirtz R, Libuda L, Hinney A, et al. Age at menarche relates to depression in adolescent girls:
Comparing a clinical sample to the general pediatric population. Journal of Affective Disorders.
2022;318:103-112. doi:10.1016/j.jad.2022.08.092
78
APPENDIX
Supplemental Figure 1: General exclusionary flowchart of complete sample prior to outcome groupings...... 79
Supplemental Figure 2: Flowchart of exclusionary criteria for emotional behavior analyses.............................. 80
Supplemental Figure 3: Violin distributions with boxplots for withdrawn/depressed symptoms analyses...... 81
Supplemental Figure 4: Violin distributions with boxplots for positive affect analyses....................................... 82
Supplemental Figure 5: Flowchart of exclusionary criteria for rs-fMRI analyses.................................................. 83
Supplemental Figure 6: Flowchart of exclusionary criteria for sMRI analyses ...................................................... 84
Supplemental Figure 7: Violin distributions with boxplots for rs-fMRI analyses.................................................. 85
Supplemental Figure 8: Violin distributions with boxplots for sMRI analyses...................................................... 86
Supplemental Figure 9: Trending effect of INR on Middle Temporal Gyrus in European-like Sample........... 87
Supplemental Figure 10: Trending effect of INR on numerous sMRI metrics in the not European-like Group
2 Sample ............................................................................................................................................................................ 88
Supplemental Table 1: Withdrawn/depressed symptoms analyses model output ................................................ 89
Supplemental Table 2: Positive affect analyses model output .................................................................................. 89
Supplemental Table 3: Resting-state functional connectivity analyses model outputs......................................... 90
Supplemental Table 4: Trending resting-state functional connectivity analyses interaction model outputs..... 92
Supplemental Table 5: Cortical analyses model outputs............................................................................................ 93
Supplemental Table 6: Subcortical analyses model outputs...................................................................................... 98
Supplemental Table 7: Rates of depression in the ABCD Study ............................................................................. 99
Supplemental Table 8: Jaccard similarity matrix for behavioral analyses.............................................................. 100
Supplemental Table 9: Jaccard similarity matrix for neuroimaging analyses........................................................ 101
79
Supplemental Figure 1: General exclusionary flowchart of complete sample prior to outcome groupings
By study design, baseline visit corresponds to ages 9-10 years, 1-year follow-up visit corresponds to ages 10-11 years, 2-year
follow-up visit corresponds to ages 11-12 years and are conducted in person, whereas 6-month follow-up and 18-month
follow-up are conducted by phone. Abbreviations: INR = income-to-needs ratio.
80
Supplemental Figure 2: Flowchart of exclusionary criteria for emotional behavior analyses
Flowchart of exclusionary criteria for Depressed/Withdrawn Symptom Scale and Positive Affect Questionnaire after
exclusionary criteria is completed for the full sample (reference Supplemental Figure 1). By study design, baseline visit
corresponds to ages 9-10 years, 1-year follow-up visit corresponds to ages 10-11 years, 2-year follow-up visit corresponds to
ages 11-12 years and are conducted in person, whereas 6-month follow-up and 18-month follow-up are conducted by phone.
Abbreviations: SES = socioeconomic status as measured by income-to-needs ratio (INR).
81
Supplemental Figure 3: Violin distributions with boxplots for withdrawn/depressed symptoms analyses
Violin distributions with boxplots of A) age, B) income-to-needs (INR), C) depression polygenic risk score (D-PRS), and
D) withdrawn/depressed symptoms score reported by caregiver across all groups and by data collection year for
depressed/withdrawn symptom scale analysis. By study design, baseline visit corresponds to ages 9-10 years, 1-year followup visit corresponds to ages 10-11 years, 2-year follow-up visit corresponds to ages 11-12 years and are conducted in-person.
Note, D-PRS values are scaled within each sample.
82
Supplemental Figure 4: Violin distributions with boxplots for positive affect analyses
Violin distributions with boxplots of A) age, B) income-to-needs (INR), C) depression polygenic risk score (D-PRS), and
D) positive affect score reported by youth across all groups and by data collection year for positive affect questionnaire
analysis. By study design, baseline visit corresponds to ages 9-10 years, 1-year follow-up visit corresponds to ages 10-11 years,
2-year follow-up visit corresponds to ages 11-12 years and are conducted in person, whereas 6-month follow-up and 18-
month follow-up are conducted by phone. Note, D-PRS values are scaled within each sample.
83
Supplemental Figure 5: Flowchart of exclusionary criteria for rs-fMRI analyses
Flowchart of exclusionary criteria for resting-state functional MRI analyses after exclusionary criteria is
completed for the full sample (reference Supplemental Figure 1). By study design, baseline visit corresponds to
ages 9-10 years and 2-year follow-up visit corresponds to ages 11-12 years; both are conducted in-person.
Abbreviations: SES = socioeconomic status; rs-fMRI = resting-state functional magnetic resonance imaging.
84
Supplemental Figure 6: Flowchart of exclusionary criteria for sMRI analyses
Flowchart of exclusionary criteria for structural MRI analyses after exclusionary criteria is completed for the full sample
(reference Supplemental Figure 1). By study design, baseline visit corresponds to ages 9-10 years and 2-year follow-up visit
corresponds to ages 11-12 years; both are conducted in-person. Abbreviations: SES = socioeconomic status; sMRI =
structural magnetic resonance imaging.
85
Supplemental Figure 7: Violin distributions with boxplots for rs-fMRI analyses
Violin distributions with boxplots of A) age, B) income-to-needs (INR), and C) depression polygenic risk score (D-PRS)
across all groups and by data collection year for resting-state functional MRI analysis. By study design, baseline visit
corresponds to ages 9-10 years and 2-year follow-up visit corresponds to ages 11-12 years; both are conducted in-person.
Note, D-PRS values are scaled within each sample. Abbreviations: rs-fMRI = resting-state functional magnetic resonance
imaging.
86
Supplemental Figure 8: Violin distributions with boxplots for sMRI analyses
Violin distributions with boxplots of A) age, B) income-to-needs (INR), and C) depression polygenic risk score (D-PRS)
across all groups and by data collection year for structural MRI analysis. By study design, baseline visit corresponds to ages
9-10 years and 2-year follow-up visit corresponds to ages 11-12 years; both are conducted in-person. Note, D-PRS values are
scaled within each sample. Abbreviations: sMRI = structural magnetic resonance imaging.
87
Supplemental Figure 9: Trending effect of INR on Middle Temporal Gyrus in European-like Sample
A) Trending effect of income-to-needs ratio (INR) on the surface area of the middle temporal gyrus in the European-like
sample with B) corresponding visual of the brain region location. Age centered at 10 years; D-PRS mean-centered in the
European-like sample; other covariates held constant (caregiver identified race/ethnicity, MRI scanner serial number, sex-atbirth, highest parents education, handedness, first 10 genetic PCs). Abbreviations: mm2
, millimeters cubed; pINR = p-value
for effect of INR.
88
Supplemental Figure 10: Trending effect of INR on numerous sMRI metrics in the not European-like Group 2
Sample
A-C) Trending effect of income-to-needs ratio (INR) on the surface area of the structural MRI metrics in the not Europeanlike Group 2 sample sample with D-F) corresponding visual of the brain region location below each graph. Age centered at
10 years; PRS mean-centered in the sample; other covariates held constant (caregiver identified race/ethnicity, MRI scanner
serial number, sex-at-birth, highest parents education, handedness). Abbreviations: pINR = p-value of INR effect; sMRI =
structural magnetic resonance imaging.
89
Supplemental Table 1: Withdrawn/depressed symptoms analyses model output
Model output of covariates of interest for each group examining the effect of income-to-needs (INR) and depression
polygenic risk score (D-PRS) on depressed/withdrawn symptom scale. Note, models for the European-like sample did not
include INR-by-Age (see Chapter 1 Methods for details). Age centered at 10 years; PRS mean-centered in each group;
uncorrected p-values; bolded numbers, p<0.05 (i.e., significant). Abbreviations: IRR = incidence rate ratio; 95% CI = 95
percent confidence interval; # = number.
Supplemental Table 2: Positive affect analyses model output
Model output of covariates of interest for each group examining the effect of income-to-needs (INR) and depression
polygenic risk score (D-PRS) on the positive affect questionnaire. Age centered at 10 years; PRS mean-centered in each
group; uncorrected p-values; bolded numbers, p<0.05 (i.e., significant). Abbreviations: Coef. = beta coefficient; 95% CI =
95 percent confidence interval; # = number.
90
Supplemental Table 3: Resting-state functional connectivity analyses model outputs
91
Supplemental Table 3 continued
Model output for covariates of interest in model examining effect of income-to-needs (INR) and depression polygenic risk
score (D-PRS) on cortical structural MRI outcomes. Note, models for not European-like Group 2 included INR, but did
not include D-PRS (see Chapter 2 Methods). Age centered at 10 years; PRS mean-centered in each group; uncorrected pvalues; left and right refers to hemisphere. No findings were trending (pFDR<0.10) or were significant (pFDR<0.05).
Abbreviations: DMN = Default Mode Network; FPN = Frontoparietal Network; SN = Salience Network; Coef. = beta
coefficient; 95% CI = 95 percentage confidence interval; # = number.
92
Supplemental Table 4: Trending resting-state functional connectivity analyses interaction model outputs
Model output for covariates of interest in for trending results examining the interactive effect of income-to-needs (INR) and
depression polygenic risk score (D-PRS) on resting-state functional MRI outcomes. Age centered at 10 years; PRS meancentered in each group; uncorrected p-values; bolded numbers, pFDR<0.1 (i.e., trend-level). No findings passed FDR
correction at pFDR<0.05 (i.e., significant). Abbreviations: DMN = Default Mode Network; FPN = Frontoparietal Network;
Coef. = beta coefficient; 95% CI = 95 percent confidence interval; # = number.
93
Supplemental Table 5: Cortical analyses model outputs
94
Supplemental Table 5 continued
95
Supplemental Table 5 continued
96
Supplemental Table 5 continued
97
Supplemental Table 5 continued
Model output for covariates of interest in each group examining effect of income-to-needs (INR) and depression polygenic
risk score (D-PRS) on structural MRI outcomes. Note, models for not European-like Group 2 included INR and Age, but
did not include D-PRS (see Chapter 2 Methods text). Age centered at 10 years; PRS mean-centered in each group; bolded
numbers, pFDR<0.1 (i.e., trend-level); red numbers passed FDR correction at pFDR<0.05 (i.e., significant). Abbreviations:
Coef. = beta coefficient; 95% CI = 95 percent confidence interval; # = number.
98
Supplemental Table 6: Subcortical analyses model outputs
Model output for covariates of interest in main effects model examining effect of income-to-needs (INR) and depression
polygenic risk score (D-PRS) on subcortical structural MRI outcomes. Note, models for not European-like Group 2
included INR, but did not include D-PRS (see Chapter 2 Methods text). Age centered at 10 years; PRS mean-centered in
each group; uncorrected p-values; bolded numbers are pFDR<0.1 (i.e., trend-level). No findings passed FDR correction at
pFDR<0.05 (i.e., significant). Abbreviations: DMN = Default Mode Network; FPN = Frontoparietal Network; SN = Salience
Network; Coef. = beta coefficient; 95% CI = 95 percent confidence interval; # = number.
99
Supplemental Table 7: Rates of depression in the ABCD Study
Rates of Major Depressive Disorder (MDD) and/or Unspecified Depressive Disorder utilizing the Kiddie Schedule for
Affective Disorders and Schizophrenia for School-Aged Children (K-SADS) caregiver report at baseline visit and 2-year
follow-up visit; this data is only presented for the European-like sample used to assess the relationship between depressive
polygenic risk scores and income-to-needs rations on depression prodromes. Abbreviations: N = sample size.
100
Supplemental Table 8: Jaccard similarity matrix for behavioral analyses
Jaccard similarity matrix across waves of data for the Withdrawn/Depressed Symptom analyses and the Positive Affect
analyses within each sample: European-like, not European-like Group 1, and not European-like Group 2. A higher Jaccard
index indicates greater overlap between the subjects for the compared waves of data; a Jaccard index ranges from [0,1].
101
Supplemental Table 9: Jaccard similarity matrix for neuroimaging analyses
Jaccard similarity matrix across waves of data for the resting-state functional MRI analyses and the structural MRI analyses
within each sample: European-like, not European-like Group 1, and not European-like Group 2. A higher Jaccard index
indicates greater overlap between the subjects for the compared waves of data; a Jaccard index ranges from [0,1].
Abbreviations: MRI = magnetic resonance imaging.
Abstract (if available)
Abstract
Depression is one of the major contributors to the global burden of disease, with the World Health Organization (WHO) ranking it as the number one non-fatal contributor. Most cases of depression appear by an individual’s third decade of life, which is classified as early onset depression. The long-term effects of early onset depression extend well into adulthood, usually leading to a high rate of recurrence and significant health concerns. Research has shown that early intervention prior to disease onset leads to the best outcomes. Therefore, detecting early markers of depression risk would help mitigate the disease. Previous investigations have looked at the effect of environmental exposures or genetic influences separately, with studies beginning to examine the interactive effects of genes and the environment on risk for depression. Though, few studies have been done examining how gene-by- environment interactions may map onto prodromal brain and behavioral biomarkers of risk for early onset depression, which could greatly assist in early detection and treatment. Specifically, select brain structure and functional networks as well as distinct emotional behaviors – such as, positive affect and withdrawal symptoms – have been consistently associated with early onset depression. Ultimately, it suggests that these may be important biomarkers in studying how gene-by-environment may contribute to risk for depression that emerges prior to disease onset. Thus, this work will examine whether the well- known environmental socioeconomic predictor of family income-to-needs may have independent and/or interactive effects with an individual’s polygenic risk score for depression on the development of emotional brain structure and function from pre- to early adolescence. To accomplish this goal, an existing longitudinal dataset will be leveraged that examines approximately 8,000 subjects 9-10 year-of- age at baseline to 11-12 year-of-age at the 2-year follow-up from across the United States as part of the larger Adolescent Brain Cognitive DevelopmentSM Study (ABCD Study®). Using up to three time points for emotional behavior outcome data and up to two time points of data for the brain imaging, we examine how genes, family income-to-needs, or their interactions affect changes in brain size and function. Historically, genetic analyses have been conducted in European-like samples based on genetic ancestry. Therefore, the first section of this research will test the independent and interactive effect of an individual’s income-to-needs and polygenic risk for depression on prodromal emotional behaviors, functional brain connectivity, and brain structure of key emotional regions previously associated with depression in youth with European-like genetic ancestry. Next, to expand the field and work towards more inclusive and generalizable findings, the same analyses will be conducted in youths who are not European-like based on their genetic ancestry. Overall, we find that depression polygenic risk scores are associated with brain and behavior within the European-like youths, but less so in youths who are not European-like. We then see a potential moderating effect of family income-to-needs ratio on the depression polygenic risk score within brain network connectivity, with further main effects of income- to-needs in youths from lower socioeconomic statuses. Ultimately, the findings from this project hold the potential to identify potential brain-behavior biomarkers that may be important to consider in establishing risk for early onset depression, in hopes of improving early detection and treatment.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Decoding the neurological and genetic underpinnings of chronic pain
PDF
Neuroimaging markers of risk & resilience to brain aging and dementia
PDF
Altered emotion perception linked to structural brain differences in youth with congenital adrenal hyperplasia
PDF
Testosterone’s effect on reward, cognition, and myelination: rat brain and behavior
PDF
Using neuroinformatics to identify genomic and proteomic markers of suboptimal aging and Alzheimer's disease
PDF
Gene-environment interactions in neurodevelopment
PDF
Genetic and environmental risk factors for childhood cancer
PDF
The association of cerebrovascular disease risk factors with brain structure and its modification by genetic variation
PDF
Novel multi-site brain imaging approaches to map HIV-related neuropathology
PDF
Genomic variant analysis in whole blood of recurrent early-onset major depression patients and differential expression in BA25 of suicide completers with major depression
PDF
Examining the neuroanatomical correlates of reading in developmental dyslexia
PDF
Cultural risk and protective factors for tobacco use behaviors and depressive symptoms among American Indian adolescents in California
PDF
The link between maternal depression and adolescent daughters' risk behavior: the mediating and moderating role of family
PDF
Data modeling approaches for continuous neuroimaging genetics
PDF
Neighborhood context and adolescent mental health: development and mechanisms
PDF
The influence of DNA repair genes and prenatal tobacco exposure on childhood acute lymphoblastic leukemia risk: a gene-environment interaction study
PDF
Using observed peer discussions to understand adolescent depressive symptoms and interpersonal interactions
PDF
Pharmacokinetics and developmental effects of exposures to selective serotonin reuptake inhibitor (SSRI) antidepressants
PDF
Reproducibility and management of big data in brain MRI studies
PDF
Genetics and the environment: evaluating the role of noncoding RNA in autism spectrum disorder
Asset Metadata
Creator
Campbell, Claire Edwards
(author)
Core Title
Independent and interactive effects of depression genetic risk and household socioeconomic status on emotional behavior and brain development
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Neuroscience
Degree Conferral Date
2024-08
Publication Date
08/08/2024
Defense Date
07/17/2024
Publisher
Los Angeles, California
(original),
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
behavior,brain,depression,Development,genetics,OAI-PMH Harvest,socioeconomic status,Youth
Format
theses
(aat)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Jahanshad, Neda (
committee chair
), Gauderman, William James (
committee member
), Herting, Megan (
committee member
), Salminen, Lauren (
committee member
)
Creator Email
claireca@usc.edu,clairecampbell310@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC113998TDU
Unique identifier
UC113998TDU
Identifier
etd-CampbellCl-13351.pdf (filename)
Legacy Identifier
etd-CampbellCl-13351
Document Type
Dissertation
Format
theses (aat)
Rights
Campbell, Claire Edwards
Internet Media Type
application/pdf
Type
texts
Source
20240813-usctheses-batch-1195
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
behavior
brain
depression
genetics
socioeconomic status