Intrinsic connectivity and family dynamics: Striato-limbic markers of risk and resilience in youth at familial risk for mood disorders

Abstract

Background:

Characterizing (1) functional connectivity (FC) markers of risk and resilience in emotion and reward networks and (2) how family dynamics in youth at high familial risk for bipolar disorder (HR-BD) and major depressive disorder (HR-MDD) are related to FC may advance our understanding of the neural underpinnings of mood disorders and lead to more effective interventions.

Methods:

139 youth (43 HR-BD, 46 HR-MDD, and 50 low risk [LR]) aged 12.9+/−2.7 years were followed for 4.5+/−2.4 years. We characterized differences in striato-limbic FC that distinguished HR-BD, HR-MDD and LR; and resilience (RES) versus conversion to psychopathology (CVT). We then examined whether risk status moderated FC-family function associations. Finally, we examined whether baseline differences in FC between HR-BD, HR-MDD and LR predicted RES versus CVT at follow-up.

Results:

HR-BD had greater amygdala-middle frontal gyrus and dorsal striatum-middle frontal gyrus FC, and HR-MDD had lower amygdala-fusiform gyrus and dorsal striatum-precentral gyrus FC (voxel-level p<0.001, cluster-level FDR-corrected p<0.05). RES had greater amygdala-orbitofrontal cortex and ventral striatum-dorsal anterior cingulate cortex FC relative to CVT (voxel-level p<0.001, cluster-level FDR-corrected p<0.05). Greater family rigidity was inversely associated with amygdala-fusiform gyrus FC across all groups (FDR-corrected p=0.017), with a moderating effect of bipolar risk status (HR-BD vs. HR-MDD p<0.001; HR-BD versus LR p=0.005). Baseline FC differences did not predict RES versus CVT.

Conclusions:

Findings represent neural signatures of risk and resilience in emotion and reward processing networks in youth at familial risk for mood disorders that may be targets for novel interventions tailored to the family context.

Introduction

Bipolar disorder (BD) and major depressive disorder (MDD) are two of the leading causes of disability among youth and are associated with a significantly increased risk of suicide, the second leading cause of death in youth (1,2). One of the strongest risk factors for the development of psychopathology in youth is having a parent with a mood disorder (3). Importantly, risk factors for developing BD may be different from risk factors for developing MDD. Elevated risk for mood and other psychiatric disorders in these youth may be attributed to a combination of genetic liability and family system factors unique to parents with BD versus MDD. Indeed, parenting plays a significant role in the child’s functioning, both individually and within the family system. Parenting characteristics that are common to parents with BD and MDD – such as critical parenting and difficulty modeling emotion regulation – may increase the risk that a child will develop psychopathology (3,4). Early and specific phenotyping is needed to inform interventions that integrate information from genetic and environmental influences for developing psychopathology in high-risk offspring (5). Characterizing neural mechanisms that underlie early pathways of risk for developing specific mood disorders is essential for identifying precise targets for preventive interventions and for learning how to promote resilience in children, particularly those at familial risk.

Significant maturation of brain circuitry subserving emotion and reward processing occurs during childhood (6,7). Appropriate connectivity and recruitment of striato-limbic circuitry is essential for adaptive emotion processing and reward functions (8,9). Mood symptoms are associated with dysregulation within reward and emotion processing networks in both BD and MDD (10–12). Further, neuroimaging studies in individuals at risk for BD and for MDD suggest disorder-specific abnormalities in functional connectivity (FC) of these networks (13–16,17). Aberrant amygdala and striatum FC are among the most consistently reported findings reported in HR-BD (13,17–19) and HR-MDD (15,20,21) samples. FC is of particular interest because of its potential to improve categorization and treatment selection in pediatric mood disorders; specifically, resting-state FC analysis provides a non-invasive means of investigating connectivity across multiple brain regions (22) and can detect plasticity within brain networks (23) that underlie specific diagnostic categories.

Parenting, at the core of the family system, profoundly affects child brain development and necessitates flexibility and adaptation (24). The amygdala and striatum are key components of the parental caregiving brain network (24,25), and FC between the amygdala, striatum and para-limbic and cortical brain regions implicated in empathy and emotion regulation (26,27) subserve parents’ ability to adaptively respond to their children (28). Researchers have posited that changes in FC within this network are a mechanism by which biobehavioral synchrony develops between parents and children, forming the dynamics within a family system (24). Disruption in striato-limbic FC in parents with a mood disorder has been linked to maladaptive dynamics within the family system (29) and to increased risk for childhood-onset psychopathology in offspring (30). At present, however, we have a limited understanding of the neurobehavioral mechanisms underlying children’s experience of family dynamics and its link to the development of psychopathology. Findings that resting state networks can change with family-focused therapy in a randomized controlled setting provide the foundation for more informed neuroscience-based treatment and prevention efforts for high-risk youth (31).Relative to low-risk offspring, offspring of a parent with BD characterize their family dynamics as having lower cohesion, adaptability and flexibility, and greater rigidity and conflict (32–34), which were significantly associated with clinical symptom severity (34,35). Offspring of a parent with MDD experience lower levels of family cohesion and greater family discord relative to low-risk offspring (36), which is prospectively associated with the development of mood symptoms (37,38). Given that family function contributes to the development of emotion regulation and other mood symptoms in youth at familial risk for mood disorders (33,34,37), family function may also influence neurodevelopment of emotion and reward circuitry in high-risk youth. Indeed, associations between family environment and connectivity of emotion and reward networks have been demonstrated, with findings suggesting increased amygdala FC in youth exposed to negative parenting styles (39,40). However, researchers have not yet examined whether familial risk moderates the association between family function and intrinsic connectivity of emotion and reward networks. Understanding the moderating influence of having a parent with BD or MDD on the association between family dynamics and striato-limbic FC in youth at high-risk for mood disorders prior to symptom onset is critical given that, as above, mood symptoms (41) and striato-limbic network connectivity (31) can be modified via family-focused psychotherapeutic approaches for high-risk youth. Evaluating youth prior to symptom onset and tracking mood outcome may aide in determining which youth might benefit from this and other evidence-based interventions to personalize treatment.

Here, we assessed FC and family dynamics in 139 psychiatrically healthy youth (M_age=12.9 years) at high or low familial risk for BD or MDD and followed-up with youth approximately 4.5 years later to assess resilience versus conversion to a psychiatric diagnosis. Our primary aims were to examine (1) whether there were common or dissociable patterns of FC in the emotion and reward networks of HR-BD, HR-MDD and LR youth; and (2) whether baseline differences in amygdala and striatal network FC distinguished subsequent resilience (RES; defined as the absence of psychiatric illness at follow-up) from conversion (CVT; defined as the presence of psychiatric illness at follow-up) among high-risk youth. We predicted that high-risk youth would exhibit atypical and distinct patterns of FC, including hyperconnectivity of the amygdala in HR-BD youth and hypoconnectivity of the striatum in HR-MDD youth (13–17). Based on prior research (14,15,42,43), we hypothesized further that greater FC between the amygdala and striatum and prefrontal cortical brain regions would distinguish RES from CVT. Our secondary aim was to examine the moderating effect of risk status on the associations between child-reported family dynamics and striato-limbic FC within youth. Based on the literature demonstrating a relation between greater amygdala FC in youth exposed to negative parenting styles (39,40), we predicted that HR-BD and HR-MDD status would have a greater moderating effect on the relation between amygdala FC and dysfunctional family dynamics including greater family rigidity in families with a parent with BD (32–34) and reduced family cohesion in families with a parent with MDD (36) relative to LR. Our final, exploratory aim was to test whether baseline differences in FC and family dynamic among HR-BD, HR-MDD and LR predicted RES versus CVT in HR youth at longitudinal follow-up.

Methods and Materials

Participants

Participants were 139 youth (43 HR-BD; 46 HR-MDD; 50 LR) aged 12.9+/−2.7 years. Youth were followed for 4.5+/−2.4 years. At study entry, participants had no lifetime history of psychopathology. HR-BD and HR-MDD youth had one biological parent with a confirmed diagnosis of BD or MDD, respectively. LR had no history of psychopathology and did not have first- or second-degree relatives with an Axis I Disorder. Exclusion criteria for all participants were: significant medical disorder, substance use disorder, pervasive developmental disorder or IQ < 80, or MRI contraindications. Participants were recruited through advertisements in the local community and clinics. 8 participants were excluded due to poor data quality. Study procedures were approved by the Stanford University Institutional Review Board. Written informed assent and consent was obtained from all youth and their parents, respectively.

Demographic, Clinical and Family Function Data

All participants were assessed for psychiatric diagnoses by trained interviewers blinded to family history status at baseline and follow-up. Mood sections of the Washington University in St. Louis Kiddie-Schedule for Affective Disorders (KSADS) (44,45) were administered to assess lifetime psychiatric diagnoses. At baseline, trained interviewers administered the Structured Clinical Interview for DSM-IV (SCID) (46) to both parents to assess whether a parent met criteria for Bipolar Disorder I or MDD for the high-risk groups, and to rule out family history of psychiatric disorders in the low-risk group. All diagnostic interviews were confirmed by a board-certified child and adolescent psychiatrist. At follow-up (4.5+/−2.4 years), youth who met diagnostic criteria for a psychiatric disorder were classified as converted (CVT), and youth who did not meet any diagnostic criteria were classified as resilient (RES).

At baseline, youth completed the Wechsler Abbreviated Scale of Intelligence (WASI) (47), the Young Mania Rating Scale (YMRS) (48), the Children’s Depressive Rating Scale-Revised (CDRS-R) (49), the Multidimensional Anxiety Scale for Children (MASC) (50), the Attention-Deficit/Hyperactivity Disorder Rating Scale (ADHD-RS) (51), and the Clinical Global Assessment Scale (CGAS) (52). Age, gender, race, ethnicity, parental education (K-SADS-PL) (45), socioeconomic status (Hollingshead Four Factor Index) (53), and handedness (Crovitz Handedness Questionnaire) (54) were also assessed.

Youth and one parent completed the Family Adaptability and Cohesion Scale (FACES-IV) (55) at baseline to assess family dynamics along 6 subscales: cohesion, flexibility, disengaged, enmeshed, rigidity and chaos (Supplementary Table 1; psychometrics information is provided in supplemental text). We used the child-reported FACES-IV scores to examine their associations with FC in emotion and reward circuitry in high-risk versus low-risk youth because we wanted to elucidate how child perceptions of family dynamics are related to the child’s clinical outcomes and related to intrinsic connectivity within the child’s reward and emotion regulation networks (56,57). Further, we used child (vs. parent) FACES-IV scores to avoid the potential negative bias in reporting from a parent with a mood disorder (58). Previous work supports our decision to use child-reported (vs. parent-reported) FACES-IV scores to predict child-related clinical outcomes, particularly for internalizing disorders (56,57). Correlations between the parent- and child-reported FACES-IV scores are presented in the supplement. Parent and child FACES-IV cohesion, flexibility, and chaos subscale scores were moderately and significantly associated. Parent and child FACES-IV disengagement, enmeshed, and rigidity subscales were modestly, but not significantly, associated.

Statistical analyses were conducted using R 3.5.1. We compared HR-BD, HR-MDD, and LR groups on demographic and clinical assessments using one-way ANOVAs with post-hoc Tukey tests for continuous variables and chi-square tests for categorical variables.

ROI Selection and Functional Connectivity Analyses

We conducted FC analysis using SPM12 and CONN (59). Methods for the acquisition of fMRI and parameters for preprocessing and for artifact correction are described in the Supplement. We selected a whole-brain seed-to-voxel analysis approach to test a priori hypotheses involving striato-limbic neural networks subserving emotion and reward processing, and to facilitate comparisons with the extant resting-state fMRI literature on pediatric BD (17,18,60) and MDD (14,15,61,62). We generated whole-brain seed-to-voxel correlation maps by extracting the residual Blood Oxygen Level Dependent (BOLD) signal time course from regions of interest (ROIs), including amygdala (emotion network) and striatum (reward network). Both ventral striatum (encompassing the nucleus accumbens) and dorsal striatum (encompassing the putamen) seeds were included due to their importance in risk for pediatric mood disorders documented in previous studies (15,18,21,63). A total of six seed regions of interest (ROI) were generated, each 5mm in diameter. ROI spheres created were centered on peak coordinates from the literature from regions in which youth at high versus low familial risk for mood disorders differed in their BOLD response (14,15,17,18,61,64,65): bilateral amygdala seeds (central coordinates: +/−22, −5, +17); bilateral ventral striatum seeds (central coordinates: +/− 9, +9, −8); bilateral dorsal striatum seeds (central coordinates: +/−24, +4, +2).

Age, gender and race were included as covariates. We computed Pearson’s correlation coefficients between the time course of each seed and all other voxels in the brain. We converted correlation coefficients to Z-scores using Fisher’s transformation and used them in second-level general linear model to examine group differences between HR-BD, HR-MDD and LR, and between RES and CVT. Significance thresholding was set to voxel-level p<0.001 and cluster-level FDR-corrected p<0.05. Lastly, we examined whether there were any significant associations between baseline FC and baseline psychiatric symptom scores on the CDRS-R, YMRS, MASC, and ADHD-RS (Supplement).

Moderating Effect of Risk Status on Family Dynamic and Functional Connectivity

Because family dynamics have been shown to differ in families with a parent with BD or MDD relative to families without a parent with a psychiatric diagnosis (33,36,37), and family dynamic influences FC (39), we probed how risk status influenced FC-family dynamic associations. First, we sought to characterize family dynamic-FC associations in our dataset. We tested the null hypothesis that there were no significant associations between any FACES domains (cohesion, flexibility, disengagement, enmeshment, rigidity, and chaos) and any FC variables. We constrained our analysis to baseline striato-limbic FC that distinguished HR-BD, HR-MDD, and LR. Linear regressions were conducted across all three groups between each FACES domain (independent variables) and FC (dependent variables). Significant findings were FDR-corrected at p<0.05 for multiple comparisons (6 FACES variables × 9 FC variables=54 tests). Compared to a multiple linear regression approach (i.e., running 9 regressions with 6 FACES domain independent variables each), our analysis approach allowed us to directly test the null hypothesis while controlling for both Type I error (via FDR correction) and Type II error (due to multicollinearity between FACES domains; see Supplement). We note, however, that this approach does not allow us to assess whether any FACES domain was associated with FC aftercontrolling for the influence of other FACES domains. Therefore, we present such a multiple linear regression analysis in the Supplement, and refer the reader there for both a description of the approach and results. Notably, findings were consistent across both approaches (running 54 simple linear regressions vs. running 9 multiple linear regressions).

Next, to determine whether significant family dynamic-FC relationships were moderated by risk status, we examined whether interactions between risk status (i.e., HR-BD vs. HR-MDD vs. LR) and FACES were significantly associated with FC. Main effects of FACES subscale scores and risk status were included as independent variables, and FC was the dependent variable.

Predictive Modeling: HR-BD and HR-MDD Baseline FC and Psychiatric Diagnosis (CVT) versus Resilience (RES) at Follow-Up

We used logistic regression to examine whether baseline FC profiles that differentiated HR-BD, HR-MDD and LR predicted RES or CVT at follow-up. To assess whether these relationships were significant over and above clinical correlates, we statistically adjusted for baseline clinical symptoms, including YMRS, CDRS-R, MASC and ADHD-RS scores, as well as current and most severe CGAS scores. We also assessed whether family dynamics (FACES-IV subscales), and interaction between family dynamics and differentiating baseline FC, were associated with RES or CVT. Time to follow-up was included as a covariate in the model.

Results

Demographic, Clinical and Family Dynamic Data

Demographic and clinical data are summarized in Table 1. Groups did not differ in age, gender, handedness, IQ, parental education, social status, YMRS scores, CDRS-R scores, MASC scores, ADHD-RS scores or CGAS most severe scores but differed in race (p=0.04). CGAS current scores were higher in the LR group (M=91.02, SD=5.40) as compared to the HR-BD (M=87.85, SD=5.88) and HR-MDD groups (M=87.55, SD=5.68). Groups did not differ in level of child-reported flexibility, engagement, enmeshment, rigidity or chaotic FACES-IV subscales, but differed in family cohesion subscale scores. A post-hoc Tukey test indicated that HR-MDD (M=48.38, SD=26.11) had lower cohesion than did LR (M=67.18, SD=26.87). A comparison of parent-reported FACES-IV scores can be found in the Supplement. At follow-up, there was a group difference in the presence of any DSM-IV diagnosis, such that more HR-BD (n=16) and HR-MDD (n=17) participants met criteria for a psychiatric diagnosis relative to LR (n=7) participants (Table 1). Clinical diagnostic data at follow-up are detailed in the Supplement (Supplementary Table 2).

Table 1.

Participant Characteristics

Characteristic	LR	HR-BD	HR-MDD	Statistical Value	P Value
	M (SD)	M (SD)	M (SD)
Demographic Information
Child Age	12.75 (2.78)	12.24 (2.63)	13.50 (2.50)	F2,136=2.58	0.79
Child IQ	117.16 (14.89)	114.30 (11.40)	112.02 (14.32)	F2,134=1.66	0.20
	n (%)	n (%)	n (%)
Gender
Male	20 (40%)	13 (30.2%)	24 (52.2%)	χ2(2)=4.46	0.11
Female	30 (60%)	30 (69.8%)	22 (47.8%)
Handedness
Right	37 (74%)	35 (81.4%)	41 (89.1%)	χ2(4)=5.28	0.26
Left	6 (12%)	3 (7%)	4 (8.7%)
Ambidextrous	4 (8%)	4 (9.3%)	0
Race and Ethnicity
African American	2 (4%)	0	1 (2.2%)	χ2(8)=16.26	0.04*
Asiand	14 (28%)	2 (4.7%)	5 (10.9%)
Whited	22 (44%)	32 (74.4%)	29 (63%)
Biracial	5 (10%)	3 (7%)	2 (4.3%)
Declined to State	7 (14%)	6 (14%)	9 (19.6%)
Hispanic	6 (12%)	3 (7%)	6 (13%)	χ2(2)=0.89	0.64
Maternal Education
Less than High School	4 (8%)	7 (16.3%)	6 (13%)	χ2(8)=5.65	0.69
High School	1 (2%)	0	3 (6.5%)
Associates	23 (46%)	16 (37.2%)	21 (45.7%)
4-Year College	18 (36%)	13 (30.2%)	13 (28.3%)
More than 4-Year College	2 (4%)	2 (4.7%)	2 (4.3%)
Paternal Education
Less than High School	4 (8%)	6 (14%)	8 (17.4%)	χ2(8)=8.97	0.34
High School	1 (2%)	2 (4.7%)	1 (2.2%)
Associates	19 (38%)	11 (25.6%)	14 (30.4%)
4-Year College	19 (38%)	16 (37.2%)	13 (28.3%)
More than 4-Year College	5 (10%)	1 (2.3%)	8 (17.4%)
Social Status Range
Lower	0	1 (2.3%)	2 (4.3%)	χ2(8)=10.97	0.20
Lower-Middle	1 (2%)	4 (9.3%)	0
Middle	6 (12%)	3 (7%)	8 (17.4%)
Upper-Middle	10 (20%)	11 (25.6%)	9 (19.6%)
Upper	26 (52%)	18 (41.9%)	21 (45.7%)
	M (SD)	M (SD)	M (SD)
Baseline Clinical Information
YMRS	0.92 (1.30)	2.05 (3.28)	1.74 (3.55)	F2,133=1.92	0.15
CDRS-R	19.00 (2.54)	20.93 (6.39)	20.07 (3.96)	F2,128=2.04	0.13
MASC	37.29 (17.57)	37.08 (14.93)	41.76 (17.90)	F2,97=0.86	0.43
ADHD-RS	2.09 (2.59)	3.72 (5.74)	4.46 (7.87)	F2,83=1.41	0.25
CGAS Currentb,c	91.02 (5.40)	87.85 (5.88)	87.55 (5.68)	F2,128=6.55	0.002*
CGAS Most Severe	85.10 (5.40)	83.15 (5.88)	83.12 (5.68)	F2,128=1.08	0.34
FACES-IV Child Report e
Cohesionc	67.18 (26.87)	55.25 (28.64)	48.38 (26.11)	F2,80=3.58	0.03*
Flexibility	67.96 (21.47)	54.71 (28.36)	53.52 (23.76)	F2,80=3.02	0.05
Disengaged	28.36 (12.96)	27.00 (14.14)	32.35 (15.30)	F2,80=1.09	0.34
Enmeshed	21.93 (7.15)	20.92 (7.97)	19.68 (9.09)	F2,80=0.56	0.57
Rigid	42.79 (11.92)	36.00 (13.07)	37.32 (16.65)	F2,80=1.74	0.18
Chaotic	23.50 (8.25)	30.13 (20.63)	27.23 (12.46)	F2,80=1.42	0.25
FACES-IV Parent Report e
Cohesionb	81.74 (15.38)	61.22 (28.97)	69.37 (27.50)	F2,95=5.19	0.007*
Flexibilityb,c	78.59 (14.03)	55.13 (26.58)	64.44 (20.79)	F2,95=9.75	<0.001*
Disengaged	19.44 (6.65)	24.09 (13.83)	22.20 (10.80)	F2,95=1.45	0.24
Enmeshed	20.29 (10.79)	25.22 (13.59)	19.34 (6.99)	F2,95=2.60	0.08
Rigida,c	40.94 (12.19)	37.04 (12.30)	30.15 (8.59)	F2,95=9.50	<0.001*
Chaoticb	20.47 (12.35)	33.35 (21.85)	25.63 (15.84)	F2,95=4.23	0.02*
Follow Up Clinical Information
Follow Up Conversion Statusb,c
Converted	7 (14%)	16 (37.2%)	17 (37%)	χ2(2)=7.98	0.02*
Resilient	42 (84%)	27 (62.8%)	29 (63%)
Years of Follow Upa,c	4.98 (2.64)	5.32 (2.55)	3.24 (1.07)	F2,136=11.58	<0.001*

Abbreviations: HR-BD, High Risk-Bipolar Disorder; HR-MDD, High Risk-Major Depressive Disorder; LR, Low Risk Healthy Control; YMRS, Young Mania Rating Scale; CDRS-R, Children’s Depression Rating Scale-Revised; MASC, Multidimensional Anxiety Scale; ADHD-RS, Attention Deficit Hyperactivity Disorder Rating Scale; CGAS, Children’s Global Assessment Scale; FACES-IV, Family Adaptability and Cohesion Scale

^aPost-hoc Tukey test results indicate HR-BD and HR-MDD are significantly different

^bPost-hoc Tukey test results indicate HR-BD and LR are significantly different

^cPost-hoc Tukey test results indicate HR-MDD and LR are significantly different

^dPost-hoc Tukey test results indicate HR-BD, HR-MDD, and LR are significantly different

^eFACES-IV values indicate percentiles.

HR-BD, HR-MDD and LR Between-group FC Differences

Results are summarized in Figure 1 and Table 2A.

Figure 1. HR-BD, HR-MDD, and LR Between-Group Functional Connectivity Differences.

A) Bilateral Amygdala Seeds (central x, y, z coordinates −/+22, −5, +17). HR-BD displayed greater functional connectivity (FC) between the right amygdala and the left middle frontal gyrus (peak x, y, z coordinates −50, +20, +26) compared to HR-MDD and LR. HR-BD displayed greater FC between the left amygdala and the right middle frontal gyrus (peak x, y, z coordinates +32, +26, +28) compared to HR-MDD and LR. HR-MDD displayed reduced FC between the left amygdala and the left fusiform gyrus (peak x, y, z coordinates −38, −62, −12) compared to HR-BD and LR. HR-BD and HR-MDD displayed greater FC between the left amygdala and the left postcentral gyrus (peak x, y, z coordinates −48, −36, +56) compared to LR. HR-BD and HR-MDD displayed reduced FC between the left amygdala and the left lingual gyrus (peak x, y, z coordinates −8, −62, −8) as well as the right lingual gyrus (peak x, y, z coordinates +30, −42, −14) compared to LR.

B) Bilateral Dorsal Striatum Seeds (central x, y, z, coordinates −/+24, +4, +2). HR-BD displayed greater FC between the right dorsal striatum and the right middle frontal gyrus (peak x, y, z coordinates +32, +28, +30) compared to HR-MDD and LR. HR-MDD exhibited reduced FC between the left dorsal striatum and the left precentral gyrus (peak x, y, z coordinates +54, +2, +48) compared to HR-BD and LR. HR-BD displayed reduced FC between the left dorsal striatum and the right lingual gyrus (peak x, y, z coordinates +30, −44, −6) compared to HR-MDD and LR.

Color bar represents F values from the one-way ANOVAs. Significant between-group FC difference corrected for multiple comparisons at p<0.001 voxel level and p<0.05 cluster level threshold.

Abbreviations: HR-BD, High Risk-Bipolar Disorder; HR-MDD, High Risk-Major Depressive Disorder; LR, Low-Risk Control; L, Left; R, Right; DS, Dorsal Striatum.

Table 2.

Baseline Between-Group Functional Connectivity Differences

Group Characteristic	L/R	Cluster Size, mm3	MNI Coordinates			Mean Fisher Z Scores (SD) e
			x	y	z
2A: Between-Group Functional Connectivity Differences in HR-BD, HR-MDD, vs. LR
						HR-BD	HR-MDD	LR
Right Amygdala Seed
Middle Frontal Gyrus a,b	L	109	−50	+20	+26	0.031 (0.110)	−0.067 (0.062)	−0.072 (0.107)
Left Amygdala Seed
Middle Frontal Gyrus d	R	172	+32	+26	+28	0.023 (0.088)	−0.046 (0.013)	−0.109 (0.091)
Lingual Gyrus b,c	L	177	−8	−62	−8	−0.008 (0.103)	−0.007 (0.096)	0.091 (0.086)
Lingual Gyrus b,c	R	477	+30	−42	−14	−0.035 (0.092)	−0.026 (0.086)	0.075 (0.095)
Fusiform Gyrus a,c	L	66	−38	−62	−12	0.028 (0.122)	−0.054 (0.090)	0.063 (0.110)
Postcentral Gyrus b,c	L	64	−48	−36	+56	−0.003 (0.131)	0.029 (0.103)	−0.084 (0.109)
Right Dorsal Striatum Seed
Middle Frontal Gyrus d	R	129	+32	+28	+30	0.033 (0.100)	−0.022 (0.096)	−0.091 (0.094)
Left Dorsal Striatum Seed
Precentral Gyrus d	L	100	+54	+2	+48	0.060 (0.114)	0.003 (0.100)	0.137 (0.123)
Lingual Gyrus a,b	R	96	+30	−44	−6	−0.045 (0.092)	0.049 (0.111)	0.076 (0.108)
2B: Between-Group Functional Connectivity Differences in CVT vs. RES
						CVT	RES
Right Amygdala Seed
Paracingulate Gyrus	L	81	−16	+34	−08	−0.141 (0.153)	−0.012 (0.101)
Orbitofrontal Cortex	R	82	+20	+24	0	−0.078 (0.162)	0.057 (0.079)
Left Ventral Striatum Seed
Ventral Anterior Cingulate Cortex	L/R	337	0	+34	+24	0.053 (0.120)	0.178 (0.110)
Left Dorsal Striatum Seed
Fusiform Gyrus	L	167	−30	−12	−46	0.095 (0.098)	−0.038 (0.091)
Postcentral Gyrus	L	113	−12	−44	+72	−0.022 (0.103)	0.105 (0.111)

Abbreviations: HR-BD, Bipolar Disorder-Risk; HR-MDD, Major Depressive Disorder-Risk; LR, Control; CVT, Converted; RES, Resilient

^aPost-hoc Tukey test results indicate HR-BD and HR-MDD are significantly different

^bPost-hoc Tukey test results indicate HR-BD and LR are significantly different

^cPost-hoc Tukey test results indicate HR-MDD and LR are significantly different

^dPost-hoc Tukey test results indicate HR-BD, HR-MDD, and LR are significantly different

^eAll reported clusters reached significance of voxel-level p<0.001 and FDR-corrected cluster-level p<0.05.

Amygdala Seeds.

HR-BD had greater FC between the bilateral amygdala seeds and the left middle frontal gyrus (MFG) compared to HR-MDD and LR. HR-MDD had lower left amygdala-left fusiform gyrus FC than HR-BD and LR. Relative to LR, HR-BD and HR-MDD had greater left amygdala-left postcentral gyrus and lower left amygdala-left lingual gyrus and left amygdala-right lingual gyrus FC.

Striatum Seeds.

HR-BD had greater right dorsal striatum-right MFG and lower left dorsal striatum-right lingual gyrus FC compared to HR-MDD and LR. HR-MDD had lower left dorsal striatum-left precentral gyrus FC compared to HR-BD and LR. There were no FC differences among HR-BD, HR-MDD, and LR groups in the ventral striatum seed.

CVT and RES FC Between-group FC Differences

Results are summarized in Figure 2 and Table 2B.

Figure 2. Resilience (RES) versus Conversion (CVT) Functional Connectivity Differences.

A) Bilateral Amygdala Seeds (central x, y, z, coordinates −/+22, −5, +17). RES had greater functional connectivity (FC) between the right amygdala and the right orbitofrontal cortex (peak x, y, z coordinates +20, +24, 0) as well as with the left paracingulate gyrus (peak x, y, z coordinates −16, +34, −8) relative to CVT.

B) Ventral Striatum Seeds (central x, y, z, coordinates +9, +9, −8). RES had greater FC between the left ventral striatum and the ventral anterior cingulate cortex (peak x, y, z coordinates 0, +34, +24) relative to CVT.

C) Dorsal Striatum Seeds (central x, y, z, coordinates −/+24, +4, +2). RES had decreased FC between the left dorsal striatum seed and the left fusiform gyrus (peak x, y, z coordinates −30, −12, −46) and greater FC between the left dorsal striatum seed and the left postcentral gyrus (peak x, y, z coordinates −12, −44, +72) relative to CVT.

Color bar represents t values from the between group paired t-tests. Significant between-group connectivity difference corrected for multiple comparisons at voxel-level p<0.001, cluster level p<0.05.

Abbreviations: RES, Resilient; CVT, Converted; L, Left; R, Right; DS, Dorsal Striatum; VS, Ventral Striatum.

Amygdala Seeds.

RES had greater right amygdala-right orbitofrontal cortex and right amygdala-left paracingulate gyrus FC relative to CVT.

Striatum Seeds.

RES had greater left ventral striatum-dorsal anterior cingulate cortex and left dorsal striatum-left postcentral gyrus FC relative to CVT. CVT had greater left dorsal striatum-left fusiform gyrus FC relative to RES.

Moderating Effect of Risk Status on Family Dynamic and Functional Connectivity

Across all three groups, there was a negative association between family rigidity and left amygdala-left fusiform gyrus FC, such that high family rigidity was associated with low FC (β =−0.386, t(81)= −3.765, FDR-corrected p=0.017). Moderation analyses demonstrated that risk status moderated the relationship between family rigidity and amygdala-fusiform gyrus FC (Figure 3), such that the negative association between family rigidity and FC was stronger in the HR-BD group compared to the LR group (β =−0.916, t(77)= −2.913, p=0.005) and compared to the HR-MDD group (β =−0.938, t(77)= −3.497, p<0.001). The negative association between family rigidity and FC was not different in the HR-MDD group compared to the LR group (β =0.026, t(77)=0.082, p=0.935).

Figure 3. Risk Status Moderates the Relation between Amygdala Functional Connectivity and Family Function.

Moderation analyses demonstrated that risk status moderated the relationship between family rigidity and amygdala-fusiform gyrus FC, such that the negative association between family rigidity and FC was stronger among the HR-BD group compared to the HR-MDD group (β =−0.938, t(77)= −3.497, p<0.001) and the LR group (β =−0.916, t(77)= −2.913, p=0.005). The negative association between family rigidity and FC was not different in the HR-MDD group compared to the LR group (β =0.026, t(77)=0.082, p=0.935).

Abbreviations: HR-BD, High Risk-Bipolar Disorder; HR-MDD, High Risk-Major Depressive Disorder; L, Left; R, Right.

Logistic Regression to Predict Psychiatric Diagnosis (CVT) versus Resilience (RES)

FC profiles that differentiated HR-BD, HR-MDD and LR did not predict RES or CVT status at follow-up. Family dynamic (i.e., FACES-IV subscales) and interactions between baseline FC profiles and family dynamic also did not predict RES or CVT at follow-up. See Supplement for detailed results.

Discussion

Few studies to date have examined the influence of family dynamics and intrinsic brain network connectivity in the context of risk for psychopathology in children with a parent with BD (HR-BD) or MDD (HR-MDD). The goal of the present study was to address this gap in the literature by characterizing functional connectivity (FC) markers of risk and resilience within emotion and reward networks in relation to a measure of family dynamics. We found that FC profiles within emotion and reward processing networks distinguished HR-BD from HR-MDD and LR youth, with a moderating effect of risk status on the association between family function and cortico-limbic FC. Although risk status did not predict RES or CVT, we identified baseline FC differences in emotion and reward processing circuitry that distinguished RES from CVT within the sample of high-risk youth. These findings have the potential to inform our understanding of the neural underpinnings of early-onset mood disorders and improve intervention approaches to mitigate risk and promote resilience in high-risk youth.

HR-BD and HR-MDD youth who did not have a psychiatric diagnosis at baseline differed in FC profiles within emotion and reward processing circuitry, suggesting that there are unique trait-level vulnerability profiles or compensatory processes in these youth. HR-BD had greater amygdala and dorsal striatum FC with the middle frontal gyrus (MFG) and reduced dorsal striatum-lingual gyrus FC relative to HR-MDD and LR, and reduced amygdala-fusiform gyrus FC relative to HR-MDD. Both HR-BD and HR-MDD had reduced amygdala-postcentral gyrus and amygdala-lingual gyrus FC relative to LR. The MFG, a site of convergence of the dorsal and ventral attention networks, is involved in goal-directed and stimulus-driven attention and flexible attention modulation (66). Reduced MFG response to inhibitory errors in children has been found to predict externalizing behaviors (67,68). Indeed, deficits in response inhibition are observed in pediatric bipolar disorder (69,70). Preliminary study of the functional benefits of working memory training in youth with BD has also found evidence of near and far transfer of working memory improvement (71). The greater FC of striatal and limbic regions of emotion and reward networks with the MFG and other visuospatial and attention-processing regions in HR-BD youth suggests greater compensatory attention regulation and behavioral control. Thus, the relative hyperconnectivity of emotion and reward networks with cognitive-control regions may serve to counteract familial vulnerability in high-risk youth. While dorsal striatum FC differentiated RES from CVT high-risk youth, these two groups of youth did not differ in ventral striatal FC. In this context, we previously found that HR-BD youth show significantly reduced putamen (i.e., dorsal striatal) activation and putamen-ventral anterior cingulate cortex FC during implicit emotion processing compared to HR-MDD and LR, but no differences in ventral striatal FC (42). The dorsal and ventral striatum subserve different reward functions (72). For instance, selective activation of the ventral striatum is found during reward anticipation, whereas dorsal striatum activation also increases proportional to the magnitude of anticipated punishment (73). The present findings suggest that the ventral striatum has shared common origins in HR-BD and HR-MDD youth that are difficult to distinguish before symptom expression, but may arise after the onset of BD or MDD. Dissociable patterns of striatal reward circuit processing and FC have been found to differentiate symptomatic patients with BD and MDD (74–77) that may represent differences in pathophysiological processes underlying BD and MDD (78). Indeed, state-dependent differences within BD (e.g., (hypo)mania versus bipolar depression) cannot be inferred from asymptomatic risk group comparisons, but point to temporal differences in neural mechanisms (79). Thus, our findings may indicate trait-level premorbid differences between HR-BD and HR-MDD youth. Future studies are needed assessing differential reward-circuit activation to assess whether aberrant reward processing, particularly with respect to loss or punishment, is characteristic of HR-BD youth relative to HR-MDD and low-risk youth.

Risk status was found to moderate the relation between family rigidity and amygdala-fusiform gyrus FC, such that the negative association between family rigidity and FC was strongest among the HR-BD group. Thus, having a parent with BD moderates the relation between family rigidity and limbic FC, providing insight about how bipolar risk and context selectively modulate connectivity in brain networks that subserve emotion regulation and reward functions in children. In particular, child-perceived family rigidity appears to distinctively characterize intrinsic amygdala FC with the fusiform gyrus in HR-BD youth. Relatedly, interaction effects between negative stressful life events and activity in the amygdala and fusiform gyrus have been previously identified in HR-BD youth during emotion processing (19). Indeed, living with a parent diagnosed with either BD or MDD has been found to contribute to a greater likelihood of experiencing repeated stressful life events and reduced parental care (19,80), and limbic networks appear to be particularly vulnerable to the effects of early adversity (81). Greater activation of the fusiform gyrus and other higher-order face-processing brain regions involved in social cognition and emotion processing have also been found to differentiate HR-BD from LR youth during emotion processing (16). Parenting requires flexibility in response to changing contextual demands; thus, greater rigidity within the family context could prevent adaptive plasticity within limbic circuitry that is needed to develop healthy self-regulation and emotion processing in offspring (24).

Family dynamic has been conceptualized as an environmental factor that can be modified via therapeutic interventions such as Family-Focused Therapy for bipolar disorder to restore family functioning and to sustain recovery from mood symptoms in youth at high risk for BD (82,83). Although our findings require replication, our results suggest that family rigidity alters the development of limbic circuitry, or conversely, that amygdala hyperconnectivity with face processing regions contributes to greater family rigidity, thereby conferring heightened risk for emotion dysregulation among HR-BD youth (32,84,85). Indeed, changes in FC within this brain network is a mechanism by which researchers have posited that biobehavioral synchrony develops between parents and children, serving as a substrate for the evolving dynamics within a family system (24). Thus, targeting limbic-associated face emotion processing in a family context also has clinical implications, given that family functioning is a known modifiable risk factor that can be targeted in preventative interventions for HR-BD youth.

Greater baseline FC with prefrontal cortical regions distinguished RES relative to CVT, including greater amygdala-orbitofrontal cortex (OFC) and ventral striatum-dorsal anterior cingulate cortex FC. These brain regions are implicated in top-down modulation of emotion and reward processing (10), which are dysregulated in HR-BD (86,87) and HR-MDD (14,20) youth and posited to contribute to mood symptoms in BD and MDD (10). In a separate investigation of adolescent females at familial risk for MDD, we found that greater amygdala-OFC FC (14) and greater striatum and anterior cingulate cortex activation during reward processing (15) distinguished resilient from remitted-depressed adolescents. Distinct FC profiles with the fusiform gyrus were also found to distinguish subsequent risk from resilience, a finding that has been proposed to be a candidate resilience (versus risk) endophenotype in youth at familial risk for BD (16,88). Thus, our findings suggest that youth at high familial risk for mood disorders have compensatory, protective connectivity characteristics within emotion and reward processing networks that confer resilience.

Lastly, we investigated whether baseline FC that differentiated HR-MDD, HR-BD and LR predicted resilience or the development of psychopathology at follow-up. Differences in baseline striato-limbic FC among these groups did not predict resilience versus conversion to psychopathology. One possible explanation for this null finding is that intrinsic FC differences in limbic circuitry that differentiate HR-BD, HR-MDD and LR are distinct from those that distinguish resilience from conversion to psychiatric diagnosis. Another possibility is that our follow-up time window, which is among the longest follow-up periods of healthy youth at high risk for BD and MDD (4.5+/−2.4 years), may still not have been sufficient to capture the neurobiological impact of conversion within striato-limbic circuitry.

Limitations

We should note four limitations of this study. First, the sample size is modest; larger samples are needed to replicate and extend findings. Nevertheless, this is the largest study to date to directly compare FC between HR-BD, HR-MDD and LR youth, and differences in baseline FC that distinguish subsequent RES from CVT. Second, neuroimaging data and measures of family dynamic were obtained at one time point; it will be informative to assess longitudinal developmental trajectories of emotion-processing and reward circuitry that may distinguish HR-BD, HR-MDD and LR youth, and resilience from conversion, as well as trajectories of family dynamics and parent-child relationships. Third, not all youth were followed longitudinally through the end of adolescence; thus, some youth categorized as resilient may have gone on to develop psychopathology. Finally, due to sample size, we were unable to compare FC differences that distinguished RES versus CVT within HR-BD and HR-MDD separately. Thus, we are limited in our ability to examine how the differential profiles of FC are related to resilience versus conversion. While longitudinal studies of larger high-risk samples are needed, the present study highlights unique markers of resilience versus conversion among otherwise healthy asymptomatic youth at high risk for a mood disorder, which is a valuable contribution to the field. Indeed, BD is often misdiagnosed as MDD in the pediatric age range studied (89). Future research that follows youth into adulthood should examine whether the intrinsic FC profiles characterized here predict onset or protect youth from developing psychopathology.

While preliminary, our findings identify potential targets and modifiable risk factors that might improve approaches to prevention and treatment selection for youth at familial risk for mood disorders. Specifically, our prior work has found that youth randomized to family-focused therapy (versus standard psychoeducation) demonstrated greater connectivity between regions of the cognitive control network (i.e., ventrolateral prefrontal cortex) and the default mode network (DMN), and a reduction in depression severity that inversely correlated with DMN connectivity (31). Moreover, family communication training, problem-solving and behavioral parenting strategies have been found to facilitate symptom remission in high-risk offspring (83). Our findings highlight the potential clinical importance of interventions that promote less rigid and more flexible parent-child interactions in HR-BD youth who are otherwise healthy and asymptomatic, and these interventions may alter striato-limbic FC, thereby promoting family resilience and reducing the likelihood of the onset of psychopathology in HR-BD youth.

Conclusions

In this study, we characterized unique profiles within emotion regulation and reward processing brain networks in youth at high familial risk for bipolar and major depressive disorder relative to youth at low familial risk for psychopathology. Our findings also demonstrate the moderating effect of BD familial risk status on the relation between family function and amygdala FC. Although larger, longitudinal studies are needed, our findings offer insights into neural signatures of risk and resilience and compensatory processes within emotion and reward processing networks in youth at familial risk for pediatric mood disorders that may be targets for early identification and prevention programs. Taking a social neuroscience approach to the study of child-onset mood disorders with a focus on the mechanisms underpinning the intergenerational transmission of mood disorders, we will continue to identify modifiable mediators and moderators of risk that may help prevent psychopathology and improve outcomes in high-risk youth.