Affective Traits and History of Depression Are Related to Ventral Striatum Connectivity

Abstract

Introduction

Studying remitted Major Depressive Disorder (rMDD) facilitates a better understanding of neural mechanisms for risk, given that confounding effects of active symptoms are removed. Disrupted functional connectivity has been reported in multiple networks in MDD. However, no study to date of rMDD has specifically examined connectivity of the ventral striatum (VS), a region highly implicated in reward and motivation. We investigated functional connectivity of the VS in individuals with and without a history of MDD, and in relation to affective personality traits.

Methods

Forty-two individuals with rMDD and 28 healthy controls across two sites completed resting-state fMRI and the Behavioral Inhibition System/Behavioral Activation System Scale. Voxel-wise, whole-brain comparisons were conducted across and between groups for four seeds: left and right inferior VS (VSi), left and right superior VS (VSs).

Results

VSs connectivity to temporal and subcortical regions including the putamen and amygdala was positive and greater in HCs compared to rMDD individuals. Across groups, VSi connectivity was positively correlated with trait reward-responsiveness in somatomotor regions. Across groups, VSs connectivity was positively correlated with trait drive, particularly in the putamen, parahippocampal, and inferior temporal gyrus, and was negatively associated with trait behavioral inhibition in the anterior cingulate, frontal gyri, and insula.

Limitations

Limitations include scanning at two sites and using multiple comparisons.

Discussion

Group connectivity differences emerged from the VSs rather than VSi. VSs showed associations with trait drive and behavioral inhibition, whereas VSi corrrelated with reward-responsiveness. Depression history and affective traits contribute meaningful and specific information about VS connectivity in understanding risk for MDD.

Introduction

Studying remitted MDD (rMDD) may facilitate a clearer understanding of disease mechanisms because symptoms related to active MDD are diminished, thus potentially allowing for the dissociation of state and trait factors. However, evidence of differences between individuals with rMDD and healthy controls (HC) is mixed, specifically with reward task performance. For example, young adults with rMDD exhibited intact reward responsiveness performance compared to HC, while active MDD individuals exhibited deficits (DelDonno et al., 2015). In another study, however, individuals with rMDD responded slower than HCs during reward anticipation (Dichter et al., 2012) and failed to develop a response bias towards rewarding stimuli (Pizzagalli et al., 2009; Pechtel et al., 2013; Whitton et al., 2016). Differences in reward task performance may challenge interpretations for task-based fMRI.

Task-related brain activation differences between rMDD and HC individuals (e.g. Morgan et al., 2015) may be more sensitive than performance markers (Admon et al., 2015; Dichter et al., 2012; Jenkins et al., 2015). Individuals with rMDD exhibited greater activation relative to HCs during reward anticipation in the bilateral anterior cingulate gyrus, right midfrontal gyrus, and right cerebellum (Dichter et al., 2012). During reward receipt, individuals with rMDD exhibited hypoactivation compared to HCs in the bilateral orbital frontal cortex (OFC), right frontal pole, left insular cortex, and left thalamus, with no performance differences (Dichter et al., 2012). In a mildly stressful negative picture-viewing task, individuals with rMDD showed hyperactivation relative to HCs in the caudate, nucleus accumbens (NAcc), and putamen (Admon et al., 2015). Moreover, studies examining reward responding in individuals with active MDD have reported weaker activation in the NAcc/VS and caudate compared to HCs (Pizzagalli et al., 2009; Smoski et al., 2009). Overall, evidence for network activation alterations at rest and during tasks in MDD is mixed, particularly regarding ventral striatum (VS) activation. Examining resting-state functional connectivity (rsFC), which may reflect stable trait-like neural characteristics that are independent of and more generalizable than task activation and performance differences, may clarify mixed findings about reward-relevant neural activity in MDD. A better understanding of neural vulnerabilities revealed by rsFC analyses of individuals not currently depressed (but who are at risk for recurrence given their prior history) may improve our understanding of disease mechanisms, novel treatment targets, and relapse prediction.

Meta-analysis of rsFC in MDD has revealed three overarching patterns that support a model of depression in which dysfunctional communication between networks relates to deficits in cognitive control, emotional regulation, and internal thought (Kaiser et al., 2015). First, individuals with MDD exhibited hypoconnectivity relative to HCs in frontoparietal regions implicated in cognitive control such as the dorsolateral prefrontal cortex (DLPFC) and posterior parietal cortex (Kaiser et al., 2015). Second, individuals with MDD relative to HC showed reduced connectivity between the medial PFC (MPFC) and limbic regions, suggesting aberrant emotion regulation in terms of diminished positivity (hypoconnectivity of MPFC and NAcc) and heightened negativity (hyperconnectivity of MPFC and amygdala) (Kaiser et al., 2015). Third, there was mixed evidence for hypoconnectivity between regions responsible for salience processing and cognitive control or internal thought in those with MDD relative to HC (Kaiser et al., 2015).

Examining rsFC in remitted versus active MDD may allow for a more accurate view of the complex neural networks at play in that we may be capturing underlying mechanisms in the absence of symptoms that represent risk for relapse. It is also possible that rsFC abnormalities seen in MDD may or may not be present in rMDD. The present study focused on the VS, a primary structure in reward, motivation, and incentive salience (Di Martino et al., 2008) within the salience and emotion network (SEN; Seeley et al., 2007), as the seed for rsFC analyses. To define the SEN, we referred to work by Seeley and colleagues, who proposed that regions responsible for salience, affect, and emotion processing belong in a single salience-emotion network (2007). The SEN includes the medial thalamus, amygdala, insula, dorsal anterior cingulate cortex (dACC), and OFC (Seeley et al., 2007), with strong connections to other limbic structures such as the hypothalamus and NAcc (Sheline et al., 2010; Wang et al., 2012). The SEN integrates sensory information with internal and hedonic signals to facilitate or impede decisions, and includes emotional responses to pain, response to pleasurable music or touch, homeostatic regulation, and response to reward (Seeley et al., 2007). In addition to the SEN, previous studies have evaluated the default mode network (DMN; Yeo et al., 2011). Compared to HCs, young adults with rMDD exhibited hyperconnectivity of the DMN (i.e. posterior cingulate cortex [PCC] and subgenual cingulate) with lateral, parietal, and frontal regions (Jacobs et al., 2014). In fact, a machine-learning algorithm predicted rMDD versus HC status based on increased rsFC between left PCC and DLPFC (Bhaumik et al., 2016). Young adults with rMDD also showed left amygdala hyperconnectivity with the right medial frontal gyrus, medial parietal lobe, rostral ACC, and left parahippocampal gyrus, which suggests that SEN connectivity may be increased in individuals with rMDD compared to HCs (Jacobs et al., 2014). In contrast, greater amygdala connectivity with the superior VS predicted rMDD versus HC classification (Bhaumik et al., 2016). Hyperconnectivity from the caudate to the amygdala and hippocampus has been reported in adults with rMDD compared to HCs (Admon et al., 2015). In late adolescent boys with rMDD, hyperconnectivity between the NAcc and MPFC relative to HCs was observed when winning a reward relative to losing (Morgan et al., 2015). Overall, these rsFC studies suggest that the DMN and SEN are disrupted in rMDD.

The present study investigated rsFC of the VS in rMDD young adults, an understudied seed and sample. We chose to separately examine the superior VS (VSs, or ventral caudate head) and the inferior VS (VSi, or NAcc; Di Martino et al., 2008), not often done in the depression and reward literature. These segregations were based upon the seeds used by Di Martino and colleagues so that the present results could be compared across that study. The more dorsal and posterior caudal seeds were excluded to limit multiple comparisons concerns and focus on reward circuitry. First, we examined VS connectivity across groups in relation to self-reported behavioral approach (reward-responsiveness and drive) and inhibition to probe whether VS connectivity dimensionally relates to affective and motivation-relevant personality traits. Reward-responsiveness, drive, and behavioral inhibition (Carver & White, 1994) may better capture dimensional risks for illness, given previous findings that these affective traits did not differ between individuals with rMDD and HCs (DelDonno et al., 2015). Second, despite mixed findings on SEN functioning in active MDD, we hypothesized increased connectivity between the VS and SEN in rMDD compared to HC individuals. In a smaller subset of the present sample, our group recently found that greater amygdala and VSs connectivity predicted rMDD status (Bhaumik et al., 2016) and we designed our second hypothesis to follow up on this finding. We also predicted hyperconnectivity from the VS to the DMN in the rMDD group relative to HCs, given the association of DMN hyperconnectivity with key clinical features of rMDD (Jacobs et al., 2014).

Methods

Participants

Participants were recruited from the community in two sites: Ann Arbor, Michigan (2010-2012) and Chicago, Illinois (2012-2015). Participants were 70 adults aged 18-23 years old with no chronic or serious medical conditions. Of the 34 participants recruited in Ann Arbor, 56% were in the rMDD group. Of the 36 participants recruited in Chicago, 64% were in the rMDD group. Remitted MDD participants identified themselves as 69% white, 10% Asian, 5% black, 5% more than one race, 2% Middle Eastern, 2% Mexican, and 7% not reported. Remitted MDD participants (n = 42, 76% female) had one to five prior episodes of MDD (M = 2.05, SD = 1.97) but none in the past month, and the average age of first episode was 16 (SD = 2.91). Remitted MDD participants could have a family history of depression or anxiety, could have a comorbid anxiety diagnosis, and were free of psychotropic medication use in the past three months. Healthy control (HC; n = 28, 50% female) participants had no personal or family history of any psychiatric disorder. HCs identified themselves as 68% white, 14% Asian, 11% Indian, 4% black, and 4% Latino. Current or past psychotic symptoms, current or past bipolar disorder or mania, family history of psychosis, history of suicidal attempts or ideation in the past six months, regular tobacco use (more than 10 cigarettes per week), and presence of alcohol or substance abuse in the last six months were exclusionary criteria for both groups. Table 1 reports participant demographics and clinical characteristics.

Table 1. Participant demographics and clinical characteristics.

	rMDD (n = 42)	HC (n = 28)	t/χ2	p
Age	21 (1.53)	21 (1.56)	-.32	.73
Percent female	76%	50%	5.12	.02
Race/ethnicity	-	-	11.01	.20
Years of education	14.40 (1.36)	14.86 (1.24)	1.41	.16
Verbal IQ estimate	107.86 (9.07)	107.00 (8.96)	.39	.70
HAM-D	4.17 (.70)	.46 (1.07)	-2.95	.004
HAM-A	2.84 (2.99)	.60 (1.35)	-2.69	.01
BAS-RR	16.89 (1.87)	17.04 (1.48)	.35	.73
BAS-D	10.61 (2.06)	10.29 (1.35)	.35	.73
BIS	19.82 (3.38)	19.08 (2.55)	-.97	.33

HC, Healthy control; rMDD, remitted major depressive disorder; HAM-D, Hamilton Rating Scale for Depression; HAM-A, Hamilton Rating Scale for Anxiety; BAS-RR, Behavioral Activation Scale – Reward-Responsiveness; BAS-D, Behavioral Activation Scale – Drive.

Procedure

All study procedures were approved by the University of Michigan (UM) and the University of Illinois at Chicago (UIC) IRBs. Participants were screened over the phone by trained research assistants and provided informed consent in accordance with the Declaration of Helsinki. History of MDD diagnosis was determined by the Diagnostic Interview for Genetic Studies (Nurnberger et al., 1994), conducted by a trained masters- or doctoral-level interviewer. The interviewer also assessed current symptoms of depression and anxiety with the Hamilton Depression Rating Scale (HAM-D; Hamilton, 1960), and Hamilton Anxiety Rating Scale (HAM-A; Hamilton, 1959). A verbal IQ estimate was obtained (Shipley et al., 2009). Participants completed the BIS/BAS questionnaire and an fMRI scan. The fMRI tasks, order of tasks, and duration of scan were the same at both sites. Participants were compensated $120 for completion of the fMRI scan with the opportunity to win up to an additional $52 based upon performance on a computerized task. The resting-state scan was acquired after all tasks had been performed.

Measures

Behavioral Inhibition Scale/Behavioral Activation Scale (BIS/BAS)

The BIS/BAS is a 20-item self-report measure that assesses trait negative affect and appetitive motivation (Carver and White, 1994). The BAS has three subscales: Reward-Responsiveness (BAS-RR), Drive (BAS-D), and Fun-Seeking (BAS-FS). The BAS-RR probed reward anticipation or receipt (e.g., “When I get something I want, I feel excited and energized”). The BAS-D assessed goal pursuit (e.g., “I go out of my way to get things I want”). BAS-FS measured impulsivity and novelty-seeking. Because BAS-FS is less relevant to reward (Taubitz et al., 2015), it was not included as a variable in the present study. The BIS measured punishment anticipation, sensitivity to anxiety-provoking circumstances (e.g., “Criticism or scolding hurts me quite a bit”), and conflict generation and resolution (McNaughton and Gray, 2000). Items are rated on a Likert scale (1 = strongly disagree; 4 = strongly agree). The BIS/BAS has appropriate divergent and convergent validity, test-retest correlations ranging from 0.59 to 0.69 (Carver and White, 1994), high internal consistency, moderate intercorrelation of the BAS subscales, and high long-term reliability in assessing stable characteristics in a depressed sample (Kasch et al., 2002). The present sample had adequate internal consistency (Cronbach's α = .75) and moderately low correlation of BAS-RR and BAS-D (r = .30).

Resting-State Functional MRI Connectivity

fMRI acquisition

Functional connectivity data were acquired during an eight-minute eyes-open resting state scan at 3.0 T. At UM, scans were acquired on a GE Signa scanner using a T2* weighted single shot reverse spiral sequence (29 4-mm thick slices, TE = 30 ms; TR = 2000 ms; 240 TRs total; flip angle = 90°; FOV = 20; matrix = 64 × 64). At UIC, scans were acquired on a GE Discovery scanner using parallel imaging with ASSET and T2* gradient-echo axial echo planar imaging sequence (44 3-mm thick slices, TE = 22.2 ms; TR = 2000 ms; 240 TRs total; flip angle = 90°; FOV = 22; matrix = 64 × 64). High-resolution T1 anatomic scans were obtained for spatial normalization. Participants were instructed to keep their eyes open and fixate on a cross on the screen.

fMRI Preprocessing

Images underwent slice-timing corrections with SPM8 (http://www.fil.ion.ucl.ac.uk/spm/doc/) and motion detection algorithms with FSL (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/). During pre-processing, images were visually inspected and subjects were removed if they had motion greater than 1.5 mm across more than three TRs. Structural and functional images were co-registered and then the co-registered T1-SPGR underwent spatial normalization (DARTEL to MNI template). The resulting normalization matrix was then applied to the slice-time-corrected, movement corrected, time series data and smoothed with a 5 mm Gaussian kernel. Resulting T2* images were 2 mm on a side with isotropic voxels.

Cross-Correlation Analysis

Time series was detrended and mean centered. Physiologic correction was performed by regressing out mean signal from white matter and cerebral spinal fluid (Behzadi et al., 2007). Motion parameters and deviations in x, y, and z translation were regressed out within first level models (Jo et al., 2013). Global signal was not regressed due to colinearity violations with gray matter signal, problematic misestimates and introductions of anticorrelations (Fox et al., 2013), and effect on distance-micromovement relationships (Jo et al., 2013). Finally, time-series were bandpass filtered over 0.01–0.10 Hz. Seeds of interest were the left VSs, right VSs, left VSi, and right VSi, using the following coordinates: -10, 15, 0; 10, 15, 0; -9, 9, -8; and 9, 9, -8, respectively (Di Martino et al., 2008; Figure 1). These four spherical ROIs comprised 19 voxels (6×6×6 mm sphere) and were validated in MNI space. Spatially averaged time course data were modeled from each ROI in cross correlation analyses. Correlation coefficients between mean time course for the four seed regions and all other voxels of the brain were calculated, producing a three-dimensional correlation coefficient r image. R images were transformed to z-scores using a Fisher transformation, and the resulting z maps were used in independent samples t tests conducted in SPM8. In MarsBaR (Brett et al., 2002), mean z values were extracted from each cluster that differed significantly between groups or correlated with BIS/BAS scores, separately for each seed and participant.

Figure 1.

Network connectivity of the VS seeds across groups. The left inferior ventral striatum (A), left superior VS (B), and right superior VS (C) seeds are shown, t = 10. Right inferior VS connectivity was similar to left inferior VS and is not shown. The seeds are highlighted in yellow and each panel shows the extent of the network.

Statistical Analyses

Linear regressions with multiple predictors were computed for each of the four seeds using SPM8. For the primary aim, predictors were BAS-RR, BAS-D, BIS. For the secondary aim, diagnostic group was the predictor. Analyses for both aims included sex and site as covariates. Z values from regions that were temporally correlated with the seed region were the dependent variables. Contrasts were made for the predictor of interest, with all other predictors included as covariates. For example, the contrast with BAS-RR as the predictor of interest included BAS-D, BIS, diagnostic group, sex, and site as covariates. Whole brain correction was achieved for each multiple regression model at p = .01 by conducting 1000 Monte Carlo simulations in 3dClustSim to determine a joint threshold of height and extent (p < .005, extent of 464 mm³). The Monte Carlo approach was intended to balance Type I and Type II error and cluster extent was determined using the bug-fixed 3dClustSim tool (Cox et al., 2016), with each of the sixteen regression models having a corrected p < .01, the FWE = .16. Post-hoc, a Bonferroni correction was applied to achieve FWE < .05; clusters of activation that did not survive Bonferroni correction are noted in the tables reporting results of the regression models (Tables 2 and 3). Two QC post-hoc motion correction strategies were employed. First, the standard deviations of movement along the x, y, and z axes were checked for significant relationships with the predictors and extracted cluster values. These movement parameters were not significantly correlated with the affective trait measures (p's > .05) and did not significantly differ between groups (t = .50, t = .65, t = -.61 for deviations in x, y, and z translation, respectively, all p's > .05). Second, framewise displacement, defined as the sum of the absolute values of the six realignment parameters' derivatives (Power et al., 2012), was calculated for each participant. By providing a measure of the degree to which head position has changed from one frame to the next, framewise displacement (Power et al., 2012) represents a conservative index of data quality. Framewise displacement did not differ between groups (t = 1.77, p = .08; Supplemental Figure 1) and was not associated with the predictor variables BIS, BAS-RR, or BAS-D (all p's > .05).

Table 2. Functional connectivity clusters relating to predictors and group – VSi seeds.

Seed	Contrast	Lobe	Gyrus	BA	MNI coordinatesa			Z	mm3
					x	y	z
LVSi	BIS negative	Frontal	Superior frontal	-	-18	54	14	3.94	624
			Precentral	6	52	-12	36	3.16	560B
		Parietal	^ Postcentral	3	-32	-34	56	3.41	728
		Temporal	^ ○ Superior temporal	38	-38	12	-40	3.92	504B
	BAS-RR positive	Frontal	Mid/posterior cingulate	24/31/5	6	-12	36	3.74	2480
			^ Superior frontal	6	12	-2	70	3.18	688
		Occipital	Lingual	17/18/1 9	-14	-94	20	3.8	3968
		Parietal	Inferior parietal	40/5	48	-18	32	3.49	800
			^ Postcentral	5/3	-12	-46	64	3.76	4152
				5/3	20	-44	70	3.75	1480
			^ Postcentral/supramarginal	43/39	56	-14	20	3.38	728
		Subcortical	Medial dorsal nucleus/pulvinar	-	8	-20	8	4.29	784
			Uvula	-	22	-84	-36	3.37	816
	HC greater than rMDD	Temporal	Inferior temporal	20	52	-36	-24	3.25	760
RVSi	BIS negative	Temporal	Middle temporal	38	-40	14	-40	3.49	576B
	BAS-RR positive	Frontal	Precentral	4	48	-16	34	3.43	568B
		Occipital	^ Inferior occipital	19/18	-44	-80	-12	4.37	1392
			^ Lingual	18/17/1 9	-8	-98	-14	5.11	3480
		Parietal	Postcentral	43	60	-10	12	3.18	592B
			Postcentral, paracentral	40/5/4/ 3/7	28	-46	62	3.65	4088
			Precuneus	19/7	28	-82	40	3.81	880
			^ Superior parietal	7	-32	-40	50	3.81	2272
		Temporal	^ Inferior temporal, fusiform	37/20	52	-40	-22	3.31	888
			Inferior temporal, middle occipital	37/19/1 8	46	-72	-14	4.64	5320
			Middle temporal	39/19	-52	-66	14	3.24	904
		Subcortical	Posterior lobe, pyramis	-	6	-80	-38	3.56	592B
			Posterior lobe, uvula, declive	-	18	-84	-34	3.7	1016
	BAS-D negative	Parietal	Inferior parietal	40	-40	-48	60	4.49	752
	HC greater than rMDD	Temporal	Inferior temporal, fusiform	37/20	52	-40	-22	3.99	1376

HC, Healthy control; rMDD, remitted major depressive disorder; BA, Brodmann area.

^ax, y, z = MNI (Montreal Neurological Institute) coordinates of significant effects.

^{^}Cluster significantly correlated with movement along the y axis.

^○Cluster significantly correlated with movement along the z axis.

^BCluster not significant after Bonferroni correction.

Table 3. Functional connectivity clusters relating to predictors and group – VSs seeds.

Seed	Contrast	Lobe	Gyrus	BA	MNI coordinatesa			Z	mm3
					x	y	z
LVSs	BIS negative	Frontal	Dorsal anterior cingulate	32/24	2	22	30	4.13	1352
		Subcortical	Caudate head	-	8	10	2	4	544B
	BAS-RR positive	Subcortical	Pons	-	4	-30	-32	4.83	608
		Temporal	*^ Parahippocampal	36/35	-14	-34	-10	3.89	488B
	BAS-D positive	Frontal	Precentral	13/6/44	-48	2	10	4.28	1544
		Parietal	*Cingulate	23/31	2	-36	26	3.2	888
			Inferior parietal	19/7/39	-38	-68	42	3.26	624
			Superior parietal	7	24	-80	48	3.09	936
			*○ Posterior lobe, declive	-	18	-86	-38	3.23	464B
		Temporal	Fusiform/parahippocampal	20/37	-40	-14	-30	3.9	1424
			^ Inferior temporal	20/21	-68	-24	-14	4.53	736
			^	20	62	-40	-20	59	472B
	HC greater than rMDD	Temporal	○ Putamen, amygdala, inferior frontal	47	28	2	-16	3.72	2936
RVSs	BIS negative	Frontal	Anterior cingulate	32	4	22	32	3.49	848
			○ Frontal inferior operculum, precentral	22/44	52	8	4	4.02	3600
		Occipital	Lingual	18/17	-22	-96	-12	3.82	824
		Parietal	^ Supramarginal	40	-52	-38	28	3.82	504B
		Temporal	^ Superior temporal	22	48	18	-6	3.56	552B
			○ Superior temporal, insula	22/1 3	-46	10	-4	3.95	4856
	BAS-RR positive	Frontal	Anterior cingulate	32	0	22	-10	3.56	496B
		Subcortical	Pons	-	4	-30	-34	4.2	632
			* Pulvinar	-	10	-26	6	3.29	560B
	BAS-D positive	Frontal	○ Dorsolateral prefrontal cortex	46	-34	44	0	3.79	488B
			○ ^ Middle frontal	44	-44	8	12	3.85	1768
		Parietal	^ ○ Precuneus	7	16	-66	54	3.77	760
	BAS-D negative	Temporal	Superior temporal/insula	22/13	36	12	-26	3.59	480B
	HC greater than rMDD	Subcortical	○ Parahippocampal, lentiform nucleus, amygdala	34	-24	-2	-18	3.84	2824
			*○ Parahippocampal, lentiform nucleus, amygdala	34	28	0	-12	3.97	4840
		Temporal	Inferior temporal	20	62	-26	-28	3.66	728
			Superior temporal	40/42	66	-26	12	3.33	1112
	HC less than rMDD	Frontal	Superior frontal	8	12	28	46	3.48	520B
		Occipit al	* ^ Lingual	18	-26	-88	-10	3.87	1136

HC, Healthy control; rMDD, remitted major depressive disorder; BA, Brodmann area.

^ax, y, z = MNI (Montreal Neurological Institute) coordinates of significant effect.

^*Cluster significantly correlated with movement along the x axis.

^{^}Cluster significantly correlated with movement along the y axis.

^○Cluster significantly correlated with movement along the z axis.

^BCluster not significant after Bonferroni correction.

Results

VSi Results Summary

BAS-RR was correlated with VSi connectivity to many regions, including the inferior temporal, middle temporal, and fusiform gyri, as well as parietal clusters such as the precuneus, postcentral, paracentral, and superior parietal gyri (Figure 2, Panels A1 and A2). The VSi seeds related to few connectivity differences between groups (Figure 2, Panels B1 and B2).

Figure 2.

VSi connectivity relating to affective traits and diagnostic group. Several regions of connectivity to the LVSi (Panel A1) and RVSi (Panel A2) were positively correlated with BAS-RR. HCs had greater connectivity than rMDDs from the LVSi (Panel B1) and RVSi (Panel B2). All images display using threshold of p < .005 and k > 55.

Left Ventral Striatum – Inferior (LVSi)

Table 2 reports regions of connectivity from the LVSi seed in relation to BIS/BAS scales and diagnosis.

Affective Traits

LVSi connectivity to the superior frontal gyrus (L), precentral gyrus (R), postcentral gyrus (L), and superior temporal gyrus (L) was negatively correlated with BIS. BAS-RR was positively correlated with LVSi connectivity to the posterior cingulate (R), superior frontal gyrus (R), lingual gyrus (L), inferior parietal gyrus (R), postcentral gyrus (B), supramarginal gyrus (R), thalamus (R), and uvula (R).

Diagnosis

Relative to rMDDs, HCs showed greater connectivity from LVSi to the right inferior temporal gyrus.

Right Ventral Striatum – Inferior (RVSi)

Table 2 reports regions of connectivity from the RVSi seed in relation to BIS/BAS scales and diagnosis.

Affective Traits

RVSi connectivity was positively correlated with BAS-RR, in regions including the precentral gyrus (R), inferior occipital gyrus (L), lingual gyrus (L), postcentral gyrus (R), paracentral gyrus (R), precuneus (R), superior parietal gyrus (L), fusiform gyrus (R), middle temporal gyrus (L), inferior temporal gyrus (R), pyramis (R), and uvula/declive (R). RVSi connectivity to the left middle temporal gyrus was negatively correlated with BIS. RVSi connectivity to the left inferior parietal gyrus was negatively correlated with BAS-D.

Diagnosis

HCs had greater connectivity from RVSi to the right inferior temporal and fusiform gyri relative to individuals with rMDD.

VSs Results Summary

BAS-D was correlated with VSs connectivity to several frontal, temporal, and parietal regions, including the middle frontal, cingulate, precuneus, fusiform, parahippocampal, and inferior temporal gyri (Figure 3, Panels A1 and A2). BIS was negatively correlated with connectivity from the VSs to the ACC, superior temporal gyrus, and insula (Figure 3, Panels B1 and B2). Groups differed in connectivity from the VSs seeds to temporal and subcortical regions such as the inferior and superior temporal gyri, putamen, and amygdala (Figure 3, Panels C1 and C2).

Figure 3.

VSs connectivity relating to affective traits and diagnostic group. Several regions of connectivity with the LVSs (Panel A1) and RVSs (Panel A2) were positively correlated with BAS-D. Several regions of connectivity with the LVSs (Panel B1) and RVSs (Panel B2) were negatively correlated with BIS. HCs had greater connectivity than rMDDs from the LVSs (Panel C1) and RVSs (Panel C2). Panel D displays the overlap (light blue) between connectivity regions that negatively related to BIS (dark blue) and were greater in HCs than rMDDs (green). All images display using a threshold of p < .005 and k > 55.

Left Ventral Striatum – Superior (LVSs)

Table 3 reports regions of connectivity from the LVSs seed in relation to BIS/BAS scales and diagnosis.

Affective Traits

LVSs connectivity to the right dorsal ACC and right caudate head was negatively correlated with BIS. Positive LVSs connectivity correlations included BAS-RR with the pons (R) and parahippocampal gyrus (L), and BAS-D with the precentral gyrus (L), cingulate (R), inferior (L) and superior (R) parietal gyri, declive (R), fusiform and parahippocampal gyri (L), and inferior temporal gyrus (B).

Diagnosis

Compared to rMDDs, HCs exhibited greater LVSs connectivity to the putamen (R), amygdala (R), and inferior frontal gyrus (R).

Right Ventral Striatum – Superior (RVSs)

Table 3 reports regions of connectivity from the RVSs seed in relation to BIS/BAS scales and diagnosis.

Affective Traits

RVSs connectivity to the inferior frontal gyrus (R), ACC (R), frontal inferior operculum/precentral gyrus (R), lingual gyrus (L), supramarginal gyrus (L), superior temporal gyrus (B), and insula (L) was negatively correlated with BIS. RVSs connectivity to the ACC (B), pons (R), and pulvinar (R) was positively correlated with BAS-RR. BAS-D was positively correlated with RVSs connectivity to the DLPFC (L), middle frontal gyrus (L), and precuneus (R). BAS-D was negatively correlated with RVSs connectivity to the right superior temporal gyrus and insula.

Diagnosis

HCs had greater connectivity than rMDDs from RVSs to, the parahippocampal gyrus (B), lentiform nucleus (B), amygdala (B), and inferior and superior temporal gyri (R). HCs showed less connectivity than rMDDs from RVSs to the right superior frontal gyrus and left lingual gyrus.

For all four seeds, effects relating to sex differences and number of depressive episodes are presented in Supplemental Table 1 and the Supplemental Results section.

Discussion

The present study investigated affective personality correlates of reward network connectivity and differences in reward network connectivity between rMDD and healthy individuals. BAS-RR was positively correlated with connectivity between the VSi and bilateral posterior visual and dorsal somatosensory regions, areas primarily involved in perception of (Zeki et al., 1991) and response to (Nii et al., 1996) the environment. These results indirectly corroborate previous findings that striatal dopamine facilitates sensory-motor interaction and responsivity to stimuli (Salamone, 1992). In rMDD individuals, VSi connectivity might confer higher trait reward-responsiveness, enabling more intact reward responding. With the exception of connectivity observed between the LVSi and mid/posterior cingulate, our findings mostly contrasted previous research suggesting that the VSi is highly connected with emotional processing regions (Di Martino et al., 2008). This difference could be due to our examination of VSi connectivity specifically in relation to BAS-RR.

Decreased integration of the VSi, visual cortex, and somatomotor regions may be related to lower reward-responsiveness and poorer performance on reward tasks. Lack of VSi-visual-somatomotor integration could lessen an individual's interest or willingness to interact with the environment, resulting in lower reward-responsiveness, decreased performance on reward-seeking tasks (e.g. less money earned; DelDonno et al., 2015), and worse depressive symptoms. VSi-visual-somatomotor connectivity may deteriorate over the course of illness and could partially explain why depressed individuals have significantly lower BAS-RR scores (DelDonno et al., 2015) than early-course rMDD individuals. It is possible that VSi-visual-somatomotor network connectivity could predict MDD diagnosis, illness severity, or likelihood of relapse, just as BAS scores are able to do (Kasch et al., 2002; Shankman & Klein, 2003; McFarland et al., 2006). For example, disease status in a sample of individuals with and without MDD was predicted at 95% accuracy using the VSi and 14 other MDD-relevant ROIs extracted from rsFC data (Craddock et al., 2009). Furthermore, although BAS-RR scores did not differ between rMDD individuals and HCs in the present study, the clinically relevant predictive power of BAS scores (Kasch et al., 2002; McFarland et al., 2006) suggests that these traits are vulnerability factors for MDD course and that the present rsFC results perhaps reflect a mechanism of this effect.

Additionally, we found that BAS-D was related to connectivity from the VSs to the precuneus and precentral regions. BAS-D has been associated with behavior that maximizes the likelihood and amount of reward (Scheres & Sanfey, 2006) and with brain activation to pictures of appetizing food (Beaver et al., 2006). These associations demonstrate that trait drive is important for pursuing and “wanting” reward, which are frequently disrupted in depression (Treadway & Zald, 2011). The VS, although not VSs specifically, has been extensively linked to drive and motivation (e.g. Nestler & Carlezon, 2006;). Previous research suggests that the VSi/NAcc is not required for goal-directed behavior or sensitivity to response-contingent rewards (Cardinal et al., 2002). Instead, the VSs, not often differentiated from VSi in the literature, may be responsible for goal-directed behavior and motivation. This potential VSs function may be why BAS-D related to VSs but not VSi connectivity, and may also explain a directional link to more frontal regions. The VSs may interact with the precuneus, which is important for self-referential processes such as the experience of agency (Cavanna & Trimble), to evaluate or generate desires. Then, desires may be translated into action via connections between the VSs and precentral gyrus, which is responsible for motor performance (Porro et al., 1996) such as pressing a response button during a task.

BIS negatively correlated with connectivity between the LVSs and dorsal ACC (dACC), a region that integrates negative affect and cognitive control (Shackman et al., 2011). Connectivity between the VSs and dACC may be part of the neural basis of behavioral inhibition and negative affectivity, especially in those with rMDD. Having less connectivity between reward and salience regions may help individuals suppress negative affect. Voluntary inhibition of negative affect has previously been associated with the activation of the dACC and the ventrolateral PFC (BA 44) and the deactivation of the NAcc (Phan et al., 2005). Our observation that decreasing connectivity from the VS to the dACC and BA 44 correlated with trait BIS suggests that less VSs-SEN connectivity may contribute to reduced distress sensitivity, fear, and anxious arousal (McNaughton & Gray, 2000).

Although groups did not differ in VSs-DMN connectivity, the VSs was hyperconnected to such SEN regions as the amygdala, putamen, and parahippocampal gyrus in HCs compared to individuals with rMDD. Although this finding ran contrary to our hypothesis, a recent meta-analysis noted that there is mixed evidence as to whether individuals with MDD exhibit greater or less connectivity from the SEN to other regions (Kaiser et al., 2015). Kaiser and colleagues suggested that connectivity between salience, attention, and emotion processing regions might be modulated by the presence of external cues that correspond with internal mentation (2015). In late-life depression, decreased rsFC was observed between the VSi and amygdala, caudate, putamen, globus pallidus, and thalamus (Alexopolous et al., 2013), a pattern of results that mirrors the present group-level VSs-SEN findings with the caveat that Alexopolous and colleagues seeded from the VSi. Somewhat consistent with our findings, a previous study found that VSs activity in HCs was correlated with the superior and lateral OFC, regions in the SEN (Di Martino et al., 2008). Although we observed hyperconnectivity between the VSs and SEN in HCs relative to individuals with rMDD, the OFC was not included. Lack of VSs connectivity to the OFC could be due to our inclusion of rMDD individuals in the sample as opposed to an entirely healthy sample. Additionally, our results contrast previous reports that VSs activity predicts activation in executive function regions such as the DLPFC (Di Martino et al., 2008). Overall, this collection of mixed findings highlights the need to further clarify the role of the SEN in MDD.

We observed a few regions of overlap where HCs showed greater connectivity between the VSs and SEN than in individuals with rMDD and where the BIS related to decreasing connectivity between the RVSs and SEN (Figure 3, Panel D). This overlap could represent a neural resistance against negative affect via greater connectivity between the VSs and SEN. Put another way, it is possible that the degradation of VSs-SEN connectivity could represent a vulnerability or scar from illness. A version of the scar hypothesis of depression (Lewinsohn, 1981) has been put forth that suggests scars develop proportionally before and after MDD episodes as a result of subthreshold or residual symptoms (Wichers et al., 2013). Whereas repetitive scar effects would be minimal because participants with rMDD were recruited early in their course of illness and had experienced only a few episodes, subthreshold symptoms, high trait negative affect, or neural vulnerability could account for rsFC differences compared to HCs.

There are several limitations of the present study. First, the study involved scanning at two sites. Although the presence of significant effects across sites indicate generalizability to multiple geographical regions and scanners, a number of site-specific effects emerged (Supplemental Figure 2) because the scanning protocols were not prospectively synchronized, contributing additional error and measurement variance. On the other hand, this renders effects across sites more conservative. Second, although each regression was corrected for whole-brain significance and FWE was reduced with cluster extent thresholding (Cox et al., 2016; Eklund et al., 2016), there is still a possibility that performing 16 group-level analyses may have resulted in false positives. Third, we did not find a difference in BIS/BAS scores between groups, weakening the argument that is evident within the literature that these would be trait risk factors for MDD. The high-functioning nature of our sample (young adults in remitted state with above-average IQs) may have obscured trait differences. As such, we utilized the RDoC dimensional framework in addition to group differences to better understand both disease and individual difference features related to VS connectivity.

Future research could examine associations between VS networks and performance-based measures of reward responsiveness, such as a monetary incentive delay task. Additionally, examining changes in VS connectivity over time in relation to affective personality traits and reward task performance could help disentangle whether these phenomena are reliable and state- or trait-like. Continuing to distinguish between the VSs and VSi might further elucidate functional specificity of the two halves of the VS. Differences in their functions could also illuminate mechanisms of and targets for existing or future treatments.

Overall, the present findings suggest that VSs and VSi networks are related to affective personality traits that have previously been demonstrated to differentiate disease status (DelDonno et al., 2015) and predict MDD risk (Kasch et al., 2002). Previously, response to rewarding stimuli has been linked to BAS-RR (Kasch et al., 2002) and VSi activation in active MDD (Epstein et al., 2006). In this study, we found a relationship between BAS-RR and the VSi across rMDD and HC groups. In contrast, group differences emerged from VSs connectivity to subcortical SEN regions. Together these findings suggest that VSi functioning may be relevant to reward deficits in active MDD whereas decreased VSs-SEN connectivity may be a marker of remission. VS connectivity may constitute an intermediate phenotype of MDD, perhaps underlying risk for onset or relapse. These neural networks may represent reliable markers that could fit within the positive valence system of the NIMH Research Domain Criteria matrix. In sum, history of MDD, BIS, and BAS-RR each contribute specific and meaningful information about how VS connectivity and other networks may relate to MDD risk or expression.

Supplemental Results

Sex Differences

Several sex differences emerged across seeds (Supplemental Table 1). Females showed greater LVSi connectivity to the inferior frontal (L), middle frontal (R), and postcentral gyri (B) but less connectivity from LVSi to the middle frontal gyrus (L), middle occipital gyrus (L), and precuneus (L) than males. Females showed greater connectivity from RVSi than males to several right subcortical regions, including the uvula, caudate, lentiform nucleus, and insula.

Females had less LVSs connectivity than males to the inferior frontal (L), middle occipital (B), middle temporal (R), and superior temporal (R) gyri. Females had greater connectivity than males from RVSs to the left uncus.

Single vs. Multiple Depressive Episodes

Compared to individuals with more than one previous depressive episode, those with a single previous depressive episode had higher connectivity between the RVSs and the left lingual gyrus, which was negatively correlated with BIS, F(1, 38) = 4.44, p < .05. No other regions of connectivity or connectivity correlations with the BIS/BAS scales were related to number of episodes.

Highlights.

• History of depression related to altered ventral striatum connectivity.

• Trait positive and negative affect related to connectivity.

• Superior and inferior ventral striatum showed different connectivity patterns.