Ventral striatal subregional dysfunction in late-life grief: Relationships with yearning and depressive symptoms

Abstract

Bereaved older adults experiencing high grief in the first year after an attachment loss is at increased risk for prolonged grief disorder (PGD) via unknown mechanisms. Yearning, a core grief symptom, is linked to the ventral striatal (VS) brain function, but the role of this neuronal system in late-life grief is poorly understood. As a first step, we examined the VS subregional abnormalities associated with multidimensional symptoms in bereaved elders during the first year post-loss. Sixty-five bereaved elders completed clinical assessments within 13 months post-loss. Ventral caudate (VCau) and nucleus accumbens (NAcc) functional connectivity (FC) was assessed using seed-based resting-state functional MRI. VCau and NAcc FC differences between high (inventory of complicated grief [ICG] score≥30; n = 35) and low (ICG score<30; n = 30) grief, and the relationships between ventral striatal subregional FC and clinical measures (yearning and depressive symptoms) were assessed after covariate adjustments (α < 0.05; 3dClustSim corrected). Relative to low grief participants, those with high grief showed higher FC between VCau and the medial prefrontal, orbitofrontal, and subgenual cingulate cortices. VCau FC abnormalities positively correlated with yearning (r² = 0.24, p < 0.001). In contrast, FC between VCau and temporoparietal junction negatively correlated with depressive symptoms, a commonly co-occurring symptom (r² = 0.37, p < 0.001). The FC between NAcc and insula/striatum positively correlated with yearning (r² = 0.35, p < 0.001); no other NAcc FC findings were seen in the full sample. In women, higher FC between the NAcc and bilateral posterior cingulate, precuneus, and visual areas were found in those with high, relative to low grief symptoms. Distinct VS subregional abnormalities associate with yearning and depressive symptoms in bereaved elders. Whether ventral striatal dysfunction correlates with PGD development and/or worsening depression remains to be elucidated.

1. Introduction

Older adults experience the death of loved ones at a higher frequency than their younger counterparts, and some form of grief is a universal response. Most bereaved older adults adapt naturally to the loss and recover to their pre-loss functioning within one year. However, in about 10–15% of older adults, grief follows a protracted and disabling course, leading to the development of prolonged grief disorder (PGD), a clinical condition with enormous adverse consequences (Shear, 2015; Simon, 2013). Although bereaved older adults at risk for pathological grief trajectories cannot be predicted at this time, emerging evidence suggests that some who experience intense symptoms during the first year post-loss could be at higher risk (Bonanno and Malgaroli, 2020; Djelantik et al., 2017). Understanding the neurobiological basis that underlies the varying intensity of early grief is an important first step to further delineate, in future studies, the neural predictors of heterogeneous grief trajectories and incident late-life PGD.

One prevailing model of PGD is based on the attachment theory (Bowlby, 1980). Close bonds developed with attachment figures play an important physiological and behavioral regulatory role in one’s life, leading to a reduction in emotional distress and the generation of pleasure. The loss of the rewarding aspects of attachment that occurs during bereavement may explain the early grief response (Hofer, 1996; Shear and Shair, 2005). From a neurobiological perspective, attachment is linked to the activity in the brain network that mediates reward (Insel, 2003). Extensive animal and human studies have identified the ventral striatum (VS) as a central hub of the reward processing brain network (Haber and Knutson, 2010; Suckling and Nestor, 2017; Wang et al., 2016; Wise, 1998).

It is hypothesized that persistent dysfunction in the reward processing brain network underlies PGD and one of its core clinical features, the intense yearning for the deceased (Kakarala et al., 2020; O’Connor, 2012). In support of this notion, greater functional activation in the VS, particularly in the nucleus accumbens (NAcc), to reminders of the deceased is reported in PGD, relative to integrated grief (O’Connor et al., 2008). Higher neural activity in this key reward processing region positively correlated with yearning across PGD and integrated grief participants (O’Connor et al., 2008). These findings, however, are not universal, and they appear to diverge based on the age of the PGD sample (McConnell et al., 2018). In contrast, data points to activity in widespread reward-related brain regions in response to grief eliciting stimuli in participants with PGD (McConnell et al., 2018; O’Connor et al., 2008) as well as in acutely grieving younger adults (Gundel et al., 2003; Kersting et al., 2009). These investigations focused on task-dependent localized activity in brain areas of the reward neural system and were not equipped to provide a task-free network-based perspective of the grieving brain. The role of the ventral striatal network functioning in explaining the symptom variance seen in bereaved older adults during the first year post-loss is unexplored.

To address this knowledge gap, the current study used the seed-based resting-state functional magnetic resonance imaging (rs-fMRI) method to map the intrinsic ventral striatal brain network architecture in bereaved older adults. Given that the VS encompasses cytoarchitecturally similar but functionally distinct anatomical subregions (Contreras-Rodriguez et al., 2017; Di Martino et al., 2008; Haber and Knutson, 2010), we performed functional connectivity (FC) analyses by focusing on 2 seeds, namely the ventral caudate (VCau) and the NAcc. First, we examined the VCau and NAcc FC differences between high and low grief symptoms among bereaved older adults within the first year post-loss. Second, we examined the relationship between these VS subregions and yearning as well as global grief severity. Since participants with high grief symptoms during the first year of bereavement are at high risk for developing PGD, we hypothesized that this group compared with participants with low grief would show increased FC within the VS subregions. We also posited that VS subregional FC would positively associate with yearning. Given the increased comorbidity of depression during the first year of bereavement(Zisook et al., 2014), we further explored the correlations of the VCau and NAcc FC with depressive symptoms.

2. Methods and materials

2.1. Study participants

In this pilot study, a total of 66 bereaved participants, aged 50 years and older (range: 51–87 years), completed cross-sectional clinical assessments and an MRI scan within 13 months of bereavement. One participant was excluded during MRI processing (see below), and thus, the final sample included 65 bereaved participants. We chose 13 months instead of 12 months post-loss to avoid potential transient symptom increases that may occur around the death anniversary. Two of the 66 grief participants completed the assessments and MRI over 12 months following the death of a loved one. The grief participants were enrolled from the community via advertisements, and through referrals from grief groups and hospice counselors. All participants provided written informed consent according to the Medical College of Wisconsin Institutional Review Board-approved protocols.

All grief participants completed clinical assessments, including the sociodemographic characteristics, medical and psychiatric histories, a medication history, a neurological examination, and the Structured Clinical Interview for DSM-5 Research Version. The 19-item Inventory of Complicated Grief (ICG) scale (Prigerson et al., 1995), 17-item Hamilton Depression Rating Scale (HAM-D), Hamilton Anxiety Scale (HAM-A) (Hamilton, 1959), Scale of Suicidal Ideation, Cumulative Illness Rating Scale-Geriatric version (CIRS-G)(Miller et al., 1992), modified Hachinski Ischemic Scale, Mini-Mental State Exam (MMSE), and Mattis Dementia Rating Scale-2 (DRS-2) (Lucas et al., 1998) were also completed. The relationship of the deceased, time since the loss in days (TSL), and psychotropic medication history were documented.

2.2. Inclusion and exclusion criteria

The inclusion and exclusion criteria have been published previously (Chen et al., 2020; Harfmann et al., 2020; Kang et al., 2021). All participants had to score <4 on the modified Hachinski Ischemic Scale, and score ≥24 on the Mini-Mental State Exam. Exclusion criteria included a lifetime history of bipolar or psychotic disorders; alcohol or substance use disorders during the past five years; acute suicidality (assessed using the Scale of Suicidal Ideation and the third HAM-D item score or judged by a clinician); a history of neurological illnesses, including seizures, stroke, dementia of any etiology, severe head injury, brain tumor, etc.; delirium/unstable medical conditions determined using the Cumulative Illness Rating Scale-Geriatric version score of 4 in any category; and MRI contraindications.

Grief participants who were diagnosed with a bereavement-related psychiatric disorder were not excluded. A bereavement-related psychiatric disorder was diagnosed if (1) the Diagnostic and Statistical Manual of Mental Disorder, 5th edition (DSM-5), criteria for depressive, anxiety, or post-traumatic stress disorders were met on the Structured Clinical Interview for DSM-5 Research Version, and (2) the onset of the episode occurred following bereavement (please refer to the results section and Table 1 for additional details). Antidepressant medications and/or low doses of benzodiazepines were allowed if doses remained stable for at least 4 weeks prior to the brain MRI scan.

Table 1.

Demographics and clinical characteristics.

	High grief (n = 35)	Low grief (n = 30)	p-value
Age (years)	64.8 ± 7.6	67.4 ± 10.3	p = 0.268a
Gender (female, n)	26	19	p = 0.340b
Race (white/black/am Indian, n)	32/3/0	27/2/1	p = 0.816c
Years of education	15.2 ± 2.9	15.9 ± 3.5	p = 0.395a
Time since loss (days)	174.0 ± 98.5 (range: 40–384)	137.1 ± 81.5 (range: 19–334)	p = 0.104a
Relationship with the deceased person (spouse/parent/child or grandchild/other, n)	13/6/10/6	14/10/2/4	p = 0.124c
Cognitive/medical screening
MMSE	28.3 ± 1.4	28.6 ± 1.5	p = 0.395a
CIRS-G	6.2 ± 4.2	5.4 ± 3.1	p = 0.394a
DRS total	139.8 ± 3.5	140.7 ± 2.9	p = 0.253a
Psychiatric measures
ICG	39.9 ± 10.7	13.3 ± 3.0	p < 0.001 a
Yearning	13.5 ± 3.2	5.6 ± 3.1	p < 0.001 a
HAM-D	17.1 ± 6.4	7.9 ± 6.2	p < 0.001 a
HAM-A	9.8 ± 4.8	5.7 ± 5.0	p < 0.001 a
SCID Depressive Disorder positive (n)d	27	4	p < 0.001 b
Antidepressant use
None (n)	21	26	n/a
SSRI monotherapy (n)	7	3	n/a
SNRI (n)	2	0	n/a
Combination antidepressant therapy (n)	5	1	n/a

Abbreviations. MMSE: Mini-Mental State Exam; CIRS-G: Cumulative Illness Rating Scale-Geriatric; DRS: The Dementia Rating Scale; ICG: Inventory of Complicated Grief; HAM-D: 17-item Hamilton Depression Rating Scale; HAM-A: Hamilton Anxiety Rating Scale; SCID: The Structured Clinical Interview for DSM-5; SSRI: selective serotonin reuptake inhibitor; SNRI: serotonin norepinephrine reuptake inhibitor; n/a: not applicable.

^a2-sample t-test.

^bChi-square test, and.

^cFischer’s Exact Test.

^dDepressive disorder only (high grief: n = 22; low grief: n = 3); Depressive disorder with anxiety (high grief: n = 3; low grief: n = 1); Depressive disorder with PTSD/other (high grief: n = 2; low grief: n = 0).

2.3. Experimental procedures

Clinical measures.

ICG is a 19-item, self-report questionnaire that has excellent psychometric properties and can reliably assess PGD symptoms (Prigerson et al., 1995). ICG has been previously utilized in multiple PGD studies (Shear et al., 2005, 2016). The total ICG scores range from 0 to 76; a score ≥30 is indicative of PGD if the death of a loved one was at least 12 months prior. For this study, the high grief group was defined by an ICG score ≥30, and the low grief group by an ICG score <30.

Factor analyses of the ICG on diverse bereaved samples have been conducted in multiple studies (Fisher et al., 2017; Simon et al., 2011). The first symptom cluster that is identified, yearning and preoccupation with the deceased, include five ICG items (see supplemental information for details). The scores for the yearning cluster range from 0 to 20, with higher scores indicating higher yearning symptoms. Yearning was the primary outcome variable of interest.

The 17-item HAM-D is a widely used, validated, clinician-administered instrument that measures depressive symptoms; each item score is summed to generate a total score from 0 to 52 (Hamilton, 1980). Total ICG and HAM-D scores were used as secondary clinical measures of interest.

MRI acquisition.

An MRI scan was obtained on a 3 T GE Discovery MR750 (Waukesha, WI) scanner with a 32-channel head coil. High-resolution anatomical images were acquired using an axial, whole-brain T1-weighted 3-dimensional fast spoiled gradient echo sequence. The T1-weighted parameters were TR (repetition time) = 8 ms; TE (echo time) = 3 ms, flip angle = 12°, FOV (field of view) = 240 × 240 mm², voxel size = 0.94 × 0.94 × 1 mm³, continuous 150 axial slices, and a scan time of approximately 6 min. A whole-brain rs-fMRI scan was acquired in a sagittal view using a single-shot gradient recalled echo planar imaging (EPI) pulse sequence. The rs-fMRI scan parameters were TR = 2000 ms, TE = 25 ms, flip angle = 70°, FOV = 240 × 240 mm², voxel size = 3.75 × 3.75 × 4 mm³, number of slices = 36, and scan time = 8 min.

2.4. Magnetic resonance image preprocessing

Structural MRI.

The T1-weighted image was skull stripped using 3dSkullStrip (AFNI; www.afni.nimh.nih.gov). Then, the brain was resampled to 2 × 2 × 2 mm and was non-linearly transformed to Montreal Neurological Institute (MNI) space using Advanced Normalization Tools. FreeSurfer was used for tissue segmentation (Desikan et al., 2006), and MNI Glasser gray matter mask (without the cerebellum) was created (Glasser et al., 2016).

To calculate each participant’s voxelwise gray matter volume (GMV), voxel-based morphometry with MATLAB 9.9 (MathWorks, Natick, MA) and statistical parametric mapping (Ashburner and Friston, 2000; Good et al., 2001) were used (see supplementary material for details).

rs-fMRI.

rs-fMRI data were preprocessed using the fMRIPrep version 20.2.1 (Esteban et al., 2019). rs-fMRI preprocessing steps are similar to those used in prior functional MRI work and are detailed in the supplementary material (Aronson Fischell et al., 2020; Faskowitz et al., 2020; Finc et al., 2020; Linke et al., 2021).

One participant’s data was removed due to brain MRI segmentation problem caused by excessive motion. Therefore, a total of 65 bereaved participants were used in our analyses.

2.5. Seed-based resting-state functional connectivity

The seed-based rs-fMRI method was used to determine the voxelwise FC for individual participants within the GM mask. The VCau (MNI coordinates: ±10, 15, 10) and NAcc (MNI coordinates: ±9, 9, −8) (Contreras-Rodriguez et al., 2017; Di Martino et al., 2008; Karcher et al., 2019) were used as the 2 non-overlapping seed regions of interest (Fig. 1). Additionally, bilateral primary motor cortex (M1; MNI coordinates: left: −38 −24 62; right: 34 −22 62) (Pool et al., 2015) and bilateral white matter (MNI coordinates: ±24 40 4) were chosen as control seeds (see Supplementary Material Fig. S4). Five-millimeter spherical seed regions of interest were drawn at the above coordinates. The preprocessed average-voxel time course of each seed region of interest was cross-correlated with the whole brain using AFNI’s 3dDeconvolve. The Pearson correlation coefficients (cc) maps were subjected to a Fisher’s z-transformation, m = 0.5 × ln{(1+cc)/(1-cc)}, which yielded variants of approximately normal distribution.

Fig. 1.

Seed regions of interest with MNI coordinates. Abbreviations: MNI: Montreal Neurological Institute; VCau: ventral caudate; NAcc: nucleus accumbens.

2.6. Statistical analysis

To compare differences between the high and the low grief groups, 2-sample t-tests were used for demographics (except gender and race) and clinical characteristics (except relationship with the deceased), a chi-square test was used for gender differences, and Fisher’s exact tests were used for differences in race and relationship with the deceased.

The AFNI program 3dttest++ (student’s t-test) was used to determine voxelwise VCau and NAcc FC differences between high and low groups, after adjusting for age, gender, TSL, and voxelwise GMV. Since a significantly higher number of high grief participants were SCID positive for a depressive disorder (Table 1), we also included HAM-D as a covariate. AFNI’s 3dClustSim with the spatial autocorrection function was used to correct for multiple comparisons, by estimating the combination of individual voxelwise probability threshold (p) and cluster size threshold to yield an overall α < 0.05 for the entire volume corrected for family-wise error, according to 10,000 iterations in Monte Carlo simulations. Thus, the findings for the group differences were considered statistically significant at 2-sided, voxelwise p < 0.05, α < 0.05, cluster size>1569 voxels.

To examine the associations between whole-brain voxelwise VCau and NAcc FC (independent variables) and the primary clinical measure of interest, i.e., the yearning score derived from the first ICG cluster (dependent variable) across all grief participants, separate multiple linear regression models were used, while adjusting for age, gender, TSL, HAM-D, and voxelwise GMV (2-sided, voxelwise p < 0.05, α < 0.05, cluster size>1003 voxels),

$$\begin{gathered} C{M}_{(s)}={\text{β}}_0+{\text{β}}_1{\mathit{Age}}_{(s)}+{\text{β}}_2{\mathit{Gender}}_{(s)}+{\text{β}}_3{\mathit{TSL}}_{(s)}+{\text{β}}_{4}GM{V}_{(s,v)}+{\text{β}}_{5}HAM-D_{(s)}+{\text{β}}_6 \\ F{C}_{(s,v)}+{\text{ε}}_{(s,v)} \end{gathered}$$

where FC = whole-brain voxelwise (either VCau or NAcc) FC, CM = clinical measure (yearning scores), TSL = time since loss, GMV = gray matter volume, s = study participant, and v = voxel. Additional secondary analyses were conducted to explore if the VCau and NAcc FC correlated with the total ICG and HAM-D scores, while adjusting for covariates used in the above model (except that in the model with HAM-D as the dependent variable, ICG was included as an independent regressor).

We explored using separate models the main and interactive effects of age and voxelwise FC on yearning or HAM-D in the full sample. Furthermore, since ~70% of our sample comprised women, FC group differences and relationships between FC and clinical measures were explored in women only (n = 45), while adjusting for covariates (2-sided, voxelwise p < 0.05, α < 0.05, cluster size>1072 voxels).

To test whether FC differences between the high and low grief groups and the linear regression findings were VS subregion-specific, additional analyses were performed using the primary motor cortex and white matter as the control seed ROIs.

The MATLAB software (error_ellipse) was used to identify outliers (see supplementary material for details) (Johnson, 2021; Johnson and Wichern, 1982).

3. Results

Demographics and clinical characteristics.

The high and low grief groups did not differ in demographics; MMSE, total DRS-2 and CIRS-G scores; and TSL and relationship with the deceased. High grief participants relative to low grief participants reported higher total ICG, yearning, HAM-D, and HAM-A scores (p < 0.001). Of the 65 grief participants, 39 (60%) had lost a spouse/life partner or a child/grandchild. A total of 33 (50.8%) grief participants met the criteria for a bereavement-related psychiatric disorder; the majority (n = 31) had a depressive disorder diagnosis; n = 2 had anxiety disorder only. Eighteen (27.7%) participants (high grief: n = 14; low grief: n = 4) were on antidepressants (Table 1).

3.1. Voxelwise functional connectivity group differences

Ventral Caudate FC. The high grief group, compared with the low grief group, showed higher FC between VCau and the bilateral frontal/cingulate cortex cluster, which included the bilateral ventromedial prefrontal (vmPFC), orbitofrontal (OFC), and dorsal anterior cingulate (dorsal ACC) cortices, as well as the left superior frontal gyrus (SFG) and subgenual ACC, after adjusting for age, gender, TSL, HAM-D, and GMV (Fig. 2; cluster size: 1760 voxels; MNI coordinates −1, −59, −6; voxelwise p < 0.05, 3dClustsim corrected). No clusters with lower VCau FC were seen in the high compared with the low grief group.

Fig. 2.

Ventral caudate functional connectivity group differences.

(A) Illustrates the frontal/cingulate cortex cluster showing significant differences in high grief compared with the low grief group. Positive m-value indicates higher functional connectivity in high compared with low grief group (cluster size: 1760 voxels; MNI coordinates −1, −59, −6). (B) represent histograms indicating individual data distribution for high (green) and low (red) grief groups, respectively. The whiskers in the histograms indicate mean ± SD of the adjusted m-value of the significant cluster. VCau: ventral caudate; FC: functional connectivity.

Nucleus accumbens FC.

No significant NAcc FC differences between the high and low grief groups were seen.

3.2. Associations between voxelwise functional connectivity and clinical measures

Yearning.

VCau FC positively correlated with yearning severity in the bilateral vmPFC, SFG, dorsal ACC; the left OFC/subgenual ACC; and the right middle frontal gyrus (MFG) (Fig. 3A). The NAcc FC positively correlated with yearning in the right insula, inferior frontal gyrus (IFG), and caudate/putamen/pallidum, after adjusting for covariates including age, gender, TSL, GMV, and HAM-D scores (Fig. 3B).

Fig. 3.

Associations between reward subregion brain network function and yearning scores.

(A) illustrates the brain regions showing significant correlations between the VCau FC and yearning scores in the frontal/cingulate cortex cluster (left) and its associated scatter plot with the mean adjusted VCau FC on the x-axis and adjusted yearning scores on the y-axis (right).

(B) illustrates the brain regions showing significant correlations between the NAcc FC and yearning scores in the right insula/IFG/subcortical cluster (left) and its associated scatter plot with the mean adjusted NAcc FC on the x-axis and adjusted yearning scores on the y-axis (right).

Color bars represent the beta coefficients from the partial-F statistic.

VCau: ventral caudate; NAcc: nucleus accumbens; FC: functional connectivity; adj: adjusted for covariates; IFG: inferior frontal gyrus; subcort: subcortical (caudate/putamen/pallidum); R: right; L: left.

Grief symptoms.

Neither VCau nor NAcc FC showed significant associations with total ICG scores, after adjusting for covariates.

Depressive symptoms.

VCau FC negatively correlated with HAM-D scores in the left supramarginal gyrus (SMG), middle temporal gyrus (MTG), and temporoparietal junction (TPJ) (Fig. 4), whereas no NAcc–HAM-D associations were found, after adjusting for covariates. Significant results from the linear regression models are summarized in Table 2.

Fig. 4.

Associations between ventral caudate functional connectivity and depressive symptoms in grief. The figure illustrates the cluster showing significant correlation between the VCau FC and HAM-D scores (left) and its associated scatter plot with the mean adjusted VCau FC on the x-axis and adjusted HAM-D scores on the y-axis (right).

The color bar represents the beta coefficient from the partial-F statistic.

HAM-D: 17-item Hamilton Depression Rating Scale; VCau: ventral caudate; FC: functional connectivity; adj: adjusted for covariates; MTG: middle temporal gyrus; SMG: supramarginal gyrus; L: left.

Table 2.

Relationship between ventral striatal subregion functional connectivity and clinical measures in grief.

Ind variable	Dep variable	Regions	BA	Side	Cluster size (voxels)	MNI Coordinates			R2	p-value
						x	y	z
VCau FC	Yearning	vmPFC	10	B/L	1472	12	−22	26	0.24	p < 0.001
		SFG	9	B/L
		dorsal ACC	32	B/L
		OFC/subgenual ACC	11	L
		MFG	46	R
VCau FC	Depressive symptoms	SMG	40	L	1043	50	22	24	0.37	p < 0.001
		MTG	21	L
NAcc FC	Yearning	Insula	13	R	1048	− 36	− 22	6	0.35	p < 0.001
		IFG	44/45	R
		caudate/putamen/pallidum		R

Results for voxelwise p < 0.05, 3dClustsim corrected.

Abbreviations. MNI: Montreal Neurological Institute; x, y, z depicts coordinates of peak voxel in the cluster; Ind variable: independent variable; Dep variable: Dependent variable; VCau: ventral caudate; NAcc: nucleus accumbens; FC: functional connectivity; vmPFC: ventromedial prefrontal cortex; SFG: superior frontal gyrus; ACC: anterior cingulate cortex; OFC: orbitofrontal cortex; MFG: middle frontal gyrus; SMG: supramarginal gyrus; MTG: middle temporal gyrus; IFG: inferior frontal gyrus; B/L: bilateral; L: left; R: right.

3.3. Control seeds

No significant differences between the high and low grief groups, or associations between FC and clinical measures were found for both M1 and WM seeds.

3.4. Exploratory analyses

Women-only findings.

In women, no VCau FC differences between high and low grief groups nor in the associations between VCau FC and clinical measures (yearning and HAM-D) were seen.

Women with high grief symptoms (n = 26) showed higher FC between the NAcc and bilateral visual, posterior cingulate cortex (PCC) and precuneus areas, relative to those with low grief symptoms (n = 19) after adjusting for covariates (Fig. 5; cluster size: 3134 voxels; MNI peak coordinates: 23, 63, 24; voxelwise p < 0.05, 3dClustsim corrected). The relationships between NAcc FC and yearning or HAM-D were not significant.

Fig. 5.

Nucleus accumbens functional connectivity differences between high and low grief groups in women.

(A) Illustrates the cluster showing significant differences in high grief compared with the low grief group. Positive m-value indicates higher functional connectivity in the high group compared with the low grief group (cluster size: 3134 voxels; MNI coordinates 23, 63, 24). (B) represent histograms indicating individual data distribution for high (green) and low (red) grief groups, respectively. The whiskers in the histograms indicate mean ± SD of the adjusted m-value of the significant cluster. NAcc: nucleus accumbens; FC: functional connectivity; pDMN: posterior default mode network.

Age.

The models that explored age × FC (VCau FC or NAcc FC) interactions on yearning or HAM-D were not significant.

4. Discussion

Our novel preliminary observations of ventral striatal subregional dysfunction in bereaved older adults suggest that distinct neurobiology underlies yearning and depressive symptoms, highly co-occurring symptoms during the first year of loss. Higher VCau FC in the reward processing and executive control brain regions was found in the bereaved older adults experiencing high grief symptoms compared with low grief symptoms. The VCau FC in these areas positively correlated with yearning for the deceased, whereas the VCau FC in the temporoparietal areas negatively correlated with depressive symptom severity. In the full sample, although no NAcc FC group differences were seen, FC between this VS subregion and the salience and striatal regions positively correlated with yearning. Interestingly, when the analyses were limited to women only, higher FC between the NAcc and the posterior default mode network (pDMN) and visual areas were found in high grief relative to low grief symptoms. These results remained after covariate adjustments, including core depressive symptoms in the yearning models and grief severity in the depressive symptom models.

Heterogeneous grief trajectories are seen in older adults during the first year after the death of an attachment figure (Bonanno and Malgaroli, 2020; Djelantik et al., 2017). Intense symptoms during the early grieving process are associated with an increased incidence of PGD, a clinical condition now included in the DSM-5 and diagnosed 12 months after the loss of a loved one (Prigerson et al., 2021a, 2021b). Despite the magnitude of the problem, the neurobiological mechanisms underlying symptom variation during the first year of later life grief are unknown. It is conceivable that the neural substrates of high grief symptoms in the first year of bereavement would resemble the neurobiology of PGD and associated symptomatology.

The activity in the reward processing neural pathways appears to underlie major forms of social attachment, including the human bond with an attachment figure (Insel, 2003). Accordingly, PGD may result from persistent dysfunction in the ventral striatal neural network (Kakarala et al., 2020). Extensive animal and human research points to the centrality of VS in the functional neuroanatomy of reward (Haber and Knutson, 2010; Robbins and Everitt, 1996; Suckling and Nestor, 2017; Volkow et al., 2012). The VS receives rich innervations from the cortical areas that mediate different aspects of reward and emotion regulation. Although there is substantial cytoarchitectural overlap in humans between the VCau and NAcc, these important VS subregions have distinct functional connections with the cognitive control, salience, and limbic areas (Di Martino et al., 2008; Haber and Knutson, 2010). Together, these observations provide a solid rationale aiming to unravel the VS subregional correlates of symptom variation seen in the first year after the death of a loved one.

The brain regions showing higher VCau FC in the high compared with low grief group—the ventromedial prefrontal cortex, OFC, and subgenual and dorsal ACC—are the same areas integral to the VCau functional subregion in humans (Di Martino et al., 2008; Haber and Knutson, 2010). The VCau receives dense projections from these frontal and cingulate areas, which are implicated in multiple functions, including reward, motivation, social pain, emotion processing, and executive control. A handful of acute grief and PGD task-based fMRI studies also have reported increased activity in these regions in response to both grief-eliciting (Gundel et al., 2003; McConnell et al., 2018; O’Connor et al., 2008) and non-grief-related (Bryant et al., 2021; Fernandez-Alcantara et al., 2020) emotional stimuli. For instance, in bereaved women who experienced the death of a first-degree relative in the past year, the vmPFC showed high activity when responding to grief-related words, while the dorsal ACC and superior frontal gyrus were activated when viewing a photo of the deceased (Gundel et al., 2003). Interestingly, participants with PGD, relative to those with post-traumatic stress disorder and major depression, had higher activity in the medial OFC during supraliminal processing of non-grief-related sad faces (Bryant et al., 2021). Collectively, these data suggest that the ventral striatal functional subregion assessments can be surrogate measures of disruptions in the brain’s attachment system following the death of a loved one. Our results further underscore the central role of the VCau subregional network in the neurobiology of late-life grief.

Higher FC between VCau and the core reward and prefrontal regulatory brain regions correlated with yearning, a core grief symptom, but not with ICG, a global measure of grief severity across all participants. These preliminary observations lead us to hypothesize that an attachment loss contributes to heightened functioning of the VCau subregion in bereaved older adults, resulting in intense yearning for the deceased in the early grieving process. This hypothesis should be tested in future studies. Whether persistent dysfunction in this VS subregion would associate with protracted, intense yearning and the development of late-life PGD remains to be elucidated and should be the focus of future experiments.

Higher FC between the NAcc and pDMN and visual regions were seen in women with high, compared with low grief symptoms. In the full sample comprising both men and women, no significant NAcc FC group differences were found, though the NAcc-salience/striatal FC positively correlated with yearning. These mixed results were perplexing at the outset, but similar inconsistencies in the NAcc have been reported previously in task-based fMRI studies of PGD. For example, heightened NAcc activation to grief-related words was found in women with PGD compared with those with integrated grief (O’Connor et al., 2008). Taking a closer look, however, the significant ventral striatal cluster also extended to the VCau. This ventral striatal activity positively correlated with the yearning for the deceased in that study (O’Connor et al., 2008). This VS hyperactivity in PGD did not emerge in a follow-up investigation, but higher yearning correlated with subgenual ACC hyperactivity across the bereaved participants (McConnell et al., 2018). Multiple factors may have contributed to the discrepancies across studies. For instance, the O’Connor et al. and McConnell et al. studies (McConnell et al., 2018; O’Connor et al., 2008) had smaller samples, were limited by the task fMRI design, and were in participants with PGD and integrated grief, whereas our study included grieving individuals who were within the first year post-loss. Also, the McConnell et al. paper, like ours, comprised a predominantly older grief sample than the O’Connor et al. paper. While we did not find an age × FC interaction on yearning, it is still possible that the VS subregion FC abnormalities may vary based on the age of the study cohort. Whether the underlying neural substrates of high grief symptoms observed here and specifically of yearning severity, is generalizable to younger bereaved cohorts remain to be examined.

Our NAcc FC group differences suggest a gender-specific variance in the underlying neurobiology and the clinical expression of grief and its core symptom of yearning in older adults. The PCC is hypothesized to play a central role in promoting adaptive behavior in response to a change in the environment (including one arising from bereavement), by mediating multiple neuronal systems associated with learning, memory, cognitive control, and reward (O’Connor, 2012; Pearson et al., 2011). Heightened activity in the pDMN regions (particularly in the PCC) and visual areas to grief-eliciting task fMRI paradigms have been reported in women who experienced acute grief after the loss of an unborn child (Kersting et al., 2009) as well as grief following the death of a first-degree relative in the past year (Gundel et al., 2003). Interestingly, higher PCC activity during grief-eliciting task was associated with lower parasympathetic activity (suggesting greater arousal) in bereaved women (O’Connor et al., 2007). A larger distributed network activity, which included the PCC, OFC, insula, and basal ganglia were found to underlie the neural circuitry of mental representation of the deceased in a grief study that consisted of mostly women. This neural pattern trained to identify mental representations of the deceased predicted the occurrence of deceased-related thinking during a sustained attention task and avoidance (Schneck et al., 2017). Together, our exploratory findings point to the possibility that a higher FC between the NAcc and pDMN in those with intense, impairing symptoms during early grief contribute to higher rumination and intrusive thoughts related to the deceased. Whether alterations in this VS subsystem will lead to a maladaptive trajectory and the development of PGD needs further study.

Our VCau and NAcc FC regression findings provide further support in conceptualizing yearning for the deceased as a symptom dimension originating from transient impairments of the VS neuronal system, which, when persistent, would lead to PGD development. Consistent with this theory, PGD is theorized as a reward system-based syndrome, akin to other disorders of the positive valence systems (Kakarala et al., 2020). In this context, yearning in grief can be considered analogous to craving in substance use disorders (Suckling and Nestor, 2017; Volkow et al., 2012), which are states mediated by dysfunctional reward processing.

The FC between VCau and TPJ areas negatively correlated with depressive symptom severity. The TPJ functions as a pivotal node in connecting brain networks that integrate internal and external information, thereby contributing to multiple emotional, attentional, and social cognitive processes(Bzdok et al., 2013; Donaldson et al., 2015; Wu et al., 2015). The TPJ findings are not commonly reported in functional imaging studies of grief. One study reported activation of the TPJ areas while viewing pictures of mourning individuals versus controls (Labek et al., 2017). In contrast, disrupted TPJ subregional activity and FC is reported in major depression (Wen et al., 2021), a condition constituting multidimensional symptoms in the emotional, cognitive, and social domains. Depressive symptoms are highly prevalent following an attachment loss. More than one-third of widow(er)s at 1 month, and about 16% at 13 months meet major depression criteria after the death of a loved one (Zisook and Shuchter, 1991). Therefore, our results provide preliminary support for distinct neurobiology that underlies yearning and depressive symptoms, two commonly co-occurring symptoms in bereaved elders during the first year of an attachment loss. Furthermore, anhedonia, a core depressive symptom, appears to result from dysfunctional reward signaling (Treadway and Zald, 2011). Future studies should examine if distinct ventral striatal neural impairments contribute to yearning and anhedonia in acute and prolonged grief samples.

This preliminary cross-sectional investigation is not without limitations. Our findings are correlational; thus, causality cannot be inferred. We did not correct for multiple comparisons when examining the brain-behavior associations due to the pilot nature of this work. Our sample was primarily comprised of white females; thus, the findings may not be generalizable. Due to the small number of men in our sample, comparing findings in women and men was not feasible. Nevertheless, our exploratory work points to the possibility of gender-specific reward processing dysfunction in late-life grief, and should be investigated in the future. About one-half of the bereaved participants had a psychiatric disorder, with the majority meeting bereavement-related major depression criteria. To account for this potential confounder, we adjusted for HAM-D scores in the regression models using yearning and ICG scores as dependent variables. Due to the observational study nature, we allowed antidepressant use. Because antidepressants may modulate VS subregional FC, we extended our regression models to include antidepressant use as a covariate; our findings were largely unchanged (Results not shown). The TSL varied among our bereaved sample. While we included TSL to account for this variance, future investigations should enroll participants closer to the proximity of their loss. We derived yearning scores from the ICG scale; future studies should utilize the Prolonged Grief Disorder-13-Revised, a newer scale that aligns well with the DSM-5 PGD criteria to assess grief symptom severity(Prigerson et al., 2021a). Similarly, the Yearning in Situations of Loss scale(O’Connor and Sussman, 2014), which has good discriminant and convergent validity, is available to measure yearning severity. There are challenges in anatomically parcellating VS subregions due to their small sizes and the coarse resolution of the single-band EPI sequence. We chose our seed coordinates for this study based on prior publications (Contreras-Rodriguez et al., 2017; Di Martino et al., 2008; Karcher et al., 2019). While these VS seeds were non-overlapping, their relatively large acquisition voxels may have led to partial overlap in some participants after data processing. Still, our findings provide sufficient data to suggest distinct VS subregional FC abnormalities in grief and its associated symptomatology. Future clustering-based approaches can be used to detect functionally distinct VS clusters for seed-based FC analysis (Chen et al., 2012). Finally, future confirmatory studies should take advantage of the recent technological advances in EPI using parallel imaging techniques and simultaneous multi-slice (SMS) acquisition methods. With these novel SMS acquisition methods, higher spatial and/or temporal resolution can be achieved (Casey et al., 2018; Smith et al., 2013; Tozzi et al., 2020; Ugurbil et al., 2013), compared with the EPI data acquired here using the single-band sequence.

This study provides novel evidence suggesting that distinct VS subregional FC abnormalities may explain the variance in yearning and depressive symptoms in older adults following an attachment loss. Whether these divergent functional brain network aberrations in the VS neural system would predict the development of PGD and worsening depression (specifically anhedonia) in bereaved elders remains to be elucidated.