Introduction

Volumetric changes in depressed individuals are well documented, particularly in frontolimbic regions of the brain (Arnone et al., 2016; Dotson et al., 2009). Fewer studies focus on depression-related changes in cortical thickness, which, together with surface area, comprises gray matter volume (Winkler et al., 2010). Individual differences in brain volumes may be more closely related to surface area rather than cortical thickness (Reuter et al., 2010; Winkler et al., 2010), while cortical thickness is thought to be more closely related to features of the cortex, such as the organization of cortical layers, synaptic connections, and the size, number, and density of cell bodies in the neurons (Kabani et al., 2001). Cortical thickness is considered a more sensitive marker of disease state in depression and has been proposed as a phenotype for depression (Peterson & Weissman, 2011).

Studies of cortical thickness in depression have yielded conflicting results, with both reductions and increases in thickness reported for regions including the orbitofrontal, prefrontal and cingulate cortices, parahippocampal and other temporal regions, as well as the insula (Grieve et al., 2013; Jarnum et al., 2011; Lim et al., 2012; van Eijndhoven et al., 2013). It has been speculated that during the early course of a depressive disorder, increased cortical thickness occurs as a compensatory response to inflammation, whereas a more prolonged illness course eventually leads to decreased cortical thickness (Rajkowska & Miguel-Hidalgo, 2007; Yang et al., 2015).

Consistent with this hypothesis, Yang and colleagues (Yang et al., 2015) found that compared with controls, people with major depressive disorder (MDD) had greater cortical thickness in the medial orbitofrontal cortex (OFC) and inferior parietal gyrus; however, a longer duration of illness was associated with a thinner medial OFC. Indirect evidence also comes from our previous work examining subthreshold depressive symptoms, which often are a prodrome for a major depressive episode (see Lee et al., 2019 for a meta-analysis). We showed that older adults with higher depressive symptoms, but without clinical depression, have greater cortical thickness in the isthmus cingulate and precuneus (Szymkowicz et al., 2016).

Structural and functional neuroimaging studies suggest possible lateralization in the neurobiology of depression. According to the valence hypothesis (Davidson & Irwin, 1999; Wittling & Roschmann, 1993), positive emotions are associated with the left hemisphere and negative emotions with the right hemisphere. Consistent with this hypothesis, the approach-withdrawal model proposes that depression is characterized by a disturbance in the normal balance of left frontal activity underlying reactions to positive stimuli, and right frontal activity underlying reactions to negative stimuli (Coan & Allen, 2003; Davidson, 1998). Functional imaging studies have shown greater reductions in left compared with right frontal activity at rest as a function of clinical depression or depressive symptoms (Bruder et al., 2017; Thibodeau et al., 2006). However, results have not been consistent, and some studies report null findings (Quinn et al., 2014; Reid et al., 1998). There also is evidence of lateralized effects of depression on cortical thickness. A review of this literature concluded that the right medial OFC is thinner in individuals with MDD compared with controls (Zhang et al., 2018). A more recent study found an effect of depression on thickness in the inferior temporal gyrus and caudal anterior cingulate that was specific to the right hemisphere (Schmaal et al., 2017).

Such studies typically make conclusions about asymmetry based on finding significant effects in one hemisphere but not the other. This approach does not shed light on the left and right hemispheres in relation to each other. Recent studies have used a different approach, calculating an asymmetry index (AI) that allows for a statistical analysis of differences in the relative size of left and right hemisphere regions using a scale that ranges from −100 to 100 (Kawasaki et al., 2008; Kurth et al., 2015; Postema et al., 2019; Shah et al., 2019). Unlike traditional approaches, the continuous nature of the AI provides a quantitative measure of the degree of right-left asymmetry, which allows for a more nuanced understanding of hemispheric differences in relation to an outcome, such as depression or depressive symptoms. Using this index, one study found depression-related differences in cortical thickness asymmetry in the caudal middle frontal, superior frontal, and rostral middle frontal cortices (Zuo et al., 2019). An ENIGMA consortium analysis found asymmetry differences in the caudal anterior cingulate, pars opercularis, and superior temporal gyrus in depressed adults older than 21 years of age, although none of these effects survived correction for multiple comparisons (de Kovel et al., 2019).

Building on previous literature, the current study used the AI to examine the relationship between dimensions of depressive symptoms (somatic symptoms, negative affect, and anhedonia) and asymmetry in cortical thickness in a community sample with primarily subthreshold depressive symptoms. Growing literature shows that different components, or symptom dimensions, of depression have distinct associations, including with cognitive functioning, brain structure and function, etiological contributors, and clinical outcomes (Brailean et al., 2016; Dotson & McClintock, 2020; Kappelmann et al., 2020; McLaren et al., 2017; Murri et al., 2018). This is in line with the National Institute of Mental Health’s Research Domain Criteria (RDoC) framework, which emphasizes research on dimensions of behavior that cut across disorders (Insel et al., 2010). Within the RDoC framework, neuroimaging studies are one of multiple units of analysis that are used to link behavioral dimensions with underlying neurobiological mechanisms that are thought to underlie those dimensions, independent of specific diagnostic categories. For example, anhedonia and somatic symptoms are symptom dimensions of depression that are also relevant to schizophrenia and anxiety disorders. Thus, by identifying neurobiological correlates of symptom dimensions within depression, we also can enhance our understanding of other disorders.

Structural neuroimaging studies of depressive symptom dimensions are limited, but the existing studies show unique structural correlates of different symptom dimensions (Bogoian et al., 2020; Lener et al., 2016; McLaren et al., 2016; Tully et al., 2017). The current study adds to this growing literature by using a more sensitive structural measure—cortical thickness—and a more sensitive measure of hemispheric differences—the AI. We focused on cingulate subregions and the OFC based on our previous work showing distinct relationships between dimensions of subthreshold depression and cingulate volumes (McLaren et al., 2016), and recent reviews that identified the OFC as showing consistent depression-related differences in cortical thickness (Suh et al., 2019; Zhang et al., 2018). We predicted that higher depressive symptoms would be associated with greater asymmetry. Based on our previous volumetric findings and the role of the OFC in reward processing (Hoflich et al., 2019; McLaren et al., 2016), we specifically expected anhedonia to be associated with orbitofrontal and anterior cingulate asymmetry, while somatic symptoms and negative affect were expected to be associated with asymmetry in the posterior and isthmus cingulate, regions associated with arousal states, internally directed cognition, and affective flattening. Given inconsistent laterality effects in the depression literature, we did not have specific predictions about the direction of asymmetry in our regions of interest (ROIs).

Methods and Materials

Participants

This study included a community-dwelling sample of 54 participants between the ages of 18 and 81 years. Demographic data are presented in Table 1. Participants were required to have at least 9 years of education, to be right-handed, native-English speakers, and to score above the suggested dementia cutoff of 30 on the Telephone Interview for Cognitive Status (Brandt et al., 1988)—a cognitive screening measure. Potential participants were excluded based on the following criteria: self-reported major medical conditions (e.g., stroke, cancer), history of head trauma, current antiepileptic or antipsychotic use, language comprehension difficulties or learning disorder, and magnetic resonance imaging (MRI) contraindications. This study was approved by the local institutional review board, and all participants gave written and verbal informed consent.

Table 1 Sample characteristics
Full size table

Depressive Symptom Assessment

Participants completed the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977), a 20-item self-report measure of depressive symptoms experienced during the preceding week. The CES-D has been well validated in both young and older adults (Caracciolo & Giaquinto, 2002; Cosco et al., 2017). Meta-analysis of the factor structure of the CES-D revealed three factors based on 14 of the 20 items: somatic symptoms, negative affect, and anhedonia. Participants in the sample had mostly subthreshold depressive symptoms based on the CES-D cutoff of 16 that is generally recommended for identifying clinical depression (Weissman et al., 1977). Five participants scored ≥16.

MRI Data Acquisition

All participants received a 3-Tesla structural MRI scan in a Phillips (Amsterdam, Netherlands) scanner located at the University of Florida’s McKnight Brain Institute. A high resolution, T1-weighted turbo field echo anatomical scan was collected using a standard 8-channel head radio-frequency coil and the following parameters: TR = 28.1 ms, TE = 2.7 ms, 170 slices acquired in a sagittal orientation, flip angle = 8 degrees, 1-mm cubic resolution.

Cortical Thickness Measurement

Cortical thickness was extracted using Freesurfer version 5.3, which has standard procedures for automated parcellation of both cortical and subcortical brain structures that are documented and freely available for download (http://surfer.nmr.mgh.harvard.edu/). Automated cortical thickness measures have been validated against histological analysis (Rosas et al., 2002) and manual measurements (Kuperberg et al., 2003; Salat et al., 2004).

The procedure includes motion correction, removing non-brain tissue (Segonne et al., 2004), applying a Talairach transformation, segmenting the subcortical structures (Fischl et al., 2002; Fischl, Salat, et al., 2004a), correcting intensity inhomogeneity (Sled et al., 1998), tessellating the gray/white matter boundary, and topology correction (Fischl et al., 2001; Segonne et al., 2007). Surface deformation followed intensity gradients to optimally place the gray/white and gray/cerebrospinal fluid (CSF) borders at the location where the greatest shift in intensity defines the transition to the other tissue class (Dale et al., 1999; Dale & Sereno, 1993; Fischl & Dale, 2000). Then, the cortex was parcellated into discrete regions with respect to gyral and sulcal structure (Desikan et al., 2006; Fischl, van der Kouwe, et al., 2004b). Using intensity and continuity information in segmentation and deformation procedures, cortical thickness was calculated as the closest distance from the gray/white boundary to the gray/CSF boundary at each vertex on the tessellated surface (Fischl & Dale, 2000). As a quality control measure, we calculated interclass correlation coefficients (ICCs) for manual adjustments in cortical thickness segmentation. ICC between raters ranged from good to excellent (.644 to .903) across ROIs in the current study (Cicchetti, 1994).

Statistical Analysis

To quantify differences between cortical thickness in the left and right hemispheres, an asymmetry index (AI) was calculated for ROIs in the OFC (medial + lateral OFC), rostral anterior cingulate, caudal anterior cingulate, posterior cingulate, and isthmus cingulate. The following formula was applied, consistent with prior studies (de Kovel et al., 2019; Zuo et al., 2019): AI = (left – right) × 100 / (left + right). Values for the AI can range from −100 to +100, representing complete right- or left-lateralized asymmetry, respectively.

Statistical analyses were performed in SAS version 9.4 (SAS Institute, Cary, NC) using the PROC GLM procedure. CES-D somatic symptoms, negative affect, and anhedonia scores were the predictors, thus allowing us to identify the effect of each symptom dimension while controlling for the other subscales. In this way, we can interpret effects as the association of each dimension of depression with cortical thickness asymmetry, controlling for the relationship of other symptom dimensions with thickness asymmetry in that region. Correlations between subscales ranged from 0.40 to 0.74. The variance inflation factors were all ≤2.55, suggesting that multicollinearity was not a significant issue (Hair, 2010). Models controlled for age and sex. Separate models were conducted for the five ROIs. All variables besides sex were continuous measures in the models. Statistical significance was set at α = 0.05. Effect sizes are reported as partial eta squared (ηp2).

For comparison with the existing literature, parallel models were conducted with left and right cortical thickness in our ROIs as the dependent variables.

Results

Table 2 and Fig. 1 summarize results of the regression analyses. Higher scores on the anhedonia subscale were associated with greater right-sided asymmetry in orbitofrontal thickness [F(5, 48) = 8.07, p = 0.006, ηp2 = 0.144]. For the somatic symptoms subscale, higher scores were associated with greater left-sided asymmetry in the posterior cingulate [F(5, 48) = 5.03, p = 0.029, ηp2 = 0.095]. None of the CES-D subscales were associated with thickness asymmetry in the rostral or caudal anterior cingulate. The Shapiro-Wilk test indicated that the residuals were normally distributed.

Table 2 Association of CES-D subscale scores with cortical thickness AI in regression analyses, b (SE)
Full size table
Fig. 1
figure 1

Least-square means showing the asymmetry index for each region of interest by Center for Epidemiologic Studies Depression Scale (CES-D) subscales scores. CES-D subscale scores were continuous measures in all analyses. The depressive symptom group are based on tertiles of CES-D scores within the study sample and are for graphical purposes only. Error bars represent standard error. *p < 0.05,**p < 0.01

Full size image

To follow up on the significant effect of anhedonia on orbitofrontal asymmetry, separate models were run for the medial and lateral OFC. The association remained significant for medial [F(5, 48) = 6.35, p = 0.015, ηp2 = 0.117] but not lateral [F(5, 48) = 2.01, p = 0.163, ηp2 = 0.040] orbitofrontal asymmetry.

None of the effects were significant when left and right ROIs were examined separately.

Discussion

Our hypotheses were largely confirmed. We found significant asymmetry in cortical thickness related to depressive symptoms. Specifically, higher scores on the anhedonia subscale were associated with right > left asymmetry in the OFC, and higher somatic symptoms of depression were associated with left > right asymmetry in the posterior cingulate. Contrary to our hypothesis, the negative affect subscale was not associated with cortical thickness asymmetry. Our findings support the growing evidence that different aspects of depressive symptomatology have distinct neural underpinnings.

In our sample, participants with higher scores on the anhedonia subscale tended to have greater cortical thickness in the right compared to left OFC, in contrast to those with lower scores, who showed a left > right asymmetry. It is not surprising that asymmetry in the OFC was specifically associated with anhedonia, considering the neurobiology of anhedonia. The OFC is a component of the reward system, the frontosubcortical network of brain regions that subserves the anticipation and experience of pleasure (Bray et al., 2010; Padoa-Schioppa, 2007). Anhedonia has a strong role in reward processing, and OFC activity is decreased in anhedonic individuals with MDD (Keedwell et al., 2005). Follow-up analyses showed that our finding was driven specifically by asymmetry in the medial OFC. This is consistent with findings of increased thickness in the right medial OFC in clinical depression (Qiu et al., 2014; Yang et al., 2015), although other research has shown the same for the left lateral OFC (Grieve et al., 2013). Nonetheless, a meta-analysis (Suh et al., 2019) showed bilateral thinning as opposed to thickening of the OFC in clinical depression.

The differences in findings across studies likely relates at least in part to differences in the severity and chronicity of depression. Our study is unique in its focus on a community sample with primarily subthreshold depressive symptoms, which could be a prodrome to more significant clinical depression in some individuals (Cuijpers & Smit, 2004; Tuithof et al., 2018). Similar to our results, the right medial OFC has been found to be thicker in people at risk for MDD (Peterson et al., 2009) and in first-episode treatment-naïve MDD (Qiu et al., 2014), while a longer duration of illness has been associated with thinning in the same region (Yang et al., 2015). Together, these findings are in line with the hypothesis that increased cortical thickness occurs as a compensatory mechanism in early stages of depression, but cortical thinning will eventually be observed after more prolonged illness (Rajkowska & Miguel-Hidalgo, 2007; Yang et al., 2015). However, it should be noted that due to the cross-sectional nature of our study, we are unable to definitively characterize subthreshold symptoms in our sample as prodromal. While prior studies do suggest that subthreshold symptoms can be a prodrome to major depression (Fava & Tossani, 2007; Lee et al., 2019), not everyone with subthreshold symptoms goes on to develop clinical depression. Moreover, subthreshold symptoms can be part of a residual phase of depression (Ji, 2012). Longitudinal studies will help to clarify these possibilities.

We also found a relationship between higher somatic symptoms of depression and asymmetry in the posterior cingulate, reflecting greater cortical thickness in the left compared to right hemisphere. Similarly, earlier work showed thickness in the left posterior cingulate was increased in first-episode medication-naïve major depression (van Eijndhoven et al., 2013). Taken with reports of orbitofrontal thinning in depressed non-remitters (Jarnum et al., 2011), this once again supports the possibility that structural changes shift from cortical thickening to cortical thinning after prolonged illness.

The link between the somatic symptoms subscale and the posterior cingulate is consistent with previous literature, as is the trend-level association of somatic symptoms with the isthmus cingulate (p = 0.09). The posterior cingulate is a highly connected region (Hagmann et al., 2008) that is a central component of the default mode network (Buckner & DiNicola, 2019). This region has a number of affective and cognitive roles, some of which align with the symptoms measured by the somatic symptoms subscale. For example, items such as ‘my sleep was restless’ and ‘I had trouble keeping my mind on what I was doing’ are consistent with the posterior cingulate’s role in sleep and insomnia (Li et al., 2018; Vogt & Laureys, 2005; Wu et al., 2020), and regulating the focus of attention (Gusnard et al., 2001; Hahn et al., 2007; Hampson et al., 2006). The isthmus cingulate, which connects the posterior cingulate to the parahippocampal gyrus, also is part of the default mode network. It has been implicated in memory and pain processing (Nielsen et al., 2005), as well as mood symptoms (Whitford et al., 2014). Thus, the specific associations we observed between dimensions of depressive symptoms and cortical thickness asymmetry are in line with the known roles of each region.

We did not find a relationship between the negative affect subscale and asymmetry in any of our ROIs, contrary to our hypotheses and to numerous studies linking depressive symptoms with structural abnormalities in the cingulate cortex, as well as the OFC (Schmaal et al., 2017; Wise et al., 2017). For example, we previously found that negative mood symptoms were associated with larger volumes in the left posterior cingulate and smaller volumes in the right isthmus cingulate (McLaren et al., 2016). However, the previous findings were in an older sample (aged 51 years or older) and the study focused on separate analysis of left and right hemisphere volumes rather than an AI of cortical thickness. Given the limited structural imaging studies of depressive symptom dimensions, the reasons for the null findings for the negative affect subscale are unclear. It is possible that the cingulate cortex and OFC have more of a role in anhedonia and somatic symptoms of depression rather than sad mood per se. However, given the dearth of research using the AI to examine the relationship between symptom dimensions of depression and cortical thickness, null results should be taken with caution.

In addition, our findings should be considered within the context of study limitations, including the relatively small sample size, homogenous sample, reliability of the CES-D subscales, and lack of information about psychiatric comorbidities for the full sample. The reliability of questionnaire subscales is particularly important to consider in studies that examine symptom dimensions of depression using measures such as the CES-D, as poor subscale reliability can limit our confidence that we are measuring the symptom dimension of interest. Nonetheless, this study contributes to a fairly small but growing literature. Though preliminary, our findings suggest that the relative size of left and right cingulate and orbitofrontal thickness might be a marker of subthreshold depressive symptoms. Considering that we did not find significant effects when examining left and right ROIs separately, results also suggest that the relative hemispheric difference, measured by the AI, is more sensitive than examining each hemisphere in isolation. Moreover, the study highlights distinct associations between symptom dimensions of depression and cortical thickness. While replication and extension of our findings is critical, these findings suggest that our understanding of the neurobiological basis of depression will be enhanced by considering symptom presentation and subtle differences in structural asymmetry.

Neuroimaging studies have shown both structural and functional asymmetries in clinical depression (Bruder et al., 2017; Thibodeau et al., 2006; Zhang et al., 2018), but much less is known about lateralization differences in the neurobiology of subthreshold depression. Subthreshold symptoms are a risk factor for the development of major depression, and are associated with cognitive, brain changes, and functional deficits, even at a subclinical level (Dotson et al., 2008; Li et al., 2017; Ludvigsson et al., 2019; Tuithof et al., 2018; Zhang et al., 2020). More work is needed to identify brain asymmetries across the depression spectrum. There is preliminary evidence that hemispheric asymmetry is related to clinical outcome after stimulation antidepressant treatments such as repetitive transcranial magnetic stimulation and electroconvulsive therapy (Downar et al., 2014; Wade et al., 2017). As such, understanding brain asymmetries related to different symptom dimensions of depression might inform targeted treatment for depression based on symptom profile. Studies in larger samples that can examine possible demographic moderators such as age, race, and sex also will provide a valuable contribution to this area of research.