Age-Related Alterations in Gray Matter Microstructure in Older People With Remitted Major Depression at Risk for Dementia

Abstract

Background: Major depressive disorder (MDD) in late life is a risk factor for mild cognitive impairment (MCI) and Alzheimer's disease. However, studies of gray matter changes have produced varied estimates of which structures are implicated in MDD and dementia. Changes in gray matter volume and cortical thickness are macrostructural measures for the microstructural processes of free water accumulation and dendritic spine loss.

Methods: We conducted multishell diffusion imaging to assess gray matter microstructure in 244 older adults with remitted MDD (n = 44), MCI (n = 115), remitted MDD+MCI (n = 61), or without psychiatric disorders or cognitive impairment (healthy control participants; n = 24). We estimated measures related to neurite density, orientation dispersion, and free water (isotropic volume fraction) using a biophysically plausible model (neurite orientation dispersion and density imaging).

Results: Results showed that increasing age was correlated with an increase in isotropic volume fraction and a decrease in orientation dispersion index, which is consistent with neuropathology dendritic loss. In addition, this relationship between age and increased isotropic volume fraction was more disrupted in the MCI group than in the remitted MDD or healthy control groups. However, the association between age and orientation dispersion index was similar for all 3 groups.

Conclusions: The findings suggest that the neurite orientation dispersion and density imaging measures could be used to identify biological risk factors for Alzheimer's disease, signifying both conventional neurodegeneration observed with MCI and dendritic loss seen in MDD.

Keywords: Corticolimbic circuit; Dementia; Diffusion-weighted imaging; Fractional anisotropy; Frontal-executive circuit; Geriatric; Gray matter; MRI; Major depressive disorder; Mild cognitive impairment; Structural covariance; T1-weighted image.

Figure 1

CONSORT (Consolidated Standards of Reporting Trials) diagram. aMCI, amnestic MCI; DWI, diffusion-weighted imaging; FOV, field of view; HC, healthy control; MCI, mild cognitive impairment; MDD, major depressive disorder; MRI, magnetic resonance imaging; naMCI, nonamnestic MCI.

Figure 2

Methods overview. Panel (A) describes the neurite orientation dispersion and density imaging (NODDI) model and derived measures (neurite density index [NDI], orientation dispersion index [ODI], and isotropic volume fraction [fISO]). Panel (B) provides a brief visual overview of the preprocessing steps and gray matter–based spatial statistics and shows (i) raw diffusion-weighted imaging data, (ii) denoised EDDY/motion/susceptibility-corrected data, (iii) NODDI model fISO estimate, and (iv) pseudo-T1 image estimated via the gray matter–based spatial statistics algorithm. Panel (C) shows the group-derived gray matter skeleton used to constrain all analyses. CSF, cerebrospinal fluid.

Figure 3

Neurite orientation dispersion and density imaging gray matter measures by age. Panel (A) is a bootstrap ratio image that represents the loadings of each voxel on the latent partial least squares variable. Bootstrap ratios are thresholded at ±2, showing regions where voxels are reliably different from 0 (roughly equivalent to p < .05). The spatial map should be interpreted alongside the correlations with age represented in (B) and (C). Panel (B) shows the correlations between age (x-axis) and brain scores (y-axis). Brain scores represent the extent to which all participants within each condition load onto the brain map. Panel (C) presents the correlations shown in panel (B) as bootstrapped distributions with 95% CIs to accentuate the differences in correlations with age. Thus, panel (A) voxels with higher intensity showed a strong positive relationship with isotropic volume fraction (fISO), a negative relationship with the orientation dispersion index (ODI), and no clear relationship with neurite density index (NDI). L, left; LV, latent variable; R, right.

Figure 4

Relationship between orientation dispersion index (ODI) and age by diagnostic group. Panel (A) shows the spatial map with reliable bootstrap ratio values (thresholded at ±2). Panels (B) and (C) show the correlation results. Panel (B) shows the raw correlations between age and brain scores for each group, and panel (C) shows the bootstrapped median correlation with a 95% confidence estimate. Panel (D) presents a table of group differences derived from the bootstrap replicates with 95% confidence bounds and a Cohen’s d estimate. Thus, in panel (B), latent variable 1 (LV1) summarizes the ODI data, to which the voxels shown in blue in panel (A) make the most robust contributions. Older age was associated with lower ODI in the bilateral caudate nuclei and higher ODI in the cerebellum. aMCI, amnestic MCI; CoV, covariance; HC, healthy control; L, left; MCI, mild cognitive impairment; MDD, major depressive disorder; naMCI, nonamnestic MCI; R, right.

Figure 5

Relationship between isotropic volume fraction (fISO) and age by diagnostic group. Panel (A) shows the spatial map with reliable bootstrap ratio values (thresholded at ±2). Panels (B) and (C) show the correlation results. Panel (B) shows the raw correlations between age and brain scores summarizing fISO in the regions highlighted in panel (A) for each group, and panel (C) shows the bootstrapped median correlation with a 95% confidence estimate. Panel (D) presents a table of group differences derived from the bootstrap replicates with 95% confidence bounds and a Cohen’s d estimate. Thus, in panel (B), latent variable 1 (LV1) summarizes the fISO data, to which the voxels shown in blue in panel (A) make the most robust contributions. aMCI, amnestic MCI; CoV, covariance; HC, healthy control; L, left; MCI, mild cognitive impairment; MDD, major depressive disorder; naMCI, nonamnestic MCI; R, right.

Figure 6

Principal component analysis (PCA) of behavioral variables. Panel (A) shows the biplot results of the PCA on cognitive variables. Panel (B) shows the 10 cognitive tests that contributed most to the first PCA dimension. The dashed red line indicates the expected average contribution for the variables under a uniform distribution (in this case, 1/12, or 8.3%). Variables exceeding this cutoff can be considered important for contributing to this component. Panel (C) shows how average orientation dispersion index (ODI) and isotropic volume fraction (fISO) correlated with the first latent variable (higher PCA scores indicate higher global cognitive performance). Labels on the x-axis are multiplied by 1000. Coding refers to digit symbol substitution, and fluency refers to semantic fluency. aMCI, amnestic MCI; BNT, Boston Naming Test; BVMT, Brief Visuospatial Memory Test; CPT, Continuous Performance Test – Identical Pairs version; CVLT, California Verbal Learning Test; EF, executive function; HC, healthy control; MCI, mild cognitive impairment; naMCI, nonamnestic MCI; PASAT, Paced Auditory Serial Addition Test; PASS, Performance Assessment of Self-Care Skills; rMDD, remitted major depressive disorder; TMT, Trail Making Test.