Fornix white matter glia damage causes hippocampal gray matter damage during age-dependent limbic decline

Abstract

Aging leads to gray and white matter decline but their causation remains unclear. We explored two classes of models of age and dementia risk related brain changes. The first class of models emphasises the importance of gray matter: age and risk-related processes cause neurodegeneration and this causes damage in associated white matter tracts. The second class of models reverses the direction of causation: aging and risk factors cause white matter damage and this leads to gray matter damage. We compared these models with linear mediation analysis and quantitative MRI indices (from diffusion, quantitative magnetization transfer and relaxometry imaging) of tissue properties in two limbic structures implicated in age-related memory decline: the hippocampus and the fornix in 166 asymptomatic individuals (aged 38-71 years). Aging was associated with apparent glia but not neurite density damage in the fornix and the hippocampus. Mediation analysis supported white matter damage causing gray matter decline; controlling for fornix glia damage, the correlations between age and hippocampal damage disappear, but not vice versa. Fornix and hippocampal differences were both associated with reductions in episodic memory performance. These results suggest that fornix white matter glia damage may cause hippocampal gray matter damage during age-dependent limbic decline.

Figure 1

Overview of two mediation models: Model (A) tests for the indirect effects of hippocampal mediator variables on the direct effect of age on the fornix (path c’). The indirect effect is the product a*b of the correlations between age and hippocampus (path a) and the partial correlation between hippocampus and fornix accounting for age (path b). Model (B) tests the indirect effects of the fornix mediator variables on the direct effect of age on the hippocampus.

Figure 2

(A) MRI modalities and maps acquired from dual-shell high angular resolution diffusion imaging (HARDI), quantitative magnetisation transfer (qMT) imaging and from T₁ and T₂ relaxometry. HARDI data were modelled with neurite orientation dispersion and density imaging (NODDI), yielding maps of intracellular signal fraction (ICSF), isotropic signal fraction (ISOSF) and orientation dispersion index (ODI). The qMT maps were the macromolecular proton fraction (MPF) and the forward exchange rate k_f. Maps from relaxometry were the longitudinal relaxation rate R₁ (1/T₁) and R₂ (1/T₂). (B) Mean indices of the metrics were extracted from left and right hippocampi, fornix and parahippocampal cinguli (PHC) tracts. Hippocampi were segmented from T₁- weighted images with FreeSurfer version 5.3 and fornix and PHC were reconstructed with damped-Richardson Lucy spherical deconvolution (dRL) based deterministic tractography on colour coded principal direction maps (RGBλ). (C) Cross-correlation matrix between white and gray matter microstructural indices that correlated with age.

Figure 3

Plots the correlations and Pearson coefficients (controlled for intracranial volume) between age and white and gray matter microstructural indices. Abbr.: ICSF = intracellular signal fraction, ISOSF = isotropic signal fraction, k_f = forward exchange rate, LHC = left hippocampus, LPHC = left parahippocampal cingulum, MPF = macromolecular proton fraction, ODI = orientation dispersion index, R = longitudinal relaxation rate, RHC = right hippocampus, RPHC = right parahippocampal cingulum ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, 5% False Discovery Rate corrected.

Figure 4

Summarises the results of the mediation analyses for Models A and B. 95% confidence intervals of the effect sizes (ES × 10⁻¹) were based on bootstrapping with 5000 replacements. Fornix mediators in Model B had significant indirect effects and fully mediated the direct age effects on right hippocampal (RHC) isotropic signal fraction (ISOSF), left hippocampal (LHC) R₁ and k_f (highlighted in red). In contrast, hippocampal mediators, although showing significant indirect effects (highlighted in bold), did not fully mediate the direct age effects on fornix MPF, R₁ and k_f. Right (R) hippocampal ISOSF fully mediated fornix ISOSF and vice versa but fornix mediators did not remove the age effect on left hippocampal ISOSF. Mediator variables that contributed significantly to the regression analyses after 5% FDR correction are highlighted in bold.