Distinct Longitudinal Brain White Matter Microstructure Changes and Associated Polygenic Risk of Common Psychiatric Disorders and Alzheimer's Disease in the UK Biobank

Abstract

Background: During the course of adulthood and aging, white matter (WM) structure and organization are characterized by slow degradation processes such as demyelination and shrinkage. An acceleration of such aging processes has been linked to the development of a range of diseases. Thus, an accurate description of healthy brain maturation, particularly in terms of WM features, is fundamental to the understanding of aging.

Methods: We used longitudinal diffusion magnetic resonance imaging to provide an overview of WM changes at different spatial and temporal scales in the UK Biobank (UKB) (n = 2678; age_{scan 1} = 62.38 ± 7.23 years; age_{scan 2} = 64.81 ± 7.1 years). To examine the genetic overlap between WM structure and common clinical conditions, we tested the associations between WM structure and polygenic risk scores for the most common neurodegenerative disorder, Alzheimer's disease, and common psychiatric disorders (unipolar and bipolar depression, anxiety, obsessive-compulsive disorder, autism, schizophrenia, attention-deficit/hyperactivity disorder) in longitudinal (n = 2329) and cross-sectional (n = 31,056) UKB validation data.

Results: Our findings indicate spatially distributed WM changes across the brain, as well as distributed associations of polygenic risk scores with WM. Importantly, brain longitudinal changes reflected genetic risk for disorder development better than the utilized cross-sectional measures, with regional differences giving more specific insights into gene-brain change associations than global averages.

Conclusions: We extend recent findings by providing a detailed overview of WM microstructure degeneration on different spatial levels, helping to understand fundamental brain aging processes. Further longitudinal research is warranted to examine aging-related gene-brain associations.

Keywords: Aging; Diffusion MRI; Magnetic resonance imaging; Microstructure; Polygenic risk; White matter.

Figure 1

Global white matter microstructure (WMM) aging trajectories. First, WMM values were standardized, mean centered, and adjusted for covariates of no interest using linear mixed models. Second, age-WMM relationships were described by linear and nonlinear functions. An overview of the utilized diffusion approaches and entailed WMM metrics can be found in Supplemental Appendix B. AD, axial diffusivity; ADC, apparent diffusion coefficient; AK, axial kurtosis; AWF, axonal water fraction; BRIA, Bayesian rotational invariant approach; DAX extra, extra-axonal axial diffusivity; DAX intra, intra-axonal axial diffusivity; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; FA, fractional anisotropy; mcSMT, multi-compartment spherical mean technique; MD, mean diffusivity; micro AX, microscopic axial diffusivity; MK, mean kurtosis; radEAD, radial extra-axonal diffusivity; RD, radial diffusivity; RK, radial kurtosis; SMT, spherical mean technique; Vcsf, cerebrospinal fluid fraction; V extra, extra-axonal water fraction; V intra, intra-axonal water fraction; WMTI, white matter tract integrity.

Figure 2

Age-stratified annual white matter microstructure change. White matter microstructure was corrected for age, sex, age × sex, and site and standardized for comparability (without mean centering). We present p values for Wilcoxon tests, which were significant at the Bonferroni-corrected ∗α < .05/(26 × 6) = 3.21 × 10⁻⁴, ∗∗α < 0.01/(26 × 6) = 6.41 × 10⁻⁵, ∗∗∗α < 0.001/(26 × 6) = 6.41 × 10⁻⁶. The red lines were added as a visual aid to identify trends of accelerated or decelerated annual change. An overview of the utilized diffusion approaches and entailed white matter microstructure metrics can be found in Supplemental Appendix B. AD, axial diffusivity; ADC, apparent diffusion coefficient; AK, axial kurtosis; AWF, axonal water fraction; BRIA, Bayesian rotational invariant approach; DAX extra, extra-axonal axial diffusivity; DAX intra, intra-axonal axial diffusivity; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; FA, fractional anisotropy; mcSMT, multi-compartment spherical mean technique; MD, mean diffusivity; micro AX, microscopic axial diffusivity; MK, mean kurtosis; NS, nonsignificant; radEAD, radial extra-axonal diffusivity; RD, radial diffusivity; RK, radial kurtosis; SMT, spherical mean technique; Vcsf, cerebrospinal fluid fraction; V extra, extra-axonal water fraction; V intra, intra-axonal water fraction; WMTI, white matter tract integrity.

Figure 3

Regional white matter microstructure changes between time points and age associations. (A) Unadjusted effect sizes (Cohen’s d) vs. Bonferroni-adjusted −log₁₀p values. Labeling was done using a medium effect size threshold of Cohen’s |d| > 0.5 (also marked with vertical lines) as well as extreme Bonferroni-adjusted p values of −log₁₀(p) > 500. (B) Adjusted white matter microstructure associations with age. Age–white matter microstructure were adjusted for sex, sex × age, scanner site, and time point (equation 2). The plot presents standardized slopes (β) vs. Bonferroni-adjusted −log₁₀p values. Labeling was done using a large association of |β| > 0.5 (also marked with vertical lines). Dotted lines were inserted as visual aid: the lower horizontal dotted line represent the significance level of α = .05 and the upper horizontal line −log₁₀(p) = 500. The vertical lines represented labeling borders based on a medium effect size of Cohen’s |d| > 0.5 (A) and large associations of |β| > 0.5 (B). Tables with test statistics are available at https://github.com/MaxKorbmacher/Long_Diffusion/. AD, axial diffusivity; AK, axial kurtosis; AWF, axonal water fraction; BodyCC, body of the corpus callosum; BRIA, Bayesian rotational invariant approach; DAX extra, extra-axonal axial diffusivity; DAX intra, intra-axonal axial diffusivity; DKI, diffusion kurtosis imaging; DTI, diffusion tensor imaging; FA, fractional anisotropy; FMIN, Forceps Minor; ICP, inferior cerebellar peduncle; MCP, middle cerebellar peduncle; mcSMT, multi-compartment spherical mean technique; MD, mean diffusivity; microAX, microscopic axial diffusivity; MK, mean kurtosis; radEAD, extra-axonal radial diffusivity; RD, radial diffusivity; RK, radial kurtosis; SCP(R), right superior cerebellar peduncle; SLTF(R), right superior longitudinal temporal fasciculus; SMT, spherical mean technique; StTer(R), right stria terminalis; Vcsf, cerebrospinal fluid fraction; V extra, extra-axonal water fraction; V intra, intra-axonal water fraction.

Figure 4

Associations of polygenic risk scores (PRSs) with the rate of white matter change. (A) The global associations between PRS and white matter microstructure change. Colors indicate the strength of association (standardized β coefficients). (B) Regional associations between PRS and white matter microstructure change. The dotted line indicates an uncorrected α < 0.001. Labels presenting the respective regions and metrics are supplied above this α threshold, as well as at |β| > 0.05. (C) Regional associations exclusively for the medial cerebral peduncle, the region where the strongest annual rate of change-PRS associations were observed. Boxes in (A) and (C) indicate the statistical significance at an uncorrected α < 0.05. All associations were adjusted for age, sex, the age × sex interaction, and site. None of the presented associations survived the adjustment of the α level for multiple comparisons. An overview of the utilized diffusion approaches and entailed white matter microstructure metrics can be found in Supplemental Appendix B. AD, Alzheimer’s disease; ADC, apparent diffusion coefficient; ADHD, attention-deficit/hyperactivity disorder; AK, axial kurtosis; ANX, anxiety; ASD, autism spectrum disorder; AWF, axonal water fraction; BIP, bipolar disorder; BRIA, Bayesian rotational invariant approach; DAX extra, extra-axonal axial diffusivity; DAX intra, intra-axonal axial diffusivity; DTI - AD, diffusion tensor imaging - axial diffusivity; FA, fractional anisotropy; lCST, left cerebrospinal tract; mcSMT, multi-compartment spherical mean technique; MD, mean diffusivity; MDD, major depressive disorder; micro AX, microscopic axial diffusivity; MK, mean kurtosis; OCD, obsessive-compulsive disorder; SCZ, schizophrenia; SMT, spherical mean technique; radEAD, radial extra-axonal diffusivity; diffusivity; RK, radial kurtosis; Vcsf, cerebrospinal fluid fraction; WMTI, white matter tract integrity.

Figure 5

Cross-sectional associations between polygenic risk scores (PRSs) and white matter microstructure (WMM) in longitudinal and cross-sectional validation data. (A) PRS associations of globally averaged WMM for time point 1 in the longitudinal sample (n = 2329) and (B) for time point 2, respectively. (C) Global WMM-PRS associations for the cross-sectional validation sample (n = 31,056). Boxes indicate significance at an uncorrected α < 0.05. For simplicity, standardized regression coefficients with |β| < 0.005 were rounded down to β = 0. (D) PRS associations of regionally averaged WMM for time point 1 in the longitudinal sample and (E) for time point 2. (F) Regional associations for the cross-sectional validation sample. The dotted line in panels (D–F) indicates an uncorrected α < 0.001. Labels presenting the respective regions and metrics are supplied above this α threshold, and |β| > 0.001. All associations were adjusted for age, sex, age × sex as fixed effects, and site as random effect, and none of the associations survived the adjustment of the α level for multiple comparisons. An overview of the utilized diffusion approaches and entailed WMM metrics can be found in Supplemental Appendix B. AD, Alzheimer’s disease; ADC, apparent diffusion coefficient; ADHD, attention-deficit/hyperactivity disorder; AK, axial kurtosis; ANX, anxiety; ASD, autism spectrum disorder; AWF, axonal water fraction; axEAD, axial extra axonal diffusivity; BIP, bipolar disorder; BRIA, Bayesian rotational invariant approach; CSF, cerebrospinal spinal fluid; DAX extra, extra-axonal axial diffusivity; DAX intra, intra-axonal axial diffusivity; DTI - AD, diffusion tensor imaging - axial diffusivity; DKI, diffusion kurtosis imaging; FA, fractional anisotropy; FMIN, Forceps Minor; mcSMT, multi-compartment spherical mean technique; MD, mean diffusivity; MDD, major depressive disorder; micro AX, microscopic axial diffusivity; MK, mean kurtosis; OCD, obsessive-compulsive disorder; radEAD, radial extra-axonal diffusivity; rCING, right Cingulum; rCST, right cerebrospinal tract; RD, radial diffusivity; RK, radial kurtosis; rSLTF, right superior longitudinal temporal fasciculus; rUF, right unicate fasciculus; SCZ, schizophrenia; SMT, spherical mean technique; Vcsf, cerebrospinal fluid fraction; V extra, extra-axonal water fraction; V intra, intra-axonal water fraction; WMTI, white matter tract integrity.