Genome-wide association study unravels mechanisms of brain glymphatic activity

Abstract

Brain glymphatic activity, as indicated by diffusion-tensor imaging analysis along the perivascular space (ALPS) index, is involved in developmental neuropsychiatric and neurodegenerative diseases, but its genetic architecture is poorly understood. Here, we identified 17 unique genome-wide significant loci and 161 candidate genes linked to the ALPS-indexes in a discovery sample of 31,021 individuals from the UK Biobank. Seven loci were replicated in two independent datasets. Genetic signals located at the 2p23.3 locus yielded significantly concordant effects in both young and aging cohorts. Genetic correlation and polygenic overlap analyses indicate a common underlying genetic mechanism between the ALPS-index, ventricular volumes, and cerebrospinal fluid tau levels, with GMNC (3q28) and C16orf95 (16q24.2) as the shared genetic basis. Our findings enhance the understanding of the genetics of the ALPS-index and provide insight for further research into the neurobiological mechanisms of glymphatic clearance activity across the lifespan and its relation to neuropsychiatric phenotypes.

Fig. 1. Schematic drawing of the ALPS-index calculation.

Analysis process from initial diffusion-weighted imaging images to calculate analysis along the perivascular space indexes. Horizontal slices at the ventricle level were selected to display the regions of projection fibers and association fibers, oriented perpendicular to the ventricular body. Color-coded FA map shows the projection fibers (blue) and association fibers (green). Four spherical ROIs, each with a radius of 4 mm, were placed in the bilateral area of the projection fibers (yellow solid circle) and the bilateral area of the association fibers (white solid circle). Then, we obtained ROI probability maps in a standard space, which were binarized into automated ROIs. Diffusivities along the x-, y-, and z-axes were extracted for each automatically binarized ROI. ALPS-index was calculated as the ratio of diffusivities perpendicular to fiber bundles and parallel to veins (Dxproj and Dxassoc) over diffusivities perpendicular to fiber bundles and perpendicular to veins (Dyproj and Dzassoc). ALPS analysis along the perivascular space, DWI diffusion-weighted images, FA fractional anisotropy, MNI Montreal Neurological Institute, ROIasso regions of interest in associative fibers, ROIproj regions of interest in projective fibers.

Fig. 2. SNP-based associations with the mean ALPS-index and functional annotation.

A Manhattan plot of the genome-wide association study (GWAS) results for the mean ALPS-index in the UKB discovery cohort. The plot shows the −log₁₀-transformed P value of each SNP on the y-axis and base-pair positions along the chromosomes on the x-axis (n = 37,327 individuals). P values shown are two-sided. The dashed line indicates genome-wide significance for the GWAS performed in linear additive regression model (P < 5 × 10⁻⁸). Significant loci are marked as yellow squares and labeled by their lead SNP. B Overview of the sizes of genomic risk loci and the number of variants associated with the mean ALPS-index. C Pie charts showing the distribution of functional consequences of SNPs in LD with genome-wide significant lead SNPs in the discovery GWAS of the mean ALPS-index, the minimum chromatin state across 127 tissue and cell types, and the distribution of RegulomeDB score (a categorical score between 1a and 7, indicating biological evidence of a SNP being a regulatory element, with a low score denoting a higher likelihood of a SNP being regulatory). SNP single nucleotide polymorphism.

Fig. 3. Genetic effect on the ALPS-index and annual rates of the ALPS-index across the lifespan.

A Fixed-effects meta-analysis of rs4665946 (2:27347931) in locus 2p23.3 for the mean ALPS-index. B and C Meta-regression of SNP effects according to age was conducted using linear and quadratic models. P values shown are two-sided and uncorrected for multiple comparisons. Plots show the results for two variants that passed the FDR correction. The x-axis represents the median age of each cohort and the y-axis denotes the genetic effect on the mean ALPS-index. The dot size indicates the −log10 p-value of the genetic effect on the mean ALPS-index. The error band around the regression line indicates the 95% CI. D Forest plots show genetic effects of rs6012259 (20:46374885) at locus 20q13.12 and rs12146713 (12:106476805) at locus 12q23.3. CI confidence interval, SNP single nucleotide polymorphism, mALPS mean ALPS-index.

Fig. 4. Genetic correlations of identified ALPS-index loci.

A Genetic correlation between the mean ALPS-index and brain volumes. All regions that are genetically correlated with the mean ALPS-index at nominal P < 0.05 are colored based on the magnitude of positive (red) or negative (blue) genetic correlation estimates. Regions with genetic correlation estimates that survived false discovery rate corrections are annotated with text. B Volcano plot displays the genetic correlation estimates (x-axis) and −log₁₀ (p-value) (y-axis) between the mean ALPS-index and white matter microstructures. The top five significant white matter microstructures were labeled in the figure. C A bubble plot shows the genetic correlation between three ALPS-indexes and putative risk factors, cognitive measures, and other MRI markers of brain aging. The dot color represents the magnitude of positive (red) or negative (blue) genetic correlation, while the dot size denotes the −log₁₀(P) of the genetic correlation. Correlation estimates with nominal P < 0.05 are annotated with dashed circles, and those that survived false discovery rate corrections are annotated with solid circles. The genetic correlations were estimated using LDSC. All p-values shown are two-tailed and are original p-values without multiple comparisons. AD Alzheimer’s disease, ADHD attention-deficit/hyperactivity disorder, ASD autism spectrum disorder, BG-PVS perivascular space in basal ganglia, BP bipolar disorder, cSVD cerebral small vessel disease, FTD frontotemporal dementia, BMI body mass index, CSF, cerebrospinal fluid, DBP diastolic blood pressure, FA fractional anisotropy, FDR false discovery rate, GC genetic correlation, HIP-PVS perivascular space in hippocampus, MRI magnetic resonance imaging, SBP systolic blood pressure, PP pulse pressure, WM-PVS perivascular space in white matter, WMH white matter hyperintensities.

Fig. 5. Polygenetic overlap between ALPS-indexes and neuropsychiatric and neurological disorders and biomarkers.

A Enhanced discovery of genetic loci for the three ALPS-indexes conditioned on neuropsychiatric and neurological disorders and biomarkers by the condFDR analysis. ConjFDR analysis detected shared genetic loci across the three ALPS-indexes and neuropsychiatric disorders and biomarkers. B Conditional Q–Q plots for the mean ALPS-index conditioned on schizophrenia, CSF pTau, HIP-PVS, WMH, and vice versa, demonstrating genetic overlap. Conditional Q–Q plots of nominal versus empirical −log10 p-values (corrected for inflation) in one phenotype below the standard GWAS threshold of p < 5 × 10⁻⁸ as a function of significance of association with another phenotype, at the level of p ⩽ 0.1, p ⩽ 0.01, p ⩽ 0.001, respectively. The blue lines indicate all SNPs. The dashed lines indicate the null hypothesis. AD Alzheimer’s disease, ADHD attention-deficit/hyperactivity disorder, ASD autism spectrum disorder, BG-PVS perivascular space in basal ganglia, BP bipolar disorder, bvFTD behavioral variant frontotemporal dementia, CSF cerebrospinal fluid, HIP-PVS perivascular space in hippocampus, mALPS mean ALPS-index, SCZ schizophrenia, WM-PVS perivascular space in white matter, WMH white matter hyperintensities.

Fig. 6. ALPS-index associated genes identified by GWGAS and gene mapping strategies.

A Manhattan plot of the GWGAS on the mean ALPS-index (n = 37,327 individuals). P values were computed by MAGMA gene-based test. The y-axis shows the −log10 p-value of each gene. The x-axis shows the chromosomal position (start position). The dashed line indicates the Bonferroni-corrected threshold for genome-wide significance of the gene-based test (P < 2.65 × 10⁻⁶; 0.05/18,878). B Venn diagram showing the number of genes associated with any of the three ALPS-indexes (mean, left, and right) implicated by positional mapping, eQTL mapping, chromatin interaction mapping, and GWGAS. C Upset plot displays the number of unique and shared genes among the three ALPS-indexes (mean, left, and right). eQTL expression quantitative trait loci, GWGAS genome-wide gene-based association analysis.