Validation of polygenic scores for longitudinal changes in brain structures

862 

Abstract Submission 

Jalmar Teeuw¹, Rachel Brouwer², Shotaro Hato¹, Sonja de Zwarte¹, Sophia Thomopoulos³, Neda Jahanshad³, Paul Thompson³, Hilleke Hulshoff Pol⁴

¹Department of Psychiatry, University Medical Center Utrecht, Utrecht, Netherlands, ²Department of Complex Trait Genetics, Center for Neurogenomics and Cognitive Research, Amsterdam, Netherlands, ³Keck School of Medicine, University of Southern California, Los Angeles, California, United States, ⁴Department of Experimental Psychology, Helmholtz Institute, Utrecht University, Utrecht, Netherlands

Jalmar Teeuw, Dr.  
Department of Psychiatry, University Medical Center Utrecht
Utrecht, Netherlands

Rachel Brouwer  
Department of Complex Trait Genetics, Center for Neurogenomics and Cognitive Research
Amsterdam, Netherlands

Shotaro Hato  
Department of Psychiatry, University Medical Center Utrecht
Utrecht, Netherlands

Sonja de Zwarte  
Department of Psychiatry, University Medical Center Utrecht
Utrecht, Netherlands

Sophia Thomopoulos  
Keck School of Medicine, University of Southern California
Los Angeles, California, United States

Neda Jahanshad  
Keck School of Medicine, University of Southern California
Los Angeles, California, United States

Paul Thompson  
Keck School of Medicine, University of Southern California
Los Angeles, California, United States

Hilleke Hulshoff Pol  
Department of Experimental Psychology, Helmholtz Institute, Utrecht University
Utrecht, Netherlands

Longitudinal changes in brain structure are phenotypically and genetically related to neuropsychiatric disorders, and may be predictive of the onset for some (Brans et al., 2008; van Haren et al., 2008; Teeuw et al, 2021). Recently, we reported on genetic variants driving brain development and brain aging in a genome-wide association study (GWAS) of changes in multiple brain volumes measured by longitudinal magnetic resonance imaging (MRI) from the ENIGMA consortium (Brouwer et al., 2022). Here we report on the validity and predictive value of polygenic scores (PGS) derived from this study.

We validated PGS derived from the GWAS in three cohorts: ABCD (N=2523; ages 9–11 years; Barch et al., 2018), UK Biobank (N=2536; ages 46–80 years; Sudlow et al., 2015), and UMCU (N=322; ages 10–65 years; 21% patients with schizophrenia and bipolar disorder). Change rates for 15 brain structures were obtained from longitudinal MRI using FreeSurfer (Fischl, 2012). Genotyped DNA was processed into PGS using PRSice-2 and PRScs (Choi and O'Reilly, 2019; Ge et al., 2019). The GWAS analysis was repeated three times leaving out each of the cohorts in turn to avoid overlap between the discovery and application sample while minimizing the impact on the power of the GWAS. Associations between PGS and rate of brain changes were determined by linear mixed effects model to derive percentage of variance of brain volume change explained by the PGS (ΔR²). Models were adjusted for sex, age, overall head size, and population stratification.

PGSs were significantly and largely positively associated with their respective changes in ABCD (ΔR² up to 0.45%; p_min=6.60E-04), UK Biobank (ΔR² up to 0.40%; p_min=1.35E-03), and UMCU (ΔR² up to 1.65%; p_min=7.43E-03) (Figure 1). The PGS for longitudinal change rate in lateral ventricles and putamen volume were consistently found across all three cohorts. Higher PGS for lateral ventricles was associated with accelerated increase in volume of the lateral ventricles in ABCD (ꞵ=+0.068, p=6.60E–04), UK Biobank (ꞵ=+0.044, p=1.24E–02), and UMCU cohorts (ꞵ=+0.137, p=7.43E–03) across the lifespan. In contrast, the association of the putamen inverts from a negative association – i.e, higher PGS predicted lower rate of change/decrease in volume of the putamen – in the childhood and adolescent ABCD cohort (ꞵ=​​–0.045, p=2.70E–02), to a positive association in the on average older participants in UK Biobank (ꞵ=​​+0.063, p=1.35E–03) and UMCU cohorts (ꞵ=​​+0.115, p=3.67E–02). Further investigation is needed to determine if PGS that are significant in a cohort-specific manner indicate possible associations with brain development and aging across the lifespan or disease-specific associations.
Although results between the PRSice-2 and PRScs method are similar in the larger ABCD (ICC(1)=+0.593, p=6.47E–03) and UK Biobank cohort (ICC(1)=+0.525, p=1.61E–02), there is still considerable variation between the two methods in the UMCU cohorts (ICC(1)=–0.252, p=0.829 [n.s.]). Despite previous reports (Ni et al., 2021), the modern PRScs method did not always outperform the traditional clumping and threshold approach from PRSice-2 in this study.

Polygenic scores for longitudinal changes in brain structures have now been validated and can be used to assess associations with other traits, such as (risk for) neuropsychiatric disorders and cognitive functioning.

Genetic Association Studies ¹

Aging ²

Lifespan Development Other

Other - longitudinal brain imaging; polygenic scores; gray/white matter brain volumes; brain development and aging