Genetic study of multimodal imaging Alzheimer’s disease progression score implicates novel loci

Marzia A. Scelsi, Raiyan R. Khan, Marco Lorenzi, Leigh Christopher, Michael D. Greicius, Jonathan M. Schott, Sébastien Ourselin, André AltmannBrain 2018 · DOI

TL;DR

This study introduces a novel "disease progression score" (DPS) derived from multimodal brain imaging (amyloid PET and hippocampal volume) to better capture Alzheimer's disease (AD) progression than single-modality measures. A genome-wide association study (GWAS) using this DPS identified a novel genetic locus, LCORL, associated with AD progression, with the minor allele showing a protective effect against conversion to AD.

What Problem Does This Paper Address?

Alzheimer's disease is a complex neurodegenerative disorder with a long prodromal phase. Traditional genetic studies often rely on simple case-control classifications, which can be imprecise and miss genetic influences on the subtle, continuous changes that occur during disease development. While imaging biomarkers like amyloid burden and hippocampal atrophy are valuable, studying them in isolation might not fully capture the dynamic, multi-faceted nature of AD progression. This paper aims to address this by developing a more integrated, quantitative phenotype (the DPS) that combines information from multiple imaging modalities over time, and then using this to identify novel genetic risk factors for AD progression.

Methods

The core innovation is the Disease Progression Score (DPS). This score is derived using the GRACE algorithm, which models the longitudinal trajectories of multiple biomarkers (here, cortical amyloid burden from PET and bilateral hippocampal volume from MRI) to estimate a subject's position along a population-level disease progression curve. The DPS is essentially a time-shift parameter that aligns an individual's biomarker data with the population model, effectively quantifying how far along the disease continuum they are.

This DPS was then used as a quantitative trait in a genome-wide association study (GWAS). Standard GWAS procedures were followed, including imputation and quality control of genetic data, and controlling for relevant covariates (age, sex, APOE4 status, education, and principal components of population structure).

To validate the findings, a competing risks regression analysis was performed on an independent cohort to assess the effect of the top identified genetic variants on the risk of conversion from healthy aging or mild cognitive impairment (MCI) to AD, accounting for death as a competing event.

Datasets

  • Alzheimer's Disease Neuroimaging Initiative (ADNI) database: This was the primary dataset used for developing the DPS and conducting the discovery GWAS. It includes longitudinal imaging (PET and MRI) and genetic data from subjects with AD, MCI, and healthy controls.
  • National Alzheimer's Coordinating Center (NACC) database: This independent dataset was used for the validation of the top genetic findings, specifically for assessing the effect of the identified SNP on clinical conversion risk.

Key Insights and Ideas

  • Multimodal DPS as a Superior Phenotype: The DPS, by integrating information from amyloid burden and hippocampal volume over time, provides a more sensitive and comprehensive measure of AD progression than single-modality biomarkers. It better separates diagnostic groups and shows stronger associations with genetic risk scores.
  • Novel Locus for AD Progression: The study identified a genome-wide significant association with the LCORL gene (specifically SNP rs6850306) on chromosome 4. This is a novel finding for AD progression, suggesting a new genetic pathway involved.
  • Protective Effect of LCORL Minor Allele: The minor allele of rs6850306 was found to be protective against conversion from MCI to AD, and also associated with a lower DPS. This suggests a potential therapeutic target or a marker for slower disease progression.
  • LCORL as an Expression Quantitative Trait Locus (eQTL): The identified SNP rs6850306 acts as a cis-eQTL for LCORL in hippocampal tissue, meaning it influences the gene's expression levels. This provides a biological mechanism linking the genetic variant to brain function.
  • Integrative Approach Potential: The paper highlights the power of combining multimodal imaging with genetic studies to uncover complex disease mechanisms, moving beyond traditional case-control designs.

Key Results

  • A genome-wide significant association was found for the DPS with SNP rs6850306 on chromosome 4, with a P-value of 1.03 × 10⁻⁸.
  • The minor allele of rs6850306 was found to be protective against conversion from MCI to AD in the NACC cohort (Hazard Ratio = 0.593, 95% CI = 0.387–0.907, PBonf = 0.032).
  • This protective effect aligns with the negative direction of effect observed in the GWAS (βdisease progression score = -0.07 ± 0.01).
  • rs6850306 was identified as a cis-eQTL for LCORL in hippocampal tissue.
  • No genome-wide significant associations were found for hippocampal volume or cortical amyloid alone.

Limitations and Future Work

  • Biomarker Selection: The DPS relies on the choice of biomarkers. While amyloid and hippocampal volume are key for AD, using a larger or different set of biomarkers could potentially yield more granular estimates and increase statistical power.
  • Validation Cohort Size: The authors acknowledge the difficulty in validating multimodal phenotypes in independent cohorts due to the need for comprehensive genotyping and longitudinal imaging data. A formal replication cohort was not secured.
  • Imaging Biomarker Variation: Imaging-derived phenotypes can introduce variability. The discrepancy in findings for a previously identified amyloid locus (rs509208) highlights the need for robust imaging biomarkers like the DPS and standardized pipelines.
  • Fine-mapping and Functional Studies: While the study implicates LCORL, further fine-mapping of the genomic region and functional studies (including animal models) are needed to definitively identify the causal gene and mechanism.
  • Future Directions: The authors suggest that this integrative approach has significant potential for identifying genetic drivers of complex diseases by combining multimodal imaging with genetic association studies. They advocate for further exploration of multimodal phenotypes in GWAS.

Summary updated

Abstract

Identifying genetic risk factors underpinning different aspects of Alzheimer's disease has the potential to provide important insights into pathogenesis. Moving away from simple case-control definitions, there is considerable interest in using quantitative endophenotypes, such as those derived from imaging as outcome measures. Previous genome-wide association studies of imaging-derived biomarkers in sporadic late-onset Alzheimer's disease focused only on phenotypes derived from single imaging modalities. In contrast, we computed a novel multi-modal neuroimaging phenotype comprising cortical amyloid burden and bilateral hippocampal volume. Both imaging biomarkers were used as input to a disease progression modelling algorithm, which estimates the biomarkers' long-term evolution curves from population-based longitudinal data. Among other parameters, the algorithm computes the shift in time required to optimally align a subjects' biomarker trajectories with these population curves. This time shift serves as a disease progression score and it was used as a quantitative trait in a discovery genome-wide association study with n = 944 subjects from the Alzheimer's Disease Neuroimaging Initiative database diagnosed as Alzheimer's disease, mild cognitive impairment or healthy at the time of imaging. We identified a genome-wide significant locus implicating LCORL (rs6850306, chromosome 4; P = 1.03 × 10-8). The top variant rs6850306 was found to act as an expression quantitative trait locus for LCORL in brain tissue. The clinical role of rs6850306 in conversion from healthy ageing to mild cognitive impairment or Alzheimer's disease was further validated in an independent cohort comprising healthy, older subjects from the National Alzheimer's Coordinating Center database. Specifically, possession of a minor allele at rs6850306 was protective against conversion from mild cognitive impairment to Alzheimer's disease in the National Alzheimer's Coordinating Center cohort (hazard ratio = 0.593, 95% confidence interval = 0.387-0.907, n = 911, PBonf = 0.032), in keeping with the negative direction of effect reported in the genome-wide association study (βdisease progression score = -0.07 ± 0.01). The implicated locus is linked to genes with known connections to Alzheimer's disease pathophysiology and other neurodegenerative diseases. Using multimodal imaging phenotypes in association studies may assist in unveiling the genetic drivers of the onset and progression of complex diseases.
BibTeX
@inproceedings{scelsi2018genetic,
  author = {Marzia A. Scelsi and Raiyan R. Khan and Marco Lorenzi and Leigh Christopher and Michael D. Greicius and Jonathan M. Schott and Sébastien Ourselin and André Altmann},
  title = {{Genetic study of multimodal imaging Alzheimer’s disease progression score implicates novel loci}},
  booktitle = {Brain},
  year = {2018},
  doi = {10.1093/brain/awy141},
}
Figure