Decreased Circulating Very Small Low-Density Lipoprotein is Likely Causal for Age-Related Macular Degeneration

Abstract

Objective: Abnormal changes in metabolite levels in serum or plasma have been highlighted in several studies in age-related macular degeneration (AMD), the leading cause of irreversible vision loss. Specific changes in lipid profiles are associated with an increased risk of AMD. Metabolites could thus be used to investigate AMD disease mechanisms or incorporated into AMD risk prediction models. However, whether particular metabolites causally affect the disease has yet to be established.

Design: A 3-tiered analysis of blood metabolites in the United Kingdom (UK) Biobank cohort to identify metabolites that differ in AMD patients with evidence for a putatively causal role in AMD.

Participants: A total of 72 376 donors from the UK Biobank cohort including participants with AMD (N = 1353) and non-AMD controls (N = 71 023).

Methods: We analyzed 325 directly measured or derived blood metabolites from the UK Biobank for 72 376 donors to identify AMD-associated metabolites. Genome-wide association studies for 325 metabolites in 98 316 European participants from the UK Biobank were performed. The causal effects of these metabolites in AMD were tested using a 2-sample Mendelian randomization approach. The predictive value of these measurements together with sex and age was assessed by developing a machine learning classifier.

Main outcome measures: Evaluating metabolic biomarkers associated with AMD susceptibility and investigating their potential causal contribution to the development of the disease.

Results: This study noted age to be the prominent risk factor associated with AMD development. While accounting for age and sex, we identified 84 metabolic markers as significantly (false discovery rate-adjusted P value < 0.05) associated with AMD. Lipoprotein subclasses comprised the majority of the AMD-associated metabolites (39%) followed by several lipoprotein to lipid ratios. Nineteen metabolites showed a likely causative role in AMD etiology. Of these, 6 lipoproteins contain very small, very low-density lipoprotein (VLDL), and phospholipids to total lipid ratio in medium VLDL. Based on this we postulate that depletion of circulating very small VLDLs is likely causal for AMD. The risk prediction model constructed from the metabolites, age and sex, identified age as the primary predictive factor with a much smaller contribution by metabolites to AMD risk prediction.

Conclusions: This study underscores the pronounced role of lipids in AMD susceptibility and the likely causal contribution of particular subclasses of lipoproteins to AMD. Our study provides valuable insights into the metabopathological mechanisms of AMD disease development and progression.

Keywords: Age-related macular degeneration; Circulating metabolites; Genome-wide association studies; Mendelian randomization.

Figure 1

Study design schematic. Colored boxes in yellow, blue, and green present our approach to answering questions (1), (2), and (3) (see text), respectively. AMD = age-related macular degeneration; GWAS = genome-wide association studies; UK = United Kingdom.

Figure 4

Metabolites that are significantly associated with age-related macular degeneration (AMD) in patients versus non-AMD controls. A circular bar plot depicting 84 metabolites, negatively or positively associated with AMD after accounting for sex and age. Only metabolites meeting a false discovery rate-corrected significance level (adjusted P value < 0.05) are displayed. Gray circles denote the circularized y-axis. Bars pointing outward of the black circle denote increased levels in AMD (effect >0), whereas bars pointing inward of the black circle indicate decreased levels in AMD (effect <0). Each estimate represents the difference in the outcome variable (i.e. AMD risk) per standard deviation increase/decrease in the scaled-transformed metabolite values. The positive and negative effect size translates to metabolic increases and decreases, respectively, in individuals with AMD in comparison to the control group. Colors indicate groups and subgroups of metabolites separated by an empty space between bars. Groups and subgroups of metabolites are described in Table S2. The prefixes indicate the size of particles: L, large; M, medium; S, small; XS, very small. The metabolite abbreviations are described in Table S1.

Figure 10

Mendelian randomization (MR) analyses on 325 metabolites using 3 MR methods. Significant metabolites resulting from corrected P values (false discovery rate [FDR] <0.05) and an effect size >0.5 (b > 0.5 or b < −0.5) are labeled. Colors represent 21 groups of metabolites. Only metabolite groups with ≥1 member with an absolute effect >0.5 are included in the legend. Positive b values indicate potential causal factors where higher metabolite plasma concentrations increase age-related macular degeneration (AMD) disease risk. Negative b values indicate metabolites where lower concentrations increase AMD risk.

Figure 16

Metabolites that are both disease biomarkers and likely causal for age-related macular degeneration (AMD). (A) A Venn diagram summarizing the findings in the 2 facets (association and Mendelian randomization [MR] analyses) of this study. A total of 84 biomarkers demonstrated statistical significance (false discovery rate [FDR] < 0.05) in the association analysis (facet 1). Nineteen metabolites reached the significance threshold (FDR < 0.05) across all 3 MR methods of which 7 biomarkers exhibited lower levels in individuals with AMD. These 7 metabolites exhibited lower levels in AMD patient plasma, in addition to having causal effects that suggest lower genetically predicted levels of these metabolites appear to increase AMD risk. (B) Effect size comparisons between AMD-risk single nucleotide polymorphisms (SNPs) and SNPs associated with metabolite levels. Each dot represents a SNP. Each panel represents a metabolite and contains the SNPs that were found to have a genome-wide significant effect on that metabolite. The x-axis captures the effect sizes of the SNPs on a given metabolite while the y-axis captures the effect of the same SNPs on AMD risk. The blue line represents a regression line between the two. Metabolites causally affecting the disease are expected to have correlated effect sizes. Metabolite descriptions as follows: XS_VLDL_PL: phospholipids in very small very low-density lipoprotein (VLDL); XS_VLDL_P: concentration of very small VLDL particles; XS_VLDL_L: total lipids in very small VLDL; XS_VLDL_FC: free cholesterol in very small VLDL; XS_VLDL_CE: cholesteryl esters in very small VLDL; XS_VLDL_C: cholesterol in very small VLDL; M_VLDL_PL_pct: phospholipids to total lipids in medium VLDL percentage.