Human Plasma Metabolomics in Age-Related Macular Degeneration: Meta-Analysis of Two Cohorts

Abstract

The pathogenesis of age-related macular degeneration (AMD), a leading cause of blindness worldwide, remains only partially understood. This has led to the current lack of accessible and reliable biofluid biomarkers for diagnosis and prognosis, and absence of treatments for dry AMD. This study aimed to assess the plasma metabolomic profiles of AMD and its severity stages with the ultimate goal of contributing to addressing these needs. We recruited two cohorts: Boston, United States (n = 196) and Coimbra, Portugal (n = 295). Fasting blood samples were analyzed using ultra-high performance liquid chromatography mass spectrometry. For each cohort, we compared plasma metabolites of AMD patients versus controls (logistic regression), and across disease stages (permutation-based cumulative logistic regression considering both eyes). Meta-analyses were then used to combine results from the two cohorts. Our results revealed that 28 metabolites differed significantly between AMD patients versus controls (false discovery rate (FDR) q-value: 4.1 × 10^-2-1.8 × 10^-5), and 67 across disease stages (FDR q-value: 4.5 × 10^-2-1.7 × 10^-4). Pathway analysis showed significant enrichment of glycerophospholipid, purine, taurine and hypotaurine, and nitrogen metabolism (p-value ≤ 0.04). In conclusion, our findings support that AMD patients present distinct plasma metabolomic profiles, which vary with disease severity. This work contributes to the understanding of AMD pathophysiology, and can be the basis of future biomarkers and precision medicine for this blinding condition.

Keywords: age-related macular degeneration; mass spectrometry; metabolomics.

Figure 1

Overview of the study. AMD: age-related macular degeneration; AMD/Control model multivariable logistic regression model considering AMD versus controls as the outcome; Stage+2 Eye: permutation-based cumulative logistic regression model considering both eyes of each patient and the severity stage of disease as the outcome (control, early, intermediate and late); ROC: receiving operating characteristic; n: number.

Figure 2

List of the metabolites differing significantly (q < 0.05) between AMD patients (AMD/Control model), and controls, and across AMD stages of both eyes (Stage + 2Eye model). AMD: age-related macular degeneration; CTL: control; ERL: early AMD; INT: intermediate AMD; LAT: late AMD. Box plots for the most statistically significant metabolites of each analysis are presented. For each box plot, yellow dots represent the mean and black horizontal lines represent the median; AMD/Control model: multivariable logistic regression model considering AMD versus controls as the outcome; Stage + 2Eye: permutation-based cumulative logistic regression model considering both eyes of each patient and the severity stage of disease as the outcome (control, early, intermediate, and late).

Figure 3

Pathway analysis of the metabolites differing significantly based on q-values (a) between AMD patients and controls or (b) across stages (controls, early AMD, intermediate AMD, and late AMD), identified by meta-analysis of the results of two study cohorts; -log(p): logarithm of the p-value.

Figure 4

Receiving operating characteristic (ROC) curve analysis from the meta-analysis of the results of the two studies. In red, the baseline model including demographic covariates alone; in orange, the model including demographic covariates plus the metabolites selected by elastic net regression with all metabolites; in green, the model including demographic covariates plus the significant metabolites identified by AMD/Control model; in blue, the model including demographic covariates plus the significant metabolites identified by Stage+2Eye model. AUC – area under the curve; CI – confidence interval; AMD – age-related macular degeneration; Stage+2 Eye – permutation-based cumulative logistic regression model considering both eyes of each patient and the severity stage of disease as the outcome (control, early, intermediate and late); AMD/Control - multivariable logistic regression model considering AMD vs controls as the outcome; All-Met+EN – elastic net regression model with all metabolites; Baseline - statistical model only considering demographic covariates alone.