Human plasma metabolomics in age-related macular degeneration (AMD) using nuclear magnetic resonance spectroscopy

Abstract

Purpose: To differentiate the plasma metabolomic profile of patients with age related macular degeneration (AMD) from that of controls, by Nuclear Magnetic Resonance (NMR) spectroscopy.

Methods: Two cohorts (total of 396 subjects) representative of central Portugal and Boston, USA phenotypes were studied. For each cohort, subjects were grouped according to AMD stage (early, intermediate and late). Multivariate analysis of plasma NMR spectra was performed, followed by signal integration and univariate analysis.

Results: Small changes were detected in the levels of some amino acids, organic acids, dimethyl sulfone and specific lipid moieties, thus providing some biochemical information on the disease. The possible confounding effects of gender, smoking history and age were assessed in each cohort and found to be minimal when compared to that of the disease. A similar observation was noted in relation to age-related comorbidities. Furthermore, partially distinct putative AMD metabolite fingerprints were noted for the two cohorts studied, reflecting the importance of nutritional and other lifestyle habits in determining AMD metabolic response and potential biomarker fingerprints. Notably, some of the metabolite changes detected were noted as potentially differentiating controls from patients diagnosed with early AMD.

Conclusion: For the first time, this study showed metabolite changes in the plasma of patients with AMD as compared to controls, using NMR. Geographical origins were seen to affect AMD patients´ metabolic profile and some metabolites were found to be valuable in potentially differentiating controls from early stage AMD patients. Metabolomics has the potential of identifying biomarkers for AMD, and further work in this area is warranted.

Fig 1. Representative ¹H NMR spectra of control plasma.

500 MHz ¹H NMR spectra of blood plasma from a control subject: a) standard 1D spectrum; b) CPMG spectrum; c) diffusion-edited spectrum. Signal assignment: 1-lactate; 2-alanine; 3 -glutamine; 4-glucose; 5-isoleucine; 6-leucine; 7-valine; 8-lysine; 9-acetate; 10-pyruvate; 11-citrate; 12-creatine; 13-creatinine; 14-dimethyl sulfone; 15-TMAO, trimethylamine-N-Oxide; 16,proline; 17-methanol; 18-glycine; 19-tyrosine; 20-histidine; 21- phenylalanine; 22-formate; 23-C18H cholesterol; 24-CH₃ lipids; 25-(CH₂)_n lipids; 26-CH₂CH₂CO lipids; 27-CH₂CH₂C = C lipids; 28-CH₂C = C lipids; 29-CH₂CO lipids; 30-C = CCH₂CH = C lipids; 31-albumin lysil groups; 32-N(CH₃)₃ choline; 33-glyceryl C1,3H; 34-glyceryl C1,3H’; 35-glyceryl C2H; 36-HC = CH lipids; 37-NH protein region.

Fig 2. Examples of PLS-DA score plots.

PLS-DA scores scatter plots and MCCV quality parameters (pairwise model Q², Q²_median (obtained through MCCV), % CR, % sens. and % spec.) obtained for variable selected CPMG NMR spectra of late AMD patients vs controls, in the a) Coimbra cohort: late AMD patients (□, n = 32), controls (∎, n = 42) and b) Boston cohort: late AMD patients (◇, n = 38), controls (♦, n = 40).

Fig 3. Effect size plots for CPMG spectra integrals.

Effect size (E.S.) plots for resonances varying in the CPMG NMR spectra across AMD evolution through different severity stages in the a) Coimbra and b) Boston cohorts. Resonances are listed alphabetically within each compound family (amino acids, organic acids, other low-M_w compounds and lipids). The dashed horizontal line refers to null E.S. and the length of the vertical segments corresponds to E.S. range. E.S. segments not intercepting the null E.S. line are considered as relevant variations (shaded rectangles). F.A.: fatty acids.

Fig 4. Boxplot graphs for metabolites varying in Coimbra cohort.

Coimbra cohort: boxplot representations of the metabolite variations found statistically relevant (* indicates p-value < 0.05) in at least one pairwise PLS-DA model. Compound names in rectangles correspond to compounds differentiating between controls and early AMD patients. C: controls, E: early AMD, I: intermediate AMD, L: late AMD. F.A.: fatty acids.

Fig 5. Boxplot graphs for metabolites varying in Boston cohort.

Boston cohort: boxplot representations of the metabolite variations found statistically relevant (* indicates p-value < 0.05) in at least one pairwise PLS-DA model. Compound names in rectangles correspond to compounds differentiating between controls and early AMD patients. C: controls, E: early AMD, I: intermediate AMD, L: late AMD. F.A.: fatty acids.