Serum metabolites associated with dietary protein intake: results from the Modification of Diet in Renal Disease (MDRD) randomized clinical trial

Abstract

Background: Accurate assessment of dietary intake is essential, but self-report of dietary intake is prone to measurement error and bias. Discovering metabolic consequences of diets with lower compared with higher protein intake could elucidate new, objective biomarkers of protein intake.

Objectives: The goal of this study was to identify serum metabolites associated with dietary protein intake.

Methods: Metabolites were measured with the use of untargeted, reverse-phase ultra-performance liquid chromatography-tandem mass spectrometry quantification in serum specimens collected at the 12-mo follow-up visit in the Modification of Diet in Renal Disease (MDRD) Study from 482 participants in study A (glomerular filtration rate: 25-55 mL · min-1 · 1.73 m-2) and 192 participants in study B (glomerular filtration rate: 13-24 mL · min-1 · 1.73 m-2). We used multivariable linear regression to test for differences in log-transformed metabolites (outcome) according to randomly assigned dietary protein intervention groups (exposure). Statistical significance was assessed at the Bonferroni-corrected threshold: 0.05/1193 = 4.2 × 10-5.

Results: In study A, 130 metabolites (83 known from 28 distinct pathways, including 7 amino acid pathways; 47 unknown) were significantly different between participants randomly assigned to the low-protein diet compared with the moderate-protein diet. In study B, 32 metabolites (22 known from 8 distinct pathways, including 4 amino acid pathways; 10 unknown) were significantly different between participants randomly assigned to the very-low-protein diet compared with the low-protein diet. A total of 11 known metabolites were significantly associated with protein intake in the same direction in both studies A and B: 3-methylhistidine, N-acetyl-3-methylhistidine, xanthurenate, isovalerylcarnitine, creatine, kynurenate, 1-(1-enyl-palmitoyl)-2-arachidonoyl-GPE (P-16:0/20:4), 1-(1-enyl-stearoyl)-2-arachidonoyl-GPE (P-18:0/20:4), 1-(1-enyl-palmitoyl)-2-arachidonoyl-GPC (P-16:0/20:4), sulfate, and γ-glutamylalanine.

Conclusions: Among patients with chronic kidney disease, an untargeted serum metabolomics platform identified multiple pathways and metabolites associated with dietary protein intake. Further research is necessary to characterize unknown compounds and to examine these metabolites in association with dietary protein intake among individuals without kidney disease.This trial was registered at clinicaltrials.gov as NCT03202914.

Keywords: chronic kidney disease; diet modification; metabolic pathways; metabolites; protein.

© 2019 American Society for Nutrition.

Figures

FIGURE 1

Scatterplot of P values for the association between known metabolites and dietary protein intake according to pathway in study A. The horizontal line represents the Bonferroni-adjusted threshold for statistical significance. P values were calculated from multivariable linear regression models testing for differences in log-transformed metabolites (outcome) according to randomly assigned intervention groups for dietary protein (exposure), with adjustment for age, sex, race, log-transformed glomerular filtration rate measured by urinary clearance of 125I-iothalamate, renal diagnosis, randomly assigned intervention group for blood pressure control, and BMI in study A (n = 482).

FIGURE 2

Scatterplot of P values and β coefficients for the association between metabolites and dietary protein intake in study A. The horizontal line represents the Bonferroni-adjusted threshold for statistical significance. P values and β coefficients were calculated from multivariable linear regression models testing for differences in log-transformed metabolites (outcome) according to randomly assigned intervention groups for dietary protein (exposure), with adjustment for age, sex, race, log-transformed glomerular filtration rate measured by urinary clearance of 125I-iothalamate, renal diagnosis, randomly assigned intervention group for blood pressure control, and BMI in study A (n = 482).

FIGURE 3

Scatterplot of P values for the association between known metabolites and dietary protein intake according to pathway in study B. The horizontal line represents the Bonferroni-adjusted threshold for statistical significance. P values were calculated from multivariable linear regression models testing for differences in log-transformed metabolites (outcome) according to randomly assigned intervention groups for dietary protein (exposure), with adjustment for age, sex, race, log-transformed glomerular filtration rate measured by urinary clearance of 125I-iothalamate, renal diagnosis, randomly assigned intervention group for blood pressure control, and BMI in study B (n = 192).

FIGURE 4

Scatterplot of P values and β coefficients for the association between metabolites and dietary protein intake in study B. The horizontal line represents the Bonferroni-adjusted threshold for statistical significance. P values and β coefficients were calculated from multivariable linear regression models testing for differences in log-transformed metabolites (outcome) according to randomly assigned intervention groups for dietary protein (exposure), with adjustment for age, sex, race, log-transformed glomerular filtration rate measured by urinary clearance of 125I-iothalamate, renal diagnosis, randomly assigned intervention group for blood pressure control, and BMI in study B (n = 192).

FIGURE 5

Top 15 metabolites in principal component 1 in study A. VIP scores are calculated as the weighted sum of squares of the partial-least-squares loadings for each component via partial-least-squares–discriminant analysis in study A (n = 482). *Compounds for which the identity is known but which have not been confirmed based on a standard. A black box for moderate protein (and white box for low protein) indicates that serum levels of the metabolite were higher between those randomly assigned to the moderate- compared with the low-protein intervention. A white box for moderate protein (and black box for low protein) indicates that serum levels of the metabolite were lower between those randomly assigned to the moderate- compared with the low-protein intervention. VIP, variable importance in projection.

FIGURE 6

Top 15 metabolites in principal component 1 in study B. VIP scores are calculated as the weighted sum of squares of the partial-least-squares loadings for each component via partial-least-squares–discriminant analysis in study B (n = 192). *Compounds for which the identity is known but which have not been confirmed based on a standard. A black box for very low protein (and white box for low protein) indicates that serum levels of the metabolite were higher between those randomly assigned to the very-low- compared with the low-protein intervention. A white box for very low protein (and black box for low protein) indicates that serum levels of the metabolite were lower between those randomly assigned to the very-low- compared with the low-protein intervention. VIP, variable importance in projection.