Visual Abstract
Keywords: AASK (African American Study of Kidney Disease and Hypertension); chronic kidney disease; metabolism; glomerular filtration rate; creatinine; Cross-Sectional Studies; phosphatidylethanolamine; Linear Models; risk factors; proteinuria; Renal Insufficiency, Chronic; hypertension; Diet
Abstract
Background and objectives
Data are scarce on blood metabolite associations with proteinuria, a strong risk factor for adverse kidney outcomes. We sought to investigate associations of proteinuria with serum metabolites identified using untargeted profiling in populations with CKD.
Design, setting, participants, & measurements
Using stored serum samples from the African American Study of Kidney Disease and Hypertension (AASK; n=962) and the Modification of Diet in Renal Disease (MDRD) study (n=620), two rigorously conducted clinical trials with per-protocol measures of 24-hour proteinuria and GFR, we evaluated cross-sectional associations between urine protein-to-creatinine ratio and 637 known, nondrug metabolites, adjusting for key clinical covariables. Metabolites significantly associated with proteinuria were tested for associations with CKD progression.
Results
In the AASK and the MDRD study, respectively, the median urine protein-to-creatinine ratio was 80 (interquartile range [IQR], 28–359) and 188 (IQR, 54–894) mg/g, mean age was 56 and 52 years, 39% and 38% were women, 100% and 7% were black, and median measured GFR was 48 (IQR, 35–57) and 28 (IQR, 18–39) ml/min per 1.73 m2. Linear regression identified 66 serum metabolites associated with proteinuria in one or both studies after Bonferroni correction (P<7.8×10−5), 58 of which were statistically significant in a meta-analysis (P<7.8×10−4). The metabolites with the lowest P values (P<10−27) were 4-hydroxychlorthalonil and 1,5-anhydroglucitol; all six quantified metabolites in the phosphatidylethanolamine pathway were also significant. Of the 58 metabolites associated with proteinuria, four were associated with ESKD in both the AASK and the MDRD study.
Conclusions
We identified 58 serum metabolites with cross-sectional associations with proteinuria, some of which were also associated with CKD progression.
Podcast
This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2019_02_07_CJASNPodcast_19_03_.mp3
Introduction
Proteinuria is one of the strongest markers of risk for adverse CKD outcomes. At higher levels of proteinuria, the risks for CKD progression, ESKD, and mortality are higher, independent of GFR, and with no obvious threshold value (1). Consequently, the 2012 Kidney Disease: Improving Global Outcomes (KDIGO) Guideline for the evaluation and management of CKD recommended incorporation of proteinuria in CKD staging (2). Although renin-angiotensin system inhibitors may lower proteinuria levels to a moderate extent (3), additional therapeutic strategies to halt kidney disease progression via controlling proteinuria are needed.
Metabolomic profiling, a high-throughput approach to the evaluation of small molecules in biospecimens, is increasingly used as a tool for clinical and epidemiologic research (4,5). The small molecules detected are between 50 and 1500 D in mass and are often referred to as metabolites. They include products of the metabolism of amino acids, peptides, carbohydrates, lipids, cofactors, vitamins, nucleotides, and xenobiotics. The levels of metabolites not only reflect upstream transcriptional and translational processes, but also provide integrated information on a broad range of environmental factors and pathophysiologic alterations. The kidney influences metabolite levels through filtration, reabsorption, secretion, synthesis, and degradation; thus, the study of the human metabolome in the setting of CKD is particularly compelling (6–8). The identification of small molecules that are strongly associated with proteinuria and independent of GFR, may reveal metabolic pathways that are instrumental in causing kidney disease, and are not elevated simply because of reduced kidney excretion.
There is little information regarding the blood metabolite associations with proteinuria in CKD. Using the African American Study of Kidney Disease and Hypertension (AASK) and the Modification of Diet in Renal Disease (MDRD) study, two rigorously conducted clinical trials with per-protocol measures of 24-hour proteinuria and GFR and involving >1500 patients with CKD, we investigated the cross-sectional associations of proteinuria and 637 named, nondrug serum metabolites identified using untargeted metabolomic profiling. To determine whether correlations with proteinuria translated to faster CKD progression, we tested the metabolites found to be significantly associated with proteinuria for associations with eGFR decline and ESKD.
Materials and Methods
Study Design and Populations
AASK was a multicenter, clinical trial that used a 3×2 factorial design to evaluate the effects of three antihypertensive agents (ramipril, metoprolol, and amlodipine) and two BP control goals (mean arterial pressure ≤92 and 102–107 mm Hg) in slowing CKD progression. Between 1995 and 1998, 1094 self-identified black Americans (18–70 years of age) with CKD attributed to hypertension, urine protein-to-creatinine ratio ≤2500 mg/g, measured GFR between 20 and 65 ml/min per 1.73 m2, and without a diagnosis of diabetes mellitus were enrolled (9). Our analysis was conducted in a sample of 962 participants who had sufficient serum for metabolomic profiling, available urine protein-to-creatinine ratio measurements, and nonmissing covariates at baseline (Figure 1).
Figure 1.
In total, 637 metabolites measured in serum samples from 1582 participants in the African American Study of Kidney Disease and Hypertension (AASK) and the Modification of Diet in Renal Disease (MDRD) Study were included in this study.
The MDRD study was a multicenter clinical trial that used a 2×2 factorial design to assess the effects of dietary protein restriction and BP control goals in slowing CKD progression. A total of 840 patients (18–70 years of age) with progressive kidney disease were enrolled between 1989 and 1991 (10). On the basis of measured GFR at enrolment, the trial was divided into two substudies. Study A included patients with GFR between 13 and 24 ml/min per 1.73 m2 who were randomized to either usual protein diet or low-protein diet (1.3 or 0.58 g of protein per kilogram of body weight per day, respectively), and study B included patients with GFR between 25 and 55 ml/min per 1.73 m2 who were randomized to either low-protein diet or very-low-protein diet (0.58 and 0.28 g of protein per kilogram of body weight per day, respectively). Participants in both substudies were randomized to usual versus low target BP (mean arterial pressure ≤92 mm Hg versus 102–107 mm Hg). Of the 746 participants followed through the 12-month postrandomization visit (1990–1992), 620 with available metabolite and urine protein-to-creatinine ratio measurements, and not missing other covariates, were included in our analysis.
All participants provided informed consent for participation in the original trials. This study was approved by the institutional review boards at the Johns Hopkins Bloomberg School of Public Health (Baltimore, MD) (number: NA_00025896).
Proteinuria, Measured GFR, and Other Variables
Log-transformed urine protein-to-creatinine ratio was used as the measure of proteinuria at baseline in the AASK and the 12-month visit in the MDRD study. The AASK and MDRD study participants were instructed to perform 24-hour urine collections 1 day before the baseline and follow-up visits. At each visit, these urine samples were aliquoted and sent to the Central Biochemistry Laboratories at the Cleveland Clinic for measurement of protein and creatinine using the TCA–Ponceau S method and the modified Jaffe reaction, respectively (11). GFR was measured by urinary clearance of 125I-iothalamate in both studies. Serum albumin and fasting serum glucose concentrations were measured with autoanalyzers. Data on all covariables were collected using standard protocols at baseline in the AASK and the 12-month visit in the MDRD study, except for history of cardiovascular disease, which was ascertained only once among MDRD study participants and thus carried forward from the baseline visit.
eGFR and ESKD during Follow-Up
In the AASK, serum creatinine was measured at baseline and every 6 months thereafter until April 2002. This was done at Central Biochemistry Laboratory at the Cleveland Clinic, with the rate-Jaffe method and an alkaline picrate assay. It was converted to eGFR using the CKD Epidemiology Collaboration equation. In both the AASK and the MDRD study, ESKD was defined as initiation of dialysis or receipt of a kidney transplant. In the AASK, participants were followed for ESKD throughout the trial phase (until April 2002) and the ensuing cohort phase, which completed in June 2007. In the MDRD study, ESKD was ascertained through linkage to the US Renal Data System on December 31, 2007.
Metabolite Profiling
Serum metabolite profiling was performed in 2017 for samples collected at baseline in the AASK and in 2015 for samples collected at the 12-month visit in the MDRD study, using untargeted mass spectrometry (MS) following standard protocols at Metabolon, Inc. (Morrisville, NC) (Supplemental Appendix 1) (12). For this study, we included only known, nondrug metabolites detected in >20% of samples and present in both the AASK and the MDRD study (n=637). These metabolites were involved in a total of 90 pathways (Supplemental Appendix 2). Across the 637 metabolites, the median number of samples with undetectable levels was 2 (0.2%) out of the 962 samples from the AASK study population (interquartile range [IQR], 0–57 [0%–6%]) and 3 (0.5%) out of the 620 samples from the MDRD study population (IQR, 0–48 [0%–8%]).
Statistical Analyses
Participant characteristics were described using summary statistics. Serum levels of all metabolites were natural-log-transformed and missing values were imputed with the minimum values. Urine protein-to-creatinine ratio and measured GFR in both studies were also natural-log-transformed to reduce right-skewedness and scaled to the natural log of 2. To discover metabolites potentially associated with proteinuria in each study, linear regression was used, regressing log-transformed urine protein-to-creatinine ratio on log-transformed metabolites, with adjustment for age, sex, body mass index, history of smoking, history of cardiovascular disease, serum albumin concentration, log-measured GFR, as well as, in the MDRD study only, race and trial arms. Metabolites with regression coefficient P values lower than the prespecified α-level of 7.8×10−5 (set by Bonferroni correction, which corrected the 0.05 α-value by 637, the total number of metabolites included for analysis) in either study, and had associations that were not in the opposite directions, were carried onto the meta-analysis. In the meta-analysis, regression coefficients were combined using a fixed-effects, inverse-variance weighted model. These metabolites were then included in a pathway analysis using pathway annotations provided by Metabolon. The Fisher exact test was used to evaluate the probability that the observed number of significant metabolites in each pathway was different than the number expected on the basis of the number of significant metabolites in all other pathways combined. The permutation test assessed the probability of obtaining the observed number of significant metabolites in a given pathway, assuming a fixed number of total significant metabolites overall (n=58, the same as observed in the main analysis), accounting for intrapathway metabolite correlations. This test was performed by randomly permuting the residuals of a regression of proteinuria on all covariates. The approach was repeated 1000 times to generate a null distribution of the number of significant metabolites within each pathway, and the permutation P value was the proportion of the permuted results equal to or more extreme than the observed data (13). Finally, in a secondary analysis, the 58 statistically significant metabolites were tested for their associations with eGFR decline (from baseline in the AASK) and ESKD (from baseline in the AASK and 12-month visit in the MDRD study) during follow-up, using mixed models with random slope, random intercept, and unstructured covariance for eGFR decline, and Cox proportional hazards models for ESKD, controlling for the same covariates as those included in the main analysis, and with and without additional adjustment for proteinuria. The Bonferroni-corrected α-level for analyses of eGFR and ESKD was 8.6×10−4 (0.05 divided by 58, which was the number of significant metabolites in the meta-analysis). MetaboAnalyst (MetaboAnalyst.ca) was used to generate a heatmap demonstrating correlations across metabolites (14). All other statistical analyses were conducted using Stata 15.1 (StataCorp, College Station, TX).
Results
Participant Characteristics
Among 962 AASK participants and 620 participants in the MDRD study, the median urine protein-to-creatinine ratio was 80 (IQR, 28–359) and 188 (IQR, 54–894) mg/g, mean age was 56 and 52 years, 39% and 38% were women, 100% and 7% were black, and median measured GFR was 48 and 28 ml/min per 1.73 m2, respectively (Table 1).
Table 1.
Characteristics of participants at baseline in AASK and at the 12-month visit in the MDRD study, overall, and by tertile of proteinuria
| Variable | AASK (Baseline) | MDRD Study (12-mo Visit) | ||||||
|---|---|---|---|---|---|---|---|---|
| Overall | Urine Protein-to-Creatinine Ratio Tertile 1 | Urine Protein-to-Creatinine Ratio Tertile 2 | Urine Protein-to-Creatinine Ratio Tertile 3 | Overall | Urine Protein-to-Creatinine Ratio Tertile 1 | Urine Protein-to-Creatinine Ratio Tertile 2 | Urine Protein-to-Creatinine Ratio Tertile 3 | |
| N | 962 | 321 | 321 | 320 | 620 | 207 | 207 | 206 |
| Mean age, yr (SD) | 56 (11) | 59 (9) | 54 (10) | 51 (11) | 52 (12) | 54 (11) | 52 (12) | 50 (13) |
| Women, n (%) | 374 (39) | 104 (32) | 159 (50) | 111 (35) | 236 (38) | 78 (38) | 84 (41) | 74 (36) |
| Black, n (%) | 962 (100) | 321 (100) | 321 (100) | 320 (100) | 46 (7) | 7 (4) | 18 (9) | 21 (10) |
| Mean BMI, kg/m2 (SD) | 30.6 (6.6) | 30.2 (6.2) | 30.0 (6.4) | 31.7 (7.0) | 26.8 (4.2) | 27.0 (3.9) | 26.5 (4.6) | 26.9 (4.2) |
| Current smoker, n (%) | 280 (29) | 74 (23) | 102 (32) | 104 (33) | 61 (10) | 16 (8) | 16 (8) | 29 (14) |
| Cardiovascular disease, n (%)a | 497 (52) | 159 (50) | 171 (53) | 167 (52) | 65 (10) | 24 (12) | 16 (8) | 25 (12) |
| Diabetes, n (%)a | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 30 (5) | 8 (4) | 8 (4) | 14 (7) |
| Median fasting serum glucose, mg/dl (25th–75th percentile)b | 92 (84–101) | 92 (84–101) | 91 (83–101) | 92 (85–100) | 89 (81–97) | 88 (81–95) | 90 (81–98) | 89 (82–98) |
| Mean serum albumin concentration, g/dl (SD) | 4.2 (0.3) | 4.3 (0.3) | 4.3 (0.3) | 4.1 (0.3) | 4.1 (0.3) | 4.2 (0.3) | 4.2 (0.3) | 4.0 (0.3) |
| Median measured GFR, ml/min per 1.73 m2 (25th–75th percentile) | 48 (35–57) | 54 (46–60) | 48 (36–57) | 37 (28–49) | 28 (18–39) | 36 (27–44) | 24 (17–37) | 21 (15–30) |
| Median urine protein-to-creatinine ratio, mg/g (25th–75th percentile) | 80 (28–359) | 21 (15–28) | 80 (54–117) | 628 (361–1264) | 188 (54–894) | 36 (24–54) | 188 (123–319) | 1366 (894–2164) |
MDRD, Modification of Diet in Renal Disease; AASK, African American Study of Kidney Disease and Hypertension; BMI, body mass index.
Data obtained at baseline (prerandomization visit) for MDRD study participants.
Fasting glucose missing in two participants in the MDRD study.
Metabolites Associated with Proteinuria in the AASK, the MDRD Study, or Both
Among the 637 named, nondrug metabolites with <80% missing data across samples, 66 were statistically significantly associated with proteinuria in either study (55 in the AASK, 15 in the MDRD Study), after controlling for covariates (all P<7.8×10−5; Table 2). All but two (N6-acetyllysine and vanillylmandelate) had consistent direction of association in the AASK and the MDRD study and were included in the meta-analysis, which identified 58 metabolites with statistically significant associations after Bonferroni correction (all P<7.8×10−4; Figure 2). Metabolites with the lowest P values include 4-hydroxychlorothalonil, 1,5-anhydroglucitol (1,5-AG), and 1-stearoyl-2-arachidonoyl-glycero-3-phosphoethanolamine (GPE) (18:0/20:4). In addition to 4-hydroxychlorothalonil and 1,5-AG, 3-carboxy-4-methyl-5-propyl-2-furanpropanoate, 3-hydroxy-5-cholestenoic acid, and oxalate also had negative combined regression coefficients. All other metabolites in the meta-analysis (n=53) were positively associated with proteinuria.
Table 2.
Regression coefficients for metabolites significantly associated with proteinuria in the AASK, the MDRD study, or both, and summary of missingness across samples
| Metabolite | AASK | MDRD | Meta-Analysis | Missingness in AASK (%) | Missingness in MDRD (%) | |||
|---|---|---|---|---|---|---|---|---|
| β Coefficient (95% CI) | P Value | β Coefficient (95% CI) | P Value | β Coefficient (95% CI) | P Value | |||
| 4-hydroxychlorothalonil | −0.33 (−0.37 to −0.29) | <1.0×10−31 | −0.36 (−0.40 to −0.32) | <1.0×10−31 | −0.35 (−0.37 to −0.32) | <1.0×10−31 | 172 (18) | 63 (10) |
| 1,5-anhydroglucitol (1,5-AG) | −0.09 (−0.12 to −0.07) | 1.6×10−14 | −0.15 (−0.19 to −0.12) | 1.4×10−15 | −0.11 (−0.13 to −0.09) | 4.2×10−28 | 0 (0) | 0 (0) |
| 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4) | 0.07 (0.05 to 0.09) | 1.3×10−11 | 0.05 (0.02 to 0.07) | 6.8×10−4 | 0.06 (0.04 to 0.08) | 5.8×10−14 | 0 (0) | 0 (0) |
| 1-stearoyl-GPE (18:0) | 0.05 (0.03 to 0.07) | 5.0×10−9 | 0.04 (0.02 to 0.06) | 2.3×10−5 | 0.05 (0.03 to 0.06) | 4.1×10−13 | 0 (0) | 0 (0) |
| 3-carboxy-4-methyl-5-propyl-2-furanpropanoate (CMPF) | −0.10 (−0.15 to −0.05) | 8.2×10−5 | −0.19 (−0.25 to −0.13) | 2.1×10−9 | −0.14 (−0.17 to −0.10) | 4.0×10−12 | 0 (0) | 0 (0) |
| 1-stearoyl-2-arachidonoyl-GPI (18:0/20:4) | 0.04 (0.03 to 0.06) | 3.9×10−9 | 0.03 (0.01 to 0.05) | 0.001 | 0.04 (0.03 to 0.05) | 1.5×10−11 | 0 (0) | 1 (0) |
| Aspartate | 0.02 (0.01 to 0.03) | 9.3×10−4 | 0.04 (0.03 to 0.05) | 2.4×10−9 | 0.03 (0.02 to 0.04) | 4.4×10−11 | 0 (0) | 0 (0) |
| 1-palmitoyl-2-arachidonoyl-GPE (16:0/20:4) | 0.07 (0.05 to 0.09) | 1.2×10−9 | 0.04 (0.01 to 0.07) | 0.007 | 0.06 (0.04 to 0.08) | 9.7×10−11 | 0 (0) | 0 (0) |
| 1-(1-enyl-palmitoyl)-2-arachidonoyl-GPE (P-16:0/20:4) | 0.05 (0.03 to 0.07) | 1.2×10−5 | 0.05 (0.03 to 0.07) | 3.4×10−6 | 0.05 (0.03 to 0.06) | 1.3×10−10 | 0 (0) | 0 (0) |
| Retinol (vitamin A) | 0.07 (0.05 to 0.09) | 2.0×10−13 | 0.02 (−0.00 to 0.03) | 0.05 | 0.04 (0.03 to 0.05) | 1.6×10−10 | 0 (0) | 0 (0) |
| 1-palmitoyl-GPE (16:0) | 0.05 (0.04 to 0.07) | 1.2×10−8 | 0.03 (0.01 to 0.05) | 0.01 | 0.04 (0.03 to 0.06) | 3.8×10−10 | 0 (0) | 0 (0) |
| 4-acetamidobutanoate | 0.08 (0.05 to 0.10) | 5.6×10−12 | 0.02 (−0.00 to 0.04) | 0.08 | 0.04 (0.03 to 0.06) | 2.5×10−9 | 2 (0) | 1 (0) |
| 1-(1-enyl-palmitoyl)-2-oleoyl-GPE (P-16:0/18:1) | 0.04 (0.02 to 0.06) | 8.3×10−4 | 0.06 (0.03 to 0.08) | 5.8×10−6 | 0.05 (0.03 to 0.06) | 2.2×10−8 | 3 (0) | 6 (1) |
| N6-carbamoylthreonyladenosine | 0.04 (0.03 to 0.05) | 5.9×10−10 | 0.01 (−0.00 to 0.02) | 0.17 | 0.03 (0.02 to 0.04) | 2.9×10−8 | 0 (0) | 0 (0) |
| N-palmitoyl-sphingosine (d18:1/16:0) | 0.03 (0.01 to 0.04) | 7.9×10−4 | 0.04 (0.02 to 0.05) | 8.7×10−6 | 0.03 (0.02 to 0.04) | 3.1×10−8 | 0 (0) | 0 (0) |
| N2,N5-diacetylornithine | 0.10 (0.07 to 0.13) | 3.0×10−11 | 0.01 (−0.02 to 0.05) | 0.42 | 0.06 (0.04 to 0.09) | 3.7×10−8 | 9 (1) | 6 (1) |
| 3-hydroxy-5-cholestenoic acid | −0.04 (−0.06 to −0.01) | 0.001 | −0.05 (−0.08 to −0.03) | 1.3×10−5 | −0.05 (−0.06 to −0.03) | 8.5×10−8 | 1 (0) | 0 (0) |
| 1-palmitoyl-2-oleoyl-GPE (16:0/18:1) | 0.07 (0.05 to 0.10) | 1.5×10−7 | 0.03 (−0.00 to 0.07) | 0.06 | 0.06 (0.04 to 0.08) | 9.4×10−8 | 4 (0) | 3 (0) |
| O-sulfo-L-tyrosine | 0.05 (0.04 to 0.07) | 1.6×10−11 | 0.00 (−0.02 to 0.02) | 0.88 | 0.03 (0.02 to 0.04) | 1.2×10−7 | 0 (0) | 0 (0) |
| 1-palmitoyl-2-arachidonoyl-GPC (16:0/20:4) | 0.03 (0.02 to 0.04) | 7.2×10−8 | 0.01 (−0.03 to 0.04) | 0.69 | 0.03 (0.02 to 0.04) | 1.3×10−7 | 0 (0) | 3 (0) |
| 6-hydroxyindole sulfate | 0.08 (0.05 to 0.11) | 5.9×10−8 | 0.03 (−0.01 to 0.07) | 0.11 | 0.06 (0.04 to 0.08) | 1.4×10−7 | 1 (0) | 0 (0) |
| Cholesterol | 0.03 (0.01 to 0.04) | 5.7×10−4 | 0.03 (0.02 to 0.05) | 7.0×10−5 | 0.03 (0.02 to 0.04) | 1.4×10−7 | 0 (0) | 0 (0) |
| 1-stearoyl-2-oleoyl-GPE (18:0/18:1) | 0.07 (0.04 to 0.09) | 3.6×10−6 | 0.04 (0.01 to 0.08) | 0.01 | 0.06 (0.04 to 0.08) | 1.7×10−7 | 1 (0) | 0 (0) |
| 3-methylglutaconate | 0.06 (0.04 to 0.08) | 9.2×10−9 | 0.01 (−0.02 to 0.04) | 0.44 | 0.04 (0.03 to 0.06) | 1.8×10−7 | 0 (0) | 0 (0) |
| 1-palmitoyl-2-linoleoyl-GPE (16:0/18:2) | 0.06 (0.04 to 0.08) | 1.9×10−6 | 0.03 (0.01 to 0.06) | 0.02 | 0.05 (0.03 to 0.07) | 2.4×10−7 | 0 (0) | 0 (0) |
| 1-stearoyl-2-linoleoyl-GPE (18:0/18:2) | 0.06 (0.04 to 0.08) | 8.1×10−7 | 0.03 (0.00 to 0.06) | 0.04 | 0.05 (0.03 to 0.06) | 2.9×10−7 | 0 (0) | 1 (0) |
| N-palmitoyl-sphinganine (d18:0/16:0) | 0.04 (0.01 to 0.07) | 0.004 | 0.06 (0.03 to 0.08) | 6.2×10−5 | 0.05 (0.03 to 0.07) | 7.5×10−7 | 47 (5) | 1 (0) |
| 1-palmitoyl-2-oleoyl-GPC (16:0/18:1) | 0.03 (0.02 to 0.04) | 2.2×10−6 | 0.01 (−0.00 to 0.03) | 0.08 | 0.02 (0.01 to 0.03) | 9.8×10−7 | 0 (0) | 0 (0) |
| N2,N2-dimethylguanosine | 0.05 (0.03 to 0.06) | 4.0×10−10 | 0.00 (−0.01 to 0.02) | 0.88 | 0.03 (0.02 to 0.04) | 1.4×10−6 | 0 (0) | 2 (0) |
| 1-stearoyl-2-arachidonoyl-GPC (18:0/20:4) | 0.03 (0.01 to 0.04) | 1.2×10−5 | 0.02 (−0.00 to 0.04) | 0.06 | 0.02 (0.01 to 0.03) | 1.9×10−6 | 0 (0) | 0 (0) |
| 1-palmitoyl-2-linoleoyl-GPC (16:0/18:2) | 0.02 (0.01 to 0.03) | 1.2×10−6 | 0.01 (−0.00 to 0.02) | 0.22 | 0.02 (0.01 to 0.02) | 2.5×10−6 | 0 (0) | 0 (0) |
| Methylsuccinate | 0.04 (0.02 to 0.05) | 1.6×10−5 | 0.03 (−0.00 to 0.06) | 0.06 | 0.04 (0.02 to 0.05) | 2.6×10−6 | 2 (0) | 23 (4) |
| 1-arachidonoyl-GPE (20:4) | 0.04 (0.02 to 0.05) | 3.6×10−6 | 0.02 (−0.00 to 0.03) | 0.11 | 0.03 (0.02 to 0.04) | 4.5×10−6 | 0 (0) | 0 (0) |
| N6-succinyladenosine | 0.05 (0.03 to 0.07) | 3.6×10−8 | 0.00 (−0.02 to 0.03) | 0.86 | 0.03 (0.02 to 0.05) | 6.4×10−6 | 9 (1) | 3 (0) |
| Kynurenate | 0.05 (0.03 to 0.07) | 1.5×10−6 | 0.01 (−0.01 to 0.04) | 0.28 | 0.04 (0.02 to 0.05) | 6.6×10−6 | 1 (0) | 1 (0) |
| 3-hydroxy-3-methylglutarate | 0.04 (0.02 to 0.05) | 2.7×10−5 | 0.02 (−0.00 to 0.04) | 0.06 | 0.03 (0.02 to 0.04) | 7.5×10−6 | 0 (0) | 0 (0) |
| 1-palmitoyl-2-arachidonoyl-GPI (16:0/20:4) | 0.05 (0.03 to 0.07) | 8.1×10−6 | 0.02 (−0.01 to 0.05) | 0.25 | 0.04 (0.02 to 0.06) | 9.5×10−6 | 3 (0) | 44 (7) |
| N-acetyl-3-methylhistidine | 0.14 (0.07 to 0.20) | 1.7×10−5 | 0.06 (−0.00 to 0.12) | 0.05 | 0.10 (0.05 to 0.14) | 9.8×10−6 | 152 (16) | 49 (8) |
| 2-stearoyl-GPE (18:0) | 0.05 (0.03 to 0.07) | 8.7×10−6 | 0.02 (−0.01 to 0.05) | 0.16 | 0.04 (0.02 to 0.06) | 1.2×10−5 | 0 (0) | 9 (1) |
| Pseudouridine | 0.03 (0.02 to 0.04) | 2.3×10−6 | 0.01 (−0.01 to 0.02) | 0.46 | 0.02 (0.01 to 0.03) | 1.7×10−5 | 0 (0) | 0 (0) |
| 3-indoxyl sulfate | 0.07 (0.04 to 0.09) | 3.8×10−7 | 0.01 (−0.02 to 0.04) | 0.56 | 0.04 (0.02 to 0.06) | 2.1×10−5 | 0 (0) | 0 (0) |
| Glycosyl-N-palmitoyl-sphingosine | 0.01 (−0.01 to 0.02) | 0.33 | 0.04 (0.03 to 0.06) | 7.4×10−7 | 0.02 (0.01 to 0.04) | 2.5×10−5 | 0 (0) | 0 (0) |
| Pyridoxate | 0.09 (0.06 to 0.13) | 7.2×10−8 | −0.02 (−0.07 to 0.04) | 0.54 | 0.06 (0.03 to 0.09) | 3.2×10−5 | 0 (0) | 0 (0) |
| 3-acetylphenol sulfate | 0.05 (−0.01 to 0.10) | 0.08 | 0.11 (0.06 to 0.17) | 5.9×10−5 | 0.08 (0.04 to 0.12) | 4.7×10−5 | 356 (37) | 107 (17) |
| Arabitol/xylitol | 0.04 (0.02 to 0.05) | 1.6×10−5 | 0.01 (−0.01 to 0.03) | 0.32 | 0.03 (0.01 to 0.04) | 6.4×10−5 | 0 (0) | 0 (0) |
| 1-palmitoyl-2-linoleoyl-GPI (16:0/18:2) | 0.04 (0.02 to 0.06) | 3.8×10−5 | 0.01 (−0.02 to 0.04) | 0.54 | 0.03 (0.01 to 0.04) | 6.4×10−5 | 0 (0) | 13 (2) |
| Indolin-2-one | 0.06 (0.03 to 0.09) | 2.1×10−5 | 0.01 (−0.02 to 0.05) | 0.46 | 0.04 (0.02 to 0.07) | 1.5×10−4 | 186 (19) | 6 (1) |
| 1-palmitoleoyl-GPC (16:1) | 0.03 (0.02 to 0.05) | 2.2×10−5 | 0.01 (−0.01 to 0.03) | 0.39 | 0.02 (0.01 to 0.03) | 1.5×10−4 | 0 (0) | 0 (0) |
| 1-methylimidazoleacetate | 0.04 (0.02 to 0.06) | 1.8×10−6 | 0.00 (−0.02 to 0.02) | 0.80 | 0.02 (0.01 to 0.04) | 1.6×10−4 | 0 (0) | 0 (0) |
| Xanthurenate | 0.09 (0.05 to 0.13) | 8.9×10−6 | 0.01 (−0.03 to 0.06) | 0.53 | 0.06 (0.03 to 0.09) | 2.2×10−4 | 209 (22) | 93 (15) |
| N-acetylphenylalanine | 0.05 (0.03 to 0.08) | 8.7×10−6 | 0.01 (−0.02 to 0.03) | 0.50 | 0.03 (0.01 to 0.05) | 2.5×10−4 | 1 (0) | 1 (0) |
| Tiglylcarnitine | 0.07 (0.04 to 0.10) | 2.0×10−7 | 0.00 (−0.02 to 0.03) | 0.74 | 0.03 (0.01 to 0.05) | 2.7×10−4 | 6 (1) | 0 (0) |
| Pantothenate | 0.05 (0.03 to 0.07) | 1.7×10−5 | 0.01 (−0.02 to 0.03) | 0.49 | 0.03 (0.01 to 0.05) | 2.8×10−4 | 0 (0) | 0 (0) |
| Serine | 0.00 (−0.01 to 0.01) | 0.96 | 0.02 (0.01 to 0.03) | 2.6×10−6 | 0.01 (0.00 to 0.02) | 3.5×10−4 | 0 (0) | 0 (0) |
| N1-methylinosine | 0.04 (0.02 to 0.06) | 6.0×10−5 | 0.01 (−0.02 to 0.03) | 0.61 | 0.03 (0.01 to 0.04) | 3.6×10−4 | 0 (0) | 7 (1) |
| C-glycosyltryptophan | 0.03 (0.02 to 0.04) | 1.1×10−5 | 0.00 (−0.02 to 0.02) | 0.88 | 0.02 (0.01 to 0.03) | 4.0×10−4 | 0 (0) | 8 (1) |
| Oxalate (ethanedioate) | −0.01 (−0.03 to 0.02) | 0.53 | −0.05 (−0.07 to −0.03) | 2.6×10−5 | −0.03 (−0.04 to −0.01) | 4.5×10−4 | 0 (0) | 0 (0) |
| N-acetylleucine | 0.03 (0.02 to 0.05) | 5.9×10−5 | 0.01 (−0.01 to 0.02) | 0.45 | 0.02 (0.01 to 0.03) | 5.4×10−4 | 1 (0) | 4 (1) |
| N-acetylglutamine | 0.05 (0.03 to 0.08) | 3.3×10−5 | 0.00 (−0.02 to 0.03) | 0.72 | 0.03 (0.01 to 0.05) | 0.001 | 0 (0) | 0 (0) |
| 4-hydroxyhippurate | 0.09 (0.05 to 0.12) | 4.4×10−6 | −0.01 (−0.05 to 0.03) | 0.70 | 0.04 (0.02 to 0.07) | 0.002 | 0 (0) | 7 (1) |
| Creatinine | 0.02 (0.01 to 0.03) | 2.0×10−5 | 0.00 (−0.00 to 0.01) | 0.50 | 0.01 (0.00 to 0.01) | 0.003 | 0 (0) | 0 (0) |
| N-acetylneuraminate | 0.03 (0.02 to 0.04) | 4.5×10−5 | 0.00 (−0.01 to 0.01) | 0.76 | 0.01 (0.00 to 0.02) | 0.004 | 0 (0) | 0 (0) |
| N-acetyl-1-methylhistidine | 0.08 (0.04 to 0.12) | 3.8×10−5 | −0.01 (−0.05 to 0.03) | 0.61 | 0.04 (0.01 to 0.07) | 0.01 | 16 (2) | 3 (0) |
| N6-acetyllysinea | 0.02 (0.01 to 0.03) | 3.0×10−5 | −0.01 (−0.03 to −0.00) | 0.05 | — | — | 0 (0) | 0 (0) |
| Vanillylmandelate (VMA)a | 0.05 (0.03 to 0.07) | 2.1×10−5 | −0.02 (−0.05 to −0.00) | 0.02 | — | — | 3 (0) | 0 (0) |
| 2-methylmalonylcarnitine | 0.10 (0.05 to 0.14) | 1.8×10−5 | −0.01 (−0.03 to 0.01) | 0.34 | 0.01 (−0.01 to 0.03) | 0.32 | 451 (47) | 4 (1) |
Metabolites are ordered by P value in the meta-analysis. Each unit change in β coefficient represents one-fold change (2β−1=100% when β=1) in proteinuria per unit higher log-transformed serum metabolite.
Statistically significant associations are shown in bold. Cut-off for statistical significance in each single study (the AASK or the MDRD study)=0.05/637=7.8×10−5; cut-off for statistical significance in the meta-analysis=0.05/66=7.6×10−4. AASK, African American Study of Kidney Disease and Hypertension; MDRD, Modification of Diet in Renal Disease; 95% CI, 95% confidence interval; GPE, glycero-3-phosphoethanolamine; GPI, glycero-3-phosphoinositol; GPC, glycero-3-phosphocholine.
Metabolites with associations opposite in direction across studies were not included in the meta-analysis.
Figure 2.
4-Hydroxychlorothalonil, 1,5-anhydroglucitol, and 1-stearoyl-2-arachidonoyl-glycero-3-phosphoethanolamine (GPE) (18:0/20:4) had the lowest P values in fixed-effects meta-analysis.
Analysis of Pathways among Metabolites Identified in the Meta-Analysis
On the basis of pathway annotations provided by Metabolon, the metabolites significant in the meta-analysis were involved in a total of 30 metabolic subpathways (eight pathways of amino acid metabolism, two pathways of carbohydrate metabolism, four pathways of cofactors/vitamins, ten pathways of lipid metabolism, four pathways of nucleotide metabolism, and two types of xenobiotics) (Supplemental Table 1), and there were moderate inter- and intrapathway correlations (Figure 3). Among these pathways, phosphatidylethanolamines (PEs), which are involved in lipid metabolism, were overrepresented (six out of six metabolites in the PE pathway had statistically significant positive associations with proteinuria; Fisher exact test P=4.5×10−7; permutation P=0.01; Table 3).
Figure 3.
Moderate Spearman correlations were found across metabolites. Spearman correlations were assessed in the pooled study populations. PCs, phosphatidylcholines; PIs, phosphatidylinositols.
Table 3.
Results from analysis of pathway overrepresentation
| Pathway | Total No. of Metabolites in Both Studies | No. of Metabolites Associated with Proteinuria | Fisher Exact P Value | Permutation P Value |
|---|---|---|---|---|
| Phosphatidylethanolamine (PE) | 6 | 6 | 4.5×10−7 | 0.01 |
| Phosphatidylinositol (PI) | 5 | 3 | 0.006 | 0.04 |
| Phosphatidylcholine (PC) | 13 | 4 | 0.02 | 0.07 |
| Ceramides | 4 | 2 | 0.04 | 0.07 |
| Sterol | 4 | 2 | 0.04 | 0.03 |
| Chemical | 16 | 4 | 0.05 | 0.07 |
| Lysophospholipid | 25 | 5 | 0.07 | 0.15 |
| Vitamin A metabolism | 1 | 1 | 0.09 | 0.10 |
| Mevalonate metabolism | 1 | 1 | 0.09 | 0.09 |
| Pantothenate and CoA metabolism | 1 | 1 | 0.09 | 0.08 |
| Vitamin B6 metabolism | 1 | 1 | 0.09 | 0.11 |
| Purine metabolism, adenine containing | 6 | 2 | 0.10 | 0.43 |
| Tryptophan metabolism | 21 | 4 | 0.10 | 0.13 |
| Purine metabolism, guanine containing | 2 | 1 | 0.17 | 0.17 |
| Pentose metabolism | 3 | 1 | 0.25 | 0.22 |
| Ascorbate and aldarate metabolism | 3 | 1 | 0.25 | 0.20 |
| Plasmalogen | 11 | 2 | 0.26 | 0.41 |
| Leucine, isoleucine and valine metabolism | 27 | 4 | 0.30 | 0.26 |
| Histidine metabolism | 12 | 2 | 0.30 | 0.30 |
| Polyamine metabolism | 5 | 1 | 0.38 | 0.36 |
| Glycolysis, gluconeogenesis, and pyruvate metabolism | 5 | 1 | 0.38 | 0.38 |
| Alanine and aspartate metabolism | 6 | 1 | 0.44 | 0.40 |
| Phenylalanine metabolism | 6 | 1 | 0.44 | 0.45 |
| Pyrimidine metabolism, uracil containing | 7 | 1 | 0.49 | 0.50 |
| Purine metabolism, (hypo)xanthine/inosine containing | 7 | 1 | 0.49 | 0.50 |
| Food component/plant | 30 | 1 | 0.51 | 0.92 |
| Glycine, serine, and threonine metabolism | 8 | 1 | 0.54 | 0.48 |
| Sphingolipid metabolism | 22 | 1 | 0.71 | 0.59 |
| Urea cycle; arginine and proline metabolism | 15 | 1 | >0.99 | 0.67 |
| Fatty acid, dicarboxylate | 16 | 1 | >0.99 | 0.65 |
Associations of Metabolites with eGFR and CKD Progression
The median follow-up in the AASK was 7.9 (IQR, 3.6–10.1) and 8.8 (IQR, 4.7–10.4) years for eGFR and ESKD, respectively. In total, 274 AASK participants developed ESKD. During a median follow-up of 5.2 (IQR, 2.7–10.7) years, 475 participants developed ESKD in the MDRD study. Among the 58 significant metabolites in the meta-analysis, 18 (31%) metabolites, including the top three significant metabolites in the main analysis (4-hydroxychlorothalonil, 1,5-AG, and 1-stearoyl-2-arachidonoyl-GPE), were associated with eGFR change during follow-up in AASK, but none was statistically significant after adjusting for proteinuria (Supplemental Table 2). The number of metabolites with statistically significant associations with ESKD (adjusted for measured GFR, trial arms, and other covariables in the main analysis) was 37 (64%) in the AASK and four (7%) in the MDRD study (Table 4). The metabolites 4-hydroxychlorothalonil, 1,5-AG, and 1-stearoyl-2-arachidonoyl-GPE were associated with ESKD in the AASK but not the MDRD study, and not independently of proteinuria.
Table 4.
Metabolites with significant cross-sectional associations with proteinuria in the meta-analysis, and their associations with ESKD
| Metabolite | AASK | MDRD | ||
|---|---|---|---|---|
| HR (95% CI) | P Value | HR (95% CI) | P Value | |
| 4-hydroxychlorothalonil | 0.56 (0.47 to 0.67) | 4.00E−10 | 0.90 (0.78 to 1.03) | 0.11 |
| 1,5-anhydroglucitol (1,5-AG) | 0.68 (0.54 to 0.85) | 8.00E−04 | 0.82 (0.72 to 0.93) | 2.40E−03 |
| 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4) | 2.94 (2.14 to 4.03) | 2.40E−11 | 1.25 (0.99 to 1.56) | 5.60E−02 |
| 1-stearoyl-GPE (18:0) | 2.21 (1.51 to 3.24) | 4.10E−05 | 1.17 (0.88 to 1.56) | 0.28 |
| 3-carboxy-4-methyl-5-propyl-2-furanpropanoate (CMPF) | 0.89 (0.79 to 1.01) | 6.50E−02 | 1.09 (1.00 to 1.20) | 5.40E−02 |
| 1-stearoyl-2-arachidonoyl-GPI (18:0/20:4) | 2.79 (1.81 to 4.31) | 3.70E−06 | 1.56 (1.19 to 2.05) | 1.40E−03 |
| Aspartate | 1.36 (0.83 to 2.22) | 0.22 | 1.00 (0.66 to 1.52) | >0.99 |
| 1-palmitoyl-2-arachidonoyl-GPE (16:0/20:4) | 2.38 (1.82 to 3.13) | 4.30E−10 | 1.12 (0.92 to 1.38) | 0.26 |
| 1-(1-enyl-palmitoyl)-2-arachidonoyl-GPE (P-16:0/20:4) | 1.51 (1.11 to 2.07) | 8.90E−03 | 1.19 (0.91 to 1.55) | 0.2 |
| Retinol (vitamin A) | 2.8 (2.01 to 3.92) | 1.50E−09 | 0.98 (0.7 to 1.36) | 0.89 |
| 1-palmitoyl-GPE (16:0) | 2.22 (1.56 to 3.16) | 8.40E−06 | 1.14 (0.88 to 1.46) | 0.32 |
| 4-acetamidobutanoate | 1.99 (1.48 to 2.66) | 4.40E−06 | 1.65 (1.23 to 2.22) | 8.30E−04 |
| 1-(1-enyl-palmitoyl)-2-oleoyl-GPE (P-16:0/18:1) | 1.36 (1.03 to 1.79) | 2.80E−02 | 0.92 (0.73 to 1.15) | 0.46 |
| N6-carbamoylthreonyladenosine | 4.22 (2.53 to 7.02)a | 3.10E−08 | 2.07 (1.35 to 3.17) | 8.20E−04 |
| N-palmitoyl-sphingosine (d18:1/16:0) | 1.93 (1.28 to 2.9) | 1.70E−03 | 1.72 (1.21 to 2.44) | 2.30E−03 |
| N2,N5-diacetylornithine | 1.76 (1.44 to 2.14)a | 2.30E−08 | 1.38 (1.16 to 1.63)a | 1.80E−04 |
| 3-hydroxy-5-cholestenoic acid | 0.75 (0.57 to 0.97) | 3.10E−02 | 1.06 (0.84 to 1.34) | 0.61 |
| 1-palmitoyl-2-oleoyl-GPE (16:0/18:1) | 1.98 (1.57 to 2.5)a | 9.30E−09 | 1.02 (0.86 to 1.2) | 0.82 |
| O-sulfo-L-tyrosine | 4.48 (2.93 to 6.85)a | 4.20E−12 | 1.66 (1.22 to 2.25) | 1.10E−03 |
| 1-palmitoyl-2-arachidonoyl-GPC (16:0/20:4) | 4.97 (2.83 to 8.75) | 2.60E−08 | 0.98 (0.83 to 1.15) | 0.81 |
| 6-hydroxyindole sulfate | 1.65 (1.3 to 2.07) | 2.50E−05 | 1.04 (0.88 to 1.21) | 0.66 |
| Cholesterol | 1.7 (1.15 to 2.51) | 8.10E−03 | 1.6 (1.12 to 2.28) | 9.00E−03 |
| 1-stearoyl-2-oleoyl-GPE (18:0/18:1) | 1.9 (1.52 to 2.39) | 2.60E−08 | 1.08 (0.91 to 1.29) | 0.38 |
| 3-methylglutaconate | 2.24 (1.65 to 3.03) | 1.80E−07 | 1.17 (0.97 to 1.42) | 0.1 |
| 1-palmitoyl-2-linoleoyl-GPE (16:0/18:2) | 2.13 (1.64 to 2.77)a | 1.60E−08 | 1.18 (0.98 to 1.43) | 8.20E−02 |
| 1-stearoyl-2-linoleoyl-GPE (18:0/18:2) | 2.3 (1.76 to 3.02)a | 1.50E−09 | 1.21 (0.98 to 1.49) | 8.10E−02 |
| N-palmitoyl-sphinganine (d18:0/16:0) | 1.54 (1.24 to 1.9) | 8.30E−05 | 1.24 (1.01 to 1.52) | 4.30E−02 |
| 1-palmitoyl-2-oleoyl-GPC (16:0/18:1) | 2.94 (1.8 to 4.81) | 1.70E−05 | 0.89 (0.63 to 1.25) | 0.5 |
| N2,N2-dimethylguanosine | 2.53 (1.65 to 3.87) | 2.00E−05 | 1.44 (1 to 2.06) | 4.90E−02 |
| 1-stearoyl-2-arachidonoyl-GPC (18:0/20:4) | 3.19 (1.79 to 5.67) | 7.80E−05 | 1.06 (0.8 to 1.41) | 0.68 |
| 1-palmitoyl-2-linoleoyl-GPC (16:0/18:2) | 7.64 (3.62 to 16.15) | 9.90E−08 | 1.74 (1.02 to 2.97) | 4.20E−02 |
| Methylsuccinate | 1.28 (0.89 to 1.84) | 0.19 | 1.28 (1.05 to 1.55) | 1.30E−02 |
| 1-arachidonoyl-GPE (20:4) | 2.3 (1.58 to 3.34) | 1.40E−05 | 0.81 (0.61 to 1.08) | 0.15 |
| N6-succinyladenosine | 3.01 (2.03 to 4.45) | 3.60E−08 | 1.2 (0.92 to 1.57) | 0.18 |
| Kynurenate | 1.78 (1.3 to 2.45)a | 3.60E−04 | 1.57 (1.29 to 1.93)a | 1.10E−05 |
| 3-hydroxy-3-methylglutarate | 1.52 (1.11 to 2.08) | 8.90E−03 | 1.29 (0.99 to 1.67) | 5.90E−02 |
| 1-palmitoyl-2-arachidonoyl-GPI (16:0/20:4) | 1.74 (1.31 to 2.32) | 1.40E−04 | 1.03 (0.87 to 1.21) | 0.76 |
| N-acetyl-3-methylhistidine | 1.3 (1.17 to 1.45)a | 1.50E−06 | 1.15 (1.05 to 1.25) | 2.30E−03 |
| 2-stearoyl-GPE (18:0) | 1.37 (1.03 to 1.81) | 2.80E−02 | 1.24 (1.01 to 1.53) | 4.10E−02 |
| Pseudouridine | 8.25 (4.5 to 15.11)a | 8.60E−12 | 1.2 (0.82 to 1.75) | 0.36 |
| 3-indoxyl sulfate | 1.88 (1.42 to 2.48) | 8.50E−06 | 1.23 (1.02 to 1.49) | 3.10E−02 |
| Glycosyl-N-palmitoyl-sphingosine | 1.27 (0.86 to 1.88) | 0.24 | 0.9 (0.64 to 1.27) | 0.55 |
| Pyridoxate | 1.43 (1.22 to 1.67)a | 7.40E−06 | 1.2 (1.06 to 1.35) | 3.00E−03 |
| 3-acetylphenol sulfate | 1.15 (0.99 to 1.35) | 7.60E−02 | 1.11 (0.99 to 1.25) | 8.50E−02 |
| Arabitol/xylitol | 2.55 (1.81 to 3.6)a | 7.70E−08 | 1.17 (0.9 to 1.53) | 0.25 |
| 1-palmitoyl-2-linoleoyl-GPI (16:0/18:2) | 1.73 (1.22 to 2.45) | 2.10E−03 | 1.11 (0.94 to 1.31) | 0.22 |
| Indolin-2-one | 1.69 (1.32 to 2.17) | 3.30E−05 | 1.11 (0.95 to 1.3) | 0.19 |
| 1-palmitoleoyl-GPC (16:1) | 1.12 (0.77 to 1.63) | 0.55 | 0.76 (0.56 to 1.05) | 9.40E−02 |
| 1-methylimidazoleacetate | 2.45 (1.76 to 3.42) | 1.30E−07 | 0.94 (0.68 to 1.29) | 0.7 |
| Xanthurenate | 1.31 (1.11 to 1.54) | 9.90E−04 | 1.19 (1.04 to 1.36) | 9.90E−03 |
| N-acetylphenylalanine | 1.25 (0.98 to 1.6) | 7.00E−02 | 1.03 (0.81 to 1.31) | 0.8 |
| Tiglylcarnitine | 2.26 (1.74 to 2.92)a | 6.50E−10 | 1.41 (1.11 to 1.81) | 5.30E−03 |
| Pantothenate | 1.51 (1.15 to 1.97) | 2.60E−03 | 1.44 (1.11 to 1.85) | 5.30E−03 |
| Serine | 0.56 (0.31 to 1) | 5.20E−02 | 0.53 (0.28 to 1.01) | 5.20E−02 |
| N1-methylinosine | 1.28 (0.95 to 1.73) | 0.11 | 1.21 (0.96 to 1.52) | 0.11 |
| C-glycosyltryptophan | 2.71 (1.7 to 4.34) | 3.10E−05 | 1.28 (0.94 to 1.75) | 0.12 |
| Oxalate (ethanedioate) | 1.08 (0.82 to 1.42) | 0.61 | 1.45 (1.12 to 1.88) | 4.60E−03 |
| N-acetylleucine | 1.71 (1.19 to 2.44) | 3.50E−03 | 1.16 (0.84 to 1.61) | 0.37 |
Statistically significant associations are shown in bold. Cut-off for statistical significance=0.05/58=8.6×10−4. Results were adjusted for baseline (AASK) or 12-month visit (MDRD) age, sex, trial arms, history of cardiovascular disease, history of smoking, body mass index, serum albumin concentration, measured GFR, and in the MDRD study, race. AASK, African American Study of Kidney Disease and Hypertension; MDRD, Modification of Diet in Renal Disease; HR, hazard ratio; 95% CI, 95% confidence interval; GPE, glycero-3-phosphoethanolamine; GPI, glycero-3-phosphoinositol; GPC, glycero-3-phosphocholine.
Hazard ratio remained statistically significant after additional adjustment for baseline urine protein-to-creatinine ratio.
Discussion
Few studies have evaluated blood metabolomic associations with proteinuria in CKD. Using data from the AASK and the MDRD study, both rigorously conducted clinical trials with per-protocol measures of 24-hour proteinuria and GFR, we identified 58 metabolites, including 4-hydroxychlorothalonil and 1,5-AG, and a metabolic pathway (PEs) that had strong cross-sectional associations with proteinuria among 1582 patients with CKD, independently of demographics, serum albumin concentration, measured GFR, and other clinical risk factors. Several of these metabolites were associated with CKD progression in the AASK, with four also associated with ESKD in the MDRD study.
Among all identified metabolites, 4-hydroxychlorothalonil and 1,5-AG showed the strongest and most consistent associations with proteinuria across studies. 4-Hydroxychlorothalonil is the primary degradation product of chlorothalonil, which is a commonly used, broad-spectrum, nonsystemic protectant pesticide that binds to sulfhydryl groups and noncompetitively inhibits glyceraldehyde-3-phosphate dehydrogenase (15). Because chlorothalonil is used on crops, lawns, and golf courses (16), serum levels of this compound and its degradants may be reflective of the amount of plant-based dietary intake or outdoor activity and thus an individual’s general health. Although low-dose exposure to chlorothalonil is considered safe in humans, its metabolism in humans has yet to be adequately elucidated. Feeding studies showed variable urinary excretion of chlorothalonil (4%–12%) (17), and animal experiments suggested that chlorothalonil may produce hyperplasia of the proximal tubular epithelium (18), altered corticosterone levels, immune cell activity, and mortality (19). Compared with its parent compound, chlorothalonil, 4-hydroxychlorothalonil may cause more damage in the liver and bone marrow, but there is less documentation of its effects on the kidney (18). Interestingly, in our study, lower serum 4-hydroxychlorthalonil was seen with higher levels of proteinuria. In the context of CKD, such an association might have been driven by impaired proximal tubule function, or reflect alterations to the metabolism and/or elimination of chlorothalonil in the kidney. Unfortunately, our assays did not identify chlorothalonil, and urine samples were not available for metabolomic profiling. Future studies using targeted assays in serum and urine may help elucidate the metabolism and excretion of these molecules.
1,5-AG is a monosaccharide derived primarily from food (20). It is filtered in the glomerulus and reabsorbed by the proximal tubule, with high levels of glucose a competitive inhibitor of tubular reabsorption of 1,5-AG (21). The urinary fractional excretion of 1,5-AG varies from 0% (healthy individuals) to approximately 10% in people with CKD (22). In people with diabetes, lower 1,5-AG has been shown to correlate cross-sectionally with microvascular complications, such as retinopathy and albuminuria (23,24), and may mark risk for incident CKD and CKD progression (25,26). The association between 1,5-AG and proteinuria may reflect poor proximal tubule function in the presence of proteinuria (27), leading to impaired reabsorption and lower serum levels of 1,5-AG, or it may be a sign of undiagnosed diabetes in our study populations (the prevalence of diagnosed diabetes was 0% in the AASK and 5% in the MDRD study). However, fasting glucose among most participants was within normal limits, the correlation between 1,5-AG and fasting glucose was only modest, and the association of 1,5-AG with proteinuria remained strong after further adjustment for fasting glucose.
In addition to 4-hydroxychlorthalonil and 1,5-AG, we also identified 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4), along with all other five metabolites in the PE pathway, as positively associated with proteinuria. There were relatively strong within-pathway correlations across the PE metabolites, and after accounting for these correlations using a permutation test, the evidence of overrepresentation was weakened (permutation P value=0.01). PEs are the second most abundant class of phospholipids in eukaryotic cell membranes, and have diverse cellular functions, such as post-translational modifications, oxidative phosphorylation, and mitochondrial biogenesis (28). They have been implicated in various diseases (including Alzheimer disease and congenital disorders) (29,30), yet the molecular basis underlying many of their biologic roles remains unclear. Abnormal lipid metabolism is common in kidney disease, most prominently in the nephrotic syndrome. Although the mechanism is not completely understood, it is possible that disorders in lipid metabolism are mediated by factors such as low oncotic pressure, hypoalbuminemia (although our findings are adjusted for serum albumin concentration, residual confounding or measurement error may exist), and resultant compensatory upregulation of lipoprotein synthesis (31,32). Further, a previous study involving urinary metabolomic profiling in patients with AKI and kidney transcriptomic analyses in a murine acute tubular necrosis model suggested that the PCYT2 gene, which encodes for CTP:phosphoenthanolamine cytidyltransferase, the rate-limiting enzyme in the Kennedy pathway of PE biosynthesis (33), was adaptively downregulated so as to preserve phosphoethanolamine, a precursor of PEs that is potentially nephroprotective by reducing mitochondrial oxidative stress in tubular epithelial cells (34,35). On the basis of these observations, the positive association between PEs and proteinuria observed in our study suggests that mitochondrial oxidative stress resulting from failure of such mechanism might play a role in the pathogenesis of proteinuria in CKD.
The strengths of this study include the broad range of compounds investigated via an unbiased approach and the combination of results from two rigorously conducted clinical trials, which involved >1500 participants with a wide range of demographic and clinical characteristics. However, there are noteworthy limitations. First, like all observational studies, causality cannot be inferred. This is a cross-sectional study. Temporality of the observed associations thus cannot be determined. Second, our data originated from clinical trials, which may limit generalizability. The AASK recruited only patients with CKD attributed to hypertension, excluding people with diabetes, and the MDRD study recruited very few patients with diabetes. Third, levels of metabolites were quantified on a relative scale, limiting comparisons across populations. Targeted assays could provide more precise and comparable assessments. Fourth, we were able to analyze data on urine protein-to-creatinine ratio only, rather than the preferred measure for CKD staging, urine albumin-to-creatinine ratio, or other specific proteins. However, urine protein-to-creatinine ratio correlates well with urine albumin-to-creatinine ratio, and both measures are valid predictors of common CKD complications (36,37). Fifth, our assessment of enriched pathways necessarily focuses on those pathways and corresponding metabolites that were included on the Metabolon platform and is thus incomplete. The platform only partially covers glycerolipids (triacylglycerols, in particular) and sterol lipids, which are important components of the human lipidome and prone to change with alterations in urine protein excretion. Sixth, metabolomic profiling was conducted in specimens at one point in time, limiting our ability to account for within-person variability in metabolite levels. Nonetheless, we were able to find strong associations that were consistent across studies and robust to adjustment for multiple covariates. Seventh, although we adopted a meta-analytic approach for the analysis of proteinuria, our results were largely driven by findings in the AASK, which had higher metabolite measurement reliability and a larger sample size.
In conclusion, untargeted metabolomic profiling demonstrated several serum metabolites (including 4-hydroxychlorthalonil and 1,5-AG) and a metabolic pathway (PEs) that were strongly associated with proteinuria in CKD. Several of these metabolites were also associated with ESKD. With replication, these findings may serve as a basis for future investigations into the pathophysiology of proteinuria, risk prediction, and the development of novel preventive or treatment measures in CKD.
Disclosures
Assay costs were discounted as part of a collaboration agreement between Metabolon, Inc., J. Coresh, L.A.I., and A.S.L. to develop a product to estimate GFR from a panel of biomarkers, for which a provisional patent was filed on August 15, 2014 (PCT/US2015/044567). The technology is not licensed in part or in whole to any company. H.I.F. reports consulting activities with Kyowa Hakko Kirin, Inc.
Supplementary Material
Acknowledgments
M.E.G. receives support from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (grant R01 DK108803). A.T. receives support from NIDDK (grant R01 DK108803). J. Coresh, A.H.A., H.I.F., S.S.W., L.A.I., A.S.L., M.E.G., and M.J.S. receive support from CKD Biomarkers Consortium (NIDDK grants U01 DK085689, U01 DK102730, U01 DK103225, and U01 DK 085660). J. Coresh, L.A.I., and A.S.L. receive support from CKD Epidemiology Collaboration panel eGFR (grant R01 DK097020). C.M.R. is supported by a Mentored Research Scientist Development Award from the NIDDK (grant K01 DK107782). The work of A.K. is supported by the German Research Foundation (DFG) 3598/3-1, DFG 3598/4-1, and DFG Collaborative Research Centre (CRC).
This work was presented at an oral abstract session on October 25, 2018 during the American Society of Nephrology Kidney Week in San Diego, CA.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
Supplemental Table of Contents
This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.10010818/-/DCSupplemental.
Supplemental Table 1. Pathways and identifiers of metabolites significantly associated with proteinuria.
Supplemental Table 2. Associations of significant metabolites with cross-sectional eGFR in the AASK and the MDRD study, and eGFR change over time in the AASK.
Supplemental Appendix 1. Metabolite profiling procedures.
Supplemental Appendix 2. Evaluated pathways and number of named, nondrug metabolites with <80% missing data.
References
- 1.Levey AS, de Jong PE, Coresh J, El Nahas M, Astor BC, Matsushita K, Gansevoort RT, Kasiske BL, Eckardt KU: The definition, classification, and prognosis of chronic kidney disease: A KDIGO Controversies Conference report. Kidney Int 80: 17–28, 2011 [DOI] [PubMed] [Google Scholar]
- 2.Kidney Disease Improving Global Outcomes : Clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl 3: 1–150, 2013 [Google Scholar]
- 3.Kunz R, Friedrich C, Wolbers M, Mann JF: Meta-analysis: Effect of monotherapy and combination therapy with inhibitors of the renin angiotensin system on proteinuria in renal disease. Ann Intern Med 148: 30–48, 2008 [DOI] [PubMed] [Google Scholar]
- 4.Patti GJ, Yanes O, Siuzdak G: Innovation: Metabolomics: The apogee of the omics trilogy. Nat Rev Mol Cell Biol 13: 263–269, 2012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Johnson CH, Ivanisevic J, Siuzdak G: Metabolomics: Beyond biomarkers and towards mechanisms. Nat Rev Mol Cell Biol 17: 451–459, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hocher B, Adamski J: Metabolomics for clinical use and research in chronic kidney disease. Nat Rev Nephrol 13: 269–284, 2017 [DOI] [PubMed] [Google Scholar]
- 7.Grams ME, Shafi T, Rhee EP: Metabolomics research in chronic kidney disease. J Am Soc Nephrol 29: 1588–1590, 2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kalim S, Rhee EP: An overview of renal metabolomics. Kidney Int 91: 61–69, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Gassman JJ, Greene T, Wright JT Jr ., Agodoa L, Bakris G, Beck GJ, Douglas J, Jamerson K, Lewis J, Kutner M, Randall OS, Wang SR: Design and statistical aspects of the African American Study of Kidney Disease and Hypertension (AASK). J Am Soc Nephrol 14[Suppl 2]: S154–S165, 2003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.The Modification of Diet in Renal Disease study: Design, methods, and results from the feasibility study. Am J Kidney Dis 20: 18–33, 1992 [DOI] [PubMed] [Google Scholar]
- 11.Pesce MA, Strande CS: A new micromethod for determination of protein in cerebrospinal fluid and urine. Clin Chem 19: 1265–1267, 1973 [PubMed] [Google Scholar]
- 12.Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E: Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small-molecule complement of biological systems. Anal Chem 81: 6656–6667, 2009 [DOI] [PubMed] [Google Scholar]
- 13.Goeman JJ, Bühlmann P: Analyzing gene expression data in terms of gene sets: Methodological issues. Bioinformatics 23: 980–987, 2007 [DOI] [PubMed] [Google Scholar]
- 14.Xia J, Wishart DS: Using MetaboAnalyst 3.0 for Comprehensive Metabolomics Data Analysis. Curr Protoc Bioinformatics 55: 14.10.1–14.10.91, 2016 [DOI] [PubMed] [Google Scholar]
- 15.Long JW, Siegel MR: Mechanism of action and fate of the fungicide chlorothalonil (2,4,5,6-tetrachloroisophthalonitrile) in biological systems. 2. In vitro reactions. Chem Biol Interact 10: 383–394, 1975 [DOI] [PubMed] [Google Scholar]
- 16.US EPA: Registration Eligibility Decision for Chlorothalonil, 1999. Available at: https://archive.epa.gov/pesticides/reregistration/web/pdf/0097red.pdf. Accessed on August 2, 2018
- 17.World Health Organization : Environ Health Criteria 183, 1996. Available at: http://www.inchem.org/documents/ehc/ehc/ehc183.htm. Accessed August 2, 2018 [Google Scholar]
- 18.Food and Agriculture Organization of the United Nations: Chlorothalonil. Available at: http://www.fao.org/fileadmin/templates/agphome/documents/Pests_Pesticides/JMPR/Report09/Chlorothalonil.pdf. Accessed on August 15, 2018
- 19.McMahon TA, Halstead NT, Johnson S, Raffel TR, Romansic JM, Crumrine PW, Boughton RK, Martin LB, Rohr JR: The fungicide chlorothalonil is nonlinearly associated with corticosterone levels, immunity, and mortality in amphibians. Environ Health Perspect 119: 1098–1103, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yamanouchi T, Tachibana Y, Akanuma H, Minoda S, Shinohara T, Moromizato H, Miyashita H, Akaoka I: Origin and disposal of 1,5-anhydroglucitol, a major polyol in the human body. Am J Physiol 263: E268–E273, 1992 [DOI] [PubMed] [Google Scholar]
- 21.Buse JB, Freeman JL, Edelman SV, Jovanovic L, McGill JB: Serum 1,5-anhydroglucitol (GlycoMark ): A short-term glycemic marker. Diabetes Technol Ther 5: 355–363, 2003 [DOI] [PubMed] [Google Scholar]
- 22.Shimizu H, Shouzu A, Nishikawa M, Omoto S, Hayakawa T, Miyake Y, Yonemoto T, Inada M: Serum concentration and renal handling of 1,5-anhydro-D-glucitol in patients with chronic renal failure. Ann Clin Biochem 36: 749–754, 1999 [DOI] [PubMed] [Google Scholar]
- 23.Selvin E, Francis LM, Ballantyne CM, Hoogeveen RC, Coresh J, Brancati FL, Steffes MW: Nontraditional markers of glycemia: Associations with microvascular conditions. Diabetes Care 34: 960–967, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kim WJ, Park CY, Park SE, Rhee EJ, Lee WY, Oh KW, Park SW, Kim SW, Park HS, Kim YJ, Song SJ, Ahn HY: Serum 1,5-anhydroglucitol is associated with diabetic retinopathy in Type 2 diabetes. Diabet Med 29: 1184–1190, 2012 [DOI] [PubMed] [Google Scholar]
- 25.Selvin E, Rawlings AM, Grams M, Klein R, Steffes M, Coresh J: Association of 1,5-anhydroglucitol with diabetes and microvascular conditions. Clin Chem 60: 1409–1418, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Rebholz CM, Grams ME, Chen Y, Gross AL, Sang Y, Coresh J, Selvin E: Serum levels of 1,5-anhydroglucitol and risk of incident end-stage renal disease. Am J Epidemiol 186: 952–960, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zoja C, Morigi M, Remuzzi G: Proteinuria and phenotypic change of proximal tubular cells. J Am Soc Nephrol 14[Suppl 1]: S36–S41, 2003 [DOI] [PubMed] [Google Scholar]
- 28.Calzada E, Onguka O, Claypool SM: Chapter two - phosphatidylethanolamine metabolism in health and disease. In: International Review of Cell and Molecular Biology, edited by Jeon KW, Cambridge, MA, Academic Press, 2016, pp 29–88 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Onodera T, Futai E, Kan E, Abe N, Uchida T, Kamio Y, Kaneko J: Phosphatidylethanolamine plasmalogen enhances the inhibiting effect of phosphatidylethanolamine on γ-secretase activity. J Biochem 157: 301–309, 2015 [DOI] [PubMed] [Google Scholar]
- 30.Dorninger F, Brodde A, Braverman NE, Moser AB, Just WW, Forss-Petter S, Brügger B, Berger J: Homeostasis of phospholipids - The level of phosphatidylethanolamine tightly adapts to changes in ethanolamine plasmalogens. Biochim Biophys Acta 1851: 117–128, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Joven J, Villabona C, Vilella E, Masana L, Albertí R, Vallés M: Abnormalities of lipoprotein metabolism in patients with the nephrotic syndrome. N Engl J Med 323: 579–584, 1990 [DOI] [PubMed] [Google Scholar]
- 32.Clement LC, Macé C, Avila-Casado C, Joles JA, Kersten S, Chugh SS: Circulating angiopoietin-like 4 links proteinuria with hypertriglyceridemia in nephrotic syndrome. Nat Med 20: 37–46, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Arthur G, Page L: Synthesis of phosphatidylethanolamine and ethanolamine plasmalogen by the CDP-ethanolamine and decarboxylase pathways in rat heart, kidney and liver. Biochem J 273: 121–125, 1991 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Martin-Lorenzo M, Gonzalez-Calero L, Ramos-Barron A, Sanchez-Niño MD, Gomez-Alamillo C, García-Segura JM, Ortiz A, Arias M, Vivanco F, Alvarez-Llamas G: Urine metabolomics insight into acute kidney injury point to oxidative stress disruptions in energy generation and H2S availability. J Mol Med (Berl) 95: 1399–1409, 2017 [DOI] [PubMed] [Google Scholar]
- 35.Kishi S, Campanholle G, Gohil VM, Perocchi F, Brooks CR, Morizane R, Sabbisetti V, Ichimura T, Mootha VK, Bonventre JV: Meclizine preconditioning protects the kidney against ischemia-reperfusion injury. EBioMedicine 2: 1090–1101, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Collier G, Greenan MC, Brady JJ, Murray B, Cunningham SK: A study of the relationship between albuminuria, proteinuria and urinary reagent strips. Ann Clin Biochem 46: 247–249, 2009 [DOI] [PubMed] [Google Scholar]
- 37.Fisher H, Hsu CY, Vittinghoff E, Lin F, Bansal N: Comparison of associations of urine protein-creatinine ratio versus albumin-creatinine ratio with complications of CKD: A cross-sectional analysis. Am J Kidney Dis 62: 1102–1108, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.