Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Feb 1.
Published in final edited form as: Kidney Int. 2021 Nov 26;101(2):369–378. doi: 10.1016/j.kint.2021.10.035

A metabolomics approach identified toxins associated with uremic symptoms in advanced chronic kidney disease.

Jiun-Ruey Hu 1,*, Leslie Myint 2,*, Andrew S Levey 3, Josef Coresh 4, Lesley A Inker 3, Morgan E Grams 4,5, Eliseo Guallar 4, Kasper D Hansen 6, Eugene P Rhee 7, Tariq Shafi 8
PMCID: PMC8792216  NIHMSID: NIHMS1766195  PMID: 34843755

Abstract

Uremic symptoms are common in patients with advanced chronic kidney disease, but the toxins that cause these symptoms are unknown. To evaluate this, we performed a cross-sectional study of the 12 month post-randomization follow-up visit of Modification of Diet in Renal Disease (MDRD) participants reporting uremic symptoms who also had available stored serum. We quantified 1,163 metabolites by liquid chromatography-tandem mass spectrometry. For each uremic symptom, we calculated a score as the severity multiplied by the number of days the symptom was experienced. We analyzed the associations of the individual symptom scores with metabolites using linear models with empirical Bayesian inference, adjusted for multiple comparisons. Among 695 participants, the mean measured glomerular filtration rate (mGFR) was 28 mL/min/1.73 m2. Uremic symptoms were more common in the subgroup of 214 patients with an mGFR under 20 mL/min/1.73 m2 (mGFR under 20 subgroup) than in the full group. For all metabolites with significant associations, the direction of the association was concordant in the full group and the subgroup. For gastrointestinal symptoms (bad taste, loss of appetite, nausea, and vomiting), eleven metabolites were associated with symptoms. For neurologic symptoms (decreased alertness, falling asleep during the day, forgetfulness, lack of pep and energy, tiring easily/weakness), seven metabolites were associated with symptoms. Associations were consistent across sensitivity analyses. Thus, our proof of principle study demonstrates the potential for metabolomics to understand metabolic pathways associated with uremic symptoms. Larger, prospective studies with external validation are needed.

Keywords: uremic symptoms, uremia, chronic kidney disease, metabolomics

Graphical Abstract

graphic file with name nihms-1766195-f0004.jpg

Introduction

Symptoms of uremia, such as lack of energy, weakness, anorexia, pruritus, nausea, and vomiting, are common causes of morbidity and poor quality of life in patients with advanced chronic kidney disease (CKD).1,2 Uremic symptoms are often the reason for initiating dialysis.3 Uremic symptoms are fully resolved by kidney transplantation, but are only partially alleviated by dialysis, suggesting that retained uremic toxins, not effectively removed by dialysis, cause uremic symptoms. However, the specific toxins causing uremic symptoms remain unknown. Knowledge of specific uremic toxins will help improve the management of uremia for patients with advanced CKD before or after the onset of treatment by dialysis.

The advent of high-throughput mass spectrometry and liquid chromatography has allowed for identifying and quantifying many metabolites in health and disease, a technique referred to as metabolomic profiling. Metabolomic profiling can identify disordered metabolic pathways and holds the potential to unravel the causes of uremic symptoms. Although several recent studies have used this technique to explore metabolic alterations in patients with kidney failure treated by dialysis,4–8 no studies have specifically investigated uremic symptoms in patients with earlier stages of kidney disease.

In this proof-of-principle study, we sought to examine the associations between the metabolomic profile and symptom profile of patients with CKD (not on dialysis). We hypothesized that untargeted metabolomics may identify metabolites associated with specific symptoms. We tested our hypothesis using data and samples from the Modification of Diet in Renal Disease (MDRD) Study, relating patient-reported specific uremic symptoms to individual metabolites.

Methods

Study Population

The MDRD study was a multi-center, randomized clinical trial investigating the effects of dietary protein restriction and blood pressure control on CKD progression.9–11 Participants were allocated to two studies based on mGFR: Participants with mGFR between 25–55 mL/min/1.73m2 were assigned to Study A, while participants with mGFR between 13–24 mL/min/1.73m2 were assigned to Study B. Within each study, participants were randomly assigned using a two-by-two factorial design to a dietary protein intervention (moderate [Diet M] vs. low [Diet L] protein and phosphorus in Study A; low [Diet L] vs. very low [Diet K] protein and phosphorus in Study B), and a blood pressure intervention (usual or low target blood pressure in both studies). At each visit, fasting blood samples were obtained, aliquoted, and frozen at −80° C. A total of 840 participants were enrolled between 1989 and 1991, among which 751 participants had available serum samples at the 12-month post-randomization visit. Of the 751 participants, we excluded 55 with missing all data on symptoms and covariates, and one with <20% metabolites measured, resulting in a final analytic cohort of 695 participants. This study’s protocol was reviewed and approved by the Johns Hopkins Institutional Review Board.

Uremic Symptoms

The MDRD study assessed uremic symptoms using the Patient Symptom Form (PSF) first described in 1997, and which has been shown to have significant positive correlations with the Symptom Checklist-90R (SCL-90R), which is a global measure of mental health, and significant negative correlations with the Quality of Well Being (QWB) scale, which is a general health-related quality of life index.12 The PSF has also been used in other large CKD cohorts, such as the Chronic Renal Insufficiency Cohort.13 From the list of symptoms, we chose bad taste, loss of appetite, nausea, vomiting, itching, lack of pep and energy, tiring easily/weakness, numbness/tingling, daytime sleepiness, decreased alertness, and forgetfulness as uremic symptoms as they are clinically well-recognized as symptoms of advanced kidney failure and are likely due to retention of uremic toxins.14 The participants rated the severity of each symptom on a scale of 0 to 3, denoting none, mild, moderate, or severe, and the number of days that the symptom was experienced in the past month. Responses were highly complete (Table S1). We examined metabolite-symptom associations in both the total group and in the subgroup of patients with mGFR <20 ml/min/1.73 m2 (mGFR<20 subgroup) because the symptoms are expected to worsen with advancing kidney failure. mGFR<20 ml/min/1.73 m2 corresponds to the lowest tertile of mGFR in the MDRD Study, and 20 ml/min/1.73 m2 is also the threshold for listing for kidney transplantation15 and referral for dialysis access planning.

Metabolomic Profiling

Fasting serum samples were sent to Metabolon, Inc. (Durham, North Carolina) for metabolite profiling. Upon receipt, samples were stored at −80° C until processing via Metabolon’s standard protocol using reverse-phase ultra-high performance liquid chromatography-tandem mass spectrometry, as previously described (see Supplemental Methods).16 For protein-bound compounds, the concentration included the total concentration (including bound and free concentrations). Metabolite relative abundances were quantified by the area under the curve. Metabolomic quality control was ensured using Metabolon’s standard procedures (Supplemental Methods). We included twenty blind duplicate pairs, which showed that 68% of all metabolites had a correlation >0.80.17 The correlation between untargeted serum creatinine on the Metabolon platform and targeted creatinine assessed by the Jaffe method was 0.96 (Figure S1). We also conducted a principal components analysis and did not find any evidence of persistent batch effects (Supplemental Methods; Figure S2). The metabolite profiling process identified 1,193 metabolites in 695 samples, excluding the one participant with <20% metabolites measured.

Covariates

We a priori identified covariates to be included in adjustment models, including age, sex, race, mGFR, history of smoking (ever versus never), diabetes, cause of kidney disease, systolic blood pressure, 24-hour urine protein and urea nitrogen, body mass index (kg/m2), Study A vs. B, and randomization to diet and blood pressure interventions. Ascertainment of these variables has been previously described.9 Cause of kidney disease was classified as glomerular disease (GD), polycystic kidney disease (PKD), or non-GD non-PKD kidney disease. In particular, participants’ mGFR was assessed by urinary iothalamate clearance.

Statistical Analysis

We described participant characteristics for the total group and the mGFR<20 subgroup, summarizing continuous and categorical variables using means with standard deviations or percentages, respectively. We calculated a uremic symptom score for each symptom as the product of the severity of the symptom [scale, 0 (none) to 3 (severe)], and the number of days the symptom was experienced (range, 0–31).12,13 We analyzed each symptom as a continuous variable. In sensitivity analyses, we also analyzed each symptom as a categorical variable with five categories, informed by symptom distribution in the cohort, including a no symptom category and 4 categories based on symptom severity level (1 or >1), and duration (≤15 days or >15 days). Missing data for covariates was <5%, and we imputed missing values using a singleimputation random forest approach implemented in the missForest R package,18 with procedures to assess the accuracy of imputation (Supplemental Section C). All statistical analyses were conducted in R. Statistical packages and functions are denoted with italic font.

For the metabolomics data, we retained only the metabolites measured in at least 10 samples with nonzero variance across samples. This removed 36 metabolites, resulting in 1,157 total metabolites available for analysis in the 695 participants (Table S2). We transformed metabolites as log base 2 (log2) of the metabolites’ abundances to normalize variances and allow comparisons between metabolites. Missing data in metabolomics studies are often due to limit of detection issues, a type of left censoring. We used a quantile regression imputation procedure (QRILC) for imputing missing metabolite data19 and empirically verified the accuracy of imputation (Supplemental Methods C).

Our analytic approach used a differential analysis framework that is best practice in the high-throughput assay literature.20 We modeled log2 metabolite levels (outcome or dependent variable) as a function of symptom scores (exposure or independent variable) and covariates, using the limma software package.21 The limma method uses an empirical Bayes estimator of metabolite-specific moderated residual variances from individual metabolite-wise linear models. This Bayes estimator provides better estimates of variability than the standard linear regression estimator. In this analysis, we used imputed symptom and covariate data to use information from as many samples as possible but used un-imputed metabolomics data to limit the potential biasing impact of imputation inaccuracies. While estimating moderated variances, we retained unidentified and xenobiotic metabolites to improve estimation but removed them after completion of modeling, resulting in 588 metabolites. We adjusted the p-values for multiple testing using the Benjamini-Hochberg procedure with a false discovery rate of 10% and used the adjusted p-values to rank the metabolites’ association with each symptom. We used the symptom score coefficient as our effect measure, which gives the adjusted log-fold change in the metabolite level per unit change in the symptom score. For presentation of the associations and visualization of the data, we grouped symptoms as gastrointestinal symptoms (bad taste, loss of appetite, nausea, and vomiting) and neurologic symptoms (decreased alertness, falling asleep during day, forgetfulness, lack of pep and energy, and tiring easily/weakness).

In sensitivity analyses, we focused on the statistical significance, effect direction (positively or negatively associated), and effect magnitude (fold change in metabolites with change in symptom) for each metabolite across all combinations of imputed and non-imputed data and across analyses in which symptoms were modeled as either a continuous or categorical variable in the overall cohort and in mGFR<20 subgroup (Supplemental Section D; Table S5). While we adjusted for study A vs B and randomization to diet and blood pressure intervention groups in the main analysis as well as for their achieved blood pressure (ascertained as systolic blood pressure) and dietary protein intake (ascertained as urine urea nitrogen excretion rate), we undertook an additional sensitivity analysis in the total group to understand if the results would be consistent in participants with different levels of dietary protein intake. We performed an as-treated analysis by defining low, medium, and high dietary protein intake subgroups based on tertiles of urine urea nitrogen, in order to assess patients by their actual protein intake rather than their assigned protein intake. We did not perform these subgroup analyses in conjunction with the mGFR<20 subgroup analysis due to severe reductions in sample size.

Results

Baseline Characteristics

Among the 695 participants in the total group, 38% were female, 86% were White, and 77% were never smokers (Table 1). They had a median BMI of 26.3 kg/m2 and median systolic blood pressure of 129 mm Hg. The cause of CKD was glomerular disease in 28%, polycystic kidney disease in 24%, and other in 43%. Among the 214 participants with mGFR <20 subgroup, 43% were female, 83% were White, and 77% were never smokers. They had a median BMI of 25.0 kg/m2 and a median systolic blood pressure of 130 mm Hg. The cause of CKD was glomerulonephritis in 30%, polycystic kidney disease in 28%, and other non-hypertensive non-diabetic etiology in 41%.

Table 1:

Summary statistics of participants in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20 subgroup”).

Total Group (N=695) mGFR<20 Subgroup (N=214)
Median (IQR) or N (%) Median (IQR) or N (%)
Age, years 52 (43, 61) 53 (40, 60)
Female Sex 267 (38.4%) 92 (43.0%)
White 596 (85.9%) 178 (83.2%)
Cause of Kidney Disease
 Glomerular disease 197 (28.3%) 64 (30.0%)
 Polycystic kidney disease 168 (24.2%) 59 (27.6%)
 Absence of one kidney 21 (3.0%) 7 (3.3%)
 Hypertensive nephrosclerosis 44 (6.3%) 15 (7.0%)
 Tubulointerstitial diseases 32 (4.6%) 9 (4.2%)
 Urinary tract diseases 27 (3.9%) 10 (4.7%)
 Other or unknown 175 (25.2%) 46 (21.5%)
Body Mass Index, kg/m2 26.3 (23.9, 28.9) 25.0 (22.8, 27.2)
Systolic Blood Pressure, mm Hg 129 (118, 141) 130 (120, 143)
mGFR, mL/min/1.73 m2 28.0 (18.0, 39.3) 14.7 (11.5, 17.4)
Hemoglobin (g/dL) 12.9 (11.6, 14.1) 11.4 (10.3, 12.7)
Serum albumin (g/dL) 4.1 (3.9, 4.3) 4.1 (3.9, 4.3)
Serum phosphorus (mg/dL) 3.6 (3.2, 4.1) 4.2 (3.6, 4.9)
Urine protein, g/day 0.2 (0.1, 1.1) 0.5 (0.2, 1.6)
Urine urea nitrogen, g/day 7.0 (5.2, 10.2) 5.5 (4.6, 7.0)
Never smoker 536 (77.1%) 165 (77.1%)
Type II Diabetes 35 (5.0%) 11 (5.1%)
Hypertension 580 (83.4%) 181 (84.6%)
Coronary artery disease 68 (9.8%) 23 (10.7%)
Hyperlipidemia 163 (23.5%) 42 (19.6%)
Cerebrovascular disease 10 (1.4%) 3 (1.4%)
Peripheral vascular disease 29 (4.2%) 5 (2.3%)
Peptic ulcer disease 26 (3.7%) 11 (5.1%)
Cancer 18 (2.6%) 7 (3.3%)
Open in a new tab

Abbreviations: IQR: interquartile range; mGFR: measured GFR

Distribution of Uremic Symptoms

Except for numbness/tingling, a higher proportion of participants in the mGFR<20 subgroup experienced all uremic symptoms of interest than did the total group. The difference was most noticeable for loss of appetite (19.6% vs 13.8%), nausea (28.8% vs 19.2%), lack of pep and energy (50.0% vs 38.2%), and tiring easily (51.9% vs 37.8%) (Table 2). The distribution of uremic symptoms was similar across the studies and randomized groups (Figure 1).

Table 2:

Distribution of uremic symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20 Subgroup”).

Tota Group mGFR<20 Subgroup
Symptom (and question phrasing) Prevalence, N (%) Symptom score (if symptomatic), median (IQR) Prevalence, N (%) Symptom score (if symptomatic), median (IQR)
Bad taste “A bad taste in your mouth?” 134/695 (19.3%) 7 (3, 20) 52/214 (24.3%) 10 (5, 30)
Loss of appetite “A loss of appetite?” 96/695 (13.8%) 6 (3, 15) 42/214 (19.6%) 5 (3, 15)
Nausea “Nausea or being sick to your stomach?” 133/693 (19.2%) 5 (2, 10) 61/212 (28.8%) 6 (2, 12)
Vomiting “Vomiting?” 52/695 (7.5%) 3 (1, 7) 25/214 (11.7%) 3 (2, 10)
Itching “Itching?” 144/694 (20.7%) 10 (4, 26) 53/213 (24.9%) 10 (3, 20)
Lack of pep and energy “Lack of pep and energy?” 265/693 (38.2%) 15 (5, 30) 106/212 (50.0%) 20 (6, 30)
Tiring easily “Tiring easily, weakness?” 262/693 (37.8%) 16 (5, 30) 110/212 (51.9%) 20 (7, 30)
Daytime sleepiness “Falling asleep during the day?” 141/693 (20.3%) 10 (3, 20) 54/212 (25.5%) 10 (3, 30)
Decreased alertness “Decreased alertness?” 88/693 (12.7%) 8 (4, 30) 28/212 (13.2%) 6 (4, 30)
Forgetfulness “Forgetfulness?” 140/695 (20.1%) 10 (3, 30) 46/214 (21.5%) 15 (4, 30)
Numbness/tingling “Numbness and tingling in your hands and feet?” 131/694 (18.9%) 10 (4, 30) 40/213 (18.8%) 8 (5, 20)
Open in a new tab

For each symptom, the original phrasing of the question on the patient symptom form is included.

Figure 1:

Figure 1:

Open in a new tab

Fraction of participants experiencing uremic symptoms in the MDRD study (“Total group”, colored red) and the subgroup of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20 subgroup”, colored blue), according to study assignment and randomized group. The dashed vertical grey lines separate related groups of symptoms by organ system.

Metabolites Associated with Uremic Symptoms

For all metabolites with significant associations with uremic symptoms, the direction of the association was concordant overall and the mGFR<20 subgroup (Figures 2 and 3). In other words, there were no metabolites for which the association was positive overall and negative in the mGFR<20 subgroup, or vice versa.

Figure 2: Comparison of metabolite-symptom strength of associations for the gastrointestinal uremic symptoms (bad taste, loss of appetite, nausea, and vomiting).

Figure 2:

Open in a new tab

(A) Total group, and (B) mGFR <20 subgroup.

Metabolites depicted have a statistically significant association [false discovery rate (FDR) < 0.1] with at least one of the four symptoms in this group. Metabolites in red font color at the top of each diagram have a significant association with at least one symptom in the total group and the mGFR<20 subgroup. Metabolite-symptom associations are plotted on the horizontal axis by their metabolite fold change per 30 unit change in symptom score. Filled circles indicate significance after accounting for an FDR of 0.1, while crossmarks (X) indicate lack of significance. Markers to the right of 1.0 (>1.0) on the horizontal axis denote a positive association whereby higher symptom score was associated with higher metabolite level. Markers to the left of 1.0 (<1.0) on the horizontal axis denote a negative association, whereby a higher symptom score was associated with a lower metabolite level.

Figure 3: Comparison of metabolite-symptom strength of associations for the neurologic uremic symptoms (loss of decreased alertness, daytime sleepiness, forgetfulness, lack of pep and energy, and tiring easily/weakness).

Figure 3:

Open in a new tab

(A) Total group, and (B) mGFR <20 subgroup.

Metabolites depicted have a statistically significant association [false discovery rate (FDR) < 0.1] with at least one of the four symptoms in this group. Metabolites in red font color at the top of each diagram have a significant association with at least one symptom in the total group and the mGFR<20 subgroup. Metabolite-symptom associations are plotted on the horizontal axis by their metabolite fold change per 30 unit change in symptom score. Filled circles indicate significance after accounting for an FDR of 0.1, while crossmarks (X) indicate lack of significance. Markers to the right of 1.0 (>1.0) on the horizontal axis denote a positive association whereby higher symptom score was associated with higher metabolite level. Markers to the left of 1.0 (<1.0) on the horizontal axis denote a negative association, whereby a higher symptom score was associated with a lower metabolite level.

For gastrointestinal symptoms (bad taste, loss of appetite, nausea, and vomiting), 11 metabolites were associated with symptoms overall and mGFR<20 subgroup (Figure 2 and Table S3) with a statistically significant positive association for 7 metabolites and a statistically significant negative association for 4 metabolites. Thirteen metabolites were associated with gastrointestinal symptoms overall but not in the mGFR<20 subgroup, whereas, 18 metabolites were associated with gastrointestinal symptoms in only the mGFR<20 subgroup, but not the total group. One metabolite (methylsuccinoylcarnitine) was positively associated with both loss of appetite and nausea overall and mGFR<20 subgroup. Hexanoylcarnitine was positively associated with both bad taste and loss of appetite in the mGFR<20 subgroup but the total group.

For the neurologic symptoms (decreased alertness, falling asleep during day, forgetfulness, lack of pep and energy, tiring easily/weakness), 7 metabolites were associated with symptoms overall and mGFR<20 subgroups (Figure 3 and Table S3) with a statistically significant positive association for 6 metabolites and a statistically significant negative association for 1 metabolite (cortisol). Twelve metabolites were associated with neurologic uremic symptoms for the total group but not in the mGFR<20 subgroup, whereas, 10 metabolites were associated with symptoms in only the mGFR<20 subgroup, but not the total group. Two metabolites (indoleacetylglutamine and 4-acetamidobutanoate) were positively associated with both lack of energy and tiring easily in the total group and mGFR<20 subgroup. In the mGFR<20 subgroup, 3 metabolites (phenylpyruvate, 6-oxopiperidine-2-carboxylic acid, and 5-hydroxylysine) were each positively associated with both lack of energy and tiring easily. There were no metabolites associated with itching (including phosphate) or numbness/tingling.

Sensitivity Analyses

The sensitivity analyses are presented in Figures S3 and S4. The direction of the association of the metabolites was generally similar to the primary analyses. In general, for loss of appetite the strength of association appeared stronger in analyses with symptoms modeled as a categorical variable, suggesting possibility of threshold effects. The associations between symptoms and metabolites were generally stronger for the gastrointestinal symptoms than the neurologic symptoms. There was no significant association of any metabolite with itching or numbness/tingling in all sensitivity analyses.

We further performed an as-treated analysis by defining low, medium, and high dietary protein intake subgroups based on tertiles of urine urea nitrogen excretion rate, in order to assess if results would differ based on participants’ actual protein intake rather than their assigned protein intake. We found that of the original set of significant metabolites, all but 2 metabolites showed identical effect directions in the original and protein subgroup analyses (Table S4).

Discussion

The metabolic alterations that underlie most uremic symptoms are not known. We conducted a proof-of-principle study to determine if a comprehensive metabolomics approach can identify metabolites associated with uremic symptoms. Among 695 MDRD study participants in our study population, we were able to identify several metabolites associated with bad taste, loss of appetite, nausea, lack of energy, and tiring easily. To our knowledge, this is the first study to explore metabolomics as a tool to evaluate specific patient-reported uremic symptoms and highlight the potential of metabolomics to identify disrupted metabolic pathways that cause uremic symptoms in patients with advanced CKD who have not yet initiated dialysis.

Uremic symptoms are common in advanced CKD and are often the reason for dialysis initiation. However, despite many advances in dialysis care delivery and dialysis technique over the last several decades, the prevalence of uremic symptoms in patients on dialysis remains unchanged.22 Uremic symptoms are also a major contributor to poor health-related quality of life (HRQOL) in patients on dialysis, for whom HRQOL is often a greater concern than survival. When asked about the possibility of improvement in HRQOL or survival by switching to intensive hemodialysis, 94% of patients would consider it to improve energy (reducing fatigue), but only 19% would consider it to improve survival.23 A recent survey of patients with kidney failure either treated by dialysis or nearing dialysis initiation and their care providers concluded that “the best ways to manage symptoms in people receiving or nearing dialysis, including poor energy (fatigue) and nausea” were top research priorities.24,25 However, since the specific mechanisms of uremic symptoms are unknown, the approaches to manage uremic symptoms are non-specific. Identifying specific toxins and biological pathways is necessary to develop novel therapeutic strategies for managing uremic symptoms and uremia.

We noted several interesting associations in our study. The metabolites methylsuccinoylcarnitine and hexanoylcarnitine were positively associated with two gastrointestinal symptoms. Several other acylcarnitines (deoxcycarnitine, methylglutarylcarnitine, adipoylcarnitine, suberoylcarnitine, propionylcarnitine, isobutyrylcarnitine, glutarylcarnitine) were also associated with symptoms in this group, suggesting a class effect rather than a single causal factor. These molecules represent short- and medium-chain carnitine species, which undergo significant kidney clearance, unlike long-chain acylcarnitines.26 Notably, a recent genome-wide association study showed that a missense variant in ACADM can also impact circulating medium-chain acylcarnitine levels in CKD, but whether this variant is associated with risk of uremic symptoms is unknown.27 Further, each of these molecules is characterized by a trimethylamine moiety, which emits a fishy odor.28

For neurologic symptoms, we noted a strong and positive association of p-cresol glucuronide with lack of pep and energy. P-cresol glucuronide shares the same precursor as p-cresol sulfate, a widely studied uremic solute.29,30 P-cresol is an endproduct of tyrosine metabolism by colonic bacteria and after absorption, it is converted to mainly p-cresol sulfate and a small extent to p-cresol glucuronide.31 P-cresol glucuronide has been associated with mortality in patients on hemodialysis.31 Its association with lack of pep and energy may suggest either a direct effect or it may be a surrogate for other gut-derived uremic toxins contributing to uremic symptoms. Second, we also noted a strong negative association between cortisol and decreased alertness in the total group and decreased alertness and forgetfulness in the mGFR<20 subgroup. The direction of association of cortisol with forgetfulness, decreased alertness, lack of pep and energy, and tiring easily/weakness is consistent with cortisol’s known effects. The association of primary adrenal insufficiency (Addison’s Disease) with fatigue, sleep disturbances, and mood disorder is well-known. In particular, sleep disturbances can contribute to cognitive impairment in Addison’s Disease, sometimes referred to as metabolic dementia.32 Third, we noted a positive association between higher indoleacetylglutamine and greater lack of pep and energy and greater tiring easily/weakness. Indoleacetylglutamine is generated from indoleacetic acid, a well-known uremic solute which is an agonist of the transcription factor aryl hydrocarbon receptor (AhR), whose activation promotes vascular inflammation, oxidative stress, and atherosclerosis.33,34 Because AhR is a receptor that is widely expressed in the central nervous system,35 elevations in indole-family uremic solutes can lead to neurologic symptoms. Other indole-family compounds have been implicated in the progression of Alzheimer’s disease and regulation of the mucosal barrier.36,37 In recent experiments in rats with CKD, experimentally-increased serum indoxyl sulfate concentrations (by increased feeding of dietary indoxyl sulfate) barrier – an effect that did not occur in non-CKD AhR−/− knockout rats.38 Fourth, we noted a positive association between higher hydantoin-5-propionic acid and greater lack of pep and energy and greater tiring easily/weakness. Hydantoin-5-propionic acid is a metabolite of histidine formed by the oxidation of imidazolonepropionic acid, and is usually lowly excreted in the urine. Patients with deficiencies of B12 or folate metabolism have been reported to produce abnormally high urinary concentrations of hydantoin-5-propionic acid.39 Thus, hydantoin-5-propionic acid has been proposed as an a surrogate index for disturbed folate or Vitamin B12 metabolism.39 Folate and B12 deficiencies have well-established neuropsychiatric consequences such as forgetfulness, cognitive change, and peripheral neuropathy.40 Because folate and vitamin B12 levels were not measured in MDRD, whether elevated hydantoin-5-propionic acid levels are a surrogate for disturbed folate or vitamin B12 metabolism is uncertain. Fifth, we noted a positive association between higher 4-acetamidobutanoate and more tiring easily/weakness and greater lack of pep and energy in the full cohort and mGFR<20 subgroup. A metabolite of the urea cycle, 4-acetamidobutanoate is associated with CKD progression.41–43 It is also reported to be a gamma-aminobutyric acid (GABA) derivative, and hence, an association with fatigue is biologically plausible.42

Our sample size was small, and so non-significant associations should not be disregarded altogether. We noted that in most cases, while a particular metabolite would be significantly associated with just one symptom in the symptom group, the remaining metabolite-symptom associations in the symptom group would share an effect in the same direction despite not meeting statistical significance. For example, methylsuccinoylcarnitine was associated with bad taste, loss of appetite, and nausea overall, and, although the association in the mGFR<20 subgroup was only significant for loss of appetite and nausea, the direction of the (non-significant) association for bad taste was similar to the total group. Thus, the remaining metabolite-symptom associations might be significant in a study with larger sample size and greater power. The congruence of the direction and strength of associations in the sensitivity analyses across imputed and non-imputed data also supports this hypothesis. The congruence of the direction and strength of associations in the main analysis and the sensitivity analysis for urine urea nitrogen reassures us that the metabolite-symptom associations reported here are not being driven by MDRD study interventions. Because the power to detect differences is decreased in the mGFR<20 subgroup compared to the total group, the metabolite-symptom associations that were significant overall but not the mGFR<20 subgroup should not be taken literally as unimportant in the mGFR<20 subgroup. Larger, better-powered studies will be necessary to address this limitation.

Other limitations of our study also deserve mention. First, in this proof-of-principle pilot study, we did not have an external validation cohort and so our results should be considered as hypothesis-generating. However, we note that the MDRD Study was a national multicenter study and thus less subject to bias than a single center or single clinical practice study that would be subject to bias by regional factors. Of note, replication of the cross-sectional association in a second cohort would not be sufficient to prove causality. The steps towards demonstrating a causal association include: 1) demonstrating biological plausibility in vitro and in animal models; 2) replicating the association in the same patients over time (intra-individual validation); and 3) demonstrating the attenuation of the association with removal or reducing the toxin load. The latter, removal or reducing the load, would require a randomized clinical trial. Our study represents the first step towards this goal. Second, the known causes of CKD in MDRD Study were predominantly polycystic kidney disease and glomerular disease. Future studies should investigate metabolite-symptom associations in CKD associated with diabetes and hypertension. Our study’s major strengths include the use of data from a well-known clinical trial with measured GFR, the use of a robust metabolomics platform, use of an unbiased discovery approach, and correction for multiple comparisons.

In conclusion, our study highlights the potential for metabolomics to unravel the disrupted metabolic pathways contributing to uremic symptoms in patients with advanced CKD. Large, well-designed studies with discovery and external validation cohorts are needed to advance our understanding of uremia and devise therapies to alleviate or cure uremic symptoms.

Supplementary Material

1

Figure S1: Correlation between untargeted serum creatinine on the Metabolon platform from samples of the F12 visit after freezing at −80°C until 2017 metabolite profiling and targeted creatinine assessed by the Jaffe method from samples of the F12 visit measured 3 decades prior

Figure S2: PCA score plot for assessment of batch effects

Figure S3: Sensitivity analyses for symptom-metabolite associations for the gastrointestinal uremic symptoms (bad taste, loss of appetite, nausea, and vomiting)

Figure S4: Sensitivity analyses for symptom-metabolite associations for the neurologic uremic symptoms (loss of decreased alertness, daytime sleepiness, forgetfulness, lack of pep and energy, and tiring easily/weakness)

Table S1: Non-metabolite variables with missing data

Table S2: Process flow for metabolite identification and selection, with number of metabolites at each stage

Table S3: Metabolites associated with different symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20”).

Table S4: For metabolites that were significantly associated with a symptom in the full group main analysis, we show the metabolite’s adjusted fold change per 30-unit increase in the symptom score in the original analysis and in the three dietary protein subgroups defined by tertiles of urine urea nitrogen

Table S5: Variations of parameters used for 16 sensitivity analyses

Table S6: Xenobiotic metabolites associated with gastrointestinal uremic symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20”)

Table S7: Xenobiotic metabolites associated with neurologic uremic symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20”)

Acknowledgments

Metabolomic profiling in the MDRD study was supported by the CKD Biomarkers Consortium (National Institute of Diabetes and Digestive and Kidney Diseases [NIDDK] U01 DK085689). JRH was supported by the Linda Kao Memorial Award from the Welch Center. TS and EPR are supported by R01NR017399. Icons for the Visual Abstract were sourced from the Noun Project and BioRender. This project was presented at the 2021 National Young Investigators’ Forum of the National Kidney Foundation.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Disclosure Statement

No financial conflicts of interest

References

  • 1.Abdel-Kader K, Unruh ML, Weisbord SD. Symptom Burden, Depression, and Quality of Life in Chronic and End-Stage Kidney Disease. Clin J Am Soc Nephrol. 2009;4(6):1057–1064. doi: 10.2215/CJN.00430109 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Goeij MCM de, Ocak G, Rotmans JI, Eijgenraam J-W, Dekker FW, Halbesma N. Course of Symptoms and Health-Related Quality of Life during Specialized Pre-Dialysis Care. PLOS ONE. 2014;9(4):e93069. doi: 10.1371/journal.pone.0093069 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Daugirdas JT, Depner TA, Inrig J, et al. KDOQI Clinical Practice Guideline for Hemodialysis Adequacy: 2015 Update – Guideline 1: Timing of Hemodialysis Initiation. American Journal of Kidney Diseases. 2015;66(5):884–930. doi: 10.1053/j.ajkd.2015.07.015 [DOI] [PubMed] [Google Scholar]
  • 4.Boelaert J, t’Kindt R, Schepers E, et al. State-of-the-art non-targeted metabolomics in the study of chronic kidney disease. Metabolomics. 2014;10(3):425–442. doi: 10.1007/s11306-013-0592-z [DOI] [Google Scholar]
  • 5.Brand JAJG van den, Mutsaers HAM, Zuilen AD van, et al. Uremic Solutes in Chronic Kidney Disease and Their Role in Progression. PLOS ONE. 2016;11(12):e0168117. doi: 10.1371/journal.pone.0168117 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Rhee EP, Souza A, Farrell L, et al. Metabolite Profiling Identifies Markers of Uremia. JASN. 2010;21(6):1041–2051. doi: 10.1681/ASN.2009111132 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Tamura MK, Chertow GM, Depner TA, et al. Metabolic Profiling of Impaired Cognitive Function in Patients Receiving Dialysis. JASN. Published online July 21, 2016:ASN.2016010039. doi: 10.1681/ASN.2016010039 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Tanaka H, Sirich TL, Plummer NS, Weaver DS, Meyer TW. An Enlarged Profile of Uremic Solutes. PLOS ONE. 2015;10(8):e0135657. doi: 10.1371/journal.pone.0135657 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Greene T, Bourgoignie JJ, Habwe V, et al. Baseline characteristics in the Modification of Diet in Renal Disease Study. JASN. 1993;4(5):1221–1236. [DOI] [PubMed] [Google Scholar]
  • 10.Klahr S The Modification of Diet in Renal Disease Study. New England Journal of Medicine. 1989;320(13):864–866. [DOI] [PubMed] [Google Scholar]
  • 11.Klahr S, Levey AS, Beck GJ, et al. The Effects of Dietary Protein Restriction and Blood-Pressure Control on the Progression of Chronic Renal Disease. New England Journal of Medicine. 1994;330(13):877–884. doi: 10.1056/NEJM199403313301301 [DOI] [PubMed] [Google Scholar]
  • 12.Rocco MV, Gassman JJ, Wang SR, Kaplan RM. Cross-sectional study of quality of life and symptoms in chronic renal disease patients: the Modification of Diet in Renal Disease Study. Am J Kidney Dis. 1997;29(6):888–896. [DOI] [PubMed] [Google Scholar]
  • 13.Zhang X, Bansal N, Go AS, Hsu C. Gastrointestinal symptoms, inflammation and hypoalbuminemia in chronic kidney disease patients: a cross-sectional study. BMC Nephrology. 2015;16:211. doi: 10.1186/s12882-015-0209-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Meyer TW, Hostetter TH. Uremia. New England Journal of Medicine. 2007;357(13):1316–1325. doi: 10.1056/NEJMra071313 [DOI] [PubMed] [Google Scholar]
  • 15.Organ Procurement and Transplantation Network (OPTN). Organ Procurement and Transplantation Network Policies, Policy 8.4: Waiting Time Published online April 6, 2017. Accessed May 4, 2017. https://optn.transplant.hrsa.gov/media/1200/optn_policies.pdf#nameddest=Policy_08
  • 16.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 smallmolecule complement of biological systems. Anal Chem. 2009;81(16):6656–6667. doi: 10.1021/ac901536h [DOI] [PubMed] [Google Scholar]
  • 17.Hu J-R, Coresh J, Inker LA, et al. Serum metabolites are associated with all-cause mortality in chronic kidney disease. Kidney International. 2018;94(2):381–389. doi: 10.1016/j.kint.2018.03.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Stekhoven DJ, Buehlmann P. MissForest - nonparametric missing value imputation for mixed-type data. Bioinformatics. 2012;28(1):112–118. doi: 10.1093/bioinformatics/btr597 [DOI] [PubMed] [Google Scholar]
  • 19.Wei R, Wang J, Su M, et al. Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data. Sci Rep 2018;8(1):663. doi: 10.1038/s41598-017-19120-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Kammers K, Cole RN, Tiengwe C, Ruczinski I. Detecting significant changes in protein abundance. EuPA Open Proteomics. 2015;7:11–19. doi: 10.1016/j.euprot.2015.02.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Smyth GK. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004;3:Article3. doi: 10.2202/1544-6115.1027 [DOI] [PubMed] [Google Scholar]
  • 22.Rhee EP, Guallar E, Hwang S, et al. Prevalence and Persistence of Uremic Symptoms in Incident Dialysis Patients. Kidney360. 2020;1(2):86–92. doi: 10.34067/kid.0000072019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Ramkumar N, Beddhu S, Eggers P, Pappas LM, Cheung AK. Patient preferences for in-center intense hemodialysis. Hemodialysis International. 2005;9(3):281–295. doi: 10.1111/j.1492-7535.2005.01143.x [DOI] [PubMed] [Google Scholar]
  • 24.Manns B, Hemmelgarn B, Lillie E, et al. Setting Research Priorities for Patients on or Nearing Dialysis. CJASN. 2014;9(10):1813–1821. doi: 10.2215/CJN.01610214 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Nissenson AR. Improving Outcomes for ESRD Patients: Shifting the Quality Paradigm. Clin J Am Soc Nephrol. 2014;9(2):430–434. doi: 10.2215/CJN.05980613 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Kalim S, Clish CB, Wenger J, et al. A plasma long-chain acylcarnitine predicts cardiovascular mortality in incident dialysis patients. J Am Heart Assoc. 2013;2(6):e000542. doi: 10.1161/JAHA.113.000542 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Li Y, Sekula P, Wuttke M, et al. Genome-Wide Association Studies of Metabolites in Patients with CKD Identify Multiple Loci and Illuminate Tubular Transport Mechanisms. JASN. 2018;29(5):1513–1524. doi: 10.1681/ASN.2017101099 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Phillips IR, Shephard EA. Primary Trimethylaminuria. In: Adam MP, Ardinger HH, Pagon RA, et al. , eds. GeneReviews®. University of Washington, Seattle; 1993. Accessed December 30, 2020. http://www.ncbi.nlm.nih.gov/books/NBK1103/ [Google Scholar]
  • 29.Shafi T, Sirich TL, Meyer TW, et al. Results of the HEMO Study suggest that p-cresol sulfate and indoxyl sulfate are not associated with cardiovascular outcomes. Kidney International. 2017;92(6):1484–1492. doi: 10.1016/j.kint.2017.05.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Vanholder R, Schepers E, Pletinck A, Nagler EV, Glorieux G. The Uremic Toxicity of Indoxyl Sulfate and p-Cresyl Sulfate: A Systematic Review. JASN. 2014;25(9):1897–1907. doi: 10.1681/ASN.2013101062 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Liabeuf S, Glorieux G, Lenglet A, et al. Does P-Cresylglucuronide Have the Same Impact on Mortality as Other Protein-Bound Uremic Toxins? PLOS ONE. 2013;8(6):e67168. doi: 10.1371/journal.pone.0067168 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Henry M, Wolf PSA, Ross IL, Thomas KGF. Poor quality of life, depressed mood, and memory impairment may be mediated by sleep disruption in patients with Addison’s disease. Physiol Behav. 2015;151:379–385. doi: 10.1016/j.physbeh.2015.08.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Dou L, Sallée M, Cerini C, et al. The Cardiovascular Effect of the Uremic Solute Indole-3 Acetic Acid. JASN. 2015;26(4):876–887. doi: 10.1681/ASN.2013121283 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Gondouin B, Cerini C, Dou L, et al. Indolic uremic solutes increase tissue factor production in endothelial cells by the aryl hydrocarbon receptor pathway. Kidney International. 2013;84(4):733–744. doi: 10.1038/ki.2013.133 [DOI] [PubMed] [Google Scholar]
  • 35.Latchney SE, Hein AM, O’Banion MK, DiCicco-Bloom E, Opanashuk LA. Deletion or activation of the aryl hydrocarbon receptor alters adult hippocampal neurogenesis and contextual fear memory. Journal of Neurochemistry. 2013;125(3):430–445. doi: 10.1111/jnc.12130 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Chyan Y-J, Poeggeler B, Omar RA, et al. Potent Neuroprotective Properties against the Alzheimer β-Amyloid by an Endogenous Melatonin-related Indole Structure, Indole-3-propionic Acid*. Journal of Biological Chemistry. 1999;274(31):21937–21942. doi: 10.1074/jbc.274.31.21937 [DOI] [PubMed] [Google Scholar]
  • 37.Pavlova T, Vidova V, Bienertova-Vasku J, et al. Urinary intermediates of tryptophan as indicators of the gut microbial metabolism. Anal Chim Acta. 2017;987:72–80. doi: 10.1016/j.aca.2017.08.022 [DOI] [PubMed] [Google Scholar]
  • 38.Bobot M, Thomas L, Moyon A, et al. Uremic Toxic Blood-Brain Barrier Disruption Mediated by AhR Activation Leads to Cognitive Impairment during Experimental Renal Dysfunction. JASN. 2020;31(7):1509–1521. doi: 10.1681/ASN.2019070728 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Duran M, Bruinvis L, Wadman SK. Quantitative gas chromatographic determination of urinary hydantoin-5-propionic acid in patients with disorders of folate/vitamin B12 metabolism. Journal of Chromatography B: Biomedical Sciences and Applications. 1986;381:401–405. doi: 10.1016/S0378-4347(00)83606-3 [DOI] [PubMed] [Google Scholar]
  • 40.Reynolds E Vitamin B12, folic acid, and the nervous system. The Lancet Neurology. 2006;5(11):949–960. doi: 10.1016/S1474-4422(06)70598-1 [DOI] [PubMed] [Google Scholar]
  • 41.Mindikoglu AL, Opekun AR, Putluri N, et al. Unique metabolomic signature associated with hepatorenal dysfunction and mortality in cirrhosis. Transl Res. 2018;195:25–47. doi: 10.1016/j.trsl.2017.12.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.The Human Metabolome Database. 4-Acetamidobutanoic acid (HMDB0003681). HMDB. Published February 26, 2020. https://hmdb.ca/metabolites/HMDB0003681 [Google Scholar]
  • 43.Rhee EP, Zheng Z, Grams M, et al. Metabolomics of CKD Progression in CRIC and AASK. Oral Presentation presented at the: American Society of Nephrology (ASN) Kidney Week; November 9, 2019; Washington, D.C. Accessed December 14, 2020. https://www.asn-online.org/education/kidneyweek/2019/programabstract.aspx?controlId=3235636 [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1

Figure S1: Correlation between untargeted serum creatinine on the Metabolon platform from samples of the F12 visit after freezing at −80°C until 2017 metabolite profiling and targeted creatinine assessed by the Jaffe method from samples of the F12 visit measured 3 decades prior

Figure S2: PCA score plot for assessment of batch effects

Figure S3: Sensitivity analyses for symptom-metabolite associations for the gastrointestinal uremic symptoms (bad taste, loss of appetite, nausea, and vomiting)

Figure S4: Sensitivity analyses for symptom-metabolite associations for the neurologic uremic symptoms (loss of decreased alertness, daytime sleepiness, forgetfulness, lack of pep and energy, and tiring easily/weakness)

Table S1: Non-metabolite variables with missing data

Table S2: Process flow for metabolite identification and selection, with number of metabolites at each stage

Table S3: Metabolites associated with different symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20”).

Table S4: For metabolites that were significantly associated with a symptom in the full group main analysis, we show the metabolite’s adjusted fold change per 30-unit increase in the symptom score in the original analysis and in the three dietary protein subgroups defined by tertiles of urine urea nitrogen

Table S5: Variations of parameters used for 16 sensitivity analyses

Table S6: Xenobiotic metabolites associated with gastrointestinal uremic symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20”)

Table S7: Xenobiotic metabolites associated with neurologic uremic symptoms in the MDRD study (“Total Group”) and the subset of patients with measured GFR (mGFR) < 20 ml/min/1.73 m2 (“mGFR<20”)

NIHMS1766195-supplement-1.pdf (983.9KB, pdf)

RESOURCES