Increased urinary osmolyte excretion indicates chronic kidney disease severity and progression rate

Increased urinary osmolyte excretion indicates chronic kidney disease severity and progression rate Abstract Background Chronic kidney disease (CKD) is a recognized global health problem. While some CKD patients remain stable after initial diagnosis, others can rapidly progress towards end-stage renal disease (ESRD). This makes biomarkers capable of detecting progressive forms of CKD extremely valuable, especially in non-invasive biofluids such as urine. Screening for metabolite markers using non-targeted metabolomic techniques like nuclear magnetic resonance spectroscopy is increasingly applied to CKD research. Methods A cohort of CKD patients (n = 227) with estimated glomerular filtration rates (eGFRs) ranging from 9.4–130 mL/min/1.73 m2 was evaluated and urine metabolite profiles were characterized in relation to declining eGFR. Nested in this cohort, a retrospective subset (n = 57) was investigated for prognostic metabolite markers of CKD progression, independent of baseline eGFR. A transcriptomic analysis of murine models of renal failure was performed to validate selected metabolomic findings. Results General linear modeling revealed 11 urinary metabolites with significant associations to reduced eGFR. Linear modelling specifically showed that increased urine concentrations of betaine (P < 0.05) and myo-inositol (P < 0.05) are significant prognostic markers of CKD progression. Conclusions Renal organic osmolytes, betaine and myo-inositol play a critical role in protecting renal cells from hyperosmotic stress. Kidney tissue transcriptomics of murine preclinical experimentation identified decreased expression of Slc6a12 and Slc5a11 mRNA in renal tissue consistent with defective tubular transport of these osmolytes. Imbalances in renal osmolyte regulation lead to increased renal cell damage and thus more progressive forms of CKD. Increases in renal osmolytes in urine could provide clinical diagnostic and prognostic information on CKD outcomes. CKD, disease progression, metabolomics, osmolytes, transcriptomics, urine INTRODUCTION Chronic kidney disease (CKD) is a term used to include a wide range of diseases that reduce kidney function. The National Kidney Foundation defines CKD as ≥3 months of kidney damage or an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2 [1]. Disease aetiology in CKD is diverse, leading to variability in morbidity and mortality [2]. Further, not all populations have an equal risk for developing CKD and progressing towards end-stage renal disease (ESRD); therefore, high-risk populations with diabetes and/or hypertension should receive targeted testing for CKD [2]. eGFR can be calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation or the Modification of Diet in Renal Disease (MDRD) study equation [1, 2]. However, both measures are relatively insensitive in early stage CKD and their prognostic value is limited. Novel biomarkers that are more sensitive in early stage CKD are needed for diagnosis and prognosis of CKD progression [3]. Metabolomics can analyse biofluids for specific metabolic signatures reflecting CKD status [4], offering promising non-invasive methods to screen, diagnose and prognose patients for CKD [5–7]. Mass spectrometry (MS) and nuclear magnetic resonance (NMR) are the two main analytical platforms used in metabolomics [8]. Here we use NMR spectroscopy as a quantitative state-of-the-art analytical platform that is widely used in metabolome studies [9]. NMR can detect a wide range of metabolites from various chemical classes, delivering robust transferable data under optimized conditions [10]. The purpose of this study was to use NMR-based metabolomic strategies to elucidate novel urinary markers supportive of current methods for diagnosing CKD. Urine samples from a cohort of 227 patients in various stages of CKD were utilized. An additional aim was to identify prognostic markers that indicate patients at risk of CKD progression towards ESRD. A nested retrospective subgroup (n = 57) from the cohort was analysed using follow-up eGFR measurements and a calculated rate of CKD progression. Results of these analyses point to increased urine concentrations of renal osmolytes as markers of reduced eGFR. These compounds were also significantly elevated in the urine of CKD patients with progressive forms of CKD. These metabolite data were supported with messenger RNA (mRNA) data generated from independent murine kidney fibrosis and injury experimentation. MATERIALS AND METHODS Study cohort and sampling The cohort consisted of patients from the Nephrology Outpatient Clinic at Ghent University Hospital, Ghent, Belgium. Samples were collected in the period 17 January 2011 to 7 July 2015. eGFR was determined using the CKD-EPI equation. Important metadata used in the analysis can be found in Table 1. Sample collection was performed in accordance with local ethics requirements and in accordance with the current revision of the Helsinki Declaration. Table 1. Urine sample cohort (n = 227) Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  This table provides an overview of cohort metadata used for biomarker discovery. Age, gender, BMI and CKD disease groups were determined to be the most important metadata in linear modelling of metabolites. For age, BMI and eGFR, the mean and standard deviation are shown with the range. a Systemic diseases includes diabetes type 2 and hypertension. Table 1. Urine sample cohort (n = 227) Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  This table provides an overview of cohort metadata used for biomarker discovery. Age, gender, BMI and CKD disease groups were determined to be the most important metadata in linear modelling of metabolites. For age, BMI and eGFR, the mean and standard deviation are shown with the range. a Systemic diseases includes diabetes type 2 and hypertension. A total of 57 patients had sufficient follow-up data to monitor their respective rate of CKD progression (i.e. ≥3 eGFR measurements, minimum of 2-years duration). To reduce confounding factors, only samples from patients with systemic diseases or glomerular diseases were used, and patients with renal transplantation or receiving renal replacement therapy were not considered (Supplementary Table S1). The percent annual eGFR slope change from baseline was calculated using the following equation, an approach that has also been previously applied [11]:   % annual eGFR slope= linear coefficient baseline eGFR×100% Sample preparation for NMR and instrumental protocol Upon collection, urine samples were immediately centrifuged (20800 g for 10 min), aliquoted per 1 mL and stored at −80°C. Following defrosting, a buffering approach designed for NMR analysis in urine was applied [12]. This included a deuterium oxide phosphate buffer (1.5 M PO4, pH 7.4; Armar Chemicals, Leipzig, Germany) with potassium fluoride (KF, 300 mM) and (trimethylsilyl)propionic acid (TSP, 0.1%). Phosphate salts, KF and TSP were purchased from Sigma Aldrich Chemie (Steinheim, Germany). Urine samples were analysed on a Bruker 800-MHz spectrometer operating at 800.35 MHz with a quadrupole inverse cryogenic probe. A standard one-dimensional pulse sequence [recycle delay (RD) 90°, t1 90°, tm 90°, acquire FID] was acquired, with water suppression irradiation during RD of 2 s, mixing time (tm) set on 200 ms and a 90° pulse set to 10.13 μs, collecting 512 scans into 64 000 data points with a spectral width of 12 ppm. All spectra were manually phased, baseline corrected and calibrated to TSP (δ 0.00) with TopSpin 3.2 (Bruker BioSpin, Rheinstetten, Germany). Data were imported to MATLAB (MathWorks, Natick, MA, USA) and further processed, i.e., water and urea region removed (δ 4.7–5.6). Daily quality control samples (i.e. pooled urine sample) were measured throughout the cohort measurement to monitor instrument performance, data processing (i.e. spectra pre-processing, normalization and alignment) and metabolite stability. The structural identity of metabolite compounds was determined by two-dimensional (2D) NMR methods; total correlation spectroscopy (TOCSY) and heteronuclear single quantum coherence (HSQC) spectroscopy. For 2D TOCSY spectra, phase-sensitive and sensitivity-improved 2D TOCSY with WATERGATE (3-9-19) and DIPSI-2 were acquired. For each spectrum 19228 × 1024 data points were collected using 32 scans per increment, 16 dummy scans and 1 s acquisition time. The spectral widths were set to 12 ppm for both the F1 and F2 dimensions. For 2D HSQC spectra, phase-sensitive ge-2D HSQC using PEP and adiabatic pulses for inversion and refocusing with gradients was used. Data points of 4788 × 1280 per spectra were collected with 256 scans per increment, 16 dummy scans and an acquisition time of 0.25 s. Spectral widths were set to 12 ppm for the F2 and 230 ppm for F1. NMR-based statistical tools Data processing and statistics were performed in MATLAB 2015b, RStudio and the MetaboAnalyst platform (http://www.metaboanalyst.ca) [13]. Probabilistic quotient normalization [14] was used for normalization of spectral data to account for biological variation in urine dilution. This method is an improved normalization technique compared with single-metabolite normalization [15], which is superior in the context of CKD, as it avoids normalization to metabolites affected by the disease. Spectral data were aligned using the recursive segment-wise peak alignment (RSPA) algorithm [16]. An orthogonal partial least squares (OPLS) regression analysis [17] using eGFR as the y variable was performed for spectral feature selection. Spectral regions most affected by CKD display high covariance and r2 values. Selected spectral features were structurally identified using 2D NMR and matched with the human metabolite database (HMDB) [18]. Identified metabolites were quantified with the R-package BATMAN [19], measuring the area under the curve. Data were then log transformed and mean centered. General linear modelling was applied to each identified metabolite with baseline eGFR or percent annual eGFR slope as y variables, adjusted for age, gender, body mass index (BMI) and disease group. Kidney transcriptomics Total RNA was isolated using the PureLink RNA Mini Kit (Invitrogen, Paisley, UK). Affymetrix transcriptomics arrays of kidney tissue (n = 3 control and n = 3 kidney injury, obtained 24 h after induction of kidney injury by a folic acid injection or injection of vehicle) were performed at Unidad Genómica Moncloa, Fundación Parque Científico de Madrid, Madrid, Spain, following the manufacturer’s protocol [20]. Image files were initially obtained through Affymetrix GeneChip Command Console software. Robust multichip analysis was performed using the Affymetrix Expression Console software. Starting from the normalized robust multichip analysis, the significance analysis of microarrays was performed using the limma package (Babelomics, http://www.babelomics.org), with a false discovery rate (FDR) of 5% to identify genes significantly dysregulated. Experimental kidney injury These studies were approved by the IIS-FJD animal ethics committee and followed Directive 2010/63/EU on the protection of animals used for scientific purposes. For experimental kidney injury, female 12- to 14-week-old C57/BL6 mice received a single intraperitoneal injection of folic acid (Sigma Aldrich) 250 mg/kg in sodium bicarbonate 0.3 mol/L (acute kidney injury, n = 3) or a vehicle alone (controls, n = 3) and were sacrificed 24 h later as previously described [20]. Kidneys were cold saline perfused in situ before removal. One kidney from each mouse was fixed in buffered formalin, embedded in paraffin and stained with haematoxylin and eosin. The other kidney was snap frozen in liquid nitrogen for ribonucleic acid (RNA) studies. Samples from this experiment were used for kidney transcriptomics. A second set of six mice per group with folic acid nephropathy or vehicle-infused controls were processed in the same manner at 24 and 72 h following the induction of kidney injury and used for real-time quantitative polymerase chain reaction (qPCR) validation. Unilateral ureteral obstruction (UUO) was performed under isoflurane-induced anaesthesia. The left ureter was ligated with silk at two locations and cut to prevent urinary tract infection (obstructed kidney) [21]. A total of five male 12- to 14-week-old C57/BL6 mice were sacrificed 14 days after surgery. UUO kidneys were compared with the healthy contralateral kidneys. Real-time qPCR validation Human metabolomics and murine transcriptomics results were validated by real-time qPCR performed using mRNA obtained from five UUO mice sacrificed on Day 14 and from folic acid nephropathy mice vehicle-infused controls obtained at 24 and 72 h following induction of kidney injury. RNA of 1 µg isolated by Trizol (Invitrogen) was reverse transcribed with a High Capacity cDNA Archive Kit and real-time qPCR was performed on a ABI Prism 7500 PCR system (Applied Biosystems, Foster City, CA, USA) using the ΔΔCt method [22]. Expression levels are displayed as ratios to glyceraldehyde-3-phosphate dehydrogenase. Pre-developed primer and probe assays were also from Applied Biosystems. Data mining The Nephroseq [http://www.nephroseq.org/ (11 October 2017, date last accessed)] database was searched for human renal biopsy transcriptomics data comparing patients with diverse aetiologies of CKD with controls. The database was searched for gene expression data for Slc5a11, Slc6a12 and Slc5A3. Supplementary Figure S1 shows the experimental workflow. RESULTS Urine metabolite markers related to eGFR An OPLS regression analysis of the entire cohort data set measured by 1H NMR revealed spectral features with significant linear relationship to eGFR (Figure 1). Spectral features with the largest absolute covariance and r2 values (Figure 1C) were targeted for identification and quantification (Figure 1A). A list of the identified metabolites can be found in Supplementary Table S2. General linear modelling of all quantified metabolites showed 11 urinary metabolites having significant linear associations with eGFR, adjusted for age, gender, BMI and disease group (Table 2). A positive association with decreased eGFR reflects a decrease in the relative concentration of each respective metabolite as eGFR declines. Table 2. Significant diagnostic metabolites of eGFR Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  This table shows general linear modelling results for diagnostic markers of baseline eGFR. These metabolites show significant linear associations with eGFR. The β coefficient, standard error (SE) and P-value for each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group. Metabolites with a negative correlation (i.e. negative β) indicate metabolites that have increased urinary concentration as eGFR declines. Table 2. Significant diagnostic metabolites of eGFR Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  This table shows general linear modelling results for diagnostic markers of baseline eGFR. These metabolites show significant linear associations with eGFR. The β coefficient, standard error (SE) and P-value for each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group. Metabolites with a negative correlation (i.e. negative β) indicate metabolites that have increased urinary concentration as eGFR declines. FIGURE 1: View largeDownload slide OPLS regression analysis for feature selection in NMR spectra. (A) Sample spectrum with identified metabolites labelled: (1) 2-furoylglycine, (2) 2-hydroxyisobutyric acid, (3) 3-hydroxyisovaleric acid, (4) 4-deoxyerythronic acid, (5) 4-deoxytheonic acid, (6) acetic acid, (7) acetone, (8) trigonelline, (9) betaine, (10) cis-aconitic acid, (11) citric acid, (12) creatine, (13) creatinine, (14) D-glucose, (15) dimethylamine, (16) dimethlyglycine, (17) ethanolamine, (18) formic acid, (19) fumaric acid, (20) glycine, (21) glycolic acid, (22) guanidoacetic acid, (23) pseudouridine, (24) hippuric acid, (25) indoxyl sulfate, (26) L-acetylcarnitine, (27) L-alanine, (28) trimethylamine, (29) L-lactic acid, (30) L-phenylalanine, (31) L-threonine, (32) trimethylamine-N-oxide, (33) L-tyrosine, (34) uracil, (35) myo-inositol, (36) N-methylnicotinamide, (37) p-cresol sulfate, (38) phenylacetylglutamine, (39) phosphorylcholine, (40) p-hydroxyphenylacetic acid, (41) proline betaine, (42) succinic acid, (43) tartaric acid and (44) taurine. (B) Loadings plot of OPLS regression analysis with eGFR in the y variable. This plot represents a ‘statistical’ pseudo-spectrum that displays the OPLS model covariance on the y-axis and chemical shift on the x-axis. The colour bar represents the r2 value for that spectral region to eGFR. Plots in (A) and (B) are aligned on the x-axis, which shows which spectral regions may best describe CKD-related changes and warrant structural identification. (C) Scores plot of the OPLS model showing data separation based on sample eGFR (colour bar). FIGURE 1: View largeDownload slide OPLS regression analysis for feature selection in NMR spectra. (A) Sample spectrum with identified metabolites labelled: (1) 2-furoylglycine, (2) 2-hydroxyisobutyric acid, (3) 3-hydroxyisovaleric acid, (4) 4-deoxyerythronic acid, (5) 4-deoxytheonic acid, (6) acetic acid, (7) acetone, (8) trigonelline, (9) betaine, (10) cis-aconitic acid, (11) citric acid, (12) creatine, (13) creatinine, (14) D-glucose, (15) dimethylamine, (16) dimethlyglycine, (17) ethanolamine, (18) formic acid, (19) fumaric acid, (20) glycine, (21) glycolic acid, (22) guanidoacetic acid, (23) pseudouridine, (24) hippuric acid, (25) indoxyl sulfate, (26) L-acetylcarnitine, (27) L-alanine, (28) trimethylamine, (29) L-lactic acid, (30) L-phenylalanine, (31) L-threonine, (32) trimethylamine-N-oxide, (33) L-tyrosine, (34) uracil, (35) myo-inositol, (36) N-methylnicotinamide, (37) p-cresol sulfate, (38) phenylacetylglutamine, (39) phosphorylcholine, (40) p-hydroxyphenylacetic acid, (41) proline betaine, (42) succinic acid, (43) tartaric acid and (44) taurine. (B) Loadings plot of OPLS regression analysis with eGFR in the y variable. This plot represents a ‘statistical’ pseudo-spectrum that displays the OPLS model covariance on the y-axis and chemical shift on the x-axis. The colour bar represents the r2 value for that spectral region to eGFR. Plots in (A) and (B) are aligned on the x-axis, which shows which spectral regions may best describe CKD-related changes and warrant structural identification. (C) Scores plot of the OPLS model showing data separation based on sample eGFR (colour bar). Metabolites from various metabolic pathways were altered with reduced eGFR. Citric acid from the tricarboxylic acid cycle, amino acids (e.g. threonine), lipid metabolism (e.g. ethanolamine), gut microbiome–derived uraemic toxins (e.g. indoxyl sulphate and p-cresol sulphate), uracil and glycolic acid were significantly lower with reduced eGFR. Many of these metabolites have also been reported as significant markers of CKD from previous studies [23–25], further strengthening them as reliable biomarkers. Additionally, two metabolites, myo-inositol and betaine, had negative coefficients, indicating a respective urinary increase with reduced eGFR. Metabolite markers related to CKD progression A nested retrospective subgroup (n = 57) of the cohort was analysed for urine metabolites predictive of CKD progression. These CKD patients were well distributed throughout the range of progression (r2 = 0.96), which was from 5% per year (i.e. improvement in eGFR) to −14.1% per year (i.e. strong deterioration) (Supplementary Figure 2). The percent annual eGFR slope was independent of baseline eGFR (P = 0.44) (Supplementary Figure S3). Linear modelling was again applied using the percent annual eGFR slope change as the y variable. A total of eight metabolites showed significant associations (Table 3). Citric acid, glycolic acid and ethanolamine were again seen to be positively associated with a declining percent annual eGFR slope. Creatinine and dimethylamine were also positively associated. The highest β coefficients in the modelling were seen for betaine and myo-inositol, which had negative correlations (Figure 2). These osmolytes again showed opposite trends from all other significant metabolites. Table 3. Significant prognostic metabolites of CKD progression Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  This table shows general linear modelling results for prognostic markers of CKD progression. These metabolites show significant linear associations with percent annual eGFR slope of the selected progression subset. The β coefficient, standard error (SE) and P-value of each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group A negative association (i.e. negative β) indicates an increase in the metabolite concentration as the percent annual eGFR slope decreases (i.e. CKD progression rate increases). Table 3. Significant prognostic metabolites of CKD progression Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  This table shows general linear modelling results for prognostic markers of CKD progression. These metabolites show significant linear associations with percent annual eGFR slope of the selected progression subset. The β coefficient, standard error (SE) and P-value of each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group A negative association (i.e. negative β) indicates an increase in the metabolite concentration as the percent annual eGFR slope decreases (i.e. CKD progression rate increases). FIGURE 2: View largeDownload slide Plots of adjusted generalized linear model for renal osmolytes. A negative percent annual eGFR slope indicates patients that are progressing in CKD. The linear model equation, 95% confidence intervals and P-values are defined in the legend. FIGURE 2: View largeDownload slide Plots of adjusted generalized linear model for renal osmolytes. A negative percent annual eGFR slope indicates patients that are progressing in CKD. The linear model equation, 95% confidence intervals and P-values are defined in the legend. Due to the longitudinal nature of the sample group, the previously mentioned metabolites may act as prognostic urinary biomarkers of CKD outcome. We focused on to examining the renal osmolytes betaine and myo-inositol, as their respective trends were contrary to the other metabolites and their physiological relevance in kidney health is unique. Using receiver operating characteristics (ROC) analysis, the values of betaine and myo-inositol were examined in the first and last quartiles of the CKD progression subgroup. Here, the first quartile (i.e. ‘stable’) showed no CKD progression, with an eGFR change of +5% to −1% per year. The last quartile (i.e. ‘rapid’ progression) showed an eGFR change of −8% to −15% per year. The area under the curve (AUC) of the ROC analysis for betaine was 0.814 and for myo-inositol was 0.781. Combined, both betaine and myo-inositol showed an AUC of 0.84 (Supplementary Figure S4). Decreased kidney expression of betaine and myo-inositol transporters in murine models To determine if osmolyte transporters were a factor for increased urine osmolyte concentrations, kidney transcriptomics of a murine CKD model were analysed for gene expression changes. The kidney gene expression of Slc5a3, Slc5a11 and Slc6a12 (see Figure 3) was assessed by real-time qPCR in experimental kidney fibrosis induced by UUO, a classic murine model of renal fibrosis and inflammation that recapitulates all the key events in CKD leading to fibrosis [26]. Significant downregulation of the genes encoding the myo-inositol transporter (Slc5a11) and the betaine transporter (Slc6a12) were observed (Figure 4A). FIGURE 3: View largeDownload slide Function of gene products in tubular cells. Slc6a12 encodes the betaine transporter bgt-1. Slc5a3 encodes the sodium myo-inositol transporter Smit. Slc5a11 encodes the sodium myo-inositol transporter Smit2. Myox encodes myo-inositol oxygenase, which is involved in myo-inositol synthesis. Transcription factor TonEBP/Nfat5 regulates gene expression of all osmolyte transporter genes; (Slc6a11, Slc5a3, and Slc5a11). FIGURE 3: View largeDownload slide Function of gene products in tubular cells. Slc6a12 encodes the betaine transporter bgt-1. Slc5a3 encodes the sodium myo-inositol transporter Smit. Slc5a11 encodes the sodium myo-inositol transporter Smit2. Myox encodes myo-inositol oxygenase, which is involved in myo-inositol synthesis. Transcription factor TonEBP/Nfat5 regulates gene expression of all osmolyte transporter genes; (Slc6a11, Slc5a3, and Slc5a11). FIGURE 4: View largeDownload slide Gene expression data. (A) Changes in experimental kidney fibrosis assessed by real-time qPCR. Kidney mRNA was measured 14 days after induction of UUO leading to kidney fibrosis. Obstructed kidneys were compared with contralateral unobstructed kidneys. (B) Changes in experimental kidney injury transcriptomics. Kidney mRNA was measured 24 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. (C) Changes in experimental kidney injury assessed by real-time qPCR. Kidney mRNA was measured 24 h and 72 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. *P ≤ 0.05, **P ≤ 0.01. FIGURE 4: View largeDownload slide Gene expression data. (A) Changes in experimental kidney fibrosis assessed by real-time qPCR. Kidney mRNA was measured 14 days after induction of UUO leading to kidney fibrosis. Obstructed kidneys were compared with contralateral unobstructed kidneys. (B) Changes in experimental kidney injury transcriptomics. Kidney mRNA was measured 24 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. (C) Changes in experimental kidney injury assessed by real-time qPCR. Kidney mRNA was measured 24 h and 72 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. *P ≤ 0.05, **P ≤ 0.01. The candidate genes (Slc5a3, Slc5a11, Slc6a12) tested in experimental CKD were also observed in a transcriptomics analysis of kidney injury using a murine model injected with folic acid. Supplementary Table S3 shows all differentially expressed genes in the transcriptomics array with a fold change increasing or decreasing in expression >25% and FDR <5%, as previously reported [27]. Our hypothesis-driven analysis of the transcriptomics database identified decreased expression of Slc6a12, Slc5a3 and Slc5a11 (Figure 4B). Downregulation of these genes did not appear to result from the downregulation of the transcription factor tonicity-responsive enhancer-binding protein (TonEBP/NFAT5), a master regulator of osmolyte transporters, since TonEBP/NFAT5 expression was preserved (fold change 0.96, FDR = ns). Additionally, Myox, the gene encoding myo-inositol oxygenase, was also downregulated in kidney injury (Supplementary Figure S5). Real-time qPCR confirmed in an independent validation experiment of the decreased expression of Slc5a3, Slc5a11 and Slc6a12 mRNA at 24 h of kidney injury induction and persisting after 72 h (Figure 4C). These data suggest that these gene expression changes are shared by both CKD and acute kidney injuries. DISCUSSION In the present study we investigated the urine metabolome signatures of patients with CKD and CKD progression. Most significant metabolites decrease in urinary concentration with reduced eGFR and in patients with progressing CKD. However, in these same patients we also observed significant linear increases of myo-inositol and betaine. Both metabolites had the highest absolute β coefficients in our modelling of percent annual eGFR slope and showed prognostic value in predicting patients who rapidly progress from those who remained stable. Kidney transcriptomics identified defective gene expressions of myo-inositol and betaine transporters in murine kidney disease models, providing convincing evidence that elevated osmolyte concentrations seen in the urine are partially a result of perturbed renal osmolyte transport. Figure 5 provides a proposed model of osmolyte dysregulation during CKD. This may be the first demonstration that links urinary osmolyte concentrations with clinical CKD outcomes using a metabolomic analysis supported with transcriptomics. FIGURE 5: View largeDownload slide Hypertonicity response mechanism in CKD. A proposed model of the osmolyte mechanism in CKD. (A) Model of osmolyte reabsorption in a healthy nephron. (B) Tubular cell response mechanism to increased hypertonicity in medullary tissue. (C) Blood and urine osmolyte concentrations in response to hypertonicity. Patients with progressive forms of CKD have increased osmolyte concentrations in urine. FIGURE 5: View largeDownload slide Hypertonicity response mechanism in CKD. A proposed model of the osmolyte mechanism in CKD. (A) Model of osmolyte reabsorption in a healthy nephron. (B) Tubular cell response mechanism to increased hypertonicity in medullary tissue. (C) Blood and urine osmolyte concentrations in response to hypertonicity. Patients with progressive forms of CKD have increased osmolyte concentrations in urine. The osmolality of mammalian blood is normally kept constant at ∼290 mosmol/kg [28] and maintained by regulatory mechanisms such as the renin–angiotensin–aldosterone and antidiuretic hormone systems. However, normal kidney function requires an interstitial osmotic gradient in renal medulla four times that of plasma osmolality. The hyperosmotic conditions created by NaCl and urea are required for urine concentration. However, these conditions can be damaging to cells and proteins [28, 29] and cause water to diffuse out of cells, resulting in cellular shrinkage [30]. Osmolytes have evolved in renal tissues as protective compounds against osmotic stress [31], and studies have explored mechanisms for how osmolytes provide protein stability [32, 33]. Five protective osmolytes are used in renal cells: sorbitol, betaine, myo-inositol, taurine and glycerophosphocholine (GPC), with a large proportion being betaine and myo-inositol [28, 29]. The concentrations of these osmolytes are regulated through the TonEBP/NFAT5 transcription factor [28, 34], which regulates the transcription of all transporter proteins responsible for osmolyte uptake in renal cells. This transcription factor is critical to kidney health and has been proven essential in knockout models [30, 35]. Mechanisms for betaine uptake in medullary cells are well studied [28, 36]. As osmolality increases, so too do medullary betaine levels [28, 37]. Uptake comes primarily from extracellular sources and not from increased local cellular synthesis [28]. Betaine is freely filtered by glomeruli and actively reabsorbed by renal tubules at levels >95% of the filtered load [38, 39] (Figure 5A). Therefore, non-perturbed reabsorption mechanisms for betaine should not allow for an increase in urinary levels as seen in our study. The primary transporter of betaine in medullary cells is betaine/GABA transporter 1 (Bgt-1) [28, 40], which is encoded by the Slc6a12 gene and regulated by TonEBP/NFAT5 [30]. Hypertonicity leads to increased TonEBP/NFAT5 expression and localization of Bgt-1 to apical and basolateral plasma membranes in tubular cells (Figure 5B). It is estimated that betaine could account for ∼25% of the total osmolyte content [40, 41], and a recent study found that betaine ranked high in protein-stabilizing properties [32]. Its high abundance and stabilizing properties provide evidence that betaine is one of the more important organic osmolytes in medullary tissues. A recent study of acute renal ischaemia found kidney tissue had reduced levels of betaine and myo-inositol when compared with controls [42, 43]. Therefore, increased betaine in the urine most likely reflects CKD-related renal stress (Figure 5C). Our data show an increase of betaine in human urine during CKD, most likely due to Bgt-1 transporter downregulation, given the lower gene expression in experimental kidney fibrosis and injury. However, the release of osmolytes into the urine following renal cell damage may also contribute. Myo-inositol also increases in healthy medullary cells when interstitial osmolality increases, and reduced Myox gene expression in experimental kidney injury argues against increased local myo-inositol production being a major driver of higher urinary levels in CKD. Sodium myo-inositol transporters (Smit and Smit2) encoded by the Slc5a3 and Slc5a11 genes are primarily responsible for renal myo-inositol uptake [28] and are also regulated by TonEBP/NFAT5 [30]. Downregulation of Slc5a11 was observed in experimental kidney fibrosis, and Slc5a3 and Slc5a11 were also downregulated during kidney injury, suggestive of a general decrease in myo-inositol uptake capacity by stressed tubular cells. Furthermore, a recent study found myo-inositol significantly elevated in the urine of mice with renal vasculitis [44]. We hypothesize that increased urinary levels of renal osmolytes reflect damage to kidney tissue from hyperosmolar stress, especially in diseases such as diabetic and hypertensive nephropathy, where osmotic and hypertensive stresses cause microvasculature damage [35, 45]. In support, gene expression data of osmolyte transporters point to a dysregulated transporter system as a major factor of osmolytes lost into the urine. Data mining of human kidney biopsy gene expression using the Nephroseq database (www.Nephroseq.org) showed a study exploring gene expression in human CKD with diverse aetiologies [46]. These data identified the downregulated kidney expression of Slc5a11 (foldchange −3.15 discovery, −1.91 validation) and Slc6a12 (foldchange −1.63 discovery, −6.98 validation) mRNA in human CKD kidneys compared with control kidneys. However, no difference in Slc5a3 mRNA expression was observed, as was similarly seen in our experimental kidney fibrosis data. The present data are consistent with the increased recognition of osmotic stress as a driving force behind CKD progression. For example, repeated dehydration and osmotic stress are thought to be a key driver of kidney injury in Meso-American nephropathy patients [47]. Moreover, tolvaptan, an antagonist of antidiuretic hormone receptors that promotes polyuria and decreases medullary hyperosmolarity, has been recently approved to treat polycystic kidney disease. Experimental evidence suggests that tolvaptan or high water intake may be protective in other nephropathies [48]. Finally, data consistent with impaired kidney responses to osmotic stress in CKD are in line with clinical observations that CKD patients are more sensitive to the toxic effects of high-osmolality iodinated contrast media [49]. There are certain limitations in our study. The number of samples in the CKD progression analysis is somewhat low, and a larger, multicentre, validation study is warranted. However, betaine and myo-inositol showed the same significant direction of alteration in the larger-scale analysis of the total cohort eGFR. Furthermore, gene expression data from independent experimental murine models provides interdiscipline validation of the pathophysiological feasibility of osmolyte abnormalities acting as CKD progression biomarkers. Therefore, the increased urinary output of betaine and myo-inositol reflects abnormal tubular transport of osmolytes and an impaired renal medullary response to osmotic stress in CKD progression. SUPPLEMENTARY DATA Supplementary data are available at ndt online. FUNDING The research leading to these results has received funding from the European Union’s Seventh Framework Programme FP7/2007-2013 under grant agreement FP7-PEOPLE-2013-ITN-608332. A.O. and M.D.S.N. were supported by Spanish government FEDER funds RETIC REDINREN RD016/0019. FIS PI16/02057, PI15/00298, CP14/00133, Sociedad Española de Nefrología, Programa Intensificacion Actividad Investigadora (ISCIII/Agencia Lain-Entralgo/CM), Miguel Servet MS14/00133. CONFLICT OF INTEREST STATEMENT A.O. reports grants from the Spanish government during the conduct of the study. REFERENCES 1 Inker LA, Astor BC, Fox CH et al.   KDOQI US commentary on the 2012 KDIGO clinical practice guideline for the evaluation and management of CKD. Am J Kidney Dis  2014; 63: 713– 735 10.1053/j.ajkd.2014.01.416. Google Scholar CrossRef Search ADS PubMed  2 Vassalotti JA, Centor R, Turner BJ et al.   Practical approach to detection and management of chronic kidney disease for the primary care clinician. Am J Med  2016; 129: 153– 162.e7 Google Scholar CrossRef Search ADS PubMed  3 Slocum JL, Heung M, Pennathur S. Marking renal injury: can we move beyond serum creatinine? Transl Res  2012; 159: 277– 289 Google Scholar CrossRef Search ADS PubMed  4 Wettersten HI, Weiss RH. Applications of metabolomics for kidney disease research. Organogenesis  2013; 9: 11– 18 Google Scholar CrossRef Search ADS PubMed  5 Weiss RH, Kim K. Metabolomics in the study of kidney diseases. Nat Rev Nephrol  2011; 8: 22– 33 Google Scholar CrossRef Search ADS PubMed  6 Hocher B, Adamski J. Metabolomics for clinical use and research in chronic kidney disease. Nat Rev Nephrol  2017; 13: 269– 284 Google Scholar CrossRef Search ADS PubMed  7 Zhao Y-Y, Xu Q. Metabolomics in chronic kidney disease. Clin Chim Acta  2013; 422: 59– 69 Google Scholar CrossRef Search ADS PubMed  8 Zhao YY, Cheng XL, Vaziri ND et al.   UPLC-based metabonomic applications for discovering biomarkers of diseases in clinical chemistry. Clin Biochem  2014; 47: 16– 26 Google Scholar CrossRef Search ADS PubMed  9 Heinzmann SS, Merrifield C. A, Rezzi S et al.   Stability and robustness of human metabolic phenotypes in response to sequential food challenges. J Proteome Res  2012; 11: 643– 655 Google Scholar CrossRef Search ADS PubMed  10 Dona AC, Jimenez B, Schäfer H et al.   Precision high-throughput proton NMR spectroscopy of human urine, serum, and plasma for large-scale metabolic phenotyping. Anal Chem  2014; 86: 9887– 9894 Google Scholar CrossRef Search ADS PubMed  11 Schanstra JP, Zürbig P, Alkhalaf A et al.   Diagnosis and prediction of CKD progression by assessment of urinary peptides. J Am Soc Nephrol  2015; 26: 1999– 2010 Google Scholar CrossRef Search ADS PubMed  12 Gil RB, Lehmann R, Schmitt-Kopplin P et al.   1H NMR-based metabolite profiling workflow to reduce inter-sample chemical shift variations in urine samples for improved biomarker discovery. Anal Bioanal Chem  2016; 408: 4683– 4691 Google Scholar CrossRef Search ADS PubMed  13 Xia J, Wishart DS. Using MetaboAnalyst 3.0 for comprehensive metabolomics data analysis. Curr Protoc Bioinform  2016; 55: 14.10.1– 14.10.91 Google Scholar CrossRef Search ADS   14 Dieterle F, Ross A, Schlotterbeck G et al.   Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in 1H NMR metabonomics. Anal Chem  2006; 78: 4281– 4290 Google Scholar CrossRef Search ADS PubMed  15 Kohl SM, Klein MS, Hochrein J et al.   State-of-the art data normalization methods improve NMR-based metabolomic analysis. Metabolomics  2012; 8: 146– 160 Google Scholar CrossRef Search ADS PubMed  16 Veselkov KA, Lindon JC, Ebbels TMD et al.   Recursive segment-wise peak alignment of biological 1H NMR spectra for improved metabolic biomarker recovery. Anal Chem  2009; 81: 56– 66 Google Scholar CrossRef Search ADS PubMed  17 Cloarec O, Dumas M, Craig A et al.   Statistical total correlation spectroscopy : an exploratory approach for latent biomarker identification from metabolic 1H NMR data sets statistical total correlation spectroscopy : an exploratory approach for latent biomarker identification from metabolic. Anal Chem  2005; 77: 1282– 1289 Google Scholar CrossRef Search ADS PubMed  18 Wishart DS, Jewison T, Guo AC et al.   HMDB 3.0—the human metabolome database in 2013. Nucleic Acids Res  2012; 41: D801– D807 Google Scholar CrossRef Search ADS PubMed  19 Hao J, Liebeke M, Astle W et al.   Bayesian deconvolution and quantification of metabolites in complex 1D NMR spectra using BATMAN. Nat Protoc  2014; 9: 1416– 1427 Google Scholar CrossRef Search ADS PubMed  20 Izquierdo MC, Sanz AB, Mezzano S et al.   TWEAK (tumor necrosis factor–like weak inducer of apoptosis) activates CXCL16 expression during renal tubulointerstitial inflammation. Kidney Int  2012; 81: 1098– 1107 Google Scholar CrossRef Search ADS PubMed  21 Ucero AC, Benito-Martin A, Fuentes-Calvo I et al.   TNF-related weak inducer of apoptosis (TWEAK) promotes kidney fibrosis and Ras-dependent proliferation of cultured renal fibroblast. Biochim Biophys Acta  2013; 1832: 1744– 1755 Google Scholar CrossRef Search ADS PubMed  22 Ortiz A, Husi H, Gonzalez-Lafuente L et al.   Mitogen-activated protein kinase 14 promotes AKI. J Am Soc Nephrol  2017; 28: 823– 836 Google Scholar CrossRef Search ADS PubMed  23 Chen DQ, Cao G, Chen H et al.   Gene and protein expressions and metabolomics exhibit activated redox signaling and wnt/β-catenin pathway are associated with metabolite dysfunction in patients with chronic kidney disease. Redox Biol  2017; 12: 505– 521 Google Scholar CrossRef Search ADS PubMed  24 Posada-Ayala M, Zubiri I, Martin-Lorenzo M et al.   Identification of a urine metabolomic signature in patients with advanced-stage chronic kidney disease. Kidney Int  2014; 85: 103– 111 Google Scholar CrossRef Search ADS PubMed  25 Raffler J, Friedrich N, Arnold M et al.   Genome-wide association study with targeted and non-targeted NMR metabolomics identifies 15 novel loci of urinary human metabolic individuality. PLoS Genet  2015; 11: e1005487 Google Scholar CrossRef Search ADS PubMed  26 Ucero AC, Benito-Martin A, Izquierdo MC et al.   Unilateral ureteral obstruction: beyond obstruction. Int Urol Nephrol  2014; 46: 765– 776 Google Scholar CrossRef Search ADS PubMed  27 Martin-Lorenzo M, Gonzalez-Calero L, Ramos-Barron A et al.   Urine metabolomics insight into acute kidney injury point to oxidative stress disruptions in energy generation and H2S availability. J Mol Med  2017; 95: 1399– 1409 Google Scholar CrossRef Search ADS PubMed  28 Burg MB, Ferraris JD. Intracellular organic osmolytes: function and regulation. J Biol Chem  2008; 283: 7309– 7313 Google Scholar CrossRef Search ADS PubMed  29 Yancey PH. Organic osmolytes as compatible, metabolic and counteracting cytoprotectants in high osmolarity and other stresses. J Exp Biol  2005; 208: 2819– 2830 Google Scholar CrossRef Search ADS PubMed  30 Lopez-Rodriguez C, Antos CL, Shelton JM et al.   Loss of NFAT5 results in renal atrophy and lack of tonicity-responsive gene expression. Proc Natl Acad Sci USA  2004; 101: 2392– 2397 Google Scholar CrossRef Search ADS PubMed  31 Yancey PH, Clark ME, Hand SC et al.   Living with water stress: evolution of osmolyte systems. Science  1982; 217: 1214– 1222 Google Scholar CrossRef Search ADS PubMed  32 Street TO, Bolen DW, Rose GD. A molecular mechanism for osmolyte-induced protein stability. Proc Natl Acad Sci USA  2006; 103: 13997– 14002 Google Scholar CrossRef Search ADS PubMed  33 Katayama H, McGill M, Kearns A et al.   Strategies for folding of affinity tagged proteins using GroEL and osmolytes. J Struct Funct Genomics  2009; 10: 57– 66 Google Scholar CrossRef Search ADS PubMed  34 Day CR, Gordon SS, Vaughn CL et al.   A single amino acid substitution in the renal betaine/GABA transporter prevents trafficking to the plasma membrane. Physiol J  2013; 2013: 1 Google Scholar CrossRef Search ADS   35 Brocker C, Thompson DC, Vasiliou V. The role of hyperosmotic stress in inflammation and disease. Biomol Concepts  2012; 3: 345– 364 Google Scholar CrossRef Search ADS PubMed  36 Burg MB. Coordinate regulation of organic osmolytes in renal cells. Kidney Int  1996; 49: 1684– 1685 Google Scholar CrossRef Search ADS PubMed  37 Nakanishi T, Turner RJ, Burg MB. Osmoregulation of betaine transport in mammalian renal medullary cells. Am J Physiol  1990; 258: F1061– F1067 Google Scholar PubMed  38 Lever M, Sizeland PCB, Bason LM et al.   Glycine betaine and proline betaine in human blood and urine. Biochim Biophys Acta  1994; 1200: 259– 264 Google Scholar CrossRef Search ADS PubMed  39 Pummer S, Dantzler WH, Lien Y-HH et al.   Reabsorption of betaine in Henle’s loops of rat kidney in vivo. Am J Physiol Renal Physiol  2000; 278: F434– F439 Google Scholar CrossRef Search ADS PubMed  40 Kempson S. A, Montrose MH. Osmotic regulation of renal betaine transport: transcription and beyond. Pflugers Arch  2004; 449: 227– 234 Google Scholar PubMed  41 Grunewald RW, Oppermann M, Schettler V et al.   Polarized function of thick ascending limbs of Henle cells in osmoregulation. Kidney Int  2001; 60: 2290– 2298 Google Scholar CrossRef Search ADS PubMed  42 Jouret F, Leenders J, Poma L et al.   Nuclear magnetic resonance metabolomic profiling of mouse kidney, urine and serum following renal ischemia/reperfusion injury. PLoS One  2016; 11: e0163021 Google Scholar CrossRef Search ADS PubMed  43 Wei Q, Xiao X, Fogle P et al.   Changes in metabolic profiles during acute kidney injury and recovery following ischemia/reperfusion. PLoS One  2014; 9: e106647 Google Scholar CrossRef Search ADS PubMed  44 Al-Ani B, Fitzpatrick M, Al-Nuaimi H et al.   Changes in urinary metabolomic profile during relapsing renal vasculitis. Sci Rep  2016; 6: 38074 Google Scholar CrossRef Search ADS PubMed  45 Fowler MJ. Microvascular and macrovascular complications of diabetes. Clin Diabetes  2008; 26: 77– 82 Google Scholar CrossRef Search ADS   46 Nakagawa S, Nishihara K, Miyata H et al.   Molecular markers of tubulointerstitial fibrosis and tubular cell damage in patients with chronic kidney disease. PLoS One  2015; 10: e0136994 Google Scholar CrossRef Search ADS PubMed  47 Martín-Cleary C, Ortiz A. CKD hotspots around the world: where, why and what the lessons are. A CKJ review series. Clin Kidney J  2014; 7: 519– 515 Google Scholar CrossRef Search ADS PubMed  48 Clark WF, Sontrop JM, Huang S-H et al.   Hydration and chronic kidney disease progression: a critical review of the evidence. Am J Nephrol  2016; 43: 281– 292 Google Scholar CrossRef Search ADS PubMed  49 Eng J, Wilson RF, Subramaniam RM et al.   Comparative effect of contrast media type on the incidence of contrast-induced nephropathy a systematic review and meta-analysis. Ann Intern Med  2016; 164: 417– 424 Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nephrology Dialysis Transplantation Oxford University Press

Increased urinary osmolyte excretion indicates chronic kidney disease severity and progression rate

Loading next page...
 
/lp/ou_press/increased-urinary-osmolyte-excretion-indicates-chronic-kidney-disease-kDNxwKHR1v
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.
ISSN
0931-0509
eISSN
1460-2385
D.O.I.
10.1093/ndt/gfy020
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Chronic kidney disease (CKD) is a recognized global health problem. While some CKD patients remain stable after initial diagnosis, others can rapidly progress towards end-stage renal disease (ESRD). This makes biomarkers capable of detecting progressive forms of CKD extremely valuable, especially in non-invasive biofluids such as urine. Screening for metabolite markers using non-targeted metabolomic techniques like nuclear magnetic resonance spectroscopy is increasingly applied to CKD research. Methods A cohort of CKD patients (n = 227) with estimated glomerular filtration rates (eGFRs) ranging from 9.4–130 mL/min/1.73 m2 was evaluated and urine metabolite profiles were characterized in relation to declining eGFR. Nested in this cohort, a retrospective subset (n = 57) was investigated for prognostic metabolite markers of CKD progression, independent of baseline eGFR. A transcriptomic analysis of murine models of renal failure was performed to validate selected metabolomic findings. Results General linear modeling revealed 11 urinary metabolites with significant associations to reduced eGFR. Linear modelling specifically showed that increased urine concentrations of betaine (P < 0.05) and myo-inositol (P < 0.05) are significant prognostic markers of CKD progression. Conclusions Renal organic osmolytes, betaine and myo-inositol play a critical role in protecting renal cells from hyperosmotic stress. Kidney tissue transcriptomics of murine preclinical experimentation identified decreased expression of Slc6a12 and Slc5a11 mRNA in renal tissue consistent with defective tubular transport of these osmolytes. Imbalances in renal osmolyte regulation lead to increased renal cell damage and thus more progressive forms of CKD. Increases in renal osmolytes in urine could provide clinical diagnostic and prognostic information on CKD outcomes. CKD, disease progression, metabolomics, osmolytes, transcriptomics, urine INTRODUCTION Chronic kidney disease (CKD) is a term used to include a wide range of diseases that reduce kidney function. The National Kidney Foundation defines CKD as ≥3 months of kidney damage or an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2 [1]. Disease aetiology in CKD is diverse, leading to variability in morbidity and mortality [2]. Further, not all populations have an equal risk for developing CKD and progressing towards end-stage renal disease (ESRD); therefore, high-risk populations with diabetes and/or hypertension should receive targeted testing for CKD [2]. eGFR can be calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation or the Modification of Diet in Renal Disease (MDRD) study equation [1, 2]. However, both measures are relatively insensitive in early stage CKD and their prognostic value is limited. Novel biomarkers that are more sensitive in early stage CKD are needed for diagnosis and prognosis of CKD progression [3]. Metabolomics can analyse biofluids for specific metabolic signatures reflecting CKD status [4], offering promising non-invasive methods to screen, diagnose and prognose patients for CKD [5–7]. Mass spectrometry (MS) and nuclear magnetic resonance (NMR) are the two main analytical platforms used in metabolomics [8]. Here we use NMR spectroscopy as a quantitative state-of-the-art analytical platform that is widely used in metabolome studies [9]. NMR can detect a wide range of metabolites from various chemical classes, delivering robust transferable data under optimized conditions [10]. The purpose of this study was to use NMR-based metabolomic strategies to elucidate novel urinary markers supportive of current methods for diagnosing CKD. Urine samples from a cohort of 227 patients in various stages of CKD were utilized. An additional aim was to identify prognostic markers that indicate patients at risk of CKD progression towards ESRD. A nested retrospective subgroup (n = 57) from the cohort was analysed using follow-up eGFR measurements and a calculated rate of CKD progression. Results of these analyses point to increased urine concentrations of renal osmolytes as markers of reduced eGFR. These compounds were also significantly elevated in the urine of CKD patients with progressive forms of CKD. These metabolite data were supported with messenger RNA (mRNA) data generated from independent murine kidney fibrosis and injury experimentation. MATERIALS AND METHODS Study cohort and sampling The cohort consisted of patients from the Nephrology Outpatient Clinic at Ghent University Hospital, Ghent, Belgium. Samples were collected in the period 17 January 2011 to 7 July 2015. eGFR was determined using the CKD-EPI equation. Important metadata used in the analysis can be found in Table 1. Sample collection was performed in accordance with local ethics requirements and in accordance with the current revision of the Helsinki Declaration. Table 1. Urine sample cohort (n = 227) Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  This table provides an overview of cohort metadata used for biomarker discovery. Age, gender, BMI and CKD disease groups were determined to be the most important metadata in linear modelling of metabolites. For age, BMI and eGFR, the mean and standard deviation are shown with the range. a Systemic diseases includes diabetes type 2 and hypertension. Table 1. Urine sample cohort (n = 227) Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  Male, n (%)  137 (60.4)  Age (years)  62.3 ± 16.8 (17–88)  Disease groups (n, as defined by ERA-EDTA Primary Renal Disease Registry)     Tubulointerstitial diseases  35   Systemic diseases affecting the kidneya  96   Familial/hereditary nephropathies  23   Miscellaneous renal disorders  41   Glomerular diseases  32  Body mass index (kg/m2)  28.0 ± 5.5 (16.3–47.9)  CKD stage (CKD-EPI, mL/min/1.73 m2), n     Stage 1 (eGFR >90)  25   Stage 2 (eGFR 89–60)  45   Stage 3 (eGFR 59–30)  115   Stage 4 (eGFR 29–15)  34   Stage 5 (eGFR <15)  8  eGFR (mL/min/1.73 m2)  51 ± 25.7 (9.4–130)  This table provides an overview of cohort metadata used for biomarker discovery. Age, gender, BMI and CKD disease groups were determined to be the most important metadata in linear modelling of metabolites. For age, BMI and eGFR, the mean and standard deviation are shown with the range. a Systemic diseases includes diabetes type 2 and hypertension. A total of 57 patients had sufficient follow-up data to monitor their respective rate of CKD progression (i.e. ≥3 eGFR measurements, minimum of 2-years duration). To reduce confounding factors, only samples from patients with systemic diseases or glomerular diseases were used, and patients with renal transplantation or receiving renal replacement therapy were not considered (Supplementary Table S1). The percent annual eGFR slope change from baseline was calculated using the following equation, an approach that has also been previously applied [11]:   % annual eGFR slope= linear coefficient baseline eGFR×100% Sample preparation for NMR and instrumental protocol Upon collection, urine samples were immediately centrifuged (20800 g for 10 min), aliquoted per 1 mL and stored at −80°C. Following defrosting, a buffering approach designed for NMR analysis in urine was applied [12]. This included a deuterium oxide phosphate buffer (1.5 M PO4, pH 7.4; Armar Chemicals, Leipzig, Germany) with potassium fluoride (KF, 300 mM) and (trimethylsilyl)propionic acid (TSP, 0.1%). Phosphate salts, KF and TSP were purchased from Sigma Aldrich Chemie (Steinheim, Germany). Urine samples were analysed on a Bruker 800-MHz spectrometer operating at 800.35 MHz with a quadrupole inverse cryogenic probe. A standard one-dimensional pulse sequence [recycle delay (RD) 90°, t1 90°, tm 90°, acquire FID] was acquired, with water suppression irradiation during RD of 2 s, mixing time (tm) set on 200 ms and a 90° pulse set to 10.13 μs, collecting 512 scans into 64 000 data points with a spectral width of 12 ppm. All spectra were manually phased, baseline corrected and calibrated to TSP (δ 0.00) with TopSpin 3.2 (Bruker BioSpin, Rheinstetten, Germany). Data were imported to MATLAB (MathWorks, Natick, MA, USA) and further processed, i.e., water and urea region removed (δ 4.7–5.6). Daily quality control samples (i.e. pooled urine sample) were measured throughout the cohort measurement to monitor instrument performance, data processing (i.e. spectra pre-processing, normalization and alignment) and metabolite stability. The structural identity of metabolite compounds was determined by two-dimensional (2D) NMR methods; total correlation spectroscopy (TOCSY) and heteronuclear single quantum coherence (HSQC) spectroscopy. For 2D TOCSY spectra, phase-sensitive and sensitivity-improved 2D TOCSY with WATERGATE (3-9-19) and DIPSI-2 were acquired. For each spectrum 19228 × 1024 data points were collected using 32 scans per increment, 16 dummy scans and 1 s acquisition time. The spectral widths were set to 12 ppm for both the F1 and F2 dimensions. For 2D HSQC spectra, phase-sensitive ge-2D HSQC using PEP and adiabatic pulses for inversion and refocusing with gradients was used. Data points of 4788 × 1280 per spectra were collected with 256 scans per increment, 16 dummy scans and an acquisition time of 0.25 s. Spectral widths were set to 12 ppm for the F2 and 230 ppm for F1. NMR-based statistical tools Data processing and statistics were performed in MATLAB 2015b, RStudio and the MetaboAnalyst platform (http://www.metaboanalyst.ca) [13]. Probabilistic quotient normalization [14] was used for normalization of spectral data to account for biological variation in urine dilution. This method is an improved normalization technique compared with single-metabolite normalization [15], which is superior in the context of CKD, as it avoids normalization to metabolites affected by the disease. Spectral data were aligned using the recursive segment-wise peak alignment (RSPA) algorithm [16]. An orthogonal partial least squares (OPLS) regression analysis [17] using eGFR as the y variable was performed for spectral feature selection. Spectral regions most affected by CKD display high covariance and r2 values. Selected spectral features were structurally identified using 2D NMR and matched with the human metabolite database (HMDB) [18]. Identified metabolites were quantified with the R-package BATMAN [19], measuring the area under the curve. Data were then log transformed and mean centered. General linear modelling was applied to each identified metabolite with baseline eGFR or percent annual eGFR slope as y variables, adjusted for age, gender, body mass index (BMI) and disease group. Kidney transcriptomics Total RNA was isolated using the PureLink RNA Mini Kit (Invitrogen, Paisley, UK). Affymetrix transcriptomics arrays of kidney tissue (n = 3 control and n = 3 kidney injury, obtained 24 h after induction of kidney injury by a folic acid injection or injection of vehicle) were performed at Unidad Genómica Moncloa, Fundación Parque Científico de Madrid, Madrid, Spain, following the manufacturer’s protocol [20]. Image files were initially obtained through Affymetrix GeneChip Command Console software. Robust multichip analysis was performed using the Affymetrix Expression Console software. Starting from the normalized robust multichip analysis, the significance analysis of microarrays was performed using the limma package (Babelomics, http://www.babelomics.org), with a false discovery rate (FDR) of 5% to identify genes significantly dysregulated. Experimental kidney injury These studies were approved by the IIS-FJD animal ethics committee and followed Directive 2010/63/EU on the protection of animals used for scientific purposes. For experimental kidney injury, female 12- to 14-week-old C57/BL6 mice received a single intraperitoneal injection of folic acid (Sigma Aldrich) 250 mg/kg in sodium bicarbonate 0.3 mol/L (acute kidney injury, n = 3) or a vehicle alone (controls, n = 3) and were sacrificed 24 h later as previously described [20]. Kidneys were cold saline perfused in situ before removal. One kidney from each mouse was fixed in buffered formalin, embedded in paraffin and stained with haematoxylin and eosin. The other kidney was snap frozen in liquid nitrogen for ribonucleic acid (RNA) studies. Samples from this experiment were used for kidney transcriptomics. A second set of six mice per group with folic acid nephropathy or vehicle-infused controls were processed in the same manner at 24 and 72 h following the induction of kidney injury and used for real-time quantitative polymerase chain reaction (qPCR) validation. Unilateral ureteral obstruction (UUO) was performed under isoflurane-induced anaesthesia. The left ureter was ligated with silk at two locations and cut to prevent urinary tract infection (obstructed kidney) [21]. A total of five male 12- to 14-week-old C57/BL6 mice were sacrificed 14 days after surgery. UUO kidneys were compared with the healthy contralateral kidneys. Real-time qPCR validation Human metabolomics and murine transcriptomics results were validated by real-time qPCR performed using mRNA obtained from five UUO mice sacrificed on Day 14 and from folic acid nephropathy mice vehicle-infused controls obtained at 24 and 72 h following induction of kidney injury. RNA of 1 µg isolated by Trizol (Invitrogen) was reverse transcribed with a High Capacity cDNA Archive Kit and real-time qPCR was performed on a ABI Prism 7500 PCR system (Applied Biosystems, Foster City, CA, USA) using the ΔΔCt method [22]. Expression levels are displayed as ratios to glyceraldehyde-3-phosphate dehydrogenase. Pre-developed primer and probe assays were also from Applied Biosystems. Data mining The Nephroseq [http://www.nephroseq.org/ (11 October 2017, date last accessed)] database was searched for human renal biopsy transcriptomics data comparing patients with diverse aetiologies of CKD with controls. The database was searched for gene expression data for Slc5a11, Slc6a12 and Slc5A3. Supplementary Figure S1 shows the experimental workflow. RESULTS Urine metabolite markers related to eGFR An OPLS regression analysis of the entire cohort data set measured by 1H NMR revealed spectral features with significant linear relationship to eGFR (Figure 1). Spectral features with the largest absolute covariance and r2 values (Figure 1C) were targeted for identification and quantification (Figure 1A). A list of the identified metabolites can be found in Supplementary Table S2. General linear modelling of all quantified metabolites showed 11 urinary metabolites having significant linear associations with eGFR, adjusted for age, gender, BMI and disease group (Table 2). A positive association with decreased eGFR reflects a decrease in the relative concentration of each respective metabolite as eGFR declines. Table 2. Significant diagnostic metabolites of eGFR Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  This table shows general linear modelling results for diagnostic markers of baseline eGFR. These metabolites show significant linear associations with eGFR. The β coefficient, standard error (SE) and P-value for each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group. Metabolites with a negative correlation (i.e. negative β) indicate metabolites that have increased urinary concentration as eGFR declines. Table 2. Significant diagnostic metabolites of eGFR Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  Metabolite  β  SE  P-value  Citric acid  0.023  0.003  3.42E-10  Uracil  0.02  0.004  3.25E-07  Formic acid  0.01  0.002  4.76E-05  L-threonine  0.016  0.005  3.15E-03  Ethanolamine  0.004  0.001  3.61E-03  myo-Inositol  −0.014  0.005  6.12E-03  Glycolic acid  0.014  0.005  1.14E-02  Indoxyl sulphate  0.009  0.004  1.53E-02  Hippuric acid  0.011  0.004  1.57E-02  p-Cresol sulphate  0.014  0.006  3.03E-02  Betaine  −0.013  0.006  4.79E-02  This table shows general linear modelling results for diagnostic markers of baseline eGFR. These metabolites show significant linear associations with eGFR. The β coefficient, standard error (SE) and P-value for each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group. Metabolites with a negative correlation (i.e. negative β) indicate metabolites that have increased urinary concentration as eGFR declines. FIGURE 1: View largeDownload slide OPLS regression analysis for feature selection in NMR spectra. (A) Sample spectrum with identified metabolites labelled: (1) 2-furoylglycine, (2) 2-hydroxyisobutyric acid, (3) 3-hydroxyisovaleric acid, (4) 4-deoxyerythronic acid, (5) 4-deoxytheonic acid, (6) acetic acid, (7) acetone, (8) trigonelline, (9) betaine, (10) cis-aconitic acid, (11) citric acid, (12) creatine, (13) creatinine, (14) D-glucose, (15) dimethylamine, (16) dimethlyglycine, (17) ethanolamine, (18) formic acid, (19) fumaric acid, (20) glycine, (21) glycolic acid, (22) guanidoacetic acid, (23) pseudouridine, (24) hippuric acid, (25) indoxyl sulfate, (26) L-acetylcarnitine, (27) L-alanine, (28) trimethylamine, (29) L-lactic acid, (30) L-phenylalanine, (31) L-threonine, (32) trimethylamine-N-oxide, (33) L-tyrosine, (34) uracil, (35) myo-inositol, (36) N-methylnicotinamide, (37) p-cresol sulfate, (38) phenylacetylglutamine, (39) phosphorylcholine, (40) p-hydroxyphenylacetic acid, (41) proline betaine, (42) succinic acid, (43) tartaric acid and (44) taurine. (B) Loadings plot of OPLS regression analysis with eGFR in the y variable. This plot represents a ‘statistical’ pseudo-spectrum that displays the OPLS model covariance on the y-axis and chemical shift on the x-axis. The colour bar represents the r2 value for that spectral region to eGFR. Plots in (A) and (B) are aligned on the x-axis, which shows which spectral regions may best describe CKD-related changes and warrant structural identification. (C) Scores plot of the OPLS model showing data separation based on sample eGFR (colour bar). FIGURE 1: View largeDownload slide OPLS regression analysis for feature selection in NMR spectra. (A) Sample spectrum with identified metabolites labelled: (1) 2-furoylglycine, (2) 2-hydroxyisobutyric acid, (3) 3-hydroxyisovaleric acid, (4) 4-deoxyerythronic acid, (5) 4-deoxytheonic acid, (6) acetic acid, (7) acetone, (8) trigonelline, (9) betaine, (10) cis-aconitic acid, (11) citric acid, (12) creatine, (13) creatinine, (14) D-glucose, (15) dimethylamine, (16) dimethlyglycine, (17) ethanolamine, (18) formic acid, (19) fumaric acid, (20) glycine, (21) glycolic acid, (22) guanidoacetic acid, (23) pseudouridine, (24) hippuric acid, (25) indoxyl sulfate, (26) L-acetylcarnitine, (27) L-alanine, (28) trimethylamine, (29) L-lactic acid, (30) L-phenylalanine, (31) L-threonine, (32) trimethylamine-N-oxide, (33) L-tyrosine, (34) uracil, (35) myo-inositol, (36) N-methylnicotinamide, (37) p-cresol sulfate, (38) phenylacetylglutamine, (39) phosphorylcholine, (40) p-hydroxyphenylacetic acid, (41) proline betaine, (42) succinic acid, (43) tartaric acid and (44) taurine. (B) Loadings plot of OPLS regression analysis with eGFR in the y variable. This plot represents a ‘statistical’ pseudo-spectrum that displays the OPLS model covariance on the y-axis and chemical shift on the x-axis. The colour bar represents the r2 value for that spectral region to eGFR. Plots in (A) and (B) are aligned on the x-axis, which shows which spectral regions may best describe CKD-related changes and warrant structural identification. (C) Scores plot of the OPLS model showing data separation based on sample eGFR (colour bar). Metabolites from various metabolic pathways were altered with reduced eGFR. Citric acid from the tricarboxylic acid cycle, amino acids (e.g. threonine), lipid metabolism (e.g. ethanolamine), gut microbiome–derived uraemic toxins (e.g. indoxyl sulphate and p-cresol sulphate), uracil and glycolic acid were significantly lower with reduced eGFR. Many of these metabolites have also been reported as significant markers of CKD from previous studies [23–25], further strengthening them as reliable biomarkers. Additionally, two metabolites, myo-inositol and betaine, had negative coefficients, indicating a respective urinary increase with reduced eGFR. Metabolite markers related to CKD progression A nested retrospective subgroup (n = 57) of the cohort was analysed for urine metabolites predictive of CKD progression. These CKD patients were well distributed throughout the range of progression (r2 = 0.96), which was from 5% per year (i.e. improvement in eGFR) to −14.1% per year (i.e. strong deterioration) (Supplementary Figure 2). The percent annual eGFR slope was independent of baseline eGFR (P = 0.44) (Supplementary Figure S3). Linear modelling was again applied using the percent annual eGFR slope change as the y variable. A total of eight metabolites showed significant associations (Table 3). Citric acid, glycolic acid and ethanolamine were again seen to be positively associated with a declining percent annual eGFR slope. Creatinine and dimethylamine were also positively associated. The highest β coefficients in the modelling were seen for betaine and myo-inositol, which had negative correlations (Figure 2). These osmolytes again showed opposite trends from all other significant metabolites. Table 3. Significant prognostic metabolites of CKD progression Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  This table shows general linear modelling results for prognostic markers of CKD progression. These metabolites show significant linear associations with percent annual eGFR slope of the selected progression subset. The β coefficient, standard error (SE) and P-value of each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group A negative association (i.e. negative β) indicates an increase in the metabolite concentration as the percent annual eGFR slope decreases (i.e. CKD progression rate increases). Table 3. Significant prognostic metabolites of CKD progression Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  Metabolite  β  SE  P-value  Creatinine  0.059  0.011  3.57E-06  Citric acid  0.079  0.027  0.005  Ethanolamine  0.039  0.014  0.010  Betaine  −0.153  0.057  0.010  Trimethylamine N-oxide  −0.053  0.022  0.022  Dimethylamine  0.025  0.011  0.025  Glycolic acid  0.09  0.041  0.031  myo-Inositol  −0.091  0.042  0.035  This table shows general linear modelling results for prognostic markers of CKD progression. These metabolites show significant linear associations with percent annual eGFR slope of the selected progression subset. The β coefficient, standard error (SE) and P-value of each metabolite are reported and are adjusted for age, gender, BMI and CKD disease group A negative association (i.e. negative β) indicates an increase in the metabolite concentration as the percent annual eGFR slope decreases (i.e. CKD progression rate increases). FIGURE 2: View largeDownload slide Plots of adjusted generalized linear model for renal osmolytes. A negative percent annual eGFR slope indicates patients that are progressing in CKD. The linear model equation, 95% confidence intervals and P-values are defined in the legend. FIGURE 2: View largeDownload slide Plots of adjusted generalized linear model for renal osmolytes. A negative percent annual eGFR slope indicates patients that are progressing in CKD. The linear model equation, 95% confidence intervals and P-values are defined in the legend. Due to the longitudinal nature of the sample group, the previously mentioned metabolites may act as prognostic urinary biomarkers of CKD outcome. We focused on to examining the renal osmolytes betaine and myo-inositol, as their respective trends were contrary to the other metabolites and their physiological relevance in kidney health is unique. Using receiver operating characteristics (ROC) analysis, the values of betaine and myo-inositol were examined in the first and last quartiles of the CKD progression subgroup. Here, the first quartile (i.e. ‘stable’) showed no CKD progression, with an eGFR change of +5% to −1% per year. The last quartile (i.e. ‘rapid’ progression) showed an eGFR change of −8% to −15% per year. The area under the curve (AUC) of the ROC analysis for betaine was 0.814 and for myo-inositol was 0.781. Combined, both betaine and myo-inositol showed an AUC of 0.84 (Supplementary Figure S4). Decreased kidney expression of betaine and myo-inositol transporters in murine models To determine if osmolyte transporters were a factor for increased urine osmolyte concentrations, kidney transcriptomics of a murine CKD model were analysed for gene expression changes. The kidney gene expression of Slc5a3, Slc5a11 and Slc6a12 (see Figure 3) was assessed by real-time qPCR in experimental kidney fibrosis induced by UUO, a classic murine model of renal fibrosis and inflammation that recapitulates all the key events in CKD leading to fibrosis [26]. Significant downregulation of the genes encoding the myo-inositol transporter (Slc5a11) and the betaine transporter (Slc6a12) were observed (Figure 4A). FIGURE 3: View largeDownload slide Function of gene products in tubular cells. Slc6a12 encodes the betaine transporter bgt-1. Slc5a3 encodes the sodium myo-inositol transporter Smit. Slc5a11 encodes the sodium myo-inositol transporter Smit2. Myox encodes myo-inositol oxygenase, which is involved in myo-inositol synthesis. Transcription factor TonEBP/Nfat5 regulates gene expression of all osmolyte transporter genes; (Slc6a11, Slc5a3, and Slc5a11). FIGURE 3: View largeDownload slide Function of gene products in tubular cells. Slc6a12 encodes the betaine transporter bgt-1. Slc5a3 encodes the sodium myo-inositol transporter Smit. Slc5a11 encodes the sodium myo-inositol transporter Smit2. Myox encodes myo-inositol oxygenase, which is involved in myo-inositol synthesis. Transcription factor TonEBP/Nfat5 regulates gene expression of all osmolyte transporter genes; (Slc6a11, Slc5a3, and Slc5a11). FIGURE 4: View largeDownload slide Gene expression data. (A) Changes in experimental kidney fibrosis assessed by real-time qPCR. Kidney mRNA was measured 14 days after induction of UUO leading to kidney fibrosis. Obstructed kidneys were compared with contralateral unobstructed kidneys. (B) Changes in experimental kidney injury transcriptomics. Kidney mRNA was measured 24 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. (C) Changes in experimental kidney injury assessed by real-time qPCR. Kidney mRNA was measured 24 h and 72 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. *P ≤ 0.05, **P ≤ 0.01. FIGURE 4: View largeDownload slide Gene expression data. (A) Changes in experimental kidney fibrosis assessed by real-time qPCR. Kidney mRNA was measured 14 days after induction of UUO leading to kidney fibrosis. Obstructed kidneys were compared with contralateral unobstructed kidneys. (B) Changes in experimental kidney injury transcriptomics. Kidney mRNA was measured 24 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. (C) Changes in experimental kidney injury assessed by real-time qPCR. Kidney mRNA was measured 24 h and 72 h after induction of kidney injury by a single injection of folic acid and in vehicle-injected controls. *P ≤ 0.05, **P ≤ 0.01. The candidate genes (Slc5a3, Slc5a11, Slc6a12) tested in experimental CKD were also observed in a transcriptomics analysis of kidney injury using a murine model injected with folic acid. Supplementary Table S3 shows all differentially expressed genes in the transcriptomics array with a fold change increasing or decreasing in expression >25% and FDR <5%, as previously reported [27]. Our hypothesis-driven analysis of the transcriptomics database identified decreased expression of Slc6a12, Slc5a3 and Slc5a11 (Figure 4B). Downregulation of these genes did not appear to result from the downregulation of the transcription factor tonicity-responsive enhancer-binding protein (TonEBP/NFAT5), a master regulator of osmolyte transporters, since TonEBP/NFAT5 expression was preserved (fold change 0.96, FDR = ns). Additionally, Myox, the gene encoding myo-inositol oxygenase, was also downregulated in kidney injury (Supplementary Figure S5). Real-time qPCR confirmed in an independent validation experiment of the decreased expression of Slc5a3, Slc5a11 and Slc6a12 mRNA at 24 h of kidney injury induction and persisting after 72 h (Figure 4C). These data suggest that these gene expression changes are shared by both CKD and acute kidney injuries. DISCUSSION In the present study we investigated the urine metabolome signatures of patients with CKD and CKD progression. Most significant metabolites decrease in urinary concentration with reduced eGFR and in patients with progressing CKD. However, in these same patients we also observed significant linear increases of myo-inositol and betaine. Both metabolites had the highest absolute β coefficients in our modelling of percent annual eGFR slope and showed prognostic value in predicting patients who rapidly progress from those who remained stable. Kidney transcriptomics identified defective gene expressions of myo-inositol and betaine transporters in murine kidney disease models, providing convincing evidence that elevated osmolyte concentrations seen in the urine are partially a result of perturbed renal osmolyte transport. Figure 5 provides a proposed model of osmolyte dysregulation during CKD. This may be the first demonstration that links urinary osmolyte concentrations with clinical CKD outcomes using a metabolomic analysis supported with transcriptomics. FIGURE 5: View largeDownload slide Hypertonicity response mechanism in CKD. A proposed model of the osmolyte mechanism in CKD. (A) Model of osmolyte reabsorption in a healthy nephron. (B) Tubular cell response mechanism to increased hypertonicity in medullary tissue. (C) Blood and urine osmolyte concentrations in response to hypertonicity. Patients with progressive forms of CKD have increased osmolyte concentrations in urine. FIGURE 5: View largeDownload slide Hypertonicity response mechanism in CKD. A proposed model of the osmolyte mechanism in CKD. (A) Model of osmolyte reabsorption in a healthy nephron. (B) Tubular cell response mechanism to increased hypertonicity in medullary tissue. (C) Blood and urine osmolyte concentrations in response to hypertonicity. Patients with progressive forms of CKD have increased osmolyte concentrations in urine. The osmolality of mammalian blood is normally kept constant at ∼290 mosmol/kg [28] and maintained by regulatory mechanisms such as the renin–angiotensin–aldosterone and antidiuretic hormone systems. However, normal kidney function requires an interstitial osmotic gradient in renal medulla four times that of plasma osmolality. The hyperosmotic conditions created by NaCl and urea are required for urine concentration. However, these conditions can be damaging to cells and proteins [28, 29] and cause water to diffuse out of cells, resulting in cellular shrinkage [30]. Osmolytes have evolved in renal tissues as protective compounds against osmotic stress [31], and studies have explored mechanisms for how osmolytes provide protein stability [32, 33]. Five protective osmolytes are used in renal cells: sorbitol, betaine, myo-inositol, taurine and glycerophosphocholine (GPC), with a large proportion being betaine and myo-inositol [28, 29]. The concentrations of these osmolytes are regulated through the TonEBP/NFAT5 transcription factor [28, 34], which regulates the transcription of all transporter proteins responsible for osmolyte uptake in renal cells. This transcription factor is critical to kidney health and has been proven essential in knockout models [30, 35]. Mechanisms for betaine uptake in medullary cells are well studied [28, 36]. As osmolality increases, so too do medullary betaine levels [28, 37]. Uptake comes primarily from extracellular sources and not from increased local cellular synthesis [28]. Betaine is freely filtered by glomeruli and actively reabsorbed by renal tubules at levels >95% of the filtered load [38, 39] (Figure 5A). Therefore, non-perturbed reabsorption mechanisms for betaine should not allow for an increase in urinary levels as seen in our study. The primary transporter of betaine in medullary cells is betaine/GABA transporter 1 (Bgt-1) [28, 40], which is encoded by the Slc6a12 gene and regulated by TonEBP/NFAT5 [30]. Hypertonicity leads to increased TonEBP/NFAT5 expression and localization of Bgt-1 to apical and basolateral plasma membranes in tubular cells (Figure 5B). It is estimated that betaine could account for ∼25% of the total osmolyte content [40, 41], and a recent study found that betaine ranked high in protein-stabilizing properties [32]. Its high abundance and stabilizing properties provide evidence that betaine is one of the more important organic osmolytes in medullary tissues. A recent study of acute renal ischaemia found kidney tissue had reduced levels of betaine and myo-inositol when compared with controls [42, 43]. Therefore, increased betaine in the urine most likely reflects CKD-related renal stress (Figure 5C). Our data show an increase of betaine in human urine during CKD, most likely due to Bgt-1 transporter downregulation, given the lower gene expression in experimental kidney fibrosis and injury. However, the release of osmolytes into the urine following renal cell damage may also contribute. Myo-inositol also increases in healthy medullary cells when interstitial osmolality increases, and reduced Myox gene expression in experimental kidney injury argues against increased local myo-inositol production being a major driver of higher urinary levels in CKD. Sodium myo-inositol transporters (Smit and Smit2) encoded by the Slc5a3 and Slc5a11 genes are primarily responsible for renal myo-inositol uptake [28] and are also regulated by TonEBP/NFAT5 [30]. Downregulation of Slc5a11 was observed in experimental kidney fibrosis, and Slc5a3 and Slc5a11 were also downregulated during kidney injury, suggestive of a general decrease in myo-inositol uptake capacity by stressed tubular cells. Furthermore, a recent study found myo-inositol significantly elevated in the urine of mice with renal vasculitis [44]. We hypothesize that increased urinary levels of renal osmolytes reflect damage to kidney tissue from hyperosmolar stress, especially in diseases such as diabetic and hypertensive nephropathy, where osmotic and hypertensive stresses cause microvasculature damage [35, 45]. In support, gene expression data of osmolyte transporters point to a dysregulated transporter system as a major factor of osmolytes lost into the urine. Data mining of human kidney biopsy gene expression using the Nephroseq database (www.Nephroseq.org) showed a study exploring gene expression in human CKD with diverse aetiologies [46]. These data identified the downregulated kidney expression of Slc5a11 (foldchange −3.15 discovery, −1.91 validation) and Slc6a12 (foldchange −1.63 discovery, −6.98 validation) mRNA in human CKD kidneys compared with control kidneys. However, no difference in Slc5a3 mRNA expression was observed, as was similarly seen in our experimental kidney fibrosis data. The present data are consistent with the increased recognition of osmotic stress as a driving force behind CKD progression. For example, repeated dehydration and osmotic stress are thought to be a key driver of kidney injury in Meso-American nephropathy patients [47]. Moreover, tolvaptan, an antagonist of antidiuretic hormone receptors that promotes polyuria and decreases medullary hyperosmolarity, has been recently approved to treat polycystic kidney disease. Experimental evidence suggests that tolvaptan or high water intake may be protective in other nephropathies [48]. Finally, data consistent with impaired kidney responses to osmotic stress in CKD are in line with clinical observations that CKD patients are more sensitive to the toxic effects of high-osmolality iodinated contrast media [49]. There are certain limitations in our study. The number of samples in the CKD progression analysis is somewhat low, and a larger, multicentre, validation study is warranted. However, betaine and myo-inositol showed the same significant direction of alteration in the larger-scale analysis of the total cohort eGFR. Furthermore, gene expression data from independent experimental murine models provides interdiscipline validation of the pathophysiological feasibility of osmolyte abnormalities acting as CKD progression biomarkers. Therefore, the increased urinary output of betaine and myo-inositol reflects abnormal tubular transport of osmolytes and an impaired renal medullary response to osmotic stress in CKD progression. SUPPLEMENTARY DATA Supplementary data are available at ndt online. FUNDING The research leading to these results has received funding from the European Union’s Seventh Framework Programme FP7/2007-2013 under grant agreement FP7-PEOPLE-2013-ITN-608332. A.O. and M.D.S.N. were supported by Spanish government FEDER funds RETIC REDINREN RD016/0019. FIS PI16/02057, PI15/00298, CP14/00133, Sociedad Española de Nefrología, Programa Intensificacion Actividad Investigadora (ISCIII/Agencia Lain-Entralgo/CM), Miguel Servet MS14/00133. CONFLICT OF INTEREST STATEMENT A.O. reports grants from the Spanish government during the conduct of the study. REFERENCES 1 Inker LA, Astor BC, Fox CH et al.   KDOQI US commentary on the 2012 KDIGO clinical practice guideline for the evaluation and management of CKD. Am J Kidney Dis  2014; 63: 713– 735 10.1053/j.ajkd.2014.01.416. Google Scholar CrossRef Search ADS PubMed  2 Vassalotti JA, Centor R, Turner BJ et al.   Practical approach to detection and management of chronic kidney disease for the primary care clinician. Am J Med  2016; 129: 153– 162.e7 Google Scholar CrossRef Search ADS PubMed  3 Slocum JL, Heung M, Pennathur S. Marking renal injury: can we move beyond serum creatinine? Transl Res  2012; 159: 277– 289 Google Scholar CrossRef Search ADS PubMed  4 Wettersten HI, Weiss RH. Applications of metabolomics for kidney disease research. Organogenesis  2013; 9: 11– 18 Google Scholar CrossRef Search ADS PubMed  5 Weiss RH, Kim K. Metabolomics in the study of kidney diseases. Nat Rev Nephrol  2011; 8: 22– 33 Google Scholar CrossRef Search ADS PubMed  6 Hocher B, Adamski J. Metabolomics for clinical use and research in chronic kidney disease. Nat Rev Nephrol  2017; 13: 269– 284 Google Scholar CrossRef Search ADS PubMed  7 Zhao Y-Y, Xu Q. Metabolomics in chronic kidney disease. Clin Chim Acta  2013; 422: 59– 69 Google Scholar CrossRef Search ADS PubMed  8 Zhao YY, Cheng XL, Vaziri ND et al.   UPLC-based metabonomic applications for discovering biomarkers of diseases in clinical chemistry. Clin Biochem  2014; 47: 16– 26 Google Scholar CrossRef Search ADS PubMed  9 Heinzmann SS, Merrifield C. A, Rezzi S et al.   Stability and robustness of human metabolic phenotypes in response to sequential food challenges. J Proteome Res  2012; 11: 643– 655 Google Scholar CrossRef Search ADS PubMed  10 Dona AC, Jimenez B, Schäfer H et al.   Precision high-throughput proton NMR spectroscopy of human urine, serum, and plasma for large-scale metabolic phenotyping. Anal Chem  2014; 86: 9887– 9894 Google Scholar CrossRef Search ADS PubMed  11 Schanstra JP, Zürbig P, Alkhalaf A et al.   Diagnosis and prediction of CKD progression by assessment of urinary peptides. J Am Soc Nephrol  2015; 26: 1999– 2010 Google Scholar CrossRef Search ADS PubMed  12 Gil RB, Lehmann R, Schmitt-Kopplin P et al.   1H NMR-based metabolite profiling workflow to reduce inter-sample chemical shift variations in urine samples for improved biomarker discovery. Anal Bioanal Chem  2016; 408: 4683– 4691 Google Scholar CrossRef Search ADS PubMed  13 Xia J, Wishart DS. Using MetaboAnalyst 3.0 for comprehensive metabolomics data analysis. Curr Protoc Bioinform  2016; 55: 14.10.1– 14.10.91 Google Scholar CrossRef Search ADS   14 Dieterle F, Ross A, Schlotterbeck G et al.   Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in 1H NMR metabonomics. Anal Chem  2006; 78: 4281– 4290 Google Scholar CrossRef Search ADS PubMed  15 Kohl SM, Klein MS, Hochrein J et al.   State-of-the art data normalization methods improve NMR-based metabolomic analysis. Metabolomics  2012; 8: 146– 160 Google Scholar CrossRef Search ADS PubMed  16 Veselkov KA, Lindon JC, Ebbels TMD et al.   Recursive segment-wise peak alignment of biological 1H NMR spectra for improved metabolic biomarker recovery. Anal Chem  2009; 81: 56– 66 Google Scholar CrossRef Search ADS PubMed  17 Cloarec O, Dumas M, Craig A et al.   Statistical total correlation spectroscopy : an exploratory approach for latent biomarker identification from metabolic 1H NMR data sets statistical total correlation spectroscopy : an exploratory approach for latent biomarker identification from metabolic. Anal Chem  2005; 77: 1282– 1289 Google Scholar CrossRef Search ADS PubMed  18 Wishart DS, Jewison T, Guo AC et al.   HMDB 3.0—the human metabolome database in 2013. Nucleic Acids Res  2012; 41: D801– D807 Google Scholar CrossRef Search ADS PubMed  19 Hao J, Liebeke M, Astle W et al.   Bayesian deconvolution and quantification of metabolites in complex 1D NMR spectra using BATMAN. Nat Protoc  2014; 9: 1416– 1427 Google Scholar CrossRef Search ADS PubMed  20 Izquierdo MC, Sanz AB, Mezzano S et al.   TWEAK (tumor necrosis factor–like weak inducer of apoptosis) activates CXCL16 expression during renal tubulointerstitial inflammation. Kidney Int  2012; 81: 1098– 1107 Google Scholar CrossRef Search ADS PubMed  21 Ucero AC, Benito-Martin A, Fuentes-Calvo I et al.   TNF-related weak inducer of apoptosis (TWEAK) promotes kidney fibrosis and Ras-dependent proliferation of cultured renal fibroblast. Biochim Biophys Acta  2013; 1832: 1744– 1755 Google Scholar CrossRef Search ADS PubMed  22 Ortiz A, Husi H, Gonzalez-Lafuente L et al.   Mitogen-activated protein kinase 14 promotes AKI. J Am Soc Nephrol  2017; 28: 823– 836 Google Scholar CrossRef Search ADS PubMed  23 Chen DQ, Cao G, Chen H et al.   Gene and protein expressions and metabolomics exhibit activated redox signaling and wnt/β-catenin pathway are associated with metabolite dysfunction in patients with chronic kidney disease. Redox Biol  2017; 12: 505– 521 Google Scholar CrossRef Search ADS PubMed  24 Posada-Ayala M, Zubiri I, Martin-Lorenzo M et al.   Identification of a urine metabolomic signature in patients with advanced-stage chronic kidney disease. Kidney Int  2014; 85: 103– 111 Google Scholar CrossRef Search ADS PubMed  25 Raffler J, Friedrich N, Arnold M et al.   Genome-wide association study with targeted and non-targeted NMR metabolomics identifies 15 novel loci of urinary human metabolic individuality. PLoS Genet  2015; 11: e1005487 Google Scholar CrossRef Search ADS PubMed  26 Ucero AC, Benito-Martin A, Izquierdo MC et al.   Unilateral ureteral obstruction: beyond obstruction. Int Urol Nephrol  2014; 46: 765– 776 Google Scholar CrossRef Search ADS PubMed  27 Martin-Lorenzo M, Gonzalez-Calero L, Ramos-Barron A et al.   Urine metabolomics insight into acute kidney injury point to oxidative stress disruptions in energy generation and H2S availability. J Mol Med  2017; 95: 1399– 1409 Google Scholar CrossRef Search ADS PubMed  28 Burg MB, Ferraris JD. Intracellular organic osmolytes: function and regulation. J Biol Chem  2008; 283: 7309– 7313 Google Scholar CrossRef Search ADS PubMed  29 Yancey PH. Organic osmolytes as compatible, metabolic and counteracting cytoprotectants in high osmolarity and other stresses. J Exp Biol  2005; 208: 2819– 2830 Google Scholar CrossRef Search ADS PubMed  30 Lopez-Rodriguez C, Antos CL, Shelton JM et al.   Loss of NFAT5 results in renal atrophy and lack of tonicity-responsive gene expression. Proc Natl Acad Sci USA  2004; 101: 2392– 2397 Google Scholar CrossRef Search ADS PubMed  31 Yancey PH, Clark ME, Hand SC et al.   Living with water stress: evolution of osmolyte systems. Science  1982; 217: 1214– 1222 Google Scholar CrossRef Search ADS PubMed  32 Street TO, Bolen DW, Rose GD. A molecular mechanism for osmolyte-induced protein stability. Proc Natl Acad Sci USA  2006; 103: 13997– 14002 Google Scholar CrossRef Search ADS PubMed  33 Katayama H, McGill M, Kearns A et al.   Strategies for folding of affinity tagged proteins using GroEL and osmolytes. J Struct Funct Genomics  2009; 10: 57– 66 Google Scholar CrossRef Search ADS PubMed  34 Day CR, Gordon SS, Vaughn CL et al.   A single amino acid substitution in the renal betaine/GABA transporter prevents trafficking to the plasma membrane. Physiol J  2013; 2013: 1 Google Scholar CrossRef Search ADS   35 Brocker C, Thompson DC, Vasiliou V. The role of hyperosmotic stress in inflammation and disease. Biomol Concepts  2012; 3: 345– 364 Google Scholar CrossRef Search ADS PubMed  36 Burg MB. Coordinate regulation of organic osmolytes in renal cells. Kidney Int  1996; 49: 1684– 1685 Google Scholar CrossRef Search ADS PubMed  37 Nakanishi T, Turner RJ, Burg MB. Osmoregulation of betaine transport in mammalian renal medullary cells. Am J Physiol  1990; 258: F1061– F1067 Google Scholar PubMed  38 Lever M, Sizeland PCB, Bason LM et al.   Glycine betaine and proline betaine in human blood and urine. Biochim Biophys Acta  1994; 1200: 259– 264 Google Scholar CrossRef Search ADS PubMed  39 Pummer S, Dantzler WH, Lien Y-HH et al.   Reabsorption of betaine in Henle’s loops of rat kidney in vivo. Am J Physiol Renal Physiol  2000; 278: F434– F439 Google Scholar CrossRef Search ADS PubMed  40 Kempson S. A, Montrose MH. Osmotic regulation of renal betaine transport: transcription and beyond. Pflugers Arch  2004; 449: 227– 234 Google Scholar PubMed  41 Grunewald RW, Oppermann M, Schettler V et al.   Polarized function of thick ascending limbs of Henle cells in osmoregulation. Kidney Int  2001; 60: 2290– 2298 Google Scholar CrossRef Search ADS PubMed  42 Jouret F, Leenders J, Poma L et al.   Nuclear magnetic resonance metabolomic profiling of mouse kidney, urine and serum following renal ischemia/reperfusion injury. PLoS One  2016; 11: e0163021 Google Scholar CrossRef Search ADS PubMed  43 Wei Q, Xiao X, Fogle P et al.   Changes in metabolic profiles during acute kidney injury and recovery following ischemia/reperfusion. PLoS One  2014; 9: e106647 Google Scholar CrossRef Search ADS PubMed  44 Al-Ani B, Fitzpatrick M, Al-Nuaimi H et al.   Changes in urinary metabolomic profile during relapsing renal vasculitis. Sci Rep  2016; 6: 38074 Google Scholar CrossRef Search ADS PubMed  45 Fowler MJ. Microvascular and macrovascular complications of diabetes. Clin Diabetes  2008; 26: 77– 82 Google Scholar CrossRef Search ADS   46 Nakagawa S, Nishihara K, Miyata H et al.   Molecular markers of tubulointerstitial fibrosis and tubular cell damage in patients with chronic kidney disease. PLoS One  2015; 10: e0136994 Google Scholar CrossRef Search ADS PubMed  47 Martín-Cleary C, Ortiz A. CKD hotspots around the world: where, why and what the lessons are. A CKJ review series. Clin Kidney J  2014; 7: 519– 515 Google Scholar CrossRef Search ADS PubMed  48 Clark WF, Sontrop JM, Huang S-H et al.   Hydration and chronic kidney disease progression: a critical review of the evidence. Am J Nephrol  2016; 43: 281– 292 Google Scholar CrossRef Search ADS PubMed  49 Eng J, Wilson RF, Subramaniam RM et al.   Comparative effect of contrast media type on the incidence of contrast-induced nephropathy a systematic review and meta-analysis. Ann Intern Med  2016; 164: 417– 424 Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Nephrology Dialysis TransplantationOxford University Press

Published: Mar 14, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off