Advertisement

Acta Diabetologica

, Volume 55, Issue 11, pp 1151–1161 | Cite as

Optimization of kidney dysfunction prediction in diabetic kidney disease using targeted metabolomics

  • Isabel Ibarra-González
  • Ivette Cruz-Bautista
  • Omar Yaxmehen Bello-Chavolla
  • Marcela Vela-Amieva
  • Rigoberto Pallares-Méndez
  • Diana Ruiz de Santiago Y Nevarez
  • María Fernanda Salas-Tapia
  • Ximena Rosas-Flota
  • Mayela González-Acevedo
  • Adriana Palacios-Peñaloza
  • Mario Morales-Esponda
  • Carlos Alberto Aguilar-Salinas
  • Laura del Bosque-Plata
Original Article
Part of the following topical collections:
  1. Diabetic Nephropathy

Abstract

Aims

Metabolomics have been used to evaluate the role of small molecules in human disease. However, the cost and complexity of the methodology and interpretation of findings have limited the transference of knowledge to clinical practice. Here, we apply a targeted metabolomics approach using samples blotted in filter paper to develop clinical-metabolomics models to detect kidney dysfunction in diabetic kidney disease (DKD).

Methods

We included healthy controls and subjects with type 2 diabetes (T2D) with and without DKD and investigated the association between metabolite concentrations in blood and urine with eGFR and albuminuria. We also evaluated performance of clinical, biochemical and metabolomic models to improve kidney dysfunction prediction in DKD.

Results

Using clinical-metabolomics models, we identified associations of decreased eGFR with body mass index (BMI), uric acid and C10:2 levels; albuminuria was associated to years of T2D duration, A1C, uric acid, creatinine, protein intake and serum C0, C10:2 and urinary C12:1 levels. DKD was associated with age, A1C, uric acid, BMI, serum C0, C10:2, C8:1 and urinary C12:1. Inclusion of metabolomics increased the predictive and informative capacity of models composed of clinical variables by decreasing Akaike’s information criterion, and was replicated both in training and validation datasets.

Conclusions

Targeted metabolomics using blotted samples in filter paper is a simple, low-cost approach to identify outcomes associated with DKD; the inclusion of metabolomics improves predictive capacity of clinical models to identify kidney dysfunction and DKD-related outcomes.

Keywords

Metabolomics Type 2 diabetes Diabetic kidney disease Filter paper Amino acids Acylcarnitines 

Abbreviations

DKD

Diabetic kidney disease

DBS

Dried blood samples

T2D

Type 2 diabetes

eGFR

Estimated glomerular filtration rate

A1C

Glycosylated hemoglobin

ACEI/ARB

Angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

U/B ratio

Ratio or urinary divided by blood concentration of measured metabolites

ARG

Arginine

CIT

Citrulline

GLY

Glycine

ALA

Alanine

LEU

Leucine + isoleucine

MET

Methionine

PHE

Phenylalanine

TYR

Tyrosine

VAL

Valine

ORN

Ornithine

PRO

Proline

SA

Succinylacetone

C0

Free carnitine

C2

Acetylcarnitine

C3

Propionylcarnitine

C4OH\C3DC

3-Hydroxybutyryl + malonyl carnitine

C5OH\C4DC

3-Hydroxyisovaleril + methylmalonyl carnitine

C5DC\C6OH

Glutaryl + 3-hydroxyhexanoyl carnitine

C6DC

Adipylcarnitine

C4

Butyrylcarnitine

C5

Isovalerylcarnitine

C5:1

Tiglylcarnitine

C6

Hexanoylcarnitine

C8

Octanoylcarnitine

C8:1

Octenoylcarnitine

C16

Decanoylcarnitine

C16:1

Decenoylcarnitine

C16:1OH

Decadienoylcarnitine

C16OH

Dodecanoylcarnitine

C10

Dedecenoylcarnitine

C10:1

Tetradecanoylcarnitine

C10:2

Tetradecenoylcarnitine

C12

Tetradecadyenylcarnitine

C12:1

3-Hydroxy-tetradecanoylcarnitine

C14

Hexadecenoylcarnitine

C14:1

Hexadecenoylcarnitine

C14:2

3-Hydroxy-hexadecanoylcarnitine

C14OH

3-Hydroxy-hexadecenoylcarnitinae

C18

Octadecanoylcarnitine

C18:1

Octadecenoylcarnitine

C18:1OH

Octadecenoylcarnitine

C18:2

3-Hydroxy-octadecanoylcarnitine

C18OH

3-Hydroxy-octadecanoylcarnitine

Introduction

Diabetic kidney disease (DKD) is a diabetes complication whose clinical diagnosis is made based on the presence of albuminuria and/or reduced estimated glomerular filtration rate (eGFR) in the absence of signs or symptoms of other primary causes of kidney damage [1]. DKD imposes significant burden in patients by increasing mortality risk and posing barriers for allotransplantation in advanced stages [2, 3]. Early DKD alterations include glomerular hypertrophy, mesangial expansion and basal membrane thickening, with late changes characterized by nodular sclerosis. Unfortunately, identification of histological alterations is costly, invasive and has not shown independent correlation with clinical outcomes [4, 5]. This has led to development of biomarkers to improve identification and screening of DKD. Existing kidney dysfunction indicators in DKD include serum creatinine and albuminuria, both of which have major limitations [6, 7, 8]. Serum creatinine concentration changes until there is an important loss of renal function. In addition, renal function is overestimated by the amount of tubular secretion of creatinine and varies per age, gender, muscle mass and metabolism, body weight, protein and water intake [5]. Albuminuria onset demonstrates established glomerular dysfunction but is not exclusive of DKD, despite being required for diagnosis [2, 7]. Additional kidney function markers in DKD include cystatin-C and the neutrophil gelatinase-associated lipocalin (NGAL), which have been shown to correlate with kidney dysfunction in T2D; nevertheless, their clinical usefulness remains to be further studied [7, 8]. Therefore, development of novel biomarkers that correlate with kidney dysfunction and improve identification of DKD at earlier stages is still an unmet necessity [2, 8].

The study of the metabolome aims to identify small-molecule profiles of complex biological samples instead of individual metabolites to improve the informative capacity of biochemical analyses [9, 10, 11, 12]. The utility of metabolomics in DKD has been proved by different groups, who have reported abnormal plasma concentrations of amino acids and acylcarnitines associated with risk of progression of kidney disease [13, 14, 15]. Targeted metabolomics is an alternative for the study of relatively high number of samples, with the advantage that it can be performed on dry biological samples blotted in filter paper, which is a low-cost approach to store and handle biological samples [15]. To the best of our knowledge, this approach has not been applied in studies of DKD. This work aims to propose the complementary use of metabolomics to improve kidney dysfunction prediction in DKD.

Materials and methods

Subjects and study setting

We performed a cross-sectional evaluation of 200 T2D individuals who were recruited from our Diabetes Clinic and healthy subjects from the Metabolic syndrome cohort of the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (INCMNSZ), which is a cohort study that includes a large number of subjects with and without T2D. We divided the study population into three groups: (1) Healthy normotensive, non-obese subjects with normal kidney function, (2) T2D subjects without DKD (T2DnonDKD) with diabetes diagnosis duration ≥ 10 years and (3) T2D with DKD (T2DDKD) with ≥ 10 years diabetes duration and any grade of diabetic retinopathy to confirm microvascular disease. We excluded subjects with other causes of kidney disease, previous acute ischemic heart disease or any condition that may alter albumin excretion or creatinine clearance. T2D was diagnosed according to standardized ADA recommendations. Albuminuria was defined by urinary albumin > 30 mg/24 h and DKD by the presence of either albuminuria and/or eGFR < 60 mL/min/1.73 m2. The Human Research Ethics Committee of the INCMNSZ approved all proceedings in the study.

Biochemical and anthropometric measurements

Twenty-four hour urine sample recollection was performed the day prior to biochemical evaluation; patients received a container with 200 μl of protease inhibitor (aprotinin protease inhibitor 500 KIU/mg, Thermo Fisher Scientific) and were instructed to collect urine during a 24 h period. After recollection, blood was drawn between 8:00–9:00 am after an overnight fast of 8–12 h. Glucose, total cholesterol, triglycerides, uric acid, serum and urine creatinine, HDL cholesterol, and albuminuria were determined by enzymatic colorimetric commercially available reagents using the Synchron CX Delta (Beckman Coulter); A1C levels were measured by high-performance liquid chromatography. For GFR estimation (eGFR), we calculated creatinine clearance using a 24 h creatinine measurement [16, 17].

In the same visit, a complete medical and family history was obtained from all subjects, including evaluation of dietary protein intake from 1 month and the day prior to urine recollection. Patients were weighed on calibrated scales and height was determined with a floor scale stadiometer. We consigned all medications used by patients; none of the subjects were receiving carnitine or fatty acid supplements during the evaluation.

Metabolomic analyses

Fasting capillary blood and 24 h urine samples obtained the day of biochemical evaluation were collected in filter paper cards (Protein Saver 903 cotton cards, Whatman-GE, USA), dried and conserved in refrigeration until analysis, performed as previously described [18, 19]. Eleven amino acids, free carnitine and 30 acylcarnitines were measured with a commercial kit (NeoBase Non-derivatized MS/MS kit; PerkinElmer Waltham Massachusetts). From each sample, a 3 mm diameter disk was punched with an automatic device (Dried blood spot punch Wallac 1296-071) into a 96-well sample plate and 190 µL of extraction solution containing a mixture of 22 stable isotope-labeled internal standards were added. The plate was covered with aluminum foil, incubated with agitation (30 °C at 650×g for 30 min). 30 µL of sample extracts were directly injected by a 2777 C Waters auto-sampler (Waters Corp., Milford, MA and HPLC pump Waters 1525 µ) to the electrospray tandem mass spectrometry equipment (Quattro Micro API tandem MS using multiple reaction monitoring (MRM) mode), with a flow rate of 1.5 mL/min and an analysis time of 1.5 min. Metabolites were quantified by reference to appropriate internal standards with the MassLynx® software. Low and high analytical controls were included in each plate in triplicate; additionally, a blank sample (extraction solution with internal standards) was included in each plate. Intra- and inter-plate variation coefficients were calculated based on repetitive measurements of the analytical control sample. Inter- and intra-assay variation coefficients ranged from 5 to 9%.

Statistical analysis

Inter-group comparisons

We used Kolmogorov–Smirnov test to explore distribution of each variable. Log and inverse transformations were applied to approximate normality in variables showing non-parametric distribution. Data are presented as mean ± SD or as median and interquartile range, where appropriate. Categorical variables are reported as frequencies and percentages; frequency distribution of categorical variables between groups was compared using chi-squared tests. To evaluate inter-group differences in individual metabolites, we compared metabolite concentrations across groups using ANOVA and Fisher’s LSD for multiple post hoc comparisons. To evaluate the association between metabolites, eGFR and albuminuria, we performed partial correlation analysis adjusted for BMI, A1C, gender, ACRI/ARB use, SBP, DBP and dietary protein intake.

Linear clinical-metabolomics models

We developed manual step-wise multiple linear regression models using eGFR and albuminuria as dependent variables to evaluate the association of clinical, biochemical and metabolomic variables adjusted for age, sex, A1C, ACEI/ARB use, protein consumption, SBP and DBP. Variables were removed from the model until the best fitting model with the maximum adjusted r2 was achieved; to confirm the improvement in the informative capacity of the model we used the Akaike information criterion (AIC) in both models. Models were also tested for multicollinearity using both tolerance and variance inflation factor (VIF). The models were validated using cross-validation derived from a training and validation samples randomly split from the original cohort to correct for over-optimism. Variables selected to enter regression analyses were those correlated significantly with albuminuria and eGFR.

Binary logistic clinical-metabolomics models

Variables associated with eGFR and albuminuria in linear regression analyses were included in binary logistic regression models to detect decreased eGFR (< 60 mL/min) and albuminuria (> 30 mg/24 h), adjusted for age, sex, A1C levels, ACEI/ARB use, protein intake, SBP and DBP. The models were also validated using cross-validation. The performance of the models to be concordant with clinical outcomes was assessed using the area under a receiver operating characteristic (ROC) curve (Harrel’s c statistic) of estimated probabilities obtained from regression analyses and goodness of fit was assessed using the Hosmer–Lemeshow test. A p < 0.05 was considered statistically significant. All statistical analyses were performed using Statistical Package for Social Sciences software (SPSS, version 21.0) and R software (Version 3.4.5).

Results

Clinical and biochemical characteristics of studied subjects

Two-hundred subjects were evaluated as follows: 43 subjects in the control group, 102 in the T2DnonDKD group and 55 in the T2DDKD group (Table 1). Recruitment process and sample size calculation are outlined in Supplementary Material. DKD subjects were significantly older, mostly male and had higher triglyceride, creatinine, A1C and lower HDL-c concentrations compared to other groups (p < 0.001). Seventy-four subjects were treated with ACEI/ARB, 19 subjects had overt DN (albuminuria > 300 g/24 h) and 24 subjects had eGFR < 60 mL/min/1.73 m2.

Table 1

Clinical and biochemical characteristics of studied patients

Parameter

Healthy individuals (N = 43)

T2DnonDKD (N = 102)

T2DDKD (N = 55)

P

Female sex (%)

28 (65.1%)

69 (67.6%)

20 (36.4%)

< 0.001

Age (years)

54.65 ± 9.08*

60.40 ± 8.22

61.71 ± 8.82

< 0.001

BMI (kg/m2)

25.21 ± 3.51*

28.17 + 3.95

27.11 + 3.87

< 0.001

Waist/hip ratio

0.89 ± 0.15*

0.92 ± 0.07

0.94 ± 0.08

0.015

Systolic BP (mmHg)

105.09 ± 13.79*

125.51 ± 16.79

130.6 ± 19.38

< 0.001

Diastolic BP (mmHg)

70.25 ± 8.87*

74.4 ± 9.98

76.71 ± 11.30

0.004

Fasting glucose (mg/dL)

95.34 ± 9.94*

160.58 ± 57.84

161.03 ± 78.83

< 0.001

A1C (%)

5.58 ± 0.36*

8.5 ± 2.03+

9.35 ± 2.24

< 0.001

Triglycerides (mg/dL)

129.0 (80.0-198.0)*

168.0 (111.3-241.3)

184.0 (126.0-230.0)

0.02

Total cholesterol (mg/dL)

203.28 ± 39.70*

184.32 ± 41.77

193.55 ± 41.21

0.08

HDL-C (mg/dL)

55.53 ± 15.61*

46.14 ± 13.08

46.24 ± 15.41

0.002

LDL-C (mg/dL)

118.35 ± 28.40*

100.23 ± 33.91

105.19 ± 30.45

0.02

Serum creatinine (mg/dl)

0.75 ± 0.18

0.73 ± 0.18

1.17 ± 0.84&

< 0.001

Albuminuria (mg/24 h)

4.02 (0.0-5.90)

7.05 (2.75–13.17)

120.0 (37.8-554.4) &

< 0.001

Uric acid (mg/dL)

5.23 ± 1.28

5.18 ± 1.17

6.21 ± 1.56&

< 0.001

Diabetes duration (yr)

16.5 ± 7.31

20.09 ± 8.82

< 0.001

eGFR (ml/min/1.73 m2)

90.44 (79.5-106.7)

99.8 (76.4-120.3)

73.3 (52.6-104.7)&

< 0.001

Values are means ± SD, unless indicated otherwise

T2D Type 2 diabetes mellitus, DKD diabetic kidney disease, BMI body mass index, BP blood pressure, A1C glycosylated hemoglobin, HDL-c high-density lipoprotein cholesterol, LDL-c low-density lipoprotein cholesterol, eGFR estimated glomerular filtration rate

*p < 0.001 Healthy vs. T2DnonDKD and T2DDKD, +T2DnonDKD vs. T2DDKD, &T2DDKD vs. healthy and T2DnonDKD

Metabolite concentrations between groups and correlation with albuminuria and eGFR

We observed significant differences between T2DDKD and other groups in concentrations of blood C0, citrulline, tyrosine and C6, urinary C10:1, U/B proline, C6, C8, C10:1 and C10:2 (p < 0.001, Table 2). Next, we explored correlation between albuminuria, eGFR and metabolomic variables, which are shown in Table 3.

Table 2

Levels of amino acids and acylcarnitines in blood, urine and the urine/blood ratio

Sample

Metabolite concentration (µM)

Healthy subjects median ± IQR

T2DnonDKD median ± IQR

T2DDKD median ± IQR

Blood

Citrulline

24.8 (20.2–30.2)*

15.2 (11.4–20.5)&

19.5 (16.2–27.0)

Methionine

5.1 (3.8–6.1)*

3.1 (2.3–4.4)

3.2 (2.6–4.5)

Phenylalanine

42.9 (38.5–47.2)*

35.8 (31.2–40.3)

36.2 (30.9–41.2)

Tyrosine

56.3 (51.0–67.2)*

50.0 (40.8–60.7)&

46.0 (39.3–53.2)

Valine

141.8 (122.1–160.1)#

119.2 (102.4-139.3)

120.0 (100.0–141.2)

Ornithine

34.0 (30.0–42.4)*

14.7 (10.0-28.7)

16.8 (12.6–27.6)

Carnitine

33.6 (28.0–38.7)#

34.1 (27.7–41.9)&

40.7 (32.1–47.6)

C4OH:C3D

0.04 (0.03–0.05)*

0.04 (0.03–0.06)

0.05 (0.04–0.06)

C5DC\C6OH

0.10 (0.09–0.12)#

0.13 (0.10–0.16)

0.14 (0.11–0.17)

C6

0.04 (0.03–0.04)#

0.03 (0.027–0.040)&

0.04 (0.03–0.05)

C8

0.08 (0.06–0.10)

0.07 (0.05–0.10)&

0.08 (0.06–0.14)

C8:1

0.12 (0.09–0.14)#

0.13 (0.09–0.17)&

0.15 (0.10–0.22)

C10:2

0.01 (0.01–0.02)#

0.01 (0.01–0.02)&

0.02 (0.01–0.02)

C14:1

0.05 (0.04–0.07)*

0.04 (0.03–0.05)

0.05 (0.03–0.06)

C14:2

0.02 (0.02–0.03)*

0.02 (0.01–0.02)

0.02 (0.01–0.02)

C18:1OH

0.02 (0.02–0.03)#

0.02 (0.02–0.03)

0.02 (0.02–0.03)

Urine

Glycine

1855.1 (1042.9-3242.4)#

1614.2 (1025.7–3048.0)

1356.4 (716.5-2229.1)

Proline

28.1 (22.9–37.5)*

51.3 (38.9–81.8)

65.1 (35.7-130.1)

C10:1

1.13 (0.81–2.09)*

1.00 (0.71–1.49)&

0.90 (0.65–1.48)

C12:1

0.76 (0.64–1.05)#

0.73 (0.55-1.00)&

0.58 (0.42–0.88)

Urine/blood

Citrulline

0.70 (0.41–1.01)*

1.12 (0.82–1.85)+

1.24 (0.78–1.90)

Methionine

3.67 (2.37–5.28)*

6.23 (4.18–9.48)

5.93 (3.83–9.66)

Ornithine

0.87 (0.62–1.19)*

2.08 (1.18–3.80)

2.05 (1.26–3.18)

Proline

0.20 (0.14–0.27)*

0.32 (0.24–0.51)&

0.40 (0.26–0.81)

C6

30.0 (18.6–75.2)*

25.6 (15.5–40.6)

16.7 (9.8–36.5)

C8

18.1 (10.1–30.9)#

17.3 (9.3–23.4)

12.3 (7.5–19.1)

C10:1

69.0 (40.5-103.9)#

64.4 (43.9–88.3)&

34.4 (21.0-61.5)

C10:2

279.3 (167.9-370.6)#

196.52 (144.9-345.5)&

148.3 (76.4–201.0)

C12:1

16.8 (11.5–21.1)#

18.6 (14.4–25.1)&

13.4 (11.1–17.6)

C14:2

15.2 (12.3–22.0)*

27.4 (17.6–37.3)

20.6 (14.6–31.5)

Post hoc analysis by Fisher LSD. p values shown are differences between the group with DKD and the other groups

T2D Type 2 diabetes mellitus, DKD diabetic kidney disease, IQR interquartile range

*p value < 0.05 for healthy vs. T2DnonDKD and T2DDKD

+p < 0.05 for healthy vs. T2DnonDKD

#p < 0.05 for healthy vs. T2DDKD

&p < 0.05 for T2DnonDKD vs. T2DDKD

Table 3

Partial correlations between microalbuminuria and eGFR with biochemical and metabolomic variables adjusted by blood pressure, A1C, gender, age, and body mass index

  

Albuminuria

Glomerular filtration rate

Blood

Variable

r

p Value

r

p Value

Citrulline

0.293

< 0.001

− 0.282

< 0.001

C0

0.272

< 0.001

− 0.234

< 0.001

C5DC/C6O

0.045

0.53

− 0.26

< 0.001

C6

0.233

< 0.001

− 0.232

< 0.001

C8

0.15

0.037

− 0.161

0.025

C8:1

0.15

0.037

− 0.438

< 0.001

C10:2

0.312

< 0.001

− 0.474

< 0.001

C14:1

0.124

0.085

− 0.173

0.016

C14:2

0.111

0.124

− 0.181

0.012

C18:1OH

0.199

0.005

− 0.133

0.065

Urine

Glycine

− 0.122

0.09

0.208

0.004

Proline

0.152

0.035

− 0.266

< 0.001

C10:1

− 0.225

0.002

0.312

< 0.001

C12:1

− 0.242

0.001

0.245

0.001

Urine/blood

Proline

0.154

0.032

− 0.257

< 0.001

C6

− 0.142

0.048

0.214

0.003

C8

− 0.165

0.022

0.22

0.002

C10

− 0.195

0.007

0.255

< 0.001

C10:1

− 0.29

< 0.001

0.439

< 0.001

C10:2

− 0.288

< 0.001

0.392

< 0.001

C12:1

− 0.312

< 0.001

0.383

< 0.001

C14:2

− 0.204

0.004

− 0.207

0.004

Linear clinical-metabolomics models

Using step-wise linear regression, we constructed a clinical model to predict albuminuria using A1C, serum creatinine and dietary protein intake, adjusted for age, sex, ACEI/ARB use, SBP and DBP (Table 4); when we introduced metabolomics, we found significant associations for serum citrulline, C0, C10:2 and urinary C12:1, which increased the explained variability of the model and decreased the AIC, thus improving informative capacity of the model. When using eGFR as the dependent variable, significant clinical associations included A1C, years of T2D exposure and uric acid; the inclusion of metabolomics increased the explained variability of the model, with significant associations for serum citrulline, C8:1, C10:2, urinary C10:1 and U/B proline (Table 4), which increased the r2and decreased the AIC of the model.

Table 4

Multiple linear regression analysis showing independent variables associated with albuminuria and glomerular filtration rate

 

Model

Parameters

Parameters

β

Standarized β

t

p value

95% CI

Albuminuria

Clinical

R2 = 0.318

F = 10.772

P < 0.001

AIC = 224.26

A1C

0.191

0.208

3.016

0.003

0.066–0.317

T2D duration

0.041

0.201

2.542

0.012

0.009–0.073

Protein intake

− 0.776

− 0.217

− 3.257

0.001

− 1.246 to − 0.306

Clinical + metabolomics

R2 = 0.460

F = 13.376

P < 0.001

AIC = 183.57

A1C

0.256

0.279

4.408

< 0.001

0.142–0.371

T2D duration

0.031

0.152

2.062

0.041

0.001–0.060

Protein intake

− 0.616

− 0.172

− 2.855

0.005

− 1.041 to − 0.190

Citrulline

2.508

0.217

3.126

0.025

0.213–3.087

C0

2.441

0.137

2.058

0.002

1.290–5.438

C10:2

2.256

0.175

2.703

0.025

0.001–0.020

C12:1 U

− 2.187

− 0.213

− 3.365

< 0.001

− 3.982 to − 1.577

eGFR

Clinical

R2 = 0.395

F = 13.146

P < 0.001

AIC =− 392.14

A1C

0.034

0.180

2.750

0.007

0.010–0.059

T2D duration

− 0.008

− 0.196

− 2.622

0.010

− 0.014 to − 0.002

Uric acid

− 0.116

− 0.372

− 5.902

< 0.001

− 0.155 to − 0.077

Clinical + metabolomics

R2 = 0.650

F = 24.056

P < 0.001

AIC = − 490.0

A1C

0.041

0.218

4.085

< 0.001

0.021–0.061

Uric acid

− 0.065

− 0.209

− 4.101

< 0.001

− 0.097 to − 0.034

Citrulline

− 0.333

− 0.136

− 2.688

0.008

− 0.577 to -0.088

C8:1

− 0.537

− 0.233

− 4.314

< 0.001

− 0.782 to − 0.291

C10:2

− 0.003

− 0.205

− 4.054

< 0.001

− 0.005 to − 0.002

C10:1 U

0.462

0.300

6.083

< 0.001

0.312–0.611

Proline U/B

− 0.329

− 0.259

− 5.400

< 0.001

− 0.449 to − 0.209

Models adjusted by age, sex, BMI, SBP, DBP, A1C, T2D duration and ACEI/ARB use

T2D Type 2 diabetes, BMI body mass index, SBP systolic blood pressure, DBP dyastolic blood pressure, A1C glycosylated hemoglobin, eGFR estimated glomerular filtration rate

Logistic clinical-metabolomics models

When evaluating specific outcomes (Table 5), decreased eGFR was significantly associated with BMI and uric acid levels. The inclusion of metabolomics (C10:2) increased the explained variability and AUC of the model. Similarly, albuminuria > 30 mg/day was associated to years of T2D exposure, A1C, uric acid, creatinine and protein intake; the inclusion of serum C0, C10:2 and urinary C12:1 increased the r2 and AUC of the models. Finally, we developed a clinical-metabolomics model for DKD, which included BMI, A1C, uric acid, age, C0, C10:2, C8:1 and urinary C12:1, which had a higher r2 and AUC compared to the clinical model.

Table 5

Logistic regression analyses using decreased GFR, albuminuria and DKD as dependent variables

 

Model

Parameters

Parameter

β

OR

95%CI

p Value

Glomerular filtration rate < 60 mL/min

Clinical

R2 = 0.344

P < 0.001

χ2 = 6.42, p = 0.600

c-statistic = 0.853

(95%CI 0.776–0.931)

BMI

− 0.170

0.844

0.719–0.990

0.037

Uric acid

0.711

2.036

1.370–3.025

< 0.001

Clinical + metabolomics

R2 = 0.547

P < 0.001

χ2 = 10.31, p = 0.244

c-statistic = 0.924

(95%CI 0.863–0.984)

BMI

− 0.197

0.821

0.682–0.989

0.038

Uric Acid

0.580

1.786

1.149–2.776

0.010

C10:2

0.063

1.065

1.033–1.098

< 0.001

Albuminuria > 30 mg/day

Clinical

R2 = 0.489

P < 0.001

χ2 = 5.69, p = 0.682

c-statistic = 0.891

(95%CI 0.845–0.938)

T2D duration

0.058

1.060

1.006–1.117

0.029

A1C

0.412

1.510

1.212–1.882

< 0.001

Uric acid

0.472

1.603

1.138–2.259

0.007

Creatinine

0.918

2.504

1.054–5.946

0.038

Protein intake

− 1.102

0.332

0.138–0.802

0.014

Clinical + metabolomics

R2 = 0.545

P < 0.001

χ2 = 7.51, p = 0.483

c-statistic = 0.908

(95%CI 0.865–0.951)

T2D duration

0.054

1.051

0.997–1.107

0.062

A1C

0.464

1.590

1.262–2.004

< 0.001

Uric acid

0.394

1.482

1.031–2.132

0.034

Creatinine

0.322

1.380

0.584–3.262

0.463

Protein intake

− 0.861

0.423

0.166–1.077

0.071

C0

0.053

1.054

0.999–1.112

0.056

C10:2

0.021

0.099

0.017–0.582

0.010

C12:1 U

− 2.309

1.021

1.000-1.043

0.049

Diabetic kidney disease

Clinical

R2 = 0.485

P < 0.001

χ2 = 11.41, p = 0.180

c-statistic = 0.879

(95%CI 0.830–0.929)

BMI

− 0.167

0.846

0.743-.963

0.012

A1C

0.353

1.424

1.153–1.758

0.001

Uric acid

0.595

1.813

1.299–2.530

< 0.001

T2D duration

0.071

1.074

1.017–1.134

0.011

Clinical + metabolomics

R2 = 0.589

P < 0.001

χ2 = 4.082, p = 0.850

c-statistic = 0.913

(95% CI 0.874–0.953)

BMI

− 0.192

0.825

0.715–0.951

0.008

A1C

0.520

1.682

1.307–2.165

< 0.001

Uric acid

0.506

1.659

1.138–2.421

0.009

Age

0.078

1.081

1.005–1.162

0.037

C0

0.055

1.056

1.000-1.115

0.050

C10:2

0.033

1.034

1.012–1.056

0.002

C8:1

− 0.157

0.854

0.747–0.977

0.021

C12:1 U

− 1.751

0.174

0.033–0.919

0.039

Models adjusted for by age, sex, BMI, SBP, DBP, A1C, T2D duration and ACEI/ARB use

BMI body mass index, SBP systolic blood pressure, DBP dyastolic blood pressure, A1C glycosylated hemoglobin, eGFR estimated glomerular filtration rate, OR Odds ratio, 95% CI 95% confidence interval, AUC area under the curve

Clinical-metabolomics model validation

We then evaluated both the linear and binary logistic regression models using training (N = 118) and validation datasets (N = 82) to correct findings for over-optimism and validate the findings. We observed an increase in r2 and a decrease in AIC with the inclusion of metabolomics without evidence of multicollinearity for all models, which was replicated in both datasets (Table 6).

Table 6

Model parameters for linear logistic regression clinical and clinical-metabolomics models using a training (N = 118) and validation (N = 82) datasets

Linear model

Model

Sample

Adjusted r2

AIC

p value

Albuminuria

Clinical

Training

0.197

133.06

0.004

Validation

0.278

118.61

< 0.001

Clinical + metabolomics

Training

0.420

110.61

< 0.001

Validation

0.411

99.66

< 0.001

eGFR

Clinical

Training

0.445

− 168.32

< 0.001

Validation

0.354

− 232.52

< 0.001

Clinical + metabolomics

Training

0.720

− 218.88

< 0.001

Validation

0.615

− 286.33

< 0.001

  

Validation

0.524

4.81

0.903 (0.835–0.970)

Logistic model

Model

Sample

Adj. r2

χ 2

c-statistic (95%CI)

Albuminuria (> 30 mg/g)

Clinical

Training

0.465

3.34

0.883 (0.819–0.947)

Validation

0.524

4.81

0.903 (0.835–0.970)

Clinical + metabolomics

Training

0.584

9.14

0.903 (0.848–0.959)

Validation

0.578

4.32

0.901 (0.830–0.971)

eGFR (< 60 mL/min)

Clinical

Training

0.323

3.35

0.845 (0.745–0.945)

Validation

0.487

15.12

0.918 (0.840–0.997)

Clinical + metabolomics

Training

0.530

3.30

0.918 (0.840–0.997)

Validation

0.717

16.14

0.931 (0.835-1.000)

DKD

Clinical

Training

0.469

3.880

0.874 (0.808–0.940)

Validation

0.540

2.948

0.889 (0.815–0.962)

Clinical + metabolomics

Training

0.673

4.193

0.930 (0.884–0.975)

Validation

0.596

6.432

0.886 (0.884–0.975)

eGFR estimated glomerular filtration rate, DKD diabetic kidney disease, AIC Akaike’s information criteria, AUC area under the curve (Harrel’s c-statistic)

Discussion

Our study highlights the development of clinical-metabolomics models related to kidney dysfunction in DKD. Here, we replicate previously reported abnormalities in metabolomics linked to DKD using a validated method of dried biological samples (DBS) blotted in filter paper. This method may facilitate sample handling and could be applied in large-scale efforts to identify new metabolomics-based biomarkers of DKD. Targeted metabolomics has recently been introduced in the study of diverse models of disease and its use in DKD has previously been reported in some animal models and more recently in human subjects. Differences in plasma and/or urine metabolomics between T2DDKD and controls suggest that this condition is associated with abnormalities in glycolysis and lipid and amino acid pathways [20, 21, 22, 23, 24, 25]. At this time, identification of specific biomarkers of DKD using metabolomics is a topic that remains largely unexplored [26].

First, we evaluated clinical-metabolomics profiles that explain the variability in identifying kidney dysfunction in our cohort. We observed that the inclusion of metabolomic and clinical variables improved the explained variability of linear models for albuminuria and eGFR and yielded predictive improvements. In addition, performances of estimated probabilities from clinical models are improved with the inclusion of metabolomics to detect decreased glomerular function, albuminuria and DKD. These observations are consistent with the expected course of kidney dysfunction in DKD, since models included A1C, BMI and years of T2D exposure, as well as protein intake for albuminuria and markers of kidney dysfunction including serum creatinine and uric acid, which is a byproduct of purine metabolism and is elevated in the setting of cellular hypoxia, oxidative stress and inflammation, processes which have been linked to kidney dysfunction, particularly albuminuria, in DKD [27, 28]. Overall, our results demonstrate that inclusion of metabolomics improves the detection threshold of glomerular dysfunction over traditional clinical variables and confirms the significance of studying metabolomics to evaluate kidney dysfunction patients with DKD [29, 30, 31, 32].

As our observations confirmed, altered metabolic pathways in amino acid biosynthesis might be relevant in DKD [33]. Citrulline concentrations are decreased in subjects with T2D regardless of the presence of DKD. However, we observed higher serum concentrations in T2DDKD compared to T2DnonDKD. Elevated levels of citrulline and other urea cycle metabolites have been shown to be related to kidney disease progression in T2D [22, 25, 26]. A possible alteration of citrulline to arginine conversion has been proposed as an explanation, given that these metabolites normally compete with endothelial nitric oxide synthase to increase nitric oxide production, stabilizing endothelial function [34]. In the case of eGFR, increased urinary and U/B proline were also identified in T2DDKD subjects, which indicates increased proline production. High proline levels have been related to insulin deprivation and products of proline metabolism have been linked to glomerular dysfunction in advanced chronic kidney disease [34, 35]. Finally, we observed altered blood concentrations of phenylalanine in patients with DKD, which is similar to previous reports of low tyrosine levels in patients with type 2 diabetes and advanced CKD [36]. Plasma elevations of acylcarnitines in patients with albuminuria has previously been described [37, 38], in our work we only found differences in U/B medium and large chain acylcarnitines between T2DDKD and the other groups. Previous reports have suggested that accumulation of various acylcarnitines in plasma demonstrates impaired metabolite clearance due to CKD and other authors have proposed that an increase of urinary acylcarnitines is associated with early kidney damage, reflecting alterations in the β-oxidation pathway, which has also demonstrated alterations in murine models of diabetic nephropathy [39, 40, 41, 42, 43, 44]. The observed progressive increase of serum carnitine concentrations in our study ranging from controls, T2DnonDKD and T2DDKD subjects has been formerly documented as related to impairment of acylcarnitine excretion and decreased carnitine clearance, indicating mitochondrial damage, which could lead to activation of oxidative stress pathways [43, 45]. Available evidence supports the benefit of carnitine supplementation in hemodialyzed T2D subjects, but the use of carnitine in DKD deserves more profound studies [40]. The contribution of acylcarnitines to our clinical-metabolomics models to identify diseased individuals with albuminuria or decreased glomerular function, indicates the elevated importance of acylcarnitines as markers of glomerular disease in DKD and its implications in identifying kidney dysfunction in subjects with T2D. Future studies should evaluate the role of metabolomics to evaluate treatment response and prediction of changes in kidney function related to acylcarnitine supplementation in individuals with DKD, as well as the role metabolomics to evaluate the impact of T2D medication in ameliorating kidney dysfunction [46]; proving such approach could be useful as a further clinical application for metabolomics research.

As shown, we were able to replicate findings from previous metabolomics approaches using DBS blotted in filter paper, which demonstrates that this technique is both viable and useful for targeted metabolomics in the study of DKD. Studies of the metabolome require collection and storage of biological samples, which is complex and costly [23]. DBS collection in filter paper is a relevant method to study complex biological samples, particularly in studies that face challenges of large sample size, longitudinal assessment or frequent sampling in which DBS collection in filter paper would reduce storage costs and facilitate sample collection and handling [47]. In general terms, any analyte that can be measured from whole blood, serum or plasma can be measured from DBS on filter paper, with the additional advantage of stabilizing and reducing the degradations of numerous analytes due to buffering by the dried blood-matrix [48, 49]. Our results indicate that the use of this approach could be helpful in studying altered metabolic pathways linked to DKD and these results could be extrapolated to other disease models.

Our study had some strengths and limitations. First, we were able to replicate previous findings in metabolomics of DKD using a low-cost approach in both training and replication datasets, which could be implemented in other studies to reduce costs associated to sample processing and storage. Second, we could collect both serum and urine samples to estimate differences in metabolite concentrations in a cohort of patients including healthy individuals, which allowed us to construct clinical-metabolomics models to identify kidney dysfunction in DKD using a targeted metabolomics approach. Furthermore, duration of T2D in our cohort had a wide range of disease exposure, with a minimum of 10 years; this reassures that kidney dysfunction is attributable to T2D. Amongst the limitations of our study is the cross-sectional design, which precluded us from estimating the role of metabolites in identifying progression of in kidney dysfunction and the relatively small number of cases with albuminuria > 300 mg/24 h, which did not allow for comparison of overt diabetic nephropathy cases. Furthermore, since variables were controlled in statistical analysis, there exists a possibility of residual confounding.

In conclusion, our study demonstrates the applications of targeted metabolomics in the study of metabolic alterations in DKD using a low-cost approach. The use of metabolomics evaluated in DBS in filter paper as a complementary method for DKD identification offers a practical alternative that could also shed light on the pathophysiology of DKD. Implementation of predictive models grouping clinical variables to identify glomerular dysfunction and albuminuria are improved with the use of recognized altered metabolites. The role of these metabolites as biomarkers of DKD remains to be studied and confirmed in independent longitudinal follow-up and replication cohorts. Targeted metabolomics in the study of DKD, performed simultaneously in blood and urine samples, is feasible and accessible in DBS collected in filter paper, which is a simple recollection device that allows the possibility of massive sampling, storage and analysis.

Notes

Acknowledgements

All authors approved the submitted version. All the authors would like to thank the staff of the Endocrinology and Metabolism Department for all their support, particularly to Lucia Guillen-Pineda, Maria Del Carmen Moreno-Villatoro, María Guadalupe López-Carrasco and Maria Del Carmen Cruz-Lopez Adriana. We are thankful to the study volunteers for all their work and support throughout the realization of the study. OYBC would like to thank PECEM and Conacyt for their support in his research.

Author contributions

Research idea and study design: IIG, ICB, OYBC, CAAS, LDBP, MVA; data acquisition: RPM, DRSN, MFST, XRF, MGA, APP, MME, OYBC; data analysis/interpretation: OYBC, ICB, IIG; statistical analysis: OYBC, IIG, MVA; manuscript drafting: IIG, ICB, OYBC, MVA, CAAS, LDBP; supervision or mentorship: CAAS, LDBP, MVA. Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008.

Informed consent

Written informed consent was obtained before the examination from each patient, as well as the approval from our institutional ethics committee.

Supplementary material

592_2018_1213_MOESM1_ESM.doc (69 kb)
Supplementary material 1 (DOC 69 KB)

References

  1. 1.
    American Diabetes Association (2001) Clinical practice recommendations 2001: diabetic nephropathy (position statement). Diabetes Care 24(suppl):S69–S72Google Scholar
  2. 2.
    Giorda CB, Carnà P, Salomone M, et al (2018) Ten-year comparative analysis of incidence, prognosis, and associated factors for dialysis and renal transplantation in type 1 and type 2 diabetes versus non-diabetes. Acta Diabetol 55(7):733–740CrossRefGoogle Scholar
  3. 3.
    Penno G, Solini A, Bonora E, Renal Insufficiency Events C (RIACE) Study Group, et al (2018) Defining the contribution of chronic kidney disease to all-cause mortality in patients with type 2 diabetes: the Renal Insufficiency And Cardiovascular Events (RIACE) Italian Multicenter Study. Acta Diabetol 55(6):603–612CrossRefGoogle Scholar
  4. 4.
    Mora-Fernández C, Domínguez-Pimentel V, de Fuentes MM, Górriz JL, et al (2014) Diabetic kidney disease: from physiology to therapeutics. J Physiol 592(18):3997–4012CrossRefGoogle Scholar
  5. 5.
    Zhang J, Wang Y, Gurung P, et al (2018) The relationship between the thickness of glomerular basement membrane and renal outcomes in patients with diabetic nephropathy. Acta Diabetol 55(7):669–679CrossRefGoogle Scholar
  6. 6.
    Susztak K, Böttinger EP (2006) Diabetic nephropathy: a frontier for personalized medicine. J Am Soc Nephrol 17(2):361–367CrossRefGoogle Scholar
  7. 7.
    Urbschat A, Obermüller N, Haferkamp A (2011) Biomarkers of kidney injury. Biomarkers 16(Suppl 1):S22–S30CrossRefGoogle Scholar
  8. 8.
    Rhee EP (2015) Metabolomics and renal disease. Curr Opin Nephrol Hypertens 24(4):371–379PubMedPubMedCentralGoogle Scholar
  9. 9.
    Suhre K, Meisinger C, Döring A, Altmaier E, Belcredi P, Gieger C et al (2010) Metabolic footprint of diabetes: a multiplatform metabolomics study in an epidemiological setting. PLoS One 5(11):e13953CrossRefGoogle Scholar
  10. 10.
    Fiehn O, Garvey WT, Newman JW, Lok KH, Hoppel CL, Adams SH (2010) Plasma metabolomic profiles reflective of glucose homeostasis in non-diabetic and type 2 diabetic obese African–American women. PLoS One 5(12):e15234CrossRefGoogle Scholar
  11. 11.
    Floegel A, Stefan N, Yu Z, Mühlenbruch K, Drogan D, Joost HG et al (2013) Identification of serum metabolites associated with risk of type 2 diabetes using a targeted metabolomic approach. Diabetes 62(2):639–648CrossRefGoogle Scholar
  12. 12.
    Niewczas MA, Sirich TL, Mathew AV, Skupien J, Mohney RP, Warram JH et al (2014) Uremic solutes and risk of end-stage renal disease in type 2 diabetes: metabolomic study. Kidney Int 85(5):1214–1224CrossRefGoogle Scholar
  13. 13.
    Li M, Wang X, Aa J, et al (2013) GC/TOFMS analysis of metabolites in serum and urine reveals metabolic perturbation of TCA cycle in db/db mice involved in diabetic nephropathy. Am J Physiol Renal Physiol 304(11):F1317–F1324CrossRefGoogle Scholar
  14. 14.
    Solini A, Manca ML, Penno G, Pugliese G, Cobb JE, Ferrannini E (2016) Prediction of declining renal function and albuminuria in patients with type 2 diabetes by metabolomics. J Clin Endocrinol Metab 101(2):696–704CrossRefGoogle Scholar
  15. 15.
    Huang T, Cao Y, Zeng J (2016) Tandem mass spectrometry-based newborn screening strategy could be used to facilitate rapid and sensitive lung cancer diagnosis. Onco Targets Ther 9:2479–2487.  https://doi.org/10.2147/OTT.S99099 CrossRefPubMedPubMedCentralGoogle Scholar
  16. 16.
    Wang TJ, Larson MG, Vasan RS, et al (2011) Metabolite profiles and the risk of developing diabetes. Nat Med 17(4):448–453CrossRefGoogle Scholar
  17. 17.
    Chace DH, Kalas TA, Naylor EW (2003) Use of tandem mass spectrometry for multianalyte screening of dried blood specimens from newborns. Clin Chem 49(11):1797–1817CrossRefGoogle Scholar
  18. 18.
    Arreola-Guerra JM, Rincón-Pedrero R, Cruz-Rivera C, Belmont-Pérez T, Correa-Rotter R, Niño-Cruz JA (2014) Performance of MDRD-IDMS and CKD-EPI equations in Mexican individuals with normal renal function. Nefrologia 34(5):591–598PubMedGoogle Scholar
  19. 19.
    Teruel Briones JL, Gomis Couto A, Sabater J, et al (2011) Validation of the chronic kidney disease epidemiology collaboration (CKD-EPI) equation in advanced chronic renal failure. Nefrologia 31(6):677–682PubMedGoogle Scholar
  20. 20.
    Aittokalio T, Schwikowski B (2006) Graph-based methods for analyzing networks in cell biology. Brief Bioinform 7(3):243–255CrossRefGoogle Scholar
  21. 21.
    Han LD, Xia JF, Liang QL, Wang Y, Wang YM, Hu P (2011) Plasma esterified and non-esterified fatty acids metabolic profiling using gas chromatography-mass spectrometry and its application in the study of diabetic mellitus and diabetic nephropathy. Anal Chim Acta 689(1):85–91CrossRefGoogle Scholar
  22. 22.
    Hirayama A, Nakashima E, Sugimoto M, et al (2012) Metabolic profiling reveals new serum biomarkers for differentiating diabetic nephropathy. Anal Bioanal Chem 404(10):3101–3109CrossRefGoogle Scholar
  23. 23.
    Mäkinen VP, Kangas AJ, Soininen P, Würtz P, Groop PH, Ala-Korpela M (2013) Metabolic phenotyping of diabetic nephropathy. Clin Pharmacol Ther 94(5):566–569CrossRefGoogle Scholar
  24. 24.
    Lanza IR, Zhang S, Ward LE, Karakelides H, Raftery D, Nair KS (2010) Quantitative metabolomics by H-NMR and LC–MS/MS confirms altered metabolic pathways in diabetes. PLoS One 5(5):e10538CrossRefGoogle Scholar
  25. 25.
    Pena MJ, Lambers Heerspink HJ, Hellemons ME, Friedrich T, Dallmann G, Lajer M (2014) Urine and plasma metabolites predict the development of diabetic nephropathy in individuals with type 2 diabetes mellitus. Diabet Med 31(9):1138–1147CrossRefGoogle Scholar
  26. 26.
    Stec DF, Wang S, Stothers C, et al (2015) Alterations of urinary metabolite profile in model diabetic nephropathy. Biochem Biophys Res Commun 456(2):610–614CrossRefGoogle Scholar
  27. 27.
    Zhang J, Wang Y, Zhang R, et al (2018) Implication of decreased serum complement 3 in patients with diabetic nephropathy. Acta Diabetol 55(1):31–39CrossRefGoogle Scholar
  28. 28.
    Feng G, Gao JL, Zhang P, et al (2017) Decreased serum extracellular superoxide dismutase activity is associated with albuminuria in Chinese patients with type 2 diabetes mellitus. Acta Diabetol 54(11):1047–1055CrossRefGoogle Scholar
  29. 29.
    Qi S, Ouyang X, Wang L, Peng W, Wen J, Dai Y (2012) A pilot metabolic profiling study in serum of patients with chronic kidney disease based on (1) H-NMR-spectroscopy. Clin Transl Sci 5(5):379–385CrossRefGoogle Scholar
  30. 30.
    Sun J, Shannon M, Ando Y, et al (2012) Serum metabolomic profiles from patients with acute kidney injury: a pilot study. J Chromatogr B Analyt Technol Biomed Life Sci 893:107–113CrossRefGoogle Scholar
  31. 31.
    Goek ON, Döring A, Gieger C, et al (2012) Serum metabolite concentrations and decreased GFR in the general population. Am J Kidney Dis 60(2):197–206CrossRefGoogle Scholar
  32. 32.
    Campion CG, Sanchez-Ferras O, Batchu SN (2017) Potential role of serum and urinary biomarkers in diagnosis and prognosis of diabetic nephropathy. Can J Kidney Health Dis 4:2054358117705371CrossRefGoogle Scholar
  33. 33.
    You H, Gao T, Cooper TK, Morris SM Jr, Awad AS (2013) Arginase inhibition mediates renal tissue protection in diabetic nephropathy by a nitric oxide synthase 3-dependent mechanism. Kidney Int 84(6):1189–1197CrossRefGoogle Scholar
  34. 34.
    Persson P, Fasching A, Teerlink T, Hansell P, Palm F (2014) l-Citrulline, but not l-arginine, prevents diabetes mellitus-induced glomerular hyperfiltration and proteinuria in rat. Hypertension 64(2):323–329CrossRefGoogle Scholar
  35. 35.
    Shah VO, Townsend RR, Feldman HI, Pappan KL, Kensicki E, Vander Jagt DL (2013) Plasma metabolomics profiles in different stages of CKD. Clin J Am Soc Nephrol 8(3):363–370CrossRefGoogle Scholar
  36. 36.
    Duranton F, Lundin U, Gayrard N, Mischak H, Aparicio M, Mourad G et al (2014) Plasma and urinary amino acid metabolomics profiling in patients with different levels of kidney function. Clin J Am Soc Nephrol 9(1):37–45CrossRefGoogle Scholar
  37. 37.
    Ahmad S (2001) L-carnitine in dialysis patients. Semin Dial 14(3):209–217CrossRefGoogle Scholar
  38. 38.
    Wanner C, Förstner-Wanner S, Rössle C, Fürst P, Schollmeyer P, Hörl WH (1987) Carnitine metabolism in patients with chronic renal failure: effect of l-carnitine supplementation. Kidney Int Suppl 22:S132–S135PubMedGoogle Scholar
  39. 39.
    Nkuipou-Kenfack E, Duranton F, Gayrard N, Argilés À, Lundin U, Weinberger KM, Dakna M et al (2014) Assessment of metabolomics and proteomic biomarkers in detection and prognosis of progression of renal function in chronic kidney disease. PLoS One 9(5):e96955CrossRefGoogle Scholar
  40. 40.
    Fouque D, Holt S, Guebre-Egziabher F, Nakamura K, Vianey-Saban C, Hadj-Aïssa A et al (2006) Relationship between serum carnitine, acylcarnitines, and renal function in patients with chronic renal disease. J Ren Nutr 16(2):125–131CrossRefGoogle Scholar
  41. 41.
    Atzler D, Schwedhelm E, Zeller T (2014) Integrated genomics and metabolomics in nephrology. Nephrol Dial Transplant 29(8):1467–1474CrossRefGoogle Scholar
  42. 42.
    van der Kloet FM, Tempels FW, Ismail N, van der Heijden R, Kasper PT, Rojas-Cherto M et al (2012) Discovery of early-stage biomarkers for diabetic kidney disease using ms-based metabolomics (FinnDiane study). Metabolomics 8(1):109–119CrossRefGoogle Scholar
  43. 43.
    Rebouche CJ (2004 Nov) Kinetics, pharmacokinetics, and regulation of l-carnitine and acetyl-l-carnitine metabolism. Ann NY Acad Sci 1033:30–41CrossRefGoogle Scholar
  44. 44.
    Rossi C, Marzano V, Consalvo A, et al (2018) Proteomic and metabolomic characterization of streptozotocin-induced diabetic nephropathy in TIMP3-deficient mice. Acta Diabetol 55(2):121–129CrossRefGoogle Scholar
  45. 45.
    Hocher B, Adamski J (2017) Metabolomics for clinical use and research in chronic kidney disease. Nat Rev Nephrol 13(5):269–284CrossRefGoogle Scholar
  46. 46.
    Chang YH, Hwu DW, Chang DM, An LW, Hsieh CH, Lee YJ (2017) Renal function preservation with pioglitazone or with basal insulin as an add-on therapy for patients with type 2 diabetes mellitus. Acta Diabetol 54(6):561–568CrossRefGoogle Scholar
  47. 47.
    Mei JV, Alexander JR, Adam BW, Hannon WH (2001) Use of filter paper for the collection and analysis of human whole blood specimens. J Nutr 131(5):1631S–1631S6SCrossRefGoogle Scholar
  48. 48.
    McDade TW, Williams S, Snodgrass JJ (2007) What a drop can do: dried blood spots as a minimally invasive method for integrating biomarkers into population-based research. Demography 44(4):899–925CrossRefGoogle Scholar
  49. 49.
    Brindle E, O´Connor KA, Garret DA (2014) Applications of dried blood spots in general human health studies. In: Dried blood spots applications and techniques, 1st edn. Wiley, Hoboken, pp 114–129Google Scholar

Copyright information

© Springer-Verlag Italia S.r.l., part of Springer Nature 2018

Authors and Affiliations

  • Isabel Ibarra-González
    • 1
    • 2
  • Ivette Cruz-Bautista
    • 3
    • 4
    • 5
  • Omar Yaxmehen Bello-Chavolla
    • 3
    • 6
  • Marcela Vela-Amieva
    • 2
  • Rigoberto Pallares-Méndez
    • 3
  • Diana Ruiz de Santiago Y Nevarez
    • 3
  • María Fernanda Salas-Tapia
    • 3
  • Ximena Rosas-Flota
    • 3
  • Mayela González-Acevedo
    • 3
  • Adriana Palacios-Peñaloza
    • 3
  • Mario Morales-Esponda
    • 3
  • Carlos Alberto Aguilar-Salinas
    • 3
    • 4
    • 5
  • Laura del Bosque-Plata
    • 7
  1. 1.Unidad de Genética de la Nutrición, Instituto de Investigaciones BiomédicasUNAM-Instituto Nacional de PediatríaMexico CityMexico
  2. 2.Laboratorio de Errores Innatos del Metabolismo y TamizInstituto Nacional de PediatríaMexico CityMexico
  3. 3.Unidad de Investigación en Enfermedades MetabólicasInstituto Nacional de Ciencias Médicas y Nutrición Salvador ZubiránMexico CityMexico
  4. 4.Departamento de Endocrinología y MetabolismoInstituto Nacional de Ciencias Médicas y Nutrición Salvador ZubiránMexico CityMexico
  5. 5.Tecnológico de MonterreyEscuela de Medicina y Ciencias de la SaludMonterreyMexico
  6. 6.MD/PhD (PECEM) Program, Facultad de MedicinaUniversidad Nacional Autónoma de MéxicoMexico CityMexico
  7. 7.Laboratorio de Nutrigenética y NutrigenómicaInstituto Nacional de Medicina GenómicaMexico CityMexico

Personalised recommendations