High-resolution mass spectrometry for exploring metabolic signatures of sepsis-induced acute kidney injury

Abstract

Sepsis is a commonly encountered scenario in an intensive care unit (ICU), and the kidney is one of the organs frequently affected. Sepsis-induced acute kidney injury (SIAKI) contributes to the high mortality and morbidity. A major reason for this is the lack of biomarkers for SIAKI in ICU patients. Metabolic phenotype signatures of SIAKI may enable metabolite biomarker discovery for diagnosis. With the progress of the Omics, currently, high-resolution mass spectrometry-based metabolomics is being used for biomarker discovery. Orthogonal partial least-squares discriminant analysis (OPLS-DA) afforded a good predictive tool to distinguish patients and detect the specific metabolic patterns. Variable importance for projection (VIP) was conducted to identify potential biomarkers for SIALI. Receiver operating characteristic analysis was performed for evaluating the diagnostic accuracy of potential metabolites. In this study, some potential biomarkers were successfully discovered by the commonly used variable selection method, VIP of OPLS-DA. Six metabolites were identified as the potential biomarkers for distinguishing SIAKI patients. Meanwhile, malonylcarnitine, D-glutamine and 3-methoxytyrosine were found to be the important variables to distinguish between SIAKI patients and healthy cases. According to the receiver operating characteristic analysis, we reported malonylcarnitine, with an area under concentration–time curve value of 0.995, maybe as a novel diagnostic biomarker associated with SIALI patients. Our results indicate that serum metabolic profiling by high-resolution mass spectrometry might be a helpful tool for determining the metabolic phenotype of SIAKI patients.

---

Introduction

Sepsis is a primary cause of mortality and morbidity worldwide.¹ Sepsis-induced acute kidney injury (SIAKI), a commonly encountered scenario in an intensive care unit (ICU), is a frequent and important complication of sepsis in critically ill patients.^2,3 Sepsis often leads to multi-organ dysfunction, and the kidney is one of the organs frequently affected.⁴ Despite extensive research and advances in diagnostic tools, SIAKI leads to high mortality rates,⁵ especially in elderly patients.^6,7 If a kidney is involved, the mortality rate will increase exponentially. Early and specific diagnosis is therefore of benefit for prevention and targeted intervention. Recently, several biomarkers for SIAKI have been detected in urine, such as kidney injury molecule-1, IL-18, and cystatin C, etc.^8–12 However, few studies on metabolites have included SIAKI patients, and this requires further study.

Early diagnosis is critical for survival, but is often challenging because the symptoms of SIAKI are subtle and become apparent only during advanced stages of this disease. Therefore, reliable biomarkers for the diagnosis of SIAKI are urgently required to reduce mortality. The identification of robust biomarkers for SIAKI is a clinical priority. We need better technologies which are convenient, rapid and sensitive. Metabolomics, metabolome analysis, involves technology to analyze the concentrations of low-molecular-weight metabolites comprehensively, and has recently rapidly developed along with improvements in analytical technology.^13–15 It is a fairly rapid approach for studying metabolic changes connected to disease progression.¹⁶ In the medical research field, especially, metabolome analysis plays an important role in discovery of the biomarkers which is considered to be closer to the disease phenotype.^17,18 Metabolic profiling is an emerging diagnostic tool for discovering effective biomarkers reflecting alterations in metabolism of disease.

Metabolomics approach offers a tremendous potential in the analysis of clinical samples, and offers the opportunity to identify diagnostic and predictive biomarkers that could translate into early diagnosis.^19,20 To date, few studies have been described by metabolomics approach for biomarkers exploration in SIAKI. A rapid test based on the stable and specific biomarkers for SIAKI would improve diagnostic accuracy and reduce costs. Value of metabolomics has been studied in the diagnosis of many diseases, and the predictability of this method suggests its potential application in the diagnosis of SIALI. The present study aimed to determine whether the metabolic profiles, based on UPLC-MS, differed between SIALI patients and healthy subjects, to identify the serum-based metabolite biomarkers, and to establish a useful tool for the diagnosis of SIALI patients.

Experimental

Chemicals and reagents

HPLC grade methanol and acetonitrile were provided by Merck (Darmstadt, Germany). Formic acid was obtained from Shanghai Jingchun Reagent Co. Ultrapure water was prepared using a Milli-Q water purification system (Millipore, Billerica, MA, USA). The standards were supplied from Sigma-Aldrich (St. Louis, MO, USA).

Patient enrolment

A prospective observational cohort study was performed in the ICU at the First Affiliated Hospital, Heilongjiang University of Chinese Medicine, in Harbin. The study was approved by Ethics Committee, and informed consent forms were signed by all of the subjects prior to participation in this study. Then, the SIALI patients and healthy subjects were divided into a training set and a test set. All participants were from the Chinese Han population. All patients admitted to the ICU who met the criteria of Consensus Conference 2001,¹³ without cardiopulmonary resuscitation, emergency origin, non-surgical, non-pregnant and non-chronic lung disease. The detailed demographic and clinical data of the participants are presented in Table S1.† The whole blood (5 mL) samples were collected on 7 days after admission to the ICU from First Affiliated Hospital of Heilongjiang University of Chinese Medicine, in the morning (8:00–9:00 am) following 10 hours of fasting.

Serum sample preparation

Fasting venous blood was obtained from all the above mentioned individuals. The blood samples were allowed to clot for 60 min in a freezer (4 °C) and then centrifuged at 5000 g for 5 min. The supernatants (serum) were separated and transferred into new vials, and immediately stored frozen at −80 °C until further metabolic profiling analysis. Prior to the analysis, serum samples were thawed and vortexed for 5 s at room temperature. Subsequently, 100 μL of supernatant was injected into 400 μL of the methanol, vortexed thoroughly and then allowed to settle at 4 °C for 5 min. Every mixture was centrifuged at 11 000g for 10 min at 4 °C, and then the final supernatant filtered through a 0.22 μm membrane before analysis.

UPLC analysis

UPLC analysis was performed on an Acquity ultra-performance liquid chromatography system (Waters corp., Milford, USA). Chromatographic separation was carried out at 40 °C on an ACQUITY UPLC HSS C₁₈ column (2.1 mm × 100 mm, 1.7 μm, Waters, Milford, MA). The column oven was set at 40 °C. The mobile phase consisted of 0.1% formic acid (A) and acetonitrile modified with 0.1% formic acid (B). The gradient elution involved 2–20% acetonitrile for 0–2.0 min, 20–60% for 2.0–4.0 min and 60–98% for 4.0–6.0 min; held at 98% for 2 min, then returned to 2% for 8–10 min and finally held at 2% for 10–12 min. The flow rate was 2 mL min⁻¹ and the injection volume was 5 μL. The blanks of pure methanol were run between every sample to limit carryover. QC samples were prepared by the pooled aliquots of all samples, and were injected to check the system's performance.

Mass spectrometry analysis

We performed MS analysis using a Waters Micromass Q-TOF/MS (Waters corp., Milford, USA) equipped with an ESI source in a positive ion mode. MS parameters were set up as follows: capillary voltage of 3000 V, sample cone voltage of 35 V, nebulizer pressure of 45 psi g, desolvation temperature of 350 °C, desolvation gas flow of 600 L h⁻¹, extraction cone voltage of 3 V, collision energy of 5 eV, source temperature of 110 °C and cone gas flow of 25 L h⁻¹. Data were collected in centroid mode and the mass range was set at m/z 50–1000 in full scan mode range, with a scan time of 0.2 s and an interscan delay of 0.1 s. Leucine enkephalin was used for optimum accuracy and reproducibility of the Q-TOF mass spectrometer system.

Multivariate statistical analysis

The datasets from the UPLC/MS were imported into MarkerLynx (V4.1, Waters corp., Milford, USA), centered, and Pareto scaled to reduce the impact of noise and artifacts on the models. Mass feature extraction by peak alignment was carried out by EZinfo software (V2, Waters corp., Milford, USA) and then used for multivariate pattern recognition analysis. For parameter settings of EZinfo software, the data set contains intensity values for all the detected retention time/mass pairs of the samples. The number of detected retention time/mass pairs depends on the peak detection parameters settings. In this study, it was opted to allow retention times between 0.5 and 8 min, and the number of detected retention time/mass pairs was between 1000 and 10 000, depending on the software package used. Principal component analysis (PCA) was performed to detect the outliers and orthogonal partial least-squares discriminant analysis (OPLS-DA) was carried out to obtain an overview of the data set after mean centering and unit variance scaling. The predictive ability of the multivariate models was evaluated and analyzed via the goodness-of-fit parameter (R²X), proportion of the variance of the response variable that is explained by the model (R²Y). The importance of mass features in the discrimination among classes was visualized by plotting the variable importance projection (VIP) which was used to select the discriminating variables. VIP value reflects the influence of every variable on the classification; only variables with a VIP value above 15 were considered. Moreover, the biomarker candidates were further confirmed by an independent t-test (P < 0.01) using SPSS 17.0 (SPSS Inc., Chicago, IL USA). The differential metabolites were then imported into SPSS software for receiver operating characteristic (ROC) analysis.

Metabolite identification

The detailed method for compound identification has been described in the previous work of authors.²¹ Briefly, we conducted UPLC/Q-TOF MS/MS experiments to obtain fragmentation patterns of selected metabolites for producing their structure information, and then calculated the accurate molecular weight. Accurate mass data and isotopic distributions for the precursor and product ions were analyzed and compared to the spectral data from freely accessible databases. The mass tolerance between measured m/z values and exact mass of the component of interest was set to within 5 ppm. At the same time, fragment ions were subjected to analysis through MS/MS to narrow the scope of target compounds. Potential molecular formulae were calculated by MassFragment application manager software (MassLynx version 4.1, Waters Corporation). Finally, the standards were adopted to support the metabolite identification.

Results and discussion

Metabolic profiling of serum samples

Representative examples of UPLC/MS spectra from patients and healthy controls are shown in Fig. 1. To reveal alterations in global metabolites associated with SIAKI, after alignment and normalization of the data sets, multivariate pattern analyses were conducted. The PCA scores plot (Fig. 2) shows a clear cluster of the QC samples, indicating the instrument has the high stability and reproducibility. Based on the PCA analysis, no samples were identified as outliers and excluded for further statistical analysis. We used OPLS-DA to describe the disease state and screen potential diagnostic biomarkers, and build to reveal specific metabolic changes in the SIALI group and improve their separation. As shown in Fig. 3, the score plot exhibited the clear separation classes between the SIALI and healthy subjects, suggesting that the metabolites were significantly altered in serum metabolome. In addition, high values of the model's goodness-of-fit metrics (R²Y = 0.883, Q² = 0.569) meant that the model was non-overfitting and reliable, and also indicated robustness.

Fig. 1 UPLC-MS typical base peak chromatograms of control subjects (A) and SIALI patients (B) from ICU.

Fig. 2 The score plot of principal component analysis of QC samples using UPLC-Q-TOF/MS. Black dot present serum samples, and samples in blue box present QCs.

Fig. 3 The metabolic profiles on OPLS-DA (R²Y = 0.883, Q² = 0.569) multivariate data analysis for discrimination between SIALI patients (red dot) and healthy cases (black dot).

Differential metabolites

To illustrate metabolite profiles between SIALI patients and healthy subjects and to identify potential biomarkers, further analysis was performed to discriminate between SIALI patients and healthy subjects. Following the OPLS-DA analysis, we performed multi-step biomarker screening processes, including VIP value, and t-test, to discover reliable biomarkers. By combining the VIP values (Fig. 4) generated from the OPLS-DA model with the results from the two-tailed Student's t-test, 6 metabolites were selected as differential metabolites for SIALI (Table S2†). The panel of metabolites includes remarkably elevated malonylcarnitine, D-glutamine, 3-methoxytyrosine, 5-hydroxykynurenine, and decreased 1-methylhypoxanthine and L-valine. These abnormal metabolite levels in serum reflected the alterations in the metabolic phenotype, which could provide insight into the underlying mechanism of SIALI.

Fig. 4 VIP-plot based on the PLS-DA model for the important variables of SIALI patients and age-matched healthy controls.

ROC analysis of potential biomarkers

To identify the serum metabolite signature that would be more practical in diagnosing SIALI, ROC curve model involving the top ranked differential metabolites were subjected to variable selection based on the training set. The relative concentrations of these serum metabolite biomarkers are presented in Fig. 5. To evaluate the capability of the potential biomarkers to discriminate between SIALI patients and healthy subjects, ROC curves were then used to evaluate the diagnostic performance of these biomarkers. As a result, listed in Table S2,† 3 metabolites including malonylcarnitine, D-glutamine, 3-methoxytyrosine yielded relatively high specificities. Of note, the highest sensitivity of these metabolites reached >0.85, could be accurately predicted in the training set. Notably, malonylcarnitine had a sensitivity of 97.0% and a specificity of 96.3% which were obtained from the ROC curve analysis, and the value of AUC was 0.995 for the validation set, exhibited good diagnostic performance. A panel of biomarkers will have more diagnostic power than one biomarker. Therefore, it was necessary to employ multiple metabolites in diagnosing SIALI patients.

Fig. 5 Changes in the relative intensity levels (mean) of marker metabolites identified by UPLC-Q-TOF/MS.

The combinational use of them has the promising clinical potential to improve the diagnostic accuracy of SIALI. More studies are still needed for further large-scale validation. We used this serum metabolomics method to discriminate between SIALI patients and healthy subjects, as well as to display the metabolic phenotype of SIALI, and thereby identified multiple types of endogenous metabolites as potential biomarkers. We developed a metabolomics method based on UPLC/MS and demonstrated that it could effectively discover reliable biomarkers, for use in diagnosis in clinical practice. This method has great potential to contribute to the next generation of serum metabolite-level diagnostic tests.

Malonylcarnitine is a metabolite that accumulates with specific disruption of fatty-acid oxidation caused by impaired entry of long-chain acylcarnitine esters into the mitochondria and failure of the mitochondrial respiratory chain and malonyl-CoA decarboxylase deficiency. D-Glutamine is a non-essential amino acid present abundantly throughout the body and is involved in many metabolic processes. It is the principal carrier of nitrogen in the body and is an important energy source for many cells. 3-Methoxytyrosine is one of the main biochemical markers for aromatic L-amino acid decarboxylase deficiency, an inborn error of metabolism that affects serotonin and dopamine biosynthesis. 5-Hydroxykynurenine is found in the tryptophan metabolism pathway. It is created from 5-hydroxy-N-formylkynurenine through the action of arylformamidase. 1-Methylhypoxanthine is a methylated hypoxanthine. Hypoxanthine is a naturally occurring purine derivative and a reaction intermediate in the metabolism of adenosine and in the formation of nucleic acids by the salvage pathway. Valine in particular, has been established as a useful supplemental therapy to the diseases.

Conclusion

In summary, high-throughput UPLC/MS based metabolomics approach has been developed to identify SIALI-related characteristic metabolites. Score plot of OPLS-DA showed significant discrimination between the SIALI and healthy groups. The malonylcarnitine, D-glutamine, 3-methoxytyrosine, 5-hydroxykynurenine, and 1-methylhypoxanthine, and L-valine in the SIALI cases were significantly different from the healthy subjects, where 3 metabolites including malonylcarnitine, D-glutamine, 3-methoxytyrosine yielded relatively high sensitivity and specificities. These findings suggest UPLC/MS serum metabolomics may possess great potential for diagnosis of SIALI patients in clinical practice. Undoubtedly, the clinical validation of the developed metabolic signatures requires further studies including large series of patients. Future work should focus on validating the results on a larger scale of serum samples.

Competing financial interests

The authors declare no competing financial interests.

Acknowledgements

This work was supported by grants from the Key Program of Natural Science Foundation of State (Grant No. 81470196, 81302905).

References

1. S. D. Glodowski and G. Wagener, New insights into the mechanisms of acute kidney injury in the intensive care unit, J. Clin. Anesth., 2015, 27(2), 175–180.
2. P. M. Honore, et al., Prevention and treatment of sepsis-induced acute kidney injury: an update, Ann. Intensive Care, 2015, 5(1), 51.
3. P. M. Honore, et al., Presepsin and sepsis-induced acute kidney injury treated with continuous renal replacement therapy: will another promising biomarker bite the dust?, J. Crit. Care, 2015, 19(1), 428.
4. P. Wang, et al., Exogenous Carbon Monoxide Decreases Sepsis-Induced Acute Kidney Injury and Inhibits NLRP3 Inflammasome Activation in Rats, Int. J. Mol. Sci., 2015, 16(9), 20595–20608.
5. M. A. Schick, et al., Sepsis-induced acute kidney injury by standardized colon ascendens stent peritonitis in rats – a simple, reproducible animal model, Intensive Care Medicine Experimental, 2014, 2(1), 34.
6. Y. Chen, et al., Panaxadiol Saponin and Dexamethasone Improve Renal Function in Lipopolysaccharide-Induced Mouse Model of Acute Kidney Injury, PLoS One, 2015, 10(7), e0134653.
7. F. Tan, et al., Dexmedetomidine protects against acute kidney injury through downregulating inflammatory reactions in endotoxemia rats, Biomed. Rep., 2015, 3(3), 365–370.
8. J. Mårtensson and R. Bellomo, Sepsis-Induced Acute Kidney Injury, Crit. Care Clin., 2015, 31(4), 649–660.
9. S. B. Mishra and A. Azim, Antimicrobial dosing in critically ill patients with sepsis-induced acute kidney injury, Indian J. Crit. Care Med., 2015, 19(5), 295–296.
10. Z. Zhang, Biomarkers, diagnosis and management of sepsis-induced acute kidney injury: a narrative review, Heart, Lung and Vessels, 2015, 7(1), 64–73.
11. I. Keir and J. A. Kellum, Acute kidney injury in severe sepsis: pathophysiology, diagnosis, and treatment recommendations, J. Vet. Emerg. Crit. Care, 2015, 25(2), 200–209.
12. A. Kumar and N. P. Singh, Antimicrobial dosing in critically ill patients with sepsis-induced acute kidney injury, Indian J. Crit. Care Med., 2015, 19(2), 99–108.
13. Q. Liang, et al., High-throughput metabolic profiling for discovering metabolic biomarkers of sepsis-induced acute lung injury, RSC Adv., 2016, 6, 11008–11013.
14. Q. Liang, et al., Discovery of serum metabolites for diagnosis of mild cognitive impairment to Alzheimer's disease progression using an optimized metabolomics method, RSC Adv., 2016, 6, 3586–3591.
15. X. Liu, et al., High-Resolution Metabolomics with Acyl-CoA Profiling Reveals Widespread Remodeling in Response to Diet, Mol. Cell. Proteomics, 2015, 14(6), 1489–1500.
16. Q. Liang, et al., Potential urine biomarkers from a high throughput metabolomics study of severe sepsis in a large Asian cohort, RSC Adv., 2015, 5, 102204–102209.
17. F. Li, et al., Metabolomics reveals that tumor xenografts induce liver dysfunction, Mol. Cell. Proteomics, 2013, 12(8), 2126–2135.
18. B. J. Blaise, et al., Metabolic phenotyping of traumatized patients reveals a susceptibility to sepsis, Anal. Chem., 2013, 85, 10850–10855.
19. J. Xu, et al., Global and targeted metabolomics of esophageal squamous cell carcinoma discovers potential diagnostic and therapeutic biomarkers, Mol. Cell. Proteomics, 2013, 12(5), 1306–1318.
20. R. J. Langley, et al., An integrated clinico-metabolomic model improves prediction of death in sepsis, Sci. Transl. Med., 2013, 5, 195ra95.
21. Q. Liang, et al., Metabolomics of alcoholic liver disease: a clinical discovery study, RSC Adv., 2015, 5(98), 80381–80387.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra01192f

---