The persistent problems regarding the rapid diagnosis of urosepsis have prompted us to explore the potential advantages of metabolomics using ¹H NMR spectroscopic based methods as a diagnostic utility based on the analysis of urine and serum for diagnosis of urosepsis in adults.

Urosepsis is an imprecise term denoting sepsis from a urinary source i.e. severe, dreaded and complicated UTI¹ with high mortality rates.² Urosepsis constitutes upto 25% of all sepsis cases; associated with prolonged hospitalization and higher complication rates. It is clinically diagnosed by estimating serum procalcitonin (PCT) levels along with time taking urine and blood cultures. However urosepsis is closely associated with UTI and extensive work has been done on the urine metabolic profiling during UTI,³ yet serum analysis through metabolomics, in these cases, still remains almost untouched. Further, discoveries of biomarker, if any, will immensely be valued in this field.
Out of 63 subjects enrolled in the study, 31 were categorised as cases of urosepsis while remaining 32 comprised the healthy control group of the present study. The study was ethically approved by Institutional Ethical Committee of Dr Ram Manohar Lohia Institute of Medical Sciences, Lucknow, India. The gold standard method for categorizing urosepsis cases was ‘Serum PCT level’ (>0.5ng/ mL), followed by urine and blood cultures for confirmation. Serum from blood samples was collected in plain vials after centrifugation, whereas urine samples were collected in labeled tubes (containing NaN₃). The samples were stored at -80⁰C until analyzed. The ¹H NMR spectra were recorded using Bruker Biospin Avance III 800 MHz NMR spectrometer (BrukerGmBH, Germany). One dimensional CPMG pulse sequence for serum and 1D NOESY- preset experiments for urine samples were performed. The spectral data, thus obtained was binned into rectangular buckets of 0.01 ppm size( using AMIX software), and subjected to unsupervised multivariate Principal Component Analysis (PCA) followed by supervised PLS-DA (Partial Least Square Discriminant Analysis) using “The Unscrambler X” software package (Version 10.0.1, CAMO, ASA, Norway). The Receiver Operating Characteristic (ROC) curve was generated using SPSS software version 21.Thirty four and thirty five endogenous metabolites were identified in serum and urine respectively. A few assignments of these metabolites in serum and urine spectra have been shown in Figure 1 and 2 respectively. The metabolomic profiling of serum identified four false positive Urosepsis cases with increased PCT levels as separate entities in score plot of PCA (Figure 3a) while in case of urine it was found to be inconclusive (Figure 3b). The metabolic profile of serum showed mainly an elevation of lactate levels and depletion of glucose (Figure 3c). This was accompanied by presence of lipids alongwith depletion of glycerol. The urine PC-1 loadings displayed an elevation of ketonic bodies, lactate, alanine, N-acetyl neuraminic acid and depletion of citrate, DMA, TMAO, creatinine, urea and hippurate (Figure 3d). The PLS-DA models (Figure 4a for serum & 4b for urine samples) generated after removing the false positive cases (as observed in PCA score plot for serum samples) showed good predictive ability with R²= 0.97 in both the biofluids and Q²= 0.87 and 0.85 for serum and urine respectively. The training set of serum samples provided precise prediction of the test set in a small cohort through random re-sampling method, while in the urine samples, the predictions were inconclusive. The diagnostic accuracy of the statistical models were also validated using ROC curve which showed an AUC (Area Under Curve) value of 0.996 and 0.989 for serum and urine samples respectively (Figure 4c & 4d). The metabolic profile of serum of the cases was found to be severely altered with hyperlactatemia and variable hypoglycemia showing different pathophysiological changes (depending upon the severity of infection) during sepsis, severe sepsis and eventually septic shock. The altered urine metabolic profile during urosepsis somewhat indicates organ dysfunction (mainly kidneys) or rather abnormal renal functioning. The appearance of malonate may be due to abnormal bacterial metabolism. This study revealed that serum PCT levels along with urine and blood cultures (which requires about 24-48 hrs) may not account as a rapid gold standard method for diagnosis. Since Urosepsis is a medical emergency and requires immediate patient care, serum PCT levels in conjunction with ¹H NMR metabolic profiling can be utilized as a rapid and efficient differential diagnostic tool for Urosepsis. Although this study had certain limitations, viz. small sample size (n=63) and short study duration (2 years), our findings were promising. Further studies on larger cohorts are needed to determine the significance of these findings on a larger patient population.