NMR metabonomics data analysis
Due to slight variations in the pH of the urine samples, the NMR peaks of some metabolites experienced pH-dependent chemical shifts, making it difficult to identify variances for these peaks by PCA using standard rectangular bucketing. Therefore, all spectra were also visually inspected to validate NMR resonances that were potentially changing between groups, and then these peaks were locally aligned to enable reliable P score calculations.
Visual inspection of the NMR spectra of class III/IV and class V LN patients led to the identification of one metabolite, citrate, that had an eight-fold higher urinary concentration (Table ) (P
score = 0.0477) in class III/IV LN patients (1.11 ± 0.97 mM) compared with class V patients (0.14 ± 0.15 mM). The concentration of citrate in the class III/IV group fell into the normal range in human urine [37
] whereas the citrate levels in the class V group were at the lower limit of published values [37
]. A receiver operator characteristic (ROC) analysis indicated the citrate had 100% specificity at 86% sensitivity and an overall 88% accuracy [see Additional data file 1
] for distinguishing between class III/IV and class V LN patients. Urinary citrate levels were also compared with the SLEDAI-R, CI, and P/C ratio (Figure ). The comparisons shown in Figure further indicated that there was a strong correlation between disease class and citrate levels; however, no correlation was observed between class and SLEDAI, CI, or P/C ratio between class III/IV and class V LN patients. Complete summaries of measured patient indices, including others not presented in Figure , are presented in Tables S1 and S2 of the Supplementary Material.
Concentrations and fold-changes of urinary metabolites measured in Class III/IV, Class V LN and FSGS patients.
Box and whisker plots. The plots shown are comparing citrate concentration (mM), taurine concentration (mM), SLE Disease Activity Index (SLEDAI), chronicity index (CI), and urine creatinine (mg/mL) for class III/IV versus class V LN patients.
Visual comparison of the NMR spectra of urine samples of patients with class III/IV versus class V LN also revealed higher concentrations (> 10 fold) of taurine in class V LN (2.29 ± 2.47 mM) compared with class III/IV LN patients, which lacked taurine altogether (Table ) (P
score = 0.00141). Taurine is normally present urine at levels in the range from 50 to 750 μM [38
]. Therefore, the complete absence of taurine in class III/IV patients indicated a renal pathology that was distinct from class V patients who appeared to have mostly normal levels of taurine, with the exception of one patient who had more than five times the normal amount of taurine in the urine (Table ). Discrimination of the two groups (class III/IV versus class V) based on urinary taurine levels was confirmed by PCA by restricting the spectral analysis just to the region of the 1
H NMR spectra that contained the taurine-specific triplet at δ 3.425 ppm (Figure ). Inspection of the PC scores plot of these samples showed a significant separation of the class III/IV from the class V populations, primarily based on higher levels of taurine in the urine of patients with class V LN (Figure ). The magnitude of the cluster separation was quantified by calculating the Mahalanobis distance between the cluster centroids and the statistical significance of the cluster separation was evaluated by calculating the F-value and comparing it with the critical F-value (Figure ). The Mahalanobis distance of 2.282 between class III/IV and class V LN group centroids and the corresponding F-value of 8.353 (critical F-value 3.982) indicated a statistically significant separation of the two groups. The PC loadings plot corresponding to the PC scores plot shown in Figure is shown in Figure . The difference in the mean intensities of the bucket at δ 3.425 ppm, which corresponded to taurine, was found to be statistically significant with a P
value of 0.00151.
Figure 2 Principal component analysis of urine samples from patients with class III/IV LN and class V LN. (a) Two-dimensional principal component analysis scores plot of urine samples from patients with class III/IV LN (green) and class V LN (black) for peaks (more ...)
ROC curve analysis indicated that taurine was a perfect predictor for discrimination between class III/IV and class V LN patients with 100% specificity, 100% sensitivity and 100% accuracy [see Additional data file 1
]. Urinary taurine levels were also compared to SLEDAI, CI, and P/C ratio (Figure ). As indicated above, the data failed to reveal any correlation between LN class and SLEDAI, CI, or urinary P/C ratios; however, a strong correlation between class III/IV and class V LN and was observed with a complete absence of urinary taurine in class III/IV patients compared with mostly normal levels of taurine in class V patients.
Unsupervised PCA was also carried out on normalized 1
H NMR spectra obtained from urine samples from Class V LN and FSGS patients. Inspection of the PC scores plot showed separation of the class V from the FSGS group based on the levels of hippurate, which were in the normal range found in human urine [37
] in the class V LN group (1.22 ± 0.62 mM), compared with the FSGS group, which completely lacked hippurate (Figure ; Table ). The Mahalanobis distance between the cluster centroids of the class V and FSGS groups was 1.781, with a corresponding F-value of 6.096 (critical F-value 3.739) indicating a statistically significant separation of these groups. NMR spectra of patients with class V LN or FSGS were visually compared to validate the differences in hippurate concentrations indicated by PCA. Figure shows a region of the NMR spectrum that contains two triplets unique to hippurate located at δ 7.64 and δ 7.55 ppm confirming that hippurate was present in the class V group but completely absent in the FSGS group. The differences in the mean bucket intensities at δ 7.55 ppm and δ 7.64 between the two groups (> 10-fold change) were found to be significant: δ 7.55 P
value = 0.000171, and δ 7.64 P
value = 0.000379, respectively. ROC curve analysis indicated that hippurate was a perfect discriminator for distinguishing between class V LN and FSGS patients with 100% specificity, 100% sensitivity, and 100% accuracy [see Additional data file 1
]. Interestingly, the complete lack of hippurate in the FSGS group indicated a unique pathology in FSGS compared with LN that leads to depletion of the normal urinary levels of hippurate.
Figure 3 Principal component analysis of urine samples from class V LN patients and focal segmental glomerulosclerosis patients. (a) Two-dimensional principal component analysis scores plot of urine samples from patients with class V LN patients (black) and focal (more ...)
Figure 4 NMR urine spectra in the region from 7.650 to 7.550 ppm of class V LN patients (black) and focal segmental glomerulosclerosis patients (red). The triplets at δ7.64 and δ7.55 belong to the metabolite hippurate, as indicated in the inset. (more ...)
Kidney biopsies are currently required to distinguish between different classes of LN, and between LN and other glomerular disorders, based on characteristic histological features. Obtaining kidney biopsies is invasive, and repetitive performance to guide day-to-day medical decisions is not practical. Although diagnostic, kidney biopsies are not suited to pinpoint altered metabolic processes or biological pathways involved in LN, which if detected, could lead to the identification of novel therapeutic targets. As the kidneys filter and reabsorb metabolites to maintain a metabolic equilibrium, existence of renal pathologies can impair the filtration of small metabolites through the glomerulus and their subsequent re-absorption in the renal tubules leading to changes in metabolic profiles [39
]. Using NMR-spectroscopy, we found the metabolites taurine, citrate, and hippurate differentially excreted in the urine of patients with proliferative LN, membranous LN, and FSGS.
Comparison of the metabolic profiles of class III/IV LN versus class V LN patients indicated that class III/IV patients had normal urinary citrate levels but low urinary taurine levels whereas class V LN patients exhibited low urinary citrate levels but elevated urinary taurine levels. Based on previous reports, citrate and taurine are both measures of tubular cell function [40
]. A possible explanation for the reduced excretion of urinary citrate in the class V LN group could be the presence of metabolic acidosis, which is known to cause decreased urinary excretion of citrate in humans [41
]. It is believed that patients experiencing low urinary citrate output may have renal tubular cells that are more acidotic than in the healthy normal populations [42
]. The proximal tubules are responsible for the regulation of re-absorption and excretion of citrate [43
]. The body's response to metabolic acidosis includes an increase of sodium/citrate co-transporter activity in the kidney, which causes increased citrate transport across the apical membrane into the tubule lumen [44
]. Up-regulation of the co-transporter activity results in the increased re-absorption of citrate and reduced excretion of the metabolite into the urine. The cytosolic citrate metabolism also plays a key role in regulating the amount of citrate excreted into the urine. During metabolic acidosis, alterations in the enzyme ATP citrate lyase also results in decreased urinary citrate excretion [44
Class III/IV LN patients had a striking absence of urinary taurine. Although the pathology leading to a complete absence of urinary taurine is not obvious, the body's store of taurine is known to be regulated by the kidneys and taurine is known to act as an anti-oxidant in a variety of in vitro
and in vivo
systems, and is used to treat renal dysfunction [45
]. Therefore, it is possible that under the conditions of the most severe LN in class III/IV patients the kidneys utilize all available taurine in an attempt to manage or repair the kidney pathology.
Although the majority of patients with class V LN have normal levels of taurine, one patient had more than five times the normal amount of taurine. A possible explanation for the elevated level of taurine in the urine of this exceptional class V LN patient may be a consequence of inadequate re-uptake of taurine into the cells [46
]. Taurine is excreted through both bile and urine, but its total body pool is primarily controlled by the kidneys via the renal tubules [47
]. Previous studies suggest that tubular dysfunction is a risk factor of taurine deficiency [48
]. Patients in renal failure often have low muscle and plasma concentrations of taurine. Although it has been suggested in the past that this was due to reduced taurine synthesis [48
], our results suggest that low taurine levels are actually the result of increased urinary taurine excretion.
Increased urinary taurine may also be a result of changes in cysteine metabolism. Hypercysteinemia is associated with alterations of the sulfur metabolism and/or sulfate transport [49
]. Taurine is known to play a critical role in these processes [49
], and patients with class V LN may be unable to adequately cope with oxidative stress and the elimination of free radicals. Interestingly, based on animal studies, the acquisition of age-related renal fibrosis can be decreased by taurine supplementation [50
], and taurine also has anti-hypertensive effects [51
Histological scoring and quantification of proteinuria are key methods used to survey disease activity and severity in patients with LN [52
]. Pirani et al
. created a scoring system that is semi-quantitative [54
], which was later adapted by Austin et al
] This system was developed to calculate the activity of LN (SLEDAI) by assessing six histological factors focusing on the severity of active lesions in the glomeruli, and the chronicity of the disease by evaluating four histological parameters focusing on the reversibility of LN [53
]. Wallace et al
. provides a table (Table 55-5 in the original article) that outlines the scoring strategies for both the AI and CI [55
]. The degree of proteinuria is determined by measuring the P/C ratio in a 24-hour urine collection. This ratio has previously proven to be a reliable predictor of proteinuria in a study of LN patients [56
]. The changes in citrate and taurine levels in the patients included in this study were plotted against the SLEDAI, CI, and P/C ratios in order to determine whether a correlation existed between any of these conventionally measured indices and biopsy-determined LN class (Figure ). The data showed that no clear correlation existed between these conventional indices or P/C ratios and the LN class; however, our data indicated strong correlations between citrate and taurine levels and LN class. The lack of correlation between renal activity and chronicity with disease class is not surprising given that other studies have shown that when applying these indices to all World Health Organization classes of LN, rather than just diffuse proliferative LN, these indices lack an association with long-term prognosis [57
]. The inability to relate P/C ratio to disease class is also expected because differing degrees of proteinuria are present throughout the patient population. Visual comparison of the above indices as they related to LN class further demonstrated a strong need for continued development of reliable biomarkers that allow for LN class differentiation, as was seen in the changes in the metabolites identified in this study.
Our pilot study also identified one urinary metabolite, hippurate, whose levels differentiated class V LN and FSGS patients. Specifically, class V LN patients had normal levels of urinary hippurate whereas FSGS patients completely lacked hippurate. Although the pathological link associated with a complete lack of urinary hippurate is not evident, the complete lack of urinary hippurate in FSGS patients is striking. A possible pathological cause could be related to having a distinct gut microbial biota, which has been linked to depleted excretion of hippurate in patients with Crohn's disease [59