Renal disease in patients with Type II diabetes is the leading cause of terminal renal failure and a major healthcare problem. Hence early identification of patients prone to develop this complication is important. Diabetic renal damage should be reflected by a change in urinary polypeptide excretion at a very early stage. To analyse these changes, we used an online combination of CE/MS (capillary electrophoresis coupled with MS), allowing fast and accurate evaluation of up to 2000 polypeptides in urine. Employing this technology, we have examined urine samples from 39 healthy individuals and from 112 patients with Type II diabetes mellitus and different degrees of albumin excretion rate. We established a ‘normal’ polypeptide pattern in the urine of healthy subjects. In patients with Type II diabetes and normal albumin excretion rate, the polypeptide pattern in urine differed significantly from normal, indicating a specific ‘diabetic’ pattern of polypeptide excretion. In patients with higher grade albuminuria, we were able to detect a polypeptide pattern indicative of ‘diabetic renal damage’. We also found this pattern in 35% of those patients who had low-grade albuminuria and in 4% of patients with normal albumin excretion. Moreover, we could identify several of the indicative polypeptides using MS/MS sequencing. We conclude that proteomic analysis with CE/MS permits fast and accurate identification and differentiation of polypeptide patterns in urine. Longitudinal studies should explore the potential of this powerful diagnostic tool for early detection of diabetic renal damage.
- capillary electrophoresis coupled with mass spectrometry (CE/MS)
- Type II diabetes
Renal disease in patients with Type II diabetes mellitus has become the leading cause of terminal renal failure and a major healthcare problem in all Western countries . Several observers have documented that, beyond genetic factors, a number of modifiable progression promotors such as hyperglycaemia and blood pressure not only predispose to the development of diabetic renal disease, but also accelerate its progression [1,2]. Early identification of patients who are prone to develop this devastating complication is therefore important, because it enables prompt medical intervention to prevent further disease progression [3–5], thus saving enormous costs in renal replacement therapy of patients who progress to terminal renal failure. The earliest putative diagnostic sign for the presence of diabetic renal damage is a stable increase in albumin excretion rate, i.e. microalbuminuria [1,6]. In contrast with patients with Type I diabetes, the presence of microalbuminuria in Type II diabetics is not specific for diabetic renal damage, because it may also reflect generalized vascular injury or renal damage from other causes, such as hypertension [1,6]. Thus a reliable non-invasive diagnostic tool for early detection of diabetic renal damage would be a major advantage in the management of patients with Type II diabetes mellitus.
Proteomic analysis is now available for large-scale studies of proteins in tissues and body fluids, and analysis of the urinary proteome may represent an important step forward in the non-invasive diagnosis of kidney diseases [7–9]. Two-dimensional PAGE is commonly used for protein separation and can be combined with MS yielding identification of individual proteins [10,11]. However, analysis of one sample generally requires more than 24 h, and, consequently, this technology is not too well suited to high-throughput. SELDI (surface-enhanced laser-desorption–ionization) MS is another tool employed for high-throughput analyses of biological samples [12,13]. However, binding to specific matrices limits the analysis to certain polypeptides. If the polypeptides and proteins could all be identified in a single time-limited step, proteomics of urine could serve as a powerful tool for clinical application. A new technology relying on CE/MS (capillary electrophoresis coupled with MS) with software designed specifically to deal with the enormous amounts of data generated is a step in this direction. This technique permits fast and accurate analysis of several hundred polypeptides simultaneously in a small volume with a high sensitivity . We have developed and tested the feasibility of such a technique for clinical application in the detection of specific urinary polypeptide patterns in patients with primary renal diseases [15,16]. In the present study, we have used this technique in order to establish specific polypeptide patterns in urine of patients with Type II diabetes mellitus and different degrees of albumin excretion rate.
After approval of the Ethics Committee of the Hannover Medical School, we obtained informed consent from all participants. In order to establish normal urinary protein patterns with CE/MS, a group of 39 non-smoking healthy individuals (18 men and 21 women) were studied. To exclude individuals with renal disease, urine and laboratory analyses were performed and only subjects with normal renal function were enrolled, i.e. with a serum creatinine concentration below 1.3 mg/dl. In addition, 112 non-smoking patients with Type II diabetes mellitus and normal serum creatinine concentration (61 men and 51 women) were examined. The duration of diabetes mellitus was at least 3 years, and none of the patients had decompensated heart failure, chronic inflammatory diseases or malignancy. Urine samples were collected from all participants on several occasions in the morning after voiding. They were instructed to refrain from unusual physical activity the day before; and none of them had fever on the study day. Urine samples were tested for presence of leucocyturia and/or haematuria, and then the samples were aliquoted and stored at −30 °C for later measurements. Albumin excretion in the collected samples was measured using nephelometry (Dade Behring GmbH, Marburg, Germany).
In all healthy subjects, urinary albumin excretion rate was below 20 mg/l. In addition, in the urine of 46 diabetic patients, no increased albumin excretion was detectable, whereas albuminuria was present in 66 patients. In this group, we found a higher grade albuminuria, i.e. an albumin excretion rate above 100 mg/l in 21 patients (409±119 mg/l). Low-grade albumin excretion (35±4 mg/l) was present in 45 patients. Clinical data of patients with Type II diabetes classified according to albumin excretion are shown in Table 1.
Aliquots of the urine samples were thawed shortly before use and prepared as described in detail previously . In brief, 2 ml of sample was applied on to a Pharmacia C2 column to enrich for protein and peptides and to remove urea, salts and other contaminating material. Polypeptides were eluted with 50% (v/v) acetonitrile in water containing 0.5% (v/v) formic acid. The eluate was subsequently freeze-dried and resuspended in 20 μl of water. CE/MS analysis was performed as described previously [14,15], using a Beckman Coulter P/ACE system coupled to an ABI Mariner ESI-TOF mass spectrometer via a sheath-flow interface (Applied Biosystems, Foster City, CA, U.S.A.). CE capillaries (inner and outer diameter, 75 and 360 μm respectively, and 90 cm in length) were from Beckman (Fullerton, CA, U.S.A.). The mobile phase contained 30% (v/v) methanol/0.5% (v/v) formic acid in water. The same solution was used for the sheath flow. Samples were injected with a pressure of 6.9 kPa for 20 s. Under these conditions, approx. 200 nl of sample could be injected. The subsequent CE/MS run was performed at +30 kV for 60 min. After each run, the CE capillary was rinsed for 5 min with 0.1 M NaOH, followed by 5 min with water and 5 min with 30% (v/v) methanol/0.5% (v/v) formic acid in water.
Analysis of CE/MS data
The MosaiquesVisu software was used for peak detection, mass deconvolution, three-dimensional data visualization and generating of the polypeptide lists, as described previously . The CE migration time and the measured amplitude were standardized against a master list containing commonly represented polypeptides. The program uses isotopic masses and conjugated masses for the determination of polypeptides. Only polypeptides with deconvoluted masses above 800 Da and charge states of at least two were accepted. All detected polypeptides were deposited and matched in a Microsoft Access database. A comparison between the samples and a search for conformity was performed. Polypeptides were considered identical if the mass deviation was less than 0.05% and the CE time deviation was less than 5 min.
For determination of relevant marker polypeptides in urine, our study population was classified into four groups: controls (healthy individuals); D+/A−, patients with Type II diabetes and albumin excretion within the normal range; D+/A±, diabetic patients positive for albuminuria, but with an albumin excretion rate below 100 mg/l; D+/A+, diabetic patients with albuminuria above 100 mg/l. Data-mining was performed by deletion of all polypeptides that were not present within at least 40% of one of the four groups. The remaining polypeptides were classified into those present in all four groups and those which were specifically predictive for the presence of diabetes or diabetic renal damage (DRD). The respective formulas are: with h(x) the frequency in group x. Accuracy and continuity of the trend over all four groups was verified manually.
Sequencing of relevant polypeptides using MALDI (matrix-assisted laser-desorption ionization) MS/MS
MS/MS analyses were performed on a MALDI–TOF (time-of-flight)/TOF instrument (Ultraflex; Bruker Daltonik, Bremen, Germany). A complete CE run was spotted on to the MALDI target (one spot every 15 s) with the matrix solution [2 mg/ml α-cyano-4-hydroxycinnamic acid in 50% (v/v) acetonitrile and 0.1% trifluoroacetic acid] added as a sheath liquid at 4 μl/min. The target was subsequently examined in the MS mode for the polypeptides of interest, based on the data from the CE/MS analyses. Polypeptides of interest were sequenced in MS/MS (LIFT) mode.
Data were analysed using a two-tailed ANOVA. When this procedure gave significant results, post-hoc comparison between treatments was carried out using Student's t test for random data and Bonferroni correction. Differences were considered significant at P<0.05. All data are presented as means±S.E.M. The SPSS package (SPSS 11.51 for Windows) was used for statistical analysis.
Establishment of a ‘normal’ urinary polypeptide pattern
Repeated analyses of identical samples did not reveal any significant differences under identical conditions of the CE/MS run for an individual sample [14,15]. The examination of urine obtained from healthy subjects led to the establishment of peaks defined by actual mass and CE retention time of the polypeptides detected, the so-called peak lists. The generation of such peak lists is described previously [14,15]. In one individual urine sample, between 300 and 1000 polypeptides with molecular masses from 800–65000 Da were detected. In a second step, the individual peak lists were deposited in a Microsoft Access database and the probability of each of the polypeptides appearing in a single sample was calculated in order to obtain group-specific polypeptide patterns for healthy subjects. A total of 168 polypeptides were present in over 90% of the urine samples examined. In addition, 159 polypeptides were present in more than 75% of the samples, and an additional 327 were found in over 50% of the urine from healthy individuals. These 654 polypeptides were found in more than 50% of all samples obtained from healthy subjects, and we used them in order to establish a ‘normal polypeptide pattern’. The electronic data manipulation and typical samples from healthy subjects are shown in the upper panel of Figure 1.
Urinary polypeptide patterns in patients with Type II diabetes mellitus
The urine from patients with Type II diabetes mellitus was examined as described above. Typical examples are shown in Figure 1. We arranged and analysed data from the individual runs in subgroups, i.e. patients with normal albumin excretion rate and patients with albuminuria, and compared values from these databases representing the typical polypeptide patterns. We found significant similarity of the polypeptide patterns present in urine samples from each patient within the same group. Typical examples of urinary polypeptide patterns obtained from patients without and with albuminuria are shown in Figure 2. Each group presented with a typical protein contour plot, revealing more than 300 polypeptides. Subsequently, we compared data from all patient groups with those obtained in healthy subjects. Figure 3 shows the occurrence of 80 polypeptides used to discriminate between the four groups. These peptides are listed in Table 2. As evident from Figure 3, significant differences in polypeptide patterns between patients without and with albuminuria were observed. Interestingly, some of the polypeptides present in healthy individuals were missing in the patient subgroups, whereas additional polypeptides appeared in the patient samples. Thus, in D+/A− patients, the polypeptide pattern in urine differed significantly from normal, characterizing a specific ‘diabetic’ pattern of polypeptide excretion. Of even greater importance was the observation that, in urine of patients with D+/A+ (albuminuria above 100 mg/l), a polypeptide pattern indicative of ‘diabetic renal damage’ was detected (Figure 3). This polypeptide pattern was also found in 35% of patients with increased albumin excretion rate and in 4% of patients without albuminuria (Figure 4). The clinical data in the group of patients with Type II diabetes from Table 1 were rearranged with respect to the presence of this ‘diabetic renal damage’ polypeptide pattern. Patients in whom this pattern was present had significantly more retinopathy than patients without such a urinary polypeptide pattern (Table 3). In fact, nine out of ten patients with diabetic retinopathy were now classified as subjects with a ‘diabetic renal damage’ polypeptide pattern.
To characterize further polypeptides that appear indicative for diabetic nephropathy, urine samples were again separated with CE and the effluent was spotted on to a MALDI target. The MALDI target was examined for polypeptides of interest in MS mode and subsequently sequenced in MS/MS mode. Three examples of indicative polypeptides identified using this approach are shown in Figure 5. Figure 5(A) shows a fragment spectrum of a peptide with a molecular mass of 1232.7 Da. A database search using Mascot Software (www.matrixscience.com) identified this sequence as being a fragment of the relaxin-like factor named INSL3 (insulin-like peptide 3). Figure 5(B) shows the spectrum and sequence of a fragment from uromodulin, i.e. THP (Tamm–Horsfall protein), with a molecular mass of 1679.9 Da, and Figure 5(C) shows the sequence of an albumin fragment with a molecular mass of 2228.2 Da.
Proteome analysis in different tissues and body fluids is being used increasingly for clinical diagnosis of various diseases. In this respect, it has advantages over genomic approaches, because co- and post-translational modifications of proteins that may be of relevance for the biological function of proteins are visualized. We have recently developed [14–16] an online combination of CE and MS for fast, sensitive and reproducible identification of (pathological) polypeptide and protein patterns in urine and other biological fluids. Using this approach, we were able to detect specific polypeptide markers in urine of patients with Type II diabetes and normal albumin excretion (D+/A−), i.e. a polypeptide pattern specific for the presence of diabetes mellitus. In this group of patients, we were already able to identify two subjects who had a polypeptide pattern that may indicate the presence of diabetic renal damage. In diabetic patients with low-grade albumin excretion (D+/A±), we could identify 35% of subjects with a typical polypeptide pattern present in the urine of patients with higher grade albumin excretion rate. Moreover, almost all diabetic patients with retinopathy, i.e. another microvascular complication of diabetes, were found in the group of patients with a urinary polypeptide pattern indicative for diabetic renal damage leading to higher grade albuminuria. Thus proteomic analysis using CE/MS applied on a large scale may provide a useful diagnostic tool for early identification of patients who have diabetic renal damage. Prospective epidemiological studies and trials with specific therapeutic interventions such as blood pressure lowering and/or inhibition of the renin–angiotensin system are, therefore, warranted to explore further the potential of this new methodology.
We have established CE/MS because other techniques (MS and HPLC/MS) yielded unsatisfactory results with respect to resolution and speed. Fractionation of the sample with a stable high-resolution method appeared essential. Since CE surpasses HPLC with respect to resolution and speed, the direct coupling of CE to a high-resolution TOF MS seemed most appropriate [17,18]. Although a limitation of CE might be the low amount of sample volume that can be loaded in comparison with LC, this is of little consideration given the high sensitivity of the TOF MS used for CE/MS, which is in the low femtomol range . Alternative approaches, such as two-dimensional electrophoresis, subsequent isolation of the proteins from the gel, digestion and analysis via MS or MS/MS, are too time consuming for routine use . In addition, smaller proteins are difficult or even impossible to visualize with this technique. Another alternative technique, i.e. SELDI MS, is well suited for high-throughput, but a serious problem might be loss of data due to the matrices selecting for particular polypeptides, leading to resolution patterns that represent only a minority of proteins and peptides present in a sample [12,13]. An elegant approach to identify urinary polypeptides could be coupling of HPLC to MS/MS . Although the technique used results in the rapid identification of urinary proteins, it does not appear well suited to obtain a representative pattern of polypeptides present in one sample. Because of tryptic digestion of pooled human urinary proteins for the purpose of protein identification, protein patterns of individual samples or comparisons between samples for diagnostic purposes are not feasible. Using a similar approach on individual urine samples, Cutillas et al.  were able to analyse several native (undigested) peptides. CE/MS represents a tool to visualize a large number of native polypeptides, resulting in proteomic patterns comparable within healthy subjects and different subsets of patients, as confirmed in the present study. The technique can be easily used for detection of polypeptide and protein patterns in other biological fluids as well and allows rapid screening for disease parameters. We have found several polypeptides which clearly discriminate between healthy individuals and diabetic patients without and with albuminuria. Furthermore, we could also establish a specific polypeptide pattern indicative of diabetic renal damage presenting with higher grade albuminuria. These polypeptides and proteins, serving as possible markers for a ‘diabetic milieu’ and/or diabetic renal damage, can be identified further using MS/MS analysis. Such an approach may lead to the establishment of new diagnostic and therapeutic targets.
Using MS/MS analysis we have identified three of the discriminating polypeptides, i.e. a fragment of the relaxin-like factor, a polypeptide from THP and an albumin fragment. These polypeptides are fragments of proteins which are known to be affected when renal disease is present. These specific fragments might even serve as better markers of (diabetic) renal damage, because the parental proteins, due to their much larger size, can only be found in urine once advanced renal damage has occurred. In contrast, these much smaller polypeptides should pass more easily through the damaged glomerular filter and, hence, could serve as very early indicators of renal disease. An alternative explanation for the appearance (or absence) of these polypeptides in urine could be changes in the proteolytic activity due to the presence of a kidney disease. Clearly, the concentration of albumin in urine is higher in patients with renal damage, but the increase in the concentration of the identified albumin fragment could also be the result of higher urinary protease activity. The frequencies of the other two identified polypeptides in urine, i.e. fragments of the relaxin-like factor (INSL3) and THP, were both decreased in patients with diabetic renal damage. The latter is one of the most abundant proteins in urine. Torffvit et al.  reported a decrease of the urinary excretion rate for THP in patients with diabetic nephropathy, which is consistent with our present results. Members of the relaxin-like hormone family are implicated in the pathobiology of tumour cell growth, differentiation, invasion and neovascularization , but INSL3 was also described to be important in gubernaculum development involved in testicular descent . In addition, INSL3 is thought to be both an agonist and an antagonist of the relaxin receptor . In this respect, the findings of a recent experimental study  are of particular interest, since accelerated progression of renal fibrosis has been reported in relaxin knockout mice. Hence, in view of our present results in patients with diabetes mellitus, it is tempting to speculate that down-regulation of INSL3 might indicate decreased (intrarenal) relaxin levels with potentially deleterious consequences for the progression of (diabetic) renal disease. Further studies on these potential targets of diabetic renal damage are certainly warranted.
In summary, proteomic analysis of urine with CE/MS is a fast and sensitive approach that could facilitate the identification of patients prone to develop diabetic renal damage at an earlier stage as this is the case with the measurement of albumin excretion. Prospective longitudinal studies are necessary in order to explore further the potential of this new technology.
This work was supported in part by grant #0312939 from BioProfil ‘Funktionelle Genomanalyse’.
Abbreviations: CE/MS, capillary electrophoresis coupled with MS; D+/A−, patients with Type II diabetes and albumin excretion within the normal range; D+/A±, diabetic patients positive for albuminuria, but with an albumin excretion rate below 100 mg/l; D+/A+, diabetic patients with albuminuria above 100 mg/l; INSL3, insulin-like peptide 3; MALDI, matrix-assisted laser-desorption ionization; SELDI, surface-enhanced laser-desorption–ionization; THP, Tamm–Horsfall protein; TOF, time-of-flight
- The Biochemical Society