Background Urine is formed in the kidney by ultrafiltration from the plasma to eliminate waste products, for instance urea and metabolites. Although the kidney accounts for only 0.5% of total body mass, a large volume of plasma (350-400 ml/100 g tissue/min) flows into the kidney, generating a large amount of ultrafiltrate (150-180 l/day) under normal physiologic conditions [1,2]. Components in the ultrafiltrate such as water, glucose, amino acids, and inorganic salts are selectively reabsorbed, and less than 1% of ultrafiltrate is excreted as urine. Serum proteins are filtered based on their sizes and charges at the glomeruli [3]. After passing through glomeruli, abundant serum proteins such as albumin, immunoglobulin light chain, transferrin, vitamin D binding protein, myoglobin, and receptor-associated protein are reabsorbed, mainly by endocytic receptors, megalin, and cubilin in proximal renal tubules [4-8]. Thus, protein concentration in normal donor urine is very low (less than 100 mg/l when urine output is 1.5 l/day), and normal protein excretion is less than 150 mg/day. This is about a factor 1000 less compared with other body fluids such as plasma. Excretion of more than 150 mg/day protein is defined as proteinuria and is indicative of glomerular or reabsorption dysfunction. Urine can be collected in large amounts fully noninvasively. Therefore, despite the low protein concentration, more than adequate amounts of material (at least 0.5 mg) can be collected from a single sample, although protein in urine must be concentrated. This advantage of urine as a body fluid for diagnosis also allows collection of samples repeatedly over lengthy time periods. Furthermore, normal urinary proteins generally reflect normal kidney tubular physiology because the urinary proteome contains not only plasma proteins but also kidney proteins [7,9-13]. Thus, urine is good material for the analysis of disease processes that affect proximal organs, such as kidney failure resulting from high blood pressure and diabetic nephropathy, which is the most frequent cause of renal failure in the Western world [14]. Urinary proteomics has been conducted by combining various protein concentration and protein separation methods as well as mass spectrometry (MS) technology. In many studies, two-dimensional gel electrophoresis was employed for protein separation. One of these studies, that conducted by Pieper and coworkers [11], identified 150 unique proteins using two-dimensional gel electrophoresis and both matrix-assisted laser desorption ionization time-of-flight MS and liquid chromatography (LC)-tandem mass spectrometry (MS/MS or MS2). However, one-dimensional and two-dimensional chromatographic approaches have been used in several recent studies, resulting in further protein identifications. Pisitkun and coworkers [9] reported identification of 295 unique proteins from the exosome fraction using one-dimensional gel electrophoresis and LC-MS/MS. Sun and colleagues [12] identified 226 unique proteins using one-dimensional gel electrophoresis plus LC-MS/MS and multidimensional liquid chromatography (LC/LC)-MS/MS. Wang and coworkers [13] applied concanavalin A affinity purification for the enrichment of N-glycoprotein in urine and identified 225 proteins using one-dimensional gel electrophoresis plus LC-MS/MS and LC/LC-MS/MS. Recently, Castagna and colleagues [10] exploited beads coated with a hexametric peptide ligand library for urinary protein concentration and equalization, and identified 383 unique gene products by LC-MS/MS using a linear ion trap-Fourier transform (LTQ-FT) instrument. These researchers combined their set of urinary proteins with others derived from the literature to yield a total of about 800 proteins. Some of these five largest urinary proteome catalogues contain proteins with single peptide identification (>30% of total identified proteins reported by Pisitkun and coworkers [9]) and lack an assessment of false-positive ratios. Moreover, proteins identified in these studies seem to be the tip of the iceberg of the urinary proteome, because nearly 1000 protein spots separated by two-dimensional gel remain unidentified [11]. These studies suggest that three steps are especially important for deep analysis: protein concentration from urine with minimal loss; protein separation to reduce the complexity of the protein mixture and remove abundant proteins; and peptide sequencing with high mass accuracy and rapid scanning. In the present study, we employed a simple and straightforward method, namely ultrafiltration, for protein concentration. For protein separation, one-dimensional gel electrophoresis or reverse phase column chromatography was used. For peptide sequencing, we employed methods recently developed in our laboratory involving the LTQ-FT and linear ion trap-orbitrap (LTQ-Orbitrap), which have extremely high mass accuracy [15,16]. The LTQ facilitates accumulation of a greater number of charges than is possible with traditional three-dimensional ion traps, and it is sufficiently fast to enable two consecutive stages of mass spectrometric fragmentation (MS/MS/MS or MS3) on a chromatographic time scale. The Fourier transform-ion cyclotron resonance (FTICR) part of the instrument provides a very high resolution of 100,000 and mass accuracies in the sub-ppm (parts per million) range using selected ion monitoring (SIM) scans. For complex protein samples, the LTQ-FT was shown to increase the number of high-confidence identifications compared with an LCQ instrument [17]. Together, high mass accuracy and MS3 result in dramatically increased confidence for peptide identification [15] and allow 'rescue' of protein identifications by single peptides. A novel hybrid mass spectrometer, the LTQ-Orbitrap [18] also provides a high mass resolving power of 60,000 and high-accuracy mass measurements (sub-ppm on average) using a lock mass strategy, even without SIM scans [15]. These techniques enabled us to identify 1543 proteins in urine from an in-depth study from a single individual and pooled urine obtained from nine individuals, while virtually eliminating false-positive identifications. In the LTQ-FTICR dataset 337 proteins (26.3% of the total identified proteins) were identified with single unique peptide using MS2 and MS3. Around a third of all characterized proteins are annotated as extracellular proteins. In the total data set we found 488 proteins to be annotated as membrane proteins (47% of all proteins with localization information). Of these proteins, 225 proteins were annotated as plasma membrane proteins (21.6%). These proteins include water, drug, sodium, potassium, and chloride transporters that are localized in the kidney and regulate homeostasis of body fluids. This high-confidence collection of proteins present in human urine can serve as a reference for future biomarker discovery. Results Identification of urinary proteins Normal total protein concentration in urine is very low and usually does not exceed 10 mg/100 ml in any single specimen (normal protein excretion is less than 150 mg/day). To concentrate and de-salt urinary proteins, various sample preparation procedures such as ultrafiltration, centrifugation, reverse-phase separation, dialysis, lyophilization, enrichment of proteins by affinity column or beads, and precipitation using organic solvents have been used [9-13,19-21]. As shown in Figure 1, we used an ultrafiltration unit, because it allows us to concentrate and desalt urine samples in a standardized way and to minimize protein loss. Furthermore, the molecular weight cut-off of the ultrafiltration membrane is 3 kDa, leading to removal of low-molecular-weight polypeptides, which are abundant in human urine samples [22,23]. Using the ultrafiltration unit, urine was concentrated about 50-fold. Concentrated protein from single urine sample was separated by one-dimensional sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and reverse phase high-performance liquid chromatraphy (HPLC). We applied crude concentrates to one-dimensional SDS-PAGE (Figure 2a) and cut the gel into 14 or 10 pieces. Protein mixtures were subjected to in-gel tryptic digestion (in-gel 1 and in-gel 2 subsets). We also applied crude concentrates to a novel macroporous reversed phase column (mRP-C18 high-recovery protein column), but resolution was poor initially (data not shown). We therefore depleted human serum albumin from the urine concentrates using an immuno-affinity column and applied the albumin-depleted protein mixture to the column, resulting in a good resolution with 22 fractions (Figure 2b). Separated proteins were denatured by 2,2,2-trifluoroethanol (TFE) [24,25] or urea and thiourea, and were subsequently digested as described in the Materials and methods section (below; in-solution 1 and in-solution 2 subsets). Concentrated urinary protein from pooled samples was separated by one-dimensional SDS-PAGE, and excised in 10 slices (pool subset). Digests from each set were desalted and concentrated on reversed-phase C18 StageTips [26] and analyzed by LC online coupled to electrospray MS. For the single urine sample sets, LC gradients lasted for either 100 or 140 min. The mass spectrometer (LTQ-FTICR) was programmed to perform survey scans of the whole peptide mass range, select the three most abundant peptide signals, and perform SIM scans for high mass accuracy measurements in the FTICR. Simultaneously with the SIM scans, the linear ion trap fragmented the peptide, obtained an MS/MS spectrum, and further isolated and fragmented the most abundant peak in the MS/MS mass spectrum to yield the MS3 spectrum. Figure 3a shows a spectrum of eluting urine peptides. A selected peptide was measured in SIM mode (Figure 3a) and fragmented (MS2; Figure 3b). The most intense fragment in the MS/MS spectrum was selected for the second round of fragmentation (Figure 3c). As can be seen in the figure, high mass accuracy, low background level, and additional peptide sequence information obtained from MS3 spectra yielded high-confidence peptide identification. Peak list files obtained from fractions in each subset were merged and the peptide sequences were identified from their tandem mass spectra using a probability based search engine, namely Mascot [27]. Database searches were performed on 15,919, 16,238, 16,312 and 12,180 MS/MS spectra from in-gel 1, in-gel 2, in-solution 1 and in-solution 2, respectively (Table 1). Identified MS3 spectra were automatically scored with in-house developed open source software, MSQUANT [15,28]. As described in Materials and methods (below), proteins were identified using criteria corresponding to a level of false positives of P = 0.0005 when at least two peptides were identified, and of P = 0.001 when one peptide was identified. We also manually checked MS2 and MS3 spectra for all proteins identified by a single peptide. To test experimentally the false-positive rate in our dataset, we performed a decoy database search [29]. In this approach peptides are matched against the database containing forward-oriented normal sequences and the same sequences with their amino acid sequences reversed. When requiring the stringent criteria mentioned above, we found no false-positive protein hits. We therefore conclude that our search criteria exclude essentially all false positives. Using the criteria established here, our analysis of four datasets, two sets employing in-gel digestion and another two sets employing in-solution digestion, resulted in the identification of 8041 unique peptides. In total, 1281 proteins were identified after the removal of contaminants (keratins, trypsin, and endoproteinase Lys-C) and redundant proteins. For the pooled urine sample, 10 slices from a one-dimensional SDS gel separation were analyzed three times per slice using the LTQ-Orbitrap. A 140 min LC gradient was employed for each analysis. The mass spectrometer was operated in the data-dependent mode. Survey full scan MS spectra (from m/z 300 to 1600) were acquired in the orbitrap and the most intense ions (up to five, depending on signal intensity) were sequentially isolated and fragmented in the linear ion trap (MS/MS). Peak list files obtained from 10 fractions were processed separately and the peptide sequences were identified as described above. Proteins were identified with criteria corresponding to a level of false positives of P = 0.0025 or 1 in 400, which is lower than the total number of proteins in each slice. In this way, independent analysis of the 10 slices allowed us to employ a lower threshold without false-positive identifications, as judged by the decoy database. Altogether, we identified 1055 proteins from 10 slices for the pooled urine sample (Table 2). Of the 8041 peptides identified from urine sample of the single person, 772 (9.6%) were found in all four datasets, 856 (10.6%) were found in three of the four datasets, 2089 (26.0%) were found in two of the four datasets, and the remaining 4324 (53.8%) were found in only one of the four input datasets (Figure 4). Overlaps between in-gel datasets and in-solution datasets were deeper than those between in-gel datasets and an in-solution datasets. Hydrophobicity value of identified peptides in each subset was calculated using the Kyte and Doolittle model [30]. Comparing in-gel specific with in-solution specific peptides, the hydrophobicity values were -0.24 versus -0.54, with an overall hydrophobicity of -0.33 in all datasets. The difference between in-gel and in-solution datasets was not significant but shows the tendency for peptides identified only in in-gel datasets to be more hydrophobic than those identified only in in-solution datasets. As described above the urinary proteome of a single person was investigated in great depth and with different methods. Because the urinary proteome is variable, even from the same individual at different time points, we wished to determine whether the individual urinary proteome was typical. Thus, we compared the overall features of the urinary proteins between single and pooled specimens. As shown in Figure 5, there was deep overlap between the two samples, and the bulk properties in terms of molecular weight and predicted cellular localization were also very similar. Characterization of the urinary proteome via Gene Ontology annotation The identified proteins were functionally categorized based on universal Gene Ontology (GO) annotation terms [31] using the Biological Networks Gene Ontology (BiNGO) program package [32,33]. In total, 1041, 1191, and 1118 proteins were linked to at least one annotation term within the GO cellular component, molecular function, and biological process categories, respectively. In total, 214 and 67 terms exhibited significance (P 24) were included. Only fully tryptic peptides with seven amino acids or longer were accepted for identification. Proteins with at least two peptides and a MS2 score of at least 24 (95% significance level) for one of the peptides and at least 31 (99% significance level) for the other were counted as identified protein. For proteins identified by a single peptide, we required the presence of an MS3 spectrum, an MS2 score of at least 34 (99.5% significance level), and a combined score for MS2 and MS3 of above 41 (99.9% significance level) and a peptide delta score (score difference between first and second candidate sequences obtained from a database search) above 5.0. MS2 and MS3 spectra for all proteins identified by a single peptide were manually checked. For LTQ-Orbitrap data, 10 fractions separated by molecular weight of proteins were analyzed independently. The 95% significance threshold in the database search was a MS2 score of 25 or 26. Proteins were considered positively identified when they were identified with at least two fully tryptic peptides of more than six amino acid length, MS2 score of at least 15 or 16, and a sum of MS2 score of at least 50 or 52 resulting in an expected false-positive rate of 0.25% or 1 in 400. For counting the number of identified proteins across each experiment, redundant protein identification was removed using Blast search function of ProteinCenter and manual check. Enrichment analysis of GO categories We used BiNGO [32,33] with the Cytoscape plugin to find statistically over- or under-represented GO categories in biologic data as the tool for enrichment analysis of our urinary proteome dataset. For enrichment analysis we needed a test dataset (which is our identified urinary proteome) and a reference set of GO annotation for the complete human proteome. As per instructions on the BiNGO webpage, the custom GO annotation for the reference set (of whole IPI human dataset) was created by extracting the GO annotations available for Human IPI IDs from EBI GOA Human 39.0 release [64]. The GOA Human 39.0 release contains annotations for 28,873 proteins compiled from different sources. The analysis was done using 'hyper geometric test', and all GO terms that were significant with P < 0.001 (after correcting for multiple term testing by Benjamini and Hochberg false discovery rate corrections) were selected as over-represented and under-represented. Additional data files The following additional data are included with the online version of this article: An Excel file containing a list of identified proteins in each experiment (Additional data file 1); an Excel file containing a list of the identified peptides in each experiment (Additional data file 2); an Excel file containing personal information on the individuals who provided urine (Additional data file 3); and a pdf file summarizing the results of the microscopic examination to confirm cell removal from urine (Additional data file 4). Supplementary Material Additional data file 1 An Excel file containing a list of identified proteins in each experiment. The spreadsheet consists of 15 worksheets containing respective proteins. Click here for file Additional data file 2 An Excel file containing a list of the identified peptides in each experiment. The spreadsheet consists of 14 worksheets containing respective peptides. Click here for file Additional data file 3 An Excel file containing personal information on the individuals who provided urine, showing sample number, age and gender. Click here for file Additional data file 4 A pdf file summarizing the results of the microscopic examination to confirm cell removal from urine. Click here for file
