Unrestrictive identification of post-translational modifications in the urine proteome without enrichment

Background Research on the human urine proteome may lay the foundation for the discovery of relevant disease biomarkers. Post-translational modifications (PTMs) have important effects on the functions of protein biomarkers. Identifying PTMs without enrichment adds no extra steps to conventional identification procedures for urine proteomics. The only difference is that this method requires software that can conduct unrestrictive identifications of PTMs. In this study, routine urine proteomics techniques were used to identify urine proteins. Unspecified PTMs were searched by MODa and PEAKS 6 automated software, followed by a manual search to screen out in vivo PTMs by removing all in vitro PTMs and amino acid substitutions. Results There were 75 peptides with 6 in vivo PTMs that were found by both MODa and PEAKS 6. Of these, 34 peptides in 18 proteins have novel in vivo PTMs compared with the annotation information of these proteins on the Universal Protein Resource website. These new in vivo PTMs had undergone methylation, dehydration, oxidation, hydroxylation, phosphorylation, or dihydroxylation. Conclusions In this study, we identified PTMs of urine proteins without the need for enrichment. Our investigation may provide a useful reference for biomarker discovery in the future.


Background
Research on urine proteomics is important for the discovery of disease biomarkers. Post-translational modifications (PTMs) of proteins regulate many physiological functions. For example, acetylation is an important PTM in metabolism regulation; phosphorylation is an important PTM in regulating enzyme activity in cellular signaling pathways; oxidation is an important marker of cellular aging; and methylation is an important PTM in the regulation of gene expression. PTMs of proteins are subject to change, and these proteins may be potential disease biomarkers. As reported previously, in patients with diabetes, there are many advanced glycation endproduct peptides in urine [1,2]. The urine glycoproteomic makeup is altered in patients with chronic kidney diseases [3]. It has been shown that changes in osteopontin PTMs in urine are related to kidney stones and ovarian cancer [4,5]. Further, 2D-gels have demonstrated that there are different molecular masses of the same protein in the urine proteome [6]. Mass spectrometric immunoassays of urine protein phenotypes have also revealed a novel glycated end product of β-2-microglobulin [7].
Previous studies of urine protein PTMs have focused primarily on glycosylation, in which the proteins were first enriched via glycosylation and then identified as glycosylated proteins [8][9][10]. With enrichment, PTMs can be detected with high sensitivity. Research on other types of PTMs has been limited by the lack of enrichment methods [11] because each method can only identify one type of PTM. In the present study, instead of enriching for any specific PTMs, conventional urine proteomics techniques were used, and unspecified PTMs of urine proteins were identified with the MODa and PEAKS 6 software. Without enrichment, sensitivity to identify the PTMs is low. Thus far, only one previous study on urine proteomics reported the identification of phosphorylated proteins without enrichment [12].
In conjunction with recent developments in PTM research, dozens of expert algorithms have been created to perform unrestrictive searches of protein PTMs that can find almost all known PTMs and even novel PTMs. In this study, the PTM algorithms in the software packages MODa and PEAKS were used. MODa enables fast "multi-blind" unrestrictive PTM searches with a speed that is an order of magnitude faster than other existing approaches. It can also identify any number of modifications on a single peptide. In contrast to alternative approaches, MODa simultaneously uses multiple sequence tags from each MS/MS spectrum and a dynamic programming algorithm to identify modifications between sequence tags matched to a database peptide [13]. PEAKS PTM is an improved software tool for peptide identification with unspecified PTMs. The improvements in this software include a default setting whereby the software considers all PTMs included in the Universal Protein Resource (Unimod) database as variable PTMs. Moreover, several search strategies are employed to reduce the search time [14]. PEAKS PTM was included in the PEAKS 6 software, which is the only commercial software that can identify unspecified variable PTMs.

PTMs identified by MODa and PEAKS 6
In this study, real in vivo PTMs were isolated from other PTMs including in vitro PTMs and amino acid substitutions by a manual search; the in vitro PTMs are mostly created during experimental processes. In all, 39,144 spectra with 6,194 unique peptides and 1,994 proteins were identified by MODa. Among these, 7,100 spectra with 1,602 unique peptides and 734 proteins contained PTMs with sizes accepted by the MODa search regardless of the modification classification in Unimod. Within these PTMs, 433 spectra with 169 unique peptides and 85 proteins had in vivo PTMs. Furthermore, 47,857 spectra with 9,878 unique peptides and 1,606 protein groups were identified by PEAKS 6. Among these, 20,329 spectra with 3,891 unique peptides and 1,578 proteins had PTMs with sizes accepted by the PEAKS 6 search regardless of the modification classification in Unimod. Within these PTMs, 880 spectra with 254 unique peptides and 182 protein groups had in vivo PTMs. These findings are summarized in Table 1.
In this search, 15 types of in vivo PTMs were identified by MODa, and 10 types of in vivo PTMs were identified by PEAKS 6 ( Table 2).
The peptides with in vivo PTMs as found by MODa and PEAKS 6 are presented in Additional file 1 and Additional file 2. The whole urine peptides identified by MODa and PEAKS 6 are presented in Additional file 3 and Additional file 4.

PTMs identified by both MODa and PEAKS 6
The peptides with in vivo PTMs identified by both MODa and PEAKS 6 were screened out because the proteins identified as containing these peptides were somewhat different between the two software packages. Table 3 shows the peptides and corresponding proteins identified by both software packages. Table 4 shows the peptides identified by both software packages and the corresponding proteins identified by either of the two. The in vivo PTMs of the proteins identified by both software packages were compared with the PTM information in Uniprot, and some new PTMs were found.
The peptides identified by both software packages had 6 types of in vivo PTMs, which are shown in Table 5. In PEAKS 6, one peptide can belong to several protein groups. In contrast, in MODa, one peptide can only belong to one protein.
The spectra of the peptides with in vivo PTMs that were identified most reliably by both software packages are listed in Additional file 5, and only one spectrum per peptide is listed.

Discussions
This is the first study of its kind to identify posttranslational modifications in the urine proteome without preferential enrichment, using a mixture of 12 human urine samples (6 males and 6 females). The pooled sample was used to identify as many PTMs as possible in a single experiment. Because the original donors that provided the urine samples may differ in gender, age and other medical conditions, the PTMs in the urine proteomes are also likely to be different among the individuals. The PTMs in individual urine samples will be studied in the future. Moreover, the reagents from the experimental procedures including protein digestion may introduce many artifact PTMs that are not endogenous to the samples. For example, urea can cause the non-enzymatic modification of carbamylation to certain proteins. The two software packages identified both artifact PTMs and in vivo PTMs.
We manually excluded all possible artifact PTMs and reported only the unequivocal in vivo PTMs.

Conclusions
In this study, we were able to identify all urine protein PTMs without enrichment. Our investigation may provide a useful reference for biomarker discovery in the future. As the technology and algorithms for conducting proteomic screens improve, more PTMs from the urine proteome will likely be identified.

Urine collection and preparation
Pooled urine was collected from 12 healthy donors (6 males and 6 females). The donors (without medical condition and eating behavior information) were between 20-40 years old. The midstream of the urine was collected, and the samples were stored at 4°C immediately. On the same day, the urine was centrifuged at 3,000 × g for 10 min at 4°C. After removing the precipitates, the supernatant was added to three volumes of cold acetone. It was then incubated at 4°C for 2 h, followed by centrifugation at 12,000 × g for 30 min. The precipitates were collected and air-dried at room temperature. Afterwards, lysis buffer (7 M urea, 2 M thiourea, 120 mM dithiothreitol, and 40 mM Tris) was added to resuspend the pellets, which were then quantified by the Bradford method.

Protein digestion and peptide preparation
The urinary proteins were digested with trypsin (Trypsin Gold, Mass Spec Grade, Promega, Fitchburg, Wisconsin) by filter-aided sample preparation [15] using 10 kD Pall filtration devices (Pall Corporation, Port Washington,  The filtration unit was centrifuged for 20 min to collect the peptides, which were then desalted using a 1 mL OASIS HLB cartridge (Waters, Milford, MA) according to the manufacturer's instructions. The elution was dried in a SpeedVac system (Thermo Fischer Scientific) and stored at −80°C until LC/MS/MS analysis.

LC/MS/MS methods
The lyophilized peptides were dissolved in 0.1% formic acid and then separated by 2D LC/MS/MS using a strong cation exchange column (150 mm × 320 mm inner diameter, strong cation exchange resins from PolyLC Inc., Columbia, USA) and a reverse phase (RP) column (150 mm × 100 mm id, Michrom Bioresources, Auburn, California). One SCX elution method was used in which the ammonia acetate pH gradients during the separation and elution steps were 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 7, 8, and 10. For RP separation, the eluted peptides were loaded onto the column with buffer A (0.1% formic acid), and the elution gradient was 5-30% buffer B (0.1% formic acid + 99.9% ACN, flow rate: 0.5 μL/min). An LTQ-Orbitrap Velos was operated in the data-dependent acquisition mode with the XCalibur software. MS survey scan data were acquired with the Orbitrap in the 300-2,000 m/z range with the resolution set to a value of 60,000. The 20 most intense ions per survey scan were selected for CID fragmentation, and the resulting fragments were analyzed with the linear trap (LTQ). Dynamic exclusion was employed within 60 s to prevent repetitive selection of the same peptide.

File conversion
The RAW files were converted to MGF files by the MM File Conversion software.

Database
The Uniprot human proteomics database released on 3/ 21/2012.

Parameters for the MODa search
According to the README instruction in the software folder, the parameters were set as follows: PeptTolerance = 2.5: This parameter indicates the parent mass tolerance in Daltons. AutoPMCorrection = [0|1]: The default parameter value is "0", whereas "1" means that the program will automatically find the optimal parent mass for the input spectrum, regardless of the specified PeptTolerance.