Abstract
Differential proteolytic 18O labeling is a cost-effective but not commonly used method in the field of quantitative proteomics based on mass spectrometry (MS). In most cases, peptide identification is performed at the MS/MS level followed by peptide quantification at the MS level. In this study, identification and quantification of 18O-labeled peptides was performed in a single step at the MS/MS level using the MASCOT 2.2 search engine, and the instrumental conditions for acquisition of ultra performance liquid chromatography electrospray MS/MS (UPLC-ESI-MS/MS) data were adapted accordingly. Using analysis of standard peptide and protein mixtures prepared by differential 16O/18O labeling, under these conditions automated MS/MS data acquisition and evaluation delivered correct data. Linearity and reproducibility of this approach indicated excellent performance. In addition, the method was applied to relative quantification of protein phosphorylation in mouse brain following treatment with ionizing radiation. The analysis led to automated quantification of 342 proteins and 174 phosphorylation sites, 24 of which were up- or down-regulated by a factor of 2 or more. The majority of these phosphorylation sites were found to be located in target sequences of known protein kinases, showing the activation of kinase-regulated signaling cascades by irradiation.
Enzymatic digestion of proteins in 18O-enriched water is an elegant, versatile and cost-effective labeling method in analytical proteomics. It has been applied for peptide sequencing (1), de novo peptide sequencing (2-4), relative peptide quantification (5-7), relative quantification of subproteomes (8-12), and relative quantification of protein phosphorylation (13). Other applications of 16O/18O labelling are the recognition of glycosylation (14-16), disulfide-linked peptides (16), isoaspartate formation (17, 18), succinimide formation (19), and deamidation (20). Methods and applications of proteolytic 18O labeling of peptides have been reviewed elsewhere (21, 22).
In the field of quantitative analytical proteomics, proteolytic 18O labeling has experienced slower methodological development compared to other chemical or metabolic labeling methods (23-27). Putative downsides of the 18O labeling method are the time required, variability of 18O incorporation, the possibility of back exchange of the label, a relatively small mass shift, and the lack of suitable data evaluation software.
The time required for 18O labeling is identical to that for the digestion procedure. Thus, 18O labeling can be achieved in 15 to 30 min by acceleration of enzymatic digestion, e.g. using immobilized trypsin (28), ultrasonic radiation or elevated temperature (29).
Several attempts have been undertaken to overcome the variable label introduction of one or two atoms of 18O, by either trying to achieve the incorporation of a single 18O atom (30), or complete labeling by two atoms of 18O (16, 31). The defined introduction of one or two 18O atoms simplifies the quantitative evaluation of a mixed (labeled + unlabeled) isotopic pattern, since in this case, the isotopic pattern consists of only two components (16O/18O or 16O/18O2). In cases of variable incorporation, the deconvolution process must consider three components (16O + 18O + 18O2), where the sum of the 18O and the 18O2 isotopomer represents the amount of the labeled species. This sum is independent of the relative composition since an interconversion of the 18O species into the 18O2 species does not change the sum of both species.
Back exchange of the 18O label can occur by an acid-catalyzed (at pH<2) or protease-catalyzed mechanism (32-34). In practice, 18O back exchange can be avoided effectively by denaturing or removal of the protease used for digestion, e.g. by using immobilized trypsin (35) and by avoiding pH values <2 subsequent to the labeling step.
Several software programs for the deconvolution of molecular ion isotopic patterns of 18O labeled peptides have been developed (e.g. (36-38)) but to our knowledge, most of these programs are currently not available as open access files or require expert software knowledge. As an alternative to the deconvolution of molecular ion patterns, the 18O content of peptides can be analyzed at the fragment ion level. By concept, evaluation at the tandem mass spectrometry (MS)/MS level should provide a better accuracy of the quantitative data compared to the evaluation of survey spectra, due to reduced background and an increased specificity (39). Recently, software for quantitative evaluation of labeling experiments with 13C- and 15N-labeled peptides at the MS/MS level was introduced (39). A version of this software that is applicable to 18O labeling was incorporated into the search engine MASCOT (version 2.2). Here, we demonstrate the performance of this software using mixtures of standard peptides/proteins. Finally, we present the first application of differential 18O labeling to one-step identification and relative quantification of site-specific protein phosphorylation in a complex protein sample using this software.
Materials and Methods
Chemicals. All chemicals were from Sigma (Deisenhofen, Germany) unless indicated otherwise. Trypsin was from Roche (Mannheim, Germany). All solvents and acids were from Biosolve (Valkenswaard, the Netherlands) in ULC grade quality suitable for ultra performance liquid chromatography. 18O-Enriched water (≥98% purity) was from Rotem (Leipzig, Germany).
Standard mixture preparation. Synthetic peptides were kindly produced in-house using Fmoc chemistry in the Central Peptide Synthesis Unit of the DKFZ by Dr. R. Pipkorn. Three differently 16O/18O-labeled mixtures of the peptide DLESQLAQSR were prepared. Fractions of the synthetic peptide were incubated either in 16O or 18O water, respectively, in the presence of trypsin for 24 h. Trypsin was removed and the samples were desalted using ZipTips (Millipore, Billerica, MA, USA). Subsequently 16O- and 18O-labeled samples were combined and the resulting ratios of nonlabeled/labeled compounds controlled by nano electrospray (ESI)-MS revealing the ratios present in the mixtures (9:1, 1:1, 1:9). Differentially 16O/18O-labeled protein digest mixtures were prepared from ovalbumin. For this purpose, 100 μg of ovalbumin were dissolved in 0.1M NH4HCO3, denatured using 4 M urea, reduced and alkylated using dithiothreitol (DTT) and iodoacetamide (IAA) and then digested in solution in 16O and 18O water, respectively, using trypsin. After digestion, the labeled and unlabeled fractions were mixed in different ratios.
Complex protein sample preparation and SCX/WAX peptide fractionation. Cerebellum tissues were prepared from untreated and irradiated (10 Gy, excised 1 h after irradiation) 10-day-old Balb-C mice. Proteins were extracted in the presence of protease and phosphatase inhibitors (both from Roche). A quantity of 3 mg protein of each fraction was precipitated using cold acetone and the precipitate was washed once using acetone. After removing the acetone, the precipitate was dried using a Speed Vac and resuspended using 0.1 M NH4HCO3 in 18O or 16O water, respectively. In the next step, in-solution digestion was carried out as described above. Resulting peptides were desalted using an XBridge BEH 130 PREP C18 250 mm × 10 mm reversed-phase column with a particle diameter of 5 μm (Waters, Milford, MA, USA). The volume of the desalted peptide fractions was reduced using a Speed Vac. Peptides were then loaded onto an SCX/WAX (1:2) 200 mm × 4.4 mm column with a particle diameter of 5 μm and 300 Å pore size (PolyLC Inc., Columbia, MD, USA) and a linear gradient from 100% A (20% acetonitrile, 0.1% formic acid), 0% B (20% acetonitrile, 0.1% formic acid, 0.5 M ammonium acetate) to 0% A, 100% B in 40 min was applied. During elution, 20 fractions were collected, the volume was reduced using a Speed Vac and the samples were desalted using the preparative reversed-phase column. Desalted fractions were again reduced in volume using a Speed Vac. 5% to 50% of each fraction was submitted to phosphopeptide enrichment by Ga3+ immobilized metal ion affinity chromatography (IMAC) (Phosphopeptide Isolation Kit, Pierce, Rockford, IL, USA). Flow-through fractions and phosphopeptide fractions were reduced in volume and citrate was added to the phosphopeptide fractions to a final concentration of 50 mM (40).
UPLC-MS/MS analyses. All LC-MS/MS analyses were carried out using a nanoAcquity UPLC system (Waters). The column used was a 150 mm × 75 μm C18 BEH column packed with 1.7 μm particles with a pore size of 130 Å. Samples were loaded directly onto the analytical column without using a trap column. After sample loading followed by 24 min of washing with 1% solvent A (water with 0.1% formic acid), a linear gradient of 60 min length was applied from 99% A, 1% B (acetonitrile with 0.1% formic acid) to 70% A and 30% B. The column temperature was 35°C and the flow was 400 nl/min. The column outlet was connected to a pico tip sprayer (Waters Micromass, Manchester, UK) using PicoTips (New Objective, Woburn, MA, USA) mounted on a Quadrupole Time-of-Flight mass spectrometer (QTOF-2, Waters Micromass, Manchester, UK). The capillary voltage applied was 2400 V. The mass spectrometer was operated in data-directed acquisition (DDA) mode with recording of 1 spectrum per second. Following one survey scan, the two most abundant signals in each scan were selected for fragmentation and up to two MS/MS scans per precursor ion were acquired. For the low and high mass resolution, a value of 2 was selected, resulting in a transmission window of 12 m/z units extending from about −3 m/z to +9 m/z relative to the selected set mass for MS/MS.
Data processing. Using MassLynx 4.1 (Waters) the MS/MS raw data were transformed into peak lists (.pkl files). For processing, no smoothing or background subtraction was applied. The individual .pkl files obtained in the analysis of the mouse brain samples were merged into a single file using the program merge.pl provided by matrixscience (www.matrixscience.com). All .pkl files were searched against the MSDB database (http://csc-fserve.hh.med.ic.ac.uk/msdb.html) using MASCOT 2.2 installed on an in-house server. As search parameters, the option ‘quantification by 18O-corrected multiplex’ was selected in combination with a precursor mass tolerance window of 6 Da and an MS/MS mass tolerance window of 0.3 Da. The instrument specification was set as ESI-QUAD-TOF. As digestion parameters, trypsin in combination with a maximum of 3 miscleavages were selected. Carbamidomethyl cysteine was selected as fixed and methionine oxidation as reversible modification. For phosphopeptide analysis, phosphorylation at serine, threonine or tyrosine was set as additional variable modification. Average values, standard deviations and correlation coefficients were calculated using Microsoft Excel.
Results
To obtain reliable and correct quantitative results, at first binary and complex peptide standard mixtures with known ratios of 16O/18O content at the C-terminus were used to optimize the LC-MS/MS settings and the search engine parameters (see Material and Methods).
16O/18O Labeling and MS/MS spectra acquisition. For correct quantification at the MS/MS level, the broad isotopic envelopes of proteolytically 18O-labeled peptides must be completely included in the precursor ion transmission window. Only under these conditions do the isotopic patterns of the y ions provide a 16O/18O ratio which is identical to that of the molecular ion. In the DDA mode, the two extreme cases that can occur are that either the most abundant peak of the nonlabeled species or that of the doubly 18O-labeled species is selected as set mass for the registration of an MS/MS spectrum. Using the 1:1 mixture of DLESQLAQSR (C16O16OH) and DLESQLAQSR (C18O18OH) both situations were studied by nanoESI-MS and manual set mass selection. Figure 1a and b demonstrate that under standard conditions, this situation is not met, since the precursor ion isolation window extends from −0.5 m/z to 2.5 m/z relative to the set mass. This results in only partial transmission of the molecular ion group, as shown in Figure 1c and d. A shift of the left border of the precursor ion window to −2 m/z would result in complete transmission for both set mass alternatives (Figure 1e, f) for molecular ions of charge state 2 or higher. To meet these conditions, the width of the precursor ion transmission window was enlarged to approximately 12 m/z units. This window extends from about −3 m/z to +9 m/z relative to the set mass.
Using this extended precursor ion transmission window, the complete isotopic pattern, including the unlabeled as well as the singly and doubly labeled form of the peptide, is transmitted. The resulting MS/MS spectra show unlabeled b ions and 18O-labeled y ions. As an example, the survey spectra and the MS/MS spectra showing the 18O-labeled y2 ion of 3 different mixtures of the synthetic peptides DLESQLAQSR (C16O16OH) and DLESQLAQSR (C18O18OH) are shown in Figure 2. This Figure proves that quantification can be performed on the MS/MS level, since the isotopic patterns of connected molecular ions and fragment ions provide identical quantitative information. For the peptide mixtures investigated, all y ions from y2 to y7 can be used for quantification, whereas some deviations were observed for the y1 and y8 (=ymax-2) ions (see Supplementary Material, Figures S1 to S3). We feel that the somewhat variable 18O content of these ions from the peptide ends might indicate a partial oxygen exchange during their formation. Exclusion of the corresponding data (y1 ion and ymax-2 ion) from the quantitative evaluation will therefore improve the accuracy of the quantitative data. However, this point was not followed in more detail in this study.
For relative quantification, there are several advantages of the MS/MS mode compared to the MS mode: i) the isotopic overlap is reduced, in particular for small y ions; ii) MS/MS spectra exhibit in general a better signal to noise ratio due to the strongly reduced background (39); iii) the quantitative results can be based on a set of y ion patterns, allowing the recognition and exclusion of interferences; iv) the MS/MS mode by principle is less prone to interferences due to its increased specificity.
For the particular application of a QTOF instrument with its limited dynamic range, the MS/MS mode also provides a somewhat extended linear dynamic range for high precursor ion signals. Using the 1:1 mixture of DLESQLAQSR (C16O16OH) and DLESQLAQSR (C18O18OH), different sample concentrations (50 fmol/μl to 50 pmol/μl), representing a dynamic range of 3 orders of magnitude, were quantified by nanoESI-MS and -MS/MS, respectively. Using quantification from the MS/MS spectra, a correct quantification over the whole range of concentrations with a maximal error of about 10% was possible. Quantification via the MS survey spectra resulted in a smaller dynamic range of only one order of magnitude (see Supplementary Material, Figure S4).
Relative quantification of standard protein mixtures by 16O/18O labeling. For evaluation of the complete workflow consisting of an automated UPLC-MS/MS analysis and subsequent quantitative data evaluation using MASCOT 2.2, a set of ovalbumin digest mixtures was prepared. Equal amounts of ovalbumin were digested in solution in either 16O or 18O water as described in Material and Methods. Measured aliquots of these samples were then mixed so that a set of samples with known ratios of 16O to 18O between 1:9 and 9:1 was obtained. These samples were analyzed by UPLC-MS/MS and the data files were transformed to peak lists and subsequently quantified using MASCOT 2.2 via its implemented 18O quantification tool. All y ions with a possible interference by other sequence ions were automatically excluded from quantification. The minimal number of y ions required for quantification was 4 and the intensity cut-off was 0.1 (all ions with a relative abundance <10% were excluded from quantification). Using these settings, 26 tryptic ovalbumin peptides were identified on average in a single experiment, of which an average of 23 fulfilled the criteria for inclusion in the quantification. All experiments were carried out in triplicate and the quantitative data are summarized in Figure 3 (with the standard deviation (S.D.) and correlation coefficient (R2)).
For all 16O/18O mixtures of ovalbumin peptides analyzed, very good agreement between the theoretical and experimental ratios, as indicated by the good R2 value for the linear regression of the data set, was achieved, as shown in Figure 3. The small S.D. values demonstrate the good reproducibility of the method. This indicates successful optimization of the instrumental parameters for data acquisition, reproducible data acquisition by the UPLC-MS/MS system and correct selection of the parameters used for the automated evaluation by MASCOT.
16O/18O Labeling and MS/MS analysis in a quantitative proteomic study. To evaluate the combination of 16O/18O labeling and MS/MS-based relative quantification for proteomic studies, we used this strategy to analyze the dynamics of the phosphoproteome in mouse cerebellum following treatment with ionizing radiation. Ten-day-old mice were treated with 10 Gy of ionizing radiation, and 1 h later cerebella were excised and proteins were extracted and digested by trypsin in either 18O or 16O water. Peptides were subsequently fractionated by SCX/WAX chromatography and IMAC as described in Materials and Methods. The resulting peptide and phosphopeptide fractions were then analyzed individually by UPLC-MS/MS using the methods described above. The MS/MS spectra of the individual fractions were transformed into peak list files and then merged into a single file. This file was searched against the MSDB database by MASCOT 2.2, which resulted in the simultaneous identification and quantification of a large set of peptides and phosphopeptides. In total, 342 proteins were quantified with an average ratio of 0.9 of irradiated to control sample. An overview of the results obtained for protein quantification is shown in Figure 4, and a detailed list of all proteins quantified is provided by Table S1 in the supplementary material.
Discussion
The data in Figure 4 imply that the ratios of 8 proteins (irradiated over control) were outside the ±3 S.D. values set as cut-off. Manual examination indicated that 4 of these proteins contained regulated phosphopeptides which were taken for quantification as well, which means that the amount of these protein was probably not regulated. The remaining 4 proteins were: hemoglobin beta major chain (Figure 4, spot #1), heat-shock protein 70 (Hsp70) (#2), Ran-specific GTPase-activating protein (RANBP1) (#5) and protein kinase C and casein kinase substrate in neurons protein 1 (PACSIN1) (#6). Due to the role of hemoglobin as major constituent of red blood cells (41), the up-regulation may indicate increased perfusion of the cerebellum following irradiation. The early up-regulation of Hsp70 (1 h) after various types of stress including DNA damage has been reported (42, 43). RANBP1 regulates the activity of the GTPase Ran, which is responsible for the transport of molecules from the nucleus to the cytoplasm, including transport of p53 which plays a crucial role in apoptosis, one of the possible reactions of cells to DNA-damage (44-46). The role of PACSIN1 in the DNA damage response is currently unclear.
To further investigate the DNA damage response at the level of posttranslational modifications, phosphorylation site analysis was performed. In total, 174 phosphopeptides were identified and simultaneously quantified. The summarized phosphopeptide data are shown in Figure 5 and a list of all phosphopeptides is presented in Table S2 in the supplementary material.
As implied by the data shown in Figure 5, the majority of the identified phosphopeptides (86%) are not regulated, while 14% of the phosphopeptides showed more than 2-fold alteration of their ratio (irradiated sample over control) between the two samples. Based on these criteria, 24 phosphorylation sites altered in their concentration were identified. The proteins carrying these sites, their sequence coverage, phosphorylation sites identified, and the extent of their regulation are summarized in Table I. For regulated phosphorylation sites, the sequence motifs ±10 residues relative to the phosphorylated residue (see Table S3) were matched to known kinase consensus sequences. Seventeen of them could be potential targets of one of the following protein kinases: protein kinase B (Akt) ([RK]X[RK]XX[ST] (47)), cAMP-dependent protein kinase (PKA) (R[X]1-2-[ST] (48, 49)), casein kinase II (CK2) ([ST]AAAAA (48)), mitogen-activated protein kinase kinase (MAPKK)/extracellular signal-regulated kinase 2 (ERK2) (both PX[S/T]P (48)), and cyclin-dependent kinase 5 (Cdk5) (X[S/T]PXK (50)) where X is any residue, A is an acidic residue and [S/T] is the phosphorylated serine/threonine residue.
The up-regulation of ERK2 phosphorylation indicates activation of the ERK1/2 MAP kinase cascade, which has been associated with DNA damage (66). ERK signalling has been specifically associated with apoptotic response to DNA damage (67). One of the substrates identified for ERK2 in that study was RRAS2 (68). RRAS2 is a Ras homolog, however, in contrast to Ras, it does not activate the ERK pathway but rather the c-Jun N-terminal kinase (JNK) and p38 MAPK pathways (69). This links ERK signalling, which is known to be activated by extracellular signals such as growth factors, to p38 and JNK signalling which is stress-related, combining 3 of the major MAPK signalling pathways (70). Phosphorylation of the other two substrates of ERK2, MARCKS and MAP1A was found to be down-regulated. This may result from further modification of these proteins upon ERK phosphorylation.
The identification of 3 up-regulated Cdk5 substrates indicates an elevated activity of this kinase as well. Cdk5 is activated by ERK1/2 (71) and interleukin 6 (IL-6), the secretion of which is initiated by thymosin 4-beta (72, 73). A thymosin 4-beta phosphorylation site was found to be up-regulated, which may result in enhanced IL-6 secretion. The activation of Cdk5 following DNA damage has been shown to result in phosphorylation of Huntingtin regulating its toxicity in neurons (74). Cdk5 also phosphorylates p53, regulating its activity and inducing neuronal cell death (75). The targets of Cdk5, which carry regulated phosphorylation sites in this study, have not yet been associated with DNA damage. Three proteins containing up-regulated phosphorylation sites (Crmp2, Map2, Mapt) are microtubule-associated proteins. Map2 and Mapt are responsible for microtubule stabilization (63), and Crmp2 is member of the Unc-33-like phospho-(Ulip) protein family whose members are associated with development and differentiation of neurons (62). The association of Cdk5 with microtubules and Cdk5-dependent phosphorylation of microtubule-associated proteins has been shown before and it was postulated that this may regulate the axonal transport of molecules in the cell (76). The activation of Cdk5 upon DNA damage may therefore be due to a elevated level of protein transport in the cell, either to start repair processes of the damaged DNA or to initiate apoptosis.
Another kinase found to be involved in signalling upon DNA damage is PKA. PKA has already been associated with DNA damage checkpoint pathways (77). It was also shown, that Cdk5 and PKA phosphorylate similar proteins while phosphorylation of a particular site by PKA influences phosphorylation of the same proteins by Cdk5 on another site (78). Two phosphorylation sites (on MAP2 and Crmp2) resemble the consensus sequence of both Cdk5 and PKA. These proteins may be phosphorylated by both kinases resulting in a crosstalk between their signalling pathways. Three other phosphorylation sites were identified to be PKA-dependent. One of them is S305 in Doublecortin, a microtubule-associated protein. Doublecortin, involved in the leading processes of migrating neurons, is a known target of PKA (S47) and Cdk5 (S297). These phosphorylations result in reduced affinity to microtubules (79, 80). The second site is located at P140CAP, which is involved in epidermal growth factor (EGF) and integrin signalling, which plays a role in actin cytoskeleton organisation and regulates Csk and Src kinase activity (58, 81). The third site was found in Rims1, which is a protein of the active zone of neurotransmitter release at synapses (61). It was shown that PKA activates the ERK signalling pathway via Ras/B-Raf and the p38 pathway via PTPs (70). Activation of ERK and p38MAPK again results in activation of the stress response of the cell.
CK2 is a constitutively active kinase with hundreds of known targets (82). For instance, it phosphorylates the checkpoint kinase Rad53 upon DNA damage and phosphorylation of MDC1 by CK2 is essential for proper accumulation of damage-response proteins at DNA double-strand breaks (83-85). CK2 is stabilized by interaction with ATR after DNA damage and phosphorylates the phosphatase and tensin homolog (PTEN) which is involved in cell cycle re-entry following DNA damage (86). Cell cycle re-entry of post-mitotic cells (e.g. neuronal cells) was reported to be involved in DNA damage-induced apoptosis (46).
Phosphorylation of Dnajc5 by Akt was shown in vitro and was linked to the late stages of exocytosis (87). The phosphorylation and activation of Akt following DNA damage was shown as well (88). In another study, Akt activation was linked to phosporylation by DNA-PK upon DNA damage, which resulted in stabilization of p53 by Akt (89). Interestingly, in this study, the phosphorylation of Dnajc5 was down-regulated by irradiation. This may be due to activation of a phosphatase, additional unexpected modifications of this region of Dnajc5 or degradation following phosphorylation.
Conclusion
Quantification of proteolytic 18O-labeled peptides using MS/MS information is a powerful and easy tool for the quantification of proteins. By the use of MASCOT 2.2, quantification is convenient and platform independent, which allows the use of different types of mass spectrometers without the need of data transformation or the use of algorithms that suit different data formats. Using a QTOF instrument in the MS/MS mede, the linear range of the applicable sample amount is superior compared to survey-based quantification. Identification and quantification are carried out simultaneously. Using this technology, a detailed insight into the regulation of protein levels and phosphorylation sites can be conveniently achieved.
Acknowledgements
This work was supported by the Cooperation Program in Cancer Research of the German Cancer Research Center (DKFZ) and the Israeli Ministry of Science and Technology (MOST).
- Received June 2, 2009.
- Revision received November 11, 2009.
- Accepted November 12, 2009.
- Copyright© 2009 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved