|Home | About | Journals | Submit | Contact Us | Français|
The success in profiling the phosphoproteome by mass spectrometry-based proteomics has been intimately related to the availability of methods that selectively enrich for phosphopeptides. To this end, we describe a protocol that combines two sequential enrichment steps. First, strong cation exchange (SCX) chromatography separates peptides by solution charge. Phosphate groups contribute to solution charge by adding a negative charge at pH 2.7. Therefore, at that pH, phosphopeptides are expected to elute earlier than their nonphosphorylated homologs. Second, immobilized metal affinity chromatography (IMAC) takes advantage of phosphate’s affinity for metal ions such as Fe3+ to uniformly enrich for phosphopeptides from the previously collected SCX fractions. We have successfully employed the SCX/IMAC enrichment strategy in the exploration of phosphoproteomes from several systems including mouse liver and Drosophila embryos characterizing over 5,500 and 13,000 phosphorylation events, respectively. The SCX/IMAC enrichment protocol requires 2 days, and the entire procedure from cells to a phosphorylation data set can be completed in less than 10 days.
Reversible protein phosphorylation is a central regulatory mechanism of protein function in eukaryotes involved in countless cellular processes such as cell growth, cell differentiation, cell division and intercellular communication. Numerous efforts have been made aiming to understand how the phosphorylation state of a protein determines its conformation, activity and function, often on a single protein basis by mutation studies, kinase assays and 32P radiolabeling techniques1. While the information obtained from such highly focused studies is invaluable, a single phosphorylation event is often just a single part of an intricate signaling network linking many proteins and multiple phosphorylation sites. Furthermore, protein phosphorylation is believed to be a ubiquitous modification targeting most cellular proteins in one particular cellular state or another. Therefore, it is of great importance to study protein phosphorylation on a large scale to obtain a global picture of signaling events within the cell. Much of this can now be accomplished through mass spectrometry (MS)-based proteomics2.
The main difficulty in large-scale phosphorylation studies is the limitation in detecting phosphorylated species directly from complex sample mixtures due to the generally low expression levels of many regulatory phosphoproteins. Furthermore, phosphorylation often occurs at low stoichiometry, where the phosphorylated form of the proteinmay account for less than 1%of the total protein3. To solve these dynamic range issues, it is essential to perform an enrichment step that specifically selects for phosphorylated species. In this regard, the ability to survey the phosphoproteome has been intimately associated with the development of phosphopeptide-enrichment strategies.
Many enrichment strategies, normally performed at the peptide level, have proven successful. Peptide immunoprecipitation uses antibodies raised against a phosphorylation motif4 to isolate a particular subset of the phosphoproteome. IMAC enrichment utilizes metal cations such as Fe3+ (ref. 5) or Ga3+ (ref. 6) to coordinate phosphate groups by affinity, whereas titanium7,8 and zirconium9 oxides benefit from Lewis acid–base interactions. Chemical approaches to modify the phosphoester have also been proposed, using β-elimination and Michael addition10,11 or phosphoramidite chemistry12. Finally, ion (cation or anion) exchange chromatography separates peptides on the basis of solution charge. For SCX chromatography, phosphopeptides at acidic pH (~2.7) retain an additional negative charge and can be generally separated from nonphosphorylated peptides due to their reduced retention in the stationary phase13.
Additionally, one can combine separation techniques with phosphopeptide enrichment to reduce sample complexity and increase coverage. We have previously used SDS–polyacrylamide gel electrophoresis (SDS-PAGE) on protein whole cell extract and combined it with either IMAC or TiO2 at the level of peptides to identify around 2,000 and 3,000 phosphorylation sites from yeast Saccharomyces cerevisiae14 and Schizosaccharomyces pombe15, respectively. However, the method that has proven most comprehensive in mapping the phosphoproteome is SCX chromatography combined with either IMAC16–19 or TiO2 (refs. 20–22), and this method is the subject of our protocol. As phosphopeptide enrichment by TiO2 has been already described in detail in other protocols23,24, we will be focusing this one on SCX/IMAC. Nevertheless any of the two IMAC or TiO2 could be chosen in combination with SCX, and, in our hands, when working from the complexity obtained by collecting 10–15 fractions, both perform similarly. We believe superior results of SCX separation over gels are due to SCX chromatography grouping peptides on total charge and local charge distribution. Thus peptide mixtures to be enriched by IMAC or TiO2 are more similar in acidic properties, and competition occurs between species that are more alike, not favoring any phosphopeptide (e.g., multiple phosphorylated peptides) in particular.
The SCX/IMAC approach has allowed us to identify more than 5,500 phosphorylation sites in mouse liver18 and over 13,000 in fly embryos19 with relatively little effort and in less than 24 h of MS analysis time. The requirements we had in mind when developing this workflow were simplicity, robustness and a reduction in the MS analysis time to manageable levels, with a future view of automation and increased throughput. The protocol described here maintains the general outline used in previous studies, with a few added improvements toward selectivity and throughput of the IMAC enrichment.
Although the protocol can be adapted to special needs and available resources, in our experience, the following points reflect the setup providing best results:
Yeast cultures should be grown under the confluence and media conditions desired for each particular experiment. Most wild-type S. cerevisiae strains yield around 100mg of protein per liter of culture at OD (600 nm)= 1.0.
Mammalian cells should be grown in the media required for each particular cell type. Most experiments are performed at 80% cell confluence, and protein amounts on the order of 1–5 mg are obtained from one 15-cm-diameter dish, depending on the cell type.
8 M urea, 75 mM NaCl, 50 mM Tris, pH 8.2, two tablets of protease inhibitors cocktail (complete mini, Roche) per 10 ml of lysis buffer, 50 mM NaF, 50 mM β-glycerophosphate, 1 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 1 mM PMSF. It is recommended to prepare the buffer fresh before use. Concentrated stocks of phosphatase inhibitors (NaF, β-glycerophosphate, sodium orthovanadate and sodium pyrophosphate) can be stored at room temperature (18–24 °C). The storage temperature for 100 mM stocks of the protease inhibitor PMSF in DMSO or MeOH is −20 °C. PMSF should be added to the lysis buffer immediately before use, as it is unstable in aqueous solutions.
8 M urea, 75 mM NaCl, 50 mM Tris, pH 8.2, one tablet of protease inhibitors cocktail (complete mini, Roche) per 10 ml of lysis buffer, 1 mM NaF, 1 mM β-glycerophosphate, 1 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 1 mM PMSF. Use same recommendations as for the yeast lysis buffer.
(A) 7 mM KH2PO4, pH 2.65, 30% ACN (vol/vol); (B) 7 mM KH2PO4, 350 mM KCl, pH 2.65, 30% ACN (vol/vol); (C) 50 mM K2HPO4, 500 mM NaCl, pH 7.5. As organic solvents affect the pH reading, the pH adjustments for buffers A and B should be performed before the addition of ACN.
Binding buffer: 40% ACN (vol/vol), 25 mM FA in H2O Elution buffer A: 50 mM K2HPO4, adjust to pH 10 with NH4OH Elution buffer B: 500 mM K2HPO4, pH 7.
Binding buffer: 0.1% TFA (vol/vol) in H2O Elution buffer: 50% ACN (vol/vol), 0.5% HAcO (vol/vol) in H2O.
Binding buffer: 1% FA (vol/vol) in H2O Elution buffer: 50% ACN (vol/vol), 0.5% HAcO (vol/vol) in H2O.
(A) 3% ACN (vol/vol), 0.125% FA (vol/vol) in H2O; (B) 0.125% FA (vol/vol) in ACN.
Valco valve for manual injection set with a 500-µl loop, quaternary pump Agilent 1100 equipped with degasser, operating at 3 ml min−1, photodiode array (PDA) detector from ThermoFisher set up to read at fixed 220- and 280-nm wavelengths.
Semipreparative SCX column: polySULFOETHYL A, 9.4-mm inner diameter × 200 mm length, 5-µm particle size, 200 Å pore size (PolyLC).
Famos autosampler (Dionex) with 10-µl loop and2.4-µl injection needle, Agilent 1100 pumps operating between 80 and 300 µl min−1 and setup with a double-split system to provide an in-column flow rate of 0.5–1 µl min−1 on a microcapillary column 125 µm × 18 cm column with hand-pulled tip packed with C18 reverse-phase material (MagicC18AQ, MichromBioResources) and LTQOrbitrap mass spectrometer equipped with a nanospray source. The method used for acquisition consists of an 80-min method with 9-min loading time, 57-min gradient from 4% to 26% B (see REAGENT SETUP for solvent composition) and 60 min of MS data collection. For each scan cycle, one full MS scan acquired in the Orbitrap mass analyzer at 60,000 resolution, 1 × 106 automatic gain control (AGC) target and 1,000 ms maximum ion accumulation time is followed by 10 MS/MS scans on the ten most intense ions in the linear ion trap (LTQ) at 3,000 signal threshold, 5,000 AGC target and 120 ms maximum accumulation time, 2.0 Da isolation width and 30 ms activation at 29% normalized collision energy. Dynamic exclusion is enabled to exclude from fragmentation ions that had been already selected for MS/MS in the previous 35 s. Ions with a charge of +1 or unassigned are also excluded. All scans (MS and MS/MS) are collected in centroid mode. To ensure proper performance of the system, we use a complex mixture of yeast peptides corresponding to 1/10 of the yeast proteome and obtained by SDS-PAGE fractionation of cell protein extract and trypsin digestion of gel regions (see TROUBLESHOOTING).
We recommend checking the sample at various stages of the protocol to ensure that every step is performed successfully. Here is a list of suggested checkpoints:
Troubleshooting advice can be found in Table 2.
To exemplify the performance of the SCX/IMAC approach for large-scale phosphorylation studies, we applied this method to budding yeast S. cerevisiae. Figure 1a shows a schematic of the described workflow. Yeast were grown to mid-log phase, harvested and lysed. Fifteen milligrams of protein were reduced, alkylated and trypsin-digested. Desalted peptides were subjected to SCX chromatography and 12 fractions were collected, desalted, enriched by IMAC and desalted in one step (Fig. 1b), and analyzed by LC-MS/MS on an LTQ Orbitrap. All MS/MS spectra collected were searched with the Sequest algorithm27 against a composite yeast database with forward and reversed sequences28.
Separation of peptides at pH 2.65 by SCX chromatography is shown in Figure 2. The chromatogram of UV absorbance detection at 220 nm is displayed in Figure 2a, and the solution charge separation across the fractions of all the phosphopeptides detected after IMAC enrichment is shown in Figure 2b. The aforementioned buffer compositions and the gradient program we designed allow us to obtain increased separation between peptides with charges +1 and +2.
To show the phosphopeptide enrichment obtained in each step, the fraction of unique phosphopeptides over the total number of unique peptides identified within each fraction is plotted in Figure 3. Figure 3a shows the result after SCX chromatography only, and Figure 3b shows the result after both SCX chromatography and IMAC. Although SCX chromatography alone provides some phosphopeptide enrichment in the three earliest fractions13, coupling SCX chromatography with IMAC increases phosphopeptide enrichment to >75% for most fractions.
We have observed that running technical duplicates (i.e., analyzing the same sample in the mass spectrometer twice) of complex peptide or phosphopeptide mixtures significantly increases the number of unique identifications19,25, as the overlap between them is 70–75% (Fig. 4a). On the other hand, the number of phosphopeptides identified repeatedly in two consecutive fractions represents only 20% of the total phosphopeptides from either of the two fractions compared, revealing good resolution from the SCX chromatography (Fig. 4b).
Finally, discrimination between right and wrong answers using Sequest XCorr score filtering alone would set the threshold at a very high value if a 1% false-discovery identification rate (FDR) is desired, causing a significant number of right but low scoring answers to be thrown away (Fig. 5a; red, reverse/wrong answers; blue, forward minus reverse/right answers). Measuring accurate precursor masses and performing searches in a mass tolerance window ~10 times bigger than the mass deviations measured (Fig. 5b) allows for an additional and powerful filtering criterion: mass error tolerance in a p.p.m. window25,29,30 that rescues many low XCorr scoring answers, while maintaining an ~1% FDR (Fig. 5c).
When generating large data sets, special attention should be paid to estimating, minimizing and reporting error rates of peptide identifications, to minimize erroneous annotations in phosphorylation databases. To this end, we introduced the target-decoy database-searching strategy28, where searches are performed against a single database that contains all proteins from the organism of interest, first in the forward and then in the reverse direction, and an FDR is estimated on the basis of the number of hits derived from flagged reverse proteins. Importantly, reverse matches can be used to guide threshold selection for mass deviation, XCorr, solution charge state and dCn′ parameters. Most laboratories performing large-scale phosphorylation studies have now adopted this simple and powerful approach for error rate determination20,31,32.
In addition to the issue of peptide false-discovery rates, analysis of phosphorylation (or other posttranslational modifications) by MS may introduce ambiguity associated with the precise position of the modification site. When more than one potential acceptor site exists in a peptide, some method of post-analysis verification should be used. The Ascore algorithm is a probability-based metric for assessing localization ambiguity29.
Besides these automated approaches for error-rate assessment and site localization, further validation for a subset of phosphopeptides might be desired, especially before follow-up biological experiments. Peptide synthesis constitutes the gold standard, where the MS/MS spectrum from the synthetic peptide is compared with the one acquired from the biological sample33. In addition, reanalysis of the sample using high-resolution MS/MS, different fragmentation techniques such as electron capture dissociation (ECD)34 and electron transfer dissociation (ETD)35 or fragmentation schemes such as neutral loss data-dependent MS3 (ref. 13) and multistage activation36 frequently bring additional confirmation for the phosphopeptide entity. Finally, computer-assisted manual inspection of MS/MS, looking for phosphopeptide diagnostic ions such as fragments derived from neutral loss of phosphoric acid, which often verify the presence of a phosphoserine or phosphothreonine in the sequence, and proline-driven fragmentation, has been regularly practiced13,16 with the drawback of being highly subjective, and correct discrimination is very much dependent on user expertise.
To perform database searches and data analysis, in our studies, RAW files were converted to the mzXML file format and imported into a MySQL relational database. MS/MS spectra were searched against a target-decoy28 S. cerevisiae ORFs database using the Sequest algorithm (version 27, revision 12), with 50 p.p.m. precursor mass tolerance, tryptic enzyme specificity with two missed cleavages allowed and static modification of cysteines (+57.02146, carboxamidomethylation). Dynamic modifications were 79.96633 Da on Ser, Thr and Tyr (phosphorylation) and 15.99491 Da on Met (oxidation). A maximum of four modifications of any one type and five total modifications were allowed per peptide. Sequest XCorr and dCn′18 score cutoffs were empirically determined for the entire data set, and mass deviation (in p.p.m.) and peptide solution charge filters were determined for each sample individually, using decoy matches as a guide28 and aiming to maximize the number of peptide spectral matches, although maintaining an estimated FDR of ≤ 1%. Identified phosphopeptides passing our filtering criteria were submitted to the Ascore algorithm29 for precise site localization.
We thank Andrew Alpert from PolyLC for kindly providing columns for SCX chromatography and Manuel Rodriguez-Falcon for the initial tests of the combined IMAC-desalting procedure. We are also grateful to Joshua T. Wilson-Grady for constructive comments on the manuscript. This work was supported by NIH grant HG3456 to S.P.G.
Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/