Global Profiling of Phosphopeptides by Titania Affinity Enrichment

Global Profiling of Phosphopeptides by Titania Affinity Enrichment Jiang Wu,* Quazi Shakey, Wei Liu, Alwin Schuller, and Maximillian T. Follettie Biological Technologies, Wyeth Research, Cambridge, Massachusetts 02140 Received July 30, 2007

Protein phosphorylation is a ubiquitous post-translational modification critical to many cellular processes. Large-scale unbiased characterization of phosphorylation status remains a major technical challenge in proteomics. In the present work, we evaluate and optimize titania-based affinity enrichment for global profiling of phosphopeptides from complex biological mixtures. We demonstrate that inclusion of glutamic acid in the sample loading buffer substantially reduced nonspecific binding of nonphosphorylated peptides to the titania while retaining the high binding affinity for phosphopeptides. The reduction in nonspecific peptide binding enhanced overall phosphopeptide recovery, ranging from 22 to 85%, and led to substantial improvement in large-scale global profiling. In addition, we observed that the overall identification of phosphopeptides was significantly enhanced by neutral loss-triggered MS3 scans and respective use of multiple charge- and mass-dependent filtering criteria for MS2 and MS3 spectra. In conjunction with strong-cation exchange chromatography (SCX) for prefractionation, a total of 4002 distinct phosphopeptides were identified from SKBr3 breast cancer cells at false-positive rates of 3.7% and 5.5%, respectively, for singly and doubly phosphorylated peptides. Keywords: phosphoproteomics • glutamic acid • titania enrichment • false-positive rate • SKBr3 cells • neutral loss • database searching

Introduction Reversible phosphorylation of serine, threonine, and tyrosine regulates a variety of biological processes such as intercellular communication, cell growth, proliferation, differentiation, and apoptosis. Characterization of phosphorylation status is critical to the elucidation of signal transduction pathways and understanding mechanisms of disease and drug actions.1–4 Multiple technical approaches, predominately relying on immunoprecipitation of pTyr or pSer/pThr motifs, have enabled interrogation of subsets of the phosphoproteome. Additionally, the phosphorylation state of individual, isolated proteins has been analyzed either by radioactive labeling followed by gel electrophoresis, phosphopeptide mapping by Edman sequencing, or site-directed mutagenesis. These studies have greatly enhanced our understanding of numerous cellular processes. While a variety of MS-based proteomic techniques have been established for the identification and relative quantitation of proteins in complex biological mixtures, large-scale analysis of phosphorylation at proteome-wide scale remains a substantial challenge. These challenges are primarily due to intrinsic limitations such as low abundance and substoichiometric concentration of phosphorylation in signaling proteins, as well as suppression of sequence-specific fragmentation of phosphopeptide backbones in the mass spectrometric analysis due to neutral loss (NL) of phosphoric acid upon CID process.5–9 An essential prerequisite for global profiling is an unbiased, highly specific method for the enrichment of phosphopeptides * To whom correspondence should be addressed. Jiang Wu, Wyeth Research, 87 CambridgePark Drive, Cambridge, MA 02140. Phone: 617-6658157. Fax: 617-665-8350. E-mail: [email protected].

4684 Journal of Proteome Research 2007, 6, 4684–4689 Published on Web 10/12/2007

from complex mixtures. A wide variety of strategies have been investigated for this purpose, among which immobilized metal affinity chromatography (IMAC) shows great promises and has been widely utilized.6–8,10–16 With this approach the negatively charged phosphopeptides are isolated by their affinity to metal ions such as Fe3+ or Ga3+. However, nonphosphorylated peptides containing multiple acidic residues tend to nonspecifically bind to the metal resin as well, compromising the unbiased enrichment and detection of phosphopeptides. Blocking the acidic group by O-methyl esterification has been shown to enhance the specificity of the phosphopeptide binding.7,11 Nevertheless, incompleteness and side reactions of the derivatization often complicate MS analysis and subsequent data interpretation, leading to decreased sensitivity. As an alternative, a promising strategy using titania (TiO2) as an adsorptive has been reported.6,8,17–20 This approach is based on the selective interaction of phosphates with porous titania beads by bidentate binding at the TiO2 surface. However, studies using titania resin have shown variable degrees of success, and the selectivity of this method is greatly compromised by the detection of vast amount of nonphosphorylated peptides that are retained on the beads.6,20 Over the past several years, considerable efforts have been taken to improve selectivity. Larsen et al. demonstrated that addition of high quantity of 2,5-dihydroxybenzoic acid (DHB) in the sample loading solution minimized the nonspecific binding.20 However, improvement in the selectivity by this approach may be accompanied by the penalty of sensitivity and phosphopeptide recovery.6 In addition, this protocol is not directly applicable to LC–MS/MS analysis, because residual DHB interferes with peptide detection and contaminates source region of the 10.1021/pr070481m CCC: $37.00

 2007 American Chemical Society

Global Profiling of Phosphopeptides instrument. Improved results were obtained by utilizing phthalic acid as a replacement for the DHB.6 During the preparation of this manuscript, Sugiyama et al. reported that aliphatic hydroxyl acid significantly enhances the titania/zirconia-based phosphopeptide enrichment.21 In present work, we report a new approach to titania-based affinity enrichment and database search/validation for global profiling of phosphopeptides in complex mixture. We demonstrate that inclusion of glutamic acid in the sample loading buffer significantly reduces the adsorption of nonphosphorylated peptides to titania beads while retaining the high binding affinity for phosphopeptides. We also show that neutral losstriggered MS3 function contributes substantially to the identification of phosphopeptides when respective filtering criteria for MS2 and MS3 searches are utilized. In conjunction with SCX, a total of 4002 distinct phosphopeptides were identified from SKBr3 breast cancer cells.

Experimental Procedures Materials. Sequence-grade modified trypsin was purchased from Promega (Madison, WI). Lysyl endopeptidase (Lys-C) was from Wako Chemicals. Toptip columns and porous titania (50 µm particle) were obtained from Glygen (Columbia, MD). Polysulfethyl aspartamide column (9.4 × 200 mm) was from PolyLC (Columbia, MD). Oasis HLC cartridge and phosphopeptide mixture (NVPLpYK, HLADLpSK, VNQIGpTLSESIK, and VNQIGTLpSEpSIK) were purchased from Waters (Milford, MA). Glutamic acid and phthalic acid were from Aldrich (Milwaukee, WI). Phosphatase inhibitor cocktail was obtained from Pierce (Rockford, IL). Benzonase was from EMD Biosciences (Madison, WI). All other chemicals and reagents were of the highest grade commercially available. Cell Lysis. Human SKBr3 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin and streptomycin. The cell pellets (1 × 108 cells) were lysed in 5 mL of lysis buffer containing 25 mM Tris (pH 8), 9 M urea, and phosphatase inhibitor cocktail. The mixture was sonicated at 10 W output with 3 bursts of 30 s each. Cell debris was removed by centrifugation at 12 000g for 15 min at 10 °C. The supernatant was transferred into a new tube, and protein concentration was measured by the BCA assay. Proteolytic Digestion. The cell lysate was treated with benzonase (250 units/mL) for 3 h at room temperature. Proteins were reduced at 50 °C for 30 min in 5 mM DTT and subsequently alkylated in the dark with 12.5 mM iodoacetamide at room temperature for 30 min. The alkylation reaction was quenched by adding DTT to 5 mM. The solution was first digested with Lys-C (1:200, w/w) at room temperature for 4 h. For subsequent digestion with trypsin, the solution was diluted with 25 mM Tris buffer (pH 8.0) to a final concentration of 2 M urea. Trypsin was added (1:50, w/w), and digestion was carried out overnight at room temperature and subsequently quenched by addition of TFA to a final concentration of 0.5%. The digests were centrifuged to remove insoluble materials, and the supernatant was loaded onto an Oasis cartridge for peptide desalting. The eluted peptides were lyophilized. Strong Cation Exchange (SCX) Fractionation. Peptides were reconstituted in 800 µL of SCX buffer A (5 mM KCl/0.2% formic acid (FA)/30% CH3CN/pH 2.6). SCX fractionation was performed on a 9.4 × 200 mm column packed with polysulfethyl aspartamide material (5 µm, 200 Å) at a flow rate of 2.5 mL/ min. Three minutes of isocratic buffer A were followed by a

research articles linear gradient from 0% to 35% buffer B (350 mM KCl/0.2%FA/ 30% CH3CN/pH 2.6) over 16 min and then to 100% B within 11 min. The column was then washed with buffer C (350 mM KCl/30% CH3CN/10 mM Tris-HCl/pH 8.5) for 5 min. A total of 15 fractions was generated for mass spectrometry analysis, each of which was desalted using an Oasis HLC cartridge. After solvent removal by vacuum evaporation, the eluted peptides were lyophilized and reconstituted in 100 µL of 60% CH3CN. Enrichment of Phosphopeptides by Titania. Titania beads (50 µm) were prepacked in 200 µL pipet toptips (25–30 mg beads per toptip). Prior to loading samples, the titania tips were rinsed with 200 µL of sample loading buffer (2% TFA/65% CH3CN solution saturated with glutamic acid). Aliquots of peptide samples (15 µL, equivalent to 1.5 × 107 SKBr3 cells) were premixed with 185 µL of the sample loading buffer and were slowly loaded onto the titania tip (10 µL/min) with the aid of a vacuum manifold. The titania beads were sequentially washed with 100 µL each of the sample loading buffer, 65% acetonitrile/0.5% TFA, and 65% acetonitrile/0.1% TFA. Bound peptides were eluted with 100 µL of 300 mM NH4OH/50% CH3CN, concentrated to ∼35 µL in a vacuum concentrator, and acidified with 5 µL of 10% formic acid for LC–MS/MS analysis. As a comparison, an equivalent amount of peptide mixture was subjected to titania enrichment using phthalic acid as a selectivity enhancer in the loading buffer. The procedure was identical, except that the sample loading buffer contained 2.5%TFA/80% CH3CN saturated with phthalic acid. Peptide Analysis by LC–MS/MS. Enriched peptide samples (10 µL) were loaded onto a PicoFrit column (New Objective) packed with reversed-phase Magic C18 material (5 µm, 200 Å, 75 µm × 10 cm) and coupled with an LTQ-MS (Thermo Electron). Peptides were separated at a flow rate of 0.2 µL/min using a 90 min linear gradient ranging from 2% to 55% B (mobile phase A, 0.1% formic acid/2% CH3CN; mobile phase B, 90% CH3CN/0.1% formic acid). The instrument method consisted of a full MS scan from 400 to 1400 m/z followed by data-dependent MS2 (MS/MS) scan on the four most intense ions, a dynamic exclusion repeat count of 1, and a repeat duration of 0.5 min Additional MS3 (MS/MS/MS) was automatically triggered for precursor ions undergoing neutral-loss of either 98, 49, or 32.7 m/z upon MS2. AGC (automatic gain control) targets of 20 000 and 10 000 with 100 and 300 ms maximum ion times were used for MS and MSn scans, respectively. The minimum signals required for MS2 and MS3 were 750 and 100, respectively; and an isolation width of 2.0 m/z and normalized collision energy of 35% were used for both. All the LC–MS/MS experiments were replicated. Peptide Identification and Validation. The MS2 and MS3 raw spectra were extracted and searched separately against the human component of the NCBI nonredundant protein database (updated on October 25, 2006) using Biowork 3.3 (Thermo Electron). The search parameters take into account the static modification of carboxamidomethylation at cysteine (+57 Da), threshold of 100, combined ion filter of 200 scans, precursor ion mass range of 600–3500 Da, (1.2 Da tolerance on precursor ions and (0.7 Da tolerance on product ions, 2 missed cleavages for trypsin, and up to two differential modifications (+80 Da) for serine, threonine, and tyrosine phosphorylation in MS2 searches. A loss of 18 Da from serine and threonine was differentially searched for MS3, which is indicative of a neutral loss of phosphoric acid from precursor ion of a pSer- or pThrcontaining peptide. The raw spectra were then searched against the reversed sequence database created by the BioWork 3.3 Journal of Proteome Research • Vol. 6, No. 12, 2007 4685

research articles

Wu et al.

program using the identical parameters to estimate falsepositive rates (FPR%) of the sequence matches. The peptide matches were filtered using the combination of criteria that take into account peptide mass, charge state, scan event (MSn), number of phosphorylation sites, final score (Sf), and Xcorr.

Results Experimental Strategy. The objective of the study is to investigate and optimize experimental conditions for large-scale proteomic profiling of peptide phosphorylation status. The strategy included: (1) development of a new method for titaniabased enrichment of phosphopeptides, (2) exploration of MS3 functions to maximize phosphopeptide identification, and (3) optimization of filtering criteria for database searches. Experimentally, lysates of human SKBr3 cells were treated with benzonase to break down DNA/RNA and subsequently proteolyze sequentially by Lys-C and trypsin digestion to generate a mixture of peptides. The peptides were further separated by SCX across an entire gradient into 15 fractions. Phosphopeptides in each SCX fraction were enriched by titania column and then subjected to LC–MS/MS analysis. The MS2 and MS3 spectra from each analysis were searched, respectively, against the forward and reversed human protein sequence databases to estimate rates of false-positive matches. Multiple filtering criteria (tryptic specificity, number of phosphorylation site, Sf value, MSn, and charge state and peptide mass-dependent Xcorr score) were established to validate search results, leading to the identification of 4002 nonredundant phosphopeptides from two replicated experiments. Phosphopeptide Enrichment by Titania. As the nonspecific binding results primarily from multiple acidic residues (i.e., Glu and Asp) in peptides, we attempted to include glutamic acid in the loading/washing solution for competitive binding. Glutamic acid forms an ion-pair with TFA, and its solubility is dependent on the concentration of TFA. The saturating concentration of glutamic acid in 2% TFA/65% CH3CN is ∼0.14 M, pH 1.5. This concentration, however, very effectively inhibited the binding of nonphosphorylated peptides to the titania. Figure 1A shows the average number of both phosphopeptides and nonphosphopeptides identified from two replicated LC–MS/MS experiments across different SCX fractions. Among the 3892 peptides identified, utilizing filtering criteria described below, 3116 (80%) were attributed to phosphopeptides. As expected, the early SCX fractions (fractions 1–7) are predominantly phosphopeptides (>93%), while the selectivity of phosphopeptide enrichment for late fractions (fractions 8–15) averages ∼63%. Because most of the tryptic peptides bear 2+ and 3+ solution charges and elute between fractions 8–15 from SCX column, the enrichment procedure described here is considered highly selective. (Supporting Information lists all the phosphopeptides that were identified, together with their Sf value, Xcorr score, and site of phosphorylation.) Further attempts to optimize the enrichment conditions by fine-tuning the concentration of TFA and organic modifier (methanol vs acetonitrile) did not significantly improve the results (data not shown). For comparison, the same SCX fractions were also subjected to titania enrichment using phthalic acid-containing loading solution under the conditions described previously (Figure 1B).6 Only 25% of the peptides identified from phthalic acid approach (fractions 1–15) were phosphopeptides, and the average number (814) of phosphopeptides from two replicate experiments was also much lower in comparison with the glutamic 4686

Journal of Proteome Research • Vol. 6, No. 12, 2007

Figure 1. The number of phospho- and nonphosphopeptides identified in different SCX fractions following enrichment with titania with loading solution containing (A) glutamic acid and (B) phthalic acid. See experimental section for detail. The phosphopeptides were identified from MS2 and MS3 with distinct sequences.

acid approach. For fractions 8–15, the nonphosphorylated peptides accounted for as high as 83% of the total peptides identified. Further optimization with phthalic acid was thus not pursued. Recovery of Spike-In Phosphopeptides. Relative recovery of phosphopeptide from the enrichment procedure is influenced by the affinity and capacity of the titania beads, peptide loss during washing, and efficiency of the peptide elution, as well as general peptide loss during sample handling. To address the effect of sample complexity on phosphopeptide recovery, we spiked-in four different synthetic phosphopeptides (400 fmol each) into selected SCX fractions, followed by titania enrichment by the described protocol and LC–MS/MS analysis. As a comparison, the mixture of the four phosphopeptides in the same amount was directly loaded for the LC–MS/MS analysis. ESI intensities (peak areas) of the phosphopeptides in each analysis were measured, respectively, from extracted ion chromatograms (considering only doubly charged peaks). The recovery was calculated as the ratio of ESI intensities of the phosphopeptide undergoing the enrichment step over that directly loaded to the column. The ratios from two independent LC–MS/MS experiments were averaged to estimate peptide recovery, and the results are summarized in Table 1. The data clearly show that all four phosphopeptides were effectively bound to titania beads and detected after elution under alkaline conditions. The recovery rate varied both with respect to fraction and individual peptide, ranging from 22 to 85%. Complexity of the peptide mixture did not significantly affect the relative recovery, in that the recovery from fractions 8, 12, and 15 were generally as good as that from fractions 1 and 4. Thus, the variations in the relative recovery are not caused by

research articles

Global Profiling of Phosphopeptides

Table 2. Filtering Criteria for Phosphopeptide Identification

Table 1. Recovery of Spike-In Phosphopeptides in the Enrichment Processa Fr01

Fr04

Fr08

Fr12

Fr15

average

MSn

NVPLpYK 812.4 26.0 HLADLpSK 862.4 36.8 VNQIGpTLSESIK 1367.7 24.9 VNQIGTLpSEpSIK 1447.6 56.6

48.0 27.8 61.5 63.8

84.0 26.4 43.0 84.6

32.1 21.6 40.7 70.4

70.6 27.8 22.7 37.5

52.1 28.1 38.6 62.6

MS2

peptide

MW

no. of phosphorylation charge

2

1 a

The phosphopeptides were spiked into the selected SCX fractions and underwent the enrichment procedure. Recovery was calculated as the peak area ratio of the phosphopeptide with and without enrichment.

MS3

Figure 2. Distributions of false-positive rates (FPR%) of peptide matches with respect to different charge and mass-dependent XCorr thresholds for MS2 and MS3 spectra of singly and doubly phosphorylated peptides. The X-axis represents charge- and peptide mass-dependent Xcorr score of Sequest assignment. The number “2.0/2.8/3.4” designates Xcorr threshold values of g2.0 (2+ ions), g2.8 (3+ ions, masses