RAPID COMMUNICATIONS IN MASS SPECTROMETRY Rapid Commun. Mass Spectrom.

2010; 24: 219231 Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/rcm.4377

Enrichment and analysis of phosphopeptides under different experimental conditions using titanium dioxide afnity chromatography and mass spectrometry
Uma K. Aryal*,y and Andrew R. S. Rossz
National Research Council, Plant Biotechnology Institute, 110 Gymnasium Place, Saskatoon, Saskatchewan, Canada S7N 0W9 Received 19 June 2009; Revised 7 November 2009; Accepted 9 November 2009

Titanium dioxide metal oxide afnity chromatography (TiO2-MOAC) is widely regarded as being more selective than immobilized metal-ion afnity chromatography (IMAC) for phosphopeptide enrichment. However, the widespread application of TiO2-MOAC to biological samples is hampered by conicting reports as to which experimental conditions are optimal. We have evaluated the performance of TiO2-MOAC under a wide range of loading and elution conditions. Loading and stringent washing of peptides with strongly acidic solutions ensured highly selective enrichment for phosphopeptides, with minimal carryover of non-phosphorylated peptides. Contrary to previous reports, the addition of glycolic acid to the loading solution was found to reduce specicity towards phosphopeptides. Base elution in ammonium hydroxide or ammonium phosphate provided optimal specicity and recovery of phosphorylated peptides. In contrast, elution with phosphoric acid gave incomplete recovery of phosphopeptides, whereas inclusion of 2,5-dihydroxybenzoic acid in the eluant introduced a bias against the recovery of multiply phosphorylated peptides. TiO2-MOAC was also found to be intolerant of many reagents commonly used as phosphatase inhibitors during protein purication. However, TiO2-MOAC showed higher specicity than immobilized gallium (Ga3R), immobilized iron (Fe3R), or zirconium dioxide (ZrO2) afnity chromatography for phosphopeptide enrichment. Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) was more effective in detecting larger, multiply phosphorylated peptides than liquid chromatography/electrospray ionization tandem mass spectrometry (LC/ESI-MS/MS), which was more efcient for smaller, singly phosphorylated peptides. Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.
Reversible phosphorylation is one of the most common mechanisms for covalent modication of proteins and is found in as many as one third of eukaryotic gene products.1,2 Although the number of cellular phosphoproteins is relatively high, the phosphorylated residues themselves are generally of low abundance due to the sub-stoichiometric nature of this modication.3 The detection and sequencing of tryptic phosphopeptides derived from such proteins has become an important aspect of biological and biomedical research. However, the prevalence of non-phosphorylated peptides in protein digests has made it necessary to develop efcient separation and enrichment methods for phosphopeptide analysis. Immobilized metal-ion afnity chromatography (IMAC) has been widely used for the selective enrichment of
*Correspondence to: U. K. Aryal, Pacic Northwest National Laboratory, P.O. Box 999, 902 Battelle Boulevard, Richland, WA 99352, USA. E-mail: Uma.Aryal@pnl.gov y Present address: Pacic Northwest National Laboratory, P.O. Box 999, 902 Battelle Boulevard, Richland, WA 99352, USA. z Present address: Fisheries and Oceans Canada, Institute of Ocean Sciences, P.O. Box 6000, 9860 West Saanich Road, Sidney, British Columbia, Canada V8L 4B2.

phosphopeptides;46 however, this method is prone to low recoveries and/or non-specic binding of non-phosphorylated peptides.7 Metal oxide afnity chromatography (MOAC) using titanium dioxide (TiO2) has recently been proposed as an alternative to IMAC.8,9 This technique is based on the selective interaction of phosphopeptides with porous TiO2 microspheres (titanospheres) via bidentate binding at the TiO2 surface.10,11 Such interactions arise from the afnity of oxygen in the phosphate groups for metal atoms in the MOAC resin.3 According to established protocols, peptide mixtures are loaded onto the column under acidic conditions and the bound phosphopeptides eluted in basic solution.8,9,12 The selectivity of different IMAC and MOAC methods for phosphopeptides has been studied extensively; however, many of the results presented in the literature are either contradictory or of limited practical use. In the case of TiO2-MOAC, peptide loading in the presence of 2,5-dihydroxybenzoic acid (DHB) or phthalic acid has been shown to increase selectivity for phosphopepetides;9,13 however, both compounds are suspected of causing interferences during the analysis of phosphopeptides by liquid chromatography/electrospray ionization tandem mass spectrometry (LC/ESI-MS/MS).12,14,15 A desire to

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

220

U. K. Aryal and A. R. S. Ross

avoid potential losses during further purication of TiO2enriched phosphopeptides prior to MS analysis has led to the search for alternative non-phosphopeptide excluders that are compatible with both matrix-assisted laser desorption/ ionization (MALDI) and ESI. Jensen and Larsen14 have proposed glycolic acid as an alternative to DHB or phthalic acid; however, Sugiyama and co-workers12 have since reported lower specicity of TiO2-MOAC in the presence of this reagent. Compared with the extensive literature on IMAC-based methods, relatively little information is available regarding optimal conditions for TiO2-MOAC.16 This makes it difcult to predict which method would be better suited to particular applications, or more compatible with the reagents used in biological studies. For example, TiO2-MOAC is known to be more tolerant than IMAC to certain reagents and buffers,14 but its compatibility with many of the protease and phosphatase inhibitors used in protein sample preparation is unknown. This information is needed to develop faster and more efcient enrichment protocols that can be applied to a wide range of biological samples. To address these uncertainties, we have evaluated the performance of TiO2-MOAC under different loading and elution conditions, using a single make and model of commercially available TiO2 column (NuTipTM NT1TIO, Glygen Corp., USA) to minimize experimental variability. We also compared the selectivity of TiO2-MOAC with that of zirconium dioxide (ZrO2)-MOAC, gallium (Ga3)- and iron (Fe3)-IMAC, and evaluated the performance of MALDI-MS and LC/ESI-MS/MS in detecting the phosphopeptides isolated using these different afnity techniques.

568C for 30 min. Each protein was alkylated using 40 mM iodoacetamide for 1 h at room temperature in the dark. The reaction was quenched by addition of DTT to a nal concentration of 5 mM, and the protein digested with trypsin at 378C overnight using an enzyme/substrate ratio of 1:50 (w/w). Each tryptic digest was dried in a vacuum centrifuge (model DNA 120; Thermo Savant, Colin Drive, NY, USA), reconstituted in 100 mL of 0.1% aqueous TFA, and desalted using PepCleanTM C18 spin columns (Pierce) according to the manufacturers instructions. Briey, each column was conditioned with 200 mL of 50% methanol and equilibrated with 200 mL of 0.5% TFA in 5% ACN, repeating each step once and centrifuging after each addition. The 100 mL of tryptic digest in 0.1% TFA was loaded, centrifuged, and the collected solution reloaded and centrifuged again. After washing twice with 200 mL of 0.5% TFA in 5% ACN, the spin column was eluted twice with 40 mL of 70% ACN. The combined eluate (80 mL) was dried, reconstituted in 100 mL of 0.1% TFA, and stored at 208C until further use. For each experiment, 1 mL of the desalted a-casein digest was combined with an equal volume of b-casein digest and 98 mL of the appropriate loading solution (see below) to produce a 10 fmol/mL combined digest solution. This was divided into ve 20 mL aliquots, each containing 200 fmol of digest, from which phosphopeptides were subsequently puried by MOAC or IMAC. Unless otherwise noted, 200 fmol of the combined a- and b-casein digests were used in all experiments.

Metal oxide afnity chromatography (MOAC)

Prior to sample loading, the disposable TiO2 NuTipsTM (Glygen) were conditioned/equilibrated with 20 mL of 1% TFA in 30% ACN using 10 aspirate/expel (A/E) cycles. To evaluate the effect of different organic acids on phosphopeptide binding, 1 mL volumes of both desalted casein digests were combined in 98 mL of 30% ACN containing 0.1, 1 or 5% AA, FA or TFA, and 20 mL of the resulting solution loaded onto a TiO2 tip using 50 A/E cycles. The effect of DHB on phosphopeptide binding was investigated by loading in 50% ACN containing 1% TFA, with or without DHB (130 mM). After loading the tips were washed once with 20 mL of 1% TFA in 30% ACN, once with 20 mL of 1% TFA in 50% ACN, twice with 20 mL of 1% TFA in 75% ACN, and, nally, twice with 20 mL of HPLC-grade water, using 10 A/E cycles each time. Bound phosphopeptides were eluted with 20 mL of 0.4 M NH4OH in 30% ACN, using 20 A/E cycles. The eluate was acidied immediately by adding 5 mL of aqueous 1% TFA and analyzed by mass spectrometry (MS). To investigate the efciency of different eluants, TiO2 tips were loaded using 1% TFA in 30% ACN, washed sequentially with various solutions as described above, and eluted with 20 mL of 0.25 M NH4HCO3 (pH 9) in 30% ACN, 0.4 M NH4OH (pH $11) in 30% ACN, 0.1 M NH4H2PO4 (pH 10) in 30% ACN, or 1% PA (with and without 130 mM DHB) in 50% ACN. The compatibility of TiO2 with reagents commonly used in protein sample preparation was also investigated by adding appropriate concentrations of okadaic acid, sodium uoride, sodium molybdate, sodium orthovanadate, sodium bglycerophosphate, imidazole, calyculin A, phenylmethylsulRapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

EXPERIMENTAL Materials
Acetonitrile (ACN) and HPLC-grade water were obtained from Merck (Darmstadt, Germany). Triuoroacetic (TFA), formic (FA), phosphoric (PA) and acetic (AA) acids, ammonium bicarbonate (NH4HCO3), phosphate (NH4H2PO4) and hydroxide (NH4OH), bovine a- and b-caseins, dithiothreitol (DTT) and iodoacetamide, mass calibrants (angiotensin 1, ACTH clip 1-17 and 18-35), PhosphoProleTM gallium silica spin columns (product no. P2873) and PHOSselectTM iron afnity gels (product no. P9740) were purchased from Sigma (St. Louis, MO, USA). Modied porcine trypsin (sequencing grade) was obtained from Promega (Madison, WI, USA) and 2,5-dihydroxybenzoic acid (DHB) from Waters (Milford, MA, USA). Porous titanium dioxide NuTipsTM (product no. NT1TIO), zirconium dioxide NuTipsTM (NT1ZRO) and empty TopTipsTM (TT2EMT) were purchased from Glygen Corp. (Columbia, MD, USA). PepCleanTM C18 spin columns (product no. 89870) were purchased from Pierce (Rockford, IL, USA), and C18 3M high-performance extraction disks from Empore (St. Paul, MA, USA).

Preparation of tryptic digests

Each of the model phosphoproteins a-casein and b-casein (100 picomoles) was dissolved separately in 100 mL of 0.1 M NH4HCO3 (pH 8.5) containing 10 mM DTT and incubated at

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Evaluation of TiO2 for phosphopeptide enrichment

221

fonyl uoride (PMSF), Sigma protease inhibitor cocktail (product no. P9599), or poly(ethylene glycol) to the loading solution (1% TFA in 30% ACN). To compare performance using the different MOAC media, TiO2 and ZrO2 NuTipsTM were equilibrated/ conditioned using 20 mL of 1% TFA in 30% ACN and loaded with 20 mL of combined digest in the same solvent, using 50 A/E cycles. After washing, as previously described, bound phosphopeptides were eluted with 20 mL of 0.4 M NH4OH in 30% ACN, using 20 A/E cycles. All these experiments were repeated at least three times to conrm the results.

Immobilized metal-ion afnity chromatography (IMAC)

For Ga-IMAC, pre-packed PhosphoProleTM gallium silica spin columns (Sigma) were rst washed/equilibrated with 50 mL of 1% TFA in 30% ACN and then loaded with 20 mL of combined digest in the same solvent, using a pipette to apply gentle pressure and ensure proper loading of the sample. After loading, columns were incubated at room temperature for 15 min before centrifuging at 1000 rpm for 1 min. The sample ow-through was then re-loaded, and the incubation and centrifugation repeated. The columns were washed with 50 mL volumes of the same washing solutions used for TiO2MOAC, with centrifugation at 1000 rpm for 2 min during each washing step. Bound phosphopeptides were eluted by loading the column with 20 mL of 0.4 M NH4OH in 30% ACN, incubating at room temperature for 5 min, then centrifuging at 1000 rpm for 1 min. The eluate was immediately acidied by adding 5 mL of 1% aqueous TFA and used for MS analysis. For Fe-IMAC, PHOS-SelectTM iron afnity gel beads were carefully stirred until completely and uniformly suspended in the stabilizing buffer supplied. The 5 mL of the resulting slurry was mixed with 5 mL of 1% TFA in 30% ACN and loaded into empty TopTipsTM (Glygen), using a regular pipette tip with about 1 mm of the end cut off to allow unrestricted ow and uniform distribution of the suspended beads. The beads were washed/equilibrated by adding 50 mL of 1% TFA in 30% ACN and spun in a microcentrifuge for 1 min at 1000 rpm. This operation was repeated two more times to ensure complete removal of the stabilizing buffer, which contains glycerol. The ow-through was discarded and the Fe-IMAC tips loaded with 20 mL of combined digest in 1% TFA/30% ACN. To ensure proper loading, the tip was again tted to a pipette and gentle pressure applied to push the sample into the gel beads. The tips were incubated at room temperature for 15 min before spinning at 1000 rpm for 1 min. The sample ow-through was reloaded and the incubation and centrifugation process repeated. Using the same procedure as for Ga-IMAC, the Fe-IMAC columns were washed, eluted, and the eluate acidied and used for MS analysis (see below).

using 1 mL of matrix solution containing 130 mM DHB in 10 mM (NH4)2HPO4 and 75% ACN and deposited directly onto the MALDI target plate. MALDI-MS analysis was performed using a Voyager-DE STR instrument (Applied Biosystems, Framingham, MA, USA) operating in the positive ion and reectron modes with delayed ion extraction. Close external mass calibration was performed using angiotensin 1, ACTH clip 1-17 and 18-35. Spectra were acquired over the m/z range 7004000 and the results of 200 laser shots combined to produce a single averaged spectrum for each sample. The efciency and selectivity of phosphopeptide binding under different loading and elution conditions were evaluated by comparing the number, type (i.e. singly, multiply, or non-phosphorylated) and relative abundance of peptide ions detected by MALDI-MS.

LC/ESI-MS/MS
For LC/ESI-MS/MS the eluates were dried, reconstituted in 10 mL of 1% aqueous TFA, and 6 mL injected onto the LC column. Analysis was performed using a nanoACQUITY UPLC system (Waters, Milford, MA, USA) interfaced to a quadrupole time-of-ight (Q-TOF) Ultima Global hybrid tandem mass spectrometer tted with a Z-spray nano-ESI source (Waters, Mississauga, ON, Canada). Chromatographic separation of peptides was accomplished using a Waters BEH130 C18 column (75 mm 100 mm, 1.75 mm) and a ow rate of 400 nL/min. Solvent A was 0.2% FA in water and solvent B was 0.2% FA in 100% ACN. The peptides were separated for 55 min by increasing the organic content of the mobile phase linearly from 0% to 45% over 45 min and then to 80% at 46 min, holding this concentration until 52 min and then reducing to 1% at 53 min. A 5 min seal wash with 10% ACN was also performed after each run. Mass calibration was performed using the product ion spectrum of Glubrinopeptide B acquired over the m/z range 501900. The instrument was run in positive ion mode with a source temperature of 808C. Analysis was carried out using datadependent acquisition, during which peptide precursor ions were detected by scanning from m/z 4001900 in TOF-MS mode. Multiply charged (2, 3, or 4) ions rising above predetermined threshold intensity were automatically selected for TOF-MS/MS analysis, and product ion spectra acquired over the m/z range 501900. The absence of singly charged peptides was subsequently veried by means of extracted ion chromatograms generated for m/z values corresponding to every theoretical, singly charged a- and b-casein peptide ion. The MS/MS data were converted into a peak list (pkl) le format using MassLynx v2.15 software (Micromass) and searched against the NCBInr database using an in-house Mascot server (v. 2.2). Phosphorylation of peptides identied as phosphopeptides via database searching was conrmed by manual inspection of the corresponding MS/MS spectra. Each validated phosphopeptide MS/MS spectrum was then reduced to single-charged, monoisotopic, centroided peaks using MaxEnt 3 (MassLynx v.4.1, Micromass), and in silico spectral fragmentation of the corresponding, databasematched phosphopeptide sequence carried out using BioLynx software. The calculated fragment ion masses were
Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

MALDI-MS
The acidied peptide eluates were vacuum-dried, reconstituted in 10 mL of 10% aqueous TFA and desalted with STAGE (STop And Go Extraction) tips prepared using EmporeTM C18 (octadecyl) extraction disks, according to established procedures.17 The bound peptides were eluted

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

222

U. K. Aryal and A. R. S. Ross

then compared with the experimental values to conrm the position(s) of the phosphorylated residue(s).

phosphopeptide coverage (Fig. 1(c)), and this loading solution was used for all subsequent experiments.

RESULTS AND DISCUSSION Effect of different organic acids on phosphopeptide binding

During afnity purication of phosphopeptides, acidic loading/washing solutions generally are effective in inhibiting non-specic binding of acidic and/or hydrophobic peptides, whereas high pH solutions are more efcient in recovering bound phosphopeptides.6 There have been several studies to determine optimal sample loading conditions for TiO2-MOAC;3,8,9,1214,18,19 however, the information presented in the literature remains contradictory. In view of these uncertainties, we decided to re-evaluate systematically the effects of different organic acids and modiers on TiO2-MOAC selectivity towards phosphopeptides. MALDI-MS analysis of the non-enriched casein digest detected only three phosphorylated peptide ions (labeled 8, 12 and 13; Fig. 1(a)), and at relatively low intensities. However, many additional phosphopeptides were detected following TiO2-MOAC enrichment (Figs. 1(b)1(d)). As in all subsequent gures and tables, phosphorylated peptides are annotated using numbers (125) and non-phosphorylated peptides using letters (AF) in Fig. 1. To evaluate the binding selectivity of the TiO2 columns for phosphorylated peptides under different loading conditions, casein digests were loaded onto TiO2 tips using 0.1, 1 or 5% AA, FA or TFA in 30% ACN and the bound peptides eluted using 0.4 M NH4OH in 30% ACN (Table 1). When samples were loaded in 0.1% concentrations of these acids, several acidic non-phosphorylated peptides were detected in the eluants. However, the number of non-phosphorylated peptides decreased as the concentration of acid in the loading solution increased (Table 1). The number of nonphosphorylated peptides detected was much lower with TFA than with either AA or FA. Repeated experiments conrmed that the ability of these organic acids to inhibit binding of non-phosphorylated peptides is in the order TFA > FA > AA. The effect of TFA concentration on the intensities of two representative non-phosphorylated peptide peaks A (m/z 1266.8) and D (m/z 1759.1) relative to phosphorylated peptide peaks was also monitored (Figs. 1(b)1(d)). The relative intensities of peaks A and D were reduced when using 1% rather than 0.1% TFA in the loading buffer, whereas using 5% TFA resulted in detection of an additional non-phosphorylated peptide F (m/z 2315.5). At the same time, multiply phosphorylated peptide 25 was enhanced in 1% relative to 0.1% TFA but absent from the 5% TFA spectrum. Although these differences are relatively minor they do suggest that a concentration of 1% TFA is generally sufcient to achieve optimum performance. Increasing the proportion of ACN from 30% to 50% or 75% did not improve TiO2 selectivity for phosphopeptides (data not shown). To summarize, loading in 1% TFA/30% ACN followed by stringent washing of the loaded peptides with 1% TFA in increasing concentrations of ACN (see Experimental section) gave best results in terms of

Effect of different eluants on phosphopeptide recovery

Conventional protocols often use basic solutions to recover phosphopeptides from afnity columns.20 However, acidic solutions containing phosphoric acid (PA) or a mixture of PA and DHB have been shown to act as efcient eluants for FeIMAC.15,20,2123 Similarly, although NH4OH is known to be an efcient eluant for both IMAC and MOAC,9,24 the suitability of other bases (e.g. NH4HCO3, NH4H2PO4) for TiO2-MOAC has yet to be explored. We decided to evaluate these acidic and basic eluants for the recovery of phosphopeptides from TiO2 using 1% TFA in 30% ACN as the loading solution (Fig. 2). A number of singly phosphorylated and multiply phosphorylated peptides (peaks 11, 18, 20 and 24/2400 ) were detected following elution with PA in 50% ACN (Fig. 2(b)); however, this was not as effective as 0.4 M NH4OH in 30% ACN (Fig. 2(a)). Furthermore, inclusion of DHB signicantly reduced the number of multiply phosphorylated peptides, as exemplied by the low signal intensity for peak 24 (Fig. 2(c)). Hence, these acidic eluants were found to be relatively inefcient for the recovery of phosphopeptides from TiO2. With regard to basic eluants, 0.4 M NH4OH in 30% ACN (Fig. 2(a)) and 0.1 M NH4H2PO4 in 30% ACN (Fig. 2(e)) gave better recovery of multiply phosphorylated peptides than did 0.25 M NH4HCO3 in 30% ACN (Fig. 2(d)). This is illustrated by the relative abundance of peak 24 in each spectrum and the fact that multiply phosphorylated peptide peaks 18, 19, 23 and 25 were detected after elution with NH4OH and NH4H2PO4, but not with NH4HCO3. Although similar results were obtained with NH4OH and NH4H2PO4 we decided to use the former as the eluant for all subsequent experiments, since optimal conditions for sample loading had already been determined using 0.4 M NH4OH in 30% ACN as the eluant (Fig. 1). In order to have practical utility, any enrichment method must enable the detection of phosphopeptides at concentrations typically found in gel spot digests, or in complex mixtures derived from whole cell lysates or subcellular fractions.25 Sensitivity was therefore evaluated by loading 100, 50 and 25 fmol of the combined casein digests onto the TiO2 tips using 1% TFA in 30% ACN, eluting with 0.4 M NH4OH in 30% ACN and analyzing the peptides by MALDIMS (data not shown). The majority of casein phosphopeptides were detected in all cases. The sensitivity and specicity of the optimized TiO2MOAC method were further evaluated using Arabidopsis thaliana leaf proteins separated by two-dimensional gel electrophoresis (2-DE) and visualized by silver staining. Two gel spots corresponding to isoforms of carbonic anhydrase 1 (CA1), which contains four phosphorylation sites (Aryal et al. unpublished data; Supplementary Table S1, see Supporting Information), were excised from the gel, combined, and subjected to trypsin digestion using a MassPREP digest station (Micromass, Manchester, UK) according to the recommended procedure. Half of the digest ($5 mL) was analyzed directly by LC/ESI-MS/MS while the remainder
Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Evaluation of TiO2 for phosphopeptide enrichment

223

(a) 100
80 A B C

2.2E + 4

In ten sity (%)

60 40 20 0 699.0

F 8 1359.4 12 13 E 2680.2 3340.6 0 4001.0 2.2E + 4 8 12

2019.8

(b) 100
80

In ten sity (%)

60 40 20 0 699.0 A 5 7 1359.4 D 11 10 9 13 2019.8 16 17 2680.2 24 25 0 3340.6 4001.0 2.2E + 4

(c) 100
In ten sity (%)
80 60 40 20 0 699.0 5 7 1359.4 8

12

13 11 9 10 2019.8 12 13 17 16 18 23 24 24 19 25 20 2680.2 3340.6

0 4001.0 2.2E + 4

(d) 100
In ten sity (%)
80 60 40 20 0 699.0 1359.4 7 5 7 8

11 10 9 F 2019.8 20 F 16 18 23 19 15 2680.2 17 24 24 0 3340.6 4001.0

m/z

Figure 1. Effect of TFA concentration on the loading of phosphopeptides during TiO2MOAC. Panels show representative MALDI mass spectra obtained from 200 fmol of a combined a- and b-casein digest (a) before TiO2-MOAC enrichment, and following TiO2MOAC enrichment using (b) 0.1%, (c) 1%, or (d) 5% TFA in 30% ACN as the loading solution. Bound phosphopeptides were eluted with 0.4 M NH4OH (pH 11) in 30% ACN. Phosphorylated peptides are annotated using numbers (125) and non-phosphorylated peptides using letters (AF), as summarized in Table 3 and Supplementary Table 2 (see Supporting Information), respectively. was vacuum-dried, re-constituted in 15 mL of 1% TFA in 30% ACN, and then puried by TiO2-MOAC prior to LC/ESIMS/MS analysis. Without prior enrichment eleven nonphosphorylated peptides and four singly phosphorylated peptides were identied; however, only one non-phosphorylated peptide (YMVFACSDSR) was detected following TiO2MOAC enrichment, whereas the ow-through fraction contained all twelve non-phosphorylated CA1 peptides
Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

224

U. K. Aryal and A. R. S. Ross

Table 1. Effect of acidic loading conditions on TiO2-MOAC enrichment of phosphorylated peptides from 200 fmol of a combined a- and b-casein digest
No. of peptides identied ( p < 0.05)b,c Phosphorylated NonLoading Concentration solution acida (%) pH Singly Multiply phosphorylated Acetic acid 0.1 1.0 5.0 0.1 1.0 5.0 0.1 1.0 5.0 3.2 2.8 1.2 2.6 2.2 1.0 2.1 1.2 0.2 7 7 7 7 7 7 7 7 7 4 6 8 5 7 8 8 10 10 7 6 1 7 5 0 4 0 0

FA

TFA

All loading solutions contained 30% ACN. Peptides eluted with 0.4 M NH4OH in 30% ACN. c Average of three independent experiments.
b

and no phosphopepetides. Furthermore, ion scores for the identied phosphopeptides were signicantly improved in the enriched sample, presumably due to enhancement of the peptide ion signal (Supplementary Table S1, see Supporting Information). Results demonstrate the utility of our TiO2MOAC method for detecting phosphopeptides and identifying sites of protein phosphorylation in biological extracts.

Use of non-phosphopeptide excluders

Loading TiO2 columns with 1% TFA in 30% ACN provides high selectivity towards phosphopeptides for relatively simple samples; however, Jensen and Larsen14 have suggested that a non-phosphopeptide excluder should be used to optimize selectivity for the enrichment of phosphopeptides from complex samples such as cell lysates. DHB and phthalic acid have been used successfully for this purpose,9,13 but are thought to inhibit LC/ESI-MS/MS analysis of the recovered phosphopeptides by contaminating the LC system and the inlet of the mass spectrometer.12,14,15 Jensen and Larsen14 have recently shown that 1 M glycolic (hydroxyacetic) acid is equally effective as a non-phosphopeptide excluder, and is compatible with both MALDI-MS and LC/ESI-MS/MS analysis. However, Sugiyama and coworkers found that using 300 mg/mL ($4 M) glycolic acid in the loading solvent resulted in non-specic binding of peptides.12 In view of this uncertainty we decided to reevaluate the use of DHB and glycolic acid as nonphosphopeptide excluders for MOAC. Based on previous investigations involving a- and bcasein,23 DHB was added to the MOAC loading solution at a concentration of 130 mM, whereas glycolic acid was included at 1, 0.5 or 0.25 M to investigate whether reducing (rather than increasing)12 the concentration of this reagent might enhance specicity. MALDI mass spectra of casein digests enriched by TiO2-MOAC with DHB or glycolic acid in the loading solution are shown in Fig. 3. The specicity obtained using 1% TFA with 30% ACN alone as the loading solution is sufcient to eliminate non-phosphorylated peptides from the

spectrum (Fig. 3(a)). DHB appeared to enhance detection of certain multiply phosphorylated peptides, as illustrated by increases in the relative intensities of peaks 6 and 24 in the presence of DHB (Fig. 3(b)). However, the relative intensities of other peaks such as 8, 10 and 17 were apparently reduced in the presence of DHB, suggesting that addition of this reagent to the loading solution offers no signicant advantage. When 1 M glycolic acid was used in the loading solution several non-phosphorylated peptides (labeled A, B, D, E, and F/F0 ) were detected at signicant levels (Fig. 3(c)). The relative abundances of these peptides decreased as the concentration of glycolic acid decreased from 1 to 0.5 or 0.25 M (Figs. 3(c)3(e)), suggesting that the addition of glycolic acid to the loading solution actually increased nonspecic binding of peptides to TiO2. No enhancement in the relative abundance of multiply phosphorylated peptides was observed using glycolic acid. We also evaluated DHB and glycolic acid as nonphosphopeptide excluders for ZrO2-MOAC. The addition of DHB to the loading solution reduced the binding of nonphosphorylated peptides to ZrO2 columns while again enhancing the relative abundance of certain multiply phosphorylated peptides (data not shown). As well, the use of glycolic acid appeared to increase non-specic binding of peptides to ZrO2 columns, and gave no enhancement in the relative abundance of multiply phosphorylated peptides (data not shown). These observations are in direct contrast with those of Jensen and Larsen,14 who found 1 M glycolic acid to be an efcient non-phosphopeptide excluder. Our results complement those obtained by Sugiyama and coworkers12 who used a single, relatively high concentration ($4 M) of glycolic acid. Hydroxy acids bind to metal oxides by forming a cyclic chelate with the metal,12,26,27 whereas phosphate anions do so by bridging two metal atoms.9 Thus, hydroxy acids such as DHB should bind to metal oxides more weakly than a phosphate group but more strongly than the carboxylic groups of acidic non-phosphorylated peptides. It is unclear why glycolic (hydroxyacetic) acid should have been ineffective in displacing non-phosphorylated peptides from TiO2 during our study. One possibility is that the structure and retention properties of titania, which are strongly dependent on the calcination temperature of the beads,28,29 may differ from one study to another. In any event, it appears that loading samples with glycolic acid may actually increase non-specic binding of peptides during MOAC, depending on the materials and conditions used.

Effect of phosphatase inhibitors on phosphopeptide enrichment

During proteomic analysis a number of reagents may be used to solubilize hydrophobic proteins and to inhibit protease and phosphatase activities. TiO2-MOAC is known to be compatible with most of the buffers and detergents used in biological experiments;14 however, the tolerance of TiO2MOAC to other reagents, including protease and phosphatase inhibitors, has not been widely explored. This information would be helpful in deciding what purication steps might be necessary prior to enrichment by TiO2-MOAC and
Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Evaluation of TiO2 for phosphopeptide enrichment

225

(a)
Intensity (%)

100 80 60 40 20 0 5 7 1359.4 8 8

12

2.2 E + 4

11 9 10

13 17 16 2019.8 18 23 24 24 19 25 20 2680.2 3340.6

(b)
Intensity (%)

699.0

100 80 60 40 20 699.0 100 80 60 40 20 0 699.0 A C 5 5 7 9 10 8 11 0 A 5 C 5 7 F F 11 12 10 E E D9 2019.8 13 13

0 4001.0 2.2E + 4

18

24 20 24 3340.6

1359.4

2680.2

(c)
Intensity (%)

0 4001.0 2.2E + 4

12 F 2019.8 12 F 2680.2 24 3340.6

1359.4

0 4001.0 2.2E + 4

(d)
Intensity (%)

100 80 60 40 20 0 699.0 A 8

11

13 5 67 1359.4 9 10 D F 2019.8 12 24 16 17 20 24 2680.2 3340.6

0 4001.0 2.2 E + 4

(e)
Intensity (%)

100 80 60 40 20 0 699.0 5 8

11

13

9 10 7 2019.8

20 24 24 17 18 19 23 25 16 2680.2 3340.6

1359.4

0 4001.0

m/z
Figure 2. Recovery of phosphopeptides from TiO2 using different eluants. Panels show representative MALDI mass spectra obtained from 200 fmol of a combined a- and b-casein digest following TiO2-MOAC enrichment and elution with (a) 0.4 M NH4OH in 30% ACN, (b) 1% phosphoric acid (PA) in 50% ACN, (c) 1% PA and 130 mM DHB in 50% ACN, (d) 0.25 M NH4HCO3 in 30% ACN, or (e) 0.1 M NH4H2PO4 in 30% ACN. Samples were loaded using 1% TFA in 30% ACN. MS analysis, especially when using gel-free proteomic approaches. Sodium uoride, sodium molybdate, sodium orthovanadate, sodium b-glycerophosphate, okadaic acid, imidazole and calyculin A are some of the more commonly used phosphatase inhibitors, and were included in our study. These reagents were added to the 1% TFA in 30% ACN loading solution at concentrations normally used in bioRapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

226

U. K. Aryal and A. R. S. Ross

(a) 100
In ten sity (%)
80 60 40 20 0 899.0 5 7 9 8

12

8397.1

11 10

13 17 16 12 2139.8

24 18 24 19 23 20 25 2760.2 3380.6

1519.4

0 4001.0 8397.1

(b)
In ten sity (%)

100 80 60 40 20 0 899.0 8 5 7 9 6 1519.4 12 D 11 10 13 24 17 18 19 23 16 20 24 25 15 2139.8 2760.2 3380.6

(c)
In ten sity (%)

100 80 60 40 20 0 899.0 A B 5 7 8 910 F 13 E F 24 24 2760.2 3380.6

0 4001.0 8397.1

1519.4

2139.8

0 4001.0 8397.1

(d)
In ten sity (%)

100 80 60 40 20 0 899.0 AB 5 6 7 8 D 11 910 13 E E F F 2139.8 12 20 24 24 3380.6 12

1519.4

2760.2

0 4001.0 8397.1

(e)
In ten sity (%)

100 80 60 40 20 0 899.0 8 5 5 7 6

13 10 11 9 2139.8 24 23 24 20 2760.2

1519.4

3380.6

0 4001.0

m/z

Figure 3. Effect of non-phosphopeptide excluders on phosphopeptide recovery using TiO2-MOAC. Panels show MALDI mass spectra obtained from 200 fmol of a combined a- and b-casein peptide digest loaded in a 30% ACN solution of 1% TFA containing (a) no additives, (b) 130 mM DHB, (c) 1 M, (d) 0.5 M, or (e) 0.25 M glycolic acid. Peptides were eluted with 0.4 M NH4OH in 30% ACN.

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

Evaluation of TiO2 for phosphopeptide enrichment

227

Table 2. Effect of different reagents on the enrichment of casein phosphopeptides using TiO2- and ZrO2-MOACa
No. of peptides identiedc,d Phosphorylated Added reagentb TiO2-MOAC None 20 mg/mL DHB 1 M glycolic acid 500 mM glycolic acid 250 mM glycolic acid 20 mM sodium uoride 2 mM sodium molybdate 2 mM sodium orthovanadate 10 mM sodium b-glycerophosphate 2 mM okadaic acid 100 mM imidazole 100 nM calyculin A 40 mM PMSF 1% PEG 0.01% Sigma protease inhibitor cocktail (product # P9599) ZrO2-MOAC None 20 mg/mL DHB 1 M glycolic acid 500 mM glycolic acid 250 mM glycolic acid
a

NonSingly Multiply phosphorylated 7 7 7 7 7 6 2 7 3 7 7 7 7 7 7 10 10 4 6 6 11 0 10 0 10 5 6 5 11 7 0 0 5 3 2 4 3 6 4 1 1 0 1 1 1

The effect of adding other reagents to the loading solution was also investigated. These included imidazole, calyculin A, PMSF, poly(ethylene glycol) (PEG), and a protease inhibitor cocktail from Sigma (product no. P9599) containing 4-(2-aminoethyl)benzenesulfonyl uoride (AEBSF), bestatin, pepstatinA, E-64, leupeptin and 1,10-phenanthroline, which are used routinely for plant protein extractions. The results of these experiments are again summarized in Table 2. We included PEG in our investigation because it has been found to improve detection of low-abundance proteins by selectively precipitating RuBisCo from plant protein extracts.3033 It is also used to purify plasma membrane proteins by aqueous two-phase partitioning.27 Our results show that TiO2 afnity chromatography is somewhat tolerant of most reagents investigated, but that the presence of imidazole, PMSF, calyculin A or the Sigma protease inhibitor cocktail reduces the number and relative intensities of multiply phosphorylated peptide peaks. The removal of certain additives prior to TiO2-MOAC may, therefore, be necessary to achieve optimum performance.

Comparison of TiO2-MOAC with other phosphopeptide enrichment methods

Both MALDI-MS and LC/ESI-MS/MS data were used to compare TiO2-MOAC with ZrO2-MOAC, Ga-IMAC and FeIMAC for phosphopeptide enrichment, and a complete summary of the phosphorylated a- and b-casein peptides detected using each technique is provided in Table 3. In all cases, samples were loaded in 1% TFA/30% ACN and eluted with 0.4 M NH4OH in 30% ACN. Both MS techniques detected more multiply phosphorylated peptides in the TiO2 eluant than in the ZrO2 eluant, although the number of singly phosphorylated peptides was similar in each case (Table 3). These results are in keeping with previous studies that also found TiO2-MOAC to be more efcient than ZrO2-MOAC at recovering multiply phosphorylated peptides.14,24 Unlike Kweon and Hakansson,24 however, we did not nd ZrO2 to be better than TiO2 for recovery of singly phosphorylated peptides. Fe-IMAC appeared to favor the enrichment of multiply phosphorylated peptides when compared with Ga-IMAC, which showed balanced recovery of singly and multiply phosphorylated peptides (Figs. 5(c) and 5(d)). However, both IMAC methods appeared less selective than TiO2MOAC, since non-phosphorylated peptides observed in the MALDI-MS spectra of IMAC-puried samples were not observed in TiO2-enriched samples. Nevertheless, relative signal intensities for multiply phosphorylated peptides were higher for IMAC than for TiO2-MOAC eluants, as shown by the corresponding MALDI mass spectra (Figs. 5(a), 5(c) and 5(d)), supporting previous claims that multiply phosphorylated peptides are recovered more efciently using IMAC methods.9 For example, peaks 20, 24 and 25 were generally of higher abundance in IMAC than in TiO2 spectra, whereas peaks 15 and 22 were detected by MALDI-MS for both IMAC methods but not for TiO2. Similarly, the tetra-phosphorylated peptide 21 was detected by LC/ESI-MS/MS in both Ga- and Fe-IMAC eluants but not TiO2 (Table 3). This could be due to the strength of the interaction between multiply phosphorylated peptides and
Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

7 7 7 7 7

4 5 1 4 4

2 0 6 3 2

Final volume of loading solution adjusted to 20 mL following addition of reagent. b All samples loaded in 1%TFA/30% ACN. c Peptides enriched from 200 fmol of a combined a- and b-casein digest. d Peptides detected using MALDI-MS.

logical experiments, as summarized in Table 2. This table also lists the number of phosphorylated and non-phosphorylated peptides detected following TiO2-MOAC enrichment in the presence of these reagents. Examples of the MALDI mass spectra from which these results were derived are shown in Fig. 4. By way of comparison, Fig. 4(a) shows the control spectrum obtained without including any of these reagents in the loading solution. Several of the reagents tested were found to be incompatible with TiO2 afnity chromatography. The most dramatic effect was observed with sodium molybdate and sodium bglycerophosphate, which completely removed specicity towards phosphopeptides, as indicated by the detection of only two phosphorylated peptides (peaks 8 and 12) but several additional non-phosphorylated peptides (Figs. 4(d) and 4(e)). Sodium orthovanadate also had a negative impact on specicity, as illustrated by the number of nonphosphorylated peptides co-puried with phosphopeptides (Fig. 4(f)). Although TiO2-MOAC appears to be somewhat tolerant of sodium uoride (Fig. 4(b)) its presence again reduced specicity and recovery of certain singly (peaks 5, 7, 8) and multiply phosphorylated peptides (peaks 24 & 25) when compared with the control (Fig. 4(a)). However, okadaic acid appeared to have negligible impact on the performance of TiO2-MOAC (Fig. 4(c)).

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

228

100

(a)
12 11 80
8 13 11 9 10 16 2019.8 2680.2 3340.6 4001.0 699.0
0

12

5387.0

100

(b)

5387.0

80

In ten sity (%)

60 17

60 40 20 AB 5 2019.8 2680.2 1359.4 0 8 9 13 E 10 16 171819 23 15 20 24 25 3340.6

40 24 23 24 18 19 20 25

U. K. Aryal and A. R. S. Ross

20

5 7

699.0

1359.4

0 4001.0 5387.0

8 5387.0 100

100

(c)
12 80 D 60 40 20 0 0 4001.0 699.0 1359.4 5387.0 100 B 12 13 2019.8 12 8 D 20 E 2680.2 3340.6 0 0 4001.0 699.0 A C5 7 1359.4 11 9 10 13 D 2019.8 2680.2 3340.6 19 24 17 16 18 23 24 25 15 20

(d)

80

60

In ten sity (%)

40 7

20

0 699.0

1359.4

2680.2

3340.6

0 4001.0 5387.0

In ten sity (%)

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

100

(e) (f)
12 60 40 13 2019.8 80

80

60 D 8

11 910 13

40

20

E F 2019.8

17 19 24 16 1820 23 24 25 2680.2 3340.6

0 699.0

1359.4

0 4001.0

m/z

Rapid Commun. Mass Spectrom. 2010; 24: 219231

Figure 4. Effect of phosphatase inhibitors on phosphopeptide purication using TiO2-MOAC. Panels show MALDI mass spectra of phosphopeptides puried from 200 fmol of a combined a- and b-casein digest loaded using 1% TFA in 30% ACN with (a) no additives, (b) 20 mM sodium uoride, (c) 2 mM okadaic acid, (d) 2 mM sodium molybdate, (e) 10 mM sodium b-glycerophosphate, or (f) 2 mM sodium orthovanadate. Peptides were eluted with 0.4 M NH4OH in 30% ACN.

DOI: 10.1002/rcm

Evaluation of TiO2 for phosphopeptide enrichment

229

TiO2, which may require a more stringent elution protocol than used in this comparative study.34 However, TiO2MOAC appeared to show even greater selectivity in our study than previously observed.9 For example, LC/ESIMS/MS detected just two non-phosphorylated peptides (peaks A and D) in TiO2-MOAC eluates whereas four nonphosphorylated peptides were detected in IMAC-enriched samples (Supplementary Table S2, see Supporting Information). Non-selective binding of peptide D (m/z 1760) to TiO2 and IMAC columns has been reported elsewhere.6,7,9,14 To summarize, TiO2-MOAC showed higher specicity than Ga-IMAC, Fe-IMAC or ZrO2-MOAC for phosphopeptide enrichment.

Comparison of MALDI-MS and LC/ESI-MS/MS for phosphopeptide analysis

LC/ESI-MS/MS analysis of the unpuried casein digest matched a total of twelve peptides (four phosphorylated, eight non-phosphorylated) to the rst (aS1) sub-unit of a-

casein, eleven peptides (ve phosphorylated, six nonphosphorylated) to the second (aS2) sub-unit, and ve peptides (two phosphorylated, three non-phosphorylated) to b-casein (data not shown). When puried using the TiO2MOAC, nine peptides (seven phosphorylated, two nonphosphorylated) were matched to aS1, eight peptides (all phosphorylated) to aS2, and three phosphopeptides to bcasein (Table 3 and Supplementary Table 2, see Supporting Information). Hence, TiO2 enrichment increased the number of matched phosphopeptides as well as their condence levels (ion scores) when compared with the unextracted samples. On the other hand, no non-phosphorylated peptides were detected in TiO2-MOAC-enriched samples using MALDI-MS (Fig. 1(b)). This difference could be due to the inherent sensitivity of LC/ESI-MS/MS and of the Q-TOF instrument, which is greater than that of the instrument used for MALDI-MS analysis. It is important to note that the two non-phosphorylated peptides (A and D) detected by LC/ ESI-MS/MS in TiO2-enriched samples were also the most

Table 3. Detection of phosphorylated casein peptides by MALDI-MS and LC/ESI-MS/MS following enrichment by TiO2-MOAC, ZrO2-MOAC, Ga-IMAC or Fe-IMAC

aS1 and aS2 refer to the rst and second subunits of a-casein, respectively. b-C represents peptides of b-casein. pI calculated according to Bjellqvist et al.36 c Hydrophobicity (BB index) determined according to Bull and Breese.37 d Samples were loaded in 1% TFA in 30% ACN and eluted with 0.4 M NH4OH in 30% ACN. Each column was loaded with 200 fmol of the combined a- and b-casein digests. e1, e2, e3 and e4 Methionine-oxidized forms of peptides 5, 7, 11 and 16, respectively. f Peptide 6 sequence according to Larsen et al.9 g Peptide 14 sequence according to Hsieh et al.38 with the N-terminal glutamine cyclized to pyroglutamic acid. h Peak due to metastable loss of phosphate from peptide 25. i Variant of b-casein with an N-terminal glycine residue, according to Jensen and Larsen.14
b

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

230

U. K. Aryal and A. R. S. Ross

(a) 100
80
In ten sity (%)

12

1.1E + 4

60 40 20 0 699.0

8 11 9 10 5 7 1359.4 2019.8 12 8 13 17 16 18 23 24 24 19 25 20 2680.2 3340.6

0 4001.0 1.1E + 4

(b) 100
In ten sity (%)

80 60 40 20 0 699.0 A

11 10 13 5 7 9 7 1359.4 2019.8 12

24 18 20 24 2680.2 3340.6

0 4001.0 1.1E + 4

(c) 100
In ten sity (%)

80 60 40 20 0 699.0 B 8 11

24 13 24 23 20 22 17 16 18 25 19 15 2019.8 11 2680.2 3340.6

10 5 7 9

1359.4

0 4001.0 1.1E + 4

(d) 100
In ten sity (%)

80 60 40 20 0 699.0 A C 7 9 12 13 2019.4 m/z 17 16 15 18 24 23 24 19 20 22 25 3339.8 0 4000.0

1359.2

2679.6

Figure 5. Comparison of TiO2-MOAC with other phosphopeptide enrichment methods. Panels show MALDI mass spectra for 200 fmol of a combined a- and b-casein digest enriched for phosphopeptides using (a) TiO2-MOAC, (b) ZrO2-MOAC, (c) Ga-IMAC, and (d) Fe-IMAC. In all cases, samples were loaded using 1% TFA in 30% ACN and eluted with 0.4 M NH4OH in 30% ACN. abundant peptides detected by MALDI-MS in the unpuried digest (see Fig. 1(a)). Regardless of which MS technique was used, fewer non-phosphorylated peptides were detected in TiO2-MOAC-enriched samples than in any other extracts (Supplementary Table S2, see Supporting Information), further conrming TiO2-MOAC as the most selective method for phosphopeptide enrichment. With regard to phosphopeptide analysis, more singly and doubly phosphorylated peptides were detected using LC/ ESI-MS/MS than MALDI-MS (Table 3). In fact, with the exception of peptide 9 all singly or doubly phosphorylated peptides observed using MALDI-MS were also detected by LC/ESI-MS/MS, while peptides 14 and 14 were only detected using LC/ESI-MS/MS. In contrast, more peptides
Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Evaluation of TiO2 for phosphopeptide enrichment

231

carrying three or more phosphate groups were detected using MALDI-MS than with LC/ESI-MS/MS. For example, multiply phosphorylated peptides 15, 1721, and 25 were only observed using MALDI-MS whereas the only multiply phosphorylated peptide detected by LC/ESI-MS/MS alone was peptide 21 (a missed-cleavage product incorporating peptide 19). Larsen et al.9 and Gruhler et al.35 also reported a bias against the detection of multiply phosphorylated peptides when using LC/ESI-MS/MS. Using our TiO2MOAC enrichment method, we were able to detect more phosphorylated peptides (including peptides 160 , 22 and 24) than reported by others using LC/ESI-MS/MS,9 and with less material (200 fmol of combined a- and b-casein digests), conrming the high efciency of the TiO2 tips. LC/ESI-MS/ MS was more efcient in detecting singly phosphorylated peptides than MALDI-MS for all four enrichment methods, whereas the opposite was true for multiply phosphorylated peptides (Table 3).

Funding for protein mass spectrometry equipment was provided by the Saskatchewan Provincial Government and the National Research Council of Canada, from which this article is contribution number 50149.

REFERENCES
1. Ahn NG, Resing KA. Nat. Biotechnol. 2001; 19: 317. 2. McLachlin DT, Chait BT. Curr. Opin. Chem. Biol. 2001; 5: 591. 3. Mazanek M, Mitulovi G, Herzog F, Stingl C, Hutchins JRA, Peters J-M, Mechtler K. Nat. Protocols 2007; 2: 1059. 4. Anderson L, Porath JO. Anal. Biochem. 1986; 154: 250. 5. Kokubu M, Ishihama Y, Sato T, Nagasu T, Oda Y. Anal. Chem. 2005; 77: 5144. 6. Posewitz MC, Tempst P. Anal. Chem. 1999; 71: 2883. 7. Ficarro SB, McCleland ML, Stukenberg PT, Burke DJ, Ross MM, Shabanowitz J, Hunt DF, White FM. Nat. Biotechnol. 2002; 20: 301. 8. Pinske MW, Uitto PM, Hilhorst MJ, Ooms B, Heck AJ. Anal. Chem. 2004; 76: 3935. 9. Larsen MR, Thingholm TE, Jensen ON, Roepstroff P, Jrgensen TJD. Mol. Cell. Proteomics 2005; 4: 873. 10. Connor PA, McQuillan AJ. Langmuir 1999; 15: 2916. 11. Dobson KD, McQuillan AJ. Acta Part A Mol. Biomol. Spectrosc. 2000; 56: 557. 12. Sugiyama N, Masuda T, Shinoda K, Nakamura A, Tomita M, Ishihama Y. Mol. Cell. Proteomics 2007; 6: 1103. 13. Thingholm TE, Jorgensen TJD, Jensen ON, Larsen MR. Nat. Protocols 2006; 1: 1929. 14. Jensen SS, Larsen MR. Rapid Commun. Mass Spectrom. 2007; 21: 3635. 15. Imanishi SY, Kochin V, Eriksson JE. Proteomics 2007; 7: 174. 16. Cantin GT, Shock TR, Park SK, Madhani HD, Yates JR III. Anal. Chem. 2007; 79: 4666. 17. Rappsilber J, Ishihama Y, Mann M. Anal. Chem. 2003; 75: 663. 18. Klemm C, Otto S, Wolf C, Haseloff RF, Beyermann M, Krause E. J. Mass Spectrom. 2006; 41: 1623. 19. Thingholm TE, Jensen ON, Robinson PJ, Larsen MR. Mol. Cell. Proteomics 2008; 7: 661. 20. Hart SR, Watereld MD, Burlingame AL, Cramer R. J. Am. Soc. Mass Spectrom. 2002; 13: 1042. 21. Stensballe A, Jensen ON. Rapid Commun. Mass Spectrom. 2004; 18: 1721. 22. Kocher T, Allmaier G, Wilm M. J. Mass Spectrom. 2003; 38: 131. 23. Aryal UK, Olson DJH, Ross ARS. J. Biomol. Techniques 2008; 19: 296310. 24. Kweon HK, Hakansson K. Anal. Chem. 2006; 78: 1743. 25. Ross ARS. Methods Enzymol. 2007; 423: 549572. 26. Tani K, Ozawa M. J. Liq. Chrom. Rel. Technol. 1999; 22: 843. 27. Tunesi S, Anderson M. J. Phys. Chem. 1991; 95: 3399. 28. Tani K, Miyamoto E. J. Liq. Chrom. Rel. Technol. 1999; 22: 857. 29. Tani K, Kitada M, Tachibana M, Koizumi H, Kiba T. Chromatographia 2003; 57: 409. 30. Kim ST, Cho KS, Jang YS, Kang KY. Electrophoresis 2001; 22: 2103. 31. Xi J, Wang X, Li S, Zhou X, Yue L, Fan J, Hao D. Phytochemistry 2006; 67: 2341. 32. Kwon SJ, Choi EY, Seo JB, Park OK. Mol. Cells 2007; 24: 268. 33. Kim SG, Kim ST, Kang SY, Wang Y, Kim W, Kang KY. Plant Cell Rep. 2008; 27: 363. 34. Imanishi SY, Kochin V, Ferraris SE, de Thonel A, Pallari HM, Corthals GL, Eriksson JE. Mol. Cell. Proteomics 2007; 6: 1380. 35. Gruhler A, Olsen JV, Mohammed S, Mortensen P, Faergeman NJ, Mann M, Jensen ON. Mol. Cell. Proteomics 2005; 4: 310. 36. Bjellqvist B, Basse B, Olsen E, Celis JE. Electrophoresis 1994; 15: 529. 37. Bull HB, Breese K. Arch. Biochem. Biophys. 1974; 161: 665. 38. Hsieh HC, Sheu C, Shi FK, Li DT. J. Chromatogr. A 2007; 1165: 128.

CONCLUSIONS
TiO2-MOAC using porous titanium (NuTipTM) media provides efcient and highly selective enrichment of phosphopeptides from protein digests. Optimal selectivity was achieved by loading peptides with 1% TFA in 30% ACN and washing sequentially with acidic (1% TFA) solutions of increasing organic content (30, 50 and 75% ACN). Glycolic acid was found to be ineffective as a non-phosphopeptide excluder in the loading solution. With the exception of okadaic acid, TiO2-MOAC was found to be intolerant of many commonly used phosphatase inhibitors, which has implications for phosphoproteomic studies that do not involve gel separation prior to MOAC. Base elution in 0.4 M NH4OH or 0.1 M NH4H2PO4 gives efcient and balanced recovery of singly and multiply phosphorylated peptides and is compatible with MALDI-MS and LC/ESIMS/MS, both of which may be necessary to ensure detection of all phosphopeptides recovered using TiO2-MOAC. GaIMAC and Fe-IMAC gave better recovery of multiply phosphorylated peptides than MOAC under the experimental conditions, but were less specic than TiO2 for phosphopeptide enrichment.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article.

Acknowledgements
We thank Doug Olson and Steve Ambrose of the Mass Spectrometry and Protein Research Group at NRC-PBI for technical support. We also thank Dr. Joan Krochko for valuable suggestions and Dr. Randy Purves for reviewing the manuscript. The Visiting Fellow and Research Associate positions awarded to UKA by, respectively, the Natural Science and Engineering Research Council (NSERC) of Canada and NRC-PBI are also gratefully acknowledged.

Copyright # 2009 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2010; 24: 219231 DOI: 10.1002/rcm