Phosphopeptide Enrichment by Aliphatic Hydroxy Acid-modified Metal Oxide Chromatography for Nano-LC-MS/MS in Proteomics Applications*S

We developed novel methods for phosphopeptide enrichment using aliphatic hydroxy acid-modified metal oxide chromatography (MOC). Titania and zirconia were successfully applied to enrich phosphopeptides with the aid of aliphatic hydroxy acids, such as lactic acid and β-hydroxypropanoic acid, to reduce the interaction between acidic non-phosphopeptides and the metal oxides. These methods removed the vast majority of non-phosphopeptides from phosphoprotein standard digests, and large numbers of phosphopeptides could be readily identified. The methods were coupled with nano-LC-MS/MS systems without difficulty. Recovery of phosphopeptides in MOC varied greatly from peptide to peptide, ranging from a few percent to 100%, and the average was almost 50%. Repeatability and linearity were satisfactory. In an examination of the cytoplasmic fraction of HeLa cells, more than 1000 phosphopeptides were identified using lactic acid-modified titania MOC and β-hydroxypropanoic acid-modified zirconia MOC, respectively. The overlap between phosphopeptides enriched by these two methods was 40%, and the combined results provided 1646 unique phosphopeptides. To our knowledge, this is the first successful application of a single MOC-based approach to phosphopeptide enrichment from complex biological samples such as cell lysates.

We developed novel methods for phosphopeptide enrichment using aliphatic hydroxy acid-modified metal oxide chromatography (MOC). Titania and zirconia were successfully applied to enrich phosphopeptides with the aid of aliphatic hydroxy acids, such as lactic acid and ␤-hydroxypropanoic acid, to reduce the interaction between acidic non-phosphopeptides and the metal oxides. These methods removed the vast majority of non-phosphopeptides from phosphoprotein standard digests, and large numbers of phosphopeptides could be readily identified. The methods were coupled with nano-LC-MS/MS systems without difficulty. Recovery of phosphopeptides in MOC varied greatly from peptide to peptide, ranging from a few percent to 100%, and the average was almost 50%. Repeatability and linearity were satisfactory. In an examination of the cytoplasmic fraction of HeLa cells, more than 1000 phosphopeptides were identified using lactic acid-modified titania MOC and ␤-hydroxypropanoic acidmodified zirconia MOC, respectively. The overlap between phosphopeptides enriched by these two methods was 40%, and the combined results provided 1646 unique phosphopeptides. To our knowledge, this is the first successful application of a single MOC-based approach to phosphopeptide enrichment from complex biological samples such as cell lysates.

Molecular & Cellular Proteomics 6:1103-1109, 2007.
Phosphorylation is a key event in cellular signaling networks (1), and an understanding of proteome-wide phosphorylation/ dephosphorylation dynamics is important not only in relation to particular signal transduction pathways but also to obtain an overview of the whole network. MS is currently one of the most powerful techniques for proteome-wide experiments (2), but further improvements in protein identification efficiency are still needed (3). One of the key requirements for phosphoproteome analysis is to remove abundant non-phosphopeptides from complex mixtures, such as cell lysates, to detect low abundance phosphopeptides. Several approaches to enrich phosphopeptides prior to MS analysis, coupled with phosphate-specific MS acquisition techniques, have been reported. For phosphotyrosine-containing proteins, antibodies such as 4G10 and pY100 work adequately at both the protein and peptide levels (4,5). In addition, precursor ion scanning specific to phosphotyrosine has been applied to filter out non-phosphopeptides (4). On the other hand, a similar approach has not been effective for phosphoserine-and phosphothreonine-containing proteins, although it has helped to identify new players in the signaling pathway in some cases (6). As more general approaches for phosphopeptide enrichment, IMAC (7-9), IMAC with methyl esterification (10,11), strong cation exchange chromatography (12,13), and metal oxide chromatography (MOC) 1 using titania (14 -18), zirconia (19), and alumina (20) have been reported. In addition, precursor ion scanning in the negative mode (21), neutral losstriggered MS 3 (22,23), pseudo-MS 3 (24), and electron transfer dissociation (25) have been developed to detect phosphopeptides selectively. By combining these methods, it is currently possible to identify hundreds or thousands of phosphopeptides (8,10,12,13,22,23,26). Nevertheless current phosphopeptide enrichment methods are still not sufficiently specific (9), and samples generally contain large amounts of non-phosphopeptides even after enrichment. This causes ambiguity in peptide identification, although high accuracy measurement of precursor ions helps to reduce false positives (22). The situation at present is that improvement of phosphopeptide enrichment processes will lead directly to higher quality phosphoproteome analysis.
Titania was first applied as a chemoaffinity medium for organophosphates including phosphopeptides by Ikeguchi and Nakamura (14) and was subsequently made commercially available for proteomics applications (15)(16)(17). Zirconia also has affinity for phosphate (27) and has been used for phosphopeptide enrichment as well (19). Larsen et al. (18) showed that inclusion of o-hydroxybenzoic acid and its derivatives, such as 2,5-dihydroxybenzoic acid (DHB), in the sample loading buffer was effective to remove acidic non-phosphopeptides during phosphopeptide enrichment with titania for MALDI-MS analysis. However, this protocol is not directly applicable to LC-MS/MS analysis because residual DHB interferes with peptide detection, and the system becomes unstable because of precipitation of DHB around the orifice and in the LC system.
We developed new approaches for phosphopeptide enrichment using MOC modified with aliphatic hydroxy acids. The conditions were optimized for phosphopeptide enrichment from cell lysates followed by LC-MS/MS analysis to identify phosphopeptides.
Digestion of Standard Phosphoproteins-␣-Casein, fetuin, and phosvitin were individually reduced with DTT, alkylated with iodoacetamide, and digested with Lys-C followed by dilution and trypsin digestion as described previously (28). These digested samples were desalted using StageTips with C 18 Empore disk membranes (29). The eluates were mixed, and the peptide concentration was adjusted to 0.5 mg/ml with 0.1% TFA, 80% acetonitrile.
Enrichment of Phosphopeptides from Standard Phosphoprotein Digests-Custom-made MOC tips were prepared using C 8 StageTips and metal oxide bulk beads (3 mg of beads/200-l pipette tip) as described for strong cation exchange (beads)-C 18 tips (30). Prior to loading samples, the MOC tips were equilibrated with 0.1% TFA, 80% acetonitrile with hydroxy acids as selectivity enhancers (solution A). As the enhancers, glycolic acid, lactic acid, malic acid, tartaric acid, and DHB were used at a concentration of 300 mg/ml, and HPA was used at 100 mg/ml. The digested standard phosphoprotein mixture (15 l) was mixed with 100 l of solution A and loaded on the MOC tip. After successive washing with solution A and solution B (0.1% TFA, 80% acetonitrile), 0.5% ammonium hydroxide was used for elution. The eluted fraction was acidified with TFA, desalted using C 18 StageTips as described above and concentrated in a vacuum evaporator followed by the addition of solution A for subsequent nano-LC-MS/MS analysis.
Analysis of HeLa Cell Cytoplasmic Fraction-HeLa cells, cultured to 80% confluence in 15-cm diameter dishes, were homogenized with a Dounce homogenizer (10 strokes) after having been spiked with protein phosphatase inhibitor mixtures 1 and 2 (Sigma) and protease inhibitors (Sigma) in 100 mM Tris buffer (pH 8.0). The homogenate was centrifuged at 1500 ϫ g for 10 min, and the supernatant was digested and desalted as shown above. Titania or zirconia MOC tips were preconditioned with 20 l of solution A. The digested sample from a total of 200 g of HeLa cytoplasmic proteins was diluted with 100 l of solution A and loaded to the MOC tip. Washing was done with 20 l each of solutions A and B, and then phosphopeptides were eluted with 20 l of 0.5% ammonium hydroxide followed by acidification with 20 l of 1% TFA. Subsequent procedures were the same as described for the standard phosphoproteins.
Nano-LC-MS System-Nano-LC-MS/MS analyses were conducted by using a QSTAR system (QSTAR XL (AB/MDS-Sciex, Toronto, Canada), Agilent 1100 nanoflow pump (Waldbron, Germany), and HTC-PAL autosampler (CTC Analytics, Zwingen, Switzerland)) or an Orbitrap system (Finnigan LTQ-Orbitrap (Thermo Electron, Bremen, Germany), Dionex Ultimate3000 pump with FLM-3000 flow manager (Germering, Germany), and HTC-PAL autosampler). Repro-Sil C 18 materials (3 m, Dr. Maisch, Ammerbuch, Germany) were packed into a self-pulled needle (150-mm length ϫ 100-m inner diameter, 6-m opening) with a nitrogen-pressurized column loader cell (Nikkyo Technos, Tokyo, Japan) to prepare an analytical column needle with "stone-arch" frit (31). A polytetrafluoroethylene-coated column holder (Nikkyo Technos) was mounted on an x-y-z nanospray interface (Nikkyo Technos), and a Valco metal connector with a magnet was used to hold the column needle and to set the appropriate spray position. The injection volume was 5 l, and the flow rate was 500 nl/min. The mobile phases consisted of 0.5% acetic acid (A) and 0.5% acetic acid and 80% acetonitrile (B). A three-step linear gradient of 5-10% B in 5 min, 10 -40% B in 60 min, 40 -100% B in 5 min, and 100% B for 10 min was used throughout this study. A spray voltage of 2400 V was applied via the metal connector as described previously (31). The MS scan range was m/z 300 -1500, and the top three (QSTAR) or top six (LTQ-Orbitrap) precursor ions were selected for subsequent MS/MS scans. A lock mass function was used for the LTQ-Orbitrap to obtain constant mass accuracy during gradient analysis (32).
Database Searching-Mass Navigator version 1.2 (Mitsui Knowledge Industry, Tokyo, Japan) was used to create peak lists on the basis of the recorded fragmentation spectra. All parameters used in this process are described in Supplemental Table S1. Peptides and proteins were identified by means of automated database searching using Mascot version 2.1 (Matrix Science, London, UK) against Uni-Prot/Swiss-Prot release 9.0 (October 31, 2006) with a precursor mass tolerance of 50 ppm (QSTAR) or 3 ppm (LTQ-Orbitrap), a fragment ion mass tolerance of 0.25 Da (QSTAR) or 0.8 Da (LTQ-Orbitrap), taxonomy of human, and strict trypsin specificity (33) allowing for up to two missed cleavages. Carbamidomethylation of cysteine was set as a fixed modification, and oxidation of methionines, and phosphorylation of serine, threonine, and tyrosine were allowed as variable modifications. Note that neutral loss products from both precursor and fragment ions were considered for Mascot scoring in this phosphorylation modification setting, although no assignment was indicated for precursor-origin neutral loss peaks in the output results according to the supplier.
Peptides were considered identified if the Mascot score was over the 95% confidence limit based on the "identity" score of each peptide and at least three successive y-or b-ions with two and more y-, b-, and/or precursor-origin neutral loss ions were observed, based on an error-tolerant peptide sequence tag concept (34). A randomized decoy database created by a Mascot Perl program estimated a 4.01% false-positive rate for identified peptides within the criteria.
Phosphorylated sites were determined based on the difference in Mascot scores of peptides with different phosphorylated sites (delta score). If the delta score was more than 5, the top ranked phosphorylated site was considered determined. All MS/MS spectra with delta score Յ5 were manually inspected as to whether the phosphorylation sites were unambiguously determined or not. Note that phosphoryl-ation can be discriminated from sulfation by LTQ-Orbitrap with the lock mass function because the error distribution is within 2 ppm (32), which is less than the mass difference between sulfation and phosphorylation (9.516 mDa) for most of detectable peptides.

RESULTS AND DISCUSSION
Effect of Hydroxy Acids on Phosphopeptide Enrichment-A digested phosphoprotein standard mixture of ␣-casein, fetuin, and phosvitin was used to evaluate the effect of hydroxy acids in the loading buffer on selective enrichment of phosphopeptides with titania, zirconia, alumina, Al(OH) 3 ⅐xH 2 O, and boehmite MOC tips (Table I). All experiments were performed under acidic conditions in the presence of 0.1% trifluoroacetic acid and 80% acetonitrile because it has been reported that acidic conditions with 50 -80% acetonitrile were effective to reduce the co-purification of acidic peptides (8,10,18,19). Without hydroxy acids, eight or nine phosphopeptides were identified with titania, zir-conia, and Al(OH) 3 ⅐xH 2 O MOC tips, and the amounts of nonphosphorylated peptides were not negligible.
Larsen et al. (18) reported that DHB, which is commonly used as a matrix in MALDI-MS analysis, reduced the number of non-phosphorylated peptides in phosphopeptide enrichment with titania. In our study, DHB was also effective to decrease non-phosphorylated peptides in the cases of zirconia and Al(OH) 3 ⅐xH 2 O as well as titania. The number of phosphopeptides, however, also decreased as DHB was added to the loading buffer. This is because we used LC-ESI-MS in which the residual DHB was eluted in the same retention time range as phosphopeptides, and significant suppression of phosphopeptide ionization occurred. Furthermore DHB caused various problems with the LC-MS system, including column clogging, precipitation around the orifice, and decrease of the sensitivity after several runs. Therefore we decided to look for an alternative. Aliphatic ␣-hydroxycarboxylic acids are well known to interact specifically with metal oxides, such as titania and zirconia (35), probably through chelation to form a five-membered ring. As summarized in Table I, most of the hydroxy acids improved the selectivity of phosphopeptide enrichment by titania and zirconia MOC tips, although they were not effective for Al(OH) 3 ⅐xH 2 O MOC tips.
Among the hydroxy acids used, lactic acid and HPA were most effective with titania and zirconia MOC tips, respectively, efficiently suppressing the identification of non-phosphorylated peptides. Fig. 1 shows that no peaks from non-phosphopeptides were observed in the total ion chromatogram of LC-MS/MS analysis after phosphopeptide enrichment with titania MOC tips modified with lactic acid. Because these aliphatic hydroxy acids are hydrophilic enough to pass through the C 18 columns under the conditions used, they can be easily removed during the desalting step using C 18 StageTips. These aliphatic acids did not appear to impair the stability or the sensitivity of our LC-MS systems.
Hydroxy acid binds to metal oxide by forming a cyclic chelate (35,36), whereas phosphate anions do so by bridging two metal ions (37). Thus, hydroxy acids such as DHB and aliphatic hydroxycarboxylic acids should bind to metal oxides more weakly than a phosphate group but more strongly than the carboxylic groups of acidic non-phosphopeptides. Interestingly HPA was more effective than glycolic acid with both titania and zirconia in this study, whereas Tani and Ozawa (35) reported the opposite results in terms of affinity strength. This may be because there are various solid states of titania, and the retention properties depend strongly on the calcination temperature of the beads not only for hydroxy acids (38) but also for mono-and disaccharides (39). We also evaluated a mixture of glutamate, aspartate, and imidazole as additives in the loading buffer (20). However, we found that several nonphosphorylated peptides were always identified in the same standard sample loaded on titania and zirconia and Al(OH) 3 ⅐xH 2 O MOC tips, and no improvement was seen compared with aliphatic hydroxy acid-modified MOC tips. Recovery of Phosphopeptides after Enrichment with Titania MOC Tip-Recovery of phosphopeptides after enrichment with lactic acid-modified titania MOC tips was examined us-ing the peak areas in extracted ion current chromatograms of phosphopeptide-enriched and non-enriched samples (Table  II). Recovery of individual phosphopeptides ranged from a few Digested peptides from ␣-casein, fetuin, and phosvitin were analyzed by nano-LC-MS/MS with QSTAR before (a) or after the enrichment step (b) using lactic acid-modified titania MOC tips. Asterisklabeled peaks in b were identified as phosphorylated peptides by Mascot search. All unlabeled peaks in b are derived from singly charged ions, and the peaks with retention time Ͼ36 min were also observed in the blank. The MS spectrum at 33.6 min and MS/MS spectrum for the precursor ion of m/z 830.7 are shown in c and d, respectively. cps, counts/s. percent to around 100% (average, 49% for 13 phosphopeptides). On the other hand, recovery of all non-phosphorylated peptides was below 0.5% (data not shown). Similar results were obtained with HPA-modified zirconia MOC tips. These results indicated that this enrichment procedure is highly specific for phosphopeptides. Variable recovery was also observed in IMAC enrichment, and the average recovery was very low (less than 3% for 20 phosphopeptides and less than 10% for 10 phosphopeptides after methyl esterification) (9). A more efficient elution procedure from MOC tips, that would be independent of peptide amount and the sequence, is clearly needed. So far we have found that the recovery was slightly improved by successive elution with ammonium hydroxide followed by 80% acetonitrile. Further study is needed on this issue.
The values of relative standard deviation of triplicate analysis for 13 peptides (Table II) ranged from 9.1 to 81% (average, 24%). Linearity was also investigated in the range from 0.125 to 25 g of total loaded amount of digested peptides per titania MOC tip, with six data points, as shown in Supplemental Fig. S1. Five phosphopeptides with the smallest peak areas in Table II were selected to avoid saturation of the MS detector. The correlation coefficients of these five phosphopeptides ranged from 0.927 to 0.999, indicating that the method is satisfactory for quantitation.
Enrichment of Phosphopeptides in HeLa Cytoplasmic Extracts-There have been only a few reports of the application of phosphopeptide enrichment methods to real complex samples such as crude cell extracts including very recent large scale studies (40,41). In those studies, ϳ10 mg of starting materials were used for strong cation exchange chromatog-raphy prior to phosphopeptide enrichment. To our knowledge there has been no report describing the successful use of a single MOC-based method for phosphopeptide analysis of cell lysates using less than 1 mg as the starting material. It is an extremely challenging task to enrich phosphopeptides from real complex mixtures with a high dynamic range. Therefore we examined the applicability of aliphatic hydroxy acidmodified MOC tips for phosphopeptide enrichment from the cytoplasmic fraction of HeLa cells with less than a 1-mg amount. Fig. 2 shows the number of identified phosphopeptides and non-phosphopeptides following enrichment with three titania MOC tips, i.e. a lactic acid-modified tip, a DHBmodified tip, and a tip without modifier. The ratio of the  number of identified phosphopeptides to that of all identified peptides was very low (ratio ϭ 0.008) when no modifier was used. The use of DHB slightly improved the enrichment selectivity as well as the number of phosphopeptides (ratio ϭ 0.111). The lactic acid-modified titania tip gave the highest selectivity (ratio ϭ 0.607) as well as the largest number of identified phosphopeptides. It is noteworthy that ϳ350 phosphopeptides were obtained in a single 2-h nano-LC-MS/MS gradient run without repetition. Similar results were also obtained for an HPA-modified zirconia tip in comparison with a DHB-modified zirconia tip and a zirconia tip without any modifier. Table III compares the results obtained after enrichment with lactic acid-modified titania MOC and HPA-modified zirconia MOC tips. In total, 1100 and 1181 phosphopeptides were identified by four replicated experiments using titania and zirconia MOC tips, respectively (Table III and Supplemental Table S2). Of 1645 distinct identified phosphopeptides, 636 were found by both methods, and the other peptides were uniquely identified by a single method. Of 1629 phosphorylated sites, 1467 phosphorylation sites from 1502 peptides were unambiguously determined. On the other hand, the exact phosphorylated sites were unknown for the other 143 phosphorylated peptides because of the lack of sufficient fragmentation ions, although these peptides were unambiguously phosphorylated. Thus, complementary use of titania and zirconia MOC tips may allow a relatively comprehensive analysis of phosphopeptides. Note that zirconia did not show any tendency to enrich singly phosphorylated peptides in our final list, contrary to the finding of Kweon and Hakansson (19). Conclusion-We developed highly selective enrichment methods for phosphopeptides using lactic acid-modified titania and HPA-modified zirconia MOC tips. Both methods were applied to the cytoplasmic fraction of HeLa cells, and over a thousand phosphopeptides were identified. The combination of these methods is expected be useful for proteome-wide experiments to study cell signaling networks, which require the enrichment of many phosphopeptides from real complex mixtures with a wide dynamic range.

TABLE III Identification of phosphoproteins, phosphopeptides, and their phosphorylated sites in the cytoplasmic fraction of HeLa cells
Nano-LC-MS/MS was performed using the LTQ-Orbitrap. These data were obtained from four replicated experimental results with four nano-LC-MS/MS runs. The number of phosphorylated peptides was based on unique sequences containing missed cleavage products, oxidization of methionine, and phosphorylation of different sites. The number of phosphorylated sites was calculated from phosphopeptides with modification sites determined unambiguously. Note that 66 peptides of a total 1645 phosphopeptides were ambiguous for either phosphorylation or sulfation.