Identification of Novel Phosphorylation Sites on Xenopus laevis Aurora A and Analysis of Phosphopeptide Enrichment by Immobilized Metal-affinity Chromatography *

Mass spectrometric analysis of proteolytically derived phosphopeptides has developed into a widespread technique for the identification of phosphorylated amino acids. Using liquid chromatography-electrospray ionization tandem mass spectrometry, 14 phosphorylation sites were identified on Xenopus laevis His6-Aurora A, a highly conserved regulator of centrosome maturation and cell division. These included seven novel phosphorylation sites, Ser-12, Thr-21, Thr-103, Ser-116, Thr-122, Tyr-155, and Thr-294, as well as the previously identified regulatory sites, Ser-53, Thr-295, and Ser-349. The identification of these novel phosphorylation sites will be important for future studies aimed at elucidating the mechanisms of Aurora A regulation by phosphorylation. Furthermore, we demonstrate that a “kinase-inactive” mutant of Aurora A, K169R, still retains 10% of activity of the wild-type enzyme in vitro along with occupancy of Thr-295 and Ser-12. However, mutation of Asp-281 to Ala completely abolishes activity of the enzyme and should therefore be used preferentially as a genuine kinase-dead construct. Because of the abundance of phosphorylated residues on His6-Aurora A, we found this protein to be an ideal tool for the characterization of immobilized metal-affinity chromatography (IMAC) as a method for phosphopeptide enrichment from complex mixtures. We present a detailed analysis of the binding and elution properties of both the phosphopeptides and unphosphorylated peptides of His6-Aurora A to Fe3+-IMAC before and after methyl esterification. Moreover, we demonstrate a significant difference in enrichment of phosphopeptides when different resins are used for Fe3+-IMAC and characterize the strengths and limitations of this methodology for the study of phosphoproteomics.

Reversible protein phosphorylation is one of the most important intracellular signal transduction events and plays a key role in regulating most aspects of cellular homeostasis, including cell proliferation, differentiation, and apoptosis (1)(2)(3)(4). It has been estimated that approximately one-third of mammalian proteins contain covalently bound phosphate, and many are subject to regulation by multisite phosphorylation (5,6).
Aurora A has been implicated in early mitotic events such as the centrosome cycle. It becomes phosphorylated and activated during mitosis and is also overexpressed in many human cancers (7). The regulation of Aurora A function is likely to be complex, and phosphorylation-dependent mechanisms for degradation are also thought to exist (8 -12). Multiply phosphorylated forms are generated during Xenopus oocyte meiotic maturation in response to progesterone (13,14), and overexpression of Aurora A in either Sf9 cells or Escherichia coli results in a form that is phosphorylated to high stoichiometry (15). One previously identified site of phosphorylation on Aurora A is the conserved T-loop residue Thr-295, which lies in the consensus RRXpT. This motif is phosphorylated in vitro by a plethora of protein kinases, including protein kinase A (9,15). Phosphorylation of this residue is reported to greatly enhance the phosphotransferase activity of this enzyme, and treatment of Thr-295-phosphorylated Aurora A with protein phosphatases abolishes activity, indicating that phosphorylation of this site is essential for kinase activity (9,15,16). More recently, mutagenesis studies identified Ser-53 (located in the non-catalytic region of the protein) as a regulatory amino acid for the Cdh1-mediated destruction of Aurora A at the end of mitosis in Xenopus (12), and phosphorylation of this residue during mitosis has recently been confirmed (17). In view of the oncogenic nature of Aurora A and its marked ability to induce polyploidy and centrosome dysregulation when overexpressed, often in the perceived absence of kinase activity (18), the identification and characterization of Aurora A autophosphorylation sites are essential for full understanding of the regulation of its kinase activity.
Although MS 1 is a sensitive analytical tool for the identification of many post-translational modifications, the detection of phosphopeptides in a complex peptide mixture can be suppressed due to the electronegativity of the phosphoryl groups, resulting in low ionization efficiency (6,19,20). Analysis of phosphorylation sites in complex mixtures can therefore be facilitated by affinity methods, which isolate and enrich phosphopeptides from proteolytic extracts prior to sequencing by MS/MS. A method that is rapid, straightforward, and relatively inexpensive involves enriching phosphopeptides using trivalent metal ions, primarily Fe 3ϩ (21)(22)(23)(24)(25), a technique referred to as immobilized metal-affinity chromatography (IMAC). However, reduced efficacy may occur with the IMAC procedure due to nonspecific binding of unphosphorylated peptides (particularly those containing high levels of glutamic and/or aspartic acid residues) (26). The nonspecific binding of unphosphorylated peptides via their carboxyl groups has recently been addressed by methyl esterification, which was reported to reduce absorption of unphosphorylated peptides by at least two orders of magnitude and thereby to enhance binding specificity (27).
Given the importance of phosphorylation site identification, we undertook to fully characterize the specificity of the binding and elution of peptides to IMAC with the aim of phosphopeptide enrichment from a complex mixture, detailed analysis of which has not been published previously. Having identified 14 phosphorylation sites on His 6 -Aurora A (from 20 tryptic phosphopeptides), we used this as a model system with which to analyze the specificity of both the binding and the elution of the phosphopeptides and the unphosphorylated peptides to two different Fe 3ϩ -chelated resins, POROS 20 MC and iminodiacetic acid-agarose (IDA-agarose), both before and after methyl esterification.

EXPERIMENTAL PROCEDURES
Cloning and Purification of Aurora A-The Xenopus protein Eg2 (subsequently referred to as Aurora A (7, 13)) was cloned from a Stage VI Xenopus cDNA library by PCR using the following primers: GAG-GACGACAAGATGGAGCGGGCTGTTAAG and GAGGAGAAGCCCG-GTCTATTGGGCGCCTGGAAG. The ligation-independent cloningready ("LIC-ready") PCR product was introduced into the bacterial expression vector pET-30 (Novagen) using the ligation-independent cloning method and sequenced over the entire coding region. Recombinant Aurora A was expressed in E. coli strain BL21(DE3) using standard techniques. Bacteria were grown to mid-log phase and induced for 16 h at room temperature with 200 M isopropyl-1-thio-␤-D-galactopyranoside. Recombinant Aurora A containing an N-terminal His 6 tag within 43 additional amino acids was purified from 2 liters of bacteria using Talon resin (Clontech). The purified protein, which was Ͼ95% pure as determined by SDS-PAGE using the Ander-son technique (14) and Coomassie Blue staining, was dialyzed into 50 mM ammonium bicarbonate buffer prior to assay or trypsinization. Approximately 2 mg of pure Aurora A was obtained per liter of bacteria.
Site-directed Mutagenesis-The mutation of Lys-169 to Arg, Asp-281 to Ala, or Thr-295 to Val was carried out using the QuikChange protocol (Stratagene). The mutated cDNAs were sequenced over the entire coding region to verify the sequence. pET-30 containing Aurora A mutants was transfected into BL21(DE3) cells and expressed and purified as wild-type protein. The mutant proteins were expressed at identical levels to wild-type Aurora A.
Determination of Aurora A Activity in Vitro-Purified recombinant Aurora A was assayed by analyzing the rate of phosphorylation of myelin basic protein (MBP). 500 ng of enzyme was incubated for 20 1 The abbreviations used are: MS, mass spectrometry; ESI, electrospray ionization; LC, liquid chromatography; IDA, information-dependent acquisition; IDA-agarose, iminodiacetic acid-agarose; IMAC, immobilized metal-affinity chromatography; MS/MS, tandem mass spectrometry; MBP, myelin basic protein; PP, protein phosphatase; TIC, total ion chromatogram.
FIG. 1. Xenopus Aurora A activity is regulated by phosphorylation. A, purified, recombinant Aurora A (4 g), treated with (ϩ) or without (Ϫ) 1 unit of PP2A catalytic subunit (PP2Ac) or 1 unit of Xenopus PP1␥, as indicated, was denatured in 2% (w/v) SDS, analyzed by SDS-PAGE on an Anderson gel, and stained for 1 h with Coomassie Brilliant Blue. The positions of phosphorylated (pAurora A) and unphosphorylated (Aurora A) Aurora A and the molecular mass markers bovine serum albumin (66 kDa) and ovalbumin (43 kDa) are indicated. B, untreated or phosphatase-treated Aurora A (500 ng) was assayed for activity using the substrate MBP (see "Experimental Procedures"). [ 32 P]Phosphate incorporation into MBP was detected by autoradiography following SDS-PAGE. C, Aurora A from reactions identical to panel B was immunoblotted with purified rabbit antibodies recognizing only the phosphoThr-295 form of Aurora A (top) or all forms of Aurora A (bottom). Similar results were seen in a separate experiment. min at 30°C with 0.5 mg/ml MBP in 50 mM Tris, pH 7.4, 0.1 mM EGTA, 0.1 mg/ml bovine serum albumin, 0.1% 2-mercaptoethanol, 0.01% Brij-35, 200 nM okadaic acid, 10 mM MgCl 2 , and 100 M [␥-32 P]ATP (specific activity 500 cpm/pmol). Incorporation of 32 P was visualized by SDS-PAGE and autoradiography.
Dephosphorylation of Aurora A in Vitro-10 g of recombinant wild-type Aurora A was incubated with 1 unit of either PP2A catalytic subunit (Upstate Biotechnology) or Xenopus PP1␥ (overexpressed and purified from E. coli) in 50 mM Tris, pH 7.4, 1 mM MgCl 2 , 0.1 mM MnCl 2 , 0.1 mg/ml bovine serum albumin, 60 mM 2-mercaptoethanol for 60 min. PP2A or PP1␥ was then irreversibly inhibited with 1 M okadaic acid and 1 M microcystin-LR, and Aurora A was assayed for activity against MBP as described above.
Western Blotting of Aurora A-A polyclonal antibody was raised in rabbit against recombinant bacterial Aurora A and affinity-purified on a column of Aurora A coupled to Sepharose beads using standard techniques. Affinity-purified antisera were used at a dilution of 1:5000 to detect recombinant Aurora A. This antibody detects both inactive and active (i.e. Thr-295-phosphorylated) protein. A comparison of the amino acids surrounding Thr-295 of both human and Xenopus Aurora A showed 100% sequence conservation in this region. We therefore used a rabbit polyclonal antibody raised against human phosphoThr-295 (Cell Signaling) (9) to analyze the phosphorylation of this residue in Xenopus Aurora A before and after phosphatase treatment.
Digestion and Methyl Esterification of Aurora A-Purified proteins in 50 mM ammonium bicarbonate were reduced using 4 mM dithiothreitol (15 min, 60°C) and then alkylated with 14 mM iodoacetamide (30 min, room temperature). Further alkylation was quenched by addition of dithiothreitol to a final concentration of 7 mM prior to digestion with 3% (w/w) trypsin (Promega) (37°C, 4 h). Lyophilized samples were methyl-esterified with 2 M methanolic HCl at 0.5 mg/ml for 30 min at room temperature (27). The reaction was quenched by addition of 2 volumes of water and was lyophilized. The resulting peptide methyl esters were dissolved in 5% (w/v) ammonium acetate (pH 9) to stabilize the pH at ϳ7. or IDA-agarose (Pierce), respectively, and loaded with 150 l of 0.2 M FeCl 3 . The columns were washed with 0.1 M NaCl prior to equilibration with 10 column volumes of either 25:74:1 acetonitrile:water:acetic acid at pH 3.0 or 5% (w/v) ammonium acetate, 5% (v/v) acetonitrile at pH 7.5. Aurora A tryptic peptides (40 pmol) (before or after methyl esterification) were applied to the columns in the relevant loading buffer. To remove nonspecific binding peptides, the columns were washed with 10 column volumes of loading buffer containing 0.1 M NaCl. Bound peptides were eluted with 5 volumes of 250 mM NaH 2 PO 4 , pH 4.2 or 250 mM Na 2 HPO 4 , pH 9.0.
Peptide Identification by LC-ESI-MS/MS-Aurora A tryptic peptides (20 pmol) (before or after methyl esterification) were acidified by addition of formic acid to 10% (w/v) and loaded onto a 20 cm ϫ 250 m inner diameter fused silica column packed with Jupiter C 18 resin (Phenomenex), which was directly coupled to the electrospray ionization interface of a QSTAR Pulsar (Applied Biosystems). Bound peptides were eluted with a gradient of 0 -40% (v/v) acetonitrile over 40 min and then 40 -80% (v/v) acetonitrile over 5 min in 0.1% (w/v) formic acid. Information-dependent acquisition (IDA) was used to acquire MS/MS data with experiments designed such that the three most abundant peptides were subject to collision-induced dissociation, using argon as the collision gas, every 15 s. Collision energies were varied as a function of the m/z and charge state of each peptide. To avoid continued MS/MS of peptides that had already undergone collision-induced dissociation, dynamic exclusion was incorporated for a further 45 s. IDA analyses were performed on tryptic digests of Aurora A before and after methyl esterification. Data from duplicate IDA experiments were examined under each experimental condition (methyl-esterified versus unmodified tryptic digests).
In all cases, the data from the IDA experiments was searched twice against His 6 -Aurora using MASCOT (28) on an internal server, first with a requirement for peptides to contain tryptic sites at both ends and then with no requirement for tryptic cleavages. The mass tolerance of both the precursor peptide ion and the MS/MS fragment ions was set at Ϯ0.1 Da, and carbamidomethyl cysteine was specified as a static modification. For methyl-esterified peptides, the data was searched specifying methyl esterification (ϩ14 Da) as both a static and a variable modification at the peptide C terminus as well as at aspartic and glutamic acid residues. Phosphorylated serine, threo-nine, and tyrosine were specified as variable modifications. All MS/MS spectra were manually checked to verify the sequence assignments.
Quantification by LC-ESI-MS-Aurora A tryptic peptides (20 pmol) (before or after methyl esterification) were resolved by reverse-phase chromatography as described; and LC-ESI-MS spectra were acquired. Having characterized the elution times and masses for each of the peptides identified by MS/MS, extract ion chromatograms were constructed by plotting ion signal intensity for each possible charge state of the relevant peptides. The area under the extract ion chromatogram curve was calculated for each charge form, and the areas were summed to yield measurements of peptide intensity (I). LC-ESI-MS experiments were performed on samples (i) prior to loading on IMAC resin ("total"), (ii) from the combined flow-through and wash, and (iii) from the IMAC eluate. To quantify binding to IMAC resins, peptide intensity in the flow-through ϩ wash (I FTW ) was quantified, subtracted from the peptide intensity in the total sample (I total ), and normalized to the intensity in the sample prior to loading on IMAC.
percentage "Binding" ϭ ((I total Ϫ I FTW )/I total ) ϫ 100% (Eq. 1) To quantify recovery after binding, peptide intensity in the eluate (I eluate ) was normalized to the intensity in the total sample.

RESULTS
Xenopus Aurora A Is Hyperphosphorylated-Recombinant His 6 -Aurora A expressed and purified from E. coli is hyperphosphorylated, evidenced by its gel mobility retardation on SDS-PAGE and the reversal of retardation by treatment with the protein phosphatase 1␥ or 2A (PP1␥ or PP2A) (Fig. 1A). Hyperphosphorylation correlates with activity because phosphatase treatment abolishes the ability of Aurora A to phosphorylate its substrate MBP (Fig. 1B). Furthermore, immunoreactivity of Aurora A with the anti-pThr-295 antibody is only observed prior to phosphatase treatment (Fig. 1C) in agreement with previous studies identifying this site in regulating kinase activity (9). The established importance of phosphorylation for regulating the function of this kinase (9,14) and its crucial role in the cell cycle make the identification of novel regulatory phosphorylation sites valuable.
Identification of Phosphorylation Sites on Aurora A-LC-ESI-MS/MS analysis identified 14 phosphorylation sites on His 6 -Aurora A, four of which were present in the N-terminal His tag ( Fig. 2A and Table I). As an example of the data quality, the MS/MS spectrum of the 2ϩ ion of the doubly phosphorylated peptide Tp10 -13 is depicted in Fig. 3. The phosphorylation sites were unambiguously assigned to Ser-116 and Thr-122 due to mass assignments of phosphorylated or ␤-eliminated b 3 , b 5 , b 6 , and b 8 and doubly ␤-eliminated b 9 , b 10 , b 11 , b 12 , and b 13 fragment ions. Of the 10 phosphorylation sites identified in Xenopus laevis Aurora A, seven are previously unreported (Ser-12, Thr-21, Thr-103, Ser-116, Thr-122, Tyr-155, and Thr-294). Thr-295, which was also identified in these studies, was known to be essential for activity (9,16), and while this paper was under review, Littlepage et al. identified Ser-349 as a phosphorylation site important for either the structure or the regulation of Aurora A (17). This group also confirmed phosphorylation of Ser-53 in vivo, a site which the authors had identified previously as being important for Aurora A degradation (12). Phosphorylation of both Ser-53 (Fig.  3) and Ser-349 was confirmed in the experiments presented here ( Fig. 2A and Table I). Of the 10 identified phosphorylated sites, six are located in the non-catalytic N-terminal region of Aurora A, which functions in centrosome binding (29), and four are located in the catalytic domain ( Fig. 2A). The stoichiometry of phosphorylation of each of the identified sites, calculated by comparing the relative levels of the phosphopeptide versus the unphosphorylated form of the same peptide, is depicted in Table I. Near stoichiometric phosphorylation occurs on Ser-12, Ser-116, Thr-295, and Ser-349, indicating that phosphorylation of these residues is likely to be important in the regulation or function of Aurora.
Quantification of Phosphorylation Sites on Aurora A Mutants-Three Aurora A mutants (K169R, D281A, and T295V) were expressed and purified in an identical manner to wildtype protein (Fig. 4A), and their activity toward MBP was assessed (Fig. 4B). Surprisingly, the K169R mutant, which disrupts a conserved ion pair interaction in the nucleotide binding domain, still possessed 10% of the activity of wildtype enzyme (Fig. 4B). However, the mutation of Asp-281 to Ala, which chelates the magnesium ion required for catalytic activity, completely abolished kinase activity in vitro. The mutation of Thr-295 to Val, which disrupts the known activating phosphorylation site, also abolished activity (Fig. 4B). The phosphorylation state of Thr-295 was also assessed in these mutants by immunoblotting. As expected, the inactive D281A and T295V mutants were not phosphorylated at this residue (Fig. 4B). In agreement with the finding that K169R had residual kinase activity, Thr-295, a residue required for activity, was phosphorylated in this mutant (Fig. 4C). The degree of phosphorylation at this residue appeared similar to wild-type Au- rora A by immunoblotting, indicating that phosphorylation at this site is not reduced in proportion to activity. Other factors must therefore play a role in the activity of Aurora A.
Given that the K169R mutant has diminished activity and is phosphorylated on Thr-295, we examined the effects of this mutation at the other phosphorylation sites. None of the phosphopeptides identified in the wild-type protein were detected in the two inactive mutants of Aurora A, D281A and T295V, as determined by LC-ESI-MS. However, the partially active K169R mutant that is phosphorylated at Thr-295 (Fig. 4, B and C) also showed 72% abundance of phosphopeptide Tp3 (Table II) (normalized to the unphosphorylated form of the same peptide), reflecting significant occupancy at Ser-12 (data not shown). Masses for the other phosphopeptides were not detected at a significant level in this mutant.
Together these findings indicate that the sites identified on Aurora A are all autophosphorylation events. The finding that Ser-12 is phosphorylated in the K169R mutant suggests that this site, in addition to Thr-295, is likely to be preferentially autophosphorylated. The phosphorylation sites identified in Aurora A fall into two major categories; Ser-12, Ser-116, Thr-294, Thr-295, and Ser-349 each lie in a putative consensus sequence for phosphorylation by Aurora A ((K/R)X(S/T)X, where X is any amino acid) (30). The other five residues (Thr-21, Ser-53, Thr-103, Thr-122, and Tyr-155) deviate from the consensus sequence but are still likely to be sites of autophosphorylation because they are absent in inactive mutants of Aurora A. Previously identified Aurora A phosphorylation sequences lack these consensus motifs in some cases, including the Aurora A-regulated mediator of mRNA translation, cytoplasmic polyadenylation element-binding protein (CPEB) (31,32).
Methyl Esterification of Aurora A Tryptic Peptides-Twenty tryptic phosphopeptides and 42 unphosphorylated peptides were identified by MS/MS in Aurora A (Table II, part A), making this an ideal tool with which to perform a comprehensive analysis of IMAC for phosphopeptide enrichment. We therefore used this complex peptide mixture to examine the characteristics of binding and elution of a range of phosphorylated and unphosphorylated tryptic peptides.
To avoid nonspecific binding to the IMAC column by carboxylate anions in the peptide digest, a methyl esterification procedure based on a recent study by Ficarro et al. was employed (27). We found that methyl esterification of the Aurora A tryptic peptides was complete when carried out at room temperature for 30 min (as determined by searching FIG. 4. Analysis of Xenopus Aurora A mutants. A, purified, recombinant wildtype (WT) Aurora A (4 g) or identically purified mutant enzymes K169R, D281A, and T295V were denatured in 2% (w/v) SDS and analyzed by SDS-PAGE. The gel was then stained for 1 h with Coomassie Brilliant Blue to determine the electrophoretic mobility of the proteins. B, the specific activity of each mutant was determined by assaying the rate of phosphorylation of MBP. WT or mutant enzyme (500 ng) was assayed in duplicate as described under "Experimental Procedures" using 100 M ATP containing [␥-32 P]ATP at a specific activity of 500 cpm/pmol of total ATP. Phosphorylation was visualized by autoradiography and quantified by Cerenkov counting of the duplicate MBP bands. C, the Aurora A mutants (4 ng) were separated as in panel A, and the position of Aurora A was determined by duplicate Western blotting with an antibody that recognizes Aurora A only when phosphorylated at Thr-295 (top) or an antibody that recognizes Aurora A in either its phosphorylated or unphosphorylated forms (bottom). Similar results were obtained in three other experiments.

TABLE II Quantification of peptide binding and recovery from Fe 3ϩ -POROS 20 MC
Samples were prepared and treated as described in the legend to Fig. 5 and under "Experimental Procedures." Peptide ion intensities were quantified, summed over the different charge forms of the same peptide, and normalized to sample load on the Fe 3ϩ -POROS 20 MC resin. Indicated are the percentages of binding for each peptide ("Binding") (average and standard deviations for 3 independent IMAC experiments) and the percentage recovery for each peptide ("Recovery") (n ϭ 1). Analyses included peptides prior to methyl esterification and peptides after methyl esterification. Negative values for "Binding" have been set to zero.
with methyl ester (C-terminal) and methyl ester (Asp/Glu) as variable modifications, data not shown), and these conditions were subsequently employed. However, complete analysis of Aurora A tryptic peptides before and after methyl esterification under these conditions revealed the loss of previously identified peptides following modification, resulting in a reduction of the total protein coverage from 78.2 to 61.1% (Fig.  2). The reason for the loss of peptides is currently unclear. Fig. 5A shows the total ion chromatogram (TIC) of methyl-esterified tryptic peptides measured during LC-ESI-MS. Following chromatography on Fe 3ϩ -chelated POROS 20 MC resin, peptides in the flow-through and wash (combined) (Fig. 5B) and in the eluate (Fig. 5C) were quantified as described under "Experimental Procedures." Binding was performed under standard conditions at pH 3.0 (27). Alternatively, binding was performed at pH 7.5 to test whether increased negative charge on the phosphate moiety may enhance adsorption of methyl-esterified phosphopeptides. Bound peptides were subsequently eluted with phosphate buffer at pH 4.2 or 9.0. To ascertain the effect of carboxyl group modification on the binding and elution of peptides from Fe 3ϩchelated resin, the same procedure was repeated with Aurora A peptides obtained prior to methyl esterification.

Analysis of IMAC for the Enrichment of Aurora A Phosphopeptides-
The average binding of both phosphorylated and unphosphorylated peptides to the POROS 20 MC column at pH 3.0 increased after methyl esterification (compare parts A and B of Table II), suggesting that neutralization of the carboxyl groups does not increase the specificity of phosphopeptide binding with peptide digests. In the absence of methyl esterification, there was a marginal increase in binding of unphosphorylated peptides to POROS 20 MC when the pH was raised from 3.0 to 7.5 (from an average of 49 -57%), as would be expected. However, there was no difference in the binding of phosphopeptides under these conditions (Table II, part A), contrary to previous reports with synthetic peptides (26). Conversely, after methyl esterification both phosphorylated and unphosphorylated peptides exhibited a moderate decrease in binding when the pH was raised from 3.0 to 7.5 (86 -63% for phosphopeptides and 72-44% for unphosphorylated peptides; Table II, part B). Nevertheless, it was surprising to note the relatively high degree of nonspecific binding at both pH 3.0 and pH 7.5, even after methyl esterification of the carboxylate groups in the peptides. Increasing the acetonitrile concentration to 25% (v/v) had no significant effect in reducing nonspecific binding to the column (data not shown). Further- more, when wash steps were performed in the presence of high molarity salt (1 M NaCl) in an attempt to reduce ionic interactions, no increase in the specificity of phosphopeptide binding was observed (data not shown).
Interestingly, the elution of methyl-esterified phosphopeptides showed bias toward the pH at which binding was performed (Table II, part B). When peptides were bound to Fe 3ϩchelated POROS 20 MC at pH 7.5, 6 of 10 phosphopeptides showed higher recovery than when the peptides were bound at pH 3.0. Elution at pH 4.2 showed increased recovery over peptides eluted at pH 9.0. These experiments demonstrate that optimal recovery of phosphopeptides from Fe 3ϩ -chelated POROS 20 MC occurred following binding at pH 7.5 and elution at pH 4.2. Overall, an average of 25% of phosphopeptide ion intensity was recovered in methyl-esterified digests when binding was performed at pH 7.5 and elution at pH 4.2, whereas recovered phosphopeptide ion intensity decreased to 9% when binding and elution were performed under standard conditions (binding at pH 3.0, elution at pH 9.0) (27). In contrast, recovery of ion intensities for both methyl-esterified and unmod- FIG. 5. Total ion chromatogram of tryptic methyl-esterified Aurora peptides before and after separation by Fe 3؉ -POROS 20 MC. Aurora A was reduced, alkylated, and digested with trypsin prior to methyl esterification as described under "Experimental Procedures." Lyophilized peptides were resuspended in 5% (w/v) ammonium acetate (pH 7.5), 5% (v/v) acetonitrile, and 40 pmol were loaded onto Fe 3ϩ -chelated POROS 20 MC resin. The flowthrough and wash were collected together, and bound peptides were eluted in 250 mM NaH 2 PO 4 , pH 4.2. Indicated is the TIC of (A) 20 pmol of tryptic, methylesterified peptides, (B) 50% of the combined flow-through ϩ wash from IMAC (equivalent to 20 pmol), and (C) 50% of the IMAC eluate, which was chromatographed by reverse-phase high performance liquid chromatography arranged in line with a QSTAR Pulsar in ESI mode set to collect LC-MS data. Peaks corresponding to the eluted phosphopeptides are annotated (C).
ified unphosphorylated peptides was significantly lower, averaging less than 1% (Table II, parts A and B). This is probably due to irreversible adsorption of the peptides to the column. Despite the low selectivity between binding of phosphorylated versus unphosphorylated peptides, methyl esterification increased the selective recovery of phosphopeptides by 2-3-fold.

TABLE III
Quantification of peptide binding and recovery from Fe 3ϩ -IDA-agarose Samples were prepared and treated as described in the legend to Fig. 5 and under "Experimental Procedures." Peptide ion intensities were quantified, summed over the different charge forms of the same peptide, and normalized to sample loaded on the Fe 3ϩ -IDA-agarose resin. Indicated are the percentages of binding for each peptide ("Binding") and the percentage recovery for each peptide ("Recovery"). Similar results were observed in duplicate experiments; the results of a single experiment are reported. Analyses included peptides prior to methyl esterification and peptides after methyl esterification. Negative values for "Binding" have been set to zero. We next repeated these experiments using Fe 3ϩ -IDA-agarose to ascertain whether varying the resin and associated chelating group would have any effect on the IMAC procedure. The results using the IDA-agarose showed a noticeable improvement. Under standard conditions, greater selectivity for binding phosphopeptides over unphosphorylated peptides was observed whether or not peptides were methylesterified (Table III, parts A and B). After methyl esterification, average recovery of the phosphopeptides was 39% compared with 3% recovery of unphosphorylated peptides. Without methyl esterification, only 16% phosphopeptide recovery was observed versus 2% recovery of unphosphorylated peptides. The recovery of methyl-esterified phosphopeptides was somewhat lower under conditions of binding at pH 7.5 and elution at pH 4.2 compared with standard conditions, whereas unmethylated peptides showed little difference in recovery between conditions for standard binding versus pH 7.5. Overall, use of Fe 3ϩ -IDA-agarose revealed that methyl esterification enhanced the selectivity of binding and recovery for Aurora A phosphopeptides (Table III, part B). DISCUSSION In this paper, a combination of chemical, biochemical, and molecular biological analyses were used to gain a compre-hensive understanding of the phosphorylation status of Xenopus Aurora A and subsequent enrichment of phosphopeptides from this protein by IMAC. We describe 14 phosphorylation sites on His 6 -Aurora A, seven of which are novel sites within the enzyme. Furthermore, we show that four of the sites in Aurora A undergo near stoichiometric phosphorylation and that site occupancy is eliminated by mutating residues that eliminate catalytic activity. This indicates that occupancy of these sites reflects autophosphorylation events either through cis or trans mechanisms. The K169R mutant of Aurora A, which retains 10% of activity of the wild-type enzyme (Fig. 4B), autophosphorylates at Ser-12 and Thr-295 but not at any of the other identified phosphorylation sites, suggesting that phosphorylation of Thr-295 and one or more of these sites may synergistically contribute to full activation. It appears likely that mutation of Lys-169 to Arg does not fully abolish ATP binding in this mutant, thus compromising its use as a dominant-negative "kinase-dead" enzyme in reports of proposed kinase-independent physiological roles of Aurora A, including the generation of tetraploidy and subsequent centrosome amplification in tissue culture cells (12,18). However, if Lys-169 is mutated to Met, there is no detectable kinase activity (33), and overexpression of this mutant of Aurora A does not induce cell cycle abnormalities or polyploidy (34). This indicates that the generation of tetraploidy following overexpression of the Lys-169 to Arg mutant may be due to aberrant kinase activity (33,35). In confirmation of this theory, it will therefore be important to assess D281A Aurora A for its ability to generate tetraploidy and centrosome amplification.
We also demonstrate that mutation of Thr-295 to Val abolishes activity in vitro (Fig. 4B). Conceivably, Thr-295 autophosphorylation is necessary although perhaps not sufficient for complete activation of Aurora A. These findings reveal greater complexity in the mechanisms of Aurora A regulation than understood previously. Ongoing studies will help to clarify the precise function of each novel phosphorylation site.
To our knowledge, quantitative analysis of the binding and recovery of peptides to IMAC from a complex mixture had not been published previously. We therefore used hyperphosphorylated His 6 -Aurora A as a model with which to test the selectivity of phosphopeptide binding and recovery to two different IMAC resins to monitor the behavior of unphosphorylated peptides under the same conditions and to test the efficacy of methyl esterification in improving the IMAC protocol. Although these studies were performed using 20 pmol of protein, it is unlikely that a given phosphoprotein in a proteomics sample would be present at these quantities. Due to the incomplete recovery of phosphopeptides following IMAC under these conditions, it is therefore unlikely that low stoichiometrically phosphorylated peptides, particularly those from low abundance proteins, would be identified.
First, our findings indicate that methyl esterification leads to the unexplained loss of both phosphorylated and unphosphorylated peptides such that the overall protein coverage decreased from 78 to 61%. We cannot explain this currently, but note that the complexity of the TIC for methyl-esterified peptides eluting from the IMAC resin exceeds the number of peptides identified (Fig. 5C). This suggests that alternative chemistries are occurring during the methyl esterification reaction, which are unaccounted for. Second, of the peptides that were observed, methyl esterification does not substantially alter the binding of phosphorylated versus unphosphorylated peptides to the IMAC column as determined by quantification of the peptides in the combined flow-through ϩ wash. However, the recovery of bound phosphopeptides under these conditions is higher than the recovery of unphosphorylated peptides, possibly due to irreversible adsorption of unphosphorylated peptides to the resin. Methyl esterification increases the recovery of bound phosphopeptides by 2-3fold using POROS 20 MC resin. Third, Fe 3ϩ -IDA-agarose shows greater selectivity in binding phosphopeptides versus unphosphorylated peptides compared with Fe 3ϩ -POROS 20 MC. In addition, the total recovery of phosphopeptides is higher with IDA-agarose. Fourth, we observe differences in the optimal pH for binding/elution, which appears best at pH 7.5/pH 4.2 for POROS 20 MC and pH 3/pH 9 for IDA-agarose.
The anomaly in pH-dependent binding of phosphopeptides with that previously published cannot be explained by the lower concentration of NaCl generally used in our experiments (26) as increasing the salt concentration had no effect. However, the differing properties of the trivalent metal ions used for chelation in some studies (Fe 3ϩ versus Ga 3ϩ ) (23) may play a role, although this was not tested in these studies. Finally, we note that the average binding of phosphopeptides to IMAC at pH 3.0 increases following methyl esterification of the carboxyl groups, and although binding of unphosphorylated peptides to Fe 3ϩ -POROS 20 MC increases following modification, binding to Fe 3ϩ -IDA-agarose is essentially unchanged. However, we did not observe significantly decreased adsorption of unphosphorylated peptides as reported by Ficarro et al. (27). Our findings suggest that the nonspecific peptide binding that occurs at pH 3.0 is not primarily due to the carboxyl groups. Interestingly, although preferential binding of phosphopeptides is not promoted by modification of the carboxyl groups, phosphopeptide recovery is preferential after methyl esterification, indicating that this modification increases the specificity of elution with phosphate buffer. Furthermore, it was noted that peptides containing two phosphorylated residues had a higher percentage recovery during elution under the majority of conditions tested compared with identical peptides containing only a single phosphorylated residue. This probably explains the high abundance of multiply phosphorylated peptides identified in previous publications using IMAC (27,36,37). Overall, the best conditions for IMAC enrichment of phosphopeptides involve methyl esterification of digests, binding to Fe 3ϩ -IDAagarose at pH 3.0, and elution at pH 9.0, although this is balanced by an unexplained loss of peptide representation in methyl-esterified reactions.
With caution, we conclude that phosphopeptide enrichment and identification appear feasible by IMAC when sampling complex peptide mixtures, although phosphoproteomic studies need to be approached with care. Given the low stoichiometry of many post-translational modifications, including phosphorylation, and the low copy number of many signaling molecules, in many cases it will be preferable to analyze phosphopeptides in mixtures using non-IMAC methods, including direct analysis of protein digests, precursor ion scanning, (38) or neutral loss scanning (39). If required, quantification of these phosphopeptides within their mixtures can subsequently be achieved using "absolute quantification," a strategy termed AQUA (40). * This work was supported in part by National Institutes of Health Grants GM48521 (to N. G. A.) and CA87648 (to K. A. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.