Optimization of Immobilized Gallium (III) Ion Affinity Chromatography for Selective Binding and Recovery of Phosphopeptides from Protein Digests

Abstract

Although widely used in proteomics research for the selective enrichment of phosphopeptides from protein digests, immobilized metal-ion affinity chromatography (IMAC) often suffers from low specificity and differential recovery of peptides carrying different numbers of phosphate groups. By systematically evaluating and optimizing different loading, washing, and elution conditions, we have developed an efficient and highly selective procedure for the enrichment of phosphopeptides using a commercially available gallium(III)-IMAC column (PhosphoProfile, Sigma). Phosphopeptide enrichment using the reagents supplied with the column is incomplete and biased toward the recovery and/or detection of smaller, singly phosphorylated peptides. In contrast, elution with base (0.4 M ammonium hydroxide) gives efficient and balanced recovery of both singly and multiply phosphorylated peptides, while loading peptides in a strong acidic solution (1% trifluoracetic acid) further increases selectivity toward phosphopeptides, with minimal carryover of nonphosphorylated peptides. 2,5-Dihydroxybenzoic acid, a matrix commonly used when analyzing phosphopeptides by matrix-assisted laser desorption/ionization mass spectrometry was also evaluated as an additive in loading and eluting solvents. Elution with 50% acetonitrile containing 20 mg/mL dihydroxybenzoic acid and 1% phosphoric acid gave results similar to those obtained using ammonium hydroxide as the eluent, although the latter showed the highest specificity for phosphorylated peptides.

INTRODUCTION

Reversible phosphorylation of proteins mediates numerous regulatory and signaling pathways in prokaryotic and eukaryotic cells.¹ Mass spectrometry (MS) is one of the most powerful tools for studying protein phosphorylation.^2–5 However, phosphoproteomic analysis by MS is still a challenging task because the phosphopeptides generated by proteolysis during such experiments are usually present at relatively low abundance due to the substoichiometric nature of protein phosphorylation, and are often detected with low efficiency due to the inhibitory effect of acidic phosphate groups on positive ionization and fragmentation.^6–7 In the last few years, several methods have been developed for the purification of phosphopeptides, and phosphopeptide enrichment by immobilized metal affinity chromatography (IMAC) prior to MS analysis haspbecome one of the most common strategies in phosphoproteome research.⁸ Although IMAC can significantly enhance the detection of phosphopeptides,^9–12 nonphosphorylated peptides, particularly those containing multiple acidic and/or histidine residues, are often retained along with phosphopeptides. This lack of selectivity can be particularly problematic when attempting to optimize conditions for the recovery of multiply phosphorylated peptides, which tend to be larger and more acidic than those bearing a single phosphate group.^13–15

Optimization of IMAC for phosphopeptide analysis has proven to be a significant challenge, with different methods producing different results with varying degrees of success.^10,16–20 Sometimes, even the results reported by two groups using the same IMAC procedure can be conflicting. For example, o-phosphoric acid has been reported as an efficient eluent for Fe (III)-IMAC,^21,22 while others report incomplete recovery of phosphopeptides using phosphoric acid, the addition of 2,5-dihydroxybenzoic acid (DHB) being necessary to optimize recovery during Fe-IMAC.⁷ More recently, it was reported that DHB did not enhance phosphopeptide recovery using phosphoric acid, and that a combination of phosphoric acid and acetonitrile was required.¹³

In view of this apparent uncertainty, we decided to take a systematic approach toward optimizing conditions for phosphopeptide recovery by IMAC, using a single make and model of commercially available Ga(III)-IMAC column (PhosphoProfile, Sigma) to minimize experimental variability. Here, we report the results of a series of experiments to evaluate the performance of this Ga-IMAC column under different loading, washing, and elution conditions, the goal being to develop an efficient and highly selective procedure for the enrichment of phosphopeptides in support of our own phosphoproteomics research program. We found that different eluents had different selectivities in terms of recovering singly and multiply phosphorylated peptides. We also investigated the effect of loading in different organic acids on the selectivity of phosphopeptide binding, as well as the effect of loading and eluting in solutions containing DHB on the specificity of phosphopeptide binding and recovery from the Ga-IMAC column.

MATERIALS AND METHODS

Materials

Acetonitrile and high-performance liquid chromatography (HPLC)-grade water were obtained from Merck (Darmstadt, Germany). Trifluoroacetic, formic, phosphoric, and acetic acids, ammonium hydroxide (NH₄OH), bovine α-and β-casein, mass calibrants (angiotensin 1, ACTH clip 1–17 and 18–35), PhosphoProfile gallium silica spin columns (product no. P2873) and Phosphopeptide Enrichment Kit (product no. PP0410; hereafter referred to as Sigma PP Kit) were purchased from Sigma (St. Louis, MO). The 2,5-dihydroxybenzoic acid (DHB) matrix compound was obtained from Waters (Milford, MA). Modified porcine trypsin (sequencing grade) was from Promega (Madison, WI). PepClean C₁₈ spin columns (product no. 89870) were purchased from Pierce (Rockford, IL), and ZipTip C₁₈ tips were obtained from Millipore (Bedford, MA).

Sample Preparation and Phosphopeptide Purification

Lyophilized α- and β-casein (100 pmol) were dissolved separately in 100 μL of 0.1 M NH₄HCO₃ (pH 8.5) and digested with 50 ng of trypsin at 37°C overnight (enzyme:substrate ratio 1:50 w/w). Tryptic digests were desalted using C₁₈ PepClean spin columns. The tryptic digests (0.5 μL) of α- and β-casein were added separately to 10 μL of 0.1% trifluoroacetic acid and used to evaluate phosphopeptide enrichment with the PhosphoProfile Ga(III)-IMAC spin columns. After drying in a DNA 120 vacuum centrifuge (Colin Drive, NY), the peptides were reconstituted in 25 μL of the bind/wash buffer (250 mM acetic acid in 30% acetonitrile) supplied with the Sigma PP Kit (product no. PP0410). The columns were washed/equilibrated twice with 50 μL of the bind/wash buffer before loading with sample (25 μL). Each column was gently pressurized by attaching the tip to a 100-μL pipette in order to push the sample into the IMAC medium. The columns were incubated for 15 min at room temperature before spinning at 1000 rpm for 1 min. The eluent was reloaded and the process repeated twice. The loaded columns were washed twice with 50 μL of bind/wash buffer and then twice with 50 μL of deionized water before eluting the bound peptides with 20 μL of elution buffer, as supplied with the Sigma PP Kit (product code E0906; hereafter called Sigma PP elution buffer). According to the product information supplied with this kit, the elution buffer contains 10% phosphoric acid, which the supplier found to be optimal for specific recovery of phosphopeptides. The columns were eluted three times and the eluents pooled. The same columns were again washed twice with 50 μL of deionized water, then eluted three times with 20 μL of 0.4 M NH₄OH. The eluents from this second step were also pooled, and neutralized with trifluoroacetic acid to prevent hydrolysis of phosphate groups. The peptide solutions were dried, reconstituted in 0.1% trifluoroacetic acid and desalted using ZipTip C₁₈ tips. The purified peptides were released in 6 μL of 75% acetonitrile, of which 0.75 μL was mixed with 0.75 μL of 20 mg/mL of DHB matrix solution for analysis by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). In another experiment, while sample loading and washing conditions remained the same as above, phosphopeptides were eluted directly with 0.4 M NH₄OH, without first eluting with Sigma PP elution buffer, to compare the efficiencies of these two eluents. Additional experiments were then performed in which different acidic solutions were substituted for the supplied bind/wash buffer, and/or different eluents substituted for NH₄OH, as detailed below.

Effect of Different Organic acids on Phosphopeptide Binding

Ga-IMAC columns were washed twice with 50 μL of 0.1% aqueous trifluoroacetic acid. Samples (25 μL) were loaded onto the column in either 0.1% or 1% aqueous acetic acid, formic acid or trifluoroacetic acid. After loading, columns were incubated at room temperature for 15 min before spinning at 1000 rpm for 1 min. The flow-through was reloaded and the process repeated twice. The columns were washed once with 50 μL of 0.1% trifluoroacetic acid in 50% acetonitrile, once with 50 μL of 0.1% trifluoroacetic acid in 75% acetonitrile, and then twice with 50 μL of HPLC-grade water to minimize nonspecific binding of nonphosphorylated peptides, before eluting with 20 μL of 0.4 M NH₄OH. The columns were eluted three times, and the eluents pooled, dried in a vacuum centrifuge, and desalted using ZipTip C₁₈ tips according to the manufacturer’s instructions. In another experiment, while other conditions remained the same as above, samples were loaded in DHB matrix solution (20 mg/mL in 0.1% trifluoroacetic acid and 50% acetonitrile) and the recovered phosphopeptides dried, reconstituted in 0.1% trifluoroacetic acid, and desalted using ZipTip C₁₈ tips.

Effect of Different Eluents on Phosphopeptide Recovery

Ga-IMAC columns were equilibrated twice with 50 μL of aqueous 0.1% trifluoroacetic acid, and loaded with sample (25 μL) in 1% aqueous trifluoroacetic acid. The columns were washed according to the procedure described above. The phosphopeptides bound to the column were recovered by eluting with the acidic Sigma PP elution buffer, 0.25 M NH₄HCO₃ (pH 9) or 0.4 M NH₄OH (pH 11).

Effect of DHB on Phosphopeptide Binding

Ga-IMAC columns were equilibrated twice with 50 μL of 0.1% trifluoroacetic acid and loaded with the sample (25 μL) in 50% acetonitrile containing 0.1% trifluoroacetic acid, with or without DHB (20 mg/mL). After incubating for 15 min at room temperature, columns were washed twice with 50 μL of 0.1% trifluoroacetic acid in 50% acetonitrile, once with 50 μL of 0.1% trifluoroacetic acid in 75% acetonitrile, and then twice with 50 μL of HPLC-grade water before eluting the bound peptides using either Sigma PP elution buffer or 0.4 M NH₄OH. In each case, the columns were eluted three times with 20 μL of the elution buffer. The three eluents were pooled together, dried, reconstituted in 0.1% trifluoroacetic acid and desalted/concentrated using ZipTip C₁₈ tips.

Effect of DHB on Phosphopeptide Recovery

Ga-IMAC columns were equilibrated twice with 50 μL of 0.1% trifluoroacetic acid and loaded with the sample (25 μL) in 0.1% trifluoroacetic acid. After incubating for 15 min at room temperature, columns were washed twice with 50 μL of 0.1% trifluoroacetic acid in 50% acetonitrile, once with 50 μL of 0.1% trifluoroacetic acid in 75% acetonitrile, and then twice with 50 μL of HPLC-grade water. The peptides were eluted in either 2.5% formic acid or 1% phosphoric acid in 50% acetonitrile, with or without DHB (20 mg/mL). In each case, the columns were eluted three times with 20 μL of the elution buffer. The three eluents were pooled, dried, reconstituted in 0.1% trifluoroacetic acid and desalted/concentrated using ZipTip C₁₈ tips.

MALDI-TOF Mass Spectrometry

A 0.75-μL aliquot of each desalted peptide solution was combined with 0.75 μL of matrix solution containing 20 mg/mL DHB and 1% phosphoric acid in 75% acetonitrile on a MALDI target plate. The plated samples were analyzed using a Voyager-DE STR mass spectrometer (Applied Biosystems, Framingham, MA) equipped with delayed ion extraction and operating in the positive ion and reflectron modes. All spectra were obtained by averaging 200 laser shots. The efficiency and selectivity of phosphopeptide binding to the Ga-IMAC column under different loading, washing, and elution conditions were evaluated by comparing the number, type (i.e., singly, multiply, or nonphosphorylated), and relative abundance of peptide ions detected by MALDI-TOF MS.

RESULTS AND DISCUSSION

Phosphopeptide Recovery Using Acid, Base and Sequential Elution

A representative MALDI-TOF mass spectrum of the combined, unextracted α- and β-casein digests is shown in Figure 1a. Only three phosphorylated peptide peaks (labeled 5, 9, and 10; see Table 1) were observed, and at relatively low intensity, in these unenriched samples. When the digests were extracted by Ga-IMAC with the Sigma PP Kit, using the recommended protocol and supplied reagents, a total of 10 phosphorylated peptides were detected in the Sigma PP eluent, of which 9 were from α-casein and one from β-casein (Figure 1b and Table 1). Of the 10 phosphopeptides detected, 7 were singly phosphorylated and 3 multiply phosphorylated peptides (peaks 8, 14, and 15), the most intense peptide ion peaks being those of the singly phosphorylated peptides. Following elution with the Sigma PP buffer, the same columns were washed twice with HPLC-grade water and eluted a second time with 0.4 M NH₄OH. Analysis of these samples (Figure 1c) showed that residual singly and multiply phosphorylated peptides were recovered from the column using NH₄OH, including larger, multiply phosphorylated peptides (peaks 17 and 21/21″ [21″ is the peak due to metastable loss of phosphate from peptide 21]) that were not detected in the Sigma PP eluent (Figure 1b). This suggests that although both singly and multiply phosphorylated peptides were retained by the Ga-IMAC column, elution with the Sigma PP buffer containing 10% phosphoric acid was incomplete and biased toward the recovery and/or detection of smaller, singly phosphorylated peptides. The increase in relative intensity of the tetraphosphorylated peptide ion at m/z 3121.3 (peak 21), which undergoes extensive metastable decomposition via neutral loss of phosphoric acid (peak 21”), is particularly evident in the second, NH₄OH elution step. When Ga-IMAC columns were eluted directly with NH₄OH, without first eluting with Sigma PP elution buffer, balanced elution of both singly and multiply phosphorylated peptides was observed, as indicated by the relative intensities of the corresponding peptide ion peaks (Figure 1d). Furthermore, multiply phosphorylated peptide ion peaks were of comparable or greater intensity (e.g., peaks 8 and 19) than those observed when eluting with NH₄OH as a second step, following initial use of the Sigma PP buffer. When flow-through and wash fractions were analyzed by MALDI-TOF MS (data not shown) only two, singly phosphorylated peptides (numbers 5 and 9) were detected, and at very low abundance, the former in both flow through and wash fractions, the latter only in the flow-through fraction. This confirms that the PhosphoProfile gallium silica spin columns efficiently retain both singly and multiply phosphorylated peptides.

FIGURE 1.

Phosphopeptide purification using Sigma PhosphoProfile Ga(III)-IMAC spin columns. The panels show representative MALDI mass spectra of a combined tryptic digest of bovine α- and β-casein (500 fmol of each protein) (a) before IMAC enrichment, (b) following IMAC enrichment using the reagents supplied with the Sigma PP Kit, (c) after washing the same column with water and eluting a second time with 0.4 M NH₄oH, or (d) after performing IMAC again using the bind/wash buffer supplied with the PP kit and eluting directly with 0.4 M NH₄oH. The reproducibility of these results was confirmed by repeating each procedure at least three times.

TABLE 1.

Phosphopeptide Ions Observed in MALDI-TOF Mass Spectra of Combined Tryptic α- and β-Casein Digests Following Ga(III)-IMAC Enrichment

Peak Label	[M+H]+ (m/z)	Peptide Sequencea	No. of Po4	pI b	BBc(×10−3)
1′	971.0	NM(ox)AINPSK(αS2:40–47)d	1	4.5	0.6
2	1237.5	TvDMESTEvF(αS2:138–147)	1	1.5	−1.5
3	1466.7	TvDMESTEvFTK(αS2:153–164)	1	3.0	−0.7
4	1593.8	TvDMESTEvFTKK(αS2:153–165)	1	4.0	−0.3
5	1660.9	vPqLEIvPNSAEER(αS1:121–134)	1	3.1	−0.3
6	1832.8	yLGEyLIvPNSAEER(αS1:104–119)e	1	3.1	−4.0
7	1847.8	DIGSESTEDqAMEDIK(αS1:58–73)	1	2.8	3.4
8	1927.6	DIGSESTEDqAMEDIK(αS1:58–73)	2	1.9	2.9
8′	1943.7	DIGSESTEDqAM(ox)EDIK(αS1:58–73)f	2	1.9	2.9
9	1951.3	yKvPqLEIvPNSAEER(αS1:119–134)	1	4.2	−1.2
10	2061.9	FqSEEqqqTEDELqDK(β-C:33–48)	1	2.8	5.7
11	2556.2	FqSEEqqqTEDELqDKIHPF(β-C:33–52)	1	3.6	3.2
12	2618.1	NTMEHvSSSEESIISqETyK(αS2:17–36)	4	1.6	0.3
13	2678.0	vNELSKDIGSESTEDqAMEDIK(αS1:52–73)	3	2.0	2.4
14	2704.5	pyroEMEAESISSSEEIvPNSvEAqK(αS1:74–94)g	5	1.0	1.8
15	2721.2	qMEAESISSSEEIvPNSvEAqK(αS1:74–94)	5	1.0	1.8
16	2746.8	NTMEHvSSSEESIISqETyKq(αS2:17–37)	4	1.6	1.3
17	2935.4	EKvNELSKDIGSESTEDqAMEDIK(αS1:50–73)	3	2.8	3.4
18	2965.2	ELEELNvPGEIvESLSSSEESITR(β-C:2–25)	4	1.2	−2.9
19	3007.8	NANEEEySIGSSSEESAEvATEEvK(αS2:61–85)	4	1.2	5.3
20	3088.9	NANEEEySIGSSSEESAEvATEEvK(αS2:61–85)	5	1.0	4.9
21	3121.3	RELEELNvPGEIvESLSSSEESITR(β-C:1–25)	4	1.6	−2.2
21″	3027.6	RELEELNvPGEIvESLSSSEESITR(β-C:1–25)h	3	–	–
22	3132.2	KNTMEHvSSSEESIISqET yKqEK(αS2:16–39)	4	3.0	2.8

^aαS1 and αS2 refer to the first and second subunits of α-casein, respectively.

^bpI calculated according to Bjellqvist et al.³⁸

^cHydrophobicity (BB index) determined according to Bull and Breese.³⁹

^dMethionine-oxidized form of peptide NMAINPSK.

^ePeptide 6 sequence according to Larsen et al.³²

^fMethionine-oxidized form of peptides 7.

^gPeptide 14 sequence according to Hsieh et al.,⁴⁰ with the N-terminal glutamine cyclized to pyroglutamic acid.

^hPeak due to metastable loss of phosphate from peptide 21.

Despite the overall effectiveness of NH₄OH in recovering phosphorylated peptides from the Ga-IMAC column, a comparison of the spectra obtained for sequential Sigma PP and NH₄OH eluents reveals a certain degree of complementarity between the two (Figure 1b, c). For example, peak 14 was visible only in the acidic Sigma PP eluent, albeit at very low intensity, while peak 10 was far more intense in the PP eluent than in NH₄OH. On the other hand, multiply phosphorylated peptide peaks 17, 18, 19, and 21/21′ were detected only in the basic eluent. For peptides detected in both eluents, peak intensities for singly phosphorylated peptides were generally similar in both cases (except peptide 10), whereas those for multiply phosphorylated peptides were usually stronger in the basic eluent, which is clearly advantageous for comprehensive mapping of protein phosphorylation sites. Furthermore (with the exception of peptide 10), the relative peak intensities of all phosphorylated peptides detected in 0.4 M NH₄OH were essentially the same regardless of whether NH₄OH was used as the second or only eluent (Figure 1c, d). Sequential acid/base elution of singly and multiply phosphorylated peptides during IMAC has recently been proposed as part of an alternative, LC/MS-based strategy in which instrumental parameters can be optimized independently for detection of singly and multiply phosphorylated peptides.²³ By using 0.4 M NH₄OH to achieve rapid, single-step elution of both singly and multiply phosphorylated peptides, our approach is more conducive to high-throughput phosphoproteome analysis using MALDI MS or MS/MS, with the option of performing LC separations prior to MALDI (or electrospray) mass spectrometry.

From our observations, it appears that peptide recovery depends not only on the extent of phosphorylation but on other physicochemical properties (e.g., hydrophobicity: see Table 1) that favour solubility and/or MS detection in either acidic or basic eluents. For example, the relative abundance of the most hydrophobic peptide peak (10) is highest in the acidic Sigma PP eluent, whereas the most hydrophilic multiply phosphorylated peptide ion (21) is also the most abundant when eluted in NH₄OH but was not detected in the acidic eluent. Another, related issue concerning IMAC is the incidental recovery of nonphosphorylated peptides, especially those containing multiple acidic residues.^14,15 Histidine-containing peptides are also known to bind strongly to Cu²⁺- or Ni²⁺-loaded IMAC columns and show increasing retention as the pH of the loading buffer increases, due to deprotonation of the side-chain (pyrrole) amino group (pKa 6.0).^24,25 Two of the peaks (labeled C and D) observed in both eluents (in Fig. 1) correspond to nonphosphorylated peptides HIQKEDVPSER (pI 5.5) and HQGLPQEVLNEN-LLR (pI 5.4), both of which contain histidine and acidic residues. This is consistent with the known tendency of acidic and/or histidine-containing peptides to be retained during IMAC. However, peak C was more intense in the Sigma PP eluent than in 0.4 M NH₄OH, demonstrating the greater selectivity of the latter for recovery of phosphorylated peptides.

Effect of Different Organic Acids on Phosphopeptide Binding

It is known that phosphopeptides are not well retained at alkaline pH values,^26–28 whereas acidic and histidine-containing peptides show increasing retention at higher pH.^9,24,25,29 In the course of these experiments, we also found that the Ga-IMAC spin columns retained nonphosphorylated peptides that contained acidic and histidine residues (see above). Acidic wash solutions are generally used to prevent binding of acidic and/or hydrophobic peptides.³⁰ The reason for selecting strongly acidic loading/washing buffers for phosphopeptide purification is to ensure that the weakly acidic (carboxyl and hydroxyl) groups found in peptides are protonated while phosphate groups (pKa 1.8; see below) remain dissociated, and therefore able to interact strongly with IMAC or metal oxide media. Earlier reports document the use of acetic acid at a concentration of 0.1–0.25 M (pH 2.7–2.9) in column loading solutions; however, a significant number of acidic nonphosphorylated peptides remained bound to either IMAC or TiO₂ columns under these conditions.12,31,32 Loading sample in 0.1% trifluoroacetic acid (pH 1.9) for the purification of phosphorylated peptides using TiO₂ increased the detection of phosphorylated peptides relative to nonphosphorylated peptides; however, a number of nonphosphorylated peptides were still recovered.³² The effects of different carboxylic acids on the selective binding of phosphopeptides to Ga-IMAC have not been discussed much in detail. Therefore, to optimize the loading condition for phosphopeptide enrichment by Ga-IMAC, we decided to evaluate the efficiency of 0.1% or 1% of acetic acid, formic acid, or trifluoroacetic acid for loading the Sigma Ga-IMAC columns. Although the Sigma PP Kit utilizes 250 mM (1.5%) acetic acid in 30% acetonitrile for sample loading, we found that loading samples in water rather than 30% acetonitrile consistently gave better results (data not shown). NH₄OH was used as the eluent in these experiments because it was found to be more efficient than the Sigma PP elution buffer for rapid, single-step elution of phosphopeptides from the Ga-IMAC spin columns (see above).

Peaks C and D in Figure 1, the two most abundant nonphosphorylated peptides observed in our initial Ga-IMAC experiments, were selected as reference peaks with which to evaluate the efficiencies of the three different organic acids in inhibiting adsorption of nonphosphorylated peptides. As shown in Figure 2, the relative abundances of these two nonphosphorylated peptides were much lower when the sample was loaded in formic acid or trifluoroacetic acid than when loading in acetic acid. Based on the results obtained from repeated experiments, we conclude that the efficacy of these carboxylic acids in inhibiting nonspecific binding of peptides to Ga-IMAC is in the order trifluoroacetic acid > formic acid > acetic acid (Table 2). Loading samples in 0.1% or 1% aqueous solutions of these acids also produced different results. In all cases, the relative abundances of peaks C and D were reduced when using 1% acid as the loading buffer. At the same time, the relative abundances of multiply phosphorylated peptides 14–17, 19, and 21/21″ increased when samples were loaded in 1% as opposed to 0.1% acid, as exemplified by comparing the intensity of peak 21 in the different spectra (Fig. 2).

FIGURE 2.

Effect of different loading solutions on the selective binding of phosphorylated peptides by Ga(III)-IMAC. The panels show representative MALDI mass spectra of tryptic α- and β-casein peptides loaded in (a) 0.1% and (b) 1% acetic acid, (c) 0.1% and (d) 1% formic acid, or (e) 0.1% and (f) 1% trifluoroacetic acid, and eluted in 0.4 M NH₄OH.

TABLE 2.

Effect of Loading in Different Acidic Solutions on the Selectivity of Phosphopeptide Enrichment from Combined Tryptic α- and β-Casein Digests Using Ga(III)-IMAC

			No. of Identified Peptides (P < 0.05)a
			Phosphorylated		Nonphosphorylated
Loading Buffers	Concentration (%)	pH	Singly	Multiply
Acetic Acid	0.1	3.2	7	4	5
	1.0	2.8	7	6	3
Formic Acid	0.1	2.6	7	8	3
	1.0	2.2	7	10	3
Trifluoroacetic Acid	0.1	2.1	7	8	2
	1.0	1.4	8	11	1b
			8	11	0c

^aBound peptides eluted in 0.4 M NH₄oH.

^b,cData summarized from Figures 2e and 3d, respectively.

There remains disagreement in the literature as to the optimum concentration of trifluoroacetic acid required when loading IMAC columns to achieve optimal specificity for phosphopeptides. For example, Kokubu et al.³³ found 0.1% trifluoroacetic acid to be the optimum concentration for maximizing phosphopeptide peak intensities and minimizing detection of nonphosphorylated peptides when compared with 0.3% or 1% trifluoroacetic acid, while others have also used 5% trifluoroacetic acid as the loading buffer for phosphopeptide enrichment.³⁴ In our hands, repeated experiments confirmed 1% trifluoroacetic acid as the optimum loading buffer for highest selectivity towards phosphopeptides, whereas loading samples in 5% trifluoroacetic acid actually reduced selectivity, increasing the relative abundances of nonphosphorylated peptide peaks (data not shown). Multiply phosphorylated peptide peak 12 was detectable when using 1% formic acid or trifluoroacetic acid as the loading buffer but absent from Ga-IMAC extracts that had been loaded with 1% acetic acid, or 0.1% of all three acids. Also, singly phosphorylated peptide peak 11 was only detected when using 1% trifluoroacetic acid as a loading solution. Substitution of an alkyl group for a proton in orthophosphoric acid (HOPO₂OH) increases acidity; for example, the pKa of phosphoric acid decreases from 1.8 to 1.1 upon methylation (i.e., CH₃OPO₂OH).32,35 Similarly, attachment of a phosphate group to a peptide may be expected to reduce significantly its pKa.³² While loading samples in higher concentrations of stronger organic acids (i.e., formic acid, trifluoroacetic acid) will completely protonate other acidic residues and inhibit their interaction with immobilized Ga³⁺ ions, peptide phosphate groups should, therefore, remain at least partially dissociated and able to bind to the Ga-IMAC column. Our results suggest that the pKa values of peptide phosphate groups are indeed significantly lower than that of free phosphate. Consequently, loading phosphoprotein digests in 1% trifluoroacetic acid (pH 1.4) proved most effective in inhibiting the adsorption of nonphosphorylated peptides during Ga-IMAC enrichment. Higher trifluoroacetic acid concentrations (>1%) appeared to reduce specificity toward phosphopeptides, possibly because the relatively high concentration of dissociated trifluoroacetic acid molecules competes with peptide phosphate groups in binding to the available immobilized Ga³⁺ ions. In any event, loading in 1% trifluoroacetic acid gave best results in terms of specificity for Ga-IMAC enrichment of phosphopeptides.

Effect of Different Eluents on Phosphopeptide Recovery

Since loading in 1% trifluoroacetic acid was found to be optimal for phosphopeptide enrichment by Ga-IMAC, we used the same loading conditions when evaluating different elution solutions for selective recovery of bound phosphopeptides from the Ga-IMAC column. Peptides loaded in 1% trifluoroacetic acid were eluted using Sigma PP elution buffer, 0.25 M NH₄HCO₃ (pH 9.0) or 0.4 M NH₄OH (pH 11) (Figure 3). As shown in Figure 3a, when columns were eluted with the Sigma PP buffer, a total of 16 phosphorylated peptides were detected from a mixture of α- and β-casein tryptic digests, of which 7 were singly phosphorylated and 9 were multiply phosphorylated peptides. In comparison, samples loaded using the bind/wash buffer supplied with the Sigma PP enrichment kit (which contains 1.5% acetic acid in 30% acetonitrile) and eluted with the same Sigma PP elution buffer recovered only 10 detectable phosphopeptides (Figure 1b), of which 7 were singly phosphorylated and 3 multiply phosphorylated. Comparing these results, it is apparent that optimizing the load/wash conditions significantly enhances the performance of Ga-IMAC for phosphopeptide enrichment.

FIGURE 3.

Effect of different eluents on the recovery of phosphorylated peptides from Ga(III)-IMAC columns. The panels show MALDI mass spectra of tryptic α- and β-casein peptides loaded in 1% trifluoroacetic acid and eluted in (a) the Sigma PP elution buffer containing 10% phosphoric acid, (b) 0.4 M NH₄OH following initial elution of the same column with the PP elution buffer, (c) NH₄HCO₃ (pH 9.0), or (d) 0.4 M NH₄OH (pH 11) without prior elution using the PP buffer.

Turning again to the eluent, use of the Sigma PP elution buffer clearly favors recovery of singly phosphorylated over multiply phosphorylated peptides, as subsequent washing of the Ga-IMAC column with 2 ×50 μL of HPLC-grade water and elution with 0.4 M NH₄OH results in a several-fold increase in the relative intensities of multiply phosphorylated peptide ions, as shown in Figure 3b. To further investigate the effect of eluent basicity on phosphopeptide recovery, Ga-IMAC columns freshly loaded with casein digests in 1% trifluoroacetic acid were eluted directly using either 0.25 M NH₄HCO₃ (pH 9) or 0.4 M NH₄OH (pH 11). Both eluents performed much better than the Sigma PP elution buffer; however, NH₄OH showed much greater specificity in the recovery of phosphorylated peptides than NH₄HCO₃, as illustrated by the significant number of nonphosphorylated peptides detected in the NH₄HCO₃ eluent (Figure 3c) and the apparent absence of such peptides in the NH₄OH eluent (Figure 3d). These results clearly demonstrate the superiority of NH₄OH for the recovery of singly and multiply phosphorylated peptides from the Ga-IMAC column.

Effect of DHB on Phosphopeptide Binding

Use of DHB in the loading buffer has been shown to enhance the selectivity of phosphopeptide enrichment by metal oxide (TiO₂) affinity chromatography.31,32 To investigate whether DHB enhances selectivity of Ga-IMAC for phosphopeptides, the combined tryptic α- and β-casein digest was loaded in a solution containing both trifluoroacetic acid and DHB (20 mg/mL DHB in 50% acetonitrile, 1% trifluoroacetic acid). After washing the column once with 50 μL of 0.1% trifluoroacetic acid in 50% acetonitrile, once with 50 μL of 0.1% trifluoroacetic acid in 75% acetonitrile, and twice with 50 μL of HPLC-grade water, the bound peptides were eluted by Sigma PP elution buffer. As shown in Figure 4a, only five phosphopeptides (peaks 5, 9, 14, 15 and 16) were detected, and with surprisingly low ion intensities. Even peptide 9 (m/z 1951.3), which usually gave the most abundant phosphopeptide ion in MALDI mass spectra of IMAC extracts, was present at very low levels. In contrast, the intensity of nonphosphorylated peptide ion C (m/z 1337.7) was unusually high.

FIGURE 4.

Effect of DHB on the binding of phosphorylated peptides during Ga(III)-IMAC. Panels show representative MALDI mass spectra of the tryptic α- and β-casein peptides loaded in 50% acetonitrile containing 0.1% trifluoroacetic acid and 20 mg/mL DHB and (a) eluted in Sigma PP elution buffer, (b) eluted in NH₄OH after initial elution of the same column with PP buffer, or (c) detected in the flow-through solution after reloading it onto the Ga-IMAC column in 1% trifluoroacetic acid and eluting with 0.4 M NH₄OH. DHB, 2,5-dihydroxybenzoic acid.

Following elution with the Sigma PP buffer, each column was again washed with HPLC-grade water and eluted with 0.4 M NH₄OH. A total of 13 phosphopeptides were detected (Figure 4b), including the 5 observed upon initial elution with the (acidic) Sigma PP buffer. Moreover, these 5 peptides (peaks 5, 9, 14, 15, and 16) were detected at significantly higher concentrations when the (basic) NH₄OH solution was used as the second eluent. Nevertheless, when compared with earlier results obtained using just 1% trifluoroacetic acid as the loading buffer (Figure 3a, b), it is apparent that the inclusion of DHB sharply decreased the performance of Ga-IMAC.

DHB in the loading buffer has been shown to enhance the selectivity of TiO₂ affinity chromatography towards phosphopeptides by inhibiting binding of nonphosphorylated peptides.³² This effect was attributed to competition between nonphosphorylated peptides and DHB molecules for TiO₂ binding sites, the larger molar excess of DHB preventing adsorption of nonphosphorylated peptides to the surface of TiO₂. The interaction between TiO₂ and phosphate groups during MOAC appears to be much stronger than between immobilized metal-ions and phosphate groups during IMAC.³² Therefore, DHB might be expected to displace certain phosphopeptides more easily from IMAC columns than from TiO₂. As a result, loading samples in the presence of DHB may prevent binding of some phosphopeptides to the IMAC column and, possibly, inhibit the binding of others. That DHB might inhibit the binding of phosphopeptides during Ga-IMAC was further suggested by the observation that several phosphopeptides, especially those bearing a single phosphate group, were present in the flow-through solution. This was verified by drying down the flow-through solution, resuspending the peptides in 1% trifluoroacetic acid, desalting with ZipTips, and performing Ga-IMAC again only this time loading sample in 1% trifluoroacetic acid and eluting by 0.4M NH₄OH. As shown in Figure 4c, the number of detected phosphopeptides clearly suggests that DHB inhibits the binding of phosphopeptides to the Ga-IMAC column. Additionally, the amount of acetonitrile (50% in this study) required to solubilize DHB in the loading solvent might also interfere with the binding of phosphopeptides to the column. Organic solvents like acetonitrile are often used to disrupt interactions between hydrophobic peptides and polymeric resins.³⁰ Consequently, acetonitrile may also disrupt interactions between certain phosphopeptides and the Ga-IMAC column, depending on the strength and nature of the interaction between peptide and stationary phase.

Effect of DHB on Phosphopeptide Recovery

The specific binding and release of phosphopeptides from IMAC and TiO₂ columns depends upon the competition between solution components for the available binding sites.^7,31 Modification of elution buffers with the MALDI matrix compound DHB, which contains one carboxyl and two hydroxyl groups, is expected to compete with and effectively displace phosphopeptides bound to these columns and, subsequently, to enhance detection and analysis by MALDI-TOF MS. However, the efficacy of using DHB in combination with phosphoric acid as an eluent for recovering phosphopeptides from IMAC media remains unclear, due to the contrasting results reported by different groups. While some workers found phosphoric acid to be an efficient eluent for Fe(III)-IMAC, ^21,22 others reported elution with phosphoric acid alone to be incomplete.⁷ In our hands, phosphopeptide recovery using the Sigma PP elution buffer supplied with the PhosphoProfile Ga-IMAC columns, which contains 10 % phosphoric acid, was also incomplete, notwithstanding the claim that this solution was formulated to optimize specific recovery of phosphopeptides. Again, some workers found DHB to be a suitable eluent for IMAC, although it biases in favor of singly phosphorylated peptides,^32,36 while others found that addition of DHB to phosphoric acid enhanced recovery of both singly and multiply phosphorylated peptides during Fe-IMAC.⁷ In contrast, Imanishi et al.¹³ report that DHB does not enhance the recovery of phosphopeptides, and that a combination of phosphoric acid and acetonitrile is more efficient.

To clarify whether or not inclusion of DHB enhances phosphopeptide recovery, we compared the MALDI-TOF MS spectra of samples obtained by using either phosphoric acid or formic acid, alone or in combination with DHB, to elute tryptic phosphopeptides from the Ga-IMAC spin columns. Columns were eluted using 1%, 2.5% or 10% formic acid or phosphoric acid in 50% acetonitrile, with and without 20 mg/mL DHB. In the absence of DHB, best results were obtained using 2.5% formic acid or 1% phosphoric acid. To clarify the effect of DHB on phosphopeptide recovery, we therefore report only those results obtained using 2.5% formic acid or 1% phosphoric acid (Figure 5). Both formic acid and phosphoric acid were capable of recovering singly and multiply phosphorylated peptides (Figure 5a, b), although several nonphosphorylated peptides (including peaks A, B, C, and D) were also detected in both eluents. The doubly phosphorylated peptide 8 gave most abundant peak when eluted in formic acid, whereas in phosphoric acid the most abundant ion was that of the singly phosphorylated peptide 9. Similarly, peptide ions 12–19 (all multiply phosphorylated and hydrophobic) were more abundant in formic acid than in phosphoric acid, whereas peptide ion 21 (also multiply phosphorylated but hydrophilic) was more abundant in phosphoric acid than in formic acid.

FIGURE 5.

Effect of DHB on the recovery of phosphorylated peptides from Ga(III)-IMAC columns. Panels show representative MALDI mass spectra of the tryptic α- and β-casein peptides loaded in 1% TFA and eluted in 50% acetonitrile containing (a) 2.5% formic acid, (b) 1% phosphoric acid, (c) 2.5% formic acid and 20 mg/mL DHB, or (d) 1% phosphoric acid and 20 mg/mL DHB. DHB, 2,5-dihydroxybenzoic acid.

When compared with the results obtained using NH₄OH (Figures 2f and 3d), however, it is clear that these acidic solutions do not match the elution efficiency of the basic eluent. When compared with results obtained using formic acid or phosphoric acid alone, addition of DHB to the elution buffer significantly enhanced peptide recovery and detection, particularly for multiply phosphorylated peptides (Figure 5c, d). In general, the intensities of singly phosphorylated peptide ions relative to those of multiply phosphorylated peptides were reduced in eluents containing DHB (cf. spectra in Figures 2, 3, and 5). Furthermore, when columns eluted with formic acid or phosphoric acid alone were eluted a second time with NH₄OH, or with the same buffer containing DHB, additional multiply phosphorylated peptides were detected (data not shown), suggesting incomplete recovery of phosphopeptides by formic acid or phosphoric acid when used alone. This was not the case, however, when columns eluted with formic acid or phosphoric acid containing DHB were subsequently eluted with NH₄OH, confirming that phosphopeptide recovery from the Ga-IMAC columns is more complete when DHB is included in the eluent.

The Ga³⁺ ion, which does not participate in redox reactions, is often used as a substitute for Fe³⁺, which does display redox activity, in applications such as the study of iron-binding proteins,³³ and the binding of phosphopeptides, since both trivalent metal ions have similar coordination geometries and a strong affinity for groups that contain oxygen.^30,37 Given these similarities, we feel it appropriate to compare the present results obtained using Ga(III)-IMAC with those reported elsewhere for Fe(III)-IMAC. Stensballe and Jensen⁷ concluded that formic acid with DHB was efficient in recovering singly but not multiply phosphorylated peptides, based on relative intensities of the two β-casein phosphopeptides (peaks 10 and 21 in our study) recovered using Fe-IMAC. However, they observed a striking increase in the relative abundance of peak 21 when eluted using phosphoric acid with DHB. We also found peak 21 to be more intense in phosphoric acid than in formic acid, the addition of DHB enhancing the relative intensity of this peak in both cases. The combination of phosphoric acid and DHB appeared to offer the best performance in terms of phosphopeptide coverage and minimal detection of nonphosphorylated peptides (Figure 5d), though not quite as good as that achieved using 0.4 M NH₄OH (Figures 2f and 3d). Hart et al.³⁶ observed a significant number of Fe-peptide adduct ions when using a DHB solution to recover phosphopeptides during Fe(III)-IMAC. No Ga-peptide adducts were observed in our experiments, demonstrating the advantage of Ga(III)-IMAC in this respect.

pi, Hydrophobicity and Phosphopeptide Coverage

We calculated the pI³⁸ and hydrophobicity (BB index)³⁹ of all phosphorylated (Table 3) and nonphosphorylated peptides (Table 2) detected following Ga-IMAC purification to help determine how these two factors influence phosphopeptide enrichment by Ga-IMAC. All six non-phosphorylated peptides are acidic, four (B, D, E, and F) being hydrophilic (negative BB index) and two (A and C) hydrophobic (positive BB index). This suggests that pI, rather than hydrophobicity, has the greater potential for discriminating against nonselective binding, recovery and/or detection of such peptides using IMAC. With regard to the phosphorylated peptides, with the exception of peptides 1, 7, 10, 11, and 21, the smaller, singly phosphorylated peptides are hydrophilic whereas larger, multiply phosphorylated peptides are hydrophobic. In general, hydrophilic peptides appeared easier to detect than hydrophobic peptides; for example, peptides 9 and 21, which usually gave the most abundant singly and multiply phosphorylated peptide ion peaks, respectively, are hydrophilic. Assuming that the combined casein digest provides a reasonable model for phosphoproteins in general, these results can be used to predict the extent of phosphoproteome coverage that might be expected using this technique. For example, Table 4 shows the number and location of theoretical phosphorylation sites in bovine α(S1)- and β-caseins,^41–43 and the MALDI MS peak labels corresponding to tryptic peptides that include these sites. All such sites are represented by the peptides detected using the optimized Ga-IMAC procedure in which peptides are loaded in 1% trifluoroacetic acid and eluted with 0.4 M NH₄OH. In the case of α(S2)-casein, all known sites^44,45 except Ser-144 and Ser-146 are also represented by peptides recovered using the optimized Ga-IMAC procedure, along with a novel phosphopeptide (peak 20) containing a phosphorylated Ser-68 residue. The diversity of these peptides in terms of size, number of phosphate groups, pI, and hydrophobicity suggests that similar results may be obtained in biological studies, with coverage somewhat biased in favor of phosphorylation sites that reside in hydrophilic phosphopeptides.

TABLE 3.

Nonphosphorylated Peptide Ions Observed in MALDI-TOF Mass Spectra of Combined Tryptic α- and β-Casein Digests Following Ga(III)-IMAC Enrichment

Peak Label	[M+H]+ (m/z)	Peptide Sequencea	pI	BB (×10−3)
A	831.4	EDvPSER (αS1:99-105)	4.0	1.8
B	1267.8	yLGyLEqLLR(αS1:106-115)	5.6	−6.5
C	1337.7	HIqKEDvPSER(αS1:95-105)	5.5	2.5
D	1760.0	HqGLPqEvLNENLLR(αS1:23-37)	5.4	−0.6
E	2187.3	DMPIqAFLLyqEPvLGPvR (β-C:184-202)	4.3	−6.9
F	2317.5	EPMIGvNqELAyFyPELER (αS1:148-166)	4.0	−4.9
F′	2333.5	EPM(ox)IGvNqELAyFyPELER (αS1:148-166)b	4.0	−4.9

^aαS1 and αS2 refer to the first and second subunits of α-casein, respectively.

^bMethionine-oxidized form of peptide F.

TABLE 4.

Sites of Phosphorylation in α- and β-Caseins and MALDI Mass Spectral Peaks Corresponding to Tryptic Peptides That Include These Sites

Protein	Phosphorylation Site(s)	Peptide Peak(s)a
β-Caseinb	Ser-15, 17, 18, 19	18, 21
	Ser-35	10, 11
α-Casein (S1)c	Ser-56	13, 17
	Ser-61	7, 8, 13, 17
	Ser-63	8, 13, 17
	Ser-79, 81, 82, 83, 90	14, 15
	Ser-114	6
	Ser-130	9
α-Casein (S2)d	Ser-23, 24, 25, 28	12, 16, 22
	Ser-46 e	1’
	Ser-71, 72, 73, 76	19, 20
	Ser-68 f	20
	Ser-144 g	2
	Ser-158	3, 4

^aSee Table 1 for peptide sequence information.

^bPhosphorylation sites for β-casein according to Ribadeau-Dumas et al.⁴¹

^cPhosphorylation sites for the S1 subunit of α-casein (S1) according to Mercier et al.,⁴² Imanishi et al.,⁴³ and Swiss-Prot accession P02662.⁴⁴

^dPhosphorylation sites for the S2 subunit of α-casein (S2) according to Imanishi et al.⁴³ and Swiss-Prot accession P02663.⁴⁵

^ePeptide 1 containing novel phosphorylation site Ser-46 detected using 2.5% formic acid as the eluent (see Figure 5a).

^fPeptide 20 containing novel phosphorylation site Ser-68 detected using 0.4 M NH₄oH as the eluent (see Figure 3d).

^gPeptides 2 containing known phosphorylation site Ser-144 detected using 50% acetonitrile containing 1% phosphoric acid and 20 mg/mL DHB as the eluent (see Figure 5d).

CONCLUSIONS

We have optimized conditions for the selective recovery of phosphorylated peptides from a combined tryptic digest of α- and β-casein with a commercially available (Sigma Phosphoprofile) Ga-IMAC column, using loading and eluting solvents other than those supplied with the column. Best results were achieved by loading the digests onto the column in 1% trifluoroacetic acid and eluting with 0.4 M aqueous NH₄OH or with 50% acetonitrile containing 20 mg/mL DHB and 1% phosphoric acid. The two eluents are somewhat complementary in terms of the relative abundances of individual phosphopeptide ions detected by MALDI MS. However, 0.4 M aqueous NH₄OH provides optimal specificity for phosphorylated peptides and is more generally compatible with MS detection methods, including those such as LC/MS that involve electrospray ionization. The extent of phosphorylation site coverage obtained for α- and β-caseins, and the physicochemical diversity of the tryptic peptides detected, suggest that the optimized Ga-IMAC procedure can successfully be applied to biological studies involving phosphoproteomic analysis.