Phosphoprotein Analysis Using Antibodies Broadly Reactive against Phosphorylated Motifs*

The substrates of most protein kinases remain unknown because of the difficulty tracing signaling pathways and identifying sites of protein phosphorylation. Here we describe a method useful in detecting subclasses of protein kinase substrates. Although the method is broadly applicable to any protein kinase for which a substrate consensus motif has been identified, we illustrate here the use of antibodies broadly reactive against phosphorylated Ser/Thr-motifs typical of AGC kinase substrates. Phosphopeptide libraries with fixed residues corresponding to consensus motifs RXRXXT*/S* (Akt motif) and S*XR (protein kinase C motif) were used as antigens to generate antibodies that recognize many different phosphoproteins containing the fixed motif. Because most AGC kinase members are phosphorylated and activated by phosphoinositide-dependent protein kinase-1 (PDK1), we used PDK1−/− ES cells to profile potential AGC kinase substrates downstream of PDK1. To identify phosphoproteins detected using the Akt substrate antibody, we characterized the antibody binding specificity to generate a specificity matrix useful in predicting antibody reactivity. Using this approach we predicted and then identified a 30-kDa phosphoprotein detected by both Akt and protein kinase C substrate antibodies as S6 ribosomal protein. Phosphospecific motif antibodies offer a new approach to protein kinase substrate identification that combines immunoreactivity data with protein data base searches based upon antibody specificity.

Progress in tracing signaling pathways has been limited by the lack of reagents and methods required to identify substrates of particular protein kinases (1). The development of broadly reactive phosphotyrosine antibodies allowed identification and characterization of many tyrosine kinase substrates involved in growth factor and cytokine signaling (2)(3)(4)(5). However, despite considerable effort, the development of comparable phosphoserine and phosphothreonine specific antibodies has proven much more difficult. Because greater than 90% of protein phosphorylation occurs at serine and threonine residues (6), methods to identify and characterize these sites are needed. Site-specific phosphoserine/threonine antibodies have proven useful in monitoring specific signaling events; however, because these reagents can be prepared only after the precise site of phosphorylation have been mapped (7), they are not useful in discovery of new sites. Identification of phosphorylation sites by chemical modification (8,9) and/or mass spectrometry (8 -11) is still in the early stages of development and, although promising, is far from routine. Early estimates suggest that the human genome encodes ϳ400 Ser/Thr protein kinases and that the majority of human proteins may be phosphorylated at multiple sites (Ͼ100,000 sites). Currently, fewer than 2,000 sites have been identified, emphasizing the need for high throughput and sensitive methods to identify, characterize, and monitor new sites of protein phosphorylation.
In general, protein phosphorylation occurs at short linear sequence motifs that regulate protein activity, location, and interaction (27)(28)(29). Consensus motifs phosphorylated by protein kinases can be determined in vitro using oriented peptide libraries (30,31) or by comparing known in vivo sites of phosphorylation (32). Here we use peptide libraries representing phosphorylated Ser/Thr protein kinase consensus motifs to two different AGC kinase subfamilies, Akt and PKC, as immunogens to produce phosphospecific antibodies that react selectively with phosphorylated motifs present in a broad array of AGC protein kinase substrates. To demonstrate that the phospho-motif antibodies could identify downstream substrates, we compared immunoreactivity of PDK1Ϫ/Ϫ ES cell extracts with wild type ES extracts. Combining one-and two-dimensional immunoblotting analysis using Akt and PKC substrate anti-bodies together with protein data base searching using epitope specificity matrices, we predicted and subsequently identified a 30-kDa protein detected in wild type but not PDK1Ϫ/Ϫ ES cells as S6 ribosomal protein. Understanding antibody specificity for phosphorylated motifs offers the possibility of using antibodyto-protein recognition features to scan protein databases to predict sites of antibody interaction. Here we show that antibodies recognizing phosphorylated motifs can serve as valuable proteomic tools, interfacing between immunoreactivity data and genome/proteome databases to identify new targets of protein phosphorylation.

EXPERIMENTAL PROCEDURES
Materials-All the primary and secondary antibodies used in this study were from Cell Signaling Technology (Beverly, MA). The specificity and characterization of primary antibody are described in the Cell Signaling Technology web site (www.cellsignal.com). Materials used in immunoblotting analysis are from Cell Signaling Technology. Proteaseinhibitor mixture tablets are from Calbiochem, and tissue culture reagents are from Invitrogen. All other chemicals were from Sigma and Calbiochem.
Antibody Production-Polyclonal antibodies to specific phosphorylation sites are produced by immunizing rabbits with a synthetic phosphorylated peptide (KLH (keyhole limpet hemocyanin) coupled) corresponding to residues surrounding phosphorylation sites. Antibodies are then purified by protein A column chromatography on an Amersham Biosciences Ä KTA fast protein liquid chromatograph to isolate the IgG antibody fraction. Affinity chromatography is then performed using peptides coupled to SulfoLink resin from Pierce according to manufacturer. Both phospho-peptide-containing resin and the corresponding non-phospho-peptide resin were prepared. Two rounds of subtractive purification were performed using the non-phospho-peptide resin; protein A eluate was incubated with non-phospho-peptide resin by rotation in a sealed column at room temperature for 1 h to remove antibodies reactive with the non-phospho version of the protein antigen. The column was drained, and the flow-through (containing the desired antibody) was incubated with fresh non-phospho-peptide resin. The flow-through from this second subtractive step was next purified by incubation with phospho-peptide resin. After the phospho-peptide column was washed twice with PBS, phospho-specific antibody (bound to the resin) was eluted with 0.1 M glycine, pH 2.7, and pooled fractions were neutralized with 1 M Tris-HCl, pH 9.5 (ϳ1-2% of the fraction volume). The eluted phospho-specific antibody was then dialyzed overnight in PBS at 4°C.
The antibody against the consensus Akt substrate was raised against the following synthetic peptide antigen CXXXRXRXXT*XXXX, where X represents a position in the peptide synthesis where a mixture of all 20 amino acids (excluding cysteine) were used, and T* represents phosphothreonine. The synthetic phospho-peptide was conjugated to keyhole limpet hemocyanin and injected into rabbits. Once rabbits showed high phospho-specific titers, serum was then purified by protein A chromatography. Phospho-Akt substrate antibody was found to be highly phospho-specific as crude serum, so that a subtraction step on a column containing the non-phospho-peptide was not necessary, and the elution from the protein A column was used directly for affinity chromatography on a phospho-peptide-containing column.
To develop antibodies against the consensus conventional PKC substrate motif, the following synthetic degenerate peptide library was constructed CXXX(K/R)(K/R)S*(F/L/V)(K/R)(K/R)XXX, where K/R means lysine or arginine are present at that position in equal moles, F/L/V means phenylalanine, leucine, or valine are present at equal mole amounts, X represents any amino acid except Trp and Cys, and S* is phosphoserine. The synthetic phospho-peptide library was conjugated to keyhole limpet hemocyanin and injected into rabbits. Phosphospecific antibodies were purified as described as above.
Determine Motif Antibody Specificity by Screening Oriented Peptide Library-The sequence specificity for polyclonal antibodies made against the consensus Akt substrate motif was determined using the orientated-peptide library approach (29 -31). The first peptide library was synthesized as AXXXXXXXT*XXXXXXXAKKK, where X stands for a mixture of 19 amino acids (Cys was omitted). Sequence analysis of this peptide library indicates that it contains all 19 amino acids with less than a 3-fold variation for each amino acid at each degenerate position (data not shown). Phospho-substrate antibodies were bound to protein A beads by incubating 1 mg of affinity-purified antibody with 200 l of 50% slurry of pre-swelled protein A beads at 4°C overnight with gentle agitation. The beads were then washed 3 times with 1 ml of PBS with 0.5% of Nonidet P-40 followed by 2 washes with 1 ml of PBS. The beads were transferred to a microspin column (Bio-Rad) and washed 3 times with 1 ml of PBS. 1 mg of peptide library (30 mg/ml) was loaded to the antibody column and incubated at room temperature for 10 min followed by 1.5 h at 4°C. The antibody column was rapidly washed twice with 1 ml of ice-cold PBS ϩ 0.5% Nonidet P-40 and twice with 1 ml of ice-cold PBS. The bound peptides were eluted with 30% acetic acid at room temperature for 10 min. Peptides were separated from antibody by passing the eluent through a Centricon (10-kDa cut off) twice with 0.4 ml of 30% acetic acid. The solution was lyophilized or evaporated to dryness on a SpeedVac apparatus. The pellet was resuspended in 80 l of water, 0.5 l of aliquots were checked by MALDI-TOF to ensure the presence of eluted peptides, and 40-l aliquots were used for sequencing. We always performed a control experiment simultaneously by performing library screening on protein A beads alone and protein A beads immobilized with another irrelevant antibody, phosphotyrosine antibody, which does not bind the peptide in library.
Peptides eluted from the antibody columns were analyzed by automated amino acid sequencing (Applied Biosystems). The abundance of each amino acid at a given cycle in the sequence of the bound peptide mixture was divided by the abundance of the same amino acid in the same cycle of the starting peptide library. These raw preference values were then summed and normalized to the total number of amino acids in the degenerate position.
A second peptide library was synthesized as sequence MAXXXRXX-T*XGGGAKK, where X is all 18 amino acids except Cys and Trp, after the first library screening resulted in a strong selection for arginine at Ϫ3 position. The same procedure was repeated with the secondary library.
ELISA Analysis of Synthetic Peptides to Determine the Optimal Binding Motif of Antibodies-ELISA was performed after the standard ELISA protocol (48). Briefly, 50 l of 1 M synthetic phospho-and non-phospho-peptides were used to coat each well in 96-well plates. Coating was carried out overnight at 4°C. Phospho-Akt substrate and phospho-PKC substrate antibodies were used at 1:1000 dilution. The plates were incubated at 37°C for 2 h after the addition of primary antibody. An alkaline phosphatase-conjugated goat anti-rabbit antibody (Cell Signaling Technology) was used as secondary antibody, and p-nitrophenyl phosphate tablets (Sigma) were used for color development. Absorbance at 405 nm was read on an ELISA plate reader (PerkinElmer Model 1420-018).
Overexpression of Active Akt by Retrovirus-mediated gene Transfer-A constitutively active form of Akt was generated using the PCR containing the Akt-coding sequence modified by the addition of the Src myristoylation signal and the hemagglutinin epitope. This form of Akt was subsequently subcloned into a pBABE-Puro retroviral vector, and viral particles were generated by co-transfection of ecotropic packaging into HEK293 cells. The MyrAkt virus was then used to infect exponentially growing NIH 3T3 cells, which then selected with puromycin (1.5 g/ml). Pools of mock-infected and MyrAkt-infected NIH3T3 cells were starved in serum-free containing medium for 24 h before cell extracts were harvested for immunoblotting.
Two-dimensional Analysis of PDK1Ϫ/Ϫ Cells-70 g of wild type or PDK1Ϫ/Ϫ ES cells lysate (2 mg/ml) was precipitated by methanol and resuspended in 125 l of urea lysis buffer (9 M urea, 0.25% Triton X-100, 5 mM CHAPS, 2% ampholytes 4 -7 or 6-11, 2% ␤-mercaptoethanol, protease inhibitor mixture, and 0.1 M calyculin A). The lysate was then used to rehydrate 7-cm IPG strips after the manufacturer's recommendations (Amersham Biosciences). The rehydration process was carried out overnight at room temperature. For the first dimension (isoelectric focus), strips were run at 200 V for 0.01 h followed by 3500 V for 1.30 h and 3500 V for 3.30 h. Strips were then incubated with SDS buffer and subjected to the second dimension on 10% SDS gel at 130 V for 1.5 h. The gel was then transferred to polyvinylidene difluoride membranes and analyzed by Western bloting as described above.

Production of Antibodies to Akt and PKC Consensus
Motifs-To identify components of the Akt-and PKC-signaling pathways, we prepared antibodies reactive with short linear motifs phosphorylated by the Akt and PKC subfamilies of AGC protein kinases. Fig. 1 shows evolutionary relationships between the Akt subfamily, including Akt, p70 S6 kinase, ribosomal S6 kinase, and serum and glucocorticoid-inducible kinase (SGK). These enzymes phosphorylate consensus motifs of the form RXRXX(T*/S*), where X represents any amino acid, R represents Arg (sometimes Lys), and S*/T* represents phosphorylated serine/threonine (24,(33)(34)(35)(36). Conventional PKC family members phosphorylate substrates at serine or threonine in the context of arginine or lysine at the ϩ2 and/or Ϫ2 positions (37). The primary sequence around the site of phosphorylation is an important determinant of downstream kinase specificity in vivo (38). Phosphopeptide libraries with fixed residues corresponding to consensus Akt motif XXXRXRXXT*-XXX and the conventional PKC motif XXXRXS*XRXXXX were used as antigens to generate polyclonal antisera in rabbits. Phosphospecific antibodies were purified from rabbit sera as described under "Experimental Procedures," and their specificity was characterized as described below.
Antibody Specificities Determined by Oriented Peptide Library and Peptide Array ELISA-Purified phosphospecific polyclonal antisera prepared against the Akt consensus substrate motif was immobilized on protein A beads and used to select peptides from a synthetic oriented-peptide library, AXXXXXXXT*XXXXXXXAKKK, where X stands for a mixture of 19 amino acids (cystine omitted). Eluted peptides were sequenced, and the relative enrichment of each amino acid at different positions over the applied peptide library is shown in Fig. 2A. The Akt substrate antibody was strongly selective for arginine at position Ϫ3 (see Fig. 2A). To further determine antibody specificity, a second peptide library was synthesized where arginine at position Ϫ3 was fixed MAXXXRXXT*XGG-GAKK, and peptides bound to the antibody were eluted and sequenced. The relative preference of each amino acid at different positions relative to the fixed phosphothreonine over the applied peptide library is shown in Fig. 2B. These data indicate some selection for arginine at positions Ϫ6, Ϫ5, and Ϫ4. The same peptide library with fixed phosphoserine at position 0 was also synthesized, and the relative preference of Both the original libraries and antibody-captured fractions were sequenced. The relative enrichment of each amino acid at each position was calculated by dividing the relative amount of that amino acid after capture by its relative amount in the original library. A, first round screening using a peptide library where only the phosphorylated amino acid is fixed shows strong selectivity for arginine at position Ϫ3. B, second round screening using a peptide library where arginine is fixed at Ϫ3 to reveal additional amino acid binding selectivity. each amino acid position over the applied peptide library (data not shown) is similar to the phosphothreonine library results shown in Fig. 2. In contrast, when phosphotyrosine antibody was used to select peptides from the same peptide libraries, no peptides were present in eluent from antibody as determined by MALDI-TOF mass spectrometry (data not shown).
Epitope mapping of the Akt consensus substrate motif antibody was further determined by ELISA analysis together with arrays of synthetic phospho-and non-phospho-peptides. ELISA reactivities relative to phospho-GSK3 peptide are shown in Table I. Non-phospho-control peptides normally score in the range of 3-5% of phospho-GSK3. Akt consensus substrate motif antibody bound only to phospho-peptides. For phosphothreonine-containing peptides, arginine at position Ϫ3 is required, although lysine can substitute for arginine with weaker binding. For phosphoserine-containing peptides, arginine appears to be required at position Ϫ3 and at position Ϫ5 or Ϫ2. Good antibody binding was associated with hydrophobic amino acids at position ϩ1, small non-charged residues at position Ϫ1, and either small residues or arginine/lysine at position Ϫ2.
ELISA analysis of synthetic peptides was also used to determine the binding specificity of the PKC substrate motif anti-body. ELISA readings for each peptide are presented in Table  II as a percentage relative to phospho-AFX peptide. The data indicate that this antibody binds only phosphoserine-containing peptides, where phosphoserine is followed by arginine or lysine at position ϩ2. In addition, the antibody appears selective for hydrophobic amino acids at position ϩ1.
Detection of Phosphoproteins Using Akt and PKC Substrate Antibodies-Because the phospho-Akt and -PKC substrate antibodies appeared to be generally reactive with Akt and PKC substrates, we next tested their reactivity with cell extracts. NIH3T3 cell extracts prepared 15 min after platelet-derived growth factor treatment showed an increase in immunoreactive proteins when analyzed by Western blotting using the Akt substrate antibody (Fig. 3A). Antibody reactivity to a subset of proteins was diminished or lost upon treatment with the phosphatidylinositol 3-kinase inhibitor, LY 294002 (Fig. 3A). Analysis with phospho-specific antibodies to Akt Ser-473 and Thr-308 indicated elevated Akt phosphorylation in the same extracts (data not shown). Similar Western analysis was carried out using the PKC substrate antibody and WEHI231 and Jurkat cells treated with TPA, an activator of PKC. TPA treatment stimulated antibody reactivity to a number of different proteins (Fig. 3B). Antibody detection was compared with 32 P incorporation using cell extracts phosphorylated in vitro with purified active Akt kinase (Fig. 3C). Western blotting sensitivity using the Akt substrate antibody (chemiluminescent detection and 3-s exposure to x-ray film) appeared comparable with a 3-day 32 P autoradiograph.
We next introduced active Akt containing the Src myristoylation sequence into NIH3T3 cells by retrovirus-mediated gene  transfer and examined cell extracts using the Akt and PKC substrate antibodies. As shown in Fig. 3D extracts from cells expressing active Akt showed enhanced immunoreactivity over control extracts with the Akt substrate antibody, whereas little change was observed with the PKC substrate antibody (Fig.  3D). The Src myristoylation sequence directs Akt to membrane, and Akt became constitutively active as Akt itself became phosphorylated on Ser-473. Two known Akt substrates, GSK3 (39) and Forkhead related transcription factors (FKHR) (40), showed increased phosphorylation at known sites of Akt phosphorylation (Fig. 3D). The expression of constitutively active Akt also stimulated p70 S6 kinase activity (41) (data not shown).
Profiling PDK1 Null Cells Using the Akt Substrate Antibody-Because many AGC kinases such as Akt, p70 S6 kinase, serum and glucocorticoid-inducible kinase (SGK), ribosomal S6 kinase are inactive in PDK1Ϫ/Ϫ cells (15), we compared Akt substrate antibody immunoreactivity in wild type and PDK1Ϫ/Ϫ ES cells treated with IGF1, LY294002, and calyculin A. Many proteins were detected only in wild type ES cells and not PDK1Ϫ/Ϫ cells (Fig. 4A). In addition, IGF1 was seen to stimulate immunoreactivity only in wild type ES cells and not PDK1Ϫ/Ϫ cells. The molecular masses of proteins detected only in IGF1-stimulated wild type ES cells are indicated by arrows. For example, immunoreactivity against a prominent 30-kDa protein is stimulated by IGF1, inhibited by LY294002, and not present in PDK1Ϫ/Ϫ ES cells (Fig. 4A). The PDK1 and LY294002 dependence of this immunoreactivity suggests that this class of phosphoproteins may be substrates of AGC kinases downstream of PDK1 and phosphatidylinositol kinase. As expected, phosphorylation of Akt at Thr-308 and p70 S6 kinase at Thr-389 was not detected in PDK1Ϫ/Ϫ cells. Treatment of wild type and PDK1Ϫ/Ϫ cells with calyculin A, a specific inhibitor of protein phosphatase 2A and protein phosphatase 1, stimulated immunoreactivity with the Akt substrate antibody and also increased phosphorylation of Akt Ser-473 and ribosomal S6 kinase Ser-380 (Fig. 4A). Phosphorylation of known Akt substrates GSK3␣ (Ser-21) and GSK3␤ (Ser-9) was not detected in PDK1Ϫ/Ϫ cells even in the presence of IGF1. Elevated phosphorylation of GSK3␣ and -␤ was coincident with the increase of Akt phosphorylation at Ser-473 after calyculin A treatment, suggesting that these proteins are subjected to phosphatase control. Phosphorylation of p44/p42 mitogen-activated protein kinase (MAPK) is not affected by PDK1 deletion (Fig. 4A), whereas p90 ribosomal S6 kinase is not activated in PDK1 null cells (15), presumably because of lack of activation loop phosphorylation; we observed that p90 ribosomal S6 kinase Ser-380 phosphorylation decreased in PDK1 null ES cells compared with the wild type ES cells, similar to the decrease seen at Akt Ser-473. Two-dimensional electrophoresis followed by Western analysis using the Akt substrate antibody in pI 4 -7 range identified many proteins where immunoreactivity was lost in PDK1Ϫ/Ϫ extracts (Fig. 4B, top panel) or after LY294002 treatment (bottom panel). Phosphoproteins a, b, c, d, and g were only present in wild type ES cells (middle panel), and LY treatment abolished the phosphorylation of these proteins (bottom panel).
Phosphoproteins e and f were present in PDK1Ϫ/Ϫ cells (top panel), and LY treatment of PDK1 wild type cells induced the appearance of e (bottom panel). Two-dimensional/Western analysis of the same ES cell wild type (WT) and PDKϪ/Ϫ-and LY-treated extracts using the phospho-Akt substrate antibody in the pI 6 -11 range indicated that the prominent 30-kDa protein detected in wild type but not PDKϪ/Ϫ-or LY-treated extracts runs with a pI of ϳ10.4 (data not shown and Fig. 6D).
Analysis of Phosphoproteins Using the PKC Substrate Antibody-Because PDK1 is also known to phosphorylate the activation loops of many PKC isoforms (16 -18), we used the PKC substrate antibody to analyze changes in downstream phospho- rylation of PKC in PDK1Ϫ/Ϫ cells. Western blot analysis failed to show major changes in antibody immunoreactivity (Fig.  5A). Few differences were detected between wild type and PDK1Ϫ/Ϫ cells except for a 30-kDa protein, which appears to be similar to that detected with the Akt substrate antibody (see Fig. 4A). A threonine-specific antibody generated against the PKC substrate motif also (T*XR) failed to detect major changes (data not shown). We also examine the phosphorylation status of several different PKCs in PDK1Ϫ/Ϫ cells. Phosphorylation at the activation loop of PKC-␦ at Thr-505 was present but reduced in PDK1 null ES cells, whereas activation loop phosphorylation of PKC-/ at Thr-410/403 was completely abolished. PKC-␣ and -␤II autophosphorylation sites at Thr-638/ A, PDK1Ϫ/Ϫ and wild type cells were treated with IGF1 or calyculin A in the absence or presence of LY as indicated. Total cell lysates were blotted with phospho-PKC substrate antibody, anti-total PKC antibody, different phospho-specific antibodies against phosphothreonine at activation loops of PKC (Thr-410), PKC (Thr-403), PKC (Thr-538), and PKC␦ (Thr-505), phospho-specific antibody against the phosphorylation site in the linker sequence between the activation loop and hydrophobic motif of PKC␣/␤II (Thr-638/641), and phospho-specific antibody against the hydrophobic motifs of PKC isoforms, P-PKC (pan), as indicated on figure. B, Western blot analysis of wild type or PDK1Ϫ/Ϫ ES cells treated with TPA in the presence or absence of PKC inhibitor, Ro318220, using phospho-PKC substrate antibody. 641 and at their C-terminal hydrophobic motifs (detected by Phospho-PKC (pan)) were phosphorylated, but the level decreased overall (Fig. 5A). Western analysis using the PKC substrate antibody shows that both wild type and PDK1 null cells respond to TPA stimulation (Fig. 5B), which induces the activity of conventional and novel PKC isoforms. Pretreating cells with the PKC inhibitor, Ro318220, reduced antibody binding to several proteins. Few differences were detected using the PKC substrate antibody comparing wild type and PDK1 null ES cells. One exception is the 30-kDa protein also detected with the PKC substrate antibody (compare Fig. 5, A and B, with Fig.  4A).
Predicting Protein-Antibody Interaction Based upon Antibody Specificity-We next attempted to use a combination of immunoreactivity data and bioinformatics to identify the 30-Da protein detected using either the Akt or PKC substrate antibodies. We used two different types of information to generate an Akt substrate matrix that can be used to search protein databases to predict antibody reactivity. First, we used the method described by Yaffe et al. (42) to generate an Akt substrate antibody matrix based upon the relative amino acid preference values from the oriented library screening (Fig. 2). Second, we incorporated into the matrix data generated from peptide ELISA (Table I and data not shown) in which favorable or unfavorable contributions from individual amino acids are weighed quantitatively. We next used the Akt substrate antibody matrix together with the motif scanning algorithm, Scansite (scansite.mit.edu), to search protein sequence databases using the method described by Yaffe et al. (42). Using the Scansite program and the Akt substrate antibody matrix and searching Genpept protein data base gave many hits. Because a strong immunoreactive spot at 30 kDa with an isoelectric point of 10.5 was seen in wild type but not PDK1 null ES cells (Fig. 6A), we limited the search to mouse proteins in the molecular mass range of 27-32 kDa, with a pI range of 10 -11. These restrictions significantly narrowed the search. The top scores of this search are shown in Table III and include S6 ribosomal protein and several putative mouse proteins. The S6 ribosomal protein is abundant, has been reported to undergo phosphorylation after growth factor treatment (43), and contains overlapping Akt and PKC sites. This site also scores very well with an antibody matrix developed for the PKC antibody (data not shown). The phosphorylation site of S6 ribosomal protein identified by Scansite is known to be phosphorylated by p70 S6 kinase in a phosphatidylinositol 3-kinase/PDK1-and rapamycin-dependent fashion, and phosphorylation at this site correlates well with increased translation (44). Because the bioinformatic analysis suggested the 30-kDa protein to be S6 ribosomal protein, we further investigated this possibility.
Identification of the 30-kDa Phosphoprotein as S6 Ribosomal Protein-We next analyzed the 30-kDa protein by two-dimensional gel electrophoresis and Western blotting using the phospho-Akt substrate antibody and S6 ribosomal protein antibodies. As shown in Fig. 6A the phospho-Akt substrate antibody reacted with a 30-kDa protein in WT ES cells, but this protein was not detected in PDK1Ϫ/Ϫ cells or after treatment with LY 294002. A protein of identical mobility was detected after the blot was stripped and reprobed with S6 antibody in PDKϪ/ ϪES cells, WT ES cells, and WT ES cells treated with LY294002. This data strongly suggests that the 30 kDa, pI 10.4 protein detected using the phospho-Akt substrate antibody is S6 ribosomal protein.
Our bioinformatic analysis suggests Ser-236 as the phosphorylated site in the S6 ribosomal protein interacting with both phospho-Akt and PKC substrate antibodies. We next generated a phospho-specific antibody that recognizes only the Ser-236-phosphorylated form of S6 ribosomal protein. The same cell extracts shown in Fig. 4A were analyzed using the control and phospho-specific S6 Ser-236 antibodies. As shown in Fig. 6B, antibodies against total S6 ribosomal protein and phosphorylated S6 Ser-236 ribosomal protein reacted with a protein of identical mobility to that detected with the Akt and PKC substrate antibodies. Also, note that phosphorylation of S6 ribosomal proteins at Ser-236 is absent in PDK1Ϫ/Ϫ ES cells and The right panel shows the same membrane that was stripped and re-blotted with an antibody specific for S6 ribosomal protein. B, Western blot analysis of IGF1treated wild type or PDK1Ϫ/Ϫ ES cells in the presence of absence of phosphatidylinositol 3-kinase inhibitor, LY294002, using antibodies against total S6 ribosomal protein or phospho-S6 ribosomal protein (Ser-235/-236). C, PDK1Ϫ/Ϫ or wild-type (ϩ/ϩ) ES cells were immunoprecipitated with phospho-Akt substrate antibody. Total cell lysate, flow-through from immunoprecipitation, or antibody-immunoprecipitated fraction (PI) were analyzed by Western blotting using total S6 ribosomal protein antibody or phospho-(Ser/Thr) Akt substrate antibody.
completely blocked by LY294002 pretreatment in wild type ES cells (Fig. 6, A and B). To further verify that this band recognized by phospho-Akt substrate antibody is S6, cell lysates from PDK1 null or wild type ES cells were immunoprecipitated with the Akt substrate antibody and analyzed by immunoblotting using the S6 control antibody or the Akt substrate antibody. As shown in Fig. 6C, the phospho-Akt substrate motif antibody specifically immunoprecipitated phospho-S6 ribosomal protein from wild type ES cells, and the phosphorylated S6 ribosomal protein was detected by immunoblotting using S6 antibody and phospho-Akt substrate antibody. DISCUSSION Here we use directed peptide libraries as immunogens to produce antibodies reactive against short-linear phosphorylated motifs present in many different proteins. In this method, the immune response is driven by the fixed residues of the peptide antigen, as variable residues present at low concentrations are unable to promote significant immune reactions. As demonstrated here, the method can be used to produce antibodies reactive against phosphorylated motifs characteristic of protein kinase substrates but is not limited to phosphorylated motifs and can be applied to a wide variety of modified protein sequence motifs. An important advantage over existing methods is the ability to produce antibodies that recognize subgroups of kinase substrates or consensus sites for other modifications including acetylation, methylation, and proteolytic cleavages. The method is an improvement over the use of modified amino acids as immunogens, first, placing the modified amino acid in the context of a variable peptide backbone tends to select antibodies that are more broadly reactive (tolerant to different amino acids), and second, incorporation of fixed residues provides additional surface contacts and binding energy for antibody-antigen interactions to generate higher affinity interactions. These two effects produce antibodies that are focused on fixed residues yet broadly tolerant at variable positions.
Although we have successfully used this method to generate antibodies recognizing motifs ranging in size from a single modified amino acid to more extended motifs involving threefour additional fixed residues, one problem is that polyclonal antibody responses against extended motifs (greater than three fixed residues) often consist of multiple specificities, each recognizing different portions of the motif. Obtaining extended specificities from polyclonal mixtures may, therefore, require purifying away antibodies reactive with shorter motifs. The method can also be extended to include mixtures of several residues at a given fixed position; however, certain immunodominant amino acids tend to bias the immune response, producing distortions in reactivity. In addition, because specificity analysis requires analysis of many different substrates, the production of monoclonal antibodies is preferred to mini-mize time spent characterizing different antibody preparations. This effort is currently ongoing, and we have successfully produced monoclonal antibodies against 14-3-3 binding motifs and antibodies directed against the hydrophobic PDK1-docking motif.
We anticipate several important applications of these antibodies. First, as demonstrated here, kinase substrate antibodies will be useful in identifying cellular phosphorylation states associated with specific deregulated signaling pathways. Deregulated protein phosphorylation is associated with many disease states including cancer and diabetes (45). Because phosphotyrosine antibodies helped to identify and characterize many proteins involved in growth factor and cytokine signaling, we anticipate that phosphoserine/threonine motif antibodies will prove equally useful in characterizing signaling pathways, remembering that greater than 90% of mammalian protein phosphorylation occurs on serine or threonine (6). For example, loss of PTEN expression in multiple tumors results in deregulated phosphatidylinositol 3-kinase signaling and upregulation of the Akt survival and anti-apopototic pathways. Analysis of Akt signaling using substrate motif antibodies will report overall pathway activity as well as the phosphorylation status of key substrates involved in survival, apoptosis, and glucose metabolism. This approach has been successfully used to identify several new Akt substrates including tuberin (26) and AS160, a Rab GAP domain containing protein (46).
Second, this class of antibody can be used together with bioinformatic analysis to predict and identify new phosphoproteins and new sites of protein phosphorylation. As demonstrated in this work, understanding the sequence requirements of the epitope recognized by the motif antibody allows construction of a specificity matrix useful in searching protein data base to predict sites that when phosphorylated will react with the antibody. Algorithms such as Scansite (42) can establish an ordered queue of proteins aligned by expectation of antibody reactivity. Here, the predicative power is determined by the length and degeneracy of the epitope recognized; motifs longer than five or six fixed residues are expected at random to be unique within a mammalian genome, whereas motifs of two or three fixed residues are expected to be unique within an average sized protein. Hence, antibodies that report phosphorylated residues in the context of two surrounding residues, for example the Akt RXRXXS* motif, are likely to predict the site of phosphorylation within a single protein but require additional information to uniquely identify phosphorylated proteins within a complex mixture as found within a cell or tissue. Additional constraints that can reduce the data base search include molecular mass, isoelectric point, and cell-specific expression information. These constraints have been recently incorporated into the Scansite algorithm, dramatically improving target predictions. Other constraints include reactivity with multiple motif-specific antibodies, as seen here with S6 ribosomal protein interacting with both Akt and PKC substrate antibodies but not phosphothreonine or T*XR specificities. As additional phospho-motif antibodies are developed and their specificities identified, we anticipate the development of a broad and increasingly powerful "tool box" of phosphorylationspecific reagents that when combined with bioinformatic data base searches, such as Scansite, will accelerate the identification of phosphoproteins. Finally, phospho-motif antibodies can be used to identify new sites of phosphorylation within a single protein. First, the protein of interest can be immunoprecipitated from activated and control cells and immunoblotted with a panel of phosphomotif antibodies. If phosphorylation is detected, the protein can be analyzed using Scansite and the motif antibody specificity matrix to predict likely antibody interactions and sites of phosphorylation. Mutation of candidate sites can identify sites of in vivo phosphorylation and site-specific antibodies developed. Alternatively, phosphoproteins can be immunoprecipitated using motif antibodies, purified by gel electrophoresis, and identified by mass spectrometry. This approach has been used by Kane et al. (46) to identify AS160 protein as an insulin-regulated phosphoprotein and to identify several other potential Akt substrates. A similar approach has also been used to identify a new PKA phosphorylation site on Src that regulates Src activity (47). As illustrated by these examples, the development of phosphoserine/threonine specific-motif antibodies provide an important new class of reagents useful in tracing signaling pathways and identifying new sites of protein phosphorylation.