Proteomic Identification of Bcl2-associated Athanogene 2 as a Novel MAPK-activated Protein Kinase 2 Substrate*

The p38 MAPK cascade is activated by various stresses or cytokines. Downstream of p38 MAPKs, there are diversification and extensive branching of signaling pathways. Fluorescent two-dimensional difference gel electrophoresis of phosphoprotein-enriched samples from HeLa cells in which p38 MAPK activity was either suppressed or activated enabled us to detect (cid:1) 90 candidate spots for factors involved in p38-dependent pathways. Among these candidates, here we identified four proteins including Bcl-2-associated athanogene 2 (BAG2) by peptide mass fingerprintings. BAG family proteins are highly conserved throughout eukaryotes and regulate Hsc/Hsp70-mediated molecular chaperone activities and apoptosis. The results of two-dimensional immunoblots suggested that the phosphorylation of BAG2 was specifically controlled in a p38 MAPK-dependent manner. Furthermore, BAG2 was directly phosphorylated at serine 20 in vitro by MAPK-activated protein kinase 2 (MAPKAP kinase 2), which is known as a primary substrate of p38 MAPK and mediates several p38 MAPK-dependent processes. We confirmed that MAPKAP kinase 2 is also required for phosphorylation of BAG2 in vivo . Thus, p38 MAPK-MAPKAP kinase 2-BAG2 phosphorylation cascade may be a novel signaling pathway for response to extracellular rehydrated into 24-cm pH 4–7 immobilized pH gradient strips followed by isoelectric focusing, resolved by 12% SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. Immunoblotting was performed using the following antibodies: anti-phospho-MAPKAPK2 (Thr 334 ), an-ti-phospho-HSP27 (Ser 82 ), anti-phospho-c-Jun (Ser 63 ), anti-phospho- p44/p42 MAPK (Thr 202 /Tyr 204 ), and anti-MAPKAPK2 (Cell Signaling Technology); anti-actin (Chemicon International); and anti-HA High Affinity (3F10) (Roche Applied Science).

macrophages with endotoxic lipopolysaccharide (4). It was also cloned as a specific target of pyridinyl imidazoles such as SB203580, which were known to inhibit biosynthesis of proinflammatory cytokines in lipopolysaccharide-stimulated monocytes (5). SB203580 is a highly selective inhibitor of p38 with no effect on ERK, c-Jun N-terminal kinase, and many other protein kinases (6).
p38 is rapidly activated by diverse stresses (hyperosmolarity, UV light, heat shock, arsenite, and anisomycin) as well as by endotoxins and cytokines (interleukin-1 and tumor necrosis factor). Activated p38 phosphorylates and regulates downstream protein kinases and certain transcription factors (7). Multiple p38-dependent inflammatory responses are mediated by a serine/threonine kinase, MAPK-activated protein kinase 2 (MAPKAPK2) (8 -10) and possibly by other MAPKAPK family members. Previous studies also showed the involvement of p38 MAPK signaling pathways in several apoptotic regulations. Undevia et al. reported a proapoptotic signaling pathway involving p38 MAPK, which is elicited by transforming growth factor-␤ 1 (11). In contrast, Varghese et al. (12) reported that activation of p38 MAPK is critical to cell survival by tumor necrosis factor in U937 cells.
Recently, two-dimensional gel electrophoresis has been applied to analyze intracellular signaling events. However, due to insufficient resolution power, the dense spots of components for cytoskeleton or housekeeping metabolic enzymes have frequently obscured low abundance proteins such as factors involved in signal transduction. To overcome this problem, prefractionation of proteins is highly desirable. Prefractionation includes purification of organelles, fractionation of proteins according to their isoelectric points (pIs), isolation of specific protein complexes, purification of glycoproteins by lectins, conventional chromatography, and so on. Another prefractionation procedure, phosphoprotein purification, is suitable for the identification of substrates of a kinase of interest, since it enriches phosphorylated substrates while reducing the number of protein species, which greatly facilitates protein identification.
In this study, we combined three key methodologies, enrichment of phosphorylated proteins, fluorescent two-dimensional difference gel electrophoresis (2D-DIGE), and mass spectrometric identification of proteins, to globally identify the factors involved in the p38 MAPK cascades. We have identified Bcl2associated athanogene 2 (BAG2) protein, splicing factor, and Ran-binding protein as candidates for targets of p38 MAPKdependent phosphorylation in response to anisomycin treatment in HeLa cells. Furthermore, we provided definite evidence that MAPKAPK2 phosphorylates BAG2 at Ser 20 in vitro and in vivo. These results demonstrate that BAG2 is a novel component of the p38 MAPK signaling pathways.

EXPERIMENTAL PROCEDURES
Cell Lines and Materials-HeLa and 293 cells were purchased from the ATCC (Manassas, VA). Both cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Transfections were performed using LipofectAMINE 2000 reagent (Invitrogen) following the manufacturer's instructions. Protein kinase inhibitors U0126, SP600125, and SB203580 were purchased from Calbiochem.
Identification of Signaling Mediators of the p38 MAPK Cascade-Total cell lysates were prepared from HeLa cells cultured in the normal condition and cells precultured in the presence or absence of 10 M SB203580 for 50 min followed by treatment with 1 g/ml anisomycin for 15 min. To enrich phosphorylated proteins, each extract was subjected to a phosphoprotein purification kit (Qiagen), and protein concentration was determined according to Bradford (20). Then 2D-DIGE was carried out as follows. Three purified samples (50 g each) were labeled with Cy2, Cy3, and Cy5 minimal dyes (Amersham Biosciences), respectively, following the manufacturer's instructions. Labeled samples were combined and passively rehydrated into a 24-cm pH 4 -7 immobilized pH gradient strips (Amersham Biosciences) for 12 h and subjected to isoelectric focusing using the Ettan IPGphor isoelectric focusing system (Amersham Biosciences) for a total of 61,500 V-h (hold at 500 V for 500 V-h, hold at 1 kV for 1 kV-h, and hold at 8 kV for 60 kV-h). Reduction and carbamidomethylation of cysteine sulfhydryls were carried out as recommended by the manufacturer. The second-dimensional SDS-PAGE was carried out with Ettan DALTtwelve (Amersham Biosciences). The Cy2, Cy3, and Cy5 signals were individually imaged with mutually exclusive excitation/emission wavelengths of 480/530, 520/ 590, and 620/680 nm, respectively, using Typhoon 9400 (Amersham Biosciences). DeCyder software (Amersham Biosciences) was used for pairwise comparisons of Cy2 and Cy3, Cy2 and Cy5, or Cy3 and Cy5 images and for quantitative determination of intensity changes of all spots among three combinations.
Determination of Proteins of Interest Using 2D-DIGE and Mass Spectrometry-Phosphoprotein-enriched sample prepared as above from anisomycin-treated cells (650 g, nonlabeled) was subjected to second dimension gel, transferred to ProBlott membranes (Applied Biosystems), and stained with Coomassie Brilliant Blue. Proteins of interest were excised, equilibrated with a reduction buffer (0.5 M Tris, pH 8.5, 8 M guanidine hydrochloride, 0.3% EDTA, 5% acetonitrile), and digested with 1 pmol of lysylendopeptidase (Wako) in 6 l of digestion buffer (18 mM Tris, pH 8.9, 70% acetonitrile) for 90 min at 37°C. Acetonitrile was added to a final concentration of 90%, and the sample was desalted/concentrated using ZipTip HPL (Millipore Corp., Bedford, MA). Peptides were eluted with 1 l of matrix solution (0.1% acetic acid, 50% acetonitrile saturated with ␣-cyano-4-hydroxycinnamic acid) and applied onto a sample plate (Applied Biosystems). Matrix-assisted laser desorption ionization time-offlight (MALDI-TOF) mass spectrometry was performed using a Voyager-DE PRO (PerSeptive Biosystems). Ions specific for each sample were then used to interrogate human protein sequences in the NCBInr data base using the MASCOT (available on the World Wide Web at www. matrixscience.com) data base search algorithms.
In Vitro Kinase Assays-Phosphorylation of recombinant GST-BAG2 or GST-BAG2-S20A by MAPKAPK2 was examined by incubation of 20 ng of active-MAPKAPK2 (Upstate Biotechnology Inc., Lake Placid, NY) or 1 unit of ERK2 (active MAPK; Calbiochem) with 1.0 g each of GST-BAG2, GST-BAG2-S20A, or GST-Hsp27 and 0.5 Ci of [␥-32 P] in 20 l of a kinase buffer (20 mM Tris, pH 7.5, 10 mM MgCl 2 , 0.1 mM ATP) for 20 min at 30°C. Reactions were terminated with Laemmli SDS sample buffer to a final volume of 30 l, halves of samples were subjected to 10% SDS-PAGE, and phosphorylation reactions were visualized by autoradiography.
Construction and Transfection of siMAPKAPK2-A short interfering RNA (siRNA) targeting endogenous MAPKAPK2 was generated using BLOCK-iT Dicer RNAi kits (Invitrogen) following recommended protocols, except that the target sequence was amplified by PCR using primers (5Ј-CAAGAAGAACGCCATCATC-3Ј and 5Ј-GAGGGTTGGAT-GCATCTT-3Ј). Purified siMAPKAPK2 (250 or 500 ng/well) was used for transfection into 293 cells cultured in 6-well plates.

Identification of Signaling Molecules in the p38 MAPK Cas-
cade-To globally identify factors involved in the p38 MAPK cascade, we developed a system consisting of phosphoprotein purification, fluorescent 2D-DIGE, and mass spectrometric identification of proteins. We prepared three lysates of HeLa cells (control untreated cells and cells treated with anisomycin in the presence or absence of SB203580), in which p38 MAPK activity was suppressed or activated, respectively. The specificity of SB203580 inhibition of p38 MAPK pathways was also confirmed as shown later. Phosphoproteins were enriched by using a commercially available purification kit. We confirmed that phosphorylated forms of p38 MAPK, MAPKAPK2, and Hsp27 bound to the column and were eluted with relatively good recoveries from the column (data not shown).
Phosphoprotein-enriched fractions from the three lysates were labeled with cyanine dyes Cy2, Cy3, and Cy5, respectively; combined; and run on the same gel. The Cy2 (control cells), Cy3 (anisomycin), and Cy5 (anisomycin ϩ SB203580) signals were individually scanned at mutually exclusive excitation/emission wavelengths, which are exhibited as blue, red, and green pseudocolors (Fig. 1, A-C). Fig. 1D shows a merge of three images. There were a large number of red spots, the amount of which was higher in p38activated cells than in the other two cells (88 and 101 spots exhibiting a 1.5-fold or higher increase among 2865 and 2891 spots in two separate experiments employing different cell lysates). Since phosphoproteins were enriched before the analysis, these spots were candidates for proteins phosphorylated in p38 signaling pathways. Most of the remaining spots were white as a merge of blue, red, and green, demonstrating their equal quantities among three samples. These spots may be the phosphoproteins unrelated to the p38 pathway or proteins nonspecifically trapped by the column. Green spots were also observed, some of which were characterized as described below.
Determination of Proteins Using 2D-DIGE and Peptide Mass Fingerprinting-Phosphoprotein-enriched sample derived from HeLa cells treated with anisomycin (650 g, nonlabeled) was separated by two-dimensional gel electrophoresis, transferred to a polyvinylidene difluoride membrane, and stained with Coomassie Blue. The spots corresponding to those of a-f in Fig. 1E (magnified view of a boxed region in D) were excised and subjected to peptide mass fingerprinting as described under "Experimental Procedures." The MALDI-TOF mass spectrum of one of these proteins is shown in Fig. 2. Using the MASCOT data base search algorithms, this protein was identified as human BAG2 (spot a). Other spots were Hsp27 (spots b-d), human splicing factor arginine/serine-rich 9 (spot e), and human Ran-binding protein (spot f). Since the activation of p38 by anisomycin caused Hsp27 phosphorylation at Ser 15 , Ser 78 , and Ser 82 (21,22), spots b-d may correspond to Hsp27 at different phosphorylation states. Ran-binding protein (spot f) exhibited a green spot suggesting a phosphorylation by a SB203580-sensitive pathway. The peptide coverages of the proteins and scores in the MASCOT analyses are shown in Table  I. The ratios of signal intensities of these spots between three sample combinations by using images shown in Fig. 1 are also shown.
BAG2 and splicing factor arginine/serine-rich 9 (spots a and e) were significantly up-regulated in a p38-dependent manner just as Hsp27, a known substrate phosphorylated in the p38-MAPKAPK2 pathway, on spots c and d, suggesting that BAG2 and splicing factor arginine/serine-rich 9 are novel factors involved in the p38-dependent phosphorylation cascades. In this study, we focused on BAG2 as a candidate of p38-signaling factor and characterized its phosphorylation in the following experiments.
BAG2 Phosphorylation Is Specifically Controlled by p38 MAPK-To confirm that BAG2 was specifically phosphorylated by the p38 cascade in response to treatment of anisomycin, we used HeLa cells transiently overexpressed with N-terminal HA-tagged human BAG2. The transfected cells were treated with or without anisomycin in the presence or absence of specific protein kinase inhibitors. Consistent with previous reports, SB203580 inhibited the activation of MAPKAPK2 by p38 and thereby inhibited Hsp27 phosphorylation (23), SP600125 inhibited phosphorylation of c-Jun (24), and U0126 inhibited the activation of ERK1/2 (25) (Fig. 3A). These inhibitors did not affect the expression of HA-BAG2.
These total lysates were used for second dimension immunoblot with anti-HA antibody (Fig. 3B). We detected four spots (indicated as arrows a-d) of HA-BAG2 in control HeLa cells. Treatment with anisomycin resulted in a remarkable increase in the two spots (c and d) with more acidic pI values and a corresponding decrease in the remaining two spots (a and b). These shifts in pI values were significantly inhibited by preincubation of cells with SB203580 but neither with SP600125 nor with U0126. These results suggested that BAG2 was specifically phosphorylated by p38 MAPK cascades in response to anisomycin treatment.
BAG2-Ser 20 Is Directly Phosphorylated by MAPKAPK2 in Vitro -To identify the phosphorylation site on BAG2 specifically phosphorylated in response to anisomycin, we scanned the whole peptide sequences of BAG2 of three distinct species (Homo sapiens, Mus musculus, and Danio rerio) for possible phosphorylation sites. Then we found a discriminative sequence in human BAG2 (residues 13-22) that completely matched with the consensus MAPKAPK2 phosphorylation site, Hyd-X-Arg-X-X-Ser, where Hyd is a bulky hydrophobic residue (Phe Ͼ Leu Ͼ Val Ͼ Ͼ Ala) and Ser is a phosphorylation site (26). As illustrated in Fig. 4 with a schematic domain structure of human BAG2, this sequence motif in human BAG2 is perfectly conserved among three species.
We therefore constructed a vector plasmid expressing a mutant human BAG2 in which Ser 20 was mutated to alanine to determine whether BAG2-Ser 20 was directly phosphorylated by MAPKAPK2. Recombinant GST-BAG2, GST-BAG2-S20A, or GST-Hsp27 was incubated with recombinant MAPKAPK2 or ERK in the presence of [␥-32 P]ATP in vitro. Fig. 5A shows that GST-BAG2-wt as well as GST-Hsp27 were strongly phospho- rylated by MAPKAPK2 and that substitution of alanine for Ser 20 completely abolished MAPKAPK2 phosphorylation of BAG2. In contrast, ERK did not phosphorylate GST-BAG2-wt at all. The kinase activity of ERK was confirmed strong phosphorylation of myelin basic protein (27) (data not shown). The result demonstrates that MAPKAPK2 directly phosphorylates BAG2 at Ser 20 .
BAG2-Ser 20 Is Phosphorylated in Vivo in a p38-dependent Manner-To confirm that the phosphorylation of BAG2-Ser 20 was the phosphorylation site in response to anisomycin in vivo, two-dimensional immunoblot was performed using HeLa cells expressing BAG2-wt or BAG2-S20A mutant. As shown in Fig.  5B, mutation at Ser 20 dramatically inhibited the pI shifts (left panel) that were observed in wild type BAG2 in response to anisomycin treatment (right). Moreover, only two forms of BAG2 with more basic pI values (spots a and b) were detectable in HeLa cells expressing mutant BAG2, although four forms of BAG2 with different pIs were detectable in HeLa cells expressing wild type BAG2 (spots a-d). Therefore, it is highly likely that spots c and d correspond to BAG2 phosphorylated at Ser 20 by p38-dependent pathways. These results demonstrated that BAG2 was phosphorylated at Ser 20 in response to anisomycin treatment in vivo. Judging from the pattern of two-dimensional immunoblot, another phosphorylation site independent of p38 activation was suggested.
MAPKAPK2 Phosphorylates BAG2 in Vivo in Response to Anisomycin Treatment-MAPKAPK2 is activated downstream of p38 MAPK, and we have proved that recombinant MAP-KAPK2 is able to phosphorylate GST-BAG2 in vitro (Fig. 5A). Next we generated siRNA targeting endogenous MAPKAPK2 to evaluate the role of MAPKAPK2 in the p38 MAPK cascade in respect to phosphorylation of BAG2. To ascertain the efficiency of siRNA, 293 cells cultured in 6-well plates were transfected with 250 or 500 ng of siMAPKAPK2. Two days after transfection, cell lysates were prepared and subjected to immunoblot with anti-MAPKAPK2 antibody (Fig. 6A). The result showed that 250 ng of siMAPKAPK2 were sufficient to suppress the endogenous expression of MAPKAPK2. Phosphorylation of BAG2 in response to anisomycin treatment was then examined by two-dimensional immunoblot in 293 cells co-transfected with 500 ng pEF-BOS/HA-BAG2 and either 500 ng of control RNA (top panel) or 500 ng of siMAPKAPK2 (second panel).  Table I. IB, immunoblot. SB, SB203580; SP, SP600125; Ani., anisomycin. a Fluorescent intensity ratios Cy3 (anisomycin)/Cy2 (nontreated), Cy5 (SB203580 ϩ anisomycin)/Cy2 (nontreated), and Cy3 (anisomycin)/Cy5 (SB203580 ϩ anisomycin) were calculated by DeCyder software using images in Fig. 1.
b Sequence coverages of MALDI-TOF peptide mass fingerprint analyses are also shown.

DISCUSSION
Cell growth, differentiation, and apoptosis are regulated by diverse extracellular signals. The MAPK cascades integrate and process various extracellular signals by phosphorylating substrates, which alters their catalytic activities and conformations or creates binding sites for protein-protein interactions. To identify proteins that are involved in MAPK cascades, diverse approaches were developed. We have shown here that components of the p38 MAPK cascade can be identified by a proteomic approach. Our approach consisted of phosphoprotein purification, 2D-DIGE, and identification of proteins by peptide mass finger printing. This method enabled us to identify many spots unique to p38 MAPK-activated cells, which may be the candidates for the factors involved in the signaling pathway. Our approach is suitable to identify components that are phosphorylated within minutes after stimulation. Previous approaches such as transient expression of the active form of kinases in the cascade may not be appropriate to detect such factors. Interaction between a kinase and its substrates is generally not strong enough to pull down its substrates by immunoaffinity purification. A yeast two-hybrid system often produces false positives.
The enrichment of phosphoproteins may be effective to identify minor components such as factors involved in signal transduction, since when total cell lysates from p38 MAPK-activated and -suppressed cells were analyzed by 2D-DIGE under the same conditions, most of the red spots detected by using phosphoprotein-enriched samples were hardly visible hidden behind other spots of nonphosphorylated proteins (data not shown). We detected numerous red or orange spots at a higher molecular weight range (Fig. 1A). By using narrower pH ranges for the first dimension and polyacrylamide gels at different concentrations for the second dimension, more candidates will be detected. We are focusing our effort to identify these proteins by peptide mass fingerprinting.
We identified here BAG2 and splicing factor arginine/serinerich 9 the phosphorylation of which varies upon stimulation of p38 MAPK cascade together with Hsp27, a known substrate for p38-activated MAPKAPK2. All of spots b-d in Fig. 1 were Hsp27. Since Hsp27 proteins are phosphorylated by MAP-KAPK2 at three sites, spots c and d may reflect different phosphorylation states. Since spot b was green (more abundant in p38-suppressed cells) and spots c and d were red (more abundant in p38-activated cells) with more acidic pI values, Hsp27 in spot b may undergo shifts to spots c and d by p38-dependent phosphorylation.
MAPKAPK2 regulates heterogenous nuclear ribonucleoprotein A0 biding to cytokine mRNAs through phosphorylation. Arginine/serine-rich 9 also binds to mRNA of heterogenous nuclear ribonucleoprotein A1 and regulates its alternative splicing. If the activity of arginine/serine-rich 9 can be changed through phosphorylation, it would provide another example of regulation of mRNA metabolism by p38/MAPKAPK2 cascade.
BAG2 was phosphorylated in a p38 MAPK-dependent manner in response to anisomycin stimulation. An in vitro kinase assay clearly demonstrated that MAPKAPK2 phosphorylated BAG2 on Ser 20 . Moreover, by utilizing a Ser to Ala mutant, we confirmed that BAG2-Ser 20 was phosphorylated in a p38 MAPKdependent manner. The results of two-dimensional immunoblot using the BAG2-S20A mutant suggested that there was another phosphorylation site on BAG2. The detail of the phosphorylation is currently unknown. siRNA-mediated inhibition of MAPKAPK2 expression significantly but not completely decreased the shifts in pI of BAG2 (Fig. 6), suggesting that p38dependent kinase(s) other than MAPKAPK2 exists. Possible candidates are other MAPKAPK family members, since their substrate specificity resembles that of MAPKAPK2 (28). Previous studies demonstrated that BAG family proteins are known as pivotal binding partner proteins of Hsc/Hsp70 molecular chaperones. All of the BAG1, BAG2, the BAG3 proteins interact with ATPase domain of Hsc/Hsp70 through the BAG domains, and suppress the chaperon activities of Hsc/Hsp70 in vitro (29). Furthermore, most of BAG family proteins play important roles in the regulation of apoptosis, cell survival, and stress response. For example, BAG4 (SODD) was identified as a binding partner for tumor necrosis factor receptor-1 (TNF-R1), death receptor 3, and Bcl-2 and was shown to act in an antiapoptotic manner (30,31). BAG3 was also reported as a regulator of stress-induced apoptosis in normal and neoplastic leukocytes (32). These previous results suggested the involvement of BAG2 in the regulation of apoptosis. We therefore extensively examined the effect of BAG2 expression using wild type and mutants of S20A and S20D (phosphomimetic). However, none of them accelerated or inhibited the anisomycininduced apoptosis of HeLa cells so far examined (data not shown).
We confirmed the previous results and found that BAG2 constitutively bound to Hsp70 irrespective of the cellular stresses in vivo and that this tight binding to Hsp70 resulted in the strong inhibition of chaperon activity of Hsp70 in the refolding of denatured luciferase (data not shown). We also tested the effect of phosphorylation on the binding of BAG2 to Hsp70 using wild type and S20A and S20D mutants. However, these BAG2 proteins did not show significant difference in the binding to Hsp70 in vivo and in the inhibition of the chaperon activity.
BAG family proteins share BAG domains at their C terminus, whereas their N-terminal regions did not show significant similarity, of which functions remain unclear. It is known that heat shock proteins form giant multimolecular complex possibly containing Hsp-binding proteins such as BAG2 (33). In this regard, it is of note that BAG2 has a coiled-coil domain near the amino terminus (Fig. 4). This coiled-coil domain starts at Ser 20 ,  ). B, twodimensional immunoblots with anti-MAPKAPK2 and anti-HA antibodies. Transfection of siMAPKAPK2 suppressed both endogenous expression of MAPKAPK2 and pI shifts of HA-BAG2 in response to anisomycin stimulation. To demonstrate the complete suppression of MAPKAPK2 expression, the blots were exposed for a rather longer period of time than in Figs. 3 and 5. which is the phosphorylation site by MAPKAPK-2. Since the coiled-coil domain functions in protein-protein interaction, it may be an interesting possibility that BAG2 interaction with other unidentified protein through the coiled-coil domain is regulated through the phosphorylation of BAG2 within the complex.
In this study, we have demonstrated the method constituting of phosphoprotein purification, 2D-DIGE, and mass spectrometric determination of the proteins for the identification of substrates for p38 MAPK cascades. Our approach could be applicable to any protein kinase of which a specific inhibitor would be available. Therefore, this method will provide a wide range of applications to diverse biological phenomena.