Targeted Proteomic Analysis of 14-3-3ς, a p53 Effector Commonly Silenced in Cancer*S

To comprehensively identify proteins interacting with 14-3-3ς in vivo, tandem affinity purification and the multidimensional protein identification technology were combined to characterize 117 proteins associated with 14-3-3ς in human cells. The majority of identified proteins contained one or several phosphorylatable 14-3-3-binding sites indicating a potential direct interaction with 14-3-3ς. 25 proteins were not previously assigned to any function and were named SIP2–26 (for 14-3-3ς-interacting protein). Among the 92 interactors with known function were a number of proteins previously implicated in oncogenic signaling (APC, A-RAF, B-RAF, and c-RAF) and cell cycle regulation (AJUBA, c-TAK, PTOV-1, and WEE1). The largest functional classes comprised proteins involved in the regulation of cytoskeletal dynamics, polarity, adhesion, mitogenic signaling, and motility. Accordingly ectopic 14-3-3ς expression prevented cellular migration in a wounding assay and enhanced mitogen-activated protein kinase signaling. The functional diversity of the identified proteins indicates that induction of 14-3-3ς could allow p53 to affect numerous processes in addition to the previously characterized inhibitory effect on G2/M progression. The data suggest that the cancer-specific loss of 14-3-3ς expression by epigenetic silencing or p53 mutations contributes to cancer formation by multiple routes.

Protein Detection by Immunofluorescence and Microscopy-DLD1-tTA cells were grown on glass coverslips and transfected with the indicated plasmids. After 24 h cells were fixed in 3.7% paraformaldehyde, PBS; permeabilized with PBS, 0.2% Triton X-100; blocked with fetal bovine serum; and stained with primary (rabbit anti-HA) and with secondary anti-rabbit IgG-Cy3 antibodies (Jackson Immunoresearch Laboratories). Images of immunofluorescence and green fluorescent protein or CFP fusion proteins were generated with an inverted microscope (Axiovert 200M, Zeiss) equipped with a CCD camera (Coolsnap HQ, Photometrics) and Metamorph software (Universal Imaging Corp.).
Multidimensional Protein Identification Technology (MudPIT)-Precipitated 14-3-3-TAP-associated protein preparations were dissolved in digestion buffer, digested with trypsin, and analyzed by LC/LC/MS/MS according to published protocols (25). Approximately 100 g of protein were used for a 12-step LC/LC/MS/MS analysis. The obtained MS/MS spectra were analyzed by SEQUEST 2.7 using a non-redundant mammalian data base (May 2003 release, NCBI). The SEQUEST outputs were then analyzed by DTASelect (26). The DTASelect filter settings were: XCorr: ϩ1 ions, 1.8; ϩ2 ions, 2.5; ϩ3 ions, 3.8; ⌬CN, 0.08; only half or full tryptic peptides were considered, and all subset proteins were removed (the "-o" option in DTASelect). Proteins with four to five peptides that passed the DTASelect filter were considered real hits. Proteins with one to three peptides that passed the DTASelect filter were manually validated.
Wounding Assay-DLD1-tTA-14-3-3-HA cells were seeded at 80% confluency into 6-well plates 24 h after removal of doxycycline and cultivated for an additional 24 h. As a control, DLD1-tTA-14-3-3-HA cells were treated with 100 nM doxycycline during the whole experiment to suppress expression of 14-3-3-HA. To prevent proliferation, cells were treated with 10 g/ml mitomycin C (Sigma) for 3 h. Subsequently the cell monolayer was scratched with a Pasteur pipette. Cells were washed, and fresh medium with or without doxycycline was added. Cell migration was monitored for 24 h using an Axiovert 200M microscope (Zeiss) integrated into a CO 2 37°C incubator (Life Imaging Services) and equipped with a CCD camera (Coolsnap HQ, Photometrics) and Metamorph software (Universal Imaging Corp.). Pictures were taken every 6 min in two different wells with 50-ms exposure time using a motorized XY precision stage (LEP).

Proteomic Analysis of 14-3-3-associated Ligands-To
comprehensively determine the subset of proteins associated with 14-3-3 in vivo we used a TAP tag approach, which allows the isolation of native protein complexes from cells ectopically expressing the tagged protein of interest (27). The TAP tag was fused to the C terminus of 14-3-3 (14-3-3-TAPc; Fig. 1A). 14-3-3-TAPc showed a cytoplasmic localization identical to the previously described localization for endogenous 14-3-3 protein (9) and to HA-tagged 14-3-3 ( Fig. 1B). Ectopic 14-3-3-TAPc transiently expressed in HEK293T cells was co-purified with proteins that contain phosphorylated 14-3-3-binding consensus motifs as determined by detection with an antibody raised against the motif (Fig. 1C). Fusion of the TAP tag to the N terminus of the 14-3-3 protein did not result in efficient co-purification of associated proteins presumably due to interference of the TAP domain with dimerization (data not shown). Subsequently a colorectal cancer cell line expressing the tTA repressor (DLD1-tTA (28)) stably expressing a conditional allele encoding a 14-3-3-TAPc fusion protein was generated (Fig. 1D). After removal of doxycycline, the expression level of the 14-3-3-TAPc protein was similar to the level of endogenous 14-3-3 expression detected after DNA damage of the wildtype p53-expressing colorectal cancer cell line HCT116 (data not shown). Stably transfected DLD-1-tTA cells were used for the subsequent protein identifications because purification of proteins associated with 14-3-3-TAPc was more efficient than in transiently transfected HEK293T cells (data not shown) presumably due to the low levels of competing endogenous 14-3-3 protein in DLD1-tTA cells (2), which express mutant p53. As a control, we generated a cell line stably expressing the TAPc protein derived from the same parental DLD1-tTA cells (Fig. 1D). To identify the proteins present in the final TAP-tagged eluates, MudPIT analyses were performed (29). To obtain a comprehensive picture of 14-3-3 interactions under different cellular conditions, DLD1-tTA cells either exponentially proliferating (untreated), treated with doxorubicin for induction of DNA damage, serum-starved, or EGF-stimulated after serum starvation were analyzed after induction of 14-3-3-TAPc expression. In each case 100 mg of protein extract were used for tandem affinity purification with the EGF-stimulated cells being analyzed in duplicate. A number of highly abundant proteins (e.g. ribosomal proteins and keratins) were also detected in the eluates obtained after purification of the TAPc tag protein (for a complete list see Supplemental Table 1). These proteins and further proteins regarded as contaminants in previously published TAP tag purifications were excluded from Table I ( and pBI-14-3-3-HA (right panel) and analyzed by immunofluorescence with a polyclonal rabbit antibody directed against the HA tag. A representative detection of the ectopic 14-3-3 proteins by indirect immunofluorescence (left image) and the corresponding phase-contrast picture (right image) is shown for each transfection. Magnification, 400ϫ. C, 14-3-3-TAPc interacts with proteins phosphorylated at 14-3-3 consensus binding sites. After tandem affinity purification of HEK293T cells transiently transfected with plasmids encoding the indicated proteins, detection of 14-3-3-binding proteins by Western blot analysis was performed using an antibody specific for the consensus 14-3-3-binding motif phosphorylated at the central serine residue. The two left lanes represent the initial lysates (input), and the two right lanes represent the eluates after tandem affinity purification. Purification of transfected TAPc served as a negative control. D, conditional expression of 14-3-3-TAPc and TAPc in DLD1 colorectal cancer cell lines. Stable cell lines were analyzed for inducible expression of 14-3-3-TAPc (upper panel) and TAPc (lower panel) by Western blot analysis after removal of doxycycline for 24 h (left lanes, in the presence of 100 ng/ml doxycycline; right lanes, 24 h after removal of doxycycline). 14-3-3-TAPc was detected by an anti-14-3-3 antibody. TAPc was detected by a Protein A-specific antibody. Detection of ␣-tubulin served as a loading control.
see Supplemental Table 2). After subtraction of contaminants, 117 protein identifications representing potential ligands of 14-3-3 were obtained (listed in Table I). 14 proteins were previously shown to associate with other 14-3-3 isoforms in detailed case-by-case studies (indicated in Table I). The detection of these 14-3-3 ligands implies that the conditions used for the TAP tag purification allow the identification of bona fide 14-3-3 ligands. However, it is possible that some of the proteins detected here indirectly associate with 14-3-3 via other 14-3-3-associated proteins or represent contaminants.
In Silico Detection of 14-3-3-binding Motifs-The presence and location of putative 14-3-3 consensus binding motifs were determined for the 117 identified 14-3-3-associated proteins using the software Scansite (30). 55.6% of these proteins contained at least one 14-3-3-binding consensus motif (RSXpSXP) when a query with the highest stringency was performed (Supplemental Table 3). When a set of proteins identified by a combined TAP-MudPIT approach using a non-14-3-3 bait was analyzed using the same settings only 15% of the identified proteins contained a RSXpSXP motif (data not shown). When medium and low stringency settings of the Scansite algorithm were applied, ϳ90% of the identified, putative 14-3-3 ligands showed 14-3-3-binding sites (Supplemental Table 3). However, the sites identified by the Scansite program using medium and low stringency settings have to be treated cautiously because ϳ90% of the proteins isolated with a non-14-3-3 bait also showed putative 14-3-3binding sites at these settings.
Confirmation of 14-3-3 Interactions-Exemplary confirmations of selected interactions between 14-3-3 and newly identified interacting proteins were performed. 14-3-3 was ectopically expressed in combination with MIG-6 or AJUBA in HEK293T cells. The interaction between 14-3-3 and MIG-6 as well as between 14-3-3 and AJUBA was confirmed by co-immunoprecipitation ( Fig. 2A). For this analysis a 14-3-3 construct N-terminally tagged with an HA tag was used. This suggests that the C-terminal TAP tag does not lead to artificial associations of proteins with 14-3-3-TAPc. In agreement with these results, MIG-6 and 14-3-3 showed a cytoplasmic co-localization (Fig. 2B). These results show that the MudPIT results obtained here may serve as the basis for detailed functional analysis.
Functional Classification of 14-3-3-associated Proteins-The 14-3-3 interactors were grouped according to the pro-cesses in which they had been implicated previously (Table I). 25 of the identified proteins represented putative open reading frames without functional annotation and were designated as SIP2-26 (for 14-3-3-interacting protein). The 14-3-3␥ isoform was co-purified with 14-3-3, which is in accordance with previous reports showing heterodimerization between different 14-3-3 isoforms (32). However, the sequence coverage for 14-3-3␥ was relatively low in the two states where it was detected, indicating that the majority of the 14-3-3-TAPc protein was present as a 14-3-3 homodimer. This might, at least in part, be due to the abundance of the ectopic 14-3-3-TAPc. At lower concentrations endogenous 14-3-3 may heterodimerize with additional 14-3-3 isoforms.
The largest functional class among the 14-3-3-associated proteins was represented by 36 proteins that affect cytoskeletal dynamics and organization (Table I). Among other processes, the proteins represented in this class control adhesion, polarity, and actin polymerization. All these processes are involved in cellular migration (for a review, see Ref. 33). To evaluate whether induced levels of 14-3-3 expression may have an effect on these processes, we analyzed cellular migration in a wounding assay. Ectopic expression of 14-3-3 inhibited closure of a wound in a confluent monolayer of DLD1 cells, whereas the cells not expressing ectopic 14-3-3 completely closed a similar wound by migration within 24 h as determined by time lapse microscopy (Fig. 3). The minimal narrowing of the wound observed after induction of ectopic 14-3-3 was due to enlargement and flattening of the cells (Fig. 3, left panel). In this experiment cell proliferation was blocked by the addition of mitomycin C. Complete inhibition of cell proliferation by mitomycin C was confirmed with and without ectopic 14-3-3 expression (data not shown). The increased number of rounded cells observed after ectopic 14-3-3 expression may be due to inhibition of cytokinesis, which has been observed previously (2). Therefore, the observed wound closure was not due to cellular proliferation in the absence of ectopic expression of 14-3-3. These results show that elevated 14-3-3 expression may influence pro-cesses involved in cellular migration.
23 interactors were assigned to functions in mitogenic signaling (Table I). As a general readout for mitogenic signaling we analyzed MAP kinase activity after activation of a conditional 14-3-3 allele (Fig. 4). In cells ectopically expressing 14-3-3 we observed a moderate increase in basal MAP kinase activity and a significant increase in stimulated MAP kinase activity as detected by phosphorylation of phospho-ERK1/2 (Fig. 4).
Many other interesting putative 14-3-3 interactors were identified (Table I). For example the detection of the adenomatous polyposis coli (APC) protein as a 14-3-3-associated protein indicates that 14-3-3 may influence the Wnt signaling pathway (Table I). We found four perfect 14-3-3-binding sites in the APC protein (Supplemental Table 3). Which aspect of APC function is affected by 14-3-3 remains to be shown. DISCUSSION As the tandem affinity purification approach allows the purification of protein complexes, it is possible that some of the 14-3-3-associated proteins identified in this study do not directly interact with 14-3-3 but with other proteins directly binding to 14-3-3. Whether the identified 14-3-3-associated proteins directly bind to the amphipathic groove of 14-3-3 in a phosphorylation-dependent manner will require the experimental confirmation of the putative 14-3-3 consensus binding sites identified here by mutational analysis.
Recently, the identification of ligands associated with 14-3-3 proteins in vitro using affinity columns coupled to the yeast 14-3-3 proteins (BMH1 and BMH2) or human 14-3-3 was reported (34,35). The overlap with the proteins identified The phosphorylation status of ERK1/ERK2 was determined by Western blot analysis. DLD1-tTA-14-3-3-HA cells expressing 14-3-3-HA under control of the Tet system were cultivated in the absence of serum and doxycycline for 24 h and then restimulated with serum (10% fetal bovine serum) for 10 min. As a control the cells were kept in the presence of 100 ng/ml doxycycline. Total cell lysates were subjected to SDS-PAGE and Western blotting. Phosphorylated ERK1/ERK2 (p-ERK1 and p-ERK2) was detected with a phospho-p44/42 ERK1/ERK2 (Thr-202/Tyr-204)-specific antibody. The membrane was reprobed with an antibody specific for ERK1/ERK2 and a mouse monoclonal anti-HA antibody to detect 14-3-3-HA.
in vivo in this study was only 6 and 17%, respectively (indicated in Table I). This disparity is presumably due to differences between the 14-3-3-ligand associations in intact cells versus those that occur between recombinant 14-3-3 and protein ligands on columns. In another recent study four tagged 14-3-3 isoforms (␥, ␤, , and ) were isolated after transient transfection of HEK293 cells, and ligands were identified after PAGE separation (36). 38 of the proteins identified by Jin et al. (36) were also detected as potential 14-3-3 ligands in our study (indicated in Table I). The relatively small overlap may result from the different experimental approaches or from isoform-specific differences in ligand binding. In the future, parallel proteomic studies comparing the ligands associated with different 14-3-3 isoforms under identical conditions may be used to address these issues.
At present it is unclear whether the observed inhibition of wound closure is due to effects of elevated 14-3-3 expression on cellular adhesion, actin polymerization, or polarity. Inhibition of any of these processes may prevent directed migration. For example down-regulation of PAR3 was shown to prevent migration in Drosophila (37). Interestingly binding of ectopic 14-3-3 to PAR3 is known to disrupt polarity of mammalian epithelial cells (38). The inhibition of migration by 14-3-3 suggests that loss of 14-3-3 expression by epigenetic silencing or mutation of p53 in cancer cells may potentially lead to increased motility and thereby promote invasion. Recently conditional ectopic expression of 14-3-3␤ was shown to enhance cellular migration in a wound closure assay (39). This opposing result suggests that inhibition of wound closure by 14-3-3 is not due to an increase in overall 14-3-3 concentration but rather due to isoform-specific interactions. Isoform-specific differences in the association with certain motifs within one 14-3-3 ligand were recently shown for the positive cell cycle regulator CDC25B, which contains five 14-3-3 consensus binding sites (23). Interestingly association of CDC25B with different 14-3-3 isoforms results in different functional outcomes (23). We recently determined the crystal structure of 14-3-3 (Protein Data Bank code 1YZ5) and detected structural differences between the 14-3-3 isoforms , , and that are likely to contribute to selective ligand interactions and specific dimerization patterns (40).
As activation of the MAP kinase pathway protects against programmed cell death induced by DNA damage (41), the previously observed sensitization toward DNA-damaging agents after loss of 14-3-3 expression (5,9) may result, at least in part, from decreased MAP kinase activity. The increase in MAP kinase activity observed after 14-3-3 expression may be due to interactions with multiple ligands. Among others, A-RAF, B-RAF, and c-RAF seem to be good candidates as previous studies of the interaction between RAF proteins and other 14-3-3 isoforms have shown that 14-3-3 proteins are critical modulators of RAF activity (42). This analysis revealed multiple new candidate interactors of 14-3-3 that may mediate its negative effects on cell cycle progression. The AJUBA protein has recently been reported to be essential for entry into mitosis as it is required for activation of nuclear Aurora A kinase (43). Cytoplasmic sequestration of AJUBA by 14-3-3 might therefore prevent Aurora A kinase activation. However, the LIM domain protein AJUBA may be multifunctional as it also contributes to the formation or strengthening of cadherin-mediated cell-cell adhesion (44). 14-3-3 was found to associate with WEE1. The activity of WEE1 kinase is required for inhibition of CDC2 by phosphorylation of Tyr-15 and thereby prevents entry into mitosis. WEE1 kinase activity is stimulated by association with 14-3-3 proteins (15,45). Another regulator of mitosis, c-TAK1, was found to associate with 14-3-3: c-TAK1 phosphorylates CDC25C on Ser-216 and thereby prevents its nuclear localization and activation of CDC2 (46). How c-TAK1 is regulated by binding to 14-3-3 protein has to be determined in the future. The WEE1, c-TAK1, and AJUBA proteins represent attractive candidates for mediating the inhibition of G 2 /M progression that has been described for 14-3-3 previously (2). The activity of the kinesin KIF23/MKLP1 is necessary for completion of cytokinesis (47). Interestingly overexpression of 14-3-3 results in abortive cytokinesis with subsequent formation of tetraploid cells (2). Potentially this effect may be related to an association between 14-3-3 and MKLP1. PTOV-1 is a mitogenic factor that migrates to the nucleus during S phase entry (48). It is therefore conceivable that association with 14-3-3 slows G 1 /S progression by cytoplasmic sequestration of PTOV-1.
Taken together the results suggest that the p53 inducibility of the 14-3-3 gene may have evolved to allow p53 to simul- 5. Model of the p53-14-3-3 pathway: induction of 14-3-3 as a nodal point of regulation, which mediates modulation of multiple cellular processes by protein-protein interactions. Expression of 14-3-3 is induced after DNA damage and during differentiation and senescence (1). p53 is a known mediator of the induction of 14-3-3 expression (2). In cancer cells expression of 14-3-3 is often lost due to mutation of p53 or CpG methylation. In normal cells, the induced 14-3-3 protein may affect a variety of cellular functions by associating with the ligands identified in this report. Whether 14-3-3 associates with these proteins is determined by their appropriate phosphorylation. Positive and negative regulation of the respective processes, which were inferred from the identity of the proteins identified by the TAP-MudPIT analysis (Table I), is indicated. Dotted lines are used for processes for which it has to be determined whether 14-3-3 association has an inhibitory or activating effect. taneously regulate a large number of different signaling processes at the protein level (Fig. 5). Future research can now focus on the identification of signaling pathways mediating the phosphorylation of novel 14-3-3 ligands. Some of the identified 14-3-3-associated proteins may be components of multiprotein complexes and therefore may not directly interact with 14-3-3. Furthermore the functional relevance of the 14-3-3-ligand associations or 14-3-3 complex formations for 14-3-3-and p53-mediated tumor suppression may now be analyzed. Pharmacological inhibition of enzymes identified in this study may be a useful therapeutic approach as they represent activities that are presumably deregulated after loss of p53 and/or 14-3-3.