Tyrosine Phosphorylation of the Bcl-2-associated Protein BNIP-2 by Fibroblast Growth Factor Receptor-1 Prevents Its Binding to Cdc42GAP and Cdc42*

Fibroblast growth factor (FGF) receptor tyrosine kinases are involved in the regulation of cell growth, development, and differentiation in a variety of tissues. To isolate potential signaling molecules in the FGF signaling pathway, we have initiated a yeast two-hybrid screening using the cytosolic domain of FGF receptor-1 (Flg). Here we report the identification of BNIP-2, a previously cloned Bcl-2- and adenovirus E1B-associated protein, as a putative substrate of the receptor. When cotransfected in 293T cells, BNIP-2 was tyrosine-phosphorylated via Flg, but their interaction was transient and could only be seen by “capture” experiments with catalytically inert kinase mutants. When responsive cells were challenged with basic FGF, endogenous tyrosine-phosphorylated BNIP-2 could be precipitated with a BNIP-2 antibody. In addition, the recombinant BNIP-2 expressed in bacteria could be phosphorylated by active Flg in vitro. BNIP-2 shares a region of homology with the noncatalytic domain of Cdc42GAP, a GTPase-activating protein for the small GTP-binding molecule, Cdc42. We show here that BNIP-2 and Cdc42GAP could directly bind to each other and they also compete for the binding to the same target, Cdc42. Unexpectedly, BNIP-2, either produced as a bacterial recombinant protein or expressed in 293T cells, could stimulate the intrinsic GTPase activity of Cdc42. In all cases, tyrosine phosphorylation of BNIP-2 severely impaired its association with Cdc42GAP and its induced GTPase-activating protein-like activity toward Cdc42. These findings should allow us to further characterize the integration of signaling between receptor tyrosine kinases, GTP-binding molecules, and apoptotic pathways.

The family of fibroblast growth factors (FGFs) 1 consists of at least 19 different growth factors that control cellular responses such as growth, differentiation, and cell migration (1). FGFs induce their biological responses by binding to and activating a family of cell surface receptors with intrinsic tyrosine kinase activity (2). Dimerization of FGF receptors is essential for kinase activation and for receptor autophosphorylation and requires the concerted action of FGF together with soluble or cell-surface heparan sulfate proteoglycans (3).
Upon activation, FGF receptor tyrosine kinases undergo rapid autophosphorylation on various tyrosine residues. Autophosphorylation sites located within the catalytic domain are crucial for stimulation of kinase activity, while autophosphorylation sites located in other regions are usually involved in the recruitment of cellular target proteins (4). FGFR-1, or Flg, contains at least seven autophosphorylation sites. Two of these are located in the catalytic domain (Tyr 653 and Tyr 654 ) and are essential for kinase activation (5). One phosphotyrosine in the C-terminal tail (Tyr 766 ) functions as a high affinity binding site for the SH2 domain of phospholipase C-␥ (6). Phosphorylation of Tyr 766 is essential for phosphatidylinositol hydrolysis but not for FGF-induced DNA synthesis in myoblasts or differentiation of PC12 cells.
The Ras/mitogen-activated protein kinase signaling pathway plays an important role in FGF-signaling (7,8). The adaptor protein Grb2 links receptor tyrosine kinases with the Ras signaling pathway in conjunction with the guanine nucleotidereleasing protein Sos (9, 10). Grb2 does not bind directly to the FGF receptors but binds to a recently cloned and characterized docker protein termed FRS-2 (11). FRS-2 is specifically and rapidly tyrosine-phosphorylated when responsive cells are treated with FGF or NGF but not significantly phosphorylated in response to other growth factors or cytokines (11,12). FGF induces a unique pattern of tyrosine phosphorylation when the lysates of cultured cells are immunoblotted and analyzed for tyrosine phosphorylation (13). Like NGF, FGF also induces the differentiation of PC12 cells, whereas epidermal growth factor and platelet-derived growth factor induce proliferation, unless overexpressed (14,15). The cellular signaling pathways of all growth factors and cytokines are initiated at the activated receptors, and the resultant signals depend on the combinations of proteins that interact at the receptor or that are substrates of the intrinsic or associated kinases. The unique nature of FGF signaling pathways raises the possibility that novel cellular proteins are either substrates of, or become associated with, the FGF receptor after ligand stimulation. Some of these proteins may be responsible for initiating signaling pathways that are separate from the mitogen-activated protein kinase pathway.
In an attempt to identify novel proteins that interact with the cytosolic domain of Flg, we used the two-hybrid protein interaction cloning system in the yeast Saccharomyces cerevisiae (16). Here we report the identification of the protein BNIP-2 as a putative substrate of Flg. BNIP-2 was first cloned as a protein that interacted with the adenovirus E1B 19-kDa protein that protects against cell death induced by viral infection and other proapoptotic stimuli (17)(18)(19)(20)(21). The Bcl-2 protein and related antiapoptotic proteins can functionally substitute for the E1B 19-kDa protein, and they too bind BNIP-2 (17,22). The physiological function of BNIP-2 is not known, and few clues can be derived from the amino acid sequence of the protein, since it does not display significant homology to any protein in the data banks except for a region away from the catalytic domain on the 50-kDa RhoGAP (Cdc42GAP) protein (17). Cdc42GAP is a GTPase-activating protein (GAP) that has highest activity for Cdc42 and low activity to Rac but negligible activity toward small molecular weight G proteins outside the Rho family (23,24).
We show that BNIP-2 binds directly to Cdc42GAP and Cdc42. BNIP-2 augments Cdc42 GTPase activity when analyzed in in vitro assays. Tyrosine phosphorylation of BNIP-2, however, inhibits its binding to each partner protein, which also results in the abrogation of its GAP-like activity. It is possible that BNIP-2 plays a role in modulating the GTPase activity of Cdc42, by acting directly on Cdc42 or indirectly on Cdc42GAP, and that tyrosine phosphorylation of the protein can act as a "switch" to reverse the associations and possible downstream effects.

MATERIALS AND METHODS
Plasmids-Full-length cDNA of BNIP-2 was amplified by reverse transcription-polymerase chain reaction from human 293T cells according to the published sequence (6) and cloned into a hemagglutinin (HA)-tagged expression vector, pXJ40 (Dr. E. Manser, IMCB, Singapore) or into pGEX4T-1 vector for producing the GST recombinant protein. Human full-length FGF receptor-1, Flg (from Dr. L. Claesson-Welsh, Biomedical Center, Sweden) was used to amplify its cytosolic domain, Flg(cyto) (amino acids 397-822) and cloned into pAS2 vector of the GAL4 yeast two-hybrid system or into pXJ40-FLAG. pGEX-Cdc42GAP and pGEX-Cdc42 (from Dr. A. Hall, University College London) were used in making GST fusion proteins and also as templates to generate pXJ40 constructs or pACT2 constructs used in the yeast two-hybrid interaction assays. Mutants of Flg(cyto) (K514R, D623E, D623N, Y653F/Y654F) were generated by site-directed mutagenesis using the Quick-Change Mutagenesis kit (Stratagene). Fulllength FRS2 cDNA was amplified by reverse transcription-polymerase chain reaction from Swiss 3T3 cells based on cDNA sequence provided by Dr. J. Schlessinger (New York University Medical Center; Ref. 11). All plasmids were purified using a Wizard miniprep kit (Promega) or Wizard Maxi/Mega-prep kit followed by ethanol reprecipitation for use in transfection experiments. Clones were confirmed correct by thermal cycle sequencing using the SequiThermal EXCEL II DNA sequencing kit (Epicentre Technologies) or mapping analyses using restriction enzymes (New England Biolabs). Escherichia coli strain DH5␣ was used as host for propagation of the clones. Reagents used were of analytical grade, and standard protocols for molecular manipulations and media preparations were from Ref. 53.
Yeast Two-hybrid Screening and Interaction Assays-Flg(cyto) was fused to the yeast DNA binding domain of GAL4 in a pAS2 vector and screened for interacting clones in a human T-lymphocyte library fused with the GAL4 DNA activation domain in the pACT vector according to the instructions of the manufacturer (CLONTECH). Putative clones were reconstituted and tested against unrelated proteins such as human lamin C and SV40 large T-antigen to eliminate false positive results.
Production of BNIP-2 Antibodies-Full-length human GST-BNIP-2 protein (1 mg) was eluted from Sepharose beads by boiling in 1% (w/v) SDS. The supernatant was dialyzed overnight at 4°C in phosphatebuffered saline, and mixed with complete adjuvant (Life Technologies, Inc.) until miscelles were formed and injected subcutaneously into female New Zealand White rabbits. Every 2 weeks, two more boosters were administered in incomplete adjuvant followed by another injection 1 month later. Ten days after the final injection, sera were collected for purification. The sera were first incubated with GST proteins immobilized on a polyvinylidene difluoride membrane in order to remove anti-GST antibodies. The supernatant, containing mostly anti-BNIP-2 antibodies, was then incubated with GST-BNIP-2 immobilized on another polyvinylidene difluoride membrane to capture anti-BNIP-2 antibodies. The bound antibodies were eluted by incubating in glycine (pH 2) and vortexed several times, and the solution was neutralized in Tris (pH 8). The antibody detected a doublet at 52 kDa in untransfected and BNIP-2-transfected lysates, and it was deemed specific by virtue of its binding inhibition by immunizing peptide.
Cell Culture and Transfection-Cells were grown in either RPMI 1640 medium (human 293T cells) or Dulbecco's modified Eagle's medium (human MRC-5 primary lung fibroblasts) supplemented with 10% (v/v) fetal bovine serum (Hyclone), 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin (all from Sigma) and maintained at 37°C in a 5% CO 2 atmosphere. 293T cells at 90% confluency in 100-mm plates were transfected for 1 h with 10 g of the indicated plasmid using Tfx-50 cationic lipids according to the manufacturer's instruction (Promega).
GAP Assay-The GAP activity toward Cdc42 was examined by determining the release of 32 P i from the [␥-32 P]GTP prebound to the molecule. GST-Cdc42 (5 g) still conjugated to the Sepharose beads, were washed twice in buffer A (50 mM HEPES, pH 7.4, 0.5 mM EDTA) and resuspended in a final volume of 10 l of the same buffer with 5 Ci of [␥-32 P]GTP (6000 Ci/mmol; NEN Life Science Products) for 10 min at room temperature. The reaction was terminated by adding 25 mM magnesium chloride. Excessive unincorporated radioactive GTP was removed by washing the beads five times in 1 ml of cold buffer B (50 mM HEPES, pH 7.4, 150 mM sodium chloride, 1.5 mM magnesium chloride, 5 mM EGTA, 10% (v/v) glycerol, 1% (v/v) Triton X-100, a mixture of proteases inhibitors, and 5 mM sodium orthovanadate), and beads were finally resuspended in 10 l of the same buffer. For in vitro GAP assays using proteins expressed in bacteria, eluted recombinants of BNIP-2 or Cdc42GAP (1 g in 100 l of buffer B) were then added to the beads suspension and mixed well. The suspension was quickly centrifuged to collect the beads and incubated at room temperature for 10 min, and aliquots of the supernatant (10 l) were then taken for counting in a scintillation counter. For GAP assays involving 293T lysates, cells transfected with BNIP-2 and/or Flg(cyto) were lysed in buffer B, and 20 l of this (approximately 40 g of total protein contents) was diluted in 100 l of buffer B before it was added to the GST-Cdc42 beads preloaded with [␥-32 P]GTP and assayed as described above.

FGFR-1 Phosphorylates BNIP-2 in Vivo and in Vitro-
The cytosolic domain of the FGF receptor-1 (Flg) was used in a yeast two-hybrid screen to identify novel interacting proteins. One of the putative clones was identified as a nearly full-length cDNA encoding BNIP-2 protein, previously known to interact with the Bcl-2 and adenovirus E1B proteins (17). To characterize the nature of the binding between BNIP-2 and Flg, the full-length BNIP-2 cDNA was isolated and transfected into 293T with the full-length receptor. Experiments were performed to see if BNIP-2 could be tyrosine-phosphorylated by Flg in these cells. The cell lysates were precipitated with Flg or HA antibodies, and the resultant Western blots were probed with PY20 antibody to detect tyrosine-phosphorylated proteins (Fig. 1). The Flg was active as indicated by the positive PY20 signal in the Flg immunoprecipitates (Fig. 1A), while HA immunoprecipitation results revealed a tyrosine-phosphorylated BNIP-2 doublet at 52 kDa (Fig. 1B). However, there was no sign of phosphorylated BNIP-2 in the Flg immunoprecipitate (possible location indicated by unlabeled arrow), suggesting that their binding in vivo may be transient in nature. To investigate this hypothesis, various "kinase-dead" mutants of the cytosolic domain of Flg, Flg(cyto), were constructed (see "Materials and Methods") and used to investigate the possible association by trapping BNIP-2 with these mutants (Fig. 1C). Previous studies had demonstrated that the mutated amino acids are necessary for the tyrosine phosphorylation of substrates by Flg(cyto) (26). 293T lysates expressing HA-BNIP-2 with either the wild type Flg(cyto) or the various Flg(cyto) mutants were precipitated with HA antibody, and the resultant Western blot was probed with PY20 antibody (Fig. 1C, upper  panel). As expected, tyrosine phosphorylation of BNIP-2 was observed only in the presence of wild type Flg(cyto). The PY20 blot was then stripped and reprobed with Flg antibody. It is apparent in Fig. 1C, middle panel, that indeed only kinasedead Flg(cyto) were co-immunoprecipitated with BNIP-2. In order to demonstrate that equivalent amounts of BNIP-2 were precipitated, the blot was again stripped and reprobed with HA antibody (Fig. 1C, lower panel).
To verify that endogenous BNIP-2 could also be phosphorylated by exogenous FGF, primary human MRC-5 fibroblasts were treated with FGF (10 ng/ml) for 10 min, and the lysates were subjected to immunoprecipitation with the BNIP-2 anti-body, prepared as described under "Materials and Methods." The resultant Western blot was then probed with PY20 antibody. In the lysates from stimulated cells, there was a tyrosinephosphorylated protein at 52 kDa (indicated by arrowheads) that was precipitated by the BNIP-2 antibody (Fig. 1D). Compared with the transfection studies in 293T, only the slower migrating form of tyrosine-phopshorylated BNIP-2 was detected here. The nature of this difference is currently unknown, although these two cell types express both isoforms (data not shown). To further confirm that BNIP-2 is a substrate of Flg-(cyto), an in vitro kinase assay was performed using GST-Flg(cyto) overexpressed and purified from 293T cells as the active enzyme. The purified GST-Flg(cyto) was active as seen by its autophosphorylation (Fig. 1E, left panel), and it could phosphorylate the purified GST-BNIP-2 in vitro (Fig. 1E, right  panel).
BNIP-2 Binds Cdc42GAP When It Is Not Tyrosine-phosphorylated-Intracellular signaling involves the association and dissociation of complexes whose formation is often controlled by phosphorylation. To understand where BNIP-2 may participate in cell signaling, it was necessary to first determine which other proteins it might associate with downstream of Flg. Sequence alignments reveal a strong homology between BNIP-2 and a noncatalytic domain of Cdc42GAP (Fig. 2). This region could represent a binding or regulatory domain, whereby both The blot was sequentially stripped and reprobed with Flg antibody (middle panel) to reveal co-immunoprecipitation of Flg(cyto) and then with HA antibody to demonstrate that equal amounts of BNIP-2 were precipitated (lower panel). D, primary MRC-5 fibroblasts were unstimulated (control) or stimulated for 10 min with basic FGF (10 ng/ml). Lysates were immunoprecipitated with BNIP-2 antibody, and equal aliquots of the precipitated proteins as well as whole cell lysates were separated by SDS-PAGE and immunoblotted with PY20 antibody. The arrowheads and arrow indicate the position of the BNIP-2 that is tyrosine-phosphorylated in whole cell lysates and immunoprecipitates, respectively. E, GST-BNIP-2 was excluded (left panel) or included (right panel) in the in vitro kinase reaction containing eluted GST-Flg(cyto) that was expressed and purified from 293T cells as described under "Materials and Methods." Y-P denotes tyrosine-phosphorylated protein.
are regulated by a common mechanism or they are both targeted to a third unidentified protein. A third potential function for this homology region was alluded to by Boyd et al. (17), who reasoned that Cdc42GAP and BNIP-2 may form a heterocomplex via this domain. To test the latter hypothesis, 293T cells were transfected with Cdc42GAP and HA-tagged BNIP-2, either singly or together. The lysates were precipitated with a BNIP-2 antibody, and the resultant Western blot was probed for the presence of Cdc42GAP (Fig. 3A). A strong signal, denoting the presence of Cdc42GAP, was detected in the immunoprecipitate derived from cells coexpressing HA-BNIP-2 and Cdc42GAP. When cells were transfected with Cdc42GAP alone, the signal was weaker. These results indicate that there is an association between BNIP-2 and Cdc42GAP in vivo.
We next investigated the effect that tyrosine phosphorylation of BNIP-2 has on its binding to Cdc42GAP. Lysates from 293T expressing HA-BNIP-2 or/and Flg(cyto) were incubated with GST-Cdc42GAP, and the associated proteins were separated, Western blotted, and probed with HA antibody (Fig. 3B). The result shows that BNIP-2 bound to Cdc42GAP, and this association decreased when BNIP-2 was tyrosine-phosphorylated by Flg(cyto). Aliquots of the lysates used in Fig. 3B were immunoprecipitated with anti-BNIP-2, and the resultant Western blots were probed with PY20 antibody (Fig. 3C, upper  panel) or with HA antibody (Fig. 3C, lower panel), verifying that BNIP-2 was tyrosine-phosphorylated and expressed equally in these experiments. Since BNIP-2 was overexpressed in the lysates and the GST-Cdc42GAP was in great excess, the association between these two molecules is most likely to be direct, as further evidenced by their positive interaction detected in a yeast two-hybrid assay (data not shown). To confirm that the dissociation of BNIP-2 and Cdc42GAP in the presence of Flg(cyto) was due to its tyrosine phosphorylation instead of other mechanisms (e.g. serine/threonine phosphorylation induced downstream of activated FGF receptors), lysates from cells expressing HA-BNIP-2 and Flg(cyto) were first treated with or without Yersinia phosphotyrosine-specific phosphatase prior to a GST-Cdc42GAP pull-down experiment similar to those shown in Fig. 3B. The result shows that when the phosphate(s) on the BNIP-2 tyrosines was/were removed, its binding to Cdc42GAP was essentially restored (Fig. 3D).
BNIP-2 Interferes with the Binding of Cdc42GAP to Cdc42-Cdc42GAP augments the intrinsic GTPase activity of Cdc42. One notable feature of the recognition of Cdc42 by Cdc42GAP is that it shows no greater preference for the GTP-bound form of Cdc42 than the GDP-bound form (24). We were therefore interested to see if BNIP-2 could modulate the binding preference of Cdc42GAP to Cdc42. 293T cells were transiently transfected with Cdc42GAP with or without HA-BNIP-2. Lysates from each of these transfected cells were divided into three equal portions and subjected to a pull-down experiment with equal amounts of GST-Cdc42 that was either preloaded with GTP␥S (a nonhydrolyzable analogue of GTP) or GDP or was not loaded with either nucleotide. The bound proteins were Western blotted to determine the presence of the GAP protein. The data in Fig. 4A demonstrate that an equal amount of Cdc42GAP bound to both the GDP-loaded and GTP␥S-loaded Cdc42 and that this was significantly more than that binding to nonloaded Cdc42. When BNIP-2 was coexpressed, there was a considerable reduction in the amount of Cdc42GAP binding to Cdc42, regardless of what form of nucleotide was bound to Cdc42. These cells expressed equal amounts of Cdc42GAP regardless of whether BNIP-2 was expressed or not (Fig. 4B).
These results demonstrate that BNIP-2 in vivo can interfere with the binding of Cdc42GAP to Cdc42. This could be due to (i) a direct inhibition on Cdc42GAP, (ii) the competition from BNIP-2 on the same target (Cdc42), or (iii) a combination of both. Since GST-Cdc42 used in the pull-down experiment was in great excess over the Cdc42GAP and BNIP-2 in the lysates, it seems most likely that the reduced binding was at least due to direct inhibition from BNIP-2 binding to Cdc42GAP. To examine the second possibility, we used GST-Cdc42GAP to precipitate BNIP-2 from the lysates in the absence or presence of Cdc42 as the potential competitor or FRS-2 as the control. As seen in Fig. 4C (upper panel), BNIP-2 binding to GST-Cdc42GAP was clearly inhibited by Cdc42 but not by FRS-2. Aliquots of lysates were Western blotted and tested for the expression of BNIP-2 (second panel), Cdc42 (third panel), and FRS-2 (lower panel) used in the experiment. Taken together, these results suggest that there is a mutual competition between Cdc42GAP, BNIP-2, and Cdc42.
Tyrosine Phosphorylation of BNIP-2 Prevents Its Binding to, and Abrogates Its GAP-like Activity toward, Cdc42-To further validate the binding of BNIP-2 to Cdc42 and to investigate the effect of tyrosine phosphorylation on the binding, a GST-Cdc42 pull-down experiment was performed on lysates from cells expressing BNIP-2 in the absence or presence of activated Flg-(cyto) (Fig. 5A). The protocol used was essentially the same as that shown for Fig. 3B. The results show that BNIP-2 binds to Cdc42, but its binding was almost completely abolished upon cotransfection with Flg(cyto). Unlike the binding of Cdc42GAP, GST-Cdc42 devoid of guanine nucleotide could readily precipitate BNIP-2 from the lysate. This suggests that BNIP-2 might bind to Cdc42 independently of the nucleotide binding status of the latter. Indeed, when GST-Cdc42 was preloaded with either GDP or GTP␥S, the binding of BNIP-2 to the nucleotide-loaded Cdc42 was the same as the nonloaded one (Fig. 5B). As in the case for Cdc42GAP, it is the tyrosine phosphorylation on BNIP-2 that causes its dissociation from Cdc42, since Yersinia phosphotyrosine-specific phosphatase treatment completely restored the binding (Fig. 5C). The collective results from Figs. 3D and 5C indicate that BNIP-2 tyrosine phosphorylation negatively modulates its binding to both Cdc42GAP and Cdc42.
The binding of BNIP-2 to Cdc42 is apparently independent of the guanine nucleotide binding status. BNIP-2 also binds weakly to RhoA and Rac1, which have close homology to Cdc42 (data not shown). Since BNIP-2 binds directly to Cdc42 and possibly competes with Cdc42GAP, we were interested to see what effect BNIP-2 has on the intrinsic GTPase activity of  Fig. 3B were immunoprecipitated with BNIP-2 antibody, separated by SDS-PAGE, blotted, and probed with PY20 antibody (C, upper panel) or HA antibody (C, lower panel) to demonstrate tyrosine phosphorylation and equal expression of transfected BNIP-2, respectively. D, 293T cells were transfected with expression vectors for HA-BNIP-2 alone or with Flg(cyto). Lysates were then treated with (in the absence of phosphatase inhibitors) or without Yersinia phosphotyrosine phosphatase (microunits/ml, as indicated) at 37°C for 30 min and then subjected to GST-Cdc42GAP pull-down. BNIP-2 binding was revealed by Western blotting with HA antibody and is expressed as the percentage relative to that of nonphosphorylated BNIP-2 as assessed by densitometric analyses. In all experiments involving GST pull-down experiments, blots were stripped and probed with GST antibody to assess the equality of loading (data not shown). Cdc42. GTPase assays were carried out as described under "Materials and Methods." It can be seen that BNIP-2 has a significant effect on the GTPase activity of Cdc42 increasing it nearly 2-fold, which was comparable with that induced by Cdc42GAP (Fig. 5D). When BNIP-2 and Cdc42GAP were added together, there was no further augmentation in their effect. Instead, the enhanced GTPase activity was less than either one added alone, probably due to the neutralizing effect of the heterocomplex. In order to investigate the effect of the tyrosine phosphorylation of BNIP-2 on the GTPase activity of Cdc42, 293T cells were transiently transfected with BNIP-2 or Flg-(cyto), alone or together. The lysates were then used in GTPase assays. In agreement with the in vitro assay, BNIP-2 overexpressed in lysates induced a 5-fold increase in the GTPase activity of Cdc42 when compared with control lysates (Fig. 5E). To ensure that this increase was not simply due to the dissociation of the GTP, Cdc42 was preloaded with the nonhydrolyzable analog 35 S-labeled GTP␥S and assayed for its release in the presence of the lysates. Both control and lysates containing overexpressed BNIP-2 did not stimulate the release of GTP␥S, whereas EDTA, used to chelate Mg 2ϩ from the nucleotide complex, caused a dramatic release of the nucleotide from Cdc42 (data not shown). The increase in GTPase activity induced by BNIP-2 was, however, attenuated upon coexpression of Flg- FIG. 5. Tyrosine phosphorylation of BNIP-2 reduces both its binding and its GAP-like activity toward Cdc42. A, 293T cells were transfected with expression vectors for HA-BNIP-2 and/or Flg(cyto). Lysates were used in a pull-down (P/D) experiment by incubating with equal amounts of GST-Cdc42 conjugated to agarose beads. The associated proteins were separated on SDS-PAGE, blotted, and probed with HA antibody. The band migrating below HA-BNIP-2 results from a cross-reactive signal to GST-Cdc42. B, an equal amount of lysates from 293T cells transfected with HA-BNIP-2 was used in a pull-down (P/D) experiment by incubating with equal amounts of GST-Cdc42 preloaded with GDP, GTP␥S (indicated as GTP), or neither, as described under "Materials and Methods." The bound protein was separated on SDS-PAGE, blotted, and probed with HA antibody. C, 293T cells were transfected with expression vectors for HA-BNIP-2 alone or with Flg(cyto). Lysates were then treated with (in the absence of phosphatase inhibitors) or without Yersinia phosphotyrosine phosphatase (microunits/ml, as indicated) at 37°C for 30 min and then subjected to GST-Cdc42 pull-down. BNIP-2 binding was then revealed by Western blotting with HA antibody. D, the effect of recombinant Cdc42GAP and/or BNIP-2 on the GTPase activity of Cdc42 was carried out as described under "Materials and Methods." The values are the means Ϯ S.D. of three separate experiments. E, 293T cells were transfected with expression vectors for HA-BNIP-2 and/or Flg(cyto), and the lysates were used in assays with recombinant Cdc42 that was preloaded with radiolabeled GTP as described under "Materials and Methods." The values shown are the means Ϯ S.D. of three replicate experiments.
(cyto). This is in agreement with previous results that show that when BNIP-2 is tyrosine-phosphorylated by Flg(cyto) its association with Cdc42 is greatly reduced. DISCUSSION We set out to discover novel proteins that interact with the cytosolic domain of Flg. We chose to employ the yeast twohybrid technique, which has been highly successful in recent years in identifying proteins in interactive complexes. We identified a protein, BNIP-2 that appears to be a "kiss and run" substrate of the receptor. The yeast two-hybrid technique is optimally effective in detecting proteins that interact without the need for covalent modification, such as phosphorylation, of either of the binding proteins. Some kinase substrates, however, have previously been discovered by this technique (27,28). BNIP-2, on the basis of detectable tyrosine phosphorylation signal, is a weak substrate of Flg when compared with previously characterized substrates such as FRS-2. The interaction of BNIP-2 with Flg is transient, and the association can only be seen with kinase-dead mutants. The use of such mutants is analogous to the use of tyrosine phosphatase catalytically dead mutants that have been employed to identify substrates of the various tyrosine phosphatases (29,30). From the evidence shown here, kinase-dead mutants may be useful in the future to detect substrates of various kinases by using the yeast two-hybrid technique.
BNIP-2 was originally discovered as a novel protein that binds to the adenovirus E1B protein. Expression of E1B protein in the host cell has been shown to suppress apoptosis. It was assumed that BNIP-2, a potentially proapoptotic protein, is sequestered by E1B. The well characterized mammalian antiapoptotic protein Bcl-2 and its related proteins also bind to BNIP-2 (17,22). Recently, expression of BNIP-2 mRNA was also shown to be down-regulated by estrogen treatment of neuroblastoma cells (31). Until now, there has been a paucity of additional information relating to its characterization, perhaps because of the lack of established domains contained within its amino acid sequence. Apart from a single EF-hand calciumbinding domain, the only homologous sequence on BNIP-2 is a stretch of amino acid sequence on the noncatalytic part of the GTPase-activating protein, Cdc42GAP. The homology between these two regions is so strong as to suggest that this region might represent a novel binding domain. In their original study, Boyd et al. (17) reasoned that the common domain could function to bind either protein to a third, as yet unknown protein, or it could function to bind Cdc42GAP and BNIP-2 to each other. We initiated studies to investigate both possibilities. Studies are still proceeding to identify other proteins that bind to this homologous domain. We therefore concentrated on evaluating the binding of BNIP-2 to Cdc42GAP. We have shown in the course of this study that these two molecules can indeed form a heterocomplex. We have also identified discrete parts within this homologous domain that are important in mediating the interaction between BNIP-2 and all of its partner proteins: Flg(cyto), Cdc42GAP, and Cdc42. 2 There is compelling evidence that BNIP-2 is tyrosine-phosphorylated by Flg(cyto) when cells are cotransfected with Flg and BNIP-2. By mutating individual tyrosine residues, we were unable to cause a significant reduction in the tyrosine phosphorylation of BNIP-2. This suggests that multiple residues are phosphorylated by the receptor kinase. We have also presented evidence that BNIP-2 is tyrosine-phosphorylated in vivo in untransfected cells when the cells are stimulated with FGF and that it can be phosphorylated by the active kinase in vitro. We have preliminary evidence that BNIP-2 is also tyrosine-phosphorylated when cells are stimulated with other growth factors such as epidermal growth factor and plateletderived growth factor (data not shown). The most noticeable feature of the tyrosine phosphorylation of BNIP-2 is that in all cases it reduces the affinity of BNIP-2 for its partner proteins. The tyrosine phosphorylation of BNIP-2 therefore has the capacity to act as a switch that alters the effects of binding to its partner proteins. With multiple binding partners it is not possible at present to predict how such a switch may operate in a physiological situation. We have not pursued the stoichiometry for the binding of BNIP-2 to Cdc42GAP and Cdc42, since we have been more intent on seeking a potential physiological function for these associations.
The observation that BNIP-2 could compete with the binding of Cdc42GAP to Cdc42 suggests that it could prevent the binding of the GAP to Cdc42 indirectly, by forming a complex with the GAP, directly competing for the same site on the latter, or both. To our surprise, we found that BNIP-2 not only interacts directly with Cdc42, but the binding, independent of the guanine nucleotide binding status, somehow results in an increase in the GTPase activity of Cdc42, and this effect can be abrogated upon the tyrosine phosphorylation of BNIP-2. The GAPlike activity that BNIP-2 exerts is rather intriguing, since it lacks the conserved RhoGAP catalytic domain. Recently, it has been documented that Cdc42 can form a homodimer where each subunit acts like a GAP toward the other subunit in augmenting the rate of GTP hydrolysis, and this effect is mediated by the polybasic C-terminal tail of the molecule (32). Interestingly, the carboxyl-terminal half of BNIP-2 appears to be polybasic also (see Fig. 2). It remains to be seen whether any of these basic residues can act like the "arginine finger" within the "cradle fold" of the ␣-helical structure employed by the conventional type of GAP in mediating GTP hydrolysis (33)(34)(35)(36).
This report has added BNIP-2 to the ever increasing list of proteins that interact directly with Cdc42. These Cdc42-binding proteins participate in a wide range of cellular effects, which include cytoskeletal rearrangement, phagocytosis, apoptosis, cell cycle progression, and transformation. Among these Cdc42-binding molecules are mitogen-activated protein kinase/ extracellular signal-regulated kinase kinase kinase 1 and mitogen-activated protein kinase/extracellular signal-regulated kinase kinase kinase 4 (37); two mitogen-activated protein kinase kinase kinases, MLK2 and MLK3 (38); PAK (a serine/ threonine kinase; Refs. 39 and 40); myotonic dystrophy kinaserelated Cdc42-binding kinases (41); ACK1 and ACK2 (nonreceptor tyrosine kinases; Refs. 42 and 43); WASP (a protein implicated in the Wiskott-Aldrich syndrome; Ref. 44); FGD1 (a Cdc42-specific guanine nucleotide exchange factor, which is also the faciogenital dysplasia gene product; Ref. 45); Cdc42GAP (23,24); IQGAP1 and IQGAP2 (25,46); CIP4 (which is homologous to the nonkinase domain of FER; Ref. 47); phosphoinositide 3-kinase p85 (48); and phospholipase C ␤2 (49). It is interesting to note that BNIP-2 binds strongly to the region on Cdc42 that contains both the effector binding site as well as the nucleotide binding site. 3 Although a more defined site on Cdc42 that BNIP-2 binds to is yet to be mapped, it is tempting to suggest that the binding to this region can compete with the binding of some of these effectors or Cdc42GAP, as seen in our competition studies. Therefore, BNIP-2 and other Cdc42-binding molecules could potentially be regulating the binding of each other, either negatively by mutual competition or positively by augmenting the complex formation through a locatory sequence. This can be achieved by temporal or spatial means or could employ tyrosine phosphorylation as a means of modulation.
The apparent lack of preferential binding to the GTP-bound form of Cdc42 would tend to argue against BNIP-2 being an effector. However, a constitutively bound protein may still serve as an effector after being activated by the conformational change that accompanies GTP hydrolysis. In addition to the regulation by phosphorylation, the apparent GAP-like activity can act as a negative feedback mechanism to terminate its effector function. The GAP-like activity of effectors has previously been noted. The protein c-Raf (a mitogen-activated protein kinase kinase kinase), a Ras effector target, has weak GAP effect on Ras (50), whereas the phopholipase C ␤ and ␥-subunit of phophodiesterase, the effectors of heterotrimeric G-protein G q and transducin, respectively, are themselves GAPs (51,52).
Although we have presented evidence that BNIP-2 acts like a GAP, we have not excluded the possibility that it might have other functions. One interesting feature to note is the presence of a well conserved EF-hand Ca 2ϩ -binding motif in BNIP-2. Although present as a single motif, this EF-hand can potentially be brought together by homo-oligomerization or interaction with other EF-hand-containing molecules to constitute functional EF-hands that might be involved in calcium/calmodulin signaling. Interestingly, we have preliminary evidence that suggests that BNIP-2 can also bind to itself and that it is also a good substrate for both protein kinase A and conventional protein kinase C in vitro. 3 Our present report has identified the Bcl-2-associated BNIP-2 as a binding partner for Cdc42GAP and Cdc42, and its binding can be abrogated by its tyrosine phosphorylation by FGF receptor tyrosine kinase. This provides a potential link between signaling of tyrosine kinase receptors, GTPases, and apoptosis. Given the complexity and the huge repertoire of Cdc42-binding partners, we suggest that BNIP-2 plays an important role in one or more of the signaling routes propagated via Cdc42. To this end, we are currently investigating the effects of BNIP-2 mutants on cellular events elicited by Cdc42.