Localization of Endogenous Grb10 to the Mitochondria and Its Interaction with the Mitochondrial-associated Raf-1 Pool*

Grb10 belongs to a small family of adapter proteins that are known to interact with a number of receptor tyrosine kinases and signaling molecules. We have recently demonstrated that the Grb10 SH2 domain interacts with both the Raf-1 and MEK1 kinases. Overexpression of Grb10 genes with mutations in their SH2 domains promotes apoptosis in cultured cells, a phenotype that is reversed by concomitant overexpression of the wild type gene. Using immunofluorescence microscopy and subcellular fractionation we now show that most of the Grb10 molecules are peripherally associated with mitochondria. Following insulin-like growth factor I or serum treatment, small pools of Grb10 can also be found at the plasma membrane and in actin-rich membrane ruffles, whereas overexpression of Grb10 leads to its mislocalization to the cytosol. Two-hybrid analysis shows that the Grb10-binding site on Raf-1 co-localizes with its Ras-binding domain. Finally, we show that the endogenous Grb10 and Raf-1 proteins can be co-immunoprecipitated from a partially purified mitochondrial extract, an interaction that is enhanced following the activation of Raf-1 by ultraviolet radiation. Thus, we infer that Grb10 may regulate signaling between plasma membrane receptors and the apoptosis-inducing machinery on the mitochondrial outer membrane by modulating the anti-apoptotic activity of mitochondrial Raf-1.

Grb10 (1) and its close homologues Grb7 (2) and Grb14 (3) are considered to be adapter proteins because they interact with several signaling proteins and lack any apparent enzymatic activity (see Ref. 4 for review). Collectively, they are termed the Grb7 family and share a carboxyl-terminal SH2 domain that binds to several activated receptor tyrosine kinases, a central Pleckstrin homology (PH) 1 domain, and a small proline-rich sequence that can interact with SH3 domains in vitro (5). The BPS domain, located between the PH and SH2 domains of Grb10 and Grb14, interacts with activated insulinlike growth factor I (IGF-I) and/or insulin receptors (6,7). Finally, sequence analysis suggests the existence of a Rasassociating-like domain (8).
Several tumors and cell lines show increased expression of various members of the Grb7 family. In addition, epidemiological and experimental evidence links Grb7 overexpression to extramucosal tumor invasion in human esophageal carcinoma (9,10). A specific role for Grb10 is less clear. It interacts with the insulin, IGF-I, epidermal growth factor, Eph family receptor, and growth hormone receptors, the Ret protooncogene, and other signaling molecules such as JAK2, BCR-Abl, and Tec (1,(11)(12)(13)(14). We have previously demonstrated that the Grb10 SH2 domain interacts with the Raf-1 and MEK1 kinases (15); both are components of the mitogenic MAP kinase signal transduction pathway that transmits hormonal signals to proliferative or differentiation events (see Refs. 16 and 17 for reviews). Stably transfected cell lines overexpressing Grb10 display moderate inhibition of cell cycle progression and alterations in other aspects of their signaling pathways (11,14,18). Finally, overexpression of full-length hGrb10 2 genes carrying SH2 domain mutations promotes apoptosis in HTC-IR and COS-7 cells. This cell death phenotype can be reversed by concomitant overexpression of the wild type hGrb10 gene (15).
The serine/threonine kinase Raf-1 transmits mitogenic signals as part of the Raf-MEK-ERK kinase cascade, although only a small proportion of total Raf-1 molecules are translocated to the plasma membrane by Ras and activated by other proteins (19,20). Recent work has revealed that Raf-1 is implicated in many other signaling events. During mitosis, cytoplasmic Raf-1 is activated independently of Ras, and its function does not include activation of the downstream MEK and ERK kinases (21). Furthermore, depending on the cell line or on its intracellular localization, Raf-1 can act either as a promoter or as an inhibitor of apoptosis (22)(23)(24)(25). For example, Raf-1 can be targeted to mitochondria by Bcl-2, resulting in increased resistance to apoptotic inducers (26,27). Such variation in function requires a complex regulatory system; Raf-1 can interact with a large number of proteins including Ras, 14-3-3, Hsp90, CSK, Ksr, MEK1, and others (20, 28 -33).
To determine how Grb10 affects the activity or intracellular localization of its ligands, we attempted to confirm the intracellular localization of the endogenous protein by immunofluorescence microscopy and cell fractionation. Previous results suggested that Grb10 is a soluble protein that becomes localized to the plasma membrane following insulin treatment (5,34). However, these investigations used antibodies that recognize several bands on immunoblots (possibly because they were raised against the well conserved SH2 domain) and also used cell lines that overexpress Grb10 and/or the insulin receptor. Using a more specific antibody, we found that, surprisingly, Grb10 is peripherally associated with the mitochondria, where it interacts with the local pool of Raf-1. These results suggest that interactions between these two proteins could be used to regulate apoptosis whose initiating steps are now known to occur on the mitochondrial outer membrane (35)(36)(37).

MATERIALS AND METHODS
Production of anti-Grb10 Antibodies-We subcloned a FLAG-tagged hGrb10(R520L) gene (15) in the EcoRI/XhoI sites of pFastBac1 and produced recombinant baculoviruses with the Bac-to-Bac Baculovirus Expression System (Life Technologies, Inc.). Expression of the wild type hGrb10 protein was found to inhibit large scale viral amplification, whereas the baculovirus protein p35 possibly protected the hGrb10(R520L)-expressing Sf9 cells from apoptosis (38). The expressed protein was purified by affinity chromatography on an anti-FLAG(M2) resin as described (39). Antibodies were raised by inoculating rabbits with 100 g of hGrb10(R520L) in Titer Max adjuvant (CytRx Corporation). Anti-Grb10 antibodies were affinity-purified with maltose-binding protein-hGrb10-coupled Activated CH Sepharose 4B (Amersham Pharmacia Biotech).
Cell Culture and Immunofluorescence Microscopy-COS-1 and HeLa cells were maintained in Dulbecco's modified Eagle's medium ϩ 10% FBS at 37°C in an atmosphere containing 5% CO 2 . Transfections were performed using LipofectAMINE according to the manufacturer's instruction (Life Technologies, Inc.). HeLa S3 cells were grown in calciumfree Dulbecco's modified Eagle's medium supplemented with 10% FBS and 0.1% Pluronic F68 (Life Technologies, Inc.). For ultraviolet treatment, cells were washed in PBS, irradiated with 100 J/m 2 UV 254 nm, and maintained in serum-free media for 4 h.
For immunofluorescence microscopy, we inoculated 5,000 -15,000 cells into the wells of an 8-chambered glass or Permanox slide (Nunc). The next day, the cells were fixed 10 min in 4% paraformaldehyde, washed twice in PBS, permeabilized for 2 min in 0.2% Triton X-100 in PBS, and washed twice more in PBS. They were then blocked for 30 min in PBS ϩ 10% FBS and incubated for 60 min with the primary antibodies in PBS ϩ 10% FBS. The samples were washed four times with PBS, incubated 30 min with the fluorescent secondary antibodies, and washed four more times in PBS. Slides were mounted in ProLong antifade reagent (Molecular Probes) and viewed with a Leitz Aristoplan microscope coupled to a Princeton Instrument CCD camera. Images were subsequently analyzed with Eclipse (Empix Imaging Inc.) and Photoshop (Adobe) software. Localization of Grb10 to the mitochondria and membrane ruffles were also observed with a Nikon Diaphot confocal microscope. The concentration of primary antibodies used were 1:100 for the preimmune and anti-Grb10 serum, 0.3 g/ml for affinitypurified anti-Grb10 and 1:50 for the anti-CoxI mAb (Molecular Probes). Specific fluorescein isothiocyanate, lissamine-rhodamine, and Texas Red-coupled secondary antibodies were from Jackson ImmunoResearch. Fluorescent phalloidin was obtained from Sigma and used at a dilution of 1:50. Although the quality and intensity of the signals were reduced, similar conclusions were reached with variations of the original protocol such as methanol fixation or using bovine serum albumin as a blocking agent.
Cell Fractionation-HeLa S3 cells were spun down, washed twice in PBS, and resuspended in 1 ml/2 ϫ 10 7 cells of MS buffer (210 mM mannitol, 70 mM sucrose, 5 mM Tris, pH 7.5, 1 mM EDTA, Roche Molecular Biochemicals Protease inhibitor pellet). The cells were lysed in a Polytron (10 s at a setting of 6.5). Nuclei and unbroken cells were eliminated by three 10-min centrifugations at 1300 ϫ g. The supernatant was then centrifuged for 20 min at 17,000 ϫ g yielding a mitochondria-enriched MP fraction. Finally the 17,000 ϫ g supernatant was spun down for 30 min at 100,000 ϫ g yielding the high speed pellet and soluble fractions. Mitochondrial-enriched fractions were also produced by the hypotonic lysis method as described (40). Purification of mitochondria by running the MP fraction through a discontinuous 1 M/1.5 M sucrose gradient was performed essentially as described (40).
Immunoprecipitations-Proteins from the MP fractions were diluted in MS buffer to 1 mg/ml, and the organelles were broken up by the addition of 0.2% Triton X-100 followed by centrifugation. We incubated 250 l of extract with either 500 ng of Santa Cruz Biotechnology Raf-1 (E10) mAb or 1 l of either the preimmune or anti-Grb10 serums. After overnight incubation at 4°C, the antibodies were precipitated with protein A-Sepharose (Amersham Pharmacia Biotech), washed extensively in MS buffer, and boiled in SDS-polyacrylamide gel electrophoresis buffer. Raf-1 activity in immunoprecipitates was determined as described (42) except that the amount of activated ERK1 was estimated with a phospho-specific antibody (New England Biolabs).

Production and Characterization of Grb10
Antibodies-The hGrb10(R520L)-FLAG protein (see "Materials and Methods") was expressed in insect cells with a baculovirus vector, purified by affinity chromatography with the FLAG mAb, and then used to raise highly specific anti-Grb10 antibodies. When used to probe immunoblots of total protein extracts from COS-1 or HeLa cells, the antibodies recognize two major bands with apparent electrophoretic mobilities of 72 and 75 kDa (Fig. 1A), which probably result from Grb10 splicing variants, of which at least four are known to exist in human cells (5,11,34,43). The two close homologues of Grb10, Grb7 and Grb14, with apparent electrophoretic mobilities of 58 -65 kDa (3,10), are not recognized by this serum. We also probed Triton-soluble extracts from a variety of human cell lines (Fig. 1B). Interestingly, the ratios between the two major immunoreactive bands varied according to the lineage of the hematopoietic cells. Cells of the erythroleukemia (K562) and T-cell (Jurkat, RPMI 8402, Peer) lineages had more of the faster migrating form, whereas the slower migrating form was more abundant in the B-cell derived Daudi cells. The U937 and HL-60 cell lines expressed the faster migrating form exclusively.
Immunolocalization of Endogenous and Overexpressed Grb10 -Serum and immunoglobulins that were affinity-purified on a maltose-binding protein-Grb10 resin were used to detect the intracellular localization of endogenous Grb10 in COS-1 and HeLa cells. As seen in Fig. 2B, most of the endogenous Grb10 is localized to a number of large vesicular struc- tures that were identified as mitochondria by co-localization with a mAb that recognize the cytochrome oxidase subunit I (CoxI) (Fig. 3A). Labeling is also seen on the nuclei and, to a lower extent, in the cytoplasm, but reaction with preimmune serum has shown these to be nonspecific signals. Mitochondrial localization was also observed in HeLa cells, the original source of the hGrb10 cDNA, although lower expression levels of endogenous Grb10 (Fig. 1A) and a much higher background made interpretation of its localization more difficult (Fig. 2, E and F). In both cell lines, a 2-min treatment with 10% serum (not shown) or 100 ng/ml of IGF-I (Fig. 2C) induced the relocalization of a small proportion of the endogenous Grb10 to the plasma membrane where it can theoretically interact with activated receptors. This translocation is very transient and is generally not observed 5 min after hormone treatment (not shown). After 30 min, some cells show a punctuate staining, which does not co-localize with the CoxI mitochondrial marker (Fig. 2D and not shown). Because this latter localization was observed only in a minority of cells, we did not study it further. Serum treatment of COS-1 cells induces the formation of actinrich membrane ruffles, and, as shown in Fig. 3, these also became a site of Grb10 localization. Finally, Fig. 4 shows that overexpression of a FLAG-tagged hGrb10 in either COS-1 or HeLa cells results in its mislocalization to the cytoplasm. Identical results were obtained with GFP-Grb10 fusions. A small proportion of the Grb10 protein is still translocated to the plasma membrane shortly following serum treatment (not shown).
Cell Fractionation-To confirm the immunocytochemistry results, HeLa S3 cells were lysed in isotonic buffer and fractionated by differential centrifugation. The MP is enriched in mitochondria, peroxisomes, and large vesicles from the Golgi and endoplasmic reticulum. The plasma membrane and most of the endoplasmic reticulum/Golgi proteins were found in the high speed pellet, whereas cytosolic proteins are found in the high speed supernatant. Most of the endogenous Grb10 is found in the MP fraction (Fig. 5A). Interestingly, unlike attached HeLa cells that express equal amounts of the 72-and 75-kDa hGrb10 variants, the suspended HeLa S3 cells express mostly the 72-kDa variant. Furthermore, overexposing the Grb10 immunoblot revealed a band of 68 kDa mobility that is restricted to the soluble fraction and whose size is consistent with the hGrb10␤ variant, which is missing most of its PH domain (11). PH domains are potential membrane-binding elements (44). As expected, the MP fraction is enriched with mitochondrial markers such as CoxI and Bcl-2, residents of the inner and outer mitochondrial membrane, respectively (45,46). Bcl-2 also localizes to the endoplasmic reticulum, which explains its presence in the high speed pellet fraction (47). Although members of the mitogenic MAP kinase pathway are mostly found in soluble form, a significant proportion of the Raf-1 and MEK1 kinases are associated with both membrane fractions. The ERK1 and ERK2 MAP kinases are almost exclusively found in the soluble fraction, thus demonstrating that our membrane fractions are not significantly contaminated by nuclear and cytoplasmic proteins. The samples used in Fig. 5A were equalized for protein content, but because the yields of soluble proteins are usually 4 -6 times higher than membrane proteins, the proportion of target protein localized to the cytosol may be underestimated. Repeating the fractionation procedure while maintaining each fraction in the same volume did not significantly change the fractionation results (Fig. 5B). We also noted that the association of Grb10 and Raf-1 with the organelles of the MP fraction is relatively weak. A commonly used cell fractionation protocol, in which cells are first lysed in hypotonic buffer followed by a rapid return to isotonic condi-tions, caused enough damage to the mitochondrial membranes to release most of the Grb10 and Raf-1 proteins to the soluble fraction even though the fractionation of intramitochondrial proteins, such as CoxI, was maintained (Fig. 5C). Finally, we further purified mitochondria by centrifugation of the MP fraction through a discontinuous sucrose gradient. Fig. 6 shows that endogenous Grb10 can easily be detected in these highly purified organelles.

FIG. 2. Immunolocalization of endogenous Grb10 in cultured cells. COS-1 (A-D) and HeLa cells (E and F) probed with either preimmune serum (A and E) or affinity-purified anti-Grb10 (B-D and F).
We next determined the nature of the Grb10-mitochondria interaction. Purified mitochondria were pelleted and resuspended in either isotonic buffer or a highly alkaline (pH 11.5) solution. The latter breaks open the organelles into linear membrane sheets, which can then be purified by high speed centrifugation (48). As shown in Fig. 6, the alkali treatment induces a relocalization of Grb10 from the pelleted to the soluble fraction indicating that it is not a transmembrane protein.
As expected, CoxI remained in the pellet fraction, whereas Raf-1 showed a distribution identical to Grb10 (not shown).
Grb10 Interacts with Mitochondrial Raf-1-In a previous report (15), we showed that Grb10 and Raf-1 can interact in a two-hybrid assay but that their co-immunoprecipitation from total protein extracts was only possible following overexpression of a tagged Grb10 protein. Using the yeast two-hybrid assay, we determined that the Grb10-binding site on the Raf-1 amino-terminal domain maps closely to the Ras-binding domain (Fig. 7). A more detailed mapping of the interaction site was impeded by the ability of many of our constructs to activate transcription nonspecifically. Furthermore, we show that a Raf-1 monoclonal antibody can efficiently co-immunoprecipitate endogenous levels of Grb10 from a Triton-solubilized MP protein extract (Fig. 8A). As mitochondrial Raf-1 becomes activated following UV irradiation, its affinity for Grb10 also appears to increase (Fig. 8B). 4 h following UV treatment, we did not observe significant changes in the amounts or intracellular localization of either Raf-1 or Grb10. As the cells enter apoptosis, degradation of these proteins becomes apparent after 20 or 24 h, respectively (not shown). Finally, Fig. 8A also shows that our anti-Grb10 antibodies are inefficient in immunoprecipitating native Grb10. DISCUSSION We demonstrate that, unexpectedly, the intracellular localization of the adapter protein Grb10 is mostly mitochondrial. Depending on growth conditions, a small proportion of Grb10 was also found on the plasma membrane as well as membrane ruffles. These observations were made possible by the production of a new highly specific antibody. Because of the low intensity of the specific labeling, we also probed our cells with preimmune sera from the same animal that was used for the production of the Grb10 antibodies. This control confirmed the nonspecificity of the nuclear and cytoplasmic fluorescence.
Mitochondrial localization of Grb10 was confirmed by cell fractionation experiments showing that most of the endogenous Grb10 is found in the mitochondria-enriched heavy membrane fraction. Grb10 was also detected in purified mitochon-  6. Grb10 is a peripheral mitochondrial protein. Highly purified mitochondria were resuspended in either isotonic buffer or an alkaline solution (100 mM Na 2 CO 3 , pH 11.5) and centrifuged. Equal amounts of the supernatant (S) and pelleted (P) fractions were then probed with either anti-Grb10 serum or CoxI mAb. dria that were isolated though a discontinuous sucrose gradient. The population of Grb10 that is found in the light membrane fraction most probably represents those molecules that localize to the plasma membrane and membrane ruffles. The latter are the site of localization for many other signaling molecules involved in regulating the mitogenic MAP kinase pathway, most significantly Ras and Grb2 (49,50). Our experiments further corroborate several reports that suggested the existence of a significant pool of Raf-1 kinase on the mitochondria (24,26,27,51). The association of endogenous Grb10, and especially Raf-1, with the mitochondria is easily lost under harsher cell disruption conditions, which explains contradictory results from other laboratories (5,26,52).
Grb10 lacks any apparent mitochondrial localization sequence. This, together with the fact the fact that overexpressing the protein leads to its mislocalization to the cytoplasm, and its peripheral association with membranes suggests that it uses a relatively low abundance lipid or protein as an anchor. One possible candidate for such an anchor is Raf-1 itself, which can be targeted to the mitochondria by Bcl-2 (26). This hypothesis was discounted because we estimate that there are approximately the same number of Grb10 and Raf-1 molecules per cell, 3 whereas only about 10% of total Raf-1 is localized to mitochondria. Furthermore, isotonic lysis of the cells completely releases Raf-1 while still retaining some Grb10 in the MP fraction. Another possibility is that Grb10 localization is mediated via its PH domain, which has been, in other proteins, shown to interact with phospholipids (44). Phosphatidylinositols have been identified in mitochondrial membranes, but there is very little published data in this area (53,54). In further support of this hypothesis, a minor Grb10 variant, whose mobility is consistent with the PH domain-deleted hGrb10␤, is found only in the soluble fraction.
In our initial report on the interaction of Grb10 with Raf-1, their co-immunoprecipitation was only possible following overexpression of a tagged Grb10 (15), most probably because only a fraction of total Raf-1 proteins co-localize with endogenous Grb10 (Fig. 5). However, both proteins were easily co-immunoprecipitated following partial purification of the mitochondria, thus demonstrating that they indeed do interact in vivo. The affinity of Grb10 for mitochondrial Raf-1 is increased following the activation of this kinase by ultraviolet light. This result is consistent with the relative co-localization of the Grb10-binding site with the Ras-binding domain of Raf-1. Studies of Ras-Raf interactions have shown this domain to be relatively inaccessible when the kinase is inactive (20,28,55).
The interaction of Grb10 with the anti-apoptotic mitochondrial Raf-1 kinase and our earlier results on the pro-apoptotic effects of Grb10 mutants (15) suggest that Grb10 could be used to regulate programmed cell death by modulating the activity of mitochondrial Raf-1. We thus propose that one of the functions of Grb10 is to serve as a link between plasma membrane receptors and the apoptosis-controlling complex on the mitochondrial outer membrane probably in collaboration with the phosphatidylinositol 3-kinase/Akt pathway, which plays a role in mediating the anti-apoptotic activity of the insulin and IGF-I receptors (56,57). The phosphatidylinositol 3-kinase/Akt path- 3 A. Nantel and D. Y. Thomas, unpublished data. FIG. 7. Two-hybrid analysis of Grb10-Raf-1 interaction. Fragments of the Raf-1 kinase were fused in frame with the LexA DNAbinding protein and tested in an "interaction trap" assay with an acidic activator domain on its own or fused with the Grb10 SH2 domain. Successful interaction conferred the ability to grow in the absence of leucine, which is indicated by the ϩ or ϩϩ symbols, depending on the growth rate of the colonies. Not shown are Raf-1 constructs 1-254, 1-212, 1-138, 1-55, and 55-138, which exhibited high background levels. way has recently been shown to affect the phosphorylation state of both Grb10 and mitochondrial Raf-1 (27,34).